PSTPIP: A Tyrosine Phosphorylated Cleavage Furrow–associated Protein that Is a Substrate for a PEST Tyrosine Phosphatase

Yeast two-hybrid screening was performed essentially as described in Chien et al. (1991) and Bartel et al. (1995). A C₂₂₁-S active site mutant of PTP HSCF (Cheng et al., 1996) was cloned in frame with the Gal binding domain in the plasmid pPC97. A library of 6 × 10⁶ individual cDNA clones was produced from Baf3 lymphoid progenitor cells in the Gal activation domain plasmid, pPC86, using standard procedures. Yeast were transformed with both plasmids and were incubated on histidine minus plates for 3 d at 30°C. Colonies that grew under these conditions were restreaked onto histidine minus plates and were tested for β-galactosidase activity (Bartel et al., 1995). Colonies that manifested various levels of β-galactosidase activity were isolated, and the cDNA inserts in the pPC86 vector were isolated by PCR and sequenced using standard procedures. Clones encoding PSTPIP were tested for dependence on the PTP interaction by transfection into cells with and without the original PTP HSCF– containing pPC97 plasmid and subsequent analysis for growth on histidine minus plates and β-galactosidase activity.

To obtain a cDNA encoding full-length PSTPIP tagged with the FLAG epitope (DYKDDDDK) at the COOH terminus, PCR was performed using the 5′ PSTPIP–specific primer 48.BAMHI.F (CGCGGATCCACCATGATGGCCCAGCTGCAGTTC) and the 3′ PSTPIP–specific primer 48.SAL.FLAG.R (GTACGCGTCGACTCACTTGTCATCGTCGTCCTTGTAGTCGAGCTT). The resulting pcr fragment was digested with BamHI and SalI, and subcloned into the BamHI and SalI sites of pRK.tkneo, an expression plasmid containing the cytomegalovirus promoter, thus creating plasmid pRK.PIP.FLAG.C. The PTP HSCF deletion mutants were derived from a construct containing the influenza hemagglutinin epitope at its NH₂ terminus and were made as follows: PCR was performed on PRK.HSCF using primers prkr (TGCCTTTCTCTCCACAGG) and 38.SPE.mid.R (CTCCTTGAGGTTCTACTAGTGGGGGCTGGTGTCCTG). The resulting pcr fragment encoding the phosphatase domain (amino acids 1–302) was digested with ClaI and SpeI and subcloned into pRK.tk.neo digested with ClaI and XbaI resulting in plasmid pRK.HSCF.PTP domain. Similarly, a PTP HSCF cDNA missing the COOH-terminal homology (CTH) domain was produced by PCR using primers prkr and 39.SPE. endR (GCGGCCGCACTAGTATCCAGTCTGTGCTCCATCTGTTAC), and the resulting fragment encoding amino acids 1–430 of PTP HSCF was digested with ClaI and SpeI and subcloned into the ClaI and XbaI sites of pRKtkneo resulting in pRK HSCFΔ24. Glutathione-S-transferase (GST) fusion proteins were prepared essentially according to the manufacturer (Pharmacia LKB Biotechnology, Inc., Piscataway, NJ) in DH5-α bacterial cells. A SalI to NotI fragment containing the full-length cDNA for PSTPIP (amino acids 2–415) was subcloned into pGEX-4T-2 (Pharmacia LKB Biotechnology Inc.) cleaved at the SalI and NotI sites. To obtain a DNA fragment encoding the coiled-coil domain of PSTPIP, PCR was performed using primers PC86F (GCGTTTGGAATCACTAC) and pip48.1706R (TTATAGTTTAGCGGCCGCTCACCGGTAGTCCTGGGCTGATG). The PCR fragment was digested with SalI and NotI and subsequently cloned into the SalI and NotI sites of pGEX-4T-2. To obtain a cDNA fragment encoding the SH3 domain of PSTPIP, PCR was performed using primers pip48.1673.F (GTACGCGTCGACCGCACTCTACGACTACACTGCACAG) and PC86R (CTCTGGCGAAGAAGTCC), and the resulting product was digested with SalI and NotI and subcloned into the SalI and NotI sites of pGEX-4T-2. To obtain a cDNA fragment encoding the PST-rich region and CTH domain of PTP HSHCF (amino acids 304–453), PCR was performed using primers PST38-RI (GATCGAATTCCCAGAACCTCAAGGAGAACTGC) and PST38-XHOI (GATCCTCGAGTTACACCCGTGTCCACTCTGCTGGAGGA). The resulting pcr product was digested with EcoRI and XhoI, and then subcloned into the EcoRI and Sal sites of pGEX-4T-2. Protein determinations were carried out according to the Couprus assay with a kit from Geno Technology (St Louis, MO). The binding was carried out according to the method of Wong and Johnson (1996). Briefly, 1 μg of plasmid with either the PSTPIP protein or HSCF PTP under the control of the Sp6 promoter was in vitro–transcribed/translated using the Promega TnT Rabbit Reticulocyte system (Promega Corp., Madison, WI). Samples were diluted in 50 mM Hepes, pH 7.2, 1% Triton X-100, 10% glycerol, 100 mM NaCl, 5 mM EDTA, and 2 μm/ml each of leupeptin, pepstatin, aprotinin, and PMSF. Samples were pre-cleared with resin for 1 h and 1 μg GST fusion protein was added along with 30 μl of GSH–Sepharose that was previously blocked in 3% BSA for 1 h. This was reacted for 1 h at 4°C and then the resin washed six times in Hepes/Triton X-100 binding buffer before SDS gel electrophoresis. The peptides were synthesized on an automated Milligen 9050 Peptide Synthesizer (Applied Biosystems, Inc., Foster City, CA) using standard solid phase chemistry with FMOC-protected amino acids on a p-alkoxybenzyl alcohol resin. Dried peptides were resuspended in the Hepes/Triton X-100 binding buffer at a concentration of 10 mg/ml. Peptide inhibition was performed by adding the peptide first to the in vitro translation product, and then to the GST fusion, followed by the GSH–Sepharose. The binding/washing steps were done as previously described. The peptides synthesized and the PTPs they were derived from were: PXXP-HSCF: ₄₃₂GFNLRIGRPKGPRDPPAEWT₄₅₁ (PTP HSCF), PXXP-PEP:₇₈₂GFG-NRFSKPKGPRNPPSAW₈₀₀ (PTP PEP), PXXP-PEST:₇₆₁GFGNRC-GKPKGPRDPPSEWT₇₈₀ (PTP PEST), PXXP-CONTROL:₃₃₄GGVLRS-ISVPAPPTLPMADT₃₅₃ (PTP HSCF).

Baf3 or PSTPIP-transfected COS cells were lysed in 1% Triton X-100, 50 mM Hepes, pH 7.2, 10% glycerol, and 5 mM EDTA containing 1 μm/ml aprotinin, PMSF, leupeptin, and pepstatin with 1 mM sodium vanadate, and 10 mM iodoacetic acid. A portion of cells were pretreated with 0.1 mM pervanadate for 4 h before lysis. Immunoprecipitations were performed in the vanadate-containing lysis buffer using 1 μg/ml anti-PSTPIP polyclonal antibody and 400 μg of lysate protein at 4°C overnight. Western blots were performed using 1 μg/ml affinity-purified anti-PSTPIP or 2 μg/ml of commercial 4G10 anti-phosphotyrosine monoclonal (Upstate Biotechnology Inc., Lake Placid, NY). Signal was detected by ECL (Amersham Corp., Arlington Heights, IL) reagents (Pierce Chemical Co., Rockford, IL). The C₂₂₁-S mutant was as previously described (Cheng et al., 1996). The PTP HSCF D₁₉₇-A mutant was generated using PCR. Mutagenesis primer D197A.F (GTATATGTCCTGGCCAGCCCATGGGGTTCCCAGCAG), corresponding to nucleotide 591, and primer D197A.R (GCAGGTCGACTCTAGATTACACCCGTGTCCACTCTG), which corresponds to the stop codon, were used in PCR to generate a fragment that could be cut with MscI and XbaI. pRK.HA.38 WT, a plasmid that encoded the wild-type enzyme under the control of the cytomegalovirus promoter (Cheng et al., 1996), was digested with ClaI and MscI and the resulting 600-bp fragment was ligated with the MscI-XbaI pcr fragment into the ClaI and XbaI sites of pRK.tkneo. A plasmid encoding the v-Src oncogene under the control of the Sv40 early promoter was the kind gift of A. Levinson (Genentech, Inc., South San Francisco, CA). National Institutes of Health 3T3 cells and COS-7 cells were cultured in high glucose DME supplemented with 10% FBS, 2 mM l-Glutamine, 10 mM Hepes, pH 7.2, and pen-strep. COS-7 cells were transfected by electroporation. Briefly, 1.5 × 10⁻⁶ COS-7 cells were mixed with 24 μg total DNA in PBS and electroporated at 960 microfarod, 0.22 V (Gene Pulsar; Bio Rad, Hercules, CA). After electroporation cells were seeded in 10-cm dishes and incubated for 3 d. 10-cm dishes of transfected COS cells were washed twice with ice-cold PBS, and then lysed in 1 ml of M-RIPA (50 mM Tris 7.4, 1% NP-40, 0.25% deoxycholate, 150 mM NaCl, 1 mM sodium ortho-vanadate, 1 mM NaF plus Complete^TM Protease Inhibitors [Boehringer Mannheim Biochemicals, Indianapolis, IN]). Lysates were incubated for 15 min with 100 μl UltraLink Immobilized Protein A/G (Pierce Chemical Co.) at 4°C, followed by centrifugation for 5 min. Supernatants were collected and stored at −70°C, or directly immunoprecipitated. 5 μg of M2 (Eastman Kodak, Inc., Rochester, NY) or 12CA5 (Boehringer Mannheim, Inc.) was added to 500 μl of lysate and incubated overnight at 4°C. Ultralink Protein A/G was added and incubation continued for 2 h at 4°C. The immune complexes were washed three times with M-RIPA. The proteins were subjected to SDS-PAGE and transferred to nitrocellulose in 2× transfer Buffer, 20% methanol (Novex, San Diego, CA). Immunoblots were blocked at room temperature for 1 h in 3% milk/ PBS. To detect FLAG-tagged PIP, blots were incubated overnight with 10 μg/ml Bio-M2 (Biotinylated anti-FLAG monoclonal Ab; Eastman Kodak), followed by incubation in 10 μg/ml streptavidin-HRP (Upstate Biotechnology Inc.). To detect hemagglutanin (HA)-tagged PTPhscf, blots were incubated in anti-HA-peroxidase (Boehringer Mannheim, Inc.) as per manufacturer's instructions. To detect phosphotyrosine, blots were incubated in HRP-conjugated 4G10 (anti-phosphotyrosine monoclonal; Upstate Biotechnology, Inc.) as per manufacturer's instructions.

Rabbit polyclonal antibodies were produced against a PSTPIP–GST fusion protein. The complete PSTPIP–GST fusion protein was purified on GSH-Sepharose and injected intramuscularly at two sites with 200 μg fusion protein, and subcutaneously at multiple sites with a total of 300 μg PSTPIP–GST fusion protein in Complete Freunds Adjuvant (Sigma Chemical Co., St. Louis, MO). Rabbits were boosted every 3 wk with 100 μg fusion protein in Incomplete Freunds. 15 ml of rabbbit sera was reacted with 0.5 mg PSTPIP-GST-GSH-Sepharose for 3 h at 4°C with gentle rotation. The resin was collected by centrifugation and washed with 10 column volumes of PBS. Immunoglobulin was eluted from the affinity matrix with 100 mM acetic acid, 500 mM NaCl, neutralized with NaOH, and then dialyzed overnight with PBS. NIH 3T3 cells were seeded at 100,000 cells per chamber slide and allowed to adhere overnight. The cells were transfected using Lipofectamine (2 μg pRK.PIP.FLAG.C/12 μl lipofectamine in 0.8 ml OPTI-MEM) for 5 h. The DNA/Lipofectamine solution was removed and fresh serum-containing medium added. 48 h after the start of transfection, the cells were fixed in 4% formaldehyde in PHEM 6.1 (60 mM Pipes, 25 mM Hepes, 10 mM EGTA, and 2 mM MgCl₂) for 20 min, and then permeabilized in 0.2% Triton X-100, 300 mM sucrose in PHEM 6.9 for 10 min. The cells were washed twice in PHEM 6.9 and then incubated with 10% FBS/PHEM 6.9 for 1 h to block nonspecific binding of the antibody. Cells were incubated for 1 h in 2% BSA/ PHEM 6.9 containing 10 μg/ml M2 (anti-FLAG monoclonal antibody; Eastman Kodak) or 10 μg/ml 12CA5 (anti-HA monoclonal antibody; Boehringer Mannheim Biochemicals) as an irrelevant antibody control. After washing cells twice with 2% BSA/PHEM 6.9, cells were incubated for 30 min with a 1:2,000 dilution of Cy3-conjugated AfinniPure sheep anti–mouse IgG and a 1:200 dilution of fluorescein-phalloidin (Molecular Probes Inc., Eugene, OR) in 2% BSA/PHEM 6.9. Cells were washed in 2% BSA/PHEM 6.9 and mounted in Vectashield Mounting Medium with DAPI. NIH 3T3 cells were seeded at 200,000 cells per chamber slide and allowed to adhere overnight. Cells were stained with 0.4 μg/ml rabbit anti-PIP or 0.4 μg/ml rabbit IgG and detected with Cy3-conjugated goat anti– rabbit. Additionally, cells were costained with a 1:200 dilution of fluorescein-phalloidin. The samples were examined in a confocal microscope (model 2001; Molecular Dynamics, Inc., Sunnyvale, CA); and analyzed with the ImageSpace software (Molecular Dynamics, Inc).

S. pombe were grown according to standard methods (Moreno et al., 1991). Cells were fixed and stained as for F-actin as described in Marks and Hyams (1985). The PSTPIP gene was expressed from nmt 1 promoter plasmid, REP3 (Basi et al., 1993). For induction, a culture in mid-exponential growth in minimal medium containing 2 mM thiamine was washed twice, and then reinoculated into fresh medium without thiamine to induce expression from the nmt1 promoter at 25°C. Cell number was determined with a CASY-1 cell counter (Coulter Immunology, Hialeah, FL).

To identify potential substrates for PTP HSCF (Cheng et al., 1996), we performed a yeast two-hybrid screen using a catalytically inactive form of the enzyme as bait and a library derived from murine Baf3 hematopoietic progenitor cells, a cell type that has been previously demonstrated to express high levels of this phosphatase (Cheng et al., 1996). The catalytically inactive form of the enzyme was used to decrease the potential toxicity of the overexpressed protein for the cells. This resulted in the isolation of ∼70 yeast clones that grew in the absence of histidine and expressed variable levels of β-galactosidase. Sequence analysis of the clones revealed that ∼40% encoded related sequences with slightly divergent 5′ in-frame fusions with the Gal 4 activating domain. The sequences of the remainder of the clones suggested that they were likely due to artifactual interactions. Analysis of histidine growth and β-galactosidase expression of all two-hybrid clones containing these related sequences revealed an absolute dependence on the inclusion of the phosphatase bait construct in the same cells (data not shown). The longest two-hybrid clone was used to isolate a full-length cDNA from the original Baf3 two-hybrid library.

Fig. 1 illustrates that the protein that interacts with PTP HSCF is a novel 415-residue molecule (predicted molecular weight of ∼47,590) with significant sequence homology to the S. pombe cell cycle protein, CDC15p, a cytoskeletal interacting protein involved with organization of the actin ring at the cleavage furrow during cytokinesis (Fankhauser et al., 1995). This homology (∼26% sequence similarity) stretches over the entire length of both molecules, with the exception of an insertion of ∼500 residues in the yeast molecule, and the yeast protein is the highest scoring homologue in the protein sequence database. A number of features are conserved in these two proteins. For example, both have an SH3 domain at their COOH termini (Pawson, 1995; Feng et al., 1995), and the mammalian SH3 domain appears to be homologous to those found in a number of known cytoskeletal regulatory proteins including myosin heavy chain, spectrin, fodrin, hematopoietic specific protein (HSP) and cortactin (Fig. 1). In addition, both the mammalian and yeast proteins contain a potential coiled-coil domain at their NH₂ termini, which is predicted both on the basis of sequence homology as well as an analysis of the mammalian sequence using the “Coil” program (data not shown). Within these coiled-coil domains is a region with an extraordinary content of acidic and basic residues (positions 99–180 of the mammalian protein). Because the mammalian protein was isolated on the basis of an interaction with a tyrosine phosphatase, it is possible that the protein is tyrosine phosphorylated (see below), and examination of the mammalian and yeast sequences revealed seven conserved tyrosine residues (positions 53, 144, 191, 287, 363, 367, and 369 of the mammalian protein) in addition to a number of non-conserved tyrosine residues. Finally, examination of the proteins for proline-rich regions that might function as SH3 binding sites (PXXP) revealed one such conserved site in these proteins (starting at position 278 of the mammalian protein) (Pawson, 1995; Feng et al., 1995). Cortactin (Wu et al., 1991) and HSP 1 (Kitamura et al., 1989) are two other mammalian proteins that contain potential coiled-coil and SH3 domains that also bear a more distant relationship to the PTP-interacting protein, although both these proteins contain homologous 37–amino acid repeats, which are absent from the interacting protein. Because the mammalian sequence was isolated based upon its ability to interact with the PEST phosphatase PTP HSCF, it has been termed PSTPIP.

Northern blot analysis of the expression of PSTPIP during embryogenesis and in adult tissues is illustrated in Fig. 2. During embryogenesis, the transcript is most highly expressed in the day 7 embryo. The transcript is expressed at relatively high levels in adult lung and spleen and at lower levels in testis, muscle, kidney, brain, and heart. However, the interacting protein is at far lower levels than actin, since the actin blots were exposed for 4 h versus the 1 wk exposure for the PSTPIP blots. Previously, we and others have demonstrated that PTP HSCF is also expressed in both adult lung and kidney (Cheng et al., 1996; Huang et al., 1996).

To characterize the regions involved with the binding between PTP HSCF and PSTPIP, a rapid and direct in vitro binding assay was performed. In this assay, various GST fusions of either the phosphatase or the interacting protein were used to precipitate in vitro translation products of the cognate binding proteins. Fig. 3 illustrates that precipitation of in vitro–translated PTP HSCF by GST fusion proteins containing various SH3 domains as well as full-length PSTPIP demonstrated a high degree of specificity in the interaction between the GST PSTPIP and the phosphatase. The figure also illustrates that at this concentration of GST fusion protein (∼1 μg/ml or ∼1.5 μM), the PSTPIP fusion protein appeared to be more efficient at precipitating the phosphatase than a polyclonal antibody directed against the enzyme or a monoclonal directed against a hemagglutinin tag at the PTP NH₂ terminus (data not shown). This result is consistent with a relatively high affinity interaction between the GST PSTPIP and the in vitro–translated PTP HSCF.

The region of PTP HSCF that interacts with PSTPIP was identified by producing deletion mutants of the enzyme missing either the 24–amino acid CTH domain, which is highly conserved in all of the PEST PTPs (Matthews et al., 1992; Yang et al., 1993; Aoki et al., 1996; Cheng et al., 1996; Dosil et al., 1996; Huang et al., 1996; Kim et al., 1996) or both this domain as well as the longer proline, serine, threonine (PST)-rich region COOH terminal to the catalytic domain. Fig. 4 reveals that deletion of the COOH-terminal 24–amino acid homology domain of PTP HSCF completely abolished the interaction between these two proteins. Because this region is conserved in all PEST PTPs, it is possible that both PTP PEST (Yang et al., 1993) as well as PTP PEP (Matthews et al., 1992) also interact with PSTPIP. To examine this possibility, and also to examine if the COOH terminal region is sufficient for this interaction, 20-residue long peptides derived from the homologous COOH terminal domains of three PEST PTPs were used to compete with the interaction between PTP HSCF and PSTPIP. In this form of the assay, a GST fusion derived from the PST-rich and COOH CTH regions of the phosphatase was used to precipitate in vitro– translated PSTPIP in the presence of varying amounts of peptides (Fig. 4,C). Fig. 4 illustrates that all three peptides effectively block the interaction at concentrations as low as ∼800 nM, while a control peptide derived from a different proline-rich region of PTP HSCF is unable to block the interaction. These data suggest that this small proline-rich region of the PEST PTPs is sufficient for mediating the high affinity interaction between the phosphatase and PSTPIP, and furthermore indicate the possibility that all of these PTPs may interact with PSTPIP via their CTH domains.

To examine the region of PSTPIP that interacts with the COOH-terminal homology region, GST fusions containing either the SH3 domain or the coiled-coil domain of the interacting protein were used to immunoprecipitate in vitro–translated PTP HSCF. The COOH terminal homology region that interacts with PSTPIP contains two overlapping consensus SH3 (PXXP) binding sites, consistent with the possibility that the phosphatase-PSTPIP interaction was an SH3-type binding event (Pawson, 1995; Feng et al., 1995). However, the affinity of the interaction as measured in the peptide experiment described above was significantly greater than many of those previously reported for SH3 domain–PXXP interactions (Feng et al., 1995), and as Fig. 5 illustrates, the interaction between these proteins was surprisingly mediated by the coiled-coil domain and not the SH3 region. This outcome is consistent with the results of the two-hybrid clones, all of which began at a site very close to the NH₂ terminus of the coiled-coil domain, suggesting that the PSTPIP site that interacts with the COOH terminal proline-rich domain requires the NH₂ terminus. Finally, Fig. 5 also illustrates that these two proteins interact in vivo in transfected COS cells, confirming the in vitro binding data. Thus, these data define a novel high affinity in vivo interaction between the COOH-terminal proline-rich domain of PTP HSCF, and the potential coiled-coil domain of PSTPIP.

The association between PTP HSCF and PSTPIP suggested that the interacting protein might be a substrate for the phosphatase. In addition, the conservation of a number of tyrosines between PSTPIP and the phosphorylated CDC15 protein was also consistent with the possibility that the interacting protein was tyrosine phosphorylated. As Fig. 6,A demonstrates, endogenous PSTPIP is indeed tyrosine phosphorylated in Baf3 cells in the presence of the tyrosine phosphatase inhibitor vanadate (Dixon, 1995), consistent with the supposition that the protein is dephosphorylated in vivo by a PTP, possibly PTP HSCF which is coexpressed in these cells (Cheng et al., 1996). In addition, tyrosine phosphorylation of transfected PSTPIP can also be demonstrated to occur in COS cells in the presence of vanadate (Fig. 6 A). A potential tyrosine kinase that might phosphorylate PSTPIP in vivo is one of the Src family kinases. Previous data suggested that the v-Src tyrosine kinase is associated with the cytoskeleton, modulates cytoskeletal elements that resulted in profound morphological changes (Cooper et al., 1993; Kaplan et al., 1994; Thomas et al., 1995), and mediates the tyrosine phosphorylation of cortactin (Wu et al., 1991; Okamura et al., 1995; Vuori et al., 1995; Dehio et al., 1995), an SH3-, coiled-coil–containing actin binding protein that bore a distant structural similarity to PSTPIP. In addition, HSP 1, another SH3-containing protein that is also structurally similar to PSTPIP, is tyrosine phosphorylated by various Src family kinases (Yamanashi et al., 1993; Nada et al., 1994; Takemoto et al., 1995; Takemoto et al., 1996). These results implied that v-Src, a constituitively active form of the enyzme, might mediate the tyrosine phosphorylation of PSTPIP, thus allowing for an analysis of the possible substrate interactions between the interacting protein and PTP HSCF. To test this possibility, PSTPIP was transfected into COS cells together with the v-Src tyrosine kinase and either wild-type or dominant-negative forms of PTP HSCF. Dominant-negative phosphatases were produced by mutating either the active site cysteine to a serine (C₂₂₉-S), which abolishes the ability of the enzyme to form a covalent transition state intermediate with the phosphate attached to the tyrosine, or by mutating a critical active site aspartate residue to alanine (D₁₉₇-A), which inhibits the catalytic removal of the phosphate (Dixon, 1995; Jia et al., 1995; Garton et al., 1996). In both cases, these mutants will tightly bind to the substrate but not dephosphorylate it, with the result that the substrate will be hyperphosphorylated. This procedure has been previously used to characterize substrates for a number of different PTPs, including PTP PEST (Garton et al., 1996) and PTP SHP-2 (Herbst et al., 1996), and it has revealed that these mutant enzymes show exquisite substrate specificity in vivo.

As can be seen from Fig. 6, B and C, both PTP HSCF as well as PSTPIP are tyrosine phosphorylated in response to v-Src cotransfection. Transfection of wild-type, C-S, and D-A mutants of PTP HSCF together with v-Src (Fig. 6,B) demonstrates that the mutant forms of the enzyme were hyperphosphorylated, while the wild-type PTP was not; this is consistent with the enzyme being a substrate for v-Src tyrosine phosphorylation. In addition, because only the dominant-negative forms of the enzyme mediated PTP HSCF hyperphosphorylation, these results also suggest that PTP HSCF is a substrate for its own catalytic activity. Fig. 6,B also illustrates that PSTPIP coprecipitates with the dominant-negative forms of the enzyme and is hyperphosphorylated, consistent with the conclusion that PSTPIP is a substrate for the catalytic activity of PTP HSCF. Fig. 6,C illustrates that transfection of the wild-type PTP HSCF into PSTPIP and v-Src–expressing cells resulted in a decreased level of tyrosine phosphate on PSTPIP, consistent with the in vivo removal of the phosphate from PSTPIP tyrosines by the phosphatase, a result that would be expected if the interacting protein were a substrate for the enzyme. Even more compellingly, Fig. 6 C also illustrates that cotransfection of either dominant-negative form of PTP HSCF into PSTPIP and v-Src–transfected cells resulted in a dramatic increase in the levels of tyrosine phosphate on the interacting protein as well as on the coprecipitating PTP HSCF. Analysis of total tyrosine-phosphorylated proteins in transfected cells demonstrated that the dominant-negative forms of PTP HSCF showed complete substrate specificity for both PTP HSCF itself and PSTPIP (data not shown). These data thus confirm the conclusion that PSTPIP interacts with PTP HSCF in vivo, and they suggest that this interaction allows the phosphatase to dephosphorylate tyrosine residues modified by the v-Src kinase. Moreover, they imply that PTP HSCF is a v-Src substrate for tyrosine phosphorylation, and that these phosphotyrosine residues are substrates for the catalytic activity of the phosphatase. In addition, because tyrosine-phosphorylated PSTPIP was only observed in cells that were transfected with v-Src, these results also suggest that the dramatic overexpression of the protein in these cells may have overwhelmed the endogenous tyrosine phosphorylation mechanism.

The in vitro mapping analysis described previously suggested that PSTPIP interacted with PTP HSCF via the COOH-terminal 24–amino acid homology domain found in all PEST PTPs. Previous work in the PEST PTP system demonstrated that p130^cas was a substrate for this enzyme, and the recognition of this substrate appeared to be predominantly dependent upon the catalytic domain (Garton et al., 1996). To test for the dependence of the PTP HSCF CTH domain on PSTPIP substrate recognition, wild-type and dominant-negative forms of the enzyme were produced that lacked the COOH terminal 24 amino acids. These deletion mutants were then tested for their ability to hyperphosphorylate PSTPIP in the presence of v-Src. Fig. 7 illustrates that the deletion of the COOH CTH domain of the two catalytically inactive, dominant-negative PTPs resulted in a complete lack of hyperphosphorylation of PSTPIP in the presence of v-Src. In addition, these experiments confirm the in vitro mapping studies by demonstrating that PSTPIP cannot interact with forms of PTP HSCF that are missing the COOH CTH domain. These data thus indicate that the interaction between PSTPIP and the dominant-negative forms of PTP HSCF that mediates the hyperphosphorylation of the interacting protein is not due to the catalytic domain, as is the case for the hyperphosphorylation of p130^cas by dominant-negative PEST PTPs, but is instead induced by the interaction between the coiled-coil domain and the COOH-terminal homology region.

S. pombe CDC15p is associated with the cleavage furrow localized F-actin during cytokinesis (Fankhauser et al., 1995). To analyze the subcellular localization of endogenous PSTPIP, 3T3 cells were stained with an affinity-purified polyclonal antibody directed against a GST fusion of the protein and were imaged using confocal microscopy. Fig. 8 illustrates that the interacting protein is colocalized to several F-actin–containing sites in the cell. A majority of the protein appears to be associated with the cortical actin cytoskeleton. The protein also appears to colocalize with the actin stress fibers as well as in lamellipodial regions of the cell. In addition, transfection of PSTPIP into CHO cells revealed expression at sites of focal contact (data not shown). These results are in contrast with the PSTPIP-related protein cortactin, which shows localization on cortical actin and at the ends of the stress fibers but not the fibers themselves (Wu et al., 1991). These data clearly suggest that PSTPIP is associated with cytoskeletal actin during interphase.

Importantly, examination of cells undergoing cytokinesis reveals that endogenous PSTPIP is predominantly associated with the cleavage furrow (Fankhauser et al., 1995; Fishkind and Wang, 1995). As Fig. 8 shows, both PSTPIP and the actin ring colocalize to this region of the dividing cells. This figure also illustrates that the PSTPIP in the cleavage furrow is predominantly associated with the membrane-bound F-actin, which acts to constrict the cleavage furrow (Fishkind et al., 1995), and examination of sections taken perpendicular to the cleavage furrow support this, showing a donutlike structure containing both PSTPIP and actin attached to the constricting plasma membrane of the cleavage furrow (data not shown). It also appears from this figure that much of the cortically associated actin and PSTPIP migrate to the cleavage furrow during cytokinesis, a result that is similar to that observed for yeast CDC15p and actin (Fankhauser et al., 1995). These subcellular localization data are thus consistent with the conclusion that PSTPIP is an actin binding protein that is potentially involved with the regulation of the cleavage furrow.

One role that PSTPIP might play in the cleavage furrow is the reorganization of polymerized actin (Cao et al., 1990a; Fishkind and Wang, 1993, 1995). To examine the possible function of PSTPIP in actin assembly, 3T3 cells were transfected with an epitope-tagged version of the protein under the control of the powerful cytomegalovirus promoter, and the transfected cells were subsequently examined for expression of transfected PSTPIP as well as F-actin. As can be seen in Fig. 9, 3T3 cells with normal morphology that expressed transfected PSTPIP showed colocalization of the protein at the cortical surface with F-actin as well as in lamellipodial structures and the F-actin stress fibers; this is in agreement with data obtained examining endogenous PSTPIP localization (Fig. 8). In addition, this figure shows that PSTPIP colocalizes with PTP HSCF in cotransfected 3T3 cells. Fig. 9 also illustrates that the overexpression of the protein often induced a remarkable morphological change in a high percentage of 3T3 cells. These cells contained extended, filopodial-like structures that were filled with polymerized actin. In many cases, the structures were up to ∼150 μm in length, and they often showed a knoblike morphology. In addition, the majority of cells contained a single extended filapodial structure. It appears that this structure was probably produced in the absence of significant cell growth or plasma membrane synthesis, since the overall size of the cell body appeared to decrease dramatically concomitant with the lengthening of the filapodial structure. This type of cell morphology is never observed with transfection of the green fluorescent protein (data not shown), and Fig. 9 illustrates that it is very different from the morphology of normally elongated, nontransfected cells. In summary, these results suggest that the unregulated expression of PSTPIP in vivo results in the induction of extended filopodial-like structures, consistent with the possibility that the overexpressed protein may induce an inappropriate polymerization of the cortical cytoskeleton.

To examine the function of PSTPIP in a readily controlled, genetically defined in vivo system, we overexpressed the protein in the fission yeast S. pombe using the thiamine inducible promoter system (Basi et al., 1993). Initial studies demonstrated that the mammalian protein was incapable of rescuing previously described CDC15 null mutants (Fankhauser et al., 1995). After induction of PSTPIP expression in wild-type cells, cell number increased more slowly than in nonexpressing cells, but division did not cease. The doubling time increased from 3 to 3.5 h. However, we noted a significant increase in the number of septated cells one generation after induction (the promoter takes three and one-half generations to induce). The number of septated cells increased from ∼12% in the uninduced control, to ∼36% (Fig. 10, 5 gen). At later times, a second phenotype appeared: cells did not cleave, but resumed growth from the tips, becoming elongated (Fig. 10, 6 gen, right-hand cell). These cells then formed a pair of septa after mitosis, producing a cell with three septa separating four nuclei (Fig. 10, 6 gen, left-hand cell). Approximately one-third of the septated cells displayed this phenotype. The block to cell cleavage was not absolute, since the number of septated cells did not exceed 36%, cell number continued to increase, though at a reduced rate, and cells were seen to cleave (Fig. 10, 8 gen, right-hand cell). Actin staining of cells overexpressing PSTPIP showed that in interphase, it is seen as dots at the tips, while in mitosis, it forms a ring, then associates with the growing septum, as expected (data not shown). FLAG staining for PSTPIP showed that the epitope localizes to the tips in interphase and to the cell equator at mitosis (Fig. 10). It thus has very similar localization to actin, in support of the data obtained for PSTPIP localization in mammalian cells. These data are thus consistent with the suggestion that mammalian PSTPIP acts as a dominant-negative inhibitor of cytokinesis in S. pombe.

We have isolated a novel member of the actin-associated protein family, PSTPIP, which binds to the PTP HSCF tyrosine phosphatase via an interaction between the proline-rich CTH domain of the PTP and the potential coiled-coil domain of the interacting protein. Like many other proteins associated with the cytoskeleton, PSTPIP is tyrosine phosphorylated in v-Src–transfected cells, and these phosphorylated residues appear to be substrates for the catalytic activity of the bound PTP HSCF. PSTPIP is localized to the cortical cytoskeleton, as well as in lamellipodia and it appears to migrate to the actin-rich cleavage furrow during cytokinesis. Overexpression of the protein in 3T3 cells induces long filopodial structures, consistent with a role for PSTPIP in the reorganization of the cytoskeleton. High level expression of the mammalian protein in the fission yeast S. pombe resulted in a dominant-negative inhibition of the completion of cytokinesis. These data suggest that PSTPIP is likely to be a cytoskeletal-associated protein, potentially involved with cytokinesis, whose physiological function is potentially regulated by its degree of tyrosine phosphorylation.

Analysis of the protein database for sequences with homology to PSTPIP suggests potential functions for this novel protein. Most of the sequences with significant homology to PSTPIP fall into the actin-associated family of proteins, and it is clear from the confocal studies reported here that PSTPIP is also associated with actin. While a number of other actin-interacting proteins, including myosin, fodrin, and spectrin, show homology to PSTPIP, the bulk of these homologies are within the SH3 domain, with little or no match in other regions of the protein. This is also true for cortactin (Wu et al., 1991), another protein that binds to the actin cytoskeleton in a similar manner, although there is additional weak homology in a small region of the coiled-coil domain in addition to the SH3 region. This is in contrast to the protein with the greatest degree of homology, the yeast S.pombe cdc15p, which shows significant sequence conservation in both the SH3 as well as the coiled-coil domains (Fankhauser et al., 1995). Cdc 15p is a highly phosphorylated protein that is absolutely required for the formation of the actin ring at the cleavage furrow of the postmitotic cell, and mutations in this protein result in an inability to assemble the actin ring over the postmitotic nucleus, thus resulting in multinucleate cells (Fankhauser et al., 1995). As with PSTPIP, cdc15p is localized to the cortical actin cytoskeleton until anaphase, when it migrates over the postmitotic nucleus and presumably mediates the reorganization of the cytoskeleton to the cleavage plane (Fankhauser et al., 1995; Simanis, 1995; Chang and Nurse, 1996). While the timing of PSTPIP migration to the cleavage furrow remains to be determined, its striking colocalization with the actin ring at this site during cytokinesis is analogous to what is observed with cdc15p (Fankhauser et al., 1995). In addition, the cdc15p is hyperphosphorylated until the onset of anaphase and the formation of the F-actin cytokinetic cleavage ring, when it becomes significantly dephosphorylated. Interestingly, the yeast protein regains its high state of phosphorylation at the conclusion of cell division, suggesting that phosphorylation regulates its association with the cleavage furrow. While the type of phosphorylation of cdc 15p has not yet been described, this suggests that tyrosine- and/or serine-threonine phosphatases must be involved with the regulation of the function of cdc 15p, and provides a mechanism whereby the binding and catalytic activity of a phosphatase such as PTP HSCF might function to control cytokinesis. Again, while the timing of tyrosine phosphorylation of PSTPIP during the cell cycle has yet to be determined, both the exact conservation of seven tyrosine residues between PSTPIP and cdc15p as well as the vanadate-sensitive tyrosine phosphorylation of the endogeneous interacting protein in Baf3 cells are suggestive of possible modulation of phosphotyrosine levels during the cell cycle. Thus, the sequence, cellular localization, and phosphorylation of both PSTPIP and cdc15p suggest that the mammalian protein is a potential homologue of cdc15p.

Phosphorylation, especially of serine and threonine residues, has been previously shown to play important roles in regulating events in cytokinesis and reorganization of the cytoskeletal (Mabuchi and Talcano-Ohmuro, 1990; Satterwhite et al., 1992; Tan et al., 1992; Egelhoff et al., 1993; Yamakita et al., 1994; Fishkind et al., 1995). To date, however, the possibility that tyrosine phosphorylation may play a role in these functions has been incompletely examined. The data reported in this paper suggest that the regulation of tyrosine phosphorylation on PSTPIP by PTP HSCF may play a role in aspects of cytoskeletal control including, possibly, cytokinesis. While the possible kinases involved in such tyrosine phosphorylation are numerous, the information described here, as well as elsewhere, suggests that a member of the Src family of tyrosine kinases may be involved with the phosphorylation of this interacting protein by either direct or indirect means. Two other PSTPIP-related proteins, cortactin and HSP 1, are both known to be tyrosine phosphorylated in v-Src–transformed cells, and cortactin, and potentially HSP 1 as well (Yamanashi et al., 1993), appear to interact with the cytoskeleton in a manner similar to PSTPIP (Wu et al., 1991; Dehio et al., 1995; Okamura et al., 1995; Vuori et al., 1995; Takemoto et al., 1996). In addition, a plethora of other proteins that are involved with the cytoskeleton are also tyrosine phosphorylated in v-Src–transformed cells (Schaller et al., 1993). Interestingly, the tyrosine phosphorylation of cortactin is also dramatically enhanced in cells isolated from mice deficient in the Csk kinase (Thomas et al., 1995), a tyrosine kinase which phosphorylates the COOH-terminal inhibitory tyrosine on c-Src, suggesting that cortactin is either a direct or indirect c-Src substrate in vivo. In addition, it has been demonstrated that HSP 1 can bind to the SH3 and SH2 domains of Src or Src family kinases in vitro, and it is also tyrosine phosphorylated by these kinases in vitro and in vivo (Takemoto et al., 1995, 1996). Although PSTPIP is only distantly related to cortactin and HSP 1, the tyrosine phosphorylation of this protein by v-Src in transfected cells may therefore have physiological relevance. In addition, previous data have demonstrated that c-Src associates with the focal adhesions and lamellipodia, as well as other actin-containing sites, consistent with the possibility that it could phosphorylate PSTPIP, which also localizes to these regions (Kaplan et al., 1994). Finally, v-Src is known to induce cytoskeletal changes in transformed cells, and it has been clearly shown that cortactin, an actin binding protein, becomes reoriented from the ends of the stress fibers to the podosomes of these v-Src–transformed cells, consistent with the possibility that phosphorylation of such actin-associated proteins might mediate changes in their cellular localization (Wu et al., 1991). Further analysis of the PSTPIP tyrosines phosphorylated in v-Src–transfected cells as well as by the endogenous tyrosine kinase will shed additional light on whether Src family kinases may mediate the phosphorylation of this interacting protein.

The use of dominant-negative forms of PTPs has previously been used to identify substrates for several enzymes, most notably PTP PEST (Garton et al., 1996) and the corkscrew PTP Src homology phosphatase-2 (SHP-2) (Herbst et al., 1996). In general, these studies have demonstrated that these dominant-negative mutants enhance the tyrosine phosphorylation of a surprisingly limited number of substrates in vivo, in contrast to the relatively promiscuous behavior of these enzymes in vitro. In the case of the dephosphorylation of p130^cas by PTP PEST, it appears that the substrate recognition is in large part due to the catalytic domain (Garton et al., 1996). The demonstration here that coexpression of two different dominant-negative forms of PTP HSCF mediates a dramatic increase in v-Src–induced PSTPIP tyrosine phosphorylation, together with the observation that this effect requires the COOH-terminal 24–amino acid homology domain, is thus consistent with several conclusions. The first is that these two proteins clearly interact in vivo, probably through the CTH domain, and the coiled-coil region interaction determined from the in vitro binding studies and the coprecipitation analyses (Figs. 5–7) supports such a physical interaction. This provides another example of the use of a noncatalytic region by a PTP to bring the catalytic domain in close proximity to the substrate (Tonks, 1993). In the case of the interaction described here, it appears that the juxtaposition of these proteins mediated by this interaction is required for substrate recognition, since the removal of the interaction domain from the PTP results in a complete lack of substrate recognition. The second is that it is possible that tyrosine-phosphorylated PSTPIP is an in vivo substrate for the PTP HSCF; and it also suggests that the enzyme inhibited by vanadate in the endogenous phosphotyrosine experiment in Baf3 cells, where both PSTPIP and PTP HSCF are expressed, is likely to be PTP HSCF (Cheng et al., 1996). While it is possible to test such a hypothesis by coprecipitation studies, we have found that the level of PTP HSCF expressed by Baf3 cells is insufficient for detection by Western blotting. Finally, if we assume that the mutant forms of PTP HSCF are endowed with the same degree of substrate specificity that has been found with other dominant-negative PTPs (Dixon, 1995; Garton et al., 1996; Herbst et al., 1996), then the v-Src cotransfection studies further suggest that either Src or a related family member may be kinases that are involved with the tyrosine phosphorylation of PSTPIP in vivo in nontransfected cells.

The v-Src–mediated tyrosine phosphorylation of PTP HSCF, together with the demonstration that dominant-negative forms of the enzyme induce a hyperphosphorylated state, strongly suggest that this PTP mediates its own autodephosphorylation. As expected, while the dominant-negative effect of the PTP on PSTPIP phosphorylation requires the COOH-terminal 24–amino acid homology domain, the dominant-negative hyperphosphorylation of the PTP does not require this region, further supporting the hypothesis that the interaction between PSTPIP and PTP HSCF is required for substrate recognition. Other PTPs, including SHP-1 (Bouchard et al., 1994; Lorenz et al., 1994), SHP-2 (Lechleider et al., 1993; Vogel et al., 1993; Feng et al., 1993, 1994; Stein-Gerlach et al., 1995), and PTPα (den Hertog et al., 1994) also appear to be tyrosine phosphorylated in response to a number of different stimuli. Interestingly, like PTP HSCF, SHP-2 appears to modulate its own phosphotyrosine levels (Stein-Gerlach et al., 1995), and these levels appear to modestly modulate enzymatic activity (Vogel et al., 1993). SHP-1 (Lorenz et al., 1994) and SHP-2 (Feng et al., 1994) are tyrosine phosphorylated in vivo by two Src family kinases, Lck and v-Src, respectively; this is consistent with results reported here that demonstrate tyrosine phosphorylation of PTP HSCF by v-Src. Moreover, tyrosine phosphorylation of SHP-2 and PTPα both result in the association of the GRB2 adaptor protein (den Hertog et al., 1994; Vogel et al., 1996). Together, these data suggest that modification of tyrosines on PTPs is a potential regulatory mechanism, and it should prove interesting to examine the role of phosphatase tyrosine phosphorylation in the system described here.

The nature of the high affinity binding between the proline-rich CTH domain and the coiled-coil region is reminiscent of that previously described for the SH3–proline-rich core interaction (Pawson, 1995; Feng et al., 1995). In this latter case, proline helices induce the formation of highly structured small peptide domains that bind with relatively high affinity and specificity to the binding pocket of the SH3 domain, and various interactions, including salt bridges, mediate the specificity and directionality of peptide binding (Feng et al., 1995). Analysis of the proline-rich CTH domains of three PEST PTPs, all of which appear to inhibit the PSTPIP-PTP HSCF binding interaction with similar 50% inhibitory concentrations (<1 μM), reveals that they share a proline-rich core region that would be predicted to form a proline helix similar to that seen for SH3 binding sites (Matthews et al., 1992; Yang et al., 1993; Cheng et al., 1996). This region contains a number of charged residues, and it appears that the potential helical nature of this domain positions these residues in an appropriate binding conformation for interaction with a site within the coiled-coil domain (Dowbenko, D., and L. Lasky, unpublished observations). Because all of the PEST PTPs are predicted to bind to PSTPIP via this proline-rich region, it is possible that the interacting protein's phosphotyrosine content is modulated by different PEST PTPs in different cell types. Along these lines, it is interesting to note that the only hyperphosphorylated protein observed in COS cells transfected with dominant-negative (Asp–Ala) PTP PEST was p130^cas (Garton et al., 1996). These results suggest that, if PSTPIP is highly expressed in COS cells, it is either not tyrosine phosphorylated or is not a substrate for this PTP in this cell line.

The mechanism by which PSTPIP migrates from the cortical actin, lamellipodia and stress fiber regions in resting cells to the cytokinetic cleavage furrow in dividing cells can only be speculated upon (Strome et al., 1993; Fishkind et al., 1995). One possibility is that this protein binds tightly to actin, and when the actin is reoriented to the cleavage plane, the PSTPIP accompanies it passively (Cao et al., 1990a,b; Fishkind et al., 1993). However, experiments in yeast where cdc15p is deleted revealed that cortical actin did not migrate to the cleavage plane in the absence of this protein, suggesting that cdc15p actively traverses to this site and mediates the assembly of the actin ring (Fankhauser et al., 1995; Simanis, 1995). Thus these data suggest that if PSTPIP is a mammalian homologue of cdc15p, that dominant-negative mutants in this protein should affect the assembly of actin at the vertebrate cleavage furrow, and we are currently testing this possibility. Interestingly, it appears that deletion mutants of cdc15p that lack the SH3 domain are incapable of rescuing the cdc15 mutants, suggesting a critical role for this COOH-terminal domain in assembling the cytokinetic actin ring (Fankhauser et al., 1995). It will therefore be important to determine the proteins which bind to this domain in PSTPIP, since they may provide additional insights into its physiological function.

A possible mechanism by which PSTPIP functions is suggested by the results of overexpression studies in murine 3T3 cells. The extended filopodial structures in many of these transfected cells are consistent with the possibility that the unregulated expression of the protein mediates an ectopic and organized assembly of actin filaments, thus resulting in a cellular protrusion containing PSTPIP and F-actin. While the striking level of lysine residues in the predicted coiled-coil domain of this protein is consistent with previously described actin binding sites (Vandekerckhove, 1990; Friederich et al., 1992), we have been unable to demonstrate a stable association between PSTPIP and F-actin in vitro (Wu, Y., and L.A. Lasky, unpublished data). Interestingly, many of the transfected cells contained a single filopodial-like structure, suggesting that this morphological feature is rapidly formed and is likely to have a negative influence on cell viability. The apparent small size of many of these cells suggests that this actin-containing spike is formed in the absence of plasma membrane synthesis, also consistent with a rapid formation of the structure. The evident heterogeneity in penetrance of this morphological entity may either be due to diverse expression levels or differences in posttranslational modifications of the transfected proteins. If these suppositions are correct, then it would appear that PSTPIP may play a role in the rapid assembly of a highly organized F-actin–containing structure.

The results of overexpression of mammalian PSTPIP in exponentially growing S. pombe provide support for a role for this protein in cytokinesis. Fission yeast cells grow mainly by elongation at their tips, and divide by binary fission after forming a centrally placed actin-rich septum. In S. pombe, F-actin is seen as patches or dots at sites of cell growth or division. During interphase, it is found at the growing ends of the cell, and, after the onset of mitosis, it relocates to form an equatorial ring whose position anticipates the site of septum formation (Marks and Hyams, 1985; for review see Robinow and Hyams, 1989). At the end of mitosis, when the daughter nuclei are well separated and the spindle begins to break down, the septum grows inwards from the cell cortex. Secondary septa are formed on either side of the primary septum, which is subsequently dissolved to effect cell separation. F-actin is then relocated to the old (preexisting) end of the cell, from which growth resumes. Some of the proteins that control the onset of septum formation and cytokinesis in fission yeast have been identified. The products of the cdc3, cdc4, cdc8, cdc12, cdc15, and rng2 genes are required for actin rearrangement and/or to stabilize the actin ring (Nurse et al., 1976; Balasubramanian et al., 1992, 1994; Fankhauser et al., 1995; McCollum et al., 1995; Chang et al., 1996). Immunofluorescence studies have shown that the products of the dmf1, cdc3, cdc4, cdc8, and cdc15 genes are associated with the medial ring (Balasubramanian et al., 1992, 1994; Fankhauser et al., 1995; McCollum et al., 1995), a cellular localization similar to that observed for PSTPIP in both mammalian cells and yeast. At the end of mitosis, the cdc7 kinase and the activities of Cdc11p and Cdc14p are required for septation (Nurse et al., 1976; Fankhauser and Simanis, 1993, 1994). The plo1 kinase appears to be required for both actin ring formation and septation, and can induce septum formation from G1 and G2, if expressed at a very high level. These data are consistent with the hypothesis that protein phosphorylation is involved with the regulation of cytokinesis. The dominant-negative phenotype observed here is similar to that seen in fission yeast that are null for the expression of the calcineurin-like phosphatase, ppb 1, although this is presumably a serine/threonine phosphatase, while PSTPIP binds to a tyrosine phosphatase (Yoshida et al., 1994). Nevertheless, the data are consistent with the possibility that PSTPIP titrates out an important component of the S. pombe cytokinetic machinery, with the result that cleavage is inhibited.

The results reported here describe PSTPIP, a novel member of the cytoskeleton-associated protein family, which is a substrate for the PTP HSCF. The homology of PSTPIP with the fission yeast cdc15 protein, together with the demonstration of in vivo tyrosine phosphorylation of the interacting protein, association with the cortical cytoskeleton and cellular cleavage furrow, the induction of cytoskeletal alterations, and the dominant-negative inhibition of S. pombe cytokinesis are all consistent with the possibility that PSTPIP is a mammalian homologue of the yeast cleavage furrow regulatory protein. Because this protein interacts with and is an enzymatic substrate for the PTP HSCF, this hypothesis suggests that the control of tyrosine phosphorylation of PSTPIP by this PTP may provide for a novel mechanism for the control of cleavage furrow formation during cytokinesis.

We thank L. Tamayo and D. Wood for help with the figures, C. Quan for peptide synthesis, and O.K.L. Lasky for helpful discussions at the beginning of this project.

CTH

COOH-terminal homology

GST

glutathione-S-transferase

HA

hemagglutanin

HSCF

hematopoietic stem cell fraction

HSP

hematopoietic specific protein

PSTPIP

proline, serine, threonine phosphatase interacting protein

PTP

protein tyrosine phosphatases