Integrin-Associated Protein Stimulates α2β1-Dependent Chemotaxis via GI-Mediated Inhibition of Adenylate Cyclase and Extracellular-Regulated Kinases

Peptides were synthesized, purified, and confirmed by mass spectrometry at the Protein and Nucleic Acid Chemistry Laboratory of Washington University School of Medicine as described previously (Kosfeld and Frazier 1992; Gao et al. 1996a,Gao et al. 1996b). The amino acid sequences of the TSP1 peptides and preparation of human platelet TSP1 were as described (Santoro and Frazier 1987). Rat tail collagen-I was obtained from Collaborative Biochemical Products. The signal transduction inhibitors and activators were purchased from CalBiochem-Novabiochem. Both anti-α2β1 antibodies (function stimulating, JBS2, and blocking, BHA2.1) were from Chemicon International. Anti-ERK antibody was from Upstate Biotechnology Inc. Anti-active (-phosphorylated) ERK antibody was from Promega. Anti-ERK polyclonal antibody for Western blotting was from Santa Cruz Biotechnology, Inc. cAMP enzyme immunoassay (EIA) system (dual range) kits and γ-[³²P]ATP were from Amersham. Myelin basic protein and anti–mouse IgG agarose were from Sigma Chemical Co. The plasmids 3CH134 encoding MAP kinase phosphatase 1 (MKP-1) and the catalytically inactive mutant of MKP-1 were generously provided by Dr. N. Tonks (Cold Spring Harbor Laboratory, Cold Spring Harbor, NY).

Human and mouse aortic SMC were isolated by the explant method and cultured as described (Ross and Kariya 1980). Cells were maintained in a humidified 37°C and 5% CO₂ environment in MEM with 20% fetal calf serum, and identified by immunostaining of α-actin (Janat and Liau 1992). Passages 2–10 were used for experiments. Cell transfection using Lipofectamine was as described (Sun et al. 1993). In brief, cells at 50–60% confluence were transiently transfected with 20 μg cDNA of each plasmid for 5 h, and recovered in growth medium for 2 d before they were harvested, tested for viability, and counted for use in assays.

Depletion of ERK from cells was performed according to the procedure of Klemke et al. 1997. In brief, SMC grown to 60% confluence in 100-mm tissue culture dishes were incubated in Opti-MEM medium (GIBCO BRL) containing Lipofectamine (10 μg/ml) and 1.5 μM ERK antisense (5′-GCC GCC GCC GCC GCC AT-3′) or control (5′-CGC GCG CTC GCG CAC CC-3′) phosphorothioate oligonucleotides for 5 h at 37°C. Cells were then rinsed, incubated for 2 d in fresh culture medium containing oligonucleotide, and then tested in migration assays as described above and processed for Western blotting with anti-ERK antibody.

Chemotaxis assays were conducted in 48-well micro-Boyden chambers (Neuroprobe) using 8-μm PVP-free, polycarbonate filters (Nucleopore). Filters were precoated by soaking in 100 μg/ml gelatin at 37°C overnight, followed by washing twice in PBS. SMC were harvested with trypsin/EDTA, washed, and diluted in MEM with 0.1% BSA to a final concentration of 3–5 × 10⁵ cells/ml. Chemoattractants were diluted in the same solution. Signal transduction inhibitors or activators were added to the cell suspension directly at the concentration indicated. The assembled chamber was incubated for 6 h at 37°C. Filters were fixed, stained, and mounted. Cells were counted in five high-power fields (HPF) in each of the triplicate wells for each condition.

Human SMC were harvested by EDTA/trypsin. After challenging with different treatments, either on matrix protein-coated 35-mm dishes or in suspension for indicated times, SMC were lysed directly at 4°C for 30 min in RIPA buffer including phosphatase inhibitors (50 mM Tris, pH 7.4, 0.15 M NaCl, 1% NP-40, 0.5% sodium deoxycholate, 1 mM EGTA, 1 mM Na₃VO₄, and 20 mM β-glycerophosphate) and protease inhibitors (10 μg/ml each of antipain, pepstatin A, chymostatin, leupeptin, soybean trypsin inhibitor, and aprotinin). The clarified cell lysates were incubated with 3 μg/tube of anti-MAP kinase antibody overnight at 4°C and then immunoprecipitated with anti–mouse IgG agarose for 1 h at 4°C. The precipitates were extensively washed with 1:3 diluted RIPA buffer three times and then two times with PBS. The immune complexes were resuspended in 60 μl kinase assay buffer (20 mM Hepes, pH 7.4, 1 mM EGTA, 20 mM β-glycerophosphate, 0.1 mM Na₃VO₄, 1 mM dithiothreitol, and 10 mM MgCl₂) including 20 μg myelin basic protein, 50 μM ATP and 5 μCi γ-[³²P]ATP, and incubated at 30°C for 30 min. The reaction was stopped by adding 20 μl of 100 mM EDTA (pH 7.0). After centrifugation to pellet the beads, aliquots (20 μl) of the supernatants were spotted onto P81 filters (Sevetson et al. 1993). The filters were washed 10 times with 1% phosphoric acid and then rinsed with 90% ethanol. The amount of ³²P incorporated into myelin basic protein was determined by scintillation counting.

Human SMC were treated as described above and lysed in RIPA buffer. Clarified cell lysates (50 μl) were added to 2× SDS-sample buffer and run on 12% Tris-glycine gels under reducing conditions. The proteins were transferred electrophoretically onto nitrocellulose membranes which were blocked with 3% BSA plus 3% dried milk in TBS buffer with 0.1% Tween-20 for at least 1 h, and subsequently incubated with antibody against ERKs overnight at 4°C. The membranes were washed extensively and incubated with peroxidase-conjugated secondary antibody for 1 h. Detection was by chemiluminescence with an ECL kit (Amersham).

Human SMC were harvested with EDTA/trypsin, resuspended in serum-free MEM at 2.5 × 10⁵/ml, and incubated with peptides or inhibitors for the indicated times. Cells were pelleted and cAMP was rapidly extracted twice with the addition of 100 μl of ice-cold ethanol (70%) for 30 min. Supernatants were collected and evaporated in a SpeedVac. The samples were dissolved in 5 ml assay buffer. The levels of cAMP were determined using a cAMP enzyme immunoassay system kit from Amersham.

We previously found that either soluble type I collagen or the TSP1 peptide 4N1K alone are relatively weak chemoattractants of human aortic SMC migrating on a matrix of gelatin. However, in combination they provoke a strong chemotactic response. This response is, in fact, chemotactic (not chemokinetic) and depends on the function of both α2β1 and IAP (Wang and Frazier 1998). To determine if the α2β1 integrin-dependent migration of SMC stimulated by 4N1K is strictly dependent on IAP, we examined the response of IAP-deficient SMC isolated from aortae of IAP^−/− mice (Lindberg et al. 1996). As shown in Fig. 1, both 4N1K and collagen can induce migration of wild-type SMC (IAP^+/+), and 4N1K plus soluble collagen together provoke a strong response, ∼200% compared with control cells, similar to the response of human SMC (Wang and Frazier 1998). However, 4N1K does not induce migration of IAP-deficient SMC. When soluble collagen is added along with 4N1K, the level of response is equal to the effect of collagen alone (Fig. 1) which reflects the action of α2β1 in the unstimulated state (Wang and Frazier 1998). This response to collagen also indicates that there is no significant defect in the ability of the IAP^−/− SMC to adhere to collagen. This was confirmed in quantitative cell adhesion assays (not shown). Thus, 4N1K stimulation of chemotaxis with or without soluble collagen is receptor mediated and is strictly dependent on the presence of IAP. This also rules out a number of trivial explanations (e.g., aggregation of collagen) for the enhanced chemotaxis.

Our previous work suggested the hypothesis that IAP stimulates α2β1 integrin function in SMC by activating the integrin in some way (Chung et al. 1997, Chung et al. 1999; Wang and Frazier 1998). If this is in fact the case, 4N1K may not need to be present during the chemotaxis assay and pretreatment of SMC with 4N1K might sensitize the cells to respond better to a gradient of soluble collagen. Thus, we compared the effects of pretreatment of SMC with 4N1K and with an anti-α2β1 mAb known to activate the integrin on other cell types (Arroyo et al. 1993; Chan and Hemler 1993; Stupack et al. 1994). As shown in Fig. 2, cells pretreated with 4N1K for 30 min and washed before placing them into the chemotaxis apparatus displayed significantly increased migration to soluble collagen. Pretreatment with the control peptide 4NGG (KRFYGGMWKK) had no effect. This effect of 4N1K is not due to residual bound peptide since direct binding assays using ¹²⁵I-4N1K reveal a very fast off rate (Dimitry, J.B., and W.A. Frazier, unpublished data). Similar results to those with 4N1K were observed when the cells were preincubated with the α2β1 function stimulating antibody: the migration to collagen increased more than threefold in three experiments (Fig. 2). An α2β1 functional blocking mAb was tested in the same experiment, and it effectively inhibited cell migration (Wang and Frazier 1998). These data support the notion that ligation of IAP activates α2β1 for a period of time and it is this activation of α2β1 that is responsible for the enhanced migration of SMC towards soluble collagen (Pickering et al. 1997; Wang and Frazier 1998).

Chemotaxis is complex process in which cell adhesion and deadhesion are regulated in a dynamic way (Lauffenburger and Horwitz 1996). In an attempt to assess the effects of 4N1K stimulation of IAP on a more direct index of integrin function, we tested the effect of 4N1K on the adhesion of SMC to gelatin and collagen-coated surfaces. As expected, collagen is a much better ligand for α2β1 than gelatin (Fig. 3). These SMC adhere to both collagen and gelatin via α2β1 as determined with a function-blocking mAb (not shown). For both immobilized ligands, 4N1K significantly decreases stable adhesion of the SMC. At submaximal densities of collagen and at all densities of gelatin, this inhibition is as much as 50%. Thus, the effect of 4N1K (and TSP1) on SMC chemotaxis, on gelatin, and collagen (Wang and Frazier 1998), may represent a case in which loosening of tightly adherent cell contacts promotes migration as pointed out by Dimilla et al. 1991 and Lauffenburger and Horwitz 1996. This deadhesion effect of 4N1K and TSP1 has been seen previously in C32 cells adhering to vitronectin (Gao et al. 1996b), even though stimulation of IAP in C32 cells causes enhanced cell spreading which uses much the same cellular apparatus as cell motility. Thus, it is not clear at this point just what the state of the integrin is when it is activated via the IAP pathway. By analogy with the platelet system, perhaps α2β1 is in a state in which it binds soluble collagen better (Shattil et al. 1994, Shattil et al. 1998). Why this should result in decreased adhesion of cells to immobilized ligand is not obvious, but may relate to changes in the mobility or clustering of the integrin on the cell surface (Lauffenburger and Horwitz 1996).

In platelets (Chung et al. 1997), C32 melanoma cells (Gao et al. 1996a), and endothelial cells (Gao et al. 1996b), IAP stimulates the activity of β3 integrins when ligated with TSP1 or 4N1K peptide. This results in platelet aggregation, C32 cell spreading, and endothelial cell chemotaxis. In all of these cases, the action of IAP is abrogated by pertussis toxin treatment of the cells (Frazier et al. 1999). Therefore, we tested the effect of pertussis toxin treatment on the response of the SMC to 4N1K and collagen. Fig. 4 shows that treatment overnight with 50 ng/ml pertussis toxin inhibited random migration to some extent, but completely eliminated the directional migration in response to 4N1K, collagen, or 4N1K plus collagen. Thus, the effects of IAP stimulation on α2β1 integrin function, like β3 integrins, appear to be mediated via a heterotrimeric Gi protein.

Since IAP signaling is mediated via activation of Gi, the alpha subunit of Gi might inhibit adenylate cyclase activity (Neer 1995). Thus, we examined the effect of IAP stimulation on intracellular levels of cAMP in SMC. As seen in Fig. 5 A, 4N1K alone evoked a rapid decrease in cAMP levels to less than half their original value in 20 min while the control peptide 4NGG had no significant effect. Interestingly, soluble collagen alone caused a substantial but transient decrease in cAMP levels while 4N1K plus collagen, the most potent chemotactic stimulus, rapidly decreased intracellular cAMP to very low levels that did not rebound by 40 min (Fig. 5 A). As seen in Fig. 5 B, pertussis toxin pretreatment of the SMC increased control (prestimulus) levels of cAMP as expected due to attenuation of inhibitory signals. The toxin also prevented the marked decrease in intracellular cAMP (at 40 min) caused by 4N1K with or without collagen. These results strengthen the conclusion that, in SMC as in other cell types, IAP acts via Gi. They also suggest that a decreased level of intracellular cAMP may be necessary for the response.

To test this idea, SMC were treated with forskolin, a direct activator of adenylate cyclase. As shown in Fig. 6, forskolin strongly inhibits the migration of SMC toward 4N1K or collagen alone and to 4N1K plus collagen. We also tested the cell-permeable cAMP derivative 8-Br-cAMP, and found that it also inhibited SMC chemotaxis to all three stimuli (Fig. 6). Thus, 4N1K/IAP induces SMC migration via activation of Gi, a necessary effect of which is decreased cytoplasmic cyclic AMP levels.

In many cell types, integrin-mediated attachment of cells to matrix proteins results in activation of MAP kinase p42/44 or ERK (Chen et al. 1994; Shattil et al. 1994, Shattil et al. 1998). Furthermore, it has been reported (Klemke et al. 1997) that ERK can regulate αvβ3-dependent haptotaxis of FG carcinoma cells on vitronectin. Thus, we asked whether IAP modulation of SMC chemotaxis might involve regulation of ERKs. When SMC adhere to gelatin (or collagen, not shown), they display a transient burst of ERK activity that peaks ∼10 min after plating and then returns to baseline levels (Fig. 7 A). 4N1K completely inhibited the increase in MAP kinase activity at 10 min and continued to decrease the baseline level of activity over time (Fig. 7 A). The control peptide 4NGG had no effect (Fig. 7 A). Western blotting using an antibody that detects the phosphorylated state of ERK confirmed this time course of activation (not shown). The inhibition of ERK activity appears to be an active process since even the basal level of kinase activity significantly declines over time with 4N1K treatment even under conditions in which cells are actively engaging the matrix (Fig. 7 A). To determine if engagement of the integrin is required for IAP-mediated inhibition of ERK activity, we harvested SMC and treated them with 4N1K in suspension. As seen in Fig. 7 B, the level of ERK activity in suspended SMC is, as expected, somewhat lower than in SMC attaching to collagen. Nonetheless, 4N1K caused a significant inhibition of ERK activity in suspended SMC while 4NGG had no effect. Significantly, collagen had no effect on ERK activity under these conditions, and collagen did not synergize with 4N1K to further reduce ERK activity (Fig. 7 B). Since it is not clear to what extent soluble collagen can bind to α2β1 under these conditions, we also treated suspended SMC with the α2β1 integrin-activating mAb used above. Neither the mAb alone nor in combination with a cross-linking anti–mouse IgG had any effect on ERK activity. Thus, ligation of IAP alone is sufficient to inhibit ERK activity and ligation and/or cross-linking of α2β1 alone is not.

To test whether ERK in SMC can be regulated by Gi, SMC were treated with pertussis toxin (50 ng/ml) overnight. These same conditions lead to a blockade of IAP-dependent chemotaxis (Fig. 4) and an increase in intracellular cAMP levels (Fig. 5). The activation status of ERK was then determined by MAP kinase immune complex kinase assays. Fig. 8 shows that pertussis toxin treatment significantly activates ERK in SMC, and the activation is concentration dependent. These results were confirmed by Western blot using an antibody that detects the phosphorylated state of ERK (not shown). Thus, blockade of a Gi-dependent pathway stimulates ERK activity indicating that an inhibitory input has been attenuated, just as in the elevation of intracellular cAMP levels by pertussis toxin (Fig. 5 B). However, elevated cAMP does not cause the increase in ERK activity. We find that treating SMC with forskolin does not elevate ERK activity (our unpublished data; Giasson et al. 1997).

Since 4N1K had a strong inhibitory effect on ERK activity, the effect on chemotaxis of inhibiting ERK by other means was tested. This was done in several ways. First, SMC were treated with the MAP kinase kinase (MEK) inhibitor PD98059 resulting in a modest but significant stimulation of chemotaxis to 4N1K and 4N1K plus soluble collagen (Fig. 9). In another approach, we downregulated the amount of ERK protein in SMC using an antisense oligonucleotide based on the ERK mRNA sequence (Klemke et al. 1997). Human SMC were treated with antisense and control (scrambled) oligonucleotides in the presence of lipofectamine for 5 h followed by a 2-d recovery period (Klemke et al. 1997). As shown in Fig. 10 A, treatment with the antisense ERK oligonucleotide resulted in increased migration to collagen, 4N1K, and 4N1K plus collagen compared with cells treated with the scrambled oligonucleotide. The same samples were subjected to Western blot analysis with an anti-ERK antibody. As shown in Fig. 10 B, the ERK protein levels were in fact substantially decreased in cells treated with the antisense ERK oligonucleotide (lane AS). ERK levels were the same in untreated cells (Fig. 10 B, lane C) and in those treated with lipofectamine and the control scrambled oligonucleotide (lane SC).

As seen in Fig. 7 A, 4N1K causes a progressive decrease in ERK activity in SMC suggesting an active suppression of ERK activity. This is normally accomplished in cells by MAP kinase phosphatases which serve to balance and attenuate the activation of MAP kinases (Keyse 1998). Thus, in a separate approach, we transiently transfected SMC with plasmid 3CH134 encoding MAP kinase phosphatase-1 (MKP-1) which can dephosphorylate MAP kinases activated by serum or v-raf (Sun et al. 1993). To determine if the MKP-1 was active in the transfected SMC, ERK activity was tested in cells replated on gelatin-coated surfaces after the cells were recovered for 48 h. As shown in Fig. 11 A, MAP kinase (primarily p42) phosphorylation was increased when control cells were attached to a gelatin-coated surface for 10 min (control). By 30 min, phosphorylation had returned to near basal levels (as in Fig. 7). Treatment with PMA (Fig. 11 A, right lane) sustained phosphorylation through 30 min. Transfection with the vector control (Fig. 11 A, VT) gave the same result as the control. However, in cells expressing wild-type MKP-1 (WT), the attachment-induced increase in ERK phosphorylation at 10 min was essentially blocked. Interestingly, in SMC transfected with the mutant MKP-1 in which the essential catalytic cysteine is replaced (Sun et al. 1993), phosphorylation of ERK is sustained through 30 min (Fig. 11 A, MT), indicating that MKPs likely play a normal role in effecting the decrease in ERK phosphorylation in SMC. The same transfected cells were also subjected to chemotaxis assays in which they were stimulated with soluble collagen alone, 4N1K alone, or 4N1K plus collagen (Fig. 11 B). In the MKP-1–transfected cells, all three stimuli were about twofold more effective chemoattractants than in the vector control or mutant MKP1 transfected SMC. The mutant MKP-1, which causes a prolonged activation of ERK (Fig. 11 A), has the same chemotactic response as the control. Thus, it appears that the transient nature of ERK phosphorylation may not be essential for the normal level of chemotaxis. The significant result here is that when ERK activation is attenuated, the chemotactic response to all three stimuli is enhanced. Thus, three different methods of decreasing ERK activity all result in enhanced chemotaxis toward collagen and magnify the effect of 4N1K.

IAP is a widely distributed membrane protein implicated in modulating integrin functions required for Ca²⁺ fluxes (Schwartz et al. 1993), phagocytosis (Blystone et al. 1995), transendothelial (Cooper et al. 1995), and transepithelial (Parkos et al. 1996) migration of PMNs, endothelial cell migration (Gao et al. 1996a), as well as integrin-mediated C32 cell spreading (Gao et al. 1996a) and platelet spreading and aggregation (Chung et al. 1997, Chung et al. 1999). Many of these functions involve modulation of β3 integrins. We have shown that TSP1, through its IAP-binding motif, the 4N1K peptide, is able to modulate the activities associated with the α2β1 integrin such that it can promote chemotaxis of human, rat (Wang and Frazier 1998), and mouse (this report) arterial SMC toward soluble collagen-I. This effect of TSP1 and 4N1K on α2β1-mediated chemotaxis requires IAP. Chemotaxis of IAP-deficient SMC toward soluble collagen cannot be stimulated by 4N1K (Fig. 1) even though these cells adhere to gelatin and collagen as well as IAP^+/+ control cells. These experiments represent the most direct proof to date that TSPs, via the 4N1K peptide, act through IAP to augment β1 integrin function.

How does this occur? The effect of 4N1K-IAP on α2β1-dependent migration on gelatin can be conceptualized in two different ways. First, IAP may alter the affinity state of α2β1, either by direct physical interaction (Wang and Frazier 1998) or via inside out signaling. Our previous data from platelets support such an affinity modulation model. The ability of 4N1K (or TSP 1 or its recombinant COOH-terminal domain, McDonald, J.F., X.-Q. Wang, and W.A. Frazier, unpublished data) to enhance SMC chemotaxis to soluble collagen is reminiscent of the activation of platelet αIIbβ3 by 4N1K/TSP-1 (Chung et al. 1997, Chung et al. 1999). This effect in platelets is also dependent on IAP since platelets from IAP deficient mice do not aggregate in response to 4N1K or TSP 1 (Chung et al. 1999). The fact that 4N1K pretreatment can stimulate migration to soluble collagen is suggestive of a change in the activation status of α2β1 that lasts long enough to be manifested during the several hours required for cells to migrate through the filters. An activating mAb against α2β1 gives the same result, i.e., enhanced chemotaxis to soluble collagen, which is a poor ligand for the basal state of the integrin (Santoro and Zutter 1995) just as soluble fibrinogen is a terrible ligand for αIIbβ3 on quiescent platelets (Shattil et al. 1994). Thus, the enhanced chemotaxis toward soluble collagen by 4N1K may be analogous to the stimulation of soluble fibrinogen binding to platelet αIIbβ3 by 4N1K (Chung et al. 1997, Chung et al. 1999) which leads to platelet aggregation. In platelets, the binding of the ligand mimetic mAb PAC-1 can be used as an index of αIIbβ3 affinity. In the case of α2β1, there is no such convenient marker for a higher affinity/avidity state of the unliganded integrin. The effect of 4N1K to decrease cell adhesion as measured in a static adhesion assay (Fig. 3) appears to argue against a simple activation of ligand binding by the integrin, since this would normally be expected to enhance not inhibit cell adhesion. This result speaks to some change in the state of the integrin's valency or affinity, but the nature of this change is unclear at the present time.

The second hypothesis, which need not be exclusive with affinity modulation, is that IAP signaling is integrated with α2β1 signaling at one or more levels inside the cell. Both collagen and 4N1K lead to a drop in intracellular cAMP levels. With collagen alone, this effect is transient, whereas 4N1K leads to the sustained suppression of cAMP which is enhanced by coadministration of collagen. This effect is Gi dependent, as shown by experiments with pertussis toxin, and independent of adhesion and integrin engagement, as shown by experiments with suspended cells. The decrease in cAMP levels and the inhibition of ERK activity can both have profound effects on the cell's motility apparatus (Tan et al. 1992; Klemke et al. 1997), and perhaps on other events upstream that connect integrin engagement to cell motility.

The precipitous drop of cAMP levels in the SMC upon treatment with 4N1K is likely to be due to a direct action of Giα_[GTP] on adenylate cyclase. In platelets and in leukocytes, high levels of cAMP appear to block the transition to a more active state of the integrins (Shattil et al. 1994, Shattil et al. 1998; Laudanna et al. 1997). Thus, our observation that a prominent effect of IAP stimulation in SMC is a profound drop in intracellular cAMP levels may well be related to removing a brake on α2β1 function. This effect is pertussis toxin sensitive and thus clearly requires Gi function (Fig. 5 B). We have found an identical, IAP-dependent, pertussis toxin-sensitive depletion of cAMP in platelets (Chung et al. 1997; Frazier et al. 1999) where other costimulators of integrin activation such as thrombin, epinephrine, and ADP act via G proteins to lower cAMP levels (Shattil et al. 1994). Soluble collagen alone also causes a transient fall in cAMP in SMC. Together, 4N1K and collagen synergize, resulting in a more profound and prolonged decrease of cAMP levels, in parallel with the more pronounced chemotactic response elicited by the combined stimuli. This suggests that the integrin and IAP work together to activate Gi. In fact, we have recently found that 4N1K or TSP1 ligation of IAP on platelets synergizes with collagen ligation of α2β1 to induce platelet activation and aggregation (Chung et al. 1999). That a heterotrimeric complex of integrin and IAP is the functional signaling unit is further indicated by the existence of a detergent-stable integrin/IAP complex in SMC (Wang and Frazier 1998), in platelets (Chung et al. 1999), and in melanoma cells which is physically associated with Gi (Frazier et al. 1999). In addition, Gi can be copurified with IAP and its integrin partner by affinity chromatography on immobilized 4N1K, and by immunoprecipitation with anti-IAP mAbs. Furthermore, 4N1K stimulates the binding of γ-[³⁵S]GTP to cell membranes and this stimulation requires IAP since it is not observed in membranes from cells of IAP knockout mice (Frazier et al. 1999). This result with isolated membranes makes it unlikely that IAP signals the synthesis of an autocrine factor which then indirectly activates Gi through its own Gi coupled serpentine receptor. If ligation of IAP is required to activate Gi, how then does collagen alone cause a decrease in cAMP levels (Fig. 5 A)? Our preliminary data suggests that cross-linking of the integrin-IAP complex by collagen may responsible.

Another immediate result of IAP stimulation with 4N1K is a suppression of ERK activity and phosphorylation (Fig. 7). That this suppression of ERK activity is causal for chemotaxis is indicated by the fact that three different methods of inhibiting ERKs all result in enhanced chemotaxis (Fig. 9,Fig. 10,Fig. 11). A mechanism of integrin modulation that involves ERKs is suggested by experiments of Hughes et al. 1997 in which a chimeric integrin expressed in CHO cells was engineered to be in an activated state as determined by the binding of the ligand-mimetic mAb, PAC-1. Then cDNAs were expressed in the CHO cells and screened for inhibition of PAC-1 binding. Several inhibitory or integrin deactivating cDNAs were identified and all encoded activators of the ERK pathway such as activated H-ras and raf-1. These results suggest that MAP kinase, specifically ERKs, inhibit integrin activation of the sort required for binding of soluble ligands. Interestingly, it has been reported that activation of αIIbβ3 and binding of soluble fibrinogen to platelets coincides with inhibition of ERKs (Nadal et al. 1997). Whether ERK inhibition might cause αIIbβ3 activation was not determined. We used three very different ways of decreasing ERK activity and all three led to an increased chemotactic response of SMC. The modulation of α2β1 activity in SMC by IAP suggested by our data may be the first physiological example of the ERK regulation of integrin affinity/avidity.

Klemke et al. 1997 have reported that α2β1-dependent migration of FG carcinoma cells on a collagen substrate required activation of MAPK/ERK. While their assay used a Boyden chamber apparatus like our chemotaxis assays, the filters on which the cells migrated were coated with fibronectin only on the bottom side. Their assay was thus a haptotaxis assay which measures increased adhesion to the matrix protein coated on the underside of the filter. Thus, their result is that activation of ERKs results in increased integrin-mediated adhesion to matrix, not increased chemotaxis as observed here for SMC. In fact, we observe (Fig. 3) that activation of IAP with 4N1K results in decreased adhesion concomitant with increased chemotactic migration. Importantly, Klemke et al. 1997 found that inhibition of ERKs did not prevent spreading of the FG cells, a cellular function closely allied to cell migration. Thus, our results are, in fact, consistent with those of Hughes et al. 1997 and Klemke et al. 1997, and reinforce the idea that activation of MAPK/ERKs inhibits binding of soluble ligands to integrins and also stimulates integrin-mediated adhesion which can inhibit cell migration. The finding that ERKs can phosphorylate MLCK (Klemke et al. 1997) may also support the second model for IAP action to regulate cell motility at downstream sites.

The mechanism by which IAP inhibits ERK activity is not clear. We find that ERK activity is stimulated by disabling Gi with pertussis toxin (Fig. 8) suggesting that under normal growth conditions, some tonic input acts through Gi to exert a negative influence on ERK activity in SMC. This negative regulation may be due to βγ since expression in SMC of the β-ARK COOH-terminal domain that sequesters βγ also stimulates ERK activity in randomly cycling, proliferating cultures of SMC (Wang, X.-Q., and W.A. Frazier, unpublished data). This result is in contrast to recently published data on the role of βγ as a positive effector of ERK activation in serum-starved SMC stimulated with serum or growth factors (Iaccarino et al. 1999). However, the effects of βγ sequestration are potentially quite complex, impacting not only downstream targets of βγ, but also receptor desensitization and down regulation (Lefkowitz 1998). Thus, it is conceivable that βγ sequestration may differ in its net result when tested in proliferating vs synchronized cells stimulated to enter the cell cycle.

Whatever the mechanism, the essential role of a Gi protein in the stimulation of chemotaxis by 4N1K is clearly demonstrated by the effect of pertussis toxin treatment of the SMC. The downstream events that lead to the effects on cell motility that we observe could be mediated by the Giα subunit or the βγ heterodimer released upon activation and dissociation of Gi (Hawes et al. 1995; Neer 1995). The inhibition of chemotaxis by forskolin and 8-Br-cAMP implies that decreased cAMP (and lower PKA activity) is necessary to allow directed migration (Fig. 6). This observation is consistent with other studies of SMC chemotaxis (Dubey et al. 1995; Palmer et al. 1998) and with the inhibitory effect of high cAMP levels on activation of leukocyte integrins (Laudanna et al. 1997) and the long known inhibitory effect of cAMP on activation of the platelet integrin αΙΙbβ3 (Shattil et al. 1994). Another well known effect of cAMP is the PKA-mediated inhibition of cell motility via phosphorylation of myosin light chain (Tan et al. 1992) which would also put a brake on chemotaxis until cAMP levels are reduced.

It has been reported (Graves et al. 1996) that adenylate cyclase is downstream of MAP kinase in SMC, since activated ERKs can increase cAMP levels. This occurs via ERK-mediated phosphorylation and activation of PLA2 leading to prostaglandin production, and autocrine adenylate cyclase stimulation via Gs (Graves et al. 1996). In addition to 4N1K, PD98059 also decreases cAMP levels in SMC (our unpublished data). Thus, when ERK is inhibited, cAMP levels follow. However, manipulation of cAMP levels does not appear to affect ERK (our unpublished data; Giasson et al. 1997), consistent with the idea that ERK regulation is upstream of adenylate cyclase in SMC (Graves et al. 1996). Thus, activation of Gi can lower cAMP levels in two ways: direct inhibition of adenylate cyclase by Giα and inhibition of ERKs leading to decreased stimulation of cyclase via the prostaglandin pathway. It appears that both decreased cyclic AMP levels and lower ERK activity are necessary for IAP modulation of SMC motility.

The data reported here indicate a novel role for TSP family members (all of which contain the IAP binding sequence) and IAP in the regulation of SMC migration. This may occur via modulation of a β1 integrin mediated by inhibition of MAP kinase, and a concomitant decrease in intracellular cyclic AMP levels. Both of these signaling events are tremendously pleiotropic, and both PKA and ERKs have been shown to impact regulation of cell motility via direct phosphorylation of myosin light chain kinase and/or myosin light chain (Tan et al. 1992; Klemke et al. 1997). Thus, IAP may act at multiple levels to modulate integrin function as well as facilitating signaling downstream of integrin ligation. There is ample literature implicating TSP1 in the proliferation and chemotaxis of cultured SMC (Majack et al. 1986, Majack et al. 1988; Yabkowitz et al. 1993; Patel et al. 1997; Wang and Frazier 1998) and in atherogenesis (Wight et al. 1985) and restenosis (Chen et al. 1999). Our data suggest the possibility that manipulation of TSP1 signaling through IAP may be of therapeutic benefit.

We thank Drs. Eric Brown and Scott Blystone for many helpful discussions and for sharing reagents with us, Dr. Nicholas Tonks for the MKP1 plasmids, and Drs. Michael Crow and David Cheresh for helpful discussions. We thank Drs. Linda Pike, Tom Mariani, and Nader Sheibani for critical reading of the manuscript, and Anna Blanchard for help in its preparation.

This work was supported by grants from the National Institutes of Health and Monsanto/Searle.