INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

The integrins constitute a large family of cell surface receptors that mediate cell-cell and cell-extracellular matrix adhesion. Each integrin is a heterodimer of an  and a  subunit, both of which are transmembrane glycoproteins. The ligand-binding specificity of a given integrin is determined by the combination of  and  subunits it comprises and by the cell type in which it is expressed (Hynes, 1992).

Integrins can transduce two types of signals: receptor conformation, affinity, and clustering are regulated by intracellular events (inside-out signaling), whereas ligand binding triggers a variety of cellular responses (outside-in signaling), including actin polymerization and cell spreading, induction of gene expression, initiation of differentiation, and suppression of apoptosis (Hynes, 1992; Juliano and Haskill, 1993; Williams et al., 1994; Hughes and Pfaff, 1998). A variety of outside-in signal transduction pathways have now been defined, many of which are also activated by growth factors and cytokines (Sastry and Horwitz, 1993; Clark and Brugge, 1995; Yamada and Miyamoto, 1995; Howe et al., 1998).

In the case of the 1 integrins, the outside-in signals that have so far been characterized involve a synergy between ligand binding and receptor aggregation; neither event alone is sufficient for signal transduction (Miyamoto et al., 1995; Yamada and Miyamoto, 1995). Aggregation of 1 integrins occurs in focal adhesions, where integrins are associated with actin bundles via cytoskeletal proteins, such as talin, vinculin, and -actinin, and with protein kinases, including focal adhesion kinase (FAK) (Schaller et al., 1992) and PKC (Jaken et al., 1989; Woods and Couchman, 1992). Thus, focal adhesions constitute integrin-signaling complexes (Schaller et al., 1994; Shattil et al., 1994).

The cytoplasmic domain of the 1 integrin subunit (reviewed by Hemler et al., 1994; Williams et al., 1994) contains sufficient information for localization to focal adhesions (LaFlamme et al., 1992) and directly binds a variety of structural and regulatory proteins, including talin (Horwitz et al., 1986), -actinin (Otey et al., 1990), and FAK (Schaller et al., 1995) (reviewed by Hemler, 1998; Howe et al., 1998). The amino acids within the 1 cytoplasmic domain that are required for localization to focal adhesions have been identified by extensive mutation and deletion analysis (Solowska et al., 1989; Hayashi et al., 1990; Marcantonio et al., 1990; Reszka et al., 1992) and include three clusters of amino acids designated cyto-1 (residues 764-774), cyto-2 (residues 785-788), and cyto-3 (residues 797-800), cyto-2 and -3 being NPXY motifs. The cyto-1, -2, and -3 clusters include amino acids that are conserved among integrin subunits 1-7 (Williams et al., 1994).

Human epidermal keratinocytes represent a unique experimental model for studies of the role of 1 integrins in regulating differentiation (reviewed by Watt and Jones, 1993; Watt and Hertle, 1994). Loss of integrin ligand-binding ability occurs on commitment to terminal differentiation (Adams and Watt, 1990; Hotchin and Watt, 1992; Hotchin et al., 1993), and this ensures that differentiation is linked to detachment of keratinocytes from the underlying basement membrane. The keratinocytes with the highest proliferative potential, the stem cells, express higher levels of 1 integrins than other keratinocytes in the epidermal basal layer (Jones and Watt, 1993; Jones et al., 1995b; Jensen et al., 1999), and reduction in 1-mediated adhesion stimulates exit from the stem cell compartment via a mechanism that involves MAPK signaling (Zhu et al., 1999). Integrin expression is normally confined to the epidermal basal layer, and suprabasal expression can result in hyperproliferation (Carroll et al., 1995). Aberrant integrin expression is a feature of squamous cell carcinomas (SCCs), and there is evidence from transfection experiments that loss of integrins in these tumors can render the cells "deaf" to positive or negative growth and differentiation signals from the extracellular matrix (Jones et al., 1993, 1995a; 1996a; Bagutti et al., 1998; and references cited therein). Finally, when normal keratinocytes are placed in suspension, they are stimulated to undergo terminal differentiation; this can be partially inhibited by extracellular matrix proteins or antibodies to 1 integrins, showing that adhesion normally suppresses terminal differentiation (Adams and Watt, 1989; Watt et al., 1993).

Although negative regulation of terminal differentiation by 1 integrins could simply be a consequence of 1-mediated adhesion, there is reason to suspect that the mechanism by which 1 integrins regulate the onset of terminal differentiation is distinct from the mechanism by which they mediate keratinocyte adhesion. Thus, keratinocyte spreading on extracellular matrix proteins involves 1 integrin clustering in focal adhesions and is abolished by CD (see, for example, Carter et al., 1990), whereas the inhibition of differentiation does not involve or require polymerization of the actin cytoskeleton and can be effected by Fab fragments of anti-integrin antibodies (Adams and Watt, 1989; Watt et al., 1993). To find out more about the differentiation-regulatory role of 1 integrins, we have introduced a series of wild-type and mutant 1 subunits into normal human keratinocytes and an undifferentiated SCC line and examined their activity in adhesion and terminal differentiation assays.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

Construction of Retroviral Vectors and Producer Cell Lines

The pRSVneo-1 vector containing the wild-type chick 1 integrin cDNA or a series of cytoplasmic domain mutants was generously provided by A. Reszka and A.F. Horwitz (University of Illinois, Urbana, IL). The following deletion and point mutations in the cytoplasmic domain were examined: 759-771, 771-790, N797I, N785I/N797I, Y788A/N797I (Reszka et al., 1992), YPRF, and YTRF (mutations of the NPIY motif at amino acids 785-788) (Lilienbaum et al., 1995). In addition, an inactivating point mutation in the extracellular domain, D154A, which is equivalent to the human D130A mutation that blocks ligand binding (Tamkun et al., 1986; Takada et al., 1992), was generated by PCR with the use of the wild-type chick 1 cDNA as template.

The chick cDNAs were removed from the parental vector as SalI fragments. The cDNAs were then cloned into the SalI site of the retroviral vector pBabe puro (Morgenstern and Land, 1990), and all mutations were confirmed by sequencing. Retroviral DNA was transfected into the ecotropic cell line GP+E via calcium phosphate-mediated transfection, and after 48 h of growth, supernatants from the transfected ecotropic cells were used to infect the amphotropic packaging cell line AM12, as described previously (Levy et al., 1998). AM12 cells with viral titers of 3 × 10⁵-5 × 10⁶ colony-forming units/ml were selected by a combination of FACS with anti-chick 1 antibodies and clonal selection in puromycin (Levy et al., 1998).

Cell Culture

Human epidermal keratinocytes were isolated from newborn foreskin and cultured in the presence of a mitomycin C-treated J2-3T3 feeder layer, as described previously (Watt, 1998). The culture medium consisted of one part Ham's F-12 medium and three parts DMEM, 1.8 × 10⁴ M adenine, 10% FCS, 0.5 µg/ml hydrocortisone, 5 µg/ml insulin, 10¹⁰ M cholera toxin, and 10 ng/ml EGF (FAD+FCS+HICE). For all experiments, cells were used at passage three or four, and 3T3 feeder cells were selectively removed with EDTA before keratinocytes were harvested. SCC4, a cell line derived from a squamous carcinoma of human tongue (Rheinwald and Beckett, 1981), was also cultured with a J2-3T3 feeder layer in FAD+FCS+HICE.

To infect keratinocytes and SCC4 with retroviral vectors, the cells were seeded onto preconfluent AM12 packaging cells that had been pretreated with 4-40 µg/ml (depending on the AM12 clone) mitomycin C. A total of 1.5 µg/ml puromycin was added after 2 d to select for infected cells. After 4-5 d, the packaging cells were removed with EDTA and replaced with puromycin-resistant J2-3T3 cells, as described previously (Zhu and Watt, 1996).

Terminal differentiation of primary keratinocytes was induced by suspending disaggregated cells in culture medium (FAD+HICE supplemented with 10% FCS from which fibronectin had been removed by affinity chromatography on gelatin Sepharose; a generous gift of K. Hodivala-Dilke [Massachusetts Institute of Technology, Cambridge, MA]) supplemented with 1.65% methyl cellulose at a density of 10⁵ cells/ml. Culture dishes were coated with 0.4% poly(2-hydroxyethyl methacrylate) to prevent cell attachment (Watt et al., 1988). The cells were recovered from suspension by diluting the methyl cellulose 10-fold with EDTA and then centrifuging, as described previously (Watt, 1994). In experiments examining the effects of the anti-1 integrin antibodies on terminal differentiation, antibodies (immunoglobulin G [IgG] and Fab fragments) were added to a final concentration of 100 µg/ml (Watt et al., 1993).

Fibroblasts were isolated from chick embryos by trypsinization or outgrowth from explants and cultured in DMEM supplemented with 5% FCS. The AM12 packaging cells were cultured in medium consisting of DMEM supplemented with 10% FCS and 1.5 µg/ml puromycin.

Antibodies and Extracellular Matrix Proteins

The mAbs used in adhesion and differentiation assays were P5D2 (anti-human 1 integrin; Dittel et al., 1993), W1B10 (anti-chick 1 integrin; Reszka et al., 1992), and JG22 (anti-chick 1 integrin; Greve and Gottlieb, 1982). Fab fragments were prepared by papain digestion of 0.5 mg of IgG with the use of a Fab preparation kit (Pierce, Rockford, IL). Involucrin was detected with the use of a rabbit antiserum (DH1; Dover and Watt, 1987). The following mAbs were used for flow cytometry: JG22, P5D2, HAS4 (human 21; Tenchini et al., 1993), VM-2 (human 31; Kaufmann et al., 1989), SAM1 (human 51; te Velde et al., 1988), l3C2 (human v; Horton et al., 1985), MP4F10 (human 6; Anbazhagan et al., 1995), and 3E1 (human 64; Ryynänen et al., 1991). For immunofluorescence staining, a rat mAb to 1 integrins (AIIB2; Werb et al., 1989) (Developmental Studies Hybridoma Bank, Iowa City, IA), rabbit anti-chick 1 integrins (Chickie; Shih et al., 1993) (a generous gift of Clayton Buck [Wistar Institute, Philadelphia, PA]), and mouse anti-vinculin (VIN-11-5; Sigma Chemical, Poole, UK) were used and detected with Alexa 488- or Alexa 594-conjugated secondary antibodies (Molecular Probes, Eugene, OR). Mouse EHS laminin and human placental type IV collagen were supplied by Sigma Chemical. Human plasma fibronectin was supplied by Bio-Products (Elstree, UK).

Flow Cytometry

Keratinocytes (5 × 10⁵ cells) were incubated with anti-integrin antibodies diluted in PBS containing 1 mM CaCl₂ and 1 mM MgCl₂ (PBSABC) on ice for 30 min with occasional agitation. After washing in the dilution buffer at 4°C, the cells were resuspended in the appropriate FITC-conjugated secondary antibody and incubated as before. The cells were washed again and then analyzed on a FACScan (Becton-Dickinson Immunocytometry Systems, Mountain View, CA), as described by Jones and Watt (1993).

Indirect Immunofluorescence Staining of Focal Adhesions

To visualize focal adhesions, cells were fixed and permeabilized simultaneously in 3.7% formaldehyde and 0.2% Triton X-100 in PBS for 10 min at room temperature. The cells were incubated with the first primary antibody for 45 min, washed extensively in PBS, incubated with Alexa-conjugated secondary antibody, washed again, incubated with the second primary antibody, washed again, incubated with the second Alexa-conjugated antibody, and washed once more. Stained cells were mounted in Gelvatol (Monsanto, St. Louis, MO) and examined under epifluorescence with the use of a Zeiss (Herts, UK) Axiophot microscope or a Zeiss LSM-500 laser scanning confocal microscope.

Adhesion Assays

Extracellular matrix proteins and anti-integrin antibody concentrations were chosen on the basis of previous experiments (Adams and Watt, 1991). Microtiter plates (Immulon II, Dynatech, Billingshurst, England) were coated with fibronectin (10 µg/ml), laminin 1 (30 µg/ml), or type IV collagen (20 µg/ml) overnight at 4°C. After washing with PBS, unbound sites were blocked by incubation with PBSABC containing 0.5 mg/ml heat-treated BSA for 1 h at 37°C. Primary keratinocytes or SCC4 cells were harvested and resuspended in serum-free growth medium. A total of 2 × 10⁴ cells were added per well (in triplicate) and incubated for 2 h at 37°C. Unbound cells were washed off with PBSABC, and bound cells were lysed with medium containing 1% Triton X-100. Quantitative measures of lactate dehydrogenase, a cytosolic enzyme that is released upon cell lysis, were performed with the use of the Cytotox 96 colorimetric kit (Promega, Madison, WI). The percentage of cells adhering was calculated with the use of a standard curve prepared by titrating known numbers of cells. For each treatment, nonspecific adhesion to BSA was <5% of cells plated. Antibodies were added for the 2-h adhesion period at a total concentration of 100 µg/ml (i.e., when two antibodies were added in combination, each was at 50 µg/ml). Results presented are the mean of triplicate determinations ± SEM and are representative of data from at least two, and in most cases four, separate experiments.

Measurement of the Proportion of Involucrin-positive Keratinocytes

Single cell suspensions of primary keratinocytes or SCC4 cells were air dried onto coverslips, fixed in 3.7% formaldehyde in PBS, permeabilized in methanol, and stained with the DH1 rabbit antiserum to involucrin and a fluorescein-conjugated anti-rabbit secondary antibody, as described previously (Read and Watt, 1988). Statistical comparisons were made with the use of Student's t test.

	RESULTS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

Expression of Chick 1 Integrin Subunits in Keratinocytes and SCC4 Cells

To distinguish mutant forms of the 1 integrin subunit from the endogenous human receptor, we introduced the chick 1 subunit, because this could be identified with species-specific antibodies. The anti-chick 1 mAbs used have been characterized previously in epitope-mapping experiments with chick/human 1 integrin chimeras (Shih et al., 1993). Figure 1 shows the amino acid sequence of the C-terminal 47 amino acids of the chick 1 subunit, which constitutes the entire cytoplasmic domain (Williams et al., 1994) and is completely conserved between chick (Tamkun et al., 1986) and human (Argraves et al., 1987). The three clusters of amino acids that contribute to focal adhesion localization, cyto-1, -2, and -3, are shown in boxes (Reszka et al., 1992). We compared the behavior of the wild-type chick subunit with the series of point and deletion mutations shown. We also generated a point mutation in the extracellular domain, resulting in substitution of aspartic acid for alanine at amino acid 154 (D154A); the equivalent mutation in the human, D130A, has been shown to inhibit ligand binding but not recruitment to focal adhesions (Takada et al., 1992).

View larger version (22K): [in this window] [in a new window]   Figure 1.   Summary of the chick 1 subunits tested and the results obtained. The sequences of the wild-type (WT) cytoplasmic domain and the point and deletion mutations within it are shown. The D154A mutation in the extracellular domain is shown in a schematic representation of the entire 1 integrin subunit, with the transmembrane domain indicated in black. Those chick 1 subunits that had activity in the keratinocyte differentiation assay were all equally effective, and the results are therefore summarized as "++" (see Figure 6). The subunits differed in activity in the SCC4 assay (see Figure 7), and a positive effect is therefore summarized as "++" or "+."

The chick 1 constructs were introduced into normal human keratinocytes and a poorly differentiated cell line, SCC4, derived from a SCC, via retroviral infection and selection for puromycin resistance, as described previously (Zhu and Watt 1996; Levy et al., 1998). The cells stably expressed each construct and could be passaged several times without loss of expression. Flow cytometry with the use of mAbs specific for the chick 1 subunit established that the level of surface expression of all of the constructs was comparable to that of the endogenous human 1 subunit (Figure 2A-C) (Levy et al., 1998; Zhu et al., 1999; our unpublished results). The proportion of cells that expressed each construct was >70% except in the case of Y788A/N797I, and values of 90% were routinely achieved (see, for example, Figure 2, B and C). The proportion of cells that expressed the Y788A/N797I construct was ~50%, possibly reflecting impaired intracellular transport (see Levy et al., 1998).

View larger version (45K): [in this window] [in a new window]   Figure 2.   Expression of the wild-type and YPRF chick 1 integrin subunits. (A-C) Flow cytometry of uninfected primary keratinocytes (A) or keratinocytes expressing the wild-type (B) or YPRF mutant (C) chick 1 subunit. Thin lines, control (second antibody alone); solid lines, anti-chick 1 antibody; y axes, cell number; x axes, fluorescence (log scale, arbitrary units). (D and E) Flow cytometry of primary keratinocytes (D) or SCC4 cells (E). Untransduced, parental cells (black bars) and cells transduced with the wild-type (speckled bars) or YPRF (white bars) chick 1 subunit were compared. The mean fluorescence (arbitrary units) of cells labeled with antibodies to the endogenous human integrins is shown.

Expression of wild-type or mutant chick 1 integrin subunits did not affect cell surface levels of the endogenous 1 and integrins or 64, as evaluated by flow cytometry (Figure 2, D and E). We previously reported immunoprecipitation data showing that the total levels of human/human and human/chick / heterodimers are similar in transduced keratinocytes, although the immature, underglycosylated chick 1 subunit is more abundant than the immature human 1 subunit, reflecting either less efficient maturation or  subunit availability being limiting (Levy et al., 1998).

To determine whether or not the chick 1 subunits localized to focal adhesions, infected keratinocytes and SCC4 cells were fixed and permeabilized and then stained with anti-chick 1 antibodies. Double labeling was performed to compare the distribution of the chick 1 constructs with the endogenous human 1 subunit (Figure 3, A-F) and with vinculin, a marker of focal adhesions (Figure 3, G-L). As predicted from earlier studies (Reszka et al., 1992; Takada et al., 1992), the wild-type chick 1 subunit, the D154A mutant, and the N797I mutant were found in focal adhesions (Figure 1 and Figure 3, A-C and G-I; our unpublished results). Again as expected, the two deletion mutations, the double point mutations in cyto-2 and -3, and the YPRF and YTRF mutants did not accumulate in focal adhesions (Reszka et al., 1992; Lilienbaum et al., 1995; our unpublished results) (Figure 1 and Figure 3, D-F and J-L). The endogenous human 1 subunit localized to focal adhesions in cells expressing each chick 1 construct (see, for example, Figure 3, D-F), and in cells expressing the wild-type chick subunit there was colocalization of human and chick 1 integrins within individual focal adhesions (Figure 3, A-C).

View larger version (75K): [in this window] [in a new window]   Figure 3.   Double-label confocal immunofluorescence microscopy of SCC4 cells (A-F) and primary keratinocytes (G-L) expressing the wild-type chick 1 subunit (A-C and G-I) or the YPRF mutant (D-F and J-L). Cells were stained with anti-chick 1 antibody (A, D, G and J; green in C, F, I and L) in combination with anti-human 1 antibody (B and E; red in C and F) or anti-vinculin (H and K; red in I and L). Images in C, F, I and L are the merged images of A and B, D and E, G and H, J and K, respectively. Bars, 20 µm.

Adhesive Function of Wild-Type and Mutant Chick 1 Subunits

The adhesive activities of the wild-type and mutant chick 1 subunits were determined by assaying the adhesion of transduced primary human keratinocytes and SCC4 cells on type IV collagen-, fibronectin-, and laminin 1-coated substrates in the presence or absence of antibodies specific for human (P5D2) or chicken (W1B10 or JG22) 1 integrins, with the use of chick embryo fibroblasts and noninfected human keratinocytes as controls (Figures 4 and 5). The adhesion of chicken embryo fibroblasts to fibronectin was partially inhibited by the blocking antibody specific for the chick 1 integrin, whereas the antibody to the human 1 integrin had no effect (Figure 4A). The adhesion of noninfected keratinocytes to fibronectin was not inhibited by the anti-chick 1 antibody but was completely inhibited by the anti-human 1 antibody (Figure 4A).

View larger version (24K): [in this window] [in a new window]   Figure 4.   Adhesion assays. CEF, chick embryo fibroblasts; NK, noninfected primary keratinocytes; WT, primary keratinocytes expressing wild-type chick 1 subunit; YPRF, primary keratinocytes expressing YPRF mutant chick 1 subunit. Cells were plated on 10 µg/ml fibronectin (A), 20 µg/ml type IV collagen (B), or 30 µg/ml laminin 1 (C). Data are means of triplicate determinations ± SEM. Adhesion of control cells (i.e., without antibody addition) is expressed as 100%. Black bars, no antibody addition; white bars, 100 µg/ml W1B10 (anti-chick 1 antibody); stippled bars, 100 µg/ml P5D2 (anti-human 1 antibody); gray bars, both antibodies in combination (each at 100 µg/ml).

Adhesion to fibronectin of keratinocytes expressing the wild-type chick 1 subunit was partially inhibited by the addition of either the chick- or the human-specific antibody (Figure 4A). In the presence of both antibodies, adhesion of the infected cells to fibronectin was inhibited completely. These observations demonstrate that both the endogenous human and wild-type chick 1 subunits contributed to the adhesion of infected human keratinocytes to fibronectin.

The adhesion to fibronectin of human keratinocytes expressing the YPRF mutant chick 1 subunit is shown in Figure 4A. In this case, the anti-chick 1 antibody had no effect, and maximal inhibition was achieved with the anti-human 1 antibody alone. The total number of cells adhering to fibronectin was lower in populations expressing YPRF (29% in the experiment shown) than in populations expressing the wild-type chick subunit (63%) or uninfected keratinocytes (50%).

Adhesion of chicken embryo fibroblasts and infected keratinocytes to type IV collagen (Figure 4B) and laminin 1 (Figure 4C) was also measured. The inhibitory effect of the anti-chick 1 antibodies on the chick fibroblasts was greater on laminin than on collagen or fibronectin. There are two probable reasons for this. First, fibronectin and collagen are "better" substrates, because the fibroblasts adhered and spread more rapidly and at lower coating concentrations than on laminin. Second, fibroblasts, unlike keratinocytes, express the additional non-1 integrin v3, which can mediate adhesion to fibronectin and collagen (Gladson and Cheresh, 1994). Adhesion of keratinocytes expressing the wild-type chick 1 subunit was inhibited by the combination of anti-human and anti-chick 1 antibodies more effectively than by either antibody alone (Figure 4, B and C), as observed for adhesion to fibronectin (Figure 4A). In contrast, the anti-chick 1 antibody had no inhibitory effect on cells infected with the YPRF mutant, and adhesion of those cells could be inhibited completely with the anti-human 1 antibody.

The complete series of constructs was screened in normal keratinocytes and SCC4 cells plated on type IV collagen in the presence of anti-human (P5D2) or anti-chick (W1B10 or JG22) 1 antibodies alone or in combination (Figures 1 and 5). As shown in Figure 5, adhesion of cells expressing the wild-type subunit or the N797I mutant was maximally inhibited by the combination of P5D2 and JG22, establishing that both the human and the chick integrins contributed to cell adhesion. Adhesion of cells expressing the D154A extracellular domain mutant or any of the other cytoplasmic domain mutants was completely inhibited with P5D2 alone, showing that the chick subunit did not contribute to adhesion under the assay conditions.

View larger version (44K): [in this window] [in a new window]   Figure 5.   Adhesion of SCC4 cells on 20 µg/ml type IV collagen in the presence of anti-chick 1 (JG22) or anti-human 1 (P5D2) antibodies alone, in combination (BOTH), or in the absence of antibodies (NONE). The number of cells that attached in the absence of antibodies is shown as 100% cell adhesion. Data are means of triplicate determinations ± SEM.

In some assays (Figures 4C and 5), the proportion of YPRF-expressing cells that adhered was increased in the presence of JG22 or W1B10, suggesting that this mutant was not only inactive in promoting cell adhesion but also could act as a weak dominant negative inhibitor of adhesion. The Y788A/N797I and 759-771 mutants had similar properties (Figure 5). None of the other mutants had any effect on the proportion of adherent cells (Figure 5; our unpublished results).

Role of Chick 1 Subunits in Regulating Suspension-induced Terminal Differentiation of Primary Human Keratinocytes

When primary human keratinocytes are disaggregated and placed in suspension in methyl cellulose for 24 h, the number of terminally differentiating keratinocytes increases approximately threefold, as measured by the number of cells expressing the cornified envelope precursor involucrin. Suspension-induced terminal differentiation can be inhibited by fibronectin alone or in combination with laminin 1 and type IV collagen, or by IgG or Fab fragments of adhesion-blocking antibodies to the 1 integrin subunit (Adams and Watt, 1989; Watt et al., 1993). The maximum inhibition is ~50%, because the starting population contains cells that are already committed to undergo terminal differentiation (Hotchin et al., 1993). To examine the role of the wild-type and mutant chick 1 subunits in regulating the onset of terminal differentiation, we tested the ability of anti-human and anti-chick 1 antibodies alone or in combination to inhibit suspension-induced differentiation of infected primary keratinocytes (Figures 1 and 6).

View larger version (29K): [in this window] [in a new window]   Figure 6.   Suspension-induced terminal differentiation of primary keratinocytes. (A) The percentage of involucrin-positive cells was determined before suspension (white bars) or after suspension in the absence of antibody (thin striped bars) or in the presence of W1B10 (anti-chick 1 antibody; black bars), P5D2 (anti-human 1 antibody; thick striped bars), or both antibodies in combination (stippled bars). Uninfected cells (NK) and cells infected with the wild-type (WT) or YPRF mutant chick 1 subunit were treated with IgG or Fab fragments of the antibodies. Results presented are the mean values ± SEM from triplicate fields of stained cells in one experiment and are representative of data from at least four separate experiments involving IgG addition. A single experiment was performed with Fab fragments. (B) The suspension-induced increase in the number of involucrin-positive cells in the absence of antibodies is expressed as 100% terminal differentiation, and the number of involucrin-positive cells in the presence of P5D2 (striped bars) or W1B10 (black bars) is expressed relative to that value. The results shown are the average of triplicate samples from individual experiments. Each construct was tested in at least two separate experiments. The effects of the two antibodies on cells expressing a given construct were significantly different (p < 0.05) for the constructs marked with asterisks.

Expression of the wild-type or mutant chick 1 subunits did not affect the proportion of involucrin-positive keratinocytes in preconfluent, adherent cultures (before suspension), and terminal differentiation in suspension was induced to the same extent in cells expressing each construct (Figure 6A; our unpublished results). Addition of anti-human or anti-chick 1 IgG inhibited terminal differentiation by 30-60% in cells expressing wild-type chick 1; the degree of inhibition was the same when the antibodies were added in combination (Figure 6A). The degree of inhibition observed with the anti-chick 1 antibody was the same in cells expressing the wild-type or YPRF mutant chick 1 subunit and when IgG or Fab fragments of the anti-chick 1 antibody were used (Figure 6A). There was no inhibition of differentiation when uninfected keratinocytes were incubated in suspension with anti-chick 1 antibodies (Figure 6A).

Figure 6B shows the results for the rest of the mutants. Because it was not possible to screen all of the constructs simultaneously in a single experiment, the data are pooled from individual experiments. The mean increase in percentage of involucrin-positive cells after suspension in the absence of antibodies, therefore, is shown as 100% terminal differentiation (Watt et al., 1993), and the effects of the anti-human or anti-chick 1 antibodies are expressed relative to this. Because of the way the data were calculated, only the mean values are shown in Figure 6B; however, a minimum of three suspension assays was carried out for each construct, and the maximum SD between triplicate determinations was 9%. In cells expressing YTRF, N797I, or the double point mutations, the inhibitory effect of the anti-human and anti-chick antibodies was not significantly different. However, in cells expressing D154A or the two deletion mutants, the anti-human 1 antibody inhibited differentiation but the anti-chick 1 antibody did not (p < 0.05 for the difference between percentage of terminal differentiation in the presence of W1B10 or P5D2). The degree of inhibition of terminal differentiation observed for all of the constructs with an inhibitory effect was similar and was the same as the effect of ligating the human integrins; therefore, their activity is summarized in Figure 1 as "++."

Role of Chick 1 Subunits in Regulating SCC4 Differentiation

We have shown previously that introduction of the integrin subunit into a poorly differentiated, -negative SCC line resulted in an increased proportion of cells that expressed involucrin (Jones et al., 1996a). Therefore, we screened a panel of SCC linesSCC4, SCC9, SCC25, SCC12B2, SCC12F2, and SCC27 (Nicholson et al., 1991)for reduced 1 expression with a view to transducing them with the chick 1 integrin constructs. Although the lines all had near-normal 1 integrin levels (Figure 2E; our unpublished results), we nevertheless investigated whether introduction of the wild-type chick 1 subunit had any effect on terminal differentiation. In the least differentiated line, SCC4, but not the other lines, expression of the chick 1 subunit led to an increase in the proportion of involucrin-positive cells (Figure 7). In uninfected postconfluent cultures of SCC4, the proportion of involucrin-positive cells was 1%, compared with ~20% in cultures of primary keratinocytes (Figure 7, A, C, and E; cf. Figure 6A). Introduction of the wild-type chick 1 subunit increased the proportion of involucrin-positive SCC4 cells in postconfluent adherent cultures to ~8% (Figure 7, B, D, and E). As in the case of involucrin-positive cells in cultures of primary keratinocytes, the involucrin-positive SCC4 cells were enlarged and stratified, overlying involucrin-negative cells attached to the substratum (Figure 7B). When SCC4 cells were suspended in methyl cellulose for 24 h, there was no further induction of terminal differentiation (our unpublished results).

View larger version (63K): [in this window] [in a new window]   Figure 7.   Involucrin expression by SCC4 cells. (A-D) Immunofluorescence staining of parental SCC4 cells (A and C) and SCC4 cells transduced with the wild-type chick 1 integrin subunit (B and D). (A and B) Adherent cells; (C and D) single cell suspensions prepared by trypsinization of adherent cultures. Bars, 60 µm (A and B) and 100 µm (C and D). (E) Percentage of involucrin-positive SCC4 cells in adherent postconfluent cultures scored after disaggregation. Data shown are means ± SEM from a minimum of three experiments.

The ability of the chick 1 integrin mutants to stimulate SCC4 differentiation was also examined (Figure 7E). Expression of the single and double point mutations in cyto-2 and -3 resulted in increased involucrin expression, although the constructs containing N797I stimulated differentiation to a lower extent than the YPRF, YTRF, and wild-type chick 1 mutants. To reflect this, the results are summarized as "++" or "+" in Figure 1. The D154A and deletion mutants were inactive, consistent with their lack of activity in suspension-induced terminal differentiation of primary keratinocytes (Figure 6B).

	DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

We have achieved stable, high-level expression of wild-type and mutant chick 1 integrin subunits in primary human epidermal keratinocytes and SCC4 cells through the use of retroviral infection. The chick subunits formed heterodimers with the endogenous human  subunits (Levy et al., 1998), and their ability to target to focal adhesions was as reported previously (Reszka et al., 1992; Takada et al., 1992; Lilienbaum et al., 1995). The wild-type and N797I constructs localized to focal adhesions and contributed to extracellular matrix adhesion, as shown by the observation that anti-chick and anti-human 1 antibodies were required in combination for maximal inhibition of keratinocyte adhesion. The D154A mutant localized to focal adhesions but did not contribute to adhesion, because the combination of anti-chick and anti-human 1 antibodies was no more effective at inhibiting adhesion than anti-human antibodies alone. The other mutants did not localize to focal adhesions or contribute to adhesion.

The ability of the chick 1 constructs to regulate keratinocyte terminal differentiation was measured in two different assays (Figure 1). Mutants that were inactive in regulating the differentiation of primary keratinocytes were also inactive in promoting differentiation of SCC4. Furthermore, constructs with activity in one assay also had activity in the other. In the experiments with primary keratinocytes, the degree of inhibition of suspension-induced differentiation achieved with the anti-chick antibodies (30-60%) was the same for all of the active constructs. However, some of the constructs promoted differentiation of SCC4 more effectively than others. Except in the case of Y788A/N797I, this could not be attributed to differences in the efficiency of expression of the individual constructs; therefore, the explanation may lie with the nature of the SCC4 differentiation defect.

Although we are reasonably confident that in normal keratinocytes ligand binding by 1 integrins serves as a negative regulator of terminal differentiation (Adams and Watt, 1989; Watt et al., 1993; the present report), it is far from obvious why introduction of the chick 1 subunit into SCC4 promoted differentiation. It is well established that in tumor cells that have lost expression of a particular integrin, introduction of the missing receptor can lead to normalization of behavior (see, for example, Giancotti and Ruoslahti, 1990; Zutter et al., 1995; Jones et al., 1996a; reviewed by Sanders et al., 1998). However, there was no difference in surface 1 levels of SCC4 compared with normal keratinocytes (Figure 2E) (Sugiyama et al., 1993; our unpublished results), and introduction of the chick 1 integrin did not affect surface expression of the endogenous integrin subunits (Figure 2E). The endogenous receptor was functional, as evaluated by adhesion assays in the presence or absence of antibodies to the human 1 subunit, and introduction of the wild-type chick 1 integrin did not affect the proportion of adherent cells. There was no further induction of SCC4 differentiation in suspension; however, we did not examine whether anchorage-independent proliferation was inhibited (cf. Jones et al., 1996a). We now need to investigate whether there is a mutation in the endogenous 1 integrin subunit of SCC4 cells or whether there is a downstream signaling defect that is corrected by increased 1 integrin expression.

Comparison of the activity of the wild-type and mutant chick 1 integrin subunits in SCC4 cells and primary keratinocytes allows us to draw some conclusions about the way in which 1 integrins regulate terminal differentiation. Because the D154A mutant was inactive, the differentiation-regulatory role of the 1 integrin subunit must depend on a functional ligand-binding domain. This is intriguing, given that the D154A mutant still bound the anti-chick 1 antibodies (W1B10 and JG22) used to inhibit suspension-induced terminal differentiation of primary keratinocytes, one of which, JG22, recognizes an epitope within the first 160 amino acids of the 1 subunit (Shih et al., 1993). That observation allows us to distinguish between two alternative differentiation signals: "do not differentiate," which would be transduced by ligand-occupied receptors, and "do differentiate," which would be transduced by unoccupied receptors. In the latter case, D154A would be functional in regulating differentiation, but in the former case, it would be inactive. Because antibodies to chick 1 did not inhibit suspension-induced differentiation of D154A-expressing cells, the differentiation signal must be "do not differentiate." Because the D154A mutant localized to focal adhesions, we can also conclude that clustering of 1 integrin cytoplasmic domains in focal adhesions is not sufficient to control differentiation.

The differentiation signal in primary keratinocytes appears to depend on the absolute number of occupied receptors rather than the proportion of occupied receptors. This is because in cells expressing a chick 1 subunit that was competent to regulate differentiation, the degree of inhibition of suspension-induced differentiation was similar whether anti-chick or anti-human 1 antibodies were added alone or in combination. This fits well with the conclusion that exit from the stem cell compartment also depends on the absolute number of occupied receptors (Zhu et al., 1999; see also Dyson and Gurdon, 1998).

The 1 integrin differentiation signal did not require focal adhesion clustering, because single and double point mutants in cyto-2 and cyto-3, the NPXY motifs, were still functional in regulating differentiation. This supports earlier conclusions based on the ability of Fab fragments of anti-1 integrin antibodies to inhibit suspension-induced differentiation (also reported here for anti-chick 1 antibodies; see Figure 6A) and the lack of a requirement for actin polymerization (Watt et al., 1993). Although the ligand-binding site must be intact for differentiation control (as shown by the D154A mutant), there does not appear to be a requirement for high-affinity ligand binding, because the cyto-2 and -3 mutants did not contribute to adhesion to immobilized extracellular matrix proteins (Figures 4 and 5), and it has been demonstrated directly that the YTRF mutant reduces ligand-binding affinity (O'Toole et al., 1995). The failure of the YPRF mutant to contribute to adhesion is in agreement with the observations of Filardo et al. (1995) on the effects of disrupting NPXY in the 3 integrin subunit.

NPXY forms a tight  turn motif that is perturbed by removal or substitution of the proline residue (Collawn et al., 1990; Haas and Plow, 1997). The NPXY motifs in the 1 cytoplasmic domain are involved in linkage to the actin cytoskeleton, e.g., via recruitment of talin (Miller et al., 1987; Tapley et al., 1989; Vignoud et al., 1997), and so our experiments suggest that differentiation regulation is unlikely to require stress fiber assembly. In addition to its importance in cytoskeleton association, NPXY is a phosphotyrosine-binding domain that is found in a number of receptor tyrosine kinases, including the EGF receptor (Van der Geer and Pawson, 1995). Law et al. (1996) have demonstrated that the second NPXY motif in the 3 integrin cytoplasmic domain is phosphorylated after receptor occupancy and, as a result, SH2-containing adaptor proteins can bind. Mutation of cyto-2 and cyto-3 in the 1 subunit, therefore, abrogates association with a variety of signaling molecules that would otherwise have been candidate components of the differentiation-regulatory pathway.

The only cytoplasmic domain mutants that failed to regulate differentiation were the deletion mutants affecting cyto-1 or both cyto-1 and -2. Proteins that are believed to bind to this part of the cytoplasmic domain include paxillin (Schaller et al., 1995), -actinin (Otey et al., 1990), and FAK (Schaller et al., 1995; see also Tahiliani et al., 1997). In addition, part of the cyto-1 motif forms a salt bridge with integrin  subunits (Hughes et al., 1996). In the 3 integrin subunit, the juxta-membrane region of the cytoplasmic domain is a conformational "hot spot," its flexibility and location making it ideal to regulate signaling (Haas and Plow, 1997). More refined mutational analysis is required within the region identified through the deletion mutants to discover events downstream of 1 in the differentiation-regulatory pathway. The contribution of the  integrin subunits to signaling (see, for example, Wary et al., 1996, 1998; Haas and Plow, 1997), the involvement of proteins that associate with the transmembrane or extracellular domains of the integrins (see, for example, Jones et al. 1996b; Wary et al., 1996, 1998; Yauch et al., 1998), and the mechanisms involving modulation of growth factor responsiveness (Renshaw et al., 1997; Wang et al., 1998) must not be ignored. It will also be important to look at MAPK signaling because of its role, in combination with 1 integrins, in differentiation of myoblasts (Sastry et al., 1999), mammary epithelium (Wang et al., 1998), and the epidermal stem to transit-amplifying cell transition (Zhu et al., 1999).

In conclusion, our data suggest that ligand binding to the 1 integrins generates at least two signals in keratinocytes. One signal, in which the NPXY motifs are involved, results in the clustering of receptors into focal adhesions and polymerization of actin filaments, providing a positive stimulus for cell adhesion and spreading. The other signal, in which sequences N terminal to the NPXY motifs play a role, is independent of receptor clustering in focal adhesions and cytoskeletal assembly and is a negative stimulus for differentiation. The challenge now is to identify the pathways required for the control of differentiation in this model.