Actin-binding protein-280 binds the stress-activated protein kinase (SAPK) activator SEK-1 and is required for tumor necrosis factor-alpha activation of SAPK in melanoma cells.

SEK-1, a dual specificity protein kinase that serves as one of the immediate upstream activators of the stress-activated protein kinases (SAPKs), associates specifically with the actin-binding protein, ABP-280, in vitro and in situ. SEK-1 binds to the carboxyl-terminal rod segment of ABP-280, upstream of the ABP carboxyl-terminal dimerization domain. Activation of SEK-1 in situ increases the SEK-1 activity bound to ABP-280 without changing the amount of SEK-1 polypeptide bound. The influence of ABP-280 on SAPK regulation was evaluated in human melanoma cells that lack ABP-280 expression, and in stable transformants of these cells expressing wild type ABP, or an actin-binding but dimerization-deficient mutant ABP (ABPΔCT109). ABP-280-deficient cells show an activation of SAPK in response to most stimuli that is comparable to that seen in ABP-280-replete cells; ABP-280-deficient cells, however, fail to show the brisk tumor necrosis factor-α (TNF-α) activation of SAPK seen in ABP-replete cells and have an 80% reduction in SAPK activation by lysophosphatidic acid. Expression of the dimerization-deficient mutant ABP-280 fails to correct the defective SAPK response to lysophosphatidic acid, but essentially normalizes the TNF-α activation of SAPK. Thus, a lack of ABP-280 in melanoma cells causes a defect in the regulation of SAPK that is selective for TNF-α and is attributable to the lack of ABP-280 polypeptide itself rather than to the disordered actin cytoskeleton that results therefrom. ABP-280 participates in TNF-α signal transduction to SAPKs, in part through the binding of SEK-1.

SEK-1 (1) (also known as MKK-4 (2) and JNKK (3) is a dual specificity kinase whose only known substrates are the ␣, ␤, and ␥ isoforms of the p54/46 SAPK 1 subfamily of ERKs, and the p38 ERK subfamily. SEK-1 phosphorylates these proteins on a tyrosine and (probably) a threonine residue in the motif TPY for the SAPKs and TGY for p38. The status of SEK-1 as a SAPK activator in situ is certain, whereas its ability to serve as an activator of p38 under physiologic circumstances is unclear, inasmuch as recombinant MEKK-1 activates SEK-1 and SAPK (3)(4)(5) but not p38 (3) in situ. Moreover, two potent activators specific for p38 have been identified at a molecular level, namely MKK-3 and MKK-6 (6). The existence of SAPK activators other than SEK-1 has been demonstrated in extracts from osmotically stressed 3Y1 rat fibroblasts; multiple chromatographic peaks of SAPK activator activity are evident, including a minor peak corresponding with immunoreactive SEK-1 (7). No information as to the molecular structure of other SAPK kinases is available. SAPKs are activated in situ by a remarkably diverse array of cell perturbations (8). Several of these stimuli, e.g. anisomycin and TNF-␣, have been shown to also activate SEK-1; however, it is not known whether all of the SAPK agonists activate SEK-1 or recruit SAPK kinases differentially.
An interesting question in the operation of these protein kinase cascades is how the signal is conveyed from each receptor and multiple upstream stimuli to the appropriate MEK. A classical view is that the specificity is inherent in the sequential interaction of each element with its upstream and downstream partners, i.e. the MEKK interacts with the MEK, which in turn interacts with the ERK. Stable, relatively high affinity interactions of this kind have been documented (9) (e.g. Raf-MEK-1). Recent work, however, has emphasized an important contribution by "targeting" proteins in signal transmission, i.e. noncatalytic polypeptides whose main function is to provide a binding site for one or more catalytic elements, which positions them proximate to their upstream regulators and/or substrates, so as to facilitate or perhaps enable a speedy, localized and productive interaction. Examples include the targeting subunits for protein phosphatase-1 catalytic subunit and the A kinase binding protein AKAP 79, which appears to bind simultaneously kinase A (through its RII regulatory subunit), the Ca 2ϩ -activated protein (Ser/Thr) phosphatase calcineurin and protein kinase C (10,11). More relevant to the ERK-based kinase cascades is the Saccharomyces cerevisiae protein Ste5p, a noncatalytic 90-kDa LIM domain polypeptide that binds the MEKK (Ste11), the MEK (Ste7), and the ERKs (Fus3/KSS1) that constitute the core kinase cassette in the pheromone signal transduction pathway of haploid S. cerevisiae (12). Each kinase is bound to Ste5p at a different site, concomitantly.
Although direct interactions between the kinase/substrate pairs in this cassette also can be demonstrated, both in vitro and by two-hybrid methods, selective ablation of the Ste5 gene abolishes the pheromone response in situ. The kinases Ste11 and Ste7 are necessary for the response of haploid cells to nutrient deprivation (called the "invasive response") as well as to pheromone, and deletion of either kinase abrogates both responses (13). In contrast, abolition of Ste5 has no effect on the response to nutrient deprivation (14). Thus, the noncatalytic function of Ste5p is indispensable for coupling the pheromone receptor to the kinase cascade but is unnecessary for signal transmission to the same kinases by the upstream elements involved in the invasive response. Although it seems likely that signal transmission in the invasive response employs a protein analogous in function to Ste5, proteins homologous in structure or function to the Ste5p have not been identified in yeast or mammalian systems.
Herein we present evidence that the actin binding protein-280 (ABP-280, non-muscle filamin) (15) provides a function analogous to yeast Ste5p (16) in at least some mammalian cells by functionally coupling TNF-␣ and perhaps lysophosphatidic acid (LPA) receptors to the activation of the mammalian SAPK pathway, concomitant with binding selectively the SAPK activator, SEK-1.

EXPERIMENTAL PROCEDURES
Two-hybrid System-Two-hybrid expression cloning was carried out as described by Durfee et al. (17). The SEK-1 cDNA sequence starting at amino acid 40 was inserted into the pAS1-CYH2A plasmid at SalI site downstream of the Gal4 DNA-binding domain. Human B cell and murine T cell cDNA libraries, inserted downstream of the Gal4 activation domain in the vector pACTII were screened by co-expression in S. cerevisiae strain Y190. The latter contains two reporter genes, HIS3 and LacZ, which allow selection of cDNAs encoding proteins that interact with the product of the gene of interest based on growth and LacZ activity.
Yeast transformed by the method of Geitz et al. (18) were grown on plates lacking at least tryptophan and leucine to select for the presence of the bait and the prey plasmid, respectively. After 4 days, a ␤-galactosidase assay was performed by a color filter assay using 5-bromo-4chloro-3-indolyl-␤-D-galactopyranoside as described previously (17).
Materials-Catalytic subunit of cAMP-dependent protein kinase (protein kinase A), protein A, LPA, and myelin basic protein were purchased from Sigma. EGF, anisomycin, TNF-␣ were obtained from Cell Biology Boehringer Mannheim. Glutathione-Sepharose 4B and protein G were purchased from Pharmacia Biotech Inc. and Life Technologies, Inc., respectively.
Binding in vitro of various polypeptides to GST or GST fusion proteins immobilized on glutathione-Sepharose was carried out at 4°C for 2 h in binding buffer containing 25 mM Tris-Cl (pH 7.5), 50 mM NaCl, 1 mM EDTA, 1 mM dithiothreitol, 1% Triton X-100, and proteinase inhibitors. The beads were washed in excess binding buffer five times and the retained polypeptides were eluted directly into SDS-containing buffer, separated by SDS-PAGE, transferred into a polyvinylidene difluoride membrane, and analyzed by immunoblot.
Transfection and Preparation of Cell Lysate-The association of polypeptides in situ was assessed during transient expression in COS M7 cells, transfected by the DEAE-dextran method (23). GST was expressed using the vector pEBG(1); GST-SEK was expressed by insertion of the cDNA encoding SEK-1 inframe downstream of the GST coding sequence. The EE-tagged MEKK was expressed using the vector pCMV5, as before (4). Subfragments derived from ABP carboxyl-terminal 411 amino acid fragment were constructed using the polymerase chain reaction and were subcloned into the cytomegalovirus FLAG vector. All transfections utilized a total of 20 g of DNA; 48 h after transfections, cells were extracted into a homogenization buffer containing 20 mM Hepes (pH 7.4), 1 mM EDTA, 2 mM EGTA, 1 mM dithiothreitol, 25 mM ␤-glycerophosphate, 1 mM sodium vanadate, 1% Triton X-100, and proteinase inhibitors.
Immunoprecipitation and Western Blotting-Extracts were equalized for protein prior to immunoprecipitation. After addition of antibody (monoclonal antibody anti-ABP-280 or monoclonal antibody for the EE-epitope) incubation was carried out for 2-4 h at 4°C. Immune complexes were harvested with protein G-Sepharose, collected by centrifugation and washed four times with homogenization buffer. Bound protein was released from the particles by boiling for 5 min in SDS sample buffer. GST fusion protein were recovered using glutathione-Sepharose beads. Immunoblots were carried our using the ECL method (Amersham Corp.) according to the manufacturers' directions.
Confluent melanoma cells were deprived of serum overnight, and treated at 37°C with anisomycin (50 g/ml, 40 min), EGF (50 ng/ml, 20 min), TNF-␣ (50 ng/ml, 20 min), LPA (10 g/ml, 10 min), or serum (20%, 20 min). Cells were rinsed and extracted into homogenization buffer as above; cleared lysates were incubated with antisera for 4 h at 4°C. Immunocomplexes were recovered by adsorption to Sepharose-protein-G; the beads were washed three times with lysis buffer, three times with lysis buffer containing 500 mM LiCl, and three washes with kinase reaction buffer (20 mM MOPS (pH 7.2), 2 mM EGTA, 10 mM MgCl 2 , 1 mM dithiothreitol, 0.1% Triton X-100). Aliquots of beads were resuspended in 30 l of kinase reaction buffer containing 1 Ci of [␥-32 P] ATP per reaction and 20 M of unlabeled ATP. Substrates used were 5 g of purified GST-c-Jun(1-135) for SAPK (19) and 3 g of myelin basic protein for MAPK and p38 kinase activity. The reactions were terminated after 20 min at 30°C by addition of 10 l of 5 times concentrated SDS gel sample buffer. One unit of activity represents transfer of 1 pmol of phosphate min Ϫ1 from ATP to protein substrate.

RESULTS
Seeking proteins that interact with the SEK-1 protein kinase, we used the SEK-1 coding region, starting at amino acid 40, to screen a human B cell cDNA library constructed in the two-hybrid system described by Durfee et al. (17). Thirteen LacZ ϩ His ϩ clones were isolated from 6.6 ϫ 10 6 transformants; 10 gave very rapid complementation of LacZ ϩ activity and each encoded isoforms of the SAPKs. Among the remaining cDNA, one encoded polypeptide sequences identical to a carboxylterminal segment of actin-binding protein-280 (20), starting at amino acid residue 2236 of this 2647-amino acid polypeptide (see Fig. 2A). In a separate screen of a mouse T cell library, a cDNA encoding a carboxyl-terminal fragment of ABP-280 was isolated independently. The specificity of the SEK-1 interaction with ABP-280 was evaluated by examining the ability of other protein kinases to bind to the carboxyl-terminal fragment of ABP-280 in the two-hybrid assay. No evidence of binding of ABP-280 to MKK-3, p54 and p46 SAPK ␣/␤, Raf, or p70 S6 kinase was detected. Weak but definite complementation was observed consistently with MEK-1, MEK-2, and p38 (Table I).
We attempted to verify the apparent interaction between SEK-1 and the carboxyl-terminal region of ABP-280, and to examine its functional significance. ABP-280 is an abundant cytoplasmic protein that contains an amino-terminal actin binding domain of approximately 275 amino acids, followed by an extended, rodlike structure created by 24 repeats of a segment that is approximately 96 amino acids in length. Each of these segments contains eight short ␤ sheet structures (A to H) interrupted by turns. Sequence insertions of 23 and 35 amino acids, probable hinge regions, precede repeats 16 and 24, respectively. The carboxyl-terminal repeat (number 24) comprises a homodimerization domain (15,20). ABP-280 efficiently cross-links actin filaments into orthogonal arrays. ABP-280 has also been shown to bind to the von Willebrand factor (GPIb ␣/␤ IX) (26) and the high affinity Fc (Fc␥R 1 ) (27) receptors through its carboxyl-terminal rod domain.
The selective binding of SEK-1 to ABP-280 as measured by the two-hybrid method was verified by examining the ability of a prokaryotic recombinant GST-SEK to bind in vitro to fulllength ABP-280 polypeptide, either in a biochemically pure form, or in cytoplasmic extracts of 293 cells, where ABP-280 is bound to endogenous actin. ABP-280, purified to homogeneity, was incubated with purified immobilized prokaryotic recombinant GST and comparable amounts of GST-SEK, GST-SAPK, and GST-ERK-1 fusion proteins; only GST-SEK bound ABP-280 (Fig. 1A). ABP-280 also bound specifically to GST-MEK-1, although weakly as compared to GST-SEK-1; in the experiment shown in Fig. 1B, GST SEK-1 retained about 90% of the input ABP-280 and GST-MEK-1 approximately 6% (Fig. 1B). GST-SEK-1 also bound ABP-280 endogenous to an extract of HEK-293 cells (Fig. 1C), wherein ABP-280 is bound to actin and perhaps other cytosolic components. Neither GST-ERK-1 nor GST-SAPK bound ABP-280, whether the latter was presented as a purified polypeptide or in an extract of 293 cells (Fig. 1C).
Specific binding in situ of recombinant GST-SEK to a carboxyl-terminal fragment of ABP-280 tagged with a FLAG epitope, was demonstrable by co-precipitation after transient expression in COS cells (Fig. 2). Truncation of the carboxylterminal 193 amino acid from the 365 amino acid carboxylterminal ABP-280 fragment did not impair its association with GST-SEK in situ (Fig. 2B); shorter fragments of ABP-280 were poorly expressed. Thus GST-SEK binds to ABP-280 between amino acids 2282 and 2454 (repeat 21 to 23C); by comparison, the von Willebrand factor receptor binds to a fragment of ABP-280 containing repeats 16 -23. Whereas the von Willebrand factor receptor appears to bind constitutively to ABP-280, ligand engagement of the Fc␥R 1 results in its dissociation from ABP-280 (27). We therefore examined the effect of SEK activation on the association of SEK with ABP-280. MEKK-1 directly phosphorylates and activates SEK-1 in vitro; when overexpressed in situ, MEKK-1 activates SEK-1 and produces a marked and preferential activation of SAPK (3)(4)(5). Vectors encoding GST-SEK or GST were cotransfected in COS cells with an EE epitope-tagged, constitutively active MEKK-1 carboxyl-terminal catalytic fragment. Purifica-tion of cell extracts on GSH-Sepharose yielded ABP-280 from cells expressing GST-SEK-1, but not GST (Fig. 3A). Despite substantial overall activation of GST-SEK (not shown), coexpression of MEKK-1 with GST-SEK did not alter the amount of ABP-280 polypeptide associated with GST-SEK-1 (Fig. 3A). The SAPK kinase activity of an ABP-280 immunoprecipitate prepared from extracts of cells expressing GST-SEK was greatly increased by coexpression of MEKK with GST-SEK (Fig. 3B). Thus MEKK-1 activates the SAPK kinase activity bound to ABP-280 (Fig. 3B) without altering the amount of associated GST-SEK-1 protein (Fig. 3A). The MEKK catalytic fragment does not associate directly with ABP-280 but can form a ternary complex with SEK and ABP-280. Thus, a catalytically active prokaryotic GST-MEKK-1 fusion protein does not bind pure ABP-280 in vitro (Fig. 3C). Nevertheless, immobilized GST-SEK-1, bound to an essentially stoichiometric amount of prokaryotic recombinant MEKK-1, still binds ABP-280 as efficiently as in the absence of MEKK-1 (Fig. 3D). Moreover, immunoprecipitates of the EE-tagged MEKK-1 catalytic fragment, expressed (singly) in COS cells, are enriched in endogenous ABP-280 as well as endogenous SEK (Fig. 3B). Thus it is likely that the MEKK-1 catalytic fragment is able to bind to, and activate, SEK-1 while the latter is bound to ABP-280, without displacement of the activated SEK-1 from ABP-280 (e.g. see Fig. 3). By contrast, SAPK immunoreactivity was not detected in association with ABP-280 (not shown). Finally, immunoprecipitation of ABP-280 from 293 cells results in the specific coprecipitation of a portion of the immunoreactive SEK-1 endogenous to these cells; activation of SEK-1 by treatment of the cells with anisomycin, prior to harvest, does not alter the amount of endogenous SEK-1 polypeptide recovered with ABP-280 (Fig. 4).
We next attempted to evaluate the biologic significance of the ABP-280/SEK association. The SAPK cascade might regulate the functions of ABP-280 through phosphorylation of ABP-280 itself or of one or more of the proteins to which it is bound or closely apposed. Although ABP-280 can be phosphorylated in vitro by numerous protein kinases (18), neither SEK (Fig. 5, lane 6) nor SAPK (Fig. 5, lane 5) phosphorylate ABP-280 in vitro under conditions wherein ABP-280 is well phosphorylated by kinase A (Fig. 5, lane 7). Nevertheless, it remains possible that as yet unidentified kinases downstream of SEK-1 other than SAPKs are responsible for a portion of ABP-280 phosphorylation in situ.
Another possibility is that ABP-280, rather than serving as a substrate or positioning SEK near its substrates, is involved in the upstream regulation of the SAPK pathway. Overexpression of the ABP-280 carboxyl-terminal 365 amino acid fragment does not abrogate activation of a cotransfected SAPK reporter induced by MEKK-1 or V12 Rac, or by treatment of cells with EGF, TNF-␣, or LPA (not shown). This negative result, however, is tempered by the knowledge that multiple, perhaps redundant, SAPK activators in addition to SEK-1 have been demonstrated in other systems (7); their expression in COS cells and the ability of ABP-280 to bind these elements is unknown.
Another approach to the role of ABP-280 in SEK-1 function is enabled by the availability of human melanoma cell lines that have spontaneously lost expression of ABP-280 (24,25). These cells show prolonged membrane blebbing after plating, poor spreading in cell culture, and a deficit in directed cell movement. Stable expression of recombinant ABP-280 in these cells such that the ratio of ABP-280 to endogenous actin (1:160) is comparable to that found in motile cells, such as macrophages, corrects these defects (24,25). Extensive membrane blebbing (over 50% of the cell circumference) in the ABP-280-replete  3. The effect of SEK-1 activation in situ on binding to ABP-280. A, endogenous ABP-280 associates with recombinant SEK. cDNAs encoding GST-SEK-1 (or GST) and an EE-tagged MEKK-1 catalytic domain (or EE vector) were co-transfected in COS cells. Extracts prepared 48 h after transfection were purified by GSH-Sepharose affinity chromatography; bound proteins were subjected to SDS-PAGE and immunoblot with cells disappears in 90% of cells over the first 12 h after plating, presumably as a consequence of the stabilization of the cortical actin network (25). In contrast the percentage of cells showing extensive membrane blebbing in the parental cells, which lack ABP-280 expression, subsides to less than 20% slowly over the first 2-3 days after plating, coincident with a progressive increase in the relative content of F-actin over that in the ABP-280-replete cells; the increase in F-actin is proposed to reflect an alternative mechanism for cortical actin stabilization. Despite the loss of ABP-280 expression, these lines express similar levels of SAPK and MAPK (ERK-1,2) polypeptides (Fig. 6A). The regulation of the SAP kinases in these two cell lines was therefore examined at 4 -5 days after plating, a time at which a comparable, low (Ͻ10%) fraction of cells continue to show blebbing. SAPK-specific activity 16 h after serum deprivation is similar in the two lines. Anisomycin, an inhibitor of protein synthesis initiation, induces a marked SAPK activation (range from 10-to 20-fold in three experiments) in both lines; EGF yields a more modest but comparable 3-3.6-fold activation (averaged over three experiments) in both lines, and in single experiments, comparable SAPK activation was elicited by 20% fetal calf serum (ABP Ϫ , 5.3-fold; ABP ϩ , 6.8-fold), 50 M sodium arsenite (ABP Ϫ , 8.3-fold; ABP ϩ , 5.8-fold), and 0.7 M NaCl (ABP Ϫ , 6.6-fold; ABP ϩ , 10.2-fold). In contrast, the average 4.7fold activation of SAPK by TNF-␣ in the ABP-280-replete line is lacking almost entirely in the ABP-280-deficient cells, and the 5.7-fold SAPK activation elicited by LPA in the ABP-280replete cells is reduced by 80% in the ABP-280-deficient, parental cells (Fig. 6B). These findings suggest that the SAPK activation in response to TNF-␣ and/or LPA involves the direct participation of ABP-280 or is influenced in a more general way by the state of cortical actin organization (or both).
The existence of specific role for ABP-280 in the regulation of SAPK, apart from indirect effects arising from of ABP-280's influence on cortical actin structure was examined by stable  lanes 1 and 3), about a 50-fold molar excess over ABP. After washing, bound proteins were subjected to SDS-PAGE and transferred to polyvinylidene difluoride membranes, which were subjected to Coomassie Blue stain (lower panel) and anti-ABP-280 immunoblot (upper panel). E, co-precipitation of recombinant EE-MEKK1 with endogenous SEK-1 and ABP-280 from COS cell extracts. COS cells were transfected with EE-tagged MEKK-1 catalytic domain (lanes 1 and 3) or EE-vector (lane 2). Cells were extracted after 48 h, and aliquots matched for protein were subjected to anti-EE immunoprecipitation. The immunoprecipitates were subjected to SDS-PAGE and immunoblot with anti-ABP (upper panel) or anti-SEK-1 (lower panel). expression in the ABP-280-deficient cells of a dimerizationdeficient of ABP-280 mutant (ABP-⌬CT109) that lacks the carboxyl-terminal 109 amino acids and is able to bind, but not cross-link, F-actin. Cells expressing ABP-⌬CT109 exhibit prolonged membrane blebbing similar to that occurring in the ABP-280-deficient parental cells, indicating (as expected) that the ability of ABP-280 to cross-link actin filaments is crucial to its ability to stabilize the cortical actin network. Expression of this ABP-280 carboxyl-terminal deletion mutant does not alter SAPK activation by anisomycin, and has little effect on the defective SAPK activation by LPA; in contrast, TNF-␣ activation by SAPK is largely restored (Fig. 7). This response indicates that, while some aspect of the overall organization of cortical actin influences SAPK regulation in response to LPA, the monomeric ABP-⌬CT109, which contains the binding sites for actin, SEK, various receptors, and perhaps other as yet unidentified elements, but lacks the ability to cross-link Factin, nevertheless contributes in a specific way to the TNF-␣ regulation of SAPK.
The effects of ABP-280 deficiency on the regulation of two other well characterized ERK isoforms was determined (Fig. 6, B and C). In striking contrast to SAPK, whose specific activity in serum-deprived cells is quite similar in the ABP-280-deficient and -replete melanoma cells, the basal activity of MAPK and the p38 kinase is 5-10-fold higher in the ABP-280-replete cells as compared to the ABP-deficient cells despite similar levels of immunoreactive MAPK polypeptide (Fig. 6A). EGF and serum increase the low basal MAPK activity in the ABP-280-deficient cells to level approaching those seen in the ABPreplete cells. TNF-␣ gives little activation of p38 in either cell line; however, the 2-3-fold increment in MAPK induced by TNF-␣ and LPA in the ABP-280-replete cell line is markedly reduced in the ABP-280-deficient cells. The abundance of the p55 TNF-␣ receptor is comparable in the ABP-280-deficient and -replete cells (Fig. 6A). Thus ABP-280 deficiency affects the regulation of the various Erk cascades in a differential way, but in the case of both the MAPKs and SAPKs, TNF-␣ activation is substantially reduced by ABP-280 deficiency. ) were grown to confluence and examined 4 -5 days after plating. A, relative content of ABP-280, SAPK, and MAPKs. Aliquots containing an equal protein content were subjected to SDS-PAGE and immunoblot using anti-ABP, anti-SAPK, and anti-Erk 1,2 antibodies. B, C, and D, regulation of protein kinase activities. Confluent melanoma cells, deprived of serum overnight, were treated with anisomycin (50 g/ml, 40 min), EGF (50 ng/ml, 20 min), TNF-␣ (50 ng/ml, 20 s), LPA (10 g/ml, 10 min), or serum (20%, 20 min). Cell lysis, immunoprecipitation, and washing was carried out as described previously (30). B, SAP kinase activity in melanoma cells lacking ABP (open bars) or stably expressing recombinant ABP-280 (filled bars) was assayed using a recombinant GST-c-Jun (1-135) substrate. Data shown are the mean Ϯ S.E. of triplicate samples from one experiment. Similar results were obtained in four independent experiments. C, MAP kinase and D, p38 kinase activity was measured after immunoprecipitation by assay of myelin basic protein phosphorylation. The MAPK and p38 activities shown represent the average (and range) of duplicate samples from one representative experiment; similar results for MAPK were obtained in four independent experiments and for p38 in one other experiment.

DISCUSSION
These data show that the dual specificity kinase SEK-1 binds directly and specifically to the actin-binding protein ABP-280. SEK-1 binding to ABP-280 occurs in vitro and in situ, and the association is not altered detectably when SEK-1 is activated in situ by anisomycin or by coexpression with an immediate upstream activator, MEKK-1. As a consequence, active SEK-1 remains associated with ABP-280, after ABP-280 immunoprecipitation, and is capable of phosphorylating and activating in vitro added bacterial recombinant SAPK. Neither SAPK itself nor MEKK-1, one of the candidate physiologic SEK activators, binds directly to ABP-280, but recombinant MEKK-1 overexpressed in COS cells coprecipitates a small fraction of cellular ABP-280 and SEK-1, presumably reflecting a complex wherein SEK-1 binds both ABP-280 and MEKK-1 concurrently. SAPK itself was not detected in immunoprecipitates of ABP-280, although we can not conclude on present evidence that SAPK is not associated (indirectly) with ABP-280 in situ, inasmuch as SEK-1 and SAPK exhibit an interaction in the two-hybrid format that is stronger than that between ABP-280 and SEK-1.
The interaction of SEK-1 with ABP-280 is specific in the sense that most of the other protein kinases we tested in the two-hybrid assay with ABP-280 failed to exhibit interaction; however, binding was detected between ABP-280 and MEK-1, and MEK-2 and the p38 kinase. The latter interactions of ABP-280 were much weaker than that observed with SEK-1 and were not characterized further, except for the demonstration that bacterial GST MEK-1 fusion protein is capable of weak but specific binding of ABP-280 in vitro.
ABP-280 appears to behave as a rodlike structure that binds actin through its amino terminus, homodimerizes at its carboxyl terminus, and contains two flexible hinges, one located two-thirds along its length, and the second situated immediately before the carboxyl-terminal homodimerization domain. As such ABP-280 is capable of cross-linking actin filaments into orthogonal arrays, and contributes substantially to the formation and structure of the actin meshwork situated imme-diately subjacent to the surface membrane. This function may be regulated, directly and indirectly, through cell surface receptors. Direct regulation of ABP-280 function may occur through the association of certain receptors with the carboxylterminal region of ABP-280. Thus the integrin GPIb-IX, which binds von Willebrand factor on its extracellular domain, associates constitutively through its intracellular extension with ABP-280, binding to carboxyl-terminal region of ABP-280 that includes the SEK-1 binding site (15,26). The likelihood that this binding is important to the organization of the platelet cytoskeleton is indicated by the observation that platelets lacking GPIb-IX (Bernard-Soulier syndrome) are enlarged and exhibit a disorganized cortical actin meshwork (28). The high affinity IgG Fc receptor (Fc␥1R) also binds to ABP-280 in situ; however, this association is inhibited when ligand binds to FcR␥ (27).
Indirect regulation of ABP-280 function may occur through its phosphorylation inasmuch as ABP-280 undergoes phosphorylation in situ in response to serum, LPA, and other stimuli (29,30). One of the candidate protein kinases responsible for a portion of ABP-280 phosphorylation in situ is the MAPK-regulated kinase, Rsk-2 (30); at this time, however, the functional consequences of ABP-280 phosphorylation in situ are not known. It is clear that neither SEK-1 nor SAPKs phosphorylate ABP-280. It is conceivable, however, that one of the elements downstream of SEK-1 (e.g. SAPK or one of its substrates) acts on a protein that also associates with ABP-280. Exploring this idea will require selective inhibition or knockout of SEK-1, an approach not yet technically attainable.
Conversely, the availability of human melanoma cells that have lost selectively expression of ABP-280 (24) enabled an examination of the possibility that the association of SEK with ABP-280 might be consequential to the regulation of SEK-1. Inasmuch as reliably immunoprecipitating anti SEK-1 antibodies are not available, we chose to monitor the activity in situ of a SEK substrate, SAPK. Our results show that, whereas SAPK activation in the serum-deprived state and in response to most perturbations is not altered significantly by the absence of ABP-280, SAPK activation by LPA is inhibited by 80%, and activation by TNF-␣ is essentially abolished. Expression of an ABP-280 fragment lacking the nomodimerization domain does not restore SAPK responsiveness to LPA, but essentially corrects the loss of TNF-␣ regulation. Thus it appears that the decrease in LPA responsiveness may be a consequence of the reorganization in cellular actin caused by the absence of ABP-280, but the loss of TNF-␣ regulation is directly attributable to the loss of the carboxyl-terminal, actin-independent region of ABP-280. Defective TNF-␣ signaling in the ABP-280-deficient cells is also evident in the lack of MAPK activation by TNF-␣, a response that is mediated by MEK-1 and MEK-2, elements completely distinct from SEK-1. Notably, specific albeit weak binding of MEK-1 to ABP-280 is detectable in vitro. The data presented provide good evidence that ABP-280 is a participant in TNF-␣ signaling to SAPKs (and probably MAPKs), but does not provide direct evidence that the loss of the ABP-280 binding site for SEK-1 (and perhaps MEK-1) underlies the loss in TNF-␣ activation of SAPK and MAPK. Convincing support for this conclusion will require the elucidation of the biochemical steps linking the TNF receptor and other upstream inputs (e.g. UV-C ionizing radiation, heat shock, and so forth) to SEK-1/ SAPK, pathways that are still poorly understood. SEK is activated by phosphorylation in catalytic subdomain VIII, and two subfamilies of protein kinases, the MEKKs and the MLKs, have been shown to be capable of catalyzing this reaction, both in vitro and in situ (by co-transfection). Regulation of the MEKKs is unclear; the MLK known as SPRK/MLK-3 contains FIG. 7. SAPK activity in human melanoma cells expressing a dimerization-deficient ABP mutant (ABP⌬CT109). Confluent, quiescent melanoma cells lacking ABP expression (empty bars), or reconstituted will full-length ABP-280 (black bars) or ABP ⌬CT109 (hatched bars), were deprived of serum overnight and treated with the agents shown as in Fig. 5. SAPK was immunoprecipitated from aliquots of extracts equalized for protein and assayed for activity. The average (and range) of duplicate assays is shown from one experiment. A second experiment gave similar results. an SH 3 domain and a binding site for the small GTPases, Rac-1 and Cdc42. Overexpression of the GTPase-deficient forms of Rac-1 and Cdc42 results in SAPK activation. It may be relevant that ABP-280 has been reported to bind a small GTPase (31,32). In unpublished experiments we have observed specific but GTP-independent binding of RhoA, Cdc42, and Rac-1 to biochemically purified ABP-280. It is not known, however, whether TNF-␣ activation of SAPK involves the participation of a small GTPase; this and other aspects of TNF-␣ signaling to SAPKs are currently under investigation. Nevertheless, the present results point to an important role for the actin cytoskeleton in TNF-␣ signaling, at least in melanoma cells, and suggest that the direct binding of one or more signaling elements to ABP-280 may be important for this role.