Cell adhesion to substratum results in the initiation of integrin signaling and an integrin-dependent organization of the cytoskeleton (cell spreading). To address the potential relationships between these events and cell proliferation, we transfected NRK fibroblasts with an antisense cDNA encoding a 1.3 kb ATG-spanning portion of α5 integrin subunit and obtained stable clones in which the surface expression of α5β1 integrin was selectively reduced. α5-antisense NRK cells are less spread than the control transfectants, have poorly defined stress fibers, and an increased amount of cortical actin. The antisense clones remained anchorage-dependent, but they proliferated very slowly. Serum dose-response curves showed that they have an impaired response to mitogens. Importantly, cell spreading and stress fiber formation could be completely restored by plating the antisense cells on collagen, but cell spreading failed to rescue proliferation. These data indicate that cell spreading can be uncoupled from cell proliferation and that cytoskeletal organization is important for NRK cell proliferation because it enforces the proliferative effect of α5β1 integrin. Our results also indicate that reduced surface expression of α5β1 integrin is not sufficient to confer the anchorage-independent phenotype to nontransformed cells.
Integrins are heterodimers consisting of noncovalently associated α and β subunits that mediate cell adhesion to the extracellular matrix (ECM). The combinatorial arrangement of α and β subunits confers ligand specificity to integrin binding. For example, α1β1 and α2β1 are receptors for collagen and laminin, α3β1 has been reported to bind to collagen and fibronectin as well as laminin-5, α5β1 is the classic receptor for fibronectin, and αvβ3 is the classic receptor for vitronectin (Hynes, 1992). Cell adhesion to ECM stimulates integrin- dependent signaling, and these signaling events are important for cell proliferation (reviewed by Giancotti, 1997; Howe et al., 1998). Integrin-mediated signaling often cooperates with the signaling events stimulated by the binding of mitogenic growth factors to receptor tyrosine kinases (RTKs). This cooperation probably explains the fact that most normal cells are both mitogen and anchorage-dependent for growth. Several studies have shown that integrin and RTK signaling are required for cells to progress through the majority of G1phase (reviewed by Assoian, 1997).
Progression through G1phase of the cell cycle is mediated by cyclin-dependent kinases (cdks), cyclin D-cdk4/6 and cyclin E-cdk2 (Sherr, 1994). RTK and integrin signals are jointly required for the activation of both enzymes in several cell types (Böhmer et al., 1996; Zhu et al., 1996; Resnitzky, 1997; Day et al., 1997; Brugarolas et al., 1998). Once activated, cyclin D-cdk4/6 and cyclin E-cdk2 phosphorylate the retinoblastoma protein (pRb) and pRb family members, allowing for the dissociation of associated E2Fs and induction of E2F-regulated genes among which is cyclin A (Zwicker et al., 1995; Schulze et al., 1995, 1996; Huet et al., 1996). Cyclin A binds to cdk2, and active cyclin A-cdk2 complexes are required for progression into and through S phase (Girard et al., 1991).
Cell adhesion may also regulate cyclin A expression through pRb-independent mechanisms. For example, NRK fibroblasts have lost their adhesion requirement for phosphorylation of pRb, but cyclin A expression remains anchorage-dependent (Guadagno et al., 1993; Zhu et al., 1996). Studies by others (Chen et al., 1996) also support the existence of pRb- independent controls on cyclin A expression, but these E2F- independent controls remain poorly understood. Overall, these and other studies (reviewed by Assoian, 1997) show that cells are mitogen- and anchorage-dependent for growth because activation of the G1phase cyclin-cdk network is jointly regulated by growth factors and the ECM.
Cell adhesion to the ECM has two conceptually distinct effects on cells. First, adhesion results in the clustering of integrins, and the clustering process is thought to be important in initiating integrin signals (Clark and Brugge, 1995). Second, adhesion results in an integrin-dependent organization of the cytoskeleton and cell spreading, and it is clear that the consequent spread cell shape is important for proliferation (Folkman and Moscona, 1978; Chen et al., 1997). Cytoskeletal organization imposes mechanical forces on cells and these forces may act as effectors of the adhesion ‘signal’ (reviewed by Chicurel et al., 1998). In principle these mechanical forces could act largely to maintain integrin clusters (hence integrin signaling) or could have a more complex effect, e.g. allowing for the generation of tensile forces that affect nuclear pore size (Feldherr and Atkin, 1993).
We and others have tried to resolve the relative contributions of direct cell adhesion and adhesion-dependent cell spreading on ECM-dependent proliferation (anchorage-dependent growth). These studies have indicated that certain G1phase events can occur in response to direct adhesion while others require a spread cell shape. For example, Hansen et al. (1994)reported that direct cell adhesion is sufficient to mediate progression through early G1phase while cell spreading was required in late G1phase in hepatocytes. We have reported that cytoskeletal organization is required for mitogen-dependent induction of cyclin D1 and phosphorylation of pRb in normal human fibroblasts (Böhmer et al., 1996). A recent study with capillary endothelial cells reports that cell adhesion per se is sufficient for activation of MAP kinase and induction of cyclin D1 mRNA while adhesion-dependent increases in cytoskeletal tension (the spread cell shape) are required for induction of cyclin D1 protein, phosphorylation of pRb and downregulation of p27kip1(Huang et al., 1998).
In this study we have examined the role of α5β1 integrin in mediating the adhesion and shape requirements for cell proliferation. Our approach was to selectively inhibit expression of α5β1 integrin in NRK fibroblasts by stable antisense transfection and assess the consequence of reduced α5β1 surface expression on cytoskeletal organization and cell proliferation. Not surprisingly, we found that cell spreading and proliferation are inhibited in α5-deficient NRK cells. However, the proliferation defect persists even when cell spreading is restored by adhesion of α5-antisense cells to collagen. The phenotype of these α5-deficient cells leads us to conclude that cell spreading can be uncoupled from cell proliferation. The extrapolation of these results is that mechanical forces generated by organization of the cytoskeletal act, at least in part, by enforcing the signaling potential of integrins that control proliferation.
MATERIALS AND METHODS
Preparation of α5-antisense and control clones
The full-length mouse integrin α5-cDNA (generously provided by Clayton Buck, Wistar Institute) was digested with SmaI and XhoI. The resulting ATG-containing upstream 1.3 kb fragment was cloned into the SmaI and XhoI sites of pcDNAII which served as a shuttle vector. The 1.3 kb fragment was recovered from pCDNAII by digestion with SpeI and XbaI, and this fragment was cloned into XbaI-linearized tet- operator expression vector. Restriction digestion was used to identify the orientation of the clones that were obtained. Identity of the final sense and antisense clones was confirmed by DNA sequencing. The α5-antisense and sense (control) vectors were transfected into NRK cells with lipofectamine. Hygromycin and G-418 resistant colonies were pooled, and clones were isolated by limiting dilution. Antisense and control NRK transfectants were maintained in 5% newborn calf serum in Dulbecco’s modified Eagle’s medium (DME) with 0.5 mg/ml geneticin and 0.25 mg/ml hygromycin. Note that the 1.3 kb murine cDNA hybridized to rat α5 mRNA in high stringency northern blots. Note also that the 1.3 kb fragment of α5 cDNA lacks the cytosolic, the transmembrane, and a large part of the extracellular domains, so NRK cells transfected with the 1.3 kb α5 cDNA in the sense orientation (control transfectants) should not express increased amounts of α5β1 integrin on the cell surface.
Analysis of integrin surface expression
Cell surface expression of integrins was assessed by surface biotinylation and radioiodination. Cells (1-2×106in 10 ml 5% FCS- DME) were seeded in 100-mm dishes and incubated overnight prior to labeling. For biotinylation, monolayer cultures were washed 3 times with 4-5 ml of ice-cold 1× bicarbonate buffer (40 mM sodium bicarbonate, pH 8.6; Amersham). After the last wash, 1 ml of labeling solution (0.5 mg of sulfo-NHS-biotin, Pierce Chemical, per 1 ml of cold 1× bicarbonate buffer) was incubated with the cells (15 minutes at 4°C). The labeled cells were rinsed 3 times with cold PBS prior to lysis in 0.2 ml of 100 mM Tris-HCl, pH 8.5, 150 mM NaCl, 0.5 mM MgCl2, 0.5% Nonidet P-40, 10 μg/ml phenylmethylsulfonyl fluoride (PMSF), 10 μg/ml leupeptin, 10 μg/ml aprotinin). A constant amount of protein (50 μg as determined by the Bio-Rad Protein Assay) was used for immunoprecipitations with anti-α5 antibody as described (Dalton et al., 1992). The immunoprecipitates were fractionated on non-reducing SDS-polyacrylamide gels (5% acrylamide), and electroeluted onto nitrocellulose. The filter was blocked for 1 hour with 5% non-fat milk in washing buffer (20 mM Tris-HCl, pH 7.5, 50 mM NaCl, 2.5 mM EDTA, 0.1% Tween-20) and then washed for at least 45 minutes in washing buffer. The washed filter was incubated for 1 hour with horseradish peroxidase-conjugated strepavidin (dilution of 1:1500, Amersham) solution. Biotinylated proteins were visualized by enhanced chemiluminescence (Amersham). Surface radioiodination in monolayer and suspension was performed as described (Dalton et al., 1992). The labeled cells were extracted in the lysis buffer described above, and aliquots of the lysates were used to determine the amount of trichloroacetic acid (TCA)-precipitable radioactivity in each sample. Aliquots of cell extracts (between 2.5×105and 2×106cpm of the TCA-insoluble radioactivity, for exact amounts see figure legends) were used for immunoprecipitation using 2-3 μl of polyclonal antibodies directed against the α3, α5, and β1 subunits, or αvβ3 integrin (Life Technologies). The samples were analyzed by non-reducing SDS-gels (5% acrylamide) and the immunoprecipitated integrin subunits were visualized by autoradiography.
Measurement of cell adhesion
To examine rates of adhesion, cells were gently detached with trypsin- EDTA, incubated with soybean trypsin inhibitor (0.5 mg/ml for 5 minutes), suspended in DME, 1 mg/ml BSA, and then seeded (1×105cells in 2 ml) in 35-mm dishes that had been coated with 20 μg type I collagen or 10-20 μg fibronectin and blocked with heat-inactivated BSA (2 mg/ml in PBS) using standard procedures. At selected times, cells were gently washed 3 times in PBS to remove unattached cells. The adherent cells were stained with 1 ml Crystal Violet solution (0.5% Crystal Violet in 20% methanol) and washed 5 times with water prior to elution of the dye and determination of absorbence at 540 nm. Background absorbence was subtracted, and the results were plotted as the percentage maximal cell attachment. To block secretion of endogenous matrix proteins, the cells were pretreated for 1-2 hours with cycloheximide (10 μg/ml) prior to trypsinization and maintained in cycloheximide throughout all subsequent incubations. In some experiments, cycloheximide-treated α5-antisense and α5-control transfectants were detached with trypsin-EDTA, and the trypsinized cells (104in 0.5 ml DME, 18 mM Hepes, pH 7.4, 1 mg/ml BSA) were pre-incubated (30 minutes at 4°C with rocking) in suspension with 100 μg of blocking antibodies specific for rat α1β1 (Pharmingen, CD49a), rat α2β1 (Pharmingen CD49b) or human α5β1 (Life Technologies, P1D6; negative control).
Measurement of cell proliferation
To examine proliferation rates, asynchronous cells were detached from culture dishes with trypsin-EDTA, reseeded at subconfluence (2.5×104cells in 2 ml per 35-mm well) in uncoated or collagen- coated tissue cultures dishes, incubated for 0-3 days with 10% FCS- DME, and stained as described above. Results were plotted as fold increase in cell number. To examine the activation of ERK or expression of G1phase cyclins and cdks, asynchronous cells or cells that had been synchronized into G0(by 3-4 day incubation of confluent cells in serum-free DME) were detached with trypsin- EDTA, reseeded in 150-mm dishes, and serum-stimulated as described in the individual figure legends. At selected times, cells were washed in PBS, collected and lysed as described (Zhu et al., 1996). Equal amounts of total cell protein from each extract (determined by Coomassie binding; Bio-Rad Protein Assay) were analyzed by western blotting using antibodies against dually phosphorylated ERK, ERK, pRb, cyclin D, cyclin E, cyclin A, or cdk4. Enhanced chemiluminescence was used to detect the immunoreactive proteins. To assess anchorage-independent growth, cells (1.25-2.5×104/ml) were cultured in DME with 10% FCS or 10% FCS, 2 nM EGF on agarose-coated dishes. Cell cycling was assessed by incubating the cultures with [3H]thymidine (2 μCi) for 24 hours prior to collection of cells and isolation of TCA-insoluble DNA as described (Han et al., 1993).
Analysis of fibronectin matrix formation
Conversion of fibronectin into a fibrillar matrix was determined by assessing the percentage of fibronectin present in deoxycholate (DOC)-soluble and DOC-insoluble fractions of cell extracts (McDonald, 1988; Schwarzbauer, 1991). α5-antisense and α5-sense transfectants were detached with trypsin-EDTA, seeded (1.5×106cells in 10 ml 10% FCS-DME) in uncoated or collagen-coated 100-mm dishes, and incubated for 24 hours. The medium was aspirated and the cells were collected by scraping into 100 μl of DOC-containing buffer (50 mM Tris-HCl, pH 8.8, 2% sodium DOC, 2 mM EDTA, 2 mM PMSF). The extracts were vortexed at full speed for 1 minute and collected by centrifugation (5200 g, 20 minutes, 4°C). The DOC- soluble supernatant was collected and the DOC-insoluble pellet was washed twice with 100 μl of DOC-containing buffer to remove residual soluble fibronectin. The washed DOC-insoluble pellet was resuspended in 100 μl of fresh DOC-containing buffer, and then both the DOC-soluble and DOC-insoluble fractions were solubilized by addition of 100 μl of 2× SDS-sample buffer containing reductant. The samples were heated (95°C, 5 minutes), fractionated on SDS- polyacrylamide gels (6% acrylamide with 37.5:1 acrylamide/bis acrylamide), and electrophoretically transferred to nitrocellulose. After blocking in 5% non-fat milk, filters were immunoblotted using a 10,000-fold dilution of mouse monoclonal antibody IC3 to fibronectin (generously provided by Jean Schwarzbauer, Princeton University).
Actin staining with fluorescein-phalloidin
Autoclaved coverslips were added to 35-mm cell culture dishes, either untreated or precoated with collagen or fibronectin. α5-antisense and α5-control NRK transfectants (5×104) were added to the wells and incubated in 10% FCS-DME. After 24 hours, the cells were fixed with formaldehyde (3.7% in PBS with Ca2+/Mg2+; 20-30 minutes), washed twice with PBS, and once with 50 mM NH4Cl in PBS before being permeabilized in 0.1% Triton X-100 in PBS. An aliquot of fluorescein-phalloidin (Molecular Probes, F-432) in methanol (2 units, Molecular Probes F-432) was dried under nitrogen and dissolved in 1 ml of PBS. Coverslips containing the permeabilized cells were placed on 40 μl droplets of fluorescein-phalloidin solution and incubated in the dark for 20-30 minutes. After staining, the coverslips were returned to the wells and washed 4 times with PBS and once with water, in the dark. Slides were mounted with antifade in glycerol (Molecular Probes), and fluorescent actin was visualized by epi-fluorescent microscopy at ×40 magnification.
Selective inhibition of α5β1 surface expression by antisense transfection of NRK fibroblasts
In an effort to inhibit the production and surface expression of the endogenous α5 subunit, we transfected NRK cells with expression vectors containing a 1.3 kb ATG-spanning fragment of α5 integrin cDNA in the antisense and sense (control) orientations. Selected clones were surface biotinylated and the level of surface α5β1 integrin was determined by immunoprecipitation with anti-α5 followed by SDS-gel electrophoresis and western blotting with HRP-streptavidin. The results from this survey (not shown) allowed us to identify two antisense clones (AS-1 and AS-2) showing strongly reduced surface expression of α5β1 integrin. As expected (see Materials and Methods), all clones transfected with the sense construct showed normal α5β1 integrin levels. Although we cloned the 1.3 kb α5 cDNA in a tetracycline-repressable vector, the antisense clones we obtained showed constitutive rather than tetracycline-regulated downregulation of surface α5β1. Nonetheless, it is unlikely that the antisense transfectants are merely poor α5β1 expressing subpopulations in the parental cell line because (i) the selected clones propagate in the appropriate selection medium and (ii) a relatively high percentage of the antisense clones (2 of 8 tested) studied showed low α5β1 surface expression while none of the control transfectants had this phenotype.
The selected α5-antisense and control transfectants were cell surface radioiodinated, and equal amounts of TCA-precipitable radioactivity were incubated with anti-α5, anti-α3, anti- vitronectin receptor (primarily anti-αvβ3), or anti-β1. The immunoprecipitated samples were analyzed by SDS-PAGE and audioradiography. Surface expression of the α5 subunit was inhibited in the α5-antisense transfectants while expression of the other α subunits (α3 and αv) was similar (Fig. 1A). Thus α5β1 integrin surface levels were selectively decreased in α5-antisense transfectants. Surface expression levels of the β1 subunit was only minimally reduced in the antisense vs control transfectants (Fig. 1B), indicating that α5β1 integrin is not a major constituent in the total β1-pool and that its depletion has only a small effect on the total integrin distribution in the cell. Note that the immunoprecipitation with anti-β1 also revealed a very faint high molecular weight band migrating at a position characteristic of the α1 subunit (Fig. 1B; arrowhead). Thus, surface expression α1β1 integrin is very low in NRK cells. We were unable to identify antibodies suitable for direct immunoprecipitation of surface-labeled rat α2β1, but studies with function blocking antibodies (see below) showed that this integrin is present and mediates adhesion to collagen in both α5-antisense and α5-control transfectants.
Different amounts of TCA-precipitable radioactivity from extracts of radioiodinated α5-control transfectants were then immunoprecipitated with anti-α5, and the resulting α5 signals were matched with an anti-α5 immunoprecipitate using a fixed amount of TCA-precipitable radioactivity from the radioiodinated α5-antisense cell extract. The factor of the difference in the amount of radioactivity between samples with comparable signals is equivalent to the fold decrease in the expression levels of α5β1 integrin in the antisense cells. This analysis indicated that the level of α5β1 integrin cell-surface expression was about 4-fold lower in the α5-antisense NRK cells as compared to the control transfectants (Fig. 1C).
Cycloheximide-treated antisense and control (sense) transfectants were used to compare the rates at which the cells attached to collagen and fibronectin. The rates of binding to collagen were similar, but the rate of binding of antisense transfectants to fibronectin lagged behind the adhesion rates of the control (sense) transfectants, especially in the early time points (Fig. 2A). We also tried plating antisense and control transfectants on vitronectin because the results in Fig. 1Aindicate that the vitronectin receptor (αvβ3) was expressed at similar levels in these cells. However, both the control and antisense transfectants spread poorly on vitronectin (not shown), consistent with the relatively low expression level of this integrin on NRK cells (refer to the amounts of radioactivity immunoprecipitated for α5 vs αv in Fig. 1A).
Blocking antibodies showed that adhesion of both α5- antisense and α5-control transfectants to collagen was strongly inhibited by anti-α2β1 integrin (Fig. 2B). We occasionally observed a small inhibition of adhesion by the α1β1 antibody (e.g. refer to the α5-antisense cells; Fig. 2B), but this effect was not consistent and always very small relative to the effect seen with anti-α2β1. The results of these binding inhibition studies agree well with those from Fig. 1which indicate that
NRK cells express minimal amounts of α1β1, the other principal collagen receptor. The near complete inhibition of binding mediated by anti-α2β1 also argues against a role for α3β1 an as important collagen receptor in NRK cells.
Reduced cell proliferation in α5-antisense NRK cells
There was a notable difference between the proliferation rates of α5-antisense and control transfectants (Fig. 3A). Conversion of these data to number of cell doublings indicate that the doubling time of parental NRK cells and the control transfectants was about twice as great as the doubling time of the antisense transfectants. Similar results were obtained when the cells were synchronized into G0prior to stimulation with serum (not shown). Dose response studies showed that the antisense transfectants responded poorly to increasing concentrations of serum as compared to either parental NRK cells or control (sense) transfectants (Fig. 3B). Thus, the loss of α5β1 integrin rendered the cells less sensitive to mitogenic stimulus. Nevertheless, the antisense cells do eventually reach confluence, and their cell density at confluence is similar to that of control (sense) transfectants (data not shown). When we examined the expression of G1phase cyclin-cdks in this system, we found that cyclin A expression, but not cyclin D1 or cyclin E expression, was reduced in the antisense transfectants (Fig. 3C). The mechanistic basis for this effect could not be resolved because cell cycle synchrony is lost in long-term cell cultures and the proliferative defect in the antisense transfectants is not readily detectable in the first cell cycle (refer to Fig. 3A).
Given its reduced sensitivity to mitogens, we considered the possibility that activation of the ERKs was reduced in the α5- antisense transfectants, and that this effect was responsible for the poor rate of proliferation. However, we found that ERK phosphorylation was strongly stimulated by serum in both α5- antisense and α5-control transfectants (Fig. 4). Thus, the reduction in surface α5β1 integrin expression did not prevent the serum-dependent activation of ERKs. This result is consistent with those in Fig. 3C showing that cyclin D1 expression (an established target of active ERK) is similarly expressed in the α5-antisense and α5-control transfectants.
Cell spreading is not sufficient to rescue growth of α5-antisense NRK cells
The α5-antisense transfectants not only had decreased proliferation rates, but also a different morphological appearance. In particular, the α5-control transfectants had the usual elongated and spread cell shape, while the α5-antisense transfectants had a rounder shape and reduced appearance of stress fibers (Fig. 5A, untreated). Since several studies have demonstrated a direct relationship between cell spreading and cell proliferation, we asked whether cell spreading would rescue the proliferation rate of the α5-antisense transfectants. α5-antisense and α5-control transfectants were plated on increasing concentrations of collagen and incubated for 3 days with serum. The α5-antisense transfectants plated on collagen developed well-defined stress fibers and appeared similarly spread to the control transfectants (Fig. 5A). However, cell spreading on collagen was completely without effect on the proliferation of α5-antisense transfectants (Fig. 5B). Thus, cell spreading did not support proliferation when α5β1 levels were depressed.
Studies by Sechler and Schwarzbauer (1998)and Bourdoulous et al. (1998)indicate that disruption of the normal fibrillar fibronectin matrix can inhibit cell proliferation, and formation of a fibrillar fibronectin matrix is thought to be mediated largely by α5β1 integrin (McDonald, 1988; Schwarzbauer, 1991). To determine if the reduced rate of proliferation seen in the serum-stimulated α5-antisense transfectants reflected impaired fibronectin matrix formation, we cultured α5-antisense and control (sense) transfectants in uncoated or collagen-coated dishes, prepared cell lysates, and assessed the percentage of fibronectin present in deoxycholate- soluble (soluble fibronectin) versus deoxycholate-insoluble (fibrillar fibronectin) fractions (McDonald, 1988; Schwarzbauer, 1991). Fig. 5Cshows that the extent of fibronectin matrix formation was similar in both the α5- antisense and α5-control (sense) transfectants, whether or not they were fully spread on collagen. These results indicate that differences in fibronectin matrix formation can not explain the reduced proliferation rate seen in α5-antisense transfectants. Since our α5-antisense transfectants are 4-fold reduced, rather than null, for α5β1 integrin surface expression, our results suggest that the threshold amount of α5β1 required for efficient proliferation is greater than that required for efficient matrix formation (refer to the Discussion).
As α5-antisense and control (sense) transfectants were maintained in culture over a 2-3 month period, the antisense cells became progressively more spread and their proliferation rates approached those characteristic of α5-control transfectants (Fig. 6) and parental NRK cells (not shown). When we analyzed α5β1 surface expression in these revertants, we found that surface expression of α5β1 integrin was now similar to that of control (sense) transfectants (Fig. 6, inset). The most straightforward interpretation of these results is that the α5-antisense cells gradually revert and that the increased rate of proliferation in the revertants allow them to overtake the culture, resulting in the eventual loss of the α5-antisense phenotype.
Reduced expression of α5β1 integrin fails to induce anchorage-independent growth
Early studies showed that transformed cells often show reduced expression of α5β1 integrin and that enforced overexpression of this integrin reverts the transformed phenotype (reviewed by Hynes, 1990). To determine if reduced expression of α5β1 integrin allowed for induction of anchorage-independent growth in nontransformed cells, we compared G1phase cell cycle progression when G0-synchronized α5-antisensese transfectants were serum-stimulated in monolayer and suspension. The expression of cyclin D1 was not affected by the presence or absence of substratum, the phosphorylation of pRb was slightly delayed, and the induction of cyclin A was significantly reduced in the suspended cells (Fig. 7). Although the expression of cyclin D1 and phosphorylation of pRb are strictly anchorage-dependent in most fibroblastic cells, only the expression of cyclin A is strongly anchorage-dependent in NRK cells (refer to the Introduction). Thus, the α5- antisense cells have retained the anchorage-dependent phenotype of parental NRK cells. α5-antisense and control (sense) transfectants, as well as parental NRK cells, failed to undergo anchorage-independent growth as assessed by incorporation of 3H-thymdine into newly synthesized DNA of suspension cultures (Fig. 8A). Moreover, even the slow rate of proliferation observed with suspended control transfectants was not detectable in the suspended antisense cells (Fig. 8B; note the different scales used for the ordinates of A and B). The latter result indicates that a reduced rate of proliferation (as observed with the antisense cells in monolayer) can not account for the inability to grow when α5- deficient cells are cultured in suspension. Similarly, control transfectants poorly formed colonies in soft agar, and the antisense transfectants were completely anchorage-dependent in this assay (not shown).
Cell spreading in the presence of reduced α5β1 integrin fails to support proliferation
Several studies indicate that a spread cell shape is necessary for efficient cell cycle progression. In agreement with these studies we find that antisense-mediated reduction in α5β1 surface expression leads to inefficient cell spreading and reduced proliferation. In our system, fibronectin (the ligand for α5β1) is being supplied by the serum, and it is also secreted by NRK cells (refer to Dalton et al., 1992). We propose that the reduction in cell surface α5β1 attenuates the effect of fibronectin on both cell spreading and proliferation. All of our results are consistent the long-standing observations that cell spreading and cell proliferation are closely related.
However, our data also indicate that cell spreading can be uncoupled from cell proliferation. In particular, we were able to rescue spreading completely by culturing the α5-antisense transfectants on collagen, and these fully spread cells still proliferated poorly. Although these results do not exclude a role for cell spreading during cell proliferation, they do demonstrate that cell spreading is not sufficient to induce cell proliferation in mitogen-treated cells. They also indicate that an essential role of NRK cell spreading involves enforcing the proliferative potential of α5β1 integrin. Rescue of cell proliferation in the α5β1-revertants provides genetic evidence supporting the role of α5β1 in proliferation. All of our observations have been reproduced with both of the α5- antisense lines we obtained. We note that α5-antisense cells show a persistent reduction in cell proliferation while their reduced adhesion to fibronectin is not persistent. We attribute this difference to the fact that the adhesion assay is performed with purified, excess fibronectin.
Why might normal surface expression of α5β1 integrin be so critical for NRK cell proliferation? Wary et al. (1996)have proposed that a subset of integrins (α1β1, α5β1 and αvβ3) support the activation of ERKs. Sastry et al. (1999)similarly attributed cell proliferation to α5β1 integrin and ERK activation. Because α1β1 and αvβ3 are expressed at low levels in NRK cells (see above), we initially thought that reduced surface expression of α5β1 might impair ERK activation. However, we found that serum-dependent ERK activation and cyclin D1 expression (an established target of ERK activation) was similar in the α5-antisense and α5-control transfectants. Others have reported that ligation of cells to fibronectin has the potential to enhance ERK-independent signaling by growth factor receptors (McNamee et al., 1993; Miyamoto et al., 1995). An attenuation of this effect in α5-antisense cells would be consistent with the reduced serum sensitivity observed in the antisense transfectants. An α5β1-dependent regulation of paxillin (Sastry et al., 1999) might also underlie the phenotype reported here.
We tried to recapitulate the growth phenotype of α5- antisense NRK cells in NIH-3T3 cells and normal fibroblasts, but have been unable to obtain stable clones in which expression of surface α5β1 is reduced. Nevertheless, the results of Wary et al. (1996)indicate that the conclusions we have reached with α5-antisense NRK cells will be generally applicable. Specifically, Wary et al. showed that human vein endothelial cells attached to laminin (via α2β1) fail to progress through G1phase when treated with purified FGF and insulin while the same cells attached to fibronectin (through α5β1) do. In our studies, the adhesion of α5- antisense NRK cells to collagen is mediated by α2β1 integrin, suggesting that the proliferative defect seen in α5-antisense transfectants plated on collagen reflects, at least in part, the poor proliferative potential of α2β1 vs α5β1 integrin. We realize that the matrix contributed by the serum (vitronectin and fibronectin in particular) and released by the cell can influence the composition of the local ECM in our experiments. However, NRK cells do not attach and spread well on purified vitronectin, and immunoprecipitation of surface-labeled cells indicated that αvβ3 levels are very low in NRK cells (not shown) and the transfectants (refer to Fig. 1). α1β1 integrin and α3β1 are also present at low levels (5- to 10-fold less than α5β1 as determined immunologically; refer to Fig. 1) and are not involved in mediating adhesion of the cells to collagen (refer to Fig. 2B). Although we have not formally excluded a role for α3β1 in binding to fibronectin, it is not a preferred fibronectin receptor in fibroblasts, it does not participate in binding of NRK cells to collagen, and we see proliferative differences between α5-antisense and α5- control transfectants despite constant expression of α3β1. For these reasons, we believe that the difference in proliferation rates of the α5-antisense and control transfectants plated on collagen speaks to the proliferative potential of α5β1 integrin.
While our results and those of Wary et al. (1996)are complementary, they are also distinguishable in several important ways. First, Wary et al. restricted their analysis to progression through first G1phase. Since our approach of reducing receptor expression circumvents the complications that arise from secretion of matrix proteins by cells, we were able to document long-term differences in the proliferation of fully spread cells expressing normal and reduced amounts of surface α5β1 integrin. Second, the experimental approach used by Wary et al. obviously necessitates that cells be spread on different matrixes (laminin vs collagen). Our approach allowed us to document α5β1-dependent proliferative effects even though the α5-antisense and α5-control transfectants were both attached to collagen. Finally, Wary et al. attributed the different proliferative effects of α2β1 and α5β1 integrin to their relative effects on ERK activation. Since serum-induced ERK activation is not impaired in the α5-antisense NRK cells, our results indicate that the growth promoting effect of α5β1 integrin extends beyond ERK activation. Although the exact growth regulatory events that are being affected in the α5- antisense NRK cells remain to be determined, our data showing normal fibronectin matrix formation and ERK activation suggest that a discrete subset of α5β1-mediated events is being targeted. A prediction of our data is that these events will be more sensitive to reductions in surface α5β1 than are either ERK activation or fibronectin matrix formation.
Reduced surface expression of α5β1 integrin fails to induce anchorage-independent growth
Several early studies showed that many transformed cells secrete reduced amounts of fibronectin and have reduced surface levels of α5β1 integrin (reviewed by Hynes, 1990). Moreover, anchorage-independent cells have been rendered anchorage-dependent by overexpression of α5β1 integrin (Giancotti and Ruoslahti, 1990). These data clearly indicate that overexpression of α5β1 integrin can confer an anchorage- dependent phenotype to transformed cells. However, the results presented here indicate that the converse is not true: anchorage- dependent cells can not be rendered anchorage-independent by reducing surface expression of α5β1 integrin. The fact that reduced expression of α5β1 fails to induce anchorage- independent growth of NRK cells (which are already partially transformed as evidenced by their anchorage-independent expression of cyclin D1) makes it even more unlikely that loss of α5β1 would be sufficient to induce anchorage-independent growth in cells that retain the full spectrum of adhesion- dependent cell cycle controls.
These results extend previous studies by us and others (Dalton et al., 1992, 1995; Hotchin et al., 1995) showing that β1-integrins are internalized and degraded when fibroblasts or keratinocytes are incubated in suspension (conditions used to study anchorage-independent growth). Together, the data argue that the reduced expression of α5β1 in transformed fibroblasts is not causal for their anchorage-independent phenotype. Rather, we suggest that the reduced α5β1 integrin expression levels characteristic of transformed cells may reflect the fact that oncogenic or spontaneous activation of integrin-dependent signaling pathways eliminates the selective pressure that keeps integrins expressed on the surface of proliferating cells. Finally, the fact that overexpression of α5β1 can restore the anchorage-dependent phenotype to transformed cells (Giancotti and Ruoslahti, 1990) indicates that induction and reversion of anchorage-independent growth are mechanistically distinct.
We thank Clayton Buck (Wistar Institute) for the murine α5 cDNA, Gene Marcantonio (Columbia University) for the α3 and β1 antisera, Jean Schwarzbauer (Princeton University) for the anti-fibronectin antibody, and Pedro Salas and Dora Vega-Salas for assistance with the epifluorescent microscopy. These studies were supported by NIH grants GM48224 and GM51878.