α3β1 integrin promotes keratinocyte cell survival through activation of a MEK/ERK signaling pathway

In normal epithelial cells, inadequate or inappropriate cell adhesion to the extracellular matrix (ECM) leads to a specialized form of apoptosis known as anoikis (Frisch and Francis, 1994; Frisch and Screaton, 2001). Integrins are the major receptors for cell adhesion to the ECM (Hynes, 2002), and they are the principle receptors for transmitting outside-in signals from the ECM that inhibit anoikis and promote cell survival (Giancotti and Ruoslahti, 1999; Stupack and Cheresh, 2002). Although a number of different integrins can promote cell survival, specific integrin-ligand interactions may be required to inhibit anoikis in a distinct cell type or under special circumstances (Wary et al., 1996; Matter and Ruoslahti, 2001). Survival signals that are transduced by a particular integrin may be regulated during tissue remodeling events through changes in the bioavailability of appropriate ECM ligands.

Adhesion-dependent survival of epithelial cells can be regulated through the extracellular signal-regulated kinase (ERK) family of mitogen-activated protein kinases (MAP kinases) (Frisch and Screaton, 2001). Upon activation of ERK by integrin-mediated signals, ERK translocates to the nucleus where it activates the transcription of genes that promote cell survival (Schulze et al., 2001). Activation of focal adhesion kinase (FAK) is an early step in many integrin-mediated survival pathways (Frisch et al., 1996; Giancotti and Ruoslahti, 1999; Hanks et al., 2003). Activated FAK can promote cell survival through interactions with several downstream signal transduction molecules, including p130 Crk-associated substrate (CAS) and phosphatidylinositol 3-kinase (PI3K) (Frisch and Screaton, 2001; Hanks et al., 2003). Some of these FAK-mediated survival pathways may involve ERK activation. For example, CAS/Crk coupling and ERK activation can suppress apoptosis in some cells (Cho and Klemke, 2000). In addition, FAK interactions with PI3K may stimulate the Raf/MEK/ERK signaling cascade through activation of p21-activated kinase (PAK) (King et al., 1998; Eblen et al., 2002). FAK has also been linked to Ras-mediated Raf/MEK/ERK signaling through direct binding interactions with the Grb2 adaptor protein (Schlaepfer et al., 1994).

Despite an increased understanding of intracellular signaling pathways that promote integrin-mediated survival in many cell types, mechanisms whereby specific integrins regulate keratinocyte survival in the epidermis, and the role of MEK/ERK signaling in this process, remain unclear. Keratinocytes in the basal layer of the stratified epidermis are adhered to the basement membrane (BM) that separates the epidermis from the dermis (Burgeson and Christiano, 1997). Laminin-5 (LN-5) is the major adhesive ligand in the cutaneous BM, and mutations in the genes that encode each of the three subunits that comprise the LN-5 trimer (α3, β3, or γ2) lead to epidermal blistering in mutant mice and in patients with junctional epidermolysis bullosa (Burgeson and Christiano, 1997; Ryan et al., 1999; Kuster et al., 1997). Keratinocytes can bind to LN-5 through two integrin receptors, α3β1 and α6β4 (Nguyen et al., 2000). Targeted null mutations in the genes that encode the subunits for these integrins also lead to epidermal blistering, although the mechanisms of blistering are distinct (van der Neut et al., 1996; Dowling et al., 1996; Georges-Labouesse et al., 1996; DiPersio et al., 1997). Epidermal adhesion to BM laminins is critical for keratinocyte survival in vivo, because mice lacking integrins α3β1 and α6β4 show increased apoptosis in regions of detached epidermis (DiPersio et al., 2000b). A specific role for LN-5 in maintaining keratinocyte survival is supported by studies in cultured keratinocytes and mice with a targeted null mutation in the LAMA3 gene (Ryan et al., 1999; Nguyen et al., 2000; Fujisaki and Hattori, 2002).

During wound healing, the cutaneous BM is broken down and keratinocytes are stimulated to migrate over a provisional ECM that is rich in fibronectin and dermal collagen (Grinnell, 1992). Keratinocytes also secrete abundant LN-5 as they migrate into the wound, which is thought to promote α3β1-mediated cell migration and provide the foundation for new BM during re-epithelialization (Goldfinger et al., 1999; Nguyen et al., 2000). Considering the pro-survival effects of LN-5 discussed above, it is possible that adhesion to newly deposited LN-5 also contributes to keratinocyte survival during migration into the provisional wound ECM. LN-5, α3β1 and α6β4 are also expressed at high levels in many invasive carcinomas, suggesting possible roles for these adhesion proteins in promoting tumor cell invasion and survival, as well (Dajee et al., 2003; Felding-Habermann, 2003; Bartolazzi et al., 1994; Natali et al., 1993; Patriarca et al., 1998; Lohi et al., 2000; Pyke et al., 1995). Indeed, previous studies have suggested that α6β4 promotes survival of normal or transformed epithelial cells (Dowling et al., 1996; Weaver et al., 2002; Bachelder et al., 1999), possibly through activation of PI3K (Shaw et al., 1997). In contrast, the role of integrin α3β1 in regulating epithelial cell survival remains unclear. Aside from overlapping ligand-binding specificities, α3β1 and α6β4 appear to have distinct and separable functions in epidermal keratinocytes (Carter et al., 1990; DiPersio et al., 2000b; Nguyen et al., 2000), and the relative contributions of these two integrins to epithelial cell survival are likely to differ both in resting epithelia and during tissue remodeling.

To determine directly whether α3β1-mediated adhesion regulates keratinocyte survival, we cultured keratinocyte cell lines derived from wild-type or α3-null mice on LN-5 ECM, and then compared them for susceptibility to apoptosis induced by serum withdrawal. We demonstrate that the presence of α3β1 inhibits proteolytic activation of caspase-3 and suppresses apoptosis in serum-starved keratinocytes. We also show that the pro-survival effects of α3β1 occur through activation of FAK and at least partly through MEK-dependent activation of ERK. Our findings therefore reveal a novel role for α3β1 in suppressing caspase-3 activation and apoptosis through a MEK/ERK signaling pathway. Although α6β4 was necessary and sufficient for cell attachment to LN-5 in the absence of α3β1, α6β4-mediated adhesion was not sufficient to suppress caspase-3 activation fully. Furthermore, blocking α6β4-mediated adhesion in cells that express α3β1 did not induce caspase-3 activation, indicating that α3β1 is more effective than α6β4 in promoting keratinocyte survival. Our findings distinguish the survival promoting functions of these two LN-5-binding integrins and illustrate the importance of specific integrin-ligand interactions for adhesion-dependent survival of epithelial cells.

The MK^+/+ cell line (MK-1.16) and MK^-/- cell line (MK-5.4.6) were derived from keratinocytes isolated from wild-type or α3 integrin knockout mice, respectively (DiPersio et al., 2000a). MK^-/- cells were stably transfected with a full-length human α3 cDNA (a gift from Martin Hemler, Dana-Farber Cancer Institute, Boston, MA); α3-transfectants express high levels of α3β1 on the cell surface, as described previously (DiPersio et al., 2000a). MK growth medium consisted of Eagle's Minimum Essential Medium (EMEM; BioWhittaker, Walkersville, MD) supplemented with 4% FBS (BioWhittaker) from which Ca²⁺ had been chelated, 0.05 mM CaCl₂, 0.4 μg/ml hydrocortisone (Sigma, St Louis, MO), 5 μg/ml insulin (Sigma), 10 ng/ml EGF (Invitrogen Corporation, Carlsbad, CA), 2×10^-9 M T3 (Sigma), 10 units/ml interferon γ (INFγ; Sigma), 100 units/ml penicillin and 100 μg/ml streptomycin (Invitrogen), and LGlutamine (Invitrogen). Stocks of MK cell lines were maintained at 33°C, 8% CO₂, on tissue culture plates coated with 30 μg/ml denatured rat tail collagen (Cohesion, Palo Alto, CA). For experiments, MK cells were sub-cultured on LN-5 ECM prepared from the human squamous cell carcinoma line SCC-25 (Rheinwald and Beckett, 1981), as described previously (DiPersio et al., 2000a).

MK cells were cultured in serum-free EMEM, 0.05 mM CaCl₂, on LN-5 ECM for two days, then fixed in 4% paraformaldehyde. Apoptotic cells were detected by TUNEL (terminal deoxynucleotidyl transferase-mediated dUTP nick-end labeling) using the Apoptosis Detection System, Fluorescein (Promega, Madison, WI) followed by direct visualization on an Olympus BX60 fluorescence microscope. To quantify the number of apoptotic cells, flow cytometry of TUNEL-positive cells was performed according to the manufacturer's protocol. For each condition, 10,000 cells were analyzed by flow and the percentage of TUNEL-positive cells was determined. As a positive control for detection of DNA fragmentation, cells were treated with DNase I before TUNEL. As a negative control, terminal transferase was omitted from the TUNEL reaction.

MK cells were trypsinized from stock plates and resuspended in serum-free EMEM, 0.05 mM CaCl₂. For most experiments, cells were pre-incubated in suspension culture for 90 minutes at 33°C, then plated on LN-5 ECM at a sub-confluent density of approximately 6.25×10⁵ cells/35 mm well, or equivalent. One hour after plating, unattached cells were removed by gentle rinsing. To induce apoptosis, cells were cultured in serum-free EMEM, 0.05 mM CaCl₂ for an additional 24 or 48 hours, as indicated in the figure legends. For experiments in which MEK was inhibited, cells were treated with the pharmacological inhibitor U0126 (Calbiochem, San Diego, CA) at a final concentration of 10 μM, or with an equivalent volume of DMSO as a control. U0126 was added to cells during the pre-incubation period before attachment to LN-5 ECM; for 48 hour time points, U0126 was replenished in the medium after the first 24 hours of culture. For experiments in which integrin α6β4 was blocked, cells were pre-treated 15 minutes before plating with the rat anti-α6 monoclonal antibody GoH3 (BD Pharmingen, San Diego, CA) or with rat IgG_2a isotype control antibody (BD Pharmingen) at a concentration of 5 μg/ml, then cultured in the presence of 5 μg/ml GoH3 or control antibody for 24 hours. For experiments in which cells were kept in suspension, cells were cultured in serum-free medium for 24 hours in wells coated with 1% agarose to prevent adhesion. Apoptotic cells that had detached during the 24 hour culture period were collected from the medium by centrifugation, and cell lysates were combined for detached and attached cells. For α6β4-blocking experiments and corresponding controls, only attached cells were lysed to focus on cells that were adhered through α3β1 or α6β4. Cell lysates were prepared in Cell Lysis Buffer (1% Triton X-100, 20 mM Tris pH 7.5, 150 mM NaCl, 1 mM EGTA, 1 mM EDTA, 2.5 mM sodium pyrophosphate, 1 mM β-glycerolphosphate, 1 mM Na₃VO₄, 1 μg/ml leupeptin, and 2 mM PMSF), then sonicated for 15 seconds and clarified by centrifugation at 20,800 g for 10 minutes. Protein concentrations were determined using the Pierce BCA Protein Assay Kit (Pierce, Rockford, IL). Caspase-3 activation was monitored by western blot, as described below.

HA-tagged Ras V12 and β-galactosidase were cloned into pAdTrack, as described (Meadows et al., 2001). Replication-defective adenovirus encoding a GFP-FRNK fusion protein was a generous gift from Dr Allen Samarel (Heidkamp et al., 2002). The day before infection, 1.7×10⁶ MK^-/- cells or 1.0×10⁶ MK^+/+ cells were seeded onto collagen-coated 10 cm dishes in serum-containing growth medium. MK cells were infected for 24 hours with adenovirus expressing either HA-tagged Ras V12 or β-galactosidase as a control (multiplicity of infection=70), or with adenovirus expressing GFP-FRNK or GFP as a control (multiplicity of infection=350). Infected cells were trypsinized and sub-cultured on LN-5 ECM at a density of approximately 1×10⁶ cells per 35 mm well, and cultured in serum-free EMEM to induce apoptosis, as described above. GFP fluorescence was visualized on an Olympus IX70 inverted microscope.

Adhesion-dependent signaling was assayed essentially as described previously (Aplin and Juliano, 1999). For most experiments, MK cell cultures were serum-starved for 4-6 hours in serum-free medium (EMEM, 0.05 mM Ca²⁺, 0.5% heat-inactivated BSA). Cells were then removed from plates with trypsin, treated with 1 mg/ml trypsin inhibitor and pelleted. Cells were washed once, resuspended in serum-free medium, and incubated in suspension at 33°C, 8% CO₂, for 30 minutes. Cells were then either kept in suspension as a control, or plated at sub-confluent densities onto LN-5 ECM and allowed to attach for times indicated in the figures. After incubation, MK cell lysates were prepared in Cell Lysis Buffer and quantitated as described above. Phosphorylation of ERK1/2 and FAK were assayed by western blot, as described below. ERK1/2 kinase activity was assayed using an in vitro kinase assay to detect ERK-mediated phosphorylation of a GST-Elk-1 recombinant fusion protein (p44/42 MAP Kinase Assay Kit; Cell Signaling Technology, Beverly, MA). To assay phosphorylation of p130^CAS, cell monolayers were lysed in modified RIPA buffer (1% NP-40, 0.25% deoxycholate, 50 mM Tris pH 7.4, 150 mM NaCl, 1 mM EDTA, 1 mM Na₃VO₄, 20 μg/ml aprotinin, 12.5 μg/ml leupeptin and 2 mM PMSF) and quantified. Aliquots of cell lysate (125 μg) were pre-cleared with anti-mouse IgG-conjugated agarose beads (Sigma), then immunoprecipitated with 3 μl of mouse monoclonal anti-p130^CAS antibody (Transduction Laboratories, Lexington, KY) followed by incubation with 30 μl of anti-mouse IgG-agarose beads overnight at 4°C. Samples were washed with RIPA buffer, resolved by reducing 10% SDS/PAGE, and transferred to PVDF membranes (Bio-Rad, Hercules, CA) for western blotting, as described below.

Equal amounts of MK cell lysates (10 μg to 20 μg) were resolved by reducing 10% SDS/PAGE and transferred to nitrocellulose membranes. For western blot, primary antibodies were used at the following concentrations: rabbit polyclonal anti-caspase-3 (Cell Signaling Technology), 1:1000; rabbit polyclonal anti-keratin 14 (Covance Inc., Richmond, CA), 1:10,000; rabbit polyclonal anti-FAK (phospho-Tyr397) (BioSource International, Camarillo, CA), 1:1000; rabbit polyclonal anti-FAK (Upstate Biotechnology, Lake Placid, NY), 1:1000; rabbit polyclonal anti-phospho-ERK1/2 (Cell Signaling Technology), 1:1000; rabbit polyclonal anti-ERK1/2 (Promega), 1:5000; mouse monoclonal anti-p130^CAS (Transduction Laboratories), 1:1000; rabbit polyclonal anti-GFP (Santa Cruz Biotechnology, Santa Cruz, CA), 1:1000; mouse monoclonal antiphospho-tyrosine 4G10 (Upstate Biotechnology), 1:1000; mouse monoclonal anti-HA-tag (Covance), 1:1000. Peroxidase (HRP)-conjugated secondary antibodies were used at the following concentrations: goat anti-rabbit IgG (Cell Signaling Technology), 1:2000; goat anti-rabbit IgG (Pierce), 1:15,000; goat anti-mouse IgG (Pierce), 1:15,000. Chemiluminescence was performed using the SuperSignal Kit (Pierce).

To determine whether α3β1 integrin plays a role in keratinocyte survival, we exploited keratinocyte cell lines derived from mice that were either wild-type (MK^+/+ cells) or homozygous for a null mutation in the gene for the α3 integrin subunit (MK^-/- cells) (DiPersio et al., 2000a). MK cells were cultured on LN-5 ECM for two days, either in the presence of serum and growth factor/hormonal supplements to promote survival, or in the absence of serum and supplements to induce apoptosis. α3-null keratinocytes adhere to LN-5 ECM through α6β4 integrin (DiPersio et al., 1997) (also shown in Fig. 3). Fluorescent staining for TUNEL-positive cells revealed very little apoptosis in cultures of MK^+/+ cells attached to LN-5 ECM under serum-free conditions (Fig. 1A,B). In contrast, TUNEL-positive cells were readily detectable in cultures of α3β1-deficient MK^-/- cells attached to LN-5 ECM under serum-free conditions (Fig. 1C,D). Restoration of α3β1 integrin expression in MK^-/- cells through stable transfection with a cDNA encoding the α3 integrin subunit reduced the number of apoptotic cells to levels seen in MK^+/+ cells (Fig. 1E,F). Calculation of the proportion of fluorescent cells from at least 400 cells per culture revealed that 7.5% of MK^-/- cells were TUNEL-positive, compared with 0.5% of MK^+/+ cells and 1.0% of α3-transfected MK^-/- cells.

To confirm our quantification of apoptotic cells, we performed flow cytometric analysis of TUNEL-positive cells in cultures of α3-null MK^-/- cells and α3-transfected MK^-/- cells. α3-transfected MK^-/- cells cultured in the presence of serum did not contain a detectable population of TUNEL-positive cells (Fig. 1G, MKα3, +serum). About 1% of untransfected MK^-/- cells were TUNEL-positive under these conditions (Fig. 1G, MK^-/-, +serum), suggesting that absence of α3β1 results in only a slight increase in apoptosis in the presence of serum. Serum deprivation caused only a small induction of apoptosis in α3-transfected MK^-/- cells (Fig. 1G, MKα3, -serum). In contrast, serum deprivation induced apoptosis in α3-null MK^-/- cells to levels that were much higher than those seen in α3-transfected MK^-/- cells (Fig. 1G, compare black bars). Almost 7% of α3-null MK^-/- cells were apoptotic after two days of adhesion to LN-5 ECM under serum-free conditions, compared with only 0.3% of α3β1-expressing MK cells (Fig. 1G), demonstrating that absence of α3β1 caused a greater than 20-fold increase in the number of apoptotic cells. Importantly, this approach provides a minimum estimate of the number of apoptotic cells, because DNA fragmentation is a relatively late event in apoptosis, and a significant proportion of cells that have committed to apoptosis may not be detected by TUNEL-staining. The increase in apoptosis seen in MK^-/- cultures is significant, because even small changes in the proportion of apoptotic cells can have devastating effects over time on tissue development or homeostasis (Jacobson et al., 1997). These data demonstrate that absence of α3β1 significantly increases the sensitivity of MK cells to apoptosis under serum-free conditions.

Caspase-3 is an effector caspase involved in the execution of apoptotic pathways in a variety of cell types, and its activation is a convenient and widely used readout of apoptosis (Nicholson, 1999). Caspase-3 is activated through proteolytic conversion of the full-length, inactive zymogen (35 kDa) to activated 17 kDa and 12 kDa subunits (Nicholson et al., 1995). To determine whether keratinocyte apoptosis caused by absence of α3β1 is accompanied by increased activation of caspase-3, we monitored cleavage of caspase-3 by immunoblot of MK cell lysates with an antiserum that recognizes the 35 kDa pro-form, as well as the 17 kDa activated fragment and cleavage intermediates. The 35 kDa pro-form of caspase-3 was detected easily in both MK^+/+ and MK^-/- cells cultured on LN-5 ECM in the presence of serum, but cleaved caspase-3 was undetectable under these conditions (Fig. 2A, +serum). MK cell culture in serum-free medium induced cleavage of caspase-3 to the 17 kDa fragment and a cleavage intermediate of ∼25 kDa, indicative of increased apoptosis. A considerably higher level of caspase-3 cleavage occurred in MK^-/- cells than in MK^+/+ cells, as indicated by the higher ratio of cleaved to uncleaved caspase-3 in the latter cells (Fig. 2A, serum-free). Stable transfection of MK^-/- cells with α3 integrin suppressed caspase-3 activation under serum-free conditions (Fig. 2A, serum-free, α3). Quantification of bands in Fig. 2A corresponding to zymogen and cleaved forms revealed that 79.0% of the total caspase-3 was cleaved in MK^-/- cells, compared with 17.2% in MK^+/+ cells and 18.4% in α3-transfected MK^-/- cells. MK cells that were cultured in suspension for 24 hours showed high levels of cleaved caspase-3 whether or not they expressed α3β1 (Fig. 2B), indicating that the protective effects of α3β1 expression require cell adhesion to LN-5 ECM. Taken together, these data demonstrate that α3β1-mediated adhesion to LN-5 suppresses caspase-3 activation and subsequent apoptosis that is induced by serum withdrawal.

A considerable proportion of MK^-/- cells that were adhered to LN-5 ECM under serum-free conditions remained negative for TUNEL staining over the time course of our survival assays (Fig. 1). Furthermore, MK^-/- cells showed considerably higher levels of activated caspase-3 when cultured in suspension than they did when attached to LN-5 ECM (Fig. 2). As adhesion of α3-null mouse keratinocytes to LN-5 ECM is completely dependent on α6β4 integrin (DiPersio et al., 1997), these observations suggest that α6β4 can also contribute to MK cell survival. Indeed, previous studies have demonstrated a role for α6β4 in promoting survival of both normal and transformed epithelial cells (Bachelder et al., 1999; Weaver et al., 2002), and increased apoptosis was reported in the epidermis of β4-deficient mice (Dowling et al., 1996). To determine whether blocking α6β4 increases apoptosis in α3β1-expressing MK cells, cells were cultured on LN-5 ECM in serum-free medium for 24 hours, then treated with either anti-α6 function blocking antibody GoH3 or with an isotype control antibody. Adhesion of α3-null MK^-/- cells to LN-5 ECM was almost completely inhibited by GoH3 (Fig. 3B), but not by control antibody (Fig. 3A), confirming that adhesion to LN-5 ECM is mediated by α6β4 in these cells. We determined that cell surface levels of α6β4 are the same in MK^+/+ and MK^-/- cells using surface iodination (DiPersio et al., 2000a) and flow cytometry (D. Choma and C. M. DiPersio, unpublished). Therefore, we used the same GoH3 treatment to block α6β4 in α3β1-expressing MK cells. Blocking α6β4 had no effect on cell adhesion or spreading of either MK^+/+ cells (Fig. 3C,D) or α3-transfected MK^-/- cells (Fig. 3E,F), indicating that α3β1 was sufficient to mediate adhesion and spreading on LN-5 ECM. To compare levels of apoptosis, cells were lysed and assayed by western blot for cleavage of caspase-3. Blocking α6β4 with GoH3 did not induce caspase-3 activation above background levels seen in IgG control-treated cells for either MK^+/+ cells or α3-transfected MK^-/- cells (Fig. 3G). Taken together, results from Figs 2 and 3 indicate that α6β4-mediated adhesion to LN-5 ECM may contribute to keratinocyte survival in the absence of α3β1-mediated adhesion, but that it is not necessary for maximal survival of cultured keratinocytes that are adhered to LN-5 through α3β1.

Focal adhesion kinase (FAK) has been shown to play an important role in integrin-dependent cell survival (Frisch et al., 1996; Frisch and Screaton, 2001). Auto-phosphorylation of FAK at Tyr-397 is a critical initiating event in FAK-mediated signal transduction pathways (Cary and Guan, 1999). To determine whether increased apoptosis in MK^-/- cells was correlated with reduced FAK signaling, we plated MK cells on LN-5 ECM in serum-free medium and assayed for FAK activation by immunoblot with an antibody specific for phosphorylation at Tyr-397 (Fig. 4A, upper panel). Control blots for total FAK showed that FAK protein levels were similar under all conditions (Fig. 4A, lower panel). As expected, FAK activation in MK^+/+ cells was adhesion-dependent because phospho-FAK was detected in adherent MK^+/+ cells (Fig. 4A, LN-5 ECM, +/+ lanes), but it was barely detectable in non-adherent MK^+/+ cells (Fig. 4A, suspended, +/+ lanes). In contrast, FAK activation was reduced markedly in α3β1-deficient MK^-/- cells adhered to LN-5 ECM (Fig. 4A, LN-5 ECM, -/- lanes). Adhesion-dependent FAK activation was completely restored in MK^-/- cells transfected with human α3 (Fig. 4A, α3 lanes). Phospho-FAK levels remained suppressed in α3-null cells, relative to α3β1-expressing cells, under serum-free conditions for 24 hours (Fig. 4B) or 48 hours (not shown), consistent with the time course of MK^-/- cell apoptosis in our cell survival assays.

Tyrosine phosphorylation of the adaptor protein CAS serves as an additional readout for FAK activation, because CAS binds directly to FAK and is subsequently phosphorylated by Src family kinases that bind to active FAK at phospho-Tyr-397 (Cary and Guan, 1999). As expected, phosphorylation of CAS was also adhesion-dependent in MK^+/+ cells (Fig. 4C, +/+ lanes). In contrast, adherent MK^-/- cells expressed barely detectable levels of phospho-CAS (Fig. 4C, -/- lanes), which were completely restored upon transfection with α3 (Fig. 4C, α3 lanes). Interestingly, we observed lower levels of total CAS protein in α3-null MK^-/- cells compared with MK^+/+ cells or α3-transfected MK^-/- cells (Fig. 4C, lower panel). Importantly, however, comparison of band intensities for total and phospho-CAS between α3-expressing MK cells and MK^-/- cells indicates that detectable CAS protein in MK^-/- cells is completely non-phosphorylated, reflecting reduced FAK function in these cells. Reduced levels of total CAS protein in MK^-/- cells may reflect an additional level at which α3β1 regulates FAK/CAS signaling, and the mechanism of this regulation is the subject of a separate study. Taken together, results in Fig. 4 show that absence of α3β1 causes decreased FAK signaling in MK cells adhered to LN-5 under serum-free conditions, consistent with loss of FAK-mediated signaling pathways that promote epithelial cell survival.

FAK-related non-kinase (FRNK) is an autonomously expressed product of the FAK gene that consists of only the C-terminal region of FAK and acts as a competitive inhibitor of FAK-mediated signal transduction from focal adhesions (Cary and Guan, 1999). To determine the effects of inhibiting FAK function on keratinocyte survival, we infected wild-type MK^+/+ cells with an adenovirus encoding a green fluorescent (GFP)FRNK fusion protein, which was shown previously to disrupt FAK signaling and induce anoikis in rat ventricular myocytes (Heidkamp et al., 2002). As a control, MK^+/+ cells were infected at the same multiplicity of infection with an adenovirus that expresses GFP only. Expression of GFP and GFP-FRNK was confirmed by immunoblotting with anti-GFP (Fig. 5A). As shown previously for ventricular myocytes (Heidkamp et al., 2002), expression of GFP-FRNK in MK^+/+ cells reduced FAK phosphorylation at Tyr-397 (data not shown). Infected cells were cultured on LN-5 ECM under serum-free conditions for 24 hours, and caspase-3 activation was assayed by immunoblot. A higher proportion of caspase-3 was cleaved in GFP-FRNK-infected cells than in GFP-infected cells or in uninfected cells (Fig. 5B), indicating that GFP-FRNK expression specifically induced apoptosis.

The apoptosis-inducing effects of GFP-FRNK expression were clearly evident upon microscopic analysis of infected MK^+/+ cells. Control GFP-infected cultures contained cells with various levels of GFP expression, as determined by relative fluorescence intensity (Fig. 5Ca). Although culture under serum-free conditions induced a rounded, apoptotic morphology in a small proportion of these cells, the majority of GFP-expressing cells maintained a spread morphology typical of healthy MK cells on LN-5, including those expressing high levels of GFP (Fig. 5Ca,b, arrowheads). In contrast, cultures infected with GFP-FRNK showed a dramatic increase in the number of cells with rounded, apoptotic morphology (Fig. 5Cc,d). Many of these cells showed membrane blebbing characteristic of apoptosis and had detached from the substrate by 24 hours. Importantly, the vast majority of GFP-FRNK expressing (i.e. fluorescent) cells were apoptotic (Fig. 5Cc,d, arrowheads). A few uninfected (i.e. non-fluorescent) cells remained well spread over the course of the experiment, providing an internal control (Fig. 5Cc,d, arrows). Taken together, results in Figs 4 and 5 suggest that α3β1-mediated FAK activation promotes keratinocyte survival.

A number of previous studies have implicated the MAP kinase ERK1/2 in the adhesion-dependent survival of keratinocytes and other epithelial cells (Jost et al., 2001; Gu et al., 2002; Frisch and Screaton, 2001). Although α3β1 has been shown to activate ERK in some epithelial cells (Gonzales et al., 1999), a role for α3β1-mediated ERK activation in keratinocyte survival has not been demonstrated. To determine whether α3β1 regulates ERK activation in keratinocytes, MK^+/+ cells, MK^-/- cells, or α3-transfected MK^-/- cells were kept in suspension or adhered to LN-5 ECM under serum-free conditions. Cell lysates were then immunoblotted for the activated forms of ERK with an antibody specific for phosphorylation on residues Thr202 and Tyr204 of p44/42 ERK (Fig. 6A, upper panels), or for total ERK protein as a control (Fig. 6A, lower panels). In addition, ERK activity was assayed using an in vitro kinase assay to detect ERK-mediated phosphorylation of an Elk-1 substrate (Fig. 6B). Neither phosphorylated ERK nor ERK activity was detected in MK cells that were held in suspension, despite high levels of ERK protein in these cells (Fig. 6A,B, susp.). However, both phosphorylated ERK and ERK activity were detected in wild-type MK^+/+ cells 15 minutes after attachment to LN-5 ECM (Fig. 6A,B; LN-5 ECM, 15 min., +/+), demonstrating adhesion-dependent activation of ERK. In contrast, α3β1-deficient MK^-/- cells contained considerably reduced levels of phosphorylated ERK and ERK activity after adhesion to LN-5 ECM for 15 minutes (Fig. 6A,B; LN-5 ECM, 15 min., -/-). Stable transfection of MK^-/- cells with α3 completely restored both ERK phosphorylation and ERK activity (Fig. 6A and B; LN-5 ECM, 15 min., α3).

Basal levels of phosphorylated ERK were observed in MK^-/- cells adhered to LN-5 ECM, and these levels increased somewhat after overnight culture. However, phospho-ERK levels remained considerably lower in MK^-/- cells than in α3β1-expressing MK cells even after 24 hours (Fig. 6A; LN-5 ECM, 24 hr.; also, see Fig. 7A). The delayed, low levels of ERK activation seen in MK^-/- cells may occur in response to growth factors or ECM ligands that are produced by the keratinocytes during culture or that are present at low levels in the LN-5 ECM preparation. Alternatively, basal levels of ERK activation in MK^-/- cells could be due to α6β4-mediated adhesion, because this integrin has been reported to activate ERK in keratinocytes (Mainiero et al., 1997). However, α6β4-mediated adhesion was not necessary to maintain high levels of ERK activation in α3β1-expressing MK^+/+ cells grown on LN-5 ECM for 24 hours, because blocking α6β4 function with GoH3 over this time course did not reduce the levels of phosphorylated ERK (Fig. 6C). These results suggest that α3β1, but not α6β4, is required for full activation of ERK in keratinocytes adhered to LN-5.

Because the absence of α3β1 from MK cells resulted in both reduced ERK activation and increased apoptosis, we next wanted to determine whether inhibition of ERK activation leads to increased apoptosis in MK cells. ERK is phosphorylated and activated by MAPK/ERK kinase (MEK), and ERK activation can be suppressed by treating cells with the MEK-specific inhibitor U0126. MK^+/+ cells or MK^-/- cells were cultured on LN-5 ECM in serum-free medium in the presence of U0126, or DMSO as a control (Fig. 7). Treatment with 10 μM U0126 was sufficient to inhibit completely the high levels of ERK phosphorylation seen in MK^+/+ cells, as well as the basal levels of ERK phosphorylation seen in MK^-/- cells (Fig. 7A). Treatment with 10 μM U0126 also caused increased cleavage of caspase-3 in MK^+/+ cells (Fig. 7B), indicating a requirement for MEK/ERK signaling in cell survival. U0126 treatment had no obvious effect on cell spreading before the development of apoptotic cell morphology (data not shown). In some experiments, we observed that caspase-3 activation was also increased slightly in MK^-/- cells treated with U0126 (Fig. 7B), suggesting that the basal levels of ERK activity observed in these cells may contribute to survival. Taken together, results in Figs 6 and 7 reveal an important role for MEK/ERK signaling in keratinocyte survival, and indicate that α3β1 mediates the majority of ERK-dependent survival.

Growth in serum dramatically reduced the amount of apoptosis that occurred in α3-null MK^-/- cells (Fig. 1G and Fig. 2A), consistent with the established importance of growth factor receptor activation in keratinocyte survival (Rodeck et al., 1997; Jost et al., 2001; Sibilia et al., 2000). Many soluble growth factors and some integrins promote cell survival through activation of the small GTPase Ras (Giancotti and Ruoslahti, 1999). Activating mutations in Ras occur frequently in epithelial cancers (Shields et al., 2000), and oncogenic forms of Ras can confer resistance to anoikis in keratinocytes and other epithelial cells (Frisch and Francis, 1994; Rosen et al., 2000; Zhu et al., 2002). To determine whether oncogenic Ras can rescue α3-null MK^-/- cells from anoikis, we expressed the constitutively active Ras-V12 mutant in MK^-/- cells and tested its ability to inhibit caspase-3 activation under serum-free conditions. MK^-/- cells were infected with an adenovirus expressing either HA-tagged Ras-V12 (Meadows et al., 2001) or β-galactosidase as a control, then cultured on LN-5 ECM under serum-free conditions for 24 hours. Ras-V12 expression was confirmed by immunoblotting for HA-tag (Fig. 8A, HARas-V12 blot, lanes 5 and 6). MK^-/- cells infected with control adenovirus showed low basal levels of phospho-ERK, similar to those seen in uninfected cells (Fig. 8A, pERK blot, lanes 2 and 3). In contrast, expression of Ras-V12 caused increased levels of phospho-ERK compared with control cells (Fig. 8A, pERK blot, lanes 3 and 5). Ras-V12-mediated ERK phosphorylation was inhibited completely by treatment with U0126 (Fig. 8A, pERK blot, lanes 5 and 6), indicating that Ras-V12 activates MEK/ERK signaling in MK cells.

Under serum-free conditions, MK^-/- cells infected with control adenovirus showed high levels of cleaved caspase-3 that were comparable to those seen in uninfected MK^-/- cells (Fig. 8B, lanes 2 and 3). In contrast, MK^-/- cells infected with Ras-V12 adenovirus showed reduced activation of caspase-3 that was comparable to background levels seen in uninfected MK^-/- cells grown in the presence of serum (Fig. 8B, compare lanes 5 and 1). These results show that oncogenic activation of Ras can suppress apoptosis in α3β1-deficient keratinocytes.

Although the presence of serum protected MK^-/- cells from apoptosis, as assayed by either TUNEL (Fig. 1B) or caspase-3 cleavage (Fig. 8B, lanes 1 and 2), only basal levels of phospho-ERK were detected after 24 hours in the presence of serum (Fig. 8A, lanes 1 and 2). These results suggest that serum growth factors, in contrast with α3β1, can promote MK cell survival through pathways that do not require sustained ERK activity. Although Ras-V12 activated ERK in MK^-/- cells, Ras has numerous effectors and can initiate multiple intracellular signaling pathways (Marshall, 1996; Shields et al., 2000). Indeed, previous studies in MDCK epithelial cells have shown that oncogenic Ras promotes survival through the PI3K pathway, but not through the Raf/MEK/ERK pathway (Khwaja et al., 1997). Therefore, we wanted to determine the effects of inhibiting MEK on the ability of Ras-V12 to suppress apoptosis in MK^-/- cells. Treatment of infected cells with 10 μM U0126 had no effect on the ability of Ras-V12 to completely inhibit caspase-3 activation (Fig. 8B, lanes 5 and 6), even though this same treatment efficiently reduced ERK phosphorylation to basal levels in the same cells (Fig. 8A, lanes 5 and 6). These results indicate that sustained MEK/ERK signaling was not required for Ras-V12-mediated survival, at least over the time course of our assay, and suggest that oncogenic Ras can stimulate keratinocyte survival through pathways that are distinct from the MEK/ERK-dependent pathways induced by α3β1.

Previous studies have demonstrated that integrins can have either pro-survival or pro-apoptotic roles in the regulation of cell survival (Stupack and Cheresh, 2002). Our findings identify a novel role for α3β1 integrin in promoting the survival of epidermal keratinocytes through a MEK/ERK signaling pathway. This pro-survival role for α3β1 in keratinocytes is clearly distinct from previously described roles in cells of distinct origin or transformation status, where α3β1 and/or its laminin ligands were shown either to promote apoptosis (Seewaldt et al., 2001; Sato et al., 1999) or to promote cell survival in a MEK/ERK-independent manner (Gu et al., 2002). Although it is known that LN-5 promotes keratinocyte survival in the epidermis (Ryan et al., 1999; Nguyen et al., 2000), the relative roles of the LN-5-binding integrins α3β1 and α6β4 and the signaling pathways involved are unclear. In the current study, we showed that while α3-null keratinocytes retained the ability to adhere to LN-5 through integrin α6β4, this adhesion did not fully protect cells from apoptosis induced by serum deprivation. These observations provide a clear example of how integrin-mediated adhesion to laminin per se may not be sufficient to inhibit anoikis of epithelial cells completely, but that appropriate adhesion through a specific integrin receptor is necessary to protect epithelial cells fully from anoikis.

In the absence of α3β1, α6β4-dependent adhesion appeared to compensate partially to promote keratinocyte survival, consistent with previous reports that α6β4 has pro-survival functions in keratinocytes and other epithelial cells (Dowling et al., 1996; Bachelder et al., 1999; Weaver et al., 2002). In support of this conclusion, the majority of α3-null MK^-/- cells that were adhered to LN-5 ECM through α6β4 remained negative for TUNEL-staining in our survival assays (Fig. 1). Furthermore, MK^-/- cells that were adhered to LN-5 through α6β4 showed lower levels of activated caspase-3 than did unattached MK^-/- cells (Fig. 2). Importantly, however, blocking α6β4-mediated adhesion with GoH3 did not increase MK cell apoptosis when α3β1-mediated adhesion was intact (Fig. 3), indicating that α3β1-mediated adhesion, but not α6β4-dependent adhesion, was sufficient for maintaining maximal levels of cell survival in our assays.

We cannot exclude the possibility that other integrins also contribute to the increased survival that we observed in adherent α3-null MK^-/- cells compared with non-adherent MK cells. Indeed, while α6β4 was clearly required for adhesion of MK^-/- cells to LN-5 ECM, it is possible that adherent MK^-/- cells subsequently interact with fibronectin or other ECM proteins that are present in the LN-5 ECM preparation or that are deposited by the MK cells themselves following attachment. This possibility is consistent with our earlier observations in mice lacking both α3β1 and α6β4, where increased apoptosis was seen only in regions of epidermis that had already detached from the basement membrane (DiPersio et al., 2000b).

Both α3β1 and α6β4 are expressed constitutively during epidermal development and in adult epidermis (Watt, 2002). While our results indicate that both of these integrins can contribute to keratinocyte survival in culture, the relative contributions of these two integrins to cell survival in quiescent epidermis remain unclear. It is possible that α6β4, rather than α3β1, is the primary LN-5 receptor for maintaining keratinocyte survival in embryonic and adult epidermis, because it is clearly the major receptor for epidermal adhesion in vivo (van der Neut et al., 1996; Georges-Labouesse et al., 1996; Dowling et al., 1996). As mentioned above, interactions with ECM ligands other than LN-5 are likely to contribute to keratinocyte survival in normal epidermis, because TUNEL-positive keratinocytes in mutant mice that lack both α3β1 and α6β4 were restricted to detached regions of epidermis (DiPersio et al., 2000b).

The potential importance of α3β1 in keratinocyte survival becomes more obvious when one considers the changes in cell-ECM interactions and the dramatic shifts in integrin function that occur during cutaneous wound healing. Activated keratinocytes at the wound edge show a redistribution of α6β4 from the basal cell surface to the baso-lateral surface, presumably reflecting a requirement to disassemble hemidesmosomes and reduce stable adhesion in migrating keratinocytes (Nguyen et al., 2000). Concurrently, α3β1 redistributes mainly to the basal surface of keratinocytes in the leading edge of the wound, where it can bind to newly deposited LN-5 and promote cell migration (Lampe et al., 1998; Nguyen et al., 2000). This switch from α6β4-LN-5 adhesion to α3β1-LN-5 adhesion in leading edge keratinocytes also occurs in in vitro scrape wounds of breast epithelial cells (Goldfinger et al., 1999) or keratinocytes (D. Choma and C. M. DiPersio, unpublished). We propose a model in which increased binding of α3β1 to LN-5 in activated keratinocytes at the wound edge becomes important for cell survival following the dissolution of α6β4-LN-5 adhesions (Fig. 9). This function may be most critical during the initial phase of keratinocyte activation, when leading edge cells have not yet engaged other ECM ligands present in the wound bed. In this model, α3β1-mediated adhesion to LN-5 would facilitate at least two processes essential to re-epithelialization of the wound: (1) keratinocyte migration into the wound bed and (2) maintenance of keratinocyte survival during migration and ECM remodeling. Importantly, our findings do not rule out pro-migratory or pro-survival roles for other integrin-ECM interactions that occur during wound healing.

Activation of ERK signaling is important for adhesion-dependent survival in a number of cell types (Frisch and Screaton, 2001). In keratinocytes, a MEK/ERK-dependent survival pathway can be induced by EGF receptor (EGFR) activation (Jost et al., 2001). In the current study, we show that α3β1 integrin can also stimulate MEK/ERK-dependent survival. Two major mechanisms whereby integrins activate the Ras-ERK signaling cascade are through FAK activation and through the Fyn/Shc pathway (Giancotti and Ruoslahti, 1999). FAK can be activated by most integrins, including α3β1. In contrast, only a subset of integrins can activating the tyrosine kinase Fyn, which activates Ras pathways through recruitment of the Shc and Grb2 adaptor proteins (Wary et al., 1996; Giancotti and Ruoslahti, 1999). Interestingly, α3β1 was among those integrins that did not activate the Fyn/Shc pathway (Wary et al., 1996), suggesting that α3β1-dependent ERK activation probably occurs through a FAK-mediated pathway. Indeed, it has been well established that integrin-mediated FAK signaling can lead to ERK activation (Schlaepfer et al., 1997). Consistent with this idea, we showed that α3β1 stimulates both FAK activation and ERK activation in keratinocytes, and that inhibition of either FAK function through FRNK over-expression or ERK signaling through MEK inhibition leads to increased apoptosis.

As we did not assay FAK or ERK activation within individual cells, we were unable to determine whether reduced FAK or ERK signaling occurred in all cells of α3-null MK^-/- cultures, or whether a subpopulation of cells retained FAK/ERK signaling despite the absence of α3β1. Indeed, a large fraction of MK^-/- cells remained TUNEL-negative for up to 48 hours (Fig. 1), suggesting that there is heterogeneity among α3-null keratinocytes regarding their sensitivity to apoptosis. Future experiments using in situ approaches to assess the phosphorylation or sub-cellular localization of FAK and ERK within individual cells should help to distinguish between these possibilities.

There are several potential mechanisms whereby α3β1-mediated activation of FAK could lead to MEK/ERK-dependent keratinocyte survival. For example, FAK/CAS interactions may play a role in cell survival, because CAS/Crk coupling and ERK activation have been linked to suppression of apoptosis in some cells (Cho and Klemke, 2000). In addition, FAK interaction with the Grb2-mSOS complex leads to activation of the Ras/Raf/MEK/ERK cascade (Schlaepfer et al., 1994). Another possibility is that activation of FAK leads to activation of PI3K, which in turn can activate PAK. Active PAK can then enable a Ras/Raf/MEK/ERK signaling cascade by phosphorylating Raf-1 at a site that is necessary for its activation by Ras (King et al., 1998; Giancotti and Ruoslahti, 1999; Eblen et al., 2002). Consistent with a role for PI3K in keratinocyte survival, treatment of α3β1-expressing MK cells with the PI3K inhibitor LY294002 also induced caspase-3 activation (data not shown). However, PI3K also promotes ERK-independent cell survival through activation of the kinase AKT (Frisch and Screaton, 2001), and further experimentation is required to determine whether PI3K and ERK promote MK cell survival through overlapping or distinct pathways.

Integrin-dependent changes in cell shape and cytoskeletal integrity can also play a key role in regulating cell survival (Giancotti and Ruoslahti, 1999; Frisch and Screaton, 2001). For example, the nuclear localization of ERK can be regulated by integrin-ECM interactions (Aplin et al., 2001) and may play a role in survival of some cell types (Lai et al., 2002). Indeed, adhesion-dependent changes in the actin cytoskeleton that reduce ERK nuclear localization can lead to decreased transcription of genes that promote cell survival (Schulze et al., 2001). Keratinocytes that lack integrin α3β1 adhere to LN-5 efficiently through integrin α6β4, but they spread poorly and display defects in the actin cytoskeleton that are evident both in cultured cells and in vivo (DiPersio et al., 1997; Hodivala-Dilke et al., 1998; DiPersio et al., 2000a). α3-null kidney collecting duct cells display similar defects in cytoskeletal organization (Wang et al., 1999). Therefore, the ability of α3β1 to activate ERK and promote cell survival may be due, at least in part, to the ability of this integrin to promote cell spreading on LN-5 and organize the actin cytoskeleton, a function which α6β4 cannot fulfill. Future experiments will directly test the importance of cell spreading for activation of MEK/ERK survival pathways in keratinocytes.

We thank Andrew Aplin for critical reading of the manuscript, and Livingston Van De Water for helpful discussions. We also thank Allen Samarel for providing adenovirus encoding GFP-FRNK fusion protein, and Patrick Bryant for preparation of adenovirus. This research was supported by a grant from the National Institutes of Health to C.M.D. (R01CA84238), and a grant from the National Institutes of Health to K.P. (R01CA081419). S. G. Shome was supported by a post-doctoral training grant from the National Institute of General Medical Sciences (NIH-T32-GM-07033). J. Lamar and V. Iyer were supported by a pre-doctoral training grant from the National Heart, Lung, and Blood Institute (NIH-T32-HL-07194).