Tcf3 and Lef1 regulate lineage differentiation of multipotent stem cells in skin

In cultured primary keratinocytes, Lef1 and Tcf3 exhibit different activities that cannot be explained by their ability to associate with and localize β-catenin to the nucleus

To express Tcf3 and Lef1 in skin epithelial cells, we engineered the vectors shown in Figure 2A. The K14 promoter/enhancer is active in primary cultured keratinocytes, in multipotent embryonic skin epithelium, in the stem cell compartment (bulge) of adult skin and its ORS and epidermal progeny (Vasioukhin et al. 1999). Because in vivo and in vitro, the skin epithelia are derived from keratinocytes in which the K14 promoter/enhancer is active, it was optimal for assessing the consequences to stem cell differentiation when Tcf3 and/or Lef1 expression are altered. For comparative purposes, we engineered mammalian Lef1/Tcf3 cDNAs to either encode or not a myc epitope tag at the amino terminus of the protein. Immunoblot analyses with an anti-myc antibody revealed roughly comparable levels of expression of all transgene products (data not shown). Both in vitro (these data) and in vivo (see below), the addition of the epitope tag did not appear to affect the activity of the proteins or their stability.

To verify that any inhibitory effects of Tcf3 would be attributable to its ability to bind DNA, we used a mutant form of Tcf3 (designated Tcf3*) harboring L383P and P407I mutations in the HMG DNA-binding domain (Fig. 2A). Flanking an α-helix required for high-affinity DNA binding, these key residues are thought to stabilize the contacts made with the minor groove of the DNA (Love et al. 1995). We confirmed this by using an electrophoretic mobility-shift assay (EMSA) performed with lysates from bacteria expressing either wild-typeTcf3 HMG or Tcf3* HMG (amino acids 331–427) (Fig. 2B). The Tcf3* mutation did not affect protein stability or interaction with β-catenin, as judged by SDS-PAGE of the expressed proteins (Fig. 2B, bottom), and like Tcf3, Tcf3* elicited nuclear localization of an epitope-tagged version of one of the Groucho proteins (Grg5) when coexpressed in keratinocytes (data not shown; see mouse data below).

Keratinocytes were cotransfected with the TOPFlash (Tcf Optimal Promoter + luciferase) reporter plasmid and K14–Lef1, K14–Tcf3*, and K14–Tcf3. Under normal conditions, that is, in the absence of exogenous β-catenin, Tcf3 repressed whereas Lef1 stimulated the basal level of TOPFlash activity in primary human keratinocytes (Fig. 2C). When cotransfected with K14–ΔN87βcat to deliver high levels of stable β-catenin, however, both Lef1 and Tcf3 stimulated TOPFlash activity (Fig. 2C, inset). The repressor and activator activity of mouse Tcf3 was dependent on its ability to bind DNA, as judged by the inactivity of Tcf3* in these assays (shown) and as judged by the failure of the proteins to transactivate or repress FOPFlash, containing mutations in the DNA-binding sites (data not shown; see Molenaar et al. 1996; DasGupta and Fuchs 1999). These data with mouse Tcf3 are consistent with those of Molenaar et al. (1996), who showed that high levels of β-catenin could activate Xenopus Tcf3 in cultured B cells.

To determine whether the results reflected an increased efficiency of Lef1 over Tcf3 to bind to and/or localize β-catenin in the nucleus, we processed transfections for immunofluorescence microscopy. Unexpectedly, ∼80% of Tcf3-expressing keratinocytes displayed nuclear anti-β-catenin staining (Fig. 2D,E) under conditions where <20% of Lef1 expressing cells showed discernable nuclear staining (Fig. 2F,G). In cells cotransfected with K14–ΔN87βcat (Gat et al. 1998), nuclear β-catenin was seen in most cells expressing either Lef1 or Tcf3 (data not shown). Therefore, although Lef1 was the more potent transactivator, Tcf3 was more efficient at concentrating β-catenin in the nucleus. To convert Tcf3 from a repressor to a transactivator in keratinocytes, very high concentrations of β-catenin seemed to be required, perhaps to titrate out or compete with some negative regulator in Tcf3 activation. Immunofluorescence staining with a panel antibody showed nuclear localization of the Groucho repressor proteins (data not shown; see in vivo data below).

Tcf3 and Tcf3* both translocate endogenous β-catenin in nuclei of transgenic mouse epidermis, ORS, and bulge, but only Tcf3 affects skin function

Normally, Tcf3 is restricted to the stem cells and the lower ORS, but is down-regulated during upward migration of cells from the bulge to the upper ORS and basal layer of the epidermis (DasGupta and Fuchs 1999). To examine the consequences of maintaining Tcf3 in this compartment, we generated mice expressing the genes shown in Figure 2A. To select mice for study that expressed comparable transgene expression levels, we used the myc-epitope tagged genes after verifying that the K14–Lef1 myc-tagged transgene gave a comparable phenotype to that reported for the untagged K14–Lef1 transgene (for details, see Zhou et al. 1995).

A total of seven K14–Tcf3, two K14–Tcf3*, and two K14–Lef1 transgenic mice were born and determined to express the epitope-tagged transgene products. Figure 3A shows anti-myc immunoblot detections of representative examples for equivalent loadings of skin protein extracts. Of our Tcf3-expressing animals, some expressed higher and some lower levels than the example shown. Over a range of ∼10×, the phenotypes were comparable. The sizes of the bands detected with the anti-myc antibody were as expected, with Tcf3 being considerably larger (∼80 kD) than Lef1 (∼55 kD) (Zhou et al. 1995; Pukrop et al. 2000). In both cultured epithelial cells and transgenic skin, Tcf3* consistently migrated slightly faster than Tcf3. Such variation in migration is relatively common even with slight variations in protein sequences, although differences in posttranslational modifications have not been ruled out.

Anti-myc antibodies did not detect the endogenous c-myc protein present in skin (Fig. 3B). Anti-myc did detect transgene expression, however, which in epidermis was targeted to the basal layer but persisted in some suprabasal layers, suggesting that the proteins were relatively stable (Fig. 3C–E). In all cases, the transgene product concentrated in the nucleus, consistent with the nuclear localization sequences located in these HMG proteins (Prieve et al. 1996). This was also true for Tcf3*, suggesting that nuclear localization was not dependent on the ability of the protein to bind DNA.

Consistent with our in vitro observations, endogenous β-catenin was detected in the nuclei of Tcf3- but not Lef1-expressing epidermis (Fig 3F–H). This behavior occurred irrespective of the ability of Tcf3 to bind DNA (data not shown). Nuclear export of β-catenin is regulated, at least in part, by the binding of APC (Henderson 2000; Rosin-Arbesfeld et al. 2000). Because the association of β-catenin with APC and Lef1/Tcf proteins is mutually exclusive, it seems likely that this increase in nuclear β-catenin reflects an increase in Tcf3–β-catenin complexes. We did not pursue this, however, because we later show that this localization is not relevant to the Tcf3-mediated changes in differentiation program that we observe.

In ORS (Tcf3+) and bulge (Tcf3+), K14–Tcf3 expression has no apparent affect, but in epidermis (Tcf3−), it suppresses barrier function and terminal differentiation

Transgenic mice expressing Tcf3 failed to thrive and they either died or were sacrificed within 48 h after birth. These features were not observed in the nontransgenic littermates, nor were they characteristics of K14–Lef1 or K14–Tcf3* transgenic mice, all of which remained healthy and fertile throughout life.

To explore the nature of the neonatal lethality, we investigated the integrity of the skin and oral epithelia, where the K14 promoter is active. Epidermis functions as a barrier to maintain body hydration and exclude harmful external agents. The barrier can be measured by the ability of skin epidermis to exclude penetration of dyes, such as 5-bromo-4-chloro-3-indoyl-β-D-galactopyranoside (Xgal) (Hardman et al. 1998; Segre et al. 1999). In wild-type littermates, the barrier initiated at embryonic day 16 (E16), and was complete by E18.5 (Fig. 4A,B; see also Hardman et al. 1998). All the K14–Tcf3 transgenic embryos examined failed to acquire barrier function by E18.5 (Fig. 4C,D). This barrier defect did not appear to be caused by a delay in development, because the Tcf3 embryos were of normal size, had opened eyelids, and exhibited a distinctive Zorro-like mask of blue dye not characteristic of the normal developmental pattern of barrier acquisition. Additionally, K14–Tcf3 embryos were as large or larger than wild-type littermates.

The skin barrier is acquired by the process of epidermal terminal differentiation. On exhaustion of their proliferative potential, basal cells exit the cell cycle, detach from the underlying basement membrane and terminally differentiate as they move toward the skin surface (Fuchs 1997). In transit, they undergo a series of transcriptional changes, which culminate in the production of dead cells. The cells consist of a proteinaceous sac, called the cornified envelope, which encases densely packed keratin filaments cemented by the protein filaggrin. These cellular ghosts are physically sealed through lipids that are secreted in the latter stages of the differentiation process.

The inner layers, that is, the basal (BL) and spinous (SP) layers, of wild-type and K14–Tcf3 epidermis appeared morphologically similar (Fig. 4E,F). In contrast, late-stage terminal differentiation was affected by Tcf3 expression. Wild-type granular layers (Gr) display prominent keratohyalin granules composed of the large precursor protein, profilaggrin. Late in differentiation, profilaggrin is extensively processed to liberate filaggrin, which then associates with and bundles keratin filaments (for review, see Fuchs 1997). In K14–Tcf3 epidermis, keratohyalin granules were either attenuated or completely absent (Fig. 4F, asterisks). Neither K14–Tcf3* nor K14–Lef1 skin displayed this defect (Fig. 4G,H), suggesting that the effect was dependent on both the Lef1/Tcf member involved and its ability to bind DNA. Similar defects were seen in the dorsal tongue epithelium of K14–Tcf3 mice, perhaps explaining why the K14–Tcf3 mice failed to feed.

Because the mice died shortly after birth, we were unable to assess the consequences of Tcf3 transgene expression to the hair cycle. However, no obvious abnormalities were seen in either the morphology or the biochemistry of the early postnatal hair follicles (data not shown). Therefore, rather than generate hair follicle-like structures as seen in K14–Lef1 and K14–ΔN87βcat mice (Zhou et al. 1995; Gat et al. 1998), K14–Tcf3 appeared to elicit defects in late-stage epidermal differentiation. To some extent, the morphological characteristics of the epithelium seen in K14–Tcf3-expressing mice resembled those of the stem cell compartment (bulge) and the lower portion of the ORS of the hair follicle, that is, the cells that normally express Tcf3. Interestingly, although K14–Tcf3 resulted in nuclear localization of β-catenin in both ORS and bulge, no obvious abnormalities were detected in these cells. Therefore, transgenic Tcf3 expression did not appear to cause deleterious effects to the keratinocytes that naturally expressed Tcf3, but it did affect the epidermal keratinocytes, which normally cease to express Tcf3 when they become committed to their epidermal fate.

Biochemical changes in K14–Tcf3 epidermis are consistent with suppression of terminal differentiation and promotion of features characteristic of the bulge/ORS

As expected from the morphology of K14–Tcf3 skin, cell proliferation markers (Ki67) and early markers of differentiation (K5, K14, K1, and involucrin) were normal (data not shown). Late stage markers, however, revealed biochemical signs indicative of partial repression of epidermal differentiation restricted to the K14–Tcf3 transgenic skin. Therefore, the paucity of keratohyalin granules was reflected by a marked reduction in anti-filaggrin staining (Fig. 5A–D), and the major constituent of the cornified envelope, loricrin, also appeared to be atypical (Fig. 5E–H). Cornified envelopes were still made, however, and transglutaminase was expressed in K14–Tcf3 epidermis (data not shown). Taken together, these results suggest that expression of Tcf3, but not Lef1, suppresses some, but not all, aspects of late-stage terminal differentiation in epidermis.

K5, K14, transglutaminase, and involucrin have been identified as components of the ORS/bulge, as well as epidermis. Therefore, we examined K17, a keratin whose expression is down-regulated in epidermis and elevated in the ORS and bulge just before birth (Fig. 5I) (McGowan and Coulombe 1998). In postnatal K14–Tcf3 epidermis, K17 expression persisted (Fig. 5I–L). Similarly, K6 and its partner K16, which are normally restricted to the inner (suprabasal) layers of the ORS, were also induced (Fig. 5M–P). Analogous to its expression pattern in the ORS and bulge, K6 was seen only in the suprabasal epidermal layers. The induction of K6 and K17 expression in K14–Tcf3 transgenic skin was not pronounced in K14–Tcf3* or K14–Lef1 transgenic skin, suggesting its specificity for Tcf3 expression. Taken together, the morphological and biochemical data suggest that transgenic expression of Tcf3 suppresses features characteristic of skin epithelial cell types lacking Tcf3 (epidermis) and promotes those typical of cell types expressing Tcf3 (ORS and bulge).

In keratinocytes, the repressive effects of Tcf3 are dependent upon its Grg-interaction domain

XTcf3 interacts with at least two different repressor proteins, CtBP and members of the Groucho family (Brannon et al. 1999; Brantjes et al. 2001), making their mouse homologs likely candidates to mediate mammalian Tcf3 repression of TOPFlash activity in keratinocytes. To investigate this possibility, we generated and tested Tcf3 mutants harboring deletions in the putative CtBP (Tcf3ΔC), Grg (Tcf3ΔGrg), and β-catenin (ΔNTcf3) interacting domains (Fig. 6). As judged by immunoblot analyses, these and other mutant forms of Tcf3 and Lef1 shown in Figure 6A were comparably expressed in cultured epithelial cells (data not shown). WhereasTcf3ΔC only modestly affected Tcf3-mediated repression of basal TOPFlash activity, Tcf3ΔGrg relieved repression. These findings suggested that of the two, Grg interactions may be more relevant in mediating Tcf3 repression in keratinocytes. Recent studies by Brantjes et al. (2001) have shown that like Drosophila Groucho, mammalian Groucho proteins associate with histone deacetylases and with vertebrate Tcfs, thereby suggesting a mechanism for the repressor activity observed here.

To further characterize the region of Tcf3 required for repression and assess the basis for the difference in activity between Lef1 and Tcf3, we engineered and tested a series of Lef1–Tcf3 fusion proteins for their effect on the TOPFlash reporter (Fig. 6). When the majority of Tcf3's amino-terminal segment (321 or 341 amino acids) was fused to the carboxy-terminal segment of Lef1 including the HMG domain (Fig. 6B, Fusions 1 or 2, respectively), these hybrid proteins effectively repressed basal TOPFlash activity in a fashion similar to that of Tcf3ΔC. With only part of the amino-terminal segment spanning the Grg domain (183 amino acids), the Tcf3–Lef1 hybrid lost most of its ability to repress (Fig. 6B, Fusion 3). With only 111 amino acids of Tcf3's unique amino-terminal segment, The Tcf3–Lef1 hybrid protein behaved as a transactivator (Fig. 6B, Fusion 4). Taken together, these findings illuminated a marked difference in the segment between the β-catenin and DNA-binding domains of these proteins.

Additional insights were obtained when the Tcf3 mutants and Tcf3–Lef1 hybrids were coexpressed with the ΔN87βcatenin. Although Tcf3ΔC and Tcf3 had repressed basal TOPFlash activity to a similar degree, Tcf3ΔC did not stimulate TOPFlash like Tcf3 did in the presence of elevated levels of stabilized β-catenin (Fig. 6B, inset). Therefore, in keratinocytes, the carboxy-terminal segment of Tcf3 seemed to be necessary to enable full-length Tcf3 to activate target genes in the presence of high levels of β-catenin. In the absence of the putative Grg-interacting domain, this requirement was alleviated, as Tcf3ΔGrg and the Tcf3–Lef1 fusion proteins lacking this segment of Tcf3 were able to transactivate TOPFlash. Based on these data, it seems that in the presence of high levels of stabilized β-catenin, the carboxy- and amino-terminal segments of Tcf3 interact either indirectly or directly to adopt a conformation that does not interfere with the β-catenin and/or HMG domain transactivating functions.

Finally, we examined the consequences to Tcf3 function on removal of its β-catenin-interacting domain (Fig. 6B“). In the presence of exogenous β-catenin, ΔNTcf3 repressed TOPFlash activity below levels of ΔN87βcatenin alone.

The transdifferentiation seen in the epidermis of K14–Tcf3 mice is dependent on the Groucho interacting domain but not on the CtBP or the β-catenin interacting domains

Our in vitro studies led to the hypothesis that in keratinocytes, Tcf3 acts as a repressor and that this effect is mediated primarily through the putative Grg-interacting domain in the amino-terminal segment of the protein. The induction of K17 and K6 and the nuclear localization of β-catenin on Tcf3 expression in skin, however, left open the possibility that in vivo, Tcf3 might function as a transactivator. To evaluate this possibility, we engineered transgenic mice expressing K14–Tcf3ΔC, K14–Tcf3ΔGrg, and K14–ΔNTcf3 (Fig. 7A).

As predicted from our in vitro studies, the overall phenotype of K14–Tcf3ΔC transgenic mice was similar to that of K14–Tcf3 mice. Two mice survived until postnatal day 3 and one died at postnatal day 4. Tcf3ΔC-expressing epidermis displayed a lack of keratohyalin granules in both newborn (data not shown) and four-day-old mouse samples (Fig. 7B). In contrast, all five independent K14–Tcf3ΔGrg founder animals survived to adulthood, produced F₁ offspring, and displayed no apparent defects in viability or fertility, although high expressing lines did show an approximately one day premature opening of eyelids. Tcf3ΔGrg-expressing mice also displayed normal skin morphology (Fig. 7C). Despite these differences between Tcf3ΔC and Tcf3ΔGrg, transgene expression levels were comparable (Fig. 7H,I; Coomassie and anti-myc immunoblot of skin proteins).

The lack of keratohyalin granules in Tcf3ΔC-expressing skin was supported by the reduction in anti-filaggrin and anti-loricrin staining observed in skin sections and by the reduction in these proteins detected by immunoblot analyses with these antibodies (Fig. 7D–G,J,K). In addition, as suggested by the bulge/ORS morphology of K14–Tcf3ΔC epidermis, it displayed K17 and K6 staining in patterns analogous to that noted for K14–Tcf3 transgenic epidermis (data not shown). In contrast, Tcf3ΔGrg-expressing skin displayed normal staining and protein expression patterns. The loss of K14–Tcf3 phenotype associated with mutation of the Grg-interacting domain suggested that Grg proteins may exist in the skin and contribute to the Tcf3-mediated abnormalities in epidermal differentiation. Consistent with this notion was the fact that a panel antibody (pan-TLE) against mammalian Grg proteins labeled the nuclei of wild-type epidermis (Fig. 7L).

As we had observed previously for the Tcf3 and Tcf3*, Tcf3ΔC and Tcf3ΔGrg expression resulted in the nuclear localization of endogenous β-catenin (data not shown). Because both Tcf3* and Tcf3ΔGrg transgenic mice exhibited generally good health, the nuclear colocalization of Tcf3 and β-catenin, per se, did not appear to be sufficient to cause the phenotypic changes associated with the K14–Tcf3 and K14–Tcf3ΔC transgenic mice. To assess whether the β-catenin interaction domain of Tcf3 was necessary for manifesting the K14–Tcf3 phenotype in vivo, we examined K14–ΔNTcf3 transgenic mice. K14–ΔNTcf3 mice displayed open eyelids at birth, were frail, and died or were sacrificed within a day. As shown in Figure 7M, the morphology of the skin of these mice was comparable to that seen in the K14–Tcf3 and K14–Tcf3ΔC animals. Furthermore, K17 and K6 expression were still seen in the K14–ΔNTcf3 skin (data not shown). Taken together, it seems most likely that the activation of genes associated with ORS/bulge differentiation in Tcf3 transgenic epidermis arises indirectly from repression of Tcf3-regulated genes. A further understanding of this process awaits identification of Tcf3 target genes.

In the absence of the β-catenin-interacting domain, transgenic Lef1 blocks hair cell differentiation and induces sebaceous cell differentiation in its place

Given that Tcf3 and ΔNTcf3 behaved similarly in vitro and in vivo, we wondered whether Lef1 might behave similarly to Tcf3 when its ability to associate with β-catenin was compromised. We first examined the difference in behavior of ΔNTcf3 and ΔNLef1 on activation of target genes in vitro. Because Lef1 normally functions in hair follicle morphogenesis, and because the hair keratin genes were shown previously to possess sequences that bind to Lef1 (Zhou et al. 1995), we used the promoter/enhancer for one of these genes (mHK1a) driving luciferase as a reporter. This gene is normally expressed in the precortical cells that exhibit TOPGAL activity and nuclear β-catenin localization (see Fig. 1).

As shown in Figure 8, neither Lef1 nor β-catenin (ΔN87βcat) alone stimulated luciferase activity appreciably when transfected in cultured keratinocytes. Activity levels rose nearly 80×, however, when Lef1 and β-catenin were coexpressed. Tcf3's effect on this promoter was considerably less potent, and in this way, was similar to what we observed with the artificial TOP promoter/enhancer (see Fig. 2). The mouse HK1a promoter/enhancer was not activated by ΔNLef1 or by ΔNTcf3, indicating that most, if not all, of the transactivating potential of Lef1 and Tcf3 are dependent on the β-catenin binding domain at the amino terminus of the protein.

We next engineered K14–ΔNLef1 transgenic mice and compared their phenotype with K14–ΔNTcf3 animals. In striking contrast to the neonatal K14–ΔNTcf3 mice (see Fig. 7M), young K14–ΔNLef1 mice exhibited no obvious epidermal defects or signs of ORS differentiation. These animals survived well into adulthood (>9 mo), and exhibited the characteristic thin epidermis of aging mice. With age, however, the hair coat became increasingly sparse. To be able to understand the basis for the defect in K14–ΔNLef1 animals, we traced the histological signs of abnormalities back to a relatively early age, at approximately the start of the first postnatal hair cycle (3 wk).

On initiation of the normal hair cycle, a stream of new cells emerges from the bulge stem cell compartment of each follicle (Fig. 9A). This stream, referred to as the secondary hair germ, soon progresses to form a mature new hair follicle (Fig. 9B). Biochemically, nuclear β-catenin localization coincides with hair keratin production in the new follicle (Fig. 9C–E; β-catenin and hair keratins in green, ORS keratin in red). This staining pattern is analogous to that which occurs during the anagen phase of the initial hair follicle (Fig. 1).

Whereas K14–ΔNLef1 follicles initiated their new hair cycle and produced a new hair germ, they soon displayed dramatic alterations in the process. Rather than a normal follicle emerging beneath the bulge, large cysts appeared (Fig. 9F). These cysts stained with oil red O, a typical marker of differentiated cells within the sebaceous gland (upper right inset). Despite their resemblance, the natural sebaceous glands of the follicle always appear above the bulge in normal follicles (Fig. 9A; see also 9G). As shown in Figure 9, F and G, the oil red O-staining cells that were specific to the ΔNLef1-expressing skin were located at the tips of the new hair follicles. In older animals, these cysts became very large, until the point where they visibly affected the skin surface and appeared as bumps.

To further pinpoint the step in the hair cycle that was defective in the younger ΔNLef1 animals, we examined hair follicles as they progressed through the cycle. As judged by immunofluorescence, the crescent of nuclear β-catenin-staining precortical cells still formed (Fig. 9H). The cells that exhibit nuclear β-catenin are those that also concentrate nuclear Lef1 (see Fig. 1), and although the K14 promoter is not active in these cells, they retained ΔNLef1 as judged by anti-myc staining (lower left inset in Fig. 9F). Most notably, despite the presence of nuclear β-catenin and Lef1 in these cells, they did not express the hair-specific keratin genes, as judged by the lack of immunofluorescence seen with a panel hair keratin antibody (Fig. 9, cf. I with E).

Taken together, these data showed that ΔNLef1 blocked the activation of hair-specific gene expression and remarkably induced what appeared to be sebum production. Because ΔNLef1 did not exhibit transactivation potential in our in vitro assays, we suspect that its effect on sebum-producing genes is indirect, resulting from the failure to activate certain key Lef1/β-catenin-regulated proteins that normally prevent sebum gene expression. The principle is remarkably similar to that argued for ΔNTcf3. Whereas ΔNTcf3 promoted ORS/bulge differentiation, however, ΔNLef1 resulted in differentiation more characteristic of sebum-producing cells. In revealing marked new differences in behavior of ΔNTcf3 and ΔNLef1, these data further underscore the importance of these factors in directing different stem cell lineages in skin.

Plasmid construction

Molecular biology techniques were performed according to conventional protocols (Ausubel et al. 2000). All constructs made with PCR were sequenced to verify their authenticity.

The K14-BG expression cassette (Gat et al. 1998) was used for construction of K14–Tcf3* and its derivatives. The myc-tagged mTcf3* cDNA was purified as a blunted EcoRV–BglII fragment from pCDNA3–myc–mTcf3* (a kind gift of Dr. Hans Clevers, University Hospital, The Netherlands) and ligated into BamHI (blunted)-cut K14-BG cassette. To make K14–Tcf3, a fragment encoding the mTcf3 HMG domain was amplified from newborn mouse epidermal cDNA, which was used to correct the mutated region of mTcf3*. To make K14–Tcf3ΔC, PCR primers were used to place a stop codon after nucleotide 1491 of mTcf3 cDNA, thereby encoding a truncated version of the protein. To make Tcf3ΔGrg, an artificial BssHII site was inserted before nucleotide 957 of mTcf3 cDNA, and a subsequent digestion with BssHII and PmlI and ligation into BssHII–PmlI-digested K14–Tcf3 resulted in a cDNA encoding an in-frame deletion of amino acids 143–318 of mTcf3. To make K14ΔNTcf3, a PCR product placing a SacII site at the 5′ end of a truncated Tcf3 cDNA was cloned into SacII–DraIII digested K14–Tcf3 such that the amino-terminal 71 amino acids of Tcf3 were deleted.

The construction of the K14–Lef1 expression plasmid with a myc-epitope tag at the 5′ end of the human Lef1 cDNA used the K14-BG–hLef1 plasmid described previously (Zhou et al. 1995). PCR primers were used to make a hLef1 fragment with a myc-epitope tag at its 5′ end. The tagged hLef1 region was digested with XhoI and EcoRI and used to replace untagged hLef1 sequence from XhoI–EcoRI (partial)-digested K14-BG–hLef1. The K14–ΔNLef1 construct was made by PCR amplifying the hLef1 sequence encoding amino acid 63 onward, using primers placing a BamHI site and myc-tag at the 5′ end of the truncated hLef1 cDNA.

To construct the mTcf3/hLef1 fusion protein expression plasmids, the K14–Lef1 and K14–Tcf3 plasmids were used as templates. For fusion 1, the 3′ end of mTcf3 was replaced with the 3′ end of hLef1 by replacing a PmlI (base pair 1021 of mTcf3)–HindIII (cuts 3′ of polyA in K14-BG cassette) deletion of K14–Tcf3 with a PmlI (base pair 841 of hLef1)–HindIII fragment from K14–Lef1. For fusion 2 and fusion 3, a PCR gene stitching method was used in which recombinant hLef1–mTcf3 PCR products were digested with BssHII and PmlI, and ligated into BssHII–PmlI-digested K14–Tcf3. For fusion 4, the 5′ end of hLef1 was deleted from K14–Lef1 by digesting with MfeI (cuts 5′ of myc-epitope tag) and XhoI (base pair 324 of hLef1; blunted), and it was replaced with the 5′ end mTcf3 with a MfeI–Asp718 (base pair 353 of mTcf3; blunted) from K14–Tcf3.

To construct bacterial expression plasmids for the HMG domains of Tcf3 and Tcf3* proteins, PCR primers were used to amplify base pairs 991–1281 from either K14–Tcf3 or K14–Tcf3* templates. PCR products were digested with BamHI and XhoI and ligated into BamHI–XhoI-cut pTRIEX (Novagen) such that the 6×His tag of the pTRIEX vector was in-frame with the HMG domain encoding PCR product. The HK1Flash plasmid was constructed by removing the 3.5-kb fragment of the mouse hair keratin 1a fragment from plasmid pmHK113.3.3R (Upjohn) with EcoRI and inserting the blunted fragment into the SmaI site of pGL3 basic (Promega).

Generation and analysis of transgenic mice

Transgenic mice were engineered as described (Vassar et al. 1989). Genomic DNA was isolated from tail or toe samples of mice and analyzed by PCR with primers specific for the human K14 promoter. For histology of mouse skin, samples of dorsal backskin were fixed overnight in freshly prepared 4% paraformaldehyde at 4°C, dehydrated, embedded in paraffin, and sectioned for hematoxylin and eosin staining. Oil Red-O staining was performed on frozen cryosections as described (Carson 1990).

Dye penetration assays

Dye penetration assays were performed as described (Hardman et al. 1998; Segre et al. 1999). Embryos from timed pregnancies were isolated at E16.5–17.5 for experiments with nontransgenic mice. Transgenic embryos at E18.5 were compared with nontransgenic littermates. Tails were removed for genotyping and for determining levels of transgene expression, and embryos were placed in a solution of containing X-Gal at pH 4.5. Embryos were allowed to incubate 5 h to overnight at 37°C, washed in PBS, and post-fixed overnight in 4% PFA at 4°C.

EMSA

Protein was extracted from BL21 codon plus bacteria (Stratagene) transformed with Tcf3HMG or Tcf3* HMG constructs in 1× DNA-binding buffer (25 mM HEPES at pH 7.9, 75 mM KCl, 1 mM EDTA, 0.1% Nonidet P-40, 10% glycerol, 0.5 mM dithiothreitol, 1× protease inhibitors; Boehringer), and levels of expression were determined by Western blot analysis using an anti-pentaHIS antibody (QIAGEN). In standard 20 μL reactions in 1× DNA-binding buffer, protein extracts were incubated with 1μg poly(dI-dC) and 1 μL ³²P-labeled TCRalpha oligo (Waterman et al. 1991) for 30 min at 4°C. The protein–DNA complexes were resolved by gel electrophoresis on 5% polyacrylamide gels, and band shifts were visualized on dried gels by autoradiography.

Immunofluorescence and antibodies

For indirect immunofluorescent detection of proteins in mouse backskin, the mouse on mouse kit (MOM kit; Vector Laboratories) was used according the manufacturers' protocol with minor variations. All immunofluorescent staining was performed on either 8–10-μm frozen sections of back skin embedded in OCT or 5 μm paraffin-embedded sections. Frozen sections were fixed in 4% PFA, and antigens were visualized with FITC- or TxR-conjugated secondary antibodies (Jackson Labs). All fluorescent images were taken with a Zeiss LSM 410 system.

Primary antibodies used were mouse anti-myc tag (Invitrogen); rabbit anti-K5; rabbit anti-filaggrin (BAbCO); rabbit anti-loricrin (BAbCO); mouse anti-β-catenin (Sigma); rat anti-E-cadherin (Zymed); rabbit anti-Lef1; guinea pig anti-Tcf3; rabbit anti-K6 (BAbCO); rabbit anti-K17 (McGowan and Coulombe 1998); rabbit anti-TLE2 (Santa Cruz); and rat anti-PanTLE (Stifani et al. 1992).

Cell culture and transient transfections

Primary human keratinocytes were grown in serum-free keratinocyte growth media (Clonetics), and transfections were performed only with cultures of five passages or less. Transfections were performed with the Fugene (Roche) reagent according to the manufacturers' protocol, and 40 h after transfection cells were harvested for luciferase assays (Promega Dual-Luciferase Kit). Firefly luciferase values (TOPFlash) were standardized to Renilla luciferase values (pRL–CMV, or pRL–TK; Promega) to account for differences in transfection efficiency between samples. In addition, FOPflash or FOPGAL (DasGupta and Fuchs 1999) was also cotransfected with expression plasmids to ensure that TOPFlash values were dependent on the Lef1/Tcf-binding sites in the promoter.