Tcf3 expression marks both stem and progenitor cells in multiple epithelia

Tcf3 (also known as Tcf7-like 1 or Tcf7l1) is a member of the Lef/Tcf family, a group of transcription factors with important roles in development, stem cell homeostasis and malignancy. Tcf3 is a crucial regulator of embryonic stem cell function, where it acts as an inhibitory regulator of the Oct/Sox/Nanog pluripotency circuit (Cole et al., 2008; Pereira et al., 2006; Yi et al., 2008,, 2011). Tcf3 also plays a key role in patterning and cell fate specification during early embryonic development (Cole et al., 2008; Dorsky et al., 2003; Houston et al., 2002; Kim et al., 2000; Merrill et al., 2004; Pereira et al., 2006; Yi et al., 2008,, 2011). Later in development, Tcf3 is involved in maintenance and specification of progenitor cells in the central nervous system (Kim and Dorsky, 2011; Kim et al., 2011). Emerging evidence also implicates Tcf3 in the pathogenesis of several types of human cancer (Ben-Porath et al., 2008; Cole et al., 2008; Pereira et al., 2006; Slyper et al., 2012; Yi et al., 2008,, 2011).

In the mammalian skin, Tcf3 is expressed throughout the primordial epithelium during development (Merrill et al., 2004; Nguyen et al., 2006), and in the adult skin it is expressed in the hair follicle bulge, a known stem cell niche (DasGupta and Fuchs, 1999; Merrill et al., 2001). Forced overexpression of Tcf3 in neonatal mouse skin blocks normal epithelial differentiation and causes epithelial cells to assume an undifferentiated, progenitor-like transcriptional state (Merrill et al., 2001; Nguyen et al., 2006). Conversely, deletion of both Tcf3 and its closely related paralogue Tcf4 in developing skin produces an epithelial ‘burnout’ phenotype, with transient hyperproliferation followed by a failure of long-term self-renewal (Nguyen et al., 2009). Based on these observations, Tcf3 has been presumed to act as a key mediator of a self-renewing undifferentiated state in skin stem cells. Despite considerable interest in the role of Tcf3 in stem cell homeostasis and malignancy, however, very little is known of its expression or function in normal adult tissues. Based on its importance in development and the results of our gain- and loss-of-function studies in skin, we hypothesized that Tcf3 might serve as a general regulator of stem cell function in adult tissues. In the present study, we set out to explore this hypothesis by examining the distribution and behavior of Tcf3-expressing cells in vivo.

We began by examining the expression pattern of Tcf3 in adult skin at different stages of the hair cycle. In telogen (resting-phase) skin, Tcf3 protein was detected throughout the hair follicle (HF) bulge, consistent with previous reports (Fig. 1A) (DasGupta and Fuchs, 1999; Merrill et al., 2001). During the anagen (growth) phase, Tcf3 expression levels were markedly increased in the HF bulge, and still higher levels were noted in cells of the outer root sheath (ORS) (Fig. 1B,C). In the bulb of the growing hair, there was a sharply defined separation between Tcf3-positive non-proliferating cells in the ORS and Lef1-positive highly proliferative cells in the transit-amplifying hair matrix (Fig. 1D,E).

To determine whether Tcf3 is a stem cell marker, we sought to observe the fate of Tcf3-expressing cells in vivo using Cre/loxP-based lineage tracing. We generated a knock-in mouse line, Tcf3-2A-eGFP-2A-CreER^T2 (hereafter Tcf3^GC), in which eGFP and CreER^T2 reporter genes were appended to the 3′ end of the native Tcf3 open reading frame (ORF) using viral 2A ‘self-processing’ peptides (Fig. 1F) (Szymczak et al., 2004). This strategy was designed to preserve all endogenous regulatory elements and to maintain a functional Tcf3 ORF. Generation of the Tcf3 knock-in mouse is described in the methods in the supplementary materials. Targeting of the knock-in construct (supplementary material Fig. S1A) was verified by Southern blotting (supplementary material Fig. S1B); subsequent genotyping was performed by PCR (supplementary material Fig. S1C). In subsequent crosses, heterozygous (Tcf3^GC/+) and homozygous (Tcf3^GC/GC) knock-in mice were born in Mendelian ratios. They were viable, fertile, and bore no obvious phenotype (supplementary material Fig. S1D), suggesting that the knock-in allele preserved normal Tcf3 gene function. In isolated skin keratinocytes from newborn mice, Tcf3 protein was detected at near wild-type levels in both heterozygous and homozygous knock-in animals, with no evidence of ‘read-through’ (incomplete processing) of the 2A peptides (supplementary material Fig. S1E). The knock-in GFP reporter was too dim for direct microscopy or cell-sorting applications, but it was easily detectable by immunostaining and colocalized with endogenous Tcf3 in neonatal back skins (Fig. 1G-I). In adult knock-in mice, the expression pattern of the knock-in GFP marker mirrored that of Tcf3 in both telogen- and anagen-phase back skins (Fig. 1J-R) as well as the other epithelial tissues we subsequently studied (supplementary material Fig. S1F-N; compare Fig. 4A,B, Fig. 5A,B and Fig. 6A,B).

For lineage-tracing experiments, we crossed Tcf3^GC with ROSA26^mTmG mice (Muzumdar et al., 2007), which express membrane-bound GFP (mGFP) following Cre activation (Fig. 1S). By treating Tcf3^GC/+; ROSA26^mTmG/+ mice with tamoxifen, we were able to mark Tcf3-expressing cells with the genetically encoded mGFP label, allowing us to monitor the fate of Tcf3-expressing cells and their progeny in vivo. After inducing these mice with tamoxifen, we observed mGFP(+) cells in Tcf3-expressing cell types throughout the skin, including the hair follicle bulge (Fig. 1T, arrowheads). CreER activity was very tightly regulated, with not a single mGFP(+) cell appearing in vehicle-treated control mice in any of the tissues we studied (n>20 across all timepoints) (Fig. 1U; supplementary material Fig. S2P,R,T). Based on these observations, we concluded that our novel Tcf3^GC allele is a faithful, tightly regulated reporter of endogenous Tcf3 expression in vivo.

Using the Tcf3^GC/ROSA26^mTmG lineage-tracing system, we first examined the fate of Tcf3-expressing cells in murine dorsal skin. We treated Tcf3^GC/+;ROSA26^mTmG/+ mice (n=5-7 per timepoint) with tamoxifen on postnatal days 21 and 22, during the first telogen (resting) phase of the hair cycle. Three days after treatment, we observed labeled cells in the HF bulge and hair germ (Fig. 2A,B). We also noted Tcf3 reporter activity in various cell types within the dermis, including (notably) the dermal papilla (Fig. 2B) and dermal blood vessels (supplementary material Fig. S2A,B).

During the succeeding anagen phase, labeled Tcf3-lineage clones expanded and incorporated into all differentiated lineages of the hair shaft (Fig. 2C-H). Six months after induction, persistent clones were seen in both telogen- and anagen-phase follicles (Fig. 2I-K). Similar results were obtained when lineage traces were initiated during the second postnatal telogen (supplementary material Fig. S2C-E). There was a moderate, nonsignificant reduction (68.4±3.7% of initial, mean±s.d.; P=0.42, Mann–Whitney U-test) in the number of labeled follicles from 3 days to 6 months post induction (supplementary material Fig. S2F), consistent with a model in which most or all HF bulge cells possess stem cell potential but are occasionally lost from the stem cell niche, as previously described by others (Hsu et al., 2011; Zhang et al., 2010).

We became curious about the fate of Tcf3-expressing cells during anagen, when Tcf3 is highly expressed throughout the outer root sheath (ORS) of the growing hair. We treated Tcf3^GC/+; ROSA26^mTmG/+ mice (n=4-6) with tamoxifen during anagen (typically at P30) and sacrificed them 3 days or 4 weeks after induction. Even with a single dose of tamoxifen, nearly all hair follicles contained multiple mGFP(+) clones in the ORS 3 days after induction (Fig. 2L), some of which were captured in transition between the ORS and matrix and thus appeared to be committing to terminal differentiation (supplementary material Fig. S2G). Labeled cells were also observed in the HF bulge at a relatively low frequency (16.5±7.2%, mean±s.d.) (Fig. 2M). Four weeks after induction, when mice had returned to telogen, a much larger proportion (70.3±16.6%, mean±s.d.; P=0.0095, Mann–Whitney U-test) of HFs contained one or more labeled cells in the bulge region, implying that Tcf3-lineage cells in the ORS survive the catagen (regression) phase to re-incorporate into the newly formed bulge (Fig. 2N,O). Cells within the new bulge have been shown to differ in their functions: cells in the outer layer of the bulge are true stem cells, while inner-layer cells serve a regulatory function and lack stem cell potential (Hsu et al., 2011). In our experiment, ORS-derived Tcf3-lineage cells incorporated into both populations in roughly equal proportions [62.6±0.2% of labeled bulges contained mGFP(+) cells in the outer bulge] (Fig. 2O), indicating that Tcf3-expressing ORS cells are capable of assuming either fate.

In the course of these experiments, the Tcf3 lineage was never observed in the upper portions of the hair follicle, the sebaceous gland (SG) or the interfollicular epidermis (IFE). Previous reports have indicated that HF bulge-derived cells can incorporate into the IFE during epithelial wound repair (Brownell et al., 2011; Ito et al., 2005; Page et al., 2013). We examined this behavior in Tcf3-expressing cells by labeling HF bulge cells at higher frequency by injecting Tcf3^GC/GC homozygous lineage-tracing mice with tamoxifen (either two or three doses) starting at P21 and subjecting them to full-thickness wounding 5 days after the last dose. One week after wounding, mGFP(+) cells were observed migrating out of wound-adjacent hair follicles and incorporating into the wound epithelium, generally in the suprabasal layers (Fig. 3A,B, arrowheads). There was a trend towards non-persistence of these cells over time, however, with the majority of cells being lost from the wound epithelium by 3-4 weeks post wounding (Fig. 3C). Some persistent clones were observed, frequently in close association with wound-edge hair follicles (Fig. 3D,E). The total number of persistent cells 4 weeks post wounding was 22.8±10.4% (mean±s.d.) of the initial timepoint (this result approached but did not reach significance at P=0.09, Mann–Whitney U-test) (Fig. 3F). This observation is consistent with prior studies that have shown a long-term persistence rate of ∼20% of bulge-derived cells (Page et al., 2013). Bulge-derived Tcf3-lineage clones formed radial tracks within the wound epithelium similar to those previously described for K15, K19 and Lgr5 (Fig. 3G,H).

Because the Tcf3 lineage was strictly excluded from the stratified IFE in normal dorsal skin, we asked whether Tcf3 might be expressed in the plantar paw skin, which, like human palmar skin, lacks hair follicles. Tcf3 was expressed in the basal epithelium of the plantar paw skin; however, rather than being confined to a rare subpopulation of stem cells, it was expressed ubiquitously throughout the basal layer (Fig. 4A,B), most highly in the thick skin of the digital pads (Fig. 4C).

To determine the stem cell potential of Tcf3-expressing cells in the paw skin, we administered tamoxifen to Tcf3^GC/+;ROSA26^mTmG/+ mice at P21 and P22, and examined the skin at various timepoints thereafter (n=5-8 mice per timepoint). Three days after induction, we observed small clones of one or two cells in the basal layer in a pattern, consistent with our Tcf3 antibody staining (Fig. 4D). By 4 weeks post induction, the clones had expanded to occupy all layers of the stratified epithelium (Fig. 4I-L), and full-thickness clones continued to be seen at subsequent timepoints (Fig. 4E-H). A proportion (although not a majority) of these clones persisted for ≥6 months in vivo, suggesting that the cells in which they arose possessed stem cell potential. We also noted extensive labeling of hair follicles in the neighboring haired skin that, as in the dorsal skin, was confined to the hair follicle (Fig. 4M).

We then inquired whether Tcf3 might be expressed in nonkeratinized as well as in keratinized epithelia. A survey of reporter activity revealed Tcf3 expression in the esophageal and lingual mucosa (discussed below), but not in more distal regions of the GI tract. In the esophagus, as in the paw skin, Tcf3 was expressed throughout the epithelial basal layer (Fig. 5A,B) and in some suprabasal cells within the lower granular layer (supplementary material Fig. S3A).

To assess the stem cell potential of Tcf3-expressing basal cells, we performed lineage-tracing experiments as described for the paw skin, inducing mice (n=6-8 per timepoint) with tamoxifen at P21 and P22 and sacrificing at various timepoints thereafter. Three days after tamoxifen treatment, clones of 1-2 mGFP(+) cells were detected in both the epithelial basal layer (Fig. 5C) and in isolated suprabasal cells (supplementary material Fig. S3C), consistent with the expression pattern seen in immunostaining. Over time, the labeled basal cells gave rise to columns of differentiated cells occupying the full thickness of the epithelium (Fig. 5D-J). As in the paw skin, some clones persisted in vivo for ≥6 months (Fig. 5H), indicating that the cells in which they arose were self-renewing stem cells. As in the paw skin, however, the majority of Tcf3-lineage clones failed to persist over time.

Previous studies have examined the self-renewal dynamics of the esophagus and paw skin; it was concluded that both tissues followed a neutral-competition homeostasis in which all basal cells have equivalent potential for long-term persistence and there is no defined subpopulation of stem cells (Doupé et al., 2012; Lim et al., 2013). Conversely, other studies have suggested the existence of a stem cell hierarchy in the interfollicular back skin (Mascré et al., 2012). Although the experiments described here were not intended to address mechanisms of homeostasis, we became curious about whether our results were compatible with a neutral competition model in which all Tcf3-expressing basal cells possess an equal chance of long-term persistence. For both tissues, we quantitated the number of mGFP(+) clones, the number of labeled basal cells within each clone and the total number of mGFP(+) basal cells in multiple random cryosections at each timepoint.

In both tissues, the majority of Tcf3-lineage clones were lost over time, with only 9.9±8.7% (paw skin) or 17.0±6.6% (esophagus) of initial clones persisting at 6 months post-induction (both mean±s.d., both P<0.001, Dunn's multiple comparison test) (supplementary material Fig. S4A,D). In both tissues, the surviving clones expanded laterally through the basal layer over time (P<0.0001 for both tissues, Kruskal–Wallis test) (supplementary material Fig. S4B,E). A key prediction of the neutral competition model is that the total number of labeled basal cells remains constant over time, as the loss of some clones is exactly balanced by the expansion of surviving clones. Interestingly, there appeared to be a general decline in the total number of basal cells over time in both tissues, though this only reached statistical significance when comparing the 3-day and 6-month timepoints (P<0.01 for paw skin, P<0.05 for esophagus, Dunn's multiple comparison test) (supplementary material Fig. S4C,F). Although this result could be interpreted to mean that Tcf3-expressing cells possess a slightly reduced capacity for long-term (>6 months) self-renewal, it is also possible that it may simply reflect under-sampling of clones at the later timepoints due to the low initial labeling frequency. Further experiments are necessary to clarify this issue, but in any event these data are incompatible with a model in which Tcf3 expression labels a master stem cell population that is preferentially fated to long-term survival.

The murine lingual epithelium is considerably more complex than the esophagus and possesses specialized appendages incorporating various differentiated cell types. The most numerous of these are the filliform papillae (FPs), hook-shaped structures that occupy the majority of the dorsal epithelial surface. Each FP consists of distinct anterior and posterior elements that are themselves distinct from the epithelium of the interpapillary pits (IPPs) (Hume and Potten, 1976). Recent reports have described a rare population of Bmi1-expressing epithelial stem cells that self-renew for extended periods and give rise to all differentiated lineages of the tongue in vivo (Tanaka et al., 2013).

We first examined the expression pattern of Tcf3 in the lingual epithelium by immunostaining. Tcf3 was expressed at varying levels throughout the epithelial basal layer, most highly in the base of the interpapillary pits (Fig. 6A,B), and also at moderate levels in some suprabasal cells (supplementary material Fig. S3B). This contrasted sharply with the previously described stem cell marker Bmi1, which occurred at a frequency of ∼1 cell/IPP (Tanaka et al., 2013). To examine the relationship between the Tcf3 and Bmi1 populations, we performed short-term (2 day) lineage traces in Bmi1^CreER/+; ROSA26^mTmG/+ mice (Sangiorgi and Capecchi, 2008). The Bmi1-expressing population was identified at a frequency of ∼1 cell/IPP, consistent with the observations of Tanaka and colleagues. Interestingly, some of these cells expressed Tcf3 at levels similar to neighboring basal cells (supplementary material Fig. S5A, arrowheads), whereas others expressed little to no Tcf3 (supplementary material Fig. S5B, arrowheads), suggesting that expression of the two factors is not directly related. Similarly, there was little correlation between Tcf3 expression and proliferative status, with cells variously expressing Tcf3 or Ki67, or both, at varying levels (supplementary material Fig. S5C).

To assess the stem cell potential of Tcf3-expressing cells in the lingual epithelium, we performed lineage-tracing experiments according to the same protocol used for the esophagus and paw skin (n=7-9 mice per timepoint). Three days after induction, we observed small clones of ∼2-4 labeled cells in the basal layer (Fig. 6C), as well as solitary labeled suprabasal cells (supplementary material Fig. S3D). Over the subsequent weeks and months, the basal clones continued to expand (Fig. 6D-G). Immunostaining at the 4-week timepoint demonstrated incorporation of Tcf3-lineage cells into multiple differentiated cell lineages in the tongue, including both the interpapillary epithelium (labeled by K13 and loricrin) and the filliform papillae (labeled by filaggrin and trichohyalin) (Fig. 6H-K). As in the esophagus and paw skin, the overwhelming majority of Tcf3-lineage cells failed to persist over time, suggesting that the majority of the Tcf3-expressing population is not fated to long-term persistence: the clone frequency at 6 months was only ∼10% of the initial timepoint (Fig. 6L). A minority of Tcf3-lineage clones did persist for ≥6 months in vivo and thus presumably arose from true stem cells.

Interestingly, those rare clones that did persist to the 6-month timepoint nearly always occupied the entirety of an interpapillary pit (IPP) and contributed to part of each of the neighboring 3-4 filliform papillae (Fig. 6M-O). This observation suggests that IPPs are separately maintained units of self-renewal and that each FP is maintained as a mosaic with contributions from at least three distinct individual stem cells, consistent with the prior results of Tanaka et al. (2013). Given the relatively low number of clones observed at this timepoint, however, we cannot rule out the possibility that there is some sharing of cells between adjacent IPPs, especially over a longer (>6 months) timeframe.

Numerous studies employing label-retention and genetic fate-mapping techniques have demonstrated that hair follicle bulge cells are self-renewing stem cells that are responsible for maintenance of the hair follicle through its cycles of growth, regression and rest, but do not contribute significantly to the IFE during normal homeostasis (Cotsarelis et al., 1990; Ito et al., 2005; Jaks et al., 2008; Morris et al., 2004; Tumbar et al., 2004). Our results from lineage tracing of telogen-phase skin are consistent with this established consensus. There is some disagreement among previous studies regarding the capacity of HF bulge cells to contribute to the sebaceous gland; fate-mapping experiments have demonstrated a contribution of K15-lineage cells to the SG during normal homeostasis (Morris et al., 2004; Petersson et al., 2011), whereas Lgr5-lineage cells do not appear to incorporate into the SG, at least in the absence of injury (Jaks et al., 2008). Consistent with the previous Lgr5 study, we never observed labeled cells in the SG in the course of our experiments, suggesting that the Tcf3 lineage is normally excluded from the SG. However, given the relatively low labeling efficiency achieved in our experiments, we cannot rule out the possibility that cells expressing low levels of Tcf3 do contribute to the SG under some circumstances.

Our wound-repair experiments are also consistent with prior observations that HF bulge cells are mobilized into the IFE during epidermal wound repair but mostly (∼60-90%) fail to persist over time (Ito et al., 2005; Page et al., 2013). Given our comparatively low labeling frequency and the low persistence of Tcf3-lineage cells in the wound epithelium, however, we cannot rule out the possibility that Tcf3 expression preferentially marks a subset of bulge cells with a reduced capacity to assume the IFE SC identity. There is some precedent for the idea that subsets of bulge cells differ in their capacity to persist in IFE, as the majority of persistent cells appear to arise from a Gli1-expressing subpopulation in the upper bulge (Brownell et al., 2011).

Of particular interest are our results from fate mapping of Tcf3-expressing cells in the anagen-phase ORS, which appear to include a mixture of both stem and progenitor cells. At the initial timepoint, clones were observed in transition between the ORS and matrix, implying that they were committing to differentiation. However, a proportion of labeled ORS cells did survive beyond the end of anagen to re-incorporate into both the inner and outer layers of the newly formed HF bulge. These observations independently confirm prior studies that used label-retention methods to demonstrate the persistence of a defined subset of ORS cells (Hsu et al., 2011); they also suggest that Tcf3 expression is not exclusive to stem cells, at least during the anagen phase. It is interesting to note that, in both telogen- and anagen-induction lineage traces, the behavior of Tcf3-expressing cells closely resembled that previously described for Lgr5-expressing cells (Jaks et al., 2008), and it seems likely that – at least in the dorsal skin – the two factors label the same population of cells.

Having observed that Tcf3-lineage cells were strictly excluded from the stratified IFE in normal dorsal skin, we were surprised to observe that Tcf3 was broadly expressed in the basal layer of several other stratified epithelia (paw skin, esophagus and tongue). In all three cases, a proportion of Tcf3-lineage clones persisted for extended periods of time in vivo and thus originated in self-renewing stem cells, but the majority of Tcf3-lineage clones failed to persist and were lost over time. Both the broad expression of Tcf3 and non-persistence of Tcf3-expressing cells were surprising in light of our initial hypothesis that Tcf3 expression would label defined, self-renewing stem cell populations.

The contrast with our expectations was especially marked in the tongue, where the majority of labeled clones were lost within the first few weeks after lineage tracing induction, implying that the cells in which they arose were short-lived committed progenitors. Over time, the Tcf3 lineage-tracing results converged with that previously seen for Bmi1-expressing SCs, with persistent Tcf3-lineage clones occupying the entirety of an IPP and contributing to the three neighboring filliform papillae. Importantly, the relative survival of Bmi1-lineage clones was not quantified in the previous study (Tanaka et al., 2013), and it is therefore uncertain whether the Bmi1 population represent classical stem cells at the apex of a stem-progenitor hierarchy or, alternately, if the Bmi1 population is in competitive homeostasis with other populations of self-renewing cells. In the former case, the persistent Tcf3-lineage clones in our study would presumably have originated in Bmi1-expressing cells. Although we cannot distinguish between these possibilities, we do conclude that Tcf3 expression in the lingual epithelium encompasses a wide range of stem, progenitor and committed cells.

The esophageal mucosa and plantar paw skin have recently been characterized as following a neutral-competition model of homeostasis, in which all basal cells have an equal probability of long-term persistence and undergo stochastic cell-fate decisions (Doupé et al., 2012; Lim et al., 2013). According to such a model, there is no stem-progenitor hierarchy and all basal cells are arguably ‘stem cells’. Progressive clone loss is an inherent feature of such systems and the observed non-persistence of Tcf3-lineage clones over time would thus be a predictable outcome. Even under the loose definition of ‘stemness’ posited by the neutral competition model, however, Tcf3 expression is not restricted to stem cells, as it is also observed in differentiating suprabasal cells that are rapidly lost from the epithelium.

Interestingly, quantitation of our lineage-tracing data suggests a decline in the total number of labeled basal cells over time (supplementary material Fig. S4), implying that Tcf3 expression labels a subset of basal cells with reduced capacity for long-term persistence. This would suggest at least a partial stem-progenitor hierarchy in these tissues and would thus conflict with the established neutral competition model. It is important to note, however, that these findings were derived from a post hoc analysis of experiments that were not originally designed to address this issue. There are also numerous differences between our methods and those of previous groups; for example, the previous work of Lim et al. (2013) analyzed only the region of the plantar epithelium lacking eccrine sweat glands (which, notably, expresses Tcf3 at very low levels), whereas our analysis included the entire plantar paw skin. Although further experiments are clearly called for in order to evaluate this fascinating discrepancy, it does not fundamentally alter our conclusion that Tcf3 is not a sole determinant of stem cell identity.

Based on the expression pattern of Tcf3 and our previous gain- and loss-of-function results (Merrill et al., 2001; Nguyen et al., 2006,, 2009), we anticipated that Tcf3 expression would serve as a marker of defined, self-renewing tissue stem cell populations in various tissues. Although the behavior of Tcf3-expressing hair follicle bulge cells was generally consistent with this hypothesis, we were surprised to observe that in the tissues we studied, Tcf3 was expressed not only in long-term persistent stem cells but also in short-lived progenitors and, in some cases, in cells undergoing early differentiation. These findings suggest that physiological Tcf3 expression alone is insufficient to maintain a cell in an undifferentiated self-renewing state in vivo and imply a more complex role for Tcf3 in epithelial homeostasis than previously appreciated. Recent studies in the hair follicle have demonstrated that Tcf3 functions as part of a complex ‘rheostat’ that, in competition with the canonical Wnt/β-catenin pathway, determines stem cell behavior (Lien et al., 2014). It seems likely that a similar mechanism, in which Tcf3 is a contributor to but not the sole determinant of cell fate, holds in the other tissues we studied. It is also important to note emerging evidence that Tcf3 plays a role in processes beyond stem-cell fate decisions: recent studies by our group have demonstrated that Tcf3 acts as a key regulator of cell migration within the wound-repair microenvironment (Miao et al., 2014). This is particularly pertinent in light of our incidental observation here that Tcf3 is expressed in a variety of non-epithelial cell types, sometimes at very high levels.

Given our findings and the observations discussed above, it is likely that Tcf3 has varying, context-dependent functions in developing and adult tissues, and in different cell types within the adult organism, probably due to the presence or absence of additional co-factors such as Groucho/TLE proteins. Findings concerning the functions of Tcf3 in development might therefore not be generalizable to normal adult tissues or to malignancy. Our results here call for further experiments to clarify the exact functions of Tcf3 in normal and disease biology, with special attention to the cellular context in which Tcf3 expression occurs.

All experiments were approved by the Baylor College of Medicine (BCM) Institutional Animal Care and Use Committee (IACUC). ROSA26-mTmG (Muzumdar et al., 2007) and Bmi1-CreER (Sangiorgi and Capecchi, 2008) mice have been described elsewhere and were obtained from The Jackson Laboratory. Tcf3 knock-in mice were generated by standard gene targeting techniques described in the methods in the supplementary material. Polymerase chain reaction (PCR) primers used in cloning and genotyping are listed in supplementary material Table S1.

For lineage-tracing experiments, Tcf3^GC/+; ROSA26^mTmG/+ mice were treated with tamoxifen (unless otherwise noted, two doses of 40 mg/ml tamoxifen in 85% corn oil/15% ethanol, 5 μl/g body weight on days P21 and P22). For wounding experiments, mice were subjected to 8 mm full-thickness punch wounds on the midline dorsal skin 5 days after induction. For quantitative comparisons of clone size and frequency over time, clones [defined as clusters of one or more closely associated mGFP(+) cells including at least one basal cell] and the number of basal cells per clone were counted in random non-serial 8 μm cryosections (five or six sagittal sections for tongue and paw skin, 15-45 transverse sections for esophagus). Graphing and statistical analysis was performed using GraphPad Prism (GraphPad Software). Statistical comparisons were performed using nonparametric tests (Mann–Whitney U-test or Kruskal–Wallis test with Dunn's multiple comparison post-test, as appropriate). Values of P<0.05 were deemed significant.

For detection of the ROSA26-mGFP lineage-tracing reporter, tissues were prefixed in 1-4% paraformaldehyde prior to processing for frozen sections. Frozen sections and immunofluorescent staining, and image acquisition, as well as western blots of neonatal keratinocytes, were performed as described previously (Leishman et al., 2013). Nuclear counterstaining was performed with Hoechst 33342 (Life Technologies) (blue channel where shown). Immunohistochemical detection of Tcf3 (in formalin-fixed, paraffin-embedded sections) or of GFP (in fixed-frozen sections) was performed using the Vectastain Elite ABC system (Vector Labs), in some cases with heat-induced epitope retrieval in 10 mM sodium citrate (pH 6.0). Antibodies used in immunofluorescence, immunohistochemistry and western blotting are listed in supplementary material Table S2.

We thank everyone involved in the preparation and publication of this work, including T. M. Shaver, and Drs H. Dierick, M. A. Goodell, M. C. V. Ngo, M. Rendl and J. Rosen for their critical reading of the manuscript. We also thank I. Lorenzo, Drs A. Rodriguez and M. Justice of the Mouse Embryonic Stem Cell Core Facility, and Dr F. DeMayo of the Genetically Engineered Mouse Shared Resource, both BCM facilities, for their technical contributions and expertise.