Signal transduction pathways regulated by prolactin and Src result in different conformations of activated Stat5b.

Stat5 is activated by a broad spectrum of cytokines, as well as non-receptor tyrosine kinases, such as Src. In this study, the DNA binding properties of the two closely related Stat5 proteins, Stat5a and Stat5b, induced either by prolactin (Prl) or by Src were analyzed by electrophoretic mobility shift assays using several different Stat5 binding sites. Src-induced Stat5b-DNA binding complexes consistently displayed a slightly faster mobility than those induced by Prl, as well as differences in their ability to be supershifted by anti-Stat5 antibodies. IP-Westerns performed using specific antibodies directed at the N and C termini of Stat5b suggested that depending on the activating stimulus, Stat5b exhibited different conformations, which influenced antibody accessibility at its C terminus. These conformational differences may in part be due to differential effects of Prl and Src on Stat5b tyrosine phosphorylation, since Src induced several additional sites of tyrosine phosphorylation of Stat5b at residues other than Tyr-699, including Tyr-724 and Tyr-679. The latter Tyr-679 is conserved in all mammalian Stat5bs, but is not present in Stat5a. A Stat 5bY679F mutant induced by Src kinase exhibited an altered pattern of nuclear localization as compared with wild-type Stat5b. Furthermore, this mutation inhibited v-Src-induced cyclin D1-luciferase reporter activity in transient transfection assays performed in Stat5a/b-deficient MEFs, suggesting that Tyr-679 phosphorylation may play a role in v-Src induced proliferation. Thus, depending on the signal transduction pathway responsible for activation, different conformations of activated Stat5 may result in selective biological responses.

Signal transducers and activators of transcription (STAT 1 proteins) form a family of seven latent cytoplasmic transcription factors, which following activation in response to many different cytokines, dimerize and translocate into the nucleus to activate gene transcription (1). Stats can be activated by a broad spectrum of cytokines interacting with specific cytokine receptors, leading to the activation of members of the Janus tyrosine kinase (JAK) family. The activated JAKs phosphorylate the recruited monomeric Stats on a specific tyrosine and induce their dimerization. The activated dimers then translo-cate into the nucleus, where they bind to specific DNA-response elements in the promoters of target genes. This linear signal transduction pathway is clearly an oversimplification and can be modulated by cross-talk with other signal transduction pathways (e.g. by steroid receptors, etc.) (2).
Most Stats bind to similar GAS (IFN-␥-activated sequence) motifs (TTCN n GAA) with slightly different affinities for optimal binding; n ϭ 4 for Stat6 and n ϭ 3 for all other Stats (3)(4)(5). Non-receptor tyrosine kinases such as Src or Abl can also directly phosphorylate Stat proteins without involving JAKs (6 -9). In addition to tyrosine phosphorylation, Stat proteins can be phosphorylated on serine residues (10 -14). While regulation of Stats by serine phosphorylation is poorly understood, it has been shown that this phosphorylation can modulate both their transcriptional activity and DNA binding (13,15).
The DNA binding region of Stat proteins is located between amino acids 350 and 500, and the activation domain is located in the C-terminal region (16 -18). C-terminal-truncated Stat proteins still retain DNA binding activity, but may in some cases suppress transcriptional activation of responsive genes in a dominant-negative manner (19 -21). There is some evidence that these naturally occurring dominant-negative Stat5 isoforms may be functionally correlated with biological responses (22), but the potential physiological significance of dominantnegative Stat isoforms is still presently unknown. The N-terminal region of Stats is required for cooperative binding of Stat dimers to form a tetrameric structure (23,24). For example, Stat5 forms tetramers on the IL-2 response element in the human IL-2Ra promoter (23)(24)(25), and the CIS gene (26), both of which contain two adjacent Stat binding sites.
Stat5 plays a key role in mammary gland differentiation, as well as in the prolactin (Prl)-induced expression of milk protein genes, such as ␤-casein (27,28), ␣s 1 -casein (29), ␤-lactoglobulin (30), and whey acidic protein (31). However, the expression of Stat5 is not restricted to the mammary gland. Both isoforms of Stat5 (Stat5a and Stat5b, which share 93% identity at the amino acid level) can be activated by many cytokines in addition to Prl, such as IL-2 (32,33), IL-3 (34,35), , IL-7, IL-15 (33), thrombopoietin (36), erythropoietin, GM-CSF (35,37), EGF (38), and platelet-derived growth factor (39) in various cells and tissues, as well as by non-receptor tyrosine kinases such as Src and Bcr-Abl (6,7). Thus, Stat5 may play a variety of regulatory roles that control different functions including cell growth, survival, and differentiation. Recent studies have demonstrated for example, that activated Stat5 may regulate cyclin D1 promoter activity resulting in cell cycle progression (40 -42). Targeted disruption of Stat5a or Stat5b genes in mice results in a distinctive tissue-specific phenotype. Stat5a knockout mice display, for example, defects in mammary gland development and lactation during pregnancy (43), while Stat5b knockout mice display sexually dimorphic liver gene expression (44). There is also evidence suggesting that closely related Stat5a and Stat5b may be differently activated (45) and exhibit distinct biochemical differences (26) and distinct DNA binding specificities (46). However, the latter studies are controversial, because they were based on an incorrectly identified single amino acid difference between murine Stat5a and 5b. Instead, recent studies on the DNA binding properties of Stat5a and Stat5b suggest that baculovirus-expressed Stat5a and Stat5b homodimers recognize a similar optimal consensus sequence (47). Significant differences, however, have been observed in the ability of Stat5a and Stat5b to form tetrameric complexes on adjacent consensus and nonconsensus GAS sites, thus in part explaining their ability to activate nonidentical sets of genes (48). Thus, the arrangement of weak affinity binding sites in pairs on target genes may help provide selectivity in response to Stat5a tetrameric complexes. Selectivity might also reflect the cooperation with other transcription factors that interact with Stat5, and their own adjacent weak affinity binding sites perhaps helping to stabilize coactivator complexes (49).
Previous studies from our laboratory have suggested that Prl and Src signaling pathways may differentially regulate Stat5a and Stat5b nuclear translocation and retention (9). Thus, we postulated that the regulation of specific gene targets involved in proliferation, differentiation, or apoptosis by these two highly related isoforms of Stat5 might also reflect their activation by different signal transduction pathways. In this study, the DNA binding properties of Stat5a and Stat5b induced by Prl and Src to a variety of GAS-like sequences, such as ␤-casein, APRE (␣2-macroglobulin), APRE-2 (double APRE site), p21 (SIE2), and p21 (SIE3) were analyzed. As expected, Stat5a activated by Prl bound to the APRE-2 (double GAS site) as a tetramer, while Stat5b activated identically bound as a dimer. However, surprisingly, differences in the mobility of DNA binding complexes of Stat5b induced by Src versus Prl were observed. Based on immunoprecipitation studies and antibody supershift experiments, we propose that Stat5b induced by Src versus Prl may exhibit a different conformation at its C terminus, possibly due to additional tyrosine phosphorylation. In contrast to Prl, which induced only Tyr-699 phosphorylation of Stat5b, Src induced phosphorylation on Tyr-699 as well as on additional tyrosines. To identify the potential Src-induced tyrosine phosphorylation sites in the C-terminal region of Stat5b in addition to Tyr-699, a series of tyrosine to phenylalanine mutants were analyzed both individually and sequentially by IP-Western blots. Based on this analysis, Tyr-724 (conserved in both Stat5a and Stat5b) and Tyr-679 (a unique site present only in Stat5b, but absent in Stat5a) were shown to contribute to the Src-induced tyrosine phosphorylation of Stat5b. Furthermore, mutation of Tyr-679 to Phe in Stat5b altered its pattern of nuclear localization upon activation by Src kinase, and decreased v-Src-induced cyclin D1 promoter activity, suggesting that phosphorylation of this tyrosine may influence proteinprotein interactions most likely of the Stat5b C-terminal transactivation domain and possibly may influence cell cycle progression. Thus, depending on the signal transduction pathway responsible for activation, different structural conformations of activated Stat5b may result in selective biological responses.
Cell Culture, Transient Transfections, and Luciferase Assays-HeLa cells were maintained in Opti-MEM media (Invitrogen), supplemented with 3% fetal bovine serum and gentamicin (50 g/ml) in a 37°C and 5% CO 2 incubator. DNA constructs (10 g) were transiently transfected into HeLa cells by using a calcium phosphate precipitation technique (5Prime-3Prime, Boulder, CO). For the co-transfection experiments, 4 g of Stat5a or Stat5b expression vector and 6 g of PrlR or Src kinase construct were utilized. Before Prl or Src induction, the cells were incubated overnight in medium containing 10% charcoal-stripped horse serum, insulin (5 g/ml), gentamicin (50 g/ml), and hydrocortisone (1 g/ml) and then (for the Prl experiments only) stimulated with ovine Prl for 30 min (1 g/ml, lot AFP-10677C, kindly provided by the National Hormone and Pituitary Program, NIDDK, National Institutes of Health, Bethesda, MD). Alternatively, for reporter assays, Stat5a/5bdeficient MEFs derived from Stat5a/5b knockout mice (66) kindly provided by Dr. James Ihle, were transiently co-transfected with 0.5 g of Stat5b or a Stat5bY679F mutant, 1 g of v-Src and 1 g of a Ϫ674CD1-Luc reporter construct. Experiments were performed in 6-well plates, and each transfection was done in triplicate. Lysates were prepared 24 h after transfection, and luciferase activity was measured using a Promega luciferase kit (Madison, WI). Luciferase values were normalized to protein amount by Western blot analysis for Stat 5b.
Antibodies-The specific, affinity-purified rabbit polyclonal antiserum generated against peptides from the unique amino acid epitopes present in the C-terminal regions of Stat5a and Stat5b have been characterized previously (50,52). The N-terminal Stat5 (N-20) antibody was purchased from Santa Cruz Biotechnology (Santa Cruz, CA). Phosphotyrosine antibody (PY-20) was purchased from Transduction Laboratories (Lexington, KY). v-Src antibody was purchased from Calbiochem (La Jolla, CA). The site-specific antibody to Ser-730-phosphorylated Stat5b was kindly provided by Dr. Robert Kirken (University of Texas Medical School, Houston, TX).
Preparation of Cells Extracts-Cells were rinsed twice with ice-cold phosphate-buffered saline (9.1 mM dibasic sodium phosphate, 1.7 mM monobasic sodium phosphate, 150 mM sodium chloride, pH 7.4) and scraped into 2 volumes of Wu buffer (51) containing 10 mM HEPES, pH 7.4, 1.5 mM magnesium chloride, 0.1 mM EGTA, 10% glycerol, 1 mM dithiothreitol, 10 g/ml aprotinin, 10 g/ml antipain, 10 g/ml benzamidine, and 5 g/ml leupeptin. After lysis in Wu buffer, cells were centrifuged for 15 min at 4°C and 14,000 rpm. Supernatant was retained as the cytoplasmic extract. After washing of remaining pellets twice with 5 volumes of Wu buffer, they were resuspended in two volumes of the same buffer with 400 mM of sodium chloride. The samples were incubated on ice for 10 min at 4°C and then centrifuged for 15 min at 4°C and 14,000 rpm. The supernatants were dialyzed in Invitrogen microdialysis units for 2 h at 4°C against 1 liter of Wu buffer with 100 mM sodium chloride and were retained as nuclear extracts. Protein concentrations were determined using a Bio-Rad protein assay.
Oligonucleotide Labeling and Electrophoretic Mobility Shift Assays (EMSA)-Five oligonucleotides were designed, annealed, and used for the EMSA experiments (Table I). ␤-Casein and APRE-2 were endlabeled using the Klenow fragment of DNA polymerase with [␣-32 P]CTP, while all others were end-labeled by T4 polynucleotide kinase with [␥-32 P]ATP. EMSA were performed as described previously (55). Briefly, nuclear extracts (10 g for each sample) were preincubated in 10 l of Wu-buffer containing 100 mM sodium chloride and 2 g of poly(dI-dC) (Amersham Biosciences) for 15 min on ice. If antibody was used for supershift analysis, it was included in the 10 l of Wu buffer. After preincubation to block nonspecific binding and allow the antibody to bind, 5 l of binding mixture (50 g/ml p(dN) 6 , 2.5 mg/ml bovine serum albumin, 4% Ficoll 400, and 10% glycerol) with labeled probe (20,000 cpm) was added to each sample and incubated on ice for 15 min. The reaction mixture was loaded onto a 4% polyacrylamide gel in 0.25ϫ Tris borate-EDTA buffer, electrophoresed, dried, and subjected to autoradiography. Immunoprecipitation and Western blot analysis were performed as described previously (9).
Indirect Immunofluorescence-For indirect immunofluorescence cells were cultured on glass coverslips coated with poly-D-lysine, fixed and stained as described in (9). Images were obtained using a DeltaVision deconvolution microscope and softWoRx imaging workstation from Applied Precision, Inc. (Issaquah, WA).

DNA Binding Specificity of Stat5a and Stat5b Induced by
Prl-To explore DNA binding specificities of rat Stat5a and Stat5b to different DNA sequences that contain the consensus Stat5 binding motif, e.g. ␤-casein GAS site, as compared with those promoters with minor differences, e.g. p21-SIE2 and APRE GAS sites (Table I), HeLa cells were transfected with expression vectors for Stat5a or Stat5b and PrlR. After stimulation with Prl for 30 min, nuclear extracts were prepared, and EMSA was performed ( Fig. 1, panel A). As expected, following Prl activation, both Stat 5a and Stat 5b formed complexes with the single GAS sites in the ␤-casein, p21-SIE2, and to APRE probes with different intensities most likely reflecting slightly different affinities to the consensus and nonconsensus Stat5 binding sites (Fig. 1, panel A, lanes 4 and 6). A slight difference in the mobility of Prl-inducible DNA binding complexes for Stat5a and Stat5b was observed (compare lanes 4 and 6, Fig. 1, panel A), which also was expected based upon previous observations (9,46). Two supershifted complexes were observed in experiments using the anti-Stat5a C-terminal-specific antibody with all of the indicated probes ( Fig. 1, panel A, lane 7), while only one supershifted complex was seen with the anti-Stat5b C-terminal-specific antibody ( Fig. 1

, panel A, lane 8).
To control for Stat5a and Stat5b activation in these transiently transfected cells, immunoprecipitation (IP) experiments were performed using an anti-Stat5 N-terminal (N-20) antibody, which recognizes both Stat5 isoforms, followed by immunoblotting with an anti-phosphotyrosine antibody ( Fig. 1, panel B). Similar levels of tyrosine phosphorylation were observed for both Stat5a (Fig. 1, panel B, lane 4) and Stat5b (Fig. 1, panel B, lane 6). Direct Western blot analysis also indicated a similar level of expression of both proteins (Fig. 1, panel C). This experiment established the basic parameters for investigating Stat5a and Stat5b-DNA interactions using the nuclear extracts from Prl-treated HeLa cells.
Stat5a Binds as a Tetramer to the Duplicated APRE Probe That Contains Two Stat Sequences, whereas Stat5b Binds as a Dimer-Since both Stat5a and Stat5b dimers exhibited decreased affinity for the APRE probe containing a single, nonconsensus Stat5 binding site, we next wanted to confirm whether the expected differences in binding of the Stat5a tetramer to two copies of the APRE element (APRE-2) ( Table I) as compared with Stat5b dimer would be observed following Prl activation. Both Stat5a and Stat5b induced by Prl formed DNA complexes with the expected mobilities and intensities (Fig. 2,

panel A).
A slower mobility and more intense Stat5a tetrameric complex was observed bound to the APRE-2 probe (Fig. 2, panel A, lane 4), in contrast to the faster mobility, weaker Stat5b dimer complex (Fig. 2, panel A, lane 10). To verify whether Stat5a bound as a tetramer to APRE-2, DNA binding analysis of N-terminal mutants, Stat5aW37A and Stat5bW37A were performed. Since the Trp-37 residue is required for efficient tetramerization of Stat5 (23), this mutation should block tetramer formation and cooperative DNA binding of Stat5a to APRE-2. As shown in Fig. 2, panel A, this N-terminal mutation of Stat5a completely inhibited its DNA binding activity to APRE-2 (lanes 7 and 8), while the comparable mutation of Stat5b did not exhibit any significant effect on DNA binding (Fig. 2, panel A, lanes 13 and 14). Thus, following Prl activation Stat5a binds as a tetramer to the duplicated APRE sequence whereas Stat5b binds as a dimer, similar to previous results obtained with the human CIS promoter (26). Again, the supershift with the anti-Stat5b-specific antibody showed only one complex (Fig. 2, panel A, lanes 11 and 14), while multiple   67. d Bolded residues represent changes compared to the consensus sequences of Stat5a or Stat5b. complexes were detected with the anti-Stat5a-specific antibody (Fig. 2, panel A, lane 5). Finally, we investigated the DNA binding activity of Stat5b and its N-terminal mutant activated by Src tyrosine kinase to the APRE-2 probe (Fig. 2, panel B).

Stat5b consensus a N (A/tg)(T/A)TTC(C/T)(T/cag)(G/a)GAA(T/A)(T/ca)N
Both Stat5b (lane 5) and the Stat5bW37A mutant (lane 6) exhibited a similar mobility DNA binding complex when induced by Src suggesting that regardless of the mechanism of activation Stat5b interacted with the APRE-2 probe as a dimer.

DNA Binding Complexes of Prl-and Src-inducible Stat5b
Have Different Electrophoretic Mobilities-Previously it had been shown that Src induction resulted in the selective translocation and retention of Stat5b as compared with Stat5a in the nucleus, whereas no differences in their nuclear localization and retention were found following Prl activation (9). While the precise mechanisms responsible for these differences were unknown, analysis of chimeric proteins suggested that the critical determinants resided in the C-terminal half of Stat5b. To examine possible differences in the DNA binding specificities of Stat5b induced by Src versus Prl, a Stat5b expression vector was co-transfected into HeLa cells along with either Src or PrlR, as described under "Experimental Procedures." Surprisingly, the DNA-bound complexes of Stat5b induced by Prl (Fig.  3, panel A, lane 5) exhibited a consistently slower mobility compared with those induced by Src with all five probes examined (Fig. 3, panel A, lane 6). Interestingly, the supershift of Stat5b-DNA complexes induced by Prl with the anti-Stat5b C-terminal specific antibody was almost complete (Fig. 3, panel A, lane 7), while only a partial supershift of Stat5b-DNA complexes induced by Src was observed in parallel (Fig. 3, panel A, lane 8). This result suggested that the C terminus of Stat5b bound to DNA following Src induction might be less accessible to the anti-Stat5b C-terminal specific antibody than it was following Prl induction.
Unexpectedly, when the IP-Western controls were performed, different levels of tyrosine phosphorylation of Stat5b activated by Prl (Fig. 4, panel A, lane 5) as compared with Src ( Fig. 4, panel A, lane 6) were detected when Stat5b was immunoprecipitated using the anti-Stat5b C-terminal-specific antibody. Remarkably, the level of total Stat5b protein, detected by IP, was also different for Stat5b activated by Prl (Fig. 4, panel B, lane 5) versus Src (Fig. 4, panel B, lane 6), while direct Western blotting indicated that roughly equivalent amounts of Stat5b protein were present in the cell extracts (Fig. 4, panel C, lanes 5 and 6). Because proteins detected by IP are more likely to exist in their native conformation, as compared with the denatured state present in SDS-PAGE and Western blotting, we hypothesized that this difference in the total levels of Stat5b detected following IP might be explained by the decreased accessibility of the C-terminal epitope following Src as compared with Prl activation. To test this hypothesis, we performed IP-Westerns using an anti-Stat5b N-terminal antibody. Interestingly, these experiments showed similar levels of Stat5b expression and tyrosine phosphorylation induced by Prl (Fig. 4, panels D and E, lanes 5 and 6) and by Src (Fig. 4, panels  D and E, lanes 5 and 6).
Remarkably, using an anti-Stat5b C-terminal antibody for immunostaining, we observed routinely that the percentage of cells with nuclear staining for Stat5b induced by Src was ϳ60 -70% less than that observed in cells induced by Prl (data not shown), confirming the results of the immunoprecipitation and antibody supershift experiments described above. Based on all these results taken together, we postulate that the DNA-bound complexes of Stat5b, stimulated with Src and Prl have different conformations of their C-terminal transactivation domains, which might be a result of differential phosphorylation.

Stat5b Induced by Src Displays Additional (to Tyr-699) Tyrosine Phosphorylation as Well as Serine Phosphorylation at
Ser-730 -To determine if there was differential phosphorylation of Stat5b as a consequence of Src as compared with Prl activation, HeLa cells were co-transfected with expression vectors for Stat5b, or a Stat5bY699F mutant, and PrlR or Src. After stimulation with Prl for 30 min, cell extracts were prepared and subjected to immunofluorescence, EMSA and immunoblotting analysis. Since Tyr-699 of rat Stat5 is the site of phosphorylation by JAK-2 and the primary regulator of Stat5 dimerization and subsequent DNA binding (27), the mutation of this Tyr to Phe should block these processes. As expected, nuclear localization and DNA binding of Stat5b induced by either Src or Prl was completely prevented by the Y699F mutation (data not shown). However, IP-Westerns performed with the anti-Stat5b C-and N-terminal-specific antibodies followed by blotting with anti-phosphotyrosine antibody revealed additional sites of tyrosine phosphorylation of Stat5b induced by the Src tyrosine kinase (Fig. 4, panels A and D, lane 9). In contrast, no additional tyrosine phosphorylation of Stat5bY699F in response to Prl was detected (Fig. 4, panels A  and D, lane 8). Equivalent amounts of Stat5b were expressed (Fig. 4, panel C) and immunoprecipitated by N-terminal Stat5b antibody (Fig. 4, panel E) from all cell lysates. Differences in the levels of Stat5b following activation by Src and Prl were once again detected when the IP was performed with the anti-Stat5b C-terminal-specific antibody (Fig. 4, panel B) as discussed above.
Ser-730 has been characterized previously as a major site of Prl-inducible serine phosphorylation of Stat5b (11). Accordingly, we wanted to determine if differential effects of Src and Prl on the phosphorylation of Ser-730 in Stat5b might also explain the apparent conformational differences observed in the C terminus. For this purpose, a Stat5bS730A mutant was co-transfected into HeLa cells along with PrlR and Src expression vectors. After stimulation with Prl, cell extracts were prepared and subjected to IP-Westerns using the anti-Stat5b N-terminal-specific antibody followed by blotting with a specific anti-phosphoserine antibody (␣-PS730) that recognizes phospho-Ser-730 (Fig. 5). Activation by both Prl (Fig. 5, panel A, lane 5) and Src (Fig. 5, panel A, lane 6) resulted in increased Stat5b phosphorylation at Ser-730 and, as expected, in both cases the S730A mutation completely prevented this increase (Fig. 6, panel A, lanes 8 and 9). These differences (Fig. 5, panel A, lanes 5 and 6 compare with lanes 8 and 9) cannot be attributed to slightly different levels of Stat5b and Stat5bS730A expressed in these experiments (Fig. 5, panel B).

Tyr-679 and Tyr-724 Are Potential Sites of Src-induced Tyrosine Phosphorylation in Stat5b in Addition to Tyr-699 -
Since Src-induced tyrosine phosphorylation appeared to influence the conformation of the C-terminus of Stat5b, it was important to determine if the naturally occurring C-terminaltruncated isoform Stat5b⌬40 exhibited any differences in tyrosine phosphorylation on tyrosine residues others than Tyr-699 relative to wild-type Stat5b. Since the truncated Stat5b⌬40 mutant does not contain any additional tyrosines as compared with wild-type Stat5b (Fig. 6, panel A), it was predicted that the 40-amino acid C-truncation of Stat5b might not affect the Src-induced tyrosine phosphorylation at residues other than Tyr-699. To test this prediction, HeLa cells were co-transfected with expression vectors for Stat5b, Stat5bY699F, Stat5b⌬40, Stat5b⌬40Y699F, and Src. Cell extracts were prepared and analyzed by IP-Westerns. As expected, both mutants Stat5bY699F and Stat5b⌬40Y699F exhibited equivalent levels of tyrosine phosphorylation at one or more additional tyrosines (Fig. 6, panels B-D).
To determine the potential involvement of specific tyrosine residues, other than Tyr-699, in the Src-induced tyrosine phosphorylation of Stat5b, mutational analysis of all the tyrosines present at its C-terminal end (Fig. 6, panel A) was then performed. Tyrosine mutations were generated on the Stat5bY699F mutant by replacing all the tyrosines from Tyr-742 to Tyr-668 with phenylalanine in a sequential manner. Specific Tyr to Phe mutants were transfected into HeLa cells activated with Src. IP-Westerns were performed using the specific anti-Stat5b N-terminal antibody for immunoprecipitation, followed by blotting with an anti-phosphotyrosine antibody. The level of tyrosine phosphorylation (Fig. 7, panel A) was normalized to the recovery of total Stat5b. No significant difference in the level of tyrosine phosphorylation of Stat5bY699F in response to Src was observed following mutation of Tyr-742 and Tyr-739 to Phe. However, the additional mutation Y724F resulted in a 2-fold decrease in phosphotyrosine levels. Interestingly, when the two neighboring tyrosines, Tyr-683 and Tyr-682, were changed to phenylalanines, an increase in the level of phosphotyrosine was observed. Additional mutation of Tyr-679, the only conserved tyrosine residue present in Stat5b but not in Stat5a, to Phe markedly reduced the total level of Src-induced tyrosine phosphorylation. These sequential Tyr to Phe mutation experiments suggested the possibility that two tyrosines, Tyr-724 and Tyr-679, in addition to the major site of tyrosine phosphorylation at Tyr-699, might contribute to the level of Src-induced tyrosine phosphorylation of Stat5b.
To confirm this result, individual tyrosine to phenylalanine mutations at the C-terminal end of Stat5bY699F were generated, and these double mutants were tested for their levels of Src-induced tyrosine phosphorylation as described above. The results of the double mutant analysis (Fig. 7, panel B) correlated well with those obtained following the sequential mutation of all the C-terminal tyrosine residues (Fig. 7, panel A). No significant changes in phosphorylation were observed for Y699F/Y742F and Y699F/Y739F mutants. In contrast, a decreased level of tyrosine phosphorylation was seen with the Y699F/Y724F mutant, and an increased level of tyrosine phosphorylation was once again observed in the Y699F/Y683-82F mutant. Finally, very low levels of tyrosine phosphorylation (Ͻ1%) were detected for mutant Y699F/Y679F. Tyrosine phos-phorylation also was not detected for the Y699F/Y668F mutant, but Tyr-668 is located within the SH2 domain of Stat5b (Fig. 6, panel A).

Mutation of Tyr-679 to Phe Resulted in an Altered Pattern of Stat5b Nuclear Localization Induced by Src Kinase, and Decreased v-Src Activation of the Cyclin D1 Promoter-Because
Tyr-679 is a unique site of tyrosine phosphorylation present in Stat5b, but not Stat5a, and the mutation of this residue resulted in a major decrease in Src-induced Stat5b tyrosine phosphorylation, the potential functional consequences of phosphorylation of this particular tyrosine were investigated using both cell and molecular approaches. Because of the previously observed differences in nuclear localization of Stat5b as compared with Stat5a following Src activation (9), the effects of mutation of Tyr-679 to Phe, on the pattern of nuclear localization of Stat5b were examined. For this study, HeLa cells were cotransfected with expression vectors for either the wild-type Stat5b or the Stat5bY679F mutant together with c-Src or PrlR and analyzed using deconvolution microscopy (Fig. 8). As expected, a similar punctate pattern of nuclear staining resistant to Triton-csk treatment was observed for Stat5b activated either by Src (Fig. 8A) or by Prl (data not shown). While the Stat5bY679F mutant was localized in the nucleus in response to both prolactin and Src activation (Fig. 8, C and B, respectively), a significantly different pattern of nuclear staining was observed for this mutant, as compared with wild-type Stat 5b. Thus, a punctate pattern of nuclear staining with a few large speckles was observed for the Stat5bY679F mutant following Prl activation (Fig. 8C) similar to that seen for wild-type Stat5b after stimulation with either Prl or Src (Fig. 8A). However, in cells co-expressing Stat5bY679F and Src, a very different nuclear pattern with numerous tiny speckles was observed (Fig. 8B).
The effect of the Stat5bY679F mutation on Src-stimulated gene transcription was evaluated in transient transfection experiments using a cyclin D1 promoter-luciferase reporter construct previously shown to respond to both v-Src and Stat5 activation (40,42,68). This is a complex promoter, which responds to a numerous signal transduction pathways, and contains multiple transcription factor binding sites, including Ϫ481Stat5 (40, 42) and Ϫ58CREB/ATF-2 (68) binding sites. To eliminate effects of endogenous Stat5b present in HeLa cells and to provide a more appropriate genetic background in which to test specific Stat5b mutants, these transfection experiments were carried out in Stat5a/5b-deficient MEFs. These studies were performed with v-Src rather than c-Src, because the increased transcriptional activation of the cyclin D1-luciferase reporter in the Stat5a/5b-deficient MEFs (6 -7-fold, as compared with 1.5-2.0-fold) provided a better model in which to compare wild-type and mutant Stat5b constructs (Fig. 9, panel A, lane 2). In cells co-expressing v-Src and wild-type Stat5b a slight increase in cyclin D1 promoter activity was observed (Fig. 9, panel A, lane 3) as compared with v-Src alone (Fig. 9,  panel A, lane 2). This is not unexpected because the majority of v-Src-induced cyclin D1 promoter activity is dependent on the CREB/ATF-2 site (68). Surprisingly, however, a significant reduction in cyclin D1 promoter activity (comparable to that seen with v-Src alone) was detected following transfection of the Stat5bY679F mutant (Fig. 9, panel A, lane 4). Western blot analysis indicated that approximately equivalent levels of the wild-type and mutant Stat5b proteins were expressed in these transient transfection assays (Fig. 9, panel B, lanes 3 and 4). DISCUSSION Previous studies from our laboratory demonstrated that ligand-dependent and -independent signal transduction pathways may differentially regulate Stat5a and Stat5b nuclear translocation and, therefore, have the potential to differentially regulate Stat5-dependent gene expression (9). Prl-dependent activation of Stat5a and Stat5b was dependent on JAK2 activation and resulted in punctate detergent-resistant nuclear complexes in COS and HeLa cells. In contrast, the selective translocation of Stat5b to the nucleus following Src and Bcr/Abl activation did not appear to be dependent on JAK2 activation. Analysis of Stat5a and Stat5b chimeric proteins suggested that the determinant responsible for these differences resides in the C-terminal half of the protein.
The present studies were initiated to determine if there might be differences in the DNA binding specificity of Stat5b activated by Prl versus Src. It had been suggested that a single amino acid in the DNA binding region of Stat5a and Stat5B could confer distinct DNA binding specificities (46). However, it is now apparent that this amino acid change (E433G) does not even normally occur in murine Stat5b, and that the DNA binding domain, and especially the amino acid residues that contact the DNA helix, are highly conserved in Stat5a and Stat5b from all mammalian species studied to date (56). Not surprisingly, therefore, careful binding site selection studies performed with baculovirus-expressed Stat5a and Stat5b have generated identical consensus high affinity binding sites (47).
One principal difference between Stat5a and Stat5b appears to be in their abilities to form tetramers and interact in a cooperative manner with tandem nonconsensus Stat5 binding sites (23,48). Tetramer formation is regulated through interactions between ␣-helices present in the N-terminal domain of Stats (18) and is critically dependent on a single tryptophan residue (Trp-37) conserved in all Stats. In light of these recent studies, it is not surprising that no major differences were observed in Prl-versus Src-activated Stat5b binding to single consensus or nonconsensus Stat5 binding sites or to tandem nonconsensus Stat5 binding sites (Figs. 1 and 2). However, differences in the mobility of Stat5b DNA complexes were observed, which were dependent on the signaling pathway used for activation (Fig. 3). These differences appear to reflect differences in antibody accessibility to the C-terminal epitope in Stat5b. This was established by the direct comparison in IP-Westerns of anti-Stat5 C-terminal and N-terminal antibodies (Fig. 4). Nuclei were permeabilized and extracted as described previously (9) following by fixing and staining with a specific C-terminal Stat5b antibody conjugated with Texas Red (red) and with DAPI (blue). Images were obtained using a DeltaVision deconvolution microscope as described under "Experimental Procedures." Differences in the accessibility of the C terminus of Stat5b may exert profound effects on Stat5b function. The C-terminal region of Stat5 has the potential to form an amphipathic ␣-helix, and this region has been shown to be required for both transcriptional activity and proteasome-dependent turnover of tyrosine-phosphorylated Stat5 (16,57). Naturally occurring isoforms of Stat5b containing a 40-amino acid C-terminal deletion result in the generation of a dominant-negative protein with sustained DNA binding (16,51). Thus, it is likely that Src-induced changes in Stat5b C-terminal accessibility will influence protein-protein interactions with co-activators and other transcription factors leading to the differential regulation of a subset of Stat5b-regulated target genes. In contrast, Prlactivation of Stat5a, which is capable of tetramer formation will selectively activate a subset of target genes containing tandem nonconsensus Stat5 binding sites.
The most likely explanation for these differences in antibody recognition is that differential phosphorylation of Stat5b by Src and Prl might result in different conformations of the C-terminal region of the protein and altered protein-protein interactions. One likely candidate was the phosphorylation of Ser-730 in Stat5b, since this has been shown to be prolactin-regulated, and, although not essential for DNA binding, may influence the duration of DNA binding following Prl stimulation (10,11). Since the inducible phosphorylation of Ser-730 in Stat5b was observed following both Prl and Src activation, this does not appear to be the basis for any conformational differences observed (Fig. 5). However, a number of other serines in Stat5 are known to be constitutively phosphorylated in mammary epithelial cells and may provide other potential targets for Srcinduced modifications (10).
Although no differences in the phosphorylation of Ser-730 were observed following Prl-and Src-activation of Stat5b, additional tyrosine phosphorylation of Stat5b was detected following Src activation of Stat5b using a Stat5bY699F mutant, in which the principal site of JAK2-dependent tyrosine phosphorylation was blocked (Fig. 4). Interestingly, EGF has been reported to induce Stat5a tyrosine phosphorylation at a site or sites in addition to Tyr-694, the major prolactin-regulated site of tyrosine phosphorylation in Stat5a (8). This effect also appears to be mediated through the activation of c-Src by the erbB receptor, since tyrosine phosphorylation was blocked by pharmacological Src kinase inhibitors and a dominant negative c-Src. Phosphorylation of Stat5a on tyrosines other than Tyr-694 has also been observed in response to interleukin-3 and thrombopoietin stimulation (58).
In the present study, all of the tyrosine residues C-terminal to the SH2 domain of Stat5b were analyzed following mutagenesis to phenylalanine, both sequentially and individually, in order to determine potential sites of Src-induced tyrosine phosphorylation. These studies demonstrated that there are several tyrosines in addition to Tyr-699 that are potential targets for Src-induced phosphorylation. Specifically, Tyr-724 (conserved in both Stat5a and Stat5b) and Tyr-679 (unique in Stat5b and not present in Stat5a) were both potential residues that contributed to the tyrosine phosphorylation of Stat5b in response to Src kinase. Interestingly, these experimental results coincided with predicted sites of tyrosine phosphorylation. Among 21 tyrosines present in Stat5b, 5 tyrosines (Tyr-548, Tyr-668, Tyr-679, Tyr-699, and Tyr-724) were identified by the prediction network to be potential sites of phosphorylation (59). Tyr-724 and Tyr-679 were identified to be the most likely potential sites with predicted scores of 0.860 and 0.832 respectively, followed by Tyr-699, with a score of 0.972. These three tyrosines (Tyr-699, Tyr-724, Tyr-679) are all located in the Cterminal portion of Stat5, in a region that was suggested to be highly disordered and extremely flexible by molecular modeling based on the x-ray crystal structures determined for both Stat1 and Stat3 (60,61). While Tyr-699 has been shown to be critical for Stat5 dimerization and interacts with the SH2 domain, the functions of Tyr-724 and Tyr-679 are less well defined. Conceivably, phosphorylation at these sites can influence the conformation of the C terminus of Stat5b and proteinprotein interactions, as illustrated by the IP-Westerns performed with the antibodies to the unique C-terminal epitope present in Stat5b as contrasted to those performed with the anti-Stat5 N-terminal antibody (Fig. 4).
The present mutagenesis studies also suggested that Tyr-668 may be a site for Src-dependent tyrosine phosphorylation of Stat5b, because no tyrosine phosphorylation was observed for the Stat5bY699F/Y668F double mutant (Fig. 7B). However, this site is located within the SH2 homology region (11,48,(61)(62) and, therefore, mutagenesis of this tyrosine in addition to Tyr-699 may influence the structure of the SH2 domain and dimerization of Stat5b, thus preventing Src-induced tyrosine phosphorylation. Similar results have been observed for IFNinduced Stat3 tyrosine phosphorylation when analyzing a Y656F mutant (63).
Tryptic peptide mapping of Stat5b following activation by EGF has been utilized in order to identify additional sites of tyrosine phosphorylation other than Tyr-699 (64). These studies identified two series of five and three peptides other than the peptide containing the major site of tyrosine phosphorylation, Tyr-699, suggesting that one or more tyrosines may be sites of additional phosphorylation. However, due to the low yields of the tyrosine phosphorylated peptides it was not possible to perform amino acid analysis to identify the specific sites of tyrosine phosphorylation induced by EGF. Instead, the authors also employed site-directed mutagenesis to further characterize the putative sites of tyrosine phosphorylation, similar to the approach employed in our study. Mutation of Tyr-725 led to a slight, but consistent, decrease in the EGFinduced tyrosine phosphorylation of Y699F Stat5b (64), in agreement with our results obtained for the Src-induced tyrosine-phosphorylated mutant, Y724F (Fig. 7, panels A and B) (The different numbering reflects the use of rat versus human Stat5b in that two studies.) However, two other tyrosines Tyr-740 and Tyr-743 (designated as Tyr-739 and Tyr-742 in our study) were also shown to be critical for EGF-induced tyrosine phosphorylation (64), but interestingly these sites were not found to play a significant role in Src-induced tyrosine phosphorylation of Stat5b (Fig. 7, panels A and B). These EGFinduced sites of tyrosine phosphorylation are present both in Stat5a and Stat5b, and Stat5a is known to be activated by the erbB family of receptor tyrosine kinases (8), suggesting that these additional tyrosines cannot be responsible for the differential activation of target genes by Stat5b. In fact, the reporter gene used in these studies was a multimerized LHRR GAS site reporter construct, and mutation of specific tyrosines resulted in an increased basal activity of this artificial reporter construct. While this construct was induced by EGF, the parental ␤-casein gene promoter containing this site in its appropriate context is actually repressed by EGF induction (65). On the other hand, our results indicated that Tyr-679, the unique site of tyrosine phosphorylation in Stat5b, appears to contribute to the of Src-induced tyrosine phosphorylation of Stat5b (Fig. 7,  panels A and B).
Using deconvolution microscopy to obtain high-resolution images, an altered pattern of nuclear staining for the Srcinduced Stat5bY679F mutant as compared with wild-type Stat5b was observed (Fig. 8, compare panels A and B). The observation that a mutation of a single tyrosine to a phenylala-nine in Stat5b resulted in a different pattern of nuclear localization helps support the biological importance of Src-induced phosphorylation of Tyr-679. This altered pattern of nuclear localization may reflect different conformations of the C-terminal transactivation domain of Stat5b as indicated by the IP-Western experiments shown in Fig. 4 and possibly its interactions with co-activators such as p300/CBP and p160 (69). More detailed co-localization experiments with different markers for splicing factors, co-activators, PML bodies, etc. will be required to determine the significance of these different patterns of nuclear localization with respect to gene regulation.
The decrease in v-Src-induced cyclin D1 promoter activity observed in the presence of the Stat5bY679F mutant ( Fig. 9) also suggests that phosphorylation of Tyr-679 may play a functional role in v-Src-induced proliferation. While Stat5 has been shown to mediate the transcriptional regulation of cyclin D1, thereby contributing to cytokine-and prolactin-dependent growth of hematopoietic and breast cancer cells, respectively (40,42), its role in v-Src induction of cyclin D1 and proliferation is less clear. The cyclin D1 promoter contains binding sites for a number of transcription factors other than Stat5, whose activity may be modulated either directly or indirectly by Src kinase (68). Preliminary studies performed using constructs containing mutations in the Stat 5 DNA binding site (Ϫ674CD1LUC/SIE1-mut), as well as truncation mutants containing only Ϫ163 bp of the cyclin D1 promoter (Ϫ163CD1LUC), suggest that only part of the inhibitory effect of the Stat5bY679F mutant on v-Src induced cyclin D1 promoter activity requires Stat5 DNA binding (data not shown). Thus, it is likely that the altered conformation of the Stat5b C-terminal region induced by phosphorylation of Tyr-679 influences cyclin D1 promoter activity in part through interaction with other transcription factors, such as LEF1 and CREB/ ATF2 and/or co-activators as discussed previously.
In conclusion, these experiments demonstrate that liganddependent signal transduction pathways exemplified by Prl activation, and ligand-independent pathways as illustrated by Src activation, may exert differential effects on the tyrosine phosphorylation of Stat5b, and influence antibody recognition of its C-terminal epitope. As demonstrated previously, Src activation may also result in the preferential nuclear translocation and retention of Stat5b. Activated Stat5b may then interact preferentially as a dimer with single consensus Stat5 DNA binding sites. These differences may, therefore, help to explain the observation from gene knockouts in mice that Stat5a and Stat5b play both "essential and nonredundant roles" not only in cytokine signaling (66), but presumably also in mediating the effects of other signaling pathways. One consequence of these effects may be the potentiation of oncogenic signals leading to malignancy. The definition of specific gene targets in different cell types that are selectively regulated by these pathways is an important area for future investigation.