ASH 50th anniversary review

Stem cell concepts renew cancer research

John E. Dick1
1Division of Cell and Molecular Biology, University Health Network, and Department of Molecular Genetics, University of Toronto, Toronto, ON

Although uncontrolled proliferation is a
distinguishing property of a tumor as a
whole, the individual cells that make up
the tumor exhibit considerable variation
in many properties, including morphol-
ogy, proliferation kinetics, and the ability
to initiate tumor growth in transplant as-
says. Understanding the molecular and
cellular basis of this heterogeneity has
important implications in the design of
therapeutic strategies. The mechanistic
basis of tumor heterogeneity has been
uncertain; however, there is now strong
evidence that cancer is a cellular hierar-
chy with cancer stem cells at the apex.
This review provides a historical over-
view of the influence of hematology on
the development of stem cell concepts
and their linkage to cancer. (Blood. 2008;
112:4793-4807)

Introduction
Tumors from different patients, whether leukemic or solid, exhibit
significant heterogeneity in terms of morphology, cell surface
markers, genetic lesions, cell proliferation kinetics, and response to
therapy. Heterogeneity of all these features is also seen within an
individual tumor that is clonal (ie, initiated from a single cell).
Although individual cells of a tumor all share common genetic
aberrations reflective of their clonal origin, single-cell analysis has
shown the existence of variation in genetic and epigenetic abnor-
malities between different cells or locations within a tumor. The
cellular and molecular basis for this heterogeneity represents a
fundamental problem that has interested cancer researchers for
many decades. One possible explanation is that all tumor cells
could be biologically equivalent and heterogeneity derives from
extrinsic or intrinsic influences that result in random or stochastic
responses. Alternatively, it is possible that the tumor is a caricature
of normal tissue development and retains hierarchical organization
with (cancer) stem cells at the apex.1 This review will focus on
50 years of hematology research that played a role in garnering
evidence to support the hierarchy model.
There is mounting evidence from cell purification studies that a
subset of cells, termed cancer stem cells (CSCs), or cancer-
initiating cells, that are distinct from the bulk of the tumor are
responsible for long-term maintenance of tumor growth in several
cancers.2 Evidence is strongest for the acute leukemias, although
recent studies have identified the existence of CSCs in an increas-
ingly longer list of solid tumors, including brain, breast, and colon.3
This conceptual shift has important implications, not only for
researchers seeking to understand mechanisms of tumor initiation
and progression but also for developing and evaluating effective
anticancer therapies. Although the clinical relevance of CSCs
beyond experimental models is still lacking, the high frequency of
relapse after conventional cytotoxic chemotherapies predicts that
CSCs survive standard treatments. Many papers, reviews, and
funding bodies are making strong predictions that targeting CSCs
more effectively will lead to improved patient outcomes. Indeed,
there has been almost breathless excitement that this “new” field of
CSC biology will solve the stubbornly high relapse rates that
continue to exist for many cancers. However, stem cell concepts
and how they apply to cancer are decades old, and many obstacles
to the development of effective anticancer therapies still exist. It is
important to understand the historical underpinnings of the CSC
field to recognize the challenges we still face. By understanding
where we came from, we will have a clearer view of the way
forward. The history of stem cell research, and cancer stem cell
research in particular, is long and better suited for a book rather
than a short review. Thus, I only touch on a few highlights that I
find significant and that played a role in guiding our own work in
this field; most of my comments revolve around studies in
hematopoiesis and acute myeloid leukemia (AML) specifically. My
apologies go to those whose work I was not able to include or could
only be treated in a superficial fashion.
Tumor heterogeneity
The notion that cancer is a disease of abnormal or uncontrolled
proliferation has governed cancer research for more than 7 decades.
Because normal tissues also undergo continuous regeneration,
albeit at rates specific to each organ, the historical thrust of cancer
research was to develop drugs that targeted the enhanced rate of
neoplastic proliferation. Clinicians used doses and schedules that
achieved many log orders of cell kill while maintaining acceptable
levels of toxicity to normal tissues. Based on a deep understanding
of the molecular basis of cancer, current therapeutic strategies now
focus on inhibiting the molecular drivers of cancer (eg, trastuzumab-
Her2/Neu, imatinib-bcr/abl, erlotonib, epidermal growth factor
receptor). With an increased appreciation of the extrinsic microen-
vironmental factors that support tumor growth, such as the
vasculature, approaches that interfere with these factors (ie,
bevacizumab, vascular endothelial growth factor) are also being
brought forward. What all these strategies, whether from the past or
present, have in common is the tacit perspective of cancer as a
homogeneous, abnormal entity. Therefore, drugs targeting molecu-
lar lesions should be equally effective against all tumor cells,
barring the emergence of resistant subclones. The goal is to ensure
that enough drug(s) is given to target the primary molecular
pathway and then combined with drugs that target other relevant or

Submitted August 11, 2008; accepted September 9, 2008. DOI 10.1182/blood- © 2008 by The American Society of Hematology
2008-08-077941.

BLOOD, 15 DECEMBER 2008 䡠 VOLUME 112, NUMBER 13 4793

4794 DICK
compensatory pathways to enable killing of every tumor cell.
Indeed, historically, even research to understand the basic biology
of cancer uses strategies that treat an individual cancer as function-
ally homogeneous. Much cancer research, even to the present day,
follows a paradigm where the entire tumor or an entire dish of
tissue culture cells is used for genetic, proteomic, or biochemical
studies, mostly because the technology used to interrogate biologic
processes (ie, proteomics, biochemistry) cannot be done on single
cells and requires large cell numbers. Clearly, this has been a
productive approach as evidenced by the thorough understanding
gained on the genetic and epigenetic basis of the neoplastic
process. However, an inability to deal meaningfully with limiting
cell numbers has forced the cancer research enterprise to operate
within a dichotomy because it is also well known that tumors are
not homogeneous but exhibit significant heterogeneity.
Since the time of the great experimental pathologists (Virchow,
Maximov, and others), it has been recognized that tumors, be they
liquid or solid, exhibit marked morphologic heterogeneity. With the
advent of transplantation of murine cancer cell lines into mice,
Furth and Kahn in 1937 were able to establish that a single cell, and
not a transmissible agent, could initiate a cancer graft.4 As others
applied transplantation assays to a variety of solid tumors and
leukemias throughout the next several decades, it became clear that
there was wide variation in the frequency of cells that initiate
cancer growth.5-7 Remarkably, human studies have been carried out
since the late 1800s (reviewed by Southam et al8) where tumors
were excised, single-cell suspensions made and autotransplanted
into the thigh at different cell doses. These studies showed that
tumor reinitiation was variable and rare, often requiring more than
106 cells. Finally, in an elegant series of experiments, Pierce et al
showed that malignant teratocarcinoma cells can spontaneously
differentiate into mature benign cells.9 Thus, a tumor can be
considered a hierarchy defined by a maturation process in the same
way that normal tissue development occurs. In 1961, Pierce and
Speers proposed the idea that tumors are caricatures of normal
development and that this maturation process could be harnessed as
a means of therapy.10
The concept of cellular heterogeneity reached its zenith with the
functional cell proliferation studies of the 1940s, 1950s, and 1960s.
The advent of radiolabeling cells, combined with autoradiography
by Belanger and Leblond in 1946, enabled precise measurements
of the proliferation, lifespan, and hierarchical relationships in
normal and neoplastic tissues.11 Clermont and Leblond showed, for
example, that the testis contained proliferating cells that took up a
pulse of label and then by tracking the label showed their
differentiation to mature cells.12; they also noted that these cells
maintained themselves. In the 1960s and 1970s, several prominent
hematologists, including Clarkson et al,13,14 Killmann et al,15
Gavosto,16 and others, carried out pioneering studies that were
(perhaps more than any other findings) the studies establishing that
tumors exhibited functional heterogeneity. These were cytokinetic
studies carried out first in cell lines and murine models of the acute
leukemias and then, remarkably by in vivo examination of
leukemia blast proliferation kinetics in human AML and acute
lymphoblastic leukemia (ALL) patients. The in vivo data could not
have been clearer; the majority of leukemic blasts were postmitotic
and needed to be continuously replenished from a relatively small
proliferative fraction17 (Figure 1). Only a small fraction (⬃ 5%) of
leukemic blasts was rapidly cycling in vivo. However, careful
analysis demonstrated that there were 2 proliferative fractions in
patients: a larger, fast cycling subset with a 24-hour cell cycle time
and a smaller, slow cycling fraction with a dormancy estimated to
BLOOD, 15 DECEMBER 2008 䡠 VOLUME 112, NUMBER 13
last from weeks to months. In particular, Clarkson, Rubinow,
Skipper and their colleagues inferred that the slow cycling fraction
was actually generating the fast cycling fraction and suggested
these represented a leukemic stem cell (LSC) population because
they had similar cytokinetic properties to those observed for
normal hematopoietic stem cells (HSCs).18 They noted that the
dormant cells were much smaller in size and had less granularity
than the rapidly proliferating fraction, which led to speculation that
these cells could represent distinct cellular fractions. Of course,
such studies could not prove whether the cycling fractions repre-
sented the same population, existing in either a proliferative or a
nonproliferative state, or whether these observations point to a
hierarchical relationship representing 2 functionally distinct cell
types. Even so, based on these observations, reviews of this era
predicted that the inability to eradicate the theoretical LSCs
represented the cause of relapse and the ultimate failure for
antiproliferative chemotherapies, which formed the main chemo-
therapeutic armament at the time.14,19 Investigators then combined
in vivo cytokinetic studies with drug treatment and tailored new
therapies based on these findings. They made the interesting
observation that LSCs respond to depletion of the leukemia cell
mass that occurs when antiproliferative drugs are administered to
AML patients. LSCs begin to go into cycle and expand in the same
way normal stem cells go into cycle after chemotherapy-induced
cytopenias. Thus, they suggested that one way to eliminate dormant
LSCs would be to find the window when they cycle to kill them
when they are most vulnerable. Indeed, the cytokinetic findings
represented the underpinning for pioneering clinical trials carried
out by Clarkson and his colleagues at Memorial Sloan-Kettering
Cancer Center that used multiple drugs.20 The fundamental prob-
lem investigators faced at this time was the inability to directly
identify and assay LSCs, making it impossible to characterize them
and harness the acquired knowledge for therapeutic approaches in the
future. Pointed comments made at least as early as 1970 predicted that
antiproliferative chemotherapy agents would not effectively eradicate
these dormant LSCs and that future progress must focus on understand-
ing their properties.17 These remarkably prescient observations retain
their potency and freshness even today, nearly 40 years later.
The cytokinetic studies drove interest away from the study of
bulk leukemias because they were mostly composed of postmitotic
blasts. Because there was little that could be done to assay the
dormant LSC, attention focused on the development of a clono-
genic assay for the rapidly proliferating leukemic cells (AML-
colony-forming unit [CFU]) that the in vivo studies had uncovered.
McCulloch et al,21,22 Moore and Metcalf et al23,24, Griffin and
Löwenberg,25 and Dicke et al26 pioneered these assays and made
detailed insights into their properties. Parallel clonal assays were
developed for solid tumor clonogenic progenitors.27 In a key series
of experiments, Sabbath et al developed a series of monoclonal
antibodies (mAb) directed against myeloid differentiation antigens
and assessed the phenotype of the AML-CFU itself by complement
mediated lysis followed by clonogenic assay.28 Parallel studies
were done with normal cells. The AML-CFU from different
patients could be classified into 3 groups: primitive (similar to
multipotent normal CFU), early lineage committed (similar to
normal day 14 CFU-GM), and late (similar to day 7 CFU-GM).
This work established that the AML-CFU were distinct from most
leukemic blasts and the leukemic cells within an individual colony
could be more differentiated than the individual AML-CFU that
originated the colony. These clonal data provided the first strong
proof for a hierarchy in AML. Moreover, this work suggested that
the different classes of AML-CFU might originate from normal
BLOOD, 15 DECEMBER 2008 䡠 VOLUME 112, NUMBER 13
Figure 1. Labeling pattern of leukemic cells in marrow of patient 1.
Patient 1, a patient with acute myelomonocytic leukemia, received a
continuous 10-day infusion of tritiated thymidine. Leukemic cells were
arbitrarily divided into types I, II, and III based on increasing levels of
morphologic maturity (type I indicates primitive blast forms; type III, most
differentiated cells). At the end of the 10-day infusion, most type II and type
III cells were labeled in both marrow (shown here) and blood (not shown),
but only 40% of type I cells were labeled, reflecting their slow proliferative
rate. Many of the type I cells remained highly labeled for over 3 weeks after
the infusion. Reprinted from Clarkson17 by permission.
cells at different stages of differentiation.25 McCulloch et al showed
additional evidence for heterogeneity with the finding that the
AML-CFU were heterogeneous in their serial replating ability,
some unable to generate secondary colonies whereas others had
high capacity.22,29,30 Because this is an indicator of self-renewal,
they predicted that some AML-CFU were more primitive than
others. Although the properties of AML-CFU (number, prolifera-
tion, drug sensitivity) on their own did not predict clinical disease
properties or response to therapy, the secondary replating capacity
had better predictive power. The AML-CFU assay was used
intensively from the late 1960s to the early 1980s and resulted in many
important insights into AML biology. However, the limitations of the
assay and the lessons that had been learned from the study of normal
hematopoiesis underscored the need for functional in vivo assays to
determine whether an LSC, earlier than AML-CFU, existed in AML.
Models of tumor heterogeneity
Collectively, the data described in “Tumor heterogeneity” firmly
established that tumors are not only morphologically heteroge-
neous as observed more than 100 years ago, but they are
functionally heterogeneous in terms of cell proliferation or even
tumor initiation based on transplant assays. Two mutually exclu-
CANCER STEM CELLS 4795
sive models could explain tumor heterogeneity. The stochastic
model predicts that a tumor is biologically homogeneous and the
behavior of the cancer cells is influenced by intrinsic (eg, levels of
transcription factors, signaling pathways) or extrinsic (eg, host
factors, microenvironment, immune response) factors. These influ-
ences are unpredictable or random and result in heterogeneity in the
expression of cell surface markers or other markers of maturation,
in entry to cell cycle, or in tumor initiation capacity. A key tenet of
this model is that all cells of the tumor are equally sensitive to such
stochastic influences and also that tumor cells can revert from one
state to another because these influences do not induce permanent
changes.1 In contrast, the hierarchy model predicts that the tumor is
a caricature of normal tissue development where stem cells
maintain normal tissue hierarchies (skin, colon, blood). In this
model, the “leukemic stem cells” are biologically distinct, they
maintain themselves and the clone by self-renewal, and they
mature to generate progeny that lack stem cell properties (Figure
2). Thus, both the stochastic and hierarchy models accommodate
the existence of “functional leukemic stem cells,” cells that behave
as leukemia stem cells. The essential difference is that, in the
stochastic model, LSCs arise randomly and every tumor cell has
the potential to behave like an LSC given the right influences;
whereas in the hierarchy model, there is only a distinct subset of
cells that has the potential to behave like LSCs. The only means to
determine which model is correct is prospective isolation of cell
4796 DICK BLOOD, 15 DECEMBER 2008 䡠 VOLUME 112, NUMBER 13

Figure 2. Models of tumor heterogeneity. Tumors are composed of phenotypically and functionally heterogeneous cells. There are 2 theories as to how this heterogeneity
arises. According to the stochastic model, tumor cells are biologically equivalent, but their behavior is influenced by intrinsic and extrinsic factors and is therefore both variable
and unpredictable. Thus, tumor-initiating activity cannot be enriched by sorting cells based on intrinsic characteristics. In contrast, the hierarchy model postulates the existence
of biologically distinct classes of cells with differing functional abilities and behavior. Only a subset of cells can initiate tumor growth; these cancer stem cells possess
self-renewal and give rise to nontumorigenic progeny that make up the bulk of the tumor. This model predicts that tumor-initiating cells can be identified and purified from the
bulk nontumorigenic population based on intrinsic characteristics.

fractions followed by clonal analysis with a functional LSC assay.
If the hierarchy model is correct, then it should be possible to sort
tumors into fractions that have LSC activity and those that do not.
In contrast, the stochastic model predicts that LSC activity may be
observed in any fraction and cannot be prospectively isolated
because all tumor cells have the same potential (even if it is low)
for tumor initiation.
Conditions for discovery
The term “leukemia stem cell” had been in use for many decades,
and a large part of the hematology community assumed that the
hierarchy model was true and the leukemia clone existed in a
hierarchy sustained by an LSC, with obvious parallels to the role
that HSCs play in sustaining the normal hematopoietic hierarchy.
However, formal proof that LSCs had stem cell attributes and were
distinct from the rest of the leukemic cells that could never behave
as LSCs was still lacking. Clearly, technologies that were not
available in the 1970s were required to move the field forward.
However, I do not think the absence of technology can fully explain
why lags existed from the mid-1970s when all the theoretical
principles pointing to the existence of LSCs and a description of
their properties were well worked out, to the mid-1990s when
direct proof for the existence of LSC was finally discovered, and
then to the most recent years when CSC concepts have been
extended to solid tumors. The explanation also rests in the time it
took for stem cell concepts to permeate the broader cancer research
community and to become more accepted. The research environ-
ment was primed with the groundbreaking studies of normal HSCs,
eventually leading to HSC transplant as therapy and to the
crossover of universal stem cell concepts into many areas of
biology. This created a backdrop in which technologies were
developing, including the use of flow cell sorters in combination
with mAb that bind cell surface markers and the development of
human xenograft assays for human HSCs and LSCs. Because these
conditions set the stage for discovery of LSCs and eventually CSCs
from solid tumors, it is instructive to review the key highlights that
culminated to identify HSCs and the universal concepts these stem
cell studies generated.
BLOOD, 15 DECEMBER 2008 䡠 VOLUME 112, NUMBER 13
Identification of normal HSCs in the mouse
The modern era of stem cell research can be traced back to the
seismic shift in the 1960s when Till and McCulloch moved HSC
research from a descriptive science to a quantitative science.
Although careful studies using microscopic analysis pointed to the
idea of a common stem cell for all blood lineages,31 equally robust
was the view that independent stem cells existed for each major
blood lineage, which was promulgated by textbooks from the
1940s.32 The experimental finding by Till and McCulloch that the
“regeneration nodules” on the spleen of lethally irradiated mice
(termed CFU-spleen [CFU-S])33 transplanted with BM cells were
clonal (based on X-ray induced chromosomal lineage tracking34)
and contained multiple blood lineages proved that the hematopoi-
etic system derives from a multipotent HSC rather than from a
series of lineage-specific stem cells. With a clonal, quantitative, and
functional assay in hand, in short order they established all of the
major stem cell principles under which we continue to operate
today: HSCs have a high capacity for self-renewal, proliferation,
and multipotential differentiation. Their observations that the
self-renewal capacity of individual CFU-S was highly variable35
led them to propose the concept that the control of HSC self-
renewal was governed by stochastic principles,36 a hypothesis that
is still being tested and still able to provoke strong opinions among
stem cell biologists.37-41 Their focus on the need for quantitative
functional assays prompted the development of in vitro clonal
progenitor assays by Pluznik and Sachs42 and Bradley and Met-
calf43 that were central to the later discovery of hematopoietic
colony stimulating factors (as reviewed in the 50th Anniversary
Review by Metcalf44). With the use of progenitor assays, it was
possible to describe precisely the lineage relationships that occur
during maturation. Despite widespread use of the spleen colony
assay as a surrogate HSC assay, by the mid 1980s it was clear that a
more primitive HSC must exist. First, CFU-S were heterogeneous
in terms of their spleen repopulation kinetics.45,46 Second, careful
lineage examination showed that the spleen colony was devoid of
lymphoid cells and therefore not multipotential.47,48 Third, after
5-fluorouracil treatment, CFU-S were killed whereas other dormant
cells capable of CFU-S replenishment survived, suggesting the
existence of a pre-CFU-S stem cell.49,50 Harrison, in particular,
drove the paradigm that cells with long-term repopulation capacity
could be distinguished from those with short-term potential.51
Using chromosomal translocation as a means of cellular marking
(first developed by Ford et al52), in 1968 the first clonal evidence
established multilineage repopulation, including CFU-S and
B cells.53 Using the same approach, Abramson et al showed that
repopulation of myeloid and lymphoid lineages could derive from a
long-term multipotential stem cell or from lineage-restricted stem
cells.54 However, the efficiency of marking was low and the need to
irradiate the bone marrow to induce chromosomal translocation
meant that it was difficult to establish whether the identified HSC
classes truly existed or were experimentally induced artifacts. With
the development of more benign methods for HSC clonal tracking
using retroviral insertion sites, it was possible to establish that
long-term, multipotential HSCs existed.55-57 Collectively, these
studies established that clonal assays are essential to uncover
stem cell properties; and when individual HSCs are assessed,
they exhibit wide variations in repopulation dynamics and
self-renewal capacity.
CANCER STEM CELLS 4797
Applying universal stem cell concepts to
other fields
While these concepts were solidifying, stem cell concepts began to
percolate into other research fields through a series of culminating
events and set the stage for further discovery. Perhaps the most
influential factor that drove stem cell concepts beyond the domain
of a small number of stem cell biologists into a wider sphere of
biomedical research was the development of HSC transplantation
for treatment of leukemia and other hematologic diseases. Through
a series of elegant studies in the 1950s, proof was provided that
cells (as opposed to a humoral factor) from the bone marrow of
mice could protect against lethal doses of radiation by effectively
regenerating all of the blood lineages in transplant recipients.52,58 In
1956, Thomas et al carried out the first human bone marrow
transplantations,59 although initially a failure, they brought a
decade of improvements in tissue typing and methods for control-
ling graft-vs-host disease. Finally, the pioneering clinical trials of
Good, Thomas, and others showed that human BM could be used to
cure immune deficiency by sibling donor transplants in 196860 and
neoplastic disease in 1969,61 beginning the modern era of human
allogeneic marrow grafting. During the same time period, the new
field of molecular biology was established. This made it possible
not only to use genetics to characterize biologic processes in a
much more rapid fashion, but also to produce biologic regulators
on an industrial scale. All of these research paths collided in the
1980s with the birth of the biotechnology industry, which was
mostly focused on the commercialization of products to stimulate
blood cells. Thus, the arcane fields of blood diseases, HSC transplanta-
tion, and stem cells in general had moved from a backwater to front
stage. The final influence in this fertile period was the discovery of
embryonic stem cells by Evans and Kaufman62 and Martin63 in 1981
with the attendant promise of regenerative medicine. Stem cell biology
had finally become a hot topic, and it had begun to capture the attention
of a broad swath of the research community.
Technology and assay developments
It was within this milieu that 2 important technologic advances
took place that directly led to the renewal of interest in the concepts
of LSC from 20 years earlier. The first was the development of
fluorescence-activated cell sorting (FACS), and the second was the
development of xenotransplantation assays for normal and leuke-
mic human stem cells. Although Miller and Phillips had developed
the first large-scale methods for cell separation based on sedimenta-
tion,64 these methods lacked the precision needed for meticulous
cell fractionation. The development of FACS by Herzenberg’s
group at Stanford in 1972,65 combined with technology to create
specific mAbs against cell surface differentiation antigens, finally
provided the means to fractionate cells with high purity. As applied
to stem cell fractionation, Visser et al were the first to use flow
sorting based on wheat germ agglutinin binding as a means to
enrich (6-fold) CFU-S compared with unfractionated bone marrow
cells.66 Later it was possible to sort on the basis of cell cycle status
using Hoechst dyes, and in this way CFU-S were distinguished
from the more actively dividing granulocyte-macrophage-CFU
progenitor cells.67 Improved cell fractionation based on cell surface
markers and longer-term functional assays culminated in an
important paper by Spangrude et al who purified mouse multipoten-
tial HSCs (although mostly those capable of relatively short-term
4798 DICK
repopulation) by combining depletion of several markers of
maturation with expression of the Sca1 and c-Kit proteins (LSK
cells).68 Improvements in markers and the use of single-cell
transplantations have now resulted in essentially pure populations
of HSCs capable of multipotential repopulation.40,69-72 However,
only 20% to 50% of these clonal HSCs are capable of long-term
repopulation; it remains unclear whether these HSC fractions are
still not pure and that additional markers may be found to
prospectively isolate HSCs with high capacity for self-renewal away
from those with more limited capacity, or whether this remaining
variability is because of stochastic influences on an otherwise pure HSC
population.39,40 The intense efforts to purify murine HSCs and their
downstream progeny established a pathway of discovery that could be
applied to characterize human hematopoiesis.
Because studies of murine hematopoiesis showed that HSCs are
distinct from CFU and that HSCs with short-term repopulation
capacity are distinct from those with long-term capacity, it was
clear that progress to understand human hematopoiesis would
require a repopulation assay. The second key technologic develop-
ment leading to the identification of LSC was the beginning of the
modern era of the xenotransplantation system, which provided the
key ability to functionally assay both normal stem cells and
LSCs.73 Although there was a long history of transplanting human
tumor cells into irradiated or immune-deficient nude mice, the
strong host resistance in these mice allowed only the most
aggressive tumors and cell lines to be engrafted. Primary human
leukemia cells could not be transplanted and most cell lines only
grew subcutaneously or as ascites, which was not a normal pattern
for hematologic proliferation. In 1988, 3 groups using severe
combined immunodeficiency (SCID) or bg/nu/xid (BNX) mice,
which are more immune-deficient than nude mice, showed that
normal human hematopoietic cells could repopulate these mice.
Mosier et al transplanted human peripheral blood leukocytes by
intraperitoneal injection into SCID mice to enable long-term
generation of mature T cells that could then be infected with HIV.74
McCune et al were also interested in developing an HIV model.
They implanted SCID mice with pieces of human fetal liver, lymph
node, and thymus,75 and later they injected human cells into fetal
bone.76 The rationale of their approach was to provide a human
tissue microenvironment that would continuously generate human
T cells for HIV studies because the fetal liver (or fetal bone) would
provide a source of stem cells. Of course, the long-term production
of human cells within the fetal bone could be used to assay and
identify HSCs, a premise that came to fruition when they showed
that purified CD34⫹Thy⫹ cells could initiate multilineage engraft-
ment of this bone fragment.77 Our approach was somewhat
different and guided by the principles of bone marrow transplanta-
tion in humans and mice: intravenous transplantation into irradi-
ated recipients resulting in human cells repopulating the hematopoi-
etic tissues of a mouse.78-80 With improvements in recipient mice,81
it was finally possible to quantify and identify the cell that was
capable of initiating a multilineage human graft, which was termed
the SCID-repopulating cell (SRC).82-85 The SRC measured in this
system generated many orders of magnitude higher cell numbers
than the fetal bone in the SCID-hu model, providing a more robust
method to detect only those cells with the highest repopulation
capacity. Many improvements to the NOD/SCID assay continue to
be made by using recipient mice that are engineered to be deficient
in both NK and macrophage activity86,87 or that have been treated
with a mAb to CD122 that depletes these components of the innate
immune system.88-90 Recent studies that identified polymorphic
alleles of Sirp-alpha governing human HSC engraftment in NOD/
BLOOD, 15 DECEMBER 2008 䡠 VOLUME 112, NUMBER 13
SCID mice establish the importance of controlling macrophages to
achieve HSC engraftment.91 Finally, direct intrafemoral delivery of
cells resulted in more sensitive SRC detection, as well as assess-
ment of progenitors that possess transient repopulation potential, or
are lineage restricted and would not read out in classic intravenous-
based HSC transplantation assays.90,92-95
Characterization of normal human HSCs
The SRC xenotransplantation assay has emerged as the “gold
standard” surrogate assay for human HSCs.96 Combined with
clonogenic and stromal-based assays, it has been possible to
identify many cell surface markers that define the developmental
hierarchy of human hematopoiesis, setting the stage to apply these
same markers to understanding leukemic hematopoiesis. A detailed
understanding of the repopulation potential, cell-surface pheno-
type, differentiation, and proliferation properties of SRCs has been
gained over the past 15 years.96,97 Most SRCs express CD34, and
distinct SRC activities can be found within the lineage-negative
(Lin⫺) CD34⫹CD38⫺ and Lin⫺CD34⫹CD38lo subfractions.88,92,98-100
Personally, the culmination of our normal HSC studies was the
use of lentivector-mediated clonal tracking101,102 to monitor greater
than 600 clones in a cohort of primary and secondary mice for a
total of 7 months.39 These studies touched on the longstanding
debate on whether the control of cell fates is rigid and tightly
organized or whether these outcomes are unpredictable and gov-
erned by stochastic processes. A single HSC can generate the entire
blood system for a lifetime,40,69-72 a result that can only occur if at
cell division the HSC self-renews to produce at least one daughter
cell with the undifferentiated characteristics of the parental stem
cell. Whether a hematopoietic clone is maintained or extinguished
ultimately rests on whether an HSC makes an asymmetric (mainte-
nance) or a symmetric self-renewing (expansion) or differentiating
(extinction) division. These outcomes are known as the “birth and
death” fates of stem cells.36 (see McKenzie et al39 for a more
detailed discussion). We clonally analyzed hundreds of individual
SRCs in primary and secondary mice; as well, we had the
opportunity to analyze the repopulation properties of SRC daughter
cells that became physically separated to different marrow loca-
tions within a single mouse.39 We found that the proliferation and
self-renewal properties varied greatly. Even with increasing SRC
purification, the variation in self-renewal capacity remained. Our
results are reminiscent of the murine CFU-S data that were first
reported approximately 45 years ago that formed the basis for the
prediction that regulation of individual HSC outcomes is stochastic
and governed by probabilistic elements.36 Our data provided
experimental in vivo evidence (albeit with all the caveats of a
xenotransplantation model) that is consistent with the hypothesis
that human HSCs are also governed at least in part according to
stochastic principles. Collectively, the identification and character-
ization of human HSCs set the stage for identification of their
leukemic counterparts. The subsequent comparisons between nor-
mal and leukemic stem cells continue to provide clues to the
cellular and molecular programs that drive their biology.
Identification of leukemia stem cells in AML
Because we were able to transplant normal human cells into
mice, it was obvious that we should attempt to grow leukemic
cells in mice. Our initial studies transplanting pediatric pre-B
BLOOD, 15 DECEMBER 2008 䡠 VOLUME 112, NUMBER 13
ALL into SCID mice showed the power of the system to
propagate ALL with great similarity to the human disease103 and
correlation to clinical outcome.104 Because we were interested in
establishing whether normal HSCs could be assayed in SCID
mice, this focus on thinking clonally impacted our leukemia
studies. At this point, we turned our attention to AML because of
the more extensive work on clonogenic assay systems and the
application of stem cell concepts to understand this disease. I
had been especially enamored by the clonogenic assay for AML
progenitors and the studies from Buick and McCulloch, which
showed that only a subset of AML-CFU had high serial replating
efficiency, pointing to heterogeneity in self-renewal capac-
ity.22,29,30 In parallel, the studies outlined in “Identification of
normal HSCs in the mouse” clearly showed that normal
clonogenic progenitors were distinct from repopulating HSCs,
raising the question of whether the same would hold true for
AML and that a more primitive LSC existed that was distinct
from AML-CFU. The question was how to assay such a cell. As
with any new line of investigation, collegial support was
essential, especially from M. Minden and T. Hoang who
provided primary AML samples and taught us how to properly
thaw cells. We initiated experiments transplanting primary AML
into SCID mice and obtained reliable engraftment from most
subtypes of AML, except for APL. The subsequent critical
experiments were performed with the prompting of my mentor
R. A. Phillips who suggested we establish that the assay was
quantitative by measuring the frequency of the initiating cell.
This suggestion launched the application of stem cell principles
to the study of the initiating cell in AML. Our first step was to
establish the frequency of the initiating cell. Although we found
it to vary by more than 1000-fold between donors, it was
exceedingly rare: on the average of 1 per 106 cells. This was the
first hint as to whether this initiating cell was actually distinct
from the AML-CFU, which has a typical frequency of 1 in 100.
Second, with the suggestion of my colleague N. Iscove, we
carried out kinetic studies to show expansion of the AML-CFU
pool during repopulation, as would be expected if they were
being generated from an earlier precursor. Of course, it was
possible that the initiating cell and the AML-CFU were the same
cell if in vivo repopulation were simply an inefficient method to
assay AML-CFU. Clearly, the only way to resolve this question
was by fractionating the cells (ie, separating AML-CFU away
from the initiating cells). CD34 was an obvious choice to
fractionate with because expression on human HSCs had already
been shown105 and adopted clinically for HSC enrichment. Prior
work from Terstappen et al had pointed to CD34 and CD38 as
being heterogeneously expressed in AML with the prediction
that the various combinations of these 2 markers reflected a
developmental hierarchy, with CD34⫹CD38⫺ being the most
primitive.106 The immediate obstacle was that I knew little about
flow cytometry, and we did not have robust sorting capabilities
in Toronto at that time. However, I had made the fortunate
acquaintance of M. Caliguiri from Roswell Park who offered us
access to their new highly modified beta test FACStar instru-
ment run by C. Stewart. I have vivid memories of traveling with
my postdoctoral fellow T. Lapidot (who deserves enormous
credit for these studies) to Buffalo, sorting cells all day and
racing back to Toronto to transplant the precious sorted cells
into mice. I still recall with amusement the difficulties that the
U.S. border agents had with his Hebrew Israeli passport. As
others had seen, CD34 expression was highly variable in the
AML samples we had evaluated and the one FAB-M2 sample we
CANCER STEM CELLS 4799
had used for sorting had greater than 40% CD34⫹ cells;
however, CD38 was a smaller fraction of the CD34⫹ cells.
AML-CFU activity was high in the CD38⫹ fraction compared
with unsorted leukemia blasts but was also present in the CD38⫺
fraction. Remarkably, despite the rarity of the CD34⫹CD38⫺
fraction and the presence of high levels of AML-CFU in the
CD34⫹CD38⫹ fraction, leukemic engraftment could only be
initiated from CD34⫹CD38⫺ fractions. In keeping with a
tradition of naming cells based on function, we termed this the
SCID leukemia-initiating cell (SL-IC). The CD34⫹CD38⫺ SL-IC
generated large numbers of CD34⫹CD38⫹cells, AML-CFU, and
mature blasts. This was proof that the cell population that
initiated AML in SCID mice was distinct from AML-CFU; we
had identified a new stem cell in AML.107
Now is probably the best time to confess that I had no idea at the
time that this experiment settled an even larger issue about the
stochastic/hierarchy controversy, at least for AML. Not having read
carefully all the literature I have outlined in this review, I did not
realize that such a controversy existed because the hierarchy model
was so well entrenched in the normal and leukemic hematopoiesis
field; it was simply a given. In my view, all we had provided was
the long sought after formal proof for the concept of a hierarchy in
AML. There was one issue that concerned us. We had only tested a
single sample in this initial study and it had a primitive FAB-M1
subtype and a high CD34 content. Perhaps the principle, or at least
the CD34⫹CD38⫺ phenotype, would not hold for AML, which
exhibited a higher degree of cellular maturation, particularly the
myelomonocytic M4/M5 subtypes. By this time, the new NOD/
SCID mouse strain was found to be a superior recipient for normal
HSC engraftment, permitting, for the first time, quantitative
assessment and purification of normal HSCs.82-85 With this more
robust assay, we studied in detail another 7 samples, 6 being M4 or
M5. Many of these samples exhibited a distinct pattern of
engraftment different from the primitive samples. The myelomono-
cytic samples disseminated widely, whereas the primitive subtypes
typically grew well only in the bone marrow and only disseminated
after long periods of time. Interestingly, although the M4 and M5
samples disseminated widely into many organs, they never infil-
trated the central nervous system. This pattern was distinct from the
pre-B ALL samples that we had studied earlier, which also
disseminated widely but often infiltrated the central nervous system
and the testis. These disease patterns that we observed in the
NOD/SCID mouse reflected the clinical features of the disease in
humans, pointing to the reliability of these new xenotransplantation
models (compared with the older subcutaneous implantations in
nude mice) to mimic the human disease. The cell purification data
from these newer experiments were clear: despite their myelomono-
cytic morphology and their low CD34⫹ and CD34⫹CD38⫺ popula-
tions, the LSCs were all found in the CD34⫹CD38⫺ fraction.108
Although some exceptions were reported in subsequent years, a
high proportion of AML patients show LSC activity in the
CD34⫹CD38⫺ fraction.109-111 As well, recent studies have sug-
gested that the ability to engraft NOD/SCID mice112 as well as
increased frequencies of LSC in AML are prognostic for poor
outcomes.113 Collectively, these studies provide conclusive evi-
dence for the hierarchy model in AML.
Cell of origin
When it came time to write the paper, D. Bonnet and I spent much
effort reading the older literature, and we realized that our data
4800 DICK
could address the “cell of origin” question: what is the normal cell
from which leukemia arises? Work from a number of directions,
including the variation in cell surface marker expression on
AML-CFU by Sabbath et al28 described earlier and the pioneering
clonality studies in humans by Fialkow et al using X-linked G6PD
polymorphisms114,115 followed by DNA-based restriction fragment
length polymorphism analysis by Fearon et al,116 had been used to
develop a theory that the maturation state of the leukemia blast
reflected the level of maturation of the cell of origin.25 In vivo
clonal studies in humans examined whether the normal nonleuke-
mic lineages (lymphocyte, erythroid cell) were or were not part of
the leukemic clone. Such an approach had definitively established
that chronic myeloid leukemia (CML) arose in a multipotential
HSC.117 If the nonleukemic cells were part of the clone, this
reflected an early multipotential cell of origin, whereas if they were
not part of the same clone, this reflected a more lineage-restricted
cell of origin. This view was also proposed based on some of the
earlier in vivo cytokinetic studies of Clarkson and others, albeit
with less evidence than these clonal analyses could bring to bear.
The alternative view, best articulated by McCulloch, was that
observed lineage restriction was indeed only an apparent restriction
and could not rule out the possibility that even the mature or lineage
restricted AMLs arose from a multipotent HSC.22,29 He argued that
transforming events could occur in an HSC, but one of the
consequences of this event was an inability of this stem cell to
generate other “normal” lineage cells, such as erythroid cells. As
discussed in the following paragraph, we proposed that the similar
cell surface phenotype of the LSC and the HSC supported the idea
that primitive cells were the cell of origin for AML regardless of the
maturation state of the leukemia blasts.108 Perhaps more than any
other, this paper has caused confusion about the nomenclature of a
“cancer stem cell,” especially when the CSC field exploded in
2003/2004 when solid tumor CSCs were identified.118,119 Our use of
this nomenclature comes from the normal blood literature where
cells are defined by function, in this case, the cancer cell that
possesses stem cell properties. Our attempts to address the “cell of
origin” question were taken by some to mean that a CSC was
derived from a normal stem cell that had become cancerous. It
finally took a blue-ribbon panel established by the American
Association for Cancer Research to strongly declare that a CSC
BLOOD, 15 DECEMBER 2008 䡠 VOLUME 112, NUMBER 13
Figure 3. Hierarchy of leukemia stem cells in AML.
Like the normal hematopoietic system, AML is organized
as a hierarchy of distinct cell classes that is sustained by
a subset of leukemia stem cells (or SCID-leukemia
initiating cells [SL-ICs], as assayed in immunodeficient
mice). Genetic tracking experiments have shown that
SL-ICs are heterogeneous in their ability to repopulate
secondary and tertiary recipients, pointing to the exis-
tence of distinct classes with differing self-renewal capac-
ity, similar to what is seen in the normal hematopoietic
stem cell compartment. Short-term (ST) SL-ICs are able
to initiate leukemia in primary but not secondary recipi-
ents, whereas long-term (LT) SL-ICs can sustain leuke-
mic growth for multiple passages. Quiescent LT SL-ICs
may not initiate a substantial graft in primary recipients
and may therefore only be detected on serial
transplantation.
simply meant the cell in a cancer that has stem cell–like properties,
especially the capacity for self-renewal.120
Although we made the argument that the cell of origin is
primitive based on the consistency of the CD34⫹CD38⫺ cell
surface phenotype of SL-IC among patients regardless of AML
subtype and their similarity to the phenotype of normal SRC, our
reasoning had serious limitations. Cell surface markers are antigens
whose expression changes with differentiation, and we know
differentiation is perturbed in cancer. Therefore, one must be
especially cautious when using a defined set of markers that rigidly
reflect the steps of lineage determination of normal cells and then
inferring that the same set will apply for neoplastic cells. Ideally,
one would like to make comparisons between normal HSCs and
LSCs using functional criteria. The only true stem cell property is
capacity for maturation combined with self-renewal. In normal
hematopoiesis, cell tracking and purification studies performed
clonally (in mice and humans) have documented that long-term
repopulating HSCs with high self-renewal capacity generate short-
term HSCs whose self-renewal capacity is more limited.101,121
Thus, we carried out lentivector-mediated clonal tracking of
AML-LSC in a manner identical to that outlined in “Characteriza-
tion of normal human HSCs” for normal HSCs. These studies
established that individual LSCs were not homogeneous but had
wide variation in repopulation potential and self-renewal capacity
similar to that observed for normal HSCs122 (Figure 3). However, a
key difference is that, during NOD/SCID repopulation, LSCs
expand to a much greater extent than normal HSCs, probably
because of increased symmetric self-renewal cell divisions.108 As
more information is gathered about the molecular machinery that
underlies the self-renewal process, it is becoming clearer that LSCs
and HSCs share at least some characteristics. For example, Bmi1123
and PTEN124 govern self-renewal of both murine HSCs and LSCs.
Together with the findings that LSC can also be quiescent like
normal HSCs,122,125-127 these discoveries add additional support to
the notion that some types of AML may arise from within the primitive
cell compartment, before full lineage commitment has occurred.Alterna-
tively, it is also possible that AML could arise from committed
progenitors if stem cell self-renewal programs are established during the
leukemogenic process. Nevertheless, this evidence is still indirect and
requires a more direct experimental approach.
BLOOD, 15 DECEMBER 2008 䡠 VOLUME 112, NUMBER 13
The most direct experiment is to introduce the same genetic
element into pure populations of HSCs or progenitor subsets to
determine which population is competent to initiate disease.
Attempts to identify the cell of origin for murine leukemia have
yielded conflicting results, demonstrating that cellular and molecu-
lar context is critically important. The fusion oncogene MLL-
GAS7 induces mixed-lineage leukemias when expressed in HSCs
or multipotent progenitors, but not in lineage-restricted progeni-
tors.128 In contrast, MOZ-TIF,129 MLL-AF9,130 and MLL-ENL131
lead to initiation of AML regardless of the target cell population
used. However, much lower numbers of transformed HSCs com-
pared with progenitors are often required for tumor initiation in
vivo and tumors arising from progenitors are monoclonal or
oligoclonal not polyclonal, suggesting that the progenitor popula-
tions were not functionally homogeneously, were not equally
susceptible to transformation, or additional genetic alterations were
required. Murine models of CML resulting from inactivation of
JunB or expressing bcr/abl must occur in HSCs and not more
restricted progenitors to induce a transplantable myeloproliferative
disorder.129,132 These murine studies have become even more
difficult to interpret with the recent findings of Chen et al who used
a knockin model of MLL-AF9.133 This approach was taken because
of concerns that, for some oncogenes, the level of expression from
exogenous promoters, such as transgenes or retroviral vectors, may
be nonphysiologic. They found that the same construct shown to
initiate disease in HSCs and progenitors by retroviral expression
only initiated leukemia from the HSC fraction when expressed
from the endogenous MLL promoter. Nevertheless, collectively,
these studies demonstrate that, under the correct circumstances and
with the correct oncogene, it is possible to transform non-HSC in
mice. The critical feature of these studies is that some component
of an HSC or self-renewal program must be established in the
non-HSC cell types.134
No direct studies of the design used for the murine studies have
been reported for human hematopoietic cells. However, the in vivo
clonality studies114,115 described earlier had pointed to variation in
the cell of origin. A more refined approach was to examine whether
phenotypically defined populations contained specific transloca-
tions and whether these exhibited normal or leukemic proliferation.
A study of patients with t(8;21) AML demonstrated that, despite the
presence of leukemia-specific AML1-ETO chimeric transcripts in
primitive CD34⫹CD90⫺CD38⫺ cells from leukemic bone marrow,
these cells give rise to normally differentiating multilineage
clonogenic progenitors, whereas more mature CD34⫹CD90⫺CD38⫹
cells form exclusively leukemic blast colonies (AML-CFC).135
Recent studies have documented that normal HSCs are
CD34⫹CD90⫹CD38⫺, whereas the CD34⫹CD90⫺CD38⫺ fraction
represents a normal multipotential progenitor with little self-
renewal capacity.136 Thus, a conclusion of this work is that,
whereas the initial t(8;21) translocation occurs in a primitive stem
cell, subsequent events occur in the committed progenitor pool,
giving rise to LSCs. However, in vivo assays are required for
formal proof that these are LSCs. A similar process of leukemic
progression recently has been proposed for CML.137 The discovery
of a marker chromosome,138 the nature of the translocation,139
followed by lineage involvement of the Ph chromosome,117 estab-
lished that CML arose from an HSC; however, analysis of blast
crisis samples suggests that additional genetic events occur at the
level of more committed progenitors, resulting in the generation of
LSCs.137 These provocative studies point to the need to understand
not only the cell of origin of the LSCs but also the steps leading to
so-called “preleukemic stem cells.”
CANCER STEM CELLS 4801
Our current understanding of the potential pathways leading to
generating CSCs and a proposed model for tumor progression are
depicted in Figure 4. Because the lifespan of non–stem cells may be
transient, any genetic alteration in their makeup may not persist
long enough to acquire additional genetic events that are needed for
full transformation, unless the initial event endows the non-stem
cell with self-renewal capacity. In addition, even when the stem cell
sustains the first hit, subsequent events may occur in the stem cell
or in the downstream progenitors. Another consideration is deter-
mining how the hierarchy model relates to leukemic progression or
tumor evolution. Figure 4 proposes that tumor progression is linked
to self-renewal capacity. For example, genetic instability and clonal
evolution can occur in both CSCs and non-CSCs, keeping in mind
that the rate of evolution could be different if CSCs proliferate
more slowly than non-CSCs. Within the CSC fraction, additional
genetic and epigenetic alterations could result in more abnormal
CSCs; and in the case of solid tumors, these could become
metastatic. The non-CSCs could also acquire genetic and epige-
netic alterations; but if these do not endow self-renewal capacity,
then these alterations in the non-CSCs are passengers and not
drivers of tumor progression. Finally, if stem cell-like properties,
including self-renewal, were endowed on non-CSCs, they could
also drive tumor progression (Figure 4). A recent report has
established in a breast cancer model that an epithelial-to-
mesenchymal transition program is linked to stem cell function and
experimental induction of epithelial-to-mesenchymal transition in
a non-CSCs can reprogram them to a CSC state, providing
evidence for this tumor progression model.140
Modeling the human leukemogenic process
A major limitation in understanding the leukemogenic process in
humans is the absence of good genetic tools and experimental
models that use primary human cells as their starting point. For
example, the first step to directly test the cell of origin hypothesis
requires an experimental model of human leukemogenesis where
oncogenes are transduced into primary human hematopoietic cells.
Several reports have demonstrated that the initial steps in the
leukemogenic process could be induced experimentally by expres-
sion of several different oncogenes, such as Ras,141 TLS-
ERG,142,143 AML-ETO,144 and bcr/abl,145 but none caused full-
blown disease (see Kennedy and Barabé146 for detailed review of
human leukemogenesis studies). Using MLL fusions, 2 recent
studies have demonstrated successful induction of AML and pre-B
ALL in NOD/SCID transplanted with transduced primary stem/
progenitor cells.147,148 These mice developed B-ALL and AML that
reflected the human disease. We found that the B-ALL in this
model maintained a germ line configuration of the IgH locus early
in disease evolution but progressed to full rearrangement on serial
passage.147 These data suggest that the LSC can evolve and that the
cell of origin was not a mature B cell, but a more primitive stem or
progenitor cell. Direct elucidation of the cell of origin will require
prospective cell sorting of pure, functionally defined HSCs, and
progenitor subsets as targets for testing.
Although the NOD/SCID studies were compelling, direct
evidence for LSC evolution in humans was lacking. In a seminal
series of studies beginning in the 1990s, Greaves and Wiemels
demonstrated that many forms of childhood leukemia arose in
utero.149 They studied twins in which one twin has leukemia
whereas the other does not; both were shown to contain the same
chromosomal translocations (MLL, Tel-AML1), a result that could
4802 DICK BLOOD, 15 DECEMBER 2008 䡠 VOLUME 112, NUMBER 13

Figure 4. Models of tumor initiation and progression. Cancer stem cells may arise through neoplastic changes initiated in normal self-renewing stem cells or downstream
progenitors, causing expansion of the stem cell and/or progenitor pool. Secondary events may occur in expanded pools of target cells. Oncogenic events acquired by
short-lived progenitors may not persist if self-renewal is not reactivated, as these cells will probably die or undergo terminal differentiation before enough mutations occur for full
neoplastic transformation. Tumor progression may be linked to ongoing genetic instability and acquisition of additional changes by cancer stem cells, or possibly by
nontumorigenic bulk cells if such changes endow self-renewal. In both cases, evolution of tumor phenotype (including genetic and epigenetic signatures) may be observed.

only occur if a single cell in one twin acquired the translocation,
expanded, and migrated to the other twin. By examination of Ig
rearrangements, they concluded that after birth additional genetic
alterations occur independently in each twin, resulting in the
appearance of leukemia at different times. This predicts that the
twin who has not yet developed leukemia must contain a preleuke-
mic stem cell that is poised to acquire additional alterations because
this twin is at high risk of developing leukemia. In an elegant series
of studies, Hong et al have identified a unique cell population, not
observed in normal donors, that contains a preleukemic stem cell
that exhibited enhanced self-renewal capacity when assessed in the
NOD/SCID xenotransplantation system.150 This preleukemic popu-
lation was clonally related to the LSC that maintained the leukemia
in the affected twin. Thus, this study provides strong proof in
humans for the concept of clonal evolution of LSC from a
preleukemic to a leukemic state.
Controversies
Recently, several reports were published that questioned the
universality of the CSC or hierarchy model.151,152 Several experi-
mentally induced murine leukemia models were found where
almost every cancer cell had LSC activity and there was little
evidence of functional heterogeneity. These and other reports153
speculated that the rarity of human CSCs observed using xenotrans-
plantation models may be the result of host resistance factors as
well as the absence of cross-species reactivity of cytokines and
other microenvironmental factors. Certainly, such factors play a
role because the use of more immune-deficient mice has improved
human CSC detection; in addition, some solid tumor studies are
now using coimplantation of human stromal cells. It is important to
note that a number of other genetically induced murine leukemia
models are heterogeneous and contain discrete and rare LSCs.154
Moreover, when human and murine leukemia models were both
induced with the same MLL fusion protein, the LSC frequency was
the same.154 Thus, it is evident that LSC frequencies can vary
widely among different cancers regardless of whether they are
quantified using xeno- or syngeneic transplantation assays. It is
important to emphasize that the fundamental concept underlying
the CSC model is not directly related to the absolute frequency of
these cells but rather on functional evidence of tumor heterogeneity
and a hierarchical organization. The CSC model is only invoked to
explain tumor heterogeneity; however, if a tumor is functionally
BLOOD, 15 DECEMBER 2008 䡠 VOLUME 112, NUMBER 13
homogeneous, the CSC model is no longer relevant. It is clear that
some murine cancer models do not follow the hierarchy model, and
it will be important to survey a much broader cross section of
human tumors to determine whether any also lack tumor
heterogeneity.
The future of CSC research
Despite the recent explosion of interest in CSCs, experimental
studies have not been translated into improved survival outcomes
for cancer patients. This presents a major question to the field: do
cancer stem cells have meaningful relevance beyond the experimen-
tal systems in which they have been defined? Increasing evidence is
pointing to CSCs having unique biologic properties (dormancy,
drug/radiation resistance) that could permit them to survive thera-
pies leading to eventual relapse. To date, CSCs have been studied in
a relatively small number of patient samples and cancer types, so it
is not known whether the CSC model is universal to all human
cancers.1 The biologic relevance of CSCs in human cancer will be
established by concentrating on the following research endeavors:
improving the assay and purification of CSC and non-CSC subsets,
carrying out detailed genomic or proteomic analysis on these
subsets to identify CSC-specific signatures, and obtaining such
signatures from a large number and wide range of tumors. Such an
approach would make it possible to determine whether the “omics”
of CSCs provide more predictive or prognostic relevance compared
with analysis of the bulk tumor. An approach like this has recently
been taken by Liu et al who showed that a breast CSC signature has
improved correlation to patient outcomes.155 Of course, the most
direct proof for the relevance of CSCs to cancer would be to use the
CSC-specific signatures to elucidate the key molecular pathways
involved, develop the means to target them effectively, and
establish in clinical trials whether targeting them enhances patient
survival.
For many types of leukemia, the long-term survival has
increased dramatically over the past 50 years; however, AML
remains one of the most difficult diseases to cure. The majority
of patients will die because of relapse, drug resistance, or from
complications associated with chemotherapies. Because LSCs
and HSCs have many similar properties, a first step to develop
novel therapies has been to identify unique properties that LSCs
possess. Through expression analysis, Guzman et al determined
that the NF-␬B pathway was active in LSCs and not in HSCs.126
They then went on to identify a natural product, parthenolide,
which targets this pathway and showed that the LSCs could be
specifically targeted using the NOD/SCID model.156,157 It should
be possible to develop a strategy that focuses on the biology of
LSC and develop assay methods that are compatible with high
throughput screening of small molecule libraries followed by
counterscreens with normal HSCs or progenitors.158 Other
approaches focus on identification of differences in cell surface
markers with concomitant development of immunoconjugates
with toxic moieties. LSCs could be targeted with a diphtheria
toxin conjugated to granulocyte colony-stimulating factor immu-
notoxin.159,160 However, not all AML samples responded to these
molecules, and regrowth or unsuccessful eradication of the
leukemic cells occurred after different treatment strategies,
suggesting that not all classes of SL-IC were killed by diphtheria
toxin conjugated to granulocyte colony-stimulating factor immu-
notoxin. Similarly a DT-IL3 conjugate was shown to target the
LSC and progressed to clinical trials. Unfortunately, this therapy
CANCER STEM CELLS 4803
resulted in the development of antibodies as well as other
toxicities.161 As well, LSCs could be targeted with a CD8⫹
cytotoxic T lymphocyte clone specific for minor histocompatibil-
ity antigens,162,163 an approach that might be useful in the
context of AML patients who undergo allogeneic transplant and
then relapse. More recently, several studies reported that LSCs
could be targeted with mAb to CD44164 or CXCR4.165 When
these mAbs were injected into NOD/SCID mice repopulated
with AML, the LSCs were markedly depleted; and in some
cases, the mice were cured. Induction of differentiation repre-
sented a partial mechanism of action. Even more noteworthy,
both CXCR4 and CD44 resulted in impaired trafficking of the
LSC, suggesting that LSCs remain dependent on niche interac-
tions. Because many other factors are involved in trafficking and
homing (see 50th Anniversary Review by Papayannopoulou and
Scadden166), this finding opens the way for testing the anti-LSC
functionality of any molecule that interferes with these pro-
cesses. Little is known of the LSC niche; however, a recent
study points to the endosteal region,167 an area that has been
proposed as the niche for murine HSCs as well. Another
LSC-specific target, which is either low or absent from normal
HSCs, is CD123, the alpha chain of the IL3R.168 A mAb
targeting CD123 impairing homing of LSC to bone marrow of
NOD/SCID mice, and it activates the innate immunity of
NOD/SCID mice resulting in eradication of LSCs.169 Collec-
tively, these studies provide clear validation for therapeutic
mAb targeting of AML-LSCs and for translation of the NOD/
SCID in vivo preclinical research findings toward a clinical
application. Finally, future therapeutic approaches will need to
be developed that target the self-renewal machinery of LSCs,
although it may be difficult to find a therapeutic window that
will spare normal tissue stem cells. As predicted from the
cytokinetic studies, perhaps the most attractive option is to
induce the quiescent LSCs into cycle. Recent studies are
pointing to specific molecular pathways that might be attractive,
including PTEN, p21, and PML. Ito et al have found that
degradation of PML induces cycling of LSC and, interestingly,
that one of the mechanisms of action of arsenic is degradation of
PML, already putting a potential drug strategy in hand.170
Although this review is focused on the premise that improve-
ments in the stubbornly high relapse rates for many cancers will
come from more research on the CSCs that drive the tumors,
attention is needed on how to move CSC therapies into clinical
trials and into the clinic (a more thorough discussion is done by
Wang171). The current approaches to preclinical drug development
and clinical evaluation of drug efficacy focus on initial require-
ments for tumor response. Therefore, a potentially effective
anti-CSC therapy might be discarded from further testing if it does
not cause a response of the non-CSCs cells rapidly enough.
Similarly, an effective anti-CSC might take time to show efficacy
or be most effective in an adjuvant setting. Much research effort
will be required to find the best algorithms for clinical evaluation of
CSC-targeted therapies.
In conclusion, stem cell concepts lie at the heart of the first
cancer therapies developed almost 40 years ago, which showed the
first tangible success in curing patients with acute leukemia. One
hopes that the renewal of interest in the linkage between stem cells
and cancer will lead to a better understanding of how the cellular
context of tumors affects cancer-specific genetic and epigenetic
pathways and will once again drive new therapeutic advances.
4804 DICK BLOOD, 15 DECEMBER 2008 䡠 VOLUME 112, NUMBER 13

Canadian Cancer Society and the Terry Fox Foundation,

Acknowledgments research contracts with CSL and Arius, and a Canada
Research Chair.
The author thanks the Dick laboratory members, Jean Wang, Bob
Phillips, Norman Iscove, Tsvee Lapidot, and Mark Minden for
advice and comments; Barney Clarkson for discussions and for
sharing a treasure trove of “hard-to-find” papers; Jean Wang for Authorship
creating the figures; and Melissa Cooper for insightful editing.
This work was supported by grants from the Canadian Contribution: J.E.D. wrote the paper.
Institutes of Health Research (CIHR), the Ontario Institute for Conflict-of-interest disclosure: The author declares research
Cancer Research and a Summit Award both with funds from the contracts with CSL and Arius.
Province of Ontario, Genome Canada through the Ontario Correspondence: John E. Dick, Toronto Medical Discovery
Genomics Institute, the Leukemia & Lymphoma Society, Tower, 101 College Street, Room 8-301, Toronto, ON, Canada
and the National Cancer Institute of Canada with funds from the M5G 1L7; e-mail: jdick@uhnres.utoronto.ca.

References
1. Wang JC, Dick JE. Cancer stem cells: lessons
from leukemia. Trends Cell Biol. 2005;15:494-
501.
2. Dalerba P, Cho RW, Clarke MF. Cancer stem
cells: models and concepts. Annu Rev Med.
2007;58:267-284.
3. Ailles LE, Weissman IL. Cancer stem cells in
solid tumors. Curr Opin Biotechnol. 2007;18:460-
466.
4. Furth J, Kahn M. The transmission of leukemia of
mice with a single cell. Am J Cancer. 1937;31:
276-282.
5. Makino S. Further evidence favoring the concept
of the stem cell in ascites tumors of rats. Ann N Y
Acad Sci. 1956;63:818-830.
6. Hewitt HB. Studies of the dissemination and
quantitative transplantation of a lymphocytic leu-
kaemia of CBA m ice. Br J Cancer. 1958;12:378-
401.
7. Bruce WR, Van Der Gaag H. A quantitative assay
for the number of murine lymphoma cells capable
of proliferation in vivo. Nature. 1963;199:79-80.
8. Southam C, Brunschwig A, Dizon Q. Autologous
and homologous transplantation of human can-
cer. In: Brennan MJ, Simpson WL, eds. Biological
Interactions in Normal and Neoplastic Growth: A
Contribution to the Tumor-Host Problem. Boston,
MA: Little, Brown; 1962:723-738.
9. Pierce GB Jr, Dixon FJ Jr, Verney EL. Teratocar-
cinogenic and tissue-forming potentials of the cell
types comprising neoplastic embryoid bodies.
Lab Invest. 1960;9:583-602.
10. Pierce GB, Speers WC. Tumors as caricatures of
the process of tissue renewal: prospects for
therapy by directing differentiation. Cancer Res.
1988;48:1996-2004.
11. Belanger L, Leblond C. A method for locating ra-
dioactive elements in tissues by covering histo-
logical sections with a photographic emulsion.
Endocrinology. 1946;39:386-400.
12. Clermont Y, Leblond C. Renewal of spermatago-
nia in the rat testis. Am J Anat. 1953;93:475-502.
13. Clarkson B, Fried J, Strife A, Sakai Y, Ota K, Okita
T. Studies of cellular proliferation in human leuke-
mia: III. Behavior of leukemic cells in three adults
with acute leukemia given continuous infusions of
3H-thymidine for 8 or 10 days. Cancer. 1970;25:
1237-1260.
14. Clarkson BD, Fried J. Changing concepts of
treatment in acute leukemia. Med Clin North Am.
1971;55:561-600.
15. Killmann SA, Cronkite EP, Robertson JS, Fliedner
TM, Bond VP. Estimation of phases of the life
cycle of leukemic cells from labeling in human
beings in vivo with tritiated thymidine. Lab Invest.
1963;12:671-684.
16. Gavosto F. The proliferative kinetics of the acute
leukaemias in relation to their treatment. Rev Eur
Etud Clin Biol. 1970;15:1042-1047.
17. Clarkson BD. Review of recent studies of cellular
proliferation in acute leukemia. Natl Cancer Inst
Monogr. 1969;30:81-120.
18. Clarkson B. The survival value of the dormant
state in neoplastic and normal populations. In:
Clarkson B, Baserga R, eds. Control of Prolifera-
tion in Animal Cells. New York, NY: Cold Spring
Harbor Laboratory; 1974;945-972.
19. Cronkite EP. Acute leukemia: is there a relation-
ship between cell growth kinetics and response to
chemotherapy? Proc Natl Cancer Conf. 1970;6:
113-117.
20. Clarkson BD, Dowling MD, Gee TS, Cunningham
IB, Burchenal JH. Treatment of acute leukemia in
adults. Cancer. 1975;36:775-795.
21. Buick RN, Till JE, McCulloch EA. Colony assay
for proliferative blast cells circulating in myelo-
blastic leukaemia. Lancet. 1977;1:862-863.
22. McCulloch E. Stem cells in normal and leukemic
hemopoiesis (Henry Stratton Lecture). Blood.
1983;62:1-13.
23. Metcalf D, Moore MA, Warner NL. Colony forma-
tion in vitro myelomonocytic leukemic cells. J Natl
Cancer Inst. 1969;43:983-1001.
24. Moore MA, Williams N, Metcalf D. In vitro colony
formation by normal and leukemic human hema-
topoietic cells: characterization of the colony-
forming cells. J Natl Cancer Inst. 1973;50:603-
623.
25. Griffin J, Löwenberg B. Clonogenic cells in acute
myeloblastic leukemia. Blood. 1986;68:1185-
1195.
26. Dicke KA, Spitzer G, Ahearn MJ. Colony forma-
tion in vitro by leukaemic cells in acute myelog-
enous leukaemia with phytohaemagglutinin as
stimulating factor. Nature. 1976;259:129-130.
27. Hamburger AW, Salmon SE. Primary bioassay of
human tumor stem cells. Science. 1977;197:461-
463.
28. Sabbath KD, Ball ED, Larcom P, Davis RB, Griffin
JD. Heterogeneity of clonogenic cells in acute
myeloblastic leukemia. J Clin Invest. 1985;75:
746-753.
29. McCulloch E, Till J. Blast cells in acute myelo-
blastic leukemia: a model. Blood Cells. 1981;7:
63-77.
30. Buick RN, Minden MD, McCulloch EA. Self-re-
newal in culture of proliferative blast progenitor
cells in acute myeloblastic leukemia. Blood. 1979;
54:95-104.
31. Maximov A. Der lymphozyt als gemeinsame
Stammzelle der verschieden Blutelemente in der
embryonalen Entwicklung und postfetalen Leben
der Sugetiere. Folia Haematologica (Leipzig).
1909;8:125-141.
32. Lichtman MA. The stem cell in the pathogenesis
and treatment of myelogenous leukemia: a per-
spective. Leukemia. 2001;15:1489-1494.
33. Till JE, McCulloch EA. A direct measurement of
the radiation sensitivity of normal mouse bone
marrow cells. Radiat Res. 1961;14:213-222.
34. Becker AJ, McCulloch EA, Till JE. Cytological
demonstration of the clonal nature of spleen colo-
nies derived from transplanted mouse marrow
cells. Nature. 1963;197:452-454.
35. Siminovitch L, McCulloch EA, Till JE. The distri-
bution of colony-forming cells among spleen colo-
nies. J Cell Comp Physiol. 1963;62:327-336.
36. Till JE, McCulloch EA, Siminovitch L. A stochastic
model of stem cell proliferation based on the
growth of spleen colony-forming cells. Proc Natl
Acad Sci U S A. 1964;51:29-36.
37. Lemischka IR. The haematopoietic stem cell and
its clonal progeny: mechanisms regulating the
hierarchy of primitive haematopoietic cells. Can-
cer Surv. 1992;15:3-18.
38. Morrison S, Uchida N, Weissman I. The biology of
hematopoietic stem cells. Annu Rev Cell Dev
Biol. 1995;11:35-71.
39. McKenzie JL, Gan OI, Doedens M, Wang JC,
Dick JE. Individual stem cells with highly variable
proliferation and self-renewal properties comprise
the human hematopoietic stem cell compartment.
Nat Immunol. 2006;7:1225-1233.
40. Sieburg HB, Cho RH, Dykstra B, Uchida N,
Eaves CJ, Muller-Sieburg CE. The hematopoietic
stem compartment consists of a limited number
of discrete stem cell subsets. Blood. 2006;107:
2311-2316.
41. Huang S, Guo YP, May G, Enver T. Bifurcation
dynamics in lineage-commitment in bipotent pro-
genitor cells. Dev Biol. 2007;305:695-713.
42. Pluznik DH, Sachs L. The cloning of normal
“mast” cells in tissue culture. J Cell Physiol. 1965;
66:319-324.
43. Bradley TR, Metcalf D. The growth of mouse
bone marrow cells in vitro. Aust J Exp Biol Med
Sci. 1966;44:287-299.
44. Metcalf D. Hematopoietic cytokines. Blood. 2008;
111:485-491.
45. Magli MC, Iscove NN, Odartchenko N. Transient
nature of early hematopoietic spleen colonies.
Nature. 1982;295:527-529.
46. Boggs DR, Boggs SS, Saxe DF, Gress LA,
Canfield DR. Hemopoietic stem cells with high
proliferative potential: assay of their concentra-
tion in marrow by the frequency and duration of
cure of W/Wv mice. J Clin Invest. 1982;70:242-
253.
47. Paige CJ, Kincade PW, Moore MAS, Lee G. The
fate of fetal and adult B-cell progenitors grafted
into immunodeficient CBA/N mice. J Exp Med.
1979;150:548-563.
48. Paige CJ, Kincade PW, Shinefeld LA, Sato VL.
Precursors of murine B lymphocytes: physical
and functional characterization, and distinctions
BLOOD, 15 DECEMBER 2008 䡠 VOLUME 112, NUMBER 13
from myeloid stem cells. J Exp Med. 1981;153:
154-165.
49. Hodgson GS, Bradley TR. Properties of hemato-
poietic stem cells surviving 5-fluorouracil treat-
ment: evidence for a pre-CFU-S cell? Nature.
1979;281:381-382.
50. Ploemacher RE, Brons RH. Separation of CFU-S
from primitive cells responsible for reconstitution
of the bone marrow hemopoietic stem cell com-
partment following irradiation: evidence for a pre-
CFU-S cell. Exp Hematol. 1989;17:263-266.
51. Harrison DE. Competitive repopulation: a new
assay for long-term stem cell functional capacity.
Blood. 1980;55:77-81.
52. Ford C, Hamerton J, Barns W, Loutet J. Cytologi-
cal identification of radiation chimeras. Nature.
1956;177:452-454.
53. Wu AM, Till JE, Siminovitch L, McCulloch EA. Cy-
tological evidence for a relationship between nor-
mal hemopoietic colony-forming cells and cells of
the lymphoid system. J Exp Med. 1968;127:455-
463.
54. Abramson S, Miller RG, Phillips RA. The identifi-
cation in adult bone marrow of pluripotent and
restricted stem cells of the myeloid and lymphoid
systems. J Exp Med. 1977;145:1567-1579.
55. Dick JE, Magli MC, Huszar D, Phillips RA,
Bernstein A. Introduction of a selectable gene into
primitive stem cells capable of long-term reconsti-
tution of the hemopoietic system of W/Wv mice.
Cell. 1985;42:71-79.
56. Keller G, Paige C, Gilboa E, Wagner E. Expres-
sion of a foreign gene in myeloid and lymphoid
cells derived from multipotent hemopoietic pre-
cursors. Nature. 1985;318:149-154.
57. Lemischka IR, Raulet DH, Mulligan RC. Develop-
mental potential and dynamic behavior of hema-
topoietic stem cells. Cell. 1986;45:917-927.
58. Lorenz E, Uphoff D, Reid TR, Shelton E. Modifi-
cation of irradiation injury in mice and guinea pigs
by bone marrow injections. J Natl Cancer Inst.
1951;12:197-201.
59. Thomas ED, Lochte HL Jr, Lu WC, Ferrebee JW.
Intravenous infusion of bone marrow in patients
receiving radiation and chemotherapy. N Engl
J Med. 1957;257:491-496.
60. Gatti RA, Meuwissen HJ, Allen HD, Hong R,
Good RA. Immunological reconstitution of sex-
linked lymphopenic immunological deficiency.
Lancet. 1968;2:1366-1369.
61. Buckner CD, Epstein RB, Rudolph RH, Clift RA,
Storb R, Thomas ED. Allogeneic marrow engraft-
ment following whole body irradiation in a patient
with leukemia. Blood. 1970;35:741-750.
62. Evans MJ, Kaufman MH. Establishment in culture
of pluripotential cells from mouse embryos. Na-
ture. 1981;292:154-156.
63. Martin GR. Isolation of a pluripotent cell line from
early mouse embryos cultured in medium condi-
tioned by teratocarcinoma stem cells. Proc Natl
Acad Sci U S A. 1981;78:7634-7638.
64. Miller RG, Phillips RA. Separation of cells by ve-
locity sedimentation. J Cell Physiol. 1969;73:191-
201.
65. Bonner WA, Hulett HR, Sweet RG, Herzenberg
LA. Fluorescence activated cell sorting. Rev Sci
Instrum. 1972;43:404-409.
66. Visser JW, Bol SJ, van den Engh G. Character-
ization and enrichment of murine hemopoietic
stem cells by fluorescence activated cell sorting.
Exp Hematol. 1981;9:644-655.
67. Baines P, Visser JW. Analysis and separation of
murine bone marrow stem cells by H33342 fluo-
rescence-activated cell sorting. Exp Hematol.
1983;11:701-708.
68. Spangrude GJ, Heimfeld S, Weissman IL. Purifi-
cation and characterization of mouse hematopoi-
etic stem cells. Science. 1988;241:58-62.
69. Osawa M, Hanada K, Hamada H, Nakauchi H.
Long-term lymphohematopoietic reconstitution by
a single CD34-low/negative hematopoietic stem
cell. Science. 1996;273:242-245.
70. Kiel MJ, Yilmaz OH, Iwashita T, Yilmaz OH,
Terhorst C, Morrison SJ. SLAM family receptors
distinguish hematopoietic stem and progenitor
cells and reveal endothelial niches for stem cells.
Cell. 2005;121:1109-1121.
71. Benveniste P, Cantin C, Hyam D, Iscove NN. He-
matopoietic stem cells engraft in mice with abso-
lute efficiency. Nat Immunol. 2003;4:708-713.
72. Kent DG, Dykstra BJ, Cheyne J, Ma E, Eaves CJ.
Steel factor coordinately regulates the molecular
signature and biologic function of hematopoietic
stem cells. Blood. 2008;112:560-567.
73. Dick JE. Normal and leukemic human stem cells
assayed in SCID mice. Semin Immunol. 1996;8:
197-206.
74. Mosier DE, Gulizia RJ, Baird SM, Wilson DB.
Transfer of a functional human immune system to
mice with severe combined immunodeficiency.
Nature. 1988;335:256-259.
75. McCune JM, Namikawa R, Kaneshima H, Shultz
LD, Lieberman M, Weissman IL. The SCID-hu
mouse: murine model for the analysis of human
hematolymphoid differentiation and function. Sci-
ence. 1988;241:1632-1639.
76. Fraser CC, Kaneshima H, Hansteen G, Kilpatrick
M, Hoffman R, Chen BP. Human allogeneic stem
cell maintenance and differentiation in a long-
term multilineage SCID-hu graft. Blood. 1995;86:
1680-1693.
77. Baum CM, Weissman IL, Tsukamoto AS, Buckle
A-S, Peault B. Isolation of a candidate human
hematopoietic stem-cell population. Proc Natl
Acad Sci U S A. 1992;89:2804-2808.
78. Kamel-Reid S, Dick JE. Engraftment of immune-
deficient mice with human hematopoietic stem
cells. Science. 1988;242:1706-1709.
79. Lapidot T, Pflumio F, Doedens M, Murdoch B,
Williams DE, Dick JE. Cytokine stimulation of
multilineage hematopoiesis from immature hu-
man cells engrafted in scid mice. Science. 1992;
255:1137-1141.
80. Vormoor J, Lapidot T, Pflumio F, et al. Immature
human cord blood progenitors engraft and prolif-
erate to high levels in severe combined immuno-
deficient mice. Blood. 1994;83:2489-2497.
81. Shultz L, Schweitzer P, Christianson S, et al. Mul-
tiple defects in innate and adaptive immunologi-
cal function in NOD/LtSz-scid mice. J Immunol.
1995;154:180-191.
82. Larochelle A, Vormoor J, Hanenberg H, et al.
Identification of primitive human hematopoietic
cells capable of repopulating NOD/SCID mouse
bone marrow: implications for gene therapy. Nat
Med. 1996;2:1329-1337.
83. Cashman JD, Lapidot T, Wang JC, et al. Kinetic
evidence of the regeneration of multilineage he-
matopoiesis from primitive cells in normal human
bone marrow transplanted into immunodeficient
mice. Blood. 1997;89:4307-4316.
84. Wang JC, Doedens M, Dick JE. Primitive human
hematopoietic cells are enriched in cord blood
compared with adult bone marrow or mobilized
peripheral blood as measured by the quantitative
in vivo SCID-repopulating cell assay. Blood.
1997;89:3919-3924.
85. Conneally E, Cashman J, Petzer A, Eaves C. Ex-
pansion in vitro of transplantable human cord
blood stem cells demonstrated using a quantita-
tive assay of their lympho-myeloid repopulating
activity in nonobese diabetic-scid/scid mice. Proc
Natl Acad Sci U S A. 1997;94:9836-9841.
86. Ito M, Hiramatsu H, Kobayashi K, et al. NOD/
SCID/gamma(c)(null) mouse: an excellent recipi-
ent mouse model for engraftment of human cells.
Blood. 2002;100:3175-3182.
87. Ishikawa F, Yasukawa M, Lyons B, et al. Develop-
ment of functional human blood and immune sys-
tems in NOD/SCID/IL2 receptor ␥ chain (null)
mice. Blood. 2005;106:1565-1573.
CANCER STEM CELLS 4805
88. Hogan CJ, Shpall EJ, Keller G. Differential long-
term and multilineage engraftment potential from
subfractions of human CD34⫹ cord blood cells
transplanted into NOD/SCID mice. Proc Natl
Acad Sci U S A. 2002;99:413-418.
89. Kerre TC, De Smet G, De Smedt M, et al.
Adapted NOD/SCID model supports develop-
ment of phenotypically and functionally mature T
cells from human umbilical cord blood CD34(⫹)
cells. Blood. 2002;99:1620-1626.
90. McKenzie JL, Gan OI, Doedens M, Dick JE. Hu-
man short-term repopulating stem cells are effi-
ciently detected following intrafemoral transplan-
tation into NOD/SCID recipients depleted of
CD122⫹ cells. Blood. 2005;106:1259-1261.
91. Takenaka K, Prasolava TK, Wang JC, et al. Poly-
morphism in Sirpa modulates engraftment of hu-
man hematopoietic stem cells. Nat Immunol.
2007;8:1313-1323.
92. Mazurier F, Doedens M, Gan OI, Dick JE. Rapid
myeloerythroid repopulation after intrafemoral
transplantation of NOD-SCID mice reveals a new
class of human stem cells. Nat Med. 2003;9:959-
963.
93. Yahata T, Ando K, Sato T, et al. A highly sensitive
strategy for SCID-repopulating cell assay by di-
rect injection of primitive human hematopoietic
cells into NOD/SCID mice bone marrow. Blood.
2003;101:2905-2913.
94. Wang J, Kimura T, Asada R, et al. SCID-repopu-
lating cell activity of human cord blood-derived
CD34-negative cells assured by intra-bone mar-
row injection. Blood. 2003;101:2924-2931.
95. Wang L, Menendez P, Shojaei F, et al. Generation
of hematopoietic repopulating cells from human
embryonic stem cells independent of ectopic
HOXB4 expression. J Exp Med. 2005;201:1603-
1614.
96. Dick JE, Lapidot T. Biology of normal and acute
myeloid leukemia stem cells. Int J Hematol. 2005;
82:389-396.
97. Wang JC, Dorrell C, Ito CY, et al. Normal and leu-
kemic human stem cells assayed in immune-defi-
cient mice. In: Zon LI, ed. Hematopoiesis: A De-
velopmental Approach. New York, NY: Oxford
University Press; 2001:99-118.
98. Pflumio F, Izac B, Katz A, Shultz LD, Vainchenker
W, Coulombel L. Phenotype and function of hu-
man hematopoietic cells engrafting immune-defi-
cient CB17-severe combined immunodeficiency
mice and nonobese diabetic-severe combined
immunodeficiency mice after transplantation of
human cord blood mononuclear cells. Blood.
1996;88:3731-3740.
99. Glimm H, Eisterer W, Lee K, et al. Previously un-
detected human hematopoietic cell populations
with short-term repopulating activity selectively
engraft NOD/SCID-beta2 microglobulin-null mice.
J Clin Invest. 2001;107:199-206.
100. Kerre TC, De Smet G, De Smedt M, et al. Both
CD34⫹38⫹ and CD34⫹38- cells home specifi-
cally to the bone marrow of NOD/LtSZ scid/scid
mice but show different kinetics in expansion.
J Immunol. 2001;167:3692-3698.
101. Guenechea G, Gan OI, Dorrell C, Dick JE. Dis-
tinct classes of human stem cells that differ in
proliferative and self-renewal potential. Nat Im-
munol. 2001;2:75-82.
102. Mazurier F, Gan O, McKenzie J, Doedens M, Dick
J. Lentivector-mediated clonal tracking reveals
intrinsic heterogeneity in the human hematopoi-
etic stem cell compartment and culture-induced
stem cell impairment. Blood. 2004;103:545-552.
103. Kamel-Reid S, Letarte M, Sirard C, et al. A model
of human acute lymphoblastic leukemia in im-
mune-deficient SCID mice. Science. 1989;246:
1597-1600.
104. Uckun FM, Sather H, Reaman G, et al. Leukemic
cell growth in SCID mice as a predictor of relapse
in high-risk B-lineage acute lymphoblastic leuke-
mia. Blood. 1995;85:873-878.
4806 DICK
105. Civin C, Strauss L, Brovall C, Fackler M,
Schwartz J, Shaper J. Antigenic analysis of he-
matopoiesis: III. A hematopoietic progenitor cell
surface antigen defined by a monoclonal antibody
raised against KG-1a cells. J Immunol. 1984;133:
157-165.
106. Terstappen LW, Safford M, Unterhalt M, et al.
Flow cytometric characterization of acute myeloid
leukemia: IV. Comparison to the differentiation
pathway of normal hematopoietic progenitor
cells. Leukemia. 1992;6:993-1000.
107. Lapidot T, Sirard C, Vormoor J, et al. A cell initiat-
ing human acute myeloid leukaemia after trans-
plantation into SCID mice. Nature. 1994;367:645-
648.
108. Bonnet D, Dick JE. Human acute myeloid leuke-
mia is organized as a hierarchy that originates
from a primitive hematopoietic cell. Nat Med.
1997;3:730-737.
109. Ailles LE, Gerhard B, Kawagoe H, Hogge DE.
Growth characteristics of acute myelogenous leu-
kemia progenitors that initiate malignant hemato-
poiesis in nonobese diabetic/severe combined
immunodeficient mice. Blood. 1999;94:1761-
1772.
110. Terpstra W, Prins A, Visser T, et al. Conditions for
engraftment of human acute myeloid leukemia
(AML) in SCID mice. Leukemia. 1995;9:1573-
1577.
111. Terpstra W, Prins A, Ploemacher RE, et al. Long-
term leukemia-initiating capacity of a CD34-sub-
population of acute myeloid leukemia. Blood.
1996;87:2187-2194.
112. Pearce DJ, Taussig D, Zibara K, et al. AML en-
graftment in the NOD/SCID assay reflects the
outcome of AML: implications for our understand-
ing of the heterogeneity of AML. Blood. 2006;107:
1166-1173.
113. van Rhenen A, Feller N, Kelder A, et al. High
stem cell frequency in acute myeloid leukemia at
diagnosis predicts high minimal residual disease
and poor survival. Clin Cancer Res. 2005;11:
6520-6527.
114. Fialkow PJ, Singer JW, Adamson JW, et al. Acute
nonlymphocytic leukemia: heterogeneity of stem
cell origin. Blood. 1981;57:1068-1073.
115. Fialkow P, Singer J, Raskind W, et al. Clonal de-
velopment, stem-cell differentiation, and clinical
remissions in acute nonlymphocytic leukemia.
N Engl J Med. 1987;317:468-473.
116. Fearon E, Burke P, Schiffer C, Zehnbauer B,
Vogelstein B. Differentiation of leukemia cells to
polymorphonuclear leukocytes in patients with
acute nonlymphoblastic leukemia. N Engl J Med.
1986;315:15-24.
117. Fialkow P, Denman A, Jacobson R, Lowenthal M.
Chronic myelogenous leukemia: origin of some
lymphocytes from leukemic stem cells. J Clin In-
vest. 1978;62:815-823.
118. Al-Hajj M, Wicha M, Morrison SJ, Clarke MF. Pro-
spective identification of tumorigenic breast can-
cer cells. Proc Natl Acad Sci U S A. 2003;100:
3983-3988.
119. Singh SK, Hawkins C, Clarke ID, et al. Identifica-
tion of human brain tumour initiating cells. Nature.
2004;432:396-401.
120. Clarke MF, Dick JE, Dirks PB, et al. Cancer stem
cells—perspectives on current status and future
directions: American Association for Cancer Re-
search Workshop on Cancer Stem Cells. Cancer
Res. 2006;66:9339-9344.
121. Jordan CT, Lemischka IR. Clonal and systemic
analysis of long-term hematopoiesis in the
mouse. Genes Dev. 1990;4:220-232.
122. Hope K, Jin L, Dick JE. Acute myeloid leukemia
originates from a hierarchy of leukemic stem cell
classes that differ in self-renewal capacity. Nat
Immunol. 2004;5:738-743.
123. Lessard J, Sauvageau G. Bmi-1 determines the
proliferative capacity of normal and leukemic
stem cells. Nature. 2003;423:455-460.
BLOOD, 15 DECEMBER 2008 䡠 VOLUME 112, NUMBER 13
124. Yilmaz OH, Valdez R, Theisen BK, et al. Pten de-
pendence distinguishes haematopoietic stem
cells from leukaemia-initiating cells. Nature.
2006;441:475-482.
125. Guan Y, Gerhard B, Hogge DE. Detection, isola-
tion, and stimulation of quiescent primitive leuke-
mic progenitor cells from patients with acute my-
eloid leukemia (AML). Blood. 2003;101:3142-
3149.
126. Guzman ML, Neering SJ, Upchurch D, et al.
Nuclear factor-kappaB is constitutively activated
in primitive human acute myelogenous leukemia
cells. Blood. 2001;98:2301-2307.
127. Terpstra W, Ploemacher RE, Prins A, et al. Flu-
orouracil selectively spares acute myeloid leuke-
mia cells with long-term growth abilities in immu-
nodeficient mice and in culture. Blood. 1996;88:
1944-1950.
128. So CW, Karsunky H, Passegue E, Cozzio A,
Weissman IL, Cleary ML. MLL-GAS7 transforms
multipotent hematopoietic progenitors and in-
duces mixed lineage leukemias in mice. Cancer
Cell. 2003;3:161-171.
129. Huntly BJ, Shigematsu H, Deguchi K, et al. MOZ-
TIF2, but not BCR-ABL, confers properties of leu-
kemic stem cells to committed murine hematopoi-
etic progenitors. Cancer Cell. 2004;6:587-596.
130. Krivtsov AV, Twomey D, Feng Z, et al. Transfor-
mation from committed progenitor to leukaemia
stem cell initiated by MLL-AF9. Nature. 2006;442:
818-822.
131. Cozzio A, Passegue E, Ayton PM, Karsunky H,
Cleary ML, Weissman IL. Similar MLL-associated
leukemias arising from self-renewing stem cells
and short-lived myeloid progenitors. Genes Dev.
2003;17:3029-3035.
132. Passegue E, Wagner EF, Weissman IL. JunB de-
ficiency leads to a myeloproliferative disorder
arising from hematopoietic stem cells. Cell. 2004;
119:431-443.
133. Chen W, Kumar AR, Hudson WA, et al. Malignant
transformation initiated by Mll-AF9: gene dosage
and critical target cells. Cancer Cell. 2008;13:
432-440.
134. Wang Y, Armstrong SA. Cancer: inappropriate
expression of stem cell programs? Cell Stem
Cell. 2008;2:297-299.
135. Miyamoto T, Weissman IL, Akashi K. AML1/ETO-
expressing nonleukemic stem cells in acute my-
elogenous leukemia with 8;21 chromosomal
translocation. Proc Natl Acad Sci U S A. 2000;97:
7521-7526.
136. Majeti R, Park C, Weissman I. Identification of a
hierarchy of multipotent hematopoietic progeni-
tors in human cord blood. Cell Stem Cell. 2007;1:
635-645.
137. Jamieson CH, Ailles LE, Dylla SJ, et al. Granulo-
cyte-macrophage progenitors as candidate leuke-
mic stem cells in blast-crisis CML. N Engl J Med.
2004;351:657-667.
138. Nowell P, Hungerford D. A minute chromosome in
human chronic granulocytic leukemia. Science.
1960;132:1497.
139. Rowley JD. A new consistent chromosomal ab-
normality in myelogenous leukemia identified by
quinacrine fluorescence and giemsa staining. Na-
ture. 1973;243:290-293.
140. Mani SA, Guo W, Liao MJ, et al. The epithelial-
mesenchymal transition generates cells with
properties of stem cells. Cell. 2008;133:704-715.
141. Darley RL, Hoy TG, Baines P, Padua RA, Burnett
AK. Mutant N-RAS induces erythroid lineage dys-
plasia in human CD34⫹ cells. J Exp Med. 1997;
185:1337-1347.
142. Pereira DS, Dorrell C, Ito CY, et al. Retroviral
transduction of TLS-ERG initiates a leukemo-
genic program in normal human hematopoietic
cells. Proc Natl Acad Sci U S A. 1998;95:8239-
8244.
143. Warner JK, Wang JC, Takenaka K, et al. Direct
evidence for cooperating genetic events in the
leukemic transformation of normal human hema-
topoietic cells. Leukemia. 2005;19:1794-1805.
144. Mulloy JC, Cammenga J, Berguido FJ, et al.
Maintaining the self-renewal and differentiation
potential of human CD34⫹ hematopoietic cells
using a single genetic element. Blood. 2003;102:
4369-4376.
145. Chalandon Y, Jiang X, Christ O, et al. BCR-ABL-
transduced human cord blood cells produce ab-
normal populations in immunodeficient mice.
Leukemia. 2005;19:442-448.
146. Kennedy JA, Barabé F. Investigating human leu-
kemogenesis from cell lines to in vivo models of
human leukemia. Leukemia. 2008 Aug 7 [Epub
ahead of print].
147. Barabé F, Kennedy JA, Hope KJ, Dick JE. Model-
ing the initiation and progression of human acute
leukemia in mice. Science. 2007;316:600-604.
148. Wei J, Wunderlich M, Fox C, et al. Microenviron-
ment determines lineage fate in a human model
of MLL-AF9 leukemia. Cancer Cell. 2008;13:483-
495.
149. Greaves MF, Wiemels J. Origins of chromosome
translocations in childhood leukaemia. Nat Rev
Cancer. 2003;3:639-649.
150. Hong D, Gupta R, Ancliff P, et al. Initiating and
cancer-propagating cells in TEL-AML1-associ-
ated childhood leukemia. Science. 2008;319:336-
339.
151. Kelly PN, Dakic A, Adams JM, Nutt SL, Strasser
A. Tumor growth need not be driven by rare can-
cer stem cells. Science. 2007;317:337.
152. Williams RT, den Besten W, Sherr CJ. Cytokine-
dependent imatinib resistance in mouse BCR-
ABL⫹, Arf-null lymphoblastic leukemia. Genes
Dev. 2007;21:2283-2287.
153. Somervaille TC, Cleary ML. Identification and
characterization of leukemia stem cells in murine
MLL-AF9 acute myeloid leukemia. Cancer Cell.
2006;10:257-268.
154. Kennedy JA, Barabé F, Poeppl AG, Wang JC,
Dick JE. Comment on “Tumor growth need not be
driven by rare cancer stem cells.” Science. 2007;
318:1722; author reply 1722.
155. Liu R, Wang X, Chen GY, et al. The prognostic
role of a gene signature from tumorigenic breast-
cancer cells. N Engl J Med. 2007;356:217-226.
156. Guzman ML, Rossi RM, Neelakantan S, et al. An
orally bioavailable parthenolide analog selectively
eradicates acute myelogenous leukemia stem
and progenitor cells. Blood. 2007;110:4427-4435.
157. Guzman ML, Li X, Corbett CA, et al. Rapid and
selective death of leukemia stem and progenitor
cells induced by the compound 4-benzyl,
2-methyl, 1,2,4-thiadiazolidine, 3,5 dione (TDZD-
8). Blood. 2007;110:4436-4444.
158. Hassane DC, Guzman ML, Corbett C, et al. Dis-
covery of agents that eradicate leukemia stem
cells using an in silico screen of public gene ex-
pression data. Blood. 2008;111:5654-5662.
159. Feuring-Buske M, Frankel A, Gerhard B, Hogge
D. Variable cytotoxicity of diphtheria toxin 388-
granulocyte-macrophage colony-stimulating fac-
tor fusion protein for acute myelogenous leuke-
mia stem cells. Exp Hematol. 2000;28:1390-
1400.
160. Terpstra W, Rozemuller H, Breems DA, et al.
Diphtheria toxin fused to granulocyte-macro-
phage colony-stimulating factor eliminates acute
myeloid leukemia cells with the potential to initiate
leukemia in immunodeficient mice, but spares
normal hemopoietic stem cells. Blood. 1997;90:
3735-3742.
161. Frankel A, Liu JS, Rizzieri D, Hogge D. Phase I
clinical study of diphtheria toxin-interleukin 3 fu-
sion protein in patients with acute myeloid leuke-
mia and myelodysplasia. Leuk Lymphoma. 2008;
49:543-553.
162. Bonnet D, Warren EH, Greenberg PD, Dick JE,
BLOOD, 15 DECEMBER 2008 䡠 VOLUME 112, NUMBER 13 CANCER STEM CELLS 4807

Riddell SR. CD8(⫹) minor histocompatibility anti-
gen-specific cytotoxic T lymphocyte clones elimi-
nate human acute myeloid leukemia stem cells.
Proc Natl Acad Sci U S A. 1999;96:8639-8644.
163. Rosinski KV, Fujii N, Mito JK, et al. DDX3Y en-
codes a class I MHC-restricted H-Y antigen that
is expressed in leukemic stem cells. Blood. 2008;
111:4817-4826.
164. Jin L, Hope KJ, Zhai Q, Smadja-Joffe F, Dick JE.
Targeting of CD44 eradicates human acute my-
eloid leukemic stem cells. Nat Med. 2006;12:
1167-1174.
165. Tavor S, Petit I, Porozov S, et al. CXCR4 regu-
lates migration and development of human acute
myelogenous leukemia stem cells in transplanted
NOD/SCID mice. Cancer Res. 2004;64:2817-
2824.
166. Papayannopoulou T, Scadden DT. Stem-cell ecol-
ogy and stem cells in motion. Blood. 2008;111:
3923-3930.
167. Ishikawa F, Yoshida S, Saito Y, et al. Chemo-
therapy-resistant human AML stem cells home to
and engraft within the bone-marrow endosteal
region. Nat Biotechnol. 2007;25:1315-1321.
168. Jordan CT, Upchurch D, Szilvassy SJ, et al. The
interleukin-3 receptor alpha chain is a unique
marker for human acute myelogenous leukemia
stem cells. Leukemia. 2000;14:1777-1784.
169. Lock R, Jin L, Lee E, et al. CD123 (IL-3 receptor
chain) neutralization by a monoclonal antibody
selectively eliminates human acute myeloid leu-
kemic stem cells. Blood. 2007;110:161.
170. Ito K, Bernardi R, Morotti A, et al. PML targeting
eradicates quiescent leukaemia-initiating cells.
Nature. 2008;453:1072-1078.
171. Wang JC. Evaluating therapeutic efficacy against
cancer stem cells: new challenges posed by a
new paradigm. Cell Stem Cell. 2007;1:497-501.

My introduction to hematology began during my postdoctoral fellowship with Alan Bernstein at

the storied Ontario Cancer Institute. This was in the mid-1980s during the early days of retrovirus-
mediated gene transfer and the breathless prediction that gene therapy was imminent and would
be the cure of many human diseases. Prior work from this group had already developed retroviral
vectors and provided the first evidence of transduction of primary murine clonogenic progeni-
tors; my project was to develop the means to transduce the hematopoietic stem cell. Although
there was intense competition to be the first group to show hematopoietic stem cell transduction,
the 2 most important lessons I learned in this period were: always look back to the past to guide
the directions of the future so as not to “reinvent the wheel,” and the drive to do research should
come from the intrinsic value and importance of the question being asked rather than from look-
ing over your shoulder to stay ahead of an imagined or real competitor. The project was a close
collaboration with Cristina Magli, a postdoctoral fellow in Dr Bob Phillips’ laboratory, who shared
a wealth of stem cell experience with me. Moreover, Bob was a leading member of the hematopoi-
esis and immunology research program at the Ontario Cancer Institute and had carried out a se-
ries of pioneering studies in stem cell biology, and clonal tracking of repopulating cells in par-
ticular. In response to some “new” idea or experimental design that Cristina and I would come up
with, our weekly laboratory meetings were often punctuated by critical observations from Bob
and Alan, including “well, in 1960-1970 Till and McCulloch, Metcalf, Dexter, we, etc, carried out
such-and-such experiments.” This prevented us from entering many a blind alley and deeply in-
stilled the principle that clonal long-term repopulation assays are the only ones that mattered.
Indeed, one could argue that my whole career has been spent repeating my postdoctoral project
over and over again, albeit in human cells. So the lesson for new trainees is that a judicious
choice of project and mentor can have benefits that last a lifetime!
Because this review already describes the pathway of discovery of our work once I established
John E. Dick my own laboratory, I will only make 2 additional points. First, time and place are essential. As the
review notes, I was fortunate to establish my laboratory at a time when key technologies were
sufficiently developed to enable our work. I was also fortunate to be surrounded by a remarkable group of colleagues in Toronto who set high standards
of excellence to aspire to but who were also committed to creating an unusually collaborative and collegial environment with few internal competitions.
Second, trainees drive research. I have been privileged to work with a wonderful group of trainees who drove all the work mentioned in the review. It was
their commitment to the projects and willingness to engage in a selfless fashion to the ferment of ideas from which research directions emerged and
evolved.