PSK, a Novel STE20-like Kinase Derived from Prostatic Carcinoma That Activates the c-Jun N-terminal Kinase Mitogen-activated Protein Kinase Pathway and Regulates Actin Cytoskeletal Organization*

Degenerate polymerase chain reaction against conserved kinase catalytic subdomains identified 15 tyrosine and serine-threonine kinases expressed in surgically removed prostatic carcinoma tissues, including six receptor kinases (PDGFBR, IGF1-R, VEGFR2, MET, RYK, and EPH-A1), six non-receptor kinases (ABL, JAK1, JAK2, TYK2, PLK-1, and EMK), and three novel kinases. Several of these kinases are oncogenic, and may function in the development of prostate cancer. One of the novel kinases is a new member of the sterile 20 (STE20) family of serine-threonine kinases which we have called prostate-derived STE20-like kinase (PSK) and characterized functionally. PSK encodes an open reading frame of 3705 nucleotides and contains an N-terminal kinase domain. Immunoprecipitated PSK phosphorylates myelin basic protein and transfected PSK stimulates MKK4 and MKK7 and activates the c-Jun N-terminal kinase mitogen-activated protein kinase pathway. Microinjection of PSK into cells results in localization of PSK to a vesicular compartment and causes a marked reduction in actin stress fibers. In contrast, C-terminally truncated PSK (1–349) did not localize to this compartment or induce a decrease in stress fibers demonstrating a requirement for the C terminus. Kinase-defective PSK (K57A) was unable to reduce stress fibers. PSK is the first member of the STE20 family lacking a Cdc42/Rac binding domain that has been shown to regulate both the c-Jun N-terminal kinase mitogen-activated protein kinase pathway and the actin cytoskeleton.

We set out to obtain a profile of kinases expressed in surgically removed prostate tumors to identify potential oncogenic signaling pathways that might be involved in progression to prostate cancer. The degenerate PCR screen of malignant prostate tissues identified 15 tyrosine or serine-threonine kinases and provides a basis for further analysis of kinase signaling pathways and function in prostate cancer development. The PCR screen for kinases also identified a novel member of the STE20-like family of protein kinases which we have called prostate-derived STE20-like kinase (PSK). PSK has a potential role as a signal transduction molecule in prostatic carcinomas and may function similarly to other STE20-like proteins discovered in mammalian tissues, which occupy pivotal roles in coupling stimuli to downstream signaling pathways. We have demonstrated that PSK specifically activates the JNK MAPK pathway and induces loss of actin stress fibers, suggesting a role for this new protein in regulating cellular stress responses and cell motility.
Cloning of PSK-For 3Ј RACE (rapid amplification of cDNA ends), RNA from a T-47D breast carcinoma cell line was reverse transcribed using the primer 5Ј-GACTACGTTAGCATCTAGAATTCTCGAG(T) 17 -3Ј. Nested PCR was carried out using two forward primers 5Ј-CT-GCGTCCATCATGGCACCTG-3Ј, followed by 5Ј-TTTGGATCCACTAG-TGTCGA-3Ј and the reverse primer 5Ј-GACTACGTTAGCATCTAGAA-TTCTCGG-3Ј and a 534-bp product obtained and cloned into pGEM-T vector (Promega). The 534-bp 3ЈRACE product was used to screen a Lambda Zap cDNA library prepared from T-47D carcinoma cells (Stratagene, kind gift from M. Crompton, ICR, London). Filters (Amersham Pharmacia Biotech, Hybond™-N ϩ ) were hybridized at 65°C in a solution containing 5ϫ SSPE, 5ϫ Denhardt's solution, and 20 g/ml salmon testis DNA (Sigma) for 48 h. Stringency of the final wash was 1ϫ SSPE, 0.1% (w/v) SDS at 65°C. CDNA clones were isolated from a screen of 1 ϫ 10 6 phage clones and sequenced on both strands (MWG-Biotech). PCR to confirm the presence of PSK in human genomic DNA was carried out using a forward primer 5Ј-CCTGAACAAGAAGCAGAC-CCA-3Ј in combination with a reverse primer 5Ј-CCAGAATCCTCT-GCTGCTTGG-3Ј. PCR conditions were 93°C/45 s, 60°C/60 s, and 68°C/2 min for 35 cycles using 1 unit of Supertaq plus DNA polymerase (HT-Biotech) per 50-l reaction.
Plasmid Construction-For mammalian cell expression, the open reading frame of PSK (758 -4463 bp) was initially cloned into EfPlink (kind gift from R. Marais, ICR, London) containing a 5Ј MYC epitope tag. The 5Ј end of PSK was generated by PCR using a forward primer 5Ј-AAGCTTGGATCCATGCCAGCTGGGGGCCGG-3Ј, to introduce a BamHI site next to the ATG start codon, with the reverse primer 5ЈAAAGAATTCGCTGCAGGCGGCTGGAGA-3Ј. After digestion with appropriate enzymes the BamHI Ϫ1103 XbaI fragment of PSK was ligated into EfPlink along with the 3Ј end of PSK (XbaI 1104 -4492 XbaI) and an annealed oligonucleotide adaptor 5Ј-CTAGAGCATTGAG-TAGATAGAATTCAAGCTTA-3Ј and 5Ј-CTAGTAAGCTTGAATTCTAT-CTACTCAATGCT-3Ј introduced at its 3Ј XbaI site. To improve protein expression in COS1 cells upon transient transfection the open reading FIG. 1. Identification of kinases expressed in three prostatic carcinoma tissues. The identity of each protein kinase is shown on the left of the alignment of the deduced amino acid sequences encoded in PCR-amplified fragments. The number of individual isolates sequenced is shown on the right. Conserved amino acid sequences are shown in boldface and the conserved subdomains described by Hanks et al. (45) are indicated by Roman numerals. The novel clones are marked (*). frame was subcloned into a cytomegalovirus-based expression vector pRK5 containing a 5Ј MYC-epitope tag. PSK (K57A) was prepared using the method of Kunkel et al. (47). Briefly, PSK (BamHI Ϫ3504 BamHI) was subcloned into pBluescript II SK and used to generate a mutated fragment using the oligonucleotide 5Ј-GAGGTGGTGGC-CATCGCGAAGATGATGTCCTACAG-3Ј. The resultant PSK (BamHI Ϫ3504 BamHI, K57A) fragment was ligated back into EfPlink-PSK (BamHI 3504 -4492 XbaI).
Northern Blotting-A human multiple tissue Northern blot (MTN™, CLONTECH) was probed with an XbaI 1104 -1800 XhoI or BamHI 3504 -4492 XbaI fragment of PSK using the method described by the manufacturer. Transient Cell Expression, Immunoblotting, Immunoprecipitation, and Kinase Assays-COS1 cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% FCS and antibiotics (5%CO 2 , 37°C). For transfection, cells (1.9 ϫ 10 6 /10 ml of medium) were seeded onto 100-mm Petri dishes and grown for 16 h before the indicated plasmids (6 g of each plasmid with the total amount of DNA made equivalent using empty vector) were co-transfected into COS1 cells for  For immunoblotting of whole cell lysates, cultures were lysed in 200 l of gel sample buffer and 100 g of each sample was separated by 10% or 15% SDS-PAGE before proteins were transferred to nitrocellulose (Schleicher & Schuell). MYC-tagged PSK or p38 proteins were detected using 9E10 anti-MYC tag antibody (Sigma), and FLAG-tagged JNK or ERK2 were detected using anti-FLAG tag M2 antibody (Sigma). Briefly, membranes were blocked with TBS (150 mM NaCl, 10 mM Tris, pH 7.4) containing 5% dried milk for 1 h at room temperature; incubated in TBS-Tween 20 (0.5%) containing either a 1/500 dilution of 9E10 or 5 g/ml M2 for 1 h (room temperature); and finally incubated in TBS-Tween 20 (0.5%) containing 1/2000 anti-mouse horseradish peroxidaseconjugated antibody (Transduction Laboratories) and 1% dried milk for 1 h (room temperature). Membranes were processed using ECL according to the manufacturer's instructions (Amersham Pharmacia Biotech).
For immunoprecipitation, 400 g of protein were taken and made up to a final volume of 1 ml in binding buffer (50 mM NaCl, 2.5 mM MgCl 2 , 0.1 mM EDTA, pH 8.0, 0.05% (v/v) Triton X-100, 40 mM Na 4 P 2 O 7 , 2 g/ml leupeptin, 2 g/ml aprotinin, 50 mM NaF, 0.1 mM Na 3 VO 4 , 1 mM phenylmethylsulfonyl fluoride, 20 mM Hepes, pH 7.6) for JNK or p38 kinase assays, or to 0.5 ml in buffer B for ERK2 and MBP phosphorylation assays. For JNK or ERK2 kinase assays, samples were incubated with 3 g of anti-FLAG mouse monoclonal antibody for 1 h at 4°C, and for MBP phosphorylation or p38 kinase assays, 3 g of anti-MYC-tag 9E10 mouse monoclonal antibody was added for 1 h at 4°C. 20 l of protein G-Sepharose beads (Sigma; 1:1 in appropriate binding buffer) were added to each sample for incubation for another 1 h (4°C), and beads were pelleted by centrifugation and washed two times in 0.5 ml of appropriate binding buffer.
For in vitro kinase assays, beads were washed in respective kinase buffers. JNK and p38 kinase assays were carried out in 30 l of kinase buffer A (20 mM MgCl 2 , 2 mM dithiothreitol, 20 mM p-nitrophenyl phosphate, 4 mM NaF, 0.1 mM Na 3 VO 4 , 20 mM Hepes, pH 7.6) containing 20 M ATP, 5 Ci of [␥-32 P]ATP (Amersham Pharmacia Biotech, 3000 Ci/mmol) and either 3 g of GST-c-Jun protein or 3 g of GST-ATF2 protein, and incubated for 20 min or 30 min, respectively, at 30°C. ERK2 or MBP kinase assays were carried out in 30 l of kinase buffer B (20 mM MgCl 2 , 2 mM MnCl 2 , 30 mM Tris, pH 8.0) containing 0.25 mg/ml MBP, 10 M ATP, and 1 Ci of [␥-32 P]ATP for 30 min at 30°C. Kinase assays were terminated in gel sample buffer, heated to 100°C for 5 min, and proteins separated by SDS-PAGE for transfer to nitrocellulose. Samples were processed by immunoblotting and analyzed by exposure to film or phosphorimaging (Fuji). Protein normalization and protein expression were confirmed for each sample by immunoblotting of cell lysates or immunoprecipitated material.
FLAG-tagged MKK3, MKK4, or MKK7 were transfected and immunoprecipitated as described for FLAG-tagged JNK except that recombinant GST-p38 (MKK3) or GST-JNK (MKK4 or -7) were used as substrate for in vitro kinase assays.
To localize MYC-tagged PSK and actin filaments, cells were incubated with 1:500 anti-Myc tag antibody 9E10 (Santa Cruz Biotechnology) in PBS, 0.1% bovine serum albumin followed by 1:400 FITCconjugated goat anti-mouse IgG (Jackson Immunoresearch) in PBS, 0.1% bovine serum albumin together with 0.1 g/ml TRITC-conjugated phalloidin (Sigma). To localize PSK and vinculin, cells were incubated with 1:200 rabbit anti-Myc tag antibody (Santa Cruz Biotechnology) and 1:50 mouse anti-vinculin antibody (Sigma). Cells were imaged with a confocal laser scanning microscope (LSM 510, Zeiss). Image files were collected as a matrix of 1024 by 1024 pixels describing the average of 8 frames scanned at 0.062 Hz, where FITC and TRITC were excited at 488 nm and 543 nm and visualized with bandpass filters of 540 Ϯ 25 and 608 Ϯ 32 nm, respectively.
Plasmids-The following plasmids were kindly provided by colleagues; pCMV-FLAG-JNK1 and pGEX-c-Jun

Kinase Expression in Human Prostatic Carcinoma Tissues-
Total RNA was prepared from three surgically removed prostate tumors (Gleason grades 6 -8) and used to obtain a profile of kinases expressed in prostatic carcinoma tissues. RNA was reverse transcribed to cDNA and amplified by PCR using degenerate oligonucleotides that correspond to the highly conserved amino acid sequences present in subdomains V1 ((I/ V)HRDL) and 1X (DVWS(F/L)G) of the kinase catalytic domain (45,46). This approach allowed us to amplify interconnecting sequences of approximately 200 base pairs between the two subdomains that were specific for each kinase. Sixty-six independent clones were sequenced. Based on this analysis, the deduced amino acid sequences (GenBank and EMBL data bases) identified 15 different tyrosine or serine-threonine kinases including six known receptor kinases (PDGFBR, IGF1-R, VEGFR2, MET, RYK and EPH-A1), six non-receptor kinases (ABL, JAK1, JAK2, TYK2, PLK-1 and EMK), and three novel kinases ( Fig. 1). Of the three tumors analyzed, one (Gleason grade 6) was more informative. All of the known kinases identified in this screen were represented at least once in this tumor, with IGF1-R being identified the most frequently (31% of clones sequenced for this tumor). Interestingly, 79% of clones sequenced from the other two tumors (Gleason grade 8) encoded PDGFBR, while IGF1-R was sequenced twice and JAK1, ABL, and MET were each sequenced once. We also detected three novel kinases. GenBank data base searches suggested that clone 1 was most related to the STE20 family of serinethreonine kinases with 52% homology to STE20 and clone 2 had most homology with the Caenorhabditis elegans predicted protein C46C2.1 (87%) and the Phycomyces blakesleeanus predicted protein PKPA (60%) (48,49). Clone 3 now appears to be the human homolog of mouse PKL12 (50). We chose to focus on the novel kinase, which had significant homology to members of the STE20 family of serine-threonine kinases, because of the role that these proteins play in the regulation of cell growth and survival. This new kinase is analyzed in detail here.
Cloning of cDNAs Encoding a Novel Member of the STE20 Kinases-3Ј RACE was carried out using a modification of the method of Frohman (51) to obtain more cDNA encoding the novel STE20-related kinase. This approach generated a PCR fragment of 534 bp, which was used to screen cDNA libraries leading to the isolation of two cDNA clones. One cDNA clone contained an insert of approximately 5 kb and was sequenced on both strands, while the second clone (approximately 3 kb) represented a C-terminally truncated version of the longer clone, lacking an in-frame stop codon.
A start codon AUG at position 758, which reflected a good match to the consensus for efficient translation initiation (52), an in-frame stop codon, and an open reading frame of 3705 bp encoding a predicted protein of 1235 amino acids were identified (Fig. 2). The kinase domain was located at the N terminus of the protein and contains all 11 subdomains. The first kinase subdomain sequence (GHGSFG) is preceded by 34 amino acids and is followed by a crucial lysine residue at amino acid 57, which is required for ATP binding. The C-terminal domain contains several proline-rich regions representing predicted SH3 domain binding regions and a leucine-rich region. We have called this new protein PSK (prostate-derived STE20-like kinase).
PSK Is Homologous to STE20-related Kinases-A search of data bases (GenBank and EMBL) revealed that PSK had significant sequence homology over the kinase domain with members of the STE20 family of kinases including TAO1 (90%), MST1 (45%), NIK (44%), HPK1 (43%), GCK (42%), and STE20 (40%) (Fig. 3A). The presence of an N-terminal kinase domain and the stretch of 22 amino acids upstream of the first of the kinase catalytic subdomains is similar to other members of the STE20 family of kinases including MST1, NIK, GCK, HPK1, GLK, KHS, and HGK. Furthermore, PSK contains the amino acid sequence GTPY/FWMAPEV in subdomain VIII characteristic for the STE20 family and there is also no evident putative Cdc42/Rac binding domain (CRIB). The C-terminal domain of PSK has no significant homology with any STE20 family member.
The catalytic domain of PSK is most similar to the recently discovered rat TAO1 (thousand and one amino acids) protein kinase (53) with 90% identity over amino acids   (Fig.  3B). The C termini of PSK and TAO1 are 68% homologous between amino acids 278 and 744 and 23% homologous between amino acids 745 and 1001 (Fig. 3B). PSK contains an additional stretch of 15 glutamic acid residues (amino acids 378 -392) that are not present in TAO1, and PSK also contains an additional 234 amino acids at the C terminus. A DNA sequence has recently been released encoding a predicted amino acid sequence for a human protein KIAA0881 (accession no. AB020688) (54) that also has homology to PSK and rat TAO1 (Fig. 3B). To confirm that PSK was distinct from TAO1 and KIAA0881, human genomic DNA was taken and subjected to PCR using a forward primer to nucleotides 2753-2774 of PSK, which are also present in KIAAO881, combined with a reverse primer that was specific for PSK (nucleotides 3180 -3200). The PCR reaction generated a product of approximately 450 bp, which was cloned and sequenced and found to be PSK (data not shown). Two expressed sequence tags representing a single clone from human testis have recently been entered into the GenBank data base (accession nos. AL048834 and AL048835) and encode 5Ј and 3Ј regions of PSK (903-1353 and 4220 -4836 bp). Furthermore, an additional cDNA sequence encoding a potential rat homolog of PSK has been submitted to the GenBank data base during review of this manuscript (TAO2, accession no. AAD39480). This cDNA, however, encodes a much shorter protein of 993 amino acids. Taken together, this information indicates that PSK and TAO1 are distinct genes encoding proteins with closely related kinase domains.
Expression of PSK-Northern blotting analysis, using an XbaI 1105-1800 XhoI DNA fragment of PSK to probe mRNA from eight different tissues, detected a 5-kb transcript in prostate. PSK was ubiquitiously expressed in all of the other tissues examined, and the strongest expression was found in testis (Fig. 4). We also used a BamHI 3504 -4492 XbaI DNA fragment representing the 3Ј end of PSK to detect the same 5-kb transcript in prostate and testis (data not shown).
PSK Is a Protein Kinase-MYC-tagged PSK was transiently transfected into COS1 cells and cell lysates were immunoblotted using antibody against the MYC-epitope tag. Two proteins of 165 and 185 kDa were specifically recognized in cells transfected with full-length PSK but absent from cells transfected with vector alone, suggesting possible post-translational processing or truncation of the protein (Fig. 5). Previous studies on other STE20-like kinases such as FLAG-tagged HGK have also revealed expression of a doublet (55).
To determine whether PSK had kinase activity, MYC-tagged PSK was transiently transfected into COS1 cells and an in vitro immune complex kinase assay was carried out on immunopre- cipitates. Although there was some background kinase activity toward MBP from cells transfected with vector alone, there was a marked increase in MBP phosphorylation in cells transfected with either full-length PSK or C-terminally truncated PSK (1-349), which retains the kinase catalytic domain (Fig. 6A). To eliminate the possibility that an associated kinase may be co-precipitating with PSK, which may account for the kinase activity, we generated kinase-defective mutants PSK (K57A) and PSK (1-349, K57A), which both lack a crucial lysine residue in the second kinase subdomain required for ATP binding. In contrast to our findings using wild type PSK, immunoprecipitates from cells transfected with mutated PSK (K57A) or mutated C-terminally truncated PSK (1-349, K57A) failed to increase levels of MBP phosphorylation above those obtained with vector alone (Fig. 6A). Additional phosphorylated proteins were also observed in the immune complex assay, at 185 and 45 kDa, and ran with the same mobility as full-length and truncated PSK, respectively, consistent with autophosphorylation of these protein kinases (Fig. 6B). This interpretation is supported by the absence of these phosphorylated proteins in cells transfected with kinase-defective PSK (K57A) or PSK (1-349, K57A) (Fig. 6B). These observations demonstrate that PSK and truncated PSK (1-349) act as functional kinases and that the kinase activity observed is attributable to PSK. Furthermore, addition of IL-1, transforming growth factor ␤, or serum to serum-starved cells expressing PSK failed to increase enzymatic activity of the immunoprecipitated protein toward MBP, suggesting that PSK is constitutively activated under these conditions (data not shown).
PSK Activates the JNK MAPK Pathway-To determine whether PSK could activate MAPK pathways, MYC-tagged PSK was co-transfected together with FLAG-tagged JNK into COS1 cells and immune complexes of JNK were prepared from cell lysates for in vitro kinase assays using recombinant c-Jun as a substrate. Fig. 7A shows that transfected wild type PSK and also truncated PSK (1-349) stimulated phosphorylation of c-Jun 17.5-and 11.1-fold, respectively, when compared with cells transfected with JNK alone. In contrast, transfection of kinase-defective PSK (K57A) or PSK (1-349, K57A) causes no significant increase in c-Jun phosphorylation levels. Normalization of samples or expression of each transfected protein was confirmed by immunoblotting either immunoprecipitated FLAG-JNK or cell lysates (MYC-PSKs) (Fig. 7A).
We also looked at effects of transfected PSK or PSK (1-349) on the ERK and p38 MAPK pathways. MYC-PSK was cotransfected with either FLAG-tagged ERK2 or MYC-tagged p38 into COS1 cells. However, immune-complexes prepared from these cells failed to demonstrate any significant effect of PSK or PSK (1-349) on ERK2 activity toward MBP phosphorylation (Fig. 7B) or p38 activity toward ATF2 (Fig. 7C) compared with V12RAS (25.4-fold stimulation) or MLK2 (20.3-fold stimulation), respectively (Fig. 7, B and C). These findings demonstrate that PSK specifically activates the JNK MAPK pathway and has no detectable stimulatory effect on ERK or p38 MAPK pathways.
PSK Activates MKK4 and MKK7-To determine whether PSK could regulate MAPKKs upstream of JNK MAPK, MYCtagged PSK was co-transfected together with FLAG-tagged MKK4 or FLAG-tagged MKK7 into COS1 cells, and immune complexes of MKK4 or MKK7 were used for in vitro kinase assays using recombinant GST-JNK as a substrate. Fig. 7D shows that transfected wild type PSK and truncated PSK (1-349) stimulated both MKK4 and MKK7 phosphorylation of GST-JNK, compared with cells transfected with kinase-dead PSK proteins. In contrast, PSK was unable to stimulate activity of MKK3 toward p38 (Fig. 7D).
PSK Is Localized to a Vesicular Compartment-To determine the intracellular localization of PSK, expression vectors encoding PSK and various mutants were microinjected into the nuclei of Swiss 3T3 cells, and the localization of MYC epitopetagged PSK determined 4 h after microinjection. Wild-type PSK localized predominantly to vesicles in the cytoplasm of the majority of injected cells (Fig. 8a), which in some cells had a tubulovesicular morphology (Fig. 8c). The localization of PSK was not detectably altered by serum starvation or serum restimulation ( Fig. 9a and c), indicating that PSK constitutively associates with this vesicular compartment. The localization of PSK was clearly different from PAK1, which showed a uniform cytoplasmic distribution (Fig. 8i). Interestingly, the C-terminally truncated version of PSK (1-349) showed a diffuse cytoplasmic localization (⌬C-t, Fig. 8e), suggesting that the C-terminal region includes sequences that are required for targeting to the vesicular compartment. Kinase-defective PSK (K57A) showed a variable intracellular localization (KD; Fig. 8g). In approximately 50% of cells, it appeared diffusely cytoplasmic, whereas in the remainder of cells it showed a less homogeneous FIG. 6. PSK has in vitro kinase activity. COS1 cells were transfected with either pRK5-MYC vector alone, V12RAS, and MYC-tagged ERK2 or various MYC-tagged PSK constructs. Proteins were immunoprecipitated with anti-MYC 9E10 antibody and immunoprecipitates subjected to an in vitro kinase assay using MBP as the substrate. A, proteins were transferred to nitrocellulose after 10% SDS-PAGE and expression of MYC-tagged ERK2 or MYC-tagged PSK proteins was confirmed by immunoblot analysis with 9E10 antibody. B, proteins were not transferred to nitrocellulose after 15% SDS-PAGE in order to detect autophosphorylated PSKs. FIG. 7. PSK activates the JNK MAPK pathway. COS1 cells were transfected with either pRK5MYC vector alone or various MYC-tagged PSK constructs, control MYC-tagged V12RAS, or control MYC-tagged MLK2 along with FLAGtagged JNK1 or FLAG-tagged ERK2 or MYC-tagged p38. A, FLAG-tagged JNK was immunoprecipitated with the M2 antibody and immunoprecipitates subjected to an in vitro kinase assay using GST-c-Jun as the substrate. -Fold increases in c-Jun phosphorylation are shown when compared with cells transfected with JNK alone. B, FLAG-tagged ERK2 was immunoprecipitated with the M2 antibody and immunoprecipitates subjected to an in vitro kinase assay using MBP as the substrate. C, MYC-tagged p38 was immunoprecipitated with the 9E10 antibody and immunoprecipitates subjected to an in vitro kinase assay using GST-ATF2 as the substrate. Normalization of immunoprecipitated FLAG-JNK, FLAG-ERK2, or MYC-p38 and expression of MYC-tagged PSK proteins was confirmed by immunoblot analysis. PSK appears as a single band after 15% SDS-PAGE. D, COS1 cells were transfected with either pRK5MYC vector alone or various MYC-tagged PSK constructs along with FLAG-tagged MKK4 (a), MKK7 (b), or MKK3 (c). FLAG-tagged MKKs were immunoprecipitated and subjected to an in vitro kinase assay using GST-JNK or GST-p38 as substrate. MEKK3 and MLK2 were used as positive controls for MKK3 and MKK4 or MKK7, respectively. distribution that nevertheless did not resemble that of wildtype PSK. This implies that autophosphorylation of PSK and/or phosphorylation of a downstream target are important for its localization. We have also analyzed localization of PSK in transfected COS1 cells and observed a similar requirement for the C terminus to localize PSK (data not shown).
PSK Induces Retraction and Actin Reorganization-We observed that cells expressing wild type PSK were clearly altered in morphology: they were more rounded, had fewer processes, and had a smaller spread area than surrounding uninjected cells or cells expressing kinase-defective PSK. To determine whether this cell shape change was reflected in altered actin cytoskeletal organization, we analyzed actin filament distribution in growing Swiss 3T3 cells following microinjection with expression vectors encoding wild-type PSK and various mutants. A marked change in the distribution of actin filaments was observed by 4 h after microinjection (Fig. 8b). Stress fibers were disorganized, and instead of being arranged in parallel bundles, were concentrated in shorter bundles at the periphery of the cell. A similar effect on actin stress fiber organization was observed both in growing Swiss 3T3 cells (Fig. 8b) and in starved cells acutely restimulated with FCS (Fig. 9b). In some cells, a significant decrease in the level of stress fibers was observed, and lamellipodia were not observed on any PSKexpressing cells (Figs. 8b and 9d, and data not shown). These changes were observed in multiple experiments, and with PSK in two different expression vectors, pRK5 and EfPlink. Staining for vinculin, a component of focal adhesions normally found at the ends of stress fibers (56), showed that vinculin-containing adhesions were ro-organized in PSK-expressing cells, such that they were mostly present as small complexes at the periphery of the cells (Fig. 8d), consistent with the reorganization of the actin cytoskeleton. The PSK-mediated loss of processes was clear from the vinculin localization, as control cells had vinculin-containing focal adhesions at the end of processes.
The alteration in cell morphology and actin organization by PSK was dependent on kinase activity, since the kinase-defective mutant PSK (K57A) had no effect (Fig. 8h). In addition, the truncated form of PSK (1-349) did not effect actin organization (Fig. 8f). Taken together with the results on PSK localization, this suggests that correct localization of PSK is important for its effects on the actin cytoskeleton.
Expression of activated PAK kinases has been reported to induce loss of stress fibers in various cell types (42)(43)(44), so we compared the effect of PSK with that of an activated form of PAK1, PAK (L107F) (57). As previously reported in REF52 fibroblasts (42), PAK (L107F) did not induce detectable loss of stress fibers at early time points (4 h) after microinjection of the expression construct, although a decrease in stress fibers was observed at 24 h after injection (Fig. 8j). PAK1 (L107F)expressing cells did not retract, indicating that the mechanism whereby PAK1 alters actin cytoskeletal organization is significantly different from that induced by PSK. DISCUSSION We have used reverse transcription-PCR to identify 15 tyrosine or serine-threonine kinases expressed in surgically removed prostate tumors. These include the receptor kinases IGF1-R, PDGFBR, VEGFR2, EPH-A1, MET, and RYK; the non-receptor kinases JAK1, JAK2, TYK2, ABL, EMK, and PLK; and three novel kinases. Many of these kinases have been shown to have oncogenic potential, and their presence in prostate tumors provides a basis for further analysis of signaling pathways involved in prostate cancer development.
PCR conditions were designed for kinase detection rather than quantitative analysis of kinase expression, but the most frequently sequenced clones using this approach were IGF1-R and PDGFBR. Others have shown that elevated levels of plasma IGF1 are associated with increased risk of prostate cancer development and that IGF1 is able to activate the ERK MAPK pathway in DU145 cells, and acts as a potent mitogen for prostate cancer cells (8,58). Moreover, an IGF1-R autocrine loop supports androgen-independent growth of DU145 prostate cancer cells and antisense to the IGF1-R suppresses growth and invasion of prostate cancer cells (6,59). The presence of IGF1-R in all three of our screened prostatic carcinoma tissues supports the notion that IGF1-R may play a functional role in prostate cancer development. We also detected a number of other receptor kinases in our degenerate PCR screen of prostatic carcinoma tissues including PDGFBR; PDGF acts as a potent mitogen for prostate cells (60) and the oncogenic potential of PDGFBRs has been well characterized in other cancers. VEGFR2 and the HGF receptor c-MET also function to regulate cell growth, migration, and the actin cytoskeleton via autocrine and paracrine loops, and increases in expression of both vascular EGF and c-MET proteins have been reported in malignant prostate cells (61)(62)(63). The EPH-A1 receptor has not yet been investigated in prostate; however, overexpression of the EPH-A1 receptor has been reported in breast, lung, liver, and colon carcinoma tissues (64).  g and h), or pRK5-MYC-PAK1 (L107F) (LF) (i and j). After 4 h, cells were fixed and co-stained with mouse anti-MYC epitope antibodies followed by FITC-labeled anti-mouse IgG to show PSK or PAK1 localization (a, e, g, and i) and with TRITC-phalloidin to show actin filaments (b, f, h, and j); or with rabbit anti-MYC epitope antibodies followed by TRITC-labeled anti-rabbit IgG to show PSK localization (c) and mouse anti-vinculin antibodies followed by FITC-labeled antimouse IgG to show vinculin localization (d). Bar represents 20 m.
Some of the non-receptor kinases detected function as transduction molecules regulating MAPK pathways. c-ABL can regulate the ERK, JNK, and p38 MAPK pathways and the BCR-ABL oncogene induces transformation by activating the STE20-like GCKR (GCK-related) kinase and JNK MAPK pathway in CML cells (65)(66)(67). The Janus family of tyrosine kinases (JAK1-3, TYK2), which are able to induce rapid stimulation of interferon-responsive genes through latent cytoplasmic transcription factors, known as signal transducers and activators of transcription, have not yet been investigated in prostate but can regulate the ERK MAPK cascade in other cell types (68).
A profile of kinases expressed in a prostate carcinoma xenograft grown in mice has recently been published and detected some of the kinases reported here, including PDG-FBR, EPH-A1, ABL, JAK1, and TYK2 (12). MAPK upstream regulators such as MKK3, MKK4, and MKK5 were also present in this malignant prostate xenograft. The different kinase profiles reported by us and by Robinson et al. (12) are most probably accounted for by the use of different degenerate primers and our use of surgically removed prostatic carcinoma tissues rather than a prostate cancer xenograft grown in mice. The number of kinases that we have detected is by no means complete and could probably be increased by using alternative degenerate primers and analyzing additional tumor tissues.
We were particularly interested in the three potentially novel kinases that were detected in our screen. Of these, one is a protein with 65% homology over its entire length to the C. elegans predicted protein C46C2.1 (49); however, none of the cDNAs that we isolated had an in-frame stop codon (data not shown) and although the predicted amino acid sequence appeared to contain all 11 kinase subdomains, a crucial lysine residue required for ATP binding was missing in the second catalytic domain (C190K). The protein could be expressed in COS1 cells, but immunoprecipitated protein was unable to phosphorylate MBP, even after a lysine residue had been introduced into the second catalytic subdomain (data not shown). The sequence of this kinase-defective protein (named KDP) has been submitted to the GenBank data base. (Another of the novel kinases turned out to encode the human homolog of the recently isolated murine protein kinase PKL12, a serine-threonine kinase, which is ubiquitiously expressed but has unknown function (50).) The third novel kinase encodes a new member of the STE20 kinase family, which we have named PSK. STE20 family members are known to be involved in regulating cell growth, transformation, differentiation, stress responses, cytoskeletal organization, cell motility, and apoptosis. They can be classified into two subfamilies according to their structure and regulation. The first group of enzymes includes the PAK and MLK proteins which have a C-terminal kinase domain and an Nterminal regulatory region containing a CRIB (69), while the second group of proteins have an N-terminal kinase domain and a C-terminal regulatory region that does not contain a CRIB. PSK is structurally related to the second group, which includes GCK, GLK, HGK, HPK1, KHS, MST1/2, NIK, and TAO1. Immune complexes of PSK phosphorylated MBP in vitro, demonstrating that it has kinase activity, and transfected PSK specifically stimulated the JNK MAPK pathway but not ERK MAPK or p38 MAPK in COS1 cells. This is similar to the response observed with GLK, HGK, HPK1, KHS, and NIK, which specifically stimulate the JNK pathway through MKK4, with GCK, GLK and NIK acting via MEKK1, HGK via TAK1, and HPK via MLK3 (28,29,37,39,41,55,70). In the case of HPK and NIK, this activity has been attributed in part to direct interaction with MLK3 and MEKK1, respectively, through homologous regions in their non-catalytic C-terminal domains (39). PSK has no homology in its C-terminal region with these kinases, but we have shown that PSK activates MKK4 and MKK7, which regulate the JNK pathway. The C-terminal region of PSK also contains proline-rich motifs, which could potentially bind to SH3 domains, and some of these sites could therefore be important in protein-protein interactions.
Among the STE20 family, PSK shows the closest homology to TAO1, with 90% amino acid identity within the kinase domain (53). Outside the kinase domain, however, homology is substantially reduced. Despite the similarity of their catalytic domains, the specificities of PSK and TAO1 for MAPK pathways appear to differ in cells. Transfected TAO1 interacts with MKK3, apparently through its C-terminal non-catalytic do- FIG. 9. Effects of PSK on serumstarved and serum-stimulated cells. Swiss 3T3 cells were starved for 16 h and then microinjected with pRK5-MYC-PSK (wt). Cells were incubated for 4 h and then fixed (a and b) or stimulated with 10% FCS for 10 min and then fixed. Cells were co-stained with mouse anti-MYC epitope antibodies, followed by FITC-labeled anti-mouse IgG to show PSK (a and c) and with TRITC-phalloidin to show actin filaments (b and d). Bar represents 20 m. main, and stimulates MKK3 but not MKK4 activity, which would normally be expected to activate p38 but not JNK. Cterminally truncated TAO1 (1-416) can, however, stimulate MKK4 activity in vitro (53), suggesting that specificity of MAPK pathway activation in cells is mediated by selective interactions of the C-terminal domain. Indeed, we observe that the C-terminal region is essential for localization of PSK to a vesicular cytoplasmic compartment, and this may be important for its function. Our finding that transfected PSK, unlike TAO1, specifically activates JNK does not, however, appear to be solely due to the different C termini of PSK and TAO1, since truncated PSK (1-349) containing the N-terminal kinase domain was also able to activate JNK, albeit less potently than full-length PSK. The inability of Hutchison et al. (53) to coexpress TAO1 with JNK or p38 in 293 cells makes further comparisons with PSK difficult at the present time.
Of the STE20-like kinases, only PAKs have so far been shown to induce changes in actin organization in mammalian cells. PAKs interact with the GTPases Cdc42 and Rac, which regulate the formation of filopodia and lamellipodia, respectively (56). Various mutated forms of PAKs can induce loss of actin stress fibers and focal complexes, stimulate formation of filopodia and lamellipodia, and/or block Cdc42 and Rac effects on the cytoskeleton (42)(43)(44)71). In contrast, MLK2 and MLK3 also interact with Cdc42 and Rac but do not regulate the actin cytoskeleton (34,40). The cell retraction, loss of processes, and actin reorganization observed with PSK is significantly different from the loss of stress fibers induced by activated PAK1. Stress fibers consist of actin and non-muscle myosin II filaments, and increased MLC phosphorylation normally correlates with an increase in the number and/or contractility of stress fibers (72). Both MLC and MLC kinase have been reported to be targets for PAKs (73)(74)(75). Expression of activated PAK1, however, can lead to either an increase or decrease in MLC phosphorylation depending on the system (44,74). We have found that immunoprecipitated PSK does not phosphorylate purified myosin light chain kinase in vitro nor does transfected PSK alter phosphorylation of MLC on serine 19 following analysis by immunoblotting of COS1 cell lysates, whereas expression of activated RhoA does induce increased MLC phosphorylation, as previously reported (76) (data not shown). An alternative possibility for how PSK acts is suggested by its localization to cytoplasmic vesicles. As its ability to induce reorganization of stress fibers correlates with this localization, it may play a role in trafficking of proteins involved in actin organization or focal adhesion formation. Indeed, there is good evidence that some focal adhesion components, including paxillin, are normally associated with intracellular vesicles and need to be transported to the plasma membrane in order for focal adhesions and associated stress fibers to form efficiently (77).
In conclusion, we have demonstrated for the first time that a STE20-like kinase which lacks a CRIB, PSK, can not only selectively activate the JNK MAPK pathway but also affects actin cytoskeletal organization. By coordinating these two responses, PSK may play a role in regulating cell migration.