Four PSM/SH2-B alternative splice variants and their differential roles in mitogenesis.

An SH2 domain originally termed SH2-B had been identified as a direct cellular binding target of a number of mostly mitogenic receptors. The complete cellular protein, termed PSM, and respective sequence variants share additional Pro-rich and PH regions, as well as similarities with APS and Lnk. A role of these mediators has been implicated in signaling pathways found downstream of growth hormone receptor and receptor tyrosine kinases, including the insulin, insulin-like growth factor-I (IGF-I), platelet-derived growth factor (PDGF), nerve growth factor, hepatocyte growth factor, and fibroblast growth factor receptors. As a result of this report a total of four PSM/SH2-B sequence variants termed alpha, beta, gamma, and delta have now been identified in the mouse and have been compared with the available rat and human sequences. Variant differences are based on alternative splicing and define distinct last exons 7, 8, and 9 that result in reading frameshifts and unique carboxyl-terminal amino acid sequences. Variant sequences have been identified from cDNA libraries and directly by reverse transcription-polymerase chain reaction. Sequence analysis predicts four distinctly sized protein products that have been demonstrated after cDNA expression. All were found phosphorylated on tyrosine specifically in response to IGF-I and PDGF stimulation. cDNA expression of the four variants caused variant-dependent levels of stimulation of IGF-I- and PDGF-induced mitogenesis. The most pronounced increase in mitogenesis was consistently observed for the gamma variant followed by delta, alpha, and beta with decreasing responses. In contrast, the mitogenic response to epidermal growth factor consistently remained unaffected. The variants are expressed in most mouse tissues, typically, most strongly in pairs of alpha and delta or beta and gamma. Our findings implicate differential roles of the PSM/SH2-B splice variants in specific mitogenic signaling pathways.

PSM has been identified as a mouse protein based on its association with the activated catalytic insulin receptor (IR) 1 domain in a yeast two-hybrid screen (1,2). PSM carries a number of domains, including Pro-rich putative SH3 domain binding regions, a pleckstrin homology region (PH), a src homology-2 (SH2) domain, and many potential Tyr and Ser phosphorylation motifs, which suggest a role as a putative signaling mediator. Such a role is also supported by the observation that PSM associates with activated but not with inactive forms of IR (2). A related sequence termed SH2-B that represents a rat homolog of PSM had been identified in the rat based on its association with the high affinity IgE receptor Fc⑀RI␥ subunit (3). A variant form of PSM/SH2-B, termed ␤, has been reported, which carries an additional 100 bp of coding region downstream of the SH2 domain and is predicted to result in a shortened protein product (2,4). This variant was described as a substrate and as a potent cytoplasmic activator of the tyrosine kinase JAK2 in response to growth hormone signaling (4 -6). The interaction with JAK2 involves the SH2 domain and amino-terminal region of SH2-B ␤ (7). An association of PSM/ SH2-B was established with the insulin-like growth factor-I (IGF-I) receptor (IGF-IR), which depends on receptor activation and was mapped to phospho-tyrosine sites that are conserved between the insulin and IGF-I receptors (8). An association with platelet-derived growth factor receptor (PDGFR) in response to PDGF activation was described for PSM/SH2-B that stimulated its phosphorylation on tyrosine, serine, and threonine (9,10). Tyrosine phosphorylation of PSM has been reported in response to insulin (11). With regard to its physiologic role PSM cDNA expression has been shown to stimulate the mitogenic response to PDGF-BB, IGF-I, and insulin whereas introduction of a cell-membrane-permeable, putatively dominant-negative SH2 domain peptide interfered with the same response (10). In addition, microinjection of the SH2 domain into transformed fibroblasts partially restored a normal actin stress fiber pattern suggesting a stimulatory role of PSM in normal and malignant cell proliferation (10). Independently, SH2-B ␤ was shown to be required for growth hormone-induced actin reorganization (12). The SH2-B gene structure and alternative splice sites have been reported in comparison with a third variant termed ␥ that specifically binds to Tyr-1146 of IR (13). A single gene has been mapped to the distal arm of mouse chromosome 7 in a region linked to obesity in mice (13).
SH2-B shares a high degree of structural similarity with adaptor proteins, APS and Lnk, that have been shown to participate in signaling events initiated by B cell receptor and T cell receptor, respectively (14 -16). Thus, the common structural relationship between PSM/SH2-B, APS, and Lnk may suggest functional similarities between these three proteins. APS is also expressed in insulin-responsive tissues such as skeletal muscle and fat and becomes phosphorylated in 3T3-L1 adipocytes in response to insulin (17). The SH2 domain of APS, which bears 80% identity with the SH2 domain of SH2-B, associates directly with tyrosyl-phosphorylated insulin receptor (17). APS also associates with activated PDGFR or with c-Cbl and inhibits PDGF-induced mitogenesis (18). SH2-B (19) and APS have been implicated as substrates of Trk nerve growth factor receptors TrkA and TrkB/C, respectively, with a role in developing neurons and neuronal differentiation (20,21). SH2-B and APS exist as homo-or heteropentamers mediated through the amino terminus and influence neurotropin signaling through direct modulation of Trk autophosphorylation (22). SH2-B interacts with the activation loop of TrkA (19) and plays a specific role in TrkA-mediated differentiation in human neuroblastoma cells (23).
The present study has focused on the identification of additional mouse sequence variants, the comparison of the involved alternative splice junctions, the predicted differences in protein primary structure when compared with human and rat, and on the implicated putative biochemical and physiologic differences. As a result of this report a total of four PSM/SH2-B sequence variants termed ␣, ␤, ␥, and ␦ have now been identified in the mouse and have been compared with the available rat and human sequences. Variant differences are based on alternative splicing and define distinct last exons 7, 8, and 9 that result in reading frameshifts and unique carboxyl-terminal amino acid sequences. Variant sequences have been identified from cDNA libraries and directly by RT-PCR. Sequence analysis predicts four distinctly sized protein products that have been demonstrated after cDNA expression. All were found phosphorylated on tyrosine specifically in response to IGF-I and PDGF stimulation. cDNA expression of the four variants caused distinct levels of stimulation of IGF-I-and PDGF-induced mitogenesis. Stimulation was inducer dose-dependent in experiments involving ecdysone-regulated expression plasmids. With two alternative sets of expression plasmids, the most pronounced increase in mitogenesis was consistently observed for the ␥ variant followed by ␦, ␣, and ␤ with decreasing responses. In contrast, the mitogenic response to EGF consistently remained unaffected. The variants are expressed in most mouse tissues, typically, most strongly in pairs of ␣ and ␦ or ␤ and ␥. Our findings implicate differential roles of the SH2-B/ PSM splice variants in specific mitogenic signaling pathways. In combination, the existing data suggest that PSM/SH2-B represents a family of adapters of several known and possibly additional unknown members that belong to a superfamily of signaling mediators, including APS and Lnk. Their physiologic roles include the control of mitogenesis, neuronal differentiation and development, and likely other mechanisms of cell regulation.

EXPERIMENTAL PROCEDURES
PSM/SH2-B cDNA Isolation-The isolation of complete PSM/SH2-B cDNA has been described earlier (2). A 529-bp fragment containing the SH2 domain of PSM/SH2-B identified by yeast two-hybrid screening served as a hybridization probe to screen a mouse brain lambda Uni-Zap XR library (Stratagene). Positive clones were isolated, and the resulting plasmids were analyzed by fluorescence-automated DNA sequencing. The 5Ј-end of the cDNA was isolated by 5Ј RACE-PCR (Life Technologies, Inc.) using mouse brain total RNA. The RACE-PCR products were inserted into a TA vector (Invitrogen), and a 1.5-kb insert was released by digestion with EcoRI. Five isolated clones were sequenced. One, TA3, served to assemble the cDNA containing the complete protein-coding region of the PSM variants. A 1.2-kb DrdI-EcoRI fragment containing part of the 5Ј-untranslated region (Ϫ130 bp) and of the coding region was isolated from TA3. The 5Ј-Drd I site was end-filled, and the resulting fragment was joined with the HincII and EcoRI cloning sites of pBluescript KS ϩ (Stratagene). The resulting plasmid (clone 2) was digested with BspEI, at a unique site in the insert, and with SmaI within the multiple cloning site region of KS ϩ for the subsequent cDNA assembly described below.
Complete cDNA Assembly of Four PSM/SH2-B Variants-Clone 5-1 was originally obtained from a mouse brain lambda Uni-Zap XR library and carries a 1.9-kb PSM cDNA insert and a unique BspE I site at its 5Ј-end, which was used to join two parts of the PSM/SH2-B cDNA. This clone also carried the 153-bp insert representing the ␥ variant. A 1.5-kb 5Ј-BspEI/3Ј-ScaI fragment containing almost half of the protein-coding region was isolated from clone 5-1 and inserted into BspEI-SmaI sites in clone 2 carrying the 5Ј-end of the PSM/SH2-B coding sequence (described earlier). With this approach a subclone PSM/SH2-B␥/KS ϩ (clone B23) was prepared, which carried the complete protein-coding cDNA for the ␥ isoform. A 2.5-kb KpnI-BamHI fragment, containing the complete PSM/SH2-B ␥ variant cDNA, was released from clone B23 and inserted into KpnI and BamHI cloning sites of the ecdysone-inducible expression vector pIND (Invitrogen). The resulting plasmid PSM/SH2-B␥/IND was used to generate the construct for the ␣ variant cDNA.
Clone 20 was obtained from the mouse brain lambda Uni-Zap XR library, which carries a 1.1-kb PSM/SH2-B cDNA insert and corresponds to the ␣ variant. An 884-bp EcoRI-BspHI fragment was isolated from clone 20, end-filled at the BspHI site, and subcloned into the EcoRI and SmaI sites of KS ϩ . From the resulting plasmid a 591-bp BstXI-NotI fragment was released and joined with a BstXI-NotI fragment of PSM/ SH2-B␥/IND. This step exchanged the region of the ␥-specific cDNA with ␣ variant sequences and resulted in an expression construct for PSM/SH2-B␣/IND. Because the PSM/SH2-B␣/IND plasmid had lost a unique BstXI site within the multiple cloning site region of the vector, the insert-specific BstXI site was used for the construction of complete protein-coding cDNA of the ␤ and ␦ variants of PSM/SH2-B.
Clones 34 and 33 were originally identified in the mouse brain lambda Uni-Zap XR library with 1.1-kb PSM cDNA inserts that corresponded to the ␤ and ␦ variants, respectively. Two BstXI-BstEII fragments, a 500-bp fragment from clone 34 containing ␤ variant-specific sequences and a 453-bp fragment from clone 33 carrying ␦ variantspecific sequences, were individually exchanged with the corresponding  (24).
RT-PCR-The region containing the variant-related inserts was amplified by reverse transcription-polymerase chain reaction (RT-PCR). Complementary DNA was synthesized from 0.5 g of mouse brain poly(A) ϩ RNA (CLONTECH) using oligo(dT) or from 2 g of total RNA from mouse tissues. Moloney murine leukemia virus reverse transcriptase (Roche Molecular Biochemicals, Mannheim) was employed in buffer containing 5 mM MgCl 2 in a total volume of 20 l with the following parameters for one cycle: 42°C for 60 min, 99°C for 5 min, and 4°C for 5 min. Following reverse transcription, PCR was carried out (where indicated at varying Mg 2ϩ concentrations between 1 and 2 mM) using Taq DNA polymerase and primers flanking the insert, 5Ј-TTCGATATGCTTGAGCACTTCCGG-3Ј, including nucleotide positions 2081-2104, and 5Ј-GCCTCTTCTGCCCCAGGATGT-3Ј, including nucleotide positions 2345-2365 (GenBank™ accession number AF020526) (2). Each cycle consisted of denaturation at 94°C for 40 s, annealing at 66°C for 40 s, and primer extension at 72°C for 1 min for 30 cycles, with a final extension step at 72°C for 10 min. For the experiment shown in Fig. 1C, the PCR products from mouse brain poly(A) ϩ RNA were cloned into pCR2.1 TA cloning vector or TOPO TA vector (Invitrogen). The resulting plasmids were analyzed by EcoRI digestion and fluorescence-automated DNA sequencing. For the experiment shown in Fig. 2, PCR products from mouse tissue RNA were separated on a DNA-agarose gel. ␤-Actin as internal control is shown using the same amplification protocol and upstream primer 5Ј-CGTCTGGACCTGGCT-GGCCGGGACC-3Ј and downstream primer 5Ј-CTAGAAGCATTTG-CGGTGGACGATG-3Ј.
Complete PSM/SH2-B cDNA Expression in Mammalian Cells-5 ϫ 10 5 mouse NIH 3T3 fibroblasts were transiently transfected with each of the four PSM/SH2-B variants in ecdysone-inducible pIND or constitutive CVN expression plasmids. Subconfluent cultures of NIH 3T3 fibroblasts were rinsed with serum-and antibiotic-free medium and incubated for 5 h at 37°C in 2 ml of transfection solution. This contained 2 g of each plasmid, the variant-specific pIND and pVgRXR of the two-plasmid expression system, 10 l of LipofectAMINE, and 8 l of Plus reagent according to the instructions of the manufacturer (Life Technologies Inc.). One milliliter of complete culture medium was added, and cells were incubated for 12 h and, subsequently, for 5 h in fresh complete medium. For subsequent biochemical and mitogenic analysis, cells were treated as described in the following sections.
Immunoprecipitation and Analysis of Tyrosine Phosphorylation-For protein expression analysis, cells were cultured in fresh complete medium for an additional 36 h in the presence of the ecdysone analog ponasterone A (10 M). Subsequently, cells were rinsed twice with PBS and harvested in ice-cold lysis buffer containing 50 mM HEPES, pH 7.4, 1% Triton X-100, 10% glycerol, 137 mM NaCl, 2 mM EDTA, 10 mM NaF, 100 mM Na 3 VO 4 , 10 mM sodium pyrophosphate, 10 g/ml leupeptin, 10 g/ml aprotinin, and 1 mM phenylmethylsulfonyl fluoride. Proteins (300 g) were first mixed with PSM antiserum (10) and co-precipitated with protein A-Sepharose for 1 h. Precipitated proteins were rinsed with lysis buffer, separated by 8% SDS-PAGE, and analyzed by immunoblotting with PSM antiserum using the ECL detection system (Amersham Pharmacia Biotech).
Alternatively, to elucidate tyrosine phosphorylation, 5 ϫ 10 5 transfected cells were incubated in normal culture medium for 24 h before reaching quiescence in serum-free medium after 20 h. During both steps the ecdysone analog ponasterone A was present at the various indicated concentrations. Subsequently, IGF-I or EGF at 100 ng/ml or, alternatively, PDGF-BB at 25 ng/ml was added for 15 min. Cells were rinsed twice with PBS and harvested in ice-cold lysis buffer. Proteins were first mixed with PSM antiserum and co-precipitated with protein A-Sepharose. Precipitates were washed with lysis buffer, separated by 8% SDS-PAGE, and immunoblotted with anti-phospho Tyr antibody PY 20 (Transduction Laboratories).
DNA Synthesis-To assay for DNA synthesis, 2 ϫ 10 6 transfected cells were split into 24-well plates and incubated in complete medium in the presence or absence of the ecdysone analog ponasterone A at various concentrations for 24 h before they were starved for 20 h. Subsequently, IGF-I or EGF at 100 ng/ml or, alternatively, PDGF-BB at 25 ng/ml was added for 18 h, and finally 0.1 Ci of [methyl-3 H]thymidine was added at 3 TBq/mmol for 5 h. Cells were rinsed three times in ice-cold PBS and incubated in 10% trichloroacetic acid for 1 h at 4°C. Precipitates were rinsed, and 0.5 ml of 0.2 N NaOH, 0.1% SDS was added at 37°C for lysis 1 h before the pH was neutralized by addition of 0.5 ml of 2 M Tris, pH 6.8. Two milliliters of mixture (ScintiSafe econo-1, Fisher) was added to quantify the incorporated radioactivity by liquid scintillation spectroscopy.
Cell Proliferation-To measure cell proliferation, 2 ϫ 10 6 transfected cells were split into 24-well plates and incubated in complete medium for 24 h. Cells were rinsed once with ice-cold PBS and incubated with 100 ng/ml IGF-I or 25 ng/ml PDGF-BB in medium with 0.5% serum. After 3 days, 200 l of 0.5 mg/ml 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl-tetrazolium bromide (MTT) was added for 4 h. The resulting formazan was dissolved in isopropyl alcohol and quantified colorimetrically at A 570 (25). The result has been presented as a measure of the cell number.

Identification of Four PSM/SH2-B Sequence Variants-The
SH2 domain-encoding region had earlier been identified in a yeast two-hybrid screen (1). It was used to screen a mouse brain Uni-Zap XR library (Stratagene) to isolate the complete PSM/SH2-B protein-coding cDNA (2). A number of independent clones were identified, and eight were sequenced and compared with the originally reported sequence termed ␣ (2-4). One clone represented the same ␣ sequence, one clone carried a 100-bp insertion representing a variant termed ␤ (2, 4), five clones carried a 153-bp insert representing a variant termed ␥ (13), and one clone carried a 53-bp insert representing a variant termed ␦. The 153-bp insert was composed of both the 100-bp and 53-bp inserts. All insertions were located downstream near the SH2 domain-coding sequence starting at nucleotide 1894 (when A of the initiation codon is defined as 1) of the complete mouse PSM/SH2-B ␣ cDNA sequence (Gen-Bank™ accession number AF421138) (2, 3). Starting from this position a comparison of the four variant sequences (Fig. 1A) suggested that the other isoforms are generated by alternative splicing of the longest primary PSM/SH2-B transcript carrying the 153-bp insertion (␥) (13). The observed sequence inserts result in reading frameshifts. All variant sequences predict unique protein carboxyl termini that carry specific motifs, including Pro-rich regions for ␣, ␥, and ␦; as well as a Ser/Trp-rich sequence and a nuclear localization signal for ␦ (Fig. 1A). A comparison with listed sequences representing human variants ␣ (GenBank™ accession number AF227967), ␤ (AF227968), and ␥ (AF227969), as well as rat variants ␣ (GenBank™ accession number U57391) and ␤ (AF047577) (2, 3), indicates highly conserved alternative splice variants and implicates functional differences of the existing isoforms (Fig. 1B).
Expression of PSM/SH2-B Variants in Mouse Brain-To address whether mRNA representing the observed variant sequences could be independently demonstrated in mouse brain, poly(A) ϩ RNA sequences were amplified by reverse transcription-polymerase chain reaction (RT-PCR) using a set of diagnostic primers (see "Experimental Procedures") to distinguish between the variant sequences by size. On agarose gels all resulting PCR products correlated well with the fragment sizes predicted for the four variants, 338 bp for ␥, 285 bp for ␤, 238 bp for ␦, and 185 bp for ␣ (Fig. 1C). Fragments were introduced into a TA cloning vector either directly or after additional PCR amplification, and numerous clones were characterized by DNA sequence analysis. These studies independently established identical variant nucleotide sequences compared with those that had been found in the cDNA library. Variant sequences were identified repeatedly in independent clones for ␦ (9), ␤ (19), ␣ (17), except for ␥. This is likely explained by the low abundance and diffuse pattern for this band shown as the largest size at the top of the gel (Fig. 1C). It may reflect a low expression level of this isoform that is also suggested in a comparison of various tissues (Fig. 2) and/or reduced stability of the respective mRNA that has not been further investigated.
Variant Exon-Intron Structure and Alternative Splicing-Based on the reported genomic organization of the SH2-B gene (13), the SH2 domain is encoded by exons 6 and 7 (Fig. 3A). In addition, exon 7 carries the 53-bp insert that is unique to the ␥ and ␦ variants (Figs. 1A and 4A). Exon 8 carries the 100nucleotide insert unique to both ␤ and ␥ variants but is absent from both ␣ and ␦ transcripts. Based on the reported genomic sequence information (13), all identified splice sites, including those of the ␦ variant (Fig. 3B), match reported consensus sequences for mammalian splice donor and acceptor sequences. When compared with the ␣ variant protein-coding region, the sequence insertions as a result of alternative splicing cause reading frameshifts. These lead to termination codons and variant proteins that are shorter than the PSM/SH2-B ␣ isoform (Figs. 1A and 4A) (2). All variants are identical through Glu-631, the last shared amino acid. The subsequent carboxylterminal sequences are unique for all variants, except for the first 18 amino acids that are shared between ␥ and ␦ (Fig. 4A). Accordingly, the 53 nucleotides at the 5Ј-end of the 153-bp insert found in ␥, constitute the ␦ insert (Fig. 1A).
Protein Variants Exhibit Different Carboxyl Termini-Based on the deduced primary structure, all four variant proteins  Expression of Variant Proteins from cDNA-To directly demonstrate that the variant sequences will be expressed into distinct protein products cDNA expression plasmids were prepared. We chose an ecdysone-inducible expression system (pIND, Invitrogen) to allow us to correlate varying doses of cDNA expression with putative physiologic responses in subsequent experiments. Mouse NIH 3T3 cell cultures were transiently transfected with each of the four variant plasmids, and expression was induced with ecdysone. Proteins from detergent cell lysates were immunoprecipitated, separated by SDS-PAGE, and identified by immunoblotting with PSM-specific rabbit antiserum (8). When compared with control-transfected cells a single distinct major protein band was observed for each variant. Relative migrations corresponded to ϳ90 kDa for ␣, 85 kDa for ␦, and 80 kDa or less for ␥ and ␤ (Fig. 4B). The ␤ variant typically migrated slightly faster than ␥, but the observed size differences were minimal. The observed relative sizes are fully compatible with the primary structures predicted by the variant sequences (Figs. 1B and 4A) and suggest that each sequence variant encodes a distinct protein product with putatively unique functional aspects.
Variant Tissue Distribution in the Mouse-To evaluate the distribution of the four variants in different tissues, total RNA from mouse tissues was analyzed by RT-PCR. Northern analysis was not feasible, because the four variants cannot be distinguished by unique DNA sequences used as hybridization probes. Primers flanking the 100-and 53-bp sequence inserts (see top of Fig. 1A), which distinguish between the four variants, were used to amplify short PCR fragments from different tissues that were compared on an agarose gel (Fig. 2). The amplification of actin sequences under the same conditions served as an internal control. Expression of ␣ and ␦ variants was clearly demonstrated in thymus, skeletal muscle, kidney, spleen, lung, and particularly strongly in brain and 10-day embryonal tissue, typically with a significantly higher signal for ␦ when compared with ␣, except in testes where the ratio was inverted. Only marginal amounts were found in ovary or heart, and none of the variants were convincingly demonstrated in liver. The ␤ and ␥ variants were prominently detected in ovary and heart, complementary to the presence of the ␦ and ␣ isoforms. In addition, ␤ and ␥ were detected in testis and minimally in skeletal muscle, brain, lung, and thymus. Typically, ␤ was significantly more prominent than ␥. Our results suggest a possible co-operative action between ␦ and ␣ as well as between ␤ and ␥; however, typically only one of either pair was prominently present in any tested tissue. The variants will likely exhibit some functional redundancy in addition to specific roles.
Ecdysone Dose-responsive Variant Protein-mediated Differential Mitogenic Effects-To address the putative functional differences between the variant proteins, we investigated their effects on three mitogenic signals. We had previously established a mitogenic role for PSM (10). For this purpose mouse NIH 3T3 fibroblasts were individually transfected with each variant plasmid, starved to quiescence, and stimulated with IGF-I, PDGF, or EGF. In parallel, increasing doses of ecdysone allowed us to raise variant protein expression levels and observe the resulting physiologic consequences. DNA synthesis was measured by exposing transfected cells to [methyl-3 H]thymidine and quantifying acid precipitated radioactivity. In the absence of ecdysone, a modest stimulation of DNA synthesis was observed for each growth factor up to about 2-fold over control cell levels (Fig. 5A). IGF-I-and PDGF-stimulated DNA synthesis was significantly increased by each variant at 2 M ecdysone and dramatically further at the highest employed dose of 10 M ecdysone. Maximum stimulation reached 9-fold over control cell levels for the ␥ variant that consistently displayed the highest level of activity (Fig. 5A). The ␦, ␣, and ␤ variants followed with decreasing activity consistently in this order. Relative activities were consistently observed in this order also at lower ecdysone concentrations of 2 M with overall reduced effects. No significant effect was observed for any of the variants in the absence of growth factor or after EGF stimulation. This implicates specific roles of the variants in the signaling mechanisms downstream of the IGF-I and PDGF receptors but not of the EGF receptor. It is consistent with earlier observations that implicated a mitogenic function of PSM in the IGF-I and PDGF receptor pathways but not in the EGF receptor pathway (10).
Variant Protein-mediated Differential Mitogenic Responses with Constitutive Expression Plasmids-To exclude any potential impact of individual expression plasmid preparations on our observations, experiments were carried out with additional preparations of ecdysone-inducible plasmids as well as with a constitutive expression plasmid. In the latter plasmid variant protein expression was controlled by early simian virus 40 transcriptional promoter elements. Upon transfection all variant proteins resulted in mitogenic responses (Fig. 5B) that were very similar compared with those observed at the highest ecdysone concentration (Fig. 5A). These experiments confirmed that the observed mitogenic responses are not plasmid-specific.
Variant Protein-mediated Differential Levels of Cell Proliferation-We evaluated whether the observed impact on DNA synthesis could also be measured at the level of cell proliferation and the resulting cell numbers. For this purpose mouse NIH 3T3 fibroblasts were individually transfected with each (constitutive) variant plasmid. Cells were subsequently stimulated with IGF-I or PDGF, and cell proliferation was measured with the MTT assay. A specific increase in cell proliferation was observed in response to IGF-I and PDGF that was differentially increased by the variant proteins. As previously shown After reverse transcription of mRNA from various mouse tissues, PCR was carried out with specific diagnostic primers (see "Experimental Procedures") to amplify variant-specific fragments of carboxyl-terminal coding regions (indicated on the right) and in parallel to amplify a control ␤-actin fragment. Products were compared on an ethidium bromide-stained agarose gel with specific size markers (in base pairs) shown on the left. The presented mouse tissues are thymus (Th), skeletal muscle (Mu), 10-day embryo (Em), kidney (Ki), ovary (Ov), testis (Te), spleen (Sp), lung (Lu), heart (He), liver (Li), and brain (Br).
in Fig. 5 (A and B) the highest level of stimulation was observed with the ␥ variant followed by ␦, ␣, and ␤ with decreasing levels (Fig. 5C). In combination, these data consistently implicate differential mitogenic roles for each variant protein.
Variant Protein Tyrosine Phosphorylation-To begin to dissect the molecular mechanisms involved in PSM function, we evaluated phosphorylation of the four variant proteins on Tyr in response to mitogenic signals. Transfected NIH 3T3 fibroblasts were incubated with various concentrations of ecdysone, starved to quiescence, and stimulated with IGF-I, PDGF, or EGF. Detergent cell lysates were immunoprecipitated with PSM antiserum, and proteins were separated by SDS-PAGE and analyzed by immunoblotting with a phospho Tyr-specific antibody. Specifically in response to IGF-I and PDGF, all four variants were detected as distinct individual protein bands (Fig. 6) of similar migration as shown in Fig. 4B. The appearance was properly correlated with the dose of ecdysone used to induce expression. In addition, the IGF-I and PDGF receptors were observed in amounts proportional to the variant proteins as expected based on co-immunoprecipitation with PSM after growth factor and ecdysone stimulation. No signal was observed in response to EGF; the presence of the EGF receptor had been confirmed in control lysates (Fig. 6). The data show that all four variants are phosphorylated on Tyr in response to IGF-I and PDGF receptor activation. Because significant differences were not detected in the signal between the variants and because putative target phosphorylation motifs are only found in the sequences shared between all variants this modification is unlikely to play a role in the differential mitogenic responses observed in Fig. 5. However, it may play an important role in PSM activation that will be the focus of future experiments. DISCUSSION In the present study we have compared the characteristic features of the four PSM/SH2-B variants, including the newly identified isoform ␦. Our data are consistent with a common gene from which all variants are derived by alternative splicing (13). We have identified unique carboxyl-terminal sequences for each variant containing specific functional motifs that are predicted to define functional differences. Compatible with these structural differences we observed distinct levels of mitogenic stimulation for the four variants similarly for PDGFand IGF-I -induced DNA synthesis and cell proliferation. Functional differences of the variants are also supported by the differential expression pattern found in many tissues.
When compared with the initially discovered variant now termed ␣, the ␤, ␥, and ␦ isoforms contain insertions of 100, FIG. 3. A, exon-intron structure of mouse variant sequences. Exon and intron boundaries are shown based on the information reported by Nelms et al. (13). All variants have been aligned for comparison with the new ␦ sequences. Exons (numbered underneath 1-9 in boldface) have been represented by heavy lines. Protein-coding regions have been indicated by heavy boxes, and their amino acid boundaries are shown by vertically oriented numbers (the initiator Met has been defined as 1). Amino acids that are encoded by two exons have been assigned to the exon that carries two nucleotides of the respective codon. Introns have been represented by light horizontal lines, and their length in nucleotides has been shown underneath. The location of termination codons has been indicated (STOP). Shared sequences, including exons 1-6, that are identical for each variant at the beginning of the gene have only been shown once at the top. As indicated by an arrow they precede each variable sequence that is shown individually only starting with exon 7. B, mouse variantspecific alternative splice boundaries. Sequence boundaries are shown based on the information reported by Nelms et al. (13). All variants have been aligned for comparison with the new ␦ sequences. Exon sequences are represented by uppercase letters and straight, horizontal, boldface lines, intron sequences by lowercase letters and tented boldface lines, and splice junctions by light vertical lines. Complete intron sizes have been shown in base pairs for each variant. All sequences start with the 3Ј-boundary of exon 7 and represent the splice events to exon 9 of ␤ or to exon 8 of the remaining variants. The length of the protein-coding region in amino acids has been shown in boldface for each variant on the right. The availability (ϩ) of sequence data for each variant from a cDNA library or from RT-PCR (brain mRNA) is indicated on the far right.
153, and 53 bp, respectively, just downstream of their shared SH2 domain-coding region (Fig. 1A). Nucleotide sequence analysis of the inserts suggests that, in addition to the longest ␥ form containing the complete 153-bp insert, alternative splicing results in the removal of the 53-bp, the 100-bp, or the complete 153-bp inserts to result in the ␤, ␦, or ␣ variants, respectively. The genomic organization of the SH2-B gene indicates that the 53-nucleotide segment of the insert is located at the 3Ј-end of exon 7 and provides the alternative splice sites to produce the ␣, ␤, and ␥ variants (13). The ␣ and ␤ transcripts (Fig. 3A) use the alternative donor splice site located 53 nucleotides upstream of the normal exon 7/intron 7 boundary, thereby excluding 53 nucleotides at the 3Ј-end of exon 7 (13). The ␣ isoform also excludes the following exon 8 that is included in the ␤ transcript (13). In contrast, both ␥ and ␦ isoforms use the downstream splice site at the exon 7/intron 7 boundary. Exon 8 is included in the ␥ transcript (13) but excluded from the ␦ variant (Fig. 1A). Following the differential use of exon 8 (termed exon 8␤/8␥) all variants use the next downstream exon (termed exon 9␤/8␣) (13). The donor splice site from exon 8 and the acceptor splice site from exon 9 ␤ create the termination codon (Fig. 3B) in the ␥ transcript (13).
The splicing of pre-mRNA for the removal of intervening introns is a crucial post-transcriptional process requiring a coordinated action of several small nuclear ribonucleoprotein particles (snRNPs) and non-snRNP factors, which assemble on pre-mRNA into a large multicomponent complex, the spliceosome (27). Among the essential non-snRNP splicing factors are the highly conserved Ser/Arg-rich SR proteins that participate in the early stages of spliceosomal assembly and affect the selection of alternative splice sites (28,29). Different members of the SR protein family display distinct functions associated with preferential use of 5Ј-splice sites in alternative pre-mRNA splicing in vitro, in addition to exhibiting differential expression in a variety of mammalian cell types and tissues (30,31). Splicing enhancers located within exons of pre-mRNA can facilitate splicing of upstream introns that are subject to alternative splicing (32)(33)(34)(35). SR proteins recognize these exonic enhancer signals with distinct specificities (29, 31, 36 -39), which may strongly effect splice site selection (31,40). The combination of SR proteins displaying different specificities for pre-mRNA subclasses and the regulated levels of SR proteins in different cell types contribute to the modulation of cellspecific splice site selection (29,31,41). We anticipated that the expression of the variants would depend on tissue and/or cell type based on the presence of alternative splicing mechanisms. To evaluate the distribution of the four variants in different tissues, Northern analysis was not feasible because specific sequences are not available that would distinguish the variants by hybridization. With RT-PCR (Fig. 2) we typically observed a complementary expression pattern of either ␦ and ␣ in most tissues or ␤ and ␥ in others. These findings suggest that both alternative splice sites at the 3Ј-end of exon 7 of the SH2-B gene (13) are used together with preferential exclusion of exon 8 (termed exon 8␤/8␥) for the generation of ␦ and ␣ isoforms. Notably, both ␤ and ␥ use exon 8 (exon 8␤/8␥) ( Fig. 3; 13). Such differential splicing and exon selection events may be regulated at the tissue and/or cell level to determine the expression of the variants. In our experiments the signal was typically significantly higher for ␦ when compared with ␣ or for ␤ when compared with ␥. Our results suggest a possible cooperative action between ␦ and ␣ as well as between ␤ and ␥; typically only one of either pair was prominently present in any tested tissue. The variants will likely exhibit some functional redundancy in addition to specific roles.
Of considerable interest is the finding that all four protein variants exhibit unique carboxyl termini (Figs. 1A and 4A), which implicate functional differences in signaling pathways. All variants share the PH and SH2 domains and are identical up to residue Gln-631 (Fig. 4A). The ␤ isoform was shown to be localized at the plasma membrane of PC12 cells (21). If this subcellular localization occurs via the shared PH domain (21), the other variants may share similar intracellular localizations. Notable is the presence of nuclear localization signals in the unique carboxyl terminus of ␦ (Fig. 1A), which may implicate a nuclear role of this variant in gene expression, a promising area of future investigation. The ␣ variant constitutively interacts with Grb2, primarily via the SH3 domain of Grb2 and the Pro-rich region located between the PH and the SH2 domain of ␣ (20), implicating the other variants in similar associations. Major features in the unique carboxyl termini of ␣ and ␥ include the presence of Pro-rich motifs (Figs. 1A and 4A) that may recruit other SH3 domain-containing signaling mediators. Other structural features within the carboxyl terminus of ␣ include a sequence SFV (aa 754 -756) (Fig. 1A) that matches the consensus recognition motif ((S/T)XV) for PDZ domains, which are found in a variety of signaling proteins (42). Alpha also contains a sequence RSTSRDP (aa 633-639), which resembles the proposed Raf consensus sequence RSXpSXP, where serine is phosphorylated, as a binding motif for 14-3-3 proteins (43). 14-3-3⑀ binds phosphorylated serine of both the IGF-I receptor and IRS-1 (44). The ␤ variant carries a sequence PDASSTLLP (aa 650 -658) (Fig. 1A) similar to the 14-3-3 binding sequence found in the carboxyl terminus of the IGF-I receptor (44). The distinct carboxyl termini result in widely differing isoelectric points for the variants, pI 8.1 for ␥, 5.6 for ␦, 5.5 for ␤, and 5.0 for ␣ (45). The fact that PSM/SH2-B contains a PH domain, an SH2 domain, and several potential phosphorylation sites as well as Pro-rich and Ser-rich regions together with unique features associated with the carboxyl termini of the four variants, suggests multiple interactions with other signaling mediators to define differential roles in various signaling pathways.
Various reports implicated different motifs of IR in the interaction with PSM/SH2-B (8,11,13). In the absence of information about the family of four SH2-B splice variants these differences remained unresolved. It is conceivable that the variant-specific carboxyl termini influence the interaction of the adjacent SH2 domain with receptor Tyr kinases. For the ␣ variant the IR activation loop was implicated in the interaction with a glutathione S-transferase fusion protein, including the SH2 domain of ␣ (11). Earlier studies in this laboratory with the (at that time unrecognized) ␤ variant found that a glutathione S-transferase fusion protein, containing the SH2 domain and carboxyl-terminal sequences of PSM ␤, no longer bound to IR mutants lacking Tyr-960 in the juxtamembrane region or Tyr-1322 at the IR carboxyl terminus (8). These findings were supported by IGF-IR mutants lacking the homologous Tyr-950 or Tyr-1316 and by competition experiments with specific IR phosphopeptides representing Tyr-960 and Tyr-1322 (8). For the ␥ variant, Tyr-1146 in the catalytic loop of IR has been implicated in the association in yeast two-hybrid interaction studies and by some biochemical data (13). An association has been reported with the catalytic loop of TrkA (19,20; which shares similarity with the catalytic loop of IR) for an unspecified SH2-B form that correlates with the reported observation for ␥ (13). It is still possible that the differing  5. A and B, variant-dependent differential stimulation of IGF-I-and PDGFmediated mitogenesis. Responses were individually tested in NIH 3T3 fibroblasts to each of the four variants, represented by distinctly labeled bars. Responses were compared after transfection with variantspecific plasmids or control plasmid. For ecdysone-inducible plasmids expression was induced with varying doses of ecdysone (0, 2, 10) indicated in M (A) or alternatively, constitutive expression plasmids were used (B). Cells were starved to quiescence, and stimulated with IGF-I, PDGF, EGF, or left untreated (Ϫ) as indicated in the figure. Acid precipitated [ 3 H]-thymidine was quantified by liquid scintillation spectroscopy. All data points were measured in duplicate in each experiment as represented by the error bar and one representative of several experiments has been shown. C, variantdependent differential stimulation of IGF-I-and PDGF-mediated cell proliferation. Individual variants are represented by distinctly labeled bars as indicated in the figure. Responses were compared in NIH 3T3 fibroblasts after transfection with constitutive variant-specific plasmids or control plasmid. Cells in minimal serum were stimulated with IGF-I, PDGF, or were left untreated (Ϫ) as indicated in the figure. Cell proliferation was evaluated by quantifying cell numbers biochemically based on mitochondrial succinate dehydrogenase activity through the colorimetric change of MTT. All data points were measured in duplicate in the experiment as represented by the error bar.
observations are in part explained by the alternative experimental approaches taken in the various reports. However, it is compelling to speculate that PSM/SH2-B ␣ and ␥ variants may indeed share target motifs in their interaction with IR and other receptor Tyr kinases such as TrkA, whereas the ␤ variant may display distinct binding preferences. The exact receptor binding preferences of all four PSM variants remain to be verified in controlled, comparative experiments. These are now feasible and should include the new variant ␦.
We have begun to address the functional differences of the variants in signal transduction by comparing their impact on mitogenesis in fibroblasts. All variants stimulated DNA synthesis in response to IGF-I or PDGF but not to EGF (Fig. 5, A  and B), consistent with earlier studies focused on ␣ (10). Gamma consistently stimulated DNA synthesis most strongly followed by ␦, ␣, and ␤ with decreasing impact. For all variants DNA synthesis was responsive to the dose of ecdysone that controlled transcription of the cDNA (Fig. 5A). To address any role of the expression plasmid preparation and the individual cDNA transfection, experiments were repeated and carried out with constitutive expression plasmids (Fig. 5B). All experiments showed the same most pronounced mitogenic stimulation by ␥ followed by ␦, ␣, and ␤ with decreasing responses. The consistent absence of any observed modulation of the mitogenic response to EGF confirmed the specificity of the experimental approach. To address whether these observations could be generalized to the level of cell proliferation, experiments were repeated by scoring cell numbers with the MTT assay. Cell proliferation was similarly stimulated in response to IGF-I or PDGF most prominently by ␥, followed by ␦, ␣, and ␤ with decreasing impact (Fig. 5C). In combination, these results implicate differential roles of the variants in IGF-I-and PDGFmediated mitogenesis that remain to be elucidated at the molecular level.
We have begun to investigate potential molecular mechanisms of PSM/SH2-B activation by evaluating phosphorylation on Tyr. We found all variants to be phosphorylated on Tyr specifically in response to IGF-I and PDGF but not to EGF (Fig.  6). However, unlike the differential effect on mitogenesis, Tyr phosphorylation was similar between all variants. This is consistent with the presence of potential receptor target motifs only in the shared (2) but not in the unique variant sequences. Whereas Tyr phosphorylation is unlikely to explain differential roles of the variants in mitogenesis, it may be important in the molecular activation of PSM, which certainly warrants further investigation. Future experiments to elucidate the specific functional differences between the variants will focus on preferences for cellular partners and downstream mediators in established mitogenic signaling mechanisms.