Current views of the identity, distribution, and regulation of small intestinal epithelial stem cells and their immediate progeny are discussed. Recent works implicating Wnt signaling in stem and progenitor proliferation, the involvement of Notch signaling in epithelial lineage specification, and the role of hedgehog and bone morphogenetic protein families in crypt formation are integrated. We had the good fortune that many of these papers came in pairs from independent groups. We attempt to identify points of agreement, reinterpret each in the context of the other, and indicate directions for continued progress.
- crypt base columnar cell
the lumen of the small intestine is lined by a single layer of cells, the epithelium. Interaction with underlying mesenchyme forms this continuous sheet of cells into a regular series of folds, the crypts and villi. The villi project into the lumen. Surrounding the base of each villus are numerous crypts, test tube-shaped indentations of the epithelium whose hemispherical bases are set deep in the mesenchyme. Production and delivery of cells to the villus is a primary crypt function. Cell production, differentiation, and migration occur at a brisk pace, yet the intense activity is structured.
COMMON ORIGIN OF DIFFERENTIATION
Proliferation is not distributed throughout the crypt. Rather, proliferating cells are largely restricted to the lower three-fourths of the crypt, whereas cells in the crypt top are postmitotic differentiating cells enroute to the villus. There is a general flow of cells from the lower toward the upper crypt and, as a result, placement along the crypt axis is a rough measure of cell age and stage of differentiation. It is as though a compass points from the crypt base toward the villus, directing cell movement and maturation processes. There are notable exceptions to this general pattern. Paneth cells, a secretory lineage, are made in pairs and die in the crypt base. It is as though for them the compass points in the opposite direction because Paneth cells migrate downward into the crypt base from their site of origin (3) in an EphB3-dependent manner (24). As a result, the crypt base also displays a differentiation gradient, but this gradient increases toward the crypt bottom. The two differentiation countergradients have a common origin just above the crypt base at approximately cell position 5 (3, 4).
CLASSIC MODEL OF STEM CELL LOCALIZATION
For over a century, the crypt has been viewed as a column of proliferative cells bounded from below by the Paneth cells and above by the terminally differentiating cells. As a result, the crypt base was usually ignored because most thought that it contained only Paneth cells. This view was so influential that cell position 1 in the crypt was originally defined to be the first cell above the Paneth cells. It was only years later that position 1 was redefined to be the lowest cell in the crypt. By the late 1950s, the idea was crystallized that stem cells, if they exist, would most likely sit at the base of the proliferative column immediately above the Paneth cells. This concept was first clearly expressed, but immediately dismissed in favor of the equivalence of all proliferative cells, by Quastler and Sherman (20a). The idea that stem cells sit just above the Paneth cells was soon resurrected by Cairnie et al. (5) and refined over the years into its current form (20) in which the stem cells are found among the ring of cells immediately above the Paneth cell, in approximately position 4, where they occupy a stem cell niche.
THERE ARE IMMATURE CELLS AMONG THE PANETH CELLS
It is important to correct the impression that the crypt base is occupied only by Paneth cells. By the late 1960s, it became clear that Paneth cells are not the only occupant of the crypt base. In fact, immature proliferating columnar cells are found in the crypt base (Fig. 1). The immature columnar cells in the base are more radiosensitive and are more actively phagocytic than are cells higher up (6). Furthermore, cycling cells in positions 1–4 cycle more slowly and do not display the circadian rhythm evident in cells in higher positions, suggesting that they represent distinct cell populations. It was originally proposed that the crypt base columnar cells were Paneth cell precursors. However, there are too many of them, and they produce too many offspring to be limited to Paneth cell production, so it was concluded that they likely include stem cells.
In the classic model, stem cells sit at the base of the column of proliferative cells just above the Paneth cells. But the proliferative column is not necessarily bounded from below by Paneth cells. Hence, the logic used suggests that the stem cells often lie deeper and are not limited to a ring of cells above the Paneth cells. Furthermore, many mammalian species have no Paneth cells and even in the mouse ∼4% of crypts in the duodenum contain no Paneth cells. Accordingly, it seems unlikely that stem cell placement is directly dependent on Paneth cell localization as is proposed in the classic scheme.
STEM CELL ZONE MODEL
In the late 1970s, we proposed that the microenvironment of cell positions 1–4 is supportive of the stem cell state, but stem cells moving up to position 5 or above are induced to commence a differentiation program (3, 4). Most of their offspring supply cells to the villus, but Paneth cells and a subset of other cell types migrate back down into the stem cell zone. Thus the postmitotic cells found in the crypt base do not arise from in situ differentiation from resident precursor progenitors but arrive there by a process of downward migration from the common origin. Subsequent work indicates that a variety of early committed progenitors situated in or above position 5 produce the differentiating cells that migrate either up or down from the common origin (2).
MOLECULAR DETERMINANTS OF STEM CELL STATE AND PROGENITOR ACTIVITY
The stem cell zone model raises a number of questions. What allows stem cells to persist in the crypt base: are they autonomous or is the microenvironment actively supportive? Why is initiation of differentiation programs largely restricted to position 5 and above? There are no definitive answers to these questions, but researchers are getting close.
The highly conserved Wnt family of secreted molecules (Fig. 2) are key regulators of progenitor behavior in the intestinal epithelium. The Wnt signal transduction pathway and the effects of its perturbation have been extensively reviewed by others. We will focus instead on the interpretation of some recent results. For the purposes of this discussion, it is helpful to view Wnts as requisite growth factors for the epithelial progenitors, although they have unrelated functions in mature cells as well. It now seems clear that Wnt signaling is necessary for driving proliferative activity, but recent results indicate that stem cell survival may be independent of Wnt signaling, at least in the short term.
EFFECTS OF DECREASED WNT SIGNALING
Dickkopf is a secreted protein that suppresses Wnt signaling by binding to and inducing the endocytosis of a component of the Wnt receptor complex (Fig. 2, C and D). Transgenic mice expressing Dickkopf in the intestine under control of a villin promoter have regions of intestine with dramatically reduced proliferative activity, absence of crypts, and fewer and stunted villi (19). Similarly, infection of adult mice with a Dickkopf-expressing adenovirus results in suppression of proliferative activity, rapid loss of crypts, and stunted villi (16). Thus the studies agree on the main point: proliferation in the crypt requires Wnt signaling. There are also discrepancies between the two studies. Most adenovirus-infected mice die from colitis and systemic infection by day 10, resulting from the Dickkopf-mediated epithelial collapse. In contrast, the transgenic mice survive into adulthood, raising the question of how areas depleted of crypts and progenitors avoid epithelial denudement and ulceration leading to death? A clue is provided by the few adenovirus-infected mice surviving beyond 10 days, because they displayed regeneration of the epithelium. Presumably, in these surviving mice, the virus is cleared, Dickkopf levels drop, Wnt signaling recovers, residual stem cells resume proliferation, crypts regenerate, and the epithelium is restored. Could survival of transgenic mice be due to a similar but recurring dynamic process?
A Dickkopf-mediated delayed negative feedback from the villus population to the crypt progenitors could cause local oscillations in the size of the epithelial population in the transgenic mice. Recall that the villin promoter is normally more active in differentiated villus cells than in crypt progenitors. Hence the use of a villin promoter to drive Dickkopf production in the transgenic mice likely introduces a delay before the Dickkopf produced by new cohorts of cells suppresses proliferation (the cells in each cohort must differentiate before making significant amounts of Dickkopf). It is well known mathematically that such delayed negative feedback can lead to oscillation. In this example, transgenic Dickkopf production suppresses local progenitor activity in the crypt (Fig. 3A), leading to depletion of crypts and then to villus stunting (Fig. 3, B and C). Local Dickkopf levels subside as the epithelium is depleted (Fig. 3C). As a result, Wnt signaling resumes, leading to regrowth of crypts and replenishing of villi (Fig. 3, A and D). Consequently, local Dickkopf production increases again, leading to suppression of Wnt signaling, and the cycle repeats, resulting in an oscillation in the size of crypts and villi as local Dickkopf levels cycle from high to low.
Local Dickkopf-driven oscillations in crypt and villus populations would help to explain the survival of the transgenic mice, despite the fact that their intestines had patches of severely affected epithelium (no apparent crypts and few stunted villi) interspersed with less affected patches (villi and a few dwarfed crypts). The authors suggested that the patchiness was due to mosaicism that may be inherent in the villin promoter used (19). In addition, we speculate that these patches are not static but rather reflect continuous cycles of Dickkopf-induced suppression of Wnt signaling leading to local epithelial collapse followed by decreased Dickkopf levels and subsequent epithelial regrowth (Fig. 3).
These two studies imply that Wnt signaling is essential for normal proliferative activity of the progenitors and that in its absence the progenitors differentiate and migrate to the villus. Stem cell proliferation also requires Wnt signaling; however, their survival and maintenance in the stem cell state seem to be Wnt independent, because stem cells survive and are able to repopulate the crypt as Dickkopf levels subside in mice recovering from virus infection (16). It should be noted, however, that our conclusion of Wnt independence is based on the assumption that Dickkopf abolishes all Wnt signaling. It is possible that low-level autonomous Wnt signaling sufficient for stem cell survival, but not for proliferation, might persist. Identification of the molecular determinants of stem cell survival and maintenance in the stem cell state is an important goal for the future.
Many colorectal cancer therapies target proliferating cells. Consequent intestinal damage is an important constraint that might be reduced, if it were possible to selectively and transiently suppress intestinal stem cell proliferation during therapy. One such approach might be suppression of Wnt signaling with a Dickkopf-like compound before treatment. Because most colorectal cancers are Wnt signaling independent [e.g., due to loss of adenomatous polyposis coli (APC) function], such an approach might allow the desired selectivity.
EFFECTS OF INCREASED WNT SIGNALING
Thus far, we have considered only the effects of diminished Wnt signaling. The effects of increased Wnt signaling may be modeled indirectly by increasing the levels of nuclear β-catenin, a key molecule in the canonical Wnt signaling pathway. APC is involved in targeting β-catenin for inactivation. Loss of APC function leads to accumulation of nuclear β-catenin (Fig. 2, A and B). Malfunction of this pathway is a key step in many cases of colorectal tumorigenesis and has been extensively reviewed. We limit our attention to a pair of recent studies in which Apc function was conditionally disrupted in adult mice (1, 22). In both studies, it was found that shortly after Apc loss was induced, the proliferation compartment expanded, differentiation was suppressed, and cell migration was retarded. In the villus, no morphological changes were evident, and the cells remained postmitotic despite the fact that the Wnt signaling target gene cyclin D1 was strongly induced in villus cells (1). This result is important because it demonstrates that postmitotic cells are not targets for the APC pathway to tumorigenesis. Rather, it is the proliferative compartment that is at risk from APC mutations. Unfortunately, these studies cast little light on the relative sensitivity of various progenitors to the effects of APC loss. Do all progenitors respond similarly, or are the stem cells and primitive progenitors more sensitive to the effects of APC deletion?
EphB3 expression is normally restricted to crypt base cells, and is involved in Paneth cell localization to the crypt base (24). Following induction of APC loss, the zone of EphB3 expression expands upward in parallel with the aberrant cells (22). In addition, cells expressing lysozyme, a Paneth cell marker, were found scattered throughout the aberrant region, suggesting an expansion of the effective crypt base (1, 22). An intriguing possibility is that the actively proliferating cells in this region are stem cells whose population expansion has been driven by the increased levels of nuclear β-catenin. Alternatively, increased nuclear β-catenin stimulates EphB3 expression in all progenitors and the aberrant cells are derived from a mixture of stem and progenitor cells. Experiments targeting Cre recombinase to the various committed progenitors might allow identification of the susceptible progenitor types.
A host of other molecules have been found to influence progenitor activity including neuro- and immunoeffectors, hormones, and peptide factors, such as GLP-2, KGF, and EGF. Some seem to target specific lineage progenitors (e.g., KGF and GLP-2 target mucous and columnar progenitors, respectively), whereas others act more generally. The elucidation of the interplay between Wnt signaling and these modulators should be interesting.
MOLECULAR DETERMINANTS OF LINEAGE SPECIFICATION
It is known that stem cells give rise to progenitors that in turn produce the myriad of cell types present in the epithelium (2, 6), but the regulatory mechanisms are only vaguely understood. Even such fundamental issues as whether stem cell mitoses directly yield cells committed to a differentiation program (through asymmetric mitosis) or whether the respective fates of the daughter cells are driven by external cues through interaction with other cells, the substratum, or secreted factors remain open questions.
Various genetic and pharmacological evidence implicates Delta-Notch signaling in intestinal epithelial lineage specification. Delta-Notch signaling is a highly conserved mechanism involved in lineage specification. Members of the Notch receptor family can be activated on binding of ligands attached to the surface of neighboring cells. Notch activation leads to increased expression of members of the Hes gene family (transcription factors known to suppress the expression of various basic helix-loop-helix transcription factors involved in lineage specification), including Hes1.
Developers of a class of drugs aimed at treating Alzheimer's disease have been frustrated by serious gastrointestinal side effects that include overproduction of secretory cells at the expense of the columnar lineage (18). The effect is due to suppression of Notch signaling by the drugs that target γ-secretase (an enzyme involved in Notch signal transduction to the nucleus), resulting in a suppression of Hes1 expression. This serendipitous finding nicely compliments the observation that Hes1−/− mice also overproduce secretory cells (13) as do mutant zebrafish lacking Delta-Notch signaling (7). Furthermore, loss of Math1 (a transcription factor repressed by Hes1 expression) function results in absence of the secretory lineages in the fetal Math1−/− intestine (25). Thus it appears that Notch activation with consequent Hes1 expression leads to specification of columnar rather than secretory cell lineages.
In the fish intestine, the Notch ligand is expressed in secretory cells (7), and this is likely also the case in mice. Extrapolating from studies of other systems, at some stage secretory cells probably express high levels of Notch ligand and, hence, inhibit neighboring cells from pursuing a secretory fate by activating their Notch receptors, inducing Hes1 expression and enforcing columnar lineage specification. The lineage specification by Notch is likely to occur in early progenitors. If late columnar progenitors were dependent on continuous Notch activation from neighboring secretory cells for enforcement of their columnar fate, then we would expect to see closely spaced secretory cells on the villus (Fig. 4A). In fact, secretory cells are often separated by numerous columnar cells (Fig. 4B). Thus it seems likely that the relevant Notch signaling is occurring near the crypt base in early progenitors, for example among the offspring of the short-lived bipotential progenitor Mix (it produces short-lived mucus and columnar progenitors), a major early progenitor derived from the stem cells (2).
A simple quantitative argument makes the point. If epithelial cells are arranged on an approximate hexagonal lattice, and if contact with one secretory cell is sufficient to enforce the columnar fate, then we would expect to observe at least one-seventh, ∼14% secretory cells (Fig. 4A). In the proximal intestine, ∼4% of villus cells are secretory, indicating that the specifying lateral inhibition is not occurring in the villus. In the crypt, columnar progenitors divide perhaps four times before terminal differentiation, whereas secretory progenitors divide once or twice (2). As a result, the columnar population expands at least four times that of the secretory population as they move up the crypt. Thus if lineage specification through lateral inhibition occurs primarily among the earliest progenitors near the crypt base (for example in a collar of cells above the crypt base), then we would expect about (1/7)/4, or ∼4% secretory cells on the villus, and they would often be separated by large expanses of columnar cells.
The pattern of Hes1 expression seems consistent with this proposal. It is expressed in columnar cells, not in secretory cells, with high levels of expression in the columnar cells just above the crypt base (15). Furthermore, time course studies following inhibition of Notch signaling with γ-secretase inhibitors show a bolus of secretory cells appearing first in the crypt (18). Thus specification of the columnar vs. secretory fate via Notch signaling most likely occurs in the lower crypt.
The fact that Math1−/− mice have no secretory cells (25) is usually interpreted as indicating that all secretory lineages share a common Math1-dependent progenitor. Alternatively, because Math1 is expressed in all secretory lineages (19, 25), they may be separately dependent on Math1 either in early progenitors or downstream in their differentiation programs (Math1 is expressed in both mature secretory cells and scattered midcrypt cells). In this alternate mechanism, the absence of secretory cells is due not to a common Math1-dependent progenitor, but to the common dependence of the secretory lineages on Math1. This contrasts with the pattern reported for Neurogenin3, a transcription factor involved in enteroendocrine specification whose expression is restricted to presumptive progenitors (12). Furthermore, clones containing only mucous, enteroendocrine, or Paneth cells were seen, but clones of secretory cells containing all three lineages were not seen (2), suggesting that either the proposed common secretory progenitor does not exist, that it is exceedingly short lived, or that it usually produces only a single lineage type at a time.
MECHANISMS OF CRYPT FORMATION
Two crypt-generating mechanisms are known: crypt neogenesis and crypt branching. As villi form during development, proliferating cells are largely restricted to the intervillus clefts. Thereafter, well-defined foci, presumably containing stem cells, appear between the villi, and the tissues are remodeled around the foci resulting in the first crypts. Henceforth, proliferation is normally restricted to the crypts. This mode of crypt neogenesis does not seem to occur normally in the adult. Instead, new crypts result from crypt branching, a process thought to be related to expansion of the crypt stem cell population. We prefer the term “crypt branching” to the more commonly used “fission” because the process is an example of branching morphogenesis, an important and well-studied mode of epithelial morphogenesis. Crypts increase in size, buds or clefts appear (usually in the base), and then a rapid and progressive remodeling up the crypt results in the formation of new smaller crypts. Because the process repeats in each new crypt, it is usefully viewed as a crypt cycle (23) (Fig. 5).
Postnatally, the crypt cycle results in a rapid expansion in crypt numbers. The cycle slows down after weaning, but can be dramatically accelerated following injury and is an important component of mucosal repair. The crypt cycle also plays a role in clonal expansion of mutant stem cells (because the mutant stem cell population expands with increasing crypt numbers), which likely is important in early tumorigenesis. Crypt neogenesis and the crypt cycle may share common mechanisms because they both likely involve morphogenesis driven by the stem cell pool.
MOLECULAR DETERMINANTS OF CRYPT FORMATION
Hedgehog and bone morphogenetic protein (BMP) signaling have long been known to play a central role in regional gut patterning during development (21). More recent work has shifted the focus to their role in establishing the crypt-villus axis. Interference with hedgehog signaling during development perturbs crypt-villus morphogenesis. Fetal Ihh−/− (Indian hedgehog) mice have fewer and dramatically stunted villi and reduced numbers of proliferating cells in the intestinal epithelium, suggesting a role for hedgehog signaling in regulating progenitor behavior (21). Similarly, villus morphogenesis is severely stunted when hedgehog signaling is suppressed by high levels of transgenic expression of a secreted form of the hedgehog-binding protein Hhip, a panhedgehog signaling inhibitor (17). In contrast to the knockout result, the Hhip-expressing mice exhibit abundant proliferation, suggesting that the absence of villi is due to more subtle morphogenetic causes than a simple lack of epithelial cell production.
Epithelium-derived PDGF-A is also implicated in regulating the outgrowth of mesenchyme responsible for formation of villus cores during villus morphogenesis; however, the effects of abrogating PDGF-A stimulation are less dramatic than those associated with disruption of hedgehog signaling. Of interest here was the observation that BMP signaling from PDGF-A-stimulated mesenchymal cells was correlated with restriction of epithelial proliferation to the intervillus cleft (14).
Introduction of BMP signaling as a mechanism of suppression of epithelial proliferation is of heightened interest because aberrant BMP signaling is responsible for many cases of familial juvenile polyposis syndrome (JPS). Transgenic mice in which the BMP-4 binding protein Noggin was expressed under control of a villin promoter develop JPS-like polyps at 3 mo (9). Polyps were also found in mice several months after the BMP receptor Bmpr1a was inactivated using inducible Cre recombinase (10). These results confirm that suppression of BMP signaling results in aberrant growth, but the mechanism is unclear. In the Noggin-expressing transgenic mice, ectopic crypts were observed in villi at 4 wk of age but not earlier. They conclude that suppressing BMP signaling allows ectopic crypt neogenesis in the villus, leading to polyp formation (9). This interpretation raises many questions. What is the source of stem cells for the new crypts? If the stem cell source is epithelial, does this mean that in the absence of BMP signaling, villus cells can be induced to become stem cells or that stem cells migrate up from the crypt to seed the villus? Alternatively, if the stem cell source is extraepithelial, why do they seed into the villus? They report that the ectopic crypts appear normal. Does this mean that the wayward stem cells induce entirely new coordinate systems in the villus? Crypt formation normally begins shortly after birth. By 4 wk, the intestine has acquired its adult form. Why does the Noggin-induced ectopic crypt neogenesis occur only in 4-wk-old mice and not earlier when normal crypt neogenesis is at its peak? Are weaning-associated signals involved in triggering the ectopic crypt neogenesis (weaning normally occurs at 3–4 wk)? Finally, given the proposed role of BMPs in early villus morphogenesis, why is villus morphology unaffected until the ectopic crypt neogenesis commences at 4 wk?
Ectopic crypt neogenesis on the villus was also observed in mice expressing relatively low levels of transgenic Hhip during development, implicating hedgehog signaling in crypt neogenesis (17). Because hedgehog is known to induce BMP signaling, these authors also constructed a transgenic mouse expressing Noggin under control of a villin promoter but, like the Clevers group (9), did not observe any abnormalities during development. Thus, on the one hand, we have evidence that BMP signaling is not involved in normal crypt neogenesis during development, yet it appears that after weaning, aberrant crypt neogenesis is induced by suppression of the BMP pathway.
An accelerated crypt cycle following disruption of BMP signaling in the adult has been proposed as an alternate mechanism leading to JPS-like polyps (10). They did not report results from mice killed at earlier stages, so it remains to be determined whether ectopic crypt neogenesis in the villus also occurs in their mice or whether polyp formation is due entirely to aberrant crypt cycling.
Crypt neogenesis and crypt cycling are also relevant to mucosal repair following injury. For example, 13 days following irradiation of a loop of gut with 1,700 rads, scattered macroscopic nodules containing numerous crypt-like structures regenerate, presumably from the few surviving stem cells (11). Such regenerating crypts exhibit numerous buds and are likely very actively cycling. Sections through these nodules appear to show ectopically placed crypts (11), giving them a superficial resemblance to the abnormal epithelium in Noggin transgenic mice. This leads us to ask whether many of the ectopic crypts observed in Noggin transgenic mice might be due to a piling up of crypts that results from rapid and irregular crypt branching, perhaps due to the onslaught of stimulation during weaning in the absence of BMP-suppression of proliferation. In Drosophila, loss of BMP-like signaling causes aberrant epithelial morphogenesis (8), suggesting that morphogenetic defects, and not just proliferative effects, play a role.
STEM CELL MARKERS
Lack of specific markers for intestinal stem cells continues to hamper progress. The stem cells are most likely in the crypt base, but there are no reliable specific means to identify them. Mushasi-1, a neural stem cell marker, is expressed in immature cells in the lower crypt and may prove useful in combination with other tools (15). A more recent study (10) introduces a variety of specific stem cell markers using localization to positions 4–5 as evidence. Caution seems warranted, however, because enteroendocrine cells are common in the crypt base and, to our eyes, their examples have characteristic enteroendocrine cell morphology. Discovery of specific and reliable stem cell markers remains a priority.
Another priority should be the development of robust in vitro culture systems for the intestinal epithelium. In their absence, progress has been slow because the field has been largely dependent on in vivo experiments for progress.
- Copyright © 2005 the American Physiological Society