Sie sind auf Seite 1von 167

Leading Edge

In This Issue

Fish Oil Turns the Tide on Insulin Resistance

PAGE 687
Inflammation mediated by macrophages promotes insulin resistance in obesity. Oh et al. now identify the G protein-coupled
receptor 120 (GPR120) on macrophages and fat cells as a receptor for omega-3 fatty acids (u-3 FAs). The authors show that
GPR120 activation by u-3 FAs inhibits inflammation pathways in macrophages and can reverse insulin resistance in mice.
These results provide a molecular basis for the anti-inflammatory effects of u-3 FAs and suggest that anti-inflammatory
treatments may ameliorate insulin resistance in obesity.

West Nile Virus Stopped at Its Source

PAGE 714
West Nile virus, a potentially deadly virus in humans, propagates in mosquitoes.
Cheng et al. now find that infection of West Nile virus triggers mosquito cells to
produce the lectin protein mosGCTL-1. This C-type lectin enhances entry of the
virus into additional mosquito cells through its interaction with a protein tyrosine
phosphatase receptor, which is homologous to human CD45. Blocking
mosGTCL-1 with an antibody disrupts the infective cycle of West Nile virus in
mosquitoes, suggesting a new strategy for controlling viral dissemination.

Antibodies Are Double-Trouble for Cancer

PAGE 699
Monoclonal antibodies are standard therapeutics for several cancers, including the
anti-CD20 antibody rituximab for B cell non-Hodgkin lymphoma (NHL). However,
antibodies are not curative and must be combined with cytotoxic chemotherapy
for clinical benefits. Now, Chao et al. identify CD47 as an antibody target in NHL
and demonstrate that combining anti-CD47 antibody with the rituximab antibody eradicates human NHL in mice. The synergistic
mechanism used by these two antibodies may be applicable to combined antibody treatments for many types of cancers.

ATAC Wears Two HATs in MAPK Signaling

PAGE 726
Extracellular cues often trigger MAP kinase (MAPK) signaling pathways, which
then activate downstream transcription factors like c-Jun. Here, Suganuma
et al. demonstrate that the ATAC histone acetyltransferase (HAT) governs the
response to MAPK signaling by serving as both a coactivator of transcription
and a suppressor of upstream signaling in the MAPK pathway. The authors
show that ATAC acetylates histone H4 at JNK target genes, which then serves
as a positive cofactor for basal transcription. In addition, ATAC directs upstream
MAPKs to the site of c-Jun binding and restricts the levels of JNK activation.

A Multitasking Leader mRNA

PAGE 737
Bacterial mRNAs often contain leader sequences that regulate transcription
of the adjacent coding region by binding metabolites and ions. For example,
the leader of the mRNA for the Salmonella Mg2+ transporter gene mgtA responds
to Mg . Now, Park et al. demonstrate that this leader also contains a short open reading frame with many proline codons, trans-
lation of which places mgtA expression under the control of cytoplasmic proline concentrations as well as Mg2+. Thus, leader
mRNAs can use distinct mechanisms to sense multiple intracellular signals.

Vivid Memories of Days Gone By

PAGE 762
Light responses and photoadaptation in the fungus Neurospora depend on the circadian transcription factor White Collar
Complex (WCC) and its negative regulator Vivid (VVD). Mazhan et al. now demonstrate how WCC and VVD cooperate to
discriminate light intensities over more than five orders of magnitude during the day. At night, previously synthesized VVD
serves as a molecular memory of the sun’s brightness during the preceding day and suppresses responses to light cues
of lower intensity, such as moonlight.

Cell 142, September 3, 2010 ª2010 Elsevier Inc. 649

Cell Polarity Goes with the Flow
PAGE 773
Planar polarization of epithelial cells allows the uniform alignment of hairs, cilia, and other cellular structures with tissue shape.
Now, Aigouy et al. combine experimental and theoretical approaches to show that polarity patterns in the Drosophila wing arise
during growth. Specifically, cell polarity is reoriented from a radial to a proximal-distal axis when mechanical stresses during
growth cause the cells to rotate or ‘‘flow’’ with respect to each other. Linking planar polarity to morphogenesis provides a simple
mechanism for coordinating the global polarity pattern with tissue shape.

Com-plexin’ with Semaphorins

PAGE 749
Semaphorins and their receptors, Plexins, are widely expressed protein families
that mediate repulsive signaling during cell guidance. Here, Liu et al. present two
X-ray crystal structures of PlexinC1, one in complex with the Semaphorin
Sema7A and another in complex with the Semaphorin mimetic A39R from the
smallpox virus. In both structures, the Semaphorin interacts with a PlexinC1
dimer in a novel edge-on, orthogonal geometry. These findings suggest that Plex-
ins are activated by ligand-induced dimerization during cell guidance.

Interspecies Organogenesis
PAGE 787
A goal of regenerative medicine is to derive organs from a patient’s pluripotent
stem cells (PSCs), but in vitro organogenesis is complex. Here, Kobayashi et al.
generate a functioning rat pancreas in mice without a pancreas by injecting rat
PSCs into mouse blastocysts. These interspecific chimeras provide proof of
principle for in vivo generation of organs derived from donor PSCs and for inter-
specific blastocyst complementation.

A Wormhole to the Origin of the Cortex

PAGE 800
The cerebral cortex or pallium controls the highest-order processing in mammals, but its evolution remains enigmatic. Now,
Tomer et al. develop an expression profiling technique to generate a gene expression map for the developing brain of the
segmented worm Platynereis dumerilii. Comparison of this map with that observed for the developing cerebral cortex
suggests a common evolutionary origin for the mammalian cortex and the worm’s mushroom body, a brain region in inver-
tebrates that processes sensory input.

Will the Real Chromosomal Proteins Please

Stand Up
PAGE 810
Proteomic analysis of large cellular structures is frequently hindered by the
presence of contaminants. In their analysis of mitotic chromosomes, Ohta
et al. overcome this problem by integrating proteomics with additional quanti-
tative and bioinformatic data—effectively adding a final purification step in silico.
This approach successfully pinpoints hitchhikers from amidst the 4,000 iden-
tified proteins and provides insight into the functional relationships among the
genuine constituents, including evidence that many more kinetochore-associ-
ated proteins exist than recognized previously.

Cell 142, September 3, 2010 ª2010 Elsevier Inc. 651

Leading Edge

Molecular Biology Select

The central importance of the tumor suppressor Retinoblastoma protein (Rb) in cell-cycle progression makes
its regulation a focal point for diverse biological processes, as evidenced by recent work described in this
issue’s Molecular Biology Select. These findings reveal new insight into Rb’s involvement in tissue regener-
ation and differentiation, as well as previously unrecognized mechanisms of Rb regulation.

Short-Term Inactivation, Lasting Benefits

Tissue regeneration in mammals is greatly restricted, whereas
many other types of vertebrates display astonishing feats of tissue
replacement, including regrowth of major structures such as
limbs. According to new findings by Pajcini et al. (2010), mamma-
lian innovations in the realm of tumor suppression are one factor
that is likely to contribute to this pronounced deficiency. The
authors show that combined inactivation of the tumor suppressors
retinoblastoma (Rb) and ARF reverses the differentiation of mouse
muscle cells, turning postmitotic cells into proliferating myoblasts,
and when these induced myoblasts are transplanted into donor
mice they successfully fuse into existing myofibers. Prior work in
newts has shown that Rb phosphorylation, which inactivates the
Dedifferentiated myocytes redifferentiate and fuse protein, promotes the reentry of myotubes into the cell cycle, a crit-
into existing muscles in vivo (visualized with green
ical initiating event in limb regeneration. The inspiration to addition-
fluorescent protein).
ally inactivate ARF was motivated by available evidence suggesting
its potential exclusivity to mammals and birds, and thus the authors
reasoned that ARF might be particularly important in mediating differences in regeneration potential between
vertebrate species. Could a similar strategy of Rb and ARF inactivation be used to promote cell-cycle entry of
endogenous cells for regeneration therapy? Although this remains to be seen, the observation that only tran-
sient inactivation of these tumor suppressors is needed for successful creation of regenerative cells may go
some way toward allaying concerns that such an approach would invariably promote cancer. Future efforts
are also likely to address whether this intervention triggers cell-cycle reentry for a range of mammalian cell
K.V. Pajcini et al. (2010). Cell Stem Cell 7, 198–213.

Fat Chance for Bone Formation

Previous in vitro studies suggested that Rb plays a critical role in the
decision of meschenchymal cells to become adipocyte or bone
cells, but in vivo evidence for this hypothesis has been lacking.
Calo et al. (2010) now show that Rb gives meschenchymal cells
the extra nudge they need to commit to becoming bone-forming
osteoblasts. Without Rb, these cells are more likely to differentiate
into brown fat cells, leading to reduced levels of calcified bones
and increased levels of brown adipose tissue in mice. To sort out Deletion of Rb in the embryo proper using Meox2-
how Rb regulates the fate of the meschenchymal cells during devel- Cre reduces the level of calcified bone as detected
opment, the authors engineered mice with the RB1 gene and/or by Alizarin Red staining. Image courtesy of J. Lees.
the p53 gene deleted only in uncommitted meschenchymal cells.
As expected, animals missing the tumor suppressor p53 develop multiple types of tumors, including osteosar-
comas. Combining the p53 mutation with a deletion of one RB1 allele increases the frequency of osteosarcomas,
whereas deletion of both RB1 alleles shifts the tumor distribution away from osteosarcomas towards brown fat
tumors. Thus, Rb regulates the fate of meschenchymal cells in a dose-dependent manner. These results are
surprising given that the majority of human osteosarcomas contain RB1 mutations. The authors speculate
that osteosarcoma tumors most likely arise from cells already committed to becoming osteoblasts, and muta-
tions in RB1 promote dedifferentiation of these cells and thus tumorigenesis.
E. Calo et al. (2010). Nature. Published online August 4, 2010. 10.1038/nature09264.

Cell 142, September 3, 2010 ª2010 Elsevier Inc. 653

Rb Gets Caught in a Custody Battle
Rb is caught in a tug-of-war between cyclin-dependent kinases
(Cdks) that inactivate it by phosphorylation to permit cell-cycle
progression and phosphatases that remove the modification
to promote cell-cycle arrest. New findings of Hirschi et al. (2010)
demonstrate that this conflict is waged over the same binding
interface of Rb. They provide structural evidence that the phospha-
tase PP1 interacts with Rb in a region previously shown critical for
the interaction of Rb with Cdks. They further show that PP1 can
suppress the activity of Cdk2-cyclin A toward Rb to block cell-
cycle progression in a human osteosarcoma cell line, and that
the complex of PP1 and Rb appears to be stable, or at least
more prevalent, at mitotic exit. Among the interesting questions
this work raises is, what factors determine the outcome of PP1
and Cdk competition? The authors propose that concentration
and subcellular localization of the competing proteins likely play
a role, but it remains unclear how this molecular dispute is settled
X-ray crystal structure of Rb (magenta) in complex under specific biological circumstances, for example after DNA
with the PP1 catalytic domain (gray). Image cour- damage.
tesy of S. Rubin.
A. Hirschi et al. (2010). Nat. Struct. Mol. Bio. Published online August
8, 2010. 10.1038/nsmb.1868.

Methylation Moves to the Front of the Line

Although first reported more than 30 years ago, methylation
of protein N termini (a-N-methylation) remains a poorly understood
protein modification. Thus, the recent identification by Schaner-Too-
ley et al. (2010) of an enzyme that catalyzes the reaction promises to
accelerate understanding of the modification’s functions. The
authors use methylation of a known target of a-N-methylation called
RCC1 (a Ran guanine nucleotide exchange factor) as an indicator of
the presence or absence of the enzymatic activity from fractionated
HeLa cell nuclear extracts. Fractions with methyltransferase activity
were subjected to mass spectrometry, leading to the identification of
the methyltransferase responsible, which the authors name NMRT
(N-terminal RCC1 methyltransferase). The NMRT crystal structure
facilitated the modeling of substrate recognition, and further analysis
defining the consensus sequence for target recognition suggested
Rb as a potential substrate. Subsequent assays provide evidence NMRT structure with RCC peptide modeled in the
active site. Image courtesy of I. Macara.
that Rb is modified by a-N-methylation, at least in some cell types.
a-N-methylation of RCC1 promotes stable association with chro-
matin, and loss of NMRT or the absence of the RCC1 methylation results in defects in mitosis. In contrast, the
purpose of Rb a-N-methylation remains a tantalizing mystery. Regardless of whether Rb a-N-methylation is rele-
vant to its roles in cell-cycle control, the discovery of NMRT opens a door through which others will likely follow.
Are there other a-N-methyltransferases? And if so, is their substrate specificity similar to NMRT? The answer may
be an indicator of whether this modification is an exotic posttranslational event or might instead be considerably
more common than previously appreciated. Another compelling question is whether there are a-N-demethylases
that reverse the modification.
C.E. Schaner-Tooley et al. (2010). Nature. Published online July 29, 2010. 10.1038/nature09343.

Robert P. Kruger

Cell 142, September 3, 2010 ª2010 Elsevier Inc. 655

Leading Edge

Phosphotyrosine Signaling: Evolving a New

Cellular Communication System
Wendell A. Lim1,2,* and Tony Pawson3,4,*
1Howard Hughes Medical Institute
2Department of Cellular and Molecular Pharmacology
University of California, San Francisco, San Francisco, CA 94158, USA
3Samuel Lunenfeld Research Institute, Mount Sinai Hospital, Toronto, Ontario M5G 1X5, Canada
4Department of Molecular Genetics, University of Toronto, Toronto, Ontario M5S 1A8, Canada

*Correspondence: (W.A.L.), (T.P.)

DOI 10.1016/j.cell.2010.08.023

Tyrosine phosphorylation controls many cellular functions. Yet the three-part toolkit that regulates
phosphotyrosine signaling—tyrosine kinases, phosphotyrosine phosphatases, and Src Homology
2 (SH2) domains—is a relatively new innovation. Genomic analyses reveal how this revolutionary
signaling system may have originated and why it rapidly became critical to metazoans.

Throughout human history, new technolo- (TyrK) phosphorylate specific target with an evolutionary process? Proteins
gies and technological platforms have tyrosine residues, phosphotyrosine phos- that bind or remove a posttranslational
constantly been invented. Only a small phatases (PTP) remove the phosphates, modification would seem useless without
fraction of these technologies go on to and Src Homology 2 (SH2) domains an enzyme to generate the modification
be widely adopted, but these can recognize the modifications (Pawson and, in principle, would not provide
ultimately have transformational conse- 1995). Together, these three modules a fitness advantage leading to its retention
quences. In the evolutionary history of form the ‘‘writer,’’ ‘‘eraser,’’ and ‘‘reader’’ and spread. The pTyr signaling platform
living organisms, we know that innovative toolkit that is common to many diverse provides a case study to look for plausible
molecular systems have appeared at key cellular information processing platforms stepwise pathways of the evolution of
points in time, and these are thought to (Figure 1A). A rich array of diverse and a multipart system.
have played a transformative role in major complex regulatory schemes can be Here, we reconstruct a possible history
evolutionary transitions in the tree of life. achieved through the dynamic interplay for the evolution of pTyr signaling. This
But how do such innovative molecular of these three modular functions (Pawson reconstruction is based on the recent
systems emerge, and how and why do et al., 1993; Pawson, 1995; Bhattachar- sequencing of the genomes of a number
some proliferate and become stably yya et al., 2006; Kholodenko, 2006). of organisms that originated both before
adopted by subsequent lineages? A combination of these modules can and after the emergence of metazoans
An example of such an innovative lead to higher-order functions (Figure 1B). from single-celled eukaryotic ancestors
molecular system is phosphotyrosine For example, there are several proteins (King et al., 2008). The genome sequence
(pTyr)-based signal transduction. This containing a combination of SH2 and of the choanoflagellate, Monosiga brevi-
molecular system for transmitting cellular kinase domains that can generate collis, has been particularly illuminating
regulatory information is estimated to positive feedback (phosphorylation of as choanoflagellates are thought to be
have appeared relatively recently in the tyrosine sites leads to SH2-mediated one of the closest single-celled relatives
history of life—600 million years ago, recruitment of the kinase, and subse- of metazoans. We present a model for
just prior to the emergence of multicellular quently, more extensive phosphorylation) how this three-part signaling system
animals (King et al., 2003; Pincus et al., (Pawson, 2004). Similarly SH2-phospha- could have plausibly evolved in a stepwise
2008; Manning et al., 2008). The pTyr sig- tase domain combinations can generate manner. We propose that once the
naling system has become an essential negative feedback (Tonks and Neel, complete three-part system was in place,
part of metazoan biology. For example, 2001). it may have rapidly taken hold in subse-
pTyr signaling plays a central role in many The three-part pTyr signaling toolkit quent lineages because it could generate
cell-to-cell communication pathways, thus raises a classic question in evolu- new regulatory behaviors without signifi-
including those that regulate proliferation, tionary biology: how do complex, interde- cant cross-interference with existing
differentiation, adhesion, hormone res- pendent systems arise? It is clear why regulatory circuits. We also discuss the
ponses, and immune defense (Hunter, a new system encompassing a writer, possible role of this new communication
2009). eraser, and reader might be extremely system in facilitating the transformative
In modern metazoans, pTyr signaling is useful. But, given their interdependence, evolutionary shift to multicellularity.
mediated by a toolkit of three distinct how could these individual components Given the incomplete record, however,
functional modules: tyrosine kinases arise in a stepwise fashion consistent such an evolutionary reconstruction is

Cell 142, September 3, 2010 ª2010 Elsevier Inc. 661

by an upstream Ser/Thr kinase on both
a Thr and Tyr residue within its activation
loop; Cdk1 is phosphorylated on Tyr14
by the inhibitory kinase Wee1). These
tyrosine modifications are clearly not
recognized by SH2 domains, but they
exert direct allosteric effects within the
proteins in which they occur. Thus, PTP
domains may have provided a fitness
benefit by negatively modulating these
rare but functionally important phosphor-
ylation events. Consistent with this model,
the proteins PTP2 and PTP3 in yeast
clearly have a functionally important role
in downregulating MAPK-mediated sig-
naling in response to pheromones or
osmolarity changes, explaining their
Figure 1. The Writer, Reader, Eraser pTyr Toolkit fitness benefit (Pincus et al., 2008). In
(A) In pTyr signaling, the tyrosine kinase (TyrK), Src Homology 2 (SH2), and phosphotyrosine phosphatase addition, PTPs may have played a general
(PTP) domains form a highly interdependent signaling platform. This platform serves as the writer, reader, role buffering against the occasional
and eraser modules, respectively, for processing pTyr marks.
(B) Components of pTyr signaling can be used to build complex circuits. For example, recruitment of an harmful stray phosphorylation of function-
SH2-TyrK protein to an initiating pTyr site can lead to amplification of tyrosine phosphorylation through ally important tyrosines.
a positive feedback loop. Where did these PTPs come from?
PTPs are likely to have arisen from
highly speculative. For example, we have tyrosine phosphatase activity. We a common ancestor of the related dual-
cannot rule out more complex paths refer to the single putative ‘‘SH2’’ domain specificity phosphatases, which are also
involving cycles of evolutionary gain and in yeast, found within the gene Spt6, as found in most single-celled eukaryotes
loss of components, nor the possibility a proto-SH2 domain because it does not (Kennelly, 2001; Alonso et al., 2004).
that similar components in distinct line- show pTyr binding (the domain has been Dual-specificity phosphatases are cata-
ages have independent origins. Nonethe- reported to show phospho-Ser/Thr lytic domains that can dephosphorylate
less, this model may provide a useful binding; Dengl et al., 2009). Thus, func- both pSer/Thr and pTyr substrates. The
framework for focusing studies of pTyr tionally, it cannot be considered a PTP and dual-specificity phosphatase
signaling origins and the origins of analo- ‘‘reader’’ domain that is part of a pTyr catalytic domains are distinct but are
gous multicomponent signaling plat- regulatory system. These observations evolutionarily related. They share a
forms. suggest a simple model: the first step in common fold and the core catalytic motif
We describe three possible stages in the evolution of the three-part pTyr HC(X)5R, in which a phosphocysteine
the emergence of the modern pTyr signaling machinery was likely to have enzyme intermediate is generated during
signaling toolkit, each represented by an been the emergence of a functional tyro- catalysis. (Sometimes, both dual-speci-
extant model organism (Figure 2). These sine phosphatase. But why would PTPs ficity phosphatases and classical PTPs
stages are representative; we do not arise in the pre-tyrosine kinase world? are referred to as PTPs; here, we use
claim to define the exact path of evolution, What functional use and fitness advan- this nomenclature only for the classical
but rather focus on identifying the domi- tage would this eraser domain provide in PTP domains that act only on pTyr). The
nant classes of stable intermediates that organisms lacking a writer domain? domains of dual-specificity phosphatases
can exist in the broader evolutionary land- The answer may lie in the fact that some have a shallower active site than classical
scape. Ser/Thr kinase domains, which are more PTPs, which may explain why they can
ancient than tyrosine kinases (dating dephosphorylate either Tyr or Thr/Ser
PTPs in a Pre-Tyrosine back close to the origins of eukaryotes), residues. In some lineages, dual-speci-
Kinase World can carry out sporadic but functionally ficity phosphatases have functionally
What came first, TyrK, PTP, or SH2 important phosphorylation of tyrosines. diverged further, giving rise to members
domains? Sequence analysis suggests Phosphoamino acid analysis of yeast that can act on lipid substrates, such as
that it was PTP domains. The genome of reveals a small but significant population the phosphoinositide phosphatases
a simple single-celled eukaryote like the of pTyr (Schieven et al., 1986). Moreover, PTEN and the myotubularins (Alonso
budding yeast Saccharomyces cerevisiae certain events, such as the activation et al., 2004). Thus, the PTPs appear to
shows no TyrK proteins and one proto- of mitogen-activated protein kinases have arisen from a somewhat promis-
SH2 domain, but a handful of PTP (MAPKs) and inhibition of the cell-cycle cuous class of multifunctional phospha-
proteins (Pincus et al., 2008) (Figure 2, kinase Cdk1, are known to involve phos- tases.
Stage 1). Most fungi have no more than phorylation of tyrosine residues (for acti- Despite the presence of PTP proteins in
five PTP proteins, and several of these vation, a MAPK must be phosphorylated fungi, there are striking differences

662 Cell 142, September 3, 2010 ª2010 Elsevier Inc.

Figure 2. Evolution of pTyr Signaling
Shown is a possible path for the emergence of phosphotyrosine (pTyr) signaling. We postulate three successive stages, each represented by what is observed in
a modern organism. The thickness of the tree reflects the approximate degree of usage of pTyr signaling (thicker lines mean more usage). Stage 1 (exemplified by
the budding yeast Saccharomyces cerevisiae) reflects the situation in early eukaryotes, in which PTPs emerged but were limited in number and complexity. They
were most likely used to reverse or process sporadic cross-phosphorylation of tyrosine residues by Ser/Thr kinases. S. cerevisiae has fewer than five PTP
proteins and no functional SH2 or TyrK domains. Stage 2 reflects systems in which functional SH2 domains emerged that were able to bind to pTyr motifs.
Together with Ser/Thr kinases with increased cross-reactivity for Tyr (such as tyrosine kinase-like or dual specificity Ser/Thr kinases), these systems may reflect
the most primitive of pTyr writer/reader/eraser systems. However, the lack of a dedicated Tyr kinase may have limited the utility and expansion of this toolkit. This
stage is potentially represented by the slime mold, Dictyostelium discoideum. Stage 3 reflects systems that evolved after the emergence of the modern TyrK
domain. We postulate that the full writer/reader/eraser system was of so much greater utility that its use expanded dramatically. This likely resulted in many
more proteins in these families, as well as much more complex, multidomain architectures than those seen in the earlier stages. This stage is represented by
both the multicellular metazoan and unicellular choanoflagellate lineages.

between these proteins and those proteins in which the PTP module has transfer enzymes; Bordo and Bork,
found in metazoans (Figure 2). For been functionally recombined with multi- 2002). Thus, fungal PTP proteins are
example, there are far fewer PTPs in ple other signaling modules. In contrast, very simple (one to two domains) and
fungi (5/genome versus 40/genome in fungi, the PTP domains are all either lack the combinatorial complexity of
in metazoans) and they are considerably in simple single-domain proteins or in metazoan PTP proteins. The simplicity
less complex in domain architecture combination with a single rhodanase-like and low number of PTPs in yeast sug-
(Pincus et al., 2008). Metazoan PTP domain (a putative regulatory domain gests that in early single-celled eukary-
proteins tend to be large multidomain that is homologous to a class of sulfur otes, PTP domains had fairly limited

Cell 142, September 3, 2010 ª2010 Elsevier Inc. 663

functional utility, especially when com- pTyr-binding SH2 domains and may kinase domain followed by an SH2
pared to their broad and complex usage therefore provide a living representative domain, a domain combination that is
in metazoans. of this second evolutionary stage (Fig- somewhat similar to metazoan cyto-
Unlike PTPs, there are no known pTyr- ure 2, Stage 2). Dictyostelium is a soil- plasmic tyrosine kinases like Src (Monia-
binding SH2 domains in fungi, although living amoeba that has a unicellular kis et al., 2001). The Shk catalytic domain,
there is one clearly homologous domain lifestyle in the presence of bacterial however, lacks motifs characteristic of
found in the yeast protein SPT6. This food. However, when food is depleted, bona fide tyrosine kinases and biochemi-
protein, which has a domain with an individual cells aggregate in response to cally displays dual specificity toward
SH2-like sequence and fold, is involved the chemoattractant cAMP to form serine/threonine and tyrosine residues.
in the regulation of transcription elonga- a multicellular structure, which then Indeed, Dictyostelium differs from
tion, and the SH2 domain binds to the develops into a fruiting body through the metazoans and choanoflagellates in that
Ser/Thr phosphorylated C-terminal tail of differentiation of stalk and spore cells. its genome does not encode any modern
RNA polymerase II. The domain does The rudimentary pTyr-SH2 system in tyrosine-specific protein kinases. For
not bind to pTyr (Dengl et al., 2009). Inter- Dictyostelium is important for aspects of example, metazoan STAT proteins are
estingly, a single SPT6 ortholog, with the this differentiation process, including usually phosphorylated by Janus tyrosine
same overall domain architecture, is intracellular responses to both cAMP kinases (JAKs), but there are no JAKs in
found in all eukaryotes, including all fungi and the morphogen differentiation in- Dictyostelium (Kay, 1997). This suggests
and metazoans (but not prokaryotes). This ducing factor or DIF (which induces the the possibility that signaling proteins con-
finding suggests that in early eukaryotes, differentiation of prestalk cells), as well taining SH2 domains such as STATs
a proto-SH2 domain emerged to perform as for transcriptional regulation in res- evolved before the modern tyrosine
a highly specialized function—one that ponse to hyperosmotic stress. These kinases with which they are associated
was unrelated to the flexible modular observations are consistent with early in metazoans. The identity of the kinase
pTyr recognition function of the modern pTyr-SH2 signaling playing a role in responsible for STAT tyrosine phosphory-
SH2 domain. This proto-SH2 domain cellular responses to changing environ- lation, and the consequent formation of
most likely did not ‘‘read’’ pTyr modifica- mental conditions. SH2-binding sites, remains mysterious.
tions, but instead recognized a special- The Dictyostelium genome specifies 13 How, then, is tyrosine phosphorylation
ized related modification. Thus, although proteins with SH2 domains (as well as of Dictyostelium proteins such as the
SPT6 is likely to represent an early a single Spt6 homolog). These 13 proteins STATs controlled? Thus far, genetic anal-
ancestor or relative that eventually gave cluster into five basic domain architec- ysis has not identified a specific relevant
rise to modern SH2 domains, it cannot tures, two of which are homologous to kinase, and it has been proposed that, in
be considered a functional part of a pTyr metazoan SH2 proteins. Notably, Dic- contrast to mammalian STATs, there
regulatory toolkit. We therefore postulate tyostelium has four STAT (signal trans- may be basal constitutive phosphoryla-
that early eukaryotes had only a pTyr ducers and activators of transcription) tion of Dictyostelium STAT tyrosine sites,
eraser function (mediated by PTPs) with proteins that are very similar to metazoan which is regulated by changes in PTP
no specialized complementary reader or STAT transcription factors (Kay, 1997; activity in response to extracellular
writer functions. Kawata et al., 1997). For example, they signals (Langenick et al., 2008). One of
In summary, the PTP domain and all have an SH2 domain juxtaposed to the PTPs in Dictyostelium, PTP3, binds
a structural ancestor of the SH2 domain a DNA-binding region; they are inducibly and dephosphorylates STATc, thereby
appear to have arisen in early single- phosphorylated on tyrosine residues in blocking SH2-mediated dimerization and
celled eukaryotes, but are likely to have response to stress or the extracellular STATc accumulation in the nucleus.
functional origins that are not directly signaling molecule DIF; they undergo Signaling induced by the DIF morphogen
related to their later function in modern pTyr/SH2-mediated dimerization and appears to transmit signals by inhibiting
pTyr regulatory systems. These compo- then translocate to the nucleus to regulate PTP3 activity and consequently boosting
nents may have provided a limited but the expression of specific genes. Dictyos- STATc tyrosine phosphorylation and
incremental fitness advantage, even in telium also has an ortholog of the STATc-dependent gene expression.
the absence of a specialized tyrosine mammalian E3 ubiquitin ligase Cbl, which Although Dictyostelium lacks true tyro-
kinase domain. uses SH2 and Ring domains to couple sine kinases, it is noteworthy that its
pTyr signals to the ubiquitination ma- genome has a significant expansion in
Toward a Write/Read/Erase System chinery (Langenick et al., 2008). The re- the number of putative dual-specificity
In the early days of a more sophisticated maining three domain architectures of protein kinases (there are 70, also known
pTyr-signaling system, we suggest that Dictyostelium SH2 proteins are distinct as tyrosine kinase-like or TKL kinases)
a proto-SH2 domain (mostly likely a from those found in other sequenced (Manning et al., 2008). This set includes
homolog of the yeast Spt6 protein) in organisms. The LrrB protein has an SH2 the Shk catalytic domain, described
a single-celled organism acquired the domain linked to a leucine-rich repeat above. It is unlikely that any of these
new and functionally beneficial ability to domain (Sugden et al., 2010), whereas kinases are precursors of modern tyrosine
bind to pTyr-containing peptide motifs. the FbxB protein has an F-box followed kinases. However, it is plausible that these
The slime mold Dictyostelium discoideum by an SH2 domain and ankyrin repeats. represent the first evolutionary form of
has the simplest repertoire of bona fide In addition, the Shk proteins have a protein the ‘‘writer’’ function in a prototype pTyr

664 Cell 142, September 3, 2010 ª2010 Elsevier Inc.

three-part regulatory system. The union of sine kinases (Lee and Jia, 2009). It is functions, thus leading to the subsequent
an SH2 domain and a dual specificity therefore probable that BY kinases expansion of the complete set. Although
kinase domain, as found in the Shk evolved separately from metazoan tyro- the PTP and SH2 domains had utility in
proteins, may be an early example of link- sine kinases and operate in a different simpler organisms, their much larger
ing ‘‘reader’’ and ‘‘writer’’ modules to fashion. functional potential was not unleashed
achieve more complex functions such as The new eukaryotic tyrosine kinase until the emergence of the TyrK domain.
positive feedback. Nonetheless, the domain appears to have been a game The rapid expansion of the pTyr
limited functionality of the dual specificity changing innovation (Figure 2, Stage 3). signaling machinery in the ancestors of
kinases in carrying out tyrosine phosphor- The total number of tyrosine kinase choanoflagellates and animals is reminis-
ylation may have limited the capabilities of proteins in both choanoflagellate and cent of how technology expands in
this early system. This may explain the metazoan species is in the range of 30– quantum jumps, especially in situations
very modest expansion of pTyr signaling 150 per genome (Pincus et al., 2008; involving codependent technologies. For
in organisms such as Dictyostelium. Manning et al., 2008). Among sequenced example, the value of the laser expanded
These observations paint the following genomes, there is a striking absence of dramatically after the later invention of
picture of Dictyostelium pTyr signaling species with only a small number of TyrK the complementary technology of fiber
and, by extension, of an early phase in proteins. This all-or-none sudden jump in optics. This codependent technology al-
the evolution of pTyr communication. the number of TyrK proteins suggests lowed lasers to be repurposed to rapidly
SH2 domains have acquired pTyr-binding their importance as they appear to have displace electrical transmission via
activity and are found in several distinct undergone rapid expansion and subse- copper wires as the backbone of global
combinations with other types of signaling quent retention. communication (Alwayn, 2004). Thus,
domains. Among these, the STAT and Cbl What is perhaps more striking is the although lasers had standalone utility,
proteins are shared with metazoans, observation that the emergence of the their major application had to await the
whereas the LrrB, FbxB, and Shk proteins TyrK domain and its rapid expansion introduction of complementary tech-
are unique to Dictyostelium. But no dedi- correlates with an equally rapid expansion nology. The expansion of molecular com-
cated modern tyrosine kinases have of PTP and SH2 domains within the same ponents in biology is likely to be similar.
been found, and the dynamic control of genomes (Pincus et al., 2008). Although A toolkit of writer, reader, and eraser
tyrosine phosphorylation may be primarily fungi and Dictyostelium have 5 PTP functions may be of full use only when all
regulated by PTPs. Although functionally proteins, metazoans, and choanoflagel- components are present. Thus, it may
important for aggregation and differentia- lates have 30–40 per genome. Similarly, be common for any system of this type
tion, the pTyr signaling system has not Dictyostelium has approximately ten to show a quantum ‘‘all-or-none’’ expan-
acquired the pervasive influence evident SH2 domain-containing proteins (fungi sion only when the final piece of the toolkit
in M. brevicollis and metazoans, perhaps have none), whereas metazoans and emerges.
because of the lack of an efficient tyrosine choanoflagellates have 100 each.
kinase. Put another way, Dicytostelium Thus, both PTP and SH2 proteins Applying the New pTyr Toolkit
has effective pTyr readers and erasers, undergo a roughly 10-fold increase in to Different Functions
but the writer is poorly developed. number per genome after the emergence Although both choanoflagellate and
of the TyrK domain. Moreover, the metazoan lineages show a large expan-
Invention of TyrK and Expansion proteins containing SH2 and PTP sion of the three-part pTyr regulatory
of the pTyr Toolkit domains become far more complex and machinery, the way in which these
Current analysis suggests that the varied (Jin et al., 2009). For example, in components are used appears to be quite
modern tyrosine kinases arose just prior yeast and Dictyostelium, SH2 and PTP different. When one examines the domain
to the evolution of the metazoans. Aside proteins normally are very simple one or types that co-occur with TyrK, SH2, or
from metazoans, canonical tyrosine two domain proteins. However, in line- PTP domains, one finds many distinct
kinases have thus far only been observed ages that have modern TyrK proteins, combinations that are unique to each
in the choanoflagellates, which appear to SH2 and PTP proteins almost always lineage (Pincus et al., 2008; Manning
be the closest known single-celled rela- comprise three to ten domains. et al., 2008). These differences in domain
tives of metazoans (King et al., 2008). These observations are consistent with combinations imply distinct functions for
The absence of significant numbers of the following model. When a far more proteins containing these domains in the
such tyrosine kinases in any other branch efficient TyrK domain—or ‘‘writer’’ func- choanoflagellate and metazoan lineages
of life suggests that this new catalytic tion—emerged, this dramatically increased (Li et al., 2009). Assuming that the evolu-
domain evolved in a recent common the functional utility of the pre-existing PTP tion of new TyrK, SH2, and PTP proteins
ancestor of choanoflagellates and meta- (eraser) and SH2 (reader) domains. As occurred by recombination with new
zoans, most likely as a branch of the older a three-part toolkit—a catalytic domain to accessory domains (Jin et al., 2009; Pei-
Ser/Thr kinases. Some bacteria do have generate pTyr, an interaction domain to sajovich et al., 2010), this observation
specialized tyrosine kinases (BY kinases), bind to these pTyr sites, and an enzyme also implies that the complete signaling
but these resemble P loop NTPases to dephosphorylate them—this domain toolkit emerged only shortly before the
(nucleotide triphosphatases) and are set could be used to encode and execute divergence of metazoans and choanofla-
structurally unrelated to eukaryotic tyro- a far wider and diverse range of regulatory gellates (i.e., shortly before the evolution

Cell 142, September 3, 2010 ª2010 Elsevier Inc. 665

of metazoan multicellularity) and that communication. Because of this valuable gous new molecular information curren-
much of the divergent expansion of these high encoding potential, there is extreme cies could have, in principle, been able
domain families occurred after the lineage pressure to quickly fill this region of the to serve as the substrate for dramatic
split. spectrum. Moreover, the exact type of phenotypic innovation. In this context,
Thus, earlier assumptions that pTyr information carried by each region of the plants make extensive use of protein
signaling is only used in metazoan cell- spectrum is flexible—for example, the phosphorylation and have numerous
cell communication are clearly incorrect. same region of the spectrum can be as- transmembrane receptor Ser/Thr kinases,
Choanoflagellates do not form the com- signed to different functions in different but they lack conventional tyrosine
plex and permanent cell-cell organization countries. We hypothesize that the new kinases, indicating that pTyr-based
that metazoans do, yet surprisingly they pTyr signaling system that emerged prior signaling is not the only mechanism of
have a comparable (if not greater) number to metazoans presented similar new information transfer through which organ-
of pTyr signaling proteins (Manning et al., opportunities to transmit more informa- isms can achieve multicellularity.
2008). Sequencing of other organisms tion. This virgin system was rapidly
that arose near the origins of metazoans exploited, though the way it has been Is pTyr Signaling Saturated?
is ongoing. Preliminary data also suggest used appears to be different in the two How close is the pTyr signaling system to
a large number of pTyr signaling proteins branches (metazoans and choanoflagel- being saturated? Is there still available en-
in other single-celled relatives of meta- lates) that emerged after the complete coding potential that could be tapped for
zoans. Thus, it may be more reasonable toolkit was established. the evolution of new pathways and
to view the pTyr signaling system as an It is tempting to speculate that the behaviors? It is difficult to answer these
innovative but generic information pro- emergence of a new signaling system questions. However, the fact that new
cessing system that could potentially be with high encoding potential may have pTyr signaling proteins appear to be asso-
used for transmitting many different types played a key role in the emergence of ciated with advanced processes like
of information. a new, complex biological function such adaptive immunity suggests that there
as metazoan multicellularity. Such large- was still some remaining encoding poten-
Orthogonal Signaling: A Platform scale phenotypic evolutionary innova- tial in the system as late as the evolution of
for Biological Innovation tions may require and coincide with mammals. The evolutionary history re-
When the three-part pTyr system first innovations in basic molecular compo- constructed here begs many questions.
emerged, it presented a new platform nents (King, 2004; Rokas, 2008). Are there new regulatory toolkits evolving
with which to transmit information that Indeed, we speculate that pTyr now or in the future? Will these new tool-
was orthogonal to pre-existing signaling signaling may provide a more general kits be the substrate required for the
systems. Because it was based on a model for the generation of multicompo- next big evolutionary innovation?
distinct covalent modification, new regu- nent biological systems, involving first a The importance of new molecular tool-
latory circuits could be assembled with limited stepwise development of elements kits is conversely also very relevant to
these components without significant that together have a rudimentary biolog- the emerging field of synthetic biology, in
cross-interference with pre-existing net- ical utility, followed by an explosive which the goal is to engineer cellular
works. Thus, this brand new signaling spread, once all of the components of systems with new functions. A major
apparatus probably had a high encoding the mature system are in place. Explora- potential limitation is how to build such
potential for evolving dramatically new tion of this concept, and further analysis new functions in a reliable fashion that
functions, such as those involved in multi- of the evolution of pTyr signaling, will be does not cross-interfere in unanticipated
cellularity. One possible problem that assisted by the increasing sequence infor- ways with existing systems (Lim, 2010).
could be caused by the expansion of the mation being gathered for both unicellular Can we develop new synthetic molecular
new pTyr signaling enzymes might be and multicellular eukaryotes (Srivastava signaling currencies that are orthogonal
excessive general phosphorylation of et al., 2010), which will no doubt yield to existing natural ones, and would these
tyrosine residues throughout the pro- surprises akin to the discovery of exten- systems dramatically facilitate our ability
teome. Interestingly, however, organisms sive pTyr signaling in M. brevicollis. More- to reliably and predictably endow cells
using pTyr signaling may have developed over, as genomic information bracketing with innovative new functions?
a simple solution to deal with this other major evolutionary transitions
problem—a decrease in the tyrosine becomes available, it will be interesting Conclusions
content of proteins across the proteome to see whether these innovations are Current data suggest that PTP and SH2
is observed to correlate with tyrosine also associated with the explosive expan- domains evolved before modern TyrK
kinase expansion (Tan et al., 2009). sion of new molecular toolkits. domains, most likely to process pTyr
A new orthogonal signaling system like A key point here is that the specific modifications sporadically catalyzed by
the pTyr signaling platform can be viewed emergence of the pTyr toolkit may not Ser/Thr kinases. However, the PTP and
as analogous to a newly opened region in have been essential for the evolution of SH2 domain protein families did not
the telecommunications spectrum. New multicellularity, but rather, any number of expand dramatically until the emergence
frequencies provide the opportunity for new orthogonal signaling toolkits with of an efficient TyrK. We postulate that
transmitting large amounts of information the same high encoding potential could only with the complete toolkit of writer
as there is little interference from existing have served a similar role. Other analo- (TyrK), reader (SH2), and eraser (PTP)

666 Cell 142, September 3, 2010 ª2010 Elsevier Inc.

domains, was the full encoding potential three-part system. One cannot help but Kennelly, P.J. (2001). Chem. Rev. 101, 2291–2312.
of this system unleashed, leading to rapid wonder what other simple pieces of Kholodenko, B.N. (2006). Nat. Rev. Mol. Cell Biol.
expansion and elaboration of these molecular machinery may be lying around 7, 165–176.
domain families. This type of explosive in today’s biological systems, of limited King, N. (2004). Dev. Cell 7, 313–325.
increase in component usage may prove utility now, but awaiting the emergence King, N., Hittinger, C.T., and Carroll, S.B. (2003).
to be common to all multipart molecular of some as yet unknown complementary Science 301, 361–363.
systems. The emergence of the modern component that will generate a complete King, N., Westbrook, M.J., Young, S.L., Kuo, A.,
TyrK maps just prior to the split between toolkit that will help to drive future evolu- Abedin, M., Chapman, J., Fairclough, S., Hellsten,
metazoans and choanoflagellates. These tionary innovation. U., Isogai, Y., Letunic, I., et al. (2008). Nature 451,
two lineages appear to have used this 783–788.
new molecular communication system in ACKNOWLEDGMENTS Langenick, J., Araki, T., Yamada, Y., and Williams,
distinct ways—multicellular metazoans J.G. (2008). J. Cell Sci. 121, 3524–3530.
We thank D. Pincus, B. Mayer, P. Beltrao, O. Hoeller,
used it for cell-cell coordination, whereas Lee, D.C., and Jia, Z. (2009). Trends Biochem. Sci.
R.Linding, G. Superti-Furga, N.King, N.Helman, L.
unicellular choanoflagellates used it for 34, 351–357.
Holt, A. Horwitz, T. Miller, G.Manning, T. Hunter,
distinct but as yet uncharacterized D.Morgan, J. Williams, H. Bourne, and I. Ernberg Li, W., Scarlata, S., and Miller, W.T. (2009).
functions. for helpful comments. This work was supported by Biochemistry 48, 5180–5186.
Thus, we are able to reconstruct a plau- the Howard Hughes Medical Institute (W.L.), the Lim, W.A. (2010). Nat. Rev. Mol. Cell Biol. 11,
sible model by which the pTyr signaling National Institutes of health (GM55040, GM62583, 393–403.
GM081879, and EY016546—W.L.), the Packard Manning, G., Young, S.L., Miller, W.T., and Zhai, Y.
machinery could have evolved in a rela-
Foundation (W.L.), the National Science Foundation (2008). Proc. Natl. Acad. Sci. USA 105, 9674–9679.
tively simple stepwise manner into what Synthetic Biology Engineering Research Center
today is a complex and highly interdepen- Moniakis, J., Funamoto, S., Fukuzawa, M.,
(W.L.), the Canadian Institutes for Health Research
dent system. In this model, evolution is Meisenhelder, J., Araki, T., Abe, T., Meili, R.,
(T.P.—MOP-6849), Genome Canada (T.P.), and
Hunter, T., Williams, J., and Firtel, R.A. (2001).
opportunistic and forward looking, the Canadian Cancer Society Research Institute
Genes Dev. 15, 687–698.
borrowing, and repurposing machinery (T.P.).
Pawson, T. (1995). Nature 373, 573–580.
that pre-exists. The first simple PTP
proteins likely arose from the more REFERENCES Pawson, T. (2004). Cell 116, 191–203.

ancient Ser/Thr phosphatase family and Pawson, T., Olivier, P., Rozakis-Adcock, M.,
Alonso, A., Sasin, J., Bottini, N., Friedberg, I., McGlade, J., and Henkemeyer, M. (1993). Philos.
may have been maintained initially as Friedberg, I., Osterman, A., Godzik, A., Hunter, Trans. R. Soc. Lond. B Biol. Sci. 340, 279–285.
a way to reverse the unavoidable occa- T., Dixon, J., and Mustelin, T. (2004). Cell 117,
sional tyrosine phosphorylation event Peisajovich, S.G., Garbarino, J.E., Wei, P., and
Lim, W.A. (2010). Science 328, 368–372.
catalyzed by a Ser/Thr kinase. In some Alwayn, V. (2004). Optical Network Design and Im-
cases, like the MAPKs, which are present plementation (Indianapolis, IN: Cisco Press). Pincus, D., Letunic, I., Bork, P., and Lim, W.A.
(2008). Proc. Natl. Acad. Sci. USA 105, 9680–9684.
in all eukaryotes, these tyrosine phos- Bhattacharyya, R.P., Reményi, A., Yeh, B.J., and
phorylation events appear to have Lim, W.A. (2006). Annu. Rev. Biochem. 75, Rokas, A. (2008). Annu. Rev. Genet. 42, 235–251.

become exploited and fixed as actual 655–680. Schieven, G., Thorner, J., and Martin, G.S. (1986).
parts of signal transmission, alongside Bordo, D., and Bork, P. (2002). EMBO Rep. 3, 741– Science 231, 390–393.
746. Srivastava, M., Simakov, O., Chapman, J., Fahey,
Ser/Thr phosphorylation events. SH2
domains also likely arose from a pre-ex- Dengl, S., Mayer, A., Sun, M., and Cramer, P. B., Gauthier, M.E., Mitros, T., Richards, G.S., Con-
(2009). J. Mol. Biol. 389, 211–225. aco, C., Dacre, M., Hellsten, U., et al. (2010).
isting fold in the SPT6 protein, which is
Hunter, T. (2009). Curr. Opin. Cell Biol. 21, Nature 466, 720–726.
found in all eukaryotes but has no pTyr
140–146. Sugden, C., Ross, S., Bloomfield, G., Ivens, A.,
binding activity. But this fold, once co-
Jin, J., Xie, X., Chen, C., Park, J.G., Stark, C., Skelton, J., Mueller-Taubenberger, A., and Wil-
opted for this function, began to expand, liams, J.G. (2010). J. Biol. Chem. 285, 22927–
James, D.A., Olhovsky, M., Linding, R., Mao, Y.,
most likely because of its ability to and Pawson, T. (2009). Sci. Signal. 2, ra76. 22935.
contribute to a wider range of modular Kawata, T., Shevchenko, A., Fukuzawa, M., Tan, C.S., Pasculescu, A., Lim, W.A., Pawson, T.,
signaling events. But the full utility of these Jermyn, K.A., Totty, N.F., Zhukovskaya, N.V., Ster- Bader, G.D., and Linding, R. (2009). Science 325,
components was only unleashed upon ling, A.E., Mann, M., and Williams, J.G. (1997). Cell 1686–1688.
the emergence of the modern TyrK 89, 909–916. Tonks, N.K., and Neel, B.G. (2001). Curr. Opin. Cell
domain, which led to the highly expanded Kay, R.R. (1997). Curr. Biol. 7, R723–R725. Biol. 13, 182–195.

Cell 142, September 3, 2010 ª2010 Elsevier Inc. 667

Leading Edge

the inhibitor (Figure S1B, lanes 3 and 4).

What Controls T Cell Receptor Time course experiments indicated that
rather than enhancing it, the Emut1+2
Phosphorylation? mutations somewhat slowed the kinetics
of CD33CD phosphorylation (data not
shown). In vivo, therefore, the wild-type
In a study published in Cell, Xu et al. (2008) To show that CD33CD is not phosphor- form of CD33CD seemed to be at least
raised the intriguing possibility that tyro- ylated in resting Jurkat T cells, Xu et al. as accessible to tyrosine kinases as
sine residues present in the immunore- used a chimeric protein consisting of the mutated, membrane nonassociating
ceptor tyrosine-based activation motifs murine CD33CD fused to the extracellular form of the protein when kinase activity
(ITAMs) of the cytoplasmic domain of the and transmembrane domains of the could be observed by inhibiting phospha-
CD33 subunit of the T cell receptor (TCR) human natural killer cell inhibitory recep- tase activity. Finally, CD33 was among the
are sequestered in the membrane. Here, tor KIR2DL3. We generated FLAG-tagged more heavily tyrosine phosphorylated
they are inaccessible to tyrosine kinases forms of this protein and a second proteins present in whole-cell lysates
that would otherwise drive signaling. chimera consisting of the extracellular from pervanadate-treated wild-type
As a consequence, the authors proposed and transmembrane domains of KIR2DL3 Jurkat T cells (Figure S1C), emphasizing
that ‘‘triggering’’ of the TCR—that is, the fused with CD33CD mutated at the set of its accessibility to kinases relative to other
events leading to phosphorylation of the clustered positively charged residues kinase targets.
receptor—must be dependent upon proposed to mediate membrane asso- How can our observations be recon-
some, as yet undefined, process that ciation, that is, the mutant Emut1+2 ciled with those of Xu et al.? Our experi-
liberates the ITAMs from the membrane. described by Xu et al. (see Experimental ments suggest that membrane associa-
In an accompanying Preview, Kuhns and Procedures). We expressed the two tion, if it occurs, has less impact on
Davis (2008) invoked the idea that mem- chimeras at similar levels in Jurkat CD33CD phosphorylation than might
brane association constitutes a ‘‘safety T cells using a lentivirus expression have been expected on the basis of the
catch’’ that prevents untoward triggering system (Figure S1A) and probed for Xu et al. work. The strongest evidence for
of the TCR. Such effects add an additional phosphorylation of CD33CD by western lipid association is, of course, the NMR
layer of complexity to the TCR triggering blotting (Figure S1B). Neither KIR2DL3/ structure of CD33CD stably bound to
problem. CD33CD nor KIR2DL3/CD33CDEmut1+2 acidic lipids. It should be borne in mind,
Using equilibrium dialysis, Xu et al. immunoprecipitated with anti-FLAG anti- however, that we know little about the
showed that the cytoplasmic domain of body showed detectable phosphorylation actual distribution of inner leaflet
CD33 (CD33CD) partitioned into acidic in resting T cells (Figure S1B, lanes 1 lipids around receptors embedded in the
rather than zwitterionic lipids. They then and 2). Even after very long exposures of membrane. Recent work by Zech et al.
went on to derive a structural model for the blots, we were unable to detect phos- (2009) shows that the acidic phospholipid
ITAMs bound by acidic lipids using phorylation over background levels of phosphatidylserine is enriched in mem-
nuclear magnetic resonance (NMR)- cross-reactivity of the antibody and we brane sheets isolated from activated
based methods. A new in vivo FRET- saw no differences in signal for the wild- T cells using beads coated with anti-CD3
based assay suggested that the associa- type versus the mutant proteins (Fig- antibodies, but the extent to which this
tion of the ITAM with the membrane was ure S1B). This was also the case for reflects the distribution of lipids in the
dependent on a cluster of positively wild-type and Emut1+2-mutated forms immediate vicinity of the TCR in resting
charged residues in the membrane-proxi- of full-length CD33 incorporated into the or activated cells is unclear. The density
mal amino-terminal region of CD33CD; TCR complex (data not shown). or properties of the acidic lipids adjacent
mutation of these residues reduced the These findings suggest either that to CD33 in resting cells might be insuffi-
association of CD33CD with the mem- the Emut1+2 mutations do not release cient to sustain an interaction of the type
brane as detected by the FRET assay. the ITAMs from the membrane or that, demonstrated in vitro by Xu et al. Whatever
The obvious question was, does mem- in vivo, the association with the mem- interaction ITAMs have with membranes
brane association prevent kinases from brane is weaker than expected and in vivo might be too weak and dynamic to
accessing ITAMs in vivo? Xu et al. showed some other process prevents CD33CD prevent access of the kinases.
that the wild-type CD33 ITAM is not phos- phosphorylation in resting cells. We sus- The reasons for the low levels of CD33CD
phorylated in resting T cells, implying that pected that tyrosine phosphatases might phosphorylation in resting T cells remain to
it may indeed be completely inaccessible be responsible for the lack of phosphory- be fully worked out. In any discussion of the
to kinases. Somewhat surprisingly, how- lation of the wild-type and mutant forms of possible constraints on tyrosine phosphor-
ever, they did not go on to clinch the argu- CD33CD. To test this, we incubated cells ylation of the TCR, some consideration
ment by showing that there was increased expressing the chimeric proteins with must be given to the protein tyrosine phos-
phosphorylation of the membrane nonas- the tyrosine phosphatase inhibitor, perva- phatases (PTPs) present at the cell surface,
sociating CD33CD mutant. The results nadate (Swarup et al., 1982). Both the such as CD45 (Hermiston et al., 2003).
of this experiment turn out to be inconsis- wild-type and mutated forms of CD33CD Leukocytes express as many as 105 CD45
tent with the predictions of Xu and were heavily phosphorylated following molecules, i.e., 2–3 for every TCR (Wil-
colleagues. incubation of the cells for 20 min with liams and Barclay, 1986), each with very

668 Cell 142, September 3, 2010 ª2010 Elsevier Inc.

broad substrate specificity (Barr et al., ACKNOWLEDGMENTS REFERENCES
2009); the catalytic activities of PTPs have
This work was funded by the Wellcome Trust, the Barr, A.J., Ugochukwu, E., Lee, W.H., King, O.N.,
been estimated to be 10- to 1000-fold
United Kingdom Medical Research Council, and Filippakopoulos, P., Alfano, I., Savitsky, P.,
higher than that of tyrosine kinases the Portuguese Fundação para a Ciência e a Tec- Burgess-Brown, N.A., Müller, S., and Knapp, S.
(discussed by Fischer et al., 1991). Another nologia. (2009). Cell 136, 352–363.
contributing factor is the activity of inhibi- Fischer, E.H., Charbonneau, H., and Tonks, N.K.
tory kinases such as Csk, which also Ricardo A. Fernandes,1,4 Chao Yu,1,4 (1991). Science 253, 401–406.
constrain T cell activation (Schoenborn Alexandre M. Carmo,2 Hermiston, M.L., Xu, Z., and Weiss, A. (2003).
et al., 2009). Progress in understanding Edward J. Evans,1 Annu. Rev. Immunol. 21, 107–137.
TCR triggering will require the teasing P. Anton van der Merwe,3 Kuhns, M.S., and Davis, M.M. (2008). Cell 135,
apart of these and other competing and Simon J. Davis1,* 594–596.
1Nuffield Department of Clinical Medicine
factors and the identification of those O’Shea, J.J., McVicar, D.W., Bailey, T.L., Burns,
factors with the largest impact on net and Medical Research Council Human
C., and Smyth, M.J. (1992). Proc. Natl. Acad. Sci.
Immunology Unit, The Weatherall Institute of
phosphorylation of the TCR. The effects USA 89, 10306–10310.
Molecular Medicine, University of Oxford,
of pervanadate on phosphorylation levels Schoenborn, J.R., Zhang, C., Shokat, K.M., and
Oxford OX3 9DU, UK
in resting T cells observed by ourselves 2Group of Cell Activation and Gene Weiss, A. (2009). J. Immunol. 182, Abstract 35.4.
(Figures S1B and S1C) and by others Expression, Institute for Molecular and Swarup, G., Cohen, S., and Garbers, D.L. (1982). Bi-
many years ago (e.g., O’Shea et al., 1992) Cellular Biology, 4150-180 Porto, Portugal ochem. Biophys. Res. Commun. 107, 1104–1109.
suggest that PTPs could be key. For us, and Abel Salazar Institute for Biomedical Williams, A.F., and Barclay, A.N. (1986). Glycopro-
the sheer weight of the numbers also Sciences, University of Porto, tein antigens of the lymphocyte surface and their
warrants their serious consideration. Porto 4099-003, Portugal purification by antibody affinity chromatography.
3Sir William Dunn School of Pathology, In Handbook of Experimental Immunology (Oxford:
The University of Oxford, Oxford OX1 3RE, Blackwell Scientific Publications), pp. 22.1–22.24.
UK Xu, C., Gagnon, E., Call, M.E., Schnell, J.R., Schwi-
4These authors contributed equally to this eters, C.D., Carman, C.V., Chou, J.J., and
work Wucherpfennig, K.W. (2008). Cell 135, 702–713.
Supplemental Information includes Experimental *Correspondence: simon.davis@ndm.ox. Zech, T., Ejsing, C.S., Gaus, K., de Wet, B., Shev-
Procedures and one figure and can be found with chenko, A., Simons, K., and Harder, T. (2009).
this article online at doi:10.1016/j.cell.2010.08.018. DOI 10.1016/j.cell.2010.08.018 EMBO J. 28, 466–476.

membrane bound in these transfected

Response nonstimulated Jurkat cells. However, we
did not claim that all CD33 cytoplasmic
Multilayered Control of T Cell domains in a given T cell are completely
bound to the plasma membrane. This
would be impossible in a biological sys-
Receptor Phosphorylation tem because binding events always
follow equilibrium conditions, with the
size of the bound versus free fractions
We would like to respond to the Corre- munoreceptor tyrosine-based activation being determined by the ratio of on and
spondence by Fernandes et al., but first motif (ITAM) partitioned into the hydro- off rates. It follows that changes in equilib-
we will summarize the data in our Cell phobic core of the bilayer in a dynamic rium can result in dissociation of CD33CD
paper (Xu et al., 2008). Our study used manner, with substantial movement of from the membrane. Our data showed
a biophysical approach to examine the the two tyrosines and other elements of a structure with substantial mobility in
binding of the CD33 cytoplasmic domain the cytoplasmic domain (Figure 6, Xu which changes in equilibrium, such as
(CD33CD) to the plasma membrane. We et al., 2008). Microscopy data yielded recruitment of the tyrosine kinase Lck,
reported a new FRET assay to examine similar FRET values for CD33CD tagged could enable phosphorylation.
CD33CD binding to the inner leaflet of with C-terminal teal fluorescent protein We wish to point out that we did not
the plasma membrane in live Jurkat cells, (TFP) and for a positive control in which claim that binding of CD33CD to the
a transformed T cell line. Furthermore, we TFP was in close proximity to the plasma membrane was the only mechanism that
developed an approach to determine membrane (there was a three amino acid prevents spontaneous T cell signaling.
the NMR structure of CD33CD bound in linker between the transmembrane and It is well established that there is
a lipid bilayer environment. The structure TFP domains). These data indicated that multilayered control of T cell receptor
showed that the two tyrosines of the im- most of the cytoplasmic domain was (TCR) signaling because complete

Cell 142, September 3, 2010 ª2010 Elsevier Inc. 669

phosphorylation of even one or a few TCRs after 2 min, whereas phosphorylation of assessed whether the negative charge of
is sufficient to result in T cell activation the mutant was delayed, with a 3.5- to the inner leaflet of the plasma membrane
(Bergman et al., 1992; Davis and van der 5-fold reduction at 10 min (Figure S1B). could be affected. Phosphatidylserine is
Merwe, 2006). Without highly effective However, both proteins were phosphory- the most abundant negatively charged
mechanisms to prevent spontaneous lated to a similar degree at a later time lipid in the inner leaflet. We used a highly
signaling, rampant chronic inflammation point (30 min), as Fernandes et al. also specific calcium-independent phosphati-
and autoimmunity would result. Sponta- showed. The greatly reduced interaction dylserine probe, the Lactadherin C2
neous signaling is inhibited by the Csk of the mutant protein with Lck as well as domain (Lact-C2), to study the dynamics
kinase, which phosphorylates Lck kinase the multiple control mechanisms regu- of phosphatidylserine distribution (Yeung
at inhibitory tyrosine 505, as well as lating T cell activation could account for et al., 2008). Jurkat cells were transduced
multiple phosphatases (Bergman et al., the absence of observable phosphoryla- with a lentivirus containing the Lact-C2
1992; Davis and van der Merwe, 2006). In tion of this particular CD33 mutant in non- probe, and the probe’s localization in live
agreement with Fernandes et al., we did stimulated Jurkat cells. It is not known cells was analyzed by confocal micros-
not detect phosphorylation of the why mutation of six basic residues in the copy (Figure S1E). Under resting condi-
CD33CD Emut1+2 mutant protein in Jurkat N-terminal part of CD33CD reduces the tions, the Lact-C2 probe predominantly
cells (Figure S1B, CD33-Mut time 0), which interaction with Lck, but conformational associated with the plasma membrane in
was a negative control for the FRET exper- changes in the cytoplasmic domain or 75% of cells. In contrast, pervanadate
iments. However, if binding of CD33 loss of long-range charge interactions treatment reduced the Lact-C2 signal
to the membrane together with other inhib- with Lck could be involved. at the plasma membrane and increased
itory mechanisms guards against sponta- Pervanadate oxidizes the active site the amount of this phosphatidylserine
neous phosphorylation, this result is not cysteines of many different phosphatases probe in internal membranes in 92%
surprising. Phosphorylation may only be and is generated by mixing hydro- of cells, suggesting that the plasma
favored with appropriate TCR localization gen peroxide with sodium orthovanadate membrane inner leaflet had a reduced
into microclusters that exclude the CD45 (Huyer et al., 1997). Given the potential negative charge (Figures S1E and S1F).
phosphatase (Varma et al., 2006), as well for unanticipated effects, we used our Phosphatidylserine at the cell surface
as simultaneous colocalization of active FRET assay to examine the conse- was also slightly increased, as measured
Lck bound to CD4/CD8 coreceptors close quences of pervanadate treatment on by annexin-V labeling (Figure S1G). We
to the clustered TCR-CD3 complexes. CD33CD membrane binding (Xu et al., conclude that pervanadate treatment
Fernandes et al. assume that the 2008). The assay measures the interac- reduces the negative charge of the inner
CD33CD Emut1+2 mutant protein would tion of a TFP domain attached to the leaflet and induces release of CD33CD
normally interact with Lck. However, the C terminus of CD33CD with a lipophilic from the plasma membrane. These find-
loss of six basic residues substantially dye R18 (octadecyl rhodamine B) incor- ings complicate the interpretation of
changes its charge properties. We porated in the plasma membrane; label- experiments using pervanadate to assess
therefore directly compared the phos- ing the cells with R18 caused a decrease whether CD33CD is bound to the plasma
phorylation of the mutant and wild-type in the TFP signal (quenching) due to membrane.
CD33CD proteins in an in vitro phosphor- energy transfer from TFP to R18 (Fig- A recent study confirmed that CD33CD
ylation assay with the purified cyto- ure S1C). In the absence of pervanadate, binds to the plasma membrane (Deford-
plasmic domains of the wild-type and the FRET efficiency was high (average Watts et al., 2009), and prior work has
Emut1+2 proteins and purified Lck in the 50%; Figure S1D), which is similar to shown that CD3z is membrane bound
absence of liposomes (Figure S1A). We the FRET signal obtained with the positive (Aivazian and Stern, 2000). Furthermore,
observed greatly reduced phosphoryla- control in which TFP is positioned close other cytoplasmic peptides have been
tion of the mutant protein, even though to the plasma membrane (a three amino shown to interact with the inner leaflet
equal amounts of wild-type and mutant acid linker between the transmembrane by similar mechanisms. McLaughlin and
proteins were used. The mutant protein domain and TFP) (Xu et al., 2008). After colleagues showed that the MARCKS
was not detectably phosphorylated with 20 min of pervanadate treatment, the peptide binds to the inner leaflet using
100 ng of Lck kinase and was only phos- FRET signal was reduced on average clusters of basic amino acids and five
phorylated at low levels with 400 ng Lck; to 15% (Figure S1D), similar to the phenylalanine residues (Zhang et al.,
the wild-type protein was robustly phos- negative control with a 50 amino acid flex- 2003). NMR measurements demon-
phorylated under both conditions. Fer- ible linker between the transmembrane strated directly that these phenylalanine
nandes et al. also mention delayed phos- domain and TFP (Xu et al., 2008). Perva- residues are located in the hydrophobic
phorylation kinetics for this mutant nadate treatment thus results in dissocia- core of the lipid bilayer, similar to the tyro-
protein in cells treated with pervanadate, tion of the CD33 cytoplasmic domain from sines of CD33CD (Zhang et al., 2003).
which inhibits a broad range of phospha- the membrane. Also, the cytoplasmic small GTPase Rit
tases (data not shown). We confirmed We next examined the potential mech- localizes to the plasma membrane using
this finding and observed substantially anisms that could contribute to this effect. three clusters of basic amino acids
delayed phosphorylation of the mutant We had previously shown that basic resi- and interspersed hydrophobic residues;
protein after pervanadate treatment: the dues of CD33CD are critical for its binding mutation of a tryptophan in this segment
wild-type protein was phosphorylated to the plasma membrane. Therefore, we results in loss of membrane binding

670 Cell 142, September 3, 2010 ª2010 Elsevier Inc.

(Heo et al., 2006). Therefore, usage of Etienne Gagnon,1 Chenqi Xu,2 Davis, S.J., and van der Merwe, P.A. (2006). Nat.
both basic and hydrophobic residues Wei Yang,2 H. Hamlet Chu,1 Immunol. 7, 803–809.
for plasma membrane binding is a more Matthew E. Call,3 James J. Chou,3
general theme that extends beyond the and Kai W. Wucherpfennig1,4,* Deford-Watts, L.M., Tassin, T.C., Becker, A.M.,
1Department of Cancer Immunology & AIDS, Medeiros, J.J., Albanesi, J.P., Love, P.E., Wulfing,
CD33 and z cytoplasmic domains of the
Dana-Farber Cancer Institute, Boston, C., and van Oers, N.S. (2009). J. Immunol. 183,
TCR-CD3 complex. 1055–1064.
MA 02115, USA
In their Correspondence, Fernandes 2Institute of Biochemistry and Cell Biology,
et al. raise questions about the functional Shanghai Institutes for Biological Sciences, Heo, W.D., Inoue, T., Park, W.S., Kim, M.L., Park,
relevance of the binding of CD33CD to the Chinese Academy of Sciences, B.O., Wandless, T.J., and Meyer, T. (2006).
plasma membrane during TCR activation. Shanghai 200031, China Science 314, 1458–1461.
It will be important to further study the 3Department of Biological Chemistry and

functional significance of the binding of Molecular Pharmacology Huyer, G., Liu, S., Kelly, J., Moffat, J., Payette, P.,
4Department of Neurology and Program in Kennedy, B., Tsaprailis, G., Gresser, M.J., and
CD33CD to the plasma membrane in
a physiological setting using primary Immunology Ramachandran, C. (1997). J. Biol. Chem. 272,
Harvard Medical School, Boston, MA 02115, 843–851.
T cells because the early events in TCR
triggering are very complex, and subtle
*Correspondence: kai_wucherpfennig@dfci. Varma, R., Campi, G., Yokosuka, T., Saito, T., and
changes may result in receptor activation Dustin, M.L. (2006). Immunity 25, 117–127.
after ligand binding. Analysis of multiple DOI 10.1016/j.cell.2010.08.019
CD33CD mutant proteins will be useful, Xu, C., Gagnon, E., Call, M.E., Schnell, J.R., Schwi-
including mutants with reduced rather eters, C.D., Carman, C.V., Chou, J.J., and
than complete loss of membrane binding. REFERENCES Wucherpfennig, K.W. (2008). Cell 135, 702–713.

Aivazian, D., and Stern, L.J. (2000). Nat. Struct. Yeung, T., Gilbert, G.E., Shi, J., Silvius, J., Kapus,
Biol. 7, 1023–1026. A., and Grinstein, S. (2008). Science 319, 210–213.
Supplemental Information includes Experimental Bergman, M., Mustelin, T., Oetken, C., Partanen,
Procedures and one figure and can be found with J., Flint, N.A., Amrein, K.E., Autero, M., Burn, P., Zhang, W., Crocker, E., McLaughlin, S., and Smith,
this article online at doi:10.1016/j.cell.2010.08.019. and Alitalo, K. (1992). EMBO J. 11, 2919–2924. S.O. (2003). J. Biol. Chem. 278, 21459–21466.

Cell 142, September 3, 2010 ª2010 Elsevier Inc. 671

Leading Edge

Fishing Out a Sensor for Anti-inflammatory Oils

Alan R. Saltiel1,*
1Life Sciences Institute, Departments of Internal Medicine and Molecular and Integrative Physiology, University of Michigan Medical School,

Ann Arbor, MI 48109, USA

DOI 10.1016/j.cell.2010.08.022

The u-3 fatty acids have anti-inflammatory and antidiabetic effects in humans. Now, Oh et al. (2010)
demonstrate that the G protein-coupled receptor GPR120 is a receptor for u-3 fatty acids on
macrophages and fat cells. Activation of GPR120 by u-3 fatty acids inhibits multiple inflammation
cascades in macrophages and reverses insulin resistance in obese mice.

Evidence providing an inflammatory link evidence indicates that u-3 fatty acids stream signaling molecules (Rajagopal
between obesity and type 2 diabetes is derived from fish oils, such as docosahex- et al., 2010). In a series of gene silencing
accumulating. In numerous animal and anoic acid (DHA) and eicosapentanoic experiments, Oh and colleagues demon-
clinical studies, obesity is associated acid (EPA), have an anti-inflammatory strate that b-arrestin2 is essential for
with a state of low-grade, chronic inflam- effect (Serhan et al., 2008). the anti-inflammatory effects of u-3 fatty
mation in liver and adipose tissue, To sort out the metabolic impact of acids in macrophage cells, but Gq is
which includes activation of the innate various types of fatty acids, Oh et al. surprisingly dispensable for this process.
immune system and the appearance of characterized the tissue expression pat- Moreover, b-arrestin2 inhibits both the
proinflammatory immune cells (Hotamisli- terns of five G protein-coupled receptors JNK and NF-kB pathways by seques-
gil, 2006; Shoelson and Goldfine, 2009). (GPCRs) known to bind and respond tering the TAK1 binding protein TAB1.
Most notably, macrophages conspire to fatty acids. Among these receptors, The inhibition of TAB1 prevents phos-
with increased levels of inflammatory GPR120 was the only one with an expres- phorylation and thus activation of IkB
cytokines to attenuate insulin action and sion profile that correlated well with a kinase upstream of NFkB and MKK4
increase lipid accumulation. potential role in regulating metabolism. (mitogen-activated protein kinase kinase 4)
Recent studies indicate that the NF-kB They found that GPR120 is highly ex- upstream of JNK.
and JNK (JUN N-terminal kinase) path- pressed in adipose tissue macrophages, These new insights into the sensing and
ways play important roles in the com- fat cells, and specialized macrophages signaling of GPR120 offered Oh and
munication among macrophages, adipo- in the liver called Kupffer cells. Moreover, colleagues a unique opportunity to study
cytes, and liver cells (Arkan et al., 2005; a high-fat diet increases the expression of the molecular mechanisms underlying
Chiang et al., 2009; Solinas et al., 2007). this receptor on macrophages, suggest- the metabolic benefits of u-3 fatty acids
However, key questions still remain about ing that GPR120 might be controlled by in vivo. First, Oh and colleagues establish
the initial establishment of this inflamma- inflammatory signals. that the high-fat diet given to mice in the
tory state. Do fat and liver cells first sense GPR120 is an orphan receptor for laboratory is low in u-3 fatty acids. They
an overload of energy and respond by which no endogenous ligands are known. then show that supplementing this diet
secreting chemokines, which then recruit Using a heterologous reporter system, Oh with DHA and EPA reverses the delete-
macrophages to the liver and fat? Or do et al. now find that the u-3 fatty acids rious effects that the high-fat diet has on
fatty acids in the diet directly initiate the DHA, EPA, and palmitoleate are agonists glucose homeostasis and lipid storage
inflammatory cascade, and, if so, which of GPR120. Furthermore, activation of (i.e., steatosis). Although the authors do
ones? In this issue of Cell, Oh et al. GPR120 by DHA antagonizes the proin- not address whether u-3 fatty acids can
(2010) address the latter question by iden- flammatory effects of TNFa and lipopoly- prevent insulin resistance or glucose intol-
tifying a sensor for u-3 fatty acids that is saccharide in a macrophage cell line. erance in mice, they do demonstrate that
upregulated in obese mice. Furthermore, DHA not only blocks the NFkB and JNK DHA and EPA reverse insulin resistance
activation of this receptor exerts potent pathways but also prevents expression caused by the high-fat diet. Moreover,
anti-inflammatory effects that improve of cytokines (Figure 1, bottom right). disruption of the GPR120 gene abolishes
insulin resistance and other symptoms of GPR120 is known to couple with the the benefits of u-3 fatty acids on glucose
metabolic syndrome in mice. Gq/11 family of G proteins. After ligand binds homeostasis and insulin sensitivity in
Previous studies suggest that saturated and Gq/11 is released, G protein receptor mice. These results demonstrate the
fatty acids promote inflammation by acti- kinases phosphorylate the receptor. This crucial role GPR120 plays in the meta-
vating Toll-like receptor 4 (TLR4) on fat generates binding sites for b-arrestins, bolic benefits of DHA and EPA. Interest-
cells and macrophages (Shi et al., 2006). which mediate internalization and down- ingly, mice receiving bone marrow
In contrast, most unsaturated fats are regulation of the receptors. However, transplants from the GPR120-deficient
metabolically neutral. However, recent b-arrestins can also interact with down- mice are also resistant to the beneficial

672 Cell 142, September 3, 2010 ª2010 Elsevier Inc.

GPR120 expression, and what is its phys-
iological role in adipose and liver tissue?
Also, does GPR120 respond to endoge-
nous ligands to control macrophage
activity? Although the new study by Oh
and colleagues explains how activation
of GPR120 inhibits inflammatory path-
ways, it is still unknown how this receptor
increases the presence of anti-inflamma-
tory M2 macrophages in adipose tissue
and from where these cells arise. Do M2
macrophages (or their precursors)
express GPR120 in order to develop or
maintain the anti-inflammatory pheno-
type, or does activation of GPR120
induce the transdifferentiation of inflam-
matory M1 macrophages to anti-inflam-
matory M2 macrophages in situ?
Finally, previous studies demonstrate
that fish oils have diverse benefits on
multiple tissues. For example, u-3 fatty
acids can inhibit the production of proin-
flammatory eicosanoids and serve as
precursors for resolvins, (i.e., protective
Figure 1. A Model for How u-3 Fatty Acids Preserve Insulin Sensitivity through GPR120 lipids that help reduce inflammation)
A high-fat diet with a disproportionate ratio of saturated fatty acids to u-3 fatty acids triggers activation of
Toll-like receptor 4 (TLR4) in adipocytes and circulating immune cells. This launches an inflammatory
(Serhan et al., 2008). In addition, fish oils
cascade that results in the recruitment of proinflammatory M1 macrophages, increased secretion of help prevent cardiovascular disease and
TNFa, and insulin resistance in adipocytes. The addition of u-3 fatty acids to the diet activates the have positive effects on many inflamma-
G protein-coupled receptor GPR120 on proinflammatory M1 macrophages (Oh et al., 2010), which in
tory disorders, such as arthritis, asthma,
turn attenuates the inflammatory response and recruits anti-inflammatory M2 macrophages to adipose
tissue. Eventually, these M2 macrophages restore secretion of interleukin-10 and improve insulin and ulcerative colitis. Is GPR120 the only
sensitivity. receptor responsible for these various
benefits? Future studies are also needed
to determine whether dietary supple-
properties of DHA and EPA. Thus, the u-3 These cells produce cytokines, such as ments and ingestion of fatty fish can
fatty acids appear to act primarily through TNFa, which further activate the macro- provide high enough concentrations of
macrophages. phages and attenuate insulin action in circulating u-3 fatty acids to promote
Taken together with previous data, adipocytes. Eventually, this leads to GPR120 activation. Nevertheless, the
these findings from Oh and colleagues local and then systemic insulin resistance. new insights presented by Oh and
support a model in which dietary fatty However, these activated M1 macro- colleagues into the anti-inflammatory
acids control the inflammatory properties phages also express elevated levels of mechanisms of u-3 fatty acids provide
of macrophages in adipose tissue by GPR120. Thus, addition of u-3 fatty acids a platform for investigating these impor-
regulating the activity and expression to the diet stimulates GPR120 and tant questions. Plus, the identification of
of opposing receptors (Figure 1). With initiates a signaling pathway through GPR120 pinpoints a new therapeutic
a normal diet containing a balanced b-arrestin2, which blocks the effects of target for treating the inflammatory state
ratio of saturated and u-3 unsaturated TLR4 and inflammatory cytokine recep- associated with obesity and type 2 dia-
fatty acids, anti-inflammatory M2 macro- tors. This reduces the inflammatory state betes. This alone is cause for a bit of
phages protect adipose cells by damp- of these cells and simultaneously pro- excitement.
ening excess inflammation and maintain- motes the return of anti-inflammatory
ing insulin sensitivity of fat cells (Lumeng M2 macrophages to adipose tissue, REFERENCES
et al., 2007). When mice are given which leads to the restoration of insulin
a high-fat diet with excess calories sensitivity. Arkan, M.C., Hevener, A.L., Greten, F.R., Maeda,
S., Li, Z.W., Long, J.M., Wynshaw-Boris, A., Poli,
and little u-3 fatty acids, TLR4 is left This model raises many new questions
G., Olefsky, J., and Karin, M. (2005). Nat. Med.
unchecked in fat cells. The activated about the crosstalk between adipose 11, 191–198.
receptor induces expression and release tissue and immune cells in obesity. First, Chiang, S.H., Bazuine, M., Lumeng, C.N., Geletka,
of chemokines, such as MCP-1 (mono- could GRP120 activation serve as L.M., Mowers, J., White, N.M., Ma, J.T., Zhou, J.,
cyte chemotactic protein-1), which then a homeostatic mechanism in macro- Qi, N., Westcott, D., et al. (2009). Cell 138,
recruit proinflammatory M1 macrophages phages to resolve inflammation in obese 961–975.
into adipose tissue (Lumeng et al., 2007). animals? If so, then what controls Hotamisligil, G.S. (2006). Nature 444, 860–867.

Cell 142, September 3, 2010 ª2010 Elsevier Inc. 673

Lumeng, C.N., Bodzin, J.L., and Saltiel, A.R. R.J. (2010). Proc. Natl. Acad. Sci. USA 107, Shoelson, S.E., and Goldfine, A.B. (2009). Nat.
(2007). J. Clin. Invest. 117, 175–184. 628–632. Med. 15, 373–374.
Oh, D.Y., Talukdar, S., Bae, E.J., Imamura, T.,
Serhan, C.N., Chiang, N., and Van Dyke, T.E.
Morinaga, H., Fan, W., Li, P., Lu, W.J., Watkins, Solinas, G., Vilcu, C., Neels, J.G., Bandyopad-
(2008). Nat. Rev. Immunol. 8, 349–361.
S.M., and Olefsky, J.W. (2010). Cell 142, this issue, hyay, G.K., Luo, J.L., Naugler, W., Grivennikov,
687–698. Shi, H., Kokoeva, M.V., Inouye, K., Tzameli, I., Yin, S., Wynshaw-Boris, A., Scadeng, M., Olefsky,
Rajagopal, S., Kim, J., Ahn, S., Craig, S., Lam, H., and Flier, J.S. (2006). J. Clin. Invest. 116, 3015– J.M., and Karin, M. (2007). Cell Metab. 6,
C.M., Gerard, N.P., Gerard, C., and Lefkowitz, 3025. 386–397.

A New Spin on Planar Cell Polarity

Patricio Olguin1 and Marek Mlodzik1,*
1Department of Developmental and Regenerative Biology, Mount Sinai School of Medicine, New York, NY 10029, USA

DOI 10.1016/j.cell.2010.08.025

The generation of planar cell polarity (PCP) and tissue shape during morphogenesis is tightly linked,
but it is not clear how. Aigouy et al. (2010) now show in the developing Drosophila wing that PCP
initially has a radial orientation that becomes realigned to the proximal-distal axis of organ shape
by mechanical forces and cell rearrangements mediated by Dachsous.

Most tissues and organs composed of and tion). The core Frizzled/PCP factors form Vang/Stbm signaling as the nonautono-
organized as epithelial cell layers display, two distinct complexes that become mous behavior of frizzled mutant cell
in addition to the common apical-basal localized asymmetrically to either the patches (clones) affecting the polarity of
epithelial polarity, a polarity within the proximal (Vang/Stbm and associated wild-type cells flanking the frizzled mutant
epithelial plane. This is commonly referred proteins) or the distal side of pupal wing cells (Vinson and Adler, 1987) is already
to as planar cell polarity (PCP). Genetic cells (Frizzled and associated proteins). observed at this stage. Strikingly, in
studies in the fruit fly Drosophila mela- These complexes are stabilized by feed- contrast to the nonautonomous effects
nogaster have established that there are back loop interactions among themselves observed at the distal side of frizzled
two molecular systems coordinating the (Seifert and Mlodzik, 2007; Strutt, 2003). In mutant cell patches in late pupal and adult
cellular asymmetries in the plane of addition, Frizzled and Vang/Stbm protein wings (Vinson and Adler, 1987), early friz-
tissues. These include the Frizzled/PCP complexes may be required earlier to zled clones influence the polarity of wild-
signaling pathway containing the Van- coordinate global tissue polarity/PCP type cells residing between the wing
Gogh (Vang, also known as Strabismus/ within the wing epithelium (Classen et al., margin and the clone within the radial
Stbm) protein and other factors (Strutt, 2005; Wu and Mlodzik, 2009). polarity axis. This confirms a ‘‘signaling
2003; Seifert and Mlodzik, 2007) and Supporting this idea, the new study by axis’’ toward the wing margin at early
a pathway mediated by the protocadher- Aigouy et al. (2010) in this issue of Cell stages. Taken together, the observations
ins Fat and Dachsous (Lawrence et al., helps to establish that subcellular asym- of Aigouy et al. (2010) indicate that (1)
2007). Although the molecular relation- metries among the Frizzled/PCP core PCP, mediated by Frizzled-Vang/Stbm
ships between these two systems are group proteins are already present at signaling, is established during late-
unclear, there is strong evidence that early pupal stages during wing develop- larval/early-pupal stages in a radial axis
they act in parallel, probably affect ment (14–15 hr after puparium formation perpendicular to the margin and (2) the
different effectors, and may compensate or earlier). Strikingly, the Frizzled/PCP polarity/PCP seen in the adult wing is
for each other in some tissues (Lawrence complexes display radial polarity that is a result of cellular rearrangements during
et al., 2007; Wu and Mlodzik, 2009). perpendicular to the wing margin wing morphogenesis within the prox-
In the developing Drosophila wing, (Figure 1A), confirming that coordination imal-distal axis that are dependent on
cellular asymmetries in the plane of the of global Frizzled/PCP signaling is estab- Dachsous.
epithelium are first detected at later pupal lished early in pupal fly wings. The authors How is polarity realigned along the
stages along the proximal-distal axis further demonstrate that these early proximal-distal axis as morphogenesis
(at around 24–30 hr after puparium forma- asymmetries indeed depend on Frizzled- proceeds? As PCP is already established

674 Cell 142, September 3, 2010 ª2010 Elsevier Inc.

at early stages, its final align- cues’’ that orient the initial
ment from the radial orienta- polarity of Frizzled/PCP
tion to the proximal-distal complexes, but this hypoth-
axis must be redirected esis has been challenged by
through active relocalization strong genetic evidence
of Frizzled and Vang/Stbm showing that the Fat/Dachs-
complexes and/or through ous and Frizzled/PCP
the shifting or moving of the systems act in parallel (Law-
cells as a whole. For rence et al., 2007). During fly
example, polarity could be larval stages, the Fat/Dachs-
achieved by rotation of the ous system is required to
cells toward the distal axis regulate the growth and
as happens with the rotation shape of the wing, the latter
of photoreceptor cell clusters (at least in part) by orienting
(ommatidia) in the Drosophila the axis of cell division
eye toward the anterior- perpendicular to the margin
posterior axis (Seifert and (Baena-López et al., 2005).
Mlodzik, 2007). But how Aigouy et al. (2010) show
would such a rotation be that high expression of
regulated? At early stages of Dachsous in the hinge region
pupal development, the is not required for its contrac-
proximal half of the wing tion, but correct Dachsous
epithelium (hinge) and the levels in the wing blade are
wing blade are similar in size Figure 1. Planar Cell Polarity in the Fly Wing required for the wing blade
(A) During fly pupal development, the initial axis of planar cell polarity (PCP) is
(Figure 1A). Subsequently, radial, that is, oriented toward the wing margin (black arrows).
to respond to the anisotropic
preceding and coinciding (B) As development proceeds, the hinge region (blue) contracts creating an aniso- mechanical stress that
with PCP realignment, the tropic mechanical stress on the wing blade, resulting in movements of wing cells orients cell elongation along
and realignment of PCP to the proximal-distal axis. Green arrows indicate the
hinge contracts and gener- the proximal-distal axis.
direction of cell movement, and red arrows show the direction of cellular rotations.
ates an anisotropic mechan- These processes take place between 14 and 24 hr after puparium formation. Moreover, cell polarity
ical stress on the blade, (C) The final orientation of PCP is in the proximal-distal axis in late pupal/adult fly defects correlate with the
which leads to its elongation wings (black arrows). inversion of local tissue rota-
in the proximal-distal axis tion in wild-type wings, with
(Figure 1B). Strikingly, epithelial cell elon- a proximal-distal axis within the wing (Fig- altered levels of Dachsous in the posterior
gation could drive the realignment of ure 1C). Interestingly, during this remodel- compartment, suggesting that Dachsous
cortical microtubules with the proximal- ing process, the global coordination/long- imbues cells with the ability to respond
distal axis, which appears essential for range coherence of PCP is diminished, coordinately to mechanical stress. Inter-
the delivery of Frizzled to the distal side which may be the reason why previous estingly, Harumoto et al. (2010) show
of cells (Shimada et al., 2006). New work studies of PCP during development have that Dachsous and Fat are required to
appearing in Developmental Cell by missed the early asymmetry/polarity of align cortical microtubules along the prox-
Uemura and colleagues (Harumoto et al., PCP core proteins. As pupal wing cells re- imal-distal axis and that Dachsous biases
2010) shows that, prior to hinge contrac- pack as they acquire a hexagonal shape, the direction of microtubule growth from
tion, cortical microtubules align perpen- the global coordination/long-range coher- high to low Dachsous levels, similar to its
dicular to the margin in the proximal ence of PCP increases again. This role in cell orientation. Whether the Fat/
region of the wing blade. This supports phenomenon is a consequence of the Dachsous system regulates cell remodel-
a general role for the orientation of cortical persistence of the Vang/Stbm and Frizzled ing by controlling the polarity of cortical
microtubules in PCP and in the realign- complexes at boundaries formed in the microtubules or vice versa remains to be
ment of PCP later in development. early stages of development and of the resolved.
Quantitative analyses of time-lapse proximal-distal alignment of new bound- In their elegant new study, Aigouy et al.
imaging of the pupal wing by Aigouy et al. aries. On the other hand, the Frizzled/ (2010) analyzed the timeline of events for
(2010) show that, in response to the aniso- PCP core factors are required for hexag- the establishment of PCP in the devel-
tropic stress, cells move with respect to onal cell packing, probably by polarizing oping Drosophila wing. They have
each other in a proximal direction and membrane trafficking along the proximal- provided evidence that the early Friz-
inwards with different velocities (Figure 1B). distal axis (Classen et al., 2005). Thus, zled/PCP core polarization toward the
This behavior causes shear and the local both early polarity and cellular packing wing margin (in a radial orientation) is real-
rotation of cells, mainly clockwise in the would feed in to one another to shape igned along the proximal-distal axis by
anterior and anticlockwise in the posterior and repolarize the epithelia. anisotropic mechanical stress and
half of the blade (Figure 1B). As a conse- It has been proposed that the Fat/ Dachsous-mediated tissue remodeling.
quence, PCP is reoriented from a radial to Dachsous system would provide ‘‘global These conclusions are consistent with,

Cell 142, September 3, 2010 ª2010 Elsevier Inc. 675

and supported by, the phenotypic PCP relation to new mechanisms that sculpt Lawrence, P.A., Struhl, G., and Casal, J. (2007).
features of Frizzled/PCP core group the shape of organs in general. Nat. Rev. Genet. 8, 555–563.
genes on one side and that of the Fat/
Dachsous system on the other. Flies REFERENCES Seifert, J.R., and Mlodzik, M. (2007). Nat. Rev.
Genet. 8, 126–138.
carrying mutations in Frizzled/PCP core
proteins exhibit defects in PCP Aigouy, B., Farhadifar, R., Staple, D.B., Sagner, A.,
Röper, J., Jülicher, F., and Eaton, S. (2010). Cell Shimada, Y., Yonemura, S., Ohkura, H., Strutt, D.,
throughout the wing. In contrast, the
142, this issue, 773–786. and Uemura, T. (2006). Dev. Cell 10, 209–222.
Fat/Dachsous system mainly affects
Baena-López, L.A., Baonza, A., and Garcı́a-Bel-
polarity in the proximal half of the wing, Strutt, D. (2003). Development 130, 4501–4513.
lido, A. (2005). Curr. Biol. 15, 1640–1644.
as this area strongly depends on cellular
Classen, A.K., Anderson, K.I., Marois, E., and Ea-
realignment and rotation during the ton, S. (2005). Dev. Cell 9, 805–817. Vinson, C.R., and Adler, P.N. (1987). Nature 329,
switch to the proximal-distal PCP axis. 549–551.
Harumoto, T., Ito, M., Shimada, Y., Kobayashi,
Together, these observations provide an T.J., Ueda, H.R., Lu, B., and Uemura, T. (2010).
exciting new framework for under- Dev. Cell 19. Published online September 2, Wu, J., and Mlodzik, M. (2009). Trends Cell Biol. 19,
standing the generation of PCP and its 2010. 10.1016/j.devcel.2010.08.004. 295–305.

Viable Rat-Mouse Chimeras:

Where Do We Go from Here?
Davor Solter1,*
of Medical Biology, A*STAR, 138648 Singapore, Republic of Singapore
DOI 10.1016/j.cell.2010.08.021

In a tour-de-force study, Kobayashi et al. (2010) describe the first viable rat-mouse chimeras and
demonstrate that rat induced pluripotent stem (iPS) cells can rescue organ deficiency in mice.
Rat iPS cells formed a fully functional pancreas when injected into mouse blastocysts lacking the
Pdx1 gene required for pancreas formation.

Experimentally produced chimeras for the mouse and capitalizing on the the geep between a sheep (Ovis aries)
between different mouse strains (Tarkow- recent isolation of rat ES and iPS cells and a goat (Capra hircus) (Fehilly et al.,
ski, 1961) have been an exceedingly (Buehr et al., 2008; Li et al., 2008). Both 1984).
useful tool for developmental biologists, mouse and rat cells were tagged with To test the possibility that viable rat-
contributing to our understanding of the different fluorescent dyes, allowing the mouse chimeras could be formed,
establishment of cell lineages, cell deter- authors to follow their distribution in the Kobayashi et al. (2010) injected fluores-
mination, and the development of the developing chimeras. The authors wanted cently labeled mouse or rat iPS cells into
immune system and other organs. In this to prove, at least in principle, that xenoge- rat or mouse blastocysts, respectively,
issue of Cell, Kobayashi et al. (2010) neic organ complementation could be and returned them to blastocyst-compat-
dramatically extend the potential of mam- achieved, that is, that donor cells of one ible pseudopregnant females (that is,
malian chimeras with their report of viable species could rescue a defect in organ foster mothers of the same species as
rat-mouse chimeras that can develop to development in a recipient of a different the blastocysts). The authors then exam-
term and become fully functional adults. species. So, as a first step, they set out ined the resulting fetuses, newborns, and
In their study, Kobayashi and col- to produce viable chimeras between rats adults and found evidence of a substantial
leagues relied on previous knowledge and mice, even though many previous contribution of donor stem cells to tis-
but also added a few new wrinkles. They efforts to make such chimeras had failed. sues and organs of the host (Figure 1A).
first derived mouse and rat embryonic The only viable intergeneric chimera— Despite a big contribution of donor cells,
stem (ES) cells and induced pluripotent that is, a hybrid between animals from the size of newborn and adult chimeras
stem (iPS) cells using standard methods different genera—reported so far is (with one exception) was determined by

676 Cell 142, September 3, 2010 ª2010 Elsevier Inc.

Figure 1. Generating Rat-Mouse Chimeras
(A) Induced pluripotent stem (iPS) cells were derived from adult mouse and rat cells and were labeled with different fluorescent proteins. Rat (blue) and mouse
(red) iPS cells were injected into reciprocal blastocysts (mouse into rat and vice versa) to produce intergeneric chimeras. From these blastocysts, several
chimeras were born and some survived to adulthood. The contribution of injected donor stem cells was observed throughout the body of the host. The size
and morphology of the newborn and adult chimeras was determined by the host blastocyst.
(B) Fluorescently labeled rat iPS cells (blue) were injected into normal mouse blastocysts (left) or blastocysts lacking the Pdx1 gene (right), which encodes the
transcription factor pancreatic and duodenal homeobox 1 that is required for pancreas development. Chimeras derived from normal or Pdx1-deficient mouse
blastocysts showed an extensive contribution of rat cells to all tissues. However, in the Pdx1-deficient chimeras, the entire pancreas was derived from donor
rat cells (inset, blue) and was fully functional, including production of insulin by b islet cells.

the species of the host blastocyst. It is not produced by combining rat stem cells are injected into tetraploid blastocysts)
clear whether it is the embryo itself or the and mouse blastocysts, the resulting to produce rat-mouse chimeras (Nagy
uterine environment that determines the ‘‘mouse-like’’ chimeras had a gall bladder et al., 1993). Tetraploid blastocyst cells
extent of chimera growth. To distinguish despite the significant contribution of rat cannot participate in formation of the
between these possibilities, one would cells to abdominal organs. Reciprocal embryo proper; thus, the resulting fetus
have to transfer chimeric embryos into chimeras were ‘‘rat-like’’ and, again, (and adult) is derived entirely from the
the uterus of pseudopregnant females of despite a significant contribution from injected cells, whereas the placenta and
the same species as the donor stem cells mouse cells to abdominal organs, the extraembryonic membranes are derived
(not the blastocysts). Previous studies gall bladder was absent. These results from the tetraploid blastocyst. It remains
suggest that such experiments would fail suggest that cells of the blastocyst inner to be seen whether this approach could
because of the need for compatibility cell mass possess a ‘‘morphogenetic’’ produce a fetus derived entirely from
between the fetal part of the placenta and capacity that controls the behavior of mouse ES cells after their injection into
the uterus (Rossant et al., 1982). injected stem cells at all developmental a rat tetraploid blastocyst that then
Besides controlling the size and growth stages. develops in the uterus of a pseudopreg-
of the chimera, the host blastocyst seems This may explain why Kobayashi et al. nant rat female.
to impose additional morphogenetic reg- were able to successfully inject rat stem A major goal of the Kobayashi et al.
ulation. The postimplantation develop- cells into mouse blastocysts, whereas study was to determine whether stem cells
ment of normal rat and mouse embryos insertion of the rat inner cell mass into from a xenogeneic donor mammal could
is very similar, but there are differences the mouse blastocyst cavity did not result correct a genetic defect in a recipient
in organ morphogenesis. One of the in viable rat-mouse chimeras (Gardner mammal of a different species. So, in their
most noticeable differences is the pres- and Johnson, 1973). This notion could next set of experiments, the authors in-
ence of a gall bladder in mice and its be tested further using tetraploid comple- jected rat iPS cells into recipient mouse
absence in rats. In all adult chimeras mentation (that is, donor ES or iPS cells blastocysts that lacked the Pdx1 gene,

Cell 142, September 3, 2010 ª2010 Elsevier Inc. 677

which encodes a transcription factor technical challenges. For example, the potential of human stem cells. Yet such
(pancreatic and duodenal homeobox 1) mouse and rat are developmentally very experiments will be complicated, time
that is essential for development of the similar (apart from size), but it is not clear consuming, difficult to interpret, and, I
pancreas and formation of insulin- that chimeras between animals belonging suspect, will never become part of the
producing b islet cells. Although they to different phylogenetic families or orders standard protocols regulating the medical
observed a substantial contribution of rat would be viable. Indeed, the only attempts use of human stem cells. Although xeno-
cells to different organs and tissues, most to make such chimeras (between a mouse geneic organ complementation is unlikely
importantly, the pancreas of the rat-mouse and a bank vole) have failed (Mystkowska, to be a viable strategy for regenerative
chimeras was composed exclusively of rat 1975). In this experiment, the mouse-bank medicine, the elegant work of Kobayashi
cells (Figure 1B). Thus, cells derived from vole chimeras were made by aggregation et al. is a boon for researchers seeking
rat iPS cells were able to completely of embryos; it is possible that injection of to better understand the biology of stem
rescue the genetic deficiency of the host bank vole stem cells into mouse blasto- cells and mammalian development.
mouse blastocyst. These rat-mouse cysts, followed by their development
chimeras developed into adult animals in the uterus of mouse foster mothers
with a normal functional pancreas, demon- might yield positive results. Successful REFERENCES
strating that xenogeneic organ comple- chimerism between members of different
mentation is achievable. This is a remark- orders (the pig and human, for example) Buehr, M., Meek, S., Blair, K., Yang, J., Ure, J.,
Silva, J., McLay, R., Hall, J., Ying, Q.-L., and Smith,
able accomplishment. seems very unlikely, and attempts to
A. (2008). Cell 135, 1287–1298.
So where do we go from here? Although produce early postimplantation human-
human ES and iPS cells offer hope for mouse chimeras have not been encour- Fehilly, C.B., Willadsen, S.M., and Tucker, E.M.
(1984). Nature 307, 634–636.
tissue and cell replacement therapies in aging (James et al., 2006). Even if we
the not too distant future, the replacement succeed in developing organ-deficient Gardner, R.L., and Johnson, M.H. (1973). Nat. New
of complex organs—lung, kidney, liver, pigs by genetic manipulation and pro- Biol. 246, 86–89.

gut, and, of course, pancreas—is likely to ducing pig-human chimeras in which the James, D., Noggle, S.A., Swigut, T., and Brivanlou,
be much more difficult. Several strategies parenchymal cells of the specific organ A.H. (2006). Dev. Biol. 295, 90–102.
for organ replacement are being tested. are entirely derived from human cells, Kobayashi, T., Yamaguchi, T., Hamanaka, S.,
These include the growth of organs immune rejection will still be a problem Kato-Itoh, M., Yamazaki, Y., Ibata, M., Sato, H.,
in vitro with mixtures of different stem cells because the human organ carried by the Lee, Y.-S.,, Usui, J.-i., Knisely, A.S., et al. (2010).
Cell 142, this issue, 787–799.
and biocompatible scaffolds or the gener- pig will contain pig-derived stromal cells
ation of ‘‘humanized’’ pigs engineered to and blood vessels. Li, P., Tong, C., Mehrian-Shai, R., Jia, L., Wu, N.,
lack certain antigens so that their organs Finally, there are huge legal and ethi- Yan, Y., Maxson, R.E., Schulze, E.N., Song, H.,
Hsieh, C.-L., et al. (2008). Cell 135, 1299–1310.
can be used for transplantation in human cal barriers to creating human-animal
patients with a reduced chance of immune chimeras and, indeed, their production is Mystkowska, E.T. (1975). J. Embryol. Exp. Mor-
rejection. Could production of human forbidden in most countries. However, it phol. 33, 731–744.

organs in, for example, human-pig is possible that injecting human ES or Nagy, A., Rossant, J., Nagy, R., Abramow-
chimeras be an alternative approach? iPS cells into a mouse blastocyst and al- Newerly, W., and Roder, J.C. (1993). Proc. Natl.
Acad. Sci. USA 90, 8424–8428.
Although production of viable rat-mouse lowing limited (early postimplantation)
chimeras could be viewed as a first step development of human-mouse chimeras Rossant, J., Mauro, V.M., and Croy, B.A. (1982).
in this direction, as Kobayashi et al. would be approved for the specific J. Embryol. Exp. Morphol. 69, 141–149.
propose, there are huge biological and purpose of testing the differentiation Tarkowski, A.K. (1961). Nature 190, 857–860.

678 Cell 142, September 3, 2010 ª2010 Elsevier Inc.

Leading Edge

‘Fore Brain:
A Hint of the Ancestral Cortex
Lora B. Sweeney1,2,3 and Liqun Luo1,2,3,*
1Neurosciences Program
2Department of Biology
3Howard Hughes Medical Institute

Stanford University, Stanford, CA 94305, USA

DOI 10.1016/j.cell.2010.08.024

By combining gene expression profiling with image registration, Tomer et al. (2010) find that the
mushroom body of the segmented worm Platynereis dumerilii shares many features with the
mammalian cerebral cortex. The authors propose that the mushroom body and cortex evolved
from the same structure in the common ancestor of vertebrates and invertebrates.

The mammalian cerebral cortex underlies identify structures in the developing brain tion factors such as Bf1 (brain factor 1,
many higher-order processes, such as of Platynereis that are possibly related also known as Foxg1) and Pax 6 (paired
perception, memory, language, and ad- to the vertebrate pallium, Tomer et al. box gene 6) (Hébert and Fishell, 2008).
vanced motor skills. With its intricate characterized the expression patterns of With this technique, Tomer et al. iden-
furrows and ridges (i.e., the sulci and many genes at different stages of neural tify a structure in Platynereis with gene
gyri), the complexity of the cerebral cortex development in the worm. The standard patterns that mirror those observed in
is evident even at its surface. Beneath the methods for characterizing gene expres- the vertebrate pallium. The authors then
surface, the cerebral cortex is separated sion are in situ hybridizations and immuno- follow this structure during the develop-
into layers of densely packed neurons staining with antibodies. However, these ment of Platynereis and find that its
with axons reaching deep into the white techniques are generally restricted to neurons develop into the mushroom
matter. These layers are divided further one or two genes at a time and thus are body, a brain region in insects and worms
into functional regions that correspond to unable to provide direct comparisons of involved in sensory processing and
the body plan. How does such a complex expression patterns for a large number of memory (Figure 1). Moreover, this embry-
structure develop such precise organiza- genes. onic structure in the worm gives rise
tion? Many researchers have approached To overcome this technical hurdle, to similar types of neurons as found
this question from a developmental Tomer et al. use advanced image registra- in the pallium (e.g., glutamatergic and
perspective by identifying and perturbing tion methods (including linear transforma- GABAergic neurons).
molecular components that contribute tion and nonlinear warping), in which Similarities between the mammalian
to the structure of the cerebral cortex multiple microscopy images are aligned cortex and the invertebrate mushroom
(Hébert and Fishell, 2008). Alternatively, to one coordinate system, standardized body have been noted previously, but
one can take an evolutionary approach to one size, and smoothed to correct arti- such clear parallels in their gene pattern-
and search for an ancestral precursor to facts due to stretches during sample ing have not been observed (Strausfeld
the cortex. In this issue of Cell, Tomer preparation (Rueckert et al., 1999; Rohlf- et al., 1998; Farris, 2008). Why not? Mush-
et al. (2010) follow the latter approach ing et al., 2001; Kurylas et al., 2008). room bodies are well characterized in in-
and identify a brain region of the seg- Specifically, Tomer et al. use the axon sects, especially in the fruit fly Drosophila,
mented worm Platynereis dumerilii (an scaffold (i.e., bundles of axons in the and choosing to study an annelid was
annelid) called the mushroom body that developing brain) as a landmark to align critical to making this new connection.
shares the same ‘‘molecular fingerprint’’ individual gene expression patterns from Based on the arrangement of the axon
as the developing mammalian cortex. multiple worms onto a standard brain fibers and the function of its neurons, the
The cerebral cortex derives from the template. This allows the simultaneous mushroom body of insects has been
pallium. From the Latin for ‘‘cloak,’’ mapping of expression profiles for an loosely compared to the mammalian
pallium refers to the outer layer of the unlimited number of genes during neural cerebral cortex, the hippocampus, or the
brain. The mammalian palium consists of development. In this manner, the authors cerebellum (Strausfeld et al., 1998).
the cerebral cortex, olfactory cortex, and define the spatial relationship for the However, insects are fast-evolving inver-
the hippocampus. The evolutionary origin expression of more than fifteen genes in tebrates and thus may deviate quickly
of the vertebrate pallium has fascinated Platynereis that were previously shown from their ancestors. By contrast, anne-
biologists for centuries because our high- to regulate the patterning of the cortex in lids evolve slowly and therefore make
est mental functions originate from it. To mammals. These include many transcrip- excellent organisms for comparing the

Cell 142, September 3, 2010 ª2010 Elsevier Inc. 679

the vertebrate pallium and invertebrate
mushroom body evolved from the same
structure in the last common ancestor
of these organisms (Figure 1). Moreover,
this structural precursor may also have
processed sensory information, as the
cortex and mushroom body do.
That said, however, the vertebrate
pallium and invertebrate mushroom body
are considerably different. First, they have
different shapes (Figure 1). This may result
in part from considerable differences
in neural developmental programs in
vertebrates and invertebrates. Second,
it is unclear whether the mushroom body
exhibits the same type of layering and
functional specialization as the cortex.
Third, gene expression patterns appear
to remain constant over time during
development of the annelid mushroom
body, but the patterns are quite dynamic
during the development of the mamma-
lian cortex. This temporal variation com-
plicates the interpretation of the homolo-
Figure 1. A Common Origin for the Mushroom Body and Cerebral Cortex? gous gene expression patterns detected
In mammals, the pallium develops into the cerebral cortex, olfactory cortex, and hippocampus. Gene for the two structures. Finally, the expres-
expression profiling combined with image registration in the developing brain of the segmented worm
Platynereis dumerilii identified a brain region with gene expression patterns similar to that of the mouse sion profiles are not identical for the
cortex (Tomer et al., 2010). This region in the worm develops into the mushroom body, which is involved two structures. For example, expression
in sensory processing and memory formation in insects and likely worms. These results suggest that the of Wnt3a was present in the vertebrate
mushroom body and the cerebral cortex evolved from the same structure in a common ancestor of verte-
brates and annelids 600 million years ago (mya). The estimates of divergence follow Peterson et al.
pallium but not in the developing
(2004). mushroom body. Such differences are
probably not surprising, given that verte-
evolution of invertebrates with that of ing olfactory inputs and for memory brates and annelids diverged from a
vertebrates. For example, an analysis formation (Heisenberg, 2003). Indeed, common precursor 600 million years
of intron abundance across the entire Tomer and colleagues found that a subset ago (Figure 1).
genomes of vertebrates, annelids, and of genes that share similar expression In summary, the data presented by
insects found that two-thirds of the patterns in the annelid mushroom body Tomer and colleagues provide a tanta-
human introns were also present in and the mammalian cortex are also lizing hint that one of the highest-order
annelids but were largely absent from expressed in the Drosophila mushroom processing centers of the human brain
insects, confirming that insects have body. shares an evolutionary origin with the
evolved faster than annelids (Raible Two distinct mechanisms can explain mushroom body of worms and insects.
et al., 2005). Moreover, annelids are ex- the molecular similarities of the inverte- In principle, the techniques presented
tremely amenable to experimental anal- brate mushroom body and the vertebrate by the authors for profiling expression
ysis because they are easy to breed and cerebral cortex. They could arise inde- patterns of numerous genes simulta-
observe in vivo. Further, they develop pendently through convergent evolution, neously can be used in other organisms
with highly stereotyped structures that or they could evolve from the same to garner further support for this hypoth-
vary little between individuals. structure in a common ancestor. Tomer esis. Furthermore, comparative struc-
Given the similarity in the gene expres- and colleagues present a statistical anal- ture-function analyses may also provide
sion patterns of the developing annelid ysis indicating that the similarities in additional insights into the function of
mushroom body and the mammalian gene expression that they observed are this ancestral structure, including how its
cerebral cortex, do these structures unlikely to occur by convergent evolution. form has adapted over the last 600 million
share similar functions? The cerebral Although these arguments suffer from years.
cortex transforms sensory information the caveat that expression patterns of
into action. The mushroom body of anne- different genes may not have evolved REFERENCES
lids also processes sensory information, independently, the number of genes
receiving input from chemosensory cells studied by Tomer and colleagues do Farris, S.M. (2008). Brain Behav. Evol. 72, 106–122.
on their mouthparts. In insects, the analo- indeed favor the possibility of a common Hébert, J.M., and Fishell, G. (2008). Nat. Rev.
gous structure is best known for process- origin. Together, their data suggest that Neurosci. 9, 678–685.

680 Cell 142, September 3, 2010 ª2010 Elsevier Inc.

Heisenberg, M. (2003). Nat. Rev. Neurosci. 4, Raible, F., Tessmar-Raible, K., Osoegawa, K., Rueckert, D., Sonoda, L.I., Hayes, C., Hill, D.L.,
266–275. Wincker, P., Jubin, C., Balavoine, G., Ferrier, D., Leach, M.O., and Hawkes, D.J. (1999). IEEE Trans.
Benes, V., de Jong, P., Weissenbach, J., et al. Med. Imaging 18, 712–721.
Kurylas, A.E., Rohlfing, T., Krofczik, S., Jenett, A., (2005). Science 310, 1325–1326.
and Homberg, U. (2008). Cell Tissue Res. 333,
Rohlfing, T., Brandt, R., Maurer, C.R., Jr., and Strausfeld, N.J., Hansen, L., Li, Y., Gomez, R.S.,
Menzel, R. (2001). IEEE Workshop on Mathemat- and Ito, K. (1998). Learn. Mem. 5, 11–37.
Peterson, K.J., Lyons, J.B., Nowak, K.S., Takacs, ical Methods in Biomedical Image Analysis,
C.M., Wargo, M.J., and McPeek, M.A. (2004). L. Staib, ed. Kauai, HI, IEEE Computer Society, Tomer, R., Denes, A.S., Tessmar-Raible, K., and
Proc. Natl. Acad. Sci. USA 101, 6536–6541. Los Alamitos, CA, pp. 187–194. Arendt, D. (2010). Cell 142, this issue, 800–809.

Cell 142, September 3, 2010 ª2010 Elsevier Inc. 681

Leading Edge

The Language of Histone Crosstalk

Jung-Shin Lee,1 Edwin Smith,1 and Ali Shilatifard1,*
1Stowers Institute for Medical Research, 1000 East 50th Street, Kansas City, MO 64110, USA

DOI 10.1016/j.cell.2010.08.011

It has been suggested that a specific pattern of histone posttranslational modifications and their
crosstalk may constitute a code that determines transcriptional outcomes. However, recent studies
indicate that histone modifications have context-dependent effects, making their interplay more
like a language within the chromatin signaling pathway than a code.

For almost two decades, a primary focus in the field of transcrip- phosphorylation of serine 10 on histone H3 stimulates the ability
tional regulation was to identify DNA elements that control the of Gcn5 to acetylate histone H3 at lysine 14 (H3K14) (Cheung
expression of genes. These efforts were in part motivated by the et al., 2000; Lo et al., 2000) (Figure 1A). Another well-character-
expectation that it would one day be possible to look at a gene ized example is the requirement of histone H2B monoubiquitina-
and its regulatory sequences and predict when and where tion for proper H3K4 methylation by the H3K4 methylase
a gene was going to be transcribed. We then learned that along complex COMPASS (Figure 1B) (Shilatifard, 2006). This process,
with the sequence-specific binding of activators and repressors, initially discovered in yeast (Shilatifard, 2006), is now known to
there is an additional world of factors that modify, interact, and be highly conserved among eukaryotes (Kim et al., 2009). Addi-
remodel chromatin to regulate gene expression. The identification tionally, a recent comprehensive mutation analysis of all histone
of a multitude of histone modifications—some correlated with residues reveals that a single point mutation in histone H3K14,
activation, some with repression—led to the proposal that the a site of acetylation, results in the specific loss of H3K4 trimethy-
modifications constitute a code that could be recognized by tran- lation, but not mono- or dimethylation (Nakanishi et al., 2008).
scription factors to determine the transcriptional state of a gene. This screen also demonstrated that H3K4 trimethylation is regu-
However, additional research has since added layers of lated by both monoubiquitination-dependent and monoubiquiti-
complexity, revealing a nuanced and intriguing language, not nation-independent processes (Nakanishi et al., 2008).
a strict code, as the basis for transcriptional regulation through Given that histone-modifying enzymes are often found in mul-
the chromatin signaling pathway. Here, we review the complex tisubunit complexes, modification of nearby residues can create
crosstalk among histone modifications, including recent studies binding sites for the components of the complex helping to
that illustrate how the context and timing of these modifications anchor an enzyme to a nucleosome. For example, the PHD finger
are critical for a particular transcriptional readout. of Yng1, a subunit of the NuA3 histone acetyltransferase
complex, recognizes methylated H3K4 and helps recruit this
Histone Crosstalk and Gene Activity histone acetyltransferase complex for acetylation of H3K14
In eukaryotic cells, gene expression can be regulated at the level (Martin et al., 2006; Taverna et al., 2006). The Yng1-related
of chromatin structure. Numerous residues within the histone ING2 can also bind methylated H3K4; however, it is present in
tails and several residues within the histone globular domains a histone deacetylase complex (Shi et al., 2006). Therefore,
can be modified in a variety of ways, including acetylation, H3K4 methylation can serve as a landing platform for a variety
phosphorylation, ubiquitination, and methylation. A well-charac- of histone-modifying enzymes with opposing activities.
terized posttranslational modification regulating chromatin Modifications of nearby residues can also prevent the recog-
structure is the acetylation of histone N-terminal tails, which is nition of a substrate by an enzyme, as recently reported to occur
thought to facilitate transcriptional activation either by charge when methylation of histone H3 arginine 2 (H3R2) interferes with
neutralization of the tails’ interaction with DNA or by forming H3K4 methylation by Set1/COMPASS in yeast and COMPASS-
a binding site for bromodomain-containing transcription factors, like complexes in mammalian cells (Guccione et al., 2007;
some of which can remodel nucleosomes. Another well-studied Kirmizis et al., 2007) (Figure 1B). Histone modifications can
histone modification is the methylation of lysine 4 of histone H3 also prevent the recruitment of factors other than enzymes.
(H3K4), a modification generally associated with transcriptionally For example, heterochromatin protein 1 (HP1), which binds
active genes and a binding site for a variety of factors that include methylated H3K9, cannot do so when the adjacent serine 10
histone-modifying and -remodeling activities (Shilatifard, 2006). (H3S10) is phosphorylated during mitosis or during gene activa-
More complex scenarios arise when histone modifications act tion (Fischle et al., 2005; Mateescu et al., 2008).
combinatorially in a context-dependent manner to facilitate or Multiple types of histone crosstalk, involving numerous histone-
repress chromatin-mediated transcription. In some cases the modifying complexes, can occur at any one gene. A major chal-
modification of one residue can alter the ability of a second lenge is to understand the events that regulate changes in gene
residue to be implemented by its modifying enzyme(s) (Figure 1). expression through these modifications/crosstalk. One strategy
The first example of histone crosstalk falls into this category: the has been to profile histone modifications genomewide, with the

682 Cell 142, September 3, 2010 ª2010 Elsevier Inc.

a gene activated in response to serum (Zippo et al., 2009) (Fig-
ure 2A). They present evidence for a transcription activation
pathway in which the phosphorylation of H3 tails leads to the
acetylation of H4 tails. In turn, acetylation of H4 tails is required
for the recruitment of the RNA Pol II positive transcription elonga-
tion factor, P-TEFb (Figure 2A). Previously, the authors found
that activation of the FOSL1 gene requires PIM1, a proto-onco-
gene whose kinase activity is activated through MAP kinase
signaling. Numerous cellular substrates of PIM1 have been iden-
tified, including H3S10. Other H3S10 kinases, such as MSK1 and
MSK2 (MSK1/2), are also implicated in the phosphorylation of
histones at serum-responsive genes, including FOSL1. Zippo
and colleagues find that the spatiotemporal pattern of H3S10
Figure 1. Examples of Histone Crosstalk phosphorylation differs for PIM1 and MSK1/2. MSK1/2 mediates
(A) The first characterized example of histone crosstalk is the stimulation of
the phosphorylation of H3S10 at the promoter of FOSL1 at early
acetyltransferase activity of GCN5 toward the histone H3 tail by prior phos-
phorylation (P) of serine 10. Acetylation, Ac. time points of gene expression, whereas PIM1 is required for
(B) Crosstalk among histone modifications can span more than one histone. H3S10 phosphorylation at a FOSL1 enhancer after the MSK1/
Monoubiquitination of histone H2B on lysine 120 of the C-terminal helix can 2-mediated phosphorylation of H3S10 (Figure 2A).
lead to the trimethylation of lysine 4 in the histone 3 tail (H3K4) by Set1/
COMPASS. However, H3K4 methylation by COMPASS and COMPASS-like Screening for other histone modifications specifically associ-
complexes can be blocked if the nearby arginine of H3 is already methylated. ated with the FOSL1 enhancer shows that the acetylation of
H4K16 coincides with H3S10 phosphorylation. RNA interfer-
expectation that a given pattern will indicate a transcriptional ence-mediated knockdown of PIM1 results in loss of H4K16
outcome due to the recruitment of specific proteins by these modi- acetylation, suggesting a trans-tail crosstalk from H3S10 phos-
fications. However, some recent examples of trans-histone cross- phorylation to H4K16 acetylation. Zippo and colleagues asked
talk illustrate that transcriptional readout depends on context and whether 14-3-3 g, 3, and z proteins, previously shown to bind
timing by which these modifications are introduced. Simply put, phosphorylated H3S10, are recruited to the promoter and the
just looking at the pattern of chromatin modifications at a locus enhancer of FOSL1 in response to serum. They find that 14-3-
is not sufficient to determine its gene expression status. These 33 and 14-3-3z are recruited to both the promoter and enhancer
studies provide new insight into the language of histone crosstalk. of FOSL1 after serum stimulation. However, 14-3-3 is required
for recruiting the H4K16 histone acetyltransferase MOF only to
From Histone Phosphorylation to Transcription the enhancer and not to the promoter of FOSL1. Recruitment
Elongation of MOF to the enhancer results in H4K16 acetylation, which
A novel form of crosstalk was recently discovered by Zippo and can be bound by the bromodomain-containing protein, Brd4.
colleagues, who studied the transcriptional control of FOSL1, Brd4 associates with P-TEFb, a kinase that phosphorylates Pol

Figure 2. Context-Dependent Outcomes of

Histone Crosstalk
(A) Zippo and colleagues (Zippo et al., 2009)
uncover a new form of histone crosstalk by
studying the transcriptional control of FOSL1,
a gene activated in response to serum. Activation
requires the binding of PIM1 to the FOSL1
enhancer. PIM1 is a kinase responsible for phos-
phorylation (P) of serine 10 on the histone H3 tail
(H3S10). Phosphorylated H3S10 creates a binding
site for 14-3-3, a phosphoserine binding protein.
Acetylation (Ac) of lysine 16 on the H4 (H4K16)
tail occurs through interaction of 14-3-3 with the
histone acetyltransferase MOF. Acetylated
H4K16 can in turn form a binding site for the bro-
modomain-containing protein Brd4, a component
of P-TEFb, a kinase that phosphorylates the
C-terminal domain of RNA Pol II to facilitate tran-
scription elongation. However, at an earlier stage
of serum stimulation, an MSK1/2 kinase is re-
cruited to the promoter where it phosphorylates
H3S10. 14-3-3 is then recruited to the promoter,
but unlike the situation at the enhancer, MOF is
not recruited to the promoter. Thus, the timing,
location, and perhaps identity of the H3 kinase, and not the H3S10 modification alone, determines downstream events.
(B) Another example of how the order of implementation of histone modifications can affect transcription comes from work from Wang et al. (2009). They report
that despite correlations between histone acetylation and H3K4 methylation, artificially increasing acetylation through treatment of cells with deacetylase inhib-
itors (HDACs) does not lead to productive transcription, despite the presence of H3K4 methylation and Pol II recruitment. Therefore, patterns of histone modi-
fications cannot simply be ‘‘read’’ but instead have distinct effects depending on the cellular context and upstream signaling events.

Cell 142, September 3, 2010 ª2010 Elsevier Inc. 683

II to facilitate transcription elongation (Figure 2A). Thus, Zippo with H3K4 methylation rarely show this increase in acetylation
and colleagues propose that crosstalk between modifications in response to deacetylase inhibitors.
on two different histone tails regulate productive transcription In order to determine whether H3K4 methylation is functionally
elongation through the sequential recruitment of proteins that linked to H4K16 acetylation, Wang and colleagues use RNA
bind these modifications. interference-mediated knockdown of WDR5, a common compo-
One question raised by this study is why H3S10 phosphoryla- nent of the Set1 and MLL (mixed-lineage leukemia) COMPASS-
tion produces different results at the enhancer than at the like H3K4 methyltransferase complexes. Upon knockdown of
promoter of FOSL1 even though 14-3-3 is recruited to both sites. WDR5, they observe reduced levels of histone acetylation at
At the enhancer, 14-3-3 recruits the histone acetyltransferase the subset of transcriptionally quiescent genes marked by
MOF. At the promoter, it does not. What is the difference H3K4 methylation. Based on this information, Wang and
between 14-3-3 at the promoter and at the enhancer? Interest- colleagues suggest that H3K4 methylation primes certain genes
ingly, 14-3-33 and 14-3-3z are thought to be regulated via lysine for an increase in H3K9 and H4K16 acetylation. Interestingly,
acetylation (Choudhary et al., 2009) and an acetyltransferase, what was not considered by Wang and colleagues is the fact
Tip60, is preferentially recruited to the promoter of FOSL1. One that WDR5 is also a component of complexes that can acetylate
possibility is that Tip60 acetylates 14-3-3 and prevents its inter- H4K16 or H3K9. WDR5 associates with the H4K16 acetyltrans-
action with MOF. ferase MOF in the NSL/MSL1v complex (Cai et al., 2010; Li
Another intriguing aspect of the study by Zippo and colleagues et al., 2009) as well as with the H3K9/K14 acetyltransferase
is the link between H3S10 phosphorylation and H4K16 acetyla- Gcn5 in the ATAC complex (Suganuma et al., 2008; Wang
tion. These two modifications were previously linked in studies of et al., 2008). As such, the effect of WDR5 knockdown could be
dosage compensation in the fruit fly Drosophila. In Drosophila a consequence of WDR5’s role as a subunit of the H3K4 methyl-
dosage compensation, MOF is recruited to the coding region ases, or WDR5’s role as a subunit of the H3 and H4 acetyltrans-
of X-linked genes in males where it mediates H4K16 acetylation ferase complexes, or a combination of the two. Given that WDR5
in a process thought to facilitate transcription elongation. Coloc- is part of the H3 and H4 acetyltransferase complexes, the exis-
alizing with MOF on the male X chromosome is the JIL-1 kinase, tence of crosstalk between H3K4 and H4K16 needs to be further
an MSK1/2-related kinase that mediates the phosphorylation of characterized.
H3S10 on this chromosome. In the case of Drosophila dosage One surprising finding of the study by Wang and colleagues is
compensation, recruitment of the JIL-1 kinase to the male X that transcription is rarely induced at the genes tested, although
chromosome occurs later than H4K16 acetylation (Wang et al., Pol II is recruited following treatment with a deacetylase inhibitor
2001), a reversal of the order of the addition of these marks at (Figure 2B). Thus, Pol II recruitment does not lead to the antici-
FOSL1 in response to serum. Concordantly, the MOF complex pated increase in transcription, despite the fact that H3K9 and
that mediates acetylation in coding regions is likely to be distinct H4K16 acetylation are increased. These modifications coincide
from the MOF complex that mediates promoter/enhancer acet- with transcriptional activation at FOSL1 upon serum treatment
ylation of genes (Li et al., 2009). Thus, by all appearances, these (Zippo et al., 2009). Together with the studies of H3S10 phos-
two examples of the coexistence of both H3S10 phosphorylation phorylation and H4K16 acetylation at the FOSL1 enhancer, it is
and H4K16 acetylation are unrelated in their order of implemen- clear that knowing the mechanism and timing of these modifica-
tation and in their biological meaning. This suggests that tions is necessary for determining the transcriptional outcome.
descriptions of histone modification patterns, without under-
standing the mechanisms leading to the implementation of these The Emerging Grammar of Histone Crosstalk
marks, should be interpreted with caution. Importantly, the study The existence of a histone modification code was proposed
by Zippo and colleagues begins to determine the role of histone 10 years ago as a way to approach the study of the quickly
modifications in the activation of FOSL1, with a spatial and growing number of histone modifications involved in the regula-
temporal dissection of how a cascade of histone modifications tion of gene expression and other DNA-templated processes,
can lead to a particular transcriptional outcome. such as replication, repair, and recombination. New ‘‘words’’ of
histone modifications are being discovered, and they continue
to appear in interesting combinations. However, discovering
Trimethylation Converses with Acetylation the exact roles these modifications play in gene expression
Another example of trans-tail crosstalk was proposed by Wang has been complicated by finding a counterexample for almost
et al. (2009). In this case the communication takes place between every example of crosstalk, such as the case of H3K4 and
the H3 and H4 tails and, like the example provided by Zippo et al. H3K36 methylation recruiting both histone acetyltransferases
(2009), involves H4K16 acetylation and FOSL1 transcription. By and deacetylases.
analyzing genomewide profiles of several histone acetyltrans- A common theme of recent research on histone crosstalk is
ferases, deacetylases, and modifications, these investigators that the order and mechanism of the addition and removal of
find a link between H3K4 methylation and H4K16 acetylation at histone modifications are important for the transcriptional
some inducible genes, including FOSL1. The authors show readout of a gene. The recent examples of histone crosstalk
that a subset of transcriptionally quiescent genes, marked by that we have addressed here illustrate this point. In one study,
the presence of H3K4 methylation, display a marked increase the implementation of H3S10 phosphorylation at two different
in histone acetylation at H3K9 and H4K16 after addition of de- locations, by two different enzymes, and at two different times
acetylase inhibitors. In contrast, quiescent genes not marked after serum stimulation had disparate effects on subsequent

684 Cell 142, September 3, 2010 ª2010 Elsevier Inc.

histone acetylation at the respective locations (Zippo et al., 2009) Choudhary, C., Kumar, C., Gnad, F., Nielsen, M.L., Rehman, M., Walther, T.C.,
(Figure 2A). Zippo and colleagues were able to propose a mech- Olsen, J.V., and Mann, M. (2009). Science 325, 834–840.
anism of gene activation by identifying the histone-modifying Fischle, W., Tseng, B.S., Dormann, H.L., Ueberheide, B.M., Garcia, B.A., Sha-
enzymes, the histone modifications, and a set of proteins that banowitz, J., Hunt, D.F., Funabiki, H., and Allis, C.D. (2005). Nature 438, 1116–
recognized these modifications on the FOSL1 gene after serum
stimulation. In another study, Wang and colleagues found Guccione, E., Bassi, C., Casadio, F., Martinato, F., Cesaroni, M., Schuchlautz,
H., Luscher, B., and Amati, B. (2007). Nature 449, 933–937.
that artificially recreating histone modifications that correlate
with gene expression could result in the recruitment of RNA Kim, J., Guermah, M., McGinty, R.K., Lee, J.S., Tang, Z., Milne, T.A., Shilati-
fard, A., Muir, T.W., and Roeder, R.G. (2009). Cell 137, 459–471.
Pol II, but this was not sufficient for transcription (Figure 2B).
Thus, simply mapping histone modification patterns without Kirmizis, A., Santos-Rosa, H., Penkett, C.J., Singer, M.A., Vermeulen, M.,
Mann, M., Bahler, J., Green, R.D., and Kouzarides, T. (2007). Nature 449,
understanding the recruitment, regulation, and interactions of
the complexes implementing these marks is not sufficient
Li, X., Wu, L., Corsa, C.A., Kunkel, S., and Dou, Y. (2009). Mol. Cell 36,
to understand the mechanisms regulating gene expression.
Genomewide profiling techniques have now become widely
Lo, W.S., Trievel, R.C., Rojas, J.R., Duggan, L., Hsu, J.Y., Allis, C.D., Marmor-
adopted, providing the ability to map histone modifications,
stein, R., and Berger, S.L. (2000). Mol. Cell 5, 917–926.
the enzymes implementing these marks, and the factors that
Martin, D.G., Baetz, K., Shi, X., Walter, K.L., MacDonald, V.E., Wlodarski, M.J.,
recruit them under different experimental conditions.
Gozani, O., Hieter, P., and Howe, L. (2006). Mol. Cell. Biol. 26, 7871–7879.
The study of the regulation of gene expression has grown from
Mateescu, B., Bourachot, B., Rachez, C., Ogryzko, V., and Muchardt, C.
identifying transcription factors and their binding sites to include
(2008). EMBO Rep. 9, 267–272.
a wide variety of other binding events associated with the modi-
Nakanishi, S., Sanderson, B.W., Delventhal, K.M., Bradford, W.D., Staehling-
fications of histones that package the DNA. Future progress will
Hampton, K., and Shilatifard, A. (2008). Nat. Struct. Mol. Biol. 15, 881–888.
require us to learn much more about how the words comprising
Shi, X., Hong, T., Walter, K.L., Ewalt, M., Michishita, E., Hung, T., Carney, D.,
the dictionary of histone crosstalk are used in a particular order Pena, P., Lan, F., Kaadige, M.R., et al. (2006). Nature 442, 96–99.
to provide the grammar of this complex language.
Shilatifard, A. (2006). Annu. Rev. Biochem. 75, 243–269.
ACKNOWLEDGMENTS Suganuma, T., Gutierrez, J.L., Li, B., Florens, L., Swanson, S.K., Washburn,
M.P., Abmayr, S.M., and Workman, J.L. (2008). Nat. Struct. Mol. Biol. 15,
We thank Dr. E. Park for the critical reading of the manuscript. We also thank 364–372.
L. Shilatifard for editorial assistance. Studies in the Shilatifard laboratory are Taverna, S.D., Ilin, S., Rogers, R.S., Tanny, J.C., Lavender, H., Li, H., Baker, L.,
supported in part by grants from the NIH: R01CA89455, R01GM069905, and Boyle, J., Blair, L.P., Chait, B.T., et al. (2006). Mol. Cell 24, 785–796.
Wang, Y., Zhang, W., Jin, Y., Johansen, J., and Johansen, K.M. (2001). Cell
105, 433–443.
Wang, Y.L., Faiola, F., Xu, M., Pan, S., and Martinez, E. (2008). J. Biol. Chem.
283, 33808–33815.
Cai, Y., Jin, J., Swanson, S.K., Cole, M.D., Choi, S.H., Florens, L., Washburn,
M.P., Conaway, J.W., and Conaway, R.C. (2010). J. Biol. Chem. 285, 4268– Wang, Z., Zang, C., Cui, K., Schones, D.E., Barski, A., Peng, W., and Zhao, K.
4272. (2009). Cell 138, 1019–1031.
Cheung, P., Tanner, K.G., Cheung, W.L., Sassone-Corsi, P., Denu, J.M., and Zippo, A., Serafini, R., Rocchigiani, M., Pennacchini, S., Krepelova, A., and
Allis, C.D. (2000). Mol. Cell 5, 905–915. Oliviero, S. (2009). Cell 138, 1122–1136.

Cell 142, September 3, 2010 ª2010 Elsevier Inc. 685

GPR120 Is an Omega-3 Fatty Acid Receptor
Mediating Potent Anti-inflammatory
and Insulin-Sensitizing Effects
Da Young Oh,1,4 Saswata Talukdar,1,4 Eun Ju Bae,1 Takeshi Imamura,2 Hidetaka Morinaga,1 WuQiang Fan,1 Pingping Li,1
Wendell J. Lu,1 Steven M. Watkins,3 and Jerrold M. Olefsky1,*
1Department of Medicine, Division of Endocrinology and Metabolism, University of California, San Diego, La Jolla, CA 92093, USA
2Divisionof Pharmacology, Shiga University of Medical Science, Tsukinowa, Seta, Otsu-city, Shiga, 520-2192 Japan
3Tethys Bioscience, 3410 Industrial Boulevard, Suite 103, West Sacramento, CA 95691, USA
4These authors contributed equally to this work

DOI 10.1016/j.cell.2010.07.041

SUMMARY and anti-inflammatory molecules within the specialized popula-

tion of proinflammatory tissue macrophages.
Omega-3 fatty acids (u-3 FAs), DHA and EPA, exert G protein-coupled receptors (GPCRs) are important signaling
anti-inflammatory effects, but the mechanisms are molecules for many aspects of cellular function. They are
poorly understood. Here, we show that the G members of a large family that share common structural motifs
protein-coupled receptor 120 (GPR120) functions as such as seven transmembrane helices and the ability to acti-
an u-3 FA receptor/sensor. Stimulation of GPR120 vate heterotrimeric G proteins, such as Gas, Gai, and Gaq.
Ligands bind specifically to GPCRs to stimulate and induce
with u-3 FAs or a chemical agonist causes broad
a variety of cellular responses via several second messenger
anti-inflammatory effects in monocytic RAW 264.7 pathways; e.g., modulation of cAMP production, the phospho-
cells and in primary intraperitoneal macrophages. lipase C pathway, ion channels, and mitogen-activated protein
All of these effects are abrogated by GPR120 knock- kinases (Ulloa-Aguirre et al., 1999; Gether, 2000; Schulte and
down. Since chronic macrophage-mediated tissue Fredholm, 2003. Recently, several groups reported that five
inflammation is a key mechanism for insulin resis- orphan receptors, GPR40, GPR41, GPR43, GPR84, and
tance in obesity, we fed obese WT and GPR120 GPR120, can be activated by free fatty acids (FFAs). Short-
knockout mice a high-fat diet with or without u-3 FA chain fatty acids (FAs) are specific agonists for GPR41 and
supplementation. The u-3 FA treatment inhibited GPR43 (Tazoe et al., 2008) and medium-chain FAs for GPR84
inflammation and enhanced systemic insulin sensi- (Wang et al., 2006a). Long-chain FAs can activate GPR40
tivity in WT mice, but was without effect in GPR120 (Itoh et al., 2003) and GPR120 (Hirasawa et al., 2005). Hirasawa
et al. showed that the stimulation of GPR120 by FFAs resulted
knockout mice. In conclusion, GPR120 is a functional
in elevation of [Ca2+]i and activation of the ERK cascade which
u-3 FA receptor/sensor and mediates potent insulin suggests interactions with the Gaq family of G proteins.
sensitizing and antidiabetic effects in vivo by repres- However, other physiological functions of GPR120 remain to
sing macrophage-induced tissue inflammation. be explored. In the current study, we found that GPR120 was
highly expressed in adipose tissue, and proinflammatory
macrophages. The high expression level of GPR120 in mature
INTRODUCTION adipocytes and macrophages indicates that GPR120 might
play an important role in these cell types. We stimulated
Chronic activation of inflammatory pathways plays an important GPR120 with a synthetic agonist (GW9508) and omega-3 fatty
role in the pathogenesis of insulin resistance and the macro- acids (u-3 FAs) and examined whether activation of GPR120
phage/adipocyte nexus provides a key mechanism underlying affected LPS- and TNF-a-induced inflammatory signaling
the common disease states of decreased insulin sensitivity responses. While SFAs are proinflammatory and unsaturated
(Schenk et al., 2008). This involves migration of monocytes/ FAs are generally neutral, we found that u-3 FAs (docosahex-
macrophages to adipose tissue (including intramuscular fat aenoic acid (C22:6n3, DHA) and eicosapentaenoic acid
depots) and liver with subsequent activation of macrophage (C20:5n3, EPA)), the major ingredients in fish oil, exert potent
proinflammatory pathways and cytokine secretion. Through anti-inflammatory effects through GPR120.
paracrine effects, these events promote inflammation and b-arrestins can serve as scaffold or adaptor proteins for a wide
decreased insulin sensitivity in nearby insulin target cells range of GPCRs, as well as a selected group of other receptor
(Shoelson et al., 2007; Schenk et al., 2008). In these studies, subtypes (Miller and Lefkowitz, 2001). After ligand binding,
we explore the interlocking biology between proinflammatory b-arrestins can associate with the cytoplasmic domains of

Cell 142, 687–698, September 3, 2010 ª2010 Elsevier Inc. 687

Figure 1. Expression Level of GPR120 and
GPR120-Mediated Anti-inflammatory Res-
ponse in RAW 264.7 Cells
(A and B) (A) The mRNA expression pattern of
various lipid sensing GPCRs is shown in adipose
tissue, (B) CD11c+ bone marrow-derived dendritic
cells (BMDCs), bone marrow-derived macro-
phages (BMDMs), IPMacs, 3T3-L1 preadipocytes,
differentiated 3T3-L1 adipocytes, RAW 264.7
cells, and L6 myocytes. Ribosomal protein S3
(RPS3) was used as internal control.
(C) Expression of GPR120 in SVF, adipocytes and
hepatic Kupffer cells from chow (NC)- or HFD-fed
mice was examined by q-PCR. Data are ex-
pressed as the mean ± SEM of at least three
independent experiments in triplicate. *p < 0.05
versus NC.
(D) RAW 264.7 cells, transfected with scrambled
(Scr) or GPR120 #2 siRNA (GPR120 KD), were
treated with 100 mM of GW9508 for 1 hr prior to
LPS (100 ng/ml) treatment for 10 min and then
subjected to western blotting. Left panel is a repre-
sentative image from three independent experi-
ments, and the scanned bar graph (right panel)
shows fold induction over basal conditions.
Knockdown efficiency of GPR120 siRNA is shown
in Figure S1.
(E) Cytokine secretion level was measured in RAW
264.7 cells by ELISA. Data are expressed as the
mean ± SEM of three independent experiments.
*p < 0.05 versus LPS treatment in scrambled
siRNA transfected cells. See also Figures S1
and S2.

GPCRs and couple the receptor to specific downstream sensing GPCR which is highly expressed in adipose tissue,
signaling pathways, as well as mediate receptor endocytosis proinflammatory CD11c+ macrophages (BMDCs), mature adipo-
(Luttrell and Lefkowitz, 2002). Here we find that b-arrestin2 asso- cytes, and monocytic RAW 264.7 cells (Figure 1A and 1B).
ciates with ligand-stimulated GPR120 and participates in down- GPR120 is induced in the stromal vascular fraction (SVF) of
stream signaling mechanisms. adipose tissue (which contains the macrophages), as well as in
Since chronic inflammation is a mechanistic feature of hepatic Kupffer cells, during high-fat diet (HFD) feeding in mice
obesity-related insulin resistance, we postulated that the anti- (Figure 1C). GPR120 is also expressed in enteroendocrine L cells
inflammatory effect of GPR120 stimulation could promote insulin with negligible expression in muscle (Figure S4C available
sensitization. In the present study, we elucidate the role of online), hepatocytes or other cell types (Hirasawa et al., 2005;
GPR120 activation in integrating anti-inflammatory and insulin Gotoh et al., 2007).
sensitizing effects in vitro and in vivo.
Ligand-Stimulated GPR120 Exerts Anti-inflammatory
It has been previously reported that GPR120 signals via
GPR120 Expression a Gaq/11-coupled pathway and can respond to long chain FAs
Fatty acids (FAs) can function as endogenous ligands modu- (Hirasawa et al., 2005). To pursue the biology of GPR120,
lating inflammatory responses, but not all FAs work in the a tool compound was needed, and, some years ago, Glaxo pub-
same way. In general, saturated FAs (SFAs) are proinflammatory, lished GW9508 as a GPR40 selective agonist. However, this
unsaturated FAs are weakly proinflammatory or neutral, and u3- compound was not specific and also stimulated GPR120 (Bris-
FAs can be anti-inflammatory (Lee et al., 2003; Calder, 2005; coe et al., 2006). Since macrophages and adipocytes do not
Solinas et al., 2007). Because of the importance of inflammation express GPR40 (this was confirmed by repeated q-PCR and
in a number of chronic human diseases including insulin resis- RT-PCR measures, Figure S1A), GW9508 is a functional
tance, obesity, and type 2 diabetes mellitus, we surveyed the GPR120 specific compound in these cell types. Using this
family of FA sensing GPCRs (GPR40, 41, 43, 84, and 120). Based approach, we found that GW9508 treatment broadly and mark-
on its tissue expression pattern, GPR120 emerged as a receptor edly repressed the ability of the TLR4 ligand LPS to stimulate
of particular interest. As seen in Figure 1, GPR120 is the only lipid inflammatory responses in RAW 264.7 cells (Figure 1D and E).

688 Cell 142, 687–698, September 3, 2010 ª2010 Elsevier Inc.

Figure 2. Omega-3 FA Stimulates GPR120
and Mediates Anti-inflammatory Effects
(A–D) GPR120-mediated SRE-luc activity after
treatment with various FAs. Results are fold
activities over basal. Each data point represents
mean ± SEM of three independent experiments
performed in triplicate. Black lines indicate SRE-
luc activities without GPR120 transfection or with
non-stimulating FAs. DHA inhibits LPS-induced
inflammatory signaling (B), cytokine secretion (C),
and inflammatory gene mRNA expression level
(D) in RAW 264.7 cells, but not in GPR120 knock-
down cells.
(E and F) GPR120 stimulation inhibits LPS-induced
inflammatory response in WT primary macro-
phage. Data are expressed as the mean ± SEM
of three independent experiments. *p < 0.05
versus LPS treatment in scrambled siRNA trans-
fected cells or WT IPMacs. See also Figure S2.

lation in RAW 264.7 cells, but only DHA-

and a-linolenic acid-mediated ERK phos-
phorylation were abolished by GPR120
knockdown (Figure S2A). These results
indicate that u-3 FAs, but not SFAs,
specifically activated ERK via GPR120.
The activation of GPR120 by u-3 FAs,
as well as its expression in adipocytes
and macrophages, led us to study
whether DHA, a representative u-3 FA,
can modulate inflammation through
GPR120 in these cells. To examine this,
we pretreated RAW 264.7 cells and
3T3-L1 adipocytes with GW9508 or DHA
for 1 hr, followed by LPS (TLR4), TNF-a,
TLR2, or TLR3 stimulation, respectively.
We found that GW9508 and, more impor-
tantly, DHA, strongly inhibited LPS-
Thus, GW9508 inhibited LPS stimulated phosphorylation of IKKb induced phosphorylation of JNK and IKKb, IkB degradation,
and JNK, prevented IkB degradation, and inhibited TNF-a and cytokine secretion and inflammatory gene expression level in
IL-6 secretion. All of these effects of GW9508 were completely RAW 264.7 cells (Figures 2B–2D) as well as TNF-a, TLR2 and
abrogated by siRNA mediated knockdown of GPR120 (Figures TLR3-induced JNK and IKKb phosphorylation in 3T3-L1 adipo-
1D and 1E and Figure S1F). cytes (Figure S2B) or RAW 264.7 cells (Figure S2C). All of the
Based on these remarkable anti-inflammatory effects of effects of GW9508 and DHA were completely prevented by
GPR120 stimulation, we established a cell based reporter GPR120 knockdown, demonstrating that these anti-inflamma-
system by transfecting HEK293 cells with constructs for tory effects were specifically exerted through GPR120 (Figure 1,
GPR120 along with a serum response element-luciferase Figure 2, Figure S1, and Figure S2). Similar results were seen in
promoter/reporter (SRE-luc). Since GPR120 is a Gaq/11- primary wild-type (WT) intraperitoneal macrophages (IPMacs)
coupled receptor, it stimulates both PKC and MAP kinase, and and GPR120 knockout (KO) IPMacs (Figures 2E and 2F). These
both of these biologic effects are detected by the SRE-driven data argue that GPR120 is an u-3 FA receptor or sensor, and
reporter system (Oh et al., 2005). The reporter cells were treated provide a molecular mechanism for the anti-inflammatory effects
with various FAs and the synthetic GW9508 ligand. We found of this class of FAs.
that GW9508, the u-3 FAs (DHA and EPA) and palmitoleate
(C16:1n7), all activated the SRE-luc reporter with an EC50 of Role of b-arrestin2 in GPR120 Signaling
1-10 mM (Figure 2A), while SFAs were without effect. GW9508 Given these potent cell selective anti-inflammatory effects,
and DHA were used at 100 mM in all subsequent studies to it was of interest to understand the specific mechanisms
achieve maximal action. The u-3 FAs (DHA and a-linolenic whereby signals from GPR120 inhibit inflammatory pathways.
acid), and SFA (palmitic acid (C16:0)) activated ERK phosphory- To further assess this, we used RNA interference to examine

Cell 142, 687–698, September 3, 2010 ª2010 Elsevier Inc. 689

Figure 3. GPR120 Internalization with
b-arrestin2 Mediates Anti-inflammatory
(A) RAW 264.7 cells were transfected with siRNA as
indicated and stimulated with or without 100 mM of
DHA 1 hr prior to LPS (100 ng/ml) treatment for
10 min and then subjected to western blotting.
(B) TNF-a secretion was measured in RAW 264.7
cell cultured media with or without RNA interfer-
ence as indicated.
(C) Phosphorylation of TAK1 and MKK4 in RAW
264.7 cells with or without siRNA transfection as
(D) HEK293 cells were cotransfected with HA-
GPR120 and b-arrestin2$GFP to analyze
GPR120 internalization after DHA stimulation for
the indicated times. GPR120 (red) and b-arrestin2
(green) were localized by confocal microscopy.
(E–H) (E) Coimmunoprecipitation between GPR120
and b-arrestin2 with DHA stimulation for 30 min
in RAW 264.7 cells and, (F) HEK293 cells (HA-
GPR120 and b-arrestin2$GFP), respectively. Lysate
indicates 1/10 input in each experiment. Interaction
between TAB1 and b-arrestin2 (G) and interaction
between TAB1 and TAK1 (H) were detected by
coimmunoprecipitation and the scanned bar graph
quantitates the association in RAW 264.7 cells.
(I) Schematic diagram of the b-arrestin2 and
GPR120-mediated anti-inflammatory mechanism.
Red colored letters and arrows indicate the DHA-
mediated anti-inflammatory effect, and black
colored letters and arrows indicate the LPS- and
TNF-a-induced inflammatory pathway. See also
Figure S2.

phorylation in a GPR120 and b-arrestin2-

dependent manner. Since TLR2/3/4 and
TNF-a signaling were inhibited by
GPR120 activation, these results indicate
that DHA signaling intersects at TAK1 and
molecules involved in generation of GPR120 signals. As seen in inhibits all upstream input activating signals via a GPR120/b-
Figures 3A and 3B, LPS signaling was not affected by b-ar- arrestin2 interaction (Figure 3I).
restin1, -2, or Gaq/11 knockdown. However, with b-arrestin2 After ligand stimulation, b-arrestin2 can translocate to
knockdown, DHA-mediated anti-inflammatory signaling was ex- a number of GPCRs where it mediates receptor internalization
tinguished, while b-arrestin1 and Gaq/11 knockdown were and signaling (Barak et al., 1997). We transfected HEK293 cells
without effect (Figure 3A). with b-arrestin2$GFP to visualize intracellular trafficking of b-ar-
Figure 3A and Figure S2 show that GPR120 stimulation inhibits restin2 following activation of GPR120 (Figure 3D). In the basal
both TLR4- and TNF-a mediated inflammatory responses. Since state, GPR120 was localized to the plasma membrane as
the TNF-a and TLR signaling cascades converge downstream of assessed by immunostaining (red fluorescence, Figure 3D),
GPR120 activation, these results indicate that the site of while b-arrestin2 exhibited a diffuse, largely cytoplasmic stain-
GPR120-induced inhibition is either at, or upstream, of JNK/ ing pattern (green, Figure 3D). Following DHA treatment for
IKKb. LPS activates inflammation through the TLR4 pathway 5 min, b-arrestin2$GFP translocated from the cytosol to the
by engaging the serine kinase IRAK, leading to phosphorylation plasma membrane and can be seen colocalized with GPR120
of transforming growth factor-b activated kinase 1 (TAK1) which (merged, right fields). After 30 min of DHA treatment, much of
is upstream of MKK4/JNK and IKKb (Kawai and Akira, 2006, the GPR120 is internalized, as visualized by punctate intracel-
Figure 3I). TNF-a and TLR2/3 also leads to stimulation of lular staining (lower left panel), and b-arrestin2$GFP is now
TAK1, resulting in activation of IKKb and JNK (Takaesu et al., colocalized with the intracellular GPR120 (lower right panel,
2003). Consequently, we determined whether DHA stimulation Figure 3D). DHA-stimulated binding of b-arrestin2 to activated
of GPR120 inhibited TAK1 and MKK4. As seen in Figure 3C, GPR120 was also detected by coimmunoprecipitation
DHA treatment abrogated LPS-induced TAK1 and MKK4 phos- (Figure 3E and F).

690 Cell 142, 687–698, September 3, 2010 ª2010 Elsevier Inc.

Figure 4. GPR120 Activation Enhances
GLUT4 Translocation and Glucose Uptake
(A) 3T3-L1 adipocytes were transfected with
a dually tagged HA-GLUT4-GFP construct. Total
GLUT4 expression was determined by GFP fluo-
rescence, and GLUT4 translocation to the cell
surface after 100 ng/ml insulin or 100 mM DHA
stimulation for 30 min was determined by indirect
immunofluorescence of the HA-conjugated with
Alexa 594 in fixed cells. Translocation following
insulin stimulation was expressed as a percentage
of the maximum response. The bar graph
represents the mean ± SEM data from four inde-
pendent experiments. *p < 0.05 versus vehicle
treatment. (B) Glucose uptake was measured in
WT and GPR120 KO mouse primary adipose
tissue and in (C–H) 3T3-L1 adipocytes ± siRNA
with the indicated treatment. Data are expressed
as mean ± SEM of three independent experiments
in triplicate. *p < 0.05 versus basal activity. The
indicated siRNA knockdown efficiency was vali-
dated by western blotting. See also Figure S3.

GPR120 Activation Enhances

Glucose Uptake in 3T3-L1
Since our data show that GPR120 is
expressed in mature adipocytes and sig-
nals through Gaq/11 in these cells, we
assessed the effects of GPR120 stimula-
tion on insulin sensitivity in primary adi-
pose tissue cultures and in 3T3-L1 adipo-
cytes. Primary adipose tissue explants
and 3T3-L1 adipocytes were pretreated
for 30 min with GW9508 or DHA, followed
by measurement of basal and insulin
stimulated GLUT4 translocation (Fig-
LPS or TNF-a signaling activate TAK1 by causing the associ- ure 4A) and 2-deoxyglucose (2-DOG) transport (Figures 4B–
ation of TAK1 binding protein 1 (TAB1) with TAK1. Figure 3H 4H). Ligand-stimulation of GPR120 led to an increase in glucose
shows that LPS stimulation of RAW 264.7 cells causes TAB1/ transport and translocation of GLUT4 to the plasma membrane
TAK1 association. DHA treatment leads to the association of in adipocytes, but was without effect in muscle cells (Figure S4D)
b-arrestin2 with TAB1 (Figure 3G) and largely blocks TAK1/ which don’t express GPR120 (Figure 1B and Figure S4C). This
TAB1 association (Figure 3H). To further examine the interaction stimulatory effect of DHA and GW9508 was blocked when
site of b-arrestin2 and GPR120 or TAB1, we pursued coimmuno- GPR120 or Gaq/11 was depleted by siRNA knockdown
precipitation with a series of b-arrestin2 truncation/deletion (Figure 4D and Figure 4F). GLUT4 knockdown also blocked the
mutants (Figure S2D). Full-length b-arrestin2 was able to bind effects of DHA and GW9508, while the effects of insulin were
to GPR120 and TAB1 only in the presence of DHA, clearly decreased by 90% (Figure 4E). This, along with the GLUT4
showing the DHA dependency of this interaction. Interestingly, translocation data provided in Figure 4A, indicates that the
only the full-length b-arrestin2 coprecipitated with GPR120 and stimulatory effects of GPR120 are indeed working through
TAB1, while a series of deletion/truncation b-arrestin2 mutants GLUT4. Further assessment of this pathway showed that DHA
did not, indicating that the interactions are dependent on the had a modest effect to stimulate phosphorylation of Akt, but
complete tertiary structure of b-arrestin2 (Figure S2D; Luttrell that this was abrogated with GPR120 knockdown (Figure S3A).
et al., 1999). Taken together, these results suggest that The effects of DHA to stimulate Akt were blocked by inhibiting
GPR120 activation leads to association of b-arrestin2 with the PI3 kinase with LY294002 (Figure S3B). Finally, DHA did not
receptor and that this complex subsequently internalizes, where- stimulate IRS-1 phosphorylation (Figure S3C), indicating that
upon b-arrestin2 can bind to TAB1. The data further suggest that its glucose transport stimulatory effects were downstream of
association of b-arrestin2 with TAB1 blocks TAB1/TAK1 binding, IRS-1. Knockdown of Gaq/11 also completely blocked the
resulting in inhibition of TAK1 phosphorylation and activation effects of DHA to stimulate glucose transport (Figure 4F), while
(Figure 3I). b-arrestin1 or -2 knockdown was without effect (Figures 4G

Cell 142, 687–698, September 3, 2010 ª2010 Elsevier Inc. 691

Figure 5. In Vivo Metabolic Studies in
GPR120 KO Mice
(A) GTT in NC-fed WT and GPR120 KO mice. n = 7
per group.
(B and C) Insulin concentration was measured at
the indicated time points and (C) area-under-curve
analysis of the insulin data shows a significant differ-
ence between WT and GPR120 KO mice on NC.
(D) Hyperinsulinemic/euglycemic clamp studies in
chow-fed WT and GPR120 KO mice.
(E) Clamp studies in HFD, u-3 FA supplemented
(+u3), and Rosiglitazone treated HFD mice
(+Rosi). n = 8 per group, *p < 0.05 compared to
HFD-fed WT group.
(F) Mean ± SEM plasma concentration (mole (%))
of DHA and EPA for each diet in WT and GPR120
KO mice. n = 7 per each group. *p < 0.05,
compared to NC, and **p < 0.05 compared to
HFD. Data are represented as mean ± SEM. See
also Figure S4, Figure S5, Figure S6, and Table S2.

(GIR) required to maintain euglycemia in

the KO mice. Since 70%–80% of total
body insulin stimulated glucose disposal
is accounted for by skeletal muscle
glucose uptake (Baron et al., 1988), the
decreased insulin stimulated (IS)-glucose
disposal rate (GDR) provides direct
evidence for skeletal muscle insulin resis-
tance in the KO mice. Likewise, the
GPR120 KO mice exhibited a marked
decrease in the ability of insulin to sup-
press hepatic glucose production (HGP),
demonstrating the presence of hepatic
insulin resistance. Thus, the decreased
and 4H). Interestingly, Gaq/11 knockdown not only inhibited GIR was 50% related to muscle and 50% due to liver insulin
DHA and GW9508 stimulated glucose transport, but it also atten- resistance, respectively. Since the chow diet contains exoge-
uated insulin stimulatory effects, and the latter is fully consistent nous u-3 FAs, we conclude that blunted u-3 FA signaling in the
with previous publication (Imamura et al., 1999) showing the role KO mice, accounts for the decreased insulin sensitivity.
of Gaq/11 in insulin signaling to glucose transport in adipocytes. Since u-3 FA administration can improve insulin sensitivity in
This scheme is shown in Figure S3D. rats (Buettner et al., 2006), we reasoned that u-3 FA supplemen-
tation could alleviate HFD/obesity-induced insulin resistance in
In Vivo Metabolic Studies in GPR120 KO Mice WT mice, but would be ineffective in GPR120 KOs. Accordingly,
Since chronic tissue inflammation can cause insulin resistance, WT and GPR120 KO mice were placed on 60% HFD for 15 weeks.
we hypothesized that deletion of GPR120 would enhance the At this point, separate groups of 15 mice each, were treated for
proinflammatory tone, promoting glucose intolerance and five additional weeks with 60% HFD or an isocaloric HFD diet con-
decreased insulin sensitivity. To test this idea, GPR120 taining 27% fish oil supplementation enriched in u-3 FAs. This diet
KO mice and WT littermates were evaluated on normal chow provided 50 and 100 mg of DHA and EPA, respectively, per
diet (NC). Body weights were similar in both groups, and as mouse, per day. Figure 5E shows that administration of the u-3
summarized in Figure 5, glucose tolerance tests (GTT) showed FA diet led to improved insulin sensitivity with increased glucose
a mild degree of impairment in GPR120 KO animals compared infusion rates, enhanced muscle insulin sensitivity (increased IS-
to WTs (Figure 5A). More impressively, insulin secretion was GDR), greater hepatic insulin sensitivity (increased HGP suppres-
more than 2-fold greater in the KO animals, and the combination sion), and decreased hepatic steatosis (Figures S6A and S6B).
of hyperinsulinemia and mild glucose intolerance indicates the Importantly, the u-3 FA diet was completely without effect in the
presence of insulin resistance (Figures 5B and 5C). This was GPR120 KO mice. A separate group of WT mice were treated
confirmed by performing hyperinsulinemic/euglycemic clamp with the insulin sensitizing thiazolidinedione Rosiglitazone, and
studies in the chow fed WT and KO mice (Figure 5D). These the effects of u-3 FAs were equal to or greater (HGP suppression)
studies revealed a 31% decrease in the glucose infusion rate than the effects of this clinically used insulin sensitizing drug.

692 Cell 142, 687–698, September 3, 2010 ª2010 Elsevier Inc.

In addition to improving hepatic insulin sensitivity, u-3 FA was minimal in both groups on HFD (Figure 6A). On the u-3 FA
treatment had a beneficial effect on hepatic lipid metabolism, diet, we observed a decrease in F4/80 staining, along with
causing decreased liver triglycerides, DAGs, along with reduced a marked increase in MGL1 positive cells in WT mice. Impor-
SFA and u-6 FA content in the various lipid classes (Figures tantly, no change in F4/80 or MGL1 staining was noted in the
S6A–S6C and Table S2). The u-3 FA supplementation was GPR120 KO mice on the u-3 FA diet. SVFs were prepared
entirely without effect, or much less effective, at reducing hepatic from adipose tissue and analyzed by flow cytometry to quanti-
lipid levels in the GPR120 KOs. tate the total number of ATMs, as well as the content of
Interestingly, in the absence of u-3 FA supplementation, CD11b+ and CD11c+ and negative macrophage subpopulations
GPR120 KO mice were just as susceptible to HFD-induced (Figure 6B). HFD led to a large but comparable increase in
insulin resistance as were the WT mice. We hypothesize that CD11b+ and CD11c+ ATM content in WT and GPR120 KO
this was because the 60% HFD is relatively deficient of exoge- mice (Figure 6B, middle panel). Treatment with the u-3 FA-en-
nous u-3 FAs, so that ligands for GPR120 were relatively absent riched HFD caused a striking decrease in CD11b+ and CD11c+
in these animals. To assess this, we performed a lipomics anal- ATMs in WT mice, but was without effect in the GPR120 KO
ysis of the various fatty acid classes in the chow and HFD-fed WT group (Figure 6B, right panel). Thus, the FACS analysis was fully
and KO mice. As predicted, circulating concentrations of u-3 consistent with the histological results. Interestingly, CD11c+
FAs were much lower on HFD compared to chow diets, and ATM content was also greater in the GPR120 KOs on the chow
the administration of the u-3 FA supplement to the HFD led to diet relative to WT consistent with the insulin resistance in the
a large increase in plasma u-3 FA content in both genotypes KO animals.
(Figure 5F). This would account for the relative lack of effect of It seemed possible that the reduction in ATM content in WT
GPR120 KO on HFD alone, since u-3 FA ligand stimulation is animals on the u-3 FA diet reflected decreased chemotaxis of
negligible, while the KO animals displayed an insulin resistant macrophages. To test this hypothesis, we measured the migra-
phenotype on chow diets when a moderate level of u-3 FAs tory capacity of IPMacs from WT and GPR120 KO mice using
was provided. Importantly, the GPR120 KO mice are completely an in vitro transwell chemotaxis assay. As seen in Figure 6C,
refractory to the insulin sensitizing effects of u-3 FA administra- macrophages from both groups readily migrated toward condi-
tion on HFD. tioned media (CM) harvested from 3T3-L1 adipocytes. Pretreat-
To address the contribution of macrophages to the overall ment of macrophages with DHA for 3 hr before exposure to CM
in vivo phenotype, we performed bone marrow transplantation led to an 80% inhibition of chemotactic capacity in WT macro-
(BMT) from GPR120 KOs into irradiated WT mice (adoptive phages, but had no significant effect on IPMacs obtained from
transfer) to generate hematopoietic cell deletion of GPR120. the GPR120 KO mice. Similar experiments were performed
The studies in the BMT WT and BMT GPR120 KO mice on using the specific chemokine, monocyte chemotactic protein-1
chow diet revealed a highly significant 20%–30% decrease in (MCP-1) as a chemoattractant, rather than CM, and these exper-
GIR in the KOs, with a more dramatic impairment in the ability iments yielded identical results (Figure 6D). These data indicate
of insulin to suppress hepatic glucose production (Figure S4A). that u-3 FAs cause decreased macrophage chemotaxis by
Thus, the studies in the BMT animals on the chow diet are acting through the GPR120 receptor, contributing to the differ-
comparable to the results (Figure 5D) observed in WT versus ences in ATM content seen in Figures 6A and 6B.
whole body GPR120 KOs on chow diet. When studied on the
HFD ± u-3 FA supplementation (Figure S4B), the u-3 FA supple- Omega-3 FAs Decrease M1 Proinflammatory Gene
mented BMT GPR120 KO animals exhibited a 30% decrease in and Increase M2 Anti-inflammatory Gene Expression
GIR compared to the u-3 FA supplemented BMT WTs. This was in Adipose Tissue
explained by skeletal muscle insulin resistance (decreased IS- As shown in Figure 7A, expression of M1 inflammatory genes
GDR) and hepatic insulin resistance (decreased HGP suppres- such as IL-6, TNF-a, MCP-1, IL-1b, iNOS, and CD11c was
sion) in the GPR120 KOs compared to the WT BMT mice on increased by HFD compared to chow diet in both genotypes,
the u-3 FA supplemented HFD. These data are fully consistent and was reduced in the u-3 FA treated WT mice, but not in
with the results in the global KOs (Figure 5E) and reinforce the the GPR120 KO mice. Even on chow diet, expression of several
concept that the in vivo phenotype we observed can be largely inflammatory genes was higher in GPR120 KOs compared to
traced to hematopoietic cells/macrophages. WT, consistent with the insulin resistance observed in the
chow-fed KO mice. Expression of the M2 anti-inflammatory
Omega-3 FAs Reduce Inflammatory Macrophages genes, arginase 1, IL-10, MGL1, Ym-1, Clec7a, and MMR
in Adipose Tissue was increased by u-3 FAs in WT, but not in the GPR120 KO
We conducted histologic examination of adipose tissue adipose tissue (Figure 7B). These results are consistent with
macrophages (ATMs) from WT and GPR120 KO mice on HFD Figure 6 and demonstrate that the dietary change from HFD
or the u-3 FA enriched HFD by immunostaining for the M1 to u-3 FA supplemented HFD resulted in an overall decreased
macrophage marker F4/80 and the M2 macrophage marker proinflammatory profile in adipose tissue from WT, but not in
MGL1 (Lumeng et al., 2008) (Figure 6A). Consistent with previous GPR120 KO mice. These changes in gene expression were
studies (Weisberg et al., 2003; Xu et al., 2003; Nguyen et al., predominantly manifested in the SVF, except for MCP-1 and
2007), HFD induced a large increase in F4/80 positive ATMs, IL-6, which are known to be readily expressed in adipocytes
which form crown-like structures (CLS) around adipocytes in (Figure S7). Qualitatively similar results were seen in the liver
both WT and GPR120 KO mice. In contrast, MGL1 staining (Figures S6D and S6E).

Cell 142, 687–698, September 3, 2010 ª2010 Elsevier Inc. 693

Figure 6. Omega-3 FA Enriched Diet Decreases
Inflammatory Macrophage Infiltration in Adipose
(A) Confocal merged images from epididymal fat pads
from HFD and u-3 FA enriched HFD (HFD+u3)-fed WT
and GPR120 KO mice, costained with anti-F4/80 (green)
and anti-Caveolin1 (blue) antibodies, left 4 panels, or
anti-MGL1 (green) and anti-Caveolin1 (red) antibodies,
right 4 panels. The image is representative of similar
results from three to four independent experiments. Scale
bar represents 100 mm.
(B) Dot plot representation of CD11b versus CD11c
expression for FACS data obtained from adipose tissue
SVF of NC, HFD or HFD+u3-fed WT and GPR120 KO.
Scattergram is representative from three independent
mice from each group.
(C and D) Migratory capacity of IPMacs from WT and
GPR120 KO mice as measured using an in vitro transwell
chemotaxis assay as described under supplemental
experimental procedures. Data are expressed as mean ±
SEM of three independent experiments in triplicate. *,
p<0.05 versus CM treatment.

of FAs to inhibit both the TLR2/3/4 and the

TNF-a response pathways and cause systemic
insulin sensitization.
GPR120 is a Gaq/11-coupled receptor, and
since it is expressed in enteroendocrine L cells,
past interest in this receptor has focused on its
potential ability to stimulate L cell GLP-1 secre-
tion. In the current study, we show that, in addi-
tion to L cells, GPR120 is highly expressed in
proinflammatory, M1-like macrophages and
mature adipocytes, with negligible expression
in muscle, pancreatic b-cells, and hepatocytes
(Gotoh et al., 2007). In the HFD/obese mouse
model, GPR120 expression is highly induced in
ATMs as well as resident liver macrophages
(Kupffer cells). To explore the biology around
GPR120, we established an artificial reporter
DISCUSSION cell assay and found that the u-3 FAs, DHA, and EPA, are ligands
for GPR120, and comparable to the effects of a non-selective
In this report we show that GPR120 functions as an u-3 FA GPR120 tool compound (GW9508), the u-3 FAs exert potent
receptor/sensor in proinflammatory macrophages and mature anti-inflammatory effects in macrophages. Our results also re-
adipocytes. By signaling through GPR120, DHA and EPA (the vealed the molecular mechanisms underlying these anti-inflam-
major natural u-3 FA constituents of fish oil), mediate potent matory effects. Thus, DHA stimulation of GPR120 inhibits both
anti-inflammatory effects to inhibit both TLR and TNF-a inflam- the TLR2/3/4 and TNF-a proinflammatory cascade. Since acti-
matory signaling pathways. The mechanism of GPR120-medi- vation of IKKb and JNK are common to TLR and TNF-a signaling,
ated anti-inflammation involves inhibition of TAK1 through a this indicates that the locus of GPR120 inhibition is at or proximal
b-arrestin2/TAB1 dependent effect. Since chronic tissue inflam- to these kinases. TAK1 activation stimulates both the IKKb/NFkB
mation is an important mechanism causing insulin resistance and JNK/AP1 pathways, and the TLR and TNF-a signaling path-
(Xu et al., 2003; Shoelson et al., 2007; Schenk et al., 2008), the ways converge at this step. Our data show that stimulation of
anti-inflammatory actions of u-3 FAs exert potent insulin sensi- GPR120 specifically inhibits TAK1 phosphorylation and activa-
tizing effects. The in vivo anti-inflammatory and insulin sensi- tion providing a common mechanism for the inhibition of both
tizing effects of u-3 FAs are dependent on expression of TLR and TNF-a signaling.
GPR120, as demonstrated in studies of obese GPR120 KO Beta-arrestins can serve as important adaptor and scaffold
animals and WT littermates. Thus, GPR120 is highly expressed molecules mediating the functions of a number of different
in proinflammatory macrophages and functions as an u-3 FA GPCRs, as well as other receptor subtypes (Miller and Lefkowitz,
receptor, mediating the anti-inflammatory effects of this class 2001). The C-terminal region of GPR120 contains several

694 Cell 142, 687–698, September 3, 2010 ª2010 Elsevier Inc.

Figure 7. M1 and M2 Inflammatory Gene Expres-
sion Levels in Adipose Tissue from WT versus
GPR120 KO Mice
Relative mRNA levels for M1 proinflammatory genes (A)
and M2 anti-inflammatory genes (B) in NC, HFD, or
HFD+u3 (+u3)-fed WT and GPR120 KO mice, as measured
by q-PCR. Data are expressed as mean ± SEM of three
independent experiments in triplicate. n=7 per group, *,
p<0.05 compared to the HFD-fed WT group. **, p<0.05
compared to the WT versus GPR120 KO on NC. See also
Figure S7. Primer sequences are shown in Table S1.

11-coupled receptor in other contexts. This

provides further evidence demonstrating the
concept that a single GPCR can independently
signal through multiple pathways. In previous
studies, we have demonstrated that Gaq/11
activation can lead to stimulation of GLUT4
translocation in adipocytes. Since GPR120
was expressed in mature adipocytes, but not
preadipocytes, we explored the potential role
of GPR120 in glucose transport control. Inter-
estingly, we found that DHA stimulation of
GPR120 in 3T3-L1 adipocytes increased
GLUT4 translocation to the cell surface with
a subsequent increase in glucose transport
into the cells. RNA interference studies showed
putative b-arrestin2 binding motifs [(S/T)X4-5(S/T); Cen et al., that the DHA effect on glucose uptake was GPR120, GLUT4, and
2001], but whether b-arrestins play any role in GPR120 function Gaq/11 dependent, but independent of b-arrestin2. This effect
was unknown. Here we find that activation of GPR120 by DHA was about 30%–50% as great as the effect of insulin and the
stimulation leads to association of the receptor with b-arrestin2, actions of DHA on glucose uptake were additive to those of
but not b-arrestin1, and that the anti-inflammatory effects of a submaximally stimulating concentration of insulin. From this,
GPR120 are completely b-arrestin2 dependent. Functional it is possible to propose that these insulinometic effects con-
immunocytochemical studies showed that DHA stimulation tribute to the overall insulin sensitizing actions of u-3 FAs.
leads to recruitment of b-arrestin2 to the plasma membrane However, muscle glucose uptake accounts for the great majority
where it colocalizes with GPR120. This is followed by receptor (70%–80%) of insulin stimulated glucose disposal. Furthermore,
and b-arrestin2 internalization, where the two are now colocal- GPR120 is not expressed in muscle, and DHA did not stimulated
ized in the cytoplasmic compartment. TAB1 is the activating glucose uptake in L6 myocytes (Figures S4C and S4D). In addi-
protein for TAK1 and our results show that following DHA-stim- tion, acute administration of DHA had no stimulatory effects on
ulated internalization of the GPR120/b-arrestin2 complex, IS-GDR (Figure S4E). This reports the conclusion that the
b-arrestin2 can now associate with TAB1, as measured in coim- in vivo stimulatory effects of DHA on GDR are related to
munoprecipitation experiments; only full-length b-arrestin2 was anti-inflammation, and that the glucose transport stimulatory
capable of interacting with GPR120 and TAB1. This apparently effects in adipocytes contribute little to the overall phenotype.
blocks the association of TAB1 with TAK1, inhibiting TAK1 Since chronic, low grade tissue inflammation is an important
activation and downstream signaling to the IKKb/NFkB and cause of obesity-related insulin resistance, we reasoned that
JNK/AP1 system. These results provide a mechanism for the the anti-inflammatory effects of GPR120 stimulation should be
b-arrestin2-mediated inhibition of TLR4, TNF-a, and TLR2/3 coupled to insulin sensitizing actions in vivo. This idea was
action. Other studies in the literature are consistent with these confirmed in studies of WT and GPR120 KO mice. On a chow
findings, since it has been shown that b-arrestin2 can inhibit diet, the lean GPR120 KO mice were glucose intolerant, hyperin-
NFkB signaling in other systems (Gao et al., 2004; Wang et al., sulinemic and displayed decreased skeletal muscle and hepatic
2006b). Furthermore, Lefkowitz’ group has recently published insulin sensitivity, as measured during glucose clamp studies.
an extensive proteomics analysis of b-arrestin2 interacting They also displayed increased ATM content, relative to WT
partners, and among the 266 proteins they identified, TAB1 mice, and a 2- to 5-fold higher expression level of the M1 proin-
was represented on the list (Xiao et al., 2007). flammatory markers, MCP-1, iNOS, and IL-6 (Figure 7A). On
Interestingly, the anti-inflammatory effects mediated by HFD, GPR120 KO and WT mice became equally obese and
GPR120 were entirely dependent on b-arrestin2, but indepen- insulin resistant. Importantly, u-3 FA supplementation markedly
dent of Gaq/11, despite the fact that GPR120 can be a Gaq/ increased insulin sensitivity in WT mice but was without effect in

Cell 142, 687–698, September 3, 2010 ª2010 Elsevier Inc. 695

the GPR120 KOs. Consistent with these results, u-3 FA treat- tion versus anti-inflammatory actions are distinct processes, and
ment led to a decrease in ATM accumulation with reduced it is certainly possible that the products derived from u-3 FA
adipose tissue markers of inflammation in WT, but not in KO metabolism work on the former but not the latter.
mice. In addition to direct anti-inflammatory effects in macro- In summary, we have found that GPR120 functions as an u-3
phages, DHA treatment inhibited the ability of primary WT FA receptor/sensor and mediates robust and broad anti-inflam-
macrophages to migrate toward adipocyte CM. This could be matory effects, particularly in macrophages. After ligand stimula-
due to DHA-induced decreased chemokine secretion or down- tion, GPR120 couples to b-arrestin2 which is followed by
regulation of chemokine receptors, or both. In addition, it is receptor endocytosis and inhibition of TAB1-mediated activation
possible that DHA, by signaling through GPR120, can mediate of TAK1, providing a mechanism for inhibition of both the TLR
heterologous desensitization of other GPCR chemokine recep- and TNF-a proinflammatory signaling pathways. Since chronic
tors. We also observed a concomitant increase in M2 markers, tissue inflammation is linked to insulin resistance in obesity, we
such as IL-10, arginase 1, MGL1, Ym-1, Clec7a, and MMR. used GPR120 KO mice to demonstrate that u-3 FAs cause
This latter finding raises the possibility that u-3 FAs can redirect GPR120-mediated anti-inflammatory and insulin sensitizing
ATMs from an M1 to an M2 polarization state. Taken together, effects in vivo. Overall, these results strongly argue that anti-
these mechanisms account for the decreased inflammatory inflammatory effects can ameliorate insulin resistance in obesity.
state. The in vivo anti-inflammatory actions of u-3 FAs are Taken together, GPR120 emerges as an important control point
consistent with the insulin sensitizing effects of these agents in the integration of anti-inflammatory and insulin sensitizing
and are fully dependent on the presence of GPR120, indicating responses, which may prove useful in the future development
a causal relationship. Finally, the adoptive transfer studies of new therapeutic approaches for the treatment of insulin resis-
showed that hematopoietic cell GPR120 deletion results in tant diseases.
a comparable insulin resistant, u-3 FA non-responsive pheno-
type as seen in the global GPR120 KOs, indicating that this EXPERIMENTAL PROCEDURES
phenotype can be traced back to inflammatory events in
macrophages. Chemicals and Reagents
GW9508 was purchased from Tocris bioscience (Ellisville, MO) and DHA was
We also performed a detailed in vivo lipidomic analysis of FAs
from Cayman chemical (Ann Arbor, MI). All other chemicals were purchased
in the different lipid classes in the liver (Table S2). The results from Sigma unless mentioned otherwise.
showed that HFD leads to an increase in total TAGs, DAGs, total
SFAs, monounsaturated FAs and u-6 FAs in WT mice, while all of Animal Care and Use
these lipid changes are ameliorated with u-3 FA treatment. In the Male C57Bl/6 or GPR120 KO littermates were fed a normal chow (13.5% fat;
GPR120 KO mice, all of these lipids are elevated on HFD to the LabDiet) or high-fat diet (60% fat; Research Diet) ad libitum for 15–20 weeks
same extent as in WT mice, but, u-3 FA supplementation was from 8 weeks of age. GPR120 KO mice and WT littermates were provided
by Taconic Inc. (Hudson, NY). After 15 weeks on HFD, WT and GPR120 KO
either ineffective or much less effective. These results are
mice were switched to an isocaloric HFD-containing 27% menhaden fish oil
consistent with the view that the reversal of steatosis/non- replacement (wt/wt; menhaden fish oil: 16% EPA (C20:5n3), 9%, DHA
alcoholic fatty liver disease (NAFLD) by u-3 FA treatment is (C22:6n3), Research Diet) (Jucker et al., 1999; Neschen et al., 2007) and fed
mediated, in large part, by GPR120 and that the GPR120 KO for 5 weeks. Mice received fresh diet every 3rd day, and food consumption
mice are predisposed toward NAFLD even in the context of and body weight were monitored. Animals were housed in a specific path-
u-3 FA supplementation. ogen-free facility and given free access to food and water. All procedures
were approved by the University of California, San Diego animal care and
Dietary DHA is rapidly esterified into chylomicrons during the
use committee. In vivo metabolic studies were performed as described under
process of gastrointestinal absorption, and is also packaged
supplemental experimental procedures.
into VLDL triglycerides by the liver. DHA can also be esterified
into phospholipids and cholesterol esters associated with Data Analysis
circulating lipopoproteins and only a small proportion (5%) of Densitometric quantification and normalization were performed using the
total plasma DHA is found in the FFA pool. Through the action ImageJ 1.42q software. The values presented are expressed as the means ±
of lipoprotein lipase bound to the luminal surface of endothelial SEM. The statistical significance of the differences between various treat-
cells, u-3 FAs are cleaved from circulating triglycerides where ments was determined by one-way ANOVA with the Bonferroni correction
using GraphPad Prism 4.0 (San Diego, CA). The p < 0.05 was considered
they can act as ligands or be taken up by peripheral tissues
(Polozova and Salem Jr., 2007). Recent studies have also indi-
cated that metabolic products derived from u-3 FAs, such as
17S-hydroxy-DHA, resolvins, and protectins may play a role in
the long term resolution of inflammation and this might attenuate Supplemental Information includes Extended Experimental Procedures, seven
insulin resistance in the context of obesity (González-Périz et al., figures, and two tables and can be found with this article online at doi:10.1016/
2009). If this proves to be correct, then this could provide an j.cell.2010.07.041.
additional mechanism for long term u-3 FA-induced anti-inflam-
matory, insulin sensitizing effects. However, in the current ACKNOWLEDGMENTS

studies, we found that these u-3 FA derivatives were unable to

We thank Jachelle M. Ofrecio and Sarah Nalbandian for their help with animal
stimulate GPR120 activation in our reporter cell assay (data not maintenance and Elizabeth J. Hansen for editorial assistance. We are grateful
shown), indicating that GPR120 functions as a receptor for u-3 to Dr. Robert Lefkowitz (Howard Hughes Medical Institute, Duke University) for
FAs and not their biochemical products. Resolution of inflamma- the gift of FLAG-tagged serial mutant b-arrestin2 constructs and to Dr. Maziyar

696 Cell 142, 687–698, September 3, 2010 ª2010 Elsevier Inc.

Saberi at NGM Bio Inc. (San Francisco, CA) for GLP-1 measurements. We Jucker, B.M., Cline, G.W., Barucci, N., and Shulman, G.I. (1999). Differential
thank the Flow Cytometry Resource and Neal Sekiya for assistance with effects of safflower oil versus fish oil feeding on insulin-stimulated glycogen
FACS analysis at the VA San Diego hospital, the UCSD Histology Core lab synthesis, glycolysis, and pyruvate dehydrogenase flux in skeletal muscle:
for technical help with processing liver specimens, and UCSD Microscope a 13C nuclear magnetic resonance study. Diabetes 48, 134–140.
Resource for microscopy analysis. This study was funded in part by the
Kawai, T., and Akira, S. (2006). TLR signaling. Cell Death Differ. 13, 816–825.
National Institutes of Health grants NIDDK DK033651 (J.M.O.), DK063491
(J.M.O.), DK 074868 (J.M.O.), and the Eunice Kennedy Shriver NICHD/NIH Lee, J.Y., Plakidas, A., Lee, W.H., Heikkinen, A., Chanmugam, P., Bray, G., and
through a cooperative agreement U54 HD 012303-25 as part of the specialized Hwang, D.H. (2003). Differential modulation of Toll-like receptors by fatty
Cooperative Centers Program in Reproduction and Infertility Research. acids: preferential inhibition by n-3 polyunsaturated fatty acids. J. Lipid Res.
44, 479–486.
Received: January 20, 2010 Lumeng, C.N., DelProposto, J.B., Westcott, D.J., and Saltiel, A.R. (2008).
Revised: May 24, 2010 Phenotypic switching of adipose tissue macrophages with obesity is gener-
Accepted: July 19, 2010 ated by spatiotemporal differences in macrophage subtypes. Diabetes 57,
Published: September 2, 2010 3239–3246.
Luttrell, L.M., Ferguson, S.S., Daaka, Y., Miller, W.E., Maudsley, S., Della
REFERENCES Rocca, G.J., Lin, F., Kawakatsu, H., Owada, K., Luttrell, D.K., et al. (1999).
Beta-arrestin-dependent formation of beta2 adrenergic receptor-Src protein
Barak, L.S., Ferguson, S.S., Zhang, J., and Caron, M.G. (1997). A beta- kinase complexes. Science 283, 655–661.
arrestin/green fluorescent protein biosensor for detecting G protein-coupled Luttrell, L.M., and Lefkowitz, R.J. (2002). The role of beta-arrestins in the
receptor activation. J. Biol. Chem. 272, 27497–27500. termination and transduction of G-protein-coupled receptor signals. J. Cell
Baron, A.D., Brechtel, G., Wallace, P., and Edelman, S.V. (1988). Rates and Sci. 115, 455–465.
tissue sites of non-insulin- and insulin-mediated glucose uptake in humans. Miller, W.E., and Lefkowitz, R.J. (2001). Expanding roles for beta-arrestins as
Am. J. Physiol. 255, E769–E774. scaffolds and adapters in GPCR signaling and trafficking. Curr. Opin. Cell Biol.
Briscoe, C.P., Peat, A.J., McKeown, S.C., Corbett, D.F., Goetz, A.S., Littleton, 13, 139–145.
T.R., McCoy, D.C., Kenakin, T.P., Andrews, J.L., Ammala, C., et al. (2006).
Neschen, S., Morino, K., Dong, J., Wang-Fischer, Y., Cline, G.W., Romanelli,
Pharmacological regulation of insulin secretion in MIN6 cells through the fatty
A.J., Rossbacher, J.C., Moore, I.K., Regittnig, W., Munoz, D.S., et al. (2007).
acid receptor GPR40: identification of agonist and antagonist small molecules.
n-3 Fatty acids preserve insulin sensitivity in vivo in a peroxisome
Br. J. Pharmacol. 148, 619–628.
proliferator-activated receptor-alpha-dependent manner. Diabetes 56,
Buettner, R., Parhofer, K.G., Woenckhaus, M., Wrede, C.E., Kunz-Schughart, 1034–1041.
L.A., Scholmerich, J., and Bollheimer, L.C. (2006). Defining high-fat-diet rat
Nguyen, M.T., Favelyukis, S., Nguyen, A.K., Reichart, D., Scott, P.A., Jenn, A.,
models: metabolic and molecular effects of different fat types. J. Mol. Endocri-
Liu-Bryan, R., Glass, C.K., Neels, J.G., and Olefsky, J.M. (2007). A subpopula-
nol. 36, 485–501.
tion of macrophages infiltrates hypertrophic adipose tissue and is activated by
Calder, P.C. (2005). Polyunsaturated fatty acids and inflammation. Biochem. free fatty acids via Toll-like receptors 2 and 4 and JNK-dependent pathways. J.
Soc. Trans. 33, 423–427. Biol. Chem. 282, 35279–35292.
Cen, B., Xiong, Y., Ma, L., and Pei, G. (2001). Direct and differential interaction Oh, D.Y., Song, J.A., Moon, J.S., Moon, M.J., Kim, J.I., Kim, K., Kwon, H.B.,
of beta-arrestins with the intracellular domains of different opioid receptors. and Seong, J.Y. (2005). Membrane-proximal region of the carboxyl terminus
Mol. Pharmacol. 59, 758–764. of the gonadotropin-releasing hormone receptor (GnRHR) confers differential
Gao, H., Sun, Y., Wu, Y., Luan, B., Wang, Y., Qu, B., and Pei, G. (2004). signal transduction between mammalian and nonmammalian GnRHRs. Mol.
Identification of beta-arrestin2 as a G protein-coupled receptor-stimulated Endocrinol. 19, 722–731.
regulator of NF-kappaB pathways. Mol. Cell 14, 303–317. Polozova, A., and Salem, N., Jr. (2007). Role of liver and plasma lipoproteins
Gether, U. (2000). Uncovering molecular mechanisms involved in activation of in selective transport of n-3 fatty acids to tissues: a comparative study of
G protein-coupled receptors. Endocr. Rev. 21, 90–113. 14C-DHA and 3H-oleic acid tracers. J. Mol. Neurosci. 33, 56–66.
González-Périz, A., Horrillo, R., Ferre, N., Gronert, K., Dong, B., Moran- Schenk, S., Saberi, M., and Olefsky, J.M. (2008). Insulin sensitivity: modulation
Salvador, E., Titos, E., Martinez-Clemente, M., Lopez-Parra, M., Arroyo, V., by nutrients and inflammation. J. Clin. Invest. 118, 2992–3002.
and Claria, J. (2009). Obesity-induced insulin resistance and hepatic steatosis
Schulte, G., and Fredholm, B.B. (2003). Signaling from adenosine receptors to
are alleviated by omega-3 fatty acids: a role for resolvins and protectins.
mitogen-activated protein kinases. Cell Signal 15, 813–827.
FASEB J. 23, 1946–1957.
Shoelson, S.E., Herrero, L., and Naaz, A. (2007). Obesity, inflammation, and
Gotoh, C., Hong, Y.H., Iga, T., Hishikawa, D., Suzuki, Y., Song, S.H., Choi,
insulin resistance. Gastroenterology 132, 2169–2180.
K.C., Adachi, T., Hirasawa, A., Tsujimoto, G., et al. (2007). The regulation of
adipogenesis through GPR120. Biochem. Biophys. Res. Commun. 354, Solinas, G., Vilcu, C., Neels, J.G., Bandyopadhyay, G.K., Luo, J.L., Naugler,
591–597. W., Grivennikov, S., Wynshaw-Boris, A., Scadeng, M., Olefsky, J.M., and
Karin, M. (2007). JNK in hematopoietically derived cells contributes to diet-
Hirasawa, A., Tsumaya, K., Awaji, T., Katsuma, S., Adachi, T., Yamada, M.,
induced inflammation and insulin resistance without affecting obesity. Cell
Sugimoto, Y., Miyazaki, S., and Tsujimoto, G. (2005). Free fatty acids regulate
Metab. 6, 386–397.
gut incretin glucagon-like peptide-1 secretion through GPR120. Nat. Med. 11,
90–94. Takaesu, G., Surabhi, R.M., Park, K.J., Ninomiya-Tsuji, J., Matsumoto, K., and
Gaynor, R.B. (2003). TAK1 is critical for IkappaB kinase-mediated activation of
Imamura, T., Vollenweider, P., Egawa, K., Clodi, M., Ishibashi, K., Nakashima,
the NF-kappaB pathway. J. Mol. Biol. 326, 105–115.
N., Ugi, S., Adams, J.W., Brown, J.H., and Olefsky, J.M. (1999). G alpha-q/11
protein plays a key role in insulin-induced glucose transport in 3T3-L1 adipo- Tazoe, H., Otomo, Y., Kaji, I., Tanaka, R., Karaki, S.I., and Kuwahara, A. (2008).
cytes. Mol. Cell. Biol. 19, 6765–6774. Roles of short-chain fatty acids receptors, GPR41 and GPR43 on colonic func-
Itoh, Y., Kawamata, Y., Harada, M., Kobayashi, M., Fujii, R., Fukusumi, S., Ogi, tions. J. Physiol. Pharmacol. 59 (Suppl 2), 251–262.
K., Hosoya, M., Tanaka, Y., Uejima, H., et al. (2003). Free fatty acids regulate Ulloa-Aguirre, A., Stanislaus, D., Janovick, J.A., and Conn, P.M. (1999).
insulin secretion from pancreatic beta cells through GPR40. Nature 422, Structure-activity relationships of G protein-coupled receptors. Arch. Med.
173–176. Res. 30, 420–435.

Cell 142, 687–698, September 3, 2010 ª2010 Elsevier Inc. 697

Wang, J., Wu, X., Simonavicius, N., Tian, H., and Ling, L. (2006a). Medium- Xiao, K., McClatchy, D.B., Shukla, A.K., Zhao, Y., Chen, M., Shenoy, S.K.,
chain fatty acids as ligands for orphan G protein-coupled receptor GPR84. Yates, J.R., 3rd, and Lefkowitz, R.J. (2007). Functional specialization of
J. Biol. Chem. 281, 34457–34464. beta-arrestin interactions revealed by proteomic analysis. Proc. Natl. Acad.
Wang, Y., Tang, Y., Teng, L., Wu, Y., Zhao, X., and Pei, G. (2006b). Association Sci. USA 104, 12011–12016.
of beta-arrestin and TRAF6 negatively regulates Toll-like receptor-interleukin 1
receptor signaling. Nat. Immunol. 7, 139–147. Xu, H., Barnes, G.T., Yang, Q., Tan, G., Yang, D., Chou, C.J., Sole, J., Nichols,
Weisberg, S.P., McCann, D., Desai, M., Rosenbaum, M., Leibel, R.L., and A., Ross, J.S., Tartaglia, L.A., and Chen, H. (2003). Chronic inflammation in fat
Ferrante, A.W., Jr. (2003). Obesity is associated with macrophage accumula- plays a crucial role in the development of obesity-related insulin resistance. J.
tion in adipose tissue. J. Clin. Invest. 112, 1796–1808. Clin. Invest. 112, 1821–1830.

698 Cell 142, 687–698, September 3, 2010 ª2010 Elsevier Inc.

Anti-CD47 Antibody Synergizes with
Rituximab to Promote Phagocytosis and
Eradicate Non-Hodgkin Lymphoma
Mark P. Chao,1,10,* Ash A. Alizadeh,1,2,3,10 Chad Tang,1 June H. Myklebust,3,9 Bindu Varghese,3 Saar Gill,5 Max Jan,1
Adriel C. Cha,1 Charles K. Chan,1 Brent T. Tan,4 Christopher Y. Park,1,4 Feifei Zhao,1 Holbrook E. Kohrt,2,3
Raquel Malumbres,6 Javier Briones,7 Randy D. Gascoyne,8 Izidore S. Lossos,6 Ronald Levy,3 Irving L. Weissman,1,4,10
and Ravindra Majeti1,2,10
1Institutefor Stem Cell Biology and Regenerative Medicine, Stanford Cancer Center, and Ludwig Center at Stanford
2Department of Internal Medicine, Division of Hematology
3Division of Oncology
4Department of Pathology
5Division of Blood and Bone Marrow Transplantation

Stanford University, Palo Alto, CA 94304, USA

6Department of Medicine, Division of Hematology-Oncology, University of Miami Miller School of Medicine, Miami, FL 33136, USA
7Department of Hematology, Hospital Santa Creu i Sant Pau, Autonomous University of Barcelona, Barcelona 08193, Spain
8Department of Pathology, British Columbia Cancer Agency, Vancouver V5Z 1L3, Canada
9Department of Immunology, Institute for Cancer Research, Oslo University Hospital/Centre for Cancer Biomedicine, University of Oslo,

Oslo 0310, Norway

10These authors contributed equally to this work

DOI 10.1016/j.cell.2010.07.044

SUMMARY modality for cancer treatment (Adams and Weiner, 2005).

Although therapies combining a mAb with chemotherapeutic
Monoclonal antibodies are standard therapeutics for agents are effective in several human cancers, antibodies alone
several cancers including the anti-CD20 antibody are not curative. Antibodies effective against cancer are believed
rituximab for B cell non-Hodgkin lymphoma (NHL). to function by several mechanisms including antibody-depen-
Rituximab and other antibodies are not curative dent cellular cytotoxicity (ADCC), stimulation of complement-
and must be combined with cytotoxic chemotherapy dependent cytotoxicity (CDC), inhibition of signal transduction,
or direct induction of apoptosis (Cheson and Leonard, 2008).
for clinical benefit. Here we report the eradication
Non-Hodgkin lymphoma (NHL) is the fifth most common
of human NHL solely with a monoclonal antibody cancer in the United States consisting of indolent and aggressive
therapy combining rituximab with a blocking anti- subtypes with a 5 year overall survival ranging from 25%–75%
CD47 antibody. We identified increased expression (The International Non-Hodgkin’s Lymphoma Prognostic
of CD47 on human NHL cells and determined that Factors Project, 1993). The anti-CD20 antibody, rituximab (Rit-
higher CD47 expression independently predicted uxan), is a standard therapy for many CD20-positive B cell
adverse clinical outcomes in multiple NHL subtypes. lymphomas and significantly improves long-term survival in
Blocking anti-CD47 antibodies preferentially enabled combination with conventional chemotherapy (Cheson and Leo-
phagocytosis of NHL cells and synergized with ritux- nard, 2008). As a single agent or in combination with chemo-
imab. Treatment of human NHL-engrafted mice with therapy, rituximab is not curative in the majority of B cell NHL
anti-CD47 antibody reduced lymphoma burden and patients and rituximab resistance has been observed (reviewed
in Cheson and Leonard, 2008). Multiple lines of evidence indicate
improved survival, while combination treatment
that rituximab acts at least in part by engaging Fc receptors
with rituximab led to elimination of lymphoma and (FcRs) on immune effector cells, such as NK cells and macro-
cure. These antibodies synergized through a mecha- phages, and stimulating effector functions such as ADCC (Glen-
nism combining Fc receptor (FcR)-dependent and nie et al., 2007; Nimmerjahn and Ravetch, 2007). Although resis-
FcR-independent stimulation of phagocytosis that tance has been reported to occur through several mechanisms
might be applicable to many other cancers. (Cartron et al., 2004), there has been limited development of
agents that can overcome this resistance.
INTRODUCTION Immune effector cells, including NK cells and phagocytes, are
critical to the efficacy of many anticancer antibodies. Phagocytic
Emerging evidence has demonstrated that monoclonal anti- cells, including macrophages and dendritic cells, express signal
bodies (mAbs) either alone or in combination are an effective regulatory protein alpha (SIRPa), which binds CD47, a widely

Cell 142, 699–713, September 3, 2010 ª2010 Elsevier Inc. 699

expressed transmembrane protein (Brown and Frazier, 2001). based on the presumed cell of origin of tumors: normal germinal
CD47-mediated activation of SIRPa initiates a signal transduc- center B cells (GCB-like), which are associated with a favorable
tion cascade resulting in inhibition of phagocytosis (reviewed in clinical outcome, or activated blood memory B cells (ABC-like),
Jaiswal et al., 2010). In identifying a role for CD47 in cancer path- which are associated with a poor clinical outcome (Alizadeh
ogenesis, we previously demonstrated that forced expression of et al., 2000; Rosenwald et al., 2002). CD47 expression was sig-
mouse CD47 on a human leukemia cell line facilitated tumor nificantly higher in ABC-like DLBCL (Figure 1F and Figure S1E).
engraftment in immunodeficient mice through the evasion of CD47 expression was not found to have independent prog-
phagocytosis (Jaiswal et al., 2009). We further demonstrated nostic value within GCB and ABC subtypes, suggesting a strong
that this mechanism could be targeted therapeutically in human association with the cell-of-origin classification of DLBCL.
acute myeloid leukemia (AML) with a blocking anti-CD47 anti- Higher CD47 expression was also associated with unmutated
body that enabled phagocytosis and eliminated AML stem cells immunoglobulin heavy chain variable regions (IgVH) in CLL
(Majeti et al., 2009). Based on this antibody mechanism, we (Figure 1G and Figure S1E) and significantly correlated with
hypothesized that the combination of a blocking anti-CD47 the proliferative index in MCL (Figure 1H), both known adverse
antibody with a second FcR-activating antibody would both prognostic factors (Katzenberger et al., 2006; Oscier et al.,
prevent an inhibitory signal and deliver a positive stimulus, 2002; Rosenwald et al., 2003). In multivariate analyses, CD47
resulting in the synergistic phagocytosis and elimination of tar- expression remained prognostic of disease progression inde-
get cells. Here, we tested this antibody synergy hypothesis by pendent from the international prognostic index in DLBCL (The
investigating the combination of a blocking anti-CD47 mAb International Non-Hodgkin’s Lymphoma Prognostic Factors
with rituximab against human NHL. Project, 1993) and two major prognostic factors in CLL: IgVH
mutation status and ZAP-70 status (Figure S1G). Within the
RESULTS small MCL cohort, a multivariate model did not find independent
prognostic value for CD47 when considering the proliferation
CD47 Expression Is Increased on NHL Cells Compared index (data not shown).
to Normal B Cells
We examined CD47 protein expression on primary human NHL Blocking Anti-CD47 Antibodies Enable Phagocytosis
samples and normal B cells by flow cytometry. Compared to of NHL Cells by Macrophages and Synergize
both normal peripheral blood and germinal center B cells, with Rituximab In Vitro
CD47 was more highly expressed on a large subset of primary We first tested the ability of anti-human CD47 antibodies to
patient samples from multiple B cell NHL subtypes (Figure 1A enable phagocytosis of human NHL cell lines, primary NHL cells,
and Figure S1A available online), including diffuse large B cell and normal peripheral blood (NPB) cells by human macrophages
lymphoma (DLBCL), B cell chronic lymphocytic leukemia in vitro. Incubation of NHL cells in the presence of IgG1 isotype
(B-CLL), mantle cell lymphoma (MCL), follicular lymphoma (FL), control or anti-CD45 IgG1 antibody did not result in significant
marginal zone lymphoma (MZL), and pre-B acute lymphoblastic phagocytosis; however, two different blocking anti-CD47 anti-
leukemia (pre-B ALL). Across NHL subtypes, we found differing bodies (B6H12.2 and BRIC126) enabled phagocytosis of NHL
levels of CD47 expression that also varied within each NHL cells but not NPB cells (Figures 2A and 2B).
subtype (Figure 1B). Next, we repeated the in vitro phagocytosis assays with
mouse macrophages. Incubation of NHL cells in the presence
Increased CD47 Expression Correlates with a Worse of IgG1 isotype control or anti-CD45 IgG1 antibody did not result
Clinical Prognosis and Adverse Molecular Features in significant phagocytosis; however, phagocytosis of NHL cells
in Multiple NHL Subtypes was observed with blocking antibodies to CD47 (B6H12.2 and
Having previously shown a correlation between CD47 mRNA BRIC126), whereas no phagocytosis was observed with a non-
and protein expression (Majeti et al., 2009), we assessed CD47 blocking antibody (2D3) (Figure 2B). Disruption of the CD47-
mRNA expression across NHL subtypes for associations with SIRPa interaction with an anti-mouse SIRPa antibody also
morphologic and molecular subgroups using gene expression resulted in significant phagocytosis (Figure 2B).
data from previously described patient cohorts (Table S1) and Given variable expression of CD47 on primary NHL, we inves-
investigated the prognostic implications of increased CD47 tigated by two independent methods whether CD47 expression
expression in disease outcome. Higher CD47 expression was levels correlated with the degree of anti-CD47 antibody-medi-
associated with adverse prognosis in DLBCL, B-CLL, and MCL ated phagocytosis. First, lentiviral shRNA vectors were used to
(Figures 1C–1E). In patients with DLBCL, whether treated with knock down expression of CD47 in Raji cells. Several clones
or without rituximab-based combination chemotherapy (Fig- were generated with a range of CD47 knockdown (Figures S2A
ure 1C and Figure S1B), higher CD47 expression was signifi- and S2B). Those clones with a greater than 50% reduction in
cantly associated with risk of death. This increased risk was CD47 expression (shCD47-1 and shCD47-2) demonstrated
largely due to disease progression (Figure S1C), a finding vali- a significant reduction in anti-CD47 antibody-mediated phago-
dated in an independent cohort of patients using quantitative cytosis (Figure S3C). In the second approach, a statistical anal-
RT-PCR on fixed archival specimens (Figure S1D). ysis demonstrated a positive correlation between CD47 expres-
We next investigated whether increased CD47 expression was sion and degree of anti-CD47 antibody-mediated phagocytosis
associated with known adverse molecular features in NHL. with both mouse and human macrophage effector cells
In DLBCL, prior studies have identified two distinct subgroups (Figure S2D).

700 Cell 142, 699–713, September 3, 2010 ª2010 Elsevier Inc.

Figure 1. CD47 Expression Is Increased on NHL Cells Compared to Normal B Cells, Confers a Worse Clinical Prognosis, and Correlates with
Adverse Molecular Features in Multiple NHL Subtypes
(A) CD47 expression on normal peripheral blood (PB) B cells (CD19+), normal germinal center (GC) B cells (CD3CD5CD20+CD10+CD38+), and NHL B cells
(CD19+) was determined by flow cytometry, and mean fluorescence intensity was normalized for cell size. Each point represents a different patient sample:
DLBCL = 2, CLL = 15, MCL = 4, FL = 6, MZL = 2, and pre-B ALL = 1 (****p < 0.0001). Normalized mean expression (and range) for each population were: normal
PB B cells 420.9 (267.3–654.0), normal GC B cells 482.5 (441.1–519.9), and NHL 888.5 (270.1–1553).
(B) CD47 expression across NHL subtypes including DLBCL (DL, n = 15), MCL (n = 34), FL (n = 28), and B-CLL (n = 14) was determined as in (A). Normalized mean
expression (and range) for each population were: DL 725.7 (261.2–1344), MCL 1055 (444.2–2196), FL 825.1 (283.6–1546), CLL 713.6 (432.8–1086) (*p < 0.05).
(C–E) Prognostic influence of CD47 mRNA expression is shown on overall (C and E) and event-free (D) survival of patients with DLBCL, B-CLL, and MCL.
For DLBCL and CLL, stratification into low and high CD47 expression groups was based on an optimal threshold determined by microarray analysis; this cutpoint
was internally validated for both DLBCL and CLL and also externally validated in an independent DLBCL cohort. For MCL, stratification relative to the median was
employed as an optimal cutpoint could not be defined. Significance measures are based on log-likelihood estimates of the p value, when treating CD47 expres-
sion as a continuous variable, with corresponding dichotomous indices also provided in Table S1.
(F–H) CD47 mRNA expression is shown in relation to cell-of-origin classification for DLBCL (F), immunoglobulin heavy chain mutation status (IgVH) for CLL (G),
and proliferation index for MCL (H). Error bars represent upper and lower quartiles (F and G). Analyses for (C)–(H) employed publicly available datasets for NHL
patients (Table S1). NGC = normal germinal center, ABC = activated B cell-like, GCB = germinal center B cell-like.
See also Figure S1 and Table S1.

Cell 142, 699–713, September 3, 2010 ª2010 Elsevier Inc. 701

Figure 2. Blocking Antibodies against CD47 Enable Phagocytosis of NHL Cells by Macrophages and Synergize with Rituximab In Vitro
(A) CFSE-labeled NHL cells were incubated with human macrophages and the indicated antibodies and examined by immunofluorescence microscopy to detect
phagocytosis (arrows). Photomicrographs from a representative NHL sample are shown.
(B) Phagocytic indices of primary human NHL cells (blue), normal peripheral blood (NPB) cells (red), and NHL cell lines (purple, orange, and green) were deter-
mined using human (left) and mouse (middle) macrophages. Antibody-induced apoptosis (right panel) was tested by incubating NHL cells with the indicated anti-
bodies or staurosporine without macrophages and assessing the percentage of apoptotic and dead cells (% annexin V and/or PI positive).

702 Cell 142, 699–713, September 3, 2010 ª2010 Elsevier Inc.

It has been reported that immobilized or crosslinked anti- planted into SCID mice (Figure S3). In addition to these cell lines,
bodies against CD47 induce apoptosis of primary human we identified a primary NHL patient specimen that engrafted in
B-CLL cells, as well as several malignant lymphoid cell lines NSG mice in the bone marrow upon intravenous transplantation
(Kikuchi et al., 2004, 2005; Mateo et al., 1999; Uno et al., (Figures S5A and S5B). As with the cell lines, ex vivo coating of
2007). Therefore, anti-CD47 antibodies might be predicted to these primary cells with anti-CD47 antibody, but not controls
directly induce apoptosis of NHL cells that are then recognized (Figure 3G), resulted in complete inhibition of bone marrow
by macrophages and phagocytosed. Contrary to this prediction, engraftment (Figure 3H).
when NHL cells were incubated with anti-CD47 antibody in
the absence of macrophages, no induction of apoptosis was Combination Therapy with Anti-CD47 Antibody and
observed when cells were incubated in suspension for either Rituximab Eliminates Lymphoma in Both Disseminated
2 hr (Figure 2B, right) or 8 hr (Figures S2E and S2F). Incubation and Localized Human NHL Xenotransplant Models
of NHL cells with immobilized anti-CD47 antibody resulted in In the second treatment strategy, mice were first engrafted with
increased apoptosis compared to controls (Figures S2G and NHL and then administered single or combination antibody
S2H), consistent with prior reports (Mateo et al., 1999). Given therapy. To best model NHL, we established disseminated and
that phagocytosis of NHL cells occurs in the presence of soluble localized xenotransplantation models in NSG mice that are defi-
anti-CD47 mAbs, it is unlikely that these mAbs induce apoptosis cient in T, B, and NK cells (Shultz et al., 2005) but retain phago-
of NHL cells that are then secondarily phagocytosed. cyte effector cells. In the disseminated model, luciferase-
Next, we tested the ability of a blocking anti-CD47 mAb to syn- expressing Raji cells were transplanted intravenously into adult
ergize with rituximab in the phagocytosis of NHL cells. We exam- NSG mice. Two weeks later, these mice were administered daily
ined CD20 expression on NHL cells and found no difference injections of either control mouse IgG, anti-CD47 antibody, ritux-
between normal B cells and NHL cells (Figures S2I and S2J). imab, or anti-CD47 antibody and rituximab. Anti-CD47 antibody
Incubation of NHL cells with rituximab in the presence of mouse treatment decreased the lymphoma burden in these mice (Fig-
or human macrophages resulted in significant phagocytosis ures 4A and 4B) and significantly prolonged survival compared
(Figures 2D and S2E). We then tested the synergy hypothesis to control IgG, although all mice eventually died (Figure 4C and
by isobologram analysis (Chou, 2008; Tallarida, 2006). Using Table S2). Similar results were seen with rituximab and were
Raji, SUDHL4, and NHL17 cells, which express varying levels not statistically different compared to anti-CD47 antibody (Fig-
of both CD47 and CD20 (Figure S2K), anti-CD47 antibody ures 4A–4C and Table S2). In contrast, combination therapy
synergized with rituximab or anti-human CD20 (mouse IgG2a) with anti-CD47 antibody and rituximab eliminated lymphoma in
antibody, as indicated by combination indices less than 1 (Fig- 60% of mice as indicated by long-term survival (Figure 4C) and
ure 2C). In a second approach, in vitro phagocytosis assays the absence of luciferase-positive lymphoma (data not shown)
were conducted with primary NHL cells incubated with either more than 4 months after the end of treatment. In humans, ritux-
anti-CD47 antibody or rituximab alone, or both in combination imab efficacy is thought to be primarily mediated by both macro-
at half of the single agent dose. NHL cells exhibited a significant phages and NK cells (Nimmerjahn and Ravetch, 2007; Taylor
increase in phagocytosis when incubated with the combination and Lindorfer, 2008). Given that NSG mice lack NK cells, we con-
compared to either antibody alone when using mouse (Fig- ducted a similar experiment in NK cell-containing SCID mice and
ure 2D) or human (Figure 2E) macrophage effectors. No phago- observed similar therapeutic responses as in NSG mice (Figures
cytosis of NPB cells was observed with either antibody alone or S4A and S4B).
in combination with human macrophages (Figure 2E). In the localized NHL model, luciferase-expressing Raji cells
were transplanted subcutaneously in the right flank of NSG
Ex Vivo Coating of NHL Cells with an Anti-CD47 Antibody mice. Once tumors were palpable (approximately 2 weeks),
Inhibits Tumor Engraftment mice were treated with antibody therapy. Mice treated with
Next, the ability of blocking anti-CD47 antibody to eliminate NHL anti-CD47 antibody or rituximab demonstrated a decreased
in vivo either alone or in combination with rituximab was explored rate of lymphoma growth compared to control IgG-treated
by two separate treatment strategies. First, the effect of ex vivo mice as measured by both luciferase signal and tumor volume
anti-CD47 antibody coating on the engraftment of human NHL (Figures 4D–4F) but, like controls, eventually had to be sacrificed
cells was tested. Luciferase-expressing Raji and SUDHL4 cell due to enlarged tumor growth. In contrast, mice treated with the
lines were precoated ex vivo with anti-CD47, IgG1 isotype combination of anti-CD47 antibody and rituximab demonstrated
control, or anti-CD45 antibody and transplanted intravenously complete elimination of lymphoma, with 86% of treated mice
into SCID mice. Coating with anti-CD47 antibody prevented having no measurable mass or luciferase-positive lymphoma
engraftment of both cell lines (Figures 3A–3F). Coating of Raji 4 weeks after the end of therapy (Figures 4D–4F and Figures
cells with rituximab also inhibited engraftment when trans- S4C–S4E). Moreover, all showed no evidence of tumor growth,

(C) Synergistic phagocytosis by anti-CD47 antibody (B6H12.2) and anti-CD20 mAbs was examined by isobologram analysis and determination of combination
indices (CI). CIobs indicates observed results, and the dashed line indicates the expected results if antibodies were additive.
(D and E) Synergy between anti-CD47 antibody and rituximab in the phagocytosis of NHL and NPB cells was assessed by determining the phagocytic index when
incubated with a combination of both antibodies compared to either antibody alone at twice the dose, with either mouse (D) or human (E) macrophages. NHL17*:
cell line derived from primary sample NHL17.
***p < 0.001, ****p < 0.0001, *****p < 0.00001. Figure 2B p values represent comparison against IgG1 isotype control antibody. See also Figure S2.

Cell 142, 699–713, September 3, 2010 ª2010 Elsevier Inc. 703

Figure 3. Ex Vivo Coating of NHL Cells with an Anti-CD47 Antibody Inhibits Tumor Engraftment
(A–F) Luciferase-expressing Raji (A) and SUDHL4 (C) cells were assessed for ex vivo antibody coating by flow cytometry. SCID mice transplanted with Raji (B) and
SUDHL4 (D) were subject to bioluminescent imaging (1-IgG1 isotype control, 2-anti-CD45, 3 and 4-anti-CD47, 5-luciferase negative control). Bioluminescence
for Raji (E) and SUDHL4 (F) engrafted mice was quantified (n = 3 per antibody condition). No tumor engraftment was observed in mice transplanted with anti-
CD47-coated cells compared to IgG-coated cells (p < 0.05) for both Raji and SUDHL4, as assessed by bioluminescent imaging. Data are represented as
mean ± standard deviation (SD).

704 Cell 142, 699–713, September 3, 2010 ª2010 Elsevier Inc.

remained relapse free, and were alive at over 197 days after different anti-CD47 antibody potencies in distinct tissue com-
tumor engraftment. Out of six mice achieving a complete remis- partments (peripheral blood versus bone marrow (BM) versus
sion, five remained relapse free whereas one mouse died of non- soft tissue compartments).
tumor-related causes (Figure 4E). For both disseminated and To assess the ability of these two primary NHL xenotransplant
localized xenograft models, expression of CD47 and CD20 in models to model the disease, we compared histological sections
transplanted Raji cells did not differ from Raji cells in culture of the primary NHL specimen and the transplanted tumor. Similar
(Figure S4F). DLBCL and FL morphology was observed for NHL7 and NHL31,
respectively (Figure S5B). We next determined whether the per-
Combination Therapy with Anti-CD47 Antibody centages of macrophages infiltrating the tumor differed between
and Rituximab Eliminates Lymphoma in Primary Human the primary patient and the xenografted tumor. For NHL31, the
NHL Xenotransplant Mouse Models percentage of infiltrating human macrophages (CD68+) in the
NHL cell lines have been valuable for the evaluation of therapeu- primary lymph node was similar to the percentage of infiltrating
tics, but they may not accurately recapitulate the heterogeneity mouse macrophages (F4/80+) in bone marrow of transplanted
of the primary disease. We report here two new mouse xenograft mice (Figure S5C). Analyzing infiltrating macrophage frequency
models for NHL in which intravenous transplantation of cells by flow cytometry, no difference was observed between the
from a DLBCL patient (NHL7/SUNHL7) and a FL patient primary specimen and xenograft for either NHL7 or NHL31
(NHL31/SUNHL31) give rise to robust lymphoma engraftment (Figure S5D).
in the bone marrow and/or peripheral blood (Figure S1A and
Figure S5A). Primary DLBCL cells were transplanted into mice, Synergy between Anti-CD47 Antibody and Rituximab
which 2 weeks later were treated with daily injections of anti- Does Not Occur through NK Cells or Complement
bodies for 14 days. Treatment with anti-CD47 antibody either Rituximab can eliminate malignant cells via apoptosis, NK cell-
alone or in combination with rituximab eliminated human mediated ADCC, and CDC (reviewed in Smith, 2003). However,
lymphoma in the bone marrow, whereas treatment with rituxi- it is not known whether anti-CD47 antibody also enables ADCC
mab resulted in a reduction of disease in 60% of mice (Figures or CDC in addition to phagocytosis. Therefore, we investigated
5A and 5B). Mouse hematopoietic cells were unaffected by anti- whether anti-CD47 antibody alone could induce antitumor
body therapy (Figure S5E). Treatment was then discontinued, effects by macrophage-independent mechanisms, and whether
and all mice were monitored for survival. Mice treated with either synergy with rituximab could occur through these modalities.
anti-CD47 antibody or rituximab alone had a significantly longer First, to investigate possible synergy in direct apoptosis, NHL
survival compared to mice treated with control IgG, but all even- cells were incubated with either anti-CD47 antibody/rituximab
tually died secondary to disease due to widespread organ alone or in combination without macrophages, and cell death
dissemination on autopsy (Figure 5C, data not shown). Most was measured. No synergistic apoptosis was observed when
significantly, 8 out of 9 mice (89%) administered combination NHL cells were incubated with soluble (Figure 6A and Fig-
antibody treatment were cured of lymphoma, as indicated by ure S6A) or immobilized (Figures S6B and S6C) antibodies.
long-term disease-free survival more than 4 months after the Furthermore, crosslinking of soluble anti-CD47 antibody alone
end of treatment (Figure 5C and Table S3) with no detectable or in combination with rituximab by macrophages did not induce
lymphoma in the bone marrow (data not shown). In a second increased apoptosis of nonphagocytosed NHL cells compared
primary model, FL cells were transplanted intravenously in a to IgG1 isotype control, whereas rituximab induced a slight
similar manner. CD20+CD10+ lymphoma engraftment in the increase in apoptosis (Figure S6D). No synergistic apoptosis
peripheral blood and bone marrow was detected after 8 weeks. was observed in this context (Figure S6D). The small increase
At this time, mice were treated with a single intraperitoneal injec- in apoptosis upon antibody treatment was not FcR dependent
tion of either control IgG, anti-CD47 antibody, rituximab, or the given that results were similar with macrophages lacking the
combination and then followed for disease progression. A single Fcg subunit (Takai et al., 1994) (Figure S6D). Second, we inves-
dose of anti-CD47 antibody alone or in combination with rituxi- tigated whether NK cells could mediate tumor elimination by
mab eliminated lymphoma both in the peripheral blood (Fig- anti-CD47 antibody alone or in synergy with rituximab. A prior
ure 5D) and in bone marrow (Figure 5E). In contrast, a single report observed increased NK cell cytotoxicity of cancer cell
dose of rituximab enabled a partial reduction in tumor burden lines with an anti-CD47 antibody in vitro, though the mechanism
that rebounded back to baseline levels in the peripheral blood of targeting was not elucidated (Kim et al., 2008). We first deter-
with no tumor reduction observed in the bone marrow. The differ- mined whether human or mouse NK cells expressed SIRPa and
ence in anti-CD47 antibody clearance of lymphoma as a single found that both human NK cells, CD3CD56+CD7+ (Milush et al.,
dose in FL-engrafted mice (Figures 5D and 5E) compared to 2009), as well as mouse NK cells, CD3DX5+, expressed minimal
multiple dose therapy in mice engrafted with DLBCL (Fig- to no SIRPa (Figure 6B). Next, the ability of anti-CD47 antibody
ure 5B) or Raji (Figure 4) may be due to cell-intrinsic differences to induce NK cell-mediated ADCC through FcRs or by CD47-
in antibody sensitivity between different NHL subtypes or due to SIRPa blockade was investigated. Utilizing human NK cells as

(G) Bulk lymphoma cells from a human NHL patient were assessed for ex vivo antibody coating by flow cytometry.
(H) Compared to IgG1 isotype control, anti-CD47 antibody pretreatment inhibited engraftment of NHL cells (p < 0.0001) whereas anti-CD45-coated cells
engrafted similarly to controls (p = 0.54). All p values were determined using Fisher’s exact test.
Error bars represent SD (E and F). See also Figure S3.

Cell 142, 699–713, September 3, 2010 ª2010 Elsevier Inc. 705

Figure 4. Combination Therapy with Anti-CD47 Antibody and Rituximab Eliminates Lymphoma in Both Disseminated and Localized Human
NHL Xenotransplant Mouse Models
(A) NSG mice transplanted intravenously with luciferase-expressing Raji cells were treated with the indicated antibodies (n = 8 per treatment group). Luciferase
imaging of representative mice from pre- and 10 days post-treatment are shown (A) and averaged for all mice in each treatment group (B).
(C) Kaplan-Meier survival analysis was performed (Table S2). p values compare IgG control to combination antibody treatment or anti-CD47 antibody/rituximab
single antibody to combination. Arrows indicate start (day 14) and stop (day 35) of treatment.
(D) Luciferase-expressing Raji cells were transplanted subcutaneously in the flank of NSG mice. When palpable tumors (0.1 cm3) formed, treatment
began with the indicated antibodies. Luciferase imaging of representative mice from each treatment group is shown before (day 0) and during (day 14)
(E) Quantified bioluminescence was determined and averaged for all mice in each treatment group (n = 7).
(F) Tumor volume was measured with average volume shown. p values were derived from a two-way ANOVA and compared to IgG treatment.
*p < 0.05, ***p < 0.001, ****p < 0.0001. Complete remission was defined as the number of mice with no evidence of tumor at the indicated date. Relapse was
defined as the number of mice achieving a complete remission that later developed recurrence of tumor growth. For (E), one mouse that achieved a complete
remission died of non-tumor-related causes. Data presented in (B), (E), and (F) are mean values ± SD. See also Figure S4 and Table S2.

706 Cell 142, 699–713, September 3, 2010 ª2010 Elsevier Inc.

Figure 5. Combination Therapy with Anti-CD47 Antibody and Rituximab Eliminates Lymphoma in Primary Human NHL Xenotransplant
Mouse Models
(A and B) Sublethally irradiated NSG mice were transplanted intravenously with cells from a primary DLBCL patient (NHL7) and treated with the indicated anti-
bodies. Flow cytometry of human lymphoma engraftment in the bone marrow of two representative mice is shown pre- and 14 days post-antibody treatment in
(A). Data from all mice are included in (B). **p < 0.01, comparing pre- and post-treatment values for each respective antibody treatment.
(C) Kaplan-Meier survival analysis (Table S3) of DLBCL-engrafted mice from each antibody treatment cohort is shown (n R 10 per antibody group), with p values
calculated comparing control IgG to combination antibody treatment or anti-CD47 antibody/rituximab single antibody to combination treatment. Arrows indicate
start (day14) and stop (day 28) of treatment.
(D and E) Mice engrafted intravenously with a primary FL patient sample (NHL31) were treated with a single dose of the indicated antibodies. Compared to IgG
control and rituximab, anti-CD47 antibody alone or in combination with rituximab eliminated tumor burden in the peripheral blood (p = 0.04, two-way ANOVA) and
bone marrow (p < 0.001, t test).
(E) Lyphoma engraftment in the bone marrow was determined 14 days post-treatment. Each antibody treatment group consisted of three mice.
For mice reported in (D) and (E), human lymphoma chimerism was between 5% and 25% in the peripheral blood and bone marrow. Error bars represent SD
(D and E). See also Figure S5 and Table S3.

Cell 142, 699–713, September 3, 2010 ª2010 Elsevier Inc. 707

Figure 6. Synergy between Anti-CD47 Antibody and Rituximab Does Not Occur through NK Cells or Complement
(A) NHL cells were incubated with the indicated soluble antibodies for 2 hr and the percentage of dead cells was calculated (% annexin V+ and/or 7-AAD+). No
statistically significant difference in % dead cells was observed with the combination of anti-CD47 antibody and rituximab compared to either anti-CD47 antibody
alone (p = 0.24) or rituximab alone (p = 0.95).
(B) SIRPa expression is shown for both human and mouse NK cells as determined by flow cytometry.
(C and D) Chromium release assays measuring ADCC were performed in triplicate with human (C) and mouse (D) at an effector:target ratio of 17.5:1, and percent-
specific lysis is reported. Antibodies were incubated at 10 mg/ml except anti-CD47 full-length or F(ab0 )2 antibody+rituximab (5 mg/ml).
(E) CDC assay with human complement was performed in duplicate. Compared to IgG1 isotype control, anti-CD47 antibody did not enable CDC (p > 0.2),
whereas rituximab did (p < 0.001) by two-way ANOVA for both SUDHL4 and NHL17*. Combination treatment with anti-CD47 antibody and rituximab did not
enable greater levels of CDC compared to rituximab (p = 0.78).
(F) CDC assay with mouse complement was performed in duplicate. Compared to IgG1 isotype control, anti-CD47 antibody did not enable CDC (p > 0.25)
whereas rituximab did (p = 0.03, p = 0.08, respectively) for both SUDHL4 and NHL17*. No difference in CDC between CD47 antibody+rituximab and rituximab
alone was observed (p > 0.13) for both SUDHL4 and NHL17*.
*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Error bars represent SD (C–F). NHL17* = primary NHL17 cells expanded in culture. See also Figure S6.

708 Cell 142, 699–713, September 3, 2010 ª2010 Elsevier Inc.

effectors, anti-CD47 antibody did not induce increased ADCC of (Takai et al., 1994), but still expressing SIRPa (Figure S7A),
Raji or SUDHL4 cells compared to IgG1 isotype control (Fig- were utilized as effector cells for phagocytosis of NHL cells incu-
ure 6C). Although rituximab did enable ADCC of these two bated with either anti-CD47 antibody, anti-SIRPa antibody, ritux-
NHL cell lines, no synergistic effect between anti-CD47 antibody imab, or anti-CD47/anti-SIRPa antibody in combination with
and rituximab was observed (p = 0.77, Figure 6C). As anti-CD47 rituximab. As predicted, anti-CD47 antibody and anti-SIRPa
antibody (B6H12.2) is a mouse IgG1 isotype, we repeated these antibody, but not rituximab, enabled phagocytosis of NHL cells,
assays with mouse NK cells. Anti-CD47 antibody caused without evidence of synergistic phagocytosis (Figure 7C).
increased ADCC of these two NHL cell lines compared to isotype Third, F(ab0 )2 fragments of both anti-CD47 antibody and
control, whereas rituximab induced ADCC to a lesser degree rituximab were generated (Figures S7B–S7I) and utilized in
(Figure 6D). To test whether anti-CD47 antibody-mediated phagocytosis assays with NHL cells and wild-type macro-
ADCC was Fc dependent, we generated a F(ab0 )2 fragment of phages. Anti-CD47 F(ab0 )2 antibody synergized with rituximab
the anti-CD47 antibody (Figures S7B–S7I). The F(ab0 )2 fragment as demonstrated by isobologram analysis (Figure 7D). Addition-
did not enable ADCC, indicating that the increased ADCC was ally, anti-CD47 F(ab0 )2, but not rituximab F(ab0 )2, enabled phago-
likely mediated through FcRs (Figure 6D). The combination of cytosis of NHL cells (Figure 7E). Consistent with the proposed
anti-CD47 antibody or F(ab0 )2 fragment with rituximab did not mechanism, synergistic phagocytosis was observed with the
induce a statistically significant increase in ADCC compared to combination of either full-length anti-CD47 or anti-CD47 F(ab0 )2
single agent therapy, indicating no synergistic effect. Third, we with full-length rituximab, but not with any combination involving
investigated the role of complement in anti-CD47 antibody- rituximab F(ab0 )2 (Figure 7E).
mediated cytotoxicity. Using either human (Figure 6E) or mouse Fourth, synergistic phagocytosis was investigated in vivo
(Figure 6F) complement, anti-CD47 antibody did not induce CDC using GFP+ Raji cells engrafted in NSG mice. As single agents,
of either an NHL cell line or a primary NHL sample, whereas anti-CD47 antibody and rituximab enabled phagocytosis of
rituximab did induce significant CDC against both of these Raji cells engrafted in the liver as evidenced by an increased
samples. Moreover, the combination of anti-CD47 antibody percentage of mouse macrophages containing phagocytosed
and rituximab did not induce increased CDC compared to ritux- GFP+ Raji cells (Figure 7F). Most significantly, combination
imab alone. Fourth, we investigated the relative contribution of anti-CD47 antibody and rituximab treatment enabled signifi-
major components of macrophages, NK cells, and complement cantly increased phagocytosis compared to either single agent
in mediating the therapeutic effects of anti-CD47 antibody and demonstrating that synergistic phagocytosis occurred in vivo
rituximab in vivo. Luciferase-labeled Raji cells were engrafted (Figure 7F).
intravenously into SCID mice, which have functional macro-
phages, NK cells, and complement. Mice were then separated DISCUSSION
into cohorts receiving selective depletion of either macrophages
by clodronate, NK cells by anti-asialoGM1 antibody, comple- In this report, we identify a distinct mechanism of synergy
ment by cobra venom factor, or a vehicle control. These cohorts between mAbs in cancer therapy leading to cures in the absence
were then treated with combination anti-CD47 antibody and rit- of chemotherapy. Specifically, we utilized a blocking anti-CD47
uximab therapy for 3 days, and tumor burden was measured antibody in combination with the anti-CD20 antibody rituximab
by bioluminescent imaging pre- and post-treatment. Compared to eradicate human NHL through a mechanism of synergy
to vehicle control, NK cell and complement depletion had no involving FcR-independent enabling of phagocytosis by anti-
effect on tumor elimination by combination antibody therapy CD47 antibody and FcR-dependent stimulation of phagocytosis
(Figure S6E). In contrast, macrophage depletion significantly by rituximab. In addition, the identification of CD47 expression
abrogated the therapeutic effect, indicating that macrophages, as a prognostic factor could be incorporated into standard clin-
and not NK cells or complement, are required for combination ical prognostic considerations across multiple subtypes of NHL
anti-CD47 antibody and rituximab-mediated elimination of NHL and may be useful in risk-adapted therapy decision-making.
in vivo. Although it is thought that many therapeutic mAbs for human
malignancies, including rituximab, function primarily through
Anti-CD47 Antibody Synergizes with Rituximab through NK cell-mediated ADCC, several lines of evidence indicate that
FcR-Independent and FcR-Dependent Mechanisms the therapeutic effect of anti-CD47 antibody alone or in combi-
We hypothesize that the observed synergy between an anti- nation with rituximab is mediated primarily through macrophage
CD47 antibody and rituximab occurs through the combination phagocytosis. First, synergistic macrophage phagocytosis was
of two separate mechanisms for stimulating phagocytosis: (1) observed with combination anti-CD47 antibody and rituximab
FcR-independent through blockade of an inhibitory phagocytic in vitro, whereas no synergy was observed for direct apoptosis,
signal by anti-CD47 antibody, and (2) FcR-dependent through ADCC, or CDC (Figures 2C–2E, Figure S6, Figure 6A, and Figures
delivery of a positive phagocytic signal by rituximab. We utilized 6C–6F). Second, phagocytosis of NHL cells in vivo was observed
four independent methods to investigate this hypothesis. First, with either anti-CD47 antibody or rituximab alone and, most
synergistic phagocytosis was observed with the combination importantly, significantly increased with combination therapy
of anti-SIRPa antibody and rituximab by isobologram analysis (Figure 7F). Third, the therapeutic effect of combination antibody
(Figure 7A), and with a large panel of primary NHL samples (Fig- treatment was similar in an NHL xenotransplant model using
ure 7B). Second, mouse macrophages lacking the Fcg receptor, complement and NK cell-deficient NSG mice (Figure 4C) as in
thereby incapable of enabling FcR-dependent phagocytosis complement and NK cell-competent SCID mice (Figures S4A

Cell 142, 699–713, September 3, 2010 ª2010 Elsevier Inc. 709

Figure 7. Anti-CD47 Antibody Synergizes with Rituximab through FcR-Independent and FcR-Dependent Mechanisms
(A) Isobologram analysis of phagocytosis induced by anti-SIRPa antibody and rituximab is shown for Raji cells and mouse macrophages.
(B and C) NHL cells were incubated in vitro with the indicated antibodies in the presence of wild-type (B) or Fcgr/ (C) mouse macrophages, and the phagocytic
index was determined.
(D) Isobologram analysis of phagocytosis induced by anti-CD47 F(ab0 )2 antibody and rituximab is shown for Raji cells and mouse macrophages.
(E) NHL cells were incubated with wild-type mouse macrophages in the presence of the indicated full-length or F(ab0 )2 antibodies (single antibodies at 10 mg/ml,
combination antibodies at 5 mg/ml each) and the phagocytic index was determined.
(F) The level of in vivo phagocytosis measured as the percentage of mouse macrophages containing phagocytosed GFP+ Raji cells (hCD45GFP+F4/80+) was
determined by flow cytometry of livers from mice engrafted with GFP+ Raji cells and then treated with the indicated antibodies (see Experimental Procedures),
with each treatment group performed in duplicate. Compared to IgG control, anti-CD47 antibody and rituximab enabled increased levels of phagocytosis.
Compared to anti-CD47 antibody alone, combination anti-CD47 antibody and rituximab enabled higher levels of phagocytosis.
*p < 0.05, **p < 0.01, ****p < 0.0001. Error bars represent SD (E and F). See also Figure S7.

710 Cell 142, 699–713, September 3, 2010 ª2010 Elsevier Inc.

and S4B), suggesting that macrophages alone are sufficient to targeting of several human cancers including AML (Majeti
mediate the therapeutic effect. Fourth, depletion of macro- et al., 2009), bladder cancer (Chan et al., 2009), and now NHL,
phages, but not complement or NK cells, abrogated the syner- leading us to speculate that CD47 targeting will be effective
gistic effect of anti-CD47 antibody in combination with rituximab against a wide range of human cancers.
(Figure S6E). These studies highlight the importance of macro-
phages as effectors of anti-CD47 antibody therapy in human EXPERIMENTAL PROCEDURES
This study describes a mechanism of antibody synergy in the Cell Lines
elimination of NHL in the absence of chemotherapy. Combina- A Burkitt’s lymphoma cell line (Raji) and a DLBCL cell line (SUDHL4) were
obtained from the American Type Culture Collection or generated in the lab.
tion antibody therapies for NHL have previously been investi-
The NHL17* cell line was generated from a patient with DLBCL by culturing
gated, mostly in combination with rituximab, with some pro- bulk cells in vitro with IMDM supplemented with 10% human AB serum for
gressing to clinical trials. These include a humanized antibody 1.5 months.
targeting the B cell antigen CD22 (epratuzumab) and galiximab,
a chimeric antibody targeting the costimulatory ligand CD80. Human Samples
Phase I/II studies with either epratuzumab or galiximab in combi- Normal human peripheral blood and human NHL samples were obtained from
nation with rituximab demonstrate relative safety and clinical the Stanford University Medical Center (Stanford, CA, USA) with informed
consent, according to an IRB-approved protocol (Stanford IRB# 13500) or
responses equal to or greater than single agent therapy alone
with informed consent from the Norwegian Radium Hospital (Oslo, Norway)
(Leonard et al., 2005, 2007, 2008). Based on these results, phase according to a Regional Ethic Committee (REK)-approved protocol (REK#
III trials are underway. Antibody combinations involving anti- 2.2007.2949). Normal tonsils for germinal center B cell analysis were obtained
CD20 antibodies and antibodies to proapoptotic receptors are from discarded tonsillectomy specimens from consented pediatric patients at
also being explored preclinically (Daniel et al., 2007; Maddipatla Stanford University Medical Center according to an IRB-approved protocol
et al., 2007). These studies highlight the clinical potential of (Stanford IRB# 13500).
combination antibody approaches in NHL patients.
Combination therapy with two or more mAbs possesses Flow Cytometry Analysis
For analysis of normal peripheral blood cells, germinal center B cells, and
several advantages compared to monotherapies in NHL or other
primary NHL cells, the following antibodies were used: CD19, CD20, CD3,
malignancies. First, therapy solely with monoclonal antibodies CD10, CD45, CD5, CD38 (Invitrogen, Carlsbad, CA, USA and BD Biosciences,
targeting cancer-specific antigens could result in decreased San Jose, CA, USA). Analysis of CD47 expression was performed with an anti-
off-target toxicity compared to current therapeutic regimens human CD47 FITC antibody (clone B6H12.2, BD Biosciences). Cell staining
that utilize chemotherapy. Second, synergy between two distinct and flow cytometry analysis was performed as previously described (Majeti
antibody effector mechanisms, FcR-independent and FcR- et al., 2009).
dependent as shown here, would result in increased therapeutic
efficacy. Third, antibody targeting of two distinct cell-surface Evaluation of Prognostic Value of CD47 in NHL
Gene expression and clinical data were analyzed for eight previously
antigens would be more likely to eliminate cancer cells with
described cohorts of adult NHL patients, including four studies of patients
pre-existing epitope variants or epitope loss, such as those with DLBCL, three with B-CLL, and one with MCL (detailed in Table S1). The
reported in rituximab-refractory/resistant NHL patients (Foran clinical end points analyzed included overall (OS), progression free (PFS),
et al., 2001; Hiraga et al., 2009; Kennedy et al., 2004). Fourth, and event-free survival (EFS), with events defined as the interval between
a bispecific FcR-engaging antibody with one arm binding and study enrollment and need for therapy or death from any cause, with data
blocking CD47 and the other binding to a validated cancer anti- censored for patients who did not have an event at the last follow-up visit.
See Extended Experimental Procedures for a detailed description of the
body target (CD20) could reduce potential antibody toxicity,
while retaining the synergy effect, especially as CD47 is
expressed in multiple normal tissue types. Although we demon-
Therapeutic Antibodies
strated that an anti-mouse CD47 antibody is relatively nontoxic Rituximab (anti-CD20, human IgG1) was obtained from the Stanford University
to wild-type mice (Majeti et al., 2009), a clinical anti-human Medical Center, mouse anti-human CD20, IgG2a from Beckman Coulter
CD47 antibody may have a different human toxicity profile that (Miami, FL, USA), and anti-CD47 antibody BRIC126, IgG2b from AbD Serotec
could be overcome by a bispecific antibody. (Raleigh, NC, USA). Other anti-CD47 antibodies were used as in Majeti et al.
In addition to its application in NHL, the reported mechanism (2009). All in vivo antibody experiments were performed using the anti-CD47
B6H12.2 antibody.
of antibody synergy provides proof-of-principle that a blocking
mAb directed against CD47 can synergize with an FcR-acti-
In Vitro Isobologram Studies
vating antibody to provide superior therapeutic efficacy. This In vitro phagocytosis assays were conducted with NHL cells incubated with
finding raises the possibility of potential synergy between an anti-CD47 antibody (B6H12.2), anti-CD20 IgG2a, or rituximab either alone or
anti-CD47 antibody and other clinically approved therapeutic in combination at concentrations from 1 mg/ml to 10 mg/ml. The concentration
antibodies that may activate FcRs on immune effector cells for of each antibody required to produce a defined single-agent effect (phagocytic
the treatment of diverse human malignancies, including trastu- index) was determined for each cell type. Concentrations of the two antibodies
zumab (Herceptin) for HER2-positive breast carcinomas, cetux- combined to achieve this same phagocytic index were then plotted on an
isobologram and the combination index (CI) determined. The CI was calcu-
imab (Erbitux) for colorectal carcinomas and head and neck
lated from the formula CI = (d1/D1) + (d2/D2), whereby d1 = dose of drug
squamous cell carcinomas, alemtuzumab (Campath) for CLL 1 in combination to achieve the phagocytic index, d2 = dose of drug 2 in
and T cell lymphoma, and others in development (Finn, 2008). combination to achieve the phagocytic index, D1 = dose of drug 1 alone to
To date, we have demonstrated effective anti-CD47 antibody achieve the phagocytic index, D2 = dose of drug 2 alone to achieve the

Cell 142, 699–713, September 3, 2010 ª2010 Elsevier Inc. 711

phagocytic index. A CI of less than, equal to, and greater than 1 indicates SUPPLEMENTAL INFORMATION
synergy, additivity, and antagonism, respectively.
Supplemental Information includes Extended Experimental Procedures, seven
Annexin V Apoptosis Assays figures, and three tables and can be found with this article online at doi:10.
Assays were performed as previously described (Majeti et al., 2009). 1016/j.cell.2010.07.044.

Preparation of Human and Mouse Immune Effector Cells, Immune
Effector Cytotoxicity Assays, and In Vivo Depletion of Immune
The authors acknowledge Dr. Christopher Contag for providing luciferase
Effector Cells
constructs, Dr. Robert Negrin for assistance, Dr. Yaso Natkunam for immuno-
See Extended Experimental Procedures.
histochemistry, Libuse Jerabek and Theresa Storm for lab management, and
Adriane Mosely for animal husbandry. We also acknowledge the patients and
Preparation of F(ab0 )2 Fragments surgeons including Drs. Wapnir, Chang, Cheng, Janisiewicz, Koltai, Liebowitz,
See Extended Experimental Procedures. and Messner for providing research specimens. M.P.C. is supported by the
HHMI and the Stanford Cancer Biology Program, A.A.A. by an NIH T32 Ruth
Generation of Luciferase-Positive Cell Lines and Luciferase Imaging L. Kirschstein National Research Service Award (HL007970), R. Malumbres
Analysis by a fellowship from Fundación Caja Madrid, I.S.L. by NIH grant CA122105,
See Extended Experimental Procedures. and R. Majeti by a grant from the AACR. R. Majeti holds a Career Award for
Medical Scientists from the Burroughs Wellcome Fund. I.L.W., M.P.C.,
A.A.A., and R. Majeti have filed U.S. Patent Application Serial No. 12/321,215
In Vivo Precoating Engraftment Assay entitled ‘‘Methods For Manipulating Phagocytosis Mediated by CD47.’’ This
Assay were performed as previously described (Majeti et al., 2009). Precoated research is supported by NIH grant P01CA139490 to I.L.W. and funding from
cells were transplanted intravenously into SCID mice or sublethally irradiated the Smith Family Fund.
(200 rads) NOD.Cg-PrkdcscidIl2rgtm1Wjl/SzJ (NSG). All experiments involving M.P.C., A.A.A., I.L.W., and R. Majeti designed the experiments, and M.P.C.,
mice were performed according to Stanford University institutional animal A.A.A., I.L.W., and R. Majeti wrote the manuscript. M.P.C., A.A.A., C.T.,
guidelines. J.H.M., S.G., M.J., A.C.C., C.K.C., B.T.T., C.Y.P, F.Z., R. Malumbres, and
I.S.L. performed the experiments and analyzed data. J.B., R.D.G., R.L., and
In Vivo Antibody Treatment in a Disseminated Lymphoma I.S.L provided patient samples and clinical data. H.E.K. provided reagents.
Xenograft Model B.V. generated F(ab0 )2 fragments of anti-CD47 antibody and rituximab. All
1.5 3 106 luciferase-labeled Raji cells were injected intravenously into the authors endorse the full content of this work.
retro-orbital sinus of 6- to 10-week-old SCID or NSG mice. Those mice with
luciferase-positive lymphoma were given daily intraperitoneal injections of Received: January 7, 2010
200 mg mouse IgG control, anti-CD47 antibody, rituximab, or 200 mg anti- Revised: April 23, 2010
CD47 antibody + 200 mg rituximab for 3 weeks. Antibody treatment was Accepted: July 6, 2010
then stopped and mice were followed for survival analysis. A complete remis- Published: September 2, 2010
sion (CR) was defined as no evidence of lymphoma by bioluminescence at the
end of treatment. A relapse was defined as evidence of lymphoma by biolumi- REFERENCES
nescence after the end of treatment in a mouse with a prior CR.
Adams, G.P., and Weiner, L.M. (2005). Monoclonal antibody therapy of cancer.
In Vivo Antibody Treatment in a Localized Lymphoma Nat. Biotechnol. 23, 1147–1157.
Xenograft Model Alizadeh, A.A., Eisen, M.B., Davis, R.E., Ma, C., Lossos, I.S., Rosenwald, A.,
3 3 106 luciferase-labeled Raji cells were injected subcutaneously into the Boldrick, J.C., Sabet, H., Tran, T., Yu, X., et al. (2000). Distinct types of diffuse
right flank of 6- to 10-week-old NSG mice. Those mice with luciferase-positive large B-cell lymphoma identified by gene expression profiling. Nature 403,
lymphoma were given daily intraperitoneal injections of 400 mg mouse IgG 503–511.
control, 400 mg anti-CD47, 200 mg rituximab, or 400 mg anti-CD47 + 200 mg rit- Brown, E.J., and Frazier, W.A. (2001). Integrin-associated protein (CD47) and
uximab for 4 weeks. Tumor volume was measured every 3–4 days using the its ligands. Trends Cell Biol. 11, 130–135.
formula (length*width)/2. Antibody treatment was then stopped and mice Cartron, G., Watier, H., Golay, J., and Solal-Celigny, P. (2004). From the bench
were followed for survival analysis. to the bedside: ways to improve rituximab efficacy. Blood 104, 2635–2642.
Chan, K.S., Espinosa, I., Chao, M., Wong, D., Ailles, L., Diehn, M., Gill, H.,
In Vivo Antibody Treatment of Primary NHL-Engrafted Mice Presti, J., Jr., Chang, H.Y., van de Rijn, M., et al. (2009). Identification, molec-
2 3 106 NHL cells were transplanted intravenously via retro-orbital plexus into ular characterization, clinical prognosis, and therapeutic targeting of human
sublethally irradiated NSG mice. Two to ten weeks later, bone marrow was bladder tumor-initiating cells. Proc. Natl. Acad. Sci. USA 106, 14016–14021.
aspirated from these mice and those mice with evidence of human lymphoma
Cheson, B.D., and Leonard, J.P. (2008). Monoclonal antibody therapy for
engraftment (hCD45+CD19/CD10+ bone marrow cells) were then treated with
B-cell non-Hodgkin’s lymphoma. N. Engl. J. Med. 359, 613–626.
the same antibody regimen as in the disseminated lymphoma model. After
14 days, bone marrow cells from these mice were aspirated and antibody Chou, T.C. (2008). Preclinical versus clinical drug combination studies. Leuk.
treatment was stopped and mice followed for survival analysis. A CR was Lymphoma 49, 2059–2080.
defined as no evidence of lymphoma in the BM at end of treatment. A relapse Daniel, D., Yang, B., Lawrence, D.A., Totpal, K., Balter, I., Lee, W.P., Gogineni,
was defined as evidence of lymphoma in the BM after end of treatment in A., Cole, M.J., Yee, S.F., Ross, S., and Ashkenazi, A. (2007). Cooperation of the
a mouse with a prior CR. proapoptotic receptor agonist rhApo2L/TRAIL with the CD20 antibody rituxi-
mab against non-Hodgkin lymphoma xenografts. Blood 110, 4037–4046.
In Vivo Phagocytosis Finn, O.J. (2008). Cancer immunology. N. Engl. J. Med. 358, 2704–2715.
In vivo phagocytosis was performed as previously described (Majeti et al., Foran, J.M., Norton, A.J., Micallef, I.N., Taussig, D.C., Amess, J.A., Rohatiner,
2009) analyzing mice transplanted with GFP+ Raji cells into adult NSG mice. A.Z., and Lister, T.A. (2001). Loss of CD20 expression following treatment with
Mice were given a single dose of antibody and analyzed 4 hr later for in vivo rituximab (chimaeric monoclonal anti-CD20): a retrospective cohort analysis.
phagocytosis. Br. J. Haematol. 114, 881–883.

712 Cell 142, 699–713, September 3, 2010 ª2010 Elsevier Inc.

Glennie, M.J., French, R.R., Cragg, M.S., and Taylor, R.P. (2007). Mechanisms tion of monoclonal antibodies targeting TRAIL-R1 and CD20. Clin. Cancer Res.
of killing by anti-CD20 monoclonal antibodies. Mol. Immunol. 44, 3823–3837. 13, 4556–4564.
Hiraga, J., Tomita, A., Sugimoto, T., Shimada, K., Ito, M., Nakamura, S., Kiyoi, Majeti, R., Chao, M.P., Alizadeh, A.A., Pang, W.W., Jaiswal, S., Gibbs, K.D.,
H., Kinoshita, T., and Naoe, T. (2009). Down-regulation of CD20 expression in Jr., van Rooijen, N., and Weissman, I.L. (2009). CD47 is an adverse prognostic
B-cell lymphoma cells after treatment with rituximab-containing combination factor and therapeutic antibody target on human acute myeloid leukemia stem
chemotherapies: its prevalence and clinical significance. Blood 113, 4885– cells. Cell 138, 286–299.
4893. Mateo, V., Lagneaux, L., Bron, D., Biron, G., Armant, M., Delespesse, G., and
Jaiswal, S., Jamieson, C.H., Pang, W.W., Park, C.Y., Chao, M.P., Majeti, R., Sarfati, M. (1999). CD47 ligation induces caspase-independent cell death in
Traver, D., van Rooijen, N., and Weissman, I.L. (2009). CD47 is upregulated chronic lymphocytic leukemia. Nat. Med. 5, 1277–1284.
on circulating hematopoietic stem cells and leukemia cells to avoid phagocy- Milush, J.M., Long, B.R., Snyder-Cappione, J.E., Cappione, A.J., 3rd, York,
tosis. Cell 138, 271–285. V.A., Ndhlovu, L.C., Lanier, L.L., Michaelsson, J., and Nixon, D.F. (2009). Func-
Jaiswal, S., Chao, M.P., Majeti, R., and Weissman, I.L. (2010). Macrophages tionally distinct subsets of human NK cells and monocyte/DC-like cells identi-
as mediators of tumor immunosurveillance. Trends Immunol. 31, 212–219. fied by coexpression of CD56, CD7, and CD4. Blood 114, 4823–4831.
Katzenberger, T., Petzoldt, C., Holler, S., Mader, U., Kalla, J., Adam, P., Ott, Nimmerjahn, F., and Ravetch, J.V. (2007). Antibodies, Fc receptors and
M.M., Muller-Hermelink, H.K., Rosenwald, A., and Ott, G. (2006). The Ki67 cancer. Curr. Opin. Immunol. 19, 239–245.
proliferation index is a quantitative indicator of clinical risk in mantle cell Oscier, D.G., Gardiner, A.C., Mould, S.J., Glide, S., Davis, Z.A., Ibbotson, R.E.,
lymphoma. Blood 107, 3407. Corcoran, M.M., Chapman, R.M., Thomas, P.W., Copplestone, J.A., et al.
Kennedy, A.D., Beum, P.V., Solga, M.D., DiLillo, D.J., Lindorfer, M.A., Hess, (2002). Multivariate analysis of prognostic factors in CLL: clinical stage,
C.E., Densmore, J.J., Williams, M.E., and Taylor, R.P. (2004). Rituximab infu- IGVH gene mutational status, and loss or mutation of the p53 gene are inde-
sion promotes rapid complement depletion and acute CD20 loss in chronic pendent prognostic factors. Blood 100, 1177–1184.
lymphocytic leukemia. J. Immunol. 172, 3280–3288. Rosenwald, A., Wright, G., Chan, W.C., Connors, J.M., Campo, E., Fisher, R.I.,
Kikuchi, Y., Uno, S., Yoshimura, Y., Otabe, K., Iida, S., Oheda, M., Fukushima, Gascoyne, R.D., Muller-Hermelink, H.K., Smeland, E.B., Giltnane, J.M., et al.
N., and Tsuchiya, M. (2004). A bivalent single-chain Fv fragment against CD47 (2002). The use of molecular profiling to predict survival after chemotherapy
induces apoptosis for leukemic cells. Biochem. Biophys. Res. Commun. 315, for diffuse large-B-cell lymphoma. N. Engl. J. Med. 346, 1937–1947.
912–918. Rosenwald, A., Wright, G., Wiestner, A., Chan, W.C., Connors, J.M., Campo,
Kikuchi, Y., Uno, S., Kinoshita, Y., Yoshimura, Y., Iida, S., Wakahara, Y., E., Gascoyne, R.D., Grogan, T.M., Muller-Hermelink, H.K., Smeland, E.B.,
Tsuchiya, M., Yamada-Okabe, H., and Fukushima, N. (2005). Apoptosis et al. (2003). The proliferation gene expression signature is a quantitative inte-
inducing bivalent single-chain antibody fragments against CD47 showed anti- grator of oncogenic events that predicts survival in mantle cell lymphoma.
tumor potency for multiple myeloma. Leuk. Res. 29, 445–450. Cancer Cell 3, 185–197.

Kim, M.J., Lee, J.C., Lee, J.J., Kim, S., Lee, S.G., Park, S.W., Sung, M.W., and Shultz, L.D., Lyons, B.L., Burzenski, L.M., Gott, B., Chen, X., Chaleff, S., Kotb,
Heo, D.S. (2008). Association of CD47 with natural killer cell-mediated cytotox- M., Gillies, S.D., King, M., Mangada, J., et al. (2005). Human lymphoid and
icity of head-and-neck squamous cell carcinoma lines. Tumour Biol. 29, myeloid cell development in NOD/LtSz-scid IL2R gamma null mice engrafted
28–34. with mobilized human hemopoietic stem cells. J. Immunol. 174, 6477–6489.
Smith, M.R. (2003). Rituximab (monoclonal anti-CD20 antibody): mechanisms
Leonard, J.P., Coleman, M., Ketas, J., Ashe, M., Fiore, J.M., Furman, R.R.,
of action and resistance. Oncogene 22, 7359–7368.
Niesvizky, R., Shore, T., Chadburn, A., Horne, H., et al. (2005). Combination
antibody therapy with epratuzumab and rituximab in relapsed or refractory Takai, T., Li, M., Sylvestre, D., Clynes, R., and Ravetch, J.V. (1994). FcR
non-Hodgkin’s lymphoma. J. Clin. Oncol. 23, 5044–5051. gamma chain deletion results in pleiotrophic effector cell defects. Cell 76,
Leonard, J.P., Friedberg, J.W., Younes, A., Fisher, D., Gordon, L.I., Moore, J.,
Czuczman, M., Miller, T., Stiff, P., Cheson, B.D., et al. (2007). A phase I/II study Tallarida, R.J. (2006). An overview of drug combination analysis with isobolo-
of galiximab (an anti-CD80 monoclonal antibody) in combination with rituxi- grams. J. Pharmacol. Exp. Ther. 319, 1–7.
mab for relapsed or refractory, follicular lymphoma. Ann. Oncol. 18, 1216– Taylor, R.P., and Lindorfer, M.A. (2008). Immunotherapeutic mechanisms of
1223. anti-CD20 monoclonal antibodies. Curr. Opin. Immunol. 20, 444–449.
Leonard, J.P., Schuster, S.J., Emmanouilides, C., Couture, F., Teoh, N., The International Non-Hodgkin’s Lymphoma Prognostic Factors Project.
Wegener, W.A., Coleman, M., and Goldenberg, D.M. (2008). Durable complete (1993). A predictive model for aggressive non-Hodgkin’s lymphoma. N. Engl.
responses from therapy with combined epratuzumab and rituximab: final J. Med. 329, 987–994.
results from an international multicenter, phase 2 study in recurrent, indolent, Uno, S., Kinoshita, Y., Azuma, Y., Tsunenari, T., Yoshimura, Y., Iida, S., Kiku-
non-Hodgkin lymphoma. Cancer 113, 2714–2723. chi, Y., Yamada-Okabe, H., and Fukushima, N. (2007). Antitumor activity of
Maddipatla, S., Hernandez-Ilizaliturri, F.J., Knight, J., and Czuczman, M.S. a monoclonal antibody against CD47 in xenograft models of human leukemia.
(2007). Augmented antitumor activity against B-cell lymphoma by a combina- Oncol. Rep. 17, 1189–1194.

Cell 142, 699–713, September 3, 2010 ª2010 Elsevier Inc. 713

A C-Type Lectin Collaborates with a
CD45 Phosphatase Homolog to Facilitate
West Nile Virus Infection of Mosquitoes
Gong Cheng,1 Jonathan Cox,1 Penghua Wang,1 Manoj N. Krishnan,1 Jianfeng Dai,1 Feng Qian,2 John F. Anderson,3
and Erol Fikrig1,4,*
1Sectionof Infectious Diseases
2Sectionof Rheumatology
Department of Internal Medicine, Yale University School of Medicine, New Haven, CT 06520, USA
3Department of Entomology, Connecticut Agricultural Experiment Station, New Haven, CT 06504, USA
4Howard Hughes Medical Institute, Chevy Chase, MD 20815, USA

DOI 10.1016/j.cell.2010.07.038

SUMMARY wide (Nir et al., 1968; Turell et al., 2000). In the United States,
the most important vectors are Culex pipiens in the east, Culex
West Nile virus (WNV) is the most common tarsalis in the midwest and west, and Culex quinquefasciatus in
arthropod-borne flavivirus in the United States; the southeast (Hayes et al., 2005). WNV has also been isolated
however, the vector ligand(s) that participate in infec- from Aedes, Ochlerotatus, and Culisetam mosquitoes (http://
tion are not known. We now show that an Aedes
aegypti C-type lectin, mosGCTL-1, is induced by Aedes aegypti, a member of the Culicinae subfamily, is a major
vector for numerous flaviviruses (Gould and Solomon, 2008).
WNV, interacts with WNV in a calcium-dependent
A. aegypti is ideal for viral pathogenesis studies because these
manner, and facilitates infection in vivo and in vitro. mosquitoes are easy to cultivate and the genome has been char-
A mosquito homolog of human CD45 in A. aegypti, acterized (Gubler, 1998; Nene et al., 2007; Halstead, 2008).
designated mosPTP-1, recruits mosGCTL-1 to A. aegypti is readily susceptibility to infection with WNV in
enable viral attachment to cells and to enhance viral the laboratory, and the virus rapidly disseminates throughout
entry. In vivo experiments show that mosGCTL-1 and most of the mosquito after the blood meal. As with Culex spp.,
mosPTP-1 function as part of the same pathway A. aegypti is a threat for the transmission of WNV to humans
and are critical for WNV infection of mosquitoes. (Vanlandingham et al., 2007).
A similar phenomenon was also observed in Culex C-type lectins are a group of carbohydrate-binding proteins
quinquefasciatus, a natural vector of WNV, further (Zelensky and Gready, 2005). Several members of this family
demonstrating that these genes participate in WNV are highly expressed by immune cells, including monocytes,
macrophages, and dendritic cells (DCs), and play a central
infection. During the mosquito blood-feeding pro-
role in activating host defenses (Robinson et al., 2006). Human
cess, WNV infection was blocked in vivo with mannose-binding lectin (MBL) is a pattern recognition molecule
mosGCTL-1 antibodies. A molecular understanding of the innate immune system that binds to sugars on the surface
of flaviviral-arthropod interactions may lead to strat- of invading pathogens, leading to opsonization, phagocytosis,
egies to control viral dissemination in nature. and activation of the complement pathway (Neth et al., 2002).
In contrast, some C-type lectins are recruited to facilitate flavivi-
INTRODUCTION ral infection. In mammals, two membrane C-type lectins,
DC-SIGN (CD209) and L-SIGN (CD209L), interact with flavivi-
West Nile virus (WNV) is maintained in a bird-mosquito transmis- ruses via high mannose glycans on viral glycoproteins (Klimstra
sion cycle and has become the most common arthropod-borne et al., 2003) and are essential host cell factors exploited by
flavivirus in the United States. Humans, horses, and other nona- dengue virus (DENV) and WNV to invade immature DCs and
vian vertebrates are incidental hosts (Monath and Heinz, 1996). macrophages (Geijtenbeek et al., 2000; Soilleux et al., 2002;
Infection in man can result in fever, meningitis, or encephalitis, Tassaneetrithep et al., 2003; Davis et al., 2006). Another
among other symptoms (Hubálek and Halouzka, 1999; Leis C-type lectin, the mannose receptor (MR), also interacts with
et al., 2002). Approved human vaccines or therapeutics are not the DENV envelope protein and may enhance viral attachment
available and preventive measures largely focus upon mosquito to phagocytes (Miller et al., 2008). A recent study identified
control (Reisen and Brault, 2007). C-type lectin domain family 5, member A (CLEC5A), as
The ability of different mosquito species to transmit WNV a DENV receptor. The association between CLEC5A and
varies widely. Culex spp. are the major vectors for WNV world- DENV does not result in viral entry, but rather stimulates the

714 Cell 142, 714–725, September 3, 2010 ª2010 Elsevier Inc.

Figure 1. Effect of WNV Infection on mosGCTL-1 Expression
(A) Viral distribution in A. aegypti. WNV (1 3 103 M.I.D50) was inoculated into the mosquito thorax.
(B–E) mosGCTL-1 messenger RNA (mRNA) was induced by viral infection in whole A. aegypti (B), salivary glands (C), hemolymph (D), and midgut (E). Total RNA
was isolated from various tissues or whole mosquitoes at 5 time points after viral infection.
(A–E) The viral load and mosGCTL-1 mRNA levels were determined by Taqman RT-QPCR and normalized with A. aegypti actin (AAEL011197). WNV (1 3 103
M.I.D50) was microinjected into each mosquito. Data are shown as the mean ± standard error (SEM). The experiments were repeated three times.
(F) Immunoblot to detect mosGCTL-1 in WNV-infected mosquitoes. Three WNV infected or control mosquitoes were pooled and homogenized. The supernatant
was then isolated, separated by SDS-PAGE, and probed with rabbit mosGCTL-1 antisera. UI, uninfected mosquitoes; I, WNV-infected mosquitoes; dpi, days
postinfection. Fifty micrograms of protein from mosquito lysates was loaded into each lane.
See also Figure S1 and Table S1.

release of proinflammatory cytokines, potentially contributing to well-known PTP, protein tyrosine phosphatase receptor type C
the pathogenesis of dengue hemorrhagic fever (Chen et al., (PTPRC, CD45), is important for thymocyte development and
2008). T cell activation (Byth et al., 1996; Trowbridge and Thomas,
Protein tyrosine phosphatases (PTPs) remove phosphate 1994) and is expressed on all nucleated cells of hemopoietic
groups from phosphorylated tyrosine residues and play critical origin (Thomas, 1989). The association between MBL and
roles in cell communication, shape, motility, proliferation, and CD45 in immature T cells influences thymocyte development
differentiation (Alonso et al., 2004; Mustelin et al., 2005). One (Baldwin and Ostergaard, 2001).

Cell 142, 714–725, September 3, 2010 ª2010 Elsevier Inc. 715

Figure 2. The Function of mosGCTL-1 in WNV Infection
(A and B) mosGCTL-1 silencing. The mock group was treated with the same amount of GFP dsRNA. mosGCTL-1- or GFP-dsRNA-treated mosquitoes were used
to isolate total RNA at several time points post dsRNA injection. mRNA levels were determined by SYBR Green RT-QPCR (A). mosGCTL-1-dsRNA-treated or
mock-treated mosquitoes were collected at 6 days after gene silencing. The supernatant was separated by SDS-PAGE and probed with rabbit mosGCTL-1
antisera (B).
(C and D) Silencing mosGCTL-1 impairs WNV (C), but not dengue virus (D), infection. The viral burden was examined at day 6 after infection. WNV or dengue virus
(10 M.I.D50) was used to challenge mosquitoes. The viral load was determined by Taqman RT-QPCR and normalized with A. aegypti actin. The result shown was
representative of four independent experiments.
(E) The role of the mosGCTL paralogues in WNV infection. The sample number was no less than 12 mosquitoes in each group. The viral burden was determined by
Taqman RT-QPCR and normalized with A. aegypti actin. *p < 0.05. The results were pooled from two independent experiments.
(F) WNV induces mosGCTL-1 homolog expression in C. quinquefasciatus. WNV-infected and mock mosquitoes were collected at 6 days after infection. Culex
mosGCTL-1 mRNA was determined by SYBR Green RT-QPCR and normalized with C. quinquefasciatus actin (CPIJ012570). The experiment was repeated three
times with similar results.
(G) Efficiency of Culex mosGCTL-1 RNA interference. The mock group was treated with the same amount of GFP dsRNA. mRNA levels were determined by SYBR
Green RT-QPCR.
(H) Silencing of Culex mosGCTL-1 decreases the WNV burden in C. quinquefasciatus. The viral burden was determined by Taqman RT-QPCR. The results were
combined from three independent experiments.

716 Cell 142, 714–725, September 3, 2010 ª2010 Elsevier Inc.

Genome-wide RNA interference (RNAi) screening studies mosquitoes. Total RNA from mock-injected and mosGCTL-1
have revealed several hundred host factors that influence WNV dsRNA-injected mosquitoes was analyzed by RT-QPCR. Com-
or DENV infection in human or Drosophila cell lines and have pared to the mock group, the target gene was silenced 6- to
identified cellular pathways that have a role in viral internaliza- 10-fold from day 3 through 9 (Figures 2A and 2B). WNV was
tion, replication, assembly, or secretion (Krishnan et al., 2008; therefore inoculated into mosquitoes on day 3 after dsRNA treat-
Sessions et al., 2009). However, the relationship between flavivi- ment, and the viral load was quantified on day 9. There was
ruses and mosquitoes is not well understood. We now examine a 3-fold reduction in the WNV burden in mosGCTL-1 silenced
WNV-mosquito interactions by using A. aegypti and Culex quin- A. aegypti (Figure 2C, p < 0.0001) compared to controls.
quefasciatus, characterize the expression profile of mosquito In contrast, the burden of a related flavivirus, dengue virus,
homologs of human susceptibility proteins (Krishnan et al., was unaffected by mosGCTL-1 silencing (Figure 2D), suggesting
2008) in response to viral infection, and identify a lectin-based that mosGCTL-1 has a specific role in WNV infection of
pathway that is critical for viral infection of mosquitoes. mosquitoes.
As mosGCTL-1 is part of a multigene family with 25 members
RESULTS and shares 22%–48% identity with many of these genes, we
examined the role of the mosGCTL paralogues in WNV and
A recent RNAi screening study characterized 283 human DENV infection. Twenty-one of these 25 genes were expressed
proteins that facilitate WNV infection (Krishnan et al., 2008). in adult female A. aegypti (Tables S1 and S2). Since AAEL014382
We have now identified 215 homologs of these genes in and AAEL014390 share greater than 95% identity with
A. aegypti. The expression of 32 genes was altered by the pres- AAEL011607 and AAEL011619, respectively, AAEL011607 and
ence of WNV in mosquitoes, potentially suggesting a role in viral AAEL011619 were selected for functional assays. We silenced
infection of the arthropod (Table S1 available online). To examine each of the expressed mosGCTL genes and then infected the
their function, we therefore silenced each of these 32 genes in mosquitoes with virus. In the WNV challenge, two additional
mosquitoes with RNAi and assessed the effect on viral load. mosGCTL subtypes (AAEL000556 and AAEL011610) showed
Knockdown of 13 genes significantly altered the viral burden in the similar phenotype to mosGCTL-1 (Figure 2E), and in DENV
the vector (Table S1 and Figure S1). One of these genes, infection, silencing of 5 mosGCTL-related genes reduced the
AAEL000563, exhibited the most dramatic reduction in viral DENV burden (Figure S2).
load after silencing and was selected as the target for further Culex mosquitoes are a common vector for WNV in nature.
investigation. AAEL000563 belongs to the A. aegypti galac- We therefore extended the studies to Culex quinquefasciatus,
tose-specific binding C-type lectin family, shares 26% amino a example of this species that is abundant in the southeast
acid identity with human mannose-binding lectin, and was U.S. We identified a mosGCTL-1 homolog (CPIJ010995, Culex
designated as mosGCTL-1 (mosquito galactose-specific binding mosGCTL-1) in C. quinquefasciatus that has a higher degree
C-type lectin). of similarity (64%) with mosGCTL-1 than other homologs.
We then determined whether Culex mosGCTL-1 had a similar
role in WNV infection. Culex mosGCTL-1 was upregulated by
mosGCTL-1 Expression Increases during WNV Infection WNV infection (Figure 2F). Silencing of Culex mosGCTL-1 in
of A. aegypti C. quinquefasciatus reduced the WNV burden compared to the
To further investigate the relationship between WNV infection control group (Figures 2G and 2H, p < 0.0001), suggesting that
and mosGCTL-1 expression, we determined the viral load and Culex mosGCTL-1 plays a role in facilitating WNV infection in
mosGCTL-1 level in selected A. aegypti tissues after the inocula- Culex mosquitoes.
tion of WNV into female mosquitoes. WNV was readily detect-
able 4 days after infection, and the viral level subsequently
increased. The salivary glands and hemolymph had the highest mosGCTL-1 Interacts with WNV
levels of WNV, while the viral load in the midgut was substantially Our results show that silencing mosGCTL-1 in A. aegypti and
lower (Figure 1A). mosGCTL-1 expression was induced in C. quinquefasciatus reduces WNV infection. The mechanism
diverse tissues over time, including the salivary glands, hemo- by which mosGCTL-1 facilitates viral infection in mosquitoes
lymph, and midgut (Figures 1B–1E). Immunoblots also demon- was, therefore, investigated. We first generated A. aegypti
strated an increased amount of mosGCTL-1 in WNV-infected mosGCTL-1 protein in a Drosophila expression system (Fig-
mosquitoes (Figure 1F). ure 3A) and investigated whether mosGCTL-1 associates with
WNV envelope (E) protein. Immunoprecipitation experiments
mosGCTL-1 Facilitates WNV Infection in A. aegypti revealed that these two proteins strongly interacted in a
Double-stranded RNA (dsRNA)-mediated gene silencing studies calcium-dependent manner (Figure 3B). We then determined,
then assessed the role of mosGCTL-1 in WNV infection of by ELISA, that mosGCTL-1 bound to WNV virions. mosGCTL-1

(A, C, D and H) Statistical analysis was done with the Mann-Whitney test in all experiments. Each dot represents the mRNA levels in an individual mosquito.
The horizontal line depicts the medians.
(E–G) The Mann-Whitney test was used for statistical analysis. Data are shown as the mean ± standard error (SEM).
See also Figure S2 and Table S2.

Cell 142, 714–725, September 3, 2010 ª2010 Elsevier Inc. 717

Figure 3. mosGCTL-1 Interacts with WNV
(A) Recombinant mosGCTL-1, produced in Drosophila cells, and purified with a Ni-His column (left). Glycosylated mosGCTL-1 was detected with a V5-HRP mAb
(right). The control was the supernatant of mock-infected S2 cells.
(B) mosGCTL-1 interacted with WNV E protein in a coimmunoprecipitation assay. The protein complex was pulled down with a flavivirus E mAb and probed with
anti-V5-HRP mAb. The experiments were repeated three times.
(C) mosGCTL-1 captured West Nile virions by ELISA. The interaction was determined with a flavivirus E mAb. Data are expressed as the mean ± standard error
from three independent experiments.
(D) Comicroinjection of the purified mosGCTL-1 with WNV enhanced virus infection in A. aegypti. The various amount of purified mosGCTL-1 were premixed with
WNV (10 M.I.D50 per mosquito) for 1 hr at 4 C. The mixture was microinjected into mosquito and compared to the control group inoculated with the same amount
WNV. The mosquitoes were collected at 3 days (i), 6 days (ii), and 9 days (iii) after WNV inoculation. Total RNA was isolated to determine the viral burden by
Taqman RT-QPCR and normalized with A. aegypti actin. Each dot represents the mRNA level in one mosquito. The horizontal line represents the medians.
n.s., nonsignificance (p > 0.05). The Mann-Whitney test was used for analysis. Three independent experiments yielded similar data.
(E) The association between mosGCTL-1 and WNV in A .aegypti hemolymph. Hemolymph was collected from WNV-infected or mock-infected mosquitoes
for immunofluorescence staining. WNV E protein was stained with anti-mouse IgG Alexa-488 (green), and mosGCTL-1 was identified with anti-rabbit
IgG Alexa-546 (red). Nuclei were stained blue with To-Pro-3 iodide. The white arrow represents the induced expression of mosGCTL-1 in WNV-infected

718 Cell 142, 714–725, September 3, 2010 ª2010 Elsevier Inc.

efficiently captured virus in the presence of calcium, and high levels of mosPTP-1 in the cytoplasm also had substantial
this interaction was inhibited by EDTA (Figure 3C). Since this mosGCTL-1 on the cell surface (Figure 4F). These results
association suggests a role during infection, we premixed suggest that mosGCTL-1 binds mosPTP-1.
mosGCTL-1 and WNV and then coinjected the combination To test whether mosPTP-1 has a conserved role in Culex spp.,
into A. aegypti and determined the viral burden. mosGCTL-1 we identified the mosPTP-1 homolog from the Culex quinquefas-
protein significantly enhanced the viral load at days 3 (Fig- ciatus genome (CPIJ014098, Culex mosPTP-1). Silencing of this
ure 3D, i, p = 0.0001) and 6 (Figure 3D, ii, p < 0.0001) after viral gene in Culex also influenced the viral burden (Figures S4B and
challenge. S4C), similar to mosPTP-1 in A. aegypti (Figure 4B), suggesting
To further characterize the association between mosGCTL-1 that mosPTP-1 in both these mosquito species have a similar
and WNV infection in vivo, we used immunofluorescence to role in WNV infection.
examine the hemolymph and salivary glands of WNV-infected
A. aegypti at different time points. Colocalization of mosGCTL-1 The mosGCTL-1/mosPTP-1 Pathway Has a Dominant
and WNV was clearly observed in both tissues at various inter- Role in WNV Infection
vals after WNV infection (Figure 3E and Figure S3). In the hemo- mosGCTL-1 and mosPTP-1 each facilitate WNV infection of
lymph, some hemocytes were highly infected by WNV, and A. aegypti in vivo and in vitro. We therefore assessed whether
mosGCTL-1 was induced in these infected hemocytes mosGCTL-1 and mosPTP-1 cooperate to enable viral infection.
(Figure 3E). We counted the number of mosGCTL-1-positive WNV E protein, mosGCTL-1, and mosPTP-1-Ex form a complex
and WNV-infected hemocytes by microscopy. All these infected in which mosGCTL-1 is the key factor linking the other two
cells stained positive for mosGCTL-1. The virus spread rapidly to proteins (Figure 5A). WNV E protein did not interact directly
the salivary glands after inoculation (Figure S3), consistent with with mosPTP-1-Ex (Figure 5A). We then examined whether
the Q-PCR data (Figure 1A). mosGCTL-1 was identified on the mosPTP-1-expressing cells could recruit more WNV in the pres-
basement membrane of salivary glands in uninfected mosqui- ence of mosGCTL-1. mosGCTL-1 and WNV were added to cells
toes, and throughout this tissue after infection, suggesting that and incubated at 4 C for membrane attachment. After washing,
mosGCTL-1 was induced by WNV infection of the salivary the cells were transferred to room temperature and collected at
glands (Figure S3). different time points to determine the viral burden (Krishnan
et al., 2007). mosPTP-1-expressing cells incorporated up to
mosPTP-1 Captures mosGCTL-1 onto the Cell Surface 5- to 10-fold more WNV in the presence of mosGCTL-1, com-
Human mannose-binding lectin (MBL) (Zelensky and Gready, pared to control groups (Figure 5B). We then examined the role
2005) and mosGCTL-1 are both secreted. Human MBL is of the mosGCTL-1/mosPTP-1 pathway for WNV infection in
thought to interact with several surface receptors to exert pleo- A. aegypti. We knocked down the mosPTP-1 gene with dsRNA
trophic effects (Baldwin and Ostergaard, 2001; Arnold et al., and then inoculated the mosGCTL-1/WNV mixture at 3 days
2006). We therefore postulated that secreted mosGCTL-1 after RNAi silencing. Silencing of mosPTP-1 interfered with the
captures WNV and presents it to a ligand on the cell surface, ability of mosGCTL-1 to facilitate WNV infection (shown in Fig-
thereby facilitating viral entry. To assess this, we identified 11 ure 3D) at 3 and 6 days after challenge (Figure 5C). The viral
A. aegypti homologs of human proteins that putatively interact burden of the mosPTP-1 RNAi/mosGCTL-1/WNV group was
with human MBL (Table S3) from the Human Protein Reference 2- to 3-fold lower than that of the mock/mosGCTL-1/WNV group
Database ( and examined their roles in (p < 0.001) and decreased to a level similar to that of the group
WNV infection of A. aegypti (Figure S4A). One mosquito CD45 that did not receive mosGCTL-1, suggesting that mosGCTL-1
homolog (AAEL013105, mosPTP-1) exhibited a phenotype and mosPTP-1 cooperate in the same pathway to enhance
similar to that of mosGCTL-1. The viral burden was significantly WNV infection. To further determine the relationship between
decreased in the mosPTP-1 dsRNA-treated mosquitoes (p < mosGCTL-1 and mosPTP-1 in infection, we silenced both of
0.002) (Figures 4A and 4B). To further determine whether these genes with dsRNA. These genes were successfully
mosPTP-1 interacts with mosGCTL-1, we cloned and expressed knocked down in mosquitoes in the cosilenced group (Figures
the extracellular region of mosPTP-1 (mosPTP-1-Ex) in a S5A and S5B). The decrease in the viral burden was similar in
Drosophila cell line (Figure 4C). mosGCTL-1 strongly interacted the cosilenced and individually silenced groups (Figure S5C),
with mosPTP-1-Ex in a coimmunoprecipitation assay (Figure 4D). suggesting that mosPTP-1 is the dominant downstream
We then expressed the mosPTP-1 gene, including the trans- receptor for mosGCTL-1 in the process of WNV infection of
membrane region (91 bp–2202 bp), in S2 cells (Figure 4C) and A. aegypti.
examined the ability of mosPTP-1 to bind mosGCTL-1 on the To better understand the association between mosGCTL-1,
cell surface. A mock DNA vector transfected stable S2 cell line mosPTP-1, and WNV in vivo, we examined the distribution of
was used as control. More than 40% of the mosPTP-1-express- mosPTP-1 in A. aegypti. mosPTP-1 was highly expressed in
ing cells bound mosGCTL-1 and showed double-positive stain- various mosquito tissues but not induced by WNV infection
ing in FACS (Figure 4E, i) compared with controls (Figure 4E, ii, (Figure S4D). The salivary glands and hemolymph were sites of
iii). Confocal microscopy demonstrated that cells expressing abundant mosPTP-1 expression, while expression in midgut

hemocytes. The yellow arrows show the infected hemocytes. Images were examined with a Zeiss LSM 510 Meta Confocal Microscope 633 objective
See also Figure S3.

Cell 142, 714–725, September 3, 2010 ª2010 Elsevier Inc. 719

Figure 4. mosPTP-1 Captures mosGCTL-1 onto the Cell Surface
(A) RNAi efficiency for the mosPTP-1 gene at 6 days after dsRNA treatment. The amount of mosPTP-1 mRNA was determined by SYBR Green RT-QPCR and
normalized with A. aegypti actin. Data are represented as the mean ± standard error.
(B) Silencing mosPTP-1 decreased WNV infection. The viral burden was measured at day 6. WNV (10 M.I.D50) was injected into each mosquito. The viral load was
determined by Taqman RT-QPCR and normalized with A. aegypti actin. Statistical analysis was done with the Mann-Whitney test. The horizontal line depicts the
medians. The result shown is a combination of three independent experiments.
(C) Expression of recombinant mosPTP-1 and mosPTP-1-Ex in S2 cells. mosPTP-1 and mosPTP-1-Ex genes were isolated from complementary DNA (cDNA)
library of A. aegypti and expressed as recombinant proteins with an HA tag at the N terminus. The left panel is mosPTP-1, and the right panel is mosPTP-1-EX,
probed by anti-HA tag mAb in western blot. The control was the products from S2 cells transfected with mock DNA vector.
(D) mosGCTL-1 interacts with mosPTP-1-Ex peptide by coimmunoprecipitation. The protein complex was pulled down with a rabbit HA antibody and probed with
a V5-HRP mAb. The experiments were repeated three times.

720 Cell 142, 714–725, September 3, 2010 ª2010 Elsevier Inc.

was comparatively lower (Figures S4E–S4G). We next generated into mosquitoes binds to secreted mosGCTL-1 in the hemo-
mosPTP-1 antibody in mice, which recognized native mosPTP-1 lymph, thereby forming a complex in the extracellular milieu
protein (Figure S5D) and the mosPTP-1-Ex expressed by S2 that has the ability to interact with the membrane protein,
cells (Figure S5E). We then determined the relationship between mosPTP-1, to facilitate cellular entry. The virus rapidly replicates
mosGCTL-1, mosPTP-1, and WNV in salivary glands by immu- in the mosquito thorax. This induces additional mosGCTL-1
nofluorescence. mosPTP-1 was copiously expressed on the expression, which accelerates formation of the mosGCTL-1/
cell surface of salivary glands (Figure 5D), thereby providing WNV complex—enabling WNV to invade different mosquito
additional data to complement and extend the initial QPCR tissues and enhancing viral spread throughout the mosquito
expression data (Figure S4E). Several regions in the salivary body. This mechanism, which involves WNV associating with
glands with substantial mosPTP-1 also demonstrated staining mosGCTL-1 and then being captured by mosPTP-1 onto the
for mosGCTL-1 and WNV (Figure 5D). cell surface in mosquitoes, suggests that an extracellular soluble
protein is an important receptor for flavivirus in arthropods.
mosGCTL-1 Antisera Interferes with WNV Infection mosGCTL-1 shares homology with human MBL. In mammals,
of Mosquitoes MBL is a pattern recognition molecule that recognizes carbohy-
Disruption of the transfer of WNV from the vertebrate to drate moieties on invading microbes (Neth et al., 2002). As exam-
arthropod host could theoretically diminish viral dissemination ples, MBL interacts with HIV envelope protein (gp120) (Saifuddin
in nature. We therefore investigated whether mosGCTL-1 anti- et al., 2000) and HBV surface antigen (HBsAg) (Chong et al., 2005)
sera reduces WNV infection in mosquitoes during the blood and has a role in the opsonization of HIV (Ezekowitz et al., 1989).
meal. We generated mosGCTL-1 antisera in rabbits and showed In these processes, MBL associates with serine proteases,
that mosGCTL-1 antisera interfered with mosGCTL-1 binding to MASPs, and activates the complement system (Neth et al.,
WNV E protein in vitro (Figure 6A). Then, we examined whether 2002). Homologs of the proteins that associate with mammalian
mosGCTL-1 antisera influenced the ability of WNV to infect MBL have not been found in A. aegypti (Table S3), suggesting that
mosquitoes during a blood meal. We mixed mosGCTL-1 antisera the A. aegypti mosGCTL-1 may have different physiological func-
and WNV with fresh whole blood and performed membrane tions than mammalian MBL. Invertebrates lack antibody- and
blood feeding with a Hemotek. Seven days later, mosquitoes interferon-based immune responses (Cheng et al., 2009). Since
were sacrificed to determine the infectivity rate. mosGCTL-1 lectin expression is significantly upregulated by microbial infec-
antisera efficiently blocked WNV infection of A. aegypti. The tion, these molecules are presumed to participate in nonself
number of infected mosquitoes was reduced in mosGCTL-1 recognition and pathogen resistance (Wilson et al., 1999; Tanji
antisera treated groups, compared to mock group, by QPCR et al., 2006). Indeed, recent studies have shown that a comple-
(Figure 6B) or a 50% tissue culture infective doses (TCID50) assay ment-like system exists in the hemolymph of Anopheles gambiae
(Figure 6C). Hence, a humoral response against mosGCTL-1 in and mediates parasite killing (Blandin et al., 2004; Povelones
a vertebrate host may alter WNV infection of mosquitoes during et al., 2009). It is possible that mosGCTL-1 and other subtypes
the feeding process. This hypothetically affords a strategy to in this family, similar to their mammalian homologs, may normally
develop a transmission blocking vaccine to control WNV recognize most pathogens and be involved in the arthropod
dissemination in nature. complement-like system. Nevertheless, our studies showed
that the expression of mosGCTL-1 is induced by WNV infection.
DISCUSSION The induced mosGCTL-1 that then binds to virus amplifies WNV
infection. Overall, these suggest a critical role for mosGCTL-1 in
As mosquitoes are prominent vectors for flaviviruses, specific WNV infection of mosquitoes.
interactions between the virus and arthropod likely enhance In the mammalian host, the association between MBL and the
pathogen survival. In the mammalian host, C-type lectins such CD45 external domain primarily occurs in immature T cells and
as DC-SIGN and the mannose receptor augment viral entry affect the development of thymocytes (Baldwin and Ostergaard,
into specific DCs and macrophages (Tassaneetrithep et al., 2001). Mammalian CD45 is expressed on the hemopoietic-orig-
2003; Davis et al., 2006; Miller et al., 2008). Our results show inated nucleated cells (Thomas, 1989); however, the mosquito
that a secreted mosquito C-type lectin, mosGCTL-1, binds to CD45 homolog, mosPTP-1, does not appear to be restricted to
WNV in a calcium-dependent manner and enhances viral infec- particular cells. As a transmembrane protein, mosPTP-1 was
tion. A mosquito homolog of human CD45 (mosPTP-1) recruits abundantly detected in the salivary glands and hemolymph of
mosGCTL-1 to facilitate viral attachment to cells. Based on our mosquitoes. The pattern of mosPTP-1 expression correlated
findings, we envision a model whereby WNV that is inoculated with the distribution of WNV in A. aegypti. In our model, after

(E) mosPTP-1 captured mosGCTL-1 to the cell surface by flow cytometry. A stable cell line was generated to express mosPTP-1 in S2 cells. The purified
mosGCTL-1 was inoculated with mosPTP-1-expressing cells at 4 C. An empty DNA vector transfected stable S2 cell line was used as the control. The interaction
between mosPTP-1 and mosGCTL-1 was investigated by FACS. mosPTP-1 was stained by Alexa-488; mosGCTL-1 was stained by Phycoerythrin (PE). Three
independent experiments yielded similar results, and one representative study is shown in this figure.
(F) Confocal microscopy to examine for mosPTP-1 and mosGCTL-1. mosGCTL-1 was stained with Alexa-488 (green) and mosPTP-1 was identified with Alexa-
546 (red). Nuclei were stained by To-Pro-3 iodide (blue). The images were collected using a Zeiss LSM 510 Meta Confocal Microscope 633 objective lens.
The arrows represent the overlap between mosPTP-1 and mosGCTL-1.
See also Figure S4 and Table S3.

Cell 142, 714–725, September 3, 2010 ª2010 Elsevier Inc. 721

Figure 5. The mosGCTL-1/mosPTP-1 Pathway in WNV Infection
(A) mosGCTL-1 associated with WNV E protein bound to mosPTP-1-Ex. The protein complex was pulled down with an HA antibody and probed with V5-HRP and
flavivirus E protein antibody. The experiment was repeated three times.
(B) mosPTP-1-expressing cells recruit WN virions in the presence of mosGCTL-1. mosGCTL-1 and WNV were added to cells and incubated at 4 C for 1 hr, for
membrane attachment. Cells were gently washed three times with cold PBS and were then moved to room temperature. At different time points—0, 15, and
60 min—cells were collected for total RNA isolation. Control cells were the empty DNA vector transfected stable S2 cell line. Viral attachment was determined
by SYBR Green RT-QPCR and normalized with Drosophila actin 5C (CG4027). Statistical analysis was done with ANOVA. Data are represented as the mean ±
standard error. The results are representative of three independent experiments.
(C) Silencing of mosPTP-1 impairs the function of mosGCTL-1. The mosPTP-1 gene was knocked down with dsRNA treatment. Then the mosGCTL-1/WNV
mixture was inoculated at 3 days after RNAi silencing. The virus burden was measured at 3 (i) and 6 (ii) days after the introduction of virus by Taqman RT-QPCR.
WNV (10 M.I.D50) and mosGCTL-1 (100 pg) were inoculated into each mosquito. Each dot represents the mRNA level in an individual mosquito. Statistical analysis
was done with the Mann-Whitney test. The horizontal line depicts the medians. The result shown is the combination of five independent experiments.
(D) Immunostaining of mosGCTL-1, mosPTP-1, and WNV in A. aegypti salivary glands. mosPTP-1 was stained with anti-mouse IgG Alexa-488 (green),
mosGCTL-1 was identified by anti-rabbit IgG Alexa-546 (red), and WNV E protein was probed with horse anti-E protein IgG and detected by anti-horse IgG
Alexa-633 (blue). The arrows show the sites of overlap between mosGCTL-1, mosPTP-1, and WNV at 6 days after infection. LL, lateral lobe; ML, median lobe
in female A.aegypti salivary glands. Images were examined with a Zeiss LSM 510 Meta Confocal Microscope 253 objective lens.
See also Figure S5.

722 Cell 142, 714–725, September 3, 2010 ª2010 Elsevier Inc.

Figure 6. mosGCTL-1 Antiserum Interferes
with WNV Infection of Mosquitoes
(A) mosGCTL-1 antisera blocked the interaction
between mosGCTL-1 and WNV E protein.
mosGCTL-1 antisera or mock sera was diluted
1000-fold. The protein complex was pulled down
with a V5 mAb, and WNV E protein was detected
with an E protein antibody.
(B and C) mosGCTL-1 antisera interrupted WNV
infection during the blood meal. The antisera or
control sera were diluted 100- or 2000-fold with
fresh whole blood containing 5 3 106 pfu/ml
WNV. Membrane blood-feeding was then per-
formed with a Hemotek. Seven days later, mosqui-
toes were sacrificed to determine the infectivity
rate by Taqman RT-QPCR (B) and TCID50 (C).
Each group included 50 mosquitoes in the QPCR
assay and 32 mosquitoes in TCID50 assay. One
dot corresponds to a mosquito. n, the number of
mosquitoes in each group. The result is represen-
tative of three independent experiments.

our understanding of flavivirus-arthropod

interactions and may aid in the develop-
ment of strategies to target selected
points in the flaviviral life cycle and inter-
fere with these pathogens in nature.


Mosquitoes and Viruses

A. aegypti and C. quinquefasciatus mosquitoes
were maintained in a sugar solution at 27 C and
80% humidity according to standard rearing
procedures (Keene et al., 2004; Xi et al., 2008).
binding to mosGCTL-1, WNV binds to membrane bound WNV strain 2471 and DENV-2 (DENV New Guinea C strain) were passaged
mosPTP-1. This implies that mosGCTL-1 and mosPTP-1 are in mosquito C6/36 cells (Hanna et al., 2005; Lin et al., 2007; Krishnan et al.,
recruited as receptors to facilitate cellular invasion by WNV. 2008). The titer of WNV for cell culture was determined by a plaque formation
assay as described previously (Bai et al., 2007). The viruses for in vivo exper-
Mosquito control is a common strategy to influence
iments were titrated in mosquitoes through thoracic microinjection. The dose
WNV numbers in nature (van der Meulen et al., 2005; Dauphin of viruses was determined by 10-fold serial dilutions (i.e., 10 4, 10 5, 10 6, and
and Zientara, 2007). The increase in viral spread and fatalities 10 7) in PBS. The mosquitoes (12 in each group) were inoculated in the thorax
over the last decade (Reisen and Brault, 2007; Lindsey et al., by microinjection with 300 nl of diluted virus. On day 6, the mosquitoes were
2009) ( sacrificed, total RNA was isolated, and the viral load determined by RT-
suggests that additional strategies could assist in combating QPCR (Figure S1). The 50% mosquito infective dose (M.I.D50) was estimated
by the Reed-Muench method (Pizzi, 1950).
WNV. For arthropod-borne microbes, vector ligands that interact
with pathogens are potential targets for interfering with Gene Silencing and Viral Challenge in Mosquitos
the successful acquisition of the microbe from the vertebrate dsRNA synthesis was performed as described previously (Brackney et al.,
host. As an example, blocking the tick gut receptor for the 2008). The primers are shown in Table S4. For silencing the target genes, adult
Lyme disease agent limits the colonization of ticks by Borrelia female mosquitoes were kept on ice for 15 min and then transferred to a cold
burgdorferi (Pal et al., 2004). Our studies show that blocking tray to receive a systemic injection of dsRNA into the hemocoele. Two micro-
grams of dsRNA/300 nl in PBS was microinjected into the thorax of each
mosGCTL-1 within A. aegypti reduced the vector competence
mosquito. After a 3 day recovery period, the mosquitoes were microinjected
for WNV and interrupted the infective cycle of WNV. These with either WNV 10 M.I.D50/ 300 nl for functional studies or 1000 M.I.D50/
results indicate that it is theoretically possible to develop a trans- 300 nl for expression profile assays.
mission-blocking vaccine to interfere with the migration of WNV
from vertebrates to mosquito, thereby restricting viral dissemi- Purification of mosGCTL-1 with the Drosophila Expression System
mosGCTL-1 was amplified by RT-PCR from adult female A. aegypti. The
nation in the environment.
primers are shown in Table S4. The PCR product was subcloned into the
In summary, we identified a lectin-based pathway to facilitate
pMT/BiP/V5-His A vector (Invitrogen, Carlsbad, CA) and transfected into
flaviviral entry, in which mosGCTL-1 and mosPTP-1 are cascade Drosophila S2 cells in combination with the hygromycin selection vector
receptors in WNV infection of A. aegypti and C. quinquefascia- pCo-Hygro for stable transfection. The cells were selected through the
tus. Characterization of mosquito ligands for WNV enhances use of 300 mg/ml Hygromycin-B (Invitrogen) for 4 weeks. The resistant cells

Cell 142, 714–725, September 3, 2010 ª2010 Elsevier Inc. 723

were grown in spinner flasks, switched to Express Five serum-free medium Received: September 12, 2009
(GIBCO, Invitrogen) for 3 days, and induced with copper sulfate at a final Revised: April 5, 2010
concentration of 500 mM for 4 days. The culture medium was cleared by Accepted: July 27, 2010
centrifugation at 1000 3 g for 5 min and collected for protein purification Published online: August 26, 2010
with the Talon metal affinity resin (Clontech, Mountain View, CA). The
protein was eluted with 150 mM imidazole, extensively dialyzed against REFERENCES
PBS (pH 7.8), and concentrated by centrifugal filtration through a 5 kDa
filter (Millipore, Bedford, MA). The protein purity was checked by SDS- Alonso, A., Sasin, J., Bottini, N., Friedberg, I., Friedberg, I., Osterman, A., God-
PAGE, and immunoblots were performed with an anti-V5-HRP mouse zik, A., Hunter, T., Dixon, J., and Mustelin, T. (2004). Protein tyrosine phospha-
mAb (Invitrogen). tases in the human genome. Cell 117, 699–711.
Arnold, J.N., Dwek, R.A., Rudd, P.M., and Sim, R.B. (2006). Mannan binding
lectin and its interaction with immunoglobulins in health and in disease. Immu-
Isolation and Image of Mosquito Tissues
nol. Lett. 106, 103–110.
Salivary glands were dissected as the previously described (Coleman et al.,
2007). Tissues were isolated, placed on sialylated slides (PGC Scientific, Bai, F.W., Town, T., Pradhan, D., Cox, J., Ashish, Ledizet, M., Anderson, J.F.,
Gaithersburg, MD), washed in PBS, and fixed in 4% PFA at 37 C for 1 hr. Flavell, R.A., Krueger, J.K., Koski, R.A., and Fikrig, E. (2007). Antiviral peptides
For hemolymph isolation, the mosquitoes were anesthetized on ice tray, and targeting the west nile virus envelope protein. J. Virol. 81, 2047–2055.
then the proboscis was removed with forceps. Hemolymph was expelled in Baldwin, T.A., and Ostergaard, H.L. (2001). Developmentally regulated
a droplet using the tip of the proboscis upon pressure to the thorax. Only clear changes in glucosidase II association with, and carbohydrate content of, the
droplets were collected, to avoid contamination by the fat body (Han et al., protein tyrosine phosphatase CD45. J. Immunol. 167, 3829–3835.
1999). The liquid was allowed to be dried on sialylated slides and fixed in Blandin, S., Shiao, S.H., Moita, L.F., Janse, C.J., Waters, A.P., Kafatos, F.C.,
4% PFA. Samples were blocked in PBS with 1% BSA and 0.1% Triton and Levashina, E.A. (2004). Complement-like protein TEP1 is a determinant
X-100 at room temperature for 2 hr before antibody incubations. After staining of vectorial capacity in the malaria vector Anopheles gambiae. Cell 116,
by primary and secondary antibodies, slides were imaged with the Multi-Track 661–670.
mode of a Zeiss LSM 510 Meta Confocal Microscope.
Brackney, D.E., Foy, B.D., and Olson, K.E. (2008). The effects of midgut serine
proteases on dengue virus type 2 infectivity of Aedes aegypti. Am. J. Trop.
Membrane Blood Feeding Med. Hyg. 79, 267–274.
Fresh whole blood was obtained from C57BL/6J mice, placed in tubes with Byth, K.F., Conroy, L.A., Howlett, S., Smith, A.J., May, J., Alexander, D.R., and
anticoagulant, and centrifuged at 3000 rpm for 10 min to separate the serum Holmes, N. (1996). CD45-null transgenic mice reveal a positive regulatory role
and cells. The serum was collected and heat inactivated at 57 C for 1 hr. for CD45 in early thymocyte development, in the selection of CD4+CD8+
The blood cells were washed in PBS three times to remove the anticoagulant. thymocytes, and B cell maturation. J. Exp. Med. 183, 1707–1718.
The cells were resuspended with serum. Five million plaque forming units per Chen, S.T., Lin, Y.L., Huang, M.T., Wu, M.F., Cheng, S.C., Lei, H.Y., Lee, C.K.,
milliliter WNV and antisera (or preimmune sera) were added to the treated Chiou, T.W., Wong, C.H., and Hsieh, S.L. (2008). CLEC5A is critical for
blood. About 1 ml of the virus/sera/blood mixture was used on each Hemotek dengue-virus-induced lethal disease. Nature 453, 672–676.
feeder. The feeder was placed on the mosquito container and the mosquitoes Cheng, G., Zhao, X., Li, Z., Liu, X., Yan, W., Zhang, X., Zhong, Y., and Zheng, Z.
were allowed to feed for 1 hr. The container was transferred to 4 C for 20 min to (2009). Identification of a putative invertebrate helical cytokine similar to the
anesthetize the mosquitoes. Carefully selected fed mosquitoes were placed ciliary neurotrophic factor/leukemia inhibitory factor family by PSI-BLAST-
into new containers and reared for 7 days. The mosquitoes were sacrificed; based approach. J. Interferon Cytokine Res. 29, 461–468.
the whole mosquitoes were homogenized in RLT buffer (RNeasy Mini kit,
Chong, W.P., To, Y.F., Ip, W.K., Yuen, M.F., Poon, T.P., Wong, W.H., Lai, C.L.,
QIAGEN) to measure the viral load by RT-QPCR, or grinded in PBS for virologic
and Lau, Y.L. (2005). Mannose-binding lectin in chronic hepatitis B virus infec-
tion. Hepatology 42, 1037–1045.
Coleman, J., Juhn, J., and James, A.A.A. (2007). Dissection of midgut and
SUPPLEMENTAL INFORMATION salivary glands from Ae. aegypti mosquitoes. J. Vis. Exp. 5, 228.
Dauphin, G., and Zientara, S. (2007). West Nile virus: recent trends in diagnosis
Supplemental Information includes Extended Experimental Procedures, five and vaccine development. Vaccine 25, 5563–5576.
figures, and four tables and can be found with this article online at doi:10.
Davis, C.W., Nguyen, H.Y., Hanna, S.L., Sánchez, M.D., Doms, R.W., and Pier-
son, T.C. (2006). West Nile virus discriminates between DC-SIGN and
DC-SIGNR for cellular attachment and infection. J. Virol. 80, 1290–1301.
ACKNOWLEDGMENTS Ezekowitz, R.A., Kuhlman, M., Groopman, J.E., and Byrn, R.A. (1989).
A human serum mannose-binding protein inhibits in vitro infection by the
This work was supported by National Institutes of Health grants AI 50031 and human immunodeficiency virus. J. Exp. Med. 169, 185–196.
AI 070343. We thank Yue Zhang, Lei Liu, Lili Zhang, and Sukanya Narasimhan Geijtenbeek, T.B., Torensma, R., van Vliet, S.J., van Duijnhoven, G.C., Adema,
for help with the biochemical studies, Debbie Beck for technical assistance, G.J., van Kooyk, Y., and Figdor, C.G. (2000). Identification of DC-SIGN, a novel
and Michel Ledizet for providing the purified WNV E protein and the flaviviral dendritic cell-specific ICAM-3 receptor that supports primary immune
E protein antibodies. G.C. is a James Hudson Brown-Alexander Brown responses. Cell 100, 575–585.
Coxe Postdoctoral Fellow at the Yale School of Medicine. P.W. is supported Gould, E.A., and Solomon, T. (2008). Pathogenic flaviviruses. Lancet 371,
by a Career Development Award from Northeast Biodefense Center (U54- 500–509.
AI057158-Lipkin). E.F. is an investigator with the Howard Hughes Medical
Gubler, D.J. (1998). Dengue and dengue hemorrhagic fever. Clin. Microbiol.
Institute. G.C. and E.F. designed the experiments and wrote the manuscript;
Rev. 11, 480–496.
G.C. performed the majority of the experiments and analyzed data; J.C. assis-
ted the mosquito techniques; J.F.A. provided the A. aegypti and C. quinque- Halstead, S.B. (2008). Dengue virus-mosquito interactions. Annu. Rev. Ento-
fasciatus mosquitoes and helped with the mosquito feeding; J.D. and F.Q. mol. 53, 273–291.
assisted in FACS and confocal microscopy; and P.W., M.N.K. and J.D. Han, Y.S., Chun, J., Schwartz, A., Nelson, S., and Paskewitz, S.M. (1999).
provided suggestions for the project. All authors reviewed, critiqued, and Induction of mosquito hemolymph proteins in response to immune challenge
provided comments on the text. and wounding. Dev. Comp. Immunol. 23, 553–562.

724 Cell 142, 714–725, September 3, 2010 ª2010 Elsevier Inc.

Hanna, S.L., Pierson, T.C., Sanchez, M.D., Ahmed, A.A., Murtadha, M.M., and Pal, U., Li, X., Wang, T., Montgomery, R.R., Ramamoorthi, N., Desilva, A.M.,
Doms, R.W. (2005). N-linked glycosylation of west nile virus envelope proteins Bao, F., Yang, X., Pypaert, M., Pradhan, D., et al. (2004). TROSPA, an Ixodes
influences particle assembly and infectivity. J. Virol. 79, 13262–13274. scapularis receptor for Borrelia burgdorferi. Cell 119, 457–468.
Hayes, E.B., Komar, N., Nasci, R.S., Montgomery, S.P., O’Leary, D.R., and Pizzi, M. (1950). Sampling variation of the fifty percent end-point, determined
Campbell, G.L. (2005). Epidemiology and transmission dynamics of West by the Reed-Muench (Behrens) method. Hum. Biol. 22, 151–190.
Nile virus disease. Emerg. Infect. Dis. 11, 1167–1173. Povelones, M., Waterhouse, R.M., Kafatos, F.C., and Christophides, G.K.
Hubálek, Z., and Halouzka, J. (1999). West Nile fever—a reemerging mosquito- (2009). Leucine-rich repeat protein complex activates mosquito complement
borne viral disease in Europe. Emerg. Infect. Dis. 5, 643–650. in defense against Plasmodium parasites. Science 324, 258–261.
Keene, K.M., Foy, B.D., Sanchez-Vargas, I., Beaty, B.J., Blair, C.D., and Olson, Reisen, W., and Brault, A.C. (2007). West Nile virus in North America: perspec-
K.E. (2004). RNA interference acts as a natural antiviral response to O’nyong- tives on epidemiology and intervention. Pest Manag. Sci. 63, 641–646.
nyong virus (Alphavirus; Togaviridae) infection of Anopheles gambiae. Proc. Robinson, M.J., Sancho, D., Slack, E.C., LeibundGut-Landmann, S., and Reis
Natl. Acad. Sci. USA 101, 17240–17245. e Sousa, C. (2006). Myeloid C-type lectins in innate immunity. Nat. Immunol. 7,
Klimstra, W.B., Nangle, E.M., Smith, M.S., Yurochko, A.D., and Ryman, K.D. 1258–1265.
(2003). DC-SIGN and L-SIGN can act as attachment receptors for alphaviruses Saifuddin, M., Hart, M.L., Gewurz, H., Zhang, Y., and Spear, G.T. (2000). Inter-
and distinguish between mosquito cell- and mammalian cell-derived viruses. action of mannose-binding lectin with primary isolates of human immunodefi-
J. Virol. 77, 12022–12032. ciency virus type 1. J. Gen. Virol. 81, 949–955.
Krishnan, M.N., Sukumaran, B., Pal, U., Agaisse, H., Murray, J.L., Hodge, Sessions, O.M., Barrows, N.J., Souza-Neto, J.A., Robinson, T.J., Hershey,
T.W., and Fikrig, E. (2007). Rab 5 is required for the cellular entry of dengue C.L., Rodgers, M.A., Ramirez, J.L., Dimopoulos, G., Yang, P.L., Pearson,
and West Nile viruses. J. Virol. 81, 4881–4885. J.L., and Garcia-Blanco, M.A. (2009). Discovery of insect and human dengue
Krishnan, M.N., Ng, A., Sukumaran, B., Gilfoy, F.D., Uchil, P.D., Sultana, H., virus host factors. Nature 458, 1047–1050.
Brass, A.L., Adametz, R., Tsui, M., Qian, F., et al. (2008). RNA interference Soilleux, E.J., Morris, L.S., Leslie, G., Chehimi, J., Luo, Q., Levroney, E., Trow-
screen for human genes associated with West Nile virus infection. Nature sdale, J., Montaner, L.J., Doms, R.W., Weissman, D., et al. (2002). Constitutive
455, 242–245. and induced expression of DC-SIGN on dendritic cell and macrophage
Leis, A.A., Stokic, D.S., Polk, J.L., Dostrow, V., and Winkelmann, M. (2002). subpopulations in situ and in vitro. J. Leukoc. Biol. 71, 445–457.
A poliomyelitis-like syndrome from West Nile virus infection. N. Engl. J. Med. Tanji, T., Ohashi-Kobayashi, A., and Natori, S. (2006). Participation of a galac-
347, 1279–1280. tose-specific C-type lectin in Drosophila immunity. Biochem. J. 396, 127–138.
Lin, C.C., Yang, C.F., Tu, C.H., Huang, C.G., Shih, Y.T., Chuang, C.K., and Tassaneetrithep, B., Burgess, T.H., Granelli-Piperno, A., Trumpfheller, C.,
Chen, W.J. (2007). A novel tetraspanin C189 upregulated in C6/36 mosquito Finke, J., Sun, W., Eller, M.A., Pattanapanyasat, K., Sarasombath, S., Birx,
cells following dengue 2 virus infection. Virus Res. 124, 176–183. D.L., et al. (2003). DC-SIGN (CD209) mediates dengue virus infection of human
Lindsey, N.P., Hayes, E.B., Staples, J.E., and Fischer, M. (2009). West Nile dendritic cells. J. Exp. Med. 197, 823–829.
virus disease in children, United States, 1999-2007. Pediatrics 123, e1084– Thomas, M.L. (1989). The leukocyte common antigen family. Annu. Rev.
e1089. Immunol. 7, 339–369.
Miller, J.L., de Wet, B.J., deWet, B.J., Martinez-Pomares, L., Radcliffe, C.M., Trowbridge, I.S., and Thomas, M.L. (1994). CD45: an emerging role as a protein
Dwek, R.A., Rudd, P.M., and Gordon, S. (2008). The mannose receptor medi- tyrosine phosphatase required for lymphocyte activation and development.
ates dengue virus infection of macrophages. PLoS Pathog. 4, e17. Annu. Rev. Immunol. 12, 85–116.
Monath, T.P., and Heinz, F.X. (1996). Flaviviruses (Philadelphia: Lippincott- Turell, M.J., O’Guinn, M., and Oliver, J. (2000). Potential for New York mosqui-
Raven Publishers). toes to transmit West Nile virus. Am. J. Trop. Med. Hyg. 62, 413–414.
Mustelin, T., Vang, T., and Bottini, N. (2005). Protein tyrosine phosphatases van der Meulen, K.M., Pensaert, M.B., and Nauwynck, H.J. (2005). West Nile
and the immune response. Nat. Rev. Immunol. 5, 43–57. virus in the vertebrate world. Arch. Virol. 150, 637–657.
Nene, V., Wortman, J.R., Lawson, D., Haas, B., Kodira, C., Tu, Z.J., Loftus, B., Vanlandingham, D.L., McGee, C.E., Klinger, K.A., Vessey, N., Fredregillo, C.,
Xi, Z., Megy, K., Grabherr, M., et al. (2007). Genome sequence of Aedes and Higgs, S. (2007). Relative susceptibilties of South Texas mosquitoes to
aegypti, a major arbovirus vector. Science 316, 1718–1723. infection with West Nile virus. Am. J. Trop. Med. Hyg. 77, 925–928.

Neth, O., Jack, D.L., Johnson, M., Klein, N.J., and Turner, M.W. (2002). Wilson, R., Chen, C.W., and Ratcliffe, N.A. (1999). Innate immunity in insects:
Enhancement of complement activation and opsonophagocytosis by the role of multiple, endogenous serum lectins in the recognition of foreign
complexes of mannose-binding lectin with mannose-binding lectin-associ- invaders in the cockroach, Blaberus discoidalis. J. Immunol. 162, 1590–1596.
ated serine protease after binding to Staphylococcus aureus. J. Immunol. Xi, Z., Ramirez, J.L., and Dimopoulos, G. (2008). The Aedes aegypti toll
169, 4430–4436. pathway controls dengue virus infection. PLoS Pathog. 4, e1000098.
Nir, Y., Goldwasser, R., Lasowski, Y., and Margalit, J. (1968). Isolation of West Zelensky, A.N., and Gready, J.E. (2005). The C-type lectin-like domain super-
Nile virus strains from mosquitoes in Israel. Am. J. Epidemiol. 87, 496–501. family. FEBS J. 272, 6179–6217.

Cell 142, 714–725, September 3, 2010 ª2010 Elsevier Inc. 725

The ATAC Acetyltransferase Complex
Coordinates MAP Kinases
to Regulate JNK Target Genes
Tamaki Suganuma,1 Arcady Mushegian,1,2 Selene K. Swanson,1,3 Susan M. Abmayr,1,3 Laurence Florens,1
Michael P. Washburn,1,4 and Jerry L. Workman1,*
1Stowers Institute for Medical Research, 1000 East 50th Street, Kansas City, MO 64110, USA
2Department of Microbiology, Molecular Genetics, and Immunology
3Department of Anatomy and Cell Biology
4Department of Pathology and Laboratory Medicine

University of Kansas Medical Center, Kansas City, KS 66160, USA

DOI 10.1016/j.cell.2010.07.045

SUMMARY complex (Lee and Workman, 2007). However, it was unknown

whether ATAC functions as a transcription cofactor like SAGA.
In response to extracellular cues, signal transduction Mitogen-activated protein kinases (MAPKs) and their target
activates downstream transcription factors like kinases in the pathway lead to phosphorylation of target
c-Jun to induce expression of target genes. We transcription factors (Edmunds and Mahadevan, 2004; Thomson
demonstrate that the ATAC (Ada two A containing) et al., 1999). Exposure of cells to increases in extracellular osmo-
histone acetyltransferase (HAT) complex serves as lality results in rapid activation of stress-activated MAPKs
(SAPKs), including c-Jun-NH2-terminal kinase (JNK/basket in
a transcriptional cofactor for c-Jun at the Jun
Drosophila) and p38 MAPKs (de Nadal et al., 2002; Edmunds
N-terminal kinase (JNK) target genes Jra and chick- and Mahadevan, 2004; Kyriakis and Avruch, 2001). Osmotic
adee. ATAC subunits are required for c-Jun occu- stress causes activation of JNK by phosphorylation which, in
pancy of these genes and for H4K16 acetylation at turn, phosphorylates c-Jun and enhances its transcriptional
the Jra enhancer, promoter, and transcribed se- activity (Kayali et al., 2000; Thomson et al., 1999). MAPKs are
quences. Under conditions of osmotic stress, ATAC more intimately involved in the regulation of downstream target
colocalizes with c-Jun, recruits the upstream kinases genes than just phosphorylation of target transcription factors
Misshapen, MKK4, and JNK, and suppresses further (Dioum et al., 2009). For example, the yeast SAPK Hog1 is
activation of JNK. Relocalization of these MAPKs recruited to target genes in chromatin and interacts with tran-
and suppression of JNK activation by ATAC are scription factors, cofactors, and RNA polymerase II (Alepuz
dependent on the CG10238 subunit of ATAC. Thus, et al., 2001, 2003; de Nadal et al., 2004; Mas et al., 2009; Pokho-
lok et al., 2006; Proft et al., 2006; Zapater et al., 2007). Recently,
ATAC governs the transcriptional response to MAP
the ERK2 MAPK was found to bind with sequence specificity
kinase signaling by serving as both a coactivator to DNA and act as a transcriptional repressor of interferon-g-
of transcription and as a suppressor of upstream induced genes (Hu et al., 2009). Moreover, multiple MAPKs,
signaling. including ERK1/2, p38, and JNK, bind to and regulate transcrip-
tion of the insulin gene (Lawrence et al., 2009). Thus, MAPKs play
INTRODUCTION important chromatin-associated functions in the regulation of
gene expression.
Histone acetyltransferase complexes have been isolated from MBIP (MAPK upstream kinase [MUK] binding inhibitory
multiple organisms and shown to be involved in nuclear events protein), a component of the human ATAC complex, was
that relate to chromatin biology (Kimura et al., 2005; Lee and identified as a MUK binding partner in a yeast two-hybrid screen
Workman, 2007). The Drosophila ATAC (Ada two A containing) (Fukuyama et al., 2000). The Drosophila ATAC component
complex consists of 13 proteins and includes two distinct CG10238 encodes the Drosophila homolog of MoaE, a subunit
histone acetyltransferases, Gcn5/KAT2 and Atac2/KAT14 (Fig- of molybdopterin (MPT) synthase, an essential enzyme involved
ure 1A). Whereas Gcn5/KAT2 preferentially acetylates histones in the synthesis of the molybdenum cofactor (Moco). Moco binds
H3K9 and H3K14, Atac2/KAT14 acetylates H4K16 (Ciurciu molybdenum in the active site of molybdenum enzymes, which
et al., 2006; Guelman et al., 2006; Suganuma et al., 2008). The catalyze redox reactions as part of ancient and conserved
Gcn5/KAT2, Ada3, and CG30390 (Sgf29 in yeast) subunits of biosynthetic pathways (Iyer et al., 2006; Leimkuhler et al.,
ATAC are shared with SAGA (Spt-Ada-Gcn5 acetyltransferase) 2003; Rudolph et al., 2001; Schwarz and Mendel, 2006).
(Suganuma et al., 2008), an important transcriptional coactivator However, CG10238 also contains C-terminal sequences that

726 Cell 142, 726–736, September 3, 2010 ª2010 Elsevier Inc.

A Figure 1. Peptides from Proteins Related to
the JNK Pathway Were Detected in Purifica-
tions of ATAC Subunits, and CG10238
CG10238 Atac2 CHRAC14 Ada2a WDS D12 Control
Proteins PEP (SC%) PEP (SC%) PEP (SC%) PEP (SC%) PEP (SC%) PEP (SC%) PEP (SC%) Inhibits JNK Activation by Osmotic Stress
ATAC CG10238 29 (72.21) 5 (21.25) 2 (9.81) 7 (25.61) 6 (30.25) 16 (53.13) 0 (0)
Gcn5/KAT 35 (46.69) 4 (6.15) 4 (6.4) 5 (8.24) 4 (7.38) 33 (33.95) 0 (0) in Drosophila Schneider’s S2 Cells
Atac1 14 (53.37) 2 (5.34) 2 (7.58) 6 (16.3) 0 (0) 9 (20.71) 0 (0) (A) Peptides of MAPK pathway proteins in addition
Ada3 21 (58.89) 1 (1.98) 1 (1.98) 2 (7.19) 1 (3.24) 10 (16.37) 0 (0)
Ada2a 20 (48.2) 2 (4.93) 2 (4.93) 9 (18.6) 0 (0) 18 (36.43) 0 (0) to ATAC subunits detected by MudPIT analysis for
HCF 66 (68.87) 11 (11.93) 16 (17.6) 17 (19.87) 19 (18.07) 54 (49.4) 1 (2.6)
D12 37 (45.86) 7 (9.08) 11 (19.81) 10 (13.62) 5 (9.7) 31 (38.29) 0 (0) affinity purifications of ATAC subunits with FLAG-
CG30390 14 (56.60) 1 (5.88) 0 (0) 6 (26.64) 0 (0) 13 (55.36) 0 (0)
Atac2/KAT14 26 (47.16) 10 (22.09) 7 (16.28) 5 (9.95) 5 (13.18) 19 (39.41) 0 (0)
HA-tagged CG10238, Atac2/KAT14, Ada2a,
CHARC14 5 (62.5) 1 (10.16) 1 (10.16) 3 (27.34) 1 (14.84) 7 (71.88) 0 (0) CHRAC14, WDS, and D12. Parental S2 cells
Atac3 14 (40.88) 1 (3.89) 0 (0) 3 (7.61) 0 (0) 9 (20.71) 0 (0)
NC2 beta 6 (60.66) 2 (14.21) 1 (7.1) 0 (0) 2 (14.21) 5 (27.32) 0 (0) were used in mock purification as a control. PEP,
WDS 18 (66.48) 7 (25.76) 5 (18.28) 8 (26.59) 20 (63.43) 13 (53.19) 1 (3.6)
Proteins Jra 2 (10.03) 0 (0) 3 (19.03) 0 (0) 0 (0) 2 (13.84) 0 (0)
peptide count; SC%, sequence coverage (%).
related to Misshapen 16 (21.34) 0 (0) 5 (6.12) 0 (0) 0 (0) 5 (5.85) 0 (0) (B) Homology domains of CG10238 and experi-
JNK pathway MKK4 4 (15.8) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0)
slik 5 (4.82) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) mental design of truncation mutants. The diagram
MEK3/MKK3 2 (8.08) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0)
Chickadee 2 (20.63) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0)
shows the domains of CG10238 that are homolo-
gous to human MOCS2B (hMoaE) and to human
MBIP, which are separate proteins. The full-length
B (FL), N-terminal that contains only MoaE domain
Drosophila melanogaster
(N), and C-terminal that contains only MBIP protein
1 137 171 367 a.a. sequences (C) were stably expressed in S2 cells
(Figure 2; see Figure S1B).
Homo sapiens
1 188 a.a.
(C) CG10238 inhibited the activation of JNK stimu-
1 344 a.a. lated by osmotic stress in S2 cells. Stably ex-
MBIP pressed FLAG-HA-tagged CG10238 was induced
Design of deletion mutants of CG10238 by 0, 50, or 150 mM CuSO4 in S2 cells. Cells were
1 367a.a.
stimulated with osmotic stress by addition of
1 137a.a.
500 mM sorbitol for 12 min before harvesting cells
137a.a. 367a.a. (C, lanes 6–10; Figure 2C, lanes 5–14; Figure 2D,
lanes 1–12; see Figures S2A–S2C). The nuclear
extracts were examined by western blot probing
with anti-HA, anti-Active JNK, and anti-JNK
C (C; Figure 2C). Parental S2 cells of stable cell lines
CuSO4 (x10µM) - 15 - 5 15 - 15 0 5 15
cultured with 0 or 150 mM CuSO4 were used as
HA-CG10238 - - - - controls (C, lanes 1, 2, 6, and 7; see Figure S3).
Parental cells + + - - - + + - - - See also Figures S1, S2, and S3.
Sorbitol - - - - - + + + + +
1 2 3 4 5 6 7 8 9 10

anti-HA 53.9



are homologous to human MBIP. This observation, coupled with genetically with the JNK pathway (Jasper et al., 2001; Morrison
the presence of MAPK signaling proteins in ATAC purifications, et al., 2000).
led us to ask whether CG10238 and ATAC play a role in MAPK
signaling. The ATAC Subunit CG10238 Functions as an Inhibitor
of JNK Activation in Response to Osmotic Stress
RESULTS We tested whether CG10238 plays a role in MAPK signaling like
MBIP (Fukuyama et al., 2000). MAPK cascades can be activated
ATAC Interacts with Proteins Related to the MAPK by osmotic stress, resulting in activation of JNK by phosphoryla-
Pathway tion (Kayali et al., 2000; Yang et al., 2003). We therefore exam-
Affinity purifications of the ATAC complex revealed proteins that ined whether expression of CG10238 affected the activation of
are part of the MAPK signaling pathway (Figure 1A; see Fig- JNK under conditions of osmotic stress (Kayali et al., 2000).
ure S1A available online). Peptides from these proteins were We first titrated the cellular response to osmotic stress
found in purifications via the CG10238, CHRAC14, and D12 stimulated by sorbitol in S2 cells by western blot (Figures S2A
ATAC subunits. Peptides were identified from the transcription and S2B). Maximum activation of JNK was observed between
factor Jra (Jun-related antigen), the Drosophila homolog of 7 and 30 min after treatment with a minimum concentration of
c-Jun, and Misshapen (MSN), the Drosophila homolog of 500 mM sorbitol (Figure S2A, lane 5; Figure S2B, lanes 2–4).
Ste-20 kinase (Figure 1A) (Morrison et al., 2000; Su et al., 1998; We then examined the level of JNK activation in cells expressing
Treisman et al., 1997). We also found peptides from other STE CG10238 in the presence or absence of osmotic stress induced
kinases, such as MKK4, slik, and MEK3/MKK3, as well as by 500 mM sorbitol for 12 min. Expression of CG10238 was
peptides from Chickadee, which has been shown to interact inducible by CuSO4, and parental S2 cells were treated with

Cell 142, 726–736, September 3, 2010 ª2010 Elsevier Inc. 727

A B Figure 2. MBIP-Related Sequences Incor-
Input nuclear extracts Elution
porate CG10238 into ATAC; However, the
- F N C F N C - (KDa)
CG10238-Domain N C MoaE Domain Inhibits JNK Activation
ATAC subunits PEP( SC%) PEP( SC%) 54.8
(A) FLAG-HA-tagged N terminus that contains only
CG10238-PA 7 (25.6) 17 (50.95)
Gcn5/KAT 0 (0) 8 (12.67)
the MoaE domain (N) and C terminus that contains
anti-HA only MBIP protein sequences (C) were stably ex-
Atac1 0 (0) 3 (9.55)
Ada3 0 (0) 3 (8.99) 29.4 pressed in S2 cells purified by FLAG affinity beads.
Ada2a 0 (0) 5 (13.09) The copurified proteins were examined by MudPIT
HCF 5 (1.47) 24 (26.8) 17.5
D12 0 (0) 15 (21.26) 97.3 analysis (see Figures S1A and S1B). The observed
CG30390 0 (0) 3 (13.15) PEP and SC (%) of each sample are indicated as in
Atac2/KAT14 0 (0) 5 (10.98) 118.2
anti-Atac2 Figure 1.
CHRAC14 0 (0) 0 (0)
97.3 (B) The copurified proteins of CG10238-full-length
Atac3 0 (0) 2 (6.55)
NC2beta 0 (0) 3 (14.21) 206.3 (F), -N, or -C (Figure 1B; Figure S1) were analyzed
WDS 1 (3.88) 6 (27.42) 118.2 by western blotting with antibodies against HA tag
and ATAC subunits. As a control, untransfected
anti-Ada2a 54.8
parental S2 cells were mock purified and analyzed
by western blotting (-).
29.4 (C and D) The MoaE domain inhibits JNK activa-
anti-NC2 beta tion. The ability to inhibit JNK activation was
54.8 compared between FLAG-HA-tagged CG10238
(FL), CG10238-N (MoaE domain), and CG10238-
anti-CG30390 C (MBIP domain) after the expression levels were
normalized (see also Figure S1B). Parental S2 cells
cultured with 150 mM CuSO4 were used as a posi-
tive control (C, lanes 1 and 5; D, lanes 1, 5, and 9).
C D Osmotic stress was induced as in Figure 1C
HA-C - - - + - - - - - - - (see also Figures S2A and S2B). For sorbitol-stim-
HA-N - - + - - - - - - - -
Active JNK ulated samples, the intensity of each band for
HA-FL - + - - - - - - - - -
Parental cells + - - - + - - - - - - - - - Active JNK in western blotting was quantified,
Sorbitol - - - - + + + + + + + + + +
1 2 3 4 5 6 7 8 9 10 11 12 13 14 140
and each sample from CG10238 (FL), CG10238-N,
and CG10238-C was individually compared with
the intensity of parental cells (P), and the ratios
Ratio (%)

(%) are shown in (D). The average of four indepen-
dent experiments is graphed. Error bars represent
40 standard deviation.
20 See also Figures S1 and S2.
anti-ActiveJNK 1 2 3 4 5 6 7 8 9 10 11 12


the same amount of CuSO4 and sorbitol as controls. We first examined the two parts of CG10238 to determine which might
confirmed by western blot that the expression level of HA- incorporate it into the ATAC complex. We generated S2 cell lines
tagged CG10238 in the stable cell line without induction was that stably expressed tagged truncated forms of CG10238 that
similar to that of endogenous CG10238 in the parental cells included only the N-terminal MoaE domain or the C-terminal
(Figure S2C). Active JNK was not detected in the absence of MBIP domain (Figure 1B). These tagged proteins were affinity
sorbitol treatment, or upon induction of CG10238 by addition purified from the stable cell lines after their expression levels
of CuSO4 (Figure 1C, lanes 1 and 2). In the presence of sorbitol, were normalized (Figure S1B). Proteins that copurified with
JNK was activated in the parental cells (Figure 1C, lanes 6 and 7). each domain of CG10238 were identified by multidimensional
However, induction of CG10238 expression by CuSO4 inhibited protein identification technology (MudPIT) analysis and
the activation of JNK in a dose-dependent manner (Figure 1C, confirmed by western blots (Figures 2A and 2B; Figure S1A).
lanes 8–10). Thus, CG10238 inhibits JNK activation by osmotic All ATAC subunits except CHRAC14 associated with the MBIP
stress in vivo. Because JNK is also activated by ultraviolet light domain. A few peptides from two of the ATAC subunits were de-
(UV), we examined whether the inhibitory activity of CG10238 tected by purification with the MoaE domain and three subunits
extended to this JNK activation mechanism (Angel et al., 1988; were detected weakly by westerns (Figure 2B). Thus, the MBIP
Devary et al., 1991; Rozek and Pfeifer, 1995). Expression of domain of CG10238 is responsible for incorporating the protein
CG10238 also inhibited activation of JNK by UV (Figure S3). into the ATAC complex.

The MBIP Domain of CG10238 Is Required The MoaE Domain of CG10238 Inhibits JNK Activation
for Incorporation into the ATAC Complex Analysis of the domain structure of MoaE and MBIP orthologs in
The finding that CG10238 prevented JNK activation when a variety of organisms indicates that many insects encode these
expressed in vivo led us to ask whether CG10238 served this two domains as a translational fusion like that of Drosophila
function in isolation or as part of the ATAC complex. We first CG10238. By contrast, orthologs in nematodes, plants, fungi,

728 Cell 142, 726–736, September 3, 2010 ª2010 Elsevier Inc.

B S2 Cells Figure 3. ATAC Inhibits JNK Signaling
A S2 Cells ta ta
nt nt c2 c2 nt ont be 2be S2 cells were transfected with dsRNA-Atac2, dsRNA-NC2
Co A-Co -Ata A-Ata Co C
A- A N b, or dsRNA-Control (dsRNA-Cont) (A and B) (see
Transfected with s RN sRN sRN sRN Transfected with ds
d d d d
- + - + (KDa) Sorbitol - + - + Figure S4). 293T cells were transfected with dsRNA-
Sorbitol (KDa)
anti-NC2beta 29.4 hAtac2/CSRP2BP, dsRNA-MBIP, or dsRNA-Control
(dsRNA-Cont) (C and D). The knockdown cells were then
54.8 54.8
anti-ActiveJNK anti-ActiveJNK stimulated with or without 500 mM (for S2 cells) or
54.8 54.8
400 mM (for 293T cells) sorbitol as in Figure 1C (see
anti-JNK anti-JNK
Figure S2D). The nuclear extracts were examined by
anti-tubulin anti-tubulin western blot probing with anti-Active JNK, anti-JNK,
54.8 54.8
anti-Atac2 (A and C), anti-NC2 b (B), anti-MBIP (D), or
anti-tubulin antibodies as a loading control (A–D, top).
dsRNA-Atac2 The intensities of each band of JNK, Active JNK, Atac2,
800 120 3000 120
NC2 b, hAtac2, or MBIP in western blotting were quantified
Ratio-JNK or Ratio-ActJNK

Ratio-JNK or Ratio-ActJNK
100 2500 100
from four independent experiments and plotted as ratios to

Ratio-NC2 beta
80 2000 80 the untreated control cells (A–D, bottom). Dark bars show

400 60 ActJNK 1500 60 ActJNK
the ratio of JNK intensities, light bars show the ratio of
Atac2 NC2beta
Active JNK intensities, and the dotted line shows the ratio
40 1000 40
200 of Atac2 (A), NC2 b (B), Atac2/CSRP2BP (C), or MBIP (D)
20 500 20
intensities. Error bars represent standard deviation.
0 0 0 0 See also Figures S2 and S4.
1 2 3 4 1 2 3 4

Sorbitol - + - + Sorbitol - + - +
dsRNA-Cont. dsRNA-Atac2 dsRNA-Cont. dsRNA-NC2beta

the inhibitory activities of the full-length protein

C 293T Cells P P D 293T Cells to that of the MoaE and MBIP domains (Fig-
2B 2B
nt nt RP RP nt nt IP BIP ures 2C and 2D). Full-length CG10238 and the
Co A-Co -CS -CS Co Co MB
A - A A A- NA- NA- NA-M
Transfected with ds ds d d Transfected with ds ds ds ds MoaE domain alone inhibited activation of JNK
Sorbitol - + - + (KDa) Sorbitol - + - + (KDa)
(Figure 2C, lanes 6–11; Figure 2D, bars 1–8).
anti-Atac2/CSRP2BP anti-MBIP
However, the MBIP domain of CG10238 failed
anti-Active JNK
anti-Active JNK
46 to prevent the activation of JNK (Figure 2C,
lanes 12–14; Figure 2D, bars 9–12). Thus, the
anti-JNK anti-JNK active portion of CG10238 in inhibition of JNK
anti-tubulin anti-tubulin
activation is the MoaE domain, whereas the
C-terminal sequences connect this activity to
the ATAC complex.

600 140 350 140

ATAC Inhibits JNK Activation

Ratio-JNK or Ratio-ActJNK

Ratio-JNK or Ratio-ActJNK

120 300 120

The MoaE domain of CG10238 was sufficient to

100 250 100

prevent JNK activation when expressed in vivo,

80 JNK 200 80 JNK

whereas the MBIP domain was required to
CSRP2BP 150 60

40 incorporate CG10238 into ATAC (Figures 2A– 100 40

20 2D). These data raised the question of whether 50 20

0 0 ATAC itself plays a role in inhibition of the JNK 0 0

1 2 3 4 1 2 3 4

Sorbitol - + - +
pathway. To address this question, we exam-
Sorbitol - + - +
dsRNA-Cont dsRNA-CSRP2BP dsRNA-Cont dsRNA-MBIP ined JNK activation in S2 cells where endoge-
nous subunits of ATAC were knocked down by
and all prokaryotes except parasitic bacteria have a MoaE dsRNA interference. The expression level of Atac2 was reduced
homology domain but are missing the MBIP domain. Sequence 60% in cells expressing dsRNA-Atac2 (Figure 3A). Interestingly,
database searches using the PSI-BLAST program (Altschul JNK was partially activated upon reduction of Atac2 even in the
et al., 1997) with human MBIP (GenBank accession number absence of osmotic stress (Figure 3A). Moreover, activation of
119586267) as query revealed sequence matches between the JNK by osmotic stress was enhanced in the Atac2 knockdown
N-terminal portion of MBIP orthologs from Metazoa and MoaE cells (Figure 3A). We observed similar results upon knockdown
proteins. Thus, the N-terminal sequences of mammalian MBIPs of NC2 b or CG10238 subunits of ATAC by dsRNA (Figure 3B;
are related to the MoaE sequences. Hence, our bioinformatic Figure S4A). Although JNK activation was not observed in D12
analysis suggests the N-terminal sequences of MBIP evolved knockdown cells without osmotic stress, its activation was
from MoaE and raises the possibility that the MoaE domain of also increased in these cells under conditions of osmotic stress
CG10238 contributes to the JNK inhibition activity. (Figure S4B). Because human MBIP was recently shown to be a
We generated S2 cell lines that stably expressed tagged component of the human ATAC complex (Fukuyama et al., 2000;
truncated forms containing the MoaE domain (N-terminal) or Wang et al., 2008), it was of interest to determine whether human
MBIP domain (C-terminal) (Figure S1B), and then compared ATAC inhibited the activation of JNK by osmotic stress in human

Cell 142, 726–736, September 3, 2010 ª2010 Elsevier Inc. 729

A B Figure 4. ATAC Is Required for the Transcription
Input IP Input IP Regulation of JNK Target Genes
IP with - IgG Jra IP with - IgG M
S (A and B) Extracts from S2 cells endogenously expressing
1 2 3 1 2 3 ATAC subunits and JNK (B) were immunoprecipitated with
anti-Atac2 anti-Atac2 antibodies to endogenous Jra (A) or Misshapen (MSN) (B).
The presence of ATAC subunits, MSN, and JNK in the
anti-CG10238 anti-CG10238 input and immunoprecipitates (IP) was detected by
anti-NC2beta western blots.
anti-NC2 beta (C) Quantitative real-time RT-PCR (qRT-PCR) of Jra
and Chickadee mRNA in S2 cells transfected with
anti-Jra anti-MSN dsRNA-LacZ (Control), dsRNA-Atac2, dsRNA-CG10238,
dsRNA-NC2 b (NC2b), dsRNA-Jra, or dsRNA-MSN (see
Figure S5A). Gene-specific mRNA levels from the cells
without sorbitol stimulation (C, top) and 30 min after
500 mM sorbitol stimulation (C, bottom) were measured
Jra Sorbitol (-) Chickadee Sorbitol (-)
and normalized to RpL-32 expression (Figure S5B), and
are represented as ratios by qRT-PCR measurements.
2 2
1.8 1.8 The average of three independent experiments is graphed.

1.6 1.6
1.4 1.4
Error bars represent standard deviation.

1.2 1.2 (D) ChIP assays were performed with an antibody against
1 1
0.8 0.8 Jra from S2 cells without sorbitol stimulation (D, left) or with
0.6 0.6
0.4 0.4
sorbitol stimulation (D, right) expressing dsRNA-LacZ
0.2 0.2 (Cont), dsRNA-Atac2, dsRNA-CG10238, or dsRNA-MSN
0 0
1 2 3 4 5 6 1 2 3 4 5 6 (D) (Figures 5B–5M). The association of Jra on the probed
RNAi b a RNAi
cZ c2 38 2 Jr MSN cZ c2 38 2b Jra MSN region of the Jra gene as indicated in Figure 5A, and input
La Ata 102 NC La Ata 102 NC
chromatin was measured by qRT-PCR and normalized to
Jra Sorbitol (+) Chickadee Sorbitol (+)
input, and the ratios of quantities are represented (see
4 1.6 Figure S6). The average of three independent experiments
is graphed. Error bars represent standard deviation.

3 1.2
See also Figures S2, S5, and S6.

2 0.8
1 0.4
0 0

1 2 3 4 5 6
in vivo. Endogenous Atac2, CG10238, NC2 b,
1 2 3 4 5 6
2 8 2b Jra SN a
cZ tac 23 cZ c2 38 2b Jr MSN
La A G10 N C M La Ata 102 NC
and D12 coimmunoprecipitated with endoge-
nous MSN (Figure 4B).
If ATAC serves as a cofactor for Jra, then
ChIP-Jra on Jra Sorbitol (-) ChIP-Jra on Jra Sorbitol (+)
Jra-dependent transcription would require
dsRNA-Cont dsRNA-Atac2 dsRNA-CG10238 dsRNA-MSN dsRNA-Cont dsRNA-Atac2 dsRNA-CG10238 dsRNA-MSN ATAC and ATAC should be localized to Jra target
genes. Because the gene encoding c-Jun, the

mammalian homolog of Jra, is positively


regulated by c-Jun protein (Angel et al., 1988),
30.00% we examined whether ATAC functions in Jra

20.00% 20.00%

10.00% transcription. We measured the levels of Jra


Probe A B C D E Probe A B C D E
transcripts by quantitative real-time RT-PCR

(qRT-PCR) of control and ATAC knockdown S2

cells. As a control, we confirmed that Jra tran-
293T cells (Figure S2D). Indeed, JNK was activated to higher script levels were reduced in cells expressing dsRNA-Jra
levels upon knockdown of MBIP or CSRP2BP (human homolog compared with cells expressing control-dsRNA-LacZ (Figure 4C,
of Atac2) in 293T cells (Figures 3C and 3D). These data indicate upper left). Importantly, Jra transcript levels were significantly
that the ATAC complex is required for complete inhibition of JNK reduced in Atac2 and CG10238 knockdown cells and slightly
activation in both Drosophila and human cells. reduced in NC2 b knockdown cells (Figure 4C, upper left). We
did not detect effects on Jra transcripts in MSN knockdown cells.
ATAC Functions as Transcriptional Cofactor We confirmed that the transcripts of atac2, CG10238, nc2 b, and
for JNK Target Genes msn were reduced by expression of each dsRNA-Atac2, dsRNA-
The presence of peptides from the Jra transcription factor in CG10238, dsRNA-NC2 b, and dsRNA-MSN by qRT-PCR (Fig-
MudPIT analyses of the affinity-purified ATAC complex (Fig- ure S5A). We also observed that the expression of Atac2 was
ure 1A) suggests that ATAC interacts with Jra and may serve increased in Jra knockdown (see below).
as a transcriptional cofactor. Indeed, endogenous Jra was able We next examined chickadee, another Jra target gene
to coimmunoprecipitate with Atac2, CG10238, and NC2 b, sub- (Jasper et al., 2001). Transcript levels of chickadee decreased
units of ATAC (Figure 4A). We also sought to confirm the interac- significantly upon knockdown of Atac2 and somewhat more
tion of MSN (Figure 1A), the Drosophila Ste-20 kinase, with ATAC modestly upon knockdown of NC2 b or Jra (Figure 4C).

730 Cell 142, 726–736, September 3, 2010 ª2010 Elsevier Inc.

Transcript levels of chickadee were not reduced significantly in knockdown of Atac2 or CG10238. We next investigated
CG10238 knockdown cells. Of note, we did not observe changes whether ATAC occupancy affected acetylation of the nucleo-
in the transcript levels of target genes of a different H4K16 some on the Jra gene by a ChIP assay for H4K16 acetylation.
acetyltransferase complex (MSL) upon knockdown of ATAC, In control cells without osmotic stress, both the enhancer and
indicating that requirement for ATAC is specific for Jra and promoter regions were enriched in H4K16 acetylation, which
chickadee (Figure S5B). was significantly reduced in the Atac2 and CG10238 knock-
The JNK cascade is activated by osmotic stress and increases down cells (Figure 5D). Thus, in the absence of osmotic
c-Jun expression in mammals. Thus, we examined Jra and stress, ATAC occupied the enhancer and promoter of Jra,
chickadee transcript levels in ATAC knockdown cells upon and H4K16 acetylation at these sites was dependent on
treatment with sorbitol. Interestingly, Jra transcripts were not ATAC.
reduced in Atac2 knockdown cells upon osmotic stress but In the presence of osmotic stress, ATAC occupancy and
were increased in CG10238 knockdown cells. The expression H4K16 acetylation of the promoter were similar but were
of Chickadee was also increased in ATAC and MSN knockdown reduced at the upstream enhancer (Figures 5C and 5E). Interest-
cells under osmotic stress. Thus, ATAC positively impacts the ingly, ATAC occupancy and H4K16 acetylation of the coding
expression of JNK target genes in the absence of osmotic stress region at primer D were stimulated by osmotic stress (Figures
but plays a predominantly negative role at JNK target genes 5C and 5E). Jra was clearly localized on this region with or
under conditions of osmotic stress. Moreover, the positive without osmotic stress (Figure 4D). Jra occupancy at primer D
effects of ATAC on expression of JNK target genes are was strongly reduced in Atac2, CG10238, and MSN knockdown
mediated primarily by Atac2, whereas its negative effects are cells in the absence of sorbitol. However, under osmotic stress,
most strongly dependent on CG10238. Jra occupancy required primarily CG10238 (Figure 4D). Thus,
under conditions of osmotic stress, ATAC occupancy at the
ATAC Functions as a Cofactor for the JNK Target Gene upstream enhancer was reduced and instead increased in the
Jra, and Coordinates Upstream MAPKs onto Jra coding region where Jra was bound.
Previous studies have shown the direct association of MAPKs We next tested whether stress-activated upstream kinases
with gene loci that they regulate (de Nadal and Posas, 2010; were recruited to the Jra gene. We probed Active JNK (phos-
Dioum et al., 2009). Thus, we measured the occupancy of pho-JNK) and phospho-MKK4/SEK1 (Ser257/Thr261) occu-
ATAC and upstream kinases directly on JNK target genes. We pancy on the Jra gene. We also tested for the occupancy of
examined the occupancy of ATAC on the Jra gene in ATAC the MSN kinase. In the absence of osmotic stress in control cells,
and MSN knockdown cells by chromatin immunoprecipitation each of these kinases could be found on the Jra gene although
(ChIP) assay. Previous studies have shown that Jun can form a their locations varied. Active JNK was found at the enhancer
homodimer or heterodimerize with ATF or Fos to form the AP-1 (primer A), the promoter (primer B), and downstream of the
transcription factor. The c-Jun promoter contains AP-1-like coding region (primer E) (Figure 5F). Phospho-MKK4 was local-
sequences (TTACCTCA and TGACATCA) in addition to an ized to the enhancer, promoter, and downstream of the coding
NF-jun binding sequence (GGAGTCTCC) (Rozek and Pfeifer, region (primer E) and MSN was localized to the promoter (primer
1993). c-Jun was found to occupy the c-Jun promoter without B) (Figures 5H and 5J). Strikingly, like ATAC upon osmotic stress,
external stimulus (Angel and Karin, 1991). We confirmed that occupancy of each of these kinases was increased in the coding
Jra actually bound to the Jra gene by ChIP assay (Figure 4D, region (primer D), the site of Jra binding (Figures 5G, 5I, and 5K).
left). Jra was found to significantly occupy a site in the middle This was most significant for Active JNK, which bound strongly
of the coding region that contains an NF-jun recognition to the primer D region upon osmotic stress (note the difference
sequence (Figure 4D, left). We analyzed regions of the Jra in scale between Figures 5F and 5G), and for MSN, which relo-
gene that are comparable to the regulatory regions of the human calized from the promoter to the coding region (primer D)
c-Jun gene (Figure 5A). We probed ATAC occupancy, with and (Figures 5J and 5K).
without osmotic stress, at an upstream enhancer (primer A), The ChIP results from the knockdown cell lines indicate a
the promoter containing the AP-1 binding sequence (primer B), dynamic interplay between ATAC and the upstream kinases.
the 50 region of the coding sequence (primer C), the middle For example, in the absence of osmotic stress, H4K16 acetyla-
of the coding sequence that contained an NF-jun binding tion at the promoter (primer B) was independent of MSN (Fig-
sequence (GGAGGCACC) (primer D), and a region 30 of the ure 5D); however, upon osmotic stress, MSN suppressed
coding sequence (primer E) (FlyBase and UCSC Genome H4K16 acetylation at the promoter but was required for
Browser) (Figure 5A) (Rozek and Pfeifer, 1993). maximum acetylation in the coding region (primer D) (Figure 5E).
In the absence of osmotic stress, Atac2 was found to In the absence of osmotic stress, the binding of Active JNK to the
occupy the enhancer and promoter regions of Jra (Figure 5B, enhancer, promoter, and downstream of the coding region
dsRNA-Cont). As expected, Atac2 occupancy of these regions required ATAC but was suppressed by MSN (Figure 5F). By
was reduced in the Atac2 knockdown cells (Figure 5B). Atac2 contrast, under osmotic stress, the strong binding of Active
occupancy of the promoter was reduced in CG10238 but not JNK to the coding region (primer D) required both ATAC and
MSN knockdown cells. As controls for the specificity of ATAC MSN (Figure 5G). In the absence of osmotic stress, binding of
occupancy of Jra, we performed an ATAC-ChIP assay on two phospho-MKK4 to the enhancer was suppressed by ATAC and
target genes of the MSL complex (Figures S5 and S6). Back- MSN (Figure 5H) but upon osmotic stress, binding to the
ground signals on these genes were not affected by the promoter required MSN and CG10238 specifically (Figure 5I).

Cell 142, 726–736, September 3, 2010 ª2010 Elsevier Inc. 731

A Figure 5. Occupancy of ATAC, H4K16
Jra gene span Jra CDS Acetylation, MAPKs, and H3S10P on the
Enhancer Promoter Coding Coding E
Jra Gene with and without Osmotic Stress
Probes 100bp
A B C D (A) Schematic representation of the Jra gene
B ChIP-Atac2 on Jra Sorbitol (-) C ChIP-Atac2 on Jra Sorbitol (+) AP-1
indicating the position of different probes tested
dsRNA-Contt dsRNA-Atac2
A ac2 dsRNA-CG10238 dsRNA-MSN dsRNA-Cont dsRNA-Atac2
A ac2 dsRNA-CG10238 dsRNA-MSN
(A–E on x axis for B–M). The gray box indicates
1.00% 1.00%

0.80% 0.80%
the coding sequence (CDS), the gray circles indi-

cate the enhancer, the white circle indicates pre-

0.60% 0.60%

0.40% 0.40%
dicted AP-1 sites, and the black square indicates

the NF-jun-like sequence.
(B–M) ChIP assays were performed with anti-
D ChIP-H4K16ac on Jra Soro
r bitol (-) E ChIP-H4K16ac on Jra Sorbitol (+) bodies against Atac2, H4K16ac, Active JNK,
dsRNA-Cont dsRNA-Atac2
A ac2 dsRNA-CG10238 dsRNA-MSN


A ac2 dsRNA-CG10238 dsRNA-MSN
phospho-MKK4 (S257T261) (pMKK4), MSN, and
0.50% 5.00% phospho-H3S10 (H3S10P) from ATAC and MSN


ChIP/Input 4.00%
knockdown S2 cells (as in Figure 4D) without

2.00% sorbitol stimulation (B, D, F, H, J, and L) (Figures

6B, 6D, 6F, 6H, and 6J) or with sorbitol stimulation
A B C D E (C, E, G, I, K, M) (Figures 6C, 6E, 6G, 6I, and 6K)
F ChIP-ActiveJNK
v JNK on Jra Sorbitol (-) G ChIP-ActiveJNK
v JNK on Jra Sorbitol (+)
on the probed region of the Jra gene as indicated
dsRNA-Cont dsRNA-Atac2
A ac2 dsRNA-CG10238 dsRNA-MSN dsRNA-Cont dsRNA-Atac2
A ac2 dsRNA-CG10238 dsRNA-MSN
in (A), and input chromatin was measured by
1.20% 70.00%

1.00% 60.00%
qRT-PCR and normalized to input. The ratios of
measured quantities are represented (see Figures



S5A and S6). The average of three independent
0.20% 10.00% experiments is graphed. Error bars represent
0.00% 0.00%
A B C D E A B C D E standard deviation.
See also Figures S2, S5, and S6.
H dsRNA-Contt
ChIP-pMKK4 on Jra Sorbitol (-)
A ac2 dsRNA-CG10238 dsRNA-MSN
I dsRNA-Cont
ChIP-pMKK4 on Jra Sorbitol (+)
A ac2 dsRNA-CG10238 dsRNA-MSN
0.30% 0.45%

0.20% 0.30%


0.10% 0.15%
0.05% 0.05%
the promoter and strongly suppressed

downstream of the coding region
J dsRNA-Contt
ChIP-Misshapen on Jra Sorbitol (-)
A ac2 dsRNA-CG10238 dsRNA-MSN
K dsRNA-Cont
ChIP-Missahpen on Jar Sorbitol (+)
A ac2 dsRNA-CG10238 dsRNA-MSN
(Figure 5L). Osmotic stress substantially
0.35% 0.30% increased H3S10 phosphorylation levels
across the Jra gene except for the





upstream enhancer (note the different

scales in Figures 5L and 5M). Induction
of H3S10 phosphorylation at the
L ChIP-H3S10P on Jra Sorbitol (-) M ChIP-H3S10P on Jra Sorbitol (+) promoter, the coding region (primer D),
dsRNA-Contt dsRNA-Atac2
A ac2 dsRNA-CG10238 dsRNA-MSN dsRNA-Cont dsRNA-Atac2
A ac2 dsRNA-CG10238 dsRNA-MSN

2.50% 20.00%
and downstream of the coding region de-
pended on ATAC (especially CG10238);
2.00% 16.00%


1.50% 12.00%

8.00% however, MSN suppressed H3S10 phos-

2.00% phorylation across the gene (Figure 5M).
0 00%

These results suggest that the relation-
ship between H3S10 phosphorylation
and H4K16 acetylation by ATAC at Jra
Finally, binding of MSN to the promoter is suppressed by ATAC differs from what was observed with H3S10 phosphorylation
in the absence of osmotic stress (Figure 5J); however, when it and H4K16 acetylation by MOF at the FOSL1 gene (Zippo
moves to the coding region where Jra is bound (primer D), et al., 2009). Acetylation of H4K16 at the enhancer of FOSL1
upon osmotic stress its binding is independent of ATAC by MOF required H3S10 phosphorylation; however, ATAC is
(Figure 5K). required for H3S10 phosphorylation in the presence of osmotic
A previous study linked the acetylation of H4K16 with stress. Moreover, ATAC-dependent H4K16 acetylation at the
H3S10 phosphorylation in the activation of the FOSL1 gene Jra enhancer in the absence of osmotic stress (Figure 5D)
(Zippo et al., 2009). At this gene, H3S10 phosphorylation was occurred independently of significant H3S10 phosphorylation
reported to recruit MOF-dependent H4K16 acetylation, which (Figure 5L).
then recruited the elongation factor P-TEFb to release paused
polymerase (Zippo et al., 2009). As ATAC is also an H4K16 ace- ATAC Acetylates Nucleosomes and Coordinates
tyltransferase, we tested for a relationship between H3S10 the JNK Upstream Kinases with Jra on chickadee
phosphorylation and H4K16 acetylation, ATAC, or MSN at Jra. The most striking aspect of our ChIP analysis of the Jra gene is the
H3S10P was enriched in the enhancer and promoter regions movement upon osmotic stress of ATAC and MAPKs to the
(primer B) in the absence of osmotic stress (Figure 5L). H3S10 coding region (primer D), which contains an NF-jun binding
phosphorylation levels were suppressed by ATAC and MSN at sequence and is occupied by the Jra transcription factor. This

732 Cell 142, 726–736, September 3, 2010 ª2010 Elsevier Inc.

A Figure 6. Occupancy of ATAC, H4K16
Chickadee gene span Acetylation, MAPKs, and H3S10P on the
chic CDS chickadee Gene with and without Osmotic
Probes ChicA ChicB ChicC Stress
1K bp
Enhancer (A) Schematic representation of the chickadee
B ChIP-Atac2 on Chickadee Sorbitol (-) C ChIP-Atac2 on Chickadee Sorbitol (+) AP-1 gene including 7 kb upstream of the transcription
dsRNA-Cont dsRNA-Atac2 dsRNA-Cont dsRNA-Atac2
A CG10238
A Jra
A MSN dsRNA-CG10238
A CG10238
A Jra
start site indicating the position of different probes
4% 1.00%
on the gene (ChicA-C on x axis for B–K). The gray

box indicates the coding sequence (CDS), the gray

8% 0.60%

0.50% circles indicate the enhancer, and the white circles
0.04% 0.30%
indicate predicted AP-1 sites.

(B–K) ChIP assays were performed with antibodies
ChicA ChicB ChicC ChicA ChicB ChicC
against Atac2, H4K16ac, pMKK4 (S257T261),
D ChIP-H4K16ac on Chickadee Sorbitol (-) E ChIP-H4K16ac on Chickadee Sorbitol (+) MSN, and phospho-H3S10 (H3S10P) from ATAC
A Cont
A CG10238
A At
A ac2
A Cont
A CG10238
A At
A ac2
A MSN and MSN knockdown S2 cells with or without
A Jra
dsRNA-Jra dsRNA-Jra
A Jra
0.90% 0.90% sorbitol stimulation (as in Figure 4D) on the probed
0.80% 0.80%
0.70% 0.70% region of the chickadee gene and upstream region

.60% 0.60%

0.50% 0.50% as indicated in (A), and input chromatin was
.40% 0.40%

0.30% 0.30%
measured by qRT-PCR and normalized to input
(see Figure S7A). The ratios of the measured
ChicA ChicB ChicC ChicA ChicB ChicC quantities are represented. The average of three
F ChIP-pMKK4 on Chicadee Sorbitol (-) G ChIP-pMKK4 on Chickadee Sorbitol (+)
independent experiments is graphed. Error bars
A Cont
A CG10238
A At
A ac2
A Cont
A CG10238
A At
A ac2
A MSN represent standard deviation.
dsRNA-Jra dsRNA-
A Jra
.60% 1.20% See also Figures S2, S5, S6, and S7.
.50% 1.00%



.30% 0.60%

.20% 0.40%

.10% 0.20%

ChicA ChicB ChicC ChicA ChicB ChicC MSN knockdown cells but not in Atac2
H ChIP-Misshapen on Chickadee Sorbitol (-) I ChIP-Misshapen on Chickadee Sorbitol (+)
knockdown cells (Figure S7A). These
A Cont
A CG10238
A Jra
A At
A ac2
A Cont
A CG10238
A Jra
A At
A ac2
A MSN two regions were co-occupied by ATAC
0.30% 0.25%
and H4K16 acetylation (Figures 6B and
0.25% 0.20%
6D). The presence of Atac2 and H4K16



0.10% acetylation at these regions was not
affected by the knockdown of Jra, sug-
0.00% 0.00%

ChicA ChicB ChicC ChicA ChicB ChicC gesting ATAC does not require Jra to
J K bind the chickadee gene in the absence
ChIP-H3S10P on Chickadee Sorbitol (-) ChIP-H3S10P on Chickadee Sorbitol (+)

A Cont
A CG10238
A At
A ac2
A Cont
A CG10238
A At
A ac2
of osmotic stress. Interestingly, in the
A Jra
dsRNA-Jra dsRNA-Jra
A Jra
2.00% 18.00%
presence of osmotic stress, ATAC occu-
1.80% 16.00%
14.00% pancy and H4K16 acetylation were


increased at the ChicA region, and this
0.40% 4.00%
increased occupancy was reduced in
0 00%

ChicA ChicB ChicC

0 00%

ChicA ChicB ChicC

Jra knockdown cells (Figures 6C and 6E).
Phospho-MKK4 and MSN also occupied
the ChicA and ChicB regions, and these
raised the question as to whether ATAC and MAPKs move to kinases were also found in the open reading frame (ChicC) in
this location because Jra is bound there or because they need the absence of osmotic stress (Figures 6F and 6H). In contrast
to associate with the coding region. Thus, we performed a ChIP to what was seen at the Jra gene, on chickadee the kinases
assay on another JNK target gene, chickadee. Chickadee does move to the upstream ChicA region upon osmotic stress and their
not have an apparent NF-jun binding site in the coding region. It occupancy is largely (MKK4) or slightly (MSN) dependent on Jra
does, however, have several AP-1-like sequences upstream of (Figures 6G and 6I). Occupancy of ChicA by the kinases under
the promoter and at a far-upstream enhancer (Figure 6A). We these conditions is slightly suppressed by Atac2 but completely
analyzed 7 kb upstream of the chickadee gene and found five dependent on CG10238. Indeed, CG10238 was crucial for the
AP-1-like sequences. We probed a region containing an AP-1- occupancy of ATAC, MKK4, and MSN on the far-upstream
like site, which was also near two CCAAT motifs that are 6.4 kb ChicA region under conditions of osmotic stress (Figures 6C,
upstream from the coding sequence (ChicA primer). We also 6G, and 6I). At the Jra gene, occupancy of the primer D region
probed adjacent to the AP-1-like site 2 kb upstream from the containing the NF-Jun binding site by ATAC, Active JNK, and
coding sequence close to the predicted promoter (ChicB primer) phospho-MKK4 upon osmotic stress was dependent on
and the middle of the coding region (ChicC primer) (Figure 6A). CG10238. As seen on Jra, H3S10P on chickadee was sup-
ChIP analysis revealed that Jra was localized at the ChicA and pressed by ATAC and MSN in the absence of osmotic stress
ChicB regions, suggesting Jra binds AP-1 sites on chickadee and, although H3S10P increased upon osmotic stress, these
(Figure S7A). Jra occupancy was reduced in CG10238 and levels were suppressed by MSN specifically (Figures 6J and 6K).

Cell 142, 726–736, September 3, 2010 ª2010 Elsevier Inc. 733

A The models presented in Figure 7 summarize the molecular
ATAC events occurring on Jra and chickadee as revealed by the ChIP
JNK data. Although differing in details having to do with the architec-
Atac2 Jra Jra-P MKK4 ture of each gene, comparison of the expression and ChIP data
Jra K16ac
Osmotic stress from Jra and chickadee reveals strikingly common features
E AP-1 NF-jun TATA-like
about the role of ATAC at these genes. First, in the absence of
MSN osmotic stress, ATAC is required for the basal expression of
these genes, which is most highly dependent on the Atac2 ace-
tyltransferase subunit. Second, upon osmotic stress, ATAC
suppresses induced transcription levels of these genes, which
CG10238 JNK-P
is most highly dependent on the CG10238 subunit. Third, upon
Atac2 osmotic stress, ATAC and the MAPKs move to the region of
K16ac K16ac the gene where Jra is bound. And fourth, the relocalization of
E AP-1 NF-jun TATA-like ATAC and MAPKs is dependent on CG10238, the MAP kinase
C MSN upstream kinase binding inhibitory protein. Thus, ATAC coordi-
P-1 nates the occupancy of Jra and upstream kinases while it
E E A 16ac
K controls the levels of JNK activation and target gene induction.

Osmotic stress
ATAC shares four subunits with the SAGA transcription coactiva-
Chickadee AP-1 tor complex and was anticipated to function as a coactivator.
MKK4-P However, preliminary studies indicated that ATAC did not
AP-1c interact with the activation domains of VP16 or p53 under
E K160aP
D S10P S10P S
conditions in which SAGA bound (Kusch et al., 2003). Prompted
by the identification of peptides from the Jra (Drosophila c-Jun)
transcription factor in ATAC affinity purifications, we demon-
strated that Jra coimmunoprecipitates with ATAC subunits. We
K16ac further demonstrate that ATAC is localized to Jra target genes
Chickadee AP-1 (Jra and chickadee) and is required for H4K16 acetylation at
the Jra and chickadee enhancer and promoter. ATAC was
required for basal levels of expression of these genes, which
Figure 7. Models Summarizing ATAC Functions and the Molecular depended strongly on the Atac2 acetyltransferase subunit.
Events Occurring on Jra and chickadee as Revealed by the ChIP Thus, to our knowledge, this is the first report to demonstrate
that ATAC functions as a transcription cofactor.
Summaries of ChIP data on the Jra gene without and with osmotic stress are
shown in (A) and (B), respectively. Summaries of ChIP data on the chickadee
Our pursuit of the functions of the CG10238 subunit of ATAC
gene without and with osmotic stress are shown in (C) and (D), respectively. and the potential role of its MoaE domain led us to discover
(A) In the absence of osmotic stress, ATAC occupies and acetylates H4K16 that ATAC is intimately integrated into MAP kinase signaling
(K16ac) at the enhancer (E) and promoter regions, containing AP-1 sites and target gene expression. CG10238 was found to inhibit
(AP-1), of the Jra gene and is required for basal levels of Jra transcription (small JNK activation by osmotic stress, an activity mediated through
arrow). ATAC blocks the recruitment of MKK4 to the enhancer and MSN to the
the MoaE domain. Thus, structural features of this protein that
promoter while inhibiting the activation of JNK. The phosphorylation of H3S10
(S10P) on the promoter region is suppressed by MSN.
are utilized for molybdopterin synthesis have been conscripted
(B) When cells are exposed to osmotic stress, the stress-activated kinases in in ATAC to function in the regulation of MAP kinase signaling.
the JNK cascade are recruited to the Jra binding motif (NF-jun) on the Jra gene The MSN kinase is the most likely direct target of CG10238, as
by ATAC, especially dependent on CG10238. In addition to interacting with the it coimmunoprecipitates with ATAC subunits. However, we
promoter region, ATAC also interacts with the NF-jun site by interaction with cannot rule out potential interactions with other kinases; for
Jra and further acetylates H4K16 at this site. ATAC continues to limit the extent example, peptides from MKK4 have been found in ATAC purifi-
of JNK activation. Phosphorylation of H3S10 is increased across the Jra gene
cations (Figure 1). Under conditions of osmotic stress, ATAC
by osmotic stress but is still suppressed by MSN. Jra transcription is rapidly
and transiently induced (large arrow). suppresses the level of target gene expression in a manner
(C) The chickadee gene does not have an apparent NF-jun site. In the absence strongly dependent on CG10238. We believe this is due to inhi-
of osmotic stress, Jra, ATAC, and acetyl-H4K16 are found at two AP-1 sites, bition of JNK activation rather than a direct negative effect of
which are 2 and 7 kb upstream from the coding sequence. Upstream kinases ATAC on transcription. In fact, ATAC may still function as a posi-
occupy these sites and the downstream coding region. tive cofactor for induced Jra expression, as the levels of H3S10P
(D) In the presence of osmotic stress, ATAC preferentially relocalizes to the
on the gene, considered a positive mark for transcription
far-upstream AP-1 site and recruits Jra to this location. Phospho-MKK4
(MKK4-P) and MSN are also relocalized to this far-upstream AP-1 site in
a CG10238- and Jra-dependent manner. The CG10238 subunit of ATAC is the levels of phosphorylated H3S10 are suppressed by MSN across the
required for ATAC, Jra, MKK4-P, and MSN relocalization to the far-upstream chickadee gene.
enhancer. Transcription of chickadee is activated (large arrow) whereas See also Figure S7.

734 Cell 142, 726–736, September 3, 2010 ª2010 Elsevier Inc.

(Zippo et al., 2009), are dependent on ATAC (Figure 5M). In Multidimensional Protein Identification Technology Analysis
a manner strongly dependent on CG10238, ATAC plays a striking MudPIT analysis was performed as described in the Extended Experimental
role in the binding of MAPKs to the Jra and chickadee genes.
Upon osmotic stress, ATAC was required for the recruitment of
the JNK, MKK4, and MSN kinases to the site of Jra binding.
The only exception was the binding of MSN to the Jra gene Supplemental Information includes Extended Experimental Procedures and
(Figure 5K). However, MSN required CG10238 for binding chick- seven figures and can be found with this article online at doi:10.1016/j.cell.
adee under inducing conditions (Figure 6I). The crucial role of 2010.07.045.
CG10238 in both the recruitment of the kinases and the inhibition
of signaling is most consistent with a model by which ATAC ACKNOWLEDGMENTS

inhibits JNK activation at the downstream target genes. The

We thank the Workman Lab members, Bjoern Gaertner in the Zeitlinger Lab,
fact that ATAC and the MAPKs all move to the site of Jra binding and Hidehisa Takahashi in the Conaway Lab for advice and support during
suggests that a localized signaling network is organized at the this project. This research was supported by the Stowers Institute for Medical
site occupied by the downstream transcription factor. Research.
Our data indicate that a positive cofactor of JNK target gene
transcription, ATAC, also serves to negatively feed back for inhi- Received: December 30, 2009
Revised: May 14, 2010
bition of the JNK signaling pathway. Why would a downstream
Accepted: July 1, 2010
activator also repress upstream signaling? The JNK pathway Published: September 2, 2010
and c-Jun have been shown to suppress p53-dependent
apoptosis and cellular senescence (Alexaki et al., 2008; Butter- REFERENCES
field et al., 1997; Das et al., 2007; Fan et al., 2010; Ham et al.,
2000; Takekawa and Saito, 1998; Yang et al., 1997). Moreover, Alepuz, P.M., Jovanovic, A., Reiser, V., and Ammerer, G. (2001). Stress-
induced MAP kinase Hog1 is part of transcription activation complexes. Mol.
c-Jun has been shown to repress transcription of the p53 gene
Cell 7, 767–777.
and hence its downstream targets (Schreiber et al., 1999).
Alepuz, P.M., de Nadal, E., Zapater, M., Ammerer, G., and Posas, F. (2003).
Thus, a delicate balance of JNK signaling and p53 expression
Osmostress-induced transcription by Hot1 depends on a Hog1-mediated
controls cell growth, proliferation, and apoptosis. If ATAC recruitment of the RNA Pol II. EMBO J. 22, 2433–2442.
suppression of JNK is involved in this process, we might expect Alexaki, V.I., Javelaud, D., and Mauviel, A. (2008). JNK supports survival in
loss of ATAC to result in decreased expression of p53 due to melanoma cells by controlling cell cycle arrest and apoptosis. Pigment Cell
increased JNK activity (Schreiber et al., 1999). Indeed, whereas Melanoma Res 21, 429–438.
knockdown of Jra increased p53 expression in S2 cells, knock- Altschul, S.F., Madden, T.L., Schaffer, A.A., Zhang, J., Zhang, Z., Miller, W.,
down of ATAC subunits resulted in a decrease in p53 expression and Lipman, D.J. (1997). Gapped BLAST and PSI-BLAST: a new generation
consistent with ATAC suppression of JNK activation (Figure S7B). of protein database search programs. Nucleic Acids Res. 25, 3389–3402.
Our results reveal a mechanism for governing the transcrip- Angel, P., and Karin, M. (1991). The role of Jun, Fos and the AP-1 complex in
tional response to signaling pathways. By serving as both a cell-proliferation and transformation. Biochim. Biophys. Acta 1072, 129–157.

downstream transcriptional cofactor and as an inhibitor of Angel, P., Hattori, K., Smeal, T., and Karin, M. (1988). The jun proto-oncogene
is positively autoregulated by its product, Jun/AP-1. Cell 55, 875–885.
upstream signaling, ATAC serves as a master regulator which
Butterfield, L., Storey, B., Maas, L., and Heasley, L.E. (1997). c-Jun
administers the appropriate level of response to the inducing
NH2-terminal kinase regulation of the apoptotic response of small cell lung
cancer cells to ultraviolet radiation. J. Biol. Chem. 272, 10110–10116.
Ciurciu, A., Komonyi, O., Pankotai, T., and Boros, I.M. (2006). The Drosophila
EXPERIMENTAL PROCEDURES histone acetyltransferase Gcn5 and transcriptional adaptor Ada2a are
involved in nucleosomal histone H4 acetylation. Mol. Cell. Biol. 26, 9413–9423.
Cell Lines, Extract Preparation, and Complex Purification Das, M., Jiang, F., Sluss, H.K., Zhang, C., Shokat, K.M., Flavell, R.A., and
The nuclear extracts from S2 cell lines were generated as described in
Davis, R.J. (2007). Suppression of p53-dependent senescence by the JNK
Extended Experimental Procedures and were used for FLAG affinity purifica-
signal transduction pathway. Proc. Natl. Acad. Sci. USA 104, 15759–15764.
tion as previously described (Suganuma et al., 2008).
de Nadal, E., and Posas, F. (2010). Multilayered control of gene expression by
stress-activated protein kinases. EMBO J. 29, 4–13.
Chromatin Immunoprecipitation Assay
de Nadal, E., Alepuz, P.M., and Posas, F. (2002). Dealing with osmostress
The crosslinked S2 cell pellets were washed, resuspended, and sonicated,
through MAP kinase activation. EMBO Rep. 3, 735–740.
and the DNA was immunoprecipitated with antibodies using Dynabeads
(Invitrogen). The bound DNA and input DNA were incubated with RNaseA de Nadal, E., Zapater, M., Alepuz, P.M., Sumoy, L., Mas, G., and Posas, F.
and the crosslinking was reversed. The ethanol-precipitated DNA pellets (2004). The MAPK Hog1 recruits Rpd3 histone deacetylase to activate
were resuspended and then were analyzed by quantitative real-time RT-PCR. osmoresponsive genes. Nature 427, 370–374.
Antibodies, buffer content, and primers for ChIP-qRT-PCR are described in Devary, Y., Gottlieb, R.A., Lau, L.F., and Karin, M. (1991). Rapid and
the Extended Experimental Procedures. preferential activation of the c-jun gene during the mammalian UV response.
Mol. Cell. Biol. 11, 2804–2811.
Osmotic Stress Response in dsRNAi Knockdown S2 Cells Dioum, E.M., Wauson, E.M., and Cobb, M.H. (2009). MAP-ping unconven-
and siRNA Knockdown 293T Cells tional protein-DNA interactions. Cell 139, 462–463.
The dsRNAi was transfected in S2 cells. The siRNA was transfected in 293T Edmunds, J.W., and Mahadevan, L.C. (2004). MAP kinases as structural
cells as described in the Extended Experimental Procedures. The cells were adaptors and enzymatic activators in transcription complexes. J. Cell Sci.
incubated with 500 mM sorbitol for 12 min before harvest. 117, 3715–3723.

Cell 142, 726–736, September 3, 2010 ª2010 Elsevier Inc. 735

Fan, Y., Lee, T.V., Xu, D., Chen, Z., Lamblin, A.F., Steller, H., and Bergmann, A. Pokholok, D.K., Zeitlinger, J., Hannett, N.M., Reynolds, D.B., and Young, R.A.
(2010). Dual roles of Drosophila p53 in cell death and cell differentiation. Cell (2006). Activated signal transduction kinases frequently occupy target genes.
Death Differ. 17, 912–921. Science 313, 533–536.
Fukuyama, K., Yoshida, M., Yamashita, A., Deyama, T., Baba, M., Suzuki, A., Proft, M., Mas, G., de Nadal, E., Vendrell, A., Noriega, N., Struhl, K., and Posas,
Mohri, H., Ikezawa, Z., Nakajima, H., Hirai, S., et al. (2000). MAPK upstream F. (2006). The stress-activated Hog1 kinase is a selective transcriptional elon-
kinase (MUK)-binding inhibitory protein, a negative regulator of MUK/dual gation factor for genes responding to osmotic stress. Mol. Cell 23, 241–250.
leucine zipper-bearing kinase/leucine zipper protein kinase. J. Biol. Chem. Rozek, D., and Pfeifer, G.P. (1993). In vivo protein-DNA interactions at the c-jun
275, 21247–21254. promoter: preformed complexes mediate the UV response. Mol. Cell. Biol. 13,
Guelman, S., Suganuma, T., Florens, L., Swanson, S.K., Kiesecker, C.L., 5490–5499.
Kusch, T., Anderson, S., Yates, J.R., III, Washburn, M.P., Abmayr, S.M., Rozek, D., and Pfeifer, G.P. (1995). In vivo protein-DNA interactions at the c-jun
et al. (2006). Host cell factor and an uncharacterized SANT domain protein promoter in quiescent and serum-stimulated fibroblasts. J. Cell. Biochem. 57,
are stable components of ATAC, a novel dAda2A/dGcn5-containing histone 479–487.
acetyltransferase complex in Drosophila. Mol. Cell. Biol. 26, 871–882.
Rudolph, M.J., Wuebbens, M.M., Rajagopalan, K.V., and Schindelin, H. (2001).
Ham, J., Eilers, A., Whitfield, J., Neame, S.J., and Shah, B. (2000). c-Jun and Crystal structure of molybdopterin synthase and its evolutionary relationship to
the transcriptional control of neuronal apoptosis. Biochem. Pharmacol. 60, ubiquitin activation. Nat. Struct. Biol. 8, 42–46.
Schreiber, M., Kolbus, A., Piu, F., Szabowski, A., Mohle-Steinlein, U., Tian, J.,
Hu, S., Xie, Z., Onishi, A., Yu, X., Jiang, L., Lin, J., Rho, H.S., Woodard, C., Karin, M., Angel, P., and Wagner, E.F. (1999). Control of cell cycle progression
Wang, H., Jeong, J.S., et al. (2009). Profiling the human protein-DNA interac- by c-Jun is p53 dependent. Genes Dev. 13, 607–619.
tome reveals ERK2 as a transcriptional repressor of interferon signaling. Cell
Schwarz, G., and Mendel, R.R. (2006). Molybdenum cofactor biosynthesis and
139, 610–622.
molybdenum enzymes. Annu. Rev. Plant Biol. 57, 623–647.
Iyer, L.M., Burroughs, A.M., and Aravind, L. (2006). The prokaryotic anteced-
Su, Y.C., Treisman, J.E., and Skolnik, E.Y. (1998). The Drosophila Ste20-
ents of the ubiquitin-signaling system and the early evolution of ubiquitin-like
related kinase misshapen is required for embryonic dorsal closure and acts
b-grasp domains. Genome Biol. 7, R60.
through a JNK MAPK module on an evolutionarily conserved signaling
Jasper, H., Benes, V., Schwager, C., Sauer, S., Clauder-Munster, S., Ansorge, pathway. Genes Dev. 12, 2371–2380.
W., and Bohmann, D. (2001). The genomic response of the Drosophila embryo
Suganuma, T., Gutierrez, J.L., Li, B., Florens, L., Swanson, S.K., Washburn,
to JNK signaling. Dev. Cell 1, 579–586.
M.P., Abmayr, S.M., and Workman, J.L. (2008). ATAC is a double histone
Kayali, A.G., Austin, D.A., and Webster, N.J. (2000). Stimulation of MAPK acetyltransferase complex that stimulates nucleosome sliding. Nat. Struct.
cascades by insulin and osmotic shock: lack of an involvement of p38 Mol. Biol. 15, 364–372.
mitogen-activated protein kinase in glucose transport in 3T3-L1 adipocytes.
Takekawa, M., and Saito, H. (1998). A family of stress-inducible GADD45-like
Diabetes 49, 1783–1793.
proteins mediate activation of the stress-responsive MTK1/MEKK4 MAPKKK.
Kimura, A., Matsubara, K., and Horikoshi, M. (2005). A decade of histone acet- Cell 95, 521–530.
ylation: marking eukaryotic chromosomes with specific codes. J. Biochem.
Thomson, S., Mahadevan, L.C., and Clayton, A.L. (1999). MAP kinase-
138, 647–662.
mediated signalling to nucleosomes and immediate-early gene induction.
Kusch, T., Guelman, S., Abmayr, S.M., and Workman, J.L. (2003). Two Semin. Cell Dev. Biol. 10, 205–214.
Drosophila Ada2 homologues function in different multiprotein complexes.
Treisman, J.E., Ito, N., and Rubin, G.M. (1997). misshapen encodes a protein
Mol. Cell. Biol. 23, 3305–3319.
kinase involved in cell shape control in Drosophila. Gene 186, 119–125.
Kyriakis, J.M., and Avruch, J. (2001). Mammalian mitogen-activated protein Wang, Y.L., Faiola, F., Xu, M., Pan, S., and Martinez, E. (2008). Human ATAC is
kinase signal transduction pathways activated by stress and inflammation. a GCN5/PCAF-containing acetylase complex with a novel NC2-like histone
Physiol. Rev. 81, 807–869. fold module that interacts with the TATA-binding protein. J. Biol. Chem. 283,
Lawrence, M.C., Shao, C., McGlynn, K., Naziruddin, B., Levy, M.F., and Cobb, 33808–33815.
M.H. (2009). Multiple chromatin-bound protein kinases assemble factors that Yang, D.D., Kuan, C.Y., Whitmarsh, A.J., Rincon, M., Zheng, T.S., Davis, R.J.,
regulate insulin gene transcription. Proc Natl Acad Sci USA 106, 22181–22186. Rakic, P., and Flavell, R.A. (1997). Absence of excitotoxicity-induced
Lee, K.K., and Workman, J.L. (2007). Histone acetyltransferase complexes: apoptosis in the hippocampus of mice lacking the Jnk3 gene. Nature 389,
one size doesn’t fit all. Nat. Rev. Mol. Cell Biol. 8, 284–295. 865–870.
Leimkuhler, S., Freuer, A., Araujo, J.A., Rajagopalan, K.V., and Mendel, R.R. Yang, S.H., Sharrocks, A.D., and Whitmarsh, A.J. (2003). Transcriptional
(2003). Mechanistic studies of human molybdopterin synthase reaction and regulation by the MAP kinase signaling cascades. Gene 320, 3–21.
characterization of mutants identified in group B patients of molybdenum Zapater, M., Sohrmann, M., Peter, M., Posas, F., and de Nadal, E. (2007).
cofactor deficiency. J. Biol. Chem. 278, 26127–26134. Selective requirement for SAGA in Hog1-mediated gene expression depend-
Mas, G., de Nadal, E., Dechant, R., Rodriguez de la Concepcion, M.L., Logie, ing on the severity of the external osmostress conditions. Mol. Cell. Biol. 27,
C., Jimeno-Gonzalez, S., Chavez, S., Ammerer, G., and Posas, F. (2009). 3900–3910.
Recruitment of a chromatin remodelling complex by the Hog1 MAP kinase Zippo, A., Serafini, R., Rocchigiani, M., Pennacchini, S., Krepelova, A., and
to stress genes. EMBO J. 28, 326–336. Oliviero, S. (2009). Histone crosstalk between H3S10ph and H4K16ac
Morrison, D.K., Murakami, M.S., and Cleghon, V. (2000). Protein kinases and generates a histone code that mediates transcription elongation. Cell 138,
phosphatases in the Drosophila genome. J. Cell Biol. 150, F57–F62. 1122–1136.

736 Cell 142, 726–736, September 3, 2010 ª2010 Elsevier Inc.

A Bacterial mRNA Leader that Employs
Different Mechanisms to Sense
Disparate Intracellular Signals
Sun-Yang Park,1,3,4 Michael J. Cromie,1,2,3,5 Eun-Jin Lee,1,2,5 and Eduardo A. Groisman1,2,5,*
1Department of Molecular Microbiology
2Howard Hughes Medical Institute
Washington University School of Medicine, 660 South Euclid Avenue, Campus Box 8230, St. Louis, MO 63110, USA
3These authors contributed equally to this work
4Present address: Section of Microbial Pathogenesis, Yale University School of Medicine, Boyer Center for Molecular Medicine,

295 Congress Avenue, New Haven, CT 06536-0812, USA

5Present address: Section of Microbial Pathogenesis, Howard Hughes Medical Institute, Yale University School of Medicine,

Boyer Center for Molecular Medicine, 295 Congress Avenue, New Haven, CT 06536-0812, USA
DOI 10.1016/j.cell.2010.07.046

SUMMARY These mRNA leaders typically control the expression of the

proteins that synthesize and/or transport the metabolites or
Bacterial mRNAs often contain leader sequences ions to which they respond. Whereas leaders known as ribos-
that respond to specific metabolites or ions by witches sense metabolites or ions directly (Henkin, 2008; Winkler
altering expression of the associated downstream and Breaker, 2005), the presence of stretches of certain nucleo-
protein-coding sequences. Here we report that the tides in a leader sequence or particular codons in a leader ORF
leader RNA of the Mg2+ transporter gene mgtA of enables cells to probe the cellular levels of the corresponding
nucleotides and amino acids indirectly. Here we identify a leader
Salmonella enterica, which was previously known
mRNA that controls gene expression by responding to Mg2+
to function as a Mg2+-sensing riboswitch, harbors directly as a riboswitch and to proline levels indirectly, via trans-
an 18 codon proline-rich open reading frame— lation of a proline-codon-rich open reading frame.
termed mgtL—that permits intracellular proline to The mRNA for the Mg2+ transporter gene mgtA from Salmo-
regulate mgtA expression. Interfering with mgtL nella enterica serovar Typhimurium includes a 264 nucleotide
translation by genetic, pharmacological, or environ- long leader sequence that functions as a Mg2+-responding ribos-
mental means was observed to increase the mRNA witch determining whether transcription proceeds into the mgtA
levels from the mgtA coding region. Substitution of CR (Cromie et al., 2006). In low Mg2+, the mgtA leader RNA
the mgtL proline codons by other codons abolished adopts a conformation (i.e., stem loop C) that favors transcrip-
the response to proline and to hyperosmotic stress tion of the full-length mgtA mRNA (Cromie et al., 2006) (Figure 1)
but not to Mg2+. Our findings show that mRNA leader and production of the MgtA protein (Cromie and Groisman,
2010). MgtA-mediated Mg2+ uptake restores cytoplasmic Mg2+
sequences can consist of complex regulatory
to prestress levels, which promotes a different conformation in
elements that utilize different mechanisms to sense
the mgtA leader RNA (i.e., stem loop B) that hinders transcription
separate signals and mediate an appropriate cellular elongation into the mgtA CR (Cromie et al., 2006) (Figure 1).
response. In addition to being regulated by a Mg2+-sensing riboswitch,
mgtA transcription is controlled at the initiation step by two
INTRODUCTION DNA binding proteins: PhoP, which activates transcription
when its cognate sensor PhoQ detects low extracytoplasmic
The leader region (LR) of many bacterial mRNAs has the ability Mg2+ (Garcia Vescovi et al., 1996), acid pH (Prost et al., 2007),
to form mutually exclusive secondary structures that determine or antimicrobial peptides (Bader et al., 2005); and Rob, the over-
whether transcription will continue into the adjacent coding expression of which stimulates mgtA transcription from a site
region (CR). Which secondary structure forms is governed located downstream of the PhoP-dependent start site (Barchiesi
by the binding of specific metabolites, ions, tRNAs, or proteins et al., 2008) (Figure 1). Thus, there is PhoP/PhoQ-dependent
to the LR, by pausing of RNA polymerase at particular sequences production of the mgtA leader mRNA following bacterial growth
within the LR and by translation of short open reading frames in media with <1 mM Mg2+ (Cromie et al., 2006) or pH 5.7 (Choi
(ORFs) located in the leader RNA (Grundy and Henkin, 2006; et al., 2009); however, due to the riboswitch action, the mRNA
Henkin, 2008; Henkin and Yanofsky, 2002; Landick et al., 1996; corresponding to the mgtA CR is observed primarily upon growth
Merino and Yanofsky, 2005; Turnbough and Switzer, 2008). in media containing very low (i.e., 10 mM) Mg2+ (Choi et al., 2009;

Cell 142, 737–748, September 3, 2010 ª2010 Elsevier Inc. 737

Figure 1. Expression of the Salmonella
Mg2+ Transporter Gene mgtA Is Regulated
by the PhoP/PhoQ System, the Rob Protein,
and the mgtA LR
When the sensor PhoQ detects low extracytoplas-
mic Mg2+, acid pH, or antimicrobial peptides,
it promotes phosphorylation of the PhoP protein,
which binds to the mgtA promoter resulting in
transcription initiation. Rob promotes mgtA tran-
scription in response to a yet unidentified signal.
Transcription elongation into the mgtA CR is regu-
lated by the mgtA leader via a Mg2+-sensing
riboswitch and by translation of an 18 codon
proline-rich ORF designated mgtL. The mgtL
ribosome binding site is denoted by RBS adjacent
to a black line. Positions and sequences of stop
codon mutations in the strains used in the experi-
ments presented in Figure 4 are denoted below the
linear mgtL RNA sequence. High-Mg2+ and high-
proline conditions promote formation of stem
loop B, hindering transcription elongation into the
mgtA CR. Low Mg2+ and/or low proline favor
formation of stem loop C, resulting in transcription
of the mgtA CR. See also Figure S1.

formants recovered on X-Gal-containing

Luria-Bertani (LB) ampicillin agar plates
that were of a darker or lighter shade of
blue than those that received a plasmid
control were purified and their plasmid
DNA extracted and used to transform
strain YS809, a derivative of strain
Cromie and Groisman, 2010; Cromie et al., 2006). This raises the YS774 with the same mgtA-lac transcriptional fusion. One of
possibility of additional signals and/or genetic elements acting on the plasmids—designated pSL55—promoted higher levels of
the mgtA leader mRNA synthesized when Salmonella experi- mgtA-lac expression than the plasmid control (Figure 2A), like
ences inducing conditions for PhoP/PhoQ and/or Rob to it did in strain YS774, indicative that the observed phenotype
promote transcription elongation into the mgtA CR. was plasmid linked. Plasmid pSL55 appears to promote mgtA-
Here we demonstrate that the mgtA LR harbors an ORF rich in lac expression specifically because it had no effect on the Lac
proline codons that enables Salmonella to control transcription phenotype of strains harboring lac transcriptional fusions in the
elongation into the mgtA CR in response to changes in cytosolic CRs of the PhoP-dependent mgtC gene or the PhoP-indepen-
proline levels. We establish that high osmolarity promotes tran- dent corA gene (data not shown). The mgtC and corA transcripts
scription of the mgtA CR in a fashion dependent on the proline also include long leaders (Lejona et al., 2003) (T. Latifi and E.A.G.,
codons present in the identified ORF. Also, we show that the unpublished results) and specify proteins participating in Mg2+
low Mg2+ signal and the low proline signal act synergistically to homeostasis (Blanc-Potard and Groisman, 1997; Maguire,
increase the mRNA levels for the mgtA CR. Our findings highlight 2006; Soncini et al., 1996).
the complexity of RNA regulatory sequences and provide Sequencing of the insert in pSL55 revealed the presence
a singular example of a leader RNA that utilizes two distinct of 1579 nucleotides corresponding to position 78347–79925 in
mechanisms to detect two different signals. the chromosome of strain 14028s, which is a portion of the
3228 nucleotide long carB gene (see Table S2 for primers).
RESULTS The insert operates in an orientation-dependent manner
because it did not alter mgtA-lac expression when cloned into
Identification of a Translated ORF in the mgtA LR pUC18 where the opposite strand is transcribed (data not
In a search for new factors that might regulate mgtA expression shown). Although this suggested that pSL55 might promote
by acting upon the mgtA leader mRNA, we introduced a plasmid mgtA-lac expression by functioning as an antisense RNA for
library made in the multicopy number vector pUC19 into strain the carB transcript, we ruled out this possibility because
YS774 (see Table S1 available online), which harbors the pSL55 upregulated mgtA transcription even in a strain deleted
PhoP-independent plac1-6 promoter driving transcription of for the carB gene (Figure 2A). Given that a transcript correspond-
a lac fusion in the mgtA CR from the normal (i.e., PhoP-depen- ing to the strand opposite the carB gene was not detected when
dent) start site (Cromie et al., 2006). Ampicillin-resistant trans- RNA was harvested from wild-type Salmonella grown in LB or in

738 Cell 142, 737–748, September 3, 2010 ª2010 Elsevier Inc.

A ing at the ninth codon, produced high levels of b-galactosidase,
whereas an isogenic strain harboring a modified plasmid with
a stop codon after the 17th mgtL codon did not (Figure S1). (There
is a potential start codon at position 26–28 in the Salmonella mgtA
leader sequence that is in frame with the mgtL ORF and is
preceded by a possible ribosome binding site [Figure 1]. However,
the AUG at this position does not appear to be a true translation
start site because first, the sequence between nucleotides 26 and
70 is not conserved in the other examined species, and second,
wild-type Salmonella harboring plasmid pmgtLD2nt-0 lacZ, a
derivative of plasmid pmgtL-0 lacZ with two nucleotides deleted
at position 29–30, which creates a stop codon at position
79–81 for the ORF starting at position 26, still expressed high
levels of b-galactosidase [Figure S1].) The strains harboring plas-
mids pmgtL-0 lacZ and pmgtLD2nt-0 lacZ displayed similar levels
B of b-galactosidase following growth in low and high Mg2+
(Figure S1). These data indicate that mgtL translation is not regu-
lated by changes in the Mg2+ concentration in the media. More-
over, they reflect that in the pmgtL-0 lacZ and pmgtLD2nt-0 lacZ
plasmids, mgtL-lac transcription is initiated from a Mg2+-blind
promoter and the Mg2+-responding riboswitch is disrupted.
Figure 2. Plasmid pSL55 Promotes mgtA-lac Transcription by Inter-
fering with mgtL Translation Inhibiting mgtL Translation Promotes Expression
(A) b-galactosidase activity (Miller units) from a chromosomal mgtA-lac tran- of the mgtA CR
scriptional fusion driven by the plac1-6 promoter was determined in wild-type Plasmid pSL55 appears to upregulate mgtA-lac expression by
(YS809), carB (SP31), and hfq (SP54) strains harboring plasmids pSL55,
producing a trans-acting RNA that interferes with mgtL transla-
pSL55-Mut, or a control plasmid. Bacteria were grown in LB medium for
4.5 hr. Shown are the mean and standard deviation (SD) from two independent tion because first, the region of complementarity with the mgtA
experiments. leader corresponds to the ribosome binding site and start codon
(B) Nucleotide sequences corresponding to position 61–75 from the mgtA LR of mgtL (Figure 2B). Second, inactivation of the hfq gene, which
including the mgtL ribosome binding sequence (RBS; horizontal black line) and codes for the RNA chaperone usually required for pairing
start codon (bold AUG) (top); the sequence of the pSL55-derived transcript between trans-acting regulatory RNAs and their targets
that is complementary to the mgtA leader mRNA (middle); and the sequence
(Brennan and Link, 2007), eliminated pSL55’s regulatory effect
of the pSL55-Mut-derived transcript that is no longer complementary to
the mgtA leader mRNA (bottom).
(Figure 2A). And third, pSL55-Mut, a pSL55 derivative with nucle-
otide substitutions in the region of complementarity with the
mgtA LR (Figure 2B), could not upregulate mgtA expression
N-minimal media (data not shown), and that mgtA-lac was not (Figure 2A).
derepressed in the DcarB strain (Figure 2A), we concluded that If plasmid pSL55 derepresses mgtA-lac transcription by
pSL55 upregulates mgtA-lac expression by a mechanism that inhibiting mgtL translation, we reasoned that disrupting mgtL
does not involve the carB gene. translation by other means might have the same effect. To test
We then examined the nucleotide sequence of the insert in this hypothesis, we introduced mutations in plasmid pYS1010,
pSL55 and identified a 15 nucleotide segment that is comple- which harbors the plac1-6 derivative of the lac promoter driving
mentary to nucleotides 61–75 in the mgtA leader RNA transcription of the full-length mgtA LR fused to a promoterless
(Figure 2B). Interestingly, this region corresponds to the potential lacZ gene and confers Mg2+-regulated synthesis of b-galactosi-
start codon and ribosome binding site for a previously unidenti- dase (Cromie et al., 2006). Wild-type Salmonella harboring
fied 18 codon ORF—hereafter designated mgtL—that begins at pYS1010 derivatives with stop codons at positions 80–82,
position 71 and ends at position 124 (Figure 1). The presence of 89–91, or 98–100 (Figure 1) produced high levels of b-galactosi-
an ORF and ribosome binding site is conserved in the mgtA LRs dase when grown in high Mg2+ (Figure 4A). These levels were 46-
from Escherichia coli, Shigella flexneri, Klebsiella pneumoniae, to 89-fold higher than those displayed by the strain carrying the
Citrobacter koseri, Enterobacter sp. 638, Dickeya zeae, Dickeya original plasmid pYS1010.
dadantii, and Serratia proteamaculans (Figure 3A). The mgtA The high expression levels exhibited by strains harboring
leader sequences from all these species have the potential ability pYS1010 derivatives with stop codon mutations in mgtL are
to form the alternative stem loop structures B and C identified in probably due to lack of translation of the full-length mgtL
the Salmonella mgtA leader (Figure 1) (Cromie et al., 2006) (data (as opposed to resulting from major alterations in the mgtA
not shown). leader RNA structure that lock the riboswitch in an ON state)
We determined that mgtL is translated in vivo because wild- because first, a plasmid expressing the amber suppressor
type Salmonella carrying the medium-copy-number plasmid supF restored normal levels of mgtA-lac transcription to the
pmgtL-0 lacZ, with the full mgtL coding sequence and its putative strain harboring a pYS1010 derivative with an amber stop codon
ribosome binding site fused in frame to the E. coli lacZ gene start- at position 98–100 whereas the plasmid vector had no effect

Cell 142, 737–748, September 3, 2010 ª2010 Elsevier Inc. 739


Figure 3. The Presence of an Open Reading Frame and Ribosome-Binding Site Is Conserved in the mgtA LR from Salmonella enterica,
Escherichia coli, Shigella flexneri, Citrobacter koseri, Klebsiella pneumoniae, Enterobacter sp. 638, Dickeya zeae, Dickeya dadantii, and
Serratia proteamaculans
(A) Alignment of the RNA sequences corresponding to the ribosome binding site and mgtL from the species listed above. Sequences in blue correspond to mgtL.
Asterisks correspond to nucleotides conserved in all species. The number of nucleotides shown in each line is indicated to the right of the nucleotide sequences.
(B) Alignment of the deduced amino acid sequences corresponding to mgtL from the species listed above. Sequences in yellow correspond to proline residues.
Asterisks correspond to positions conserved in all species. The number of sense codons in the mgtL ORFs is indicated to the right of the deduced amino acid

(Figure 4C). Furthermore, the supF-promoted suppression was promoted by protein synthesis inhibitors is not limited to a partic-
specific because the supF-carrying plasmid did not reduce ular mechanism of action. Tetracycline exerts its effect on the
expression in a strain carrying a pYS1010 derivative with an mgtA LR (as opposed to the mgtA promoter) because
opal stop codon at the same position (Figure 4C), nor did it increased lacZ mRNA levels in wild-type Salmonella harboring
it modify it in a strain with the original pYS1010 plasmid carrying plasmid pYS1010 (data not shown). Furthermore, tetracycline
the wild-type mgtA leader sequence (Figure 4C). Second, acts by inhibiting mgtL translation because it did not promote
the structures of the wild-type and stop codon mutant mgtA a significant increase in lacZ mRNA levels in wild-type Salmo-
leaders were comparable when examined by in-line probing nella carrying the pYS1010 derivative with a stop codon at posi-
(Figure S2A). And third, in vitro transcription assays demon- tion 98–100 (data not shown). Taken together, our results
strated that Mg2+ regulated transcription elongation beyond indicate that interfering with mgtL translation by genetic or phar-
the mgtA LR similarly in DNA templates corresponding to the macological means increases the mRNA levels for the mgtA CR.
wild-type and stop codon mutant mgtA leaders (Figure S2B). The mgtL-encoded peptide does not appear to act in trans
We next tested whether inhibiting mgtL translation by means because a plasmid carrying mgtL DNA failed to restore normal
other than mutation of the mgtA LR affected the mRNA levels mgtA-lac expression in strains harboring pYS1010 derivatives
for the mgtA CR. The protein synthesis inhibitor tetracycline with stop codons at different positions in mgtL (Figure S2C).
promoted an 28-fold increase in the mgtA CR mRNA 15 min This suggested that mgtL exerts its regulatory effect in cis and
after its addition to wild-type Salmonella (Figure 4D). Thus, tetra- raised the question as to the physiological signal controlling
cycline could overcome the transcriptional silencing of the mgtA mgtL translation.
CR that normally takes place when wild-type Salmonella experi-
ences 500 mM Mg2+ (Cromie and Groisman, 2010; Cromie et al., Proline Limitation Enhances Transcription
2006). Tetracycline appears to affect the mRNA levels produced of the mgtA CR
from the mgtA CR specifically because there was little change in The Salmonella mgtL sequence includes four proline codons
the mRNA levels of the mgtA LR or the phoP CR, which were (Figure 1), which is a disproportionately high frequency for an
examined as controls (Figure 4D). Addition of chloramphenicol 18 codon ORF. The number and location of the proline codons
also increased the mRNA levels of the mgtA CR in wild-type (at the third, fifth, seventh, and ninth positions) are largely
Salmonella (data not shown), implying that the derepression conserved in the nine examined species (Figures 3A and 3B).

740 Cell 142, 737–748, September 3, 2010 ª2010 Elsevier Inc.

A B Figure 4. Translation of mgtL Governs
10000 10000 Transcription Elongation beyond the
mgtA LR
-galactosidase activity

-galactosidase activity
8000 8000
(A) b-galactosidase activity (Miller units) produced
(Miller units)

(Miller units)
6000 by wild-type Salmonella (14028s) harboring 6000
plasmid pYS1010 with the wild-type mgtA leader
4000 4000 (Cromie et al., 2006), or derivatives with stop
2000 2000
codons at different positions in mgtL and/or
mutations that hinder stem loop C formation:
0 0 pstop 80–82; pstop 89–91; pstop 98–100; pstop









107–109; pstop 110–112; pstop 107–109 D110–


-1 0



















112; pYS1010-G120C; pstop 80–82-G120C;






to p
to p






pYS1010-C145G; and pstop 98–100-C145G.










to p



Bacteria were grown in N-minimal medium with





10 mM Mg2+ for 4 hr. Shown are the mean and

SD from at least three independent experiments.
C D (B) b-galactosidase activity produced by the
pstop 98-100 pstop 98-100 pYS1010
strains listed in (A). Bacteria were grown in
-galactosidase activity

(amber) (opal) N-minimal medium with 10 mM Mg2+ for 4 hr.
Shown are the mean and SD from at least three
(Miller units)

Fold change

(T15 / T0 )

independent experiments.
20 (C) b-galactosidase activity produced by wild-type
2000 Salmonella (14028s) harboring plasmid pUHsupF
or the plasmid vector pUH21-2lacIq, and
pYS1010 or its derivatives with UAG (pstop
98–100; amber) or UGA (pstop 98–100; opal)











stop codons at position 98–100 in the mgtA leader.










Bacteria were grown in N-minimal medium with


500 mM Mg2+ for 3 hr and with 1 mM IPTG for

E F 1 hr. Shown are the mean and SD from three inde-
2.0 pendent experiments.
(no proline / proline)

(no proline / proline)

(D) Fold change in the mRNA levels of the mgtA LR

Fold change

Fold change

8 and the mgtA and phoP CRs produced by wild-

6 1.0
type Salmonella (14028s) treated with the protein
synthesis inhibitor tetracycline (25 mg/ml). Expres-
0.5 sion levels of target genes were normalized to that
of the 16S ribosomal RNA rrs gene. Fold change
0 0 was calculated by dividing the mRNA levels from




samples taken 15 min after treatment with tetracy-








cline (T150 ) by that of samples taken before addi-







tion of the antibiotic (T00 ). Shown are the mean



and SD from three independent experiments.

(E) Fold change in the mRNA levels of the mgtA LR and the mgtA and phoP CRs produced by a proline auxotrophic strain (EG19886) grown in modified N-minimal
media with 500 mM Mg2+ in the presence of 1 mM proline for 1 hr, and then grown for 15 min in media containing or lacking proline. Expression levels of target
genes were normalized as described in (D). Fold change was calculated by dividing the mRNA levels of cells grown in the absence of proline by that of cells grown
in the presence of proline. Shown are the mean and SD from three independent experiments.
(F) Fold change in the mRNA levels of the mgtA LR and the mgtA and phoP CRs produced by wild-type Salmonella (14028s) grown and analyzed as described in
(D). Shown are the mean and SD from two independent experiments.
See also Figure S2.

Because the proline codons are located in the mgtL region where The mRNA level for the mgtA CR was 8-fold higher in organisms
introduction of stop codons heightens expression downstream of grown in the absence of proline than in those grown in its pres-
the mgtA leader (Figure 1), we hypothesized that a drop in the ence (Figure 4E). This effect was unique to the mgtA CR, as no
cytosolic proline levels might decrease the availability of proline- differences were detected in the mRNA levels corresponding
charged tRNAs, causing ribosome stalling at mgtL proline to the mgtA LR or the phoP CR (Figure 4E). Furthermore,
codons, leading to formation of stem loop C and an increase in it was specific for proline, as limitation for histidine, which does
the levels of the mgtA CR mRNA. By contrast, when cytosolic not have codons in the mgtL sequence (Figure 1), failed to
proline is abundant, coupling of mgtA leader transcription and increase the mRNA level for the mgtA CR (data not shown).
mgtL translation would be restored, thereby favoring formation Indeed, despite the presence of three arginine codons in mgtL
of stem loop B and reducing the production of mgtA CR mRNA. (Figure 1), arginine limitation did not promote an increase in the
To explore this hypothesis, we grew a proline auxotroph in the mRNA levels of the mgtA CR in an arginine auxotroph (data not
presence of 1 mM proline for 1 hr, washed the cells, split shown). Perhaps this reflects the location of the arginine codons,
the culture into two different media—one containing and one one of which is in stem loop C, and that the remaining arginine
lacking proline—and then harvested the mRNA 15 min later. codons may not be sufficient to mediate mgtA derepression.

Cell 142, 737–748, September 3, 2010 ª2010 Elsevier Inc. 741

A Proline limitation promoted a 2-fold increase in the mRNA
mgtA leader mgtA coding phoP coding
levels of the mgtA CR in wild-type (i.e., prototrophic) Salmonella
(Figure 4F). Although this is lower than the increase observed in
12 the proline auxotroph (Figure 4E), it is similar to the gene expres-
(no proline / proline)

10 sion changes promoted by branched-chain amino acids in the

Fold change

attenuation-regulated ilvGMEDA operon (Chen et al., 1991).
Moreover, it is higher than what has been reported for the atten-
uation-regulated trp operon, where starvation for tryptophan
4 does not alter trp expression in a tryptophan prototroph
2 (Yanofsky and Horn, 1994).




Pr o3

7 ,9
P r 5 ,7 ,

Pr 5,7,
Pr 5,7,

The mgtL Proline Codons Are Necessary






,5 ,






for the Response to Proline Limitation

We determined that proline limitation failed to promote an
B increase in the mRNA levels for the mgtA CR in a derivative of
80 mgtA leader mgtA coding phoP coding the proline auxotroph with all four mgtL proline codons replaced
by codons specifying other amino acids (Figure 5A). As found
(10 µ M Mg2+ / 500 µM Mg2+)


60 with the isogenic mgtL+ strain, proline limitation had little effect
on the mRNA levels for the mgtA leader and phoP CRs
Fold change


(Figure 5A). Importantly, the mutant still responded to changes
in the Mg2+ concentration (Figure 5B) (albeit not as well as the
isogenic strain with a wild-type mgtA leader), indicating that
the proline codon substitutions did not lock the mutant into an
10 inactive conformation. These data indicate that the mgtL proline
0 codons are essential for the response to proline limitation, and


Pr o3


7 ,9
Pr o3
o3 ,9

o3 ,9
7 ,9
Pr o3

they suggest that this ability is distinct from the mgtA leader’s
P r 5 ,7,





,5 ,




role as a Mg2+-sensing device.



We hypothesized that the proline codon at the third position of

mgtL might not be required for regulation of mgtA expression by
C proline because D. dadantii does not have a proline codon at this
mgtA leader mgtA coding phoP coding position (Figures 3A and 3B). Indeed, a Salmonella strain in which
the mgtL proline codon at the third position was replaced by
Relative mRNA levels

60 a leucine codon derepressed the mgtA CR mRNA when limited

50 for proline, like the strain with the wild-type mgtL (Figure 5A).
40 However, this mutant still responded to changes in the Mg2+
concentration, but to a smaller degree than the mutant
substituted in all four mgtL proline codons (Figure 5B).
The results presented above suggested that replacement of
10 the mgtL proline codons at the fifth, seventh, and ninth positions
0 by codons specifying other amino acids might be sufficient to
H H L L H H L L H H L L [Mg2+]
+ - + - + - + - + - + - Proline eliminate the response to proline. As predicted, there was little
derepression of the mgtA CR mRNA upon proline limitation in
Figure 5. The mgtA Leader Responds to Proline and Mg2+ Indepen-
a derivative of the proline auxotroph with the last three mgtL
dently and Requires the mgtL Proline Codons for Regulation of the proline codons substituted by the same codons as in the mutant
mgtA CR by Changes in the Proline Concentration with substitutions in all four mgtL proline codons (Figure 5A). As
(A) Fold change in the mRNA levels of the mgtA LR and the mgtA and phoP with the other isogenic strains, proline limitation did not affect the
CRs produced by a proline auxotroph (SP1; mgtL+), or derivatives in which mRNA levels of the mgtA leader and phoP CRs (Figure 5A).
the mgtL proline codon at the third position was substituted (SP2; Pro3), the
The response to Mg2+ of the mgtL derivative with substitutions
last three mgtL proline codons were substituted (SP61; Pro5,7,9), or in which
in the last three proline codons was intermediate to that
all mgtL proline codons were substituted (SP8; Pro3,5,7,9). Bacteria were
grown and mRNA was analyzed as described in Figure 4E. Shown are the
mean and SD from three independent experiments.
(B) Fold change in the mRNA levels of the mgtA LR and the mgtA and phoP (C) Relative mRNA levels of the mgtA LR and the mgtA and phoP CRs
CRs produced by the strains described in (A). Bacteria were grown in modified produced by a proline auxotroph (EG19886) that experienced 500 mM Mg2+
N-minimal medium with 10 mM or 500 mM Mg2+ for 3.5 hr. Expression levels of and 1 mM proline, 500 mM Mg2+ and no proline, 5 mM Mg2+ and 1 mM proline,
target genes were normalized as described in (A). Fold change was calculated or 5 mM Mg2+ and no proline for 15 min. Relative mRNA levels were calculated
by dividing the mRNA levels of cells grown in N-minimal medium with 10 mM by dividing the mRNA levels of cells grown in a given condition by those
Mg2+ by those present in cells grown in 500 mM Mg2+. Shown are the mean present in cells grown in 500 mM Mg2+ and 1 mM proline. Shown are the means
and SD from three independent experiments. and SD from three independent experiments.

742 Cell 142, 737–748, September 3, 2010 ª2010 Elsevier Inc.

exhibited by the one with the third proline codon substituted and A mgtL+
the one with all four proline codons substituted (Figure 5B).
mgtA leader mgtA coding phoP coding
Synergism between the Low Mg and the Proline

Relative mRNA levels

Limitation Signals
To examine whether direct Mg2+ sensing by the mgtA leader RNA 5

affects indirect proline sensing by the mgtL ORF and vice versa, 4
we grew the mgtL+ proline auxotrophic strain in the presence of 3
500 mM Mg2+ and 1 mM proline for 1 hr, washed the cells, and split
the culture into four different media containing either 500 mM
Mg2+ and 1 mM proline, 500 mM Mg2+ and no proline, 5 mM 1
Mg2+ and 1 mM proline, or 5 mM Mg2+ and no proline. Fifteen 0
minutes later, we determined the mRNA levels corresponding - G N NG NP - G N NG NP - G N NG NP
to the mgtA and phoP CRs as well as to the mgtA LR. We found
that the mRNA levels corresponding to the mgtA CR were higher
in cells experiencing both low Mg2+ and low proline than in cells Pro3,5,7,9
limited only for Mg2+ or for proline (Figure 5C). As expected, the mgtA leader mgtA coding phoP coding
mRNA levels corresponding to the mgtA LR and the phoP CR 7

were higher in cells experiencing 5 mM Mg2+ than in those 6

Relative mRNA levels
exposed to 500 mM Mg2+ (Figure 5C), reflecting that the amount 5
of activated PhoP protein increases as the Mg2+ concentration
in the media decreases (Shin et al., 2006). Our results support
the notion that Mg2+ and proline are sensed independently 3
by the mgtA leader RNA. Furthermore, they suggest synergism 2
between the two inducing signals because the mRNA levels for
the mgtA CR were higher in cells experiencing both low Mg2+
and low proline than the sum of the mRNA levels produced by 0
cells experiencing only low Mg2+ or low proline (Figure 5C).
In agreement with this notion, wild-type Salmonella harboring
Figure 6. Hyperosmotic Shock Promotes Transcription of the mgtA
plasmid pYS1010 derivatives with early stop codons in the
CR in a Process that Requires the mgtL Proline Codons
mgtL sequence produced more b-galactosidase when grown in Relative mRNA levels of the mgtA LR and the mgtA and phoP CRs produced
low Mg2+ than the sum of the b-galactosidase activities produced by wild-type Salmonella (YS957) (A), or a derivative in which all mgtL proline
by the same strains grown in high Mg2+ plus the b-galactosidase codons were substituted (EG19870) (B). Bacteria were grown for 1 hr in modi-
activity produced by wild-type Salmonella carrying pYS1010 fied N-minimal medium without casamino acids containing 500 mM Mg2+ and
grown in low Mg2+ (Figures 4A and 4B). either no additional supplements (-), 1 mM glycine betaine (G), 0.3 M NaCl (N),
0.3 M NaCl and 1 mM glycine betaine (NG), or 0.3 M NaCl and 1 mM proline
(NP). Relative mRNA levels were calculated by dividing the mRNA levels of
Hyperosmotic Shock Promotes an Increase in the mgtA
cells grown under the specified condition by the mRNA levels present in cells
CR mRNA Levels grown in 500 mM Mg2+ (i.e., with no additional supplement). Shown are the
Proline plays two major functions in bacterial cells: it is a compo- mean and SD from three independent experiments.
nent of proteins and it can function as an osmoprotectant
(Csonka and Leisinger, 2007). Thus, we hypothesized that
when bacteria experience hyperosmotic shock, the increased the induction of the mgtA CR mRNA promoted by NaCl but
requirement for proline in osmoprotection might decrease its had negligible effects when added in the absence of NaCl
availability to charge proline tRNAs. This could potentially lead (Figure 6A). Proline had a similar (albeit not as strong) effect as
to ribosome stalling at the mgtL proline codons and result in glycine betaine (Figure 6A), presumably because it is not as
derepression of the mgtA CR. We tested this hypothesis by effective as glycine betaine in osmoprotection (Cayley et al.,
comparing the mRNA levels produced by wild-type Salmonella 1992). The increase in the mgtA CR mRNA levels provoked by
experiencing hyperosmotic shock in the presence or absence hyperosmotic shock requires an intact mgtL ORF, because it
of osmoprotectants. was not observed in an isogenic strain substituted in all four
We determined that the mRNA levels corresponding to the mgtL proline codons (Figure 6B). Cumulatively, these data
mgtA CR were 6-fold higher when Salmonella experienced demonstrate that hyperosmotic shock promotes transcription
500 mM Mg2+ + 0.3 M NaCl for 1 hr than in organisms grown in of the mgtA CR in an mgtL-dependent manner.
500 mM Mg2+ (Figure 6A). By contrast, the mRNA levels for the
mgtA LR and the phoP CR were similar under the two growth Formation of Stem Loop C Is Necessary
conditions (Figure 6A), indicative that the induction of the mgtA for Transcription of the mgtA CR
CR mRNA levels promoted by high osmolarity is not mediated Because transcription and translation are coupled in bacteria,
by the PhoP-dependent mgtA promoter. Addition of the osmo- when cytosolic proline is abundant, a ribosome translating the
protectant glycine betaine together with NaCl compromised complete mgtL sequence is likely to occlude the left arm of

Cell 142, 737–748, September 3, 2010 ª2010 Elsevier Inc. 743

A mgtA leader region B trp leader region Figure 7. Regulation of Transcription Elon-
gation into the mgtA and trp CRs by ORFs
Located within Their Respective LRs
(A) Regulation of transcription elongation by the
Stem-loop C mgtA leader RNA. Top: when proline levels are
limiting in the cytosol, a ribosome translating
Ribosome is stalled at Ribosome Ribosome mgtL stalls at proline codons, which favors forma-
STOP tion of stem loop C and results in transcription of
the mgtA CR. Middle: when proline levels are not
limiting in the cytosol, a ribosome can translate
the complete mgtL ORF, which favors formation
ON state ON state of stem loop B and hinders transcription elonga-
tion into the mgtA CR by an unknown mechanism.
Stem-loop B Bottom: a ribosome loads but cannot translate
a mutant mgtA leader with nucleotide substitutions
in the mgtL start codon, which favors formation of
stem loop C and results in transcription of the
Leader ORF is mgtA CR.
completely translated RBS RBS
(B) Regulation of transcription elongation by the trp
leader RNA. Top: when tryptophan levels are limiting
in the cytosol, a ribosome translating trpL stalls at
OFF OFF state
tryptophan codons, which favors formation of an
antiterminator structure and results in transcription
of the trp operon. Middle: when tryptophan levels
Stem-loop C
are not limiting in the cytosol, a ribosome can trans-
late the complete trpL ORF, which favors formation
of an intrinsic transcription terminator and there is no
Ribosome loads but transcription of the trp operon. Bottom: a ribosome
there is no mgtL or loads but cannot translate a mutant trp leader with
trpL translation RBS RBS UUUUUUU nucleotide substitutions in the trpL start codon,
which favors formation of both the anti-antitermina-
ON state OFF state tor and terminator structures and there is no tran-
scription of the trp operon. Note the different posi-
tion of the mgtL and trpL ORFs relative to the
sequences forming stem loop C and the anti-antiter-
minator structures, respectively, which determines
the different phenotypes resulting from mutation of
the mgtL and trpL start codons.

stem loop C (Figure 1 and Figure 7A). This would favor formation tive with a stop codon at position 107–109 no longer conferred
of stem loop B, which has been shown to hamper transcription high levels of b-galactosidase upon wild-type Salmonella when
elongation beyond the mgtA LR (Cromie et al., 2006). nucleotides 110–112 were deleted (Figure 4A), which brought
By contrast, conditions that reduce the levels of free cytosolic the mgtL stop codon only six nucleotides away from the left
proline would promote ribosome stalling at the mgtL proline arm of stem loop C. Cumulatively, these results indicate that
codons, thereby advancing formation of stem loop C and result- the distance between the ribosome translating mgtL and the
ing in transcription elongation into the mgtA CR (Figure 1 and nucleotides forming the left arm of stem loop C (as opposed to
Figure 7A). Therefore, the position that a translating ribosome the size of the translated mgtL product) is critical for mgtL-medi-
reaches in the mgtL ORF should determine whether transcription ated gene control.
continues into the mgtA CR (Figure 1 and Figure 7A). If the lack of translation of the full-length mgtL stimulates tran-
We tested our model by investigating the phenotype of wild- scription elongation beyond the mgtA LR by favoring formation of
type Salmonella harboring pYS1010 derivatives with stop stem loop C (Figure 1 and Figure 7A), hindering formation of stem
codons at different positions within mgtL. A derivative with loop C should abolish this stimulation. As predicted, the G120C
a stop codon at position 110–112, which is only six nucleotides and C145G substitutions in the mgtA leader (Figure 1), which
upstream of the left arm of stem loop C (Figure 1), produced were previously shown to impede formation of stem loop C
very low levels of b-galactosidase when grown in 10 mM Mg2+ (Cromie et al., 2006), thwarted derepression in constructs
(Figure 4A), like the isogenic strain with plasmid pYS1010 harboring stop codons at either of two mgtL positions (Figure 4A).
harboring the wild-type mgtA leader (Figure 4A). By contrast,
there were high levels of b-galactosidase in a pYS1010 derivative DISCUSSION
with a stop codon at position 107–109 (Figure 4A), which is nine
nucleotides from the left arm of stem loop C (Figure 1), similar to The LR of many mRNAs can respond to specific nutritional
the behavior of strains with stop codons at positions 80–82, and/or physical signals by modifying expression of the associ-
89–91, or 98–100 (Figure 4A). Furthermore, the pYS1010 deriva- ated downstream coding sequences. Some of these leader

744 Cell 142, 737–748, September 3, 2010 ª2010 Elsevier Inc.

RNAs rely on the same mechanism to detect more than one some is too close to the sequences that make up stem loop C,
signal whereas others utilize different means of controlling then the alternative stem loop B would form (Figure 1 and
gene expression in response to one signal. For example, two Figure 7A), and transcription would not continue into the mgtA
riboswitches located in tandem, one responding to S-adenosyl- CR. For instance, stop codons at positions 80–82, 89–91,
methionine and the other to coenzyme B12, control transcription 98–100, or 107–109 in the mgtA leader resulted in derepression
of the metE gene in Bacillus clausii (Sudarsan et al., 2006). In whereas a stop codon at position 110–112 did not (Figure 4A).
other cases, a cis-acting riboswitch can also function as Yet, the stop codon at position 107–109 no longer derepressed
a trans-acting regulatory RNA to modulate the expression of expression if nucleotides 110–112 were also deleted (Figure 4A),
genes located somewhere else in the genome (Loh et al., which moved the mgtL stop codon only six nucleotides away
2009). The mgtA leader described here constitutes a singular from the left arm of stem loop C (Figure 1).
example of an mRNA leader that utilizes different mechanisms It was recently reported that a chromosomal C98T mutation in
to sense different signals. the leader RNA causes constitutive high expression of the mgtA
We determined that the mRNA leader corresponding to the CR (O’Connor et al., 2009). In the original description of this
Mg2+ transporter gene mgtA harbors a translated ORF rich in mutation, it was not clear why this nucleotide substitution should
proline codons designated mgtL (Figure 1; Figure S1), which is alter mgtA expression, but we now see that it created a stop
conserved in other bacterial species (Figures 3A and 3B) and codon in mgtL that is identical to the mutation we had previously
enables Salmonella to regulate transcription elongation into the constructed, which resulted in constitutive high expression of
mgtA CR in response to the levels of cytoplasmic proline. mgtA (Figures 4A and 4B).
Whereas the mgtA leader RNA senses Mg2+ directly (Cromie The mgtL proline codons are located at positions 77–79,
et al., 2006), it monitors proline levels in an indirect fashion, via 83–85, 89–91, and 95–97 (Figure 1), which correspond to the
mgtL translation (Figure 4A), using an attenuation-like mecha- region where placing a stop codon results in derepression (Fig-
nism (Landick et al., 1996; Henkin and Yanofsky, 2002). ure 4A). Therefore, we propose that when cytosolic proline levels
The dual sensing ability of the mgtA leader allows Salmonella are low, there would be less proline-charged tRNAs, resulting in
to transcribe the mgtA CR not only when Mg2+ levels drop below ribosome stalling at mgtL proline codons. This would favor
a certain threshold (Cromie and Groisman, 2010; Cromie et al., formation of stem loop C and transcription elongation into the
2006) but also under conditions resulting in a decrease in cyto- mgtA CR (Figure 7A). In contrast, when cytosolic proline levels
solic proline levels (Figures 4E and 4F). are high, the ribosome would translate the complete mgtL
The discovery of a translatable ORF in the mgtA leader sequence, thereby occluding the left arm of stem loop C and,
(Figure 1; Figure S1) suggests a possible explanation for the in this manner, favor formation of stem-loop B (Figure 7A) and
recent observation that the mRNA levels corresponding to the hinder transcription of the mgtA CR. Interestingly, despite
50 -most 127 nucleotides of the mgtA leader are much higher utilizing different mechanisms to sense proline and Mg2+, the
than those for the 30 -most 94 nucleotides of the mgtA leader mgtA leader relies on formation of the same structure—stem
when Salmonella is grown in high-Mg2+ media (Spinelli et al., loop C—to promote transcription elongation into the mgtA CR,
2008). This is because the RNA-processing enzyme RNase E, because mutations that hinder formation of stem loop C pre-
which has been implicated in mgtA mRNA degradation during vented the derepression caused by stop codons in mgtL
growth in high Mg2+ (Spinelli et al., 2008), can target nascent (Figure 4A) and by low Mg2+ (Figure 4B) (Cromie et al., 2006).
transcripts as well as complete messages (Hammarlof and The regulation of mgtA by changes in the cytosolic proline
Hughes, 2008). Therefore, the ribosome translating mgtL, which concentration is reminiscent of class I transcription attenuation,
extends to position 124 in the mgtA leader (Figure 1), could where certain amino acid biosynthetic operons are regulated by
protect the 50 region of the mgtA leader from RNase E action, re- translation of a short ORF in the LR. These ORFs are rich in
sulting in higher levels of this portion of the RNA. codons specifying the amino acid(s) synthesized by the enzymes
encoded in the operon (Landick et al., 1996). Yet, there are signif-
mgtL Translation Regulates Transcription icant differences between the two systems. First, mgtA codes for
of the mgtA CR a transporter rather than for an amino acid biosynthetic enzyme.
We determined that interfering with mgtL translation increased This provides experimental support for the proposal, based on
the mRNA levels for the mgtA CR (Figures 4A and 4D–4F). This genomic analysis, that classical transcriptional attenuators
appears to result from the formation of a particular secondary might direct the expression of operons mediating functions other
structure (i.e., stem loop C) in the mgtA leader (Figure 1 and than nutrient biosynthesis (Merino and Yanofsky, 2005). Second,
Figure 7A) that promotes transcription elongation into the mgtA the position of the leader ORF relative to the potential RNA
CR, as nucleotide substitutions impeding formation of stem secondary structures that can be adopted differs between the
loop C (Cromie et al., 2006) prevented derepression provoked mgtA leader (Figure 7A) and the leaders of attenuation-regulated
by stop codon mutations in mgtL (Figure 4A). Because transcrip- amino acid biosynthetic operons such as the trp operon
tion and translation are coupled in bacteria and because the (Figure 7B). This provides a plausible explanation for the oppo-
ribosome occupies 30 nucleotides, covering 12–15 nucleo- site effects that mutation of the leader ORF start codon has in
tides from the P site (Laursen et al., 2005), the position that these two classes of leaders. It results in superattenuation for
a translating ribosome reaches in mgtL relative to the sequences the trp operon (Landick and Yanofsky, 1987) because there
that make up stem loop C would determine whether transcription would be no ribosomes to stall at Trp codons, which would result
continues into the mgtA CR (Figure 1 and Figure 7A). If the ribo- in formation of a terminator structure (Figure 7B). However,

Cell 142, 737–748, September 3, 2010 ª2010 Elsevier Inc. 745

it gives rise to derepression of the mgtA CR (our unpublished and ribosomes, which could be affected during hyperosmotic
results), because formation of stem loop C would be favored stress. On the other hand, CorA-mediated Mg2+ uptake may
when mgtL is not translated (Figure 7A). Third, an intrinsic be compromised because it is driven by the membrane potential
terminator structure forms in the trp leader when the levels of (Froschauer et al., 2004), which decreases in cells experiencing
charged tRNATrp are high and the ribosome does not pause at high osmolarity (Csonka, 1989). Yet, this stress would not affect
the consecutive Trp codons in the leader ORF (Landick and Mg2+ uptake by the P-type ATPase MgtA protein because it is
Yanofsky, 1987) (Figure 7B). This is in contrast to the absence energized by ATP hydrolysis (Maguire, 1992).
of an intrinsic transcription terminator in the mgtA leader, where That proline limitation promotes expression of the MgtA Mg2+
sequences located downstream of the stem loop B structure transporter suggests that there is a physiological connection
(Figure 1) are necessary to control transcription elongation into between proline and Mg2+. In agreement with this notion, tran-
the mgtA CR (Cromie et al., 2006) by an unknown mechanism. scription of the proline transporter gene proP is promoted during
And fourth, apart from utilizing a classical transcription attenua- growth in low Mg2+ by the PhoP protein (Eguchi et al., 2004),
tion-like mechanism to mediate the response to proline, the which directs transcription initiation of the mgtA gene (Garcia
mgtA leader senses Mg2+ directly functioning as a riboswitch Vescovi et al., 1996).
(Cromie et al., 2006). This raises the intriguing possibility of clas-
sical transcriptional attenuators also sensing physical and/or EXPERIMENTAL PROCEDURES
chemical signals directly.
Bacterial Strains, Plasmids, Primers, and Growth Conditions
2+ Bacterial strains and plasmids used in this study are listed in Table S1. Primers
High Osmolarity Enhances the mRNA Levels of the Mg
used in this study are listed in Table S2. Unless otherwise stated, bacteria were
Transporter Gene mgtA grown at 37 C in LB broth or in N-minimal medium (pH 7.4) (Snavely et al.,
Why does proline limitation promote transcription of the Mg2+ 1991) supplemented with 0.1% casamino acids, 38 mM glycerol, and the indi-
transporter gene mgtA? We contemplated the possibility of the cated concentration of MgCl2.
MgtA protein being a conduit for proline whereby cells experi-
encing low proline would derepress expression of a proline Construction of Strains with Chromosomal Mutations
import system. However, we found that [14C]proline uptake Deletion strains were constructed using the one-step inactivation method
(Datsenko and Wanner, 2000). See Extended Experimental Procedures for
was similar between wild-type and mgtA Salmonella, and
detailed protocols.
between a triple mutant lacking all known proline uptake
systems (ProP, ProU, and PutP) (Csonka and Leisinger, 2007) Treatment with Bacteriostatic Protein Synthesis Inhibitors
and the isogenic mgtA derivative (our unpublished results). We An overnight bacterial culture grown in N-minimal medium with 10 mM MgCl2
also considered that MgtA-mediated Mg2+ uptake might be was used to inoculate 125 ml flasks containing 10 ml of the same medium
necessary for proline biosynthesis, because the activity of the (1:50 dilution) and grown for 3 hr at 37 C with shaking. To promote PhoP-
proline biosynthetic enzyme glutamate 5-kinase is Mg2+ depen- dependent mgtA transcription initiation, bacteria were washed and resus-
dent (Perez-Arellano et al., 2005). Yet, inactivation of the mgtA pended in 10 ml of N-minimal medium with 500 mM MgCl2. Following growth
for 1 hr, a 500 ml aliquot was removed to determine the mRNA levels before
gene did not render Salmonella auxotrophic for proline (our
treatment. Subsequently, tetracycline (to 25 mg/ml final concentration) or
unpublished results). chloramphenicol (to 200 mg/ml final concentration) was added and bacteria
We determined that hyperosmotic shock promoted an were grown for 15 min, when they were harvested and their RNA was isolated
increase in the mRNA levels corresponding to the mgtA CR for analysis.
(Figure 6A). This appears to result from a transient decrease in
the levels of charged-proline tRNAs taking place when proline Effect of Proline Limitation
is used as an osmoprotectant, as there would be less cytosolic Bacteria were grown overnight in modified N-minimal medium containing
0.2% glucose, 10 mM MgCl2, and 1 mM proline. The overnight culture was
proline available to charge the tRNAPros. Indeed, the increase
used to inoculate 125 ml flasks containing 10 ml of the same medium (1:50 dilu-
in mgtA mRNA levels was severely compromised if the osmotic tion) and grown for 3 hr at 37 C with shaking. The harvested bacteria were
shock was alleviated by the osmoprotectant glycine betaine, washed in the modified N-minimal medium containing 500 mM MgCl2, and
which did not influence mgtA expression when added in the grown in 10 ml of the same medium with 1 mM proline for 1 hr. After removing
absence of NaCl (Figure 6A). The osmotic shock-promoted a 250 ml aliquot to determine the mRNA levels before treatment, the harvested
increase in the mRNA levels for the mgtA CR requires the mgtL bacteria were washed with the modified N-minimal medium containing 500 mM
MgCl2, and suspended in 100 ml of the same medium. To see the effect of
proline codons, because a mutant with substitutions in all four
proline limitation, the suspended bacterial cells were split into two flasks con-
proline codons in mgtL lost the ability to derepress the mgtA taining the modified N-minimal media with 500 mM MgCl2 and a mixture of
CR mRNA in response to hyperosmotic shock (Figure 6B). each of the 19 essential amino acids (50 mM final concentration for each of
Why does hyperosmotic shock induce transcription of the them): one flask contained 1 mM proline and the other had no proline added.
Mg2+ transporter gene mgtA when Salmonella also harbors Bacteria were grown for 15 min, when they were harvested and their RNA was
the constitutively expressed Mg2+ transporter CorA? On the isolated for analysis.
one hand, high osmolarity promotes excretion of putrescine
(Schiller et al., 2000), which constitutes the major organic diva- Testing Effect of Growth in High versus Low Mg2+
Bacteria were grown overnight in modified N-minimal medium containing
lent cation in bacterial cells and is normally bound to nucleic
0.2% glucose, 10 mM MgCl2, and 1 mM proline. The harvested bacteria
acids (Wortham et al., 2007). This might create an increased were washed in the modified N-minimal medium containing 500 mM MgCl2,
need for Mg2+ in order to neutralize the negatively charged and this culture was used to inoculate 125 ml flasks containing 10 ml of the
DNA and RNA, and to stabilize structures such as membranes modified N-minimal medium containing either 500 mM MgCl2 or 10 mM

746 Cell 142, 737–748, September 3, 2010 ª2010 Elsevier Inc.

MgCl2 (1:50 dilution) and grown for 3.5 hr at 37 C with shaking. The bacteria REFERENCES
were then harvested and their RNA was isolated for analysis.
Bader, M.W., Sanowar, S., Daley, M.E., Schneider, A.R., Cho, U., Xu, W.,
Klevit, R.E., Le Moual, H., and Miller, S.I. (2005). Recognition of antimicrobial
Testing Effect of Mg and/or Proline Limitation Treatment peptides by a bacterial sensor kinase. Cell 122, 461–472.
Bacteria were grown overnight in modified N-minimal medium containing
Barchiesi, J., Castelli, M.E., Soncini, F.C., and Vescovi, E.G. (2008). mgtA
0.2% glucose, 10 mM MgCl2, and 1 mM proline. The overnight culture
expression is induced by rob overexpression and mediates a Salmonella en-
was used to inoculate 125 ml flasks containing 20 ml of the same medium
terica resistance phenotype. J. Bacteriol. 190, 4951–4958.
(1:50 dilution) and grown for 3 hr at 37 C with shaking. The harvested bacteria
were washed in the modified N-minimal medium containing 500 mM MgCl2, Blanc-Potard, A.B., and Groisman, E.A. (1997). The Salmonella selC locus
and grown in 20 ml of the same medium with 1 mM proline for 1 hr. The contains a pathogenicity island mediating intramacrophage survival. EMBO
harvested bacteria were washed with the modified N-minimal medium J. 16, 5376–5385.
containing 500 mM MgCl2, and suspended in 200 ml of the same medium. Brennan, R.G., and Link, T.M. (2007). Hfq structure, function and ligand
This cell suspension was used to inoculate 125 ml flasks containing 5 ml of binding. Curr. Opin. Microbiol. 10, 125–133.
media containing either 500 mM MgCl2 and 1 mM proline, 500 mM MgCl2
Cayley, S., Lewis, B.A., and Record, M.T., Jr. (1992). Origins of the osmopro-
and no proline, no MgCl2 and 1 mM proline, or no MgCl2 and no proline
tective properties of betaine and proline in Escherichia coli K-12. J. Bacteriol.
(1:100 dilution). Bacteria were grown for 15 min, when they were harvested
174, 1586–1595.
and their RNA was isolated for analysis.
Chen, J.W., Bennett, D.C., and Umbarger, H.E. (1991). Specificity of attenua-
tion control in the ilvGMEDA operon of Escherichia coli K-12. J. Bacteriol. 173,
Testing Effect of Hyperosmotic Shock 2328–2340.
Bacteria were grown overnight in modified N-minimal medium containing Choi, E., Groisman, E.A., and Shin, D. (2009). Activated by different signals, the
0.2% glucose, 0.1% casamino acids, and 10 mM MgCl2. The overnight culture PhoP/PhoQ two-component system differentially regulates metal uptake.
was used to inoculate 125 ml flasks containing 20 ml of the same medium J. Bacteriol. 191, 7174–7181.
(1:50 dilution) and grown for 3.5 hr at 37 C with shaking. The harvested
Cromie, M.J., and Groisman, E.A. (2010). Promoter and riboswitch control of
bacteria were washed in modified N-minimal medium containing 500 mM
the Mg2+ transporter MgtA from Salmonella enterica. J. Bacteriol. 192,
MgCl2 without casamino acids and suspended in 250 ml of the same medium.
This cell suspension was used to inoculate 125 ml flasks containing 5 ml of the
modified N-minimal medium without casamino acids containing 500 mM MgCl2 Cromie, M.J., Shi, Y., Latifi, T., and Groisman, E.A. (2006). An RNA sensor for
and either no additional supplements, 1 mM glycine betaine, 0.3 M NaCl, 0.3 M intracellular Mg2+. Cell 125, 71–84.
NaCl and 1 mM glycine betaine, or 0.3 M NaCl and 1 mM proline. Bacteria were Csonka, L.N. (1989). Physiological and genetic responses of bacteria to
grown for 1 hr at 37 C with shaking, when they were harvested and their RNA osmotic stress. Microbiol. Rev. 53, 121–147.
was isolated for analysis. Csonka, L.N., and Leisinger, T. (2007). Biosynthesis of proline. In EcoSal—Es-
cherichia coli and Salmonella: Cellular and Molecular Biology, G. Cohen, ed.
(Washington, DC: ASM Press),
RNA Isolation and Determination of Transcript Levels
Total RNA was extracted using an RNeasy Mini Kit (QIAGEN). cDNA was Datsenko, K.A., and Wanner, B.L. (2000). One-step inactivation of chromo-
synthesized using TaqMan reverse-transcription reagents (Applied Biosys- somal genes in Escherichia coli K-12 using PCR products. Proc. Natl. Acad.
tems) following the manufacturer’s instructions. Quantification of transcripts Sci. USA 97, 6640–6645.
was performed by real-time PCR using Fast SYBR Green Master Mix (Applied Eguchi, Y., Okada, T., Minagawa, S., Oshima, T., Mori, H., Yamamoto, K., Ish-
Biosystems) in a 7500 Sequence Detection System (Applied Biosystems). A list ihama, A., and Utsumi, R. (2004). Signal transduction cascade between EvgA/
of primers used for quantitative RT-PCR is presented in Extended Experi- EvgS and PhoP/PhoQ two-component systems of Escherichia coli. J. Bacter-
mental Procedures. See Extended Experimental Procedures for a detailed iol. 186, 3006–3014.
Froschauer, E.M., Kolisek, M., Dieterich, F., Schweigel, M., and Schweyen,
R.J. (2004). Fluorescence measurements of free [Mg2+] by use of mag-fura 2
b-Galactosidase Assays in Salmonella enterica. FEMS Microbiol. Lett. 237, 49–55.
b-galactosidase activity was determined as described (Cromie et al., 2006). Garcia Vescovi, E., Soncini, F.C., and Groisman, E.A. (1996). Mg2+ as an extra-
cellular signal: environmental regulation of Salmonella virulence. Cell 84,
Grundy, F.J., and Henkin, T.M. (2006). From ribosome to riboswitch: control of
gene expression in bacteria by RNA structural rearrangements. Crit. Rev. Bio-
Supplemental Information includes Extended Experimental Procedures, two
chem. Mol. Biol. 41, 329–338.
figures, and two tables and can be found with this article online at doi:10.
1016/j.cell.2010.07.046. Hammarlof, D.L., and Hughes, D. (2008). Mutants of the RNA-processing
enzyme RNase E reverse the extreme slow-growth phenotype caused by
a mutant translation factor EF-Tu. Mol. Microbiol. 70, 1194–1209.
ACKNOWLEDGMENTS Henkin, T.M. (2008). Riboswitch RNAs: using RNA to sense cellular metabo-
lism. Genes Dev. 22, 3383–3390.
We thank Laszlo Csonka, Kerry Hollands, Robert Landick, and Charles
Henkin, T.M., and Yanofsky, C. (2002). Regulation by transcription attenuation
Yanofsky for comments on the manuscript; Henry Huang for the plasmid
in bacteria: how RNA provides instructions for transcription termination/anti-
harboring supF; and John Roth for strains. This work was supported, in part,
termination decisions. Bioessays 24, 700–707.
by grant AI49561 from the NIH to E.A.G., who is an Investigator of the Howard
Hughes Medical Institute. Landick, R., and Yanofsky, C. (1987). Transcription attenuation. In Escherichia
coli and Salmonella Cellular and Molecular Biology, J.L. Ingraham, ed. (Wash-
Received: March 19, 2010 ington, DC: ASM Press).
Revised: May 24, 2010 Landick, R., Turnbough, C.L., and Yanofsky, C. (1996). Transcription attenua-
Accepted: July 14, 2010 tion. In Escherichia coli and Salmonella Cellular and Molecular Biology, F.C.
Published: September 2, 2010 Neidhardt, ed. (Washington, DC: ASM Press).

Cell 142, 737–748, September 3, 2010 ª2010 Elsevier Inc. 747

Laursen, B.S., Sorensen, H.P., Mortensen, K.K., and Sperling-Petersen, H.U. Shin, D., Lee, E.J., Huang, H., and Groisman, E.A. (2006). A positive feedback
(2005). Initiation of protein synthesis in bacteria. Microbiol. Mol. Biol. Rev. 69, loop promotes transcription surge that jump-starts Salmonella virulence
101–123. circuit. Science 314, 1607–1609.
Lejona, S., Aguirre, A., Cabeza, M.L., Garcia Vescovi, E., and Soncini, F.C. Snavely, M.D., Miller, C.G., and Maguire, M.E. (1991). The mgtB Mg2+ trans-
(2003). Molecular characterization of the Mg2+-responsive PhoP-PhoQ regu- port locus of Salmonella typhimurium encodes a P-type ATPase. J. Biol.
lon in Salmonella enterica. J. Bacteriol. 185, 6287–6294. Chem. 266, 815–823.
Loh, E., Dussurget, O., Gripenland, J., Vaitkevicius, K., Tiensuu, T., Mandin, P.,
Soncini, F.C., Garcia Vescovi, E., Solomon, F., and Groisman, E.A. (1996).
Repoila, F., Buchrieser, C., Cossart, P., and Johansson, J. (2009). A trans-
Molecular basis of the magnesium deprivation response in Salmonella typhi-
acting riboswitch controls expression of the virulence regulator PrfA in Listeria
murium: identification of PhoP-regulated genes. J. Bacteriol. 178, 5092–5099.
monocytogenes. Cell 139, 770–779.
Maguire, M.E. (1992). MgtA and MgtB: prokaryotic P-type ATPases that Spinelli, S.V., Pontel, L.B., Garcia Vescovi, E., and Soncini, F.C. (2008). Regu-
mediate Mg2+ influx. J. Bioenerg. Biomembr. 24, 319–328. lation of magnesium homeostasis in Salmonella: Mg2+ targets the mgtA tran-
script for degradation by RNase E. FEMS Microbiol. Lett. 280, 226–234.
Maguire, M.E. (2006). Magnesium transporters: properties, regulation and
structure. Front. Biosci. 11, 3149–3163. Sudarsan, N., Hammond, M.C., Block, K.F., Welz, R., Barrick, J.E., Roth, A.,
Merino, E., and Yanofsky, C. (2005). Transcription attenuation: a highly and Breaker, R.R. (2006). Tandem riboswitch architectures exhibit complex
conserved regulatory strategy used by bacteria. Trends Genet. 21, 260–264. gene control functions. Science 314, 300–304.
O’Connor, K., Fletcher, S.A., and Csonka, L.N. (2009). Increased expression of Turnbough, C.L., Jr., and Switzer, R.L. (2008). Regulation of pyrimidine biosyn-
Mg2+ transport proteins enhances the survival of Salmonella enterica at high thetic gene expression in bacteria: repression without repressors. Microbiol.
temperature. Proc. Natl. Acad. Sci. USA 106, 17522–17527. Mol. Biol. Rev. 72, 266–300.
Perez-Arellano, I., Rubio, V., and Cervera, J. (2005). Dissection of Escherichia
Winkler, W.C., and Breaker, R.R. (2005). Regulation of bacterial gene expres-
coli glutamate 5-kinase: functional impact of the deletion of the PUA domain.
sion by riboswitches. Annu. Rev. Microbiol. 59, 487–517.
FEBS Lett. 579, 6903–6908.
Prost, L.R., Daley, M.E., Le Sage, V., Bader, M.W., Le Moual, H., Klevit, R.E., Wortham, B.W., Patel, C.N., and Oliveira, M.A. (2007). Polyamines in bacteria:
and Miller, S.I. (2007). Activation of the bacterial sensor kinase PhoQ by acidic pleiotropic effects yet specific mechanisms. Adv. Exp. Med. Biol. 603,
pH. Mol. Cell 26, 165–174. 106–115.

Schiller, D., Kruse, D., Kneifel, H., Kramer, R., and Burkovski, A. (2000). Poly- Yanofsky, C., and Horn, V. (1994). Role of regulatory features of the trp operon
amine transport and role of potE in response to osmotic stress in Escherichia of Escherichia coli in mediating a response to a nutritional shift. J. Bacteriol.
coli. J. Bacteriol. 182, 6247–6249. 176, 6245–6254.

748 Cell 142, 737–748, September 3, 2010 ª2010 Elsevier Inc.

Structural Basis of Semaphorin-Plexin
Recognition and Viral Mimicry from
Sema7A and A39R Complexes with PlexinC1
Heli Liu,1 Z. Sean Juo,2,3,4 Ann Hye-Ryong Shim,1 Pamela J. Focia,1 Xiaoyan Chen,1 K. Christopher Garcia,2,3,4
and Xiaolin He1,*
1Department of Molecular Pharmacology and Biological Chemistry, Northwestern University Feinberg School of Medicine, Searle 8-417,

303 East Chicago Avenue, Chicago, IL 60611, USA

2Howard Hughes Medical Institute
3Department of Molecular and Cellular Physiology
4Department of Structural Biology

Stanford University School of Medicine, Beckman B171B, 279 Campus Drive, Stanford, CA 94305, USA
DOI 10.1016/j.cell.2010.07.040

SUMMARY organogenesis, and immune responses (Kruger et al., 2005;

Suzuki et al., 2008). Semaphorins are also associated with tumor
Repulsive signaling by Semaphorins and Plexins is progression and other diseases (reviewed in Kruger et al., 2005;
crucial for the development and homeostasis of the Tamagnone and Comoglio, 2000). Within the neural system,
nervous, immune, and cardiovascular systems. Semaphorin-Plexin signaling is implicated well beyond axon
Sema7A acts as both an immune and a neural Sem- guidance, ranging from axon pruning to synaptic formation,
aphorin through PlexinC1, and A39R is a Sema7A specificity, and plasticity (reviewed in He et al., 2002; Waimey
and Cheng, 2006). While Plexins are the predominant receptors
mimic secreted by smallpox virus. We report the
for Semaphorins, the alternative Semaphorin receptors, Neuro-
structures of Sema7A and A39R complexed with pilin-1 and -2, differ from Plexins in structure and function
the Semaphorin-binding module of PlexinC1. Both and appear to serve as obligate coreceptors for either Sema-
structures show two PlexinC1 molecules symmetri- phorin signaling or VEGF signaling in a variety of specific cellular
cally bridged by Semaphorin dimers, in which the (neuronal, vascular, etc.) contexts (Pellet-Many et al., 2008;
Semaphorin and PlexinC1 b propellers interact in Uniewicz and Fernig, 2008).
an edge-on, orthogonal orientation. Both binding Semaphorins have been grouped into eight classes on the
interfaces are dominated by the insertion of the Sem- basis of primary sequence and source, where classes 3–7 are
aphorin’s 4c-4d loop into a deep groove in blade 3 of vertebrate Semaphorins and class V are viral Semaphorins
the PlexinC1 propeller. A39R appears to achieve (Semaphorin Nomenclature Committee, 1999). The Semaphor-
Sema7A mimicry by preserving key Plexin-binding ins are characterized by an N-terminal 500 amino acid (aa)
Sema domain that is essential for signaling through Plexins
determinants seen in the mammalian Sema7A
(Koppel et al., 1997). The crystal structures of the Sema domains
complex that have evolved to achieve higher affinity of Sema3A and Sema4D (also known as CD100) have been
binding to the host-derived PlexinC1. The complex determined, showing that the Sema domain is a seven-bladed
structures support a conserved Semaphorin-Plexin b propeller, and appear to exist as a homodimer by virtue of
recognition mode and suggest that Plexins are acti- conserved structural elements (Antipenko et al., 2003; Love
vated by dimerization. et al., 2003). There are four subfamilies of vertebrate Plexins
(A, B, C, and D), which are all type I transmembrane glycopro-
INTRODUCTION teins featuring an extracellular segment containing an N-terminal
Sema domain, followed by variable numbers of PSI (‘‘found in
Semaphorins and their receptors, Plexins, are two families of Plexins, Semaphorins and Integrins’’) domains, immunoglob-
widely expressed proteins whose respective structures and ulin-like (Ig) domains (Bork et al., 1999), and an intracellular
functions are conserved across the animal kingdom. Originally GTPase-activating (GAP) domain that regulates Rho family
identified as ligand-receptor pairs that control axon guidance GTPases (Oinuma et al., 2004). The structures of the intracellular
by repulsion during central nervous system (CNS) development domains of two Plexins have been solved (He et al., 2009; Tong
(Kolodkin et al., 1993; Luo et al., 1993; Tamagnone et al., et al., 2009), but three-dimensional structural information does
1999; Tessier-Lavigne and Goodman, 1996), Semaphorins and not exist for the Plexin extracellular segment or for the Sema-
Plexins have been shown to serve as path-finding controls for phorin-Plexin interaction.
a diverse array of additional functions in physiology, including In the immune system, a subset of Semaphorins are active
vascularization and angiogenesis (normal and pathological), and are designated as ‘‘immune Semaphorins.’’ These include

Cell 142, 749–761, September 3, 2010 ª2010 Elsevier Inc. 749

Sema4D, Sema4A, and Sema7A, which appear to modulate (Figures 1B–1E). In HEPES-buffered saline (HBS), Sema7A
a variety of immune responses ranging from thymic selection bound PlexinC1-SemaPSI with a KD of 290 nM (Figure 1B).
to B cell homing (Suzuki et al., 2008). Sema7A is a GPI-anchored The ITC measurements for Sema7A/PlexinC1 were complicated
cell-surface glycoprotein expressed on activated lymphocytes by the small enthalpy change ( 2 kcal/mol), which resulted in
and thymocytes (Xu et al., 1998; Yamada et al., 1999), and its shallow titration curves with poor signal-to-noise ratio. There-
receptor is PlexinC1 (also named VESPR or CD232) (Tamagnone fore, we used high concentrations of the samples and injected
et al., 1999). The Sema7A/PlexinC1 interaction was originally large amounts of protein so as to maximize the heat per injection,
shown to induce the activation of monocytes (Holmes et al., and also to fully saturate the titration curve. The Sema7A/Plex-
2002), but recently this receptor-ligand pair has been shown to inC1 KD measured here by calorimetry using soluble proteins,
regulate melanocyte adhesion, and silencing of PlexinC1 is albeit in the nanomolar range, is weaker than the reported KD
seen during the development and progression of melanoma of 2.1nM from Scatchard analysis using intact receptors in the
(Scott et al., 2009). Sema7A also has been suggested to have cell membrane (Tamagnone et al., 1999). Previously reported
neuronal functions, by promoting axon growth through integrins Semaphorin/Plexin affinity measurements are primarily cell
(Pasterkamp et al., 2003). Underscoring the importance of Sem- based and generally in the low nanomolar KD range. We attribute
aphorin/Plexin interactions in the immune system, viruses have the lower-affinity values of the soluble recombinant proteins to
evolved proteins that engage the immune Semaphorin/Plexin the lack of membrane confinement in the ITC format, not to
system, presumably to enhance virus survival in the host. The structural differences in the recombinant proteins. Similar differ-
most well-characterized examples derive from Vaccinia ences between solution-based and cell-based affinities have
(smallpox) virus and Alcelaphine herpesvirus, which encode been seen for ligand binding to other receptors, such as SCF/
secreted Semaphorin homologs A39R and AHVsema (Comeau KIT (55 nM versus 2 nM) (Lemmon et al., 1997; Lev et al.,
et al., 1998; Kolodkin et al., 1993). These viral Semaphorins 1992), CNP/NPRB (2.7 nM versus 0.03 nM) (He et al., 2006; Kol-
show sequence similarity to Sema7A and can also bind to and ler and Goeddel, 1992), and MHC/TCR (micromolar versus nano-
activate PlexinC1 (Comeau et al., 1998). The viral Semaphorins molar) (Huppa et al., 2010; Krogsgaard et al., 2003). We also
probably play an immunomodulatory role by mimicking Sema7A; determined that the affinity of Sema7A for the truncated Plex-
an antibody against PlexinC1 inhibits A39R-induced induction of inC1-SemaPSI and full-length PlexinC1 ECD were nearly iden-
inflammatory cytokines by monocytes (Comeau et al., 1998). tical (Figures 1B and 1C), indicating the Ig and PSI domains C-
We present here the crystal structures of both the Sema7A/ terminal to the PlexinC1-SemaPSI module are not significantly
PlexinC1 complex and the A39R/PlexinC1 complex, revealing involved in Semaphorin recognition. Thus, for structural studies
the basic architecture of a Semaphorin/Plexin recognition com- we proceeded with crystallization of the Sema7A/PlexinC1-
plex. Our data point to a conserved mode of recognition of SemaPSI complex.
Plexins by Semaphorins and provide insights into Plexin activa- The viral-derived A39R bound PlexinC1-SemaPSI with a KD of
tion and viral mimicry. 9.4 nM (Figure 1D). The large enthalpy change ( 14.6 kcal/mol)
facilitated the obtainment of high-quality binding curves. We also
RESULTS characterized A39R for the effect of various PlexinC1 ECD trun-
cations on affinity. A39R affinity and thermodynamic parameters
Biochemical Analysis of PlexinC1 Binding for full-length PlexinC1-ECD (KD 8.9 nM) were nearly identical
by the Mammalian Semaphorin Sema7A to that of A39R binding to PlexinC1-SemaPSI (Figures 1D and
and the Viral Semaphorin A39R 1E). Thus, similar to Sema7A/PlexinC1 binding, this indicated
The extracellular domains of Semaphorins and Plexins are large that the N-terminal Sema-PSI domains of PlexinC1 were suffi-
and highly glycosylated; thus we expressed recombinant forms cient for Semaphorin binding and represented an appropriate
of PlexinC1, Sema7A and A39R using baculovirus-mediated form for cocrystallization of both the Sema7A/PlexinC1 and
mammalian cell gene transduction (BacMam) (Dukkipati et al., A39R/PlexinC1 complexes.
2008) and purified the secreted proteins from HEK293H cells
or N-acetylglucosaminyltransferase I-deficient HEK293 (GnTI- Structures of the Sema7A/PlexinC1 Complex, Free
HEK293) cells (Reeves et al., 2002). Our construct designs A39R, and the A39R/PlexinC1 Complex
were informed by prior structural analysis of Sema4D (Love To obtain diffraction quality crystals of the heavily glycosylated
et al., 2003), which showed that the first membrane-distal PSI Sema7A/PlexinC1-SemaPSI complex (four N-linked glycosy-
domain is integrally associated with the Sema domain. Deletion lated sites on Sema7A and seven on PlexinC1-SemaPSI), we
of the PSI domains in either PlexinC1 or Sema7A destabilized the used Endo-H to trim the N-linked glycans attached to the GnTI
proteins. Ultimately, after testing a variety of constructs, for HEK293-expressed Sema7A and PlexinC1-SemaPSI proteins.
PlexinC1, we expressed the Sema domain plus the first PSI This treatment leaves a single N-acetylglucosamine (GlcNAc)
domain (SemaPSI), and the entire extracellular segment (ECD). residue remaining at each occupied N-linked glycan site,
For Sema7A, we expressed the entire extracellular segment preserving the core glycosylation of the proteins. The glycan-
excluding the GPI anchor; for A39R, we expressed the full-length trimmed proteins exhibited the same solution behavior to wild-
protein, which is a ‘‘minimal’’ secreted Semaphorin that does not type proteins, e.g., the Semaphorins existed exclusively as
contain a PSI domain (Figure 1A). dimers after glycan-trimming, as assessed by gel filtration chro-
To analyze the solution binding of PlexinC1 to its A39R and matography (Figure S1 available online). The Sema7A/PlexinC1-
Sema7A ligands, we used isothermal titration calorimetry (ITC) SemaPSI complex structure was determined to a resolution of

750 Cell 142, 749–761, September 3, 2010 ª2010 Elsevier Inc.

Figure 1. Characterization of the Binding of Sema7A and A39R to PlexinC1
(A) Domain diagram of PlexinC1, Sema7A, and A39R and the constructs used in characterization. The cartoon key for the diagrams is shown to the right.
(B–E) Profiles of the binding of PlexinC1 to Sema7A and A39R as measured by isothermal titration calorimetry. The binding parameters are indicated below the
See also Figure S1 for gel filtration profiles showing that Sema7A and A39R exist as dimers.

2.4 Å via the method of single isomorphous replacement with propellers are 50 Å for Sema7A-Sema7A, 55 Å for Sema7A-Plex-
anomalous scattering (SIRAS) (Table S1, Figure S2). The asym- inC1, and 80 Å for PlexinC1-PlexinC1.
metric unit of the crystal contains one dimeric complex in a 2:2 Interestingly, the Sema domains of Sema7A and PlexinC1
stoichiometry, comprised of a central Sema7A dimer and two interact ‘‘edge-on’’ using their sides to contact one another,
monomeric PlexinC1-SemaPSI molecules (Figure 2A). rather than the top or bottom faces, as was generally expected.
Overall, the shape of the complex resembles a crab where The planes of the Sema7A and PlexinC1 b propellers are orthog-
Sema7A comprises the body, and the two PlexinC1 molecules onally related. The orientation shown in Figure 2 places the
resemble pincers emanating radially from the body at four and C-termini of Sema7A close to the cell membrane, leading to
eight o’clock. The approximate dimensions of the complex are the GPI anchor. The C terminus of the PSI domain of PlexinC1,
80 Å 3 140 Å 3 160 Å (Figure 2A). The head-to-head docking which is 160 Å from the Semaphorin C terminus, leads to the
architecture of the complex is in an ideal orientation for interac- opposing cell membrane, albeit through one additional PSI and
tion of two cell surface-associated proteins across a cell-cell four Ig domains not present in the current structure. Because
junction. In the complex, both Sema7A and PlexinC1 contain the remaining PSI and Ig domains of PlexinC1 would appear to
large, disk-shaped Sema domains composed of seven-bladed emanate away from the central dyad axis of the complex, they
b propellers, each of which is intimately associated with its are unlikely to contact Sema7A.
respective PSI domain and, in the case of Sema7A, a single Ig We expressed A39R in baculovirus and determined the
domain that would lead to the membrane. In the complex, the unliganded A39R structure, to a resolution of 2.0 Å, by SIRAS
Sema domains mediate the vast majority of the receptor-ligand (Table S1). The structure of A39R shows a dimer of minimal
contact. The center-to-center distances between the Sema Sema domains with no PSI or Ig domains (Figure 2D). The

Cell 142, 749–761, September 3, 2010 ª2010 Elsevier Inc. 751

Figure 2. Structures of the Sema7A/PlexinC1-SemaPSI and A39R/PlexinC1-SemaPSI Complexes
(A) Ribbon models of the Sema7A/PlexinC1-SemaPSI complex in front view (left) and side view (right), with the Sema7A protomers colored in cyan and blue, and
the PlexinC1-SemaPSI protomers in pink and magenta. The N-linked glycans are depicted as sticks with carbon atoms colored in green. A cartoon of a membrane
is drawn above and below the complex to indicate where the respective proteins would be attached to the cell surfaces.
(B) The structure of an individual PlexinC1-SemaPSI molecule from the complex in two orthogonal views, with each of the seven b -propeller blades, the extrusion,
the flap, and the PSI domain individually colored.
(C) Ribbon model of the A39R/PlexinC1-Sema-PSI complex in front view (left) and side view (right), with the A39R protomers colored in yellow and wheat and the
other components colored similarly to (A).
(D) Ribbon model of an A39R protomer from the free A39R dimer, with the structural modules colored in the same format as PlexinC1 shown in (B).
See also Table S1 and Figure S2 for crystallographic statistics and structural comparisons.

Sema domains of A39R and Sema7A are clearly built on the domains mediate all the A39R-PlexinC1 contact. The orienta-
same scaffold, with a root-mean-square deviation (rmsd) of tions of the Sema propellers in the complex are nearly identical
1.5 Å for matching Ca atoms (Figure S3). Major structural differ- to those in Sema7A/PlexinC1, as is the ‘‘edge-on’’ interaction
ences are located only at the N-terminal segment and several mode between the ligand and receptor b propellers.
loops remote from the Plexin-binding site of Sema7A. Sema7A,
like Sema3A and Sema4D, has a long N-terminal segment that Dimerization of Sema7A and A39R Are Mediated by
provides an additional, fifth b strand on the outer edge of blade Conserved Structural Elements with Varied Chemistry
6, whereas A39R lacks this segment (Figure S3). Dimerization appears to be a general and probably important
We followed the same approach as Sema7A/PlexinC1- property of Semaphorins (Antipenko et al., 2003; Love et al.,
SemaPSI to obtain the crystals of the A39R/PlexinC1-SemaPSI 2003) that we can now examine within the context of a receptor
complex, which we solved by molecular replacement (Table complex. The Sema7A molecules in the complex, consisting of
S1). The architecture of the A39R/PlexinC1 complex is similar Sema, PSI, and Ig domains, form a dimer roughly similar to the
to that of Sema7A/PlexinC1, except that as a result of the lack Sema4D dimer (Love et al., 2003) and the Sema3A dimers that
of PSI and Ig domains in A39R, it is shorter in its longest dimen- only contain the Sema domains (Antipenko et al., 2003)
sion (Figure 2C). As in the Sema7A/PlexinC1 complex, the Sema (Figure S2C). This is consistent with our gel filtration analysis

752 Cell 142, 749–761, September 3, 2010 ª2010 Elsevier Inc.

(Figure S1). The Sema7A dimer interface buries 2860 Å2 of Plexins is not a central part of the interface. It appears, then,
solvent-accessible surface area. Two sets of three loops on that Plexins have lost the prominence of this structural element
the top face of the Sema domain, 4b-4c, 4d-5a, and 5b-5c are as the relative functions of the Sema domain in Semaphorins
intertwined at the dimer interface. In comparison, both Sema4D versus Plexins specialized over the course of evolution
and Sema3A use an additional loop for dimerization (Figure S2E) (Antipenko et al., 2003; Love et al., 2003; Stamos
(Figure S2C). The Sema7A Sema-Sema interface is primarily et al., 2004).
hydrophobic. The Ig domain of Sema7A is also involved in the The PSI domain of PlexinC1 is a small cysteine-rich domain
dimer interface but only contributes 15% to the total buried similar to that of Semaphorins and integrins, but its orientation
surface area. The Ig domain of Sema7A, unlike the Sema4D Ig relative to the Sema domain is different from that in Semaphorins
domain, which is canonical, is an unusual variation of the Ig by an 20 rotation (Figure S2E). The PSI domain is intimately
fold, with a 5-on-2 topology in which strand D switches from packed against the Sema domain by a broad array of primarily
the BED b sheet to join the AFGC b sheet (Figure S2D). hydrophobic interactions, which would facilitate the sensitive
The viral Semaphorin A39R also exists as a dimer both in solu- structural transmission of Sema binding from the membrane-
tion and in the crystal (Figure S1). Dimerization of A39R is exclu- distal to the membrane-proximal regions of PlexinC1.
sively mediated by its Sema domain, since this protein does not
have PSI or Ig domains, and the dimer interface is considerably The Sema7A-PlexinC1 Interaction Features
smaller (1990 Å2) than that seen on the canonical Semaphorins a ‘‘Loop-in-Groove’’ Recognition Mode
(Figure S2C). A39R uses the same set of loops as the Sema7A The edge-on, orthogonal stacking of the respective b propellers
Sema for dimerization, but in contrast to the hydrophobic in the Sema7A-PlexinC1 interface (Figure 3A) buries a total of
Sema7A dimer interface, the A39R dimer interface is composed 2100 Å2 solvent-accessible surface area. The extensive inter-
almost exclusively of hydrophilic interactions involving six salt face can be divided into three principal regions: (1) The 4c-4d
bridges. loop of Sema7A inserting into the groove on PlexinC1 that is
The Sema7A and A39R structures, together with the previous bounded by walls composed of the PlexinC1 loop 3b-3c and
Sema3A and Sema4D structures, indicate that although the the bulged strand 3d—this PlexinC1 groove has an open and
dimer interfaces of Semaphorins can vary significantly in a closed (obstructed) end (Figure 3B); (2) the extrusion helix 2
sequence and chemical nature, the general structural mode of of Sema7A contacting the loop 3b-3c of PlexinC1 at one side
domain dimerization is conserved, and this is likely to be impor- of the groove; and (3) a small area of contact between the
tant for the appropriate dimerization geometry of the bound Sema7A blade 3, and the PlexinC1 strand 3d, that forms the
Plexin receptors for signaling. The preservation of the dimeriza- opposing wall of the groove (Figures 3C–3E).
tion geometry despite vastly different interface chemistries The central ‘‘loop-in-groove’’ interaction (Figure 3D) features
suggests that there is evolutionary pressure to achieve this ten hydrogen bonds (all hydrogen bonds discussed are pre-
dimerization mode for Semaphorin function, presumably to dicted from geometry) between the protruding 4c-4d loop of
orient bound Plexins for signaling. Sema7A and the PlexinC1 groove (Table S2). The polar interac-
tions are supplemented by two hydrophobic residues (Leu276
The Structure of PlexinC1-SemaPSI Has Unique and Val278) in the Sema7A 4c-4d loop, which contact several
Features hydrophobic residues in the PlexinC1 groove through van der
Given that the extracellular segments of Plexins have not been Waals interactions. Overall, the PlexinC1/Sema7A loop-in-
structurally characterized, the fold of the PlexinC1 Sema-PSI groove interaction includes a mixture of hydrophobic residues
domains merits some description. The Sema domain of PlexinC1 lining the groove wall, interspersed with hydrophilic residues,
in the complex is generally similar to other seven-bladed to presumably provide ligand specificity through polar interac-
b propeller domains in topology but has unique features that tions. Importantly, at the edge of this groove in PlexinC1,
are distinct from Semaphorins or MET (Stamos et al., 2004) Sema7A Lys280 forms a salt bridge with PlexinC1 Asp200, and
(Figure 2B). After the common nomenclature used by other this interaction appears to be a key component of A39R mimicry
propeller proteins, blade 7 of the PlexinC1 Sema domain is (discussed below).
C-terminally adjacent to blade 1; each of the seven blades is At the obstructed (closed) end of the groove (Figure 3B), the
formed by four antiparallel b strands with strands a-d from the Sema7A extrusion helix 2 interacts in a roughly parallel manner
inside to the outside of the b propeller (Figure 2B). The surface with the PlexinC1 3b-3c loop, also showing several ancillary
bears the loops linking strands b and c (e.g., loop 4b-4c) and link- long-range interactions with the tips of the PlexinC1 2b-2c and
ing strands d and a (e.g., loop 5d-6a) on the top face. Loop 1d-2a 2d-3a loops (Figure 3C). The interaction is intimate only at the
traverses the top of the propeller like a flap, sealing the central PlexinC1 Ala197-Ala198-Ser199 bulge. The obstructed end of
channel of the propeller, which is hollow in Semaphorins, MET, the PlexinC1 groove is largely due to the steric bulk sidechain
and integrins (Figure S2E) (Antipenko et al., 2003; Love et al., of Arg131, which forms a salt bridge with Sema7A Glu376
2003; Stamos et al., 2004; Xiao et al., 2004; Xiong et al., 2002). (Figure 3C).
A long insertion, termed the ‘‘extrusion’’ (Love et al., 2003), is In the third region of the interface, at the open end of the Plex-
located between strands 5c and 5d. A major structural difference inC1 groove (Figure 3B), the outer edge of Sema7A blade 3 forms
is that the extrusion of PlexinC1 is shorter than that of Semaphor- extended interactions with the bulged strand 3d of PlexinC1
ins. The structure of the complex shows that this extrusion in (Figure 3E). There are two spatially separate clusters of salt
Semaphorins is critical for binding Plexins, but the extrusion in bridges in this region of the interface. The first cluster involves

Cell 142, 749–761, September 3, 2010 ª2010 Elsevier Inc. 753

Figure 3. The Interface between Sema7A and PlexinC1-SemaPSI
(A) The overall structure of an interacting pair of Sema7A and PlexinC1-SemaPSI protomers, highlighting the structural elements involved in their binding, colored
cyan for Sema7A and pink for PlexinC1-SemaPSI.
(B) Close-up view of the interface with Sema7A depicted as ribbons and PlexinC1-SemaPSI in a surface representation, showing how the protruding loop 4c-4d
of Sema7A inserts into a groove in PlexinC1 blade 3, and is flanked by the contacts between the Sema7A extrusion helix 2 and the PlexinC1 loop 3b-3c, and the
contacts between the blade 3 of Sema7A and the bulged strand 3d of PlexinC1.
(C) The interactions of residues near the obstructed end of the PlexinC1 groove, with Sema7A in chocolate and PlexinC1 in pink.
(D) The interaction between residues of the 4c-4d loop of Sema7A (cyan) and the PlexinC1 groove (pink).
(E) The interaction between the residues of blade 3 of Sema7A (green) and strand 3d of PlexinC1 (pink). Note that residues 218–220 of PlexinC1 comprise a bulge
from strand 3d.
See also Table S2 and Figure S4 for a list of the interactions and for mutagenesis/binding data on this interface, respectively.

Sema7A residues Arg204 and Arg202 both forming salt bridges binding (Figure S4). In the complex structure, the location of the
with PlexinC1 residue Glu219. Sema7A Tyr213 occupies the ridges flanking the groove in the Semaphorin binding site is
space between this salt bridge and a salt bridge involving the consistent with a previous mutagenesis study implicating
important Sema7A residue Lys280 and PlexinC1 Asp200 (Fig- the region between residues 166–235 of the Sema domain of
ure S2A). The second cluster of charged interactions involve Sema3A in its Plexin-binding specificity (Koppel et al., 1997). In
Sema7A Asp216, which forms bifurcated salt bridges with summary, the Sema7A/PlexinC1 interface is extensive and varied
Arg222 and Lys224 of PlexinC1. Mutagenesis data support that in chemical and structural character and is dominated by the
the interactions at this region are important for Sema7A-PlexinC1 insertion of a long loop in Sema7A into a deep groove in PlexinC1.

754 Cell 142, 749–761, September 3, 2010 ª2010 Elsevier Inc.

Figure 4. The Interface between A39R and PlexinC1-SemaPSI
(A) Structural superposition of the A39R/PlexinC1-SemaPSI and Sema7A/PlexinC1-SemaPSI complexes indicating the closely related docking modes of the viral
and mammalian Semaphorins to the PlexinC1 receptor. The complexes have been aligned on the PlexinC1 component in order to visualize the respective overlap
of the two different Semaphorins. The A39R is in yellow, Sema7A in cyan.
(B) Close-up view of the interface with A39R depicted as ribbons and PlexinC1-SemaPSI in a surface representation.
(C) The interactions of residues near the obstructed end of the PlexinC1 groove, with A39R in chocolate and PlexinC1-SemaPSI in pink.
(D) The interaction between residues of the 4c-4d loop of A39R (blue) and the PlexinC1 groove (pink).
(E) The interaction between the residues of blade 3 of A39R (green) and strand 3d of PlexinC1-SemaPSI (pink).
See also Table S3 for a list of the interactions.

The A39R-PlexinC1 Interaction Globally Resembles accommodations at the periphery of the interface (Figure 4A).
the Sema7A-PlexinC1 Interaction These small movements result in remodeled pairwise interac-
With strikingly similar orientations of the respective b propellers tions by these surrounding structural elements, relative to the
in the two complexes, the A39R-PlexinC1 interface involves Sema7A/PlexinC1 interface, but still preserving several key
the same set of structural elements as the Sema7A-PlexinC1 contacts that presumably are important for the cross-reactivity
interface (Figure 4). The A39R-PlexinC1 interface buries a (discussed below).
total of 1890 Å2 solvent-accessible surface area, slightly smaller The ‘‘loop-in-groove’’ interaction in A39R/PlexinC1 (Fig-
than the Sema7A-PlexinC1 interface. The protruding loop 4c-4d ure 4D) has six hydrogen bonds between the A39R 4c-4d loop
of A39R is in an almost identical conformation as that of Sema7A, and the PlexinC1 groove, four less than in Sema7A/PlexinC1
inserting into the groove of the blade 3 surface of PlexinC1 (Table S3). Only one of these hydrogen bonds is conserved
(Figures 4A, 4B, and 5). The neighboring segments of this loop, between Sema7A/PlexinC1 and A39R/PlexinC1. The diversity
including the extrusion helix 2 of A39R that contacts the loop of amino acid contacts with PlexinC1 formed by the Sema7A
3b-3c of PlexinC1, and the A39R blade 3 that contacts the Plex- versus A39R 4c-4d loop indicates that the PlexinC1 pocket
inC1 strand 3d, have undergone some relatively minor structural has the capacity for highly degenerate interactions, which

Cell 142, 749–761, September 3, 2010 ª2010 Elsevier Inc. 755

Figure 5. The Conformation of the 4c-4d Loop of Semaphorins Is Central for Plexin Recognition
(A–D) Stick models of the isolated 4c-4d loops of Sema7A (cyan), A39R (yellow), Sema4D (green), and Sema3A (pink), showing its conserved conformation at the
bases (top) and the midpoints of the loops, and variable conformation at the tips of the loops (bottom).
(E) Sequence comparison of the 4c-4d loop of Semaphorins, with the key conserved structural determinants highlighted.
See also Figure S5 for the RGD motif at the base of the Sema7A 4c-4d loop.

facilitates cross-reactivity. At the edge of this groove in PlexinC1, PlexinC1 residues as in Sema7A/PlexinC1, but the pattern
A39R presents an arginine (Arg207), as opposed to a lysine of interaction is altered. While there are four salt bridges in
(Lys280) in Sema7A, to form a salt bridge with PlexinC1 two clusters in Sema7A/PlexinC1, there are only two (A39R
Asp200 (discussed below). Arg132/Arg134 to PlexinC1 Glu219) in A39R/PlexinC1 at this
Peripheral to the 4c-4d loop, the structural chemistry of the region (Figure 4E). In A39R/Sema7A, the loss of the other cluster
A39R/PlexinC1 interactions at the obstructed (closed) end of of salt bridges seen in Sema7A/PlexinC1 (PlexinC1 Arg222/
the PlexinC1 groove (Figures 4B and 4C) are quite different Lys224 to Sema7A Asp216) is due to the raised A39R blade 3.
than what is seen for Sema7A/PlexinC1 as a result of a slight While PlexinC1 Lys224 is not directly bonded to A39R in the
rigid-body repositioning (Figure 4A). Compared to Sema7A, the A39R/PlexinC1 complex, PlexinC1 Arg222 manages to form
A39R extrusion helix 2 is shifted outwards relative to the center hydrogen bonds with the hydroxyl of A39R Tyr145. The loss
of the interface and tilted away from PlexinC1 (Figures 4A of these two salt bridges from the mammalian complex may
and 4C). Consequently, the A39R/PlexinC1 interaction is not as be further compensated by the strengthening of the A39R
intimate as the Sema7A/PlexinC1 interaction at this region. Arg132-PlexinC1 Glu219 salt bridge, which is more deeply
However, the shifting of the helix also allows A39R Asp300, buried than in Sema7A/PlexinC1, because of the neighboring
corresponding to Sema7A Gln379 (Figures 3C and 4C), to hydrophobic interaction between A39R Ile125 and PlexinC1
move into the position to form a salt bridge with PlexinC1 Leu220 (Figure 4E). Collectively, The A39R/PlexinC1 interface
Arg131, which is not present in Sema7A/PlexinC1. shows variations from the Sema7A/PlexinC1 interface due to
The A39R/PlexinC1 interactions at the open end of the some intermolecular repositioning around the central 4c-4d
PlexinC1 groove (Figures 4B and 4E) involve the same set of loop. Nevertheless, the general scheme of inserting the long

756 Cell 142, 749–761, September 3, 2010 ª2010 Elsevier Inc.

4c-4d loop into the blade 3 groove in PlexinC1 is identical for both Sema7A and A39R 4c-4d loops (Figure 5) ensures similar
Sema7A and A39R. loop conformations presented to PlexinC1. There is mimicry of
an array of interactions peripheral to the tip of the 4c-4d loop,
as is clear from comparison of the respective binding surfaces.
The Similarities and Variations of the Central
For example, the Lys280 in Sema7A that salt bridges to
Recognition Loops of Semaphorins
Asp200 in PlexinC1 is mimicked by Arg207 in A39R that also
The structural features of the 4c-4d interaction loop of Sema7A
salt bridges to Asp200 on PlexinC1. Also found in both Sema-
and A39R are similar in all Semaphorin structures determined
phorins is a cluster of salt bridges on blade 3 that interact with
to date (Figure 5). The conformation of this loop is stabilized
residues on strand d of PlexinC1. Here, Arg202 of Sema7A salt
by an extensive intraloop hydrogen-bonding network, which
bridges with Glu219 in PlexinC1, and this interaction is mimicked
involves residues highly conserved in most Semaphorins (Fig-
by Arg134 of A39R salt bridging to PlexinC1 Glu219. Arg202
ures 5A–5E). At the base of the loop, Asp269 (Sema7A
and Asp216 in Sema7A are spatially mimicked by Arg132 and
numbering), a buried aspartate conserved in all Semaphorins
Asp148 in A39R, but the hydrogen bonding network with the
(highlighted in Figures 5A–5E), forms a network of four hydrogen
same PlexinC1 residues is remodeled. Key to this cluster is
bonds with main chain amides, which play a central role in orga-
a Tyrosine residue (Tyr213 in Sema7A, Tyr145 in A39R) (Fig-
nizing the loop conformation. At the midpoint, a serine (274 in
ure 6F) that is in a nearly identical position in both complexes,
Sema7A and 201 in A39R) hydrogen bonds with two main-chain
in the center of the cluster of salt bridges.
amides on the opposing strand to narrow the loop. The kinking
In order to ask whether the residues A39R uses to ‘‘mimic’’
of the loop is also facilitated by the presence of two consecutive
Sema7A were energetically important for the viral Semaphorin
glycines (Gly271-Gly272) conserved in all Semaphorins (Fig-
binding, we mutated several of them and tested binding by ITC
ure 5E). At the tip of the loop, the Sema7A Ser274 main-chain
(Figure 6). Mutations of A39R Arg132 or Tyr145 (corresponding
carbonyl forms hydrogen bonds with the main-chain amides
to Sema7A Arg202 and Tyr213) to glutamate or serine abolished
of Ser277 and Val278, a pattern reminiscent of reverse-turns,
PlexinC1 binding (Figures 6C and 6D), whereas mutation of A39R
which may help to maintain a rigid and protruding conformation.
Arg207 (corresponding to Sema7A Lys280) to glutamate
Similar conformation is observed for the tip of this A39R loop
reduced PlexinC1 binding by >60-fold (Figure 6E). The A39R
(Figures 5A and 5B).
Arg207 has a stronger charge and a larger head group than
In Sema3A and Sema4D, however, the tip of this loop has
a Lys residue, and its aliphatic stem is kinked, contacting the
a variation from Sema7A/A39R (Figures 5C and 5D). Class 3–5
neighboring hydrophobic moieties more intimately. Arg207 is a
Semaphorins, including Sema3A and Sema4D, have a one-
more complementary fit in the A39R/PlexinC1 interface com-
residue deletion at the Ser274 position. Sema4D and Sema3A
pared to the Lys280 in the Sema7A-PlexinC1 interface, where
also lack the main-chain hydrogen bonds at the tip seen in
there is more solvent-occupied space adjacent to the side chain
Sema7A and A39R. Consequently, the tips of the 4c-4d loops
(Figures S2A and S2B). Our data (Figure 6E) suggest that the
of these Semaphorins can adopt different main chain conforma-
difference at this position could be an important contributing
tions, probably reflecting the structural requirements of their
factor to the higher-affinity binding to PlexinC1 achieved by
specific Semaphorin-Plexin interactions, yet all appear to
A39R, although additional interactions no doubt also contribute.
contain a polar residue at the loop tips for hydrogen bonding.
We also mutated PlexinC1 Arg222, which is hydrogen bonded to
Because the residues that reside at the apex of the loop are
A39R Tyr145 and serves as shared contact with both Semaphor-
the most deeply embedded in the base of the Plexin groove,
ins. We found a 9-fold affinity loss in A39R-PlexinC1 binding
as exemplified by Sema7A/PlexinC1, it is likely that these are
(Figure 6G). While Sema7A and A39R have exactly the same resi-
major determinants for Semaphorin-Plexin binding specificity
dues at the local region contacting PlexinC1, PlexinC1 Arg222
between classes. Indeed, there is extensive sequence diver-
adopts very different rotamers in the two complexes (Figure 6F).
gence between different classes of Semaphorins at these
The 9-fold affinity loss upon PlexinC1 Arg222Ser mutation
positions, but few within the same class of Semaphorins
suggests that this is an important determinant in A39R/PlexinC1
(Figure 5E).
binding (Figure 6G). Therefore, the residues we mutated, which
are strikingly coincident in the viral and mammalian Sema
The Mimicry of the Mammalian Semaphorin Sema7A complexes, while certainly not giving us a comprehensive ener-
by the Viral Semaphorin A39R getic map of the A39R/PlexinC1 interface landscape, are indeed
The mammalian and viral Semaphorins have arrived at nearly energetically important in ligand-receptor binding.
identical binding modes despite substantial variations in Collectively, our structures suggest that A39R, built on a
sequence (31% identity between Sema7A and A39R). Visualiza- smaller scaffold, has used a limited set of key, highly coincident
tion of the PlexinC1 binding surfaces of Sema7A and A39R amino acid contacts to evolve a more efficient binding interac-
(Figures 6A and 6B) indicates that there are several key corre- tion with PlexinC1 through both the enhancement of particularly
sponding regions of structural mimicry that likely serve as the important polar interactions, such as the substitution of A39R
foundation for their cross-reactivity with a common receptor. Arg207 for the Sema7A Lys280, and slight adjustments of side
There is obvious structural conservation of the centrally located rotamer and main chain positions throughout the binding
4c-4d loop and the presence of identical residues at the tips of surface. These adjustments sum to a substantial optimiza-
the loop (Ser-Leu) that engage the deepest portion of the tion of binding energetics and a concomitant gain in binding
PlexinC1 groove. The preservation of the di-Glycine pair in affinity.

Cell 142, 749–761, September 3, 2010 ª2010 Elsevier Inc. 757

Figure 6. Viral Mimicry of Sema7A by A39R
(A and B) The respective PlexinC1-binding surfaces of Sema7A (cyan, left) and A39R (yellow, right). The residues mapped at the interface are colored red for acidic
residues, blue for basic residues, gray for polar, noncharged residues, and green for apolar residues.
(C–E) Mutations of the A39R residues mapped to the Sema7A-PlexinC1 interface abolished or dramatically reduced A39R-PlexinC1 binding.
(F) The different conformation of PlexinC1 Arg222 in binding A39R and Sema7A.
(G) Calorimetric measurement of the binding between A39R and the Arg222Ser mutant of PlexinC1-SemaPSI.
See also Figure S3 for a structural and sequence comparison of Sema7A and A39R.

758 Cell 142, 749–761, September 3, 2010 ª2010 Elsevier Inc.

DISCUSSION the knowledge that Plexins are active as dimers should help
refine our thinking about possible downstream effectors. The
We present here the structures of two Semaphorin-Plexin quiescent PlexinC1 possibly exists as preformed oligomers,
complexes, establishing the general recognition paradigm given the evidence that some other Plexins, e.g., plexin B, exist
between two protein families with important and wide-ranging as stable homodimers in the absence of ligands (Tong et al.,
physiological functions involving deliverance of repulsive guid- 2007). We have also found that unliganded PlexinC1 ECD exists
ance cues. Unexpectedly, the two b propellers of the Sema- as a mixture of oligomeric states, as assessed by gel filtration
phorin and Plexin Sema domains contact each other edge-on. chromatography and crosslinking, although we have not quanti-
In comparison, MET uses the bottom face of the propeller for tated the exact stochiometries of these states. We propose that
HGF-b chain interaction (Stamos et al., 2004), whereas integrins constitutively dimeric ligands such as Sema7A and A39R orient
bind their ligands with the top face of the propeller (Xiao et al., the inactive PlexinC1 into a dimerized state, in which the
2004; Xiong et al., 2002). Thus, as for other recognition modules distance, geometry and conformation enables the activation of
in biology, such as leucine-rich repeats, immunoglobulin folds, the intracellular GAP activity. This PlexinC1 activation scheme
and even MHC-folds, the b propeller reveals itself to be capable is simpler than a previously proposed model for Sema3A-
of a diverse array of ligand interaction modes. The moderate-to- induced PlexinA1 activation. That model proposed that the
low sequence identity between different classes of Semaphorins Plexin Sema domain associates intramolecularly with its addi-
and Plexins, especially at the binding segments, suggests that tional extracellular domains in an autoinhibited conformation,
there is variability in the specific interatomic contacts in Sema- analogous to the LDL receptor (Rudenko et al., 2002), and that
phorin/Plexin interfaces between families that is buttressed by Sema3A disrupts this intramolecular association (Antipenko
a framework of conserved interactions. This variability may be et al., 2003; Takahashi and Strittmatter, 2001). Although we
necessary for coding the complex pattern of specificity in the have no unequivocal data to support or refute such a model,
growth- or motility-guiding activities. Nevertheless, the orienta- our ITC experiments of A39R binding to full-length versus trun-
tion of each component in the complex, the Sema-Sema dock- cated PlexinC1 ECD show identical binding energetics and
ing geometry, and the global usage of structural elements should monophasic association, indirectly implying that PlexinC1 is in
be generally conserved for Semaphorins and Plexins, as has the same conformation in the full-length ECD and truncated
been found in other ligand-receptor families with similar extents forms (Figures 1D and 1E). An important caveat to our experi-
of sequence identity. ments is that thermodynamic measurements would not detect
The poxvirus-derived A39R has evolved a more efficient a potential biphasic binding kinetics indicative of a two-step
PlexinC1 binding surface, yet it is smaller than that of Sema7A, binding event involving a Plexin conformational change. Never-
with fewer overall contacts. Our structural analysis showed theless, whether Plexin undergoes a conformational change
that the central binding elements of Sema7A are rigorously main- prior to Semaphorin binding, clearly a Semaphorin-induced
tained in the viral complex; therefore, how does A39R achieve arrangement of Plexin into a specifically orientated dimer is the
higher affinity binding? This is very difficult to clearly answer, means of receptor activation. Future structural and mechanistic
as the sources of free energy differences between interacting studies can now address whether the preactivated Plexin is
proteins are not obviously identified by inspection of structures involved in an intramolecular association that contributes to
and are contributed by many subtle effects such as intramolec- the maintenance of an autoinhibited state.
ular side chain cooperativity, solvent restructuring at the binding The Sema7A/PlexinC1 and A39R/PlexinC1 complexes evoke
surface, and enthalpy-entropy compensation. Thus, we suspect several important questions. First, where is the Neuropilin-
that it is due to a collective reduction of energetic penalties docking site in a class 3 Semaphorin-Plexin complex? Although
across the entire Semaphorin/Plexin interface, perhaps in predictions could be made on the basis of our structures
combination with an optimization of protein dynamics, such as and the previous mutagenesis mapping data (Antipenko et al.,
lowering overall temperature factors (Table S1), to gain favor in 2003), accurate insights may require a complex structure
entropic change. The high similarity of Sema7A and A39R encompassing Neuropilin. Second, does Sema7A bind integrins
suggests the likelihood that the poxviruses have acquired the directly? It has been suggested Sema7A promotes axon growth
A39R gene by hijacking and mutating the Sema7A gene, rather and initiates T cell-mediated inflammation responses through
than through convergent evolution where the virus would have integrins, but not PlexinC1, using an RGD-dependent mechanism
evolved a structurally distinct scaffold for PlexinC1 binding (Pasterkamp et al., 2003; Suzuki et al., 2007). The RGD motif of
(Gewurz et al., 2001). Thus, this does not seem to be a case of Sema7A (Arg267-Gly268-Asp269) is buried (Figure S5) and is
molecular ‘‘mimicry’’ as much as simply modifying an existing unlikely to be recognized by integrins. Finally, the complex struc-
binding scaffold with subtle improvements. Given the lack of tures can now be used as a basis to interrogate Sema/Plexin
involvement of the PSI and Ig domains in Plexin interactions, interactions in vivo with high precision. While mutation of the
these domains may have been lost in A39R over the course of key determinants of both Sema7A-PlexinC1 and A39R-PlexinC1
evolution. interfaces clearly compromise binding (Figure 6, Figure S4), fur-
The dimerized Sema7A/PlexinC1 and A39R/PlexinC1 com- ther work will be necessary to demonstrate the consequences
plexes support a model of PlexinC1 activation via dimerization of such perturbation in a cellular or organismal context. Such
akin to many other cell-surface receptors. This, in itself, is an functional validation should determine whether modulation of
important advance since the downstream mechanisms of Sem- Semaphorin-Plexin recognition holds clinical promise for applica-
aphorin/Plexin repulsive signaling are not well understood, and tions involving cell guidance, e.g., directional nerve regeneration.

Cell 142, 749–761, September 3, 2010 ª2010 Elsevier Inc. 759

EXPERIMENTAL PROCEDURES Isothermal Titration Calorimetry
The proteins used for calorimetry were all expressed from HEK293H cells. ITC
Cloning, Cell Culture, and Baculovirus Generation was carried out on a VP-ITC calorimeter (MicroCal, Northhampton, MA) at
Human Sema7A, PlexinC1-SemaPSI, and PlexinC1-ECD, were expressed in 30 C. All samples were thoroughly degassed before titration. The titration
the BacMam system as described previously (Dukkipati et al., 2008), with data were processed with MicroCal Origin software, Version 5.0.
a 7-His tag directly attached to the C terminus of each expression construct.
Full-length Vaccinia virus A39R was expressed in both the BacMam system
and the baculovirus-insect cell system. Sf9 and hi5 cells were maintained in ACCESSION NUMBERS
the HyQ-SFX (HyClone) and InsectXpress (Lonza) media. HEK293H cells
(Invitrogen) and GnTI- HEK293 cells (Reeves et al., 2002) were maintained in The structure factors and coordinates for the Sema7A/PlexinC1-SemaPSI
CDM4HEK293 media (HyClone). For generation of recombinant viruses, the complex, free A39R and the A39R/PlexinC1-SemaPSI complex have been
expression constructs and the BacVector-3000 baculovirus DNA (EMD deposited in the Protein Data Bank (accession codes 3NVQ, 3NVX, and
Chemicals) were used to cotransfect sf9 cells in 6-well plates. After 5 days, 3NVN).
the resulted low titer virus stock was harvested and was used to infect Sf9 cells
at 2 3 106 cell/ml for amplification. SUPPLEMENTAL INFORMATION

Protein Preparation Supplemental Information includes Extended Experimental Procedures, five

The amplified baculoviruses were used to infect 1-6 liters of HEK293 cells figures, and three tables and can be found with this article online at doi:
or hi5 cells at a density of 1.0–1.5 3 106 cells/ml. After 72 hr, the cells were 10.1016/j.cell.2010.07.040.
pelleted and the supernatants were concentrated. The recombinant proteins
in the supernatant were captured by Ni-NTA metal affinity resin (QIAGEN)
and eluted with 300 mM imidazole (pH 7.5). The eluted proteins were further ACKNOWLEDGMENTS
purified with gel filtration.
We thank the staff at the LS-CAT beamlines at APS and the staffs of ALS and
Glycan Trimming, Crystallization, and Data Collection SSRL for the support in X-ray data collection. We also thank L. Colf, and
and Processing N. Goriatcheva for help with the expression of A39R. X.H. is supported by
For the Sema7A-PlexinC1 complex, GnTI HEK293-expressed proteins were the NIH grant 1R01GM07805, and K.C.G. is an Investigator of the Howard
trimmed with carboxypeptidase-A and Endo-H. The digested products were Hughes Medical Institute and is supported by NIH-R01-AI51321. The Struc-
further purified with size exclusion columns and concentrated to 10 mg/ml. tural Biology Facility is supported by the R.H. Lurie Comprehensive Cancer
Crystallization was performed via the sitting-drop vapor-diffusion method. Center of Northwestern University.
The Sema7A and PlexinC1-SemaPSI proteins, at 1:1 molar ratio, were mixed
with an equal volume of reservoir solution and equilibrated against the reser- Received: March 19, 2010
voir solution containing 12% PEG1000, 0.1 M Tris (pH 7.5), and 0.2 M Li2SO4. Revised: June 8, 2010
The A39R used for crystallization was expressed from hi5 cells using the Accepted: July 20, 2010
pAcGP67A vector. A39R was His-tagged at the N-terminus and expressed Published online: August 19, 2010
in the presence of 10 mM kifunensine in the expression media (InsectXpress).
The His-tag was removed from A39R with TEV protease, the glycans were REFERENCES
trimmed with Endo-H, and the protein was further purified by gel filtration.
A39R crystallized in the following condition: 50 mM sodium citrate (pH 5.0), Antipenko, A., Himanen, J.P., van Leyen, K., Nardi-Dei, V., Lesniak, J., Barton,
0.2 M Li2SO4, 15% PEG3350, and 20% glycerol. W.A., Rajashankar, K.R., Lu, M., Hoemme, C., Püschel, A.W., and Nikolov,
For the A39R-PlexinC1-SemaPSI complex, the A39R protein was expressed D.B. (2003). Structure of the semaphorin-3A receptor binding module. Neuron
from the BacMam system following the same procedure for PlexinC1-Sema- 39, 589–598.
PSI. The complex was crystallized in the following condition: 6% PEG8000,
Bork, P., Doerks, T., Springer, T.A., and Snel, B. (1999). Domains in plexins:
0.1 M CaCl2, 0.1 M HEPES (pH 7.0), and 0.2 M NaCl.
links to integrins and transcription factors. Trends Biochem. Sci. 24, 261–263.
The Sema7A/PlexinC1 and A39R/PlexinC1 data were measured at the
Life Sciences Collaborative Access Team (LS-CAT) beamlines 21-ID-D and Bricogne, G., Vonrhein, C., Flensburg, C., Schiltz, M., and Paciorek, W. (2003).
21-ID-G at the Advanced Photon Source (APS) and were processed with Generation, representation and flow of phase information in structure determi-
HKL2000 (Otwinowski and Minor, 1997) (Table S1). The A39R data were nation: recent developments in and around SHARP 2.0. Acta Crystallogr. D
measured at both the Stanford Synchrotron Research Lab (SSRL) and the Biol. Crystallogr. 59, 2023–2030.
Advanced Light Source (ALS) and were processed with Mosflm and scaled Brünger, A.T., Adams, P.D., Clore, G.M., DeLano, W.L., Gros, P.,
with Scala of the CCP4 program suite (Collaborative Computational Project, Grosse-Kunstleve, R.W., Jiang, J.S., Kuszewski, J., Nilges, M., Pannu, N.S.,
Number 4, 1994). The statistics on the data are presented in Table S1. et al. (1998). Crystallography & NMR system: a new software suite for macro-
molecular structure determination. Acta Crystallogr. D Biol. Crystallogr. 54,
Structure Determination and Model Refinement 905–921.
For structure determinations of Sema7A/PlexinC1-SemaPSI and free A39R,
Collaborative Computational Project, Number 4. (1994). The CCP4 suite:
experimental phases were calculated using heavy atom derivatives by the
programs for protein crystallography. Acta Crystallogr. D Biol. Crystallogr.
SIRAS method, as implemented in the program SOLVE (Terwilliger and
50, 760–763.
Berendzen, 1999) and SHARP (Bricogne et al., 2003). For structure determina-
tion of the A39R/PlexinC1-SemaPSI complex, the components from the above Comeau, M.R., Johnson, R., DuBose, R.F., Petersen, M., Gearing, P.,
two structures were used as search models in molecular replacement. The VandenBos, T., Park, L., Farrah, T., Buller, R.M., Cohen, J.I., et al. (1998). A
models were refined with CNS (Brünger et al., 1998). The final models include poxvirus-encoded semaphorin induces cytokine production from monocytes
seven N-linked glycans per PlexinC1-SemaPSI molecule (Asn86, Asn141, and binds to a novel cellular semaphorin receptor, VESPR. Immunity 8,
Asn149, Asn241, Asn252, Asn386, and Asn407), four per Sema7A (Asn105, 473–482.
Asn157, Asn258, and Asn330), and one per A39R (Asn51). Each glycan Dukkipati, A., Park, H.H., Waghray, D., Fischer, S., and Garcia, K.C. (2008).
was modeled with a single GlcNAc residue, as expected, due to the Endo-H BacMam system for high-level expression of recombinant soluble and
treatment of the proteins. A summary of the refinement statistics is given in membrane glycoproteins for structural studies. Protein Expr. Purif. 62,
Table S1. 160–170.

760 Cell 142, 749–761, September 3, 2010 ª2010 Elsevier Inc.

Gewurz, B.E., Gaudet, R., Tortorella, D., Wang, E.W., and Ploegh, H.L. (2001). N-acetylglucosaminyltransferase I-negative HEK293S stable mammalian cell
Virus subversion of immunity: a structural perspective. Curr. Opin. Immunol. line. Proc. Natl. Acad. Sci. USA 99, 13419–13424.
13, 442–450. Rudenko, G., Henry, L., Henderson, K., Ichtchenko, K., Brown, M.S.,
He, Z., Wang, K.C., Koprivica, V., Ming, G., and Song, H.J. (2002). Knowing Goldstein, J.L., and Deisenhofer, J. (2002). Structure of the LDL receptor extra-
how to navigate: mechanisms of semaphorin signaling in the nervous system. cellular domain at endosomal pH. Science 298, 2353–2358.
Sci. STKE 2002, re1. Scott, G.A., McClelland, L.A., Fricke, A.F., and Fender, A. (2009). Plexin C1,
He, X.L., Dukkipati, A., and Garcia, K.C. (2006). Structural determinants of a receptor for semaphorin 7a, inactivates cofilin and is a potential tumor
natriuretic peptide receptor specificity and degeneracy. J. Mol. Biol. 361, suppressor for melanoma progression. J. Invest. Dermatol. 129, 954–963.
698–714. Semaphorin Nomenclature Committee. (1999). Unified nomenclature for the
He, H., Yang, T., Terman, J.R., and Zhang, X. (2009). Crystal structure of the semaphorins/collapsins. Cell 97, 551–552.
plexin A3 intracellular region reveals an autoinhibited conformation through Stamos, J., Lazarus, R.A., Yao, X., Kirchhofer, D., and Wiesmann, C. (2004).
active site sequestration. Proc. Natl. Acad. Sci. USA 106, 15610–15615. Crystal structure of the HGF beta-chain in complex with the Sema domain of
Holmes, S., Downs, A.M., Fosberry, A., Hayes, P.D., Michalovich, D., the Met receptor. EMBO J. 23, 2325–2335.
Murdoch, P., Moores, K., Fox, J., Deen, K., Pettman, G., et al. (2002). Sema7A Suzuki, K., Okuno, T., Yamamoto, M., Pasterkamp, R.J., Takegahara, N.,
is a potent monocyte stimulator. Scand. J. Immunol. 56, 270–275. Takamatsu, H., Kitao, T., Takagi, J., Rennert, P.D., Kolodkin, A.L., et al.
Huppa, J.B., Axmann, M., Mörtelmaier, M.A., Lillemeier, B.F., Newell, E.W., (2007). Semaphorin 7A initiates T-cell-mediated inflammatory responses
Brameshuber, M., Klein, L.O., Schütz, G.J., and Davis, M.M. (2010). TCR- through alpha1beta1 integrin. Nature 446, 680–684.
peptide-MHC interactions in situ show accelerated kinetics and increased Suzuki, K., Kumanogoh, A., and Kikutani, H. (2008). Semaphorins and their
affinity. Nature 463, 963–967. receptors in immune cell interactions. Nat. Immunol. 9, 17–23.
Koller, K.J., and Goeddel, D.V. (1992). Molecular biology of the natriuretic Takahashi, T., and Strittmatter, S.M. (2001). Plexina1 autoinhibition by the
peptides and their receptors. Circulation 86, 1081–1088. plexin sema domain. Neuron 29, 429–439.
Kolodkin, A.L., Matthes, D.J., and Goodman, C.S. (1993). The semaphorin Tamagnone, L., and Comoglio, P.M. (2000). Signalling by semaphorin recep-
genes encode a family of transmembrane and secreted growth cone guidance tors: cell guidance and beyond. Trends Cell Biol. 10, 377–383.
molecules. Cell 75, 1389–1399.
Tamagnone, L., Artigiani, S., Chen, H., He, Z., Ming, G.I., Song, H., Chedotal,
Koppel, A.M., Feiner, L., Kobayashi, H., and Raper, J.A. (1997). A 70 amino A., Winberg, M.L., G