Although there is an expanding literature to support the existence of CSC, important caveats of these studies continue to provoke controversy and debate. The current definitive test of a CSC is the capacity to propagate tumors as xenografts in immunocompromised mice¹¹; however, it has been argued that species differences alone might account for the selective growth of subpopulations of cells in these assays. Indeed, the great majority of cells in a mouse lymphoma were shown recently to possess tumor initiating capacity when allografted into syngenic mice¹². Further controversy has dogged the use of stem cell surface markers to isolate CSC. Notable among these is PROM1, a five-transmembrane domain containing glycoprotein that is expressed on the surface of a variety of normal stem cells^13-15. CSC of brain^7,16, gut^3,4 and pancreatic¹⁷ tumors have been isolated using PROM1 antibodies; however, most PROM1 antibodies recognize glycosylation-dependent epitopes that vary with the differentiation and transformation status of the cell¹⁸, complicating the use of these reagents to fractionate stem cells. Without better understanding of normal tissue stem cells and their susceptibility to neoplastic transformation, it will be difficult to conduct definitive studies of the existence and origins of CSC.

To understand better the identity of normal tissue stem cells and their role in the cancer process, we generated a knock-in allele in which we integrated a creERT2-IRES-nLacZ cassette at the first ATG codon of Prom1 (Prom1^C-L); thereby creating a null allele (Figure 1a and Supplemental Figure 1). Prom1^C-L/C-L mice were born and aged normally, indicating that Prom1 is dispensable for gross embryonic and postnatal development. nLacZ expression in Prom1^+/C-L embryos that retained one copy of the wild-type allele, was restricted to the central nervous system, kidney, intestine and developing skeletal system (Figure 1b). The tissue distribution of Prom1 expression increased with subsequent development and was detected ultimately in several organs of adult Prom1^+/C-L mice (Figure 1c, Supplemental Figure 2 and Supplemental Table 1). Analysis of nLacZ and Prom1 protein expression from the modified and wild-type alleles, respectively, as well as tissue specific markers, detected Prom1 in Nestin⁺ cells of the ependymal layer and subventricular zone of the adult brain; Clara Cell Specific Protein (CCSP)⁺ cells at the bronchoalveloar junction in the lung; and cells lining the pancreatic ducts (Figure 1c). Each of these cell populations has been reported to include adult stem cells^19-22. nLacZ and Prom1 expression were also observed in numerous differentiated cells in Prom1^+/C-L mice, including proximal renal tubule cells, photoreceptors in the retina (both Figure 1c), neurons in the adult brain, acinar and islet cells in the pancreas, and goblet and columnar epithelial cells lining the colon (Supplemental Figure 2). This pattern of Prom1 expression is in agreement with recent studies of the Prom1 locus in mouse tissues²³, and indicates that Prom1 expression is not confined to stem cells, but is expressed also in multiple types of differentiated cells.

In contrast to most tissues, Prom1 expression was relatively restricted in the small intestine of Prom1^+/C-L adult mice. Less than three percent of cells covering the villi expressed nLacZ (n=82/3305), versus 26% (n=106/402) of cells at the crypt base (β-galactosidase staining, P<0.0001, Chi-square test, Figure 2a and Supplemental Figure 3). Prom1 mRNA expression was similarly distributed in the small intestine and overlapped closely with that of Lgr5, which was reported recently to mark intestinal stem cells² (Figure 2b and c, and Supplemental Figure 3). Therefore, as a first step to test if Prom1 might be expressed by Lgr5⁺ crypt stem cells, we performed concurrent Lgr5 in situ hybridization and nLacZ immunofluorescence analyses of adult Prom1^+/C-L mouse small intestine. Over 75% of Lgr5⁺ crypt cells contained nLacZ⁺ nuclei (n=390/500 cells counted), suggesting that Prom1⁺ crypt cells are predominantly stem cells (Figure 2d). Further, of the 4.1 LacZ positive nuclei (average, range 0 to 8 cells in 210 crypts counted) observed in each Prom1^+/C-L crypt, 23%, 30% and 17% were located at the 0, 1′ and 2′ cell positions, respectively, that are the sites most frequently occupied by Lgr5⁺ cells² (Figure 2e). Together, these data suggest strongly that Lgr5⁺ small intestinal stem cells express Prom1.

To test more directly if Prom1 marks stem cells in the small intestine, we used the Cre-dependent Rosa26-Yellow Fluorescence Protein (RosaYFP) reporter allele to trace the lineage of Prom1⁺ crypt cells. Two month old Prom1^+/C-L ; RosaYFP mice were treated with tamoxifen to activate CreERT2 expressed from the Prom1^C-L locus; thereby irreversibly activating YFP expression in Prom1⁺ cells and their progeny. Mice were sacrificed 2, 10 and 60 days following induction. Two days following induction, an average of 2.5 YFP-labelled cells was seen in 25% of small intestinal crypts (>50 labelled crypts counted; Figure 2f inset). By 10 days, trains of contiguous YFP⁺ cells were observed emanating in single file from the base of crypts in a manner identical to that reported for the progeny of Lgr5⁺ stem cells² (Figure 2g). YFP cells remained present at 60 days, extended to the tip of the villi, and included the four differentiated cell types of the intestinal lining: enterocytes, Paneth, goblet, and enteroendocrine cells (Figure 2f and Supplemental Figure 4). Thus, we conclude that the stem cells of the mouse small intestine are Prom1⁺. In contrast, immediately following induction, YFP-labelling in the colon of these mice displayed the same broad distribution as Prom1 expression, and gradually disappeared within 60 days. Therefore, Lgr5⁺/Prom1⁺ stem cells of the small intestine are distinct from Lgr5⁺/Prom1⁻ stem cells of the large intestine^2,23. YFP labelling of the brain, kidney, lung, and pancreas remained relatively static between 2 and 60 days post tamoxifen induction (Supplemental Figure 5 and data not shown). Since these organs have low rates of cell turn-over, it remains possible that some Prom1⁺ cells in these tissues are quiescent adult stem cells.

Mutations that activate aberrant WNT signaling, including mutations in Beta-Catenin (CTNNB1), have been identified in over 80% of human colonic cancers^5,24 and this tumor type was reported recently to contain Prom1⁺ CSC^3,4. Intestinal tumorigenesis caused by aberrant Wnt signaling has been modeled successfully in genetically engineered mice²⁵. Although tumors in these animals develop predominantly in the small intestine, the disease process mimics much of the cellular and molecular characteristics of human colonic polyposis^25-29. Therefore, to determine if intestinal tumors might arise directly from Prom1⁺ stem cells, we activated endogenous WNT signaling in two month old Prom1^+/C-L mice using a Cre-dependent mutant allele of Ctnnb1 (Ctnnb1^lox(ex3))²⁹ and traced the lineage of these cells using the Rosa26YFP allele. Two days following tamoxifen-induction, we observed a marked increase in the number of YFP⁺ cells at the base of small intestinal crypts, but at no other site above crypt cell position 4, in Prom1^+/C-L ; RosaYFP ; Ctnnb1^{+/ lox(ex3)} mice relative to Prom1^+/C-L ; RosaYFP ; Ctnnb1^+/+ control mice (Figure 3a). Two-photon laser scanning microscopy demonstrated that these cells had a slim and elongated morphology compatible with them being crypt stem cells² (Supplemental Movie 1). Counting of cells isolated from the entire small intestinal mucosae of mice two days following tamoxifen-induction confirmed the marked expansion in YFP⁺ cells in mice containing the Ctnnb1^lox(ex3) allele (P<0.0001, Wilcoxon test, Figure 3b). Remarkably, cultures of these cells yielded four times as many, and much larger, YFP⁺ clonogenic colonies when isolated from Prom1^+/C-L ; RosaYFP ; Ctnnb1^{+/ lox(ex3)} mice than those from control animals (P<0.0001, Wilcoxon test, Figure 3c and d). Thus, activation of Wnt signalling in small intestinal Prom1⁺ stem cells induces first a marked expansion of these cells in the crypt base.

Ten days following tamoxifen-induction, Prom1^+/C-L ; RosaYFP ; Ctnnb1^+/lox(ex3) crypts were markedly disorganized and contained cells that were uniformly nuclear Ctnnb1 immunoreactive, highly proliferative and disproportionately Prom1⁺ (nLacZ immunopositive) relative to controls (Figure 4a and b and Supplemental Figure 6a to c). Contiguous streams of YFP⁺ cells were now seen emanating from the crypt and migrating up the surface of the villi (Figure 4c). In contrast to Prom1^+/C-L ; RosaYFP ; Ctnnb1^+/+ control mice, cells emanating from the crypts of Prom1^+/C-L ; RosaYFP ; Ctnnb1^+/lox(ex3) mice were both hyperplastic and grossly dysplastic, forming a carpet of contiguous YFP⁺ cells from the base of the crypt to the tip of the villus (compare Figures ​Figures2g,2g, and ​and4c).4c). All Prom1^+/C-L ; RosaYFP ; Ctnnb1^+/lox(ex3) mice allowed to age following tamoxifen induction succumbed to their disease within 90 days (n=15/15). Sixty days following tamoxifen induction, the small intestine of Prom1^+/C-L ; RosaYFP ; Ctnnb1^+/lox(ex3) mice was twice the width of that of control animals (Supplemental Figure 6d). Gross inspection of the duodenal, jejunal and ileal lining revealed a thickened, rugous and YFP-fluorescent mucosa; no regions of normal tissue were identified (Supplemental Figure 6e and f). Microscopic analysis revealed loss of the normal villus architecture across the entire small intestine that was replaced by dysplastic tissue, characterized by focal high-grade intraepithelial neoplasia and crypt adenoma formation (Figure 4d and Supplemental Figure 6f). In contrast, no gross or microscopic abnormalities were seen in the colon which lacks Prom1⁺ stem cells (Supplemental Figure 6g). The absence of tumors in the colon in which Ctnnb1 was activated in Prom1⁺ differentiated cells, points to progenitor cells as the source of tumors in this tissue. All cells in the neoplastic small intestine of Prom1^+/C-L ; RosaYFP ; Ctnnb1^+/lox(ex3) mice expressed YFP and nuclear Ctnnb1 and therefore arose from within the Prom1-lineage; however, only 7%±15.5 (n=36/502 YFP⁺ cells counted from five separate regions) of tumor cells retained expression of Prom1 (Figure 4 e to g). Dual Ki67 and nLacZ co-immunoflourescence analysis of tumors demonstrated that approximately 10% of Prom1⁺ tumor cells were proliferating (Figure 4h). These data are compatible with the hypothesis that tumors in these animals contain a fraction of transformed stem cells.

Our data demonstrate that Prom1 marks stem cells in the small intestine and that activation of endogenous Wnt signaling in these cells disrupts normal tissue maintenance that begins first in the crypt, expanding aberrantly the Prom1⁺ stem cell population, resulting ultimately in neoplastic transformation of the small intestinal mucosa. Therefore, we show for the first time that Prom1 marks an adult solid tissue stem cell that is susceptible to neoplastic transformation, forming a model of a human tumor that contains Prom1⁺ CSC. Further analysis will determine if the fraction of Prom1⁺ cells observed in the neoplastic small intestine of Prom1^+/C-L ; RosaYFP ; Ctnnb1^+/lox(ex3) mice represents a remnant of mutated stem cells that may function as CSC. Our Prom1^C-L knock-in provides the research community with an extremely useful tool to explore further the relationship between normal and malignant stem cells in the lung, kidney, brain, pancreas and other tissues.

1