α₃β₁ Integrin is moved into focal contacts in kidney mesangial cells

The regulated interaction of cells with the extracellular matrix (ECM) is critical during embryonal developmental and aberrant interactions that can accompany pathological changes. Matrix components in combination with soluble growth factors can influence cell attachment, motion, differentiation and proliferation. In the kidney, for example, during diseases like glomerulosclerosis, inflammatory processes accompany alteration of matrix and proliferation of the smooth muscle-like cells of the kidney glomerular mesangium. Glomerular cells such as the mesangial cells, epithelial podocytes and their matrices are thought to influence fluid flow through the glomerulus (Morel-Maroger-Striker et al., 1984; Michael and Kim, 1988; Sterzel and Lovett, 1988; Mene et al., 1989). However, little is known about how these cells interact with matrix molecules. We have, therefore, been studying mesangial cell (MC) interactions.

Integrins are heterodimeric transmembrane proteins implicated in interactions with the ECM (Albelda, 1991; Hynes, 1992). The β₁-family of integrins have defined specificity for their ligands, excepting α₃β₁, where the specificity is more controversial. For the β₁-series integrins, α₁ binds collagens (Turner et al., 1987; Turner et al., 1989) and laminin derived from the murine Engelbreth-Holm-Swarm tumour (EHS-laminin) (Hall et al., 1990; Goodman et al., 1991). The α₂ integrin is a collagen and laminin receptor (Languino et al., 1989; Staatz et al., 1991; Chan et al., 1992); α₄ (Elices et al., 1990) and α₅ (Pytela et al., 1985) bind fibronectin; α₆ and α₇ bind laminin (Sonnenberg et al., 1990; von der Mark et al., 1991). In situ, mesangial cells express integrins, in particular α₁β₁ (Cosio et al., 1990; Korhonen et al., 1990; Simon and McDonald, 1990), and α₅β₁ (Cosio et al., 1990). Organization into focal contacts and the matrix partners for these integrins have not been widely studied, since MCs are available only as slowgrowing secondary mortal strains from normal human kidneys, which restricts classical biochemical and cell biological investigations.

The functional specificity of the integrins has been established by ligand-affinity chromatography, by blocking of cell attachment with anti-integrin antibodies or by investigating the integrin distribution in focal contacts. There is often good correlation between the integrin specificities obtained by these techniques, as illustrated for α₁β₁ (Goodman et al., 1991), α₂β₁ (Elices and Hemler, 1989), α₅β₁ (Pytela et al., 1985) and α₆β₁ (Sonnenberg et al., 1991). In addition, mammalian β₁-series integrins require specific divalent cations for activity. All can be activated by Mg²⁺ and Mn²⁺ (Hynes, 1992), while α₅β₁ (Gailit and Ruoslahti, 1988) and α₃β₁ (on collagen; Elices et al., 1991) but not α₆β₁ can also be activated by Ca²⁺ (Sonnenberg et al., 1988). The binding specificity of the α₃β₁ integrin is controversial (Elices et al., 1991; Gehlsen et al., 1992) and it behaves in a manner unusual for a molecule implicated in cell-substrate interactions. Its ever expanding range of ligands includes collagens, laminin and fibronectin (Wayner and Carter, 1987; Wayner et al., 1988), epiligrin (Carter et al., 1991), entactin (Dedhar et al., 1992) and recently the α₂β₁ integrin (Symington et al., 1993). The mercurial phenotype of α₃β₁ was pinned down by the discovery that it had adhesive function only in the absence of other functioning integrins (Tomaselli et al., 1990; Elices et al., 1991; Sonnenberg et al., 1991). In affinity chromatography, it bound on pepsin-digested human placental laminin (Gehlsen et al., 1988) but not on EHS-laminin, which binds α₇β₁, α₆β₁ and α₁β₁ integrins (Kramer et al., 1990; Goodman et al., 1991; von der Mark et al., 1991; Sonnenberg et al., 1991).

In cultured cells, integrins involved in adhesion are found at focal contacts, adhesive organelles involved in signal transduction (Burridge et al., 1992; Juliano and Haskill, 1993) where the cell membrane comes to within 15-20 nm of the substrate (Burridge et al., 1988). By contrast, α₃β₁ has a ‘diffuse’ intracellular distribution (Carter et al., 1990; Elices et al., 1991) or is at intracellular contacts (Kaufmann et al., 1989; Marchisio et al., 1991). It has been seen in focal contacts (FCts) only weakly at the periphery of keratinocytes plated onto laminin (Carter et al., 1990). The apposition of plasma membrane and substrate at focal contacts (FCts) as revealed by interference reflection microscopy coincides with the distribution of the cytoskeletal protein talin (Burridge and Connell, 1983; Burridge et al., 1988; Izzard, 1988). Talin probably participates in seeding cytoskeletal organization at the FCts (DePasquale and Izzard, 1987; Izzard, 1988; Burridge et al., 1988; Burridge and Connell, 1983; Beckerle and Raymond, 1990) where it concentrates together with paxillin (Turner et al., 1990), vinculin (Burridge and Feramisco, 1980) and α-actinin (Otey et al., 1990), and interacts with integrin β-chains and the cytoskeleton (Horwitz et al., 1986; Otey et al., 1990; Otey and Burridge, 1990).

In this study we investigate the dynamics of integrin assembly into FCts of mesangial cells. In talin-containing FCts, the majority of mesangial cells rapidly organize α₂β₁ only on collagen, α₅β₁ only on fibronectin and αVβ₃ only on vitronectin, while α₆β₁ was not organized on any substrate. Control cells organized α₆β₁ well on laminin but do not organize α₃β₁. By contrast, MCs organized α₃β₁ into FCts on collagen, fibronectin and on laminin, but not on vitronectin. The data contribute to our understanding of how MCs may interact with the surrounding ECM in vivo, interactions that are important in the normal physiology of the glomerulus and in pathological conditions.

HT1080 (human fibrosarcoma) and Rugli (rat glioblastoma) cells have been described elsewhere (Brown and Goodman, 1991; Goodman et al., 1991; von der Mark et al., 1991). HT1080 express VLA-2,3,5,6 and α_Vβ₃. Rugli express the rat homologues of VLA-1,3,5, α₇β₁ and α_Vβ₃. HT1080 employs α₂β₁ to attach on collagen types I and II, α₅β₁ on fibronectin and α₆β₁ on EHS-laminin (E8-fragment) as adhesion receptors. α₃β₁ function is not known in HT1080. Rugli use α₁β₁ on collagen, α₅β₁ on fibronectin (unpublished observations) and α₇β₁ on EHS-laminin (E8). α₃β₁ has been reported as a receptor in Rugli for a sequence in laminin (E3) (Gehlsen et al., 1988, 1992).

Culture of mesangial cells obtained from human kidneys not suitable for transplantation and from Spragg-Dawley male rat kidneys has been described (Lovett et al., 1983; Ishimura et al., 1989). In brief, kidney cortex was sieved, glomeruli were retained on 75 μm mesh and were seeded in DMEM supplemented with 20% heat-inactivated FCS, 2 mM glutamate, 5 ng/ml insulin, 100 i.u./ml penicillin, 1000 μg/ml streptomycin. After 3-6 weeks, mesangial cells grew out from the glomeruli and were passaged with trypsin/EDTA (0.05%/5 mM). The subcultured cells bear the mesangial markers thy1.1, muscle-specific myosin and the muscle intermediate filament protein, desmin (Lovett et al., 1983; Ishimura et al., 1989). The cells were used between passages 4 and 11, and the results described here were independent of passage number.

Human vitronectin (Yatohgo et al., 1988) and fibronectin (Engvall and Ruoslahti, 1977) were purified from fresh blood. Laminin was isolated from the murine EHS-tumour (Paulsson et al., 1987). Human laminin from fresh post-term placenta (Brown and Goodman, 1991) was the kind gift of Dr J. Brown (Max-Planck Institute for Biochemistry, Martinsried). Collagen type I from human skin was the kind gift of Dr K. von der Mark (Max-Planck Institute, Erlangen). All matrix components were >95% pure as judged by SDS-PAGE, and by their attachment specificities (Goodman et al., 1987; Brown and Goodman, 1991), with the exception of human placental laminin, which was a mixture of laminin isoforms (J. Brown, personal communication). Cell attachment assays (Goodman et al., 1987; Goodman et al., 1991) performed in A-buffer (DMEM, 0.5% BSA, 20 mM HEPES, pH 7.4) confirmed that both rat and human mesangial cells attached and spread rapidly on matrix molecules.

C-buffer was A-buffer with DMEM substituted by EDTA-saline (5 μM EDTA, 140 mM NaCl, 5 mM KCl, 2 mM glucose). Cells in exponential growth were washed in PBS, removed from the substrate with trypsin/EDTA (0.025%/10 mM; 3 min, 37°C), harvested by centrifugation (1000 g, 5 min) and washed with culture medium. Divalent cations were removed by resuspending the cells in 5 mM EDTA in PBS. After excess EDTA had been eliminated by resuspension in C-buffer (3×), the cells were plated into C-buffer on 96-well plates coated with matrix (10 μg/ml) and blocked with heat treated BSA as described (Goodman et al., 1987; Goodman et al., 1991). The C-buffer was supplemented with 1 mM of MnCl₂, CaCl ₂, MgCl ₂ or EDTA, and after 1 h at 37°C the attached cells were counted as described (Goodman et al., 1991). Attachment to BSA-coated wells or in the presence of EDTA was <0.2% of the added cells.

Mouse anti-human integrin alpha chains P1E6 (α₂), P1B5 (α₃), P1D6 (α₅) (Wayner and Carter, 1987), rabbit anti-human vitronectin receptor (α_Vβ₃) and rabbit anti-fibronectin were from Telios (La Jolla, CA). Other antibodies were the generous gifts of our colleagues: 7F7F7C2 (mouse anti-human α₄; Dr S. Vekemans, Leuven), GOH3 (rat anti-mouse α₆, cross reacts in human; Sonnenberg et al., 1988), 3A3 (mouse anti-rat α₁; Turner et al., 1989), AIIB2 (mouse anti-human β₁, Hall et al., 1990), rabbit anti-talin (Burridge and Connell, 1983), mouse anti-paxillin (Turner et al., 1990), rabbit anti-collagen type I (Dr K. von der Mark, Erlangen), and rabbit anti-laminin prepared by immunization with EHS-laminin.

When necessary, second layer antibodies were adsorbed overnight at 4°C against rat liver acetone powder (for staining rat cells) or swine liver acetone powder (for human cells) before centrifugation (15 min, 14000 g max.) and filtration (0.22 μm) to reduce non-specific nuclear and stress fibre staining. Second layer antibodies were fluoresceinisothiocyanate (FITC)-F(ab′)₂ sheep anti-mouse-IgG (Sigma), 5(6)-carboxyfluorescein-N-hydroxysuccinimido-sheep anti-rabbit-IgG (Boehringer-Mannheim), FITC-dichlorotriazinylaminofluorescein (DTAF)-F(ab′)₂ goat anti-rat-IgG (Dianova), DTAF-donkey anti-goat-IgG (Dianova), Texas Red donkey anti-mouse (Amersham) and Texas Red goat anti-rat (Dianova).

Plastic 8-well multichamber slides (Nunc) were coated with the matrix proteins (20 μg/ml in PBS; 1 h, 37°C) before blocking for 1 h with BSA (2% heat-denatured, in PBS) and washing with PBS. Cells in exponential growth were harvested as described above and plated in A-buffer (5000 cells per well in 50 μl). Preliminary experiments using rat MCs showed that maximum and stable adhesion under these conditions was reached between 1-2 h after plating. After 1-6 h the cells were rinsed (PBS, 37°C), and fixed with paraformaldehyde (3% w/v in PBS; 10 min, 0°C). Residual aldehyde groups were blocked with ammonium chloride (50 mM in PBS; 20 min, 20°C) and the cells were permeabilized with Triton X-100 (1% w/v in PBS; 10 min, 20°C). The fixed and permeabilized cultures were treated with FCS (100%; 20 min, 20°C) and primary antibody was added (in 1% FCS in PBS). After 2 h at 37°C, the cultures were washed, and second antibody was added (1 h, 37°C). After a final wash, the cultures were mounted in Trisbuffered Moviol (Hoechst) (pH 8.6) and examined under epifluorescence illumination on a Leitz Aristoplan microscope with a ×63 Planapo objective, using FITC (490 nm emission) and Texas Red (580 nm emission) filter combinations.

Controls, where either the primary antibody was omitted or inappropriate secondary antibodies were used, showed no cross reaction between the labelled reagents. Cells were photographed using T-MAX (3200 ASA) film (Eastman-Kodak). To allow semiquantitative comparison of focal contact structures, exposure times returned by darkfield settings of the exposure meter (Wild-Leitz MPSY6) on talin-stained control cultures were used in photographing other cultures within a given experiment. Negatives were printed using similar exposure times (Ilford Multigrade paper). All Figs show cells after 4 h of attachment.

In some experiments cycloheximide (30 μg/ml) was added 3 h before harvesting for immunofluorescence, and was present in all solutions until the cells were fixed. Cycloheximide abolished endogenous fibronectin and laminin production and deposition, as judged by immunofluorescence, but had no effect on cell attachment, spreading or focal contact formation.

Focal contacts (FCts) are arrowhead or elongated talin-containing structures (Burridge and Connell, 1983; Burridge et al., 1988). We term the movement of integrins and talin into FCts ‘focalization’. Cells assembled FCts and actin stress fibres on all the substrata tested. Talin focalized strongly in most cells by 2-4 h, and was used as an internal reference for the intensity and organization of the integrin staining, which was scored in some 100 cells per staining condition. Cells with well organized integrins and staining intensities as high as for talin were scored 3+. Cells with highly organized but more weakly staining FCt arrays scored 2+, while cells with any detectable FCt organization scored 1+. Cells with no visible focal contacts scored 0. Fig. 6 has been scored in the Fig. legend to clarify the evaluation. The data sets 2+ and 3+ have been pooled as ‘positive’, and 1+ and 0 pooled as ‘negative’. Cells lacking classical FCts or with weak staining are thus eliminated from the positive count, resulting in a conservative estimation of FCt staining.

To test the specificity of the staining techniques, we examined FCt formation in cells whose integrin usage and expression, but not their ultrastructural localization, was characterized. Rugli and HT1080 cells attach to collagen type I (via α₁β₁ or α₂β₁), plasma fibronectin (via α₅β₁), EHS-laminin (via α₁β₁, α₆β₁ or α₇β₁), plasma vitronectin (via αVβ₃) and placental laminin (for references see Materials and Methods). The kinetics of FCt formation for HT1080 are shown in Fig. 1. On fibronectin and collagen >95% of the cells had talin-staining FCts within 1 h. On laminin and vitronectin the process took up to 4 h to reach a maximum and ≈20% of the cells did not form FCts. In order to confirm the formation of FCts, paxillin was stained. Paxillin focalized in >95% of cells on all substrata at 4 h. On collagen and laminin both paxillin and talin were less focalized at 6 h than at 4 h, the reason for this is not clear. The β₁-integrin chain also focalized rapidly. HT1080 focalized β₁-integrin in ≈90% of cells on collagen and fibronectin within 1 h. On laminin only ≈80% of the cells focalized β₁-integrin and the process did not go to completion. β₁-integrin was not found in FCts on vitronectin.

Individual β₁-series integrins co-localized with talin in FCts (Fig. 2). HT1080 focalized α₂β₁ only on collagen, α₅β₁ only on fibronectin and α₆β₁ only on laminin. The α₃ integrin formed no FCts, but occasional punctate talin-integrin aggregates were visible; α₄ did not form FCts. Rugli concentrated α₁β₁ in FCts on collagen and laminin but not on fibronectin or vitronectin. Thus, the presence of integrins in the FCt was substratum-dependent. The integrins that were well focalized have been shown to mediate Rugli and HT1080 attachment (Brown and Goodman, 1991; Goodman et al., 1991; von der Mark et al., 1991), and only those integrins which mediate attachment focalized. FCts formation in the control cells is summarized in Table 1.

Having confirmed the specificity of the method, we examined FCt organization in mesangial cells. In MCs, talin, paxillin and integrins moved into FCts on collagen, fibronectin, laminin and vitronectin, and strong and extensive paxillin and β₁-integrin FCts were present at 1 h (Fig. 3). FCt formation was slower in MCs than in the control cells, consistent with their slower attachment and spreading. Paxillin was organized more rapidly than both talin and β₁-integrins. Double staining with talin and with antibodies against specific integrin α-chains was used to monitor the movement of integrins into FCts in response to various matrix substrates. In MCs the concentration of α₁, α₂, α₅, and αVβ₃ into FCts closely followed the pattern of the control cells, Rugli and HT1080, while α₃ and α₆ behaved distinctly differently. The percentage of cells showing talinintegrin FCts was counted at various times after plating and the reader is invited to compare these semi-quantitative kinetic data (Fig. 4) with the immunofluorescence images which follow.

The α₁β₁ integrin concentrated within 2 h in rat MCs into talin positive-FCts on collagen and laminins, but not on fibronectin and weakly on vitronectin. The FCts were often large, pericellular and plaque-like (Fig. 5). Focalization of α₂β₁ was rapid in >80% of human MCs on collagen (Fig. 6) and moderate to weak α₂β₁ staining in FCts on human laminin was seen in ≈25% of cells at 2 h but by 6 h had dispersed; the reason for this is not clear.

The α₃β₁ integrin focalized strongly into talin-containing FCts in human MCs (Fig. 7). On collagen, fibronectin, EHS-laminin and human placental laminin, but not on vitronectin, up to 50% of the population had α₃β₁ FCts within 4 h. The α₃β₁ persisted in FCts on fibronectin and EHS-laminin at 6 h, by which time it had largely dispersed from FCts on collagen and human laminin. Outside the FCts there was a marked diffuse granularity to the α₃β₁ staining (Fig. 7); by contrast, integrin α-chain staining was generally sharply focalized. The α₄β₁ integrin did not focalize in any of the cells tested.

The α₅β₁ integrin rapidly focalized on fibronectin and ≈90% of the population had strong α₅β₁-staining FCts by 4 h. On collagen, laminin and vitronectin, 10-15% of the cells had formed weak α₅β₁ FCts by 6 h (Fig. 8). By marked contrast with the control cells, α₆β₁ was absent from FCts in mesangial cells (Fig. 9). In 80% of cells attached to vitronectin, αVβ₃ was found in prominent FCts (Fig. 9), while αVβ₃ FCts formed transiently in ≈10% of cells on fibronectin and laminin.

Although we were unable to perform double-labelling for pairs of integrin α-chains, it is implicit from the kinetic data that significant numbers of human MCs on fibronectin had both α₃β₁ and α₅β₁ in FCts, while on collagen both α₂β₁ and α₃β₁ were simultaneously present in FCts. This might explain the routine observation, especially for α₃β₁ staining, that not all the talin positive FCts within a given cell labelled for the integrin. In addition to FCt staining patterns, α-chain antibodies often gave rise to perinuclear staining, reminiscent of golgi, ER or vesicular structures, patterns which persisted in cells where integrins did not focalize.

The kinetics of incorporation of integrin α-chains into FCts in general closely followed the kinetics of attachment and spreading for MCs, however, small numbers of cells showed anomalous distribution of integrins to FCts at longer times of culture. For example α₅β₁ was focalized in 10-15% of cells on collagen and laminin at 6 h, and α₁β₁ was focalized in 20% of cells on vitronectin at 6 h, which probably reflects the deposition of matrix by these cells.

Since endogenous ECM deposition might affect FCt formation, we performed attachment and FCt studies in the presence of cycloheximide (Fig. 10). Cell attachment and spreading were remarkably little affected by cycloheximide (Fig. 10A), except on vitronectin, where spreading was inhibited (not shown). The actin cytoskeleton appeared normal (Fig. 10B,C) and talin-integrin focal contacts formed with unchanged specificity (Fig. 10D-G) even after 8 h of incubation with cycloheximide (3 h pretreatment, harvest, and 4 h attachment). Staining for fibronectin and laminin, strongly synthesized in untreated MCs, was essentially abolished by cycloheximide treatment (Fig. 10E,G). This, together with the low numbers of anomalously distributed integrins in untreated MCs strongly suggests that formation of FCt, particularly in initial adhesions, is dependent solely on the substratum offered and not on ECM proteins synthesized by the cells.

We investigated the divalent cation dependency of MCs attachment to the substrates. This indicates integrin usage because it is known that integrins α₁β₁, α₂β₁ and α₆β₁ cannot be activated by calcium ions, while α₅β₁ and αVβ₃ can, and all are activated by magnesium and manganese ions. The attachment of Rugli, HT1080 and MCs in the presence of 1 mM Mg²⁺, Ca²⁺ or Mn²⁺ is shown in Fig. 11. The cells use Mg²⁺ or Mn²⁺, but not Ca²⁺ for attachment to collagen, while Ca²⁺ can be used for attachment to vitronectin and fibronectin. Both Rugli and MCs attach strongly to laminin in the presence of Ca²⁺, whereas HT1080 can attach only in the presence of Mg²⁺. Experiments with laminin fragments show that it is the E8 fragment to which Ca²⁺-dependent attachment occurs, E1-X attachment is dependent only on Mg²⁺ (data not shown). HT1080 uses α₆β₁ to attach to E8 and Rugli use α₇β₁ (von der Mark et al., 1991). The known integrin usage of HT1080 and Rugli for attachment is summarized in Table 2, together with the ion dependency of attachment for HT1080, Rugli and MCs.

Cell attachment and spreading on extracellular matrix involves the specific relocation of integrins and intracellular components to form focal contacts. Organization of integrins into FCts can be useful in the analysis of cell-matrix interactions, especially when the availability of a cell type is limited, as with the mesangial cell of the normal human kidney (Sterzel and Lovett, 1988; Mene et al., 1989; Sterzel et al., 1992), which we have analyzed in this study. The glomerular basement membrane is produced both by MCs and glomerular podocytes, and ultimately defines the permeability of the glomerulus both to fluids and to invading leukocytes during inflammation. Although the precise mechanisms leading to ECM expansion during glomerular nephritis and other pathological conditions are not known, the resident MCs are likely to play a significant role. The ECM is known to modify the biology of a wide range of cell types, thus the interactions between MCs and the ECM are of considerable interest in normal and altered renal physiology. In the present study, several integrins, which to date have only been shown to be expressed on MCs are shown to organize into FCts, suggesting a functional significance. Our principle findings for MCs are (a) that movement of talin and paxillin into FCts is essentially independent of substrate, (b) that cell attachment and spreading, and the movement of integrins into FCts is dependent on the substrate and is independent of endogenous protein synthesis, (c) that MCs can rapidly organize α₃β₁ but not α₆β₁ integrin into FCts and (d) that a Ca²⁺-dependent integrin is used during MC attachment to laminin.

FCts habitually contain talin (Izzard, 1988; Horwitz et al., 1986; Burridge et al., 1988; Beckerle and Raymond, 1990), thus, simultaneous staining for talin and integrins allows integrin incorporation into FCts to be assessed. To test for the specificity of our techniques we followed the incorporation of integrins into FCts in the HT1080 and Rugli cell lines. Talin, paxillin (another FCt marker), and the β₁-integrin chain move rapidly into focal contacts on all substrates studied (β₁ not on vitronectin), while the αintegrin chains moved into FCts only on those substrata known to be their ligands, i.e. α₁β₁ and α₂β₁ on collagen, α₅β₁ on fibronectin and α₆β₁ on laminin, αVβ₃ on vitronectin (Goodman et al., 1987; Brown and Goodman, 1991; Gehlsen et al., 1988; Goodman et al., 1991; Wayner and Carter, 1987). We found that α₃β₁ and α₄β₁ were essentially absent from FCts.

In cultured mesangial cells and on kidney sections the expression but not the structural organization of integrins has been examined (Cosio et al., 1990; Korhonen et al., 1990; Simon and McDonald, 1990). Cultured MCs rapidly attached, spread and organized talin, paxillin and β₁-series integrins into FCts on all substrata tested (β₁ not on vitronectin). They organized integrin α-chains depending on the substratum they are offered: α₁β₁ on collagen and laminins; α₂β₁ on collagen; αVβ₃ on vitronectin; and, supporting earlier studies on MCs (Cosio et al., 1990), α₅β₁ on fibronectin. The organization of these integrins in MCs thus correlates directly with the extensive literature on other cell types (see Introduction), and also complements their organization in the control cells tested here.

Our results suggest that integrins organize because they are functioning as MC receptors for these extracellular matrix components, because in MCs: (a) the integrins organize with unchanged specificity when endogenous protein synthesis is blocked. (b) In the overwhelming majority of cells particular integrins organize only on particular substrates (e.g. α₂β₁ on collagen, α₅β₁ on fibronectin, α₆β₁, not organized). The counter hypothesis, ‘MC integrins are not specific to the substratum’ would require all integrins organized on all substrata. (c) The MC integrins that organized on particular substrata are those shown in other systems to be receptors for those substrata. (d) The divalent cation dependency of initial attachment of MCs to the various substrates matches the reported divalent cation dependencies of the integrins subsequently seen to organize into focal contacts on those substrata.

The presence of high concentrations of cycloheximide had astonishingly little effect on attachment, spreading, and focal contact organization. Synthesis and deposition of fibronectin, laminin and collagen were stopped by cycloheximide. This suggests that matrix molecules produced and deposited by MCs are not influencing cell attachment and integrin organization. Furthermore, it suggests that the entire attachment system is resilient to proteolytic degradation and turnover, and that de novo synthesis is not required to drive integrins or talin into focal adhesions. Burridge and colleagues have also recently shown a similar lack of effect of protein synthesis on FCt generation and organization in fibroblasts (Burridge et al., 1992).

In adult kidney, previous immunohistochemical studies have shown α₁β₁ to be the major β₁-series integrin in the mesangium, α₂β₁ and α₃β₁ are minor components, and α₄β₁, α₅β₁ and α₆β₁ are absent. In culture, MCs also expressed α₅β₁, and α₃β₁ is prominent (Cosio et al., 1990; Korhonen et al., 1990; Simon and McDonald, 1990). Here we show that α₂β₁ is also expressed and organized in cultured MCs. MCs express mRNA for α₁, α₂, α₃, α₅ and, weakly, α₆, while HT1080 expressed α₂, α₃, α₄, α₅, α₆ and Rugli α₁, α₃, α₅ mRNAs, which supports the immunocytochemical data we present here (Petermann, A., Grenz, H., Fees, H., Goodman, S. L. and Sterile, R. B., unpublished).

To our great surprise we found that MCs organized α₃β₁ into FCts. The α₃β₁ integrin is an enigmatic receptor and, despite its widespread distribution and multiple activities, has been reported in FCts only in keratinoctyes (Carter et al., 1990; Elices et al., 1991; Marchisio et al., 1991). It acts as a receptor for fibronectin, collagen (Elices et al., 1991) laminin (Tomaselli et al., 1990; Elices et al., 1991), epiligrin (Carter et al., 1991) and entactin (Dedhar et al., 1992), but only in the ‘absence of primary integrins’ (i.e. α₁β₁, α₂β₁, α₅β₁, α₆β₁ and α₇β₁), but its absence from FCts has been notable. By contrast to keratinocytes, which focalize α₃β₁ slowly and mainly in response to laminin (Carter et al., 1990), in MCs α₃β₁ focalizes rapidly on collagen, fibronectin and EHS-laminin despite the presence of α₂β₁ and α₅β₁. As α₃β₁ organized more slowly and incompletely compared with α₂β₁ and α₅β₁, this appears to contradict the suggestion that it has a role in initial adhesion and spreading (Carter et al., 1990). Approx. 80% of human MCs had α₂β₁ and α₅β₁ in FCts (on collagen and fibronectin, respectively), thus some cells simultaneously focalized both α₅β₁ or α₂β₁, and α₃β₁. One explanation for these results is that α₃β₁ is acting as a subsidiary receptor of broad specificity, as has been previously suggested (Wayner and Carter, 1987; Takada et al., 1987; Elices et al., 1991). Alternatively, MCs may be secreting a matrix component on which they subsequently organize α₃β₁ (e.g. epiligrin; Carter et al., 1991), although our cycloheximide blocking data suggests this is unlikely. What is more, although in the absence of cycloheximide MCs produce and deposit quantities of fibronectin and laminin, they rarely use it to organize integrins over the time-course of our assays; α₂β₁, α₅β₁ and αVβ₃ organize more or less only on the germane substrates (and α₆β₁ does not organize).

In this study, two forms of laminin have been used, EHS-laminin (the A-B1-B2 isoform) and a preparation from human placenta also containing the alternative A-chain, merosin (the M-B1-B2 isoform) (Ehrig et al., 1990). MCs attach to both isoforms (and use both the E1-X and E8 regions of EHS-laminin; unpublished observations). α₁β₁ is used to attach to E1-X, but the receptor on MCs for E8 is not clear. In MCs the α₆β₁ laminin E8 receptor (Sonnenberg et al., 1988; von der Mark et al., 1991) does not focalize, while in HT1080 it does, and α₆β₁ is not found in kidney mesangium or on MCs (Cosio et al., 1990; Korhonen et al., 1990; Simon and McDonald, 1990). In addition, the ion-dependency for MC attachment to EHS-laminin, Mg²⁺ and Ca ²⁺, also suggests that α₆β₁ is not used -α₆β₁ can only be activated by Mg²⁺ (Sonnenberg et al., 1988; and our unpublished observations). In MCs, FCts on both laminins contained α₁β₁ and α₃β₁ (this study). The α₁β₁ integrin is Mg²⁺-but not Ca²⁺-dependent (Turner et al., 1989) and cannot bind well to the E8 fragment (Hall et al., 1990; Goodman et al., 1991). Thus, either α₃β₁ is acting in the absence of α₆β₁ as a primary laminin receptor in MCs (Elices et al., 1991; Sonnenberg et al., 1991), or another integrin (Ramos et al., 1990; von der Mark et al., 1991), or a non-integrin laminin binding molecule (Davis et al., 1991; Rapraeger et al., 1987; Wewer et al., 1986) may be involved in MC attachment to laminin. We have recently shown that the major laminin binding integrin on skeletal muscle myoblasts (and the Rugli cell line) is α₇β₁ (von der Mark et al., 1991; Song et al., 1992). MCs, Rugli and myoblasts can all be activated to bind laminin E8 by Ca²⁺. MCs are smooth-muscle-like cells and it is conceivable that they use α₇β₁ as a laminin receptor.

In summary, we have investigated the incorporation of integrins into focal contacts in cultured mesangial cells. MCs rapidly organize integrins into FCts as directed by matrix components. Notably, on laminin MCs organize α₃β₁ but not α₆β₁ into focal adhesion structures. As mesangium-matrix interactions may play a pivotal role in kidney disease, the mesangial cell is an interesting cell type in which to study the interaction of matrix receptors in response to the substratum, and immunocytochemical investigation of focal contact components provides a suitable analytical tool with which to do so.

We thank the referees for suggesting the cycloheximide experiments. Elke Pausch and Ute Zimmermann assisted in establishing mesangial cultures. Drs K. Burridge (Chapel Hill, NC), C. Turner (Syracuse, NY), C. Damsky (UCSF), A. Sonnenberg (Amsterdam), and S. Vekemans (Leuven) generously provided antibodies. The constructive criticism of Dr L. Sorokin considerably improved the manuscript. H.G. and S.L.G. thank Dr R. B. Sterzel for the laboratory space. This work was completed as part of H.G.s doctoral studies at the University of Erlangen-Nürnberg. The authors gratefully acknowledge the financial support of the Deutsche Forschungsgemeinschaft, Sonderforschungsbereich 263/B5 and of the Max-Planck Society.