Inappropriate α6β4 integrin expression correlates with a high risk of tumour progression in stratified squamous epithelia. Targeted expression of α6β4 in the suprabasal layers of transgenic mouse epidermis dramatically increased the frequency of papillomas, carcinomas and metastases induced by chemical carcinogenesis, independent of the β4 cytoplasmic domain. Suprabasal α6β4 also perturbed transforming growth factor β (TGFβ) signalling as demonstrated by decreased nuclear Smad2 in transgenic epidermis and tumours. In cultured keratinocytes, suprabasal α6β4 relieved TGFβ-mediated growth inhibition and blocked nuclear translocation of activated Smad2/3. Responsiveness to TGFβ could be restored by inhibiting cadherin-mediated cell-cell adhesion or phosphoinositide 3-kinase (PI3-K) activity, but not by inhibiting mitogen-activated protein kinase (MAPK) activity. These data suggest that suprabasal α6β4 promotes tumourigenesis by preventing TGFβ from suppressing clonal expansion of initiated cells in the epidermal basal layer.
The epidermis is a constantly renewing tissue that is maintained by a highly coordinated balance between keratinocyte proliferation and terminal differentiation. Extracellular matrix receptors of the integrin family play an important role in regulating normal epidermal homeostasis (Watt, 2002). The major keratinocyte integrins are α2β1 (collagen receptor), α3β1 (laminin receptor) and α6β4 (laminin receptor) (van der Flier and Sonnenberg, 2001; Watt, 2002). The β1 integrins have a pericellular distribution in the basal layer of the epidermis and are associated with the actin cytoskeleton. The α6β4 integrin is concentrated at the basement membrane zone and is associated with hemidesmosomes and the intermediate filament system.
Whereas integrin expression is normally confined to the basal layer of the epidermis, expression is frequently perturbed in keratinocyte tumours. The alteration that is most heavily implicated in epithelial carcinogenesis is upregulated expression of the α6β4 integrin (Rabinovitz and Mercurio, 1996). Suprabasal expression of α6β4 in keratinocytes that are not adjacent to the tumour stroma correlates with poor prognosis in human squamous cell carcinomas (SCCs) (Rabinovitz and Mercurio, 1996; van Waes et al., 1995). In mouse skin, suprabasal expression of α6β4 is observed in those benign papillomas with a high risk of conversion to SCCs (Tennenbaum et al., 1993).
Mechanistic studies of α6β4 have emphasized its role in promoting invasion by stimulating epithelial cell motility (Mercurio et al., 2001). In invasive carcinomas, α6β4 can be found associated with the actin cytoskeleton in the absence of hemidesmosomes and it is becoming clear that α6β4 is capable of ligand-independent signal transduction (Mercurio et al., 2001). Mobilization of α6β4 from hemidesmosomes occurs in response to chemotactic factors and is correlated with phosphorylation of the β4 cytoplasmic domain (Mainiero et al., 1996). α6β4 cooperates with growth factor receptors to activate phosphoinositide 3-kinase (PI3-K), and PI3-K activation is necessary for α6β4-mediated invasiveness (Mercurio et al., 2001). Human keratinocytes lacking α6β4 are resistant to the tumourigenic effects of Ras and NF-κB (Dajee et al., 2003).
Whereas there is strong evidence that α6β4 promotes invasion of carcinoma cells, little is known about how integrin overexpression influences the early stages of tumourigenesis. It is also unclear how inappropriate expression of α6β4 in the differentiated compartment of a tumour could influence the growth and metastatic potential of undifferentiated cells in the basal layer. To address these questions, we have generated transgenic mice in which α6β4 is expressed in the suprabasal layers of the epidermis under the control of the involucrin promoter (Carroll et al., 1995). We demonstrate that suprabasal α6β4 has a profound and positive influence on the susceptibility of keratinocytes to forming tumours, relieving the growth-inhibitory effects of TGFβ via a mechanism that requires E-cadherin-mediated cell-cell adhesion and PI3-K activity.
Materials and Methods
Generation of transgenic mice
Human full-length α6 and β4 (Giancotti et al., 1992) and truncated β4 (β4Δ 1-2329) (Mainiero et al., 1997) integrin subunit cDNAs, kindly provided by F. Giancotti (Sloan Kettering Cancer Center, New York), were subcloned into a previously reported expression cassette containing the involucrin promoter (Carroll et al., 1995). All constructs were microinjected into fertilized mouse oocytes and at least three transgene-positive founders per construct were identified as previously described (Carroll et al., 1995).
The Invα6 transgenics, described previously (Romero et al., 1999), and the Invβ4 mice were derived in an F1 hybrid (C57Bl/6×CBA) mouse strain, then backcrossed for at least seven generations onto a homogeneous FVB/N background. The Invβ4Δ mice were generated directly in the FVB/N strain. Invα6β4 and Invα6β4Δ double-transgenic and wild-type (wt) experimental mice were generated by crossing heterozygous α6 single transgenics to heterozygous β4 or β4Δ single transgenics. Animal husbandry was as described previously (Owens and Watt, 2001).
Seven-week-old female Invα6β4, Invα6β4Δ and wt littermate mice (25 animals/group) were shaved once on the dorsal surface with electric clippers. After one week, all animals that did not show signs of hair regrowth received one topical application of 100 nmol (25 μg) 7,12-Dimethylbenz[a]anthracene (DMBA; Acros Organics) in 200 μl acetone or 200 μl acetone alone. One week later, mice received twice weekly applications of 6 nmol (3.7 μg) 12-O-tetradecanoylphorbol-13-acetate (TPA; LC Laboratories) in 200 μl acetone or 200 μl acetone alone for 15 weeks.
Benign and malignant skin tumours were recorded once weekly for up to 52 weeks after the start of promotion. The statistical significance of differences in papilloma and SCC formation between transgenic and wt mice was determined with a Student's t-test. To confirm weekly papilloma and SCC counts, tumour sections were graded as described previously (Owens and Watt, 2001). The experiment comparing Invα6β4 and wt mice was performed twice, as was the experiment comparing Invα6β4, Invα6β4Δ and wt mice.
Endogenous and transgenic integrin expression was examined in frozen skin, tumour and lymph node sections, essentially as described previously (Owens and Watt, 2001), using antibodies to human α6 (MP4F10), human β4 (3E1; Life Technologies) or an antibody that detects mouse α6β4 (CD49f; Serotec). 5-bromo-2′-deoxyuridine (BrdU) incorporation was examined in formalin-fixed sections (Owens and Watt, 2001) from mice that received an i.p. injection of 100 mg/kg BrdU 1 hour prior to sacrifice. For keratin 14 and Smad staining, formalin-fixed sections were microwaved in Citra Plus antigen retrieval solution (Biogenex) for 8 minutes (keratin14) or 21 minutes (Smad). Slides were cooled, blocked in 10% normal goat serum and probed overnight at 4°C with rabbit anti-keratin 14 (Babco), anti-Smad2/3 (Transduction Laboratories) or anti-phosphoSmad2 (kind gift of C. Heldin, Ludwig Institute, Uppsala, Sweden).
TGFβ responsiveness of cultured keratinocytes
Spontaneously immortalized mouse keratinocyte lines were derived from Invα6β4 and wt adult mouse epidermis and maintained with a feeder layer as described previously (Romero et al., 1999). In some experiments, cells were transduced with retroviral vectors encoding a dominant-negative E-cadherin (H-2Kd-E-cad) or a control construct in which the β-catenin-binding site in the cadherin cytoplasmic domain was deleted (H-2Kd-E-cadΔC25) (Zhu and Watt, 1996).
To measure TGFβ responsiveness, transgenic and wt mouse keratinocytes were plated onto glass coverslips and grown to form confluent monolayers or for up to 5 days post-confluence. Cells were treated with 2 ng/ml TGFβ1 (Pepro Tech EC) for 1 hour to induce Smad2/3 nuclear translocation. Cells were labelled with the anti-Smad2/3 antibodies (Transduction Laboratories) (Pierreux et al., 2000) or anti-human α6 integrin antibodies (Romero et al., 1999) described above.
Stratified cultures were also reconstituted as follows. Transgenic and wt mouse keratinocytes were induced to differentiate in suspension (Romero et al., 1999) and then seeded onto confluent monolayers of transgenic or wt mouse keratinocytes and allowed to attach for 18 hours. Cultures were treated with 2 ng/ml TGFβ1 for 1 hour and stained for Smad2/3 (Pierreux et al., 2000) or involucrin (Owens and Watt, 2001).
For inhibitor studies, cells were pre-incubated with 10 μM U0126 MEK inhibitor (Promega), 50 nM LY 294002 PI3-K inhibitor (Sigma) or dimethylsulphoxide for one hour before treatment with TGFβ1. In some cases, medium conditioned for 6 or 24 hours by 5-day postconfluent stratified Invα6β4 or wt keratinocyte cultures was incubated with monolayer cultures of keratinocytes.
To measure BrdU incorporation, keratinocytes were serum starved for 24 hours, transferred to complete medium ±2 ng/ml TGFβ1 for 19 hours and then pulsed with 80 μg/ml BrdU for 1 hour. Cells were fixed in 3.7% formaldehyde and permeabilized in 2 M HCl/0.5% Triton X-100 followed by treatment with 0.1 M NH4Cl, then immunolabelled with a monoclonal BrdU antibody (Becton Dickinson).
Confluent monolayers and 5-day post-confluent stratified cultures were treated with 2 ng/ml TGFβ1 or PBS for 1 hour, washed with cold PBS and scraped from the dish in RIPA lysis buffer. After 10 minutes incubation on ice, the lysates were centrifuged at 13,000 rpm at 4°C. The supernatants were recovered and subjected to electrophoresis on 10% Tris-glycine pre-cast polyacrylamide gels (Zaxis). Proteins were transferred to Immobilon-P PVDF membranes (Millipore) and probed with antibodies against total Smad2/3 (Transduction Laboratories), phosphorylated Smad2 (courtesy of C. Heldin), TGFβRI (Santa Cruz Biotech.), Ser473 phosphorylated Akt (Cell Signalling), total Akt (Upstate Biotech.) or actin (Sigma). Bands were visualized by incubation with horseradish peroxidase (HRP)-linked secondary antibodies (Amersham) followed by chemiluminescence in Western Lightning (Perkin Elmer).
Generation of transgenic mice
Transgenic mice expressing the single α6 and β4 integrin subunits under the control of the involucrin promoter (Invα6 and Invβ4) were crossed to create mice expressing suprabasal α6β4 (Invα6β4). Endogenous α6β4 expression was confined to the basolateral surface of basal keratinocytes in wt and transgenic mouse epidermis (Fig. 1A and data not shown). Transgenic α6β4, detected with antibodies specific for the human integrin subunits, was expressed in the suprabasal epidermal layers (Fig. 1A). The α6 founder line 1374C+D and the β4 founder line 1376A were used for all subsequent experiments.
Suprabasal expression of α6β4 integrin increases epidermal sensitivity to chemical carcinogenesis
None of the mice expressing suprabasal α6β4 displayed any gross skin phenotype or developed any spontaneous tumours. In addition, the histological appearance of wt and transgenic skin was indistinguishable (acetone treatment; Fig. 1B,E). A single application of the phorbol ester tumour-promoter TPA resulted in a similar increase in the number of epidermal nucleated cell layers and degree of dermal inflammatory infiltrate in transgenic and wt skin (Fig. 1C,F). However, visualizing S-phase cells by BrdU incorporation revealed that the presence of α6β4 in the differentiated epidermal layers resulted in a greater increase in proliferation in the basal (transgene-negative) epidermal layers in response to TPA (Fig. 1D,G). This indicates that the presence of suprabasal α6β4 did not induce differentiated cells to proliferate, but enhanced the proliferative response of basal cells to TPA.
To determine whether the transgenics had altered sensitivity to tumour formation, Invα6β4 and wt mice were subjected to a classical two-stage carcinogenesis protocol in which DMBA induces Ha-Ras mutations and repeated TPA treatments cause tumour promotion (Owens and Watt, 2001). Invα6β4 mice developed 3-4 times as many benign papillomas as wt mice (Invα6β4: 13.1 papillomas/mouse; wt: 4.0 papillomas/mouse; P=0.001) (Fig. 2A). 100% of Invα6β4 mice developed papillomas, compared with 80% of wt animals. No skin tumours were observed in wt or Invα6β4 mice either initiated with acetone vehicle and promoted with TPA or initiated with DMBA and promoted with acetone vehicle (data not shown).
To determine whether the papillomas in Invα6β4 mice were phenotypically distinct from those of wt animals, their histopathology and growth fraction were compared. As illustrated in Fig. 2C, the two groups of papillomas tended to be well differentiated and could not be distinguished by the extent of dysplasia and cellular atypia. There were no differences in the number or location of BrdU-positive S-phase cells in wt and transgenic papillomas (data not shown).
Invα6β4 mice went on to develop 3-4 times more SCCs than wt mice (Invα6β4: 2.16 SCCs/mouse; wt: 0.65 SCCs/mouse; P=0.0001) (Fig. 2B). The proportion of Invα6β4 mice that developed SCCs was also greater than that of wt mice (Invα6β4: 100%; wt: 40%). Both wt and transgenic SCCs were well differentiated and no spindle cell carcinomas were observed (Fig. 2C and data not shown).
The total number of SCCs observed in Invα6β4 mice was probably an underestimate because these mice had a higher morbidity rate than wt animals (Fig. 2B). By 27 weeks, a time when the frequency of SCCs was still rising, 50% of the transgenics were dead, compared with only 4% of wt mice (data not shown). The reason for the increased morbidity was that Invα6β4 mice had a higher frequency of metastases. As shown in Table 1, 68% of Invα6β4 mice developed metastases compared with 8% of wt mice. In addition, Invα6β4 mice developed 6-7 times as many metastases per mouse, on average, than wt mice (Table 1).
Whereas the higher frequency of SCCs in Invα6β4 mice is likely to result from the higher frequency of papillomas (Fig. 2A,B), it does not explain the increased number of metastases. As shown in Table 1, the metastasis (secondary tumour) to SCC (primary tumour) ratio was 4 times higher in Invα6β4 mice, indicating that Invα6β4 SCCs were 4 times more likely to spread than those from wt animals. Only those mice that developed SCCs were found to contain secondary tumours.
Histological sections of Invα6β4 metastases were examined to confirm the presence of keratinocytes (Fig. 2D). All of the metastases stained positive for keratin 14 and the α6 transgene. Expression of the α6 transgene indicates that the metastatic keratinocytes were still capable of some degree of terminal differentiation.
The tumour response in Invα6β4 mice does not depend on the β4 cytoplasmic domain
Signalling via the cytoplasmic domain of the β4 integrin subunit is critical for the pro-invasive and migratory effects of α6β4 on tumour cells (Mercurio et al., 2001). To test whether it was also required for the tumourigenic effects of suprabasal α6β4, we generated Invα6β4Δ (tailless) double-transgenic mice. The β4Δ subunit is truncated immediately after Lys734, the boundary between the transmembrane and intracellular domains (Mainiero et al., 1997). Like the full-length β4 subunit, β4Δ was expressed on the surface of suprabasal keratinocytes (Fig. 1A). The β4Δ founder line 2417C was used for all subsequent experiments.
Remarkably, deletion of the β4 subunit cytoplasmic tail did not suppress, but rather enhanced, the proliferative effects of suprabasal α6β4 expression. Acetone-treated Invα6β4Δ skin was mildly hyperplastic and contained a higher dermal infiltrate than wt or Invα6β4 skin (Fig. 1H). The number of nucleated epidermal cell layers dramatically increased after a single application of TPA and was considerably greater than in Invα6β4 or wt skin (Fig. 1I). Greater than 90% of basal keratinocytes in TPA-treated Invα6β4Δ epidermis were BrdU-positive (Fig. 1J) compared with approximately 15% of basal keratinocytes in TPA-treated wt epidermis (Fig. 1D) and 40% in Invα6β4 epidermis (Fig. 1G). Suprabasal BrdU-labelled cells were rarely observed in Invα6β4Δ epidermis (Fig. 1J).
Invα6β4Δ mice developed even more papillomas than Invα6β4 mice (Fig. 2A), averaging 27.8 papillomas/mouse compared with 13.1 papillomas/mouse in Invα6β4 mice (P=0.009). Invα6β4Δ mice also developed significantly more SCCs than wt mice (Fig. 2B) (Invα6β4Δ: 2.33 SCCs/mouse; wt: 0.65 SCCs/mouse; P=0.0001). The numbers of SCCs (Fig. 2B) and metastases (data not shown) in Invα6β4 and Invα6β4Δ mice were not significantly different. Unlike Invα6β4 and wt mice, Invα6β4Δ animals developed tumours with DMBA treatment alone (4.33 papillomas/mouse; 0.67 SCCs/mouse). Therefore, the β4 cytoplasmic tail was not required for the tumourigenic effect of suprabasal α6β4, and removal of the cytoplasmic domain actually enhanced tumour formation.
Suprabasal α6β4 expression disrupts TGFβ signalling in vivo and in culture
Since suprabasal α6β4 expression correlated with increased proliferation of keratinocytes in the underlying basal layer (Fig. 1D,G,J), we postulated that it might alter growth factor signalling or responsiveness. TGFβ negatively regulates keratinocyte proliferation and the early stages of epidermal tumour promotion (Derynck et al., 2001; Wakefield and Roberts, 2002) and also mediates the effects of the αvβ6 and αvβ8 integrins on epithelial homeostasis (Mu et al., 2002; Morris et al., 2003). TGFβ signalling is mediated by receptor activation of the Smad proteins Smad2 and Smad3 (Moustakas et al., 2001). Phosphorylated Smad2/3 forms a complex with Smad4 and translocates from the cytoplasm to the nucleus to activate gene transcription (Moustakas et al., 2001). To determine whether TGFβ signalling was perturbed in Invα6β4Δ mice, we stained sections of skin with an antibody to phosphoSmad2 (Fig. 2E). In wt epidermis treated once with TPA or acetone vehicle, the majority of basal and suprabasal nuclei were positively stained. By contrast, very few nuclei stained positive in acetone- or TPA-treated transgenic epidermis.
Reduced Smad2/3 activity is known to be associated with progression of skin papillomas to SCCs (He et al., 2001). The number of phosphoSmad2-positive nuclei was indeed significantly lower in wt papillomas and SCCs than in wt epidermis (Fig. 2E). Nevertheless, the number of positive nuclei was further reduced in the transgenic tumours (Fig. 2E).
We next developed in vitro models to examine the mechanism of disruption of TGFβ signalling (Fig. 3A,B). Keratinocytes from wt and transgenic epidermis were grown either to confluent monolayers in which very few differentiated cells were present, or to 5-day post-confluent cultures that contained stratified, involucrin-positive suprabasal cells. In stratified Invα6β4 cultures, the involucrin-positive cells were also transgene-positive (Romero et al., 1999). The effects of the transgene on TGFβ responsiveness could then be monitored by translocation of receptor-activated Smad2/3 from the cytoplasm to the nucleus (Wakefield and Roberts, 2002) in monolayers and postconfluent, stratified cultures of transgenic and wt keratinocytes (Fig. 3A).
In untreated wt and transgenic monolayers (Fig. 3C,D) or stratified cultures (Fig. 3E,F), Smad2/3 was found almost exclusively in the cytoplasm. Upon treatment with TGFβ1, Smad2/3 translocated to the nucleus in wt and transgenic monolayers (Fig. 3G,H). Nuclear Smad2/3 was also observed in the basal layer of post-confluent stratified cultures of wt cells (Fig. 3I). However, it did not occur in basal cells of Invα6β4 post-confluent stratified cultures (Fig. 3J). The presence or absence of the transgene was confirmed by immunofluorescence staining with an antibody to the human α6 integrin subunit (Fig. 3K,L and data not shown). These experiments suggest that suprabasal expression of α6β4 suppresses the responsiveness of basal keratinocytes to TGFβ.
No differences in the total levels of Smad2/3 or TGFβRI were observed in wt and transgenic keratinocytes, whether grown as monolayers or stratified cultures and whether treated with TGFβ or untreated (Fig. 3M). Activation of Smad2, measured with antibodies specific for phosphoSmad2, increased on TGFβ treatment and was greater in monolayers than stratified cultures, possibly reflecting the greater proportion of basal cells (Fig. 3M). However, no major differences in the level of phosphorylated (activated) Smad2 were seen in transgenic versus wt cultures (Fig. 3M). We conclude that suprabasal α6β4 does not inhibit TGFβ-mediated Smad2 phosphorylation (Fig. 3M) but prevents translocation of phosphoSmad2 to the nucleus (Fig. 3J). This is consistent with in vivo immunolocalization data obtained with an antibody to total (data not shown), as opposed to phosphorylated (Fig. 2E), Smad2.
Disruption of TGFβ signalling by α6β4 requires cell-cell contact and PI3-K activity
To investigate the mechanism by which suprabasal keratinocytes expressing α6β4 inhibited the TGFβ responsiveness of basal keratinocytes, monolayers of wt or Invα6β4 keratinocytes were combined with suprabasal cells that had been induced to undergo terminal differentiation in suspension (Romero et al., 1999) (Fig. 3B). When wt or transgenic monolayers were combined with wt suprabasal cells, Smad2/3 underwent nuclear translocation in response to TGFβ (Fig. 4A,B,E,F). When suprabasal transgene-positive cells were attached to wt or transgenic monolayers, Smad2/3 translocation was inhibited (Fig. 4C,D,G,H). These results support the conclusion from post-confluent cultures (Fig. 3) that suprabasal α6β4 inhibits Smad2/3 translocation.
Wild-type keratinocytes incubated with conditioned medium from post-confluent stratified Invα6β4 keratinocytes remained completely responsive to TGFβ, suggesting that cell-cell contact was required for the effect (Fig. 4I,J). This was confirmed by examining incompletely stratified cultures of Invα6β4 keratinocytes treated with TGFβ (Fig. 4K,L). In basal cells underneath transgene- and involucrin-positive suprabasal cells, there was no Smad2/3 translocation; however, adjacent basal cells that were not in contact with suprabasal cells had nuclear Smad2/3.
Keratinocyte intercellular adhesion is mediated by adherens junctions and desmosomes. When adherens junction formation is inhibited by overexpression of a dominant-negative E-cadherin construct (dnEcad; consisting of the transmembrane and cytoplasmic domains of E-cadherin coupled to the extracellular domain of an irrelevant protein), intercellular adhesion is inhibited in keratinocyte monolayers; however post-confluent cells are able to stratify through intercellular adhesion mediated by desmosomal junctions (Zhu and Watt, 1996). Expression of dnEcad relieved suprabasal α6β4-mediated inhibition of Smad2/3 translocation in response to TGFβ, whereas a control construct in which the β-catenin-binding site of E-cadherin is deleted (dnEcadΔC25) did not (Fig. 5A). The proportion of Smad2/3-positive nuclei was significantly increased in TGFβ-treated Invα6β4 cells expressing dnEcad compared with untransduced Invα6β4 cells (Fig. 5A).
The ability of the α6β4 integrin to promote carcinoma invasion depends on activation of PI3-K (Shaw et al., 1997), and α6β4 can also signal to the Ras-MAPK (mitogen-activated protein kinase) pathway (Mainiero et al., 1997). In addition, Smad2/3 nuclear translocation can be inhibited by MAP kinase phosphorylation and may also involve PI3-K activity (Kretzschmar et al., 1999). We therefore tested the effects of PI3-K and ERK/MAPK kinase (MEK) inhibitors in our assays. As shown in Fig. 5B, the percentage of Smad2/3-positive nuclei was not increased in Invα6β4 cells pre-treated with the MEK inhibitor U0126 compared with DMSO controls, whereas pre-treatment of cells with the PI3-K inhibitor LY 294002 prior to addition of TGFβ increased the number of Smad2/3-positive nuclei by 4-5-fold. In view of the effect of the PI3-K inhibitor, we examined whether PI3-K activity was elevated in Invα6β4 keratinocytes, using phosphorylation of Akt as a read-out (Cantley, 2002) (Fig. 5C). The basal level of phosphoAkt was higher in Invα6β4 than wt cells and could be completely abolished by pre-incubation with LY 294002.
Collectively, these results show that the repression of TGFβ-induced Smad2/3 translocation by suprabasal α6β4 expression is not due to release of diffusible factors from suprabasal cells but is dependent on cadherin-mediated cell-cell adhesion and requires PI3-K activity.
Suprabasal α6β4 expression relieves the growth-inhibitory response of keratinocytes to TGFβ
TGFβ inhibits proliferation of primary keratinocytes and overexpression of TGFβ in transgenic mouse epidermis reduces papilloma formation during chemical carcinogenesis (Derynck et al., 2001; Shipley et al., 1986). TPA-induced proliferation of basal keratinocytes is enhanced in Invα6β4 epidermis (Fig. 1D,G,J). These observations suggested that suprabasal α6β4 might overcome TGFβ-mediated growth inhibition. We tested this by measuring BrdU incorporation in cultured keratinocytes in the presence or absence of suprabasal α6β4 and TGFβ (Fig. 5D). Monolayers of wt or Invα6β4 keratinocytes responded to TGFβ1 with similar reductions (40-50%) in BrdU uptake (Fig. 5D, left panel). When wt suprabasal keratinocytes were combined with wt basal keratinocytes, the same suppression of BrdU uptake was observed (Fig. 5D, right panel). However, when suprabasal Invα6β4 keratinocytes were attached to wt monolayers, the number of BrdU-positive cells was not decreased by TGFβ1 treatment. Furthermore, addition of suprabasal transgene-positive cells in the absence of TGFβ stimulated BrdU incorporation relative to the addition of wt cells (Fig. 5D, right panel).
Our results indicate that suprabasal expression of α6β4 can enhance proliferation of the underlying basal cells both in vivo and in culture, and can overcome TGFβ-mediated growth inhibition in culture. This might provide the explanation for the increased susceptibility of Invα6β4 epidermis to chemical carcinogenesis.
Carcinogenesis is a complex multi-step process that requires the accumulation of multiple genetic and epigenetic events (Hahn and Weinberg, 2002). There are many factors that can influence this process, such as dose and frequency of exposure to mutagens, target cell phenotype and changes in the tumour cell microenvironment. As long-term residents of the basal layer, epidermal stem cells have the capacity to sustain multiple mutations and can thereby found a tumour (Owens and Watt, 2003). However, suprabasal or differentiated cells have the potential to influence the course of the disease by exerting a positive or negative influence over the underlying basal cells that either encourages or restricts the proliferation of mutant clones (Owens and Watt, 2003).
Suprabasal expression of the α6β4 integrin did not result in spontaneous epidermal tumours, but greatly increased the susceptibility of the epidermis to chemical carcinogenesis. Nuclear phosphoSmad2 levels were markedly decreased in both control and TPA-treated transgenic epidermis, suggesting that the elevated proliferation of keratinocytes in the basal layer relative to wt epidermis could be due to lack of TGFβ-mediated growth inhibition. It is striking that suprabasal expression of α6β4 led exclusively to increased proliferation of basal keratinocytes, whereas suprabasal expression of α2β1 or α3β1 leads to suprabasal BrdU incorporation in TPA-treated epidermis (Owens and Watt, 2001). Suprabasal α6β4 relieved TGFβ-mediated growth inhibition in cultured keratinocytes and the effect was at the level of nuclear translocation of phosphorylated Smads. We have thus demonstrated a link between α6β4 overexpression and loss of TGFβ signalling in the progression of chemically induced tumours.
Since α6β4 signalling in carcinoma invasion is dependent on the β4 cytoplasmic domain (Mercurio et al., 2001), it is surprising that the phenotype of Invα6β4Δ mice was more severe than that of Invα6β4 mice. Mice with the truncated β4 subunit had spontaneous epidermal hyperproliferation, an increased incidence of papillomas on treatment with DMBA and TPA, and even developed papillomas when treated with DMBA alone. There is good evidence for signalling via integrin α subunit cytoplasmic domains and for modulation of integrin signalling through interactions between α and β cytoplasmic domains (Hughes and Pfaff, 1998; Hynes, 2002). One possibility is therefore that the ability of α6β4 to inhibit TGFβ responsiveness depends on the α6 cytoplasmic domain and that, whereas the β4 cytoplasmic domain is permissive, its loss leads to an enhancement of the α6 signal (Han et al., 2001; Zhang et al., 2001).
The major adhesive contacts between basal and suprabasal keratinocytes are adherens junctions and desmosomes, with E-cadherin being the main adhesive receptor in the adherens junctions. Cell-cell attachment was required for suprabasal α6β4-mediated disruption of TGFβ signalling. TGFβ responsiveness was restored to Invα6β4 cells by overexpressing the E-cadherin cytoplasmic domain, which inhibits adherens junction formation but not desmosome formation (Zhu and Watt, 1996). E-cadherin is linked to the actin cytoskeleton via β-catenin, and deletion of the β-catenin-binding site prevented the E-cadherin cytoplasmic domain from inhibiting cell-cell adhesion and interfering with suprabasal α6β4 (Fig. 5A). We conclude that cadherin-mediated intercellular adhesion is required for the suppression of TGFβ signalling by α6β4.
Two observations suggest that the effect of suprabasal α6β4 is mediated by PI3-kinase. PI3-K inhibitors relieved the block in TGFβ-mediated Smad translocation, and phosphoAkt levels were higher in transgenic than wt keratinocytes, indicative of increased PI3-K activity (Fig. 5B,C). PI3-K activation is required for the pro-invasive and survival effects of α6β4 (Mercurio et al., 2001; Shaw et al., 1997); however, this depends on a tyrosine residue, Y1494, which is deleted in Invα6β4 mice (Shaw, 2001). Instead, the increase in PI3-K activity in transgenic keratinocytes could potentially involve cross-talk between α6β4 and E-cadherin (Hodivala and Watt, 1994; Arregui et al., 2000), since homotypic E-cadherin interactions can activate PI3-K (Kovacs et al., 2002).
The inhibition of TGFβ signalling was at the level of translocation of phosphorylated Smad2/3 into the nucleus (Fig. 3). In contrast to an earlier report (Dong et al., 2000), we found no evidence that sequestration of Smads in the cytoplasm required intact microtubules (data not shown). Instead, we favour a model in which reorganization of the actin cytoskeleton by TGFβ plays a role in Smad translocation. TGFβ can rapidly rearrange actin filaments by a Smad-independent process that involves the Rho GTPases Cdc42 and RhoA, and it has previously been proposed that TGFβ activation of small GTPases is required for Smad translocation (Edlund et al., 2002). In keratinocytes, suprabasal α6β4 was linked to actin filaments (data not shown) and actin-associated α6β4 is known to activate RhoA (O'Connor et al., 2000). E-cadherin-mediated cell-cell adhesion activates Rac1, which is upstream of PI3-K (Nakagawa et al., 2001). We propose that, in the presence of suprabasal α6β4 and E-cadherin, TGFβ is unable to effect the actin reorganization required for Smad translocation. dnEcad may restore the subcellular location of components of the PI3-K signalling cascade (Khayat et al., 2000) by reduced activation of Rho GTPases.
Whereas previous work has emphasized the positive effect of α6β4 on the behaviour of carcinoma cells (Rabinovitz and Mercurio, 1996; Mercurio et al., 2001), our experiments establish that changes in expression of this integrin in otherwise normal epithelium can have a major impact on the initiation and course of the disease. We have uncovered a novel mechanism by which aberrant integrin expression enhances the tumour microenvironment by altering growth regulation of neighbouring cells.
We thank R. Rudling, B. Cross and J. Groeninger for their assistance in conducting the mouse tumour studies and C. Hill for her advice. We are grateful to C. Heldin and F. Giancotti for reagents. D.M.O. was supported by a National Research Service Award from the NCI (CA75638). F.M.W. was supported by Cancer Research UK and a European Union Quality of Life Network grant.
↵* Present address: MRC Mammalian Genetics Unit, Harwell OX11 ORD, UK
- Accepted June 24, 2003.
- © The Company of Biologists Limited 2003