New insights into cadherin function in epidermal sheet formation and maintenance of tissue integrity

Abstract

Co-expression and gene linkage have hampered elucidating the physiological relevance of cadherins in mammalian tissues. Here, we combine conditional gene ablation and transgenic RNA interference to uncover new roles for E- and P-cadherins in epidermal sheet formation in vitro and maintenance of epidermal integrity in vivo. By devising skin-specific RNAi technology, we demonstrate that cadherin inhibition in vivo impairs junction formation and intercellular adhesion and increases apoptosis. These defects compromise epidermal barrier function and tissue integrity. In vitro, with only E-cadherin missing, epidermal sheet formation is delayed, but when both cadherins are suppressed, defects extend to adherens junctions, desmosomes, tight junctions and cortical actin dynamics. Using different rescue strategies, we show that cadherin level rather than subtype is critical. Finally, by comparing conditional loss-of-function studies of epidermal catenins and cadherins, we dissect cadherin-dependent and independent roles of adherens junction components in tissue physiology.

Adherens junctions (AJs) function in the dynamic regulation of intercellular adhesion (1). The epithelial AJ transmembrane core is composed of E-cadherin (Ecad), whose ectodomain binds Ca²⁺ to mediate transcadherin interactions between neighboring cells (2, 3). Its intracellular domain binds directly to p120-catenin and β-catenin, which in turn binds to α-catenin. Together, catenins regulate cadherin stability and coordinate associated actin dynamics to ensure efficient cell adhesion (1, 4).

E-cadherin, P-cadherin, α-catenin, p120-catenin and β-catenin have all been conditionally targeted for ablation in mouse epidermis, making it an excellent system to probe the physiological importance of AJ components (5–11). Given the well-established connection between cadherin/catenin mutations and cancers, it was not initially surprising to find that loss of AJ components predisposed skin to cancers. Curiously, however, obvious disruptions of intercellular adhesion are not always among the most striking defects arising from catenin deficiencies in skin (5, 6, 8, 12, 13). Thus, although loss of α-catenin disrupts epidermal adhesion, the hyperproliferation and inflammation associated with this loss do not rely upon the severing of junctions, but rather on perturbations in Ras-MAPK and NF-κB signaling. Moreover, epidermal adhesion is seemingly unaltered when β-catenin or p120-catenin is absent, and instead, respective defects in Wnt signaling and inflammation prevail.

Ascertaining classical cadherin function(s) in epidermis has been complicated by the up-regulation of P-cadherin (Pcad) in the basal layer when Ecad is absent, the tight linkage of Ecad and Pcad genes, and the presence of cadherin-catenin complexes at cell borders in the other AJ mutant mouse models (10, 11). Using a combined knockout/RNAi strategy, we have now overcome these difficulties and uncovered defects not present in single loss of function mutations for any of the AJ components. Our studies reveal new insights into the functional significance of Ecad and Pcad in the formation of cell adhesion complexes in keratinocytes in vitro and in maintenance of epidermal integrity in vivo.

Results

Generation of Cadherin-Deficient Epidermis in Mice.

Ecad and Pcad genes are tightly (approximately 45 kb) linked on chromosome 8 (14) and separated by a predicted ORF of unknown function (Fig. 1A). Since this precludes conventional strategies, we first engineered transgenic (Tg) mice expressing P-cadherin short hairpin RNA (shRNA) driven by the epidermal keratin 14 (K14) promoter (Fig. 1B). Three independent PcadRNAi Tg lines were established, each with similar levels of Pcad knockdown.

Fig. 1.

Generation of Pcad RNAi transgenic mice. (A) Schematic of the Pcad and Ecad chromosomal locus. (B) The K14-Pcad RNAi construct used to generate transgenic (Tg) mice. (C) Images of WT and PcadRNAi Tg littermates and WT and Pcad KO littermate adult mice. (D and F) Newborn (P0) tail skins of indicated mice were processed for indirect immunofluorescence with indicated Abs. Arrows denote downregulation of Ecad at sites of HF downgrowth in WT, but not Pcad KO or PcadRNAi Tg mice. (E) Immunoblot analyses of P0 epidermal lysates with indicated Abs; α-tub (α-tubulin). epi, epidermis; der, dermis; hf, hair follicle. Dotted lines represent junction between epidermis and dermis. (Scale bars, 10 μm.)

PcadRNAi Tg mice were viable and displayed no obvious skin phenotype even though epidermal P-cadherin was significantly reduced (Fig. 1 C–E). Histologically and biochemically, Tg skin appeared normal except that Ecad was not downregulated at sites of hair follicle (HF) downgrowth (15, 16) (Fig. 1F). This alteration was also seen in mice lacking Pcad (Pcad KO) (7), underscoring the specificity of the PcadRNAi transgene in vivo. Each of these PcadRNAi Tg lines were then mated to our Ecad conditional knockout (cKO) mice (10) to generate mice deficient for epidermal cadherins.

Novel Phenotypic and Morphological Perturbations in the Absence of Epidermal Cadherins.

K14-Cre/Ecad(fl/fl)/K14-PcadRNAi (cKO/Tg) mice displayed dramatic phenotypic abnormalities which distinguished them from their single loss-of-function counterparts (Fig. 2A). cKO/Tg mice were atypically small and died 1–2 h after birth. Their skin was shiny like that of Ecad cKO mice, but it was also taut and inflexible and showed considerable ventral flaking. Additionally, their skin blistered on the paws and around the mouth, umbilicus, and tail (arrows). Moreover, newborn (P0) cKO/Tg pups aberrantly adsorbed dye through their paws, facial skin, ear buds and lower belly, reflective of a defective epidermal barrier (Fig. 2B).

Fig. 2.

Inhibition of classical cadherins in mouse epidermis results in cell dissociation, blistering skin lesions, and defective epidermal barrier. PcadRNAi Tg and K14-Cre/Ecad cKO mice were bred to generate P0 mice WT or conditionally null for the Ecad allele and either negative (Ecad cKO) or positive (cKO/Tg) for the PcadRNAi transgene. Arrows in A denote blistered skin lesions, seen only in cKO/Tg mice. Barrier function analyses in B assessed by exclusion of an XGAL-containing solution. Arrows indicate areas, found only in cKO/Tg mice, where skin barrier function is disrupted and endogenous β-galactosidase converts dye to blue. Immunofluorescence in C shows absence of P-cadherin in cKO/Tg backskin epidermis (epidermal Ecad is completely absent in Ecad cKO mice) (10). H&E staining in D reveals that such regions lacking both cadherins exhibit gross perturbations in tissue integrity.

Pcad was not detected over large regions of cKO/Tg skin (Fig. 2C). Despite germline transmission, and for reasons unclear, F1 pups from different Tg lines displayed patches of Pcad(+) skin (≈5% of total backskin, data not shown) providing an internal control for our analyses. Areas devoid of cadherins displayed severe histological abnormalities not evident in Ecad cKO epidermis or in cKO/Tg regions lacking only Ecad. HFs were fewer and were either Pcad(+) or severely stunted (see below). Most notable was a loss of epidermal integrity accompanied by an apparent hyperthickening within areas lacking both cadherins (Fig. 2D).

Ultrastructurally, distortions in cellular morphology were sufficiently severe to account for the observed epidermal hyperthickening [Fig. 3A and supporting information (SI) Fig. S1A]. Notably, the typical columnar orientation of cells within the basal layer was largely lost in cadherin-deficient epidermis as was the flattened squamous morphology typical of WT suprabasal cells. Although most intercellular membranes appeared to be sealed, focal gaps compromised the continuity of the epithelial sheets (Fig. 3 A and B).

Fig. 3.

Ultrastructural abnormalities in cadherin-deficient epidermis. (A) 0.8 μm semithin sections of P0 backskins were stained with toluidine blue. Asterisks denote intercellular gaps between keratinocytes of cKO/Tg epidermis. Also note epidermal hyperthickening and marked disorganization and altered cuboidal morphology of basal cells in cKO/Tg skin. (B and C) Transmission electron microscopy. Asterisks in (B and B′) and double arrows in (B′) denote intercellular gaps. Despite gaps and signs of cellular degeneration, double membranes appear to persist between cKO/Tg keratinocytes. Note also that desmosomes (Dm), which were not reduced in number, appear to be intact even in areas where intercellular gaps occurred (B′). Region in C contrasts the normal desmosome-keratin filament network of suprabasal cells of WT epidermis versus the irregular aggregates of keratin filaments (Kf) that were frequently observed in cKO/Tg epidermis. Such alterations in keratin organization frequently reflect defects in mechanical integrity. Additional abbreviations: BL (basal layer); Sp (spinous layer); Gr (granular layer); SC (stratum corneum). Dotted lines in A and B represent junction between epidermis and dermis. (Scale bars: A, 5 μm (A); B, 2 μm; B′, 500 nm; C, 5 μm.)

Interestingly, desmosomes (Dms) still formed in normal numbers (Fig. 3B′ and data not shown). Intercellular gaps seemed to arise not from splits within Dms, but rather from degeneration of one of the two neighboring cells (asterisks in Fig. 3B′). Such perturbations increased Dm density in some areas and decreased it in others, suggestive of a collapse in cellular integrity. Dissociated suprabasal cells occasionally displayed condensed nuclei, indicative of apoptosis (data not shown). Finally, suprabasal keratin intermediate filament (IF) organization was perturbed (Fig. 3C and Fig. S1B). Overall, these features suggested an increased mechanical fragility within cKO/Tg epidermis.

Global Junctional Perturbations Without Epidermal Cadherins.

Classical cadherins appeared to be effectively targeted in cKO/Tg epidermis, since all other AJ components failed to localize to cell borders (Fig. 4 A and G and Fig. S2 A and B). Actin organization was also aberrant particularly at sites of suprabasal cell-cell contacts (Fig. 4B). The disorganized and often discontinuous cortical actin belts could explain why suprabasal cells failed to adopt the flattened squamous shape of their WT counterparts. Signs of actin disorganization were also evident in the basal layer, as reflected by altered cellular organization, and by discontinuous basal and atypical spinous layer immunolocalization of hemidesmosomal integrin β4 (Fig. 4B).

Fig. 4.

Perturbations in intercellular junctions, cytoskeleton organization, and tight junction function in epidermis lacking E- and P-cad. (A–F) P0 skins from tail (D) and back (all others) were processed for fluorescence microscopy with indicated Abs or TRITC-Phalloidin (actin, red). Additional Ab abbreviations: α-cat (α-catenin); β4 (β4 integrin), hemidesmosomal component restricted to the base of the basal layer; DP (desmoplakin); K1 (keratin 1), component of suprabasal IF network; ClaI (claudin 1); Occl (occludin). Asterisk in A denotes nonspecific 2⁰Ab staining of cornified layer. Abnormalities unique to cKO/Tg skin are denoted by: Arrows in B and D, gaps between basal and suprabasal cells; arrowhead in B, absence of cortical actin network between two suprabasal keratinocytes; asterisk in B, expansion of β4 integrin localization into spinous layers (cornified cell staining is nonspecific as per above); arrowhead in D, non-uniform staining patterns of keratin in some suprabasal keratinocytes. (F) Inside-out permeability assessed by monitoring impedance to biotin flow at TJs in granular layer. Arrows indicate occludin-based TJs. Arrowheads denote biotin flow past TJs in cKO/Tg epidermis only. (G) Immunoblot analysis of total P0 epidermal lysates with indicated Abs; Pan Cad (pan cadherin); β-cat (β-catenin); p120 (p120-catenin); Afad (afadin); PG (plakoglobin). (Scale bars: 10 μm.)

Dm cadherins (desmocollin-2 and desmoglein) and the keratin IF-Dm linker protein (desmoplakin) still localized to cell borders, and total levels of Dm proteins were unchanged (Fig. 4G and data not shown). Consistent with our ultrastructural findings, however, immunolabeling was discontinuous (Fig. 4C and Fig. S2 C and D). Similarly, suprabasal keratin 1 (K1) immunolocalization was often perinuclear (Fig. 4D), reflective of the abnormal intracellular aggregates of keratin IFs observed ultrastructurally.

Cadherin deficiency also resulted in defects in intercellular border localization but not overall levels of tight junction (TJ) proteins claudin 1, occludin, and ZO-1 (Fig. 4 E and G and Fig. S2 E and F). TJs appeared to be functionally compromised, since upward biotin flow was not blocked at the granular layer (Fig. 4F), i.e., where TJs are normally assembled (17). Although a TJ defect was noted in a different strain of Ecad cKO mice (11), the epidermis from our Ecad cKO strain showed biotin flow restriction. Strain-specific differences aside, the striking differences in single versus double cadherin inhibition underscored the importance of cadherins in overall formation and/or stability of TJs.

Recent studies suggest that cadherin-based cell adhesion may function not only in Dm and TJ assembly but also in epithelial cell polarity (18). Interestingly, cadherin-deficient epidermis displayed defective localization of Par3, aPKC, and Scribble (Fig. S3 A–C). By contrast, loss of Ecad alone did not alter Par3 (11) or Scribble distribution (Fig. S3 A and C). While our current understanding of polarity-regulating complexes in stratified epithelia is limited, these findings suggest that cadherins are required for their proper localization. These findings lend physiological relevance to in vitro connections previously identified between AJs and polarity proteins of the Scribble complex (19).

Additional Insights from in Vitro Studies.

The defects in intercellular junction assembly and actin dynamics were further characterized in vitro by infecting Ecad cKO primary mouse keratinocytes (1⁰MK) with a retrovirus expressing short hairpin (sh) RNAs against Pcad (Fig. 5 C and D). Loss of Ecad alone delayed Ca⁺²-induced AJ formation and assembly of 1⁰MK into a continuous sheet (Fig. 5A and Fig. S4A). However, without Pcad and Ecad, catenins weren't recruited to cell contacts and sheets failed to form even after 24 h exposure to high Ca²⁺ (Fig. 5E and Fig. S4B). Additionally, F-actin was not organized properly and Dm and TJ constituents no longer localized to cell borders (Fig. 5E). Each of these defects was rescued by a P-cadherin-cherry mRNA that was silencing-resistant, underscoring the specificity of the shRNA (Fig. 5E and Fig. S4 C and D).

Fig. 5.

Overall cadherin level governs epidermal sheet formation in vitro. (A–E) Confluent monolayers of WT and Ecad KO primary mouse keratinocytes (1⁰MKs) alone or stably expressing indicated constructs were shifted from low (0.05 mM) to high (1.5 mM) Ca²⁺ media for indicated times before processing for fluorescence microscopy with Abs or Alexa 647-phalloidin (actin, red) as indicated and immunoblot analysis of total lysates (D). Actin in A and B represents epifluorescence from K14-GFPactin (28) 1⁰MKs WT or null for Ecad. Arrows in all images indicate sites of cell-cell interactions. (A) Despite increased Pcad at puncta, Ecad KO 1⁰MKs display delayed kinetics of catenin localization and actin organization. (B) Expression of either Ecad-GFP or Pcad-GFP rescues the early delay in epidermal sheet formation in Ecad KO 1⁰MKs. Asterisks denote sites of interactions between Ecad KO cells that do not express cadherin-GFP. (C and D) Immunofluoresence and immunoblot analysis of PcadRNAi in vitro. (E) Inhibition of Ecad and Pcad blocks epidermal sheet formation in vitro and this is rescued by expression of silencing resistant Pcad-cherry fusion protein. Anti-RFP antibody was used to detect mutant Pcad protein. Asterisks specifically indicate sites of interactions between Ecad KO + PcadRNAi cells lacking rescue construct. (Scale bars: 10 μm.)

Based upon these data, we conclude that sustained junctional defects and failure in sheet formation are a consequence of loss of cadherin function. Moreover, two lines of evidence suggested that the delay in epidermal sheet formation observed with Ecad KO 1⁰MK was due to a reduction in overall cadherin level rather than Ecad loss per se. First, endogenous Pcad levels rose steadily in both WT and Ecad KO 1⁰MK following Ca⁺² exposure, and this correlated with sheet assembly (Fig. S4 F and E). Second, Ecad and Pcad GFP fusion proteins both rescued the delayed kinetics of Ecad KO 1⁰MKs (Fig. 5B).

Dissecting AJ-Dependent vs. AJ-Independent Defects.

The adhesion and junctional defects associated with cadherin loss resembled those of α-catenin cKO skin (20). Additionally, although we previously noted mild alterations in Ecad cKO-associated epidermal differentiation (10), such defects were not detected in the Ecad cKO or cadherin-deficient strains used here (Fig. S3 D–F) nor in α-catenin cKO mice (6).

By contrast, in both cadherin and α-catenin deficient epidermis, apoptosis was increased (Fig. 6 A–D), a finding also documented in Ecad-deficient mammary gland (21). Cells positive for TUNEL and activated caspase 3 were most often suprabasal and dissociated from neighbors (Fig. 6 A and B; Fig. S5 A and B). Since apoptosis was not enhanced in desmoplakin cKO epidermis (Fig. S5B) (22), the effects appeared to be specific to AJ formation rather than intercellular adhesion per se.

Fig. 6.

In vivo cadherin and catenin inhibition results in an increase in epidermal apoptosis, yet loss of epidermal cadherins does not perturb proliferative and inflammatory responses. (A and B) P0 backskins and E18.5 embryos, respectively, were processed for labeling of fragmented DNA via TUNEL. (C and D) Quantification of active-caspase 3 (pCasp3) immunofluorescence of indicated littermate samples. Data were collected from two independently processed sets of animals. Results are shown as percent anti-pCasp3 immunoreactive cells of the total epidermal cells counted. Asterisks denote significant difference from WT cells determined by t test: (C): P < 1 × 10⁻⁵; (D): P < 1 × 10⁻¹⁰. Error bars represent SD. (E and H) P0 backskins were processed for indirect immunofluorescence and immunohistochemistry with indicated Abs. (F) Quantification of BrdU-labeling experiments. P0 mice were injected s.c. with 50 μg/g wt BrdU and killed 2 h. later. Data were collected from two independently labeled sets of animals. Results are shown as percent BrdU-labeled cells of the total epidermal cells counted. Error bars represent SD. (G) Total P0 epidermal lysates were processed for immunoblot analysis with indicated Abs. (Scale bars: 10 μm.)

The hyperthickened epidermis associated with cadherin loss suggested that the skin might be hyperproliferative as well. Surprisingly, however, incorporation of the thymidine analog BrdU revealed no significant differences in the number or basal location of epidermal cells actively in S-phase (Fig. 6 E and F). Similarly, immunofluorescence with the mitosis marker phospho-histone H3 revealed no differences in basal location or numbers of positive cells, and no changes were detected in activated MAPK levels (Fig. S5C and Fig. 6G). Although keratin 6 was induced (Fig. S5D), this is only a broad indicator of perturbed epidermal biology and not hyperproliferation per se (23).

The lack of proliferative defects in cadherin-deficient skin contrasted with α-catenin and/or p120-catenin cKO skins (5, 6). An additional perturbation arising from α-catenin and p-120 catenin loss (5, 12) was a striking inflammatory cell infiltrate and epidermal NF-κB activation. Surprisingly, however, neither inflammatory cell recruitment nor NF-κB activation were features of cadherin-deficient epidermis (Fig. S5 E and F and Fig. 6H).

Discussion

Conditional targeting of a specific shRNA in vivo represents a technological advance that should be broadly applicable in the future. By removing cadherins in vivo, subsequent to intercellular junction formation, we have uncovered a dependency of TJs on classical cadherins that has not been evident from in vitro studies (24, 25). In addition, even though Dms formed, the Dm-keratin IF network appeared no longer able to provide mechanical integrity. Interestingly, a collapse in epidermal architecture rather than a disruption in intermembrane sealing appeared to be at the root of the defects observed in cadherin-deficient epidermis. It is tempting to speculate that the early perturbations in AJ-cortical actin cytoskeleton might alter the stability of the underlying membrane, which in turn could compromise overall cellular architecture. Alterations in actin dynamics may also be responsible for the failure of cadherin-deficient 1⁰MK to assemble intercellular junctions de novo.

The loss of both Ecad and Pcad unveiled defects in intercellular adhesion, survival and epidermal integrity that were not present in skin lacking only one of these cadherins. To some extent, these newfound defects resembled those of α-catenin-deficient skin (6), and it was notable that cortical α-catenin was selectively diminished only when both cadherins were missing. That said, the Pcad/Ecad and α-catenin mutant mice differed in whether they localized cadherin-β-catenin complexes at cell borders, and this difference may account for other distinctions in the observed phenotypes.

Thus, although epidermal hyperthickening was observed in both mutants, the hyperthickening arising from cadherin inhibition was not accompanied by significant changes in proliferation or MAPK activation. Additionally, while intercellular adhesion was compromised in both cases, only the α-catenin cKO mice displayed an inflammatory cell infiltrate and enhanced epidermal NF-κB activation (12). Together, our in vivo studies suggest that proliferative and proinflammatory defects arising from loss of α-catenin within epidermis are independent of cadherin-mediated adhesion and provide strong support for our prior in vitro analyses (6, 12).

In summary, we have identified new roles for epidermal cadherins in mediating effective intercellular junction formation in keratinocytes in vitro and in maintaining epidermal tissue integrity in vivo. Although further studies will be necessary to fully appreciate the functional parallels between cadherins and α-catenin, our results suggest that these AJ components function coordinately in mediating keratinocyte adhesion, yet differ in their ability to influence proliferative and inflammatory responses in skin. These observations are interesting in light of the many cancers involving alterations in the expression of both cadherins and catenins and suggest that the loss of one may not functionally equate to the loss of the other.

Materials and Methods

shRNA, Constructs and Generation of Mice.

Pcad shRNAs were designed by J. Silva and cloned into pMLP vector (J. Silva, Cold Spring Harbor, Cold Spring Harbor, NY) containing microRNA 30 adaptor sequences (see SI Materials and Methods) (26). Ecad-GFP (A. Vaezi and E. Fuchs, Rockefeller University, New York, NY) was PCR amplified and cloned into pMSCVpuro (Clontech). Pcad cDNA (M. Takiechi, RIKEN, Kobe, Japan) was amplified by PCR, cloned into pEGFP-N1 (Invitrogen) and subsequently subcloned into pMSCVpuro. 4 bp changes were introduced in Pcad cDNA to generate silent mutations within the region targeted by PcadRNAi through Quikchange site directed mutagenesis kit (Stratagene). This was PCR amplified and cloned into a modified retroviral vector pMSCVhyg (Clontech) expressing the cDNA of cherry fluorescent protein (R. Tsien, University of California at San Diego, La Jolla, CA) to generate a C-terminal tagged mutant Pcad-cherry protein. All constructs were confirmed by DNA sequencing.

Generation of mice expressing K14-Pcad shRNA are described in SI Materials and Methods. Three independent PcadRNAi Tg lines were maintained and mated to Ecad cKO mice (10) to generate epidermal-specific cadherin-deficient mice. K14-Cre Tg (27), K14-GFPactin Tg (27, 28), α-catenin cKO (6), desmoplakin cKO (22) and P-cadherin^−/− (7) mice have been previously described.

Tissue Processing and Analyses.

Tissues were frozen, fixed, sectioned and subjected to immunofluorescence (10). Additional Abs: P-cadherin (1:300; R&D systems), N-cadherin (Invitrogen), pan-cadherin (Sigma), l/s-afadin (Sigma), pan-desmoglein (1:300; Fitzgerald), desmocollin 2 (1:300; Fitzgerald), occludin (Invitrogen), claudin 1 (Invitrogen), ZO-1 (Invitrogen), Par3 (Upstate), aPKCλ (Santa Cruz), Scribble (Santa Cruz), RFP (MBL International), phospho-histone H3 (1:300; Millipore), BrdU (1:300; Abcam), active caspase 3 (R&D systems), CD3 (Chemicon), CD11b (1:25; BD Biosciences), active MAPK (Sigma), phospho-NF-κB (Cell Signaling), α-tubulin (1:1,000; Serotec). Unless otherwise stated, 1⁰Abs were used at 1:100. 2⁰Abs and dilutions were: Alexa Fluor 488 (1:1,000; Invitrogen), rhodamine RedX and cy5 (1:200; Jackson Labs). For immunohistochemistry, biotin-conjugated 2⁰Abs (1:100) were used and developed using the Vectastain ABC kit and DAB substrate (Vector Labs). Additional reagents were TRITC/Alexa Fluor 647-Phalloidin (Sigma/Invitrogen), ApopTag fluorescein direct in situ apoptosis detection kit (Chemicon), and 4′6′-diamidino-2-phenylindole (DAPI) to label nuclei.

For transmission EM, tissues were fixed, processed and visualized as described (29).