REVIEW doi:10.

1038/nature09781

Cells of origin in cancer

Jane E. Visvader1,2

Both solid tumours and leukaemias show considerable histological and functional heterogeneity. It is widely accepted
that genetic lesions have a major role in determining tumour phenotype, but evidence is also accumulating that cancers
of distinct subtypes within an organ may derive from different cells of origin. These cells acquire the first genetic hit or
hits that culminate in the initiation of cancer. The identification of these crucial target cell populations may allow earlier
detection of malignancies and better prediction of tumour behaviour, and ultimately may lead to preventive therapies
for individuals at high risk of developing cancer.

R
emarkably, the oldest description and surgical treatment of cancer Tumour heterogeneity
dates back to 1600 bc in Egypt, where the papyrus described eight Phenotypic and functional heterogeneity are hallmarks of cancers aris-
cases of tumours occurring on the breast and their treatment by ing in several organs. Variability can occur between tumours arising in
cauterization. Today, cancer is a leading cause of death worldwide, with the same organ (intertumoral heterogeneity), leading to the classifica-
the number of deaths from cancer projected to increase markedly, owing tion of discrete tumour subtypes. These subtypes are typically charac-
partly to an ageing global population. This immense cancer burden terized by their molecular profile, together with their morphology and
demands strategies that permit earlier detection, better stratification of expression of specific markers (such as hormone and growth-factor
tumours to guide therapy and the development of effective preventive receptors). Variation also occurs within individual tumours (intra-
therapies. Targeted therapeutic strategies that suppress the progression tumoral heterogeneity), in which the tumour cells often have a range
of preneoplastic cells towards the malignant state hold great promise of functional properties and a diverse expression of markers. For exam-
for circumventing the huge challenges associated with the treatment of ple, the proportion of cells that express the oestrogen receptor within
late-stage disease. a patients breast tumour can vary extensively, from 1% to 100%. The
Tumours show marked heterogeneity in their cellular morphology, CSC and clonal-evolution models have been put forward to account
proliferative index, genetic lesions and therapeutic response. The for intratumoral heterogeneity and intrinsic differences in tumour-
molecular and cellular mechanisms underpinning tumour hetero- regenerating capacity (reviewed in refs 1 and 2). Interestingly, despite
geneity remain central questions in the cancer biology field. Key the heterogeneous nature of tumours, the histopathology and gene-
issues include whether the different subtypes of cancer reflect a dis- expression profiles of tumours arising in patients often remain relatively
tinct cell of origin, the extent to which the genetic mutational profile stable during progression from localized disease to metastatic and even
contributes to tumour phenotype and the nature of the relationship end-stage disease3,4.
between the cell of origin and the cancer stem cell. This Review Two main mechanisms have been conceptualized to explain
focuses on the strategies used to identify cells of origin, the impact intertumoral heterogeneity: different genetic or epigenetic muta-
of these cells on cancer cell fate and behaviour, and the implications tions occurring within the same target cell result in different tumour
for the development of improved prognostic markers and preven- phenotypes (Fig. 2a), and different tumour subtypes arise from
tive therapies. distinct cells within the tissue that serve as the cell of origin (Fig. 2b).
It is important to note that these cellular and molecular mechanisms
Cell-of-origin and cancer stem-cell concepts are distinct are not mutually exclusive, but can act together to determine tumour
It is important to note that the cell of origin, the normal cell that histopathology and behaviour. In addition, extrinsic mechanisms may
acquires the first cancer-promoting mutation(s), is not necessar- be involved in generating tumour heterogeneity, because interactions
ily related to the cancer stem cell (CSC), the cellular subset within between tumour cells and the stromal micro-environment are a crucial
the tumour that uniquely sustains malignant growth. That is, the determinant of malignant growth5. Several studies on human cancers
cell-of-origin and CSC concepts refer to cancer-initiating cells and and mouse models have highlighted the importance of specific genetic
cancer-propagating cells, respectively (Fig. 1). Although the tumour- aberrations in contributing to tumour behaviour. Many oncogenes
initiating cell and the CSC have been used interchangeably, the and tumour-suppressor proteins, most prominently phosphatidyl-
tumour-initiating cell more aptly denotes the cell of origin. There inositol-3-OH kinase (PI(3)K), MYC, RAS, p53, PTEN, p16Ink4a and
is considerable evidence that several diverse cancers, both leukae- retinoblastoma protein (RB), are frequent culprits in diverse cancers,
mias and solid tumours, are hierarchically organized and sustained although the overall mutational profiles of different cancer types can
by a subpopulation of self-renewing cells that can generate the full vary considerably. Tumour maintenance undoubtedly depends on the
repertoire of tumour cells (both tumorigenic and non-tumorigenic continued expression of certain oncogenes a phenomenon known
cells)1. The cell of origin, the nature of the mutations acquired, and/ as oncogene addiction6. Lineage-dependency oncogenes that have key
or the differentiation potential of the cancer cells are likely to deter- survival roles, in which genetic changes may be predetermined by the
mine whether a cancer follows a CSC model. In most instances, the lineage programs inherent in the tumour precursor cell7, are also likely
phenotype of the cell of origin may differ substantially from that of to contribute. There is mounting evidence, however, that the nature of
the CSC. the cellular target has an important influence on tumour cell fate and
1
Stem Cells and Cancer Division, The Walter and Eliza Hall Institute of Medical Research, Parkville, Victoria 3052, Australia. 2Department of Medical Biology, The University of Melbourne, Parkville,
Victoria 3010, Australia.

3 1 4 | N AT U R E | VO L 4 6 9 | 2 0 JA N UA Ry 2 0 1 1
2011 Macmillan Publishers Limited. All rights reserved
 Review insight

No tumour

Stem cell Oncogenic Second Tumour

hit(s) hit

Common progenitor Cell of origin CSC CSC

Expansion Progression
CSC

Committed progenitor

Mature cells

Figure 1 | The cell of origin and evolution of a cancer stem cell. Normal of further epigenetic mutations by a cell within the aberrant population (in this
cellular hierarchy comprising stem cells that progressively generate common case expanded) during neoplastic progression may result in the emergence of
and more restricted progenitor cells, yielding all the mature cell types that a CSC. In this model, only the CSCs (and not other tumour cells) are capable
constitute a particular tissue. Although the cell of origin for a particular tumour of sustaining tumorigenesis. Thus, the cell of origin, in which tumorigenesis is
could be an early precursor cell such as a common progenitor, the accumulation initiated, may be distinct from the CSC, which propagates the tumour.

pathology. Indeed, activation of the same oncogenic pathway in differ- the hierarchy provides an important framework for understanding cells
ent cellular compartments or contexts may profoundly influence malig- of origin in cancer, if tumour cells show phenotypic plasticity or ded-
nant potential8. For example, transgenic mouse models have shown ifferentiate during neoplastic progression, then lineage markers and
that mutant Hras targeted to the hair follicle region highly predisposed molecular signatures of tumour cells may not precisely reflect the true
mice to squamous carcinomas, whereas its targeting to more differ- cell of origin in normal tissue.
entiated interfollicular or suprabasal cells resulted in papillomas with
low malignant potential9,10. Moreover, transformation of distinct breast Strategies to investigate the cellular origins of cancers
epithelial cells in vitro has indicated that the target cell is an important Genetically engineered mouse models have proven indispensable
determinant of tumour phenotype11. in addressing the cellular origin of cancers (Fig. 3). Two primary
Understanding the normal cellular hierarchy within a given tissue approaches have been used to tackle this question: one, transgenic
is an important prerequisite to identifying the cells of origin of can- or conditionally targeted gene technologies to explore the effects
cers. Organ development proceeds in a hierarchical manner from stem of oncogenes and tumour suppressors in different cellular con-
cells to committed progenitor cells, which in turn yield differentiated texts; and, two, genetic alteration of cells ex vivo before evaluating
cells that constitute the bulk of the tissue or organ (Fig. 1). The most their tumorigenic capacity in mice. The first approach requires
primitive cells, stem cells, have been favoured candidates for targets cell-specific promoters that direct expression of an oncogene, or
of transformation because of their inherent capacity for self-renewal Cre-mediated deletion of a tumour-suppressor gene, in a specific
and their longevity, which would allow the sequential accumulation of subset of cells in vivo (Fig. 3a). Ideally, such studies should use at
genetic or epigenetic mutations required for oncogenesis. Nevertheless, least two promoters with different cell-type specificity to reveal the
any cell in the hierarchy with proliferative capacity could serve as a cell tumorigenic susceptibility of distinct cell subpopulations within
of origin in cancer, if it acquires mutations that re-instigate self-renewal that tissue. In this model, targeting of only one cell subpopula-
capacity and prevent differentiation to a post-mitotic state. tion is expected to reveal tumours that recapitulate the phenotype
The normal lineage hierarchy can serve as a framework to probe of the human cancer being modelled. Although this approach has
potential targets of carcinogenesis by comparison of lineage markers been increasingly used to study cells of origin, particularly in brain
expressed on the surface of normal and neoplastic cell subsets. More tumours, it is often hampered by a lack of established cell-lineage-
accurate correlations, however, depend on comparisons of the expres- specific promoters, given that unique markers of stem and progeni-
sion signatures of normal cell populations with those of the different tor cells do not exist for the overwhelming majority of organs and
tumour subtypes arising within that organ. Notably, histologically indis- tissues.
tinguishable glial-cell tumours from different parts of the central nerv- A further refinement of this in vivo targeting approach involves
ous system have distinct molecular gene signatures and chromosomal lineage tracing of cells as they undergo transformation. In this system,
abnormalities, suggesting that they originated in different subpopula- the main oncogenic event is activated conditionally in a limited number
tions of site-restricted progenitor cells12,13. In a recent integrated genom- of cells rather than simultaneously in all cells that express the promoter.
ics approach to studying tumour heterogeneity, the transcriptomes of For example, a tamoxifen-inducible Cre recombinaseoestrogen recep-
human brain tumours were matched to those of mouse neural stem cells tor fusion protein (CreER)18,19 driven by a cell-type-specific promoter
(NSCs) from different cellular compartments within the central nervous allows inducible gene expression, in which the dose and number of
system. Embryonic cerebral NSCs and adult spinal NSCs were revealed pulses can be fine-tuned to ensure single-cell tracking. Lineage tracing
as the potential cells of origin for supratentorial and spinal ependymo- at the clonal level is the current gold standard for delineating the target
mas, respectively14. In breast cancer, the different molecular subtypes15,16 cell of transformation in mouse models (Table 1).
have also been linked to normal epithelial subpopulations by the inter- In the second approach, defined cell subpopulations are genetically
rogation of specific gene-expression signatures17. Such observations manipulated ex vivo and subsequently transplanted orthotopically into
remain correlative, however, until the tumorigenic potential of specific mice to assess their predisposition to tumour initiation (Fig. 3b). The
cells is proven in vivo by clonality or lineage-tracing studies. Although strategy is applicable to cells from both human and mouse tissues, and

2 0 JA N UA Ry 2 0 1 1 | VO L 4 6 9 | N AT U R E | 3 1 5
2011 Macmillan Publishers Limited. All rights reserved
 insight Review

a Genetic mutation model Tumour Cells of origin in haematopoietic malignancies

In different leukaemias, both normal stem and committed progeni-
tor cells have been implicated as cellular targets of transformation. In
Subtype x
chronic myeloid leukaemia (CML) one of the first disorders to be
Oncogenic defined by a dominant genetic mutation the long-term haemat-
event A
opoietic stem cell (HSC) containing the BCRABL mutation has been
Oncogenic
event B Subtype y established as the cell of origin by in vivo clonality studies in humans22.
Oncogenic
Although the HSC maintains the chronic phase of the disease, analysis
event C of samples from patients in blast crisis the acute and advanced stage
of disease has indicated that subsequent genetic events occurring in
Subtype z downstream progenitor cells give rise to leukaemia stem cells, high-
lighting the dynamic state of the tumorigenesis process23. The cells of
origin for acute leukaemias, including myeloid, lymphoid and mixed-
lineage, have not been definitively established. Human acute myeloid
leukaemia (AML) may originate within the primitive haematopoietic
cell compartment, on the basis of the similar cell-surface phenotypes of
the leukaemia-initiating cell and the HSC, as well as lentivirus-mediated
clonal-tracking studies24. A primitive human haematopoietic cell may
also be the primary target of MLL fusion genes25,26. Moreover, in vivo
evidence has implicated a human HSC-like cell as the initiating cell in
a case of childhood leukaemia arising in utero27.
b Cell-of-origin model Tumour Several studies have addressed potential cells of origin in mouse leu-
kaemia models by transducing primary haematopoietic cell populations
Oncogenic with oncogenes before transplantation, but these have yielded variable
Subtype x
event A results. For mouse models of CML, only BCRABL targeted to HSCs, but
not to committed progenitor cells, induced myeloproliferative disease28,
consistent with findings for human CML. Interestingly, the MLLGAS7
Oncogenic
Subtype y fusion protein produced mixed lymphoid leukaemia when transduced
event A
into HSCs or multipotential progenitor cells but not when introduced
into lineage-restricted progenitors. However, the MOZTIF2 (ref. 28),
MLLAF9 (ref. 29) and MLLENL30,31 fusion proteins all initiated AML
Subtype z
irrespective of the cell subtype transduced. Although HSCs generally
Oncogenic
event A
appeared more susceptible to transformation than committed progeni-
tors, a self-renewal program seemed to be reactivated in the latter cells
during leukaemogenesis. In a knock-in mouse model of MLLAF9,
only HSCs that expressed high levels of the fusion product and not the
granulocytemacrophage progenitors were transformed, but the latter
could be efficiently transformed by a higher dose of MLLAF9 after
retroviral transduction32. Thus, oncogene dosage affects transformation
susceptibility, emphasizing the importance of using models that permit
oncogene expression at levels relevant to human disease.
Figure 2 | Two models of intertumoral heterogeneity. a, In the genetic (and
Further evidence that cancer can be initiated in cells other than stem
epigenetic) mutation model, mutations primarily determine the phenotype
of the tumour, such that different mutations result in different tumour cells has emerged from cell-fate mapping studies in transgenic mice
morphology. b, In the cell-of-origin model, different cell populations in the overexpressing Lmo2: preleukaemic T-cell progenitors that had acquired
lineage hierarchy serve as cells of origin for the different cancer subtypes self-renewal potential were identified as the cell of origin for T-cell acute
arising within that organ or tissue. lymphoblastic leukaemia (T-ALL) in this model33. Pertinently, mice
lacking three pathways commonly repressed in cancer (p53, p16Ink4a
relies on the reproducible sorting of functionally defined populations and p19Arf ) contain cells that phenotypically resemble haematopoietic
that can serve as targets for the introduction of relevant oncogenic multipotential progenitor cells but have long-term reconstituting ability,
lesions. This approach has been widely exploited to identify potential indicating that they have acquired self-renewal capacity34.
cells of origin in human leukaemias, many of which contain character-
istic chromosomal translocations. Cells of origin in solid tumours
An alternative way of exploring early cellular changes that occur Evidence is increasing that either stem or progenitor cells can act as tar-
before the onset of overt disease is to dissect the cellular components gets for tumour initiation in a range of solid tumours (Table 1). Lineage-
of preneoplastic tissue from individuals in families at high risk of tracing studies (shown schematically in Fig. 4a) have identified probable
cancer. These include carriers of germline mutations in the adenom- cells of origin of intestinal, prostate and basal cell carcinomas, as well as
atous polyposis coli (APC) gene, hereditary non-polyposis colorectal pancreatic ductal adenocarcinoma, in mouse genetic models. Several
cancer (HNPCC) genes (such as MSH2 and MLH1), and BRCA1 other reports have used cell-specific promoters to drive Cre-mediated
or BRCA2 genes. Carriers of mutated APC and HNPCC genes are expression of the oncogenic event(s) in different cellular compartments
predisposed to developing colorectal cancer20; female BRCA1- or of the mouse, whereas genetic manipulation of discrete cellular subsets
BRCA2-mutation carriers are prone to breast and ovarian cancer21; has provided valuable insight into cell types prone to the initiation of
and male BRCA2-mutation carriers often develop prostate cancer. carcinogenesis. It is crucial to note that although many studies have
This strategy has proven insightful in the case of BRCA1-mutation clearly identified the lineage in which the cancer originates, the precise
carriers (see Histopathology does not necessarily reflect cell of ori- cell type in the hierarchy (the cell of origin) in which transformation
gin). Combined with transplantation and clonality studies, cell sub- occurs remains elusive in most cases. Nevertheless, in mouse mod-
sets predisposed to neoplastic progression can thus be identified. els of intestinal and prostate tumours, it seems clear that the cancers

3 1 6 | N AT U R E | VO L 4 6 9 | 2 0 JA N UA Ry 2 0 1 1
2011 Macmillan Publishers Limited. All rights reserved
 Review insight

a Genetically modified mice

A Oncogene B Oncogene
or IoxP IoxP or IoxP IoxP
A Cre x Suppressor B Cre x Suppressor

Glioblastoma Medulloblastoma

b Transduction of cellular subsets

Subset 1 Subset 1 Transduced subset 1

Sort Sort GFP+

Lentiviral cells
Remove cells transduction Tumour
in vitro
Subset 2
Single-cell
suspension
Subset 2 Transduced subset 2

Sort GFP+
cells
No tumour

Figure 3 | Strategies used to identify cells of origin in cancer. a, A and B are active in different cell subpopulations within the brain. b,
Genetic mouse models can be used to either activate an oncogene or Potential cells of origin in cancer can be addressed through the sorting
inactivate a tumour-suppressor gene in a discrete subpopulation of cells of defined subpopulations from human or mouse tissue and their genetic
using cell-type-specific promoters. Comparative mouse models in which manipulation ex vivo. Cell subpopulations are first transduced with genes
different promoters (A or B) drive expression of the same oncogenic encoding the oncogenic lesion(s), together with a fluorescent marker, then
lesion (either oncogene activation or tumour-suppressor inactivation) in transplanted orthotopically into immunocompromised mice to evaluate
the brain. Initiation of a glioblastoma is observed in the case of promoter the tumorigenic potency of the different subpopulations. GFP, green
A and a medulloblastoma in the case of promoter B, where promoters fluorescent protein.

originate in a bona fide stem cell that is capable of self-renewal and evidence argues for stem or multipotential neural progenitors in the
multilineage differentiation. Although the lineage in which the can- subventricular zone (SVZ) as the primary cellular target for gliob-
cer originates has been revealed for skin, pancreatic, brain and breast lastoma development, the cell of origin remains elusive owing to the
tumours, the precise cells of origin have yet to be defined. complexity of this zone43. Many studies on brain tumorigenesis have
Two distinct crypt stem cells have been identified as the cells of origin used the nestin (Nes) or Gfap promoter regions to direct expression
for intestinal cancers. The vast majority of colorectal cancer is caused or inactivation: it is important to note that although both promoters
by mutations in the WNT signalling pathway, including loss of the drive expression in neural precursor cells, the Gfap promoter is also
negative regulator APC and activating mutations in -catenin, both active in mature astrocytes41. Nonetheless, nestin-positive precursors
of which result in constitutive WNT activation35. Apc deletion in long- were more susceptible to transformation by RAS and AKT than the
lived stem cells (LGR5+) but not in short-lived transit-amplifying cells GFAP-positive population, and produced high-grade glioblastomas,
(using AhCre) revealed stem cells as a cell of origin for intestinal cancer consistent with tumours originating in the stem/progenitor popula-
in mice (Fig. 4b)36. This target cell is also marked by CD133 (also known tion40. Moreover, neural precursor cells in the SVZ of the adult brain
as PROM1 or prominin 1)37. A novel intestinal stem cell located in the efficiently initiated glioblastomas after conditional inactivation of p53
+4 or +5 position from the base of the crypt and therefore distinct from (also known as Trp53), Pten and/or Nf1 tumour-suppressor genes41,
the LGR5+ stem cell was also shown to be susceptible to tumorigenesis and presymptomatic mice exhibited a premalignant cell population. By
by deregulated WNT signalling using a BMI1CreER knock-in model38. contrast, the more differentiated cell types in non-neurogenic areas of
LGR5+ stem cells are likely to be the target population for WNT-driven the adult brain proved less susceptible to malignant transformation41,42.
tumorigenesis in the stomach, where these cells seeded small adeno- Similarly, mice deficient in varying combinations of p53, Pten and/
mas39. The intestinal tumour load, however, precluded further lineage or Rb (also known as pRb and Rb1) developed tumours only in the
tracing of stem cells during the development of stomach cancer. SVZ and not from mature peripheral astrocytes44. Interestingly, the
The cell of origin for brain cancers has been investigated using sev- same stem/progenitor population seemed to initiate either gliomas
eral mouse genetic models that have differed in design and the nature or medulloblastomas, depending on the nature of the genetic lesions.
of the initiating oncogenic lesions. The models have predominantly More restricted progenitor cells may also initiate glioma development.
included conditionally targeted mice, the RCASTVA system in Single-cell tracking studies of cells expressing mutant p53 implicate
which gene transfer is mediated by an oncogene-carrying RCAS ret- transit-amplifying cells in the SVZ45. Although oligodendrocyte pro-
rovirus to somatic cells in TVA transgenic mice40 and cell-culture- genitors and cells within the astrocyte compartment may also have the
based analyses. Stereotactic injection of viruses into different areas potential to seed glioblastomas46,47, culturing cells before manipulation
of the brain41,42 has also been used for the introduction of oncogenes may not accurately reflect the in vivo situation, and the presence of
or Cre recombinase to mimic focal tumorigenesis, but this approach more primitive cells within the cell cultures cannot be excluded. Thus,
cannot be used to identify cells of origin of cancer unequivocally definitive evidence that mature astrocytes can serve as cells of origin
as the transduced cell types are unknown. Although the available for brain tumours awaits further experimentation.

2 0 JA N UA Ry 2 0 1 1 | VO L 4 6 9 | N AT U R E | 3 1 7
2011 Macmillan Publishers Limited. All rights reserved
 insight Review

Table 1 | Cells of origin (proven and candidate) identified in solid tumours by targeting distinct cellular subsets
Tumour type Genetic model PromoterCre construct Lineage Cell of origin
tracing

Mouse models
Brain: Glioblastoma RAS, AKT activation (RCASTVA system: NA Neural progenitor cell40
nestin, Gfap promoters)
p16Ink4a/p19Arf, BMI1 inactivation; NA Neural progenitor and astrocyte46,47
mutant EGFR
p53, NF1 and/or PTEN inactivation NestinCreERT2, AdenoCre Multipotent progenitor41
PDGFB activation (RCASTVA system) NA Oligodendrocyte progenitor85
RAS, AKT activation; p53 inactivation GFAPCre Multipotent progenitor42
Mutant p53 expression GFAPCre Neural progenitor or transit-amplifying cell45
PTEN, p53 inactivation GFAPCre Multipotent progenitor44
Medulloblastoma Patched inactivation MATH1Cre, GFAPCre Multipotent progenitor and granule neuron
progenitor49
Smoothened activation GFAPCre, MATH1Cre, OLIG2 Multipotent progenitor and granule neuron
TVACre progenitor48
RB, p53, PTEN inactivation GFAPCre Multipotent progenitor44
RB, p53 inactivation AdenoCre Neural progenitor cell50
-catenin mutant, p53 inactivation BLBPCre, ATOH1Cre Dorsal brainstem progenitor51
Ependymoma p16Ink4a/p19Arf inactivation; EPHB2 NA Embryonic cerebral stem/progenitor cell14
(supratentorial) activation
Intestine APC inactivation AhCre, LGR5CreERT2 + Stem cell36
Mutant -catenin CD133CreERT2 + Stem cell37
Mutant -catenin BMI1CreER + Stem cell38
Lung Kras activation AdenoCre Bronchioalveolar stem cell77
Mammary NOTCH1 activation in cell subsets NA Luminal progenitor65
BRCA1, p53 inactivation BLGCre, K14Cre Luminal progenitor63
Pancreas Kras activation, inflammation RIPCreER + Endocrine cell69
Prostate PTEN inactivation NKX3.1CreERT2 + Luminal stem cell54
ERG1, PI(3)K and/or AR expression NA Basal progenitor59
PTEN inactivation PBCre Basal progenitor58
PTEN inactivation PSACre Luminal cell56,57
Skin/basal cell Smoothened activation K14CreER + Interfollicular epidermal progenitor72
carcinoma
Stomach APC inactivation LGR5CreERT2 + Stem cell39
Human tissue
Breast (basal-like Preneoplastic BRCA1+/ cell subsets NA Luminal progenitor17
subtype)*
Prostate PI(3)K, ERG, AR into cell subsets NA Basal progenitor60
Adeno, adenoviral; Ah, cytochrome P450 1A1 gene (also known as Cyp1a1); AR, androgen receptor; BLG, -lactoglobulin; K14, cytokeratin 14; NA, not applicable;
PB, probasin (prostate-specific); PSA, prostate-specific antigen; RIP, rat insulin promoter.
*Refers to analysis of specific subsets from normal versus premalignant human breast tissue, leading to identification of a candidate cell of origin.

Unipotent cells within the mouse brain have been identified as the distinct cellular origins. In mouse models of malignant peripheral
cell of origin for medulloblastoma. Constitutive hedgehog signalling nerve sheath tumours, tumours may initiate from differentiated glial
(due to loss of Ptc or activation of Smo) in either the stem-cell com- cells in the adult brain52,53.
partment or granule neuron progenitor cells could initiate medullob-
lastomas but not astrocytomas or oligodendrogliomas48,49. Targeted Prostate cancer can originate from distinct cell lineages
deletion of a different set of genes (p53 and Rb) also supports the Prostate cancers have been widely presumed to originate from mature
notion that cerebellar stem cells or lineage-restricted granule progeni- luminal cells as these cancers are characterized by an expansion of
tor cells can give rise to medulloblastomas50. Extending these observa- luminal cells and the absence of basal cells. Recent findings, however,
tions further, it was shown that the cells must transition to the granule implicate distinct stem cells in the basal and the luminal cellular com-
progenitor stage for the initiation of medulloblastomas, indicating partments, each of which can be targeted for oncogenesis by the loss
that the true cell of origin for medulloblastomas that exhibit hedgehog of PTEN or PI(3)K activation. Rare luminal epithelial stem cells that
pathway activation48 is a unipotent progenitor cell48. A distinct cell type express NKX3.1 and are castration resistant were identified54, and
within the dorsal brainstem has recently emerged as the cell of origin these cells could initiate high-grade prostate intraepithelial neoplasia
for medulloblastomas that harbour activating mutations in the WNT (PIN) and carcinomas. It is not yet clear whether these cells, termed
pathway51, indicating that different subtypes of medulloblastoma have CARNs, exist within normal mouse or human prostate (that is, the

3 1 8 | N AT U R E | VO L 4 6 9 | 2 0 JA N UA Ry 2 0 1 1
2011 Macmillan Publishers Limited. All rights reserved
 Review insight

a Normal crypt Preneoplastic lesion Tumour Figure 4 | Identification of crypt stem cells as
the cell of origin in intestinal cancer by lineage
tracing. a, Schematic depiction of lineage tracing in
a colonic crypt, in which a single dose of tamoxifen
can be used to activate CreER specifically in a stem
cell to drive expression of a given oncogenic lesion in
this cell population. The promoter that drives CreER
expression will determine the cells in which this
Tagged occurs. In the case of a stem cell, the incorporated
stem cell reporter gene, such as lacZ, will mark all progeny
Single dose of the stem cell. b, Schematic representation
of tamoxifen summarizing the data from ref. 36, in which lineage
tracing of either stem or transit-amplifying cells
deficient in Apc shows the initiation of intestinal
b tumours from stem cells only.

Absorptive Goblet cell Oncogenic No tumour

epithelial cell hit in transit-
amplifying cell

Enteroendocrine cell

Transit-amplifying cell

+4 cell
Paneth cell
Oncogenic
hit in stem cell

Stem cell

Tumour-initiating cell Tumour

non-castrated state), but these bipotent, self-renewing cells may be in precancerous breast tissue from BRCA1-mutation carriers dem-
mobilized as facultative stem cells during prostate regeneration after onstrated expansion of luminal progenitor cells that showed altered
androgen withdrawal. Indeed, only the basal population isolated from growth properties and aberrantly produced nodules when transplanted
normal mouse prostate has been demonstrated to contain stem cells into mice (ref. 17 and F. Vaillant and J.E.V., unpublished data). Moreo-
with prostate-regenerative potential55. The question of whether CARNs ver, there are significant similarities between the gene-expression pro-
exist in patients with prostate cancer is difficult to address as castration files of normal breast luminal progenitors, preneoplastic tissue from
is implemented only after the development of advanced disease. The BRCA1-mutation carriers and basal-like breast cancers17. Indeed, inac-
identification of a luminal cancer-initiating cell is consistent with find- tivation of Brca1 (and p53) in either luminal or basal cells of the mouse
ings that deletion of Pten in luminal cells of the mouse prostate leads to mammary gland showed that only the luminal cell population initiated
prostatic hyperplasia56,57. basal-like cancers reminiscent of those arising in BRCA1-mutation
Conversely, basal cells have been demonstrated as an efficient target carriers63. The presence of ALDH1+ lobules in pathologically normal
of tumorigenesis in a Pten-deficient mouse model58 or when genetically tissue from BRCA1-mutation carriers64 is compatible with a luminal
manipulated ex vivo to overexpress Erg, androgen receptor (Ar) and/ cell of origin, because luminal progenitors exhibit ALDH activity (F.
or PI(3)K, resulting in PIN lesions and carcinomas59. The tumorigenic Vaillant and J.E.V., unpublished data). NOTCH1 activation also targets
susceptibility of purified basal and luminal subpopulations from human luminal progenitor cells, generating an aberrant, self-renewing progen-
prostate tissue was recently evaluated 60. When the cells were transduced itor cell that yields mammary hyperplasia and, eventually, tumours65.
with relevant oncogenic lesions, together with a fluorescent marker, and Accordingly, high NOTCH1 levels occur in basal-like breast cancers
transplanted into immunocompromised mice, only the basal cells could and predict poor prognosis66. The cells of origin for most other breast
initiate the development of prostate cancer reminiscent of the luminal- cancers have yet to be defined. In particular, the role of the mammary
like cancers that arise in humans. stem cell in breast oncogenesis is unclear, although WNT pathway
activation primarily targets this population67, and the claudin-low
Histopathology does not necessarily reflect cell of origin subtype of breast cancer, which is characterized by low expression of
Tumours have largely been classified on the basis of their histological genes involved in tight junctions and cellcell adhesion, shares a similar
appearance and expression of markers (such as ER and HER2 in breast molecular profile to that of the stem-cell subset17,68.
cancer) that predict the response of the tumour to a given treatment. Pancreatic ductal adenocarcinoma (PDAC) and premalignant duc-
However, the histological and cell-surface marker profiles of tumours tal lesions (termed pancreatic intraepithelial neoplasia) have a ductal
do not necessarily predict the cell of origin, as illustrated above for pros- morphology, suggesting that they develop from pancreatic duct cells69.
tate cancer. Other examples that underscore this point include breast, Unexpectedly, however, premalignant lesions were shown to derive from
pancreatic and basal cell carcinomas, as discussed below. differentiated acinar cells that were reprogrammed to a duct-like pheno-
Individuals that harbour mutations in the BRCA1 tumour-suppres- type6971. Moreover, targeting a Kras oncogenic signal to insulin-positive
sor gene develop breast cancers that usually resemble the basal-like endocrine cells induced PDAC. Notably, the ductal reprogramming of
subtype, typically associated with poor clinical prognosis61,62. The acinar cells required inflammatory tissue damage, highlighting a role
basal stem cell has therefore been presumed to be the transforma- for non-genetic factors in contributing to tumour phenotype. Ductal
tion target for this tumour subtype, but the luminal progenitor has adenocarcinoma can also arise from other pancreatic cell lineages in the
instead emerged as the likely cell of origin. Analysis of cellular subsets absence of tissue injury, for example PDX1-expressing cells69.

2 0 JA N UA Ry 2 0 1 1 | VO L 4 6 9 | N AT U R E | 3 1 9
2011 Macmillan Publishers Limited. All rights reserved
 insight Review

Lineage-tracing studies have shown that basal cell carcinomas origi- these patients. In other families at high risk of colorectal cancer, with
nate in progenitor cells resident in the interfollicular epidermis of the mutations in the MLH1 or MSH2 mismatch repair genes, inhibition
skin rather than from stem cells as originally postulated72. Conditional (short-term and intermittent) of selective DNA polymerases may be a
activation of hedgehog signalling in different cellular compartments, potential chemopreventive strategy, as these agents have been shown
combined with cell-fate mapping, showed that long-lived progenitors to elicit tumour cell death in patients with HNPCC80.
in the interfollicular epithelium, but not the hair follicle bulge stem cell Tamoxifen and aromatase inhibitors (inhibitors of oestrogen action
or the transit-amplifying cells, produced tumours. The block in differ- and biosynthesis, respectively) are the prototypes for chemopreven-
entiation evident in interfollicular epidermal cell clones with constitu- tive agents in hormone-receptor-positive breast cancer, as they mark-
tive hedgehog signalling correlated with the expression of basal lineage edly reduce the rate of disease recurrence and more than halve the
markers (such as P-cadherin and keratins 7 and 15) and may have led incidence of new cancers in patients81. Recent findings have clari-
to the notion that basal cell carcinomas arise from hair follicle bulge fied how ovarian hormone exposure enhances breast cancer risk by
stem cells8,73. showing that mammary stem cells, despite lacking receptors for these
hormones, are highly responsive to steroid hormone signalling in
Potential relationships between cells of origin and CSCs vivo82. As the paracrine signals relayed to these stem cells seem to
Although a stem cell may sustain the first oncogenic hit, subsequent involve the receptor activator of NF-B (RANK) signalling pathway,
alterations required for the genesis of a CSC can occur in descendent an exciting corollary of these findings is that it should be possible
cells (Fig. 1). This is exemplified by CML, in which the HSC is the cell to prevent some forms of breast cancer by driving stem cells into a
of origin in the more indolent phase of the disease but in patients with dormant state for example, by blockade of the RANK pathway
CML blast crisis, granulocytemacrophage progenitors acquire self- for which inhibitors are already in clinical trial for bone metastases.
renewal capacity through a -catenin mutation and emerge as the prob- There may also be prophylactic benefit for BRCA1/2-mutation car-
able CSC23. riers in the use of poly(ADP-ribose) polymerase (PARP) inhibitors83,
In some instances, particularly in early-stage cancers, the CSC may which are being evaluated for the treatment of BRCA1/2-associated
closely resemble the cell of origin, although this remains to be proven. breast cancers84, if the early lesions in these individuals prove to be
For example, the leukaemia-initiating cell in AML74 may prove to be defective in DNA repair.
the same as the leukaemia stem cell that propagates the disease. In a
mouse model of intestinal cancer, despite all neoplastic cells arising from Perspective
CD133+ stem cells, only a small fraction of the tumour cells retained It seems intuitive that both the cell of origin and the pattern of
CD133 expression. It is tempting to speculate a hierarchical model of acquired mutations determine tumour fate and phenotype. The close
tumour progression in which this small subset of CD133+ cells might association between cell lineage and cancer phenotype suggests that
generate the full repertoire of tumour cells and thereby correspond lineage-restricted mechanisms that normally operate during develop-
to CSCs. This notion is compatible with the observation that CD133 ment may contribute to tumorigenesis. The cell of origin may often
marks CSCs in certain human colorectal tumours75,76. Nevertheless, it correspond to the normal tissue stem cell, exploiting its intrinsic self-
remains to be determined whether these CD133+ or LGR5+ cells have renewal ability. This may particularly apply to tissues with very high
tumour-propagating ability. Bronchioalveolar stem cells (BASCs) have turnover, such as the gut, because progenitor cells may not live long
been implicated as the cell of origin for lung adenocarcinomas induced enough to acquire the full set of mutations required for malignancy.
by mutant Kras77 in mice and may be closely related to the CSC, because The stem cell or an early progenitor cell has also emerged as a likely
the BASC marker Sca1 was recently shown to identify CSCs in certain cell of origin in certain leukaemias, glioblastomas and prostate cancer.
mouse models of non-small-cell lung cancer78. In prostate cancer, if the In other malignancies, however, the initiating cell can be a restricted
oncogenic transformation of CARNs leads to the formation of CSCs in progenitor, as in the case of medulloblastomas, basal cell carcinomas
prostate cancer, then this might explain how early events occurring in and BRCA1-associated breast cancer. Indeed, in cell types that retain
the cell of origin can contribute to the emergence of hormone-refractory high proliferative potential, such as some differentiated lymphoid cells,
disease54. Although the relationships between tumour cells of origin and the cell of origin could even be a mature cell type. Notably, there are
CSCs are not well understood, comprehensive cellular analyses of the several examples indicating that tumour phenotype may not directly
preneoplastic and neoplastic states of different tumour subtypes should reflect tumour histology or lineage marker expression, thus highlight-
eventually shed light on this issue. ing the requirement for in vivo studies to assess the propensity of cell
populations to act as cells of origin.
Therapeutic and diagnostic implications Mouse models of oncogenesis have been pivotal in uncovering the
Identification of the cell of origin has important implications for new cellular origins of cancer and the impact of specific mutations on tumor-
preventive therapeutic approaches to suppress or reverse the initial igenesis. Arguably, the choice of the genetically modified mouse model
phase of disease. Cancer chemoprevention will be most applicable to and the promoter/enhancer to recapitulate the effects of the oncogenic
individuals within families at high risk of cancer such as BRCA1/2- lesion has a major influence on tumour phenotype and behaviour. More
mutation carriers. Cell-surface markers or proto-oncogenic kinases specific promoters to drive expression of an initiating event within a
such as c-KIT17 that show altered expression in cell subsets in preneo- definitive cellular compartment are likely to evolve as the normal lineage
plastic tissue can be evaluated as prognostic markers, and for their abil- hierarchies within tissues are further refined. For studies on the cell of
ity to eradicate or modulate aberrant cells in either preneoplastic or origin in human tissues, genetic and cellular analyses of tumour cell
established disease. populations, at the single-cell level, from patients at different stages of
In principle, individuals that carry a defect in the APC gene, and disease should provide substantial insight into the relationships among
are thus highly susceptible to colorectal cancer, could benefit from normal cells, cells of origin and CSCs.
prophylactic treatment that targets APC-deficient cells for apoptosis. Identification of the cell of origin may permit a more systematic
Tumour-necrosis-factor-related apoptosis-inducing ligand (TRAIL) in analysis of the genetic lesions involved in tumour initiation and pro-
combination with all-trans retinoic acid selectively induced apoptosis in gression, and serve as a platform for the identification of early disease
APC-deficient premalignant cells and intestinal polyps, thus inhibiting biomarkers. It may also have important implications for preventing
tumour growth79. Furthermore, treatment of biopsy samples of human relapse, particularly in cases in which relapse results from a prema-
colonic polyps from patients with familial adenomatous polyposis lignant clone (perhaps the cell of origin itself) that persists in the
showed selective apoptosis of polyps, whereas normal tissue was unaf- patient before acquiring a mutation that renders it malignant. If so,
fected, providing a potentially effective method of chemoprevention in even patients with cancer who have a profound regression may require

3 2 0 | N AT U R E | VO L 4 6 9 | 2 0 JA N UA Ry 2 0 1 1
2011 Macmillan Publishers Limited. All rights reserved
 Review insight

maintenance therapy to reduce the chance of relapse. Finally, the gene translocations can induce lineage reassignment in tumorigenesis. EMBO J. 24,
31363146 (2005).
signature of the cell of origin may elucidate key molecular pathways 32. Chen, W. et al. Malignant transformation initiated by Mll-AF9: gene dosage and
and driver mutations that could lead to new therapeutic approaches to critical target cells. Cancer Cell 13, 432440 (2008).
prevent or target early-stage disease. This study highlights the importance of gene dosage when assessing the
effects of oncogene expression in candidate cells of origin.
1. Visvader, J. E. & Lindeman, G. J. Cancer stem cells in solid tumours: 33. McCormack, M. P. et al. The Lmo2 oncogene initiates leukemia in mice by
accumulating evidence and unresolved questions. Nature Rev. Cancer 8, inducing thymocyte self-renewal. Science 327, 879883 (2010).
755768 (2008). 34. Akala, O. O. et al. Long-term haematopoietic reconstitution by Trp53/ p16Ink4a/
2. Marusyk, A. & Polyak, K. Tumor heterogeneity: causes and consequences. p19Arf/ multipotent progenitors. Nature 453, 228232 (2008).
Biochim. Biophys. Acta 1805, 105117 (2010). 35. Clevers, H. Wnt/-catenin signaling in development and disease. Cell 127,
3. Ma, X. J. et al. Gene expression profiles of human breast cancer progression. 469480 (2006).
Proc. Natl Acad. Sci. USA 100, 59745979 (2003). 36. Barker, N. et al. Crypt stem cells as the cells-of-origin of intestinal cancer. Nature
4. Weigelt, B. et al. Gene expression profiles of primary breast tumors maintained 457, 608611 (2009).
in distant metastases. Proc. Natl Acad. Sci. USA 100, 1590115905 (2003). Using lineage-tracing studies, this paper demonstrates that colonic stem cells
5. Tlsty, T. D. & Coussens, L. M. Tumor stroma and regulation of cancer can act as the cell of origin for colon cancer. See also references 37 and 38.
development. Annu. Rev. Pathol. 1, 119150 (2006). 37. Zhu, L. et al. Prominin 1 marks intestinal stem cells that are susceptible to
6. Weinstein, I. B. Addiction to oncogenesthe Achilles heal of cancer. Science neoplastic transformation. Nature 457, 603607 (2009).
297, 6364 (2002). 38. Sangiorgi, E. & Capecchi, M. R. Bmi1 is expressed in vivo in intestinal stem cells.
7. Garraway, L. A. & Sellers, W. R. Lineage dependency and lineage-survival Nature Genet. 40, 915920 (2008).
oncogenes in human cancer. Nature Rev. Cancer 6, 593602 (2006). 39. Barker, N. et al. Lgr5+ve stem cells drive self-renewal in the stomach and build
8. Perez-Losada, J. & Balmain, A. Stem-cell hierarchy in skin cancer. Nature Rev. long-lived gastric units in vitro. Cell Stem Cell 6, 2536 (2010).
Cancer 3, 434443 (2003). 40. Holland, E. C. et al. Combined activation of Ras and Akt in neural progenitors
9. Bailleul, B. et al. Skin hyperkeratosis and papilloma formation in transgenic induces glioblastoma formation in mice. Nature Genet. 25, 5557 (2000).
mice expressing a ras oncogene from a suprabasal keratin promoter. Cell 62, 41. Alcantara Llaguno, S. et al. Malignant astrocytomas originate from neural stem/
697708 (1990). progenitor cells in a somatic tumor suppressor mouse model. Cancer Cell 15,
10. Brown, K., Strathdee, D., Bryson, S., Lambie, W. & Balmain, A. The malignant 4556 (2009).
capacity of skin tumours induced by expression of a mutant H-ras transgene 42. Marumoto, T. et al. Development of a novel mouse glioma model using lentiviral
depends on the cell type targeted. Curr. Biol. 8, 516524 (1998). vectors. Nature Med. 15, 110116 (2009).
11. Ince, T. A. et al. Transformation of different human breast epithelial cell types 43. Merkle, F. T., Mirzadeh, Z. & Alvarez-Buylla, A. Mosaic organization of neural
leads to distinct tumor phenotypes. Cancer Cell 12, 160170 (2007). stem cells in the adult brain. Science 317, 381384 (2007).
12. Sharma, M. K. et al. Distinct genetic signatures among pilocytic astrocytomas 44. Jacques, T. S. et al. Combinations of genetic mutations in the adult neural stem
relate to their brain region origin. Cancer Res. 67, 890900 (2007). cell compartment determine brain tumour phenotypes. EMBO J. 29, 222235
13. Taylor, M. D. et al. Radial glia cells are candidate stem cells of ependymoma. (2010).
Cancer Cell 8, 323335 (2005). 45. Wang, Y. et al. Expression of mutant p53 proteins implicates a lineage
14. Johnson, R. A. et al. Cross-species genomics matches driver mutations and cell relationship between neural stem cells and malignant astrocytic glioma in a
compartments to model ependymoma. Nature 466, 632636 (2010). murine model. Cancer Cell 15, 514526 (2009).
This paper demonstrates integrated genomic and cell-based approaches to 46. Bachoo, R. M. et al. Epidermal growth factor receptor and Ink4a/Arf: convergent
identify cells of origin in ependymomas. mechanisms governing terminal differentiation and transformation along the
15. Sorlie, T. et al. Gene expression patterns of breast carcinomas distinguish neural stem cell to astrocyte axis. Cancer Cell 1, 269277 (2002).
tumor subclasses with clinical implications. Proc. Natl Acad. Sci. USA 98, 47. Bruggeman, S. W. et al. Bmi1 controls tumor development in an Ink4a/Arf-
1086910874 (2001). independent manner in a mouse model for glioma. Cancer Cell 12, 328341
16. Sotiriou, C. et al. Breast cancer classification and prognosis based on gene (2007).
expression profiles from a population-based study. Proc. Natl Acad. Sci. USA 48. Schuller, U. et al. Acquisition of granule neuron precursor identity is a
100, 1039310398 (2003). critical determinant of progenitor cell competence to form Shh-induced
17. Lim, E. et al. Aberrant luminal progenitors as the candidate target population medulloblastoma. Cancer Cell 14, 123134 (2008).
for basal tumor development in BRCA1 mutation carriers. Nature Med. 15, This study demonstrates that descendants of stem cells can act as crucial
907913 (2009). cellular targets of transformation. See also reference 49.
Identification of an aberrant cell population in preneoplastic tissue, and 49. Yang, Z. J. et al. Medulloblastoma can be initiated by deletion of Patched in
discovery that mutant-BRCA1 tissue and basal cancers share a gene lineage-restricted progenitors or stem cells. Cancer Cell 14, 135145 (2008).
signature with normal luminal progenitors. 50. Sutter, R. et al. Cerebellar stem cells act as medulloblastoma-initiating cells in a
18. Hayashi, S. & McMahon, A. P. Efficient recombination in diverse tissues by mouse model and a neural stem cell signature characterizes a subset of human
a tamoxifen-inducible form of Cre: a tool for temporally regulated gene medulloblastomas. Oncogene 29, 18451856 (2010).
activation/inactivation in the mouse. Dev. Biol. 244, 305318 (2002). 51. Gibson, P. et al. Subtypes of medulloblastoma have distinct developmental
19. Metzger, D. & Chambon, P. Site- and time-specific gene targeting in the mouse. origins. Nature 468, 10951098 (2010).
Methods 24, 7180, (2001). 52. Joseph, N. M. et al. The loss of Nf1 transiently promotes self-renewal but not
20. Hewish, M., Lord, C. J., Martin, S. A., Cunningham, D. & Ashworth, A. Mismatch tumorigenesis by neural crest stem cells. Cancer Cell 13, 129140 (2008).
repair deficient colorectal cancer in the era of personalized treatment. Nature 53. Zheng, H. et al. Induction of abnormal proliferation by nonmyelinating Schwann
Rev. Clin. Oncol. 7, 197208 (2010). cells triggers neurofibroma formation. Cancer Cell 13, 117128 (2008).
21. Narod, S. A. & Foulkes, W. D. BRCA1 and BRCA2: 1994 and beyond. Nature Rev. 54. Wang, X. et al. A luminal epithelial stem cell that is a cell of origin for prostate
Cancer 4, 665676 (2004). cancer. Nature 461, 495500 (2009).
22. Fialkow, P. J., Denman, A. M., Jacobson, R. J. & Lowenthal, M. N. Chronic This report establishes a new luminal stem cell as a target of prostate
myelocytic leukemia. Origin of some lymphocytes from leukemic stem cells. carcinogenesis and indicates a hierarchy of stem cells in this tissue.
J. Clin. Invest. 62, 815823 (1978). 55. Leong, K. G., Wang, B. E., Johnson, L. & Gao, W. Q. Generation of a prostate from
23. Jamieson, C. H. et al. Granulocytemacrophage progenitors as candidate a single adult stem cell. Nature 456, 804808 (2008).
leukemic stem cells in blast-crisis CML. N. Engl. J. Med. 351, 657667 (2004). 56. Ma, X. et al. Targeted biallelic inactivation of Pten in the mouse prostate leads to
The first functional evidence that cells of origin and cancer-propagating cells prostate cancer accompanied by increased epithelial cell proliferation but not
in a given malignancy are likely to be distinct. by reduced apoptosis. Cancer Res. 65, 57305739 (2005).
24. Hope, K. J., Jin, L. & Dick, J. E. Acute myeloid leukemia originates from a 57. Korsten, H., Ziel-van der Made, A., Ma, X., van der Kwast, T. & Trapman, J.
hierarchy of leukemic stem cell classes that differ in self-renewal capacity. Accumulating progenitor cells in the luminal epithelial cell layer are candidate
Nature Immunol. 5, 738743 (2004). tumor initiating cells in a Pten knockout mouse prostate cancer model. PLoS
25. Barabe, F., Kennedy, J. A., Hope, K. J. & Dick, J. E. Modeling the initiation and ONE 4, e5662 (2009).
progression of human acute leukemia in mice. Science 316, 600604 (2007). 58. Mulholland, D. J. et al. LinSca-1+CD49fhigh stem/progenitors are tumor-
26. Wei, J. et al. Microenvironment determines lineage fate in a human model of initiating cells in the Pten-null prostate cancer model. Cancer Res. 69,
MLL-AF9 leukemia. Cancer Cell 13, 483495 (2008). 85558562 (2009).
27. Hong, D. et al. Initiating and cancer-propagating cells in TEL-AML1-associated 59. Lawson, D. A. et al. Basal epithelial stem cells are efficient targets for prostate
childhood leukemia. Science 319, 336339 (2008). cancer initiation. Proc. Natl Acad. Sci. USA 107, 26102615 (2010).
28. Huntly, B. J. et al. MOZ-TIF2, but not BCR-ABL, confers properties of leukemic 60. Goldstein, A. S., Huang, J., Guo, C., Garraway, I. P. & Witte, O. N. Identification of a
stem cells to committed murine hematopoietic progenitors. Cancer Cell 6, cell-of-origin for human prostate cancer. Science 329, 568571 (2010).
587596 (2004). This study demonstrates through transduction of isolated cell subsets that
29. Krivtsov, A. V. et al. Transformation from committed progenitor to leukaemia basal stem/progenitor cells are an important target cell.
stem cell initiated by MLLAF9. Nature 442, 818822 (2006). 61. Foulkes, W. D. BRCA1 functions as a breast stem cell regulator. J. Med. Genet.
30. Cozzio, A. et al. Similar MLL-associated leukemias arising from self-renewing 41, 15 (2004).
stem cells and short-lived myeloid progenitors. Genes Dev. 17, 30293035 62. Turner, N., Tutt, A. & Ashworth, A. Hallmarks of BRCAness in sporadic cancers.
(2003). Nature Rev. Cancer 4, 814819 (2004).
31. Drynan, L. F. et al. Mll fusions generated by Cre-loxP-mediated de novo 63. Molyneux, G. et al. BRCA1 basal-like breast cancers originate from luminal

2 0 JA N UA Ry 2 0 1 1 | VO L 4 6 9 | N AT U R E | 3 2 1
2011 Macmillan Publishers Limited. All rights reserved
 insight Review

epithelial progenitors not from basal stem cells. Cell Stem Cell 7, 403417 capable of initiating tumour growth in immunodeficient mice. Nature 445,
(2010). 106110 (2007).
This study provides in vivo functional evidence that luminal progenitors rather 77. Kim, C. F. et al. Identification of bronchioalveolar stem cells in normal lung and
than basal cells are an important target in the genesis of BRCA1-associated lung cancer. Cell 121, 823835 (2005).
breast tumours. 78. Curtis, S. J. et al. Primary tumor genotype is an important determinant in
64. Liu, S. et al. BRCA1 regulates human mammary stem/progenitor cell fate. Proc. identification of lung cancer propagating cells. Cell Stem Cell 7, 127133 (2010).
Natl Acad. Sci. USA 105, 16801685 (2008). 79. Zhang, L. et al. Chemoprevention of colorectal cancer by targeting APC-deficient
65. Bouras, T. et al. Notch signaling regulates mammary stem cell function and cells for apoptosis. Nature 464, 10581061 (2010).
luminal cell-fate commitment. Cell Stem Cell 3, 429441 (2008). 80. Martin, S. A. et al. DNA polymerases as potential therapeutic targets for cancers
66. Lee, C. W. et al. A functional Notchsurvivin gene signature in basal breast deficient in the DNA mismatch repair proteins MSH2 or MLH1. Cancer Cell 17,
cancer. Breast Cancer Res. 10, R97 (2008). 235248 (2010).
67. Vaillant, F. et al. The mammary progenitor marker CD61/3 integrin identifies 81. Forbes, J. F. et al. Effect of anastrozole and tamoxifen as adjuvant treatment for
cancer stem cells in mouse models of mammary tumorigenesis. Cancer Res. early-stage breast cancer: 100-month analysis of the ATAC trial. Lancet Oncol.
68, 77117717 (2008). 9, 4553 (2008).
68. Prat, A. et al. Phenotypic and molecular characterization of the claudin-low 82. Asselin-Labat, M. L. et al. Control of mammary stem cell function by steroid
intrinsic subtype of breast cancer. Breast Cancer Res. 12, R68 (2010).
hormone signalling. Nature 465, 798802 (2010).
69. Gidekel Friedlander, S. Y. et al. Context-dependent transformation of adult
83. Ashworth, A. Drug resistance caused by reversion mutation. Cancer Res. 68,
pancreatic cells by oncogenic K-Ras. Cancer Cell 16, 379389 (2009).
1002110023 (2008).
70. De La, O. J. et al. Notch and Kras reprogram pancreatic acinar cells to ductal
84. Tutt, A. et al. Oral poly(ADP-ribose) polymerase inhibitor olaparib in patients
intraepithelial neoplasia. Proc. Natl Acad. Sci. USA 105, 1890718912
(2008). with BRCA1 or BRCA2 mutations and advanced breast cancer: a proof-of-
71. Habbe, N. et al. Spontaneous induction of murine pancreatic intraepithelial concept trial. Lancet 376, 235244 (2010).
neoplasia (mPanIN) by acinar cell targeting of oncogenic Kras in adult mice. 85. Lindberg, N., Kastemar, M., Olofsson, T., Smits, A. & Uhrbom, L. Oligodendrocyte
Proc. Natl Acad. Sci. USA 105, 1891318918 (2008). progenitor cells can act as cell of origin for experimental glioma. Oncogene 28,
72. Youssef, K. K. et al. Identification of the cell lineage at the origin of basal cell 22662275 (2009).
carcinoma. Nature Cell Biol. 12, 299305 (2010).
73. Owens, D. M. & Watt, F. M. Contribution of stem cells and differentiated cells to Acknowledgements I am grateful to J. Adams, G. Lindeman and A. Strasser
epidermal tumours. Nature Rev. Cancer 3, 444451 (2003). for critical review of the manuscript, P. Dirks for discussion and P. Maltezos
74. Bonnet, D. & Dick, J. E. Human acute myeloid leukemia is organized as a for preparation of figures. I apologize to authors whose work could not be
hierarchy that originates from a primitive hematopoietic cell. Nature Med. 3, cited owing to space limitations. J.E.V. is supported by the National Health
730737 (1997). and Medical Research Council and the Victorian Breast Cancer Research
The first report indicating that early stem/progenitor cells are targeted for Consortium.
transformation in AML.
75. Ricci-Vitiani, L. et al. Identification and expansion of human colon-cancer- Author Information Reprints and permissions information is available at
initiating cells. Nature 445, 111115 (2007). www.nature.com/reprints. The author declares no competing financial interests.
76. OBrien, C. A., Pollett, A., Gallinger, S. & Dick, J. E. A human colon cancer cell Correspondence should be addressed to the author ([email protected]).

3 2 2 | N AT U R E | VO L 4 6 9 | 2 0 JA N UA Ry 2 0 1 1
2011 Macmillan Publishers Limited. All rights reserved