Stress-activated Protein Kinase-3 Interacts with the PDZ Domain of α1-Syntrophin

Mechanisms for selective targeting to unique subcellular sites play an important role in determining the substrate specificities of protein kinases. Here we show that stress-activated protein kinase-3 (SAPK3, also called ERK6 and p38γ), a member of the mitogen-activated protein kinase family that is abundantly expressed in skeletal muscle, binds through its carboxyl-terminal sequence -KETXL to the PDZ domain of α1-syntrophin. SAPK3 phosphorylates α1-syntrophin at serine residues 193 and 201 in vitro and phosphorylation is dependent on binding to the PDZ domain of α1-syntrophin. In skeletal muscle SAPK3 and α1-syntrophin co-localize at the neuromuscular junction, and both proteins can be co-immunoprecipitated from transfected COS cell lysates. Phosphorylation of a PDZ domain-containing protein by an associated protein kinase is a novel mechanism for determining both the localization and the substrate specificity of a protein kinase.

A third class of SAPK consists of the more recently identified SAPK3 (also called ERK6 and p38␥) (20 -23) and SAPK4 (also called p38␦) (10,11,24,25). The mRNAs encoding these enzymes are present in all mammalian tissues examined, with the mRNA encoding SAPK3 being most abundant in skeletal muscle (20 -22). SAPK3 and SAPK4 are not inhibited by SB 203580 (10,23), and consequently only little is known about their substrates. The transcription factor ATF2 is a good substrate of SAPK3 in vitro (23), whereas stathmin has been proposed as a physiological substrate of SAPK4 (26). Here we identify ␣1-syntrophin as a substrate for SAPK3 and show that phosphorylation is dependent on the interaction of the carboxyl-terminal sequence -KETXL of SAPK3 with the PDZ domain of ␣1-syntrophin. In skeletal muscle SAPK3 and ␣1-syntrophin were found to co-localize at the neuromuscular junction and throughout the sarcolemma.
Yeast Two-hybrid System Screening-Yeast two-hybrid screening (27) was performed using an adult human brain expression library (CLONTECH) containing cDNAs fused to the GAL4 transactivation domain of pACT2 and rat SAPK3 DNA (20) subcloned into vector pAS2-1, which contains the GAL4 DNA binding domain. The plasmids were transformed into Y190 yeast cells, and positive clones were selected on triple minus plates (Leu Ϫ , Trp Ϫ , His Ϫ ) ϩ 25 mM 3-aminotriazole and assayed for ␤-galactosidase activity. Two million clones were screened, and two positives were obtained. Positive clones were cotransformed with either the bait vector or the original pAS2-1 (used as a control) into yeast to confirm the interaction. All the constructs that were used in other interaction experiments were from polymerase chain reaction products subcloned into pAS2-1 or pACT2 and were confirmed by DNA sequencing.
ELISA-GST fusion proteins of PDZ domain-containing proteins were bound to 96-well Micro Test plates (Falcon) at 10 g/ml in 50 mM Tris-HCl (pH 7.9). Plates were incubated overnight at 4°C, washed three times in phosphate-buffered saline (PBS) and blocked with 1% bovine serum albumin in PBS for 1 h at 37°C. After washing four times in PBS, serial 1:3 dilutions (starting at 200 g/ml) of thioredoxin-SAPK3(1-367) or thioredoxin-SAPK3(1-363) in 1% bovine serum albumin/PBS ϩ 0.1% Tween 20 (w/v) were added and allowed to bind for 1 h at 37°C. Plates were washed four times in PBS ϩ 0.1% Tween 20, incubated with anti-thioredoxin antibody (1:3,000, Invitrogen) for 1 h at 37°C, washed four times in PBS ϩ 0.1% Tween 20, and incubated with goat anti-mouse IgG-conjugated peroxidase (1:2,000, Bio-Rad) for 1 h at 37°C. Plates were washed three times in PBS, followed by the addition of 100 l of 50 mM citrate-phosphate buffer (pH 5.0) ϩ 0.5 mg/ml o-phenylenediamine (Sigma). After 5 min the color reaction was stopped by addition of 20 l of 8 N H 2 SO 4 and absorbance at 450 or 490 nm determined using a microplate reader (Molecular Devices).
Identification of Phosphorylation Sites-GST-␣1-syntrophin (0.5 M) was incubated at 30°C for 1 h with activated GST-SAPK3 (2 units/ml) (23), 10 mM magnesium acetate, and 100 M [␥-32 P]ATP in a total volume of 200 l of 50 mM Tris-HCl (pH 7.5), 0.1 mM EGTA, 0.1 mM sodium orthovanadate, and 0.1% (v/v) 2-mercaptoethanol. After SDSpolyacrylamide gel electrophoresis and autoradiography, the band corresponding to 32 P-labeled ␣1-syntrophin was excised and digested with trypsin, and the phosphopeptides generated were chromatographed on a Vydac 218TP54 C 18 column equilibrated with 0.1% (v/v) trifluoroacetic acid, and the column was developed with a linear acetonitrile gradient. The flow rate was 0.8 ml/min, and fractions of 0.4 ml were collected. The two peaks of 32 P radioactivity were analyzed by solid and gas phase sequencing (28) and also by electrospray mass spectrometry to identify the peptide sequences and sites of phosphorylation. SAPK3 was assayed routinely with MBP as substrate (23). Phosphorylation of ␣1-syntrophin by wild-type GST-SAPK3, GST-SAPK3(1-363), and GST-L367VSAPK3 was carried out in the same manner. Reactions were stopped by the addition of 1 ml of 10% (w/v) trichloroacetic acid, and after centrifugation for 10 min at 13,000 ϫ g, the supernatants were discarded. The pellets were washed three times with 1 ml of 25% (w/v) trichloroacetic acid, and 32 P incorporation was measured by Cerenkov counting. Incorporation of phosphate into substrate was kept below 0.1 mol phosphate/mol substrate in all experiments to ensure that initial rate conditions were met.
Immunofluorescence-Pectoral and semitendinous muscles were dissected from five adult Sprague-Dawley rats and kept at Ϫ70°C until use. Cryosections (10 m) were dipped in acetone, air-dried, and fixed in 2% paraformaldehyde (w/v). Following a 5-min wash in PBS, sections were incubated overnight at 4°C in 10 Ϫ7 M tetramethylrhodamine ␣-bungarotoxin (Molecular Probes, Inc.) diluted in PBS. Tissue sections were then washed for 15 min in PBS and fixed for 5 min in ethanol. For double staining, tissue sections were further incubated overnight with anti-SAPK3 serum R5 (diluted 1:200). R5 was raised in a rabbit against the synthetic peptide KPPRQLGARVPKETAL (corresponding to residues 352-367 of rat SAPK3) conjugated to keyhole limpet hemocyanin. After a 30-min wash in PBS, tissue sections were incubated for 2 h at room temperature with biotinylated anti-rabbit secondary antibody (diluted 1:200, Vector Laboratories), and, following a further 30-min wash in PBS, they were incubated for 1 h at room temperature with fluorescein-avidin D (diluted 1:200, Vector Laboratories). Sections that were triple stained were washed in PBS, blocked using the Vector blocking kit, and incubated overnight at 4°C with anti-␣1-syntrophin serum SYN17 (diluted 1:50) (29). Incubation with biotinylated secondary antibody and washings were done as for double staining and sections were then incubated for 1 h at room temperature with 7-amino-4-methylcoumarin-3-acetic acid-streptavidin (diluted 1:50, Boehringer Mannheim). Sections were mounted using Vectashield mounting medium. Immunofluorescence was observed using a Leitz DMRD fluorescence microscope using filters for rhodamine, fluorescein, and 7-amino-4-methylcoumarin-3-acetic acid. In parallel experiments, muscle sections were single stained with tetramethylrhodamine ␣-bungarotoxin, antiserum R5, or antiserum SYN17. As a control for the specificity of staining, diluted antiserum R5 was incubated with 10 M recombinant GST-SAPK3 prior to staining. Moreover, in double or triple stainings, tetramethylrhodamine ␣-bungarotoxin and antibodies R5 or SYN17 were alternatively omitted.
Transfection and Immunoprecipitation-Full-length rat SAPK3 and human ␣1-syntrophin cDNAs were subcloned into the eukaryotic expression vector pSG5 and COS cells transiently transfected with 10 g/ml plasmid DNA using DEAE-dextran chloroquine. After 48 h cells transfected with SAPK3 alone and double transfected with SAPK3 and ␣1-syntrophin were lysed in 300 l of buffer (20 mM Tris acetate, pH 7.5, 0.27 M sucrose, 1 mM EDTA, 1 mM EGTA, 1% Triton X-100, 10 mM ␤-glycerophosphate, 0.1% 2-mercaptoethanol (v/v), 1 mM benzamidine, 0.2 mM phenylmethylsulfonyl fluoride, and 5 g/ml leupeptin). Aliquots (100 l) of cell lysates were incubated for 90 min at 4°C on a shaking platform with 20 l of protein A-Sepharose conjugated to 10 l of anti-␣1-syntrophin serum TROPHA. TROPHA was raised in a rabbit against the synthetic peptide ASGRRAPRTGLLELRAG (corresponding to residues 2-17 of human ␣1-syntrophin) coupled to keyhole limpet hemocyanin. The suspensions were centrifuged for 1 min at 13,000 rpm, and the immunoprecipitates were washed twice with 1 ml of lysis buffer containing 0.5 M NaCl and once with 1 ml lysis buffer, followed by resuspension in gel loading buffer. Immunoprecipitates were detected with anti-␣1-syntrophin serum TROPHA and anti-SAPK3 serum R5.

RESULTS
To identify SAPK3 substrates, we performed a yeast twohybrid screen of a human brain cDNA library using rat SAPK3 as bait. This screen yielded two independent clones encoding residues 85-505 of ␣1-syntrophin. ␣1-Syntrophin is a peripheral membrane protein that comprises two pleckstrin homology domains, a PDZ domain and a unique carboxyl-terminal domain, with the PDZ domain being inserted into the first pleckstrin homology domain (30 -33). The related proteins ␤1syntrophin and ␤2-syntrophin share a similar domain organization (31)(32)(33)(34). Syntrophins are believed to function as modular adapters that recruit signaling proteins to the dystrophin-glycoprotein complex at the plasma membrane (35). The yeast two-hybrid system was used to examine the domains that are responsible for the ␣1-syntrophin-SAPK3 interaction (Fig.  1). Full-length ␣1-syntrophin interacted with SAPK3. The shortest construct that was positive when paired with SAPK3 contained the PDZ domain (residues 78 -179) of ␣1-syntrophin. By contrast, a construct extending from the end of the PDZ domain to the carboxyl terminus of ␣1-syntrophin (residues 174 -505) failed to interact with SAPK3, establishing that the PDZ domain of ␣1-syntrophin mediates the binding to SAPK3. PDZ domains are known to interact with the carboxyl termini of proteins that have the consensus sequence -E(S/T)XV (36,37). The carboxyl terminus of rat SAPK3 (amino acid sequence -ETAL) (20) is similar to this consensus sequence. Deletion of the last four amino acids of SAPK3 prevented its association with ␣1-syntrophin, demonstrating that this sequence is necessary for the interaction (Fig. 1). The syntrophin constructs were also expressed as GST fusion proteins and their binding to thioredoxin-SAPK3 assessed by ELISA (Fig. 1). As in the yeast two-hybrid system, SAPK3 bound through its carboxylterminal four amino acids to the PDZ domain of ␣1-syntrophin. Similarly, SAPK3 interacted with the PDZ domain of ␤1syntrophin (Fig. 1), whereas it failed to bind to the PDZ domain of neuronal nitric-oxide synthase (Fig. 1), which forms homotypic interactions with the PDZ domain of ␣1-syntrophin and PDZ domains 1 and 2 of postsynaptic density protein 95 (PSD95/SAP90) (38). The PDZ domain of neuronal nitric-oxide synthase bound to ␣1-syntrophin both in the yeast two-hybrid system and as judged by ELISA (not shown).
Human ␣1-syntrophin contains nine (S/T)P sites located outside the PDZ domain that are potential sites of phosphorylation by SAPKs (30,31). Activated GST-SAPK3 phosphorylated GST-␣1-syntrophin to 2 mol phosphate/mol protein in vitro, and two 32 P-labeled tryptic peptides were identified that corresponded to residues 198 -207 and 178 -197, respectively ( Fig.  2A). Solid and gas phase sequencing, as well as electrospray mass spectrometry were used to identify the phosphorylated residues as serines 193 and 201, which are located between the PDZ domain and the second half of the first pleckstrin homology domain (Fig. 1A). Initial rates of phosphorylation showed that relative to myelin basic protein ␣1-syntrophin is a good substrate for SAPK3 but not for other SAPKs or for p42 MAPK (Table I). SAPK3 phosphorylated ␣1-syntrophin at approximately the same rate as it phosphorylated MBP, its standard substrate ( Table I). Phosphorylation of ␣1-syntrophin by SAPK3 was dependent on the carboxyl-terminal four amino acids of SAPK3, as demonstrated by the following three separate lines of evidence (Fig. 2, B-D). ␣1-Syntrophin was a poor substrate for GST-SAPK3(1-363), which lacks the carboxylterminal four amino acids, whereas MBP was an equally good substrate for both GST-SAPK3(1-363) and GST-SAPK3(1-367) (Fig. 2B). Furthermore, preincubation of wild-type rat GST-SAPK3 with an antibody raised against its carboxyl-terminal 16 amino acids prevented phosphorylation of ␣1-syntrophin but not MBP (Fig. 2C). Finally, preincubation of ␣1syntrophin with synthetic peptides corresponding to the carboxyl-terminal 6 or 8 amino acids of rat SAPK3 prevented phosphorylation of ␣1-syntrophin by GST-SAPK3 (Fig. 2D).
The carboxyl-terminal sequence -KETAL of mouse, rat, rabbit, and zebrafish SAPK3 (20) 2 or -KETPL of human SAPK3 (10,21,22) is the most conserved sequence in the carboxylterminal region of SAPK3 and differs from the prototypical consensus PDZ domain-binding sequence (36, 37) by replacement of the terminal valine with leucine. We therefore investigated the ability of rat GST-L367VSAPK3 to bind and phosphorylate GST-␣1-syntrophin. By ELISA, the binding of wildtype GST-SAPK3 to ␣1-syntrophin was similar to that of mutant GST-L367VSAPK3 (Fig. 3A). The rate of phosphorylation of ␣1-syntrophin by GST-L367VSAPK3 was slightly faster than by wild-type GST-SAPK3 (Fig. 3B). However, both mutant and wild-type SAPK3 phosphorylated ␣1-syntrophin to the same extent (Fig. 3B). The phosphorylation of MBP by FIG. 1. Interactions between SAPK3 and ␣1-syntrophin. A, interaction of SAPK3 with the PDZ domain of ␣1-syntrophin. Binding of GAL4 fusion constructs of human ␣1-syntrophin, the PDZ domain of human ␤1-syntrophin, and the PDZ domain of human neuronal nitric-oxide synthase (nNOS) to rat SAPK3 was tested in the yeast two-hybrid system. Interactions were measured by the activity of the reporter genes HIS3 and ␤-galactosidase. HIS3 activity was judged by growth in medium lacking histidine in the presence of 25 mM 3-aminotriazole and ␤-galactosidase activity was determined from the time taken for the colonies to turn blue in 5-bromo-4-chloro-3-indolyl-␤-D-galactopyranoside filter lift assays performed at room temperature: ϩ, 90 -240 min; Ϫ, no significant ␤-galactosidase activity. In vitro binding of SAPK3 to the PDZ domain-containing proteins was tested by ELISA. The two SAPK3-interacting clones isolated in the yeast two-hybrid screen (shown as pACT2) encoded residues 85-505 of human ␣1-syntrophin. B, interaction of human ␣1-syntrophin with full-length rat SAPK3(1-367) but not with SAPK3(1-363). SAPK3 was unaffected by the L367V mutation (Fig. 3C).
If the association of SAPK3 with ␣1-syntrophin is physiologically relevant, the two proteins should be co-localized in vivo. Both SAPK3 and ␣1-syntrophin are expressed at highest levels in skeletal muscle (20 -22, 30, 31), where ␣1-syntrophin is associated with the sarcolemma and concentrated at the neuromuscular junction (39). We used immunofluorescence to examine the localization of SAPK3 in rat skeletal muscle. SAPK3 was found throughout the sarcolemma and was concentrated at the neuromuscular junction, as indicated by its co-localization with ␣-bungarotoxin, which visualizes nicotinic acetylcholine receptors at the neuromuscular junction (Fig. 4). Moreover, double staining for SAPK3 and ␣1-syntrophin showed extensive co-localization, both at the neuromuscular junction and throughout the sarcolemma (Fig. 4). The staining was specific, because it was abolished by incubation of diluted SAPK3 antiserum with 10 M recombinant SAPK3 (not shown). For an independent assessment of the ␣1-syntrophin-SAPK3 interaction, the ability of SAPK3 to co-immunoprecipitate with ␣1syntrophin was examined in extracts from mammalian cells co-transfected with both proteins. ␣1-Syntrophin and SAPK3 were co-expressed transiently in COS cells. Immunoprecipitation was carried out using an anti-␣1-syntrophin antibody, and proteins present in the pellet were immunoblotted using anti-␣1-syntrophin and anti-SAPK3 antibodies. The strong signal seen for SAPK3 upon immunoprecipitation with the anti-␣1syntrophin antibody indicates that ␣1-syntrophin existed in a complex with SAPK3 in COS cell lysates (Fig. 5). DISCUSSION SAPK3 is a protein kinase whose phosphorylation of ␣1syntrophin depends on the interaction between its carboxylterminal sequence and the PDZ domain of this substrate. The carboxyl-terminal sequence of SAPK3 thus provides a mechanism both for its selective targeting to subcellular sites and for determining its substrate specificity. During vulval induction  in Caenorhabditis elegans, the PDZ domain-containing protein LIN-7 is essential for localizing the epidermal growth factor receptor-like tyrosine kinase LET-23 to cell junctions by binding through its PDZ domain to the carboxyl-terminal sequence -KETCL of LET-23 (40 -42). Similarly, protein kinase C ␣ is a protein kinase that is targeted to subcellular sites through the interaction of its carboxyl-terminal sequence -QSAV with the PDZ domain of the protein kinase C ␣-binding protein (PICK1) (43). Moreover, p70 S6 kinase has been shown to bind through its carboxyl-terminal sequence to the PDZ domain of neurabin, suggesting a mechanism for localizing p70 S6 kinase to nerve terminals (44). The ␣1-subunits SkM1 and SkM2 of voltage-gated sodium channels from skeletal muscle and heart (45,46) have recently been shown to bind to the PDZ domain of ␣1-syntrophin through their carboxyl-terminal sequences -KESLV (SkM1) or -RESIV (SkM2) (47,48). In skeletal muscle the interaction between SkM1 and ␣1-syntrophin has been proposed as a mechanism for anchoring voltage-gated sodium channels in the depths of the junctional folds of the post-synaptic membrane (46,47). At the neuromuscular junction SAPK3 is therefore likely to be anchored in close proximity to voltage-gated sodium channels.
The carboxyl-terminal sequences of voltage-gated sodium channels closely resemble the carboxyl-terminal -KETAL or -KETPL of SAPK3 from different species, except that the terminal leucine is replaced by valine. However, binding of L367VSAPK3 to the PDZ domain of ␣1-syntrophin was found to be similar to that of wild-type SAPK3. Phosphorylation of ␣1-syntrophin by L367VSAPK3 was also similar to that of wild-type SAPK3. This indicates that proteins with a leucine residue at position 0 of the consensus sequence of PDZ domainbinding proteins will bind to ␣1-syntrophin. Mammalian  Lysates from COS cells transfected with rat SAPK3 alone (marked SAPK3) or double transfected with SAPK3 and human ␣1-syntrophin (marked SAPK3 ϩ ␣1-Syntrophin) were immunoprecipitated with anti-␣1-syntrophin serum TROPHA. Total cell lysates and immunoprecipitates (marked IP) were immunoblotted with anti-␣1-syntrophin and anti-SAPK3 antibodies. ␣1-Syntrophin-immunoreactive bands were present in double transfected cell lysates and immunoprecipitates (arrows). SAPK3-immunoreactive bands were detected in single and double transfected cell lysates (arrowhead). SAPK3 was detected as an immune complex with ␣1-syntrophin in double transfected cell lysates (arrowhead) but not in cells transfected with SAPK3 alone. The strong band in the lanes marked IP corresponds to the IgG heavy chain. type-II activin receptors are transmembrane serine/threonine protein kinases of the transforming growth factor ␤ receptor superfamily with the carboxyl-terminal sequences -KESSL or -KESSI (49,50), suggesting that they may also be PDZ domainbinding proteins and bind to ␣1-syntrophin.
Although SAPK3 is expressed at highest levels in skeletal muscle, it is expressed at lower levels in many other tissues (20). It is likely that SAPK3 will be found to interact with the PDZ domains of proteins other than ␣1-syntrophin. Possible candidates include the PDZ domains of proteins whose binding partners have a leucine residue at position 0, such as the recently identified Veli proteins, the vertebrate homologues of LIN-7 (51). SAPK3 is unique among members of the MAPK family in having a carboxyl-terminal PDZ domain-binding sequence. It therefore probably serves distinct physiological functions and is not a mere isoform of SAPK2a/p38. Inactivation of endogenous SAPK3 by gene targeting and/or the use of specific inhibitors will help to identify its specific functions.
Many proteins with PDZ domains localize to specialized cell junctions, such as synapses and tight junctions, where they bind to the carboxyl termini of transmembrane proteins, thereby creating a mechanism for positioning and clustering these proteins and for connecting them to the cytoplasmic network (52). The finding that SAPK3 co-localizes with ␣1syntrophin in skeletal muscle, that it binds to the PDZ domain of ␣1-syntrophin, and that phosphorylation of ␣1-syntrophin depends on this interaction identifies a novel mechanism for targeting a protein kinase to its substrates. Protein phosphorylation may be important for modulating the interactions between PDZ domain-containing proteins and their binding partners. It is also likely that additional protein kinases that interact with PDZ domains through a carboxyl-terminal targeting sequence remain to be discovered.