, 1670 (2009);
et al.Jeffrey M. Rosen,
The Increasing Complexity of the Cancer Stem Cell
www.sciencemag.org (this information is current as of June 26, 2009 ):
The following resources related to this article are available online at
version of this article at:
including high-resolution figures, can be found in the online
Updated information and services,
, 12 of which can be accessed for free:
cites 37 articles
This article appears in the following
in whole or in part can be found at:
permission to reproduce
of this article or about obtaining
Information about obtaining
registered trademark of AAAS.
is aScience2009 by the American Association for the Advancement of Science; all rights reserved. The title
CopyrightAmerican Association for the Advancement of Science, 1200 New York Avenue NW, Washington, DC 20005.
(print ISSN 0036-8075; online ISSN 1095-9203) is published weekly, except the last week in December, by the Science
on June 26, 2009
The Increasing Complexity of the
Cancer Stem Cell Paradigm
Jeffrey M. Rosen1and Craig T. Jordan2
The investigation and study of cancer stem cells (CSCs) have received enormous attention over the
past 5 to 10 years but remain topics of considerable controversy. Opinions about the validity
of the CSC hypothesis, the biological properties of CSCs, and the relevance of CSCs to cancer
therapy differ widely. In the following commentary, we discuss the nature of the debate, the
parameters by which CSCs can or cannot be defined, and the identification of new potential
therapeutic targets elucidated by considering cancer as a problem in stem cell biology.
originates from a primitive hematopoietic cell, as
shown in Fig. 1 (1, 2). This report became the
similar model existed for solid tumors with can-
cer stem cells (CSCs) at the top of a hierarchical
pyramid (3). These studies were based on a sim-
bodies directed at defined cell-surface markers
followed by limiting dilution transplantation, usu-
ally into an orthotopic site in immunocompro-
mised mice (the xenograft model). Thus, the CSC
cancer cells to initiate tumorigenesis by under-
going self-renewal and -differentiation, like nor-
mal stem cells, whereas the remaining majority of
the cells are more “differentiated” and lack these
Why Is There a Debate?
that CSCs arise as an intrinsic property of tumor
biology and development (Fig. 2). However, the
adipocytes, and endothelial cells, as well as the
extracellular matrix) and the immune system are
known to play important roles in cancer progres-
model in the context of a xenograft is the lack of
an appropriate microenvironment because of dif-
ferences between the mouse and human and the
lack of an intact immune system when evaluating
the tumor-initiating capacity of these human can-
of cells that appeared nontumorigenic might
actually be tumorigenic in the presence of the ap-
propriate microenviroment. In other words, tumor
n 1994, John Dick and colleagues published
leukemia is organized as a hierarchy that
cells might be functionally homogeneous, with
heterogeneous potential arising as a consequence
of extrinsic cues or the lack thereof (Fig. 2).
Strasser and colleagues attempted to test the origi-
nal CSC hypothesis by using an alternative ap-
proach to the xenograft system. They used two
was used to express either the c-myc or N-ras on-
cogenes to induce B or Tcell lymphomas, respec-
tively (6). Upon the analysis of transplants, these
authors concluded that “tumor growth need not
be driven by rare cancer stem cells” based upon
>10% of the transplanted cells, giving rise to
tumors in syngeneic mice. Alternatively, recent
studies by Guo et al. used a mouse model in
which deletion of the Pten tumor suppressor
gene in hematopoietic stem cells resulted in a
myeloproliferative disorder followed by acute T-
lymphoblastic leukemia (7). Using this model of
a human leukemia, these investigators demon-
strated by limiting dilution transplantation that a
rare population of leukemia stem cells (LSCs)
Center, University of Rochester School of Medicine, Rochester,
NY 14642, USA.
Normal blood cells
Fig. 1. Initialstudies in leukemia provided the paradigm for the generalCSCmodel. As shown on the left
the leukemia stem cell (LSC). The LSC retains some degree of developmental potential, generating the
leukemia progenitor and leukemic blast cells, which differ in their biological properties from the parent
LSC. As in normal hematopoiesis, the stem cell maintains the ability to undergo self-renewal and thereby
perpetuate the leukemia population.
26 JUNE 2009 VOL 324
on June 26, 2009
together, these findings indicate thattherelative
frequency and role of malignant stem cells can
vary considerably as a function of the specific
experimental system as well as perhaps the
particular oncogenes or tumor suppressor genes
used. However, for at least some forms of leu-
kemia, stem cell properties reside within a rela-
tively rare (<1%) subset of the tumor population,
a finding that is consistent in both the xenograft
and syngeneic models (8, 9).
What Is the Situation in Solid Tumors?
In 2003, Michael Clarke and colleagues adapted
to breast cancer the methods (10) used in leu-
kemias and demonstrated the existence of a sub-
frombreast cancerpleuraleffusions by limitingdi-
lution transplantation of CD44+/CD24–/lo/lineage–
cells into the mammary fat pad of immunocom-
promised nonobese diabetic–severe combined
immunodeficiency (NOD-SCID) mice. A year
tification of human brain tumor–initiating cells
again with FACS analysis but using a different
and orthotopic intracranial transplantation into
of similar studies in other solid tumors [reviewed
FACS and xenotransplantation of viable sin-
gle cells from solid tumors require modifications
of the approaches used for hematologic cancers
single cells, which are larger and more fragile
than the hematopoietic cells usually isolated by
these methods. Such procedures require several
manipulations that may affect cell viability or
in limiting dilution experiments can be markedly
influenced by these parameters as well as the du-
ration of the experiment performed to assess
tumorigenicity. In addition, the nature of the
role in the efficiency of tumor formation. This is
illustrated by the recent studies on human mela-
noma cells in which modified xenotransplanta-
tion conditions, including the use of more highly
immunocompromised NOD/SCID interleukin-2
receptor g chain null mice, increased the detec-
tion of tumorigenic melanoma cells by several
orders of magnitude (12). Approximately one in
four of the unselected melanoma cells formed
tumors in these studies, which argues that the
CSC population is not necessarily always rare
and leads investigators to question, “Have we
highlight that the relative frequency of CSCs may
vary as a function of both the tumor type and the
specific experimental system used.
Once again, the use of syngeneic mouse mod-
els for analysis of CSCs has helped clarify the
role of the microenvironment for solid tumors.
As an example, two studies using mouse breast
cancer models, the p53 null transplantation and
the MMTV-Wnt1 models, have demonstrated by
limiting dilution transplantation analysis that a
small subpopulation of tumor cells are CSCs,
whereas the bulk of the tumor cells are non-
tumorigenic (13, 14). The secondary tumors that
arise from isolated CSCs phenocopy the original
tumors and give rise to the heterogeneous cell
cell heterogeneity is not observed in all geneti-
cally engineered mouse models of breast
cancer, so it is probable that in some models
to the observations made by Strasser and his
colleagues using the Em lymphoma models.
Studies using the same p53 null mouse breast
cancer model and improved FACS protocols
found tumors formed with a more than 10-fold
higher frequency than previously reported (14),
thus illustrating that the absolute CSC frequen-
cies are highly dependent on the experimental
conditions used. Although the calculated fre-
quencies of CSCs may therefore vary depending
on the methods used to isolate and transplant
the CSCs and bulk tumor population was quite
similar. Most importantly, >90% of the bulk
tumor cells were nontumorigenic even after
more than a year after transplantation. Thus, like
the findings from hematologic cancers, the data
from solid-tumor analyses demonstrate a high
degree of variability depending on the specific
experimental system. However, despite concerns
noted for the use of xenografts, data from syn-
geneic models demonstrate the existence of sub-
populations of cells in at least some solid tumors,
satisfying the functional criteria of CSCs.
Perhaps the most challenging issue facing the
field is the fact that CSCs in primary tumors do
not always display the properties classically used
to define normal stem cells, cells with the ability
to self-renew and -differentiate into multiple cell
types. The cell-surface immunophenotype of pri-
mary tumors, as well as the frequency of func-
tionally defined CSCs, can vary dramatically
among different patients. In some cases, CSCs
are relatively rare, whereas in others CSCs can
constitute a substantial proportion of the tumor
mass (12). So, why are tumor CSCs so variable?
In normal steady-state systems, one can expect
reasonably well-conserved development behav-
ior, but upon any kind of substantial genetic or
epigenetic perturbation, the rules that define cell
and tissue behavior are not easily predicted and
must be defined empirically. Thus, in the con-
text of inherently unstable conditions such as
cancer, the fact that CSCs display varying be-
haviors is no surprise. Moreover, the properties of
CSCs appear to be influenced by both the spe-
cific genetic aberrancies in a given tumor as well
as the stage of disease progression and the types
of drugs used to challenge tumor growth (Fig. 3).
Consequently, for any particular type of cancer
the patient-to-patient variability of CSCs may be
quite substantial. Taken together, these issues
make any consistent definition of CSC proper-
ties difficult and suggest that being overly rigid
in how CSCs are defined is not realistic. Further-
more, the variability in CSC properties introduces
problems when developing new therapies.
A genetic program that might account in part
for the diversity of abundance of CSCs in solid
tumors is the ability of cells to undergo an
epithelial-to-mesenchymal transition (EMT).
Recent studies have suggested that induction of
cells results in cells with stem-like properties,
such as “ an increased ability to form mammo-
spheres, a property associated with mammary
epithelial stem cells” (15). A number of studies
Fig. 2. CSC models. (A) The intrinsic model suggests that specific subpopulations within a tumor (pink
cells) possess the functional properties of CSCs. (B) The extrinsic model proposes that all tumor cells are
functionally equivalent and display heterogeneous behaviors as a function of extrinsic (microenviron-
VOL 32426 JUNE 2009
on June 26, 2009
have suggested that cells at the leading invasive
edge of solid cancers, such as colon, breast, and
pancreatic tumors, exhibit more mesenchymal
features and are characterized by the expression
of CSC markers (16–18). Thus, there has been a
convergence of the well-established concept that
EMT is associated with tumor progression with
the more recent hypothesis of migratory cancer
stem cells (19, 20). So how might this process be
regulated in CSCs? Recent studies have impli-
cated transcription factors, such as Zeb1 and -2
and Twist, which are known inducers of EMT
(21), and a negative feedback pathway involving
transforming growth factor–b (TGFb) (16, 22).
So, changes in the surrounding microenviron-
ment that influence expression of TGFb family
members as well as other cytokines expressed by
mesenchymal stem cells or other cells in the
microenvironment may influence both EMTand
the reverse process of mesenchymal-to-epithelial
transition, which is most likely critical for
metastasis and colonization at distant sites (23).
If true, then the CSC state may be transitory in
certain circumstances, with tumor cells acquiring
more of a stem-like phenotype upon stimulation
with the appropriate environmental cues.
Studies of putative tumor stem cells have
also served to highlight the potential clinical
importance of the relationship between EMT
and CSCs. For example, using the same sub-
population of tumorigenic breast cancer cells
identified in (10), we reported the intrinsic re-
sistance of these cells to chemotherapy studied
in paired breast cancer biopsies (24). We also
identified a TGFb-like tumorigenic gene signa-
ture in a molecular subtype of human breast
tumors characterized by expression of many
mesenchymal-associated genes. Tumors resistant
to conventional treatments were enriched for cells
bearing this signature, and increased mesenchy-
mal markers were observed in the posttreatment
specimens. These data support a growing body of
evidence for a mesenchymal-like phenotype in
breast and possibly other solid tumors that may
be responsible for invasion, metastases, and even
To understandthe relationshipbetweenCSCs
pared the transcriptional programs in embryonic
stem cells (ESCs) with adult tissue stem cells and
human cancers (21). The ESC-like transcriptional
program shown to be activated in diverse human
epithelial cancers was a predictor of metastasis
and death. The oncogene c-Myc, but not other
oncogenes, appeared to be sufficient to activate
this ESC-like program and could increase the
fraction of CSCs. These authors concluded that
“Activation of an ESC-like transcriptional pro-
gram in differentiated adult cells may induce
pathologic self-renewal characteristic of cancer
stem cells.” Further, Weinberg and colleagues
have reported that a subset of ESC-associated
transcriptional regulators are more frequently
overexpressed in poorly differentiated tumors
(25). These authors concluded “that these genes
many tumors.” Such data imply that some poor-
New Directions Resultingfrom the Study ofCSCs
The controversy about CSCs has had the un-
expected benefit of stimulating research in areas
that previously were not the focus of cancer ther-
apeutics. Pathways known to be important for
stem-cell self-renewal, such as the Wnt, Notch,
tified as relevant to cancer as well as develop-
mental biology), are now of increased interest
because of their potential role in CSCs. For exam-
ple, the first clinical trial using gamma-secretase
inhibitors to block the Notch pathway in combi-
nation with chemotherapy in the neoadjuvant
setting for breast cancer has recently been ini-
tiated. Another example is the recent report that
Hhsignaling isessential for maintenance of CSCs
in myeloid leukemia (26).
populations of cells within tumors with stem-like
properties, which could be isolated using a va-
to identify the specific signaling pathways in
these cells as compared with the bulk tumor cells
and their normal counterparts. Studies have sug-
gested there might be selective effects of inhib-
iting the Pten/Akt and nuclear factor kB (NF-kB)
atopoietic stem cells (27–29). These studies have
been extended to solid tumors with the observa-
tion that brain (and possibly breast) CSCs may
be preferentially sensitive to Akt inhibitors (30).
Studies of the mechanisms by which parthenolide,
an NF-kB pathway inhibitor, might induce apo-
ptosis in LSCs have suggested a role for increased
reactive oxygen species (ROS) (31). In fact,
recent studies have suggested an association of
ROS levels and radioresistance in CSCs (32).
Variability in the response of CSCs to conven-
tional chemotherapy and radiation therapy have
led to investigations of the differences in cell-
cycle checkpoint and DNA-repair pathways in
CSCs versus the bulk of tumor cells (14, 33, 34).
An exciting new approach is the use of chemical
genetic screens of drug libraries against CSCs
with the possibility of discovering drugs to target
CSCs that may already be approved for alter-
native clinical use (35). Another interesting ap-
proach will be to use synthetic lethal screens to
search for new agents that may sensitize resistant
tumor-cell subpopulations to conventional chemo-
and radiation therapies (36). Finally, pharma and
Normal cell type
properties due to
De novo mutations
Initial events leading to
Additional genetic and epigenetic
changes due to chemotherapy
and disease progression
Fig. 3. StagesofCSCevolution.Formanytumortypes,thedenovomutationsleadingtoprimaryCSCare
varied. Thus, one would expect that the biology of primary CSCs may also be heterogeneous. Properties
such as CSC frequency, cell-surface phenotype, and drug sensitivity may vary as a function of the specific
progression may occur either as a consequence of intrinsic tumor pathogenesis and/or challenge with
chemotherapy. Selective pressures associated with neoplastic progression may lead to a higher frequency
of functionally defined CSC in secondary or metastatic stages as well inter-patient and intra-patient
variability of CSC properties.
26 JUNE 2009VOL 324
on June 26, 2009
biotech companies now have active programs to
Whether the CSC model is relevant to all can-
to target tumor cells that are resistant to current
therapies and give rise to recurrence and treat-
ment failure. Notwithstanding our ability to
sequence the cancer genome(s) and to create
personalized targeted therapies, it is apparent
that combination therapies, which target the
CSC subpopulation as well as the bulk of the
tumor cells, will be required to effectively man-
age cancer treatment (37). One pressing need is
the development of improved preclinical mod-
els to test these therapies because determining
the appropriate doses and combinations as well
as the order of addition of these agents will be
critical for success in the clinic. The fact that
these concepts are steadily making their way
into the clinic is exciting and suggests that the
recent interest in CSCs may yield beneficial
outcomes in potentially unexpected ways.
References and Notes
1. T. Lapidot et al., Nature 367, 645 (1994).
2. J. E. Dick, Blood 112, 4793 (2008).
3. J. E. Visvader, G. J. Lindeman, Nat. Rev. Cancer 8, 755
4. M. J. Bissell, M. A. Labarge, Cancer Cell 7, 17 (2005).
5. A. Mantovani, Nature 457, 36 (2009).
6. P. N. Kelly, A. Dakic, J. M. Adams, S. L. Nutt, A. Strasser,
Science 317, 337 (2007).
7. W. Guo et al., Nature 453, 529 (2008).
8. D. Bonnet, J. E. Dick, Nat. Med. 3, 730 (1997).
9. S. J. Neering et al., Blood 110, 2578 (2007).
10. M. Al-Hajj, M. S. Wicha, A. Benito-Hernandez, S. J. Morrison,
M. F. Clarke, Proc. Natl. Acad. Sci. U.S.A. 100, 3983
11. S. K. Singh et al., Nature 432, 396 (2004).
12. E. Quintana et al., Nature 456, 593 (2008).
13. R. W. Cho, M. F. Clarke, Curr. Opin. Genet. Dev. 18, 48
14. M. Zhang et al., Cancer Res. 68, 4674 (2008).
15. S. A. Mani et al., Cell 133, 704 (2008).
16. U. Burk et al., EMBO Rep. 9, 582 (2008).
17. C. Ginestier et al., Cell Stem Cell 1, 555 (2007).
18. P. C. Hermann et al., Cell Stem Cell 1, 313 (2007).
19. J. P. Thiery, Nat. Rev. Cancer 2, 442 (2002).
20. T. Brabletz, A. Jung, S. Spaderna, F. Hlubek, T. Kirchner,
Nat. Rev. Cancer 5, 744 (2005).
21. D. J. Wong et al., Cell Stem Cell 2, 333 (2008).
22. P. A. Gregory, C. P. Bracken, A. G. Bert, G. J. Goodall,
Cell Cycle 7, 3112 (2008).
23. A. E. Karnoub et al., Nature 449, 557 (2007).
24. X. Li et al., J. Natl. Cancer Inst. 100, 672 (2008).
25. I. Ben-Porath et al., Nat. Genet. 40, 499 (2008).
26. C. Zhao et al., Nature 458, 776 (2009).
27. Q. Xu, S. E. Simpson, T. J. Scialla, A. Bagg, M. Carroll,
Blood 102, 972 (2003).
28. M. L. Guzman et al., Blood 98, 2301 (2001).
29. O. H. Yilmaz et al., Nature 441, 475 (2006).
30. C. E. Eyler et al., Stem Cells 26, 3027 (2008).
31. M. L. Guzman et al., Blood 105, 4163 (2005).
32. M. Diehn et al., Nature 458, 780 (2009).
33. S. Bao et al., Nature 444, 756 (2006).
34. W. A. Woodward et al., Proc. Natl. Acad. Sci. U.S.A. 104,
35. P. Diamandis et al., Nat. Chem. Biol. 3, 268 (2007).
36. A. W. Whitehurst et al., Nature 446, 815 (2007).
37. W. A. Woodward, M. S. Chen, F. Behbod, J. M. Rosen,
J. Cell Sci. 118, 3585 (2005).
38. The authors thank W. Woodward, M. Zhang, Y. Li,
M. Guzman, and J. Dick for their helpful comments and
apologize to investigators for the failure to cite many
other important contributions in this field because of
space limitations. These studies were supported by NIH
grants R37-CA16303 and R01-CA122206.
Growth Factors, Matrices, and Forces
Combine and Control Stem Cells
Dennis E. Discher,1David J. Mooney,2Peter W. Zandstra3
Stem cell fate is influenced by a number of factors and interactions that require robust control
for safe and effective regeneration of functional tissue. Coordinated interactions with soluble
factors, other cells, and extracellular matrices define a local biochemical and mechanical niche
with complex and dynamic regulation that stem cells sense. Decellularized tissue matrices and
synthetic polymer niches are being used in the clinic, and they are also beginning to clarify
fundamental aspects of how stem cells contribute to homeostasis and repair, for example, at
sites of fibrosis. Multifaceted technologies are increasingly required to produce and interrogate
cells ex vivo, to build predictive models, and, ultimately, to enhance stem cell integration in vivo
for therapeutic benefit.
lenging. In studies of animal models and humans
where stem cell engraftment has been quantified
after injection, only a few percent of cells remain
after several days or weeks [e.g. (1)]. Many clin-
ical trials are nonetheless under way, particularly
ontrol over stem cell trafficking, survival,
proliferation, and differentiation within a
with adult bone marrow–derived mesenchymal
stem cells (MSCs), which are being investigated
as treatments for diseases of nonhematopoietic
tissues—primarily, myocardial infarction and pe-
ripheral ischemia (2). Although U.S. Food and
Drug Administration approval for human testing
of cells differentiated from embryonic stem cells
(ESC) is a recent landmark for the field (3), two
widely reported clinical cases highlight some of
the technical opportunities and challenges with
stem cells in soft tissue repair. One patient in
Spain was successfully transplanted with a re-
engineered trachea in 2008; donor trachea was
first decellularized by using a detergent (without
denaturing the collagenous matrix), and then,
this scaffold was recellularized in a rotating bio-
reactor using MSC-derived cartilage-like cells (4).
Long-term safety and efficacy will be important
to monitor and understand. Indeed, in a second
case, the cerebellum of a boy with ataxia telangi-
brain tumor of stem cell origin was found (5).
Upon implantation, stem cells and their derived
lineages encounter a multitude of cues that can
influence cell fate. Efforts to parse the molecular
mechanisms for translation from bench to clinic
will increasingly benefit from a wide range of new
and established technologies. Here we briefly re-
view salient features of microenvironments, me-
chanics, and material systems that are being
pursued to control stem cells for both basic
insight and application.
Niche Interactions and in Vitro Designs
The niche is the in vivo microenvironment that
regulates stem cell survival, self-renewal, and dif-
ferentiation. Key niche components and interac-
tions include growth factors, cell-cell contacts,
and cell-matrix adhesions (Fig. 1A). The interplay
of these niche factors is particularly important to
comprehend if any desired stem cell response is
to be made robust for therapy, i.e., resistant to the
many types of perturbations encountered by cells
delivered in vivo.
Growth factors added to culture or secreted
by stem cells and nearby niche cells are often
potent in their effects on cell fate, and so, in em-
bryonic development, growth factors are tightly
regulated in space and time (6). In culture, one
means of controlling niche interactions in two
dimensions is with micropatterns of extracellular
matrix (ECM) islands, which limit diffusion of
secreted growth factors within and between islands
and also limit the modulating effects of matrix and
cell contacts. With human ESCs (hESCs), for ex-
1Biophysical Engineering and Nanobiopolymers Laboratory,
School of Engineering and Applied Science, University of
Pennsylvania, Philadelphia, PA 19104, USA.E-mail: discher@
seas.upenn.edu2Laboratory for Cell and Tissue Engineering,
School of Engineering and Applied Sciences, Harvard Univer-
sity, Cambridge, MA 02138, USA. E-mail: mooneyd@seas.
harvard.edu3Stem Cell Bioengineering Laboratory, Institute
of Biomaterials and Biomedical Engineering, Centre for Bio-
medical and Biomolecular Research, University of Toronto,
Toronto, ON M5S 3E1, Canada. E-mail: peter.zandstra@
VOL 32426 JUNE 2009
on June 26, 2009