Hydrophobic Residues Phe 751 and Leu 753 Are Essential for STAT5 Transcriptional Activity*

One facet of cytokine signaling is relayed to the nucleus by the activation, through tyrosine phosphorylation, of latent cytoplasmic signal transducers and acti-vators of transcription (STAT) family members. It has been demonstrated that the C termini of STATs contain the transactivation domain and are essential for the transactivation of target genes. To better understand the function of the STAT C terminus, we have generated a series of C-terminal mutants in STAT5a and examined their effects on transactivation, tyrosine phosphorylation, and DNA binding. Using GAL4 chimerae with the C terminus of STAT5, we have defined a 12-amino acid region essential for STAT5 transactivation. Surpris-ingly, deletion of these 12 amino acids in the context of the native STAT5 backbone preserved the overall transcriptional activity of the protein. Further analysis revealed that deletion of this region resulted in hyper-DNA binding activity, thus compensating for the weakened transactivation domain. Using site-directed mutagenesis, we show that within this 12-amino acid region the acidic residues were non-essential for transactivation. In contrast, the non-acidic residues were crucial for transactivation. Mutating either Phe 751 or Leu 753 to alanine abolished transactivation suggesting that these residues were essential for connecting STAT5 to the basal transcriptional machinery. Cytokines initiate their pleiotropic effects on cells by binding to specific transmembrane receptor proteins. This association induces a conformational change in the receptor (1, 2) that signals to the nucleus via a direct

One facet of cytokine signaling is relayed to the nucleus by the activation, through tyrosine phosphorylation, of latent cytoplasmic signal transducers and activators of transcription (STAT) family members. It has been demonstrated that the C termini of STATs contain the transactivation domain and are essential for the transactivation of target genes. To better understand the function of the STAT C terminus, we have generated a series of C-terminal mutants in STAT5a and examined their effects on transactivation, tyrosine phosphorylation, and DNA binding. Using GAL4 chimerae with the C terminus of STAT5, we have defined a 12-amino acid region essential for STAT5 transactivation. Surprisingly, deletion of these 12 amino acids in the context of the native STAT5 backbone preserved the overall transcriptional activity of the protein. Further analysis revealed that deletion of this region resulted in hyper-DNA binding activity, thus compensating for the weakened transactivation domain. Using site-directed mutagenesis, we show that within this 12-amino acid region the acidic residues were non-essential for transactivation. In contrast, the non-acidic residues were crucial for transactivation. Mutating either Phe 751 or Leu 753 to alanine abolished transactivation suggesting that these residues were essential for connecting STAT5 to the basal transcriptional machinery.
Cytokines initiate their pleiotropic effects on cells by binding to specific transmembrane receptor proteins. This association induces a conformational change in the receptor (1, 2) that signals to the nucleus via a direct pathway involving the activation of receptor-associated Janus tyrosine kinase (JAK), 1 and the subsequent phosphorylation and nuclear translocation of members of the signal transducer and activator of transcription (STAT) family (3,4). Once activated, JAK phosphorylates the receptor on cytoplasmic tyrosine residues providing sites for STAT proteins to bind to via their Src homology 2 (SH2) do-mains (5)(6)(7)(8). The juxtaposed JAK phosphorylates STAT on a tyrosine residue C-terminal to its SH2 domain (9 -12). Phosphorylated STAT dissociates from the receptor and homo-or heterodimerize through a reciprocal interaction between the phosphotyrosine of one STAT and the SH2 domain of the other STAT (13,14). These active dimers translocate to the nucleus, possibly through a nucleoprotein interactor-mediated process (15), where they bind specific DNA elements to activate transcription of target genes (16,17).
Seven mammalian STATs, ranging in size from ϳ90 to ϳ115 kDa, as well as several alternatively spliced variants have been documented (3,4). The most conserved feature about the STAT family is their SH2 domain (13,14). As already alluded to, this region is indispensable for activation and dimerization of the STAT proteins. N-terminal to this domain, but separated by a linker region, is the DNA-binding domain (17)(18)(19). The crystal structures for DNA-bound STAT-1 and -3 homodimers, lacking their C termini, have revealed that STAT dimers bind DNA in a saddle-like structure (18,19).
There are several lines of evidence to suggest that the Nterminal region of STATs is involved in protein-protein interaction. The N terminus of STAT-1 is required for the tyrosine dephosphorylation of STAT-1 and has been shown to regulate the nuclear translocation of and to mediate the interaction between STAT dimers (20 -23). Furthermore, the transcriptional co-activator, p300/CBP, can associate with the N terminus of STAT-1 (24). Recently, it was shown that Nmi can interact with the N-terminal coiled-coil domain of several STATs and enhance their function (25). In particular, association of Nmi with STAT1 and STAT5 facilitated the recruitment of p300/CBP to the complex and enhanced their transactivation following interferon-␥ and IL-2 stimulation, respectively.
The least conserved feature of STATs is their C termini. Alternatively spliced variants lacking the C-terminal region as well as mutants in which the C terminus was truncated has provided evidence that this region acts as the transactivation domain (11,26,27). In the case of STAT5, these mutants show sustained tyrosine phosphorylation and exhibit increased levels of DNA binding activity compared with the wild-type protein but remain transcriptionally inert (26,28). In addition, the C-terminal region of STATs has been described to interact with various co-activators, such as MCM5 (29) and p300/CBP (24, 30 -34).
These observations suggest that the C terminus of STATs has pleiotropic functions and plays an important regulatory role. To better understand the function of the C termini of STATs, we have generated a series of C-terminal mutants in STAT5a and examined their effects on transactivation, tyrosine phosphorylation, and DNA binding. Using GAL4 chimerae with the C terminus of STAT5, we have identified a 12-amino acid region essential for STAT5 transactivation. Surprisingly, deletion of these 12 amino acids in the context of the native STAT5 backbone preserved the overall transcriptional activity of the protein. Further analysis revealed that deletion of this region resulted in hyper-DNA binding activity, thus compensating for the weakened transactivation domain. Using sitedirected mutagenesis, we show that, within this 12-amino acid region, acidic residues were non-essential for transactivation. In contrast, the non-acidic residues, especially the hydrophobic residues Phe 751 and Leu 753 , were crucial for transactivation.

EXPERIMENTAL PROCEDURES
Chemicals, cDNAs, and Antibodies-Recombinant murine IL-3 was purchased from R&D (Minneapolis, MN). Recombinant murine prolactin was obtained from the National Hormone and Pituitary Program (NIDDK, National Institutes of Health, Bethesda, MD). Staurosporine was obtained from Sigma, dissolved in dimethyl sulfoxide (Me 2 SO), and used at a final concentration of 500 nM. Murine HA-JAK2 and HA-STAT5a cDNAs were obtained from Alan D. D'Andrea (Dana-Farber Cancer Institute, Boston, MA). Murine prolactin receptor cDNA was generously provided by Bernd Groner (Institute for Experimental Cancer Research, Tumor Biology Center, Freiburg, Germany). An internal ribosome entry site/green fluorescent protein (IRES/GFP) cDNA was obtained from Richard Mulligan (Children's Hospital, Boston, MA). Anti-HA antibody (12CA5) was purchased from Roche Molecular Biochemicals. Mouse anti-phosphotyrosine-specific STAT5 monoclonal antibody (ST5P-4A9) was purchased from Zymed Laboratories Inc. (South San Francisco, CA).
cDNA Plasmids-The truncated STAT5 mutant (⌬737) was generated by introducing a stop codon at amino acid 737. A Y694S mutation in STAT5a was introduced into its cDNA by the polymerase chain reaction, generating a BamHI site. This BamHI site was used to subclone the 3Ј end of the STAT5a coding region (a 306-base pair BamHI/ EcoRI fragment) into pBluescript (pBS-CT-STAT5). All subsequent mutants were generated by polymerase chain reaction on this construct using the Quickchange site-directed mutagenesis kit (Stratagene, La Jolla, CA). All mutations were confirmed by sequencing and the NsiI/ EcoRI fragment subcloned back into the full-length, wild-type cDNA in the mammalian expression vector pcDNA3. To generate the GAL4-Cterminal STAT5 fusion constructs, the BamHI/EcoRI fragment of the various mutants in pBS-CT-STAT5 were subcloned into the BamHI/ EcoRI sites of the GAL4 DNA-binding domain fusion expression vector, pFA-CMV (Stratagene). All cDNAs were purified using Qiagen Maxiprep kits (Qiagen Inc., Valencia, CA).
COS-7 Transfections and Extracts-Confluent COS-7 cells were split one into five the day prior to transfection. Approximately 50% confluent plates were transiently transfected for 5 h at 37°C with 1 g of JAK2 and 5 g of STAT5 cDNAs per dish using LipofectAMINE/Opti-MEM (Life Technologies, Inc.) according to the manufacturer's directions. Cellular extracts were prepared at either 24 or 48 h after transfection. Cells were washed with phosphate-buffered saline, resuspended in lysis buffer (1% Nonidet P-40, 50 mM Tris.HCl, pH 7.5, 150 mM NaCl, 0.5% sodium deoxycholate, 50 mM NaF, 0.5 mM dithiothreitol, 0.2 mM phenylmethylsulfonyl fluoride, 1 mM Na 3 VO 4 , 2 g/ml aprotinin C, and 0.5 g/ml leupeptin), and incubated on ice for 30 min. Extracts were centrifuged (4°C) for 10 min at 13,000 rpm, and the resulting supernatants were used for subsequent immunoprecipitations, direct immunoblotting, or electrophoretic mobility shift assay (EMSA). For some experiments, STAT5 EMSA was performed using nuclear extracts prepared as described previously (38).
Immunoprecipitations and Immunoblotting-Anti-HA antibody (1 g) was added to cellular extracts and incubated overnight at 4°C in the presence of protein A-Sepharose (Amersham Pharmacia Biotech). Immune complexes were washed thrice with Nonidet P-40 lysis buffer before addition of 2ϫ Laemmli sample buffer. Bound proteins were eluted by boiling for 10 min and separated by sodium dodecyl sulfatepolyacrylamide gel electrophoresis (SDS-PAGE) using 6 -7% gels. For direct immunoblotting, extracts (25-50 g) were mixed with 2ϫ sample buffer, boiled, and separated by SDS-PAGE.
Electrophoresed proteins were transferred to Immobilon-P PVDF membrane (Millipore, Bedford, MA) and blocked with 3% BSA in TBST (Tris-buffered saline plus 0.05% Tween 20). Anti-Tyr(P)-STAT5 (3 g) was diluted in 1% BSA/TBST and incubated for 2 h at room temperature. Anti-HA antibody (4 g) was added to blocking solution and incubated overnight at 4°C. Membranes were washed four times with TBST and incubated with a 1:5000 dilution of horseradish peroxidaseconjugated anti-mouse Ig (Amersham Pharmacia Biotech) in 1% BSA/ TBST for 30 min at room temperature. After four washes with TBST, immunoreactive proteins were detected using enhanced chemiluminescence (ECL) reagent (Amersham Pharmacia Biotech). RC20 blotting was performed as described previously (35). Where appropriate, membranes were stripped with a solution containing 2% SDS, 62.5 mM Tris, and 0.7% ␤-mercaptoethanol for 30 min at 55°C, washed extensively with H 2 O, twice with TBST, and re-blocked with 3% BSA/TBST before addition of primary antibody.
Electrophoretic Mobility Shift Assay-Samples (5 g) of extracts were used for STAT5 EMSA. EMSA was performed with a STAT5 oligonucleotide probe from the ␤-casein promoter element (top strand: 5Ј-GTAGATTTCTAGGAATTCAAA-3Ј) as described previously (35). Association and dissociation of the STAT5-DNA complex was determined using modifications to an established protocol (39). Briefly, to measure association, 5 g of extracts were incubated at 0°C with labeled probe before quenching with a 100-fold excess of specific competitor. For dissociation experiments, a steady-state complex was initially achieved by pre-incubating 5 g of extracts with labeled probe for 30 min at 0°C, then a 100-fold excess of unlabeled probe was added before incubating at 30°C. Dissociation was dependent on the addition of unlabeled probe since poly(dI)⅐poly(dC), a nonspecific competitor DNA, failed to compete with the STAT5-DNA complex (data not shown). For both association and dissociation experiments, samples were immediately loaded at the indicated times onto continuously running 6% non-denaturing polyacrylamide, 0.5ϫ TBE gels. For more details, see legend to Fig. 6. All quantitation was done by phosphorimage analysis using a GS-525 molecular imager system (Bio-Rad).
STAT5 Transactivation Assays-Confluent NIH3T3/ProR cells were split one into three the day prior to transfection. Approximately 50% confluent plates were transiently transfected for 5 h at 37°C with 2 g of (STAT5RE) 6 -61pF* and 3 g of STAT5 cDNAs per dish using Lipo-fectAMINE/Opti-MEM. 100 ng of a Renilla luciferase construct was included to normalize for transfection efficiency. After 24 h, cultures were split and cultured in either the presence or absence of 10 g of recombinant murine prolactin for 16 -20 h. Cells were harvested and assayed for luciferase activity using a dual luciferase assay kit (Promega, Madison, WI). Activity was normalized with Renilla luciferase activities and expressed relative to that obtained with prolactin-stimulated wild-type STAT5 (100%).
For GAL4 fusion experiments, NIH3T3/ProR cells were prepared as described earlier before lipofection with 6 g of the GAL4-fusion cDNA, 4 g of the GAL4 reporter gene pFR-LUC (Stratagene), and 100 ng of Renilla luciferase construct. After 16 -20 h cells were harvested and luciferase activities determined. BaF 3 cells (5 ϫ 10 6 ) were prepared as described previously (36) before being electroporated at 350 V and 960 microfarads with 5 g of the GAL4-fusion cDNA, 7.5 g of pFR-LUC, and 100 ng of Renilla luciferase construct. After 16 -20 h cells were harvested and luciferase activities determined. Activity was normalized with Renilla luciferase activities and expressed as a percentage of GAL4-WT STAT5 activity.

RESULTS
Internal Deletions within the STAT5 C Terminus Abolish the Transcriptional Activity of GAL4 Chimerae-Truncation analysis of the C terminus of STAT5 had indicated that the region between amino acids 750 and 772 played an important role in transactivation (26). To further understand the contribution of this region to transactivation, we generated a series of Cterminal mutants of STAT5 (Fig. 1A). The C termini of these mutants, starting from residue 694, was fused to the GAL4 DNA-binding domain (DBD) (Fig. 1B) and the relative ability of these constructs to transactivate a GAL4-responsive luciferase reporter gene was examined by ectopically expressing these chimerae in BaF 3 cells. As shown in Fig. 2A, fusing the wildtype (WT) STAT5 C terminus to the GAL4-DBD resulted in luciferase expression 10 -20-fold higher than the GAL4-DBD alone. A C-terminal truncation of the last 57 amino acids (⌬737) removed all of STAT5's ability to transactivate confirming that the STAT5 transactivation domain is localized within its C terminus. Internal deletions of 29 (TD⌬29) and 12 (TD⌬12) amino acids also dramatically decreased the ability of these mutant chimerae to transactivate. However, unlike the truncated mutant, these internally deleted mutants, especially TD⌬12, were not completely inert. A potential SH3 binding site within the 29-amino acid region was destroyed by mutating four proline residues to alanine (P 4 /A 4 ). This mutation had no effect on the ability of the GAL4-chimera to transactivate compared with the WT fusion protein, suggesting the structural contributions of these proline residues were non-essential for transactivation. Identical results were obtained when we repeated these experiments in NIH3T3 cells (Fig. 2B). Since the P 4 /A 4 mutant had 100% WT activity in BaF 3 cells, we did not test this mutant in NIH3T3 cells.
Internal Deletions within the C Terminus Preserve Transactivation by STAT5 Mutants-To extend these observations, we placed these mutations in the context of the STAT5 backbone and we examined the relative ability of these constructs to transactivate a STAT5-responsive promoter in NIH3T3 cells stably expressing the prolactin receptor (NIH3T3/ProR cells). STAT5 constructs were co-transfected into these cells with a STAT5 luciferase reporter gene, and luciferase expression was measured following activation of the ectopically expressed STAT5 proteins in response to prolactin treatment. As shown in Fig. 3, expression of WT STAT5 resulted in a 4-fold increase in luciferase expression following prolactin treatment. This increase was specific as the Y694S mutant, which cannot be phosphorylated (and therefore cannot dimerize and bind DNA), failed to transactivate the reporter gene. Furthermore, the transfection of pcDNA3 failed to induce luciferase expression following prolactin treatment (data not shown), indicating that prolactin-induced luciferase expression was dependent on ectopically expressed STAT5. The P 4 /A 4 mutant also transactivated the reporter gene, although in this case to 70% of WT levels in response to prolactin treatment. Prolactin activation of the ⌬737 mutant failed to significantly increase luciferase expression above that of the untreated sample, confirming our earlier experiments that the C terminus contains the transactivation domain. Unexpectedly, we observed that both the TD⌬29 and TD⌬12 mutants transactivated the STAT5 reporter gene to 75% and 100% of WT levels following prolactin treatment, respectively. This was in contrast to our data using GAL4 fusion proteins, which clearly demonstrated that these deletions affect the activity of the transactivation domain per se (Fig. 2) and suggested that deletion of this region of the C terminus, in the context of the native protein, had additional effects on STAT5 function.
Internal Deletion of the STAT5 C Terminus Enhances DNA Binding Activity-In an effort to explain the discrepancy in the previous experiments, we expressed the C-terminal mutants in COS-7 cells and examined whether the mutations altered the function of STAT5 by affecting DNA binding or tyrosine phosphorylation. Co-expression of the ⌬737 mutant and JAK2 in COS-7 cells resulted in a marked increase in DNA binding activity compared with cells expressing WT STAT5 (Fig. 4A). Transfection of either WT STAT5 or JAK2 cDNAs alone did not result in detectable STAT5 DNA binding activity, indicating that the co-expression of both STAT5 and JAK2 is necessary for measurable STAT5 DNA binding activity in COS-7 cells (data not shown). Since STAT5 is normally phosphorylated on Tyr 694 upon activation (40), we confirmed that phosphorylation of Tyr 694 was essential for STAT5 DNA binding activity, as mutation of this residue to serine (Y694S) rendered the protein incapable of being tyrosine phosphorylated and abrogated DNA binding ( Fig. 4A; data not shown). We also examined the status of Tyr 694 in the ⌬737 mutant using a Tyr(P) 694 -specific-STAT5 antibody. We observed an increase in Tyr 694 phosphorylation, which largely explained the enhanced DNA binding activity seen with the ⌬737 mutant.
Next we examined whether the TD⌬29 and TD⌬12 mutations affected DNA binding or tyrosine phosphorylation. We observed a markedly enhanced level of DNA binding activity for both mutants (Fig. 4B). The increase was comparable to that observed with the ⌬737 mutant. As with ⌬737, this effect was largely due to an increase in Tyr 694 phosphorylation in both TD⌬29 and TD⌬12 mutants. Tyrosine 694 was essential for the enhanced DNA binding activity and tyrosine phosphorylation observed with these mutants, since mutation of this residue to serine (Y694S) abolished all phosphorylation and DNA binding activity in the double mutant ( Fig. 4B; data not shown). Although the 29-amino acid region loosely corresponded to a PEST sequence (41), deletion of this region did not affect the level of protein expression (Fig. 4B, lower panel), suggesting this region of STAT5 is not involved in protein stability. In the case of the P 4 /A 4 mutant, we observed a slight reduction in DNA binding activity compared with WT STAT5. This correlated well with the decrease in Tyr 694 phosphorylation in this mutant, suggesting that the prolines may be required for optimal phosphorylation of Tyr 694 by JAK. These results confirm that the internal C-terminal deletions of 29 and 12 amino acids mimicked the effects of the C-terminal truncation (⌬737 mutant). In addition, they demonstrated that the extreme C terminus of STAT5 is not essential for normal levels of DNA binding or Tyr 694 phosphorylation.
Internal Deletion of the C Terminus of STAT5 Delays Tyrosine Dephosphorylation-To determine the mechanism of the deletion mutant's enhanced DNA binding activity, we investigated the effects of the 29-amino acid deletion (TD⌬29) on tyrosine dephosphorylation of STAT5. We stably expressed the WT or TD⌬29 STAT5 proteins in BaF 3 cells and established clonal lines from single cells expressing high levels of the HA-tagged proteins. After stimulation with IL-3, the mutant protein was dephosphorylated at a slower rate compared with that of WT STAT5 (Fig. 5A). Interestingly, an associated phosphoprotein (pp59) was consistently observed to co-precipitate with HA-tagged STAT5 proteins expressed in BaF 3 cells. Its phosphorylation and association profile closely parallels that of either mutant or WT STAT5 phosphorylation. Although we have not yet identified pp59, our data suggest that it is unlikely to be a degradation product of STAT5 (data not shown). To further confirm that the 29-amino acid deletion affects the rate of dephosphorylation rather than that of phosphorylation, WT or TD⌬29 STAT5 cDNAs were co-transfected into COS-7 cells with JAK2. After 24 h, staurosporine was added to inhibit JAK2 activity required for phosphorylation of the ectopically expressed STAT5 proteins. Although WT STAT5 DNA binding activity could no longer be detected after 1 h of staurosporine treatment, the DNA binding activity of TD⌬29 was unaffected up to 4 h after treatment, indicating a slower rate of tyrosine dephosphorylation (Fig. 5B). Together, these experiments indicate that TD⌬29 is dephosphorylated on Tyr 694 more slowly than WT STAT5. Furthermore, the resulting increase in steady state STAT5 dimers due to prolonged phosphorylation partially explains the enhanced DNA binding activity of the mutant protein.
Careful examination of mutants which exhibit a markedly enhanced DNA binding characteristic (i.e. ⌬737, TD⌬29, TD⌬12) revealed that they do so in a manner that is disproportionately greater than their respective enhanced levels of Tyr 694 phosphorylation would predict (Fig. 4). This suggested that, in addition to altering the kinetics of dephosphorylation, these mutations were in some way affecting DNA binding, perhaps affecting their DNA binding affinity. To explore this possibility, we determined the respective "on" and "off" rates of DNA binding for TD⌬12 and WT STAT5 proteins. Surprisingly, neither the on or off rate of DNA binding was significantly affected by the 12 amino acid deletion (Fig. 6). Thus, it appears that the mutations do not affect DNA binding per se. We believe that the mutations favor the rate of dimerization once the mutant proteins are phosphorylated. This would effectively raise the concentration of DNA-binding complex relative to what is normally observed with WT STAT5.
Non-acidic Residues, but Not Acidic Residues, within the 12-Amino Acid Region Are Crucial for Transactivation by STAT5-To better understand how this 12-amino acid region affects STAT5 function, we made additional mutations within this region. The 12-amino acid domain, which is essential for transactivation, contains one serine (Ser 756 ). Since serine phosphorylation has been shown to be important for maximal transactivation by STATs (29,42), we mutated this residue to glycine (S756G; Fig. 7A). Additionally, since acidic residues have been reported to be important for transactivation (43)(44)(45)(46), we made two additional mutants in which we mutated either all of the acidic (TDacid/Ala) or all of the non-acidic (TDnon-acid/Ala) residues within this region to alanine residues (Fig. 7A).
Using the GAL4 chimera transactivation assay, we observed that the S756G mutation resulted in levels of transactivation 45% and 70% of WT when tested in BaF 3 and NIH3T3 cells, respectively (Fig. 7, B and C). The acidic mutant (TDacid/Ala), although inactive in BaF 3 cells, retained nearly 50% of WT activity when expressed in NIH3T3 cells. We favor the explanation that this difference reflects differential expression of the acidic mutant in the two cell lines. Indeed, it is the only mutant for which we see reduced levels of protein expression when expressed in COS-7 cells (see below), suggesting it is less stable than the WT or other mutant proteins.
When placed in context of the STAT5 backbone and expressed in fibroblasts, the S756G mutation resulted in 75% of WT levels of transactivation following prolactin treatment (Fig.  7D), similar to that observed in fibroblasts using the GAL4 chimera. Interestingly, despite weakening the transactivation domain in the GAL4 chimera, the TDacid/Ala mutation resulted in greater than WT levels of transactivation when placed in the STAT5 backbone, suggesting that this mutation might also affect DNA binding activity. Somewhat to our surprise, we found that the TDnon-acid/Ala mutation abolished transactivation either in the context of the GAL4 chimera or in the context of the STAT5 molecule (Fig. 7, B-D).

Loss of Acidic Residues within the 12-Amino Acid Region
Enhances DNA Binding Activity-Ectopic expression of the S756G mutant in COS-7 cells affected neither DNA binding activity nor Tyr 694 phosphorylation when compared with WT STAT5 (data not shown). In contrast, expression of the nonacidic mutant resulted in an increased level of DNA binding activity compared with WT STAT5. However, this increase was less dramatic than that observed with the TD⌬12 mutant (Fig.  8). Furthermore, it correlated very well with an increase in phosphorylation of Tyr 694 . Expression of the acidic mutant also resulted in enhanced DNA binding activity, similar to that observed with the non-acidic mutant. However, unlike the nonacidic mutant, the acidic mutant had significantly decreased levels of protein expression and Tyr 694 phosphorylation compared with WT STAT5 (Fig. 8). Despite its reduced expression and phosphorylation, the acidic mutant was still capable of binding more DNA than the WT protein, indicating that loss of the acidic residues within the 12-amino acid domain significantly augments DNA binding activity (possibly through an increased rate of dimer formation).
Hydrophobic Residues Phe 751 and Leu 753 Are Essential for STAT5 Transactivation-The loss of transcriptional activation associated with the TDnon-acid/Ala mutation led us to more closely investigate the role of hydrophobic residues within the 12-amino acid region. When we compared the minimal transactivation domain of the herpes simplex virus transcription factor, VP16, with our 12-amino acid region of STAT5, we found that there was a homologous sequence EFDLD common to both proteins (Fig. 9A). Phenylalanine and leucine residues in this region were shown to be critical for transactivation by VP16 (47,48). Interestingly, the same residues in STAT5 had been mutated as part of the non-acidic mutation that abolished transactivation. We therefore tested whether both Phe 751 and Leu 753 of STAT5 might also be critical for transactivation. We made three additional mutants in which Phe 751 or Leu 753 or both were mutated to alanine residues (Fig. 9B). Since the non-acidic mutant was transcriptionally inert using either a GAL4 chimera or the STAT5 backbone, we only tested the activity of these mutants using GAL4 chimerae in BaF 3 cells. As seen in Fig. 9C, mutating either Phe 751 or Leu 753 to alanine reduced transactivation to 25% and 30% of WT STAT5, respectively. The double mutant (F751A/L753A) did not result in any further decrease in transcriptional activity compared with the F751A mutation, suggesting that Phe 751 may be more essential than Leu 753 . These results indicate that the hydrophobic residues within the 12-amino acid region of STAT5, especially Phe 751 and Leu 753 , are crucial for transactivation.

DISCUSSION
Using deletion and truncation mutants, we have provided further evidence that the STAT5 transactivation domain is localized within its C terminus. Specifically, GAL4-chimerae have defined a crucial 12-amino acid region, which, when deleted, resulted in only residual levels of transactivation. Surprisingly, when we deleted these 12 amino acids from the entire STAT5 molecule, we found that transactivation by this mutant STAT5 was preserved. Biochemical analysis of this mutant in COS-7 cells revealed that deletion of these 12 amino acids resulted in a marked increase in DNA binding activity. Therefore, this 12-amino acid deletion has two effects. First, it weakens, but does not abolish all transcriptional activity, and second, it markedly enhances DNA binding activity. The combination of these two effects compensate for each other, resulting in a STAT5 mutant with net activity similar to the WT protein. Hence, the function of the C terminus of STAT5 is pleiotropic with somewhat "plastic" properties such that considerable mutation in this region can be tolerated before an overall loss of function is observed.
The enhanced DNA binding activity seen with several STAT5 mutants in this study is largely due to an increase in phosphorylation of Tyr 694 . This increase in DNA binding activity was critically dependent on Tyr 694 , since mutation of this residue to serine rendered the double mutant proteins incapable of being phosphorylated and binding DNA. We have shown previously that dephosphorylation of Tyr 694 is the most likely mechanism for the inactivation of STAT5 (35). In this regard, we found that the TD⌬29 mutant had a slower rate of tyrosine dephosphorylation. which is consistent with that seen with variants of STAT5 lacking the C terminus (28). Thus, a delay in dephosphorylation causes a slower inactivation of TD⌬29 and consequently a steady-state accumulation of active dimers leading to an increase in DNA binding activity. Furthermore, the DNA binding activity and Tyr 694 phosphorylation of both our truncated (⌬737) and deletion (TD⌬12) mutants are similarly affected. Hence, the extreme C terminus of STAT5 is not involved in the normal regulation of DNA binding or Tyr 694 phosphorylation; rather, the essential 12 amino acids are. Whether or not this domain recruits a phosphatase or merely modulates its activity remains to be determined.
We noted that an increase in Tyr 694 phosphorylation alone does not entirely account for the enhanced DNA binding characteristic for several of our mutants. The ⌬737, TD⌬29, and TD⌬12 mutants all bound more DNA than WT STAT5, yet they FIG. 6. Deletion of the essential 12-amino acid domain does not affect DNA binding affinity. COS-7 cells were co-transfected for 24 h with HA-JAK2 and either WT or TD⌬12 HA-STAT5a cDNAs. A, to measure dissociation of STAT5-DNA complex, a 100-fold excess of unlabeled probe was added to a pre-formed, steady-state complex (5 g of extracts) prior to incubating at 30°C for the indicated time. Gels were dried, quantitated, and the fraction bound at time zero was plotted against time. B, to measure association of STAT5-DNA complex, 5 g of extracts were incubated at 0°C with labeled probe as indicated before quenching with a 100-fold excess of unlabeled competitor. Gels were dried and quantitated, and the amount of bound probe (% max) was plotted against time. Solid squares, WT; open squares, TD⌬12. did so in a manner that was disproportionately greater than their respective levels of Tyr 694 phosphorylation would predict (Fig. 4). For example, the TD⌬12 mutant showed a 2-3-fold increase in Tyr 694 phosphorylation, yet bound more than 5 times as much DNA as WT STAT5. Therefore, we examined the possibility that these mutations might also be affecting the affinity of DNA binding. However, we found that neither the rates of association or of dissociation of the TD⌬12 mutant are affected when they are compared with those of WT STAT5 (Fig. 6).
How does the C terminus modulate DNA binding activity? We favor the possibility that the C terminus affects the rate of dimer formation. Since only phosphorylated STAT dimers can bind DNA it is possible that certain C-terminal mutations (e.g. TD⌬12), although not affecting the DNA binding affinity per se, result in a shift in the equilibrium toward the phosphorylated dimer, effectively increasing the concentration of STAT5 dimers capable of binding DNA. One potential mechanism might be through the interaction of protein inhibitor of activated STAT (PIAS) proteins (49,50). PIAS proteins associate with the phosphorylated tyrosine of STATs via their SH2 domains and reduce STAT transactivation through inhibition of STAT dimerization and DNA binding. Although a PIAS protein has not been identified for STAT5, it is possible that the mutations that result in enhanced DNA binding activity do so by decreasing the association between PIAS and STAT5. The clon-ing of a STAT5-specific PIAS protein would allow us to directly test this hypothesis.
The S756G mutation within the critical 12-amino acid region resulted in an overall reduction in STAT5 transactivation. Attempts to show phosphorylation at this site in vivo have failed (data not shown), consistent with a previous study on serine phosphorylation of STAT5 proteins (51). Indeed, phosphorylated serine residues other than Ser 756 have been reported for both STAT5a and STAT5b (51). Contrary to what has been reported for STAT-1 and -3 (29,42), mutation of these serine residues had no effect on the transcriptional activity of STAT5a and only a modest effect on STAT5b (51). Since both DNA binding and phosphorylation of Tyr 694 were unaffected by the S756G mutation, the effect of this mutation on transactivation is most likely due to structural changes in the 12-amino acid region.
Although the acidic nature of transactivation domains is important for the activity of many transcription factors (43)(44)(45)(46), we were surprised that the acidic residues within the 12-amino acid region of STAT5 were not essential for transactivation, as mutation of these residues to alanines only reduced transactivation of the GAL4 chimera to about 50% of WT activity in NIH3T3 fibroblasts. Although the mutant GAL4 chimera failed to elicit activity in BaF 3 cells, we believe that this is likely a consequence of reduced protein expression since the acidic mutant was the only mutant for which we have seen FIG. 7. Non-acidic residues, but not the acidic residues within the 12-amino acid region are essential for STAT5 transactivation. A, Amino acid sequence of the 12-amino acid region from WT STAT5. Residues in boldface indicate residues mutated in STAT5 mutants. GAL4-STAT5a cDNAs and pFR-Luc were co-transfected into either BaF 3 (B) or NIH3T3 (C) cells as indicated. Cells were harvested as described in Fig. 2. D, NIH3T3/ProR fibroblasts were co-transfected with either WT or mutant STAT5a cDNAs with (STAT5RE) 6 -61pF* as indicated. 24 h later the culture was divided and continued for 16 -20 h in either the presence or absence of prolactin (1 g/ml). Cells were harvested as described in Fig. 3. B-D, each bar represents the mean Ϯ S.E. of four independent experiments. differential expression. When the identical mutation was introduced in the context of the STAT5 backbone, we found that it led to increased activity in fibroblasts compared with the WT protein. Despite having a reduced Tyr 694 phosphorylation, this mutant exhibited an increased capacity to bind DNA compared with WT STAT5, possibly through augmented dimer formation. Again, a partially active transactivation domain combined with enhanced intrinsic DNA binding result in a net level of activity equivalent to that of WT STAT5.
Our finding that acidic residues are non-essential for transactivation is not without precedent. The same was reported for the glucocorticoid receptor, GAL4, VP16, C1, and v-myb (47,(52)(53)(54)(55). In those cases, hydrophobic residues within the transactivation domain were found to be more critical for transactivation (47,48,54). Consistent with these observations, we found that mutating all the non-acidic residues within the 12-amino acid region resulted in significantly reduced levels of transactivation. When we compared the amino acid sequence of the minimal essential transactivation domain of VP16 with that of 12 amino acids essential for STAT5 transactivation, we found the two sequences to be highly similar, with the conserved sequence EFDLD. Previous studies had shown that both the phenylalanine and leucine in this sequence were essential for transactivation by VP16 (47,48). Mutation of either F751A or L753A in STAT5 severely compromised the ability of the mutant proteins to transactivate, indicating that these residues are crucial for transactivation. Mutating both residues did not result in any further decrease in transcriptional activity, suggesting that Phe 751 may be more critical than Leu 753 .
Transactivation domains are highly divergent regions of transcription factors and tend to be acidic in nature (43,44).
One common feature shared by several acidic transactivation domains, including that of VP16, is that under physiological conditions these regions adopt an unordered structure (55)(56)(57)(58)(59)(60)(61). However, when exposed to more hydrophobic or mildly acidic conditions, they have the propensity to transform into regions rich in secondary structure (55,(57)(58)(59)(60)(61). Only the crystal structure of STATs lacking their C termini has been reported (18,19), suggesting that the transactivation domain of STATs also adopts an unordered structure. Despite this prediction, it is clear from our experiments that a single hydrophobic point mutation in or near the transactivation domain of STAT5 is highly destructive. It is likely that these mutations disrupt secondary structure(s) and affect the ability of this region to contact or stabilize contact with components of either co-activator complexes or of the basal transcriptional machinery. In this regard, these point mutants will help to elucidate the mechanism by which STATs transactivate target genes. Furthermore, these transcriptionally weak mutants will be useful in identifying novel interactors with the C terminus of STAT5 by both biochemical and genetic approaches.  9. Hydrophobic residues Phe 751 and Leu 753 are critical for STAT5 transactivation. A, alignment of the amino acid sequence of the 12-amino acid region from STAT5 with the minimal essential transactivation domain of VP16. Asterisks (*) indicate residues in STAT5 mutated in the TDnon-acid/Ala mutant. Residues underlined in VP16 are essential for transactivation. B, residues in boldface indicate residues mutated in STAT5 mutants. C, GAL4-STAT5 cDNAs and pFR-Luc were co-transfected into BaF 3 cells as indicated. Cells were harvested as described in Fig. 2. Each bar represents the mean Ϯ S.E. of four independent experiments.