Research ArticleCancer

α-Catenin Is a Tumor Suppressor That Controls Cell Accumulation by Regulating the Localization and Activity of the Transcriptional Coactivator Yap1

See allHide authors and affiliations

Science Signaling  24 May 2011:
Vol. 4, Issue 174, pp. ra33
DOI: 10.1126/scisignal.2001823

Abstract

The Hippo pathway regulates contact inhibition of cell proliferation and, ultimately, organ size in diverse multicellular organisms. Inactivation of the Hippo pathway promotes nuclear localization of the transcriptional coactivator Yap1, a Hippo pathway effector, and can cause cancer. Here, we show that deletion of αE (α epithelial) catenin in the hair follicle stem cell compartment resulted in the development of skin squamous cell carcinoma in mice. Tumor formation was accelerated by simultaneous deletion of αE-catenin and the tumor suppressor–encoding gene p53. A small interfering RNA screen revealed a functional connection between αE-catenin and Yap1. By interacting with Yap1, αE-catenin promoted its cytoplasmic localization, and Yap1 showed constitutive nuclear localization in αE-catenin–null cells. We also found an inverse correlation between αE-catenin abundance and Yap1 activation in human squamous cell carcinoma tumors. These findings identify αE-catenin as a tumor suppressor that inhibits Yap1 activity and sequesters it in the cytoplasm.

Introduction

The proliferation and differentiation of adult stem and progenitor cells are tightly controlled to generate and then maintain an appropriate number of cells, which is necessary for proper organ function while preventing tumor formation. How stem and progenitor cells measure and accordingly control their rates of accumulation and differentiation is a key question in stem cell biology. It has been proposed that cell-cell adhesion structures called the adherens junctions can be used by cells to measure and regulate the local cell “crowdedness” (1, 2). At the core of the adherens junctions are the transmembrane cadherins, which form intercellular bridges that are bound intracellularly to β-catenin and functionally linked to the actin cytoskeleton through α-catenin (3). β-Catenin, a structural component of adherens junctions and a transcriptional coactivator in the canonical Wnt signaling pathway, represents one putative link between adherens junctions and the regulation of cell density (4). The constitutive activation of the Wnt pathway is oncogenic and causes tumor development in various organs (5); therefore, sequestering β-catenin to adherens junctions during conditions of high cell density may normally function to decrease rates of cell proliferation by inhibiting Wnt pathway signaling. Here, we propose a mechanism that couples a different adherens junction protein, α epithelial (αE) catenin, to the regulation of cell accumulation and cancer.

αE-catenin (which is encoded by Ctnna1) is required for adherens junction formation in multiple epithelial cell types (68), where it functionally links the cell membrane–localized cadherin-catenin adhesion complexes to the actin cytoskeleton (9). Intriguingly, αE-catenin is often not detectable or shows decreased abundance in various primary human tumors and tumor-derived cell lines through deletion, inactivating mutations, or promoter methylation (10, 11). Skin keratinocytes lacking αE-catenin display loss of contact-mediated inhibition of cell proliferation (12), and xenografts containing these keratinocytes result in the formation of skin lesions in nude mice that resemble squamous cell carcinoma (13). Moreover, reexpression of αE-catenin in αE-catenin–negative cell lines attenuates their growth (11, 14). Tumor development in animals lacking a particular protein suggests a role for that protein as a tumor suppressor; however, due to the crucial role of αE-catenin in tissue integrity, previously used gene knockout strategies that targeted αE-catenin for deletion in embryonic progenitor cells resulted in embryonic or neonatal lethality (1, 8, 12). Therefore, the question of tumor suppressor activity of αE-catenin has remained unresolved.

The Hippo signaling pathway was discovered in Drosophila, where it plays a crucial role in regulating the size of the organs in developing embryos (15). Many genes involved in the Hippo pathway are evolutionarily conserved, and analysis of Hippo signaling in mammalian organisms has revealed its role in regulation of stem and progenitor cell accumulation and development of cancer (1622). The Hippo pathway senses local cell densities and controls tissue growth through a kinase cascade that culminates in phosphorylation of the transcriptional coactivator Yap1, a posttranslational modification that prevents its nuclear localization, thus blocking its activity (23). However, the molecular mechanisms that link the Hippo pathway to focal cell accumulation, particularly in the mammalian system, are not well understood.

Here, we show that genetic deletion of αE-catenin in the hair follicle stem cell compartment leads to development of squamous cell carcinoma. We also demonstrate that αE-catenin interacts with Yap1 and regulates its nuclear localization. Finally, our examination of human keratoacanthoma tumors revealed frequent loss of αE-catenin abundance, which correlated significantly with nuclear Yap1 localization.

Results

Conditional deletion of αE-catenin in the hair follicle stem cell compartment results in the formation of inflammatory skin lesions and squamous cell skin tumors

We generated and analyzed mice with a conditional deletion of αE-catenin in the hair follicle stem and progenitor cell niche (Fig. 1 and fig. S1). In the GFAP-Cre mouse line (24), Cre was activated in the bulge region of the hair follicle at postnatal day 2 (P2) (Fig. 1, A and B). Because the progeny of P2 bulge stem cells expand throughout the hair follicle, the gene-targeting event occurred in most of the epithelial cells lining the hair follicle at P6, except for hair matrix cells (Fig. 1C). In adult GFAP-Cre animals, all epithelial cells of the hair follicle that originate from the bulge stem cell niche were targeted for Cre-mediated recombination (Fig. 1D). GFAP-Cre/αE-cateninfl/fl mice were viable and fertile; however, they either were completely bald or displayed only patchy hair growth on their backs due to partially penetrant Cre expression in some of the animals (Fig. 1E). Sox9 is a specific marker of hair follicle stem and early progenitor cells (25). In GFAP-Cre/αE-cateninfl/fl mice, Sox9-positive hair follicle stem and early progenitor cells displayed loss of αE-catenin (Fig. 1, F and G). Wild-type hair follicles undergo cycles of growth (anagen), degeneration (catagen), and rest (telogen), which together are known as the hair cycle (26). Histological examination of skin from wild-type and GFAP-Cre/αE-cateninfl/fl mice at different time points after birth revealed defective hair follicle morphogenesis and formation of disorganized hair follicles that were unable to produce hair (fig. S2). Sox9-positive hair stem and progenitor cells persisted in disorganized hair follicles of GFAP-Cre/αE-cateninfl/fl mice, and significantly more Sox9-positive cells incorporated BrdU (bromodeoxyuridine), indicating that αE-catenin−/− stem and early progenitor cells cycle more actively than their αE-catenin+/+ counterparts (fig. S3). We concluded that deletion of αE-catenin in the hair follicle stem cell compartment resulted in abnormal hair follicle maintenance but did not affect viability, thus enabling the analysis of the long-term consequences of αE-catenin ablation.

Fig. 1

Conditional deletion of αE-catenin in hair follicles. (A) Model of growing hair follicle. Hair follicle stem cells localize to the bulge region. Matrix contains committed progenitors that differentiate and give rise to hair and inner root sheath. (B to D) Staining for LacZ activity in frozen sections from newborn P2, P6, and P30 GFAP-Cre/ROSA26Cretest mice. Red is nuclear fast red counterstain. (E) General appearance of 6-month-old αE-cateninfl/fl (Ctrl) and GFAP-Cre/αE-cateninfl/fl (α-cat cKO) mice. (F) Immunofluorescent staining of skin sections from P60 Ctrl and α-cat cKO mice with anti–E-cadherin (red) and anti–α-catenin (green) antibodies. Note loss of αE-catenin in hair follicles of α-cat cKO skin. (G) Immunofluorescent staining of skin sections from P35 Ctrl and α-cat cKO mice with antibodies against Sox9, a stem and early progenitor marker (red), and α-catenin (green). α-Catenin is absent in Sox9+ cells in α-cat cKO skin. Blue in (F) and (G) is nuclear DAPI (4′,6-diamidino-2-phenylindole) stain. Scale bars, 70 μm in (B); 190 μm in (C) and (D); and 47 μm in (F) and (G).

GFAP-Cre/αE-cateninfl/fl mice developed extensive skin lesions over time and needed to be euthanized with a half-survival time of ~10 months (Fig. 2, A and B). Histological analyses of the skin lesions showed skin inflammation and tumors with squamous cell differentiation that resembled human squamous cell carcinoma of the keratoacanthoma type (Fig. 2, C to P, and fig. S4, A to C and G to I). Tumors displayed expansion of the keratinocyte population with prominent signs of cellular atypia, intercellular bridges, and extensive extracellular keratinization. In addition, pearls of keratin that localized to the middle of concentric layers of squamous cells in tumor cell masses were prevalent (Fig. 2, F and H). The outer edges of tumor cell masses contained nondifferentiated proliferating cells, which stained for keratins 5 and 6 (Fig. 2, I to N). The inner layers of tumor cell masses were positive for involucrin, a marker of differentiated keratinocytes (Fig. 2P). Overall, the results obtained from immunostaining of tumor sections with cell type–specific markers were consistent with the histological diagnosis of keratoacanthoma squamous cell carcinoma. Tumors in GFAP-Cre/αE-cateninfl/fl animals formed in skin areas that displayed prominent inflammation; however, the onset of tumor formation varied among animals (fig. S4). Moreover, both the tumor and the surrounding areas of inflammation were negative for αE-catenin, indicating that they were derived from the αE-catenin−/− keratinocytes of the hair follicles (fig. S5).

Fig. 2

αE-catenin is a tumor suppressor in skin keratoacanthoma. (A) General appearance of 2-month-old αE-cateninfl/fl (Ctrl) and tumor-bearing GFAP-Cre/αE-cateninfl/fl (α-cat cKO) mice. (B) Kaplan-Meier survival curves of Ctrl (n = 19) and α-cat cKO (n = 59) mice. Statistical significance was determined by the log-rank test (P ≤ 0.0001). (C to H) H&E staining of skin sections from 2-month-old Ctrl and tumor-bearing α-cat cKO mice. Images in (E) to (H) show higher magnifications of the sections shown in (C) and (D). (I to P) Immunofluorescent staining of skin sections from 2-month-old Ctrl and tumor-bearing α-cat cKO mice with anti-BrdU (red), anti–keratin 5 [green in (I) to (L)], anti–keratin 6 [green in (M) and (N)], and anti-involucrin [green in (O) and (P)] antibodies. Blue in (I) to (P) is nuclear DAPI stain. Scale bars, 500 μm in (C) and (D); 214 μm in (E) and (F); 53 μm in (G) and (H); 75 μm in (I) and (J); 19 μm in (K) and (L); and 75 μm in (M) to (P).

GFAP-Cre/αE-cateninfl/fl/p53fl/fl mice display development of an early-onset multifocal keratoacanthoma

Because the onset of skin lesions varied considerably, we hypothesized that in addition to loss of αE-catenin, other genetic or epigenetic events had to take place before tumor development occurred in these animals. The inactivation of the gene encoding the tumor suppressor p53 occurs frequently in human keratoacanthoma (27, 28). We found that the abundance of p16Ink4A and p53 was increased in the skin of GFAP-Cre/αE-cateninfl/fl mice before they develop skin lesions (Fig. 3A). p53 elicits its tumor-suppressing function by activating apoptotic cell death or cell senescence (29, 30). Indeed, we found that a fraction of αE-catenin−/− cells was positive for apoptotic TUNEL (terminal deoxynucleotidyl transferase–mediated deoxyuridine triphosphate nick end labeling) staining (fig. S6A). To determine whether increased p53 signaling attenuates tumor development in GFAP-Cre/αE-cateninfl/fl mice, we generated and analyzed GFAP-Cre/αE-cateninfl/fl/p53fl/fl mice. Apoptotic cell death was significantly decreased in double-mutant GFAP-Cre/αE-cateninfl/fl/p53fl/fl animals (fig. S6, B and C). Furthermore, these mice developed completely penetrant, early-onset, multifocal keratoacanthomas (Fig. 3, B to F). Unlike GFAP-Cre/αE-cateninfl/fl mice, skin tumors in GFAP-Cre/αE-cateninfl/fl/p53fl/fl mice appeared at multiple sites and displayed no skin inflammation preceding tumor initiation (Fig. 3C and fig. S4). Therefore, it is likely that the deletion of p53 is completely sufficient to bypass the inflammatory response required for tumor development in GFAP-Cre/αE-cateninfl/fl mice. Although the exact cellular origin of human keratoacanthoma is unknown, this cancer is thought to originate in the upper portion of the hair follicle (31), which is the site of the hair follicle stem cell compartment targeted in our GFAP-Cre mice.

Fig. 3

Early-onset keratoacanthoma tumors in GFAP-Cre/αE-cateninfl/fl/p53fl/fl mice and loss of αE-catenin abundance in human keratoacanthomas. (A) Western blot analysis of total protein extracts from skins of αE-cateninfl/fl (Ctrl) and GFAP-Cre/αE-cateninfl/fl mice with anti-p53, anti-p16Ink4A, and anti–β-tubulin antibodies. n = 3. (B) Kaplan-Meier survival curves for GFAP-Cre/αE-cateninfl/fl/p53fl/fl (n = 30), GFAP-Cre/p53fl/fl (n = 19), and wild-type (Ctrl) (n = 19) mice. Curves for GFAP-Cre/αE-cateninfl/fl/p53fl/fl and GFAP-Cre/p53fl/fl mice are significantly different. P < 0.0001 (log-rank test). Because of GFAP-Cre activity in the brain, GFAP-Cre/p53fl/fl mice die from brain tumors; however, they do not have a skin phenotype at the time of euthanasia. (C) General appearance of 4-month-old GFAP-Cre/p53fl/fl (p53 cKO) and GFAP-Cre/αE-cateninfl/fl/p53fl/fl-cat + p53 cKO) mice. (D to F) H&E staining of skin tumor from 4-month-old GFAP-Cre/αE-cateninfl/fl/p53fl/fl-cat + p53 cKO) mouse. Images in (E) and (F) show higher magnifications of the section shown in (D). (G to L) Immunofluorescent staining of human keratoacanthoma tumor with anti–α-catenin (red) and anti–pan-cytokeratin (green) antibodies. αE-catenin is present in uninvolved areas but is not detectable in tumor cells. Blue in (I) to (L) is nuclear DAPI stain. Scale bars, 1 mm in (D); 0.4 mm in (E); 100 μm in (F); 30 μm in (G) to (I); and 7.5 μm in (J) to (L).

αE-catenin is lost or decreased in abundance in human keratoacanthoma

To determine whether the abundance of αE-catenin changes in human keratoacanthomas, we performed immunofluorescent staining with anti–αE-catenin antibodies on human tumors. Abundance of αE-catenin in the tumors was compared to its abundance in uninvolved areas of the epidermis in the same tissue sections. Of the 29 tumors analyzed, 3 (~10%) had αE-catenin abundance similar to that in uninvolved epidermis. Nineteen tumors (~66%) displayed reduced staining, and 7 tumors (~24%) showed staining for αE-catenin, comparable to background (Fig. 3, G to L). These data suggest that αE-catenin may be frequently decreased in abundance or lost in human keratoacanthomas, a finding consistent with the potential role for αE-catenin as a tumor suppressor in these neoplasms.

An siRNA screen reveals a functional connection between αE-catenin and the transcriptional coactivator Yap1

To elucidate the tumor-suppressing mechanisms of αE-catenin, we isolated and analyzed mouse αE-catenin−/− keratinocytes. These cells displayed loss of cell density–dependent inhibition of cell accumulation, a phenotype that was rescued by reexpression of αE-catenin (Fig. 4, A to B, and figs. S7 and S8).

Fig. 4

Yap1 is necessary for αE-catenin–mediated loss of contact inhibition and hyperplasia. (A) Growth curves of αE-cateninfl/fl (Ctrl) and αE-catenin−/− (α-cat−/−) cells expressing vector or full-length αE-catenin (Ctrl + α-cat, α-cat−/− + α-cat). Cell accumulation is inhibited in confluent Ctrl keratinocytes, but not in α-cat−/− keratinocytes, a phenotype that is rescued by reexpression of full-length αE-catenin. (B) αE-cateninfl/fl or αE-catenin−/− keratinocytes were plated at high density in siRNA-Lipofectamine mixture in triplicate and cultured for 5 days. Cell numbers at the end of culture were determined by MTT assay. Under these conditions, αE-catenin−/− cells were not contact-inhibited and show increased cell accumulation. This phenotype was replicated in αE-cateninfl/fl cells by transfection of siRNAs targeting αE-catenin. Bar graph shows means ± SD. n = 3. **P < 0.01 by t test. (C) Yap1 is necessary for the loss of contact inhibition in αE-catenin–null cells. Validation of the functional connection of Yap1 to αE-catenin identified by an siRNA screen (fig. S10), with siRNA oligos targeting disparate regions of Yap1 (fig. S11). Bar graph shows means ± SD. n = 3. **P < 0.01. (D) Validation of the specificity of Yap1 siRNA hit with reexpression of human Yap2, which is not targeted by anti-mouse Yap1 siRNAs. **P < 0.01; ***P < 0.001. (E to J) Representative images of confluent αE-cateninfl/fl (Ctrl) and αE-catenin−/−-cat−/−) keratinocytes stained with anti-Yap1 and anti–α-catenin antibodies. (K) Quantification of Yap1 localization illustrated in (E) to (J) and fig. S12. Bar graph shows means ± SD. Number of cells counted was >50 per condition. P value was determined by Mann-Whitney test. (L to O) H&E and anti-Yap1 immunohistochemical staining of skin sections from 8-month-old αE-cateninfl/fl (Ctrl) and GFAP-Cre/αE-cateninfl/fl-cat cKO) mice. (P to S) H&E and anti-Yap1 immunohistochemical staining of keratoacanthoma tumors from GFAP-Cre/αE-cateninfl/fl-cat cKO) and GFAP-Cre/αE-cateninfl/fl/p53fl/fl-cat + p53 cKO) mice. Blue in (G) and (J) is nuclear DAPI stain. Scale bars, 21 μm in (E) to (J); 35 μm in (L) to (O); and 50 μm in (P) to (S).

To identify signaling pathways involved in αE-catenin–dependent inhibition of cell proliferation, we performed a targeted small interfering RNA (siRNA)–mediated gene knockdown screen for 70 genes previously identified as encoding αE-catenin–interacting proteins or otherwise implicated in contact inhibition or cancer (figs. S8 to S10). siRNA targeting of most genes, including β-catenin (β-cat), resulted in minimal phenotypes; however, knockdown of Yap1 inhibited the proliferation of αE-catenin−/− cells (fig. S10). The specificity of the Yap1 siRNA was validated with siRNAs from a different supplier (Dharmacon) and that targeted against a different portion of the Yap1 gene (Fig. 4C and fig. S11) and by reexpression of human Yap2, which was not targeted by mouse Yap1–directed siRNAs (Fig. 4D). We conclude that Yap1 is necessary for the contact inhibition–defective phenotype of αE-catenin−/− keratinocytes.

αE-catenin negatively regulates Yap1 nuclear localization

The Hippo signaling pathway, which has a central role in the regulation of contact inhibition, organ size determination, and tumor suppression (22, 32), impinges on the nuclear localization of the transcriptional cofactor Yap1 (15). Therefore, we examined whether the activity of Yap1 was increased in αE-catenin−/− keratinocytes. In nonconfluent cells, Yap1 localized primarily to the nucleus in both αE-catenin+/+ and αE-catenin−/− cells (fig. S12). In contrast, confluent αE-catenin+/+ cells displayed a diffuse cytoplasmic and partially junctional localization of Yap1, whereas confluent αE-catenin−/− keratinocytes showed predominantly nuclear Yap1 staining (Fig. 4, E to K). Similarly, cellular fractionation experiments demonstrated a significant increase in the amount of nuclear Yap1 in αE-catenin−/− keratinocytes (fig. S13). Reexpression of αE-catenin in αE-catenin−/− cells rescued the cytoplasmic localization of Yap1 (fig. S14). In epithelial cells, αE-catenin localizes to both cell-cell junctions and cytoplasm (33). Confocal microscopy analysis revealed partial colocalization between cytoplasmic αE-catenin and Yap1 in confluent keratinocytes (fig. S15).

If Yap1 constitutively localizes to the nucleus in proliferating cells and is cytoplasmic in nonproliferating cells, the differences in nuclear Yap1 localization between αE-catenin+/+ and αE-catenin−/− cells could be due to the differences in their proliferation status. To address this possibility, we analyzed the proliferation status and Yap1 localization in serum-starved subconfluent αE-catenin+/+ and αE-catenin−/− cells by immunostaining with anti-Ki67, which is a marker of proliferative cells, and anti-Yap1 antibodies, respectively. We found that Ki67-negative keratinocytes maintained nuclear Yap1 localization (fig. S16). Therefore, nuclear Yap1 was not a marker of cellular proliferation, and differences in nuclear Yap1 between αE-catenin+/+ and αE-catenin−/− cells were not due to the differences in their proliferation status.

Yap1 phosphorylation on Ser127 by the MST1-LATS kinase cascade promotes Yap1 cytoplasmic retention and decreases its transcriptional activity (22). Analyses of total protein extracts with phosphospecific antibodies revealed that phosphorylation of Ser127 in Yap1 was decreased in αE-catenin−/− keratinocytes (fig. S17). However, when this was normalized to total Yap1 abundance, we found that the changes in specific phosphorylation were not significant (fig. S18; n = 5). Western blot and quantitative reverse transcription–polymerase chain reaction (qRT-PCR) analyses revealed that the abundance of total Yap1 protein and mRNA in αE-catenin−/− keratinocytes was decreased compared to that of αE-catenin+/+ cells (figs. S17 to S20), which may indicate activation of the Yap1-mediated negative feedback loop (34). In addition, Western blot analysis of αE-catenin−/− keratinocytes did not reveal a decrease in the phosphorylation and thus activity of LATS, the kinase that phosphorylates Yap1, to prevent its nuclear localization (fig. S17). Together, these data suggest that αE-catenin does not regulate nuclear localization of Yap1 by affecting the canonical Hippo kinase cascade, which converges on phosphorylation of Ser127 in Yap1.

We next analyzed the localization of Yap1 in vivo. Yap1 was predominantly localized in nuclei in hair follicle cysts and tumors in GFAP-Cre/αE-cateninfl/fl and GFAP-Cre/αE-cateninfl/fl/p53fl/fl mice (Fig. 4, L to S). We also found that Yap1 was prominently nuclear in αE-catenin−/− neural progenitor cells in Nestin-Cre/αE-cateninfl/fl mice (1), suggesting that the connection between αE-catenin and nuclear Yap1 is not restricted to keratinocytes (fig. S21).

We reasoned that if αE-catenin inhibits nuclear localization of Yap1, human keratoacanthoma tumors with low abundance of αE-catenin should show increased nuclear Yap1. Indeed, immunostaining of sections of human keratoacanthoma tumors with anti-Yap1 antibodies revealed a significant correlation between low αE-catenin abundance and nuclear Yap1 localization (Fig. 5, A to I).

Fig. 5

αE-catenin interacts with Yap1 and inhibits its nuclear localization and transcriptional activity. (A to H) H&E and anti-Yap1 immunohistochemical staining of human keratoacanthoma tumors with high and low abundance of αE-catenin. (I) Box-and-whisker plot showing quantification of differences in Yap1 abundance in keratoacanthoma tumors with low (n = 5) and high (n = 5) αE-catenin abundance. Cells were counted in three randomly selected images per tumor. Number of cells counted per image was >170. The plots show percentage of tumor cells displaying nuclear staining for Yap1. Lines within boxes, median values. Upper and lower borders of the boxes, 25th and 75th percentiles. Upper and lower bars, maximum and minimum values. Statistical significance was assessed by the Mann-Whitney test. Scale bars, 155 μm in (A), (B), (E), and (F); 50 μm in (C), (D), (G), and (H). (J) Coimmunoprecipitation of endogenous αE-catenin and Yap1. Proteins were extracted from cultured keratinocytes (IN) and immunoprecipitated (IP) with IgG controls, anti-Yap1, or anti–α-catenin antibodies and analyzed by Western blotting (WB) with anti–α-catenin or anti-Yap1 antibodies. (K and L) αE-catenin inhibits Yap1 transcriptional activity. Indicated constructs with or without siRNAs were cotransfected with a GAL4-TEAD firefly luciferase reporter and a Renilla luciferase control plasmid into HEK293FT cells for gain-of-function experiments (K) or wild-type mouse keratinocytes for loss-of-function experiments (L). Reporter luciferase activity was normalized to Renilla luciferase activity. Bar graph shows means ± SE. Statistical significance was determined by ANOVA with the Bonferroni post test. (M) Model showing the role of αE-catenin in regulating Yap1 localization and transcriptional activity.

αE-catenin interacts with Yap1 and inhibits its transcriptional activity

To analyze the potential mechanisms responsible for αE-catenin–mediated regulation of Yap1 localization, we performed an additional siRNA screen to target 34 known members of the Hippo pathway. However, except for Yap1, none of these genes affected cell accumulation (fig. S22). Therefore, we hypothesized that αE-catenin–mediated regulation of Yap1 nuclear localization may occur downstream, perhaps at the level of Yap1 itself. Because αE-catenin is normally localized to the cytoplasm and is required for the cytoplasmic distribution of Yap1, we asked whether αE-catenin could regulate its intracellular localization through direct binding. Coimmunoprecipitation experiments revealed an interaction between endogenous αE-catenin and Yap1 proteins in keratinocytes (Fig. 5J). Quantification of Western blots indicated that 11.6 ± 1.78% (n = 3) of endogenous Yap1 was associated with αE-catenin.

Yap1 interacts with the transcription factors TEAD, RUNX2, p73, and ErbB4 and activates transcription of various genes involved in regulating cell proliferation and apoptosis (35). To determine whether the interaction between Yap1 and TEAD is required for increased proliferation in αE-catenin−/− keratinocytes, we performed cell density–dependent inhibition of cell accumulation experiments with cells expressing wild-type or Ser94→Ala (S94A) forms of human Yap2. The S94A mutation prevents Yap2 from interacting with the TEAD family of transcription factors, but not other transcription factors (35). We found that wild-type, but not the S94A mutant, human Yap2 supported increased proliferation in αE-catenin−/− keratinocytes (fig. S23). Therefore, TEAD transcription factors may be critical partners of Yap1 in αE-catenin−/− keratinocytes.

To analyze whether αE-catenin inhibits Yap1 transcriptional activity, we performed Yap1 transcription coactivation assays with the TEAD family of transcription factors. In gain-of-function experiments in human embryonic kidney (HEK) 293 cells, overexpression of αE-catenin significantly decreased Yap1-mediated transcriptional activity in a dose-dependent manner (Fig. 5K). The extent of Yap1 signaling inactivation by αE-catenin was similar to inactivation caused by LATS1, a pivotal inhibitor of Yap1 signaling. In complementary loss-of-function experiments, knockdown of αE-catenin in wild-type keratinocytes resulted in an increase of Yap1-mediated transcriptional activity (Fig. 5L). Endogenous transcriptional targets of Yap1 are tissue-specific. Recently, Cyr61 was identified as a Yap1 transcriptional target in mouse skin keratinocytes (36). We found that the abundance of Cyr61 mRNA was significantly increased in skins and tumors of GFAP-Cre/αE-cateninfl/fl mice (fig. S24). Therefore, we conclude that αE-catenin inhibits Yap1-mediated transcription by interacting with Yap1 and promoting its cytoplasmic localization (Fig. 5M).

Discussion

αE-catenin is a tumor suppressor protein

We demonstrate that conditional deletion of αE-catenin in the skin hair follicle stem cell compartment results in the development of keratoacanthoma and, correspondingly, that αE-catenin is frequently absent in human keratoacanthoma tumors. These data provide genetic evidence of the tumor-suppressing function of αE-catenin. Although αE-catenin is frequently missing in various human epithelial and hematopoietic system tumors (10, 11, 37), it has been difficult to provide genetic, causal evidence of its tumor suppression in mice. Conditional deletion of αE-catenin during development in embryonic progenitor cells results not only in hyperplasia of the αE-catenin–null epidermis and brain, but also neonatal or perinatal lethality (1, 12, 38). In contrast, conditional deletion of αE-catenin in differentiated cells results in mild phenotypes with no reported tumor development (39, 40). Precisely targeting the deletion of αE-catenin to the bulge stem cell population after the major developmental events were already completed enabled us to establish direct genetic proof of the tumor-suppressing function of αE-catenin.

Similar to other known tumor suppressors, deletion of αE-catenin in stem and progenitor cells is not sufficient for immediate tumor development. We found that αE-catenin−/− hair follicle stem and progenitor cells could persist many months without signs of tumor initiation. Because we detected skin tumors in GFAP-Cre/αE-cateninfl/fl mice only in the areas displaying a prominent inflammatory response, inflammation may be a critical factor that cooperates with the loss of αE-catenin to allow direct tumor development. Indeed, inflammation plays a central role in human cancer development and progression (41).

Unlike the mice lacking only αE-catenin, the double-mutant mice lacking both αE-catenin and p53 displayed rapid and multifocal tumor development that initiated directly in the hair follicle, without signs of skin inflammation before tumor initiation. Therefore, it is likely that the deletion of p53 is sufficient to bypass the inflammatory response, which may be required for tumor development in GFAP-Cre/αE-cateninfl/fl mice. p53 attenuates tumor development by inducing cell cycle withdrawal and senescence or by activating programmed, apoptotic cell death (29). Although skin cells from GFAP-Cre/αE-cateninfl/fl and Cre/αE-cateninfl/fl/p53fl/fl mice did not show differences in cellular senescence, cells from Cre/αE-cateninfl/fl/p53fl/fl mice showed a significant decrease in apoptotic cell death (fig. S6). Hence, the canonical role of p53 in regulation of apoptotic cell death is the most likely explanation for the tumor-promoting function of p53 in GFAP-Cre/αE-cateninfl/fl/p53fl/fl mice.

Molecular mechanisms of αE-catenin function in regulation of cell accumulation

αE-catenin is required for contact-mediated inhibition of cell proliferation. Intriguing correlations between αE-catenin membrane localization, cellular crowdedness, and contact-mediated control of cell proliferation have resulted in the formulation of a “cell crowd control” hypothesis, which postulates that in progenitor cells, the adherens junctions link information about the local progenitor cell density to signaling pathways that influence proliferation or cell cycle withdrawal (1, 2). This could be a mechanism underlying tissue homeostasis, a process that is necessary to ensure that an appropriate number of cells are generated for adult tissue maintenance. If αE-catenin is involved in connecting information about stem cell niche crowdedness with cell cycle regulation, it is important to determine the molecular mechanisms responsible for this function, because they are likely to be linked to the tumor-suppressing function of this protein. To reveal these mechanisms, we used an unbiased approach and performed an siRNA screen for genes that may be involved in αE-catenin function in contact inhibition. This screen identified Yap1 as a critical component necessary for increased proliferation in αE-catenin−/− cells. Further analyses revealed that Yap is constitutively nuclear in αE-catenin−/− cells. In addition, we showed that αE-catenin bound to Yap1 and determined its subcellular (cytoplasmic or nuclear) distribution. Thus, we conclude that αE-catenin is a critical regulator of Yap activity, and abnormal activation of Yap1 signaling may be responsible for increased proliferation of αE-catenin−/− cells.

The Hippo pathway determines organ size in Drosophila and in mammalian organisms; it is implicated in the regulation of stem cell self-renewal and differentiation, as well as in cancer development (16, 20, 32, 35, 42). Similarly to αE-catenin, the Hippo pathway was previously implicated in the regulation of contact inhibition; however, the mechanisms connecting extracellular signals to Hippo signaling are not well understood (22, 43). Our study functionally and physically connects Yap1 with αE-catenin and suggests that Hippo signaling is regulated downstream of the canonical Hippo kinases by the binding of αE-catenin to Yap1 to promote the cytoplasmic retention of Yap1. The tight junction protein Amot has also been implicated in regulating contact-mediated proliferation and nuclear Yap1 localization by tethering Yap1 outside of the nucleus (44, 45). This function of Amot is similar to the function of αE-catenin described in this study. We found that siRNA-mediated targeting of Amot or Amot-like genes in keratinocytes did not affect contact-mediated inhibition of cell accumulation (fig. S22). Therefore, it appears that the function of Amot and αE-catenin in keratinocytes may be different.

To summarize, we show here that αE-catenin suppresses tumor development in murine skin epidermis and functions to regulate Yap activity. Because αE-catenin is an adherens junction protein, and adherens junctions are regulated by cell crowdedness, the connection between αE-catenin and Yap may potentially provide progenitor cells with a mechanism that can measure the local cell accumulation and adjust their rates of proliferation accordingly to ensure normal tissue homeostasis and to protect from tumor development.

Materials and Methods

Animals and labeling experiments

Homozygous αE-cateninfl/fl mice (12) were crossed with GFAP-Cre mice [Jackson Laboratory, FVB-Tg(GFAP-cre)25Mes/J] (24). To monitor Cre-mediated excision, we crossed mice with the reporter strain ROSA26Cretest [B6.129S4-Gt(ROSA)26Sortm1Sor/J; Jackson Laboratory] (46). To generate GFAP-Cre/αE-cateninfl/fl/p53fl/fl mice, we crossed GFAP-Cre/αE-cateninfl/fl animals with p53fl/fl mice [National Cancer Institute (NCI) Mouse Repository, 01XC2]. To identify cycling Sox9+ stem and progenitor cells, we injected adult mice intraperitoneally with BrdU (50 μg/g) and euthanized them after 1 hour. BrdU+ cells were detected in skin sections by immunofluorescent staining with anti-BrdU and anti-Sox9 antibodies.

Histology, immunofluorescent staining, and immunohistochemistry

Tissues for histology were fixed in 4% paraformaldehyde in phosphate-buffered saline, processed, and embedded in paraffin. Sections (5 μm) were stained with hematoxylin and eosin (H&E) and photographed with an Olympus BX41 microscope and a Microfire camera (Optronics) or a Zeiss LSM 510 confocal microscope. Sections were deparaffinized and processed for immunofluorescent staining and immunohistochemistry as previously described (47).

Antibodies, LacZ, and apoptosis staining

The antibodies used were anti-GFAP (glial fibrillary acidic protein) (Sigma, G3893), anti–β-catenin (Sigma, C2206), anti–α-catenin (Epitomics, 2028-1, or Sigma, C2081), anti–E-cadherin [Invitrogen, 13-1900, or rabbit polyclonal immunoglobulin Gs (IgGs) developed against the cytoplasmic domain of dog E-cadherin], anti-GFAP (Advanced ImmunoChemical, 031223), anti-Sox9 (Santa Cruz Biotechnology, sc-20095), anti-Ki67 (Novocastra, NCL-Ki67p), anti-BrdU (Developmental Studies Hybridoma Bank), anti–α6-integrin (BD Pharmingen, CD49f, clone GoH3), anti–keratin 5, anti–keratin 6, anti-involucrin (gift from J. Segre, National Institutes of Health–National Human Genome Research Institute), anti–pan cytokeratin (Sigma, C1801), anti-p120 catenin (BD Transduction Labs, 610133), anti–β-actin (Sigma, A5441), anti-Yap1, anti–phospho-Yap1 (Santa Cruz Biotechnology, sc-101199; Cell Signaling, 4912, 4911), anti-LATS1 (Cell Signaling, 3477), anti-LATS (S909) (Cell Signaling, 9157), anti-MST1 (T183)/MST2 (T180) (Cell Signaling, 3681), and anti-p21 (Calbiochem, OP76). LacZ staining was performed on 7-μm frozen skin sections as described (48). Apoptosis was detected with ApopTag Plus Peroxidase In Situ kit (Millipore). To determine senescence, we stained frozen skin sections for senescence-associated β-galactosidase with a senescence detection kit (Calbiochem) as previously described (49).

Primary keratinocyte culture, 96-well siRNA screen for contact inhibition, and cell fractionation

Keratinocytes were isolated as described (50). αE-cateninfl/fl keratinocytes were first established in low-calcium medium, and then later adapted and maintained in normal-calcium E medium (51). αE-catenin−/− cells were generated with AD5-CMV-Cre or Ad5-CMV-GFP (green fluorescent protein) adenoviruses (Vector Development Laboratory, Baylor College of Medicine). For αE-catenin rescue experiments, keratinocytes were infected with retrovirus containing full-length mouse αE-catenin and selected with G418. For the siRNA screen, 104 keratinocytes were plated directly into Lipofectamine 2000–siRNA mixture on 96-well plates. Each gene was targeted with a mixture of four independent siRNA oligos (Qiagen, 5 pmol per well, each). Cells were allowed to grow for 5 days after seeding; medium was changed daily, and final cell numbers were determined by manual cell counting or by MTT [3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide] assay (Promega). For validation, SMARTpool siRNA oligos targeting Yap1 were purchased from Dharmacon. Cytoplasmic and nuclear fractions were isolated with the NE-PER Nuclear and Cytoplasmic Extraction reagents (Pierce).

DNA constructs

Retroviral αE-catenin expression vector was generated by cloning full-length mouse αE-catenin into the multiple cloning site (MCS) of pLNCX2-GFP vector with Bgl II–Eco RI. pLNCX2-GFP vector was constructed by inserting the MCS-IRES-eGFP fragment from pBMN-I-GFP vector (G. Nolan, Addgene) into the Bam HI–Sal I sites of pLNCX2 vector. Retroviruses were produced with the Phoenix system (provided by G. Nolan, Stanford University, Stanford, CA) (52). Plasmids carrying Flag-tagged human Yap1 and Lats1 were generated by M. Sudol and obtained from Addgene (53). Retroviral human Yap2 expression construct was generated by J. Brugge and obtained from Addgene (54). GAL4-TEAD4 and glutathione S-transferase (GST)–Yap2 plasmids were generated by K.-L. Guan and obtained from Addgene (35). pUAS-Luc2 was generated by L. Luo and obtained from Addgene (55). Retroviral expression construct encoding S94A mutant human Yap2 was generated with a site-directed mutagenesis kit (Stratagene) and the following oligos: 5′-ggaagctgcccgacgccttcttcaagccg-3′ and 5′-cggcttgaagaaggcgtcgggcagcttcc-3′.

Luciferase assay

HEK293FT cells were seeded in 96-well plates. Firefly luciferase reporter, cytomegalovirus (CMV)–Renilla luciferase, and the indicated plasmids were cotransfected. Forty-eight hours after transfection, cells were lysed and luciferase activity was assayed with the Dual-Glo Luciferase Assay System (Promega) following the manufacturer’s instructions. Firefly luciferase activity was normalized to Renilla luciferase activity. All assays were done in quadruplicate.

RNA extraction and qRT-PCR

Total RNA was extracted with TRIzol (Invitrogen) and complementary DNA was prepared with SuperScript III First-Strand Synthesis kit (Invitrogen). qPCR was performed with Prism 7900HT (Applied Biosystems), platinum qPCR mix (Invitrogen), and Universal ProbeLibrary kit using the primers, probes, and PCR conditions recommended by the Universal ProbeLibrary assay center (http://www.roche-applied-science.com/sis/rtpcr/upl/adc.jsp). qPCR data were normalized to ribosomal protein Rps16.

Data analysis

Statistical significance was determined by the unpaired Student’s t test, analysis of variance (ANOVA) with the Bonferroni post test, or the Mann-Whitney test. P value is indicated by asterisks in the figures: *P < 0.05; **P < 0.01; ***P < 0.001. Differences at P = 0.05 and lower were considered statistically significant.

Supplementary Materials

www.sciencesignaling.org/cgi/content/full/4/174/ra33/DC1

Fig. S1. Generation of GFAP-Cre/αE-cateninfl/fl mice.

Fig. S2. Skin histology of GFAP-Cre/αE-cateninfl/fl mice.

Fig. S3. Increased rates of cell proliferation in αE-catenin−/− stem and early progenitor cells.

Fig. S4. Inflammatory response in hair follicle cysts and tumors from mice with conditional knockout of αE-catenin or of αE-catenin and p53.

Fig. S5. Tumors in GFAP-Cre/αE-cateninfl/fl mice are derived from αE-catenin−/− keratinocytes.

Fig. S6. Apoptosis in hair follicle cysts from mice with conditional knockout of αE-catenin or of αE-catenin and p53.

Fig. S7. αE-catenin is necessary for contact-mediated inhibition of cell accumulation.

Fig. S8. Ninety-six–well plate assay for contact inhibition of cell accumulation.

Fig. S9. Efficient knockdown of gene targets after siRNA transfection in 96-well plate contact inhibition assay.

Fig. S10. siRNA screen for genes required for αE-catenin–mediated contact inhibition.

Fig. S11. Efficient knockdown of Yap1 after transfection of keratinocytes with siRNA oligos.

Fig. S12. Nuclear localization of Yap1 in subconfluent αE-cateninfl/fl and αE-catenin−/− keratinocytes.

Fig. S13. Increased nuclear localization of Yap1 in confluent αE-catenin−/− keratinocytes.

Fig. S14. Rescue of cytoplasmic localization of Yap1 in αE-catenin−/− cells transduced with retroviruses expressing full-length αE-catenin.

Fig. S15. Partial colocalization between αE-catenin and Yap1 in confluent keratinocytes.

Fig. S16. Yap1 is nuclear in both dividing and nondividing subconfluent keratinocytes.

Fig. S17. Western blot analysis of Hippo pathway proteins in cultured keratinocytes.

Fig. S18. Quantitation of Yap1-specific phosphorylation on Ser127.

Fig. S19. Decreased abundance of Yap1 mRNA in αE-catenin−/− keratinocytes.

Fig. S20. Acute knockdown of αE-catenin reduces total Yap1 protein abundance.

Fig. S21. Nuclear localization of Yap1 in αE-catenin−/− cortical neural progenitor cells.

Fig. S22. Hippo pathway siRNA screen on cultured αE-cateninfl/fl (Ctrl) and αE-catenin−/− (α-cat−/−) keratinocytes.

Fig. S23. Wild-type, but not S94A mutant, human Yap2 protein is necessary for increased cell proliferation in αE-catenin−/− keratinocytes.

Fig. S24. qRT-PCR analysis of Cyr61 expression in 4-month-old skin and tumors from αE-cateninfl/fl (Ctrl) and GFAP-Cre/αE-cateninfl/fl-cat cKO) mice.

References

References and Notes

  1. Acknowledgments: We thank E. Fuchs for encouragement and helpful advice; all members of our laboratory for suggestions and comments; H. Denny Liggitt for help with skin pathology analysis; E. Herman, L. Nguyen, N. Ramirez, and E. Stepniak for help with the maintenance of mutant mice; and J. Segre, S. J. Kaufman, and the Developmental Studies Hybridoma Bank for gifts of antibodies. Funding: This work was supported by NCI grants R01 CA098161 and R01 CA131047 to V.V. M.R.S. was partially supported by the Chromosome Metabolism and Cancer Training grant NIH T32 CA09657. Author contributions: M.R.S., B.T.K., W.-H.L., O.K., and G.M.R. designed and performed the experiments and analyzed the data. F.D.C. contributed to interpretation. D.M.L. and J.T.S. performed histological analysis. V.V. designed the experiments, analyzed the data, and wrote the manuscript. All authors contributed to manuscript revision. Competing interests: The authors declare that they have no competing interests.
View Abstract

Navigate This Article