Sustained Expression of Snail Results in Epidermal Hyperproliferation and Differentiation Defects in Tg Mouse Skin

The striking spike of Snail expression coincident with hair bud formation and enhanced proliferation prompted us to examine the consequences of ectopically expressing Snail elsewhere in mouse skin epidermis. To distinguish Tg from endogenous Snail, we used the HA-epitope, shown previously not to alter Snail's transcriptional activity [12]. Of 20 K14-Snail[HA] Tg animals generated, three expressed the transgene and all exhibited analogous phenotypes. Mice that integrated the transgene at the single-cell stage died at or shortly after birth. The three surviving full-Tg founder mice harbored transgene integrations that gave stable transmission of mosaic Snail gene expression through the germline. Progressively poor health necessitated our sacrificing most offspring from these lines within a year of birth.

As Snail Tg animals grew, they became distinguished by their small size, short tails, and flaky skin (Figure 2A). Histological analyses of 3-d old (P3) mice revealed mosaic patches marked by epidermal thickening (Figure 2B). The mosaic morphology was reflected at the level of Tg Snail protein, with only the hyperthickened regions expressing nuclear HA-tagged Snail (Figure 2C). These hyperthickened areas were marked by excessive proliferation, as revealed by antibodies against the proliferating nuclear antigen Ki67 (Figure 2D and ​and2E).2E). Activated, pMAPK-positive cells were also prevalent in these areas (Figure 2F and ​and2G),2G), as were cells expressing keratin 6, a keratin induced in the suprabasal layers of hyperproliferative skin (Figure 2H and ​and22I).

Open in a separate window

Figure 2

Misexpression of Snail in Mouse Skin Epidermis Results in Hyperproliferation

Three different surviving Tg founder mice harbored a K14-Snail transgene that was integrated into a locus that resulted in inheritable, mosaic expression of the transgene in skin epidermis. All displayed similar abnormalities, as did their offspring.

(A) P16 WT and K14-Snail Tg mice. Insets denote magnified tail segments, which displayed a mosaic, flaky appearance in Tg mice. Size differences appeared with age, and are likely due to K14-promoter activity in the tongue and oral epithelium, resulting in progressive defects and reduced food intake. Hence, skin sections from young (P3) mice were analyzed (B–I).

(B) Hematoxylin- and eosin-stained Tg skin section. Double arrows demarcate the border of mosaic histology, with seemingly normal epidermis (epi) and a mature hair follicle (hf) at left and hyperthickened epidermis at right.

(C) Immunofluorescence of Tg skin section labeled with antibodies as color-coded on frame. Double arrows demarcate the border of mosaic anti-Snail (green), revealing Snail expression coincident with regions of hyperthickened epidermis (at left) and absent in regions of normal epidermis (at right).

(D–I) Sections of P3 WT or Tg skin (affected region) subjected to either immunofluorescence (D, E, H, and I) or immunohistochemistry (F and G) with antibodies as indicated on the panel. Anti-keratin 5 indicates K5, normally restricted to the basal layer of the epidermis; anti-keratin 6 detects keratin 6, expressed in postnatal epidermis under conditions such as wounding, in which hyperproliferation occurs. All other antibodies are as in the legend to Figure 2. Comparison of D and E provide representative examples that illustrate that pMAPK is found in only a subset of all proliferating (Ki67-positive) cells. Note also the presence of Ki67- (E) and pMAPK-positive (G) cells in some suprabasal areas; Ki67-positive cells colabeled with anti-Snail (E).

Expression of the Snail transgene did not block terminal differentiation in the hyperproliferative epidermis, but it distorted it markedly (Figure 3A–3H). Typical of most hyperproliferating conditions, Snail expression led to a large expansion in layers with spinous and granular morphology. Additionally, however, was a marked and variable expansion of keratin 5 (K5), normally restricted to the innermost basal layer (see Figure 3). Although the failure of Snail-null mice to develop past gastrulation [21] precluded our ability to study the loss of Snail function in skin development, a good correlation emerged between the expression of Snail protein and the extension of K5, Ki67, and pMAPK suprabasally (compare data in Figures 2 and ​and33).

Open in a separate window

Figure 3

Alterations in the Differentiation Program and Basement Membrane Organization in Snail-Expressing Tg Epidermis

(A–H) Immunofluorescence of skin sections from P3 WT and Tg mice. Shown are affected areas of Tg skin; in areas where Snail protein was not expressed, stainings were normal. Sections were labeled with antibodies as indicated and color-coded on each frame. Antibodies are against markers of normal epidermal differentiation, and include K5 (a basally expressed keratin), K1 (a suprabasal keratin, expressed in spinous layer cells), involucrin (Inv; a suprabasally expressed cornified envelope protein found in upper spinous and granular layer cells), loricrin (Lor; a cornified envelope protein expressed in the granular layer), and filaggrin (Fil; a protein that bundles keratin filaments in the granular layer and stratum corneum). Note abnormal extension of anti-K5 suprabasally, often present in anti-K1 positive suprabasal Tg cells.

(I–N) Immunohistochemistry (I and J) or immunofluorescence (K–N) of sections of P30 Wt (I, K, and M) and Tg (J, L, and N) (affected areas) skins using the antibodies indicated. Note that with age, affected areas of the Tg epidermis became increasingly undulating, often exhibiting papilloma-like invaginations (J). Insets in I and J are magnified views of the boxed areas, illustrating the absence (Wt) or presence (Tg) of nuclear anti-cyclin D staining. With age, affected areas of the Tg epidermis also displayed perturbations within the basement membrane, as judged by antibody labeling against either basement membrane (K and L) or hemidesmosomal (M and N) components. Double arrows in L demarcate mosaic zones, revealing that perturbations were restricted to hyperthickened, i.e., Snail-positive zones (to left of double arrows). Other abbreviations are as noted in the legend to Figure 2.

The changes in hyperproliferation and differentiation were not initially accompanied by gross signs of epithelial invaginations. With age, however, epidermal folds and undulations developed in areas where Snail was expressed, and proliferative markers persisted in these regions (Figure 3I and ​and3J;3J; anti-cyclin D staining). The undulations were accompanied by partial dissolution of the underlying basement membrane (Figure 3K and ​and3L).3L). Aberrant staining was also observed with antibodies against components of the hemidesmosomes, which provide strong adhesion of basal epidermal cells to the underlying basal lamina (Figure 3M and ​and3N).3N). Interestingly, similar types of alterations occur in the basement membrane in the hair bud of embryonic and newborn mice when Snail is normally expressed. The fact that the basement membrane separating the epidermis from the dermis is altered only in the adult Tg animals suggests the involvement of intermediary factors not as readily available in the epidermis as they are in the follicle.

Possible Links between Epidermal Hyperproliferation and Down-regulation of AJ Proteins in Snail Tg Mice

Given that the E-cadherin promoter is a direct target for Snail-mediated repression in vitro [4,12,13], and that E-cadherin was down-regulated in Snail-expressing hair buds, we examined the status of E-cadherin and other AJ proteins within regions of hyperproliferative epidermis where Tg Snail was present (Figure 4A). In these regions, immunofluorescence staining of E-cadherin and α-catenin were markedly diminished. In contrast, the intensity of antibody staining for two other AJ proteins, β-catenin and Ajuba, was still strong. Interestingly, however, despite appreciable immunofluorescence, localization of β-catenin and Ajuba appeared to be largely cytoplasmic rather than at cell-cell borders (Figure 4A insets).

Open in a separate window

Figure 4

Snail-Mediated Remodeling of AJs Contributes to Hyperproliferation

(A) Immunofluorescence of skin sections from P30 Wt and Tg mice. Shown are affected areas of Tg skin; in areas where Snail protein was not expressed, stainings were normal. Antibodies used are against AJ proteins and include E-cadherin (E-cad), the transmembrane core protein; β-catenin (β-cat), which binds E-cadherin at AJs and which can also participate as a transcription cofactor when associated with LEF-1/TCF proteins in the nucleus; α-catenin (α-cat) which binds to both β-catenin and Ajuba, a close relative of zyxin; and Ajuba, which can associate with proteins that bind to the actin cytoskeleton, as well as with Grb-2, a mediator of the GTP nucleotide-exchange protein Sos, involved in activation of the Ras-MAPK signaling cascade. In Snail-expressing Tg regions, there was a reduced staining with anti-E-cad and anti-α-cat and a more diffuse staining with anti-Ajuba. Insets in the panels for β-catenin and Ajuba staining are magnified views of the boxed areas. Arrows mark membrane localization of the protein and asterisks mark cells with elevated levels of cytoplasmic β-catenin or Ajuba.

(B) Western blot analyses of protein extracts from P30 Wt and Tg back and ear skins. Antibodies are as in (A) except anti-P-cad, which detects P-cadherin, whose expression in the hair follicle was not affected, and anti-tubulin, which detects tubulin, a control for equal protein loadings. Note that the reductions seen in E-cadherin and α-catenin are likely to be underestimates of the actual differences in affected regions, since the Tg skin expressed Snail mosaically.

(C) In the presence of elevated Snail, α-catenin levels can be restored by overexpression of E-cadherin. Keratinocytes were transfected with either HA-tagged Snail (Snail[HA]; images on the left) or Snail(HA) and Ecad(HA) (images on the right). 2 d after transfection, cells were switched from low-calcium growth medium to high-calcium medium for 6 h to induce AJ formation. Cells were stained with antibodies as indicated on the panels. Arrowheads point to sites of intercellular contact between a Snail-transfected keratinocyte and its neighboring untransfected cell.

(D) Reintroduction of E-cadherin in keratinocytes expressing Snail returns pMAPK to basal levels. Keratinocytes were transfected with control vector (K14), or Snail(HA), or Snail(HA) + E-cad(HA). After 2 d, cells were serum starved for 4 h and whole cell lysates were made and Western blotted with antibodies to pMAPK, HA to recognize the HA-tagged Snail and E-cadherin protein, 20or tubulin as a loading control.

(E) Ajuba interacts with Grb-2 under conditions where α-catenin levels are reduced. Protein extracts were made from skins of P30 Wt and K14-Snail Tg P30 mice (blots on the left) and of newborn Wt and K14-Cre/α-catenin (fl/fl) conditionally null animals (blots on the right) [7]. Equal amounts of protein extracts were treated with anti-Grb-2 antibody (+) or control isotype antibody (–), and following centrifugation, immunoprecipitates were subjected to SDS-PAGE and Western blot analysis with anti-Ajuba and anti-Grb-2 antibodies. Note the presence of Ajuba only under conditions where levels of α-catenin and other AJ proteins were aberrantly low or absent.

(F) Transgene expression of excess Ajuba or the Grb-2-interacting domain (pre-LIM) of Ajuba in keratinocytes results in the activation of the Ras-MAPK pathway. Primary newborn mouse keratinocytes were transfected with either the empty K14 expression vector (K14), or the expression vector driving Snail, full length Ajuba, or the pre-LIM domain of Ajuba in the absence or presence of a peptide inhibitor (inh) that disrupts the interaction between Grb-2 and Sos. 48 h posttransfection, protein extracts were prepared and subjected to SDS-PAGE and Western blot analyses with antibodies against pMAPK, total MAPK, Ajuba (also recognizing the smaller, pre-LIM domain), and Snail.

Architectural differences in the epidermis made Western blot analyses somewhat difficult to gauge. However, in regions such as ear skin, where the highest levels of Snail protein were expressed, the effects were accentuated. In both back skin and ear skin, overall levels of E-cadherin and α-catenin were reduced, under conditions where β-catenin and Ajuba levels remained unchanged relative to controls (Figure 4B). Taken together, these data were consistent with our results obtained from immunofluorescence microscopy.

A priori, the decrease in α-catenin levels could be due to either direct transcriptional repression by Snail or perturbations in AJ formation caused by the decrease in E-cadherin gene expression. To distinguish between these possibilities, we tested whether α-catenin levels could be restored by exogenous expression of E-cadherin in Snail-expressing keratinocytes. As shown in Figure 4C, transiently transfected keratinocytes expressing HA-tagged Snail displayed a loss of E-cadherin and α-catenin at cell-cell borders. Coexpression of exogenous HA-tagged E-cadherin not only enabled cell-cell border localization of E-cadherin protein, but also rescued the cell-cell border staining of α-catenin (Figure 4C). The ability to restore α-catenin expression and localization under these conditions argues against the notion that Snail transcriptionally represses α-catenin. Rather, the findings are consistent with a previous report that E-cadherin is required for the translation of α-catenin mRNA [22].

Despite the reductions in AJ markers, Tg skin still displayed sealed membranes and intercellular junctions that were largely intact, as judged by ultrastructural analyses (unpublished data). In this respect, the skin epithelium resembled that of the hair bud, where the down-regulation in junction proteins is permissive for cell-cell remodeling without abrogating intercellular adhesion.

The similarities between Snail Tg epidermis and hair buds extended to the hyperproliferative state, leading us to wonder whether the down-regulation of AJ proteins might contribute to this condition. Given the increase in pMAPK staining in Snail Tg epidermis (see Figure 2G), we used pMAPK levels as our assay to test whether the loss of E-cadherin contributed to the Snail-mediated increase in proliferation. Consistent with our in vivo observations, transfected keratinocytes expressing Snail exhibited a substantial increase in pMAPK levels relative to control cells (Figure 4D). Coexpression of E-cadherin with Snail appeared to abrogate this effect. Together, these findings raised the possibility that an AJ-associated protein that is normally sequestered at the plasma membrane may participate in a proliferation signaling pathway when AJs are deconstructed.

Numerous studies have correlated a down-regulation of E-cadherin with a translocation of β-catenin to the nucleus and a transactivation of genes that are regulated by the LEF-1/T cell factor (TCF) family of DNA binding proteins [23,24,25]. The presence of nuclear cyclin D in hyperproliferative Snail Tg epidermis was particularly intriguing since prior studies have reported cyclin D gene as a direct target of TCF/β-catenin transcription [26]. This said, we did not detect nuclear β-catenin in our Tg epidermis, and mating the Snail Tg mice against the TOPGal reporter mouse [20] gave no signs of ectopic LEF-1/Tcf/β-catenin activity (unpublished data).

We next turned to the presence of cytoplasmic Ajuba for a possible mechanistic link to the proliferative increase in our Snail Tg epidermis. In addition to its documented ability to bind α-catenin [10], Ajuba can also associate with growth factor receptor-bound protein-2 (Grb-2)/son of sevenless (Sos), the nucleotide exchange factor for Ras, which is upstream from activation of MAPK [9]. Given the increase in pMAPK staining in Tg skin, we examined the possibility that Ajuba might have changed its binding partner in Snail-expressing epidermis. Interestingly, Ajuba was readily detected in anti-Grb-2 immunoprecipitates of protein lysates from skins of Snail Tg mice but not from the corresponding wild-type (WT) animals (Figure 4E). When these experiments were repeated with α-catenin-null epidermis, a similar Grb-2-Ajuba association was detected, and again, this interaction was not detected in the protein extracts from control littermate skin (Figure 4E). Together, these data demonstrate that the reduction in α-catenin levels, either by Snail-mediated down-regulation of E-cadherin or by α-catenin conditional targeting, allows Ajuba to interact with Grb-2/Sos.

If the competition between Grb-2/Sos and α-catenin for Ajuba is functionally relevant to the hyperproliferative state of a keratinocyte, then overexpression of Ajuba would be expected to bypass the competition and promote activation of the Ras-MAPK pathway in WT keratinocytes. Indeed, when serum-starved keratinocytes were transiently transfected with an Ajuba expression vector, the levels of pMAPK were not only elevated but also comparable to those transfected with the K14-HASnail transgene (Figure 4F). This activation was abolished when cells were treated with a small peptide inhibitor that specifically interrupts the Grb-2/Sos interaction (Figure 4F; see lanes marked “inh”) [27].

Ajuba's pre-LIM domain is the segment that associates with Grb-2's Src-homology 3 domain [9]. When this domain was overexpressed in serum-starved keratinocytes, a comparable elevation in pMAPK was observed (Figure 4F). As expected, the small peptide inhibitor that interrupts the Grb-2/Sos association blocked the effects. These data suggested that by elevating cytosolic Ajuba levels, Ajuba's pre-LIM domain may associate with Grb-2/Sos in a manner that stimulates its nucleotide exchange activity and leads to activation of the Ras-MAPK pathway. Although this pathway provides one mechanism by which Snail expression and proliferation may be coupled in skin epithelium, proliferative circuitries involving AJs are known to be complex and often interwoven. Future studies will be needed to systematically dissect these putative intricacies at a molecular level.

Signaling during Early Hair Follicle Morphogenesis

Recent studies on hair bud morphogenesis suggest that Wnt signals likely from the epithelium and BMP inhibitory signals from the underlying mesenchyme converge to produce an active transcription factor complex involving β-catenin and LEF-1, which in turn plays a key role in specifying the hair follicle fate [4,29,30,36,37]. Sonic hedgehog (Shh) and TGF-β2 signaling also play essential roles in follicle morphogenesis, but in contrast to β-catenin null skin, in which follicle invaginations are absent [30], some hair buds still form in the absence of LEF-1, Shh, or TGF-β2 [32,38]. These likely reflect the first wave of follicle (i.e., guard hair) morphogenesis, which accounts for a small number (fewer than 5%) of hairs and is under distinct regulatory control. Guard hairs form in the absence of LEF-1 and TGF-β2, and we have found that they also fail to express Snail at the budding stage of development (unpublished data). How E-cadherin is regulated in guard hairs remains to be determined. Several candidates include other Snail family members such as Slug or twist, or alternatively, transcription factors involving β-catenin and a different member of the LEF-1/TCF/Sry-type HMG box (commonly known as SOX) family [39,40]. Further investigation will be required to determine whether the signaling pathway we have elucidated here is a theme with multiple variations.

TGF-βs are known to promote withdrawal of keratinocytes from the cell cycle [41]. Hence, when TGF-β2 protein was detected at the transition between the growing and destructive phases of the adult hair cycle, research initially and naturally focused on a role for this family member in cessation of growth and/or triggering apoptosis ([42] and references therein). However, in contrast to TGF-β1-null skin, which exhibits an extended growing phase of postnatal hair follicles, TGF-β2-null skin displays an embryonic block in follicle bud progression [32]. Although this phenotype is consistent with TGF-β2's embryonic expression patterns [33], about 50% of TGF-β2 null buds appear unable to progress to the down-growth phase, a feature that cannot be explained readily on the basis of previously established effects of TGF-βs.

Our finding that TGF-β2 is upstream from Ki67 expression and MAPK activation lends further support to the notion that hair follicle keratinocytes at this early stage of development react to TGF-β2 signaling in a fashion opposite to that typically expected for TGF-β factors. This said, based upon pSMAD2 immunohistochemistry, the immediate steps of downstream signaling appeared to be intact. Thus, we surmise that the proliferative outcome is likely to be rooted in differences in the repertoire of activated SMAD target genes. In this regard, the positive effects of TGF-β2 on proliferation within the hair bud may be more analogous to what has been seen in progression of squamous cell carcinoma to metastatic carcinoma [43] rather than that typically observed for keratinocytes [44,45,46].

The prior identification of the Snail gene as a potential target of TGF-β signaling [15] was intriguing, given the temporal wave of Snail gene expression that occurs in the developing hair bud. The additional correlation between epithelial hyperproliferation and Snail transgene expression further fostered our interest in a possible link between TGF-β2 and Snail. Our functional studies demonstrate that without TGF-β2, Snail expression is abolished in the mutant hair buds, and conversely, in K14-Smad2 skin, Snail is ectopically activated. Moreover, our in vitro studies indicate that even sustained TGF-β2 exposure may cause only a transient induction of Snail, offering a possible explanation as to why Snail is so briefly expressed during hair follicle morphogenesis. An additional point worth mentioning is that prolonged expression of Tg Snail in postnatal skin resulted in morphological and biochemical signs of epithelial to mesenchymal transitions (unpublished data), underscoring why transient Snail expression may be so important during normal hair follicle morphogenesis [18].

At first glance, the sparsity in hair coat of K14-Snail Tg mice seemed indicative of a defect in follicle formation (see Figure 2A). Closer inspection, however, revealed that not all hairs penetrated the hyperthickened Tg epidermis. Several factors may contribute to the seemingly normal follicle development in these mice. One obvious factor is the K14 promoter, which is elevated in the basal layer of the epidermis and the outer root sheath (ORS) of the hair follicle, but is markedly down-regulated in developing embryonic hair buds as well as in the postnatal hair progenitor cells. The K14 promoter is also less active in the ORS than epidermis and hence this might also account for the lack of apparent response of the ORS to ectopic Snail. Additional contributing factors could be the multiplicity of Snail family members and their differential expression, the saturation, and/or diversity of regulatory mechanisms that govern AJ formation, migration, and proliferation in the follicle ORS. Distinguishing between these possibilities must await the generation of mice harboring skin-specific loss-of-function Snail mutations.

We thank M Takeichi, P Coulombe, G Longmore, and L Gold for sharing their antibodies, and AG de Herreros for the Snail construct. L Degenstein and L Polak from the Fuchs laboratory provided outstanding technical assistance with transgenic work and animal husbandry. T Doetschman and Y Chai are acknowledged for the TGF-β2 KO and K14-Smad2 mice, respectively. We thank additional members of the Fuchs lab for generously sharing their reagents and ideas. CJ was partially supported by a fellowship from the Helen Hay Whitney Foundation. EF is an investigator at the Howard Hughes Medical Institute. This work was supported by the Howard Hughes Medical Institute and a grant from the National Institutes of Health.

Competing interests. The authors have declared that no competing interests exist.