s inA Sponsored Supplement to Science

arch.
Optimizing
your live-cell
microscopy:
Tricks and
trade-offs

Produced by the
Sponsored by Science/AAAS Custom
Publishing Office
 TA B L E O F C O N T E N T S

Optimizing
Introductions
your live-cell
2 A life well lived in the age of microscopy
microscopy: Sean Sanders, Ph.D.
Science/AAAS

Tricks and 3 Shedding light on life

trade-offs Bernhard Zimmermann, Ph.D.
Senior Director Segment Marketing Life Sciences
Carl Zeiss Microscopy GmbH

ScienceCareers.org
White paper: Superresolution microscopy
4 Bringing cellular dynamics to light with live-cell microscopy
Jeffrey M. Perkel

Research articles
8 Detryrosinated microtubules buckle and bear load in
contracting cardiomyocytes
Patrick Robison, Matthew A. Caporizzo, Hossein Ahmadzadeh et al.

17 Complement and microglia mediate early synapse loss in

Alzheimer mouse models
Soyon Hong, Victoria F. Beja-Glasser, Bianca M. Nfonoyim et al.

22 Bioengineering a 3D integumentary organ system from

iPS cells using an in vivo transplantation model
Ryoji Takagi, Junko Ishimaru, Ayaka Sugawara et al.

Application note

ABOUT THE COVER:

Learn more and don’t let your job search
leave you washed up.
30 The Fast module for ZEISS LSM 880 with Airyscan:
Confocal superresolution imaging with four times the
Choanoflagellate rosette colony isolated from Mono
Lake, California, stained for DNA, microtubules, and
F-actin. Acquired using a ZEISS LSM 880 with Airyscan.
speed and improved signal-to-noise ratio
Joseph Huff and Annette Bergter
Credit: Kayley Hake, University of California, Berkeley.

This booklet was produced by the Science/AAAS

Custom Publishing Office and sponsored by Carl Zeiss
Microscopy GmbH.

Editor: Sean Sanders, Ph.D.

● Search thousands of job postings Proofreader/Copyeditor: Bob French
Designer: Amy Hardcastle
● Create job alerts based on your criteria
BILL MORAN, GLOBAL DIRECTOR
● Get career advice from our Career Forum experts Custom Publishing
bmoran@aaas.org
● Download career advice articles and webinars +1-202-326-6438

● Complete an individual development plan at “myIDP” ROGER GONCALVES, SALES MANAGER

Target your job search using relevant resources
on ScienceCareers.org.
Custom Publishing
Europe, Middle East, and India
rgoncalves@science-int.co.uk
+41-43-243-1358
© 2016 by The American Association for the Advancement
of Science. All rights reserved. 13 May 2016

1
O PT IM IZING YOUR LI V E -CE L L M I CRO S CO P Y: TR I C K S A N D T R A D E - O F F S
I A
n the world of cellular biology, the microscope is king. If you
want to know what’s going on at a cellular or subcellular level,
A life well one of the few direct ways of doing so is with some flavor of
microscopy. From its invention in the early 1600s (the origin of
lived in
the first light microscope is somewhat in dispute), inventors and
researchers have been constantly improving on the original design
through hardware and software upgrades, pushing the boundaries
the age of of resolution, speed, and depth of imaging. Seldom does a year pass
without a significant advance in this field, most especially in the realm
microscopy of fluorescent microscopy (which garnered a Nobel Prize in Chemistry
in 2014).
For the vast majority of the time since the invention of the micro-
scope, in order to view a specimen, it has needed to be prepared
through a permanent fixation procedure. Also, samples are often spe-
cially treated to reveal the location of specific cellular components,
a process that requires the fixation step. However, this methodol-
ogy provides only a single instant in time for researchers to analyze.
Sometimes this is sufficient. But in order to study the kinetics of a
cellular event or track changes in a cell over time, the ability to image
a live cell is paramount. Additionally, the process of preparing fixed
cells has been found to cause artifacts that can confound the interpre-
tation of the data.
Caveat emptor. Together with the clear advantages of live-cell
imaging come new challenges. Dead cells can be treated relatively
harshly without much concern. Live cells on the other hand, frequently
require careful handling that includes a constant temperature, stable
CO2 levels, and sufficient nutrients. Moreover, the lasers used in many
fluorescent microscopes to create the beautiful and colorful images
we have come to know are so powerful that they can easily damage
cells even during short-term illumination. Live-cell imaging is, in its
most fundamental form, a constant balance between achieving the
necessary data and not damaging or killing the cells.
A variety of clever solutions have been developed recently that
surpass the methods and materials of yesteryear, including brighter
fluorescent dyes that require lower-energy activation light, more
sensitive detectors that can image both more quickly and in lower
light, and techniques and technologies that use hardware and
software tricks to minimize the amount of light bathing the sample.
None of these provides a complete solution, but these advances
and others have allowed researchers to better balance the best
resolution, fastest imaging speed, maximum imaging depth, and best
cell health to obtain the optimal results. The content provided in this
new supplement gives you an indication of how they’re succeeding
in these efforts. We’ve provided three recent Science and Science
Advances articles and an exclusive white paper in this timely update
on live-cell imaging. We hope you enjoy reading it.
Sean Sanders, Ph.D.
Editor, Science/AAAS Custom Publishing Office
2 sciencemag.org SCIENCE
confocal imaging. S E C T I O N O N E | I N T RO D U C T I O N S
ZEISS LSM 880 with Airyscan
“
picture is worth a thousand words,” goes the saying. As
such, the evolution of humanity is tracked and studied
Shedding by the drawings and pictures created by generations
who sought to depict their world, and the accuracy and
light on life
frequency of the images mark how well we were and are
able to describe and understand the phenomena of the world.
Over the last two centuries the rapid progression of innovative mi-
croscopy techniques has allowed unprecedented visualization and
description of microscopic phenomena, making a light microscope
an indispensable tool in every biological or medical laboratory. As
the light microscope has developed, it has given rise to hundreds of
different imaging techniques, from brightfield and differential interfer-
ence contrast to modern fluorescence and superresolution microsco-
py, which all have one thing in common: They are based on light and
its interaction with the specimen. Nowadays light microscopes cannot
match the resolution of electron microscopes or the penetration of X-
ray microscopes, but they remain unchallenged when it comes to the
imaging of dynamic processes in living cells, tissues, and organisms—
observing life as it happens. Traditionally, light microscopes were
divided into two distinct groups: widefield microscopes, which can
acquire an entire image in a single exposure, and laser scanning mi-
croscopes, that image a specimen sequentially, point-by-point. While
the latter produce crisper images of thick specimens, the former re-
cord images much faster and with less damage to the specimen.
Since Carl Zeiss, Ernst Abbe, and Otto Schott laid the foundations
of modern microscopy 170 years ago, the company they founded
delivered most of the radical microscope innovations, from the simple
// INNOVATION achromatic and apochromatic lenses of the 19th century, to the com-
MADE BY ZEISS plex imaging instruments of today. The latest innovation from ZEISS
is a novel confocal detector named Airyscan. It blurs the traditional
divide between widefield and laser scanning microscopes and offers
users the best of both worlds: high-speed, gentle imaging combined
with superb resolution and contrast, capabilities unmatched by stan-
dard confocal microscopes.
We hope you will find this collection of publications and peer-
reviewed articles interesting and inspiring, an example of how in-
novative light microscopy continues to enable researchers to push
the boundaries of human knowledge. We believe that modern light
Your new standard for fast and gentle confocal imaging
microscopy can make a difference in your research as well. We are
Discover ZEISS LSM 880 with Airyscan – the new excited
confocaltolaser
contribute to microscope
scanning your success.
that offers high sensitivity, improved resolution inBernhard
x, y and z, Zimmermann,
and high speed.Ph.D.
All in one system. Find out more and book a hands-on demonstration in one
Senior Director Segment of
Marketing Life Sciences
our ZEISS Microscopy Labs now. Carl Zeiss Microscopy GmbH
microscopy@zeiss.com
www.zeiss.com/lsm880
SCIENCE sciencemag.org 3
O PT IM IZING YOUR LI V E -CE L L M I CRO S CO P Y: TR I C K S A N D T R A D E - O F F S
Bringing cellular
dynamics to light
with live-cell
I
microscopy
By Jeffrey M. Perkel
L
ight microscopy has countless applications in today’s
life sciences laboratory, from documenting cellular and
organismal structure and mapping RNA localization, to
tracking cellular dynamics. First, though, researchers must an-
swer a fundamental question: Are they going to image the cells
while they’re alive, or after they’re dead?
For many projects, the answer is: both. Yet the question
defines the kind of data a microscope provides, and thus, the
questions researchers can ask of it. Dead men tell no tales, as
the saying goes, but dead cells do. Immunohistochemistry, for
instance, reveals cellular and tissue morphology and protein
distribution in fixed samples, while in situ hybridization reports
the abundance and subcellular location of specific DNA or
RNA sequences—data that pathologists can use to diagnose
and stratify disease, among other things. Confocal and super-
resolution microscopy methods likewise have yielded up daz-
zling images of organismal and subcellular architecture from
fixed cells, such as the 2013 discovery (using stochastic optical
reconstruction microscopy, or STORM, superresolution micros-
copy) by Harvard University’s Xiaowei Zhuang of a “periodic
cytoskeletal structure” undergirding long axonal filaments—a
subcellular protein scaffold that researchers never previously
knew existed (1).
In all these cases, the data produced are static representa-
tions of once-living systems—the equivalent of reducing a
4 sciencemag.org SCIENCE
R
esearchers have no shortage of options for livecell
microscopy, from standard brightfield and differential
interference contrast microscopy, to confocal and super-
resolution approaches. Here, we consider some of the more
popular fluorescence-based strategies.
The most straightforward approach, perhaps, is basic wide-
field fluorescence microscopy, in which an entire field of view is
illuminated with excitation energy, and total fluorescence from
the area is recorded at once using an area detector, such as a
CCD (charge-coupled device) or CMOS (complementary metal-
Airyscan image of oxide semiconductor) camera. The simplest microscopy meth-
od available, widefield imaging also yields the blurriest images,
the central nervous
as there’s no mechanism to exclude out-of-focus light from
system of a Drosophila
melanogaster embryo.
reaching the detector. Thus, according to a review on the Carl
Zeiss Microscopy Online Campus, “widefield imaging achieves
the optimum results when the features of interest are either
large (such as an organelle) or highly punctate in nature” (4).
One strategy for overcoming those limitations is confocal
movie to a single frame. And for many questions, the resulting
data are good enough to provide the answers. But some ques-
tions simply cannot be answered from single images. In those
microscopy. A bright light source (usually a laser) is focused
onto a particular point in the sample. The resulting fluores-
cence is then captured after it passes through a small pinhole
cases, researchers turn to live-cell imaging. aperture, which blocks out-of-focus light, producing a sharp
image. The focal plane is then stepped perpendicular to the
n theory, live-cell microscopy is no different from any other slide (i.e., along the z axis), creating multiple optical planes
microscopy method. But live cells do pose significant techni- that can be virtually assembled to create a 3D representation
cal issues, and researchers pursuing live-cell methods have of the original sample.
devised strategies for dealing with them. Confocal systems generally take one of two approaches
The sine qua non of live-cell microscopy is, obviously, live to reconstruct each optical slice: Either the diffraction-limited
cells, and both microscope vendors and third-party firms offer illumination point is raster-scanned across the sample (as in
tools (such as climate-controlled growth chambers) to facili- laser-scanning confocal microscopy), or multiple points are im-
tate such research. But it’s not enough that the cells be alive; aged simultaneously (as in spinning-disk confocal microscopy).
generally they also have to be healthy, and not all microscopy The difficulty is that whatever z position is being imaged, the
techniques are compatible with that state. Fluorescence-based entire sample volume is illuminated, increasing the likelihood of
IMAGE: COURTESY OF JULIA SELLIN, MICHAEL HOCH, LIMES INSTITUT, UNIVERSITY OF BONN
approaches, especially, often require extended and intense phototoxicity and photobleaching. The spinning-disk approach
sample irradiation with high-powered light sources, which can is generally considered faster and less phototoxic, but both can
induce phototoxicity. Fixed cells can easily stand this assault—so be applied to live cells, assuming the experiment is designed
long as the fluorescent dyes or proteins being imaged do not appropriately. Newer resonant scanning heads, such as that in
succumb to negative effects such as photobleaching. But live the Nikon A1R+ confocal system can accelerate scanning-based
cells can become stressed and apoptotic, or at the very least microscopy to 30 frames per second or higher, depending on
IMAGE: COURTESY OF BALAZS ERDI, MAX F. PERUTZ LABORATORIES,
alter their behavior, under the blinding glare of such scrutiny. the size of the scanned region.
“As part of their normal life cycle, most tissues and cells are
UNIVERSITY OF VIENNA AND MEDICAL UNIVERSITY OF VIENNA
never exposed to light, and it is known that ultraviolet (UV)
light damages DNA, focused infrared (IR) light can cause lo-
calized heating, and fluorescence excitation causes phototox-
icity to tissues and cells,” wrote Clare Brown and colleagues of
McGill University in a 2009 commentary in the Journal of Cell
Science (2).
Live cells also move and change shape, as do their subcel-
lular components, potentially complicating image collection
and data analysis. Researchers are also limited in the types
and nature of stains they can use to visualize those cells, and
many fluorescent dyes are incompatible with live cells. As a
result, live-cell imaging studies often employ fluorescent pro- A still from an 11-hour time lapse of Drosophila melanogaster
teins rather than inorganic dyes or fluorophore-conjugated embryogenesis imaged with Airyscan Fast. Mictrotubules
antibodies for staining purposes, or enzyme-based tags such labeled with green fluorescent protein (GFP); depth-coded
as HaloTag, which allow researchers to label specific proteins maximum intensity projection.
in live cells (3).
SCIENCE sciencemag.org
S E C T I O N T WO | WHITE PAPER
One alternative to confocal microscopy is pairing widefield
microscopy with deconvolution strategies. Widefield images
typically subject cells to lower light intensity than confocal. But,
since the resulting pictures tend to be blurrier, especially with
thick specimens, they can be cleaned up using computational
algorithms to produce a sharper reconstruction.
“Superresolution microscopy” is a
catch-all term describing a half-dozen
or so different methods.
Another alternative strategy, also compatible with live-cell
imaging, is confocal microscopy using ZEISS’s Airyscan detec-
tor. In a typical confocal, one or more pinholes are used to re-
ject out-of-focus light; in-focus emission light is captured using
a point detector (such as a photomultiplier tube) or spectral
array. As a result, the researcher must balance signal with clar-
ity: By making the pinhole smaller, image sharpness increases,
but less light is captured, meaning more intense excitation light
must be used. Airyscan circumvents that problem by eliminat-
ing the pinhole and projecting the emission light onto a 32-el-
ement array of hexagonal detectors. “Each detector element
functions as a single, very small pinhole,” according to compa-
ny literature, leading to stronger signals, increased resolution,
and decreased imaging speed.
“Superresolution microscopy” is a catch-all term describing
a half-dozen or so different methods. In general, they also fall
into two categories, the first of which includes localization-
based approaches such as PALM (photoactivated localization
microscopy) and STORM (available from ZEISS and Nikon,
respectively). Normally, all fluorescent molecules in a diffrac-
tion-limited spot will fluoresce, making it impossible to dif-
ferentiate fluorophores separated by less than about 200 nm.
Localization-based approaches circumvent that problem by
keeping most fluorophores dark and allowing only a stochastic
handful to fluoresce. The positions of those fluors can then be
mapped with subdiffraction resolution, at which point the pro-
cess repeats until sufficient molecules have been mapped to re-
construct the original image. The second category of methods,
which includes STED (stimulated emission depletion, available
from Leica Microsystems), RESOLFT (reversible saturable opti-
cal fluorescence transitions), and SIM (structured illumination
microscopy, available from ZEISS, Nikon, and GE Healthcare
Life Sciences), use patterned light sources to define specifically
where fluorophores are on and off.
Live-cell images have been collected with many of these
different techniques. In one 2013 study, for instance, Nobel
laureate Stefan Hell and colleagues at the Max Planck Institute
for Biophysical Chemistry used a homemade RESOLFT system
capable of producing 116,000 simultaneous “doughnuts” of
light to capture neurite growth over nearly 4 minutes (5). In
early 2016, Joerg Bewersdorf and colleagues at continued>
5
O PT IM IZING YOUR LI V E -CE L L M I CRO S CO P Y: TR I C K S A N D T R A D E - O F F S S E C T I O N T WO | WHITE PAPER

F
Dual-color PALM image of Chinese Hamster Ovary cells using
fast sequential laser switching (Dronpa-Paxillin, red; tdEOS-
Vinculin, green).
Superresolution SIM image of synapses at the neuromuscular
or those specifically interested in membrane dynamics, Conversely, an imaging approach that is very fast must, by
TIRF (total internal reflection fluorescence) microscopy is necessity, not linger over any one position, meaning that fewer
a popular live-cell imaging technique. In TIRF, excitation photons will be collected, causing signal-to-noise to suffer—
energy is directed at the glass-sample interface at an oblique unless brighter excitation light is used.
angle such that it creates an evanescent wave parallel to the Microscopists are no stranger to this tug-of-war—it’s one of
slide surface, extending only about 100 nm into the sample. the reasons most research institutions contain microscopy core
As a result, only those fluorophores relatively near the slide sur-
facilities stocked with multiple instruments (economic consider-
face will fluoresce, producing a sharper image. In one recent
ations being another key factor, of course). And in many cases,
example, Lynn Enquist and colleagues at Princeton Univer-
sity used live-cell TIRF microscopy to image the exit of newly researchers can obtain sufficient quality data out of whatever
formed viral particles from cells (9). instrument they have available, even if it isn’t the top-of-the-line
A conceptually related idea is light-sheet microscopy. Here, hardware.
rather than illuminating the entire sample volume (as in bright- The good news for researchers is this: With so many tools
field and confocal microscopy), specimens are hit with a thin pla- in the toolbox, finding the right one for your imaging needs is
nar “sheet” of light from the side and imaged from an orthogonal easier than ever.
direction. By translating that sheet up and down, the system can
then capture 3D volumes repeatedly over time, producing a 4D
Jeffrey M. Perkel is a freelance science writer based in Pocatello, Idaho.
dataset. Commercial light-sheet implementations are available,
but researchers can build their own using the detailed instruc-
tions laid out by the OpenSPIM (selective plane illumination mi- References
croscopy) project (openspim.org). 1. K. Xu, G. Zhong, X. Zhuang, Actin, spectrin, and associated proteins
Some researchers, though, need to up the ante. In late 2015, form a periodic cytoskeletal structure in axons. Science 339, 452–456

junction in Drosophila melanogaster. Triple staining for the for instance, Philipp Keller and colleagues of the Janelia Re- (2013).

presynaptic active zone marker Brp (green), postsynaptic search Campus described a custom microscopy system called 2. M. M. Frigault et al., Live-cell microscopy—tips and tools. J. Cell Sci.

glutamate receptors (red), and the presynaptic membrane (blue).
Yale University described a method for two-color live-cell STED
imaging using HaloTag- and SNAP-tag-based fluorophores,
“IsoView,” which exploits two perpendicular light sheets to cap- 122, 753–767 (2009).
ture four separate images of a sample simultaneously. That strate- 3. J. B. Grimm et al., A general method to improve fluorophores for live-
gy is particularly useful when imaging relatively opaque samples, cell and single-molecule microscopy. Nat. Methods 12, 244–250
such as Drosophila embryos, and the team used it to (among oth- (2015).
er things) track cellular motion during fruit fly gastrulation over a 4. M. E. Dailey et al., Live-cell imaging techniques. Carl Zeiss Microscopy
Online Campus, available at http://zeiss-campus.magnet.fsu.edu/
3-hour period, collecting 3D volumes at 0.25 Hz (10).

which they used to differentiate the lumen and membrane of
the endoplasmic reticulum (6). And in 2012, Susan Cox and col-
IMAGES: (FROM TOP) COURTESY OF H. SHROFF, H. HESS, HHMI JANELIA FARM, ASHBURN, VIRGINIA, USA;
articles/livecellimaging/techniques.html.
Also at Janelia, in 2014 Betzig developed a higher-resolution
Light-sheet image of hepatocyte spheroids showing nucleus 5. A. Chmyrov et al., Nanoscopy with more than 100,000 ‘doughnuts.’
light-sheet-based strategy called “lattice light-sheet microscopy.”
(HNF4a/HNF3b) and cytoskeleton (CK18) staining. Nat. Methods 10, 737–740 (2013).
This approach differs from IsoView essentially in the thickness of

leagues of King’s College London developed a computational
strategy called “Bayesian analysis of blinking and bleaching”
(3B) that accelerated PALM to the point that the team could
IMAGE: COURTESY OF JOHN TNG WEIQUAN, NG HUCK HUI, GENOME INSTITUTE OF SINGAPORE
6. F. Bottanelli et al., Two-colour live-cell nanoscale imaging of
the generated sheets and the resolution of the resulting images,
intracellular targets. Nat. Commun. 7, 10778 (2016).
and Betzig used it to image subcellular events in 3D over extend-

COURTESY OF JAN PIELAGE, FRIEDRICH MIESCHER INSTITUTE, BASEL, SWITZERLAND

7. S. Cox et al., Bayesian localization microscopy reveals nanoscale
image structures called “podosomes” for about 100 seconds at ed periods, including mitosis in HeLa cells (11).

4-second resolution (7).
In general, however, the localization techniques are consid-
ered too slow for many live-cell applications (as many thou-
sands of “frames” must be collected to reconstruct a single
image), and often require light of such high intensity as to be
phototoxic in many instances. The one exception is SIM, thanks
to its speed and relatively gentle illumination conditions. But
SIM also yields the lowest resolution of any superresolution
method, about 100 nm.
In a 2015 report in Science detailing technical modifications
that increase SIM resolution to between 84 nm and 45 nm,
Janelia Research Campus investigator and Nobel Prize winner
Eric Betzig detailed the difficulty of applying superresolution
methods to live cells. Many such techniques, he wrote, “place
extraordinary demands on the photon budget, represented
by the product of the number of fluorescent molecules in the
specimen and the number of photons each can emit before
bleaching irreversibly. They also require specialized photo-
switchable labels and excitation intensities of 103 to 108 W/cm2,
which are orders of magnitude greater than the 0.1 W/cm2 un-
der which life evolved” (8).
W
Those researchers interested in imaging cells relatively deep hen it comes to microscopy, there is no one perfect
within tissue sections and even live animals may benefit from two- instrument for every occasion. Confocal microscopy
photon or multiphoton microscopy. Like confocal and light-sheet excels at removing out-of-focus light, but that may
microscopy, multiphoton microscopy creates thin optical sections not be absolutely necessary if you’re imaging only very thin
in tissue (12). But it does so via a different mechanism, requiring specimens. Superresolution microscopy yields magnificent
the fluorophore to almost instantaneously absorb two low-energy detail, but that really only matters if the objects being studied
photons to fluoresce rather than one high-energy photon. are small enough or close enough together to be otherwise
This strategy offers several potential benefits. First, because unresolvable.
multiple absorption events occur only in the focal plane, out-of- Microscopy, in other words, is a balancing act. In a 2015 sup-
focus fluorophores remain dark in this method, not to mention plement to Science, Betzig described that balance in terms of
unaffected by the incident light, thereby increasing signal-to- a tetrahedron, with vertices representing spatial resolution, cell
noise while minimizing photobleaching and phototoxicity. And, viability, imaging depth, and speed (14). Science writer Mike
because the technique uses lower-energy, higher-wavelength May, who interviewed Betzig, explains it thusly:
light, that light can penetrate deeper into the tissue than most Imagine a point inside that tetrahedron as a representation of
microscopy methods are capable of reaching—several hundred an imaging method’s combination of features. If it improves one
microns at least. That said, multiphoton microscopy traditionally feature, say spatial resolution, it moves closer to that vertex and
suffers from a relatively low scan speed, slowing image capture. away from the others, thereby reducing its capabilities in the re-
In one recent report, researchers at the University of Strathclyde maining three characteristics. To attain higher spatial resolution,
in Scotland describe a widefield-based image-capture strategy for example, an imaging method requires more pixels, and that
to circumvent that problem, which they used to image calcium requires more measurements, which takes more time and might
signaling events in rat neurons at up to 100 Hz (13). damage the cells by exposing them to more light.
podosome dynamics. Nat. Methods 9, 195–200 (2012).
8. D. Li et al., Extended-resolution structured illumination imaging
of endocytic and cytoskeletal dynamics. Science 349, aab3500
(2015).
9. I. B. Hogue et al., Cellular mechanisms of alphaherpes virus egress:
Live cell fluorescence microscopy of pseudorabies virus exocytosis.
PLOS Pathog. 10, 1004535 (2014).
10. R. K. Chhetri et al., Whole-animal functional and developmental
imaging with isotropic spatial resolution. Nat. Methods 12, 1171–1178
(2015).
11. B.-C. Chen et al., Lattice light-sheet microscopy: Imaging molecules to
embryos at high spatiotemporal resolution. Science 346, 1257998
(2014).
12. R. K. P. Benninger, D. W. Piston, Two-photon excitation microscopy
for the study of living cells and tissues. Curr. Protoc. Cell Biol. 59,
4.11.1–4.11.24 (2013).
13. R. Amor et al., Widefield two-photon excitation without scanning: Live
cell microscopy with high time resolution and low photo-bleaching.
PLOS ONE 11, e0147115 (2016).
14. M. May, Top tips from the superresolution microscopy pros, in
Microscopy now: Getting the most from your imaging, pp. 13–16.
Sponsored supplement to Science, Oct. 2 (2015).

6 sciencemag.org SCIENCE SCIENCE sciencemag.org 7

T
OIT
PTLIM
E IZING YOUR LI V E -CE L L M I CRO S CO P Y: TR I C K S A N D T R A D E - O F F S R ES E A RC H | R E S EA R C H A R T I C LE S WO || RESEARCH
R ES E A RC H SEEC
CTT IION
ON T
T WO RESEARCH ARTICLES
ARTICLES

◥
◥ MTs not anchored to other cytoskeletal or sarco- Confocal active frame, rest frame 0ms 1.0
◥ MTs not anchored to other cytoskeletal or sarco-
RE
RE SE
RESEA
SEAA RC
RC HH AA RT
RT IC
IC LL EE meric proteins could rearrange or slide passively
meric proteins could rearrange or slide passively
with the surrounding medium; anchored MTs 0.8
with the surrounding medium; anchored MTs

Pearson’s
could directly experience contractile force and
MICROTUBULES could directly experience contractile force and
MICROTUBULES themselves deform under load; or the MTs could subsarcolemmal SiR tubulin
themselves deform under load; or the MTs could 0.6
break and/or disassemble and reform. These pos- 35ms
break and/or disassemble and reform. These pos-
Detyrosinated
Detyrosinated microtubules
microtubules
sibilities offer divergent mechanisms for the reg-
sibilities offer divergent mechanisms for the reg-
ulation of mechano-signaling and the overall
ulation of mechano-signaling and the overall
0.4

mechanics of the myocyte. Without direct obser-

buckle
buckle and
and bear
bear load
load in
in
mechanics of the myocyte. Without direct obser-
vation, however, this behavior has been difficult
vation, however, this behavior has been difficult
to quantify.
interior
1µm

rest 70ms
0 1 2
Time (s)
3 4

to quantify.
contracting
contracting cardiomyocytes
cardiomyocytes
Standard confocal imaging, although capable
Standard confocal imaging, although capable
of resolving MTs in living cells (17), suffers from
of resolving MTs in living cells (17), suffers from
Airyscan rest

limitations in the signal-to-noise ratio when

limitations in the signal-to-noise ratio when
Patrick Robison,111 Matthew A. Caporizzo,222 Hossein Ahmadzadeh,222 Alexey I. Bogush,111 pushed to speeds that can resolve events on the
Patrick Robison, Matthew A. Caporizzo, Hossein Ahmadzadeh, Alexey I. Bogush, pushed to speeds that can resolve events on the
Christina Yingxian Chen,111 Kenneth B. Margulies,333 Vivek B. Shenoy,222 Benjamin L. Prosser111* time scale of cardiomyocyte contraction (Fig. 1B contract
Christina Yingxian Chen, Kenneth B. Margulies, Vivek B. Shenoy, Benjamin L. Prosser * time scale of cardiomyocyte contraction (Fig. 1B 105ms
and movie S1). Consequently, we turned to a high-
and movie S1). Consequently, we turned to a high- EMTB
The microtubule (MT) cytoskeleton can transmit mechanical signals and resist speed, subdiffraction limit technique called Airy-
The microtubule (MT) cytoskeleton can transmit mechanical signals and resist speed, subdiffraction limit technique called Airy- contract Amp
compression in contracting cardiomyocytes. How MTs perform these roles remains unclear scan (see fig. S1). This technology maintains high
compression in contracting cardiomyocytes. How MTs perform these roles remains unclear scan (see fig. S1). This technology maintains high
because of difficulties in observing MTs during the rapid contractile cycle. Here, we used signal-to-noise ratios at the required temporal
because of difficulties in observing MTs during the rapid contractile cycle. Here, we used signal-to-noise ratios at the required temporal 1µm
high spatial and temporal resolution imaging to characterize MT behavior in beating mouse resolution, while offering a 1.7-fold improvement
high spatial and temporal resolution imaging to characterize MT behavior in beating mouse resolution, while offering a 1.7-fold improvement
myocytes. MTs deformed under contractile load into sinusoidal buckles, a behavior in spatial resolution beyond the standard diffrac- 175ms
myocytes. MTs deformed under contractile load into sinusoidal buckles, a behavior in spatial resolution beyond the standard diffrac-
dependent on posttranslational “detyrosination” of a-tubulin. Detyrosinated MTs tion limit.
dependent on posttranslational “detyrosination” of a-tubulin. Detyrosinated MTs tion limit.
associated with desmin at force-generating sarcomeres. When detyrosination was reduced, Using the MT-binding fluorogenic dye SiR tu- ~ 1.65µm
associated with desmin at force-generating sarcomeres. When detyrosination was reduced, Using the MT-binding fluorogenic dye SiR tu- 30 2 µm

MT amplitude (µm)
MTs uncoupled from sarcomeres and buckled less during contraction, which allowed bulin, based on the fluorophore silicon rhodamine
MTs uncoupled from sarcomeres and buckled less during contraction, which allowed bulin, based on the fluorophore silicon rhodamine
sarcomeres to shorten and stretch with less resistance. Conversely, increased (18) (Fig. 1C and movie S2), we imaged internal 1µm
sarcomeres to shorten and stretch with less resistance. Conversely, increased (18) (Fig. 1C and movie S2), we imaged internal
detyrosination promoted MT buckling, stiffened the myocyte, and correlated with impaired MTs during contractions triggered by a 1 Hz elec-
detyrosination promoted MT buckling, stiffened the myocyte, and correlated with impaired MTs during contractions triggered by a 1 Hz elec- 20
function in cardiomyopathy. Thus, detyrosinated MTs represent tunable, compression- trical field stimulation in isolated mouse cardio-
function in cardiomyopathy. Thus, detyrosinated MTs represent tunable, compression-

resistant elements that may impair cardiac function in disease.
resistant elements that may impair cardiac function in disease.
long with its well-defined transport func-
modify their mechanical properties and binding
Counts
trical field stimulation in isolated mouse cardio- 3.3µm
myocytes. We were able to capture MT behavior ~ 3.3µm
myocytes. We were able to capture MT behavior
during contraction and found that longitudinally
during contraction and found that longitudinally
oriented MTs frequently deformed and developed 10
oriented MTs frequently deformed and developed
sinusoidal buckles. Because SiR tubulin may poly-

A
long with its well-defined transport func- modify their mechanical properties and binding sinusoidal buckles. Because SiR tubulin may poly-
tions, the microtubule (MT) network serves interactions. Detyrosination, a PTM of a-tubulin, merize MTs (18), we also generated adenovirus
tions, the microtubule (MT) network serves interactions. Detyrosination, a PTM of a-tubulin, merize MTs (18), we also generated adenovirus ~ 4.7µm
multiple mechanical roles in the beating has recently been shown to augment MT-dependent encoding a small fragment of the MT-binding pro-
multiple mechanical roles in the beating has recently been shown to augment MT-dependent encoding a small fragment of the MT-binding pro- 0
cardiomyocyte. MTs function as mechano- mechanotransduction in dystrophic cardiac and tein ensconsin fused to three copies of GFP (EMTB-
cardiomyocyte. MTs function as mechano- mechanotransduction in dystrophic cardiac and tein ensconsin fused to three copies of GFP (EMTB- 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
transducers, converting changing contract- skeletal muscle (12). This specific PTM is also in- 3xGFP) to decorate MTs. We achieved similar Buckle Wavelength, λ ( µm) rest contracted
transducers, converting changing contract- skeletal muscle (12). This specific PTM is also in- 3xGFP) to decorate MTs. We achieved similar 1.65µm
ile forces into intracellular signals (1, 2). MTs may creased in animal models of heart disease (1, 13, 14), results (table S1) but with improved signal-to-
ile forces into intracellular signals (1, 2). MTs may creased in animal models of heart disease (1, 13, 14), results (table S1) but with improved signal-to-
also act as compression-resistant elements, which which raises a mechanistic question: If mechano- noise ratios (Fig. 1D and movie S3).
also act as compression-resistant elements, which which raises a mechanistic question: If mechano- noise ratios (Fig. 1D and movie S3).
could provide a mechanical impediment to cardio- signaling is altered, have the mechanical properties We measured blindly selected MTs for deforma- Fig. 1. MTs reversibly buckle in contracting cardiomyocytes. (A) The sub- cycles. Initial drop to ~0.96 is due to imaging noise. (G) Quantification of
could provide a mechanical impediment to cardio- signaling is altered, have the mechanical properties We measured blindly selected MTs for deforma-
myocyte contraction (3–5). If so, they must bear of the myocyte changed? tion with two parameters—amplitude and wave- surface (top) and interior (bottom) cardiomyocyte MT network. (B) High- buckling amplitude (measured from centerline to edge) and l (wavelength
myocyte contraction (3–5). If so, they must bear of the myocyte changed? tion with two parameters—amplitude and wave-
some of the compressive and tensile load of a Although the idea that a proliferated (and per- length (Fig. 1G). Where possible, the same MT was speed confocal imaging of MTs at rest (top) and during contraction (bottom) measured as twice the distance between consecutive inflection points).
some of the compressive and tensile load of a Although the idea that a proliferated (and per- length (Fig. 1G). Where possible, the same MT was
working myocyte. Unfortunately, little is known haps modified) MT network may mechanically followed through contraction. Amplitude rose labeled with SiR tubulin with brightness increased for comparison with (C). (H) Amplitude of MTs labeled with EMTB-3xGFP in resting (black) and con-
working myocyte. Unfortunately, little is known haps modified) MT network may mechanically followed through contraction. Amplitude rose
about MT behavior during the contractile cycle. interfere with contraction is attractive, the “MT quickly from resting to contracted levels (Fig. 1H), (C) Airyscan imaging of the same MTs as in (B) at rest and during con- tracted (red) cardiomyocytes. The threshold to determine buckling occur-
about MT behavior during the contractile cycle. interfere with contraction is attractive, the “MT quickly from resting to contracted levels (Fig. 1H),
During this cycle, Ca2+ 2+
-mediated actin-myosin hypothesis” has remained controversial [for re- with clearly visible buckles. Using a threshold of traction. (D) Wider view of MTs labeled with EMTB-3xGFP at rest (top) and rence (blue line) was two standard deviations above the mean resting value.
During this cycle, Ca2+-mediated actin-myosin hypothesis” has remained controversial [for re- with clearly visible buckles. Using a threshold of
interaction first shortens repeating contractile views, see (15, 16)]. Two important limitations two standard deviations above resting amplitude, during contraction (bottom). (E) MTs imaged throughout a contractile (I) Distribution of buckling wavelengths in cardiomyocytes shows a dom-
interaction first shortens repeating contractile views, see (15, 16)]. Two important limitations two standard deviations above resting amplitude,
units called sarcomeres, which are then stretched have hindered our understanding: (i) a reliance we found that two-thirds of MTs buckle under cycle (cyan) were overlaid onto the network configuration from the initial inant population between 1.6 and 1.7 mm, and a second population at 3.3 mm.
units called sarcomeres, which are then stretched have hindered our understanding: (i) a reliance we found that two-thirds of MTs buckle under
as the heart fills with blood during diastole. on blunt pharmacological tools (colchicine/Taxol) control conditions (Fig. 1H). frame at rest (red). (F) Colocalization analysis of (E) shows that MTs re- (J) A representative MT demonstrating buckles with wavelengths that cor-
as the heart fills with blood during diastole. on blunt pharmacological tools (colchicine/Taxol) control conditions (Fig. 1H).
Although an isolated MT would present minimal that have off-target consequences; (ii) a lack of MT buckles quickly reversed during relaxation, peatedly return to the same position. Pearson’s coefficient is used to esti- respond to the distance between one (1.65 mm) or two (3.3 mm) adjacent
Although an isolated MT would present minimal that have off-target consequences; (ii) a lack of MT buckles quickly reversed during relaxation,
resistance to myocyte compression, the stiffness direct observation of MTs under the stress and and the configuration of the MT network between mate goodness of fit to original MT configuration over several contractile sarcomeres.
resistance to myocyte compression, the stiffness direct observation of MTs under the stress and and the configuration of the MT network between
of the network within a living cell, with MT- strain of the contractile cycle. Here, we charac- contractions tightly colocalized with the network
of the network within a living cell, with MT- strain of the contractile cycle. Here, we charac- contractions tightly colocalized with the network
associated proteins and other cytoskeletal ele- terized MTs under contractile loads using a configuration from previous cycles (Fig. 1, E and
associated proteins and other cytoskeletal ele- terized MTs under contractile loads using a configuration from previous cycles (Fig. 1, E and indicative of ordered
ordered geometric
geometric constraints
constraints on
on ily
ments, can change by orders of magnitude (6, 7). high-resolution imaging technique and directly F), with a minimal mean reduction in Pearson’s indicative of ily conserved
conserved across
across eukaryotes
eukaryotes (20)
(20) and
and appears
appears ing this
ing this construct
construct in in isolated
isolated cardiomyocytes
cardiomyocytes
ments, can change by orders of magnitude (6, 7). high-resolution imaging technique and directly F), with a minimal mean reduction in Pearson’s the buckling
buckling MT.
It is in this context that MTs are proposed to act tested how MT detyrosination may regulate colocalization of 0.01 per contractile cycle (n = the MT. This
This was
was observed
observed inin certain
certain required
required for for life
life (21),
(21), yet
yet its
its functional roles are
functional roles are could effectively
could effectively reduce
reduce the the level
level of
of detyrosina-
detyrosina-
It is in this context that MTs are proposed to act tested how MT detyrosination may regulate colocalization of 0.01 per contractile cycle (n = cells
as compression-resistant elements (6, 8) that may load-bearing and the mechanical properties of 18 runs). The rapid and precise reversibility of the cells where faint transverse
where faint transverse staining
staining at
at the
the Z-disc
Z-disc still
still poorly
poorly understood.
understood. Because
Because detyrosination
detyrosination tion as
tion as shown
shown by by both
both immunofluorescence
immunofluorescence (Fig. (Fig.
as compression-resistant elements (6, 8) that may load-bearing and the mechanical properties of 18 runs). The rapid and precise reversibility of the shows
impair sarcomere shortening and thus cardiac the myocyte. network deformations suggested tight coupling shows MTs
MTs buckling
buckling between
between sarcomeric con-
sarcomeric con- can
can protect
protect MTsMTs from
from disassembly
disassembly (22, 23) and
(22, 23) and 2, A
2, A and
and B)B) and
and immunoblot
immunoblot (Fig. (Fig. 2,
2, C
C and
and D),D),
impair sarcomere shortening and thus cardiac the myocyte. network deformations suggested tight coupling straints
function, particularly in disease states associated to the contractile apparatus, and it argues against straints (movie
(movie S4).
S4). can
can facilitate
facilitate their
their cross-linking
cross-linking with
with intermediate
intermediate which resulted
which resulted inin aa 71%
71% reduction
reduction inin the
the amount
amount
function, particularly in disease states associated to the contractile apparatus, and it argues against filaments
with MT proliferation (8–10). Posttranslational MTs buckle under contractile load
MTs buckle under contractile load MT breakage and regrowth contributing to me- Detyrosination filaments (IFs)(IFs) (24, 25), we
(24, 25), we hypothesized
hypothesized thatthat the
the of polymerized,
of polymerized, detyrosinated
detyrosinated MTs, MTs, with
with aa con-
con-
with MT proliferation (8–10). Posttranslational
modification (PTM) of MTs (11, 12) could also MT networks in cardiomyocytes have two major
MT breakage and regrowth contributing to me-
chanical properties and signaling over the time Detyrosination regulates
regulates MT
MT buckling
buckling in
in high
high proportion
proportion of of detyrosination
detyrosination maymay confer
confer the
the comitant up-regulation
comitant up-regulation of of tyrosinated
tyrosinated tubulin
tubulin
modification (PTM) of MTs (11, 12) could also MT networks in cardiomyocytes have two major chanical properties and signaling over the time the heart
the heart resilient
features (Fig. 1A): an orthogonal grid just beneath scale of myocyte contraction. resilient load-bearing
load-bearing capabilities
capabilities ofof the
the cardiac
cardiac (Fig. 2,
(Fig. 2, C
C and
and D,
D, and
and fig. S2). Overexpression
fig. S2). Overexpression of of
features (Fig. 1A): an orthogonal grid just beneath scale of myocyte contraction. This
the membrane that wraps the myofibrils and a A notable feature of the MT buckles was the This robust
robust buckling
buckling behavior of the
behavior of the MT
MT network
network cytoskeletal
cytoskeletal network.
network. TTL
TTL also resulted in
also resulted in aa modest
modest (10%)
(10%) reduction
reduction in in
the membrane that wraps the myofibrils and a A notable feature of the MT buckles was the may
11 deeper network composed primarily of longitu- emergence of subpopulations of buckle wavelength may be be aa result
result of
of aa particularly
particularly high
high abundance
abundance Using
Using antibodies
antibodies specific
specific to
to detyrosinated
detyrosinated a- a- the density
the density of of the
the polymerized
polymerized MT MT network
network (Fig.
(Fig.
1Department
Department ofof Physiology,
Physiology, Pennsylvania
Pennsylvania Muscle
Muscle Institute,
Institute, deeper network composed primarily of longitu- emergence of subpopulations of buckle wavelength
Department Pennsylvania
University
University of
of Physiology, Pennsylvania
of Pennsylvania Perelman
Perelman School
Muscle
School of
Institute,
of Medicine,
Medicine, dinal elements that interdigitate the myofibrils. centered at ~1.65 mm, 3.3 mm, and perhaps even of “detyrosinated”
of “detyrosinated” MTs MTs in in adult
adult cardiomyocytes
cardiomyocytes tubulin,
tubulin, we we found
found aa high
high abundance
abundance of of detyrosi-
detyrosi- 2B), consistent
2B), consistent with
with an an increased
increased disassembly
disassembly of of
dinal elements that interdigitate the myofibrils. centered at ~1.65 mm, 3.3 mm, and perhaps even (19).
University of PA
Philadelphia, Pennsylvania
19104, USA.Perelman
22 SchoolofofMaterials
Department
Philadelphia, PA 19104, USA. 2Department of Materials
Medicine, Longitudinal MTs often run many sarcomeres in 4.7 mm (Fig. 1I). These corresponded closely to the (19). Detyrosination
Detyrosination is is aa PTM
PTM of a-tubulin where
of a-tubulin where nation
nation in the a-tubulin
in the a-tubulin network
network of adult myocytes
of adult myocytes tyrosinated MTs
tyrosinated MTs (22, 23). We
(22, 23). We complemented
complemented this this
PA 19104, USA. Department
Philadelphia,Engineering, of Materials Longitudinal MTs often run many sarcomeres in 4.7 mm (Fig. 1I). These corresponded closely to the the
Science
Science and
and Engineering, University
Science and Engineering,
University of
of Pennsylvania
Pennsylvania School
University Philadelphia,
School
of Pennsylvania School length but do not span the full cell. Given that length of one, two, or three contracted sarcomeres, the C-terminal
C-terminal tyrosine residue has
tyrosine residue has been
been cleaved
cleaved (Fig.
(Fig. 2, A and B), as expected (12, 19). To test
2, A and B), as expected (12, 19). To test the
the genetic
genetic strategy with aa pharmacological
strategy with approach
pharmacological approach
length but do not span the full cell. Given that length of one, two, or three contracted sarcomeres, by
of
of Engineering
Engineering and
of Engineering
33
and Applied
Applied Science,
andofApplied
Science, Philadelphia, PA
Science, Philadelphia,
PA 19104,
19104,
PA 19104, cardiomyocytes change shape during contraction, respectively. Although MTs buckling outside of by aa tubulin
tubulin carboxypeptidase (TCP); this
carboxypeptidase (TCP); this process
process role
role ofof detyrosinated
detyrosinated MTs,MTs, wewe generated
generated adeno-
adeno- to
to inhibit TCP using
inhibit TCP using parthenolide
parthenolide (PTL)(PTL) (26).
(26). PTL
PTL
USA. Department Medicine, University of Pennsylvania
USA. 3Department of Medicine, University of Pennsylvania cardiomyocytes change shape during contraction, respectively. Although MTs buckling outside of can
USA. Department
Perelman ofMedicine,
Medicine,Philadelphia,
University ofPAPennsylvania
Perelman School of Medicine, Philadelphia, PA 19104, USA.
School of 19104, USA. the MT cytoskeleton must accommodate this these populations could be found in our data with- can be readily reversed by tubulin tyrosine ligase
be readily reversed by tubulin tyrosine ligase virus
virus encoding
encoding TTL (AdV-TTL) with
TTL (AdV-TTL) with aa Discosoma
Discosoma treatment also reduced the fraction
treatment also reduced the fraction of detyrosi-of detyrosi-
the MT cytoskeleton must accommodate this these populations could be found in our data with- (TTL)
Perelman School of Medicine, Philadelphia, PA 19104, USA.
*Corresponding author. Email: bpros@mail.med.upenn.edu
*Corresponding author. Email: bpros@mail.med.upenn.edu change. There are three apparent possibilities: out difficulty, these subpopulations were strongly (TTL) (11).
(11). This
This tyrosination
tyrosination cycle
cycle is
is evolutionar-
evolutionar- red fluorescent
red fluorescent protein
protein (DsRed)
(DsRed) reporter.
reporter. Express-
Express- nated MTs, albeit
nated MTs, albeit to
to aa lesser
lesser extent (42%) than
extent (42%) than
*Corresponding author. Email: bpros@mail.med.upenn.edu change. There are three apparent possibilities: out difficulty, these subpopulations were strongly

8
8 Originally
Originally published
published 22on
online April 2016 in
22 April SCIENCE
2016 in SCIENCE sciencemag.org SCIENCE
sciencemag.org
sciencemag.org SCIENCE
SCIENCE aaf0659-2
SCIENCE sciencemag.org
22 APRIL 2016 • VOL 352 ISSUE 6284 9
sciencemag.org SCIENCE
SCIENCE sciencemag.org 22 APRIL 2016 • VOL 352 ISSUE 6284 aaf0659-1 SCIENCE sciencemag.org 9
OIT
T PTLIM
E IZING YOUR LI V E -CE L L M I CRO S CO P Y: TR I C K S A N D T R A D E - O F F S RE S E ARCH | R E S E A R C H A R T I C L E R ES E A RC H | R E S EA R C H A R T I C LE S E C T IION
O N T WO | RESEARCH ARTICLES

Ctrl TTL Ctrl TTL ability of detyrosinated MTs to bear compres-

1.80 0.40
sive load.
***

Sarcomere Length (µm)

0.30 40
0.35 1.9
1.70 ** Detyrosination regulates myocyte
***
0.30 30
mechanical properties

∆SL (µm)
1.8 0.20
We next measured mechanical resistance directly
1.60
0.25 20 using atomic force microscopy (AFM). AFM mea-
1.7
0.10
surements of transverse stiffness were performed
0.20
1.50 1.6 10 across a range of indentation rates. Myocyte stiff-
0.15
ness changed substantially with indentation rate
1.40
1.5 0.00 0 and was well fit by a standard linear solid model
0.0 1.0 2.0 3.0 Resting SL ∆SL Shortening
0 1 2 3 4 Log TTL Fluorescence time (ms)
(SLSM) [methods in supplementary materials (SM)]
Time (s) (µm) (µm)
(fig. S4 and Fig. 4A). At low rates (100 nm/s), the
Ctrl ** DMSO **
stiffness of the cardiomyocyte was essentially elas-
4 4 TTL 4 PTL tic, reported as E1, and was reduced by PTL treat-

Contractile Velocity (µm/s)

ment and TTL overexpression (Fig. 4B). At higher

Peak Velocity (µm/s)

2

Peak Velocity (µm/s)

2 2 rates, the modulus increased by E2, which reflects
0 cross-linked material inside the cell that slips on
0 0
-2
the time scale of slower measurements but “turns
-2 -2 on” (stiffens) at faster time scales (> 2 mm/s) (28).
-4 The viscosity derived by the SLSM defines the
Ctrl -4 -4 rate above which these cross-links are engaged.
-6
TTL TTL overexpression significantly decreased E2 and
-8 -6 -6 viscosity (Fig. 4B), which suggested that reducing
*** ***
0 1 2 3 4 Contraction Relaxation Contraction Relaxation detyrosination decreases the number of cross-
Time (s) links engaged at physiological strain rates in the
6 0.20
Ctrl cardiomyocyte.
Ctrl TTL
6 The fact that MTs deform under load and re-
5 TTL

Ca2+ decay Tau (s)

0.15 sist sarcomere shortening implies a transfer of

[Ca2+]i (F/F0 )

Peak Ca2+ (F/F0 )

4 force between MTs and the sarcomere. If MTs
4
0.10 resist longitudinal compression, they could also
3 confer a tensile resistance when the sarcomeres
2
are stretched, as occurs during diastolic filling. To
2 0.05 test this idea, we measured passive stiffness direct-
1 ly along the longitudinal axis of TTL-overexpressing
0 0.00 myocytes. We attached cardiomyocytes to laser-
0 1 2 3 4 Null TTL Null TTL etched cell holders (Fig. 4C, fig. S5, and movie S7)
Time (s)
via a biological adhesive (1) and subjected them
Fig. 3. Detyrosinated MTs impede contractility. (A) Sarcomere shortening (DSL) during contraction is
to steplike changes in length, while simultaneously
increased in TTL-overexpressing myocytes. This change is (B) dose-dependent [P = 1.2 × 10−5, correlation
measuring sarcomere length and passive force
coefficient (r2) = 0.23] and (C) associated with a faster shortening time without affecting resting
with a high-sensitivity transducer (Fig. 4D). A typ-
ical force response (Fig. 4D, blue) showed a rapid
sarcomere length. (D) First derivative of traces in (A) demonstrate contractile velocities in control and TTL-
rise to peak force during the high-velocity stretch
overexpressing myocytes. (E) TTL-overexpressing myocytes demonstrate an increase in the peak velocity of
(Fpeak), containing both elastic and viscous ele-
both contraction and relaxation. (F) PTL-treated cells display similar behavior. (G to I) Despite the significant
ments, followed by a relaxation to a steady-state
changes in contractility, no changes in the peak or kinetics of the global calcium transient were observable.
force (Fs.s.) that largely reflects the elastic stiff-
F/F0,the change in fluorescence from the original fluorescence before stimulation. Data are presented as
means ± SEM; *P < 0.05, **P < 0.01, ***P < 0.001. Further statistical details are available in table S2.
ness of the myocyte. For a given step size, TTL-
overexpressing myocytes exerted significantly
Fig. 2. Detyrosination underlies MT buckling. (A) The MT cytoskeleton tractile axis rather than deforming. (F) Comparison of MTs in resting (top) and reduced peak forces during physiological length
(blue and aqua) in rat adult cardiomyocytes (top) is heavily detyrosinated contracted (bottom) cardiomyocytes in control, TTL, and PTL groups. In TTL When MT buckling was observed, the mean ingful resistance to myocyte contraction. We thus changes, with modest reductions in steady-state
(orange). TTL overexpression (bottom) reduces detyrosination dramatically but and PTL groups, some MTs slide (orange arrows) relative to others that deform wavelength was not significantly different between tested directly if MT detyrosination affects contrac- force (fig. S5F). The TTL-overexpressing cells also
makes comparatively small changes in the overall MT network. (B) Quantifi- (white arrows). Additional examples are provided in fig. S3. (G) Buckling occur- control and TTL-overexpressing cells (table S1). tility in beating cardiomyocytes. After overexpression underwent significantly larger changes in sarco-
cation of the fraction of cell area covered by a-tubulin and detyrosinated rence and amplitude are reduced by overexpression of TTL or treatment with However, the majority of MTs in TTL-overexpressing of TTL, we found significant enhancements in mere length with any given step (fig. S5G), which
tubulin (dTyr-tubulin) in null (n = 14) and TTL-overexpressing (n = 13) cells, as PTL. (H) Buckling wavelength distribution in control and TTL-overexpressing myo- myocytes no longer buckled at the wavelength of both the magnitude (Fig. 3, A to C) and peak rate indicated increased sarcomere compliance and
determined from thresholded images as shown in fig. S2E. (C and D) Western cytes and (I) the difference between these distributions. Overexpression of TTL a single sarcomere, and no subpopulations at (Fig. 3, D and E) of sarcomere shortening (table which suggested that stiffer sarcomeres in con-
blots from cardiomyocyte lysates show the effects of viral overexpression of causes MTs to buckle more often at wavelengths between 2 and 3 mm, and MTs multiples of the sarcomeric period were observed S2). PTL had a similar effect on contractility (Fig. trol cells distribute the length change to other
TTL. GAPDH, glyceraldehyde-3-phosphate dehydrogenase. (E) Time course of are far less likely to buckle at distinct sarcomeric wavelengths (1.7 and 3.3 mm) (Fig. 2H). Instead, the majority of these MTs 3F and table S2). Peak relaxation rates were also compliant components in series. As can be sur-
MTs in a contracting cardiomyocyte (cyan) transduced with AdV-TTL overlaid on when detyrosination is reduced. Data are presented as means ± SEM; *P < 0.05, buckled in a single population at wavelengths increased, which could be because of a decrease mised from fig. S5, F and G, TTL overexpression
the resting MT configuration (red). MTs appear to translocate along the con- **P < 0.01, ***P < 0.001. Further statistical details are available in table S1. between 2 and 3 mm, which suggested that MT in cellular viscosity (see below) or may reflect the decreased tension across the physiological range
buckling was less constrained by a sarcomeric increased magnitude of shortening and, there- of sarcomere lengths achieved during diastolic
interaction after detyrosination was reduced fore, compression of internal elastic elements filling (Fig. 4E), which indicated a role for de-
AdV-TTL and with no effect on MT network simply slide in the moving cell (Fig. 2, E and F, (P = 0.87, n = 19 runs). The occurrence of buckling (Fig. 2I). (e.g., titin) that develop restoring force (27). These tyrosinated MTs as tensile-resistant elements.
density (fig. S2). orange arrows; movies S5 and S6; and fig. S3), in TTL-overexpressing and PTL-treated cells fell contractile changes were not associated with Visual evidence supporting such a relation was
The load-bearing behavior of the MT network rather than buckling under load. This behavior significantly (Fig. 2G, left), whereas amplitude Detyrosinated MTs resist any significant difference in global calcium tran- seen in MT networks in a control cell at resting
in cardiomyocytes overexpressing TTL or treated was again reversible, with a minimum reduction changes observed on the same MT between rest contractile compression sients (Fig. 3, G to I) or resting sarcomere length and stretched length (fig. S5 and Fig. 4C, cyan).
with PTL was dramatically different from control in Pearson’s colocalization over successive con- and contraction also dropped significantly (Fig. The energy required to deform detyrosinated MTs (Fig. 3C), which suggested a change in intrin- At resting length, there was some inherent slack in
myocytes. Tyrosinated MTs frequently seemed to tractions that was not different from controls 2G, right, and table S1). under compressive load could confer some mean- sic mechanical resistance associated with the the MT network, whereas the same MTs became

10
SCIENCE sciencemag.org sciencemag.org
22 APRIL 2016 • VOL 352 ISSUE 6284 SCIENCE
aaf0659-3 SCIENCE sciencemag.org
aaf0659-4 22 APRIL 2016 • VOL 352 ISSUE 6284 11
sciencemag.org SCIENCE
OIT
T PTLIM
E IZING YOUR LI V E -CE L L M I CRO S CO P Y: TR I C K S A N D T R A D E - O F F S RE S E ARCH | R E S E A R C H A R T I C L E R ES E A RC H | R E S EA R C H A R T I C LE S E C T IION
O N T WO | RESEARCH ARTICLES

800 Cell Holder Cell Holder 20 Sarcomere Length (L)

Ctrl DMSO

Step (µm)
ctrl MT titin

MT
PTL (Force) (Length) 10 ε δ
TTL

rceM
Force
0
TTL medium z-disk
Elastic Modulus (Pa)

600 E2 2.3

SL (µm)
actin myosin actin
2.1 contractile element
E1 1.8 sarcomeric anchor
400
0 = 0.1 µm
0.4 Fpeak
titin
rest

Force (µN)
200 Fs.s.
0.2
medium

5 µm 0
0 anchor anchor
0.1 1 10 100 Resting Sarcomere Length - 1.84µm 0 10 20 30 40 50 60
Velocity ( µm/s) Time (s)
Stretch 0.5 microtubule
Ctrl - F peak
E1(Pa) E2 (Pa) Viscosity (Pa*s)
500 400
TTL - F peak γ/2 ξ γ/2
100 0.4
Ctrl - F s.s.
400 TTL - Fs.s. contracted

Force (µN)
300 80 0.3 0.0
0.15

Buckling Amplitude ( µm)

**
300 ***
60
200 0.2
** * -0.1 ctrl ctrl

∆ SL (µm)
200 ** *** TTL
0.10
TTL
40
0.1 ***
100
100 20
***
-0.2 0.05
0.0

0 0 0 Stretch Sarcomere Length - 2.07µm
Ctrl TTL DMSO PTL Ctrl TTL DMSO PTL
Fig. 4. Detyrosinated MTs regulate the viscoelasticity of cardiomyocytes.
(A) Elastic modulus of cardiomyocytes measured by AFM at various inden-
tation velocities and fit to SLSM (see methods in SM). (B) Quantification of
velocity-independent (E1) and velocity-dependent (E2) components of the
elastic modulus, and SLSM fit–derived viscosity. Both TTL overexpression and
PTL treatment significantly reduced elasticity and viscosity. There were no
significant differences in these parameters between dimethyl sulfoxide (DMSO)
(gray) and AdV-null (black) transduced cells (P = 0.28, 0.34, and 0.33, re-
spectively). Reductions in stiffness due to TTL overexpression are also apparent
in cells under stretch along the longitudinal axis. (C) Myocytes were attached
via glass cell holders (C, top, and fig. S5) to a force transducer and length
taut when the cell was stretched and held at long
sarcomere lengths (movie S8).
Model of MT contribution to
cardiac contractility
We next sought to develop a mathematical mod-
el to recapitulate the experimentally measured
changes in MT buckling and contractility when
detyrosination is reduced. Previous work mod-
eling MT buckles (6) suggests that three critical
variables determine buckling behavior; MT stiff-
ness, stiffness of the surrounding medium, and
force incident on the long axis of the MT. How
these three variables are predicted to alter MT
behavior and myocyte mechanics is described in
fig. S6. Of the three, only a decrease in incident
force can explain our experimental observations
after suppressing detyrosination. If MT anchor-
ing to the sarcomere is disrupted, the reduced
incident force on the MT may drop below the
critical force required for buckling, which results
in simultaneous decreases in buckling amplitude
(Fig. 2) and viscoelasticity (Figs. 3 and 4). The
1.8 1.9
Ctrl TTL DMSO PTL
controller and were subjected to stretch. MTs visualized by EMTB-3xGFP
(C, blue and aqua) at rest (top) and at a stretched length (bottom). (D) Rep-
resentative force versus length protocol. A series of stepwise stretches (red) in
4-mm increments are applied to an isolated myocyte, which increases sarco-
mere length (SL, black). Passive tension (blue) generated by the step relaxes
quickly from a peak value to a new steady state. (E) Force measurements
binned according to measured change in sarcomere length with a given step
size. TTL-overexpressing cells exert reduced peak passive tension during step
changes in length, with a more modest reduction in steady-state tension. Data
are presented as means ± SEM; *P < 0.05, **P < 0.01, ***P < 0.001. Further
statistical details are available in tables S3 and S4.
sarcomeric periodicity of buckles (Fig. 1I) also increase the average distance between detyrosi-
suggests an underlying structural constraint that nated tubulins that could interact strongly with
changes in MT or medium stiffness alone cannot MT anchoring points. This disruption of MT-
explain. We thus chose to model MT buckling sarcomere coupling produced model outputs (Fig.
within a contractile model that includes a MT 5, C and D) that closely recapitulated our experi-
compression-resistive element that’s interaction mental contractility and buckling results.
with the sarcomere can be varied (see model in An alternate possibility to the sliding anchor is
SM for details). that the anchor is completely uncoupled by sup-
Using the mechanical scheme detailed in pressing detyrosination and reverts to buckling
Fig. 5A, we fitted the contraction resulting from a behavior governed by local viscoelasticity rather
log-normal force input to derive both contractile than underlying structure, as proposed for less
and buckling parameters. By modifying the in- rigidly organized cell types, including developing
cident force applied to a MT for a given amount myocytes (6). In either case, the coupling of MTs
of sarcomere shortening (Fig. 5A and model in to the sarcomere is reduced, which impairs their
SM), we simulated the effect of a sarcomeric an- ability to resist contraction.
chor sliding and then catching at detyrosinated
regions of the MT. Inclusion of a 100-nm slide Potential role for desmin as a
(50 nm at each anchor, see model) before MTs sarcomeric MT anchor
engage with the rest of the sarcomere is rea- The putative characteristics of the anchor—a me-
sonable, given the ~80% reduction in detyrosi- chanically stiff protein capable of forming complexes
nated area observed by immunofluorescence with MTs and restricted to a spatially defined
with TTL overexpression (Fig. 2C), and reflects region of the sarcomere—suggested the interme-
the fact that reductions in detyrosination would diate filament desmin as an immediate candidate.
2.0 2.1 2.2 2.3
SL (µm) -0.3 0.00
contracted, detyrosination
0.0 0.2 0.4 0.0 0.2 0.4
Time (s) Time (s)
null TTL
Fig. 5. Modeling MTs in the contracting sarcomere. (A) Mechanical sche- sarcomere length at peak contraction and buckling amplitude. (C) and
matic of the modeled sarcomere. A force-generating contractile arm (top) is (D) recapitulate experimental observations for TTL-overexpressing myocytes
coupled in parallel at the Z-disc to a spring element representing titin (orange), after this change. (E and F) The cardiac sarcomere, shown with MTs with
a viscoelastic medium (yellow spring and dashpot), and a MT (green) with putative stiff anchors to the sarcomere, here, at the Z-disc. Contraction
anchors (fuschia pink) to the Z-disc (gray). The anchor to the Z-disc is only reduces the distance between anchor points, which requires the MTs either to
engaged at regions of MT detyrosination. (B) TTL overexpression is modeled buckle (G) if the anchors are engaged or to slide (H), if the anchors are not
by allowing the anchors to slide for 50 nm at each end before engaging and engaged and force incident on the MT remains low. Mathematical model
transmitting force to the MT at a detyrosinated subunit. (C) The change in parameters are available in table S5.
Desmin forms structural bundles that form a The two populations of MTs show similar overall removal of desmin should both decrease cyto-
complex with the Z-disc (29), and intermediate patterning in WT myocytes, except for a specific skeletal stiffness and prevent tyrosination-
filaments can form detyrosination-dependent accumulation of detyrosinated (and not tyrosi- dependent changes in viscoelasticity. Blind studies
cross-links with MTs (30, 31). nated) tubulin in transverse bands at the Z-disc in WT and KO myocytes revealed that desmin
We first sought to determine whether desmin that colocalized with desmin (Fig. 6, D and E, KO myocytes were significantly less stiff than WT
preferentially associates with detyrosinated MTs. and fig. S7). Note that KO animals lacked this counterparts (Fig. 6, F and G), and that treatment
Cosedimentation of cardiomyocyte lysates showed transverse pattern completely (Fig. 6D and fig. with PTL no longer reduced viscoelasticity (Fig. 6,
that desmin forms pellets with polymerized MTs S7H), although the Z-disc itself remained intact F and G).
(Fig. 6A) in direct proportion to their level of (fig. S7C). In addition, KO myocytes had a denser
detyrosination (Fig. 6, B and C), which indicated (fig. S7F) and more disorganized MT network MT detyrosination is sufficient to impair
a specific and sensitive interaction. We also co- (Fig. 6D and fig. S7, B and E), which suggested cardiomyocyte contractility
stained cardiomyocytes from desmin-depleted that desmin is required for proper MT network Increasing detyrosination correlates with impaired
[knockout (KO)] and wild-type (WT) mice for organization. function in animal models of heart disease (13, 14).
desmin and both tyrosinated and detyrosinated If desmin cross-links with detyrosinated MTs Thus, we next tested whether increasing detyr-
tubulin to observe any preferential interaction. to structurally reinforce the network, then the osination could directly impair cardiac contractility.

12
SCIENCE sciencemag.org sciencemag.org
22 APRIL 2016 • VOL 352 ISSUE 6284 SCIENCE
aaf0659-5 SCIENCE sciencemag.org
aaf0659-6 22 APRIL 2016 • VOL 352 ISSUE 6284 13
sciencemag.org SCIENCE
O PT IM IZING YOUR LI V E -CE L L M I CRO S CO P Y: TR I C K S A N D T R A D E - O F F S RE S E ARCH | R E S E A R C H A R T I C L E R ES E A RC H | R E S EA R C H A R T I C LE S E C T I O N T WO | RESEARCH ARTICLES

Fig. 6. Desmin associ-

ates with detyrosinated
DMSO PTL A B 1200 C ctrl shTTL
MT Pellet Super 1.1 myocytes

Normalized ratio in MT pellet

MTs to increase cardio- 1.0 600 140
PTL DMSO PTL DMSO C2C12s AdV-Scramble AdV-shTTL ** **
1000 ***
myocyte stiffness. (A) 1.0 ctrl 500

desmin in MT pellet
b 120
MT cosedimentation -tu 0.8 L shTTL 500
yr TT
shows the interaction dT *** *** 0.9 800 400

Modulus (Pa)
100
between polymerized b 400
tu H

Modulus (Pa)
0.6

Modulus (Pa)
MTs (pellet) and desmin. α- 0.8 PD
GA 600 80

Pa*s
slope = 1.21 300
(B) Quantification of the in 300
sm 0.4 0.7 r2 = 0.88 tub 60
amount of detyrosination
de yr- 400
and desmin (relative to dT 200
200
40
the total amount of P DH 0.2 0.6
tub
α-
tubulin) in the MT pellet GA 0.5
200 100 100
20
from cardiomyocyte 0.0 0.5 0.6 0.7 0.8 0.9 1.0
lysates with and without dTyr desmin dTyr in MT pellet 0 0 0 0
PTL treatment. Data
0.1 1 10 100 E1 E2 Viscosity
Des KO desmin Velocity (µm/s)
were normalized to WT desmin
D E 0.3 F

desmin
DMSO level. (C) The ctrl shTTL Donor Hypertrophy DCM HCM
amount of desmin asso-
ciated with the MTs after 5 ***
*** dTyr-tub

Peak velocity (µm/s)

2µm
PTL treatment is directly
0.2
proportional to the
α-tub
amount of MT detyrosi-

dTyr-tub

0.1 µm
5µm
nation across several 0
GAPDH
experiments in rat cardi- dTyr-tub dTyr-tub
omyocytes and C2C12 0.1
TTL
cells. (D) Immuno- ctrl
fluorescence of desmin, shTTL -5 tyr-tub
***

merge
dTyr, and Tyr-tubulin 1s
shows dTyr-specific 0.0 GAPDH
contraction relaxation SL (µm)
transverse pattern in WT,
but not desmin KO, myo-
G 3.5
α-tubulin
H
cytes. (E) Overlay of dTyr- tyr-tub tyr-tub
tubulin and desmin. (See D 0.3 dTyr:des
dTyr-tubulin *** 80

Mander’s coefficient
tyr:des
fig. S7 for more 3.0 tyr-tubulin
* Healty donor

Fold change (norm. to donors)

examples.) (F and G) AFM 70
0.2 ***
TTL *** HCM
measurements show a Linear Fit
2.5 60
PTL-dependent reduction

LVEF (%)
0.1
in myocyte stiffness and
viscosity in WT, but not 2.0 * 50

desmin, KO myocytes. 0.0 *** ** 40 p = 1.056 x 10-6

Viscoelasticity in desmin
2000 WT E1(Pa) 1200 E2 (Pa) Viscosity (Pa*s) 1.5 * * *** r2 = 0.59
KO myocytes is not sta- 800
tistically different from WT
WT+PTL 120 * 30
Des KO 1000
with PTL treatment. Data 1.0 20
1500 Des KO +PTL 600 100
are presented as means ±
Modulus (Pa)

800
SEM; *P < 0.05, **P < * p = 0.2 10
80 0.5
0.01, ***P < 0.001 with
### 600 ###
respect to DMSO treat- 1000 400 ###
60 n=9 n = 17 n = 10 n = 10 n = 15
ment; ###P < 0.001 with n = 17 0 1 2 3 4 5 6 7
0.0
respect to untreated WT 400 *** dTyr-tubulin (fold change)
40 Healthy Hypertrophy DCM Ischemic PostVad HCM
myocytes. Further statis- 500 200 Donors DCM
tical details are available 200 20
in table S6. Fig. 7. Increasing detyrosination impairs contraction and is associated human heart lysates. (G) Data from pooled analysis; n = 17 healthy donors,
0 0 0 0 with human heart failure. (A) Western blot shows that shRNA against 9 hypertrophy, 17 DCM, 11 ischemic, 10 with DCM following ventricular
0.1 1 10 100 WT WT KO KO WT WT KO KO WT WT KO KO
+PTL +PTL
TTL selectively increases dTyr-tubulin without changing overall levels of assist device support (VAD DCM), and 15 HCM hearts. (H) There was a
Velocity ( µm/s) +PTL +PTL +PTL +PTL
a-tubulin. (B) Elastic modulus of control and shTTL-expressing myocytes negative correlation between LVEF and dTyr-tubulin expression in control
at various indentation rates. (C) shTTL myocytes demonstrate increases and hypertrophic cardiomyopathy patients. Data are presented as means ±
in E1, E2, and viscosity. (D and E) TTL suppression significantly reduces SEM; *P < 0.05, **P < 0.01, ***P < 0.001. Further statistical details are
Using an adenoviral construct expressing short and suppressed contractile velocity and mag- S7). Detyrosinated tubulin was significantly in- contractile magnitude and velocity. (F) Representative Western blots from available in tables S7 to S9.
hairpin RNAs (shRNAs) against TTL (shTTL), we nitude (Fig. 7, D and E). creased in patients with clinically diagnosed
suppressed TTL expression, which enhanced We next examined whether this modification hypertrophic and dilated cardiomyopathies (HCM
detyrosination (Fig. 7A and fig. S8). shTTL- correlated with functional deficits in human heart and DCM, respectively), along with a modest in- indicator of cardiac contractility (Fig. 7H). gested a specific link between detyrosination TTL was unchanged in all patient populations,
transduced myocytes were then tested for their disease. To this end, we analyzed left ventricular crease in total tubulin content (Fig. 7, F and G). There was no such correlation detected be- and LVEF. Myocardium from patients with which showed that a decreased expression of the
viscoelastic and contractile properties. The ex- tissue samples from healthy patient donors and Blind analysis of HCM patient data showed that tween LVEF and total or tyrosinated tubu- DCM all demonstrated considerably depressed tyrosinating enzyme does not explain the in-
cess detyrosination alone was sufficient to in- from patients exhibiting varying degrees of heart detyrosination inversely correlated with left ven- lin levels, nor any correlation between heart LVEF and variable, but increased, detyrosi- crease in detyrosinated tubulin in patients with
crease viscosity and stiffness (Fig. 7, B and C) disease due to several underlying causes (table tricular ejection fraction (LVEF), a primary weight and detyrosination (fig. S9), which sug- nated tubulin. heart disease (Fig. 7G). Because the molecular

14
SCIENCE sciencemag.org sciencemag.org
22 APRIL 2016 • VOL 352 ISSUE 6284 SCIENCE
aaf0659-7 aaf0659-8
SCIENCE sciencemag.org
22 APRIL 2016 • VOL 352 ISSUE 6284 15
sciencemag.org SCIENCE
 Growing Gap in Life Expectancy by Income: Implications for 22 December 2015; accepted 17 March 2016
Princeton Center for Translational Research on Aging
Federal Programs and Policy Responses” (2015). Published online 21 April 2016
grant 2P30AG024928. Data and code are available at
10.1126/science.aaf1437
5. J. Pijoan-Mas, J. V. Ríos-Rull, Demography 51, 2075–2102 http://dx.doi.org/10.7910/DVN/C2VYNM.
T IT
O PTLIM
E IZING YOUR LI V E -CE L L M I CRO S CO P Y: TR I C K S A N D T R A D E - O F F S R ES (2014).E A RC H | R E PO R TS S E C T I O N T WO | RESEARCH ARTICLES
6. H. Waldron, Soc. Secur. Bull. 67, 1–28 (2007).
7. H. Waldron, Soc. Secur. Bull. 73, 1–37 (2013).
identity of TCP is unknown, it is unclear if up- appears to return in the form of a restoring force 9. M. R. Zile et al., J. Am.8. Coll.
J. Wilmoth,33,
2. B. Cardiol.
Bosworth, K.250–260 (1999). inofInternational
C. Boe, M. Barbieri,
Zhang, “Evidence
Differences
Increasing Differential NEURODEVELOPMENT
ACKN OWLED GMEN TS SUPPLEMENTARY MATERIALS ing that elevated levels at 1 mo are likely
10. G. Cooper 4th, Am. J. Physiol.in Mortality
HeartACirc.at Older
Physiol.Ages:
291, Dimensions
H1003– and Sources,
regulation of the detyrosinating enzyme may un- that would quicken sarcomere extension. This Mortality: Comparison of the
E. M. Crimmins, S. H. Preston, B. Cohen, Eds.
HRS and SIPP,” Center for We thank M. Barbieri, A. Case, A. Deaton, J. Goldstein, www.sciencemag.org/content/352/6286/708/suppl/DC1 not a developmental artifact. C1q was also
H1014 (2006).
derlie this effect. implies that energy used to deform MTs under- Retirement Research at Boston College Working Paper 2015-13
(National Academies Press, Washington, DC, 2011), I. Kuziemko, R. Lee, and K. Wachter, as well as seminar Materials and Methods similarly increased in the hippocampus of

Discussion
goes substantial loss. Buckling of the MT exerts
compressive force on the surrounding matrix and
11. C. Janke, J. C. Bulinski, Nat.
(2011). pp.
9. G. K. 6,
12. J. P. Kerr et al., Nat. Commun.
Rev. Mol.
(2015).
337–372.

4. National8526
Singh, M.
Cell Biol. 12,

Research
773–786
3. R. Chetty et al., JAMA 10.1001/jama.2016.4226 (2016).
Siahpush, Int. J. Epidemiol. 35, 969–979
(2015). Council, Committee on the Long-Run
Complement and microglia mediate
participants at Berkeley, the Chicago Federal Reserve,
Fundação Getúlio Vargas São Paulo, Bonn University,
University of Munich, Princeton University, ETH Zurich,
Figs. S1 to S8
Tables S1 to S4
References (36–38)
another model of AD, the APP/PS1 (prese-
nilin 1) mice (15) (fig. S2C). Notably, SIM
13. H. Sato et al., J. Cell Biol.(2006).
Our findings demonstrate a regulatory pathway
for MT load–bearing and myocyte mechanics
deforms the cytoplasm, which, due to its intrinsic
viscosity, can act as an energy sink during each
10. M.
14. S. Belmadani, C. Poüs, R.Growing
Méry, Mol. Cell. Biochem.Med.
139, 963–973 (1997).
Macroeconomic
Ezzati,
237,5,
Federal
A.
Ventura-Clapier, B.
Gap in Life
e66 (2008).
39–46
Programs
Effects of the Aging U.S. Population, “The
Friedman,
(2002).
S. C. Kulkarni, C. J. Murray, PLOS
R. Fischmeister,
Expectancy P.-F. by Income: Implications for
and Policy Responses” (2015).
early synapse loss in Alzheimer
and the University of Zurich for comments. Supported by
Princeton Center for Translational Research on Aging
22 December 2015; accepted 17 March 2016
Published online 21 April 2016
demonstrated colocalization of C1q with
PSD95-positive puncta in 1 mo J20 hip-
through posttranslational detyrosination of tu- cycle. This is reflected in the large viscous com- grant 2P30AG024928. Data and code are available at pocampus (Fig. 1C). A higher percentage
5. 462,
mouse models
15. E. White, Pflugers Arch. 11. C. J. Murray et
177–184 (2011). al., PLOS Med. 3, e260 (2006). 10.1126/science.aaf1437
J. Pijoan-Mas, J. V. Ríos-Rull, Demography 51, 2075–2102 http://dx.doi.org/10.7910/DVN/C2VYNM.
bulin. Detyrosinated MTs buckle under load in ponent of the MT contributions to myocyte me- 16. G. Cooper 4th, Heart12. Fail.H.Rev.
Wang,
(2014). 5, 187–201 (2000).
A. E. Schumacher, C. E. Levitz, A. H. Mokdad, of PSD95 colocalized with C1q in the hip-
contracting cardiomyocytes, which confers mechan- chanical properties observed at deformation rates 17. L. Cassimeris, P. Tran, Methods
C. J. Murray,Cell Biol. 97, xvii–xviii
Popul. (2010).
Health Metr. 11, 8 (2013). pocampus of J20 mice than in that of WT
6. H. Waldron, Soc. Secur. Bull. 67, 1–28 (2007).
18. G. Lukinavičius et al.,13.Nat.J. Methods
S. Olshansky11, 731–733 (2014).Aff. 31, 1803–1813 (2011).
et al., Health
ical resistance to contraction and regulating the consistent with contractile velocities both in this 7. H. Waldron, Soc. Secur. Bull. 73, 1–37 (2013). littermates (Fig. 1D and fig. S3), suggesting
19. S. Belmadani, C. Poüs, 14. R.E.Fischmeister,
R. Meara, S.P.-F. Méry, Mol.
Richards, D. M.Cell.
Cutler, Health Aff. 27, 350–360 Soyon Hong,1 Victoria F. Beja-Glasser,1* Bianca M. Nfonoyim,1* Arnaud Frouin,1
viscoelastic properties of the myocyte. and previous work (34). However, we do note a 8. J. Wilmoth, C. Boe, M. Barbieri, in International Differences
Biochem. 258, 35–48 (2004). that the C1q-associated synapses may be
(2008).
in Mortality at Older Ages: Dimensions and Sources, Li,2 Saranya Ramakrishnan,1 Katherine M. Merry,1 Qiaoqiao Shi,2
NEURODEVELOPMENT
Shaomin
The observation that MTs normally buckle, slight prolongation of the late phase of relaxation 20. C. Janke et al., Science 15. 308,
D. M.1758–1762
Cutler, F.(2005).
Lange, E. Meara, S. Richards-Shubik, marked for elimination.
E. M. Crimmins, S. H. Preston, B. Cohen, Eds.
21. C. Erck et al., Proc. Natl. Acad. Sci. U.S.A. 102, 7853–7858 Arnon Rosenthal,3,4,5 Ben A. Barres,6 Cynthia A. Lemere,2
rather than break or slide, strongly indicates that in TTL-overexpressing myocytes, which may rep- C. J. Ruhm, J. Health Econ. 30, 1174–1187 (2011).
(National Academies Press, Washington, DC, 2011), Punctate Aβ was found deposited in J20
Dennis J. Selkoe,2,7 Beth Stevens1,8†
they bear load and store elastic bending energy
during the cardiac contractile cycle. This has im-
resent the loss of a MT contribution to restoring
force. We consider it probable that the restoring
(2005). 16. J. K. Montez, L. F. Berkman, Am. J. Public Health 104, e82–e90
pp. 337–372.
185, 1159–1166 (2009).
22. L. Peris et al., J. Cell Biol.(2014).
9. G. K. Singh, M. Siahpush, Int. J. Epidemiol. 35, 969–979
23. M. Sirajuddin, L. M. Rice, R. D. Vale,
17. Human Biol. 16, 335–344
Nat. CellDatabase;
Mortality www.mortality.org.
Complement and microglia mediate
Synapse loss in Alzheimer’s disease (AD) correlates with cognitive decline. Involvement
hippocampus at 1 mo (fig. S4), long before
Aβ plaques deposit (11, 12), raising the ques-
(2006). tion of whether C1q increase in these pre-
plications for MT-dependent mechano-signaling
in muscle and other tissues (1, 32, 33) but also
force of other internal elastic elements, such as
titin, are likely to play a more dominant role, at
(2014). 18. D. D. Reidpath, P. Allotey, J. Epidemiol. Community Health 57,
10. M.
24. G. Kreitzer, G. Liao, G. G.344–346
1118 (1999).
Ezzati,
Gundersen, A. B.
(2003).
Med. 5, e66 (2008).
Friedman, S.
Mol. Biol. Cell 10, 1105– C. Kulkarni, C. J. Murray, PLOS
early synapse loss in Alzheimer
of microglia and complement in AD has been attributed to neuroinflammation,
prominent late in disease. Here we show in mouse models that complement and microglia
plaque brains is dependent on soluble Aβ
has direct implications for contractility. Our mod- least in the initial return toward resting sarco- 19. A. Case, A. Deaton, Proc. Natl. Acad. Sci. U.S.A. 112, levels. To test this hypothesis, we injected
mouse models
11. C. Res.
25. R. A. Whipple et al., Cancer J. Murray et al., PLOS
68, 5678–5688 Med. 3, e260 (2006).
(2015). (2008). mediate synaptic loss early in AD. C1q, the initiating protein of the classical
el of myocyte contractility demonstrates how mere length (27). Thus, an increase in detyrosina- 26. X. Fonrose et al., Cancer
15078–15083
12. H. Wang, A. E. Schumacher,
67, 3371–3378 C. E. Levitz, A. H. Mokdad, the mice with compound E, a γ-secretase
20. J.Res. Bound, A. Geronimus, (2007). J. Rodriguez, T. Waidman, “The complement cascade, is increased and associated with synapses before overt plaque
changing MT load–bearing and force transfer tion may increase myocyte stiffness and impair C.H.J.Granzier,
27. M. Helmes, K. Trombitás,Implications Murray,Circ.Popul.Res.Health Metr. 11, 8 (2013).
79, 619–626
of Differential Trends in Mortality for Social inhibitor that rapidly decreases Aβ produc-
(1996). 13. J. S. Olshansky et al., Health Aff. 31, 1803–1813 (2011). deposition. Inhibition of C1q, C3, or the microglial complement receptor CR3 reduces the
with the sarcomeres can substantially alter con- contraction by acting as an energy sink, without Security Policy,” University of Michigan Retirement Research 1 1 tion (12). Compound E markedly reduced
28. R. E. Mahaffy, C. K. Shih,
14. E. F. C.
R.MacKintosh, J. Käs,2014-314
Meara, S. Richards, Phys.
D. M.Rev.Cutler, Health Aff. 27, 350–360 number of phagocytic
Soyon Hong, Victoria microglia, as well
F. Beja-Glasser, * as the extent
Bianca of early1synapse
M. Nfonoyim, * Arnaudloss. C1q1 is
Frouin,
tractile properties. Our experimental data show providing significant energetic return during Center Working Paper (2014). soluble Aβ levels in J20 mice; there was a
Lett. 85, 880–883 (2000). (2008). necessary
Shaomin Li, 2
for Saranya
the toxicRamakrishnan,
effects of soluble 1
Katherine
b-amyloid M.(Ab)
Merry, 1
oligomers
Qiaoqiaoon synapses
Shi, 2
and
that such changes in MT load–bearing can be relaxation. 21. J. B. Dowd, A. Hamoudi, Int. J. Epidemiol. 43, 983–988
15. Biol. 181, 667–681 (2008).
29. P. Konieczny et al., J. Cell(2014).
D. M. Cutler, F. Lange, E. Meara, S. Richards-Shubik, 3,4,5
corresponding reduction of C1q deposition
achieved by posttranslational modifications of Consistent with this, an increase in detyrosi-
hippocampal
Arnon Rosenthal, long-term Ben A. Barres,6Finally,
potentiation. Cynthiamicroglia
A. Lemere, 2
in adult brains engulf synaptic
(Fig. 1E), suggesting that Aβ upregulates
30. G. Liao, G. G. Gundersen,C. J. Biol. Chem.J.273,
J. Ruhm, 9797–9803
Health Econ. 30,(1998).
1174–1187 (2011).
22. T. Goldring, F. Lange, S. Richards-Shubik, “Testing for Changes material
Dennis J.inSelkoe, 2,7
a CR3-dependentBeth Stevens 1,8
process † when exposed to soluble Ab oligomers. Together,
the MTs themselves, particularly detyrosination. nation was associated with clinical contractile 31. G. Gurland, G. G. Gundersen, Cell Biol.L.131,
16. J. K.J.Montez, 1275–1290
F. Berkman, Am.(1995).
J. Public Health 104, e82–e90 C1q. To further address whether the in-
in the SES-Mortality Gradient When the Distribution of
Signal. 5, ra56 (2012).
32. R. J. Khairallah et al., Sci.(2014). these findings suggest that the complement-dependent pathway and microglia that
The measured reductions in buckling and visco- dysfunction in human hearts. Our cellular studies Education Changes Too,” National Bureau of Economic crease of C1q is dependent on soluble
33. M. Prager-Khoutorsky, 17.A.Human
Khoutorsky,MortalityC. W.Database;
Bourque, Neuron www.mortality.org. prune
Synapseexcess
loss synapses
in Alzheimer’sin development
disease (AD) arecorrelates
inappropriately activateddecline.
with cognitive and mediate
Involvement
elasticity and the increase in contractile speed of demonstrate that acute reduction of detyrosina- Research Working Paper 20993 (2015).
83, 866–878 (2014).18. D. D. Reidpath, P. Allotey, J. Epidemiol. Community Health 57, Aβ, and if so, which species, we injected
23. A. S. Hendi, Int. J. Epidemiol. 44, 946–955 (2015). synapse lossand
of microglia in AD.complement in AD has been attributed to neuroinflammation,
PTL-treated, TTL-overexpressing myocytes can tion with genetic or pharmacologic approaches 34. H. Tagawa et al., Circ. Res.344–34680, 281–289 (2003). (1997). soluble Aβ oligomers or monomers into
24. A. Aizer, J. Currie, Science 344, 856–861 (2014). prominent late in disease. Here we show in mouse models that complement and microglia
all be attributed to changing the way MTs in- can boost contractility and reduce mechanical 19. A. Case, A. Deaton, Proc. Natl. Acad. Sci. U.S.A. 112,
25. D. Brown, A. Kowalski, I. Lurie, “Medicaid as an Investment in
lateral ventricles of WT mice. Hippocam-
teract with the sarcomere and impairing their stiffness. Additionally, these approaches are 15078–15083 (2015). mediate synapticassociation
enome-wide
enome-wide loss early in
association AD. C1q,
studies
studies the initiating
impli- brain (1).protein
synapse During of thewe
loss occurs, classical
development, used
C1qsuperresolu-
and C3 local- pus contralateral to the injection site was
ACKNOWLEDGMENTS

G
Children: What Is the Long-Term Impact on Tax Receipts?”
ability to act as compression resistors. It is also able to induce large changes in detyrosination, 20. J. Bound, A. Geronimus, J. Rodriguez, T. Waidman, “The
The authors thank C. W. Ward National
and E. L. Bureau
Holzbauroffor kindly providing
Economic Research Working Paper 20835 complement
cate cascade,
catemicroglia
microglia andiscomplement-related
and increased and associated
complement-related ize to with
tion synapses
structured
synapses before
illumination
and mediate overt
synapse plaque
microscopy
elimination examined to avoid any surgery-related
Implications of Differential Trends in Mortality for Social deposition. Inhibition of C1q,disease
C3, or(AD)
the microglial complement receptor CR3 reduces
possible that detyrosinated MTs anchored to one while only slightly altering the overall MT cyto- TTL and EMTB-3xGFP plasmids, (2015).respectively, as well as for
Security Policy,” University of Michigan Retirement Research pathways
pathways in Alzheimer’s
Alzheimer’s disease (AD)
(1). (SIM) (10) tomicroglia
by phagocytic quantify synapse
(5–7). density the
We hypothesized in effects. Oligomeric Aβ (oAβ), which is
sarcomere form bundles with MTs anchored to skeleton, which minimizes off-target consequences. 26. S. Cahodes,
discussion; R. Bloch for providing CenterdesminWorking
S. Kleiner,
KO mice
Paper
M.littermate
F. Lovenhem, M. Grossman,
and2014-314 (2014).
number
(1). of phagocytic
Previous
Previous research
research microglia,
has has as well
demonstrated
demonstrated bothas the extent
hippocampal
that of early
this normal CA1 synapse
stratumloss.
developmental C1q is
radiatum
synaptic of fa-
pruning prefibrillar in nature and acts as a me-
controls; Carl Zeiss Microscopy of Child Health Insurance
“EffectAiryscan Access on Schooling,”
adjacent sarcomeres. If so, disrupting bundling Thus, interfering with detyrosination may repre- 21. J. for B. Dowd, A.instrumentation
Hamoudi, Int. J. use;Epidemiol. 43, 983–988 necessary
both for the
beneficial andtoxic
beneficial effectsroles
and detrimental
detrimental of soluble
roles
of com- ofb-amyloid
milial
pathway (Ab) oligomers
AD-mutant
is activated oninsynapses
human
early amyloid
the AD brain and
precur-
and diator of synapse loss and dysfunction in
IonOptix for technical supportNational
and
(2014). cell Bureau
holder of Economic
materials; and Research
Y. E. Working Paper
hippocampal long-term potentiation. Finally, microglia in (hAPP)
adultloss.
brains engulf synaptic
would also effectively uncouple MTs from force- sent an attractive and novel therapeutic strategy Goldman, E. L. Grishchuk,22. and20178 (2014). for discussion and
R.Goldring,
J. Composto complement
plement and
and microglia microglia in plaque-related
in plaque-related neuro- sor protein
mediates synapse (“J20”) transgenic mice AD (4), but not the relatively innocuous
T. F. Lange, S. Richards-Shubik, “Testing for Changes material in
generating structures. Regardless of the mecha- for increasing contractility. 27. S. Miller, L. R. Wherry, “The Long-Term Health
guidance. B.L.P., P.R., M.A.C., A.I.B.,
in the and K.B.M. wereGradient
SES-Mortality
Effects
responsible forthe Distribution of
of Early pathology (2,a3);CR3-dependent
neuropathology (2, 3); however,
however, process
their
their roles in roles when
synapse in exposed
(11). to soluble
TheQuantification
degree Abofoligomers.
colocalized
of region-specific Together,
synapsepre-loss
and is monomeric Aβ or vehicle, induced C1q
Coverage,” SocialWhen these findings a suggest that the of complement-dependent pathway ofand microglia thatinand
Life Medicaid Science Research Network
nism, disrupting coupling to sarcomeres would In conclusion, our data show that MTs exhibit experimental design; B.L.P., P.R., M.A.C., Changes
Education C.Y.C., andToo,”
Working Paper 2466691 (2015).
A.I.B.National
carried Bureau of Economic synapse
loss, loss,pathological
a major major pathological
correlate correlate
cognitive postsynaptic puncta
a stronger correlate [synaptophysin
cognitive decline AD deposition (Fig. 2A and fig. S5). A higher
reduce the incident force on the MT, and buckling divergent mechanical behavior because of the out experiments and analyzedResearch data; H.A.Working
and V.B.S. designed
Paper 20993Teens:
(2015).Using and Eligibility prune
of
decline excess
cognitive
in (4),synapses
AD decline remain toin
in AD development
be(4), to are
remainEmerg-
identified. be inappropriately
postsynaptic activated
than counts ofdensity and
plaques,95tangles, mediate
(PSD95) and(Fig. 1A);
neuronal percentage of PSD95 colocalized with C1q
28. L. R. Wherry, B. Meyer, “Saving
occurrence would drop. differences in how they couple to the rest of the and executed the mathematical A. model.
23. Discontinuity
S. Hendi, B.L.P.to and
Int. P.R. cowrote
J. Epidemiol.
Estimate 44, 946–955
the Effects (2015).
of Medicaid Eligibility,”
synapse
identified.
ing loss
research in AD. research
Emerging
implicates microgliaimplicates
and immune- mi- synaptotagmin and homer
loss (8, 9). To determine how early(fig.synapse
S1, Alossto in oAβ-injected versus monomer-injected
the manuscript. All authors 24.participated
National J.inCurrie,
A. Aizer, Bureau the critical
of Science review
Economic 344, and856–861
Research (2014).Paper 18309
Working
The striking periodicity of buckles in untreated cardiac cytoskeleton. The tyrosinated portions of revision of the manuscript. This
25. (2013).work was supported by funding
D. Brown, A. Kowalski, I. Lurie, “Medicaid as an Investment in
croglia and immune-related
related mechanisms in brain wiring mechanisms
in the healthy in D)]
occurs,revealed
we useda superresolution
significant lossstructured of synapses illu- mice (Fig. 2B), in a manner similar to this
myocytes lends further support to the idea of a the network, moving readily with the myocyte brain wiring
enome-wide in the association
healthy brain (1). During
studies impli- in J20(1).hippocampus
mination
brain microscopy
During at 3(10)
(SIM)
development, toC1q 4 months
to quantify old
syn-
and C3 local- colocalization in J20 mice. Together, these

G
from National Institute for29. Arthritis
L.Children:and What
R. Wherry, Musculoskeletal
S.IsMiller,
the Long-Termand SkinImpact
R. Kaestner, B. D. on Tax “Childhood
Meyer, Receipts?”
sarcomeric anchor. The preferential association of during contraction, provide little contractile re- Diseases (NIH) (T32AR053461-09 National
Medicaid to P.R.
Bureauand T32HL007954
Coverage of and
Economic to HealthWorking
Research
Later Life Paper 20835
Care Utilization,” development, C1q andand
cate microglia C3 localize to synapses
complement-related (mo),
apse an age
ize todensity
synapses that
in hippocampalprecedes
and mediate CA1 plaque
stratum
synapse deposi-
radiatum
elimination findings show an early and aberrant in-
desmin with detyrosinated tubulin and the insen- sistance. Conversely, the detyrosinated portions M.A.C.) and from National Heart, NationalLung,Bureau
(2015). and Blood Institute (NIH)
of Economic Research Working Paper 20929 and
1
mediate
F.M. Kirby synapse
Neurobiology
pathways elimination
Center, by (AD)
Boston Children’s
in Alzheimer’s disease phago-
Hospital(1). tion
of (11, 12).
byfamilial
phagocytic Synapse
AD-mutant
microglia loss
human inamyloid
(5–7). preplaque
We J20
precursor
hypothesized crease and synaptic localization of C1q in
sitivity of desmin KO animals to changes in de- of the MT network, forming complexes with des- (R00-HL114879 to B.L.P.).26. The(2015).
S.VIVA experiments
Cahodes, usedM.
S. Kleiner, shared
F. Lovenhem, M. Grossman, (BCH) and
cytic Harvard Medical
microglia
Previous (5–7).School
research has(HMS),
We Boston, MA that
hypothesized
demonstrated both CA1
proteinwas
that this confirmed
(hAPP)
normal (“J20”) by
developmentalelectron
transgenic micemicroscopy
synaptic (11).pruning
Quan- multiple AD model systems. Furthermore,
of Child Health Insurance 2
experimental facilities from 30.the
J.“Effect
Nano
Ludwig, Science
D. L. andMiller, J. Econ. Access
Engineering
Q. on Schooling,
122, 159–208 (2007).” 02115, USA. Ann Romney Center for Neurologic Diseases,
tyrosination strongly implicates desmin as at least min IFs, produce a cross-linked MT-IF network National Bureau of Economic Research Working Paper this normal
Departmentbeneficialdevelopmental
andBrigham
of Neurology, detrimental
andsynaptic pruning
rolesHospital
Women’s of com- (fig.
pathwayS1G).
tification Confocal
isofactivated
colocalized imaging
early pre- andalso
in the AD showed
postsynaptic
brain and fluorescent in situ hybridization (FISH)
Center on Molecular Function 31. H. at Hoynes,
the Nano-Bio Interface at the
D. Whitmore-Schanzanbach, D. Almond,
one component of a sarcomeric anchoring com- that confers robust resistance to contraction. This 20178 Run
“Long
University of Pennsylvania supported (2014).
Impacts
by NSF underofgrant
Childhood
DMR08- Access to the Safety Net,” pathway
plement
(BWH) is activated
and and
HMS, microglia
Boston, MA early inUSA.
the 3AD
in plaque-related
02115, brain
Alector and
neuro-
Inc., 953 synapse
puncta
mediates loss
synapsein CA1,
[synaptophysin loss.CA3,andand dentate gyrus
postsynaptic den- demonstrated up-regulated C1qa expres-
plex of detyrosinated MTs. Note that myocytes orthogonal MT-IF grid requires tightly periodic 32802. We thank R. Composto, 27. National
S.who
Miller, L. R. Wherry,
Bureau
acknowledges of Economic NSFLong-Term
the“The Health Effects
Research Working Paper of Early
18535 mediates
Indiana synapse
Street,
pathology San3);
(2, loss. CA their
Francisco,
however, USA. 4in
94107, roles Annexon
synapse of 3 95
sityThemo J20 hippocampus
(PSD95)
degree of (Fig.
region-specific butsynapse
1A); synaptotagminnot in loss stria-
and is sion in microglia (fig. S6), implicating mi-
Life Medicaid Coverage,” Social Science Research Network Biosciences, 280 pathological
Utah Avenue Suite 110, South
lacking desmin have decreased viscoelasticity, MT deformations to accommodate myocyte mor- Polymers Program under grant (2012).
DMR09-07493. Procurement of The
loss, degree
a major of region-specific
correlate ofSan
synapse loss
cognitive tum
homer (fig.
a stronger (fig.S1E).
S1, A Synapse
correlate levelsa significant
to D)]ofrevealed
cognitive were
decline not al-
inloss
AD croglia as a major source of C1q in these
32. A. Working
Isen, M. Paper 2466691 (2015).
Rossin-Slater, R. Walker, “Every Breath You Take Francisco, CA 94080, USA. 5Department of Anatomy,
human heart tissue was enabled by grants from NIH (HL089847 is
University of California San Francisco (UCSF), San decline
a stronger correlate of cognitive tered in 1 mo J20 brains versus
despite a denser MT network, which supports phology changes during contraction. These defor- 28. Every
L. R. Wherry,
Dollar You’llB. Meyer,
Make:“Saving Teens: Using
The Long-Term and Eligibility
Consequences of decline in AD (4), remain to be identified. Emerg-
Francisco,
of synapses
than counts inofJ20 hippocampus
plaques, tangles, 3 towild
at and type
4neuronal
months preplaque brains.
and HL105993) to K.B.M. H.A. and V.B.S. are supported by the in AD than
the idea that MT network organization and cross- mations require a considerable amount of energy Discontinuity
the Clean Air Act to Estimate
of 1970,”the EffectsBureau
National of Medicaid Eligibility,”
of Economic ing94143,
CA USA. 6counts
research implicates
Department ofofplaques,
microglia tangles,
Neurobiology,and immune-
Stanford and (WT)
old (8, littermates
loss(mo), 9).anToage determine (fig.how
that precedes S1F),
plaque
early suggesting
deposition
synapse loss To test whether C1q and oAβ act in a
National Institute of Biomedical Imaging and Bioengineering
linking is a stronger determinant of myocyte me- to form and dissipate a large fraction of that en-
(NIH) under award number
National Bureau
Research WorkingofPaper Economic 19858Research
(2014). Working Paper 18309 neuronal
University
related loss
School
mechanisms of(8, 9). To
Medicine,
in determine
Palo
brain Alto,
wiring how
CA 94305,
in the early
USA.
healthy that
(11,
occurs, the
12). we hippocampal
Synapse
used superresolution synaptic
loss in preplaque J20 loss
structured CA1atwas 3
illu- common pathway to eliminate synapses,
33.R01EB017753
A.(2013).
Fenelon, S. andH.NSF grant CMMI-
Preston, Demography 49, 797–818 7 8
Prothena Biosciences, Dublin, Ireland. Stanley Center for mo is likely not a result of
chanical properties than network density per se. ergy due to viscous interactions. This has im- 1312392. The data are included in
29. (2012). the main manuscript and in the
L. R. Wherry, S. Miller, R. Kaestner, B. D. Meyer, “Childhood
confirmed
mination by electron
microscopy microscopy
(SIM) (10)abnormal
to(fig. S1G). syn-
quantify Con-
syn- we injected oAβ into lateral ventricles of
Psychiatric Research, Broad Institute of MIT and Harvard,
Both the desmin and MT networks have ele- portant implications for MT load–bearing across supplementary materials34. online.
D.Medicaid
de Walque, CoverageJ. Hum.andResour.
Later Life45, Health
682–717Care Utilization,”
(2010). Cambridge, MA 02142, USA. aptic
focal development.
imaging
apse density in also showed synapse
hippocampal CA1 stratum lossradiatum
in CA1, C1qa knockout (KO) mice (16). Soluble
ments perpendicular to their typical orientation, cell biology, as well as for the altered mechanical 35. C. National
E. Finch,Bureau E. M.ofCrimmins,
Economic ScienceResearch305, Working Paper 20929
1736–1739
1
F.M. Kirby
*These Neurobiology
authors contributed Center,
equally Boston Children’s
to this work. Hospital
†Corresponding of We
CA3, asked
and
familial whether
dentate
AD-mutant gyrus the
human of classical
3amyloid
mo J20 comple-
hippo-
precursor oAβ induced a significant loss of colocal-
particularly near the sarcolemma, which may stiffness and mechano-signaling in cardiac disease. SUPPLEMENTARY MATERIALS
(2015).
(2004). (BCH) and
author. Harvard
Email:
2
Medical School (HMS), Boston, MA
beth.stevens@childrens.harvard.edu ment
proteincascade
campus but not
(hAPP) isinup-regulated
(“J20”)striatum
transgenic in
(fig.mice preplaque
S1E). Synapse
(11). Quan- ized synapsin- and PSD95-immunoreactive
30. J. Ludwig, D. L. Miller, Q. J. Econ. 122, 159–208 (2007).
www.sciencemag.org/content/352/6284/aaf0659/suppl/DC1 02115, USA. Ann Romney Center for Neurologic Diseases,
alter how those elements interact with the cyto-
Department of Neurology, Brigham and Women’s Hospital
brains
tification when synapses pre-
of colocalized are and
already vulner-
postsynaptic puncta in WT mice within 72 hours (Fig.
Materials and Methods 31. H. Hoynes, D. Whitmore-Schanzanbach, D. Almond,
skeleton and plasma membrane. However, we REFERENCES AND NOTES (BWH) and HMS, Boston, MA 02115, USA. 3Alector Inc., 953 able.
puncta C1q immunoreactivity
[synaptophysin (13) (antibody
and postsynaptic den- 3A, left panel) (17). In contrast, oAβ failed
REFERENCES
1. B. L. Prosser, C.AND NOTES
W. Ward, W. J. Lederer, Science 333, 1440–1445
Supplementary Text 7 12“Long6Run MAY Impacts
2016 of Childhood
• VOL Access to
352 ISSUE the Safety Net,”
6286 sciencemag.org SCIENCE
believe that the preponderance of the contractile
Figs. S1 to S11
National Bureau of Economic Research Working Paper 18535 Indiana Street, San Francisco, CA 94107, USA. 4Annexon now
sity 95 available
(PSD95) at Abcam)
(Fig. was elevatedand
1A); synaptotagmin in to induce synapse loss in C1qa KO mice (Fig.
1. B.(2011).
L. Prosser, C. W. Ward, W. J. Lederer, X-ROS signaling: Rapid
resistance that results from detyrosination is due Tables S1 to S9
(2012). Biosciences, 280 Utah Avenue Suite 110, South San J20
homer brains
(fig. S1, asAearly
to D)]as 1 mo and
revealed preceding
a significant loss 3A, right panel), suggesting that C1q is re-
2. mechano-chemo
B. L. Prosser, C. W. Ward, W. J. Lederer,
transduction in heart.Cardiovasc. Res. 98,
Science 333, 32. A. Isen, M. Rossin-Slater, R. Walker, “Every Breath You Take Francisco, CA 94080, USA. 5Department of Anatomy,
to longitudinal MTs in an orthogonal grid with 307–314 (2013).
1440–1445 (2011). doi: 10.1126/science.1202768; References (35–42)
Every Dollar You’ll Make: The Long-Term Consequences of University of California San Francisco (UCSF), San Francisco,
synapse
of synapses lossin (Fig. 1B and fig. S1).
J20 hippocampus at 3 to C1q eleva-
4 months quired for oAβ-induced synapse loss in vivo.
transverse desmin IFs because of the simple nu- 3. pmid:
M. R. Zile et al., Circulation 98, 567–579 (1998).
21903813 Movies S1 to S8 the Clean Air Act of 1970,” National Bureau of Economic CA 94143, USA. 6Department of Neurobiology, Stanford tion was an
old (mo), region-specific,
age that precedes particularly
plaque depositionin the To determine whether local, acute inhibi-
merical majority of cytoskeletal elements in this 2.4. B.S.L.
Nishimura
Prosser, etC. al.,
W.Circ.
Ward, W.98,
Res. J. 81–87 (2006).
Lederer, X-ROS signaling is Research Working Paper 19858 (2014). University School of Medicine, Palo Alto, CA 94305, USA. hippocampus
(11, 12). Synapseand loss frontal
in preplaque cortex, J20 twoCA1 wasre- tion of C1 activation could similarly blunt
configuration. 5. enhanced
H. L. Granzier, T. C. Irving,
and graded Biophys.
by cyclic J. 68, 1027–1044
cardiomyocyte (1995).
stretch. 12 December 2015; accepted 33. 29 January 2016 7
Prothena Biosciences, Dublin, Ireland. 8Stanley Center for
A. Fenelon, S. H. Preston, Demography 49, 797–818 gions
confirmed vulnerable
by electron to microscopy
synapse loss (fig.(14)
S1G).(Fig.
Con- the synaptotoxic effects of oAβ, we used an
6. Cardiovasc.
C. P. Brangwynne
Res. 98,et al., J. Cell Biol.
307–314 173,doi:
(2013). 733–741 (2006).
10.1093/cvr/cvt066 10.1126/science.aaf0659 (2012). Psychiatric Research, Broad Institute of MIT and Harvard,
Despite the fact that detyrosinated MTs store 3.7. M.
M.R.Mehrbod, M. R.
Zile et al., K. Mofrad, PLOS
Constitutive ONE 6,
properties ofe25627 (2011).
adult mammalian 1B and
focal fig. S2A).
imaging also C1q
showedimmunoreactivity
synapse loss in was CA1, antibody against C1q (anti-C1q) (ANX-M1,
34. D. de Walque, J. Hum. Resour. 45, 682–717 (2010). Cambridge, MA 02142, USA.
energy during sarcomere contraction, which pro- 8. cardiac
H. Tsutsui, K. Ishihara,
muscle G. Cooper 4th,
cells. Circulation Science 260,
98, 567–579 682–687
(1998). 35. C. E. Finch, E. M. Crimmins, Science 305, 1736–1739 *These authors contributed equally to this work. †Corresponding comparable
CA3, and dentate between gyrus J20of and3 mo WT J20mice hippo-at Annexon Biosciences), which blocks the
vides compression resistance, little of this energy (1993).
doi: 10.1161/01.CIR.98.6.567; pmid: 9714115 (2004). author. Email: beth.stevens@childrens.harvard.edu postnatal
campus butday not21 in (P21)
striatum (fig.(fig.
S2B),
S1E).suggest-
Synapse classical complement cascade (see fig. S7

16 7 12 sciencemag.org SCIENCE
6 MAY 2016 • VOL 352 ISSUE 6286 SCIENCE sciencemag.org Originally published sciencemag.org
6 May 2016 in SCIENCE
SCIENCE 17
O PT IM IZING YOUR LI V E -CE L L M I CRO S CO P Y: TR I C K S A N D T R A D E - O F F S RE S EAR CH | R E P O R T S R ES E A RC H | R E PO R TS S E C T I O N T WO | RESEARCH ARTICLES

Fig. 2. Oligomeric Ab increases C1q and microglial phagocytic activity. (A and B) Soluble Ab oligomers in WT mice led to elevation of C1q (green) (A)
and a higher percentage of PSD95 (red) colocalization with C1q versus monomers (B). (C and D) oAb induced high levels of CD68 (green) immunoreactivity
in Iba1-positive (red) microglia in WTmice (C), but not in those of C1qa KO mice (D). Both had negligible changes in morphology. See fig. S10. Scale bar, 10 mm (A),
5 mm (B), or 20 mm (C). Means ± SEM; n = 3 to 5 mice per treatment group per genotype. *P < 0.05 using two-tailed t test (B) or *P < 0.05, **P < 0.01 versus
control-treated or ##P < 0.01 versus Ab monomer–treated using two-way ANOVA followed by Bonferroni posttest (C).

Fig. 3. Complement is necessary for synapse

loss and dysfunction in AD models. (A) Ab oli-
gomers induced loss of colocalized synapsin- and
PSD95-immunoreactive puncta in the contralateral
hippocampus of 3 mo WTmice (left panel); however,
they failed to do so in C1qa KO mice (right panel).
(B) Coinjection of Ab oligomers with the function-
blocking antibody against C1q, ANX-M1, but not with
Fig. 1. C1q up-regulation and deposition onto synapses precede pre- image showing colocalization of C1q (green) and PSD95 (red). (D) Higher its IgG isotype control, prevented synapse loss in
plaque synapse loss in J20 mice. (A) Superresolution SIM images of percentage of PSD95 colocalized with C1q in 1 mo J20 dentate gyrus versus WT mice. (C) Pretreatment of hippocampal slices
synaptophysin (green)– and PSD95 (red)–immunoreactive puncta in WT. (E) Compound E reduces deposited soluble Ab (red) and C1q (green) in with the anti-C1q antibody, ANX-M1, prevented Ab-
stratum radiatum of 3 mo J20 or WT hippocampus (CA1). Quantification 1 mo J20 dentate gyrus, with minimal effect on C1q levels in WT mice. Scale mediated LTP inhibition (green) versus IgG (red).
of synaptic puncta or their apposition using Imaris indicates selective loss bar, 2 mm (A, C, and D) or 10 mm (B and E). Means ± SEM; n = 3 or 4 mice per IgG alone had a minimal effect (blue) versus artificial
of PSD95 in J20 hippocampus as compared to their WT littermate controls. genotype or per treatment group per genotype. *P < 0.05, **P < 0.01, or cerebrospinal fluid (aCSF) vehicle (black). n = 6 to
See fig. S1. (B) Region-specific up-regulation of C1q (green) in 1 mo J20; DG, ***P < 0.001 using two-way analysis of variance (ANOVA) followed by 11 slices per group. (D) Percentage of PSD95 co-
dentate gyrus; FC, frontal cortex; STR, striatum; CRB, cerebellum; DAPI, Bonferroni posttest (A and B), two-tailed one-sample t test (D), or two- localized with C3 is increased in APP/PS1 hippo-
4′,6-diamidino-2-phenylindole. See fig. S2. (C) Orthogonal view of SIM tailed unpaired t test (E). campus versus that of WT mice. (E and F) Genetic
deletion of C3 prevents synapse loss in 4 mo APP/
PS1 mice. Quantification of colocalized immuno-
levelssupplementary
were not alteredmethods).
in 1 mo J20Coadminis-
brains ver- vulnerable reactive puncta for synaptotagmin and homer in
and We testedtothe synapse loss (14) effects
functional (Fig. 1B and fig.
of the that the C1q-associated
support C1q as a key synapses may beofmarked
mediator oAβ-
sus wild-type dentate gyrus (E) or synaptophysin and PSD95 in
tration of the(WT) littermates
ANX-M1 (fig. antibody,
anti-C1q S1F), sug- S2A).
ANX-M1 C1q immunoreactivity
anti-C1q antibody was in
comparable
acute hip- be- for elimination.
induced synaptic loss and dysfunction.
gesting CA1 stratum radiatum (F) of WT, APP/PS1, APP/
but notthatits the hippocampal synaptic
immunoglobulin G (IgG) loss at
iso- tween J20 and WT mice at postnatal
pocampal slices treated with oAβ. IgG day 21 (P21) Punctate Ab was found
In the healthy depositedbrain,
developing in J20 C1q
hip-
3 mo control,
is likely not a result of abnormal synaptic (fig. S2B), suggesting that elevated levels at 1 in-
mo pocampus at 1 mo (fig.ofS4), PS1xC3 KO, and C3 KO hippocampi. Means ± SEM;
type prevented oAβ from inducing alone had negligible effects on LTP promotes activation C3,long
whichbefore
opso-Ab
development. are likely innotWTa developmental artifact. C1q was plaques depositof (11, 12), raising n = 3 to 5 mice per genotype or per treatment
synapse loss in WT mice (Fig. 3B). Thus, duction mouse hippocampal slices nizes subsets synapses forthe question of
elimination,
We asked group per genotype. *P < 0.05, **P < 0.01, or ***P <
blocking C1whether the by
activation classical
eithercomplement
genetic or also
and similarly increased
on the ability of in
oAβthetohippocampus
inhibit LTP; of whether
a processC1qthat
increase in these preplaquein
is down-regulated brains
the
cascade is up-regulated means
in preplaque brainsoAβ’s
when another model of AD, the APP/PS1 (presenilin 1) is dependent on soluble 0.001 using two-tailed one-sample t test (D), one-
antibody-mediated lessened however, pretreatment of hippocampal mature brain (5, 6). Ab levels. TooAβ
However, test this
in-
synapses are already vulnerable. C1q immuno- mice way (A, C, E, F) or two-way (B) ANOVA followed by
synaptotoxic effects. slices(15) (fig.the
with S2C). Notably, antibody
anti-C1q SIM demonstrated
signifi- hypothesis,
duced we injected the
a significant C3 mice with compound
deposition in WT
reactivity (13) (antibody now available Bonferroni posttest. ns, not significant.
To determine whether C1q at isAbcam)
asso- colocalization
cantly prevented of C1q with PSD95-positiveofpuncta
the impairment LTP E, a g-secretase
adult mice (fig.inhibitor
S7A, that
upper rapidly decreases
panel). This
was elevated
ciated within J20 brains as
synaptic early as 1 mo and
dysfunction, we in
by1oAβ mo (Fig.
J20 hippocampus
3C). Neither(Fig. ANX-M11C). Anorhigher
its Ab production
was (12). Compound
significantly reduced Einmarkedly
both the re-
preceding
asked synapsethe
whether lossestablished
(Fig. 1B and fig. S1). C1q
ability of percentage
IgG controlofalteredPSD95 basal
colocalized
synapticwithneuro-
C1q in duced KO
C1qa soluble
(fig.AbS7A,
levelslower
in J20panel)
mice; there
and was
the
elevation
oAβ was region-specific,
to potently particularly
inhibit long-term poten- in the hippocampus
transmission of S8).
(fig. J20 mice than in that
Collectively, these of a corresponding
ANX-M1 anti-C1q reduction of C1q deposition
antibody–treated WT
the hippocampus
tiation (LTP) (4) andwas
frontal cortex, twoon
dependent regions
C1q. WT littermates
results (Fig. 1D andslices
in hippocampal fig. S3),
andsuggesting
in mice (Fig. 1E),
mice suggesting
(fig. that Ab up-regulates
S7B), suggesting that the C1q.C3

18
SCIENCE sciencemag.org 6 MAY 2016 • sciencemag.org 7 13
SCIENCE
VOL 352 ISSUE 6286 7 14 6 MAY
SCIENCE 2016 • VOL 352 ISSUE 6286
sciencemag.org sciencemag.org SCIENCE
19
 RE S EAR
O PT IM IZING YOUR CH
LI V E -CE|L LR E
MPIOCRO
R T S S CO P Y: TR I C K S A N D T R A D E - O F F S
Fig. 4. Microglia engulf synapses via CR3 upon
oligomeric Ab challenge. (A) Orthogonal view of
high-resolution confocal image shows colocalization
of homer-GFP and Iba1 (red). (B) Three-dimensional
reconstruction and surface rendering using Imaris
demonstrate larger volumes of homer-GFP puncta
inside microglia of oAb-injected contralateral hippo-
campus versus those of monomer-injected. (C) Mi-
croglia of homer-GFPxCR3 KO mice (right panel)
show less engulfment of homer-GFP when chal-
lenged with oAb versus those of homer-GFP mice
(left panel). (D) Ab oligomers failed to induce syn-
apse loss in the contralateral hippocampus of CR3
KO mice (right panel) as they did in WT mice (left
panel). Scale bar, 5 mm (A and B). Means ± SEM; n
= 3 mice per treatment group per genotype (n = 6
to 17 microglia analyzed per mouse). *P < 0.05,
**P < 0.01, or ***P < 0.0001 using two-tailed t test
(B) or two-tailed one-sample t test (C and D). ns,
not significant.
deposition in this model is downstream of
s, block- synapses (figs. S2C and S9), to C3-deficient mice
the classical complement cascade. Consis-
ntibody- (18). Quantification of colocalized pre- and post-
tent with these findings, a higher percent-
ptotoxic synaptic puncta demonstrated synapse loss in
age of PSD95 colocalized with C3 in J20
4 mo APP/PS1 hippocampus as compared to WT;
and APP/PS1 brains (Fig. 3D and fig. S9).
ed with however, APP/PS1xC3 KO mice did not display
To determine whether C3 is necessary for
her the this synapse loss (Fig. 3, E and F). Together, our
early synapse loss in AD genetic models,
inhibit data indicate that genetic deletion of C3 amelio-
we crossed APP/PS1 mice, which, similar
depen- rates synapse loss in APP/PS1 mice, providing
to the J20 mice, had a significant increase
ffects of further evidence that the classical complement
and localization of C1q and C3 onto hip-
e hippo- cascade mediates early synapse loss in AD mouse
pocampal synapses (figs. S2C and S9), to
one had models.
C3-deficient mice (18). Quantification of
T mouse Microglia express complement receptors and
colocalized pre- and post-synaptic puncta
f oAb to mediate synaptic pruning in the developing brain
demonstrated synapse loss in 4 mo APP/
hippo- (1, 6), raising the question of whether this normal
PS1 hippocampus as compared to WT;
y signif- developmental pruning pathway could be acti-
however, APP/PS1xC3 KO mice did not
P by oAb vated to mediate synapse loss in the preplaque
display this synapse loss (Fig. 3, E and F).
control AD brain. Consistent with this hypothesis, mi-
Together, our data indicate that genetic
(fig. S8). croglia had increased amounts of the lysosomal
deletion of C3 ameliorates synapse loss in
al slices protein CD68 in J20 hippocampus compared
APP/PS1 mice, providing further evidence
iator of to WT and less so in striatum, a less vulnerable
that the classical complement cascade
on. region (figs. S1C and S10). Furthermore, in WT
mediates early synapse loss in AD mouse
romotes mice challenged with oAb, microglia had sig-
models.
bsets of nificantly increased levels of CD68 immuno-
Microglia express complement recep-
s down- reactivity (Fig. 2C). However, in C1qa KO mice
tors and mediate synaptic pruning in the
owever, in which synapse loss was rescued, oAb failed
developing brain (1, 6), raising the ques-
n in WT to induce such an increase (Fig. 2D), suggesting
tion of whether this normal developmen-
was sig- that microglia eliminate synapses through the
tal pruning pathway could be activated
fig. S7A, complement pathway.
to mediate synapse loss in the preplaque
ntibody– To directly test whether phagocytic microg-
AD brain. Consistent with this hypoth-
that the lia engulf synaptic elements, we adapted our
esis, microglia had increased amounts
ream of in vivo synaptic engulfment assay (19) using in-
of the lysosomal protein CD68 in J20
nsistent tracerebroventricular injections of Ab in homer-
hippocampus compared to WT and less
f PSD95 GFP (green fluorescent protein) mice (20) (Fig.
so in striatum, a less vulnerable region
1 brains 4A). oAb induced a significantly higher volume
(figs. S1C and S10). Furthermore, in WT
ther C3 of internalized homer-GFP in microglia than
mice challenged with oAβ, microglia had
genetic monomeric Ab controls did at the contralateral
significantly increased levels of CD68
h, simi- hippocampus (Fig. 4B), indicating that microglia
immuno-reactivity (Fig. 2C). However,
ncrease engulf synaptic elements when challenged with
in C1qa KO mice in which synapse loss
ocampal oAb. Internalized homer-GFP often colocalized
was rescued, oAβ failed to induce such
20 6 MAY 2016 • VOL 352 ISSUE 6286 7 15
RE S EAR CH | R E P O R T S S E C T I O N T WO | RESEARCH ARTICLES
We propose
Fig. 4. Microglia engulf synapsesavia model in which C1q and
CR3 upon pathway also plays a role in plaque-as- ACKNOWLEDGMENTS
oAβ operate
oligomeric Ab challenge. in a common
(A) Orthogonal view ofpathway to ac- sociated synapse loss or in other syn- We thank B. Sabatini (HMS), T. Bartels (BWH), and
tivateimage
the complement cascade and drive aptopathies, including tauopathies and members of the Stevens laboratory for critical reading of
high-resolution confocal shows colocalization the manuscript; L. Dissing-Olesen (BCH) for help with the
of homer-GFP andsynapse
Iba1 (red).elimination by microglia through
(B) Three-dimensional Huntington’s disease. If so, our findings
conceptual figure (fig. S12), M. Ericsson [HMS electron
reconstruction and CR3 (fig. rendering
surface S12). Thisusingcould occur in multiple
Imaris may suggest complement and microglia microscopy (EM) facility] for EM imaging, K. Kapur (BCH)
demonstrate larger ways: Soluble
volumes oAβ associates
of homer-GFP puncta with synaptic as potential early therapeutic targets in for advice on statistics, D. M. Walsh (BWH) for Ab oligomers
(S26C), S. Okabe (University of Tokyo) for homer-GFP mice,
inside microglia ofmembranes and other synaptic
oAb-injected contralateral hippo- markers (4, AD and other neurodegenerative dis-
and M. Leviten and T. Yednock (Annexon Biosciences) for
campus versus those26);ofthus, oAβ bound to
monomer-injected. (C)synapses
Mi- may an- eases involving synaptic dysfunction and
characterization and advice on the ANX-M1 anti-C1q antibody;
chor C1q directly. Alternatively,
croglia of homer-GFPxCR3 KO mice (right panel) oAβ bind- memory decline. D. Richardson (Harvard Center for Biological Imaging), A. Hill
ing toof synapses
show less engulfment homer-GFPmay when weaken
chal- the synapse BCH Intellectual and Developmental Disabilities Research
(4) and expose a C1q receptor. Although Center Cellular Imaging Core NIH-P30-HD-18655, and H.
lenged with oAb versus those of homer-GFP mice REFERENCES AND NOTES Elliot and T. Xie (HMS Image and Data Analysis Core) for
specific
(left panel). (D) Ab oligomersreceptors for syn-
failed to induce C1q at synapses 1. S. Hong, L. Dissing-Olesen, B. Stevens, Curr. Opin.
assistance with imaging and data analysis; and S. Kim (BWH),
are not yet known, we have shown that Neurobiol. 36, 128–134 (2016).
apse loss in the contralateral hippocampus of CR3 K. Colodner (BCH), and S. Matousek (BWH) for assistance
C1q as binds synapses 2. T. Wyss-Coray, J. Rogers, Cold Spring Harb. Perspect. Med.
KO mice (right panel) they did in WT micein(left
vulnerable re-
2, l a006346 (2012).
with mice. The J20 mice, C1qa KO mice, P2RY12 antibody,
and the ANX-M1 C1q function-blocking antibody are available
panel). Scale bar, 5 mm (A and B). Means ± SEM; n loss (5, 27). It
gions undergoing synapse 3. M. E. Benoit et al., J. Biol. Chem. 288, 654–665 (2013).
from L. Mucke, M. Botto, O. Butovsky, and A. Rosenthal
is also
= 3 mice per treatment plausible
group that (n
per genotype oAβ= 6 and C1q may 4. L. Mucke, D. J. Selkoe, Cold Spring Harb. Perspect. Med. 2,
under material transfer agreements with UCSF Gladstone,
to 17 microglia analyzed per mouse). *P < 0.05, synapse loss
work indirectly to mediate a006338 (2012).
Imperial College London, BWH, and Annexon Biosciences,
through 5. B. Stevens et al., Cell 131, 1164–1178 (2007).
**P < 0.01, or ***P < 0.0001cytokines sucht test
using two-tailed as transforming
6. D. P. Schafer et al., Neuron 74, 691–705 (2012).
respectively. A.R. is a cofounder, consultant, and chairman of
growth factor–β (7), through microglial the board of directors; B.A.B. is a cofounder and chairman of
(B) or two-tailed one-sample t test (C and D). ns, 7. A. R. Bialas, B. Stevens, Nat. Neurosci. 16, 1773–1782 the scientific advisory board; and B.S. serves on the scientific
not significant. or astrocytic activation, or through other (2013). advisory board of Annexon LLC. A.R., B.A.B., and B.S. are
mechanisms, including major histocom- 8. S. T. DeKosky, S. W. Scheff, Ann. Neurol. 27, 457–464 minor shareholders of Annexon LLC. All other authors declare
patibility complex class I (MHCI)–PirB, (1990). no competing financial interests related to this project.
another immune pathway critical for 9. R. D. Terry et al., Ann. Neurol. 30, 572–580 (1991). The following patents related to this project have been
10. S. Hong, D. Wilton, B. Stevens, D. S. Richardson, granted or applied for: PCT/2015/010288 (S.H. and B.S.),
synapse elimination in development and
Structured Illumination Microscopy for the Investigation of US14/988387 and EP14822330 (S.H., A.R., and B.S.), and
AD (28–30). Synaptic Structure and Function. In Methods in Molecular US8148330, US9149444, US20150368324, US20150368325,
Finally, our studies show that resident Biology; Synapse Development: Methods and Protocols. US20150368326, and US20120328601 (B.S. and B.A.B.).
microglia in the adult central nervous 11. L. Mucke et al., J. Neurosci. 20, 4050–4058 (2000). This work was funded by an Edward R. and Anne G. Lefler
12. S. Hong et al., J. Neurosci. 31, 15861–15869 (2011). Fellowship (S.H.), Coins for Alzheimer’s Research Trust
an increase (Fig. 2D), suggesting that mi- synaptic material when challenged by system phagocytose synapses when chal-
13. A. H. Stephan et al., J. Neurosci. 33, 13460–13474 (B.S.), Fidelity Biosciences Research Initiative (F-Prime)
To further address whether the increase of C1q synapse loss in WT mice (Fig. 3B). Thus, block- synapses (figs. S2C and S9),
lenged by to C3-deficient mice
synaptotoxic oAβ, implicating
croglia eliminate synapses through the oAβ through a complement-dependent (2013). (B.S. and C.A.L.), JPB Foundation (B.A.B.), the National
is dependent on soluble Ab, and if so, which ing C1 activation by either genetic or antibody- (18). Quantification of colocalized
microglia pre- and post-
as potential cellular media- Institutes of Health AG000222 (S.H.), National Institute
complement pathway. mechanism. 14. J. A. Harris et al., J. Neurosci. 30, 372–381 (2010).
species, we injected soluble Ab oligomers or mediated means lessened oAb’s synaptotoxic synaptic punctators demonstrated
of synapse synapse loss in
loss. Although microglia 15. J. L. Jankowsky et al., Hum. Mol. Genet. 13, 159–170 of Neurological Disorders and Stroke–NIH R01NS083845
To directly test whether phagocytic Synaptic deficits occur in early AD and (D.J.S.), National Institute on Aging–NIH 1RF1AG051496A
monomers into lateral ventricles of WT mice. effects. 4 mo APP/PS1 hippocampus
and complement as compared to WT; are promi-
activation (2004).
microglia engulf synaptic elements, we mild cognitive impairment before onset of (B.S.). Supplementary materials contain additional data,
Hippocampus contralateral to the injection site To determine whether C1q is associated with however, APP/PS1xC3 16. M. Botto et al., Nat. Genet. 19, 56–59 (1998).
adapted our in vivo synaptic engulfment plaques and are some of the first signs of nentlyKOinvolved
mice did in notplaque
display maintenance including materials and methods. S.H. and B.S. designed
17. D. B. Freir et al., Neurobiol. Aging 32, 2211–2218 (2011).
was examined to avoid any surgery-related ef- synaptic dysfunction, we asked whether the this synapse lossand (Fig. related
3, E and F). Together, our
periplaque neuropathology, the study and wrote the manuscript, with help from all
assay (19) using intracerebroventricular the neuronal degenerative process (4, 23– 18. M. R. Wessels et al., Proc. Natl. Acad. Sci. U.S.A. 92,
fects. Oligomeric Ab (oAb), which is prefibrillar established ability of oAb to potently inhibit data indicate that genetic
their deletion
roles haveofheretofore
C3 amelio- been largely 11490–11494 (1995). authors. S.H. performed most experiments and data analysis;
injections of Aβ in homer-GFP (green flu- 25). Here we identify critical synaptotoxic V.F.B.-G. and B.M.N. performed microglial activation and
in nature and acts as a mediator of synapse loss long-term potentiation (LTP) (4) was depen- rates synapse loss in APP/PS1
regarded as mice, providing event related
a secondary 19. D. P. Schafer, E. K. Lehrman, C. T. Heller, B. Stevens, J. Vis. Exp.
orescent protein) mice (20) (Fig. 4A). oAβ roles of complement and microglia in AD engulfment experiments along with immunohistochemistry;
and dysfunction in AD (4), but not the relatively dent on C1q. We tested the functional effects of further evidencetothat the classical complement
neuroinflammation (2). Our studies 88, 51482 (2014).
induced a significantly higher volume models before plaque formation and neu- 20. T. Ebihara, I. Kawabata, S. Usui, K. Sobue, S. Okabe, J. S.R. and K.M.M. performed C1q immunohistochemistry; A.F.
innocuous monomeric Ab or vehicle, induced the ANX-M1 anti-C1q antibody in acute hippo- cascade mediatesdirectly
early synapse loss in ADthis
challenge mouseview and sug- performed FISH; S.L. performed electrophysiology; Q.S. and
of internalized homer-GFP in microglia roinflammation, in regions of the hippo- Neurosci. 23, 2170–2181 (2003).
C1q deposition (Fig. 2A and fig. S5). A higher campal slices treated with oAb. IgG alone had models. gest that microglia and immune-related 21. A. Coxon et al., Immunity 5, 653–666 (1996). C.A.L. assisted with design and collection of APP/PS1 tissue;
than monomeric Aβ controls did at the campus undergoing synapse loss. Using A.R. and B.A.B. designed and characterized the ANX-M1
percentage of PSD95 colocalized with C1q in negligible effects on LTP induction in WT mouse Microglia express complement
pathways can actreceptors and mediators of
as early 22. O. Butovsky et al., Nat. Neurosci. 17, 131–143 (2014).
contralateral hippocampus (Fig. 4B), in- multiple experimental approaches, we antiC1q antibody; and D.J.S. contributed in the discussions
oAb-injected versus monomer-injected mice (Fig. hippocampal slices and on the ability of oAb to mediate synapticsynapse
pruning inlossthe developing brain
and dysfunction that oc- 23. D. J. Selkoe, Science 298, 789–791 (2002).
dicating that microglia engulf synaptic demonstrate a region-specific increase of 24. S. W. Scheff, D. A. Price, F. A. Schmitt, E. J. Mufson, and experimental design.
2B), in a manner similar to this colocalization in inhibit LTP; however, pretreatment of hippo- (1, 6), raising the cur
question
in AD of whether
models this normalplaques form.
before
elements when challenged with oAβ. In- phagocytic microglia and accumulation Neurobiol. Aging 27, 1372–1384 (2006).
J20 mice. Together, these findings show an early campal slices with the anti-C1q antibody signif- developmental pruning
Although pathway could be acti- pathway may
the complement SUPPLEMENTARY MATERIALS
ternalized homer-GFP often colocalized of C1q and C3 on synapses in preplaque 25. S. W. Scheff, D. A. Price, F. A. Schmitt, S. T. DeKosky, E. J.
and aberrant increase and synaptic localization icantly prevented the impairment of LTP by oAb vated to mediatenot synapse loss in the
be involved preplaque
in all pathological routes Mufson, Neurology 68, 1501–1508 (2007). www.sciencemag.org/cgi/content/full/science.aad8373/DC1
with CD68 (fig. S11A), suggesting that the brains. Microglia in the adult brain, when
of C1q in multiple AD model systems. Further- (Fig. 3C). Neither ANX-M1 nor its IgG control AD brain. Consistentto AD, with this hypothesis,
including mi-
plaque-associated syn- 26. S. Hong et al., Neuron 82, 308–319 (2014). Materials and Methods
engulfed synapses are internalized into ly- challenged with synaptotoxic, soluble Aβ
more, fluorescent in situ hybridization (FISH) altered basal synaptic neurotransmission (fig. S8). croglia had increased
apse amounts
loss, the of work
the lysosomal
reported here pro- 27. A. H. Stephan, B. A. Barres, B. Stevens, Annu. Rev. Figs. S1 to S12
sosomal compartments in a manner similar oligomers, engulf synapses in the absence
demonstrated up-regulated C1qa expression in Collectively, these results in hippocampal slices protein CD68 invides J20 hippocampus
new insights compared
into how synapses Neurosci. 35, 369–389 (2012).
to that of developmental synaptic pruning of plaque aggregates; deletion of CR3
microglia (fig. S6), implicating microglia as a major and in mice support C1q as a key mediator of to WT and less so arein striatum,
lost in AD. a lessItvulnerable
will be important in 28. A. Datwani et al., Neuron 64, 463–470 (2009). 10 November 2015; accepted 18 March 2016
(6). Notably, oAβ failed to increase synaptic blocks this process. Finally, inhibiting
source of C1q in these preplaque brains. oAb-induced synaptic loss and dysfunction. region (figs. S1Cfuture
and S10). Furthermore,
studies in WT whether this
to examine 29. T. Kim et al., Science 341, 1399–1404 (2013). Published online 31 March 2016
engulfment in microglia lacking CR3 (21), C1q, C3, or CR3 activity rescues synaptic
To test whether C1q and oAb act in a common In the healthy developing brain, C1q promotes mice challengedmicroglia
with oAb, or microglia had sig- 30. H. Lee et al., Nature 509, 195–200 (2014). 10.1126/science.aad8373
a high-affinity receptor for C3 expressed on loss and dysfunction. the complement-dependent
pathway to eliminate synapses, we injected oAb activation of C3, which opsonizes subsets of nificantly increased levels of CD68 immuno-
macrophages [homer-GFPxCR3 KO versus Our data suggest a local activation of a
into lateral ventricles of C1qa knockout (KO) mice synapses for elimination, a process that is down- reactivity (Fig. 2C). However, in C1qa KO mice
homer-GFP mice, which received tail vein developmental pruning pathway (5, 6) as
(16). Soluble oAb induced a significant loss of co- regulated in the mature brain (5, 6). However, in which synapse loss was rescued, oAb failed
injections of phosphate-buffered saline a key mechanism underlying oAβ-induced
localized synapsin- and PSD95-immunoreactive oAb induced a significant C3 deposition in WT to induce such an increase (Fig. 2D), suggesting
(PBS) or oAβ (Fig. 4C)]. These data demon- synapse loss in preplaque AD brain. C1q
puncta in WT mice within 72 hours (Fig. 3A, left adult mice (fig. S7A, upper panel). This was sig- that microglia eliminate synapses through the
strate that CR3 is necessary for oAβ-depen- is aberrantly increased by diffusible oAβ
panel) (17). In contrast, oAb failed to induce syn- nificantly reduced in both the C1qa KO (fig. S7A, complement pathway.
dent engulfment of synapses by microglia. in a region-specific manner and deposits
apse loss in C1qa KO mice (Fig. 3A, right panel), lower panel) and the ANX-M1 anti-C1q antibody– To directly test whether phagocytic microg-
To test whether inhibition in microglial onto synapses, triggering the activation of
suggesting that C1q is required for oAb-induced treated WT mice (fig. S7B), suggesting that the lia engulf synaptic elements, we adapted our
engulfment leads to protection against downstream classical complement path-
synapse loss in vivo. To determine whether local, C3 deposition in this model is downstream of in vivo synaptic engulfment assay (19) using in-
oAβ-induced synapse loss, we performed way and phagocytic microglia. Blocking
acute inhibition of C1 activation could similarly the classical complement cascade. Consistent tracerebroventricular injections of Ab in homer-
tail vein injections of oAβ into WT and Aβ production in J20 mice significantly
blunt the synaptotoxic effects of oAb, we used with these findings, a higher percentage of PSD95 GFP (green fluorescent protein) mice (20) (Fig.
CR3 KO mice. oAβ induced synapse loss ameliorated C1q deposition in the hip-
an antibody against C1q (anti-C1q) (ANX-M1, colocalized with C3 in J20 and APP/PS1 brains 4A). oAb induced a significantly higher volume
in the hippocampus of WT mice but not pocampus, and genetic or antibody-
Annexon Biosciences), which blocks the classical (Fig. 3D and fig. S9). To determine whether C3 of internalized homer-GFP in microglia than
in that of CR3 KO mice (Fig. 4D). All CR3- mediated inhibition of complement
complement cascade (see fig. S7 and supplemen- is necessary for early synapse loss in AD genetic monomeric Ab controls did at the contralateral
positive microglia were P2RY12-positive blocks oAβ from inducing microglial syn-
tary methods). Coadministration of the ANX-M1 models, we crossed APP/PS1 mice, which, simi- hippocampus (Fig. 4B), indicating that microglia
(fig. S11), indicating that they are resi- aptic engulfment, synapse loss, and LTP
anti-C1q antibody, but not its immunoglobulin G lar to the J20 mice, had a significant increase engulf synaptic elements when challenged with
dent cells (22). Altogether, these results inhibition. These complementary find-
(IgG) isotype control, prevented oAb from inducing and localization of C1q and C3 onto hippocampal oAb. Internalized homer-GFP often colocalized
suggest that resident microglia engulf ings have direct therapeutic relevance.
SCIENCE sciencemag.org sciencemag.org SCIENCE 6 MAY 2016 • VOL 352 ISSUE 6286
SCIENCE sciencemag.org 7 15 21
O PT IM IZING YOUR LI V E -CE L L M I CRO S CO P Y: TR I C K S A N D T R A D E - O F F S S E C T I O N T WO | RESEARCH ARTICLES

the regulation of complex patterning signals
BIOMEDICAL ENGINEERING
during embryogenesis and self-formation of
pluripotent stem cells in three-dimensional
Bioengineering a 3D (3D) stem cell culture (2, 3, 18–22). However,
ectodermal organs, such as the skin, hair folli-
per well of a 96-well plate), iPS cells formed
embryoid bodies (EBs) in 3 days, and the
EBs were allowed to grow for 1 week (Fig.
1B). The relative mRNA expression levels of
undifferentiated iPS markers (Nanog) were
Bioengineered 3D IOS in iPS cell– sors in DP cells (34, 35). The follicle variable
derived explants generated using the
region mediates the hair cycle, which depends
CDB transplantation method
on the activation of Lgr5-positive follicle epi-
In the iPS cell–derived explants generated
via the CDB transplantation method, we
thelial stem cells during the telogen-to-anagen
transition. In addition, the follicle epithelial

integumentary organ system

cles, teeth, and exocrine organs, have not been significantly reduced over the culture period observed that Wnt10b, which regulates the stem cells in the bulge region connect to the
generated with sufficient reproducibility from (Fig. 1C and table S1). By contrast, neural crest dermal papilla (DP) (28–30) and the subcuta- calponin-positive arrector pili muscles (Fig.
pluripotent stem cells in 3D stem cell culture markers, such as Nestin and Pax3, were sig- neous adipose tissue (31) derived from neural 2F). We next investigated whether various

from iPS cells using an in vivo or by transplantation (3).

The integumentary organ system (IOS) is
nificantly increased on day 7 in culture (Fig.
1C). In an immunohistochemical analysis
crest origin, induced the bioengineered hair
follicles and developed the stage of hair fol-
stem/progenitor cells and their niches were
reconstructed in the EGFP-labeled bioengi-

transplantation model
a complex system that plays important roles
in waterproofing, cushioning, protecting
deeper tissues, excreting waste, and thermo-
regulation. The integumentary organs include
Ryoji Takagi,1∗ Junko Ishimaru,1∗ Ayaka Sugawara,1* Koh-ei Toyoshima,2,3,4
the skin and its appendages (hair, sebaceous
Kentaro Ishida,5 Miho Ogawa,2,3,4 Kei Sakakibara,1 Kyosuke Asakawa,2 glands, sweat glands, feathers, and nails)
Akitoshi Kashiwakura,1 Masamitsu Oshima,5 Ryohei Minamide,2 Akio Sato,4 (23). The integumentary organs arise from
Toshihiro Yoshitake,4 Akira Takeda,4 Hiroshi Egusa,6 Takashi Tsuji2,3,5† organ germs through reciprocal epithelial-
mesenchymal interactions in the skin field (4).
The integumentary organ system is a complex system that plays important roles During embryogenesis, the skin field forms
in waterproofing, cushioning, protecting deeper tissues, excreting waste, and through a regulated process of pattern for-
thermoregulation. We developed a novel in vivo transplantation model designated mation, and its appendage organs are then
as a clustering-dependent embryoid body transplantation method and generated induced through epithelial-mesenchymal in-
a bioengineered three-dimensional (3D) integumentary organ system, including teractions according to a typical Turing model
appendage organs such as hair follicles and sebaceous glands, from induced pluripotent of activator and inhibitor signals (23). Regen-
stem cells. This bioengineered 3D integumentary organ system was fully functional eration of the 3D IOS could contribute to re-
following transplantation into nude mice and could be properly connected to surrounding generative therapies for patients with burns,
host tissues, such as the epidermis, arrector pili muscles, and nerve fibers, without scars, and alopecia, and could be used as a
tumorigenesis. The bioengineered hair follicles in the 3D integumentary organ system novel assay system for nonanimal safety test-
also showed proper hair eruption and hair cycles, including the rearrangement of ing of cosmetics and quasi-drugs (5, 24). How-
follicular stem cells and their niches. Potential applications of the 3D integumentary ever, it is difficult to generate the complex 3D
organ system include an in vitro assay system, an animal model alternative, and a IOS using in vitro stem cell culture or in vivo
bioengineered organ replacement therapy. transplantation models and to recapitulate
experience organ dysfunction as a result of the physiological functions of the skin using a
INTRODUCTION disease, injury, or aging (8). Our recent stud- bioengineered 3D IOS.
ies provided the proof of concept that fully Here, we generated a bioengineered 3D
of EBs after 7 days in culture, we observed
that the iPS cell–derived EBs had differenti-
ated into both epithelial and mesenchymal
cells (Fig. 1D). EBs expressed Sox2 and p63
as integumentary markers and expressed
Sox17 as an endodermal epithelial marker.
EBs also expressed proteins for the neural
progenitor marker Pax6 and for neural crest
markers such as Snail and Twist (Fig. 1D).
These results indicated that both epithelial
and mesenchymal cells were generated in
iPS cell–derived EBs after 7 days in culture.
We next transplanted these EBs under
various conditions into the subrenal capsule
of severe combined immunodeficient (SCID)
mice in vivo. Both single iPS cells and single
EB transplants formed teratoma-like tissues,
which contained three germ layers, including
neural tissue, muscle, cartilage, and bronchial
epithelia, as reported previously (Fig. 1E, top
and middle) (26). By contrast, in multiple EBs
transplanted using collagen gel, which con-
tained more than 30 EBs cultured for 7 days
under nonadhesive conditions (Fig. 1F), cys-
tic epithelia were observed in the CDB trans-
plants (Fig. 1E, bottom). The tissue weight of
licles compared with the not-stimulated con-
dition (Fig. 2A). The morphology of EBs and
the expression of Sox2, p63, Sox17, Pax6, Snail,
and Twist were not different with or without
Wnt10b signaling (Fig. 1D and fig. S3). The
hair shaft in the explant is black (Fig. 2A), al-
though the hair shaft of the SCID mouse is
white, indicating that hair follicles contain
melanocytes that originate from iPS cell–de-
rived neural crest cells. Although the frequen-
cies in the explants with or without Wnt10b
signaling were not different (Fig. 2B), the
number of bioengineered hair follicles in the
CDB explants treated with Wnt10b was drasti-
cally higher than that of the explants without
Wnt10b (Fig. 2C). In the explants treated with
Wnt10b, the bioengineered hair shaft, which
was measured from the tip of the hair shaft to
the DP, was longer than that in the explants
without Wnt10b (Fig. 2D). Moreover, the hair
follicles in explants treated with Wnt10b were
abundant, and mature hair follicle structures
included the dermal sheath, DP, sebaceous
glands, and subcutaneous adipose tissue (Fig.
2E). The expression of Shh, Msx2, Wnt10b,
β-catenin, and Lef1 in the epithelium, and
neered hair follicles of the CDB transplants,
which were produced using the EGFP-labeled
iPS cell clone. All cells composed of bioengi-
neered 3D IOS showed green fluorescence
(Fig. 2G, left panels). Sox9 and CD34 dou-
ble-positive epithelial stem cells were also
observed in the bulge region of the normal iPS-
derived bioengineered hair follicles (Fig. 2G).
Lrig1-positive epithelial stem cells and Lgr5-
positive epithelial cells were also observed in
the proper anatomical regions, which were de-
tected in the upper and lower portions of the
CK15-positive bulge region, respectively (Fig.
2G). The calponin-positive arrector pili muscle
was connected to the bulge region epithelial
cells of both the natural hair follicle and the
bioengineered hair follicle (Fig. 2H). These
results indicate that various follicle stem cells
and the arrector pili muscles were correctly
arranged in the bioengineered hair follicles
constructed using the bioengineered 3D IOS,
which was derived from iPS cells.
Orthotopic transplantation of the iPS
cell–derived bioengineered 3D IOS
We next investigated whether the iPS cell–

O
rganogenesis is a complex process in-
volving tissue self-organization, cell-cell
interactions, regulation of signaling mol-
ecules, and cell movement (1–3). Almost
all organs arise from organ germs, which
are induced by reciprocal epithelial-mesen-
chymal interactions in each organ-forming
field (4). Current regenerative therapy uses
tissue stem cell transplantation to restore
damaged tissues and organs in a wide vari-
ety of diseases (5–7). Next-generation regen-
erative therapy consists of organ replacement
regenerative therapy, which aims to repro-
duce the reciprocal epithelial-mesenchymal
interactions during embryogenesis. This or-
gan regeneration therapy represents a fun-
damental approach to treating patients who
1
Department of Biological Science and Technology, Graduate
School of Industrial Science and Technology, Tokyo University of
Science, Noda, Chiba 278-8510, Japan. 2Laboratory for Organ
Regeneration, RIKEN Center for Developmental Biology, Kobe,
Hyogo 650-0047, Japan. 3Organ Technologies Inc., Minato-ku,
Tokyo 105-0001, Japan. 4Department of Regenerative Medicine,
Plastic and Reconstructive Surgery, Kitasato University of Medicine,
Sagamihara, Kanagawa 252-0374, Japan. 5Research Institute
for Science and Technology, Tokyo University of Science, Noda,
Chiba 278-8510, Japan. 6Division of Molecular and Regenerative
Prosthodontics, Tohoku University Graduate School of Dentistry,
Sendai, Miyagi 980-8575, Japan.
*These authors contributed equally to this work.
†Corresponding author. Email: t-tsuji@cdb.riken.jp
functional regeneration of ectodermal or-
gans, such as teeth, hair follicles, and salivary
and lachrymal glands, could be achieved with
the transplantation of bioengineered organ
germs using the organ germ method (9–13).
However, organ-inductive stem cells in a wide
variety of organs, except hair follicles, exist
only during embryonic organogenesis (2).
Thus, we must develop techniques to recon-
stitute a wide variety of bioengineered organ
germs using pluripotent stem cells, such as
embryonic stem (ES) cells and induced plu-
ripotent stem (iPS) cells (2, 14).
Pluripotent stem cells, such as ES and iPS
cells, can be induced to differentiate into spe-
cific somatic cell lineages using cytokines that
mimic the patterning and positioning signals
during embryogenesis (15). In the embryo,
patterning signals indicating body axis and
organ-forming fields are strictly controlled by
signaling centers according to the embryonic
body plan (16). These complex pattern forma-
tion signals in local areas of the body may
lack centralized organizing signals, as ob-
served in the generation of teratomas, which
include disorganized neural tissues, cartilage,
muscle, and bronchial epithelia (17). Although
it is difficult to control the development of
specific types of compartmentalized tissues,
several groups recently reported neuroecto-
dermal and endodermal organs generated via
IOS from iPS cells, which includes appendage
organs such as hair follicles and sebaceous
glands, using a novel in vivo transplantation
model designated as the clustering-depen-
dent embryoid body (CDB) transplantation
method. We first developed the conditions for
the induction of various epithelial germ layers
in in vivo explants of iPS cells. Wnt signaling
controlled the frequency of the 3D IOS, calcu-
lated by the number of hair follicles in the ex-
plants. After orthotopic transplantation of the
bioengineered IOS into nude mice, the bioen-
gineered hair follicles showed full functional-
ity, including the ability to undergo repeated
hair cycles through the rearrangement of vari-
ous stem cell niches. Our current study thus
demonstrates the development of a functional
bioengineered 3D IOS and the realization of
organ replacement therapy using iPS cells.
RESULTS
In vivo induction of epithelial tissue
using a CDB transplantation method
To generate various epithelia from iPS cells,
we first developed a CDB transplantation
method as an in vivo transplantation model.
To form the complete three-germ-layer epi-
thelial tissue from iPS cells, we established
iPS cells from murine gingiva (Fig. 1A) (25).
In an in vitro culture of iPS cells (3000 cells
CDB transplants was heavy compared with
single EB transplants, and organ formation
was observed in 4 to 9 g of explants (Fig.
1G). The ratio of cystic tissue area in CDB
transplants was significantly larger than that
in explants of single iPS cells and single EB
transplants (Fig. 1H and fig. S1).
We next characterized the types of epithe-
lium generated in the in vivo explants using
the CDB transplantation method, such as
ectodermal epithelium, nerve tissue, integu-
ment, and endodermal epithelium, includ-
ing the respiratory tract and gastrointestinal
tube, by histochemical analyses with specific
antibodies for CK5, CK10, Muc2, Cdx2, Pax6,
villin, Tuj1, CC10, and E-cadherin (Fig. 1I and
fig. S2). In the explants, we identified various
types of epithelium arising from the three
germ layers, including ectodermal epithelium
(CK5-positive), integument (CK10-positive),
and endodermal epithelium, including the re-
spiratory tract (Muc2-positive, CC10-positive,
and CK10-negative) and gastrointestinal tube
(villin-positive and Cdx2-positive). Although
the transplantation of single ES or iPS cells
induced epithelium at low efficiency (27), the
incidence ratio of epithelium was statistically
constant in explants generated via the CDB
transplantation method (Fig. 1J). These re-
sults indicate that our CDB transplantation
method reproduced various epithelia.
of Bmp4 and Notch1 in the mesenchyme of
the bioengineered hair follicles in the CDB
explants cultured with Wnt10b, was similar
to that observed in natural skin on embry-
onic day 18.5 (fig. S4) (11). Thus, in the iPS
cell–derived explants generated via the CDB
transplantation method, we observed the
bioengineered 3D IOS, including the skin,
hair follicles, dermis, sebaceous glands, and
subcutaneous adipose tissue. We confirmed
these structures using other iPS cell clones,
including a gingiva-derived iPS cell clone
(ectodermal origin), the stomach-derived
iPS-Stm-FB/gfp-99-1 clone (endodermal ori-
gin), and the hepatocyte-derived iPS-Hep-
FB/Ng/gfp-103C-1 clone (endodermal origin)
(32). The gingiva- and stomach-derived iPS
cell clones, but not the iPS-Hep-FB/Ng/gfp-
103C-1 clone, generated the bioengineered 3D
IOS, including hair follicles. These results in-
dicate that the generation of a bioengineered
3D IOS is not dependent on the origin of iPS
cell clones from different germ layers because
heterogeneity exists among iPS cell clones.
In hair follicles, various stem cell types
are maintained in particular regions, such as
CD34-, CK15-, and Sox9-positive follicle epi-
thelial cells in the follicle stem cell niche of the
bulge region (33, 34), Lrig1-positive epithelial
stem cells in the upper bulge outer root sheath
region, and multipotent mesenchymal precur-
derived IOS could be transplanted into the
cutaneous environment of adult nude mice
(Fig. 3A, upper panel). We isolated male cys-
tic tissues with hair follicles in the explants,
cut them into small specimens containing 10
to 20 follicular units, and transplanted them
onto the backs of female nude mice using a
follicular unit transplantation method (Fig.
3A, lower panels). At 14 days and thereafter,
we observed the eruption and growth of black
hair shafts (n = 216; Fig. 3B). After 3 months,
we did not observe tumorigenesis in the trans-
plantation areas (n = 171). To confirm that the
transplanted 3D IOS was of iPS cell origin,
we performed Y-chromosome fluorescence in
situ hybridization (Y-FISH) with male mouse–
specific DNA probes (Fig. 3C and fig. S5).
Y-FISH–positive cells were observed in the
transplantation area (including the skin epi-
thelium, dermis, sebaceous glands, intracuta-
neous adipose tissue, and hair follicles) of the
cervical skin of recipient female mice. These
results indicated that the bioengineered IOS
could engraft in the cervical skin and revealed
that it is of iPS cell origin. The engrafted hair
follicle must establish proper connections
with the surrounding tissues, such as the ar-
rector pili muscles and nerve fibers. The hair
follicles in the bioengineered 3D IOS had
connected to calponin-positive arrector pili
muscles in the correct polarization to iPS cell–

22 Originally published 1 April 2016 in SCIENCE ADVANCES sciencemag.org SCIENCE SCIENCE sciencemag.org 23
 O PT IM IZING YOUR LI V E -CE L L M I CRO S CO P Y: TR I C K S A N D T R A D E - O F F S S E C T I O N T WO | RESEARCH ARTICLES
RESEARCH ARTICLE RESEARCH ARTICLE

A Day 0 7 30 C 4.0
Nanog
1.6
Nestin
1.0
Pax3
in vitro in vivo

Relative expression
0.8
3.0 1.2
Sorting Transplantation In vivo 0.6
iPS cells by CDB method organogenesis 2.0 0.8
0.4
1.0 0.4 0.2

0.0 0.0 0.0

Analysis Collagen gel SCID mouse

D0
D3
D5
D7

D0
D3
D5
D7

D0
D3
D5
D7
Cells EBs Cells EBs Cells EBs
B iPS cells EB formation
Day 0 Day 3 Day 5 Day 7

D H&E IHC F
Sox2/p63 Sox17 Pax6 Snail/ Twist Top Side

Downloaded from http://advances.sciencemag.org/ on April 5, 2016

Downloaded from http://advances.sciencemag.org/ on April 5, 2016
E H&E I H&E IHC
Macroscopy Low High Low High CK5/NC CK10/NC

Integumental
iPS cells

CL
CL
CL

Muc2/E-cad Cdx2/E-cad
Single EB

Gastrointestinal

CL CL

CL

Villin/Tuj1 CC10/E-cad
CDB

Respiratory

CL
CL

G H J
14 30
60
*
*
12
25
Area occupancy of epithelial cyst

* 50
(% of each epithelial type)
Frequency of epithelium

10
20
Tissue weight (g)

40
(% of specimen)

8
15 30
6
10 20
4

2 5 10

0 0 0
Single CDB iPS Single CDB Neural Skin/ Upper Lower Unclassified
EB cells EB rosette mucosa endodermal endodermal
tissues tissues

Fig. 1. Induction of epithelial tissues via the CDB transplantation method. (A) Schematic representation of EB cultures and the CDB transplantation
method. (B) Phase-contrast
Fig. 1. Induction images
of epithelial of iPSvia
tissues cellstheandCDB
the transplantation
formation of EBs, which method. were cultured
(A) Schematicin nonadherent
representationplastic wells
of EB for 3, 5,and
cultures or 7thedays.
CDB Scale bars, 100 mm.
transplantation Fig. 2. Analysis of the bioengineered hair follicle induced from iPS cells via the CDB transplantation method. (A) Macroscopic (left panels) and
(C) mRNA(B)
method. expression levels ofimages
Phase-contrast undifferentiated
of iPS cellsiPS andcell
the markers
formation(Nanog)
of EBs,andwhich
neural crest
were cell markers
cultured (Nestin andplastic
in nonadherent during
Pax3)wells forEB 3, formation.
5, or 7 days. *PScale
< 0.001 by
bars, microscopic (H&E staining; right panels) examination of the hair follicles and shafts in the CDB transplants. EBs were stimulated without Wnt10b (upper
Fig. 2. Analysis of the bioengineered hair follicle induced from iPS cells via the CDB transplantation method. Macroscopic (left panels) and
Student’s
100 µm. (C) t test.
mRNA(D) Hematoxylin and eosin
expression levels (H&E) staining iPS
of undifferentiated andcellimmunostaining
markers (Nanog) of EBs,
andusing
neural antibodies
crest cellof epithelial
markers (Sox2/p63
(Nestin and Sox17),
and Pax3) during neural progenitor
EB formation. *P panels) or with Wnt10b (lower panel) on day 7. Scale bars, 100 mm. (B) Frequency of the epithelial tissue, including hair follicle, in CDB transplants. The data
microscopic (H&E staining; right panels) examination of the hair follicles and shafts in the CDB transplants. EBs were stimulated without Wnt10b (upper
(Pax6),
< 0.001and neural crest
by Student’s markers
t test. (Snail and Twist)
(D) Hematoxylin after 7(H&E)
and eosin days. staining
The nuclei andwere stained usingofHoechst
immunostaining EBs, using33258 (white).ofScale
antibodies bars, 50
epithelial mm. (E) Macroscopic
(Sox2/p63 and Sox17), are presented as the means ± SEM from individual experiments; n = 13 (single iPS injection), n = 49 (CDB transplants without Wnt10b), and n = 15 (CDB
panels) or with Wnt10b (lower panel) on day 7. Scale bars, 100 µm. (B) Frequency of the epithelial tissue, including hair follicle, in CDB transplants. The
photographs
neural progenitor (left panels)
(Pax6),andand microscopy
neural crest(H&Emarkersstaining,
(Snailcenter and right
and Twist) afterpanels)
7 days.ofThein nuclei
vivo transplants
were stained under various
using transplantation
Hoechst 33258 (white). conditions.
Scale bars,The50in µm.
vivo transplants with Wnt10b). (C) Number of hair follicles in the CDB transplants. The data are presented as the means ± SEM from individual experiments; n = 13
data are presented as the means ± SEM from individual experiments; n = 13 (single iPS injection), n = 49 (CDB transplants without Wnt10b), and n =
transplants
(E) Macroscopic of 3000photographs
dissociated iPS
(leftcells (upper),
panels) andsingle EBs (middle),
microscopy and morecenter
(H&E staining, than 30 andEBs (lower)
right wereofplaced
panels) in vivointransplants
the subrenal capsule
under for 30transplantation
various days and then (single iPS injection), n = 74 (CDB transplants without Wnt10b), and n = 4 (CDB transplants with Wnt10b). *P < 0.001 by Student’s t test. (D) Comparative
15 (CDB transplants with Wnt10b). (C) Number of hair follicles in the CDB transplants. The data are presented as the means ± SEM from individual
analyzed.
conditions. (F)The
Macroscopic photographs
in vivo transplants of multiple
of 3000 EB in aiPS
dissociated collagen gel before
cells (upper), transplantation.
single EBs (middle), Scale
andbars,
more 1 mm.
than(G)30Weight of thewere
EBs (lower) in vivo transplants.
placed The data
in the subrenal analysis of the length of hair shafts from the hair tip to the DP in CDB explants treated with or without Wnt10b. *P < 0.05 by Student’s t test. (E) Histological
experiments; n = 13 (single iPS injection), n = 74 (CDB transplants without Wnt10b), and n = 4 (CDB transplants with Wnt10b). *P < 0.001 by Student’s t test.
are presented
capsule for 30as the and
days median
then±analyzed.
maximum(F) orMacroscopic
minimum from individual experiments;
photographs of multiple EBnin = a8 collagen
and n = 48 gelper experiment.
before Red circles
transplantation. indicate
Scale bars, the
1 mm.cyst, including
(G) Weight analysis of the hair follicles and their surrounding tissues in iPS cell–derived bioengineered 3D IOSs. The isolated cystic structures with hair follicles were
(D) Comparative analysis of the length of hair shafts from the hair tip to the DP in CDB explants treated with or without Wnt10b. *P < 0.05 by Student’s t
hair follicles,
of the in vivointransplants.
the explants. *Pdata
The < 0.001
are by Student’s
presented test.median
ast the (H) The±area occupancy
maximum of the cystic
or minimum fromlumen in the
individual whole specimens
experiments; n = 8 andof innvivo
= 48transplants of the
per experiment. observed macroscopically (left panels). In the H&E analyses, the boxed areas in the low-magnification macroscopic views are shown at a higher magnification
test. (E) Histological analysis of the hair follicles and their surrounding tissues in iPS cell–derived bioengineered 3D IOSs. The isolated cystic structures
three types of
Red circles conditions
indicate wasincluding
the cyst, compared. *Pfollicles,
hair < 0.001 in bytheStudent’s
explants. t test.
*P <(I)0.001
Histochemical
by Student’sandtimmunohistochemical
test. (H) The area occupancy analyses ofofthethe cystic
cystic epithelia
lumen inwhole
in the the in in other panels. CL, cystic lumen; EL, epidermal layer; DL, dermal layer; SD, subdermal tissue; SG, sebaceous gland; HO, hair opening; v, vessel. Scale bars,
with hair follicles were observed macroscopically (left panels). In the H&E analyses, the boxed areas in the low-magnification macroscopic views are shown
vivo explants
specimens ofof
in multiple iPS cell–derived
vivo transplants EBs.types
of the three Boxedof areas in thewas
conditions left compared.
panels show *PH&E staining.
< 0.001 To identify
by Student’s epithelial
t test. types, suchand
(I) Histochemical as ectodermal epithelium,
immunohistochemical 500 mm (H&E; upper left) and 100 mm (others). (F) Schematic representation of stem/progenitor cells in the hair follicle and surrounding tissues on the skin at
at a higher magnification in other panels. CL, cystic lumen; EL, epidermal layer; DL, dermal layer; SD, subdermal tissue; SG, sebaceous gland; HO, hair
integument
analyses of (top panels),epithelia
the cystic and endodermal
in the in epithelium,
vivo explants including
of multiple the gastrointestinal
iPS cell–derived tube
EBs.(middle
Boxedpanels)
areas inandtherespiratory
left panels tract
show(bottom
H&E panels),
staining.we Toanalyzed
identify the anagen and telogen phases. APM, arrector pili muscle. (G) Immunohistochemical analyses of stem/progenitor cells in the follicles of natural pelage and
opening; v, vessel. Scale bars, 500 µm (H&E; upper left) and 100 µm (others). (F) Schematic representation of stem/progenitor cells in the hair follicle
CDB transplants
epithelial by immunostaining
types, such with specificintegument
as ectodermal epithelium, antibodies(top for CK5, CK10,
panels), and Muc2, Cdx2, villin,
endodermal CC10, Tuj1,
epithelium, and E-cadherin.
including The nuclei
the gastrointestinal were
tube stained
(middle using
panels) enhanced green fluorescent protein (EGFP)–labeled iPS cell–derived bioengineered hair follicles. These samples were immunostained with anti-Sox9 (red),
and surrounding tissues on the skin at the anagen and telogen phases. APM, arrector pili muscle. (G) Immunohistochemical analyses of stem/progenitor
Hoechst 33258 (blue).
and respiratory To identify
tract (bottom the nonspecific
panels), we analyzedfluorescence
CDB transplants signals in these immunohistochemical
by immunostaining analyses,for
with specific antibodies weCK5,
performed
CK10, Muc2, the experiments under
Cdx2, villin, CC10, anti-CD34 (green), anti-CK15 (red), anti-Lrig1 (white), and anti-Lgr5 (red) antibodies. Arrows, epithelial cells of bulge regions; arrowheads, Lrig1-positive
cells in the follicles of natural pelage and enhanced green fluorescent protein (EGFP)–labeled iPS cell–derived bioengineered hair follicles. These samples
the conditions
Tuj1, withoutThe
and E-cadherin. specific antibodies
nuclei against
were stained antigens
using Hoechst [negative
33258control
(blue).(NC)]. Scale bars,
To identify 1 mm (low-magnification
the nonspecific fluorescence images)
signals inand mm (high-magnification
100 immunohistochemical
these cells; *, background fluorescence of hair shafts. Scale bars, 200 mm (upper panels) and 50 mm (lower panels). (H) Histological and immunohistochemical
were immunostained with anti-Sox9 (red), anti-CD34 (green), anti-CK15 (red), anti-Lrig1 (white), and anti-Lgr5 (red) antibodies. Arrows, epithelial cells of
images).
analyses, (J)weThe frequencythe
performed of experiments
epithelial typesunderin CDB
thetransplants. Epithelialspecific
conditions without types inantibodies
CDB transplants
againstwere classified
antigens basedcontrol
[negative on the(NC)].
cell morphology
Scale bars,and 1 mmnumber
(low- analyses of the iPS cell–derived hair follicles and their surrounding tissues. Natural pelage and iPS cell–derived hair follicles were stained with calponin.
bulge regions; arrowheads, Lrig1-positive cells; *, background fluorescence of hair shafts. Scale bars, 200 µm (upper panels) and 50 µm (lower panels).
count. The dataimages)
magnification are presented
and 100asµm the means ± SEM fromimages).
(high-magnification individual (J)experiments;
The frequency n= of5.
epithelial types in CDB transplants. Epithelial types in CDB transplants Arrowheads indicate calponin-positive arrector pili muscles. Scale bars, 200 mm.
(H) Histological and immunohistochemical analyses of the iPS cell–derived hair follicles and their surrounding tissues. Natural pelage and iPS cell–derived
were classified based on the cell morphology and number count. The data are presented as the means ± SEM from individual experiments; n = 5. hair follicles were stained with calponin. Arrowheads indicate calponin-positive arrector pili muscles. Scale bars, 200 µm.
Takagi et al. Sci. Adv. 2016; 2 : e1500887 1 April 2016 3 of 10 Takagi et al. Sci. Adv. 2016; 2 : e1500887 1 April 2016 4 of 10

24 sciencemag.org SCIENCE SCIENCE sciencemag.org 25

O PT IM IZING YOUR LI V E -CE L L M I CRO S CO P Y: TR I C K S A N D T R A D E - O F F S RESEARCH ARTICLE S E C T I O N T WO | RESEARCH ARTICLES

derived follicles, which were identical to natu-
ral pelage follicles (Fig. 3D). Nerve fibers had
also formed proper connections to the bulge
region of the bioengineered hair follicles (Fig.
3D, lower panels). These results indicate that
the iPS cell–derived hair follicles in the bio-
engineered IOS were correct in structure and
established the proper connections to the sur-
rounding tissues.
Analysis of iPS cell–derived hair species
and distribution
We next investigated the hair species and
their distribution in the bioengineered IOS
in the cutaneous environment of recipient
adult nude mice. The iPS cell–derived IOS
produced all types of pelage hairs, including
zigzag and awl/auchene and guard, in a simi-
mentalized organized tissues derived from
three germ layers, such as neural tissue (ec-
todermal), cartilage (mesenchymal), and
gut/bronchial epithelia (endodermal), by an
in vivo injection assay (37). Several cutane-
ous types of mature teratomas, such as cystic
dermoids in the ovary and orbital regions,
generate ectodermal organs, such as teeth
and hair follicles (17). However, the molecu-
lar mechanisms underlying the development
of these dermoids are still unknown. In an
in vivo injection assay of ES and iPS cells,
ectodermal organs could not be detected at
a considerable frequency in pluripotent stem
cell–derived teratomas. In 3D stem cell cul-
ture, bioengineered 3D tissues are limited to
neuroectodermal and endodermal tissues (2).
Thus, to generate a 3D IOS, we developed a
the replacement of skin physiological func- Fig. 3. Transplantation of the bioengineered 3D IOS.
tions, such as the protection of deeper tis- (A) Schematic representation of the methods used for the
sues, waterproofing, and thermoregulation generation and transplantation of iPS cell–derived hair follicles.
(5). However, it is well recognized that current Cystic tissue with hair follicles was isolated and divided into small
therapies using cultured epithelial tissues suf- pieces containing 10 to 20 hair follicles (lower left panels). The
fer from critical issues, including aesthetics small pieces were transplanted into the back skin of nude mice
and the inability to excrete sweat and lipids using a follicular unit transplantation (FUT) method developed
from exocrine organs (44). In addition, artifi- in humans (lower right panels). Scale bars, 500 µm (lower left
cial skin consisting of cultured epidermis and panels) and 1 mm (lower right panels). (B) Macromorphological
dermis without appendage organs and pores observations of two independent engraftments into the
is often used for testing during the develop- dorsoventral skin of nude mice showing the eruption and growth
ment of cosmetics and quasi-drugs, although of iPS cell–derived hair follicles. Scale bars, 1 mm. (C) Y-FISH
this method has critical limitations related to analysis of the CDB transplants using male mouse–specific DNA
the permeability of cosmetics and drugs and probes. H&E and differential interference contrast (DIC) images
the reliability of physiological responses (45). are shown in the left panels. FISH images are shown in the upper
By contrast, it is expected that bioengineered left panel. Green and blue signals indicate Y-chromosome–
3D IOS can overcome these issues (46). Here, positive cells and nuclei, respectively. Boxed areas in the FISH
we provide evidence for the generation of a image are shown at a higher magnification in the right panels.

Downloaded from http://advances.sciencemag.org/ on April 5, 2016

lar population to the back skin of adult mice novel CDB transplantation method to induce 3D IOS from iPS cells using an in vivo trans- Boxed areas U1 and U2 indicate the skin epithelium and upper
(Fig. 4, A and B) (36). We found that the dis- ectodermal tissues at a high frequency and plantation model. The explants were fully region of the hair follicle. U3 indicates the bulge region. Boxed
tance between bioengineered hair follicles successfully generated a 3D IOS in in vivo functional and included hair follicles and area sU3 indicates the Y-FISH–positive sebaceous gland isolated
was not significantly different from that on explants. In general, cultured EBs form an sebaceous glands with proper connections to from fig. S5. Boxed areas L1 and L2 indicate the lower hair bulb
the skin of normal mice (Fig. 4C). We further outer layer epithelial progenitor (38, 39). the surrounding tissues, such as the epithe- region. Boxed area L3 indicates the intracutaneous adipose
analyzed the distribution of hair shaft spe- Epithelium induction in our method may lium, dermis, fat, arrector pili muscles, and tissue. Broken lines in the differential interference contrast
cies in the transplanted iPS cell–derived IOS be due to the connection of the outer layer nerve fibers. Although the origin of nerve image indicate the outermost limit of the dermis. Scale bars, 200
and compared the result with the natural epithelium of each EB by self-assembly and fibers is still unclear, nerve fibers would be µm (whole images) and 100 µm (high-magnification images).
skin of adult murine trunk region, including self-formation in the explants. Furthermore, innervated from host tissues. Our previous (D) Analysis of the integration with surrounding tissues, such as
the body center and lumbar region of the side the frequency of hair follicles with sebaceous studies of ectodermal organ transplantations the arrector pili muscles and nerves, in the cervical skin of nude
trunk lining on linea axillaris media (fig. glands in the in vivo explants of our bioen- demonstrated that the innervation into the mice. The differential interference contrast images show black hair
S6). We also observed that the distribution gineered 3D IOS could be controlled by Wnt explants had been indicated from host tissues shafts derived from the bioengineered 3D IOS (origin: C57BL/6
of these hair species in the transplanted IOS signaling. Hence, our CDB transplantation (10–13). mice) among the white hairs of nude mice. The arrector pili
was not significantly different from that in method can be used to generate a bioengi- In conclusion, we generated a fully func- muscles and nerve fibers were analyzed by immunohistochemical
normal mice skin (fig. S6C). We further ana- neered 3D IOS and contributes to our under- tional 3D IOS by the self-assembly of epithe- staining using specific antibodies against calponin (red) for
lyzed the hair cycles, which comprise alter- standing of the molecular mechanisms of the lial and mesenchymal stem cells from iPS smooth muscle and neurofilament H (NF-H; white). The nuclei
nating growth (hair growing) and regression onset of dermoid formation in vivo. cells using the CDB transplantation method. (Nuc) were stained using Hoechst 33258 (blue). The boxed
(nongrowing) phases, of the iPS cell–derived Wnt signaling pathways are essential for Our study contributes to the development of areas in the left panels are shown at a higher magnification in the
hair follicles for 90 days (Fig. 4D). The iPS body patterning during embryogenesis and bioengineering technologies that will enable right panels. The arrows and arrowheads indicate nerve fibers
cell–derived follicles repeated the hair cycle organ induction through epithelial-mesen- future regenerative therapies for patients and muscles connected with hair follicles, respectively. Broken
at least three times during the transplanta- chymal interactions (40). Wnt family mol- with burns, scars, and alopecia. Further opti- lines indicate the outermost limit of each hair follicle. Scale bars,
tion period (Fig. 4D), and no significant dif- ecules play important roles in forming an mization of our technique using in vitro stem 200 µm (low-magnification photographs) and 100 µm (high-
ferences in the hair cycle periods were found organ-forming field and the initial organ bud cell culture and humanization will contrib- magnification photographs).
between natural and bioengineered follicles in both epithelial and mesenchymal tissues ute to the development of bioengineered IOS
(both zigzag and awl/auchene and guard; (41). Wnt10b signaling is known to play a key therapy as a prominent class of organ system
Fig. 4E). These results suggest that the CDB role in the progression of an initial organ replacement regenerative therapy and as a
transplantation method induced a skin field bud to an organ germ, and also regulates the novel nonanimal assay system for cosmetics
during early development by the subrenal reciprocal epithelial-mesenchymal interac- and quasi-drugs in the future.
capsule transplantation and generated the tions of ectodermal organs, such as teeth,
bioengineered IOS that behaved like natural hair follicles, and exocrine organs (28). In METHODS
mouse skin. hair follicle organogenesis, Wnt10b signal-
ing regulates the differentiation from neu- Cell culture of mouse iPS cells
ral crest–derived mesenchyme into DP and
Fig. 3. Transplantation of the bioengineered 3D IOS. (A) Schematic representation of the methods used for the generation and transplantation of iPS
The mouse iPS cell line “gingiva-derived
DISCUSSION cell–derived hair
the surrounding fat tissue (30, 42, 43). Here, iPS” was established by Egusa et al. (25) at follicles. Cystic tissue with hair follicles was isolated and divided into small pieces containing 10 to 20 hair follicles (lower left panels). The
Our current findings reveal the generation the frequency of hair follicle formation, but small pieces were transplanted into the back skin of nude mice using a follicular unit transplantation (FUT) method developed in humans (lower right panels).
Tohoku University. The iPS-Stm-FB/gfp-99-1
of the bioengineered 3D IOS from iPS cells, not epithelial tissue formation, in in vivo ex- and iPS-Hep-FB/Ng/gfp-103C-1Scale lines weremm (lower left panels)Induction
bars, 500 and 1 mmof(lower
mouse iPS
right cell (B) Macromorphological
panels). Gene expression
observations analysis
of two byindependent engraftments into the(PCR) are listed in table S1. For
chain reaction
including appendage organs such as hair fol- plants was controlled by Wnt10b signaling in purchased from the RIKEN dorsoventral skin of nude mice showing
BioResource the eruption and growth of iPS cell–derived hair real-time
differentiation PCR
follicles. Scale bars, 1 mm. (C) Y-FISH analysis of the quantitative
CDB transplants
analysis, 32 to 48 EBs were ana-
licles and sebaceous glands, with proper con- EBs. After treatment with Wnt10b on day 6 in usingmaintained
Center. All iPS cell lines were male mouse–specific DNA probes. H&E iPS
Mouse and cells
differential interference
were harvested bycontrast
dissocia- (DIC) Total
images are was
RNA shown in thefrom
isolated left panels.
mouseFISH
iPS images
cells are shown
lyzed in the in each experiment, which
in triplicate
nections to surrounding tissues, including the culture, we detected the expression of neural upper left
in the presence of a mitomycin panel. Green and blue signals
C–treated indicate
tion usingY-chromosome–positive cells and nuclei,orrespectively.
0.25% trypsin-EDTA (Invitrogen) EBs using Boxed
TRIzolareas in the
reagent FISHTechnolo-
(Life image are shown at a higher
was repeated at least three times.
epidermis, dermis, arrector pili muscles, fat, crest markers, including Nestin, Pax3, Snail, magnification
SNLP feeder layer on gelatin-coated in the right panels. Boxed
dishes. withareas U1 cells.
feeder and U2 indicate
Then, the skincells
dissociated epithelium
were and gies)upper regiontoof the
according the manufacturer’s
hair follicle. U3 indicates
proto- the bulge region.
and nerve fibers, in an in vivo transplantation and Twist. Therefore, our findings suggest The culture medium for theseBoxedmouse
area sU3iPSindicates the Y-FISH–positive sebaceous
separated into iPSgland isolated
cells and from
feeder fig.using
cells S5. Boxedcol.
areas L1 and
Reverse L2 indicate the
transcription waslower hair bulb
performed with Animals
region. Boxed area L3
model—the CDB method. These findings sig- that Wnt10b signaling plays important roles cells consisted of Dulbecco’sindicates
modified the intracutaneous
Ea- adipose tissue.
FeederBroken
Removal linesMicroBeads
in the differential
(MACS;interference
Miltenyi contrast image indicate
the PrimeScript II 1st the outermost
strand limit of the dermis.
cDNA Synthesis Scale bars,
C.B-17/lcr-scid/scidJcl mice were purchased
nificantly advance the technological develop- in the formation of hair follicles in the skin gle’s medium (DMEM) (Nacalai 200 mm (whole
Tesque)images) and 100 mm (high-magnification images).
Biotec). For EB culture, (D)iPS
the Analysis of there-
cells were integration with surrounding
Kit (TaKaRa tissues, expression
Bio). The mRNA such as the arrector
lev- pili muscles
from CLEA andJapan Inc., and C57BL/6NCrSlc
ment of bioengineered 3D IOS. field, neural crest–derived melanocyte differ- supplemented with 15% fetal bovine
nerves, serum
in the aggregated
cervical skin of nude mice. using Nunclon
The differential interferenceSphera round-bot-
contrast images showels black
were hair
determined usingfrom
shafts derived SYBR thePremix Ex
bioengineered and Balb/c
3D IOS nu/nu mice were purchased
(origin:
Studies of pluripotent stem cells have entiation, and progression of organogenesis (Biosera), 2 mM L-glutamine (Invitrogen),
C57BL/6 1
mice) among tom
the white hairs of 96-well
nude mice. plates (Thermo
The arrector Fisher Scientific
pili muscles Taq were
and nerve fibers II (TaKaRa Bio),
analyzed and the products were
by immunohistochemical from
staining Japan
using SLC Inc. Animal care and han-
specific
demonstrated the ability to generate whole through reciprocal epithelial-mesenchymal × 10−4 M nonessential aminoantibodies
acids (Invitro- Inc.) inmuscle
against calponin (red) for smooth a culture medium (3000Hcells/100
and neurofilament analyzed
µl The
(NF-H; white). with were
nuclei (Nuc) an Applied
stainedBiosystems
using HoechstQuant- dling The
33258 (blue). conformed
boxed to the National Institutes
embryos by tetraploid complementation in interactions during morphogenesis. gen), 1 × 10−4 M 2-mercaptoethanol (Invitro-
areas in the per well).
left panels are shown at a higher Fresh medium
magnification was
in the added
right to the
panels. arrowsStudio
Thecul- 12K Flex (Life
and arrowheads Technologies).
indicate nerve fibersThe musclesof
anddata Health guidelines,
connected with the requirements of the
blastocysts and revealed that these cells can For regenerative therapy in patients with gen), penicillin (50 U/ml), and streptomycin ture on day 3, and Wnt10b (500 ng/ml; R&D were normalized to β-actin expression. The
hair follicles, respectively. Broken lines indicate the outermost limit of each hair follicle. Scale bars, 200 mm (low-magnification photographs) and 100 mm (high-Tokyo University of Science Animal Care and
form teratomas, which include compart- burns, a cultured epithelial sheet is useful for (50 µg/ml; Invitrogen) (25). magnification photographs). Systems) was added on day 6. primer pairs used for real-time polymerase Use Committee (permit no. N13008), and the

26 sciencemag.org SCIENCE SCIENCE sciencemag.org 27

Takagi et al. Sci. Adv. 2016; 2 : e1500887 1 April 2016 6 of 10
R E SR E A
S ER A
C RHCAHR A
T IRCTLI EC L E
O PT IM IZING YOUR LI V E -CE L L M I CRO S CO P Y: TR I C K S A N D T R A D E - O F F S S E C T I O N T WO | RESEARCH ARTICLES

CC10 (T-18) (1:100, goat; Santa Cruz Biotech-
nology), Tuj1 (TU-20) (1:200, mouse; Chemi-
con), and E-cadherin (36/E-Cadherin) (1:200,
mouse; BD Biosciences) in blocking solution.
The primary antibodies were detected using
highly cross-absorbed Alexa Fluor 488 don-
key anti-mouse immunoglobulin G (IgG; H
+ L) (1:500; Life Technologies), Alexa Fluor
488 goat anti-rat IgG (H + L) (1:500; Life
Technologies), Alexa Fluor 488 donkey anti-
rabbit IgG (H + L) (1:500; Life Technologies),
Alexa Fluor 594 donkey anti-mouse IgG (H
+ L) (1:500; Life Technologies), Alexa Fluor
594 donkey anti-goat IgG (H + L) (1:500; In-
vitrogen), Alexa Fluor 594 donkey anti-rabbit
IgG (1:500; Invitrogen), Alexa Fluor 594 goat
explants, were intracutaneously transplanted
into the back skin of 6-week-old Balb/c nu/nu
mice, as described above. The iPS cell–derived
skins after transplantation were dissected at
the first or second anagen phase. The natural
skins and iPS cell–derived skin were fixed and
placed horizontally on a plastic dish before
macroscopic photographs were taken. To ex-
amine the hair types and distribution of the
species of hair shafts, we observed all samples
using the SteREO Lumar.V12 and AxioCam
microscope system (Carl Zeiss). The distance
between hair pores was measured with Axio-
Vision Vs40 Image processing software (Carl
Zeiss). The position of natural and iPS cell–
derived hair pores, hair growth direction, and
35. C. A. B. Jahoda, C. J. Whitehouse, A. J. Reynolds, N. Hole,
Exp. Dermatol. 12, 849–859 (2003).
36. R. R. Driskell, A. Giangreco, K. B. Jensen, K. W. Mulder, F. M.
Watt, Development 136, 2815–2823 (2009).
37. K. Takahashi, S. Yamanaka, Cell 126, 663–676 (2006).
38. C. M. Metallo, L. Ji, J. J. de Pablo, S. P. Palecek, Stem Cells
26, 372–380 (2008).
39. T. Kadoshima et al., Proc. Natl. Acad. Sci. U.S.A. 110, 20284–
20289 (2013).
40. A. Stathopoulos, M. Levine, Dev. Cell 9, 449–462 (2005).
41. L. Jiménez-Rojo, Z. Granchi, D. Graf, T. A. Mitsiadis, Front.
Physiol. 3, 107 (2012).
42. M.-X. Lei, C.-M. Chuong, R. B. Widelitz, J. Invest. Dermatol.
133, 7–9 (2013).
43. X. Lim, R. Nusse, Cold Spring Harb. Perspect. Biol. 5,
a008029 (2013).
44. C. Niemann, F. M. Watt, Trends Cell Biol. 12, 185–192 (2002).
45. E. E. Kandyba, M. B. Hodgins, P. E. Martin, J. Invest.
Dermatol. 128, 1039–1049 (2008).
46. E. Bellas, M. Seiberg, J. Garlick, D. L. Kaplan, Macromol.

Downloaded from http://advances.sciencemag.org/ on April 5, 2016

Downloaded from http://advances.sciencemag.org/ on April 5, 2016
anti-rabbit IgG (H + L) (1:500; Life Technolo- species of hair shafts (classified as zigzag and Biosci. 12, 1627–1636 (2012).
gies), or Alexa Fluor 633 goat anti-rat IgG awl/auchene and guard) in 1-mm2 portions 47. K. Asakawa et al., Sci. Rep. 2, 424 (2012).
Fig. 4. Analysis of iPS cell–
(H + L) (1:500; Invitrogen) with Hoechst were analyzed by the methods described pre-
derived hair types and ACKNOWLEDGMENTS
33258 dye (1:500; Dojindo) for 1 hour at viously (36).
hair cycle. (A) Microscopic We are grateful to T. Irie for the invaluable comments. Funding: This
room temperature. All fluorescence micros- work was partially supported by a Grant-in-Aid for KIBAN (A) from
observation of the iPS cell–
copy images were obtained with an LSM 780 REFERENCES AND NOTES the Ministry of Education, Culture, Sports, Science, and Technology
derived bioengineered hair
confocal microscope (Carl Zeiss) or an Axio 1. M. Takeichi, Dev. Cell 21, 24–26 (2011). (grant 25242041) and by a collaboration grant (to T.T.) from
shafts showing zigzag, awl/ 2. Y. Sasai, Cell Stem Cell 12, 520–530 (2013). Organ Technologies Inc. This work was partially funded by Organ
Scan.Z1 (Carl Zeiss).
auchene, and guard hairs. The 3. Y. Sasai, Nature 493, 318–326 (2013). Technologies Inc. Author contributions: T.T. designed the research
hair shafts were analyzed 4. J. Pispa, I. Thesleff, Dev. Biol. 262, 195–205 (2003). plan; R.T., J.I., A. Sugawara, K.-e.T., K.I., M. Ogawa, K.S., K.A., A.K., M.
Mouse Y-FISH 5. B. K. Sun, Z. Siprashvili, P. A. Khavari, Science 346, 941–945 Oshima, and R.M. performed the experiments; A. Sato, T.Y., A.T., H.E.,
using light microscopy.
To detect male-origin iPS cell–derived 3D (2014). and T.T. discussed the results; and T.T., K.-e.T., and M. Ogawa wrote
Whole (left) and high- 6. T.-T. Sun, H. Green, Cell 9, 511–521 (1976). the paper. Competing interests: This work was performed under
IOSs engrafted into female mouse skin, we
magnification (right) views 7. A. Atala, Lancet 385, 487–488 (2015). an invention agreement between the Tokyo University of Science,
performed mouse Y-FISH on 10-µm paraffin 8. P. T. Sharpe, C. S. Young, Sci. Am. 293, 34–41 (2005).
are shown. Scale bars, 2 mm RIKEN, and Organ Technologies Inc. T.T. is a director at Organ
sections using fluorescein isothiocyanate– 9. K. Nakao et al., Nat. Methods 4, 227–230 (2007). Technologies Inc. Data and materials availability: All data needed
(low magnification) and 20
conjugated murine Y-chromosome probes 10. E. Ikeda et al., Proc. Natl. Acad. Sci. U.S.A. 106, 13475– to evaluate the conclusions in the paper are present in the paper
µm (high magnification). 13480 (2009). and/or the Supplementary Materials. Additional data related to this
(Chromosome Science Labo Inc.). All fluo-
(B) Analysis of hair types, 11. K.-e. Toyoshima et al., Nat. Commun. 3, 784 (2012). paper may be requested from the authors.
rescence microscopy images were captured 12. M. Ogawa et al., Nat. Commun. 4, 2498 (2013).
such as zigzag (Z), awl/
on an LSM 780 confocal microscope (Carl 13. M. Hirayama et al., Nat. Commun. 4, 2497 (2013).
auchene (A/Au),and guard (G), of natural mouse pelage and iPS cell–derived hair shafts. The hair SUPPLEMENTARY MATERIALS
Zeiss). 14. S. M. Wu, K. Hochedlinger, Nat. Cell Biol. 13, 497–505
http://advances.sciencemag.org/cgi/content/full/2/4/
types were determined by microscopic observations. The data are presented as the means ± SEM (2011).
e1500887/DC1
from six individual experiments; n = 104 (natural skin) and n = 248 (iPS cell–derived transplants). In situ hybridization 15. D. E. Cohen, D. Melton, Nat. Rev. Genet. 12, 243–252 (2011).
Supplementary Text
(C) Distance between hair shafts in natural pelage and iPS cell–derived bioengineered hair shafts. 16. E. Walck-Shannon, J. Hardin, Nat. Rev. Mol. Cell Biol. 15,
Fig. Fig.
4. Analysis
4. Analysis of iPS of cell–derived
iPS cell–derivedhair hair
typestypes
and and
hair hair
cycle.cycle.
(A) Microscopic
(A) Microscopicobservation
observationof the
of iPS
thecell–derived
iPS cell–derived
bioengineered
bioengineered hair hair
shafts shafts
showing
showing In situ hybridizations were performed using Fig. S1. Analysis of the area of the cystic lumen.
34–48 (2014).
Theanalyzed
distances were calculated by microscopic observations of H&E-stained sections. The data are Fig. S2. Histochemical and immunohistochemical analyses of the
zigzag,
zigzag,
awl/auchene,
awl/auchene, and guard
and guard
hairs.hairs.
The hair
The shafts
hair shafts
werewere
analyzed usingusing
lightlight
microscopy.
microscopy.
Whole Whole
(left)(left)
and high-magnification
and high-magnification(right)
(right)
viewsviews
are shown.
are shown.Scale Scale 10-µm frozen sections as described previ- 17. F. A. Tavassoll, P. Devilee, Eds., Pathology and Genetics of
cystic epithelia in the in vivo explants of multiple iPS cell–derived
Guidelines
bars,bars,
2 mm 2 mm for magnification)
(low (low Use of Laboratory
magnification) andmm
and 20 20Animals
mm (high
(high presented
ofmagnification).
magnification). as the
(B) Analysis
(B) means
Analysis of ±
of hair SEM
types,
hair fromsuch
types,
such individual
as zigzag
as zigzagexperiments;
(Z), awl/auchene n=
(Z), awl/auchene 6 (natural
(A/Au),(A/Au), skin)guard
and guard
and and
(G), nof=natural
(G), 6of(iPS cell–
natural ously (11). Briefly, digoxigenin-labeled probes Tumours of the Breast and Female Genital Organs. World
EBs.
mouse RIKEN
mouse
pelage and
andUse
pelage iPS Committee
andcell–derived (AH26-02-3).
iPS cell–derived
hair shafts.
hair shafts.
The hair
The types derived
hair types
werewere transplants).
determined
determined (D)
by Macromorphological
by microscopic
microscopic
observations. observations
observations.
The data
The data at the
are presented
are anagen
presented
as theasphase
means of
the means the±hair
± SEM fromcycles
SEM from for specific transcripts were prepared by PCR
Health Organization Classification of Tumours (IARC Press,
Fig. S3. Culture of iPS cells for EB formation.
Lyon, France, 2003), 432 pp.
six individual
six individual
experiments;
experiments; n = 104
n = (natural
104 (natural
skin)skin)
and nand nin=iPS
= 248 248cell–derived
(iPS cell–derived bioengineered
(iPS cell–derived
transplants). (C)hair.
transplants). (C)Scale
Distance bars,between
Distance
between 1 mm.
hair(E) Assessment
shafts
hair shafts
in natural of
in naturalthepelage
pelagehair
andgrowth
iPS
andcell– (closed
iPS cell– with published primers. The mRNA expres- 18. M. Eiraku et al., Nature 472, 51–56 (2011).
Fig. S4. Gene expression in iPS cell–derived bioengineered hair
iPS cell–derived EB transplantation circles) and regression (open circles) follicle germs in a 3D IOS generated via the CDB transplantation
derived
derived
bioengineered
bioengineered hair shafts.
hair shafts.
The distances
The distances
werewere
calculated
calculated
by microscopic
by microscopicobservations
observations ofphases
of H&E-stained of
H&E-stained thesections.
bioengineered
sections. The data are hair,
The data including
presented
are presented
as thezigzag
as means and
the means ± awl/± sion patterns were visualized according to 19. H. Suga et al., Nature 480, 57–62 (2011).
and extraction method.
SEMSEMfromfromindividual
individualexperiments;
experiments;
n = 6n(natural
= 6 (natural andauchene
skin)skin) and
n = 6and
= 6n(iPS (iPSguard
cell–derivedhairtransplants).
cell–derivedtypes. The data
transplants). are
(D)presented
(D) Macromorphological as the means
Macromorphological ± SEM;atnthe
observations
observations = at
5 anagen
(zigzag)
the anagen andphase
phase n = 10 the immunoreactivity with anti-digoxigenin 20. T. Nakano et al., Cell Stem Cell 10, 771–785 (2012).
Fig. S5. Y-FISH analysis of the distribution of iPS cell–derived
of the EBs
of thecultured
hair cycles infor
hair cycles iPSin7cell–derived
days
iPS were selected,
cell–derived and 48
bioengineered
bioengineered (awl/auchene
hair. hair.
Scale Scale
bars,bars,
1 mm. 1and
(E)guard).
mm. Assessment
(E) Assessmentof theof hair
the hair
growthgrowth
(closed
(closed
circles)
circles)
and and
regression
regression
(open (open
circles)
circles) alkaline phosphatase–conjugated Fab frag-
21. K. R. Koehler, A. M. Mikosz, A. I. Molosh, D. Patel, E. Hashino,
cells among the 3D IOSs grafted to the natural murine skin.
Nature 500, 217–221 (2013).
phases(notof stimulated)
phases theof bioengineered and 32hair,
the bioengineered (Wnt10b
hair,
including stimulated)
including zigzagzigzag
and and
awl/auchene
awl/auchene
and and
guard guard
hair hair
types.types.
The data
The dataare presented
are presented as theas means
the means ± SEM; ± SEM;
n = 5n(zigzag)
= 5 (zigzag) ments (Roche) according to the manufactur- 22. C. L. Watson et al., Nat. Med. 20, 1310–1314 (2014).
Fig. S6. Distribution of hair species of iPS cell–derived hairs.
small pieces containing 10 to 20 hair follicles, (Carl Zeiss) or an Axio Scan.Z1 (Carl Zeiss). Table S1. Primer sequences used in real-time PCR.
nEBs
and and nwere
= 10 10 sterilely
=(awl/auchene
(awl/auchene placed
and and in guard).
guard). a 20-µl collagen er’s instructions. 23. T.-X. Jiang et al., Int. J. Dev. Biol. 48, 117–135 (2004).
gel drop (Nitta Gelatin Inc.). The collagen were intracutaneously transplanted into the For fluorescent immunohistochemistry, fro- 24. Y. Fukano et al., Wound Repair Regen. 14, 484–491 (2006).
25. H. Egusa et al., PLOS One 5, e12743 (2010). Submitted 6 July 2015; Accepted 29 February 2016
gel drop was transplanted into the subrenal back skin of 6-week-old Balb/c nu/nu mice, zen sections (10 and 100 µm) and paraffin Analyses of the distance between hair Published 1 April 2016
markers,
markers,
including
capsules including Nestin,
of 6-week-old Nestin,
Pax3, Pax3,
Snail,Snail,
and and
C.B-17/lcr-scid/scidJcl Twist.
Twist.
Therefore,
Therefore,
our findings
our findings
as previously development
described development of cosmetics
(47). Shallow of cosmetics
stab andsections
and
quasi-drugs,
quasi-drugs,
(10 µm)although
although
were this this
method
prepared method
andhas has
immu- follicles and the distribution of hair
26. K. Okita, M. Nakagawa, H. Hyenjong, T. Ichisaka, S.
10.1126/sciadv.1500887
Yamanaka, Science 322, 949–953 (2008).
suggest
suggest
thatThirty
mice. that
Wnt10b Wnt10b
days signaling
signaling
after theplays plays
important
important
transplantation roles
ofroles
in thein formation
woundsthe formation
were of
made ofcritical
on critical
thelimitations
backlimitations
skin related related
of nude to thetonostained
permeability
the permeability
as ofpreviously
cosmetics
of cosmetics and and
drugs
described drugs
and
(11,and
12). shaft species 27. J. T. Fujii, G. R. Martin, J. Embryol. Exp. Morphol. 74, 79–96
hair hair
follicles
follicles
iPS-EBs, inthethein
inskin
the
vivoskin
field, field,
neural
transplants neural
crest–derived
werecrest–derived
extracted, melanocyte
melanocyte
mice differentia-
differentia-
using the reliability
a 20-gauge the reliability
Ophthalmic of physiological
of physiological
V-Lance responses
responses
Before (45).(45).
By contrast,
immunostaining, By contrast,it is expected
frozen it issections
expectedof Skin samples from C57BL/6NCrSlc mice were (1983). Citation: R. Takagi, J. Ishimaru, A. Sugawara, K.-e. Toyoshima, K.
tion,tion,
and and
divided progression
progression
into of organogenesis
five equal ofparts,
organogenesis
and placed through
through
intoreciprocal
reciprocal
(Alconepithelial-
epithelial-
Japan). ThethatiPS that
bioengineered
bioengineered
cell–derived 3D3D IOS
3D IOS
IOSs can can
overcome
overcome
EGFP-labeled these these
issues
samples issues
(46).treated
were (46).
Here, Here,
we
withweor 28. S. Reddy et al., Mech. Dev. 107, 69–82 (2001). Ishida, M. Ogawa, K. Sakakibara, K. Asakawa, A. Kashiwakura, M.
dissected from three different regions, in- 29. J. Fu, W. Hsu, J. Invest. Dermatol. 133, 890–898 (2013). Oshima, R. Minamide, A. Sato, T. Yoshitake, A. Takeda, H. Egusa,
DMEM (WAKO)
mesenchymal
mesenchymal interactions containing
interactions during 10%
during fetal
morphogenesis. calf se-
morphogenesis. were held so that provide the provide
hairevidence
protruded
evidence from
for theforgenerationwithout
the generation of aSuperfix
of
3DaIOS3D IOS(KURABO)
from from iPS for
iPS cells cells
using30 minanat
using
an cluding the thorax, body center, and lumbar 30. Y. Ouji, F. Nakamura-Uchiyama, M. Yoshikawa, Biochem. T. Tsuji, Bioengineering a 3D integumentary organ system from iPS
rum
For For(GIBCO),
regenerative
regenerative penicillin
therapy
therapyin (100
patients U/ml;
in patients withSigma),
with
burns,burns, the
a cultured skin
a cultured surface.
epithelialThe
epithelial intransplantation
in vivo vivo
transplantation
transplantation sites
model.model.room
The The temperature.
explants
explants
werewere The
fully fully primary
functional
functional
andantibodies
and
in- in- region of the side trunk lining on murine Biophys. Res. Commun. 438, 493–499 (2013). cells using an in vivo transplantation model. Sci. Adv. 2, e1500887
streptomycin
sheetsheet
is useful
is usefulfor the(100 mg/ml;
for replacement
the Sigma),
replacement of skin and
of skin
physiological were
20 physiological
mM then covered
functions,
functions,
suchsuch with
cluded surgical
cluded
hair hairbandage
follicles andtape
follicles and
sebaceous used
sebaceous were
glands as follows:
glands
with with
proper CK5connections
proper (1:100, to
connections rabbit;
the Co-
to the linea axillaris media. The iPS cell–derived 31. S. E. Ross et al., Science 289, 950–953 (2000). (2016).
Hepes (Invitrogen) on ice. They were treated (Nichiban). To examine the engraftment of vance), CK10 (DE-K10) (1:100, rabbit; Abcam), 32. T. Aoi et al., Science 321, 699–702 (2008).
as the
asprotection
the protection of deeper
of deepertissues,tissues,
waterproofing,
waterproofing, and and
thermoregulation
thermoregulationsurrounding
surrounding tissues,
tissues,
suchsuch as the as epithelium,
the epithelium, dermis,
dermis,fat, arrector
fat, arrectorpili pili IOSs, which were divided into small pieces 33. R. J. Morris et al., Nat. Biotechnol. 22, 411–417 (2004).
with collagenase (100 U/ml; Worthington) for the transplants and to study the hair cycles Muc2 (H-300) (1:100, rabbit; Santa Cruz Bio- containing 10 to 20 hair follicles from in vivo
(5). However,
(5). However, it is well
it is well
recognized
recognized that thatcurrentcurrent
therapies
therapies
usingusing
cultured
culturedmuscles,
muscles, and and
nerve nerve
fibers.
fibers.
Although
Althoughthe origin
the origin
of nerve
of nervefibersfibers
is still
is still 34. F. M. Watt, Science 346, 937–940 (2014).
10 min at 37°C and washed with medium to of the bioengineered hairs, we observed all technology), Cdx2 (EPR2764Y) (1:250, rab-
epithelial
epithelial
tissues tissues
suffersuffer
fromfrom criticalcritical
issues, issues,
including
including
aesthetics
aesthetics
and and
the theunclear,
unclear,
nerve nerve
fibersfibers
would would
be innervated
be innervated from from
host host
tissues.
tissues.
Our Our
pre- pre-
separate the organs for in vivo transplants. transplanted sites using the SteREO Lumar. bit; Abcam), Pax6 (1:200, rabbit; Covance),
inability
inability
to excrete
to excretesweat sweat
and and
lipids
The iPS cell–derived 3D IOS was dissected us- lipids
from from
exocrine
exocrineorgans
organs
(44).(44).
In ad-
In ad-vious vious
studies
V12 and AxioCam fluorescent stereoscopic studiesof ectodermal
of ectodermal organ organ
transplantations
transplantations demonstrated
calponin (EP798Y) (1:250, rabbit; Abcam), demonstrated
dition,
dition,
ingartificial
artificial
a surgical skin skin
consisting
knife consisting
under aofmicroscope.
cultured
of cultured epidermis
epidermis
and and
dermis
microscopedermis
with-with-that
system (Carlthat
the innervation
the innervation
Zeiss). Observationsintointothe CK15
explants
the explantshad had
(1:2000, been been
indicated
chicken; indicated
Abcam),from from
host(1:100,
Sox9 host
out appendage
out appendage organs organs
and and pores pores
is oftenis often
usedusedfor testing
forwere
testing
during
made during
the the
every tissues
2 to 3 tissues
days.(10–13).
(10–13). mouse; Abcam), CD34 (MEC 14.7) (1:100, rat;
Engraftment of the iPS cell–derived 3D Abcam), Lrig1 (1:400, goat; R&D Systems),
IOS Immunohistochemistry Lgr5 (1:100, rabbit; Abcam), neurofilament
Takagi
Takagi
et al.etSci.
al.Adv.
Sci. Adv.
2016;2016;
2 : e1500887
2 : e15008871 April 1 April
20162016 7 of 710of 10
For hair follicle regeneration, the iPS cell– Paraffin sections (10 µm) were stained with H (TA51) (1:500, rat; Chemicon), villin (C-
derived hair follicles, which were divided into H&E and observed using an Axio Imager A1 19) (1:200, goat; Santa Cruz Biotechnology),

28 sciencemag.org SCIENCE SCIENCE sciencemag.org 29

 Application Note Application Note

The Fast module for ZEISS LSM 880 with Airyscan: Improving Resolution and SNR on a Confocal
The Airyscan detector design simultaneously improves SNR
Confocal superresolution imaging with four times the speed and resolution with a hexagonally packed 32 channel GaAsP
PMT array that is positioned in the pinhole plane (Fig. 1) and
and improved signal-to-noise ratio removing the traditional pinhole-PMT scheme. This concept
combined two preferable but opposed / conflictive settings
of the pinhole in the traditional LSM; an open pinhole would
Author: Joseph Huff, PhD
Carl Zeiss Microscopy, LLC, Thornwood, NY, USA allow a lot of light to pass and thus increasing signal to
noise, while a small pinhole opening will improve the resolu-
Dr. Annette Bergter tion of the system. Further, each of the 32 detector elements
Carl Zeiss Microscopy GmbH, Germany
behaves as a small independent pinhole where not only the
Date: April 2016 optical resolution is improved but also spatial distribution of
the light is recoded to additionally improve the spatial fre- Figure 1 Schematic of Airyscan detector for LSM 880
1. Mirror 2. Emission filters 3. Zoom optics 4. Airy disk 5. Airyscan detector
quency contrast of the LSM system beyond what a tradition-
First introduced in August 2014, the Airyscan detector from ZEISS represents a new detector concept for laser- al system provides. To use Airyscan for super resolution im-
scanning microscopy (LSM) that enables a simultaneous resolution and signal-to-noise (SNR) increase over aging, the adjustable zoom optics can be arranged to project
traditional LSM imaging. The Airyscan detector design substitutes the conventional LSM detector and pinhole 1.25 Airy units onto the detector where each detector ele-
scheme for an array of 32 sensitive GaAsP detector elements, arranged in a compound eye fashion that resides ment represents 0.2 Airy units yielding a 1.4 x optical resolu-
in the pinhole-plane while still generating an optical section. The new detection geometry allows for the tion improvement and collect up to 50 % more light than an
collection of the spatial distribution of light originating from every point of a microscopic fluorescent object at traditional LSM system with a 1 AU pinhole. Once combined
the pinhole allowing access to higher frequency information and while additionally collecting more light for with a linear deconvolution, a 1.7 x resolution improvement
ultra-efficient imaging. Based on the Airyscan detection concept, the next innovation from ZEISS has been in all three dimensions is achieved and the additional collec-
developed with the introduction of the Fast mode for Airyscan. The Fast mode concept utilizes the Airyscan tion efficiency of the detector provides a 4x – 8x SNR improve-
detector technology in combination with an illumination shaping approach to enhance acquisition speeds by ment affording researchers the ability to better quantify and
four times while simultaneously increasing SNR and resolution overcoming the traditional compromises of LSM study biological processes like microtubule buckling as a mea-
imaging. sure of measure of mechanical stress on a cell. (Fig. 2, [3]).

Introduction
The Confocal Laser Scanning Microscope (LSM) has become
one of the most popular instruments in basic biomedical
research for fluorescence imaging. The main reason LSM has
become so popular is that the technique affords researchers
25 years, all major developments of commercially available
systems focused on novel methods to increase image contrast,
instrument versatility and acquisition speed. These innova-
tions have provided the LSM with the flexibility to address
Improving Speed on a Point Scanning Confocal
Traditionally in order to improve acquisition speeds on a
point scanning confocal system, the speed at which the laser
spot rasters the field of view is increased resulting in a
decrease the amount of time the laser spot spends on one
pixel. For the maximum speed increase, resonance scanning
Figure 2 Application example of standard confocal compared to Airyscan
detection using cardiomyocyte Cells with SiR-tubulin to measure microtubule
buckling. (Left Panel) Traditional confocal still images from a timeseries
acquired at 30 fps with low SNR. (Right Panel) Airyscan still images from a
time series acquired at 30 frames/sec with high SNR and resolution.
Each series of stills represents the resting tubules, contracted tubules, and
resting tubules respectively. Images courtesy of Ben Prosser, University of
Pennsylvania [3]

images with high contrast and a versatile optical sectioning
capability to investigate three dimensional biological struc-
tures [1]. The optical sectioning ability of an LSM is a product
of scanning a diffraction limited spot, produced by a focused
laser spot, across a sample to create an image one point at a
time. Traditionally, the generated fluorescence for each point
is collected and focused back through an aperture (pinhole)
onto a unitary detector (typically a PMT). The traditional
detection scheme allows the pinhole to reject the out of
focus light and the PMT to turn the remaining light from the
focal plane into a digital signal to form an optical sectioned
image [2]. As confocal systems developed over the past
research applications in many areas, providing data of cell
differentiation, cell tracking, live cell kinetics, protein expres-
sion, protein localization, neural network connectivity, tissue
structure, gene expression and protein / gene function (list is
not meant to be inclusive). However, until the introduction
of the Airyscan detection concept, the LSM system flexibility
has always forced a researcher into making a compromise on
a performance metric of a system in order to increase another
(i.e. speed for resolution, resolution for SNR etc) to meet
application requirements. The introduction of both the
Airyscan detection concept in 2014 and now the Airyscan
Fast mode represents a fundamental change where tradi-
approaches are utilized in LSM systems allowing frame rates
of around 30 Hz for a full field of view. However, to achieve
the higher acquisition rates a researcher will have to sacrifice
the resulting SNR of an image. The reason for the image SNR
decrease is that as the acquisition speed is increased, the
laser spot will spend less time per pixel and thus fewer
photons will be available resulting in a deterioration of Figure 3 Application example of a resonance scanning time series
compared to a time series from the Airyscan Fast acquisition mode using
image SNR (Fig. 3, upper panel). At some point the SNR can
cardiomyocyte Cells with tubulin-EMTB to measure microtubule buckling at
get so low, that it becomes impossible to discriminate high frame rates. (Top panel) Still images from a resonance scanning time
individual features of the sample and structural information series acquired at 80 frames/sec. (Bottom panel) Still images from an Airyscan
Fast mode time series acquired at 96 frames/sec. Each series of stills
is lost in the process to reach the necessary temporal represents the resting tubules, contracted tubules, and resting tubules
resolution, rendering the method non-usable for the aspired respectively. Images courtesy of Ben Prosser, University of Pennsylvania [3]

tional imaging compromises no longer exist and all metrics scientific goal.
can be improved simultaneously.

30 31
2 3
 Application Note
Application Note

To overcome the trade-off between acquisition speed and

Standard Airyscan Airyscan in Fast Mode
To overcome
image theAiryscan
SNR, the trade-offdetector
between acquisition
concept speedcombined
has been and
Standard Airyscan Airyscan in Fast Mode
image
with anSNR, the Airyscan
illumination detector
shaping concept
approach for ahas
newbeen
fastcombined
acquisi- Illumination
with an illumination shaping approach for a new fast acquisi-
tion mode. The innovation of the new acquisition mode utilizes Shape
Illumination
Shape
tion mode. The
the concept innovationand
of excitation of the new acquisition
detection modeto
parallelization utilizes
in-
the
crease frame rates to avoid sacrificing SNR. Subsequently, in-
concept of excitation and detection parallelization to in
crease
the newframe
Fast rates
modetoapproach
avoid sacrificing SNR. Subsequently,
the excitation beam is shapedin to
the
illuminate four lines in parallel and the Airyscan area detectorto
new Fast mode approach the excitation beam is shaped
illuminate four
collects data linesallinfour
from parallel
linesand the Airyscan(Fig.
simultaneously area4).
detector
By em-
collects adata from all four lines simultaneously (Fig. 4).are
By in-
em-
ploying
ploying
parallelization approach, acquisition speeds
To get ahead in your research you may want to image the smallest structures,
creased abyparallelization
a factor of 4 approach, acquisition
while keeping speeds
high pixel dwellare in-
times
creased
resultingby
in aa factor of 4 while
high image SNR. keeping high pixel
Hence, while dwell imag-
increasing times catch the faintest signal or track the fastest processes – or do all of that
resulting y y
ing speedinnotably,
a high image SNR. to
Hence,
noisewhile
ratiosincreasing imag-
at once.
high signal can be preserved
y y
ing speedinnotably,
resulting high data.
meaningful signalMoreover,
to noise ratios
sincecan
the be preserved
Fast mode
resulting
approachinis meaningful data.
based on the Moreover,
Airyscan sincethe
detector, theadditional
Fast mode x x
x x
approach is based
simultaneous on the
increase Airyscan
in SNR detector, the
and resolution additional
remain. Therefore
sAiryscan
imultaneous increase
in Fast in SNRaand
mode offers resolution
researcher theremain. Therefore
unprecedented
Detection
Shape
Detection
When it comes to getting accurate data from live cells
Airyscan in Fast
of amode
1.5 x offers a researcher
increase the
andunprecedented or other weakly-labeled samples, there is no such
Shape
combination resolution 4 x SNR increase
combination
at four times of
thea acquisition
1.5 x resolution
speedincrease
(Fig. 3, and 4 xpanel).
lower SNR increase
at four times the acquisition speed (Fig. 3, lower panel). thing as too much sensitivity, resolution or speed.
Summary Figure 4 (Left Panel) The illumination and detection scheme of the standard
Each photon of emission light is precious.
Summary
In summary, the Airyscan detection module dispenses with Airyscan4mode
Figure (Left where
Panel) aThediffraction limited
illumination andspot is projected
detection scheme into
of the sample
standard
and the associated
Airyscan mode where Airy disc from the
a diffraction collected
limited fluorescence
spot is is projected
projected into onto
the sample
In
thesummary, the Airyscan
classical physical detection
pinhole module
and unitary dispenses
detector with
assembly all 32the
and elements of the
associated AiryAiryscan
disc fromdetector. (Right Panel)
the collected The illumination
fluorescence is projectedspot is
onto
the classical physical pinhole and unitary detector assembly
With ZEISS Airyscan you use a combination
and utilizes a new pinhole-plane image detection assembly elongated in theofy the
all 32 elements dimension
Airyscantodetector.
cover four linesPanel)
(Right simultaneously and 16spot is
The illumination
channels ofinthe
elongated theAiryscan
y dimensiondetector are utilized
to cover collect
four lines the emission
simultaneously andfrom
16 all
and utilizes
based a new32
on a novel pinhole-plane image detection
channel GaAsP-PMT assembly
area detector. four lines of
channels simultaneously
the Airyscan detector are utilized collect the emission from all
based on a novel 32 channel GaAsP-PMT area detector.
In 2014 Airyscan was implemented to provide a flexible and four lines simultaneously of fast superresolution and sensitive confocal
In 2014method
Airyscan
towas implemented to provide a flexible
and and
robust substantially
robust method compared
to substantially
increase resolution
increase confocal
resolutionLSM.
and Using image acquisition.
signal-to-noise to traditional
signal-to-noise compared
the Airyscan detector to traditional
design confocal LSM.
for new innovation, Using
Fast mode
the Airyscan
for Airyscan provides
detector adesign for new
substantial innovation,
increase Fastwhile
in speed mode
for Airyscan provides
simultaneously a substantial
providing increases increase in speed
in resolution and while
signal-
Cover Image Courtesy of: Dr. Pawel Pasierbek, BIOOPTICS, IMBA – Institute This collection of publications gives you
of Molecular
Cover ImageBiotechnology, IMPPawel
Courtesy of: Dr. - Institute of Molecular
Pasierbek, Pathology,
BIOOPTICS, IMBA – Institute
simultaneously providing increases in resolution
to-noise compared to traditional confocal LSM. and signal-
in-depth information about the technology

IMAGE: ARABIDOPSIS ROOT. COURTESY OF T. PASTERNAK, INSTITUTE OF BIOLOGY,

Vienna
of Biocenter
Molecular Biotechnology, IMP - Institute of Molecular Pathology,
to-noise compared to traditional confocal LSM. Vienna Biocenter
behind Airyscanning.
References:
References:
[1] Conchello, J.-A. and J. W. Lichtman, Optical sectioning microscopy. Nature methods, 2005. 2(12): p. 920 – 931.
[1]
[2] Conchello, J.-A.
J.R.and J. W. Lichtman, Optical sectioning microscopy. Nature methods, 2005. 2(12): at
p. different
920 – 931.
Neu, T.R. and
[2] Trends
Neu, T.R. and J.R.
Lawrence,
Lawrence,
Innovative
Innovative
techniques,
techniques,
sensors,
sensors,
and approaches
and approaches
for imaging
for imaging
biofilms
biofilms at different
scales.
scales.
Go to: http://bit.ly/1SSGfc3
in Microbiology, 2015.

ALBERT LUDWIGS UNIVERSITY FREIBURG, GERMANY

[3] Trends
RobisoninetMicrobiology, 2015. microtubules buckle and bear load in contracting cardiomyocytes”, Science April 2016
Al, “Detyrosinated
[3] Robison et Al, “Detyrosinated microtubules
[5] Sheppard, C.J., Super-resolution in confocal buckle andOptik,
imaging. bear load
1988.in80(2):
contracting
p. 53 –cardiomyocytes”,
54. Science April 2016
[5]
[6] Sheppard,
Sheppard, C.J.,
C.J., Super-resolution in Heintzmann,
S.B. Mehta, and R. confocal imaging. Optik, 1988.
Superresolution by 80(2):
imagep.scanning
53 – 54.microscopy using pixel reassignment.
[6] Opt
Sheppard, C.J., S.B.
Lett, 2013. Mehta,
38(15): and R.
p. 2889 Heintzmann, Superresolution by image scanning microscopy using pixel reassignment.
– 2892.
[7] Opt
Huff,Lett, 2013.
J., The 38(15):
Airyscan p. 2889from
detector – 2892.
ZEISS: confocal imaging with improved signal-to-noise ratio and super-resolution.
[7] Huff,
NatureJ.,methods,
The Airyscan detector
2015. 12. from ZEISS: confocal imaging with improved signal-to-noise ratio and super-resolution.
[8] Nature methods,
Weisshart, K., The2015.
basic 12.
principle of Airyscanning. 2014. ZEISS Technology Note
[8]
[9] Huff, J.; Bathe, W.; Netz,principle
Weisshart, K., The basic of T.;
R.; Anhut, Airyscanning. 2014.
Weisshart, K., TheZEISS Technology
Airyscan detectorNote
from ZEISS.
[9] Huff, J.; Bathe,
Confocal W.;with
imaging Netz,improved
R.; Anhut, T.; Weisshart,ratio
signal-to-noise K., The
andAiryscan detector 2015.
superresolution. from ZEISS.
ZEISS Technology Note
Confocal imaging with improved signal-to-noise ratio and superresolution. 2015. ZEISS Technology Note
32 4
4
 Revolutionizing your
confocal imaging.
ZEISS LSM 880 with Airyscan

Researchers and microscope makers alike are constantly striving to push

the boundaries of sensitivity, resolution, and speed. Their determination has
produced a steady stream of advances in microscope technology,
and researchers armed with these ever-improving tools and
techniques are squeezing ever more and better data
out of their samples. This poster leads the reader
through the past, present, and future of confocal
microscopy. Additionally, you can go online
to find an interactive version of the poster
with additional information and graphics.

// INNOVATION
Go to poster.sciencemag.org/confocal MADE BY ZEISS

This poster is brought to you

by the Science/AAAS Custom
Publishing Office.
Your new standard for fast and gentle confocal imaging
Discover ZEISS LSM 880 with Airyscan – the new confocal laser scanning microscope
that offers high sensitivity, improved resolution in x, y and z, and high speed.
All in one system. Find out more and book a hands-on demonstration in one of
our ZEISS Microscopy Labs now.

www.zeiss.com/lsm880
 Finding answers in
demanding research.
ZEISS LSM 800

// INNOVATION
MADE BY ZEISS

Your compact confocal for high-end imaging

With ZEISS LSM 800 you are choosing a flexible and compact confocal laser scanning
microscope, complete with highly sensitive GaAsP detector technology and fast linear
scanning. Add Airyscan, the revolutionary detection concept from ZEISS, and you will
gain 1.7× higher resolution in all three dimensions – resulting in a 5× smaller confocal
volume. And you will be pushing sensitivity beyond the limits of all conventional
confocals.

www.zeiss.com/lsm800