Gastrointestinal stem cells in development and cancer


  • No conflicts of interest were declared.


An enormous body of knowledge about the biology of stem cells and their role in development, tissue homeostasis and cancer formation has been gained in the last 20 years. This review gives a comprehensive overview on knowledge about localization and regulation of normal gastrointestinal stem cells and links it to our understanding of gastrointestinal tumourigenesis and malignant progression in the light of the cancer stem cell concept. The focus is on intestinal stem cells and newly identified stem cell factors, such as the β-catenin target gene Lgr5. The basis of intestinal stem cell regulation is a permanent crosstalk between epithelial and underlying mesenchymal cells in the intestinal stem cell niche. This crosstalk is mediated by crucial pathways, including the Wnt, Hedgehog (HH), Notch, PI3K and BMP pathways. Disturbances in this fine-regulated interaction can both initiate intestinal tumours and, in association with additional genetic alterations or environmental activation of embryonic processes such as epithelial–mesenchymal transition (EMT), lead to tumour invasion and metastasis. Copyright © 2008 Pathological Society of Great Britain and Ireland. Published by John Wiley & Sons, Ltd.


The intestine is one of the most investigated organ systems, in terms of both normal intestinal development and cancer formation. Development and tumourigenesis of the small intestine have been mainly studied in mouse model systems, whereas most data of colorectal adenomas and carcinomas are derived from human tumours. This review summarizes the knowledge about the localization and regulation of normal gastrointestinal stem cells and links it to the understanding of gastrointestinal tumourigenesis and malignant progression in the light of the cancer stem cell concept. We will show that there is an enormous increase in the knowledge about specific aspects of normal development, tissue homeostasis and carcinogenesis in the gastrointestinal (GI) tract. The focus is on intestinal stem cells. The basis of intestinal development from gastrulation to adult tissue homeostasis is a permanent crosstalk between epithelial and underlying mesenchymal cells. This crosstalk is mediated by crucial pathways, including the Wnt, Hedgehog (HH), Notch, PI3K and BMP pathways. Disturbances in this fine-regulated interaction can lead to tumour formation. But many questions, for instance about the link between normal GI stem cells and potential GI cancer stem cells, still remain open.

Intestinal stem cells in normal development and tissue homeostasis

After gastrulation, a stratified intestinal epithelium is derived from endoderm 1. Epithelial–mesenchymal interactions induce an evagination of villi in the small intestine and later, after birth, an invagination of crypts 2, 3. The regulation of these morphogenetic processes within the whole gut wall is poorly understood, but Wnt and HH signals seem to play a major role. In parallel, epithelial cells in the gut mucosa differentiate to specific subtypes. Finally, the small intestine is composed of villi and crypts, lined by a single epithelial layer which consists of stem cells, transit-amplifying cells and the differentiated subtypes of secreting Paneth cells, enteroendocrine cells, goblet cells and absorbing enterocytes (Figure 1). The colorectal epithelium lacks villi and Paneth cells 4. Two ways of crypt generation are proposed: a de novo crypt formation only directly after birth, and a crypt fission postnatally and in adulthood, for instance in regenerative processes 5.

Figure 1.

Architecture and signalling in the small intestinal mucosa. Terminally differentiated intestinal epithelial cells (red) are found in the villi; only Paneth cells are located in the crypt base. Green indicates location of putative intestinal stem cells (dark green) and the transit-amplifying cells. Crypt base columnar cells (CBCs) selectively express the Wnt target gene Lgr5/GPR49. Gradients of pathways (or components) that induce stemness or proliferation are indicated in green. Triggers of differentiation are shown in red

Localization of intestinal stem cells

Based on studies showing that intestinal crypts are monoclonal, it was proposed that each crypt is derived from their own intestinal stem cells (ISCs) 6–9. Like all stem cells, ISCs should also be able to self-renew by asymmetric cell division, proliferate in transit-amplifying units and give rise to all four differentiated types of intestinal epithelial cells. Since specific markers were lacking, the localization of putative ISCs was only possible by indirect means. Based on a postulated slow cycling, Potten et al used the long-term label-retaining technique to mark putative ISCs in the small intestine. They detected long-term label-retaining cells (LRCs) in an annulus four cells up from the crypt base (+4 LRCs) 10–12. These cells were long considered as ISCs. However, this hypothesis was recently challenged by the work of Barker et al13. Based on the assumption that Wnt signalling is an important determinant of intestinal stemness, they identified a Wnt target gene, Lgr5/GPR49, which is expressed exclusively in crypt base columnar cells (CBCs) at crypt positions 1–4. CBCs were initially described by Cheng and Leblond in 1974 and are located between Paneth cells 14, and later were also proposed to be the ISCs 15 (Figure 1). Using lineage tracking experiments in Lgr5–GFP–IRES–Cre–ERT X RosaLacZ mice, Barker et al now elegantly showed that Lgr5-expressing CBCs fulfil all criteria of putative ISCs: they can persist for a long time, self-renew, give rise to all mature intestinal epithelial cells and in addition are apoptosis-resistant. However, in contrast to current views, these cells are highly proliferative. Notably, CBCs also express Musashi (Msi-1) 16. Msi-1 is important for Notch signalling 17 by inhibiting expression of the Notch repressor Numb 18, and is also considered by some to be a stem cell factor in ISCs.

How could the discrepancy between the + 4 LRC model and the CBC model be explained? The obvious possibility would be that + 4 LRCs are not ISCs, since they were never functionally characterized as stem cells. But there are still other options to integrate both cell types in an ISC model. Scoville et al recently proposed that there are two types of ISCs 4: a ‘reserve’ pool of quiescent ISCs, corresponding to the + 4 LRCs, and the active cycling CBCs, which are ready to receive regulating stimuli from the underlying stromal cells (see below). Indeed, this could be necessary for stem cells in high-turnover tissues, such as the gut epithelium. Very recently a publication by Sangiorgi and Capecchi 19 supported this hypothesis: they found that the polycomb factor Bmi1, known to be involved in self-renewal of haematopoietic and neural stem cells, is exclusively expressed in the + 4LRCs. By constructing mice with inducible Cre from the Bmi1 locus, they elegantly showed that these cells can self-renew, proliferate, expand and, like the CBCs, give rise to all the differentiated lineages of the small intestine epithelium. They postulated that + 4LRCs and CBCs are stem cells in different stem cell niches, which can migrate from one niche to the other. However, the lack of established protocols to cultivate putative ISCs in vitro is still preventing further functional characterization of such cells. Nevertheless, the Lgr5/GPR49 data are so convincing that CBCs should be considered as ISCs. Moreover, lineage-tracking experiments revealed that Lgr5/GPR49 may be also expressed in the stem cell compartment of other epithelial tissues, including hair follicle, mammary gland, stomach epithelium and, importantly, also in the basal regions of the colon crypts 13, 20. This makes Lgr5 the hottest candidate of a universal epithelial stem cell marker.

Regulation of intestinal stem cells and differentiation

The status and function of all stem cells, including ISCs, has to be strictly controlled in order to maintain tissue homeostasis and to prevent tissue degeneration or tumour growth. A fine-tuned interaction of epithelial cells and underlying mesenchymal/stromal cells seems to be crucial. Thereby, a spatial and temporal restriction of secreted ligands and corresponding receptor expression generates specific morphogenetic areas, including the stem cell niche at the crypt base (Figure 1). A few embryonic pathways were shown to be central for the regulation of ISCs and subsequent tissue development and homeostasis. Their specific function and interaction to control ISC activity is briefly described.

Wnt pathway

The discovery that mutations in the APC gene, the most important tumour suppressor in intestinal tumourigenesis, affects the control of Wnt signalling, boosted the knowledge about molecular components and the function of the Wnt pathway within the last 10 years 21–28. Very briefly explained, binding of Wnt ligands to their receptors, Fzd/LRP5/6, prevents degradation of the main effector β-catenin by a destruction complex containing APC and Axin1/2. β-Catenin translocates to the nucleus, where it acts as a transcriptional activator after binding to TCF/LEF family members (for a more detailed description of the Wnt pathway, see eg 29–31). Aberrant activation of the Wnt pathway, due to mutations in regulatory components or effectors such as APC or β-catenin, respectively, is crucial for the development of many different human tumours 32. A causal connection with tumourigenesis was confirmed in many mouse model systems 33–37. The canonical Wnt pathway turned out to be the major regulator of stem cells in many organ systems, including intestinal, epidermal, hair follicle, haematopoietic and neural systems 38–41. In the intestine, Wnt signalling is essential for endoderm formation 42, but also central for subsequent intestinal development. In a seminal work, Korinek et al showed that lack of the intestine specific β-catenin partner TCF4 results in the depletion of the epithelial stem cell compartment in the small intestine 43. The identification of many different Wnt/ β-catenin target genes indicated how Wnt signalling is involved in ISC regulation, but also showed that Wnt signalling has different effects in different cell types, also depending on its localization along the crypt axis. The best stemness-specific Wnt/ β-catenin known today is Lgr5/GPR4913. Other Wnt targets are associated with proliferation in transit-amplifying cells and include c-myc and cyclin D1 44–46. The identification of EphB/ephrinB as Wnt targets furthermore proved that Wnt signalling is involved in the control of intestinal morphogenesis 47. Wnt targets such as MMP7 47–50, lamc2 51 or Snai2 (Slug) 52 showed that Wnt signalling also induces specific epithelial differentiation and is involved in epithelial–mesenchymal transition (EMT) and tumour invasion. For a comprehensive list of Wnt target genes, see the Wnt homepage by R. Nusse ( The differential expression of Wnt target genes indicates that the effects of Wnt signalling are coordinated through interaction with other pathways.

PTEN–PI3K–Akt pathway

A second pathway that plays a role in ISC regulation is the PTEN–PI3K–Akt pathway (referred to here as the PI3K pathway). PI3K, composed of the catalytic subunit p110 and the regulatory subunit p85, is the central player. The p85 subunit can bind to receptor tyrosine kinases (RTKs). Subsequent to ligand binding, the p110 subunit is activated and phosphorylates its substrate. This leads to phosphorylation and activation of the kinase Akt by PDK1. In contrast to many other pathways, the PI3K pathway has only one major negative regulator, the phosphatase PTEN. Active, phosphorylated PTEN reconverts PIP3 to PIP2, thereby inhibiting Akt function (for review, see 53).

PI3K turned out to be activated in many different human tumours, which may be attributed to its important role in activating cell survival, growth and proliferation. Forty percent of all human colorectal carcinomas show an activated PI3K pathway, mainly due to PTEN inactivation 54. Moreover, inherited mutations in PTEN (Cowden syndrome) lead to intestinal hamartomatous polyps. A role of the PI3K pathway in ISC regulation was indicated by the observation that PI3K signalling enhances ISC self-renewal 55. This might be explained by a link between the PI3K signalling and the Wnt pathway: p-Akt can phosphorylate β-catenin, the main effector of the canonical Wnt pathway at S552 56. Thus, p-Akt may induce nuclear accumulation and thus enhance the transcriptional activity of β-catenin 55, 57. PTEN is expressed in a crypt–lumenal gradient, with strongest expression in lumenal epithelial cells, and thereby might be involved in the restriction of strong Wnt signalling to the crypt base (Figure 1) 58.

BMP pathway

Recent evidence indicates that inhibition of BMP signalling is another requirement to confer intestinal stemness, by enhancing and specifying the effects of both the Wnt and the PI3K pathways. Bone morphogenetic proteins (BMPs) bind to BMP receptor types I or II (BMPR1 or BMPR2). This leads to phosphorylation of SMAD1, 5 or 8, which then form a heterodimer with SMAD4, translocate to the nucleus and act as transcriptional activators 59. In the intestine, BMP4 is secreted by intervillus stromal cells and BMPR1 is expressed in all intestinal epithelial cells 55, 60. Active BMP signalling, indicated by phosphorylated SMADs, is found predominantly in differentiated intestinal epithelial cells. Interestingly, it was shown that BMP pathway inhibition by conditional mutation of BMPR1A resulted in de novo crypt formation and a juvenile polyposis phenotype, indicating that the BMP pathway may control the ISC number and self-renewal 55. Moreover mutation of Bmpr1a led to reduced differentiation into the three secretory cells lines, Paneth cells, goblet cells and enteroendocrine cells in mice 61. Also, human juvenile polyposis has been shown to be associated with mutations in the SMAD4/DPC4 or BMPR1A genes 62, 63.

Physiological inhibitors of the BMP pathway are Noggin and Gremlin, which bind and inactivate BMPs 64, 65. Like genetic inactivation of BMPR1A, over-expression of Noggin leads to de novo crypt formation and a juvenile polyposis phenotype in mice 60. Consistent with the postulated inhibitory role of BMP signalling on ISC self-renewal, these BMP antagonists were found to be expressed in intestinal subepithelial myofibroblasts at the crypt base 66 (Figure 1). How can the inhibition of BMP signalling result in the formation of ISCs? Recent results again point to the Wnt and PI3K pathways. Inhibition of BMP signalling in intestinal epithelial cells by Gremlin activated Wnt signalling 66. Moreover, BMP stabilizes PTEN, thereby leading to reduced Akt activity and subsequent reduction of nuclear β-catenin accumulation 57, 67, 68. These results are supported by the phenotype of Pten mutant mice, which show similar tumours to those of Bmpr1a-mutant or Noggin-transgenic mice, characterized by aberrant crypt fission and increased de novo crypt formation 55, 56, 60.

Notch pathway

Notch signalling is known to control cell fate decisions in the development of many tissues. The ligands Delta or Jagged bind to the Notch receptor, thereby inducing its proteolytic cleavage by γ-secretase. A cleavage fragment of Notch, NCID, translocates to the nucleus, where it acts as a transcription factor after dimerization with RBP-jκ/CSL. This induces expression of a bHLH transcription factor termed hairy/enhancer of split (Hes), which finally activates factors involved in the control of proliferation and differentiation 69. Manipulations in different components of the Notch pathway in mice revealed its role in intestinal epithelial differentiation. Knocking out RBP-jκ or Hes1 led to increased numbers of secretory epithelial cells 70, 71. A similar effect was seen after treating mice with inhibitors of γ-secretase 72. Consistent with these effects, mutation in Atoh1, a transcription factor repressed by Notch signalling, led to depletion of all three secretory lineages 73. Using inducible gut-specific Notch-mutant mice, it was shown that Notch signalling is important to maintain the proliferative crypt compartment 74. In summary, Notch signalling seems to trigger proliferation of crypt progenitor cells in the transit-amplifying units and a regulated reduction of Notch signalling in cooperation with activation of specific bHLH factors, such as Atoh1 and neuroD 73, 75, induces specific differentiation into the intestinal epithelial lineages.

Hedgehog pathway

In contrast to the other signal pathways described above, the direction of hedgehog (HH) signalling is reversed. The morphogens sonic hedgehog (Shh) and Indian hedgehog (Ihh) are secreted by epithelial cells and their receptor, Patched (PTCH) is expressed in the subepithelial myofibroblasts. Consequently, HH signalling is not directly involved in the fate of the epithelial cells but is important to build up the correct overall structure into crypts and villi of the intestinal mucosa 76 (for further reading about HH signalling and its role in development, see the excellent review by van den Brink 77). Transgenic over-expression of the pan-HH inhibitor Hhip in the intestinal epithelium resulted in reduced villus formation and mislocalization of subepithelial stromal cells. Knowing the important role of the stromal–epithelial interaction in the regulation of the epithelial cell fate, disturbed HH signalling also has strong secondary effects on intestinal epithelial cells. Wnt signalling is enhanced, proliferation is increased and atypical crypt structures are formed in the villi. These effects can be attributed to reduced BMP expression by stromal cells, which is normally triggered by HH 55. Moreover, Ihh was shown to down-regulate expression of TCF4 and β-catenin and thus its expression pattern, with a maximum at the lumenal crypt site in the colon epithelium, restricts Wnt signalling to the crypt base 78.

Shaping the intestinal stem cell niche

It has become clear that different pathways interact to specify ISCs. Thereby, Wnt signalling is the most dominant force in controlling ISCs, proliferation and differentiation, but the other pathways are important modulators. The increasing knowledge about molecular components and functions of these pathways as well as the spatial restrictions in the expression of their ligands and receptors suggests a model of the intestinal epithelial stem cell niche (Figure 2).

Figure 2.

Putative intestinal stem cell niche. Scheme of the architecture and signalling in the mucosa of the large intestine (left). Red indicates differentiated cells and pathway components inducing differentiation. Green indicates stem cells (CBCs), transit-amplifying cells or signalling inducing it. A simplified view of the stem cell niche, which might turn out be much more complex (right): putative signalling crosstalks defining the stem cell niche and the differentiating cell niche. In the upper crypt regions, indian hedgehog (Ihh) triggers BMP expression in stromal cells, which then activates PTEN expression in epithelial cells. All three factors directly or indirectly inhibit Wnt signalling. In the stem cell niche, surrounding stromal cells secrete the BMP inhibitors Noggin and Gremlin, leading to a relief of repression of Wnt signalling. In addition, the stroma provides Wnt ligands to induce Wnt signalling through Fzd receptors in stem cells and transit-amplifying cells. Additional unknown signals specify Wnt signalling towards expression of stem cell-associated target genes (X) or proliferation-associated target genes (Y)

Intestinal stem cells and cancer

How important is gastrointestinal stem cell biology for gastrointestinal cancer? A parallel evolution of knowledge from studying normal and tumour development in mouse models as well as human intestinal cancer allowed a compilation of the most relevant data. These data pointed out that cancer may be considered, at least partially, a consequence of dysregulated stem cell control. Research in intestinal development and cancer was among the most successful examples. The rationale supporting the stem cell origin of intestinal tumours is:

  • 1.In contrast to the short-living differentiated intestinal epithelial cell, in this high-turnover tissue ISCs are long-lived, thus allowing accumulation of critical genetic alterations.
  • 2.In colorectal cancer the most common, rate-limiting mutations, such as those occurring in the APC gene, increase Wnt signalling, which turns out to be the crucial regulator of ISCs.
  • 3.Typical human colorectal adenocarcinomas and corresponding metastases are heterogeneous and show many differentiation stages within an individual tumour, which is in line with a stem cell origin.
  • 4.The cancer stem cell concept, originally developed for haematopoietic neoplasia, is becoming accepted for solid cancers, including colorectal cancer.

The second part of this review discusses the role of dysregulated pathways controlling ISCs for the initiation and malignant progression of intestinal cancer, as well as recent progress in the operational identification of intestinal cancer stem cell markers.

Functional connection to known ISC regulators

Fundamental knowledge was gained by the identification of APC as the crucial tumour suppressor mutated in familial adenomatosis polyposis coli (FAP) 22, 23 and subsequent results showing that APC is a negative regulator of the Wnt pathway 21, 26–28, 79. The critical role of aberrant Wnt signalling was further supported by the fact that almost all human colorectal adenomas and carcinomas show genetic alterations in one of the Wnt pathway components, mainly APC loss-of-function- or β-catenin-activating mutations 80. Mouse models using different APC-deletion mutants or activating β-catenin mutants further demonstrated a direct role of Wnt signalling in intestinal tumourigenesis 33–37. Thus, alterations in Wnt signalling and also in the other important ISC regulatory pathways were shown to affect both early and late steps in intestinal cancer.

Tumour initiation and benign tumour growth

In FAP, which is associated with germline mutations in APC, the earliest visible change before the development of small, dysplastic adenomas, is an upward shift in the proliferative compartment 81. Mathematical models proposed an increase of ISCs as the origin of the observed hyperproliferation 82–84. Using non-oncogenic mitochondrial DNA mutations as markers, it was elegantly shown that mutant clones initially expand by crypt fission 85. Wnt signalling enhances expression of proliferation-associated genes, such as c-myc and CyclinD1, and, together with Notch signalling, regulates the switch of ISCs to transit-amplifying crypt progenitor cells. This supports the idea that APC mutations affect the regulation of ISCs. Furthermore, aberrant, asymmetric and increased crypt fission is detectable in FAP 86 (Figure 3). This is also in line with an increase of ISCs and led to the proposal that APC mutations lead to a shift from an asymmetrical to a symmetrical division of ISCs 87, 88. Other models predicted increased stem cell survival after APC mutation, which can be explained by enhanced expression of the Wnt target gene survivin, also considered to be a putative stem cell marker 84, 89, 90.

Figure 3.

Aberrant crypt fission in adenoma of FAP patients. (A) Monocryptal adenoma (arrow) between normal colon crypts. (B) Aberrant crypt fission in a monocryptal adenoma. (C, D) Microdissected crypts, showing normal mucosa with symmetrical crypt fission (arrow) (C) and a monocryptal adenoma with multiple, asymmetrical crypt fission (D) (adapted from 86)

In the subsequent adenoma stage, the epithelium shows large areas of an immature phenotype, resembling the restricted transit-amplifying zone in the lower parts of normal crypts 91 and indicating an expansion of the proliferative cell population 92. Moreover, nuclear β-catenin, a hallmark of active Wnt signalling and normally only visible in putative ISCs at the crypt base, is increasingly detectable and associated with increased dysplasia in human adenomas 93 and mouse tumour models 35. This might again reflect an increase in the ISC population. A dysregulated control of ISCs in early stages of intestinal tumourigenesis is further indicated by the fact that alterations in other ISC regulatory pathways induce or influence intestinal tumour growth: 40% of all human colorectal carcinomas show an over-activated PI3K pathway, mainly due to reduced PTEN function 54. Moreover, inherited mutations in PTEN (Cowden syndrome) lead to intestinal hamartomatous polyps. Inhibition of the BMP pathway by either conditional mutation of BMPR1A or over-expression of the BMP inhibitor Noggin resulted in de novo crypt formation and juvenile polyposis in mice 60. Removal of the common Notch pathway transcription factor CSL/RBP-J or blocking the Notch cascade with a γ-secretase inhibitor induced goblet cell differentiation and attenuated growth of adenomas in mice carrying a mutation of the APC tumour suppressor gene 72.

Malignant progression towards invasive carcinomas and metastases

Analyses in human colorectal cancer indicated that aberrant Wnt signalling not only triggers early steps of intestinal carcinogenesis, but is also involved in malignant tumour progression towards invasive carcinomas and metastases 29. An initial observation was that, in typical colorectal adenocarcinomas, strong nuclear β-catenin accumulation is predominantly found in undifferentiated tumour cells at the invasive front. In contrast the differentiated central tumour areas often lacked nuclear β-catenin 94, 95 (Figure 4A). Interestingly, most metastases again showed the differentiated phenotype of the primary tumour, indicated by lack of nuclear and retention of membranous β-catenin (Figure 4B). These changes were accompanied by a loss of E-cadherin at the invasive front and its regain in metastases, which indicated that invasive tumour cells undergo an EMT and a subsequent mesenchymal–epithelial retransition (MET) in the metastases 96–98. Moreover, the observed transient EMT phenotype of the invading tumour cells points out that, in addition to genetic alterations, these phenotypic changes, including the strong nuclear accumulation of β-catenin, are triggered by signals from the tumour environment 98. Nuclear β-catenin might be directly involved in the observed EMT, since Wnt signalling not only regulates stemness and organ development, but was also shown to mediate EMT, eg in gastrulation 99, 100. It is proposed that a differentiated activation of Wnt signalling regulates different functions in a dose-dependent fashion 101. Thereby, initiating mutations (APC or β-catenin) lead to a weak activation of Wnt signalling, enough to disturb normal ISC regulation. Additional somatic mutations, for instance as shown for KRAS in a mouse model 102, or environmental triggers, as suggested for the most common human colorectal adenocarcinomas 98, further boost nuclear β-catenin accumulation and Wnt signalling, which now activates additional Wnt-associated processes, such as EMT.

Figure 4.

Nuclear accumulation of β-catenin and EMT at the invasive fronts of primary tumours and metastases. (A) Immunohistochemical staining of a colorectal carcinoma for β-catenin (red, nuclear counterstaining in blue). Towards central areas (left), the tumour exhibits differentiated qualities, characterized by tubular structures built up of polarized epithelial tumour cells lacking nuclear β-catenin (arrows). In invasive regions, tumour cells change their morphology and undergo EMT: dissociation towards single, disseminating tumour cells with strong accumulation of nuclear β-catenin (arrowheads). Scale bar = 75 µm (adapted from 95). (B) Nuclear β-catenin (brown staining) in few epithelial cells at the base of a normal colon crypt. Nuclear β-catenin accumulates in dedifferentiated tumour cells at the tumour–host interface, which have undergone epithelial–mesenchymal transition (EMT) and are potential migrating cancer stem cells (arrows). Towards central tumour areas (arrowheads), tumour cells differentiate and polarize, building up tubular structures, and lose nuclear β-catenin. Of note, this heterogeneous pattern of the primary tumour is recapitulated in corresponding metastases, eg in the liver. Inserts show magnifications of the corresponding central and invasive areas

These observations led to the model of migrating cancer stem cells (MCS cells) (Figure 5) 103: in addition to the acquisition of stemness, the low-cycling 104, strong nuclear β-catenin-expressing tumour cells at the invasion front activate EMT and now combine two features necessary to make metastases—the ability to both disseminate in the body and restart the growth and differentiation programme of stem cells at the metastatic site. This model combines many aspects of human cancers, such as the initiation of carcinogenesis by mutations and triggering malignant progression by the tumour environment. Moreover, it can explain why cancers, in particular colorectal adenocarcinomas, are heterogeneous and often include all differentiation stages of the derived tissue, and why primary tumour and metastases are so similar in their overall appearance. The proposed functional connection of EMT and stemness in invading MCS cells is supported by a recent finding in human mammary epithelial cells, where it was shown that the EMT activators, eg FoxC2, can also confer stem cell properties on epithelial cells 105.

Figure 5.

The migrating cancer stem cell (MCS cell) concept. (A) Normal stem cells (expressing nuclear β-catenin) are located at the crypt base of normal colon mucosa. Stationary cancer stem (SCS) cells are embedded in benign adenomas and might still be detectable in differentiated central areas of carcinomas and metastases. A crucial step towards malignancy is the induction of an epithelial–mesenchymal transition (EMT) in tumour cells, including SCS cells, which now become mobile, migrating cancer stem cells (MCS) cells. This step could be activated by aberrant environmental signals [in the example of colorectal cancer, after crossing the border of mucosa to submucosa (lamina muscularis mucosae) in late adenomas]. (B) Detailed view on MCS cells in carcinomas and metastases. MCS cells divide asymmetrically; one daughter cell starts proliferation and differentiation (1). The remaining MCS cell either migrates a short distance before new asymmetrical division, thereby adding mass to the primary tumour (2), or eventually starts long-range dissemination through the blood or lymphatic vessels and builds up a metastasis after subsequent asymmetrical divisions (3). Thus, the basic mechanisms are the same for primary carcinomas and metastases (adapted from 103)

Colorectal cancer stem cells—operational definition

Independent from the knowledge-based linkage of dysregulated regulatory pathways and subsequent stem cell activation as origin of cancer, an operational definition has led to the identification of cancer stem cells (CSCs) 106–108. Initially identified in haematopoietic tumours 109, CSCs have now also been discovered in many solid tumours, including brain, breast, pancreas and colon cancers 110. The operational definition is based on fundamental traits of stem cells—self-renewal by asymmetric division, proliferation and differentiation to all possible phenotypes and resistance to apoptosis 111. The required combinations of these traits make CSCs rare and difficult to detect but highly tumourigenic in nude mice xenografts. Recently, different marker combinations were published, allowing the isolation of CSCs also from human colorectal carcinomas. Two publications showed that a selection of tumour cells expressing CD133 from colorectal cancer led to a strong enhancement of tumourigenicity after xenografting the cells in immunodeficient mice 112, 113. Moreover, these cells were able to self-renew and to re-establish the morphology of the primary tumour. CD133 (also known as prominin-1) was initially identified as marker of Drosophila neuroblasts 114. Although it is expressed on different types of stem cells and may play a role in cell polarity, a significant role in key stem cell function is unknown. CD133 is described as potential marker of tumour-initiating/cancer stem cells in different tumours, including neural, prostate, renal and hepatocellular tumours 115. However, recently its role as a marker of cancer stem cells in colorectal carcinoma was questioned by Shmelkov et al116, who showed that CD133 expression is not restricted to stem cells and that both CD133-positive and -negative metastatic colon cancer cells can initiate tumours. Another promising combination of three markers to select for potential colorectal CSCs was described by Dalerba et al117. Selection of tumour cells for high expression of CD44 and EpCAM enriched for initiating- cells. Further enrichment for CD166 expression enhanced the cancer stem cell properties of the tumour cells. Although CD166 is a mesenchymal stem cell marker, its direct function in cancer stem cells is not known.

Until now it has been unclear whether all the published markers really mark CSCs or only lead to their enrichment. Also it is unclear whether these markers are directly linked to stem cell functions. Therefore, the search for the perfect CSC marker, also useful as a therapeutic target, will go on. So far, the ISC marker Lgr5 (GRP49) is the most promising candidate 13, 20 and its role and expression in CSCs of different human tumours should be investigated.

Cancer stem cells in other GI tumours

Recently, also for other human GI cancers, marker combinations to select for CSCs were described (Table 1). Two independent publications showed that CD133 is also a marker to enrich CSCs in hepatocellular carcinomas 118, 119. Moreover, a combination of CD133 and the chemokine receptor CXCR4 was able to isolate putative CSCs in pancreatic adenocarcinomas 120. A second protocol using CD24, CD44 and CD326/ESA to enrich for such cells in pancreatic carcinomas was recently published by two groups 121, 122. In a mouse model, Houghton et al elegantly showed that gastric cancer can originate from bone marrow-derived cells 123. However, this is not yet confirmed for human gastric cancer. No markers to enrich for CSCs in human gastric, oesophageal and biliary duct cancer have yet been described.

Table 1. Cancer stem cell markers in GI cancers
GI cancerMarkersReferences
ColorectalCD133112, 113
 CD44, EpCAM, CD166117
HepatocellularCD133118, 119
PancreaticCD133, CXCR4120
 CD24, CD44, CD326/ESA121, 122

Concluding remarks

An enormous body of knowledge about the biology of intestinal stem cells has been gained over the last 20 years. In particular, Wnt signalling turns out to be the most potent regulator of ISCs. It is also Wnt signalling that indicated the link between ISC biology and intestinal cancer. In addition, it also became clear that intensive crosstalk between stromal cells and epithelial cells through additional pathways plays an important role in specifying ISCs and controlling intestinal morphogenesis and differentiation. However, important questions are still unanswered and should be addressed in the future:

  • What is the link between the Bmi1-expressing + 4LRCs and the Lgr5-expressing CBCs?

  • What is the exact nature of the intestinal stem cell niche; are there two niches?

  • Why is the Wnt target Lgr5 so strictly restricted to CBCs, in contrast to many other Wnt targets?

  • Is Lgr5/GPR49 also a marker for CSCs?

  • Can CSCs only derive from normal ISCs or also from their differentiated descendants?

  • What environmental factors trigger dissemination and redifferentiation of CSCs in metastases?

  • Are currently identified CSC markers functionally connected to stemness?

  • Finally, despite the enormous gain of knowledge, what is the benefit for the patient, in particular for cancer patients? The present knowledge is a broad basis for developing new strategies to diagnose and treat cancer and metastasis. Lgr5/GPR49, if also confirmed to be a CSC marker, holds enormous potential as a diagnostic and therapeutic target in GI and other cancers.


This work was supported by grants to TB from the EU (MCSC Contract No. 037297), the DFG (Grant No. BR 1399/4-3) and the Deutsche Krebshilfe (Grant No. 106958).

Teaching Materials

Power Point slides of the figures from this Review may be found in the supporting information.