Cyclic GMP-dependent protein kinase is required for thrombospondin and tenascin mediated focal adhesion disassembly

Cells function in the context of their extracellular environment. Regulation of cell-matrix interactions and of cell shape are essential for control of cellular processes such as mitosis, migration, and tissue morphogenesis. Cell cycle progression and gene expression are affected by cell shape, which is dependent on cytoskeletal organization and cell-matrix interactions (Ingber, 1993; Juliano and Haskell, 1993; Sims et al., 1992). Interactions of extracellular matrix molecules with their integrin and proteoglycan receptors localized in focal adhesion plaques not only promote organization of the actin-containing microfilaments, but also stimulate intracellular signaling events (Hynes, 1992; Woods and Couchman, 1992, 1994; Lo and Chen, 1994; Damsky and Werb, 1992; Schwartz, 1992; Rosales et al., 1995). Phosphorylation of regulatory proteins in adhesion plaques, such as talin, paxillin, tensin, integrins and focal adhesion kinase (FAK), appears not only to control cellular architecture, but to modulate nuclear events as well: ligand binding to integrin induces nuclear translocation of mitogen activated protein kinase (Schaller et al., 1992; Juliano and Haskill, 1993; Turner et al., 1989; Turner, 1991; Chen et al., 1994).

In previous work, we identified and characterized the antiadhesive activity of several matrix proteins: thrombospondin (TSP) (reviewed by Murphy-Ullrich, 1995; Murphy-Ullrich and Höök, 1989, 1993), tenascin (Murphy-Ullrich et al., 1991), and SPARC (Murphy-Ullrich et al., 1991, 1995). These proteins induce disassembly of focal adhesion structures from a subpopulation of spread endothelial cells and fibroblasts (Murphy-Ullrich, 1995). These changes are characterized by a loss of vinculin from plaques, redistribution of actin-containing stress fibers to a submembranous web, and retention of the α_vβ₃ integrins in adhesion plaques. The effects of these anti-adhesive proteins on cell adhesion are variable and context-dependent. Immobilized forms of TSP can mediate cell attachment and in some cases, cell spreading, although it has not been observed to promote focal adhesion formation (Adams and Lawler, 1994; Adams, 1995; reviewed by Murphy-Ullrich, 1995). SPARC substrates are generally non-adhesive, whereas, tenascin substrates can support cell attachment and/or spreading of certain cell types (Lane and Sage, 1994; Lotz et al., 1989; Prieto et al., 1992; Sriramarao et al., 1993). Furthermore, each of these proteins can inhibit cell attachment under certain conditions (Murphy-Ullrich and Mosher, 1987; Lahav, 1988; Spring et al., 1989; Lightner and Erickson, 1990; Riou et al., 1991; Sage et al., 1989). The regions of these proteins active in focal adhesion disassembly have also been identified (Murphy-Ullrich et al., 1991, 1993, 1995): a 19 amino acid heparin-binding peptide from the amino terminus of TSPs1 and 2 (hep I); the alternatively spliced fibronectin type III repeats of tenascin (TNfnA-D); and the follistatin-like (peptide 2.1) and the Ca²⁺-binding E-F hand (peptide 4.2) of SPARC. Although TSP, tenascin, and SPARC are structurally distinct, they are similar in that expression of these proteins is induced by growth factors and they are abundant at foci of migrating or dividing cells (e.g. in wounds, during embryogenesis, and in subconfluent cultures; Penttinen et al., 1988; Majack et al., 1987; Chiquet-Ehrismann et al., 1989a,b; Chiquet-Ehrismann, 1991; Adams and Lawler, 1993; Frazier, 1991; Lahav, 1993; Bornstein and Sage, 1994; Erickson 1993; Lane and Sage, 1994). The temporally-restricted expression of these proteins is consistent with the fact that focal adhesions are disassembled in migrating cells and reform when cells become stationary (Couchman and Rees, 1979; Huttenlocher et al., 1995). Furthermore, these proteins can potentially modulate cellular processes such as angiogenesis, proliferation, and metastasis (Sage and Bornstein, 1991; Lahav 1993; Bornstein and Sage, 1994; Erickson, 1993; Lane and Sage, 1994).

The mechanism(s) whereby TSP, SPARC, and tenascin mediate focal adhesion disassembly have not been previously described. There is accumulating evidence that phosphorylation of cytoskeletal components plays a role in focal adhesion integrity. For example, activation of protein kinase C (PKC) in epithelial cells, smooth muscle cells, and fibroblasts by TPA or PDGF results in loss of stress fibers, dispersal of vinculin, and loss of focal adhesions (Tidball and Spencer, 1993; Schliwa et al., 1984; Herman et al., 1986, 1987; Herman and Pledger, 1985). Elevation of cyclic AMP-dependent protein kinase (PKA) activity in fibroblasts or mesangial cells by selective agonists or by microinjection of the kinase also induces loss of stress fibers and focal adhesions (Turner et al., 1989; Lamb et al., 1988; Glass and Kreisberg, 1993). The GTPase rho stimulates cytoskeletal organization and focal adhesion formation (Ridley and Hall, 1994), whereas microinjection of the GTPases rac and cdc42 induce actin filament reorganization and the formation of lamellipodia and filopodia, respectively (Nobes and Hall, 1995).

In order to determine whether focal adhesion disassembly mediated by the anti-adhesive matrix proteins occurs via intracellular signaling events, we tested for the involvement of various serine/threonine kinases in mediating focal adhesion disassembly. Through the use of kinase inhibitors and kinase transfected cell lines, we now show that the counter-adhesive matrix proteins TSP and TN, but not SPARC, require the involvement of cyclic GMP-dependent protein kinase (PKG) in order to stimulate cytoskeletal reorganization.

Thrombospondin was prepared from fresh human platelets purchased from the Birmingham American Red Cross and purified by heparin-affinity chromatography and gel permeation chromatography as previously described (Murphy-Ullrich et al., 1993). Peptides (hep I: ELTGAARKGSGRRLVKGPD) were prepared as previously described (Murphy-Ullrich et al., 1993) or purchased from Quality Controlled Biochemicals, inc., Hopkinton, MA. Peptides were purified by reversed-phase high pressure liquid chromatography and subjected to amino acid analysis.

A bacterially expressed recombinant fragment corresponding to the alternatively spliced segment (TNfnA-D) of tenascin-C (M_r= 70×10³) was provided by Dr Harold Erickson, Duke University (Murphy-Ullrich et al., 1991; Chung and Erickson, 1994). Synthetic peptides corresponding to SPARC domains 2.1 and 4.2 were generous gifts from Dr E. Helene Sage, University of Washington (Lane and Sage, 1990; Murphy-Ullrich et al., 1995).

Bovine aortic endothelial (BAE) cells were isolated and cultured in DMEM containing 4.5 g/l glucose (Cell-Gro, Mediatech, Herndon, VA), and 20% fetal bovine serum (Hyclone, Logan, UT) as previously described (Murphy-Ullrich et al., 1993). Cells were used between passages 5 and 12. Cells were routinely tested for mycoplasma.

Rat aortic smooth muscle cells (RASMC) were isolated and cultured as described previously (Cornwell and Lincoln, 1989). Cells in early passage (i.e. passage 1-3) contained PKG when examined by western blot analysis whereas late passage cells (passage >5) were PKG-deficient as reported by Cornwell and Lincoln (1989). Expressed PKG in both normally passaged cells and cells stably-transfected with PKG constructs was measured by western blotting using specific antibodies which recognize PKG and by specific PKG phosphorylation assays. To create cloned RASMC expressing constitutively active PKG, RASMC in passage 3 were transfected with the catalytic domain of PKG Iα (Boerth and Lincoln, 1994). The cDNA for bovine cGMP-kinase Iα was cloned into pcDNA1-neo (CMV promoter driving expression of kinase, SV40 poly(A) tail, neomycin resistance gene) at the BamHI site. Passage 3 RASMC were transfected with the cDNA construct using Transfectam (Promega, Madison, WI) for 6 hours. The transfection was terminated by adding 10% FBS. After 48 hours, the cells were split in a ratio of 1:15 and were allowed to attach overnight. The cells were placed in medium containing 500 µg/ml G418. Clonal cell lines were selected by antibiotic resistance and after two weeks were replated in medium containing 250 µg/ml G418. Clonal cell lines were assessed for PKG expression by western blot analysis using a polyclonal antibody to bovine lung PKG Iα and by cGMP-dependent phosphotransferase activity. The colony chosen for this study expressed similar levels of PKG as freshly isolated RASMC. All cell lines were maintained in DMEM supplemented with 5% FBS, 5% calf serum, and 250 µg/ml G148. Control transfected cells were isolated in a similar fashion as PKG-expressing cells, except that the former were transfected with the empty vector only. All experiments were performed on stably-transfected RASMC cultures between passages 6 and 8.

Focal adhesion assays were performed as described (Murphy-Ullrich and Höök, 1989; Murphy-Ullrich et al., 1993). BAE cells or RASMC were grown on glass coverslips overnight in DME with 5% FBS until nearly confluent. Cells were then pretreated for 1 hour with 10 µg/ml cycloheximide, rinsed once with warm DMEM and then incubated with serum-free DMEM, or TSP peptides for 1 hour at 37°C in the continued presence of cycloheximide. In some experiments, cycloheximide was omitted with no change in results. In some experiments, protein kinase inhibitors were pre-incubated with the cells at 37°C 2030 minutes prior to the addition of peptides, a time that was empirically determined to be sufficient to allow the KT and Rp drugs to enter the cells. After these incubations, cells were fixed with 3% warmed glutaraldehyde for 30 minutes at 37°C, washed, mounted on glass slides, and examined by interference reflection microscopy (IRM) for the presence of focal adhesions using a Zeiss Axiovert 10 microscope. A minimum of 300 cells/condition was evaluated for the presence of focal adhesions. A cell was scored positive if it contained at least 3 adhesion plaques. Each experiment was repeated a minimum of 3 times with n=3-8. Statistical significance of data was analyzed using Student’s t-test.

Intracellular cGMP levels were assayed by a specific radioimmunoassay (Harper and Brooker, 1975). Assays were performed on extracts of BAE cells treated with 1 µM hep I in the presence or absence of the phosphodiesterase inhibitor 3-isobutyl-1-methylxan-thine (IBMX). Cells treated with ANP served as positive controls for elevation of cGMP levels.

The following kinase inhibitors were purchased: Rp-8-Br-cGMPS, Rp-8-Br-cAMPS (BioLog, LaJolla, CA), HA1004 (Seikagaku America, Inc., Rockville, MD), KT5720, KT5823, KT5926 (Kamiya Biomedical Company, Thousand Oaks, CA). The following cyclic nucleotide-dependent kinase agonists were purchased from BioLog, LaJolla, CA: Sp-5,6-DCI-cBIMPS, Sp-8-Br-cGMPS. Atrial natriuretic peptide (ANP), S-nitroso-N-acetylpenicillamine (SNAP), and forskolin were purchased from Sigma Chemical, St Louis, MO.

Protein phosphorylation has been shown to modulate the assembly and disassembly of focal adhesion structures (Juliano and Haskill, 1993; Schaller et al., 1992; Turner et al., 1989; Burridge et al., 1992). In order to determine whether specific protein kinases regulate focal adhesion disassembly by the antiadhesive matrix protein, TSP, BAE cell monolayers were treated with the active sequence from TSP1 (the hep I peptide) in the presence of increasing concentrations of serine/threonine kinase inhibitors. Cells were preincubated with the kinase inhibitors for 20-30 minutes prior to addition of hep I, a time we determined to be sufficient for their entry into cells. As shown in Fig. 1, HA1004, an isoquinolinesulfonamide derivative selective for the cyclic nucleotide-dependent protein kinases (PKA, PKG), significantly blocked loss of focal adhesions mediated by hep I. The concentrations that inhibit hep I-mediated focal adhesion disassembly were in the range that selectively inhibit cAMPdependent protein kinase (PKA) and cGMP-dependent protein kinase (PKG), but below that required to inhibit PKC activity (Fig. 1). These data suggest that PKA and/or PKG are required to mediate loss of focal adhesions by induced hep I. Because the selectivity of HA1004 for PKG and PKA is poor, additional experiments using other protein kinase inhibitors were performed in order to test for their effects on hep I-mediated loss of focal adhesions. The kinase inhibitors KT5823 and KT5720 have been reported to be competitive inhibitors of ATP binding to the catalytic domains of PKG and PKA, respectively (Kase et al., 1987). In the experiments described in Fig. 2, hep I-stimulated loss of focal adhesions was highly sensitive to inhibition by KT5823 at concentrations near the K_i for PKG (Fig. 2a), whereas the selective inhibitor of PKA, KT5720, had no effect on hep I activity at concentrations effective for inhibiting PKA (Fig. 2b). However, hep I activity was blocked by KT5720 at concentrations which were effective for inhibiting PKG and PKC. These results suggest that of the two cyclic nucleotide dependent protein kinases, PKG may play a role in the hep I-induced loss of focal adhesions in BAE cells.

Another class of cyclic nucleotide dependent kinase inhibitors, i.e. the cyclic nulceotide phosphorothioates (‘Rp compounds’) act as membrane permeant cGMP/cAMP antagonists (Butte et al., 1990; Moretto et al., 1993; Cornwell et al., 1994a). These compounds were tested as antagonists to the hep I-mediated loss of focal adhesions in BAE cells. As shown in Fig. 3, the PKG antagonist Rp-8-Br-cGMPS completely blocked hep I-mediated loss of focal adhesions (Fig. 3a), consistent with the data obtained from the studies with the KT compounds. However, the PKA antagonist Rp-8-Br-cAMPS also blocked hep I activity. This may have been due to the fact that the phosphorothioates may accumulate to levels in intact cells that ‘cross over’ and bind to both kinases (i.e. Rp-8-Br-cAMPs may inhibit PKG and vice versa) even though these compounds are partially selective for the inhibition of their respective kinases in vitro (Butte et al., 1990; Hofmann et al., 1985). In any event, the cyclic nucleotide phosphorothioate compounds are highly specific for the inhibition of cyclic nucleotide dependent kinases in comparison to other classes of kinases (Hofmann et al., 1985). Taken together, the results from experiments using KT drugs and the Rp compounds demonstrate that PKG activity is necessary for the focal adhesion dissociating activity of hep I.

In previous studies, we demonstrated that the alternatively spliced segment of tenascin induced a loss of focal adhesions from spread endothelial cells, resulting in cytoskeletal changes similar to those observed in cells treated with TSP or its active sequence, hep I (Murphy-Ullrich and Höök, 1989; Murphy-Ullrich et al., 1991). These findings have been confirmed by others (Hahn et al., 1995). Therefore, the alternatively spliced segment of tenascin (TNfnA-D) was examined for sensitivity of its effects on focal adhesion disassembly to protein kinase inhibitors. As shown in Fig. 4, the ability of TNfnA-D to stimulate loss of focal adhesions was blocked by KT5823 at concentrations that selectively inhibit PKG (Fig. 4a). Rp-8-Br-cGMPS also blocked the effects of TNfnA-D on focal adhesion disassembly (Fig. 4c). On the other hand, KT5720 did not inhibit TNfnA-D-mediated loss of focal adhesions at concentrations that inhibit PKA, although KT5720 was inhibitory near the K_i for PKG and PKC (Fig. 4b). These data indicate that both TSP and tenascin require PKG in the pathway to signal focal adhesion disassembly.

The specificity of the requirement for PKG in mediation of focal adhesion disassembly in response to extracellular matrix proteins was studied by examining the effect of PKG inhibitors on the anti-adhesive properties of SPARC (Murphy-Ullrich et al., 1995). As shown in Fig. 5, SPARC sequences (peptides 2.1 and 4.2) that mediate focal adhesion disassembly were insensitive to inhibition with either KT5823 or Rp-8-Br-cGMPS (Fig. 5), suggesting that SPARC mediates focal adhesion disassembly via an alternate pathway and that only certain anti-adhesive matrix proteins exhibit a requirement for PKG activity. Furthermore, these experiments show that these kinase inhibitors do not constitutively block focal adhesion disassembly.

The possible role of PKG in focal adhesion disassembly was lso examined using rat aortic smooth muscle cells (RASMC). RASMC are known to express high levels of PKG in primary and early passage cells (i.e. 1-3 passages), but expression of the kinase is rapidly down-regulated with repeated passage (and dedifferentiation) in vitro, such that by passage 8-10, there is a 10-to 20-fold reduction in PKG protein levels in these cells as compared to primary cultures (Cornwell and Lincoln, 1989; Lincoln et al., 1990; Cornwell et al., 1994a,b). The expression of PKA, on the other hand, is not affected by passage of cells in vitro (Lincoln et al., 1990). These biological differences provided a means to test the involvement of PKG in hep I-mediated focal adhesion disassembly independent of the use of ither kinase inhibitors or agonists. Specifically, we examined whether early (passage 2-3) and/or late (passage 17-23) passage RASMC were sensitive to the effects of hep I on focal adhesion disassembly. Early passage cells that express PKG show a loss of focal adhesion plaques in response to hep I, whereas late passage cells that lack PKG do not exhibit these changes in response to hep I (Fig. 6). The interference reflection images of hep I-treated RASMC show a loss of focal adhesion positive cells with a morphology similar to that observed in hep I-treated BAE cells (Fig. 7). The broad grey area in the center of Fig. 7b is a close contact and the shorter, fuzzy dark gray areas are characteristic of focal adhesions that are in the process of ‘breaking up’ as viewed by time lapse interference reflection microscopy (J. E. Murphy-Ullrich, unpublished results), whereas intact focal adhesions appear as sharp, precise, slitlike structures by IRM (arrowheads in Fig. 7c,d). Hep I-treated smooth muscle cells immunostained for vinculin have primarily a disperse distribution of vinculin, consistent with our previously published studies (data not shown). The lack of response to hep I by late passage RASMC does not appear to be due to a loss in the ability of these cells to interact with hep I, since late passage cells bound radiolabeled hep I at levels similar to either early passage RASMC or BAE cells (data not shown).

To further examine the role of cGMP and PKG in focal adhesion disassembly, we examined the effects of the expression of the constitutively-active domain of PKG in late passage RASMC. The active catalytic domain of PKG Ia (amino acids 336-671) has been thoroughly characterized in baculovirus expression systems (Boerth and Lincoln, 1994).

This construct has proven useful in defining specific cGMP-mediated actions since the enzyme is active in the absence of cGMP elevation, yet retains the substrate specificity of the intact enzyme. Stable-transfection of the catalytic domain of PKG Iα in RASMC restores responsiveness of the cells to hep I and TNfnA-D, whereas RASMC stably-transfected with vector alone remain refractory to the effects of hep I and TNfnA-D (Fig. 8a,b). Consistent with the data from the inhibitor studies, the ability of SPARC to induce focal adhesion disassembly was independent of PKG expression. On the other hand the expression of active PKG alone in the absence of hep I and TNfnA-D did not reduce the percentage of cells positive for focal adhesions, suggesting that PKG activity in the absence of an additional signal does not by itself trigger focal adhesion disassembly. This conclusion is supported by the observation that the guanylate cyclase agonists atrial natriuretic peptide and SNAP fail to induce focal adhesion disassembly (Fig. 9). In contrast the cAMP elevating agent, forskolin, readily stimulated loss of focal adhesions. Moreover, hep I treatment failed to elevate cGMP levels in BAE cells even in the presence of the phosphodiesterase inhibitor IBMX (untreated cells, 3.6 fmol cGMP/60 mm² dish; 1 µM hep I, 1.4 fmol; 1 µM hep I + 100 µM IBMX, 2.1 fmol, whereas ANP treatment elevated cGMP levels (100 nM ANP, >128 fmol/dish).

These results together with data from the inhibitor studies with BAE cells demonstrate that PKG activity is necessary, but not sufficient by itself, to accomplish the events needed to effect focal adhesion disassembly by either TSP or tenascin.

Anti-adhesive extracellular matrix proteins exhibit temporal and spatial patterns of expression: the expression of these proteins at foci of cell locomotion and tissue morphogenesis is consistent with their role in inducing cytoskeletal reorganization and weakening of cell-matrix interactions (Sage and Bornstein, 1991; Murphy-Ullrich, 1995; Tucker, 1993; Huttenlocher et al., 1995). In these studies, we now show that the anti-adhesive proteins TSP and tenascin similarly require intracellular signals to stimulate cytoskeletal reorganization and loss of focal adhesion plaques. Interactions of the active regions (hep I and TNfnA-D) of these molecules with cells stimulate focal disassembly via a process that requires PKG activity. This was demonstrated through the use of several protein kinase inhibitors with differential selectivities for specific protein kinases and by the lack of response to hep I and TNfnA-D in PKG-deficient cells and restoration of responsiveness in PKG-transfectants. This is the first demonstration of a role for PKG in the intracellular signaling pathways involved in cell adhesion that are regulated by either TSP or tenascin.

Both hep I and TNfnA-D decrease focal adhesion assembly in BAE cells through a process that requires PKG activity. These data, however, suggest that hep I and TNfnA-D do not directly stimulate activation of PKG. Neither hep I nor TNfnAD increases intracellular cGMP levels as measured by radioimmunoassays in whole cell extracts (data not shown), although these data do not rule out the possibility that there may be locally elevated levels of cGMP. This does not appear to be the likely explanation, however, in that the constitutively active PKG catalytic domain does not cause focal adhesion disassembly in the absence of hep I or TNfnA-D. We propose that hep I and TNfnA-D trigger focal adhesion disassembly through interactions with their respective receptors, activating signaling pathways that require PKG activity to facilitate events that alter cytoskeletal protein interactions necessary for focal adhesion disassembly and cytoskeletal reorganization. A similar situation has been described in neutrophils in which PKG was necessary for both formyl peptide and A23187-induced morphological changes and degranulation (Wyatt et al., 1991, 1993; Pryzwansky et al., 1995). However, PKG activation alone was insufficient to regulate these neutrophil events. In the vascular smooth muscle and endothelial cells, PKG is clearly required for TSP and TN-dependent focal adhesion disassembly, but PKG by itself did not apear to be sufficient to trigger these events. This is consistent with our observations that the guanylate cyclase agonists, ANP and SNAP, did not stimulate focal adhesion disassembly.

PKG has only recently been reported to be present in large and small vessel endothelium (MacMillan-Crow et al., 1994; Diwan et al., 1994), although it is well-established as a mediator of nitric oxide-induced vascular smooth muscle relaxation (reviewed by Lincoln, 1994). One of the reasons for the difficulty encountered in identifying PKG in endothelial cells is that the kinase is localized primarily in the cytoskeletal (membrane) fractions of BAE cells and not in the cytosolic fraction (MacMillan-Crow et al., 1994). Similarly, PKG has been shown to be transiently associated with the intermediate filament protein, vimentin, in activated neutrophils (Wyatt et al., 1991). Targeting of PKG to specific structures/regions within a cell may confer substrate specificity on PKG activity by co-localizing the kinase with its potential substrate (Lincoln and Cornwell, 1993). It is possible that hep I and TNfnA-D also enable PKG to be co-localized with certain substrate proteins important in the regulation of focal adhesion disassembly. Although it has not been directly demonstrated, it is reasonable that PKG is localized to focal adhesion plaques, since a major substrate of PKG (and PKA) in platelets, VASP, is present in the focal adhesions of endothelial cells and fibroblasts (Reinhard et al., 1992) and binds profilin and a zyxinrelated protein (Reinhard et al., 1995a,b). The role of VASP phosphorylation in hep I mediated focal adhesion disassembly remains to be clarified.

The observation that talin, vinculin, and integrins were more highly phosphorylated in transformed cells as compared to normal adherent cells prompted investigation of kinasedependent mechanisms of focal adhesion disassembly. There are several lines of evidence suggesting the involvement of serine/threonine kinases in regulating cytoskeletal reorganization. Activation of PKC by TPA in epithelial cells and by TPA and/or PDGF in smooth muscle cells and fibroblasts leads to loss of stress fibers and of focal adhesions (Schliwa et al., 1984; Herman and Pledger, 1985; Herman et al., 1986, 1987; Tidball and Spencer, 1993). Microinjection of PKA into rat embryo fibroblasts or treatment of mesangial cells with cAMPelevating agents resulted in loss of stress fibers and focal adhesions (Turner et al., 1989; Lamb et al., 1988; Glass and Kreisberg, 1993). We have similarly observed that treatment of BAE cells with cAMP agonists (forskolin, 5,6-DC1-BIMPS) results in loss of focal adhesions (Fig. 9; J. E. Murphy-Ullrich, unpublished data). These data do not address the role of PKA, however, in that it is known that cAMP and cAMP analogues ‘cross over’ and activate PKG in cells (Lincoln et al., 1990; Jiang et al., 1992). The intracellular pathways mediating focal adhesion disassembly appear to have some cell type specificity, since PKC activation resulted in loss of focal adhesions from BSC-1 epithelial cells, but had no effect on rat embryo fibroblasts; conversely, dibutyryl-cAMP induced focal adhesion loss from fibroblasts but not from epithelial cells (Turner et al., 1989). Little is known concerning focal adhesion stability in large vessel endothelial cells, although it has been shown that activation of PKC by lipoxygenase metabolites of arachidonic acid in microvascular endothelial cells decreases vinculin and α_vβ₃-containing adhesion plaques (Tang et al., 1993). There may be differences between large and small vessel endothelium, since with some inhibitors we observed a decrease in the number of focal adhesion positive cells with kinase inhibitors in the effective range for PKC.

With the exception of PDGF, interleukin 1β (Qwarnström et al., 1991), interleukin 8 (Dunlevy and Couchman, 1995) and the arachidonic acid metabolites, most of these studies have utilized non-physiologic molecules such as drugs, to examine signaling pathways involved in triggering focal adhesion disassembly. TSP and tenascin, on the other hand, are physiologically relevant ligands that mediate alterations in cytoskeletal organization and cell adhesiveness. These proteins play a role in cell growth and motility and have been localized to regions of cell migration (wound healing) and tissue morphogenesis. Each of these proteins interacts with cells via multiple receptors (Frazier, 1991; Adams and Lawler, 1993; Erickson, 1993; Yost and Sage, 1993). A cell surface form of annexin II has been shown to function as a receptor for the alternatively spliced domain of tenascin (Chung and Erickson, 1994) and antibodies against annexin II block the ability of this domain to induce focal adhesion disassembly (Chung et al., 1996). Since annexin II is not an integral membrane protein, it is not known whether actual signaling by TNfnA-D occurs via an annexin-associated membrane protein. The receptor for the TSP active sequence, hep I, is currently under investigation in our laboratory. Although both hep I and TNfnA-D exhibit a requirement for PKG, the extracellular domains of their respective receptors appear to be distinct, since polyclonal antibodies to annexin II have no effect on hep I activity (Chung et al., 1996). It remains to be determined whether a common cytoplasmic motif exists on the hep I and TNfnA-D signaling receptors.

The further elucidation of the signaling pathways triggered by the anti-adhesive matrix proteins is warranted. It will be interesting to determine how these pathways compare to signaling cascades induced by adhesive extracellular matrix proteins.

This work was supported by NIH grants HL44575 (J.E.M.U.), HL34646 (T.M.L.) and NSF grant IBN9119404 (T.M.L.) and a grantin-aid to T.L.C. from the Alabama affiliate of the American Heart Association. J.A.G. was supported by an NIH post-doctoral traineeship (T32 AR07450) from the Training Program in Rheumatic Diseases Research at UAB. The authors thank Drs Harold Erickson, Duke University, and Helene Sage, University of Washington, for the gifts of recombinant TNfnA-D and SPARC peptides 2.1 and 4.2, respectively, and Dr Anne Woods for critical reading of the manuscript.