A functional DNA binding domain is required for growth hormone-induced nuclear accumulation of Stat5B.

The mechanisms regulating the cellular distribution of STAT family transcription factors remain poorly understood. To identify regions of Stat5B required for ligand-induced nuclear accumulation, we constructed a cDNA encoding green fluorescent protein (GFP) fused to the N terminus of Stat5B and performed site-directed mutagenesis. When co-expressed with growth hormone (GH) receptor in COS-7 cells, GFP-Stat5B is tyrosyl-phosphorylated, forms dimers, and binds DNA in response to GH in a manner indistinguishable from untagged Stat5B. In multiple cell types, laser scanning confocal imaging of GFP-Stat5B co-expressed with GH receptor shows that GFP-Stat5B undergoes a rapid, dramatic accumulation in the nucleus upon GH stimulation. We introduced alanine substitutions in several regions of Stat5B and assayed for GH-dependent nuclear localization. Only the mutation that prevented binding to DNA (466VVVI469) abrogated GH-stimulated nuclear localization. This mutant fusion protein is tyrosyl-phosphorylated and dimerizes in response to GH. These results suggest that either high affinity binding to DNA contributes to nuclear accumulation of Stat5B or that this region is crucial for two functions, namely accumulation of Stat5B in the nucleus and DNA binding. Thus, we have identified a mutant Stat5 defective in nuclear localization despite its ability to be tyrosyl-phosphorylated and to dimerize.

The Signal Transducers and Activators of Transcription (STAT) 1 family of transcription factors provide a crucial signaling link between complexes of cytokine/hematopoietin receptors and Janus (JAK) family tyrosine kinases at the plasma membrane and gene transcription in the nucleus (1). Seven mammalian STAT genes have been identified and mouse genetics have revealed functions for most that are well supported by the in vitro experiments that led to their discovery (1,2). A general model for cytokine activation of STATs has been proposed, based primarily on Stats 1-3 (2,3). In this model, STATs exist as monomers located in the cytoplasm prior to receptor activation, although evidence for preassociation prior to activation exists (4). Upon cytokine stimulation, phosphotyrosine residues in the receptor recruit STATs through an SH2 domain interaction. Activated Janus kinases phosphorylate STATs on a carboxyl tyrosine, promoting STAT dimer formation by intermolecular SH2 domain interaction and dissociation from the receptor complex. Once dimerized, STATs translocate to the nucleus, bind DNA, and regulate gene transcription (2,3). Recent studies have shown that in some cells some STATs are present in the nucleus prior to activation (5)(6)(7)(8)(9). Thus, it appears that beyond the general mechanism outlined above, there may be cell type-specific mechanisms that could further regulate some of the STATs. Finding STATs in the nucleus of some cells prior to activation suggests that constitutive nuclear import and export exist in these cells.
The mechanism by which activated STATs accumulate in the nucleus is unknown. STATs have a mass above the upper limit for diffusion through the nuclear pore (ϳ45 kDa) (10 -12) and thus are assumed to be actively transported into the nucleus. The best characterized nuclear import pathway involves binding of the transported protein to the heterodimer protein complex importin ␣/␤ followed by energy-dependent transport through the nuclear pore complex requiring the GTPase activity of Ran/TC4 (10 -12). STATs appear to lack a conventional nuclear localization sequence, the single or dual stretch of basic amino acids that bind to importin ␣. Yet, Stat1 dimers have recently been shown to associate with the importin ␣-homologue NPI-1 (13). Furthermore, IFN-␥-stimulated Stat1 nuclear accumulation requires NPI-1 (13) and the GTPase activity of Ran/TC4 (14). Perhaps STAT dimers possess the structure required for binding to nuclear import proteins. In cells where STATs are found in the nucleus prior to activation, cytokine-induced nuclear accumulation might also arise from down-regulation of nuclear export. Insight into how STAT proteins interact with proteins needed for nuclear transport awaits identification of a mutant STAT that can dimerize but is unable to localize to the nucleus.
Stat5 was first identified as a mammary gland factor required for prolactin (PRL)-stimulated gene transcription (15,16) but was found to be activated by many hormones, growth factors, and cytokines (1,17,18). We now know that in humans and rodents Stat5 is actually two distinct proteins arising from different genes (Stat5a and Stat5b) (18 -22). Stat5A and Stat5B are highly homologous, differing mainly in their C terminus, and are expressed in most tissues (18,19). Genetic disruption of the Stat5a gene results in mice deficient in mammopoiesis and lactogenesis (23). Knock-out of the Stat5b gene removes the sexual dimorphism of body growth and liver gene expression induced by growth hormone (GH) (24,25). The simultaneous deletion of both genes results in the most severe growth and reproductive defects, revealing functional redundancy of the Stat5 proteins in physiological processes mediated by GH and PRL (25).
To probe the mechanisms regulating Stat5B localization within cells, we constructed a green fluorescent protein (GFP)-Stat5B fusion protein. We find that the fusion protein is tyrosyl-phosphorylated, dimerizes, accumulates in the nucleus, and binds DNA upon cytokine receptor stimulation in a manner indistinguishable from untagged Stat5B. We identify by site-directed mutagenesis residues required for DNA binding that are also needed for GH-dependent nuclear localization. These studies identify for the first time a region of Stat5 required for nuclear localization independent of those regions needed for dimerization.

EXPERIMENTAL PROCEDURES
Materials-Human fibrosarcoma 2C4 and 2C4-GHR cells were provided courtesy of G. Stark, Y. Han (Cleveland Clinic, Cleveland, OH), and I. Kerr (Imperial Cancer Research Fund, London, UK). Rat Stat5A cDNA (26) was kindly provided by J. Rosen (Baylor College of Medicine, Houston, TX); rat GH receptor (GHR) cDNA (27) was from G. Norstedt (Karolinska Institute, Stockholm, Sweden), and recombinant human GH was the gift of Lilly. Prestained molecular weight standards were from Life Technologies, Inc. All chemicals were reagent grade or better.
Construction of GFP-Stat5B cDNA-pEGFP-C1 (CLONTECH), which encodes a red-shifted variant of GFP optimized for fluorescence intensity and high expression in mammalian cells, was used to construct a cDNA encoding a GFP-Stat5B fusion protein. The BglII-EcoRI fragment of rat Stat5B cDNA (28) was first subcloned into pEGFP-C1 (designated pEGFP-⌬Stat5B). Polymerase chain reaction with Pfu DNA polymerase (Stratagene) and oligonucleotides 5Ј-GAAGATCTAT-GGCAATGTGGATACAG-3Ј and 5Ј-GGAGCTGCGTGGCATAG-3Ј as primers was used to engineer a BglII site and remove an in-frame stop codon upstream of the start codon of Stat5B. The polymerase chain reaction product was purified, digested with BglII, and inserted into pEGFP-⌬Stat5B, yielding cDNA encoding GFP fused to the N terminus of full-length Stat5B by a five amino acid linker (SGLRS). DNA sequencing (Sequenase 2.0; U. S. Biochemical Corp.) was performed to verify the region created by polymerase chain reaction and all junctions.
Cell Culture and Transfections-2C4, 2C4-GHR, COS-7, and NIH-3T3 cells were cultured at 37°C in Dulbecco's modified Eagle's medium supplemented with 1 mM L-glutamine, 100 units/ml penicillin, 100 g/ml streptomycin, 0.25 g/ml amphotericin B, and either 10% fetal bovine serum (2C4, 2C4-GHR, COS-7) or 9% calf serum (NIH-3T3). Medium for all lines except COS-7 was further supplemented with 5 mM sodium pyruvate. COS-7, 2C4, and 2C4-GHR cells were transiently transfected by calcium phosphate precipitation (29) by incubating subconfluent cultures with DNA precipitates for 6 -16 h. For transfection of 2C4 and 2C4-GHR cells, 20 g/ml CalPhos maximizer (CLONTECH) was used. The cultures were then rinsed twice with Dulbecco's modified Eagle's medium and fed with culture medium. To equalize the amount of DNA transfected in non-imaging experiments, empty pEGFP-C1 was used. NIH-3T3 cells were transfected with Lipofectin (Life Technolo-gies, Inc.) as described by the manufacturer. Cells were harvested or used for imaging between 48 and 72 h post-transfection. 2C4 cells stably transfected with a mammalian expression vector containing the cDNA for human GHR (2C4-GHR) have been described previously (30).
Phosphatase Treatment-Cell lysates were immunoprecipitated with ␣Stat5B and incubated at 37°C for 60 min in 100 l of dephosphorylation buffer (50 mM Tris-HCl, pH 8.5, 0.1 mM EDTA, 0.2 mM MgCl 2 , 0.02 mM ZnCl 2 ) containing 40 units of calf intestinal alkaline phosphatase (AP; Boehringer Mannheim). As controls, 10 mM Na 3 VO 4 was added to the dephosphorylation buffer or AP was omitted. The reaction was terminated, and proteins were eluted by boiling in a 4:1 mixture of lysis buffer and SDS-polyacrylamide gel electrophoresis sample buffer. The resultant dephosphorylated proteins were separated by SDS-polyacrylamide gel electrophoresis and immunoblotted with ␣Stat5B as described above.
Electrophoretic Mobility Shift Assays (EMSAs)-EMSAs were performed using nuclear extracts of COS-7 cells transfected with cDNAs encoding rat GHR and either Stat5B or GFP-Stat5B (wild-type or VVVI mutant). Forty-eight hours after transfection and 16 h after serum deprivation, cells were treated with 500 ng/ml GH for 5-60 min and nuclear extracts prepared (31). The extracts were incubated with or without 1 g of ␣Stat5B (C-17 10x, Santa Cruz Biotechnology) and then with a probe corresponding to the PRL response element of the ␤-casein promoter (5Ј-AGATTTCTAGGAATTCAA-3Ј; 40,000 cpm, 5 ϫ 10 Ϫ15 mol) (32). Samples were analyzed on a non-denaturing polyacrylamide gel and subjected to autoradiography. The EMSA experiments were performed twice with similar results.
Confocal Fluorescence Microscopy-Confocal imaging was performed with a Noran OZ laser scanning confocal microscope equipped with a 60ϫ Nikon objective. GFP was excited at 488 nm by a krypton-argon laser, and fluorescence above 500 nm was captured. Cells were grown on glass coverslips attached to the bottom of a 60-mm culture dish, transfected with cDNA, incubated in serum-free medium for 6 -16 h, and then imaged at room temperature in Krebs-Ringer phosphate buffer (128 mM NaCl, 7 mM KCl, 1 mM CaCl 2 , 1.2 mM MgSO 4 , 1 mM NaHPO 4 , 10 mM glucose, pH 7.4) containing 0.1% bovine serum albumin. The contribution of cellular autofluorescence was judged to be less than 1%. Preliminary experiments revealed that repeated laser exposure inhibited GH-dependent migration of GFP-Stat5B into the nucleus, presumably because of phototoxicity. Thus, the following protocol was adopted. Once control images were obtained, cell location was recorded by capturing a low-power image, and the cells were stimulated with GH in a 37°C incubator. Following the times indicated in the figures, the same cells were found and imaged a second time. The presented images are representative of at least three separate experiments during which at least 20 cells were imaged. For quantitative analysis of fluorescence distribution, cells were fixed (3.7% paraformaldehyde in phosphate-buffered saline for 10 min) following stimulation. The mean intensities of neighboring cytosolic and nuclear regions (approximately 5 m 2 each) were calculated using Adobe Photoshop TM , corrected for background, and expressed as nuclear-to-cytosol fluorescence ratios.

RESULTS
Characterization of GFP-Stat5B-As a tool for the study of the cellular localization of Stat5B, we constructed a GFP-Stat5B fusion protein. We first examined whether this protein, when expressed in COS-7 cells, is regulated by GH in a manner similar to untagged Stat5B. To examine whether GH stimulates tyrosyl phosphorylation of GFP-Stat5B, cells transfected with cDNAs encoding rat GHR and either Stat5B or GFP-Stat5B were treated with or without GH for 15 min. Proteins were immunoprecipitated with ␣Stat5B and analyzed by Western blotting using ␣Stat5B, anti-phosphotyrosine antibody (␣PY), or ␣GFP (Fig. 1A). Antibodies directed against the C terminus of Stat5B (␣Stat5B) and the N-terminal GFP tag (␣GFP) both recognized a protein with an apparent molecular mass appropriate for GFP-Stat5B (ϳ120 kDa) in immunopre-cipitates from cells transfected with GFP-Stat5B cDNA (Fig.  1A, lanes E-H). Neither antibody detected a protein of smaller size (except for endogenous Stat5B in the case of ␣Stat5B; Fig.  1A, lanes E-H, and data not shown) indicating that the vast majority of expressed protein was full-length fusion protein. In immunoprecipitates from cells transfected with Stat5B cDNA, ␣Stat5B recognized an approximately 90-kDa protein (Fig. 1A, lanes A-D), appropriate for untagged Stat5B. GH stimulated to similar extents tyrosyl phosphorylation of GFP-Stat5B (Fig.  1A, lane H) and untagged Stat5B (Fig. 1A, lane D). Hence, GFP-Stat5B expressed in COS-7 cells is full-length (not truncated at either the N or C terminus), is recognized by both ␣Stat5B and ␣GFP, and like untagged Stat5B is tyrosyl-phosphorylated in response to GH.
We next sought to determine whether GFP-Stat5B is capable of forming dimers in response to GH-induced tyrosyl phosphorylation. To analyze dimer formation, we utilized the ability of Stat5B to form dimers with Stat5A in response to many cytokines, including GH (33). COS-7 cells were transiently transfected with cDNAs encoding GHR, Stat5A, and either GFP, Stat5B, or GFP-Stat5B. Stat5A was immunoprecipitated with Stat5A-specific antibody, and precipitated proteins were probed with Stat5B-specific antibody. GH induced co-immunoprecipitation of Stat5B (Fig. 1B, lanes C and D) and GFP-Stat5B (Fig. 1B, lanes E and F) with Stat5A. No proteins were detected when Stat5A was co-expressed with GFP alone (Fig.  1B, lanes A and B), showing that ␣Stat5B did not cross-react with Stat5A. The Stat5B which co-immunoprecipitated with Stat5A in response to GH migrates as three distinct bands (Fig.  1B, lane D). Similarly, three GFP-Stat5B bands also co-immunoprecipitate with Stat5A in response to GH (Fig. 1B, lane F), although the relative amounts differ when compared with Stat5B. The lower level of GFP-Stat5B compared with Stat5B bound to Stat5A in Fig. 1B is assumed to be due to a lower level of expression of GFP-Stat5B since in all experiments where it was examined, GFP-Stat5B was expressed at lower levels than untagged Stat5B. Interestingly, co-expression of Stat5A with either Stat5B or GFP-Stat5B but not with GFP leads to a low level of constitutive tyrosyl phosphorylation of Stat5A (Fig. 1B, lanes A, C, and E). In sum, these experiments show that activated GFP-Stat5B, like untagged Stat5B, forms heterodimers with Stat5A.
Based on studies of Stat5B regulation by GH in liver (34), we predicted that the multiple Stat5B and GFP-Stat5B bands seen in immunoprecipitates from GH-treated cells reflect differential phosphorylation of Stat5B on serines/threonines and tyrosines. To test for differential phosphorylation, we treated Stat5B immunoprecipitates with the general phosphatase alkaline phosphatase (AP). AP reduced the three Stat5B and GFP-Stat5B bands seen in immunoprecipitates from GHtreated cells to predominantly the fastest migrating band, with a variable amount of a slower migrating band (Fig. 2, lanes B  and C). The changes in band pattern produced by AP treatment are the result of dephosphorylation and not protein degradation since sodium orthovanadate inhibited the mobility changes (Fig. 2, lane D). The incomplete dephosphorylation of Stat5B seen in immunoprecipitates from GH-treated cells is thought to result from limited access of AP to one of the phosphorylated sites in the Stat5B dimer. In support of Stat5B being phosphorylated on serines/threonines as well as tyrosines, we have found that the serine/threonine-specific phosphatase PP2A condenses the three Stat5B bands from GHtreated cells to two, tyrosyl-phosphorylated bands (data not shown). Further support for the multiple GFP-Stat5B bands not being truncated forms of Stat5B is the finding that all bands are recognized by both antibody to the N-terminal GFP Stat5B DNA Binding Domain Required for Nuclear Localization tag and antibody to the 10 amino acids at the C terminus of Stat5B which are unique to the B isoform (28). Also, the bands do not arise from adventitious proteolysis during the immunoprecipitation since the same bands are seen in blots of the cell lysates (Fig. 2, lane A). Overall, these data indicate that GFP-Stat5B, like untagged Stat5B, is phosphorylated at multiple sites.
To assess the ability of GFP-Stat5B to bind DNA, we performed EMSAs with a probe corresponding to the GAS-like element of the ␤-casein promoter and nuclear extracts from COS-7 cells overexpressing GHR with or without either Stat5B or GFP-Stat5B (Fig. 3). GH produced a similar time-dependent increase in the formation of a DNA-binding complex in nuclear extracts from cells expressing Stat5B (Fig. 3, lanes A-E) or GFP-Stat5B (Fig. 3, lanes G-K) but not in extracts from cells expressing GHR alone (Fig. 3, lanes M and N). This DNAbinding complex contains Stat5B or GFP-Stat5B since pretreatment of the nuclear extracts with ␣Stat5B results in a supershifted complex (Fig. 3, lanes F and L). We find that the DNA complex containing GFP-Stat5B migrates slightly slower than the complex containing untagged Stat5B, consistent with the addition of the GFP tag. Overall, Figs. 1 and 3 illustrate that the critical cytokine-regulated events of tyrosyl phosphorylation, dimerization, and DNA binding are functionally intact for GFP-Stat5B.
GH-dependent Nuclear Accumulation of GFP-Stat5B-The EMSAs indicated that GFP-Stat5B was present in COS-7 cell nuclei following GH treatment (Fig. 3). We next sought to verify the presence of GFP-Stat5B in nuclei by directly visualizing GFP-Stat5B in single, living cells. For these experiments we employed human fibrosarcoma 2C4 cells because they possess a more uniform morphology than COS-7 cells and are easily transfected. Cells transiently expressing GHR and GFP, GFP-Stat5B alone, or GHR and GFP-Stat5B were imaged by confocal microscopy prior to and following stimulation with GH (Fig. 4). In cells expressing both GFP and GHR (Fig. 4, A and  B), GFP was found throughout the cytoplasm and nucleus, and this distribution did not change upon GH treatment. Likewise, in cells expressing GFP-Stat5B but not GHR (Fig. 4, C and D), GFP-Stat5B was present both in the cytoplasm and nucleus prior to and following GH addition. A similar subcellular distribution of GFP-Stat5B was seen in cells expressing GFP-Stat5B and GHR in the absence of GH (Fig. 4E). GFP-Stat5B was also present in both the cytosol and nucleus prior to activation when expressed in COS-7, NIH-3T3, and 2C4-GHR cells (Fig. 5, A, C, and E). The presence of some GFP-Stat5B in the nucleus as well as in the cytoplasm is unlikely to be the result of the GFP tag or overexpression since a similar distribution  N) were treated without (lanes A, G, and M) or with 500 ng/ml GH for 5 (lanes B and H), 10 (lanes C and I), 30 (lanes D, F, J, L, and N), or 60 min (lanes E and K). EMSAs were performed using nuclear extracts of these cells and a GAS-like PRL response element from the ␤-casein promoter. Where indicated, nuclear extracts were preincubated without (Ϫ) or with (ϩ) ␣Stat5B for 20 min. D), or GHR and GFP-Stat5B (panels E and F) were imaged by laser scanning confocal microscopy prior to (A, C, and E) and following (panels B, D, and F) 40 -60 min treatment with 500 ng/ml GH. Images in panels A, C, and E are of the same cell as B, D, and F, respectively. Scale bar (F) represents 15 m.

Stat5B DNA Binding Domain Required for Nuclear Localization
was seen for the following: 1) untagged Stat5B expressed in COS-7 cells and endogenous Stat5B in CHO-GHR cells (both detected by immunocytochemistry); and 2) GFP-tagged Stat5B expressed at levels barely above the limits of detection of GFP fluorescence (data not shown).
GH stimulation of cells co-expressing GHR and GFP-Stat5B resulted in a dramatic nuclear accumulation of GFP-Stat5B in the majority of cells (Ͼ90%; Fig. 4F). Nuclear accumulation was detectable by 10 -15 min of GH treatment and persisted for several hours with continuous exposure to GH (data not shown). Interestingly, GH stimulated the formation of intense, punctate patterns of GFP-Stat5B fluorescence in the nucleus (Fig. 4F) and cytoplasm (Fig. 5B) of ϳ80% and ϳ15% of cells, respectively. GH also stimulated nuclear accumulation of GFP-Stat5B in COS-7, NIH-3T3, and 2C4 cells stably expressing GHR (2C4-GHR) (Fig. 5, A-F), indicating that GH regulation of GFP-Stat5B localization is not restricted to a single cell type. Immunofluorescent detection of untagged Stat5B expressed with GHR in COS-7 cells showed similar GH-regulated nuclear accumulation (data not shown), confirming that the GFP tag does not influence Stat5B nuclear accumulation.
Determinants of Stat5B Nuclear Localization-We next sought to identify regions of Stat5B required for nuclear accumulation. We performed site-directed mutagenesis on GFP-Stat5B, transiently expressed the mutant proteins with GHR in 2C4 cells, and assayed for GH-dependent nuclear accumulation. Based on sequence similarity, the region between the DNA-binding and SH2 domains was thought to be an SH3 domain (35). Since SH3 domains are known to be important in regulating the localization of signaling proteins within cells (36), we reasoned that this region of Stat5B might be important for nuclear localization. We mutated to alanine a single amino acid (Tyr 568 ) that, based on the structure of SH3 domains, would be predicted to abrogate binding to proline-rich sequences. This mutant protein (GFP-Stat5B Y ) was present in the cytosol and nucleus prior to stimulation and localized to the nucleus in response to GH (Fig. 6, A and B), similar to wild-type GFP-Stat5B. The recent report of the crystal structures of Stat1 and Stat3 homodimers bound to DNA have revealed that this region does not have the architecture of an SH3 domain and is now referred to as a linker domain (37,38).
Clusters of basic amino acids in STATs might constitute a functional nuclear localization sequence when present in the context of dimers. Three basic regions in the N terminus of Stat1 have been mutated without any effect on nuclear accumulation in response to interferon-␥ (13). We identified a different basic region consisting of a cluster of three lysines between the linker and SH2 domains (Lys 582 , Lys 583 , and Lys 586 ) of Stat5B and mutated the residues to alanine. This protein (GFP-Stat5B KKK ) accumulated in the nucleus in response to GH in a manner indistinguishable from wild-type GFP-Stat5B (Fig. 6, C and D), ruling out this region as a nuclear localization determinant.
High affinity binding to DNA has been suggested to contribute to nuclear accumulation of some transcription factors (39,40). Hence, we reasoned that disruption of Stat5B DNA binding might inhibit its nuclear accumulation. A stretch of four amino acids ( 466 VVVI 469 ) in the DNA binding domain of Stat5B is critical for binding to the IRF-1 promoter (28). In the context of GFP-Stat5B, we mutated to alanine these four amino acids (GFP-Stat5B VVVI ). Unlike wild-type GFP-Stat5B and the other mutant GFP-Stat5Bs, GFP-Stat5B VVVI localized predominantly to the cytoplasm prior to GH stimulation (Fig. 7, A and C) in Ͼ95% of 2C4 cells. Furthermore, GH did not cause nuclear accumulation of this protein (Fig. 7, B and D) at any stimulus duration tested (5-90 min; data not shown). Yet, GH stimulated the formation of punctate GFP-Stat5B VVVI fluorescence in the cytoplasm in approximately 40% of cells (Fig. 7D), similar to wild-type GFP-Stat5B. GFP-Stat5B VVVI failed to localize to the nucleus in COS-7 (Fig. 7, E and F) and 2C4-GHR cells (data not shown), indicating that the failure of mutant GFP-Stat5 VVVI to localize to the nucleus in response to GH is not specific to a single cell type. Like GFP-Stat5B VVVI , untagged Stat5B VVVI detected by immunocytochemistry was located primarily in the cytoplasm, and its distribution did not change with GH (data not shown). Thus, mutation of residues 466 VVVI 469 creates a genuine nuclear localization defective Stat5B, independent of cell type and the presence of the GFP tag.
To quantify our observations with GFP-Stat5B and its mutants, we co-expressed wild-type or mutant GFP-Stat5B with GHR in COS-7 cells, stimulated with GH for 40 min where appropriate, fixed the cells, and measured the fluorescence intensity of the cytosol and nucleus in 20 cells per condition. Fig. 8 shows the nuclear-to-cytosol fluorescence ratios (N/C) from a representative experiment. Wild-type GFP-Stat5B, the linker domain tyrosine mutant, and the lysine series mutants show similar cellular distribution, with close to equal fluorescence in the nucleus and cytosol prior to stimulation (N/C Ϸ1.2) and prominent GH-dependent nuclear accumulation with GH (N/C Ϸ2.5-3.0). DNA-binding mutant GFP-Stat5B VVVI is mostly cytoplasmic (N/C Ϸ0.4), and this distribution does not change following treatment with GH.
Lack of GH-dependent nuclear accumulation of mutant GFP-Stat5B VVVI could arise from a defect in some crucial upstream signaling event. To test this idea, we co-expressed GHR with GFP-Stat5B or GFP-Stat5B VVVI , treated the cells with or without GH, and prepared cytosol and nuclear extracts. GFP-Stat5B or GFP-Stat5B VVVI were immunoprecipitated with ␣Stat5B and probed with ␣PY. Like wild-type GFP-Stat5B, GFP-Stat5B VVVI extracted from the cytosol is tyrosyl-phosphorylated in response to GH (Fig. 9A, lanes A-F). The lower level of tyrosyl-phosphorylated GFP-Stat5B VVVI in the cytosol compared with wild-type GFP-Stat5B can be attributed to lower expression of GFP-Stat5B VVVI compared with wild-type, as seen by reprobing the membrane with ␣Stat5B (Fig. 9A, lower  panel, lanes A-F). Unlike wild-type GFP-Stat5B, tyrosyl-phosphorylated GFP-Stat5B VVVI is undetectable in the nucleus (Fig. 9A, upper panel, lanes G-L). A small amount of GFP-Stat5B VVVI is detected in the nuclear fraction when the membrane is reprobed with ␣Stat5B (Fig. 9A, lower panel, lanes  J-L); it may represent contamination from the cytosolic fraction or a small amount of GFP-Stat5B VVVI in the nucleus. In support of it being a contaminant, GFP-Stat5B VVVI was undetectable in the nuclear fraction of a second experiment (data not shown).
By using the same cell extracts, we next sought to determine if mutant GFP-Stat5B VVVI can bind DNA. As expected, GH-dependent DNA binding is present in both cytosol and nuclear extracts from cells expressing wild-type GFP-Stat5B and GHR (Fig. 9B, lanes A-C and G-I). Both cytosolic and nuclear extracts from cells expressing GFP-Stat5B VVVI and GHR lack DNA binding activity (Fig. 9B, lanes D-F and J-L). Thus, although GFP-Stat5B VVVI in the cytosol is tyrosyl-phosphorylated in response to GH, it cannot bind DNA, as predicted (28).
To test whether the failure of GFP-Stat5B VVVI to accumulate in the nucleus and bind DNA in response to GH might stem from an inability to form dimers, we examined the ability of GFP-Stat5B VVVI to dimerize with Stat5A. COS-7 cells expressing GHR, Stat5A, and either GFP, wild-type GFP-Stat5B, or GFP-Stat5B VVVI were treated with or without GH (Fig. 10). Immunoprecipitation with ␣Stat5A and probing with ␣Stat5B or ␣PY reveals that tyrosyl-phosphorylated mutant GFP-Stat5B VVVI co-immunoprecipitates with Stat5A in response to GH in a manner similar to wild-type GFP-Stat5B (Fig. 10, lanes C-F). In this experiment, mutant GFP-Stat5B VVVI was present in lysates at about 30% of the level of wild-type GFP-Stat5B. When this difference in expression is considered, the DNA-binding mutant GFP-Stat5B VVVI forms dimers with Stat5A better than wild-type GFP-Stat5B. Interestingly, we find that the slowest migrating band of wild-type GFP-Stat5B is absent in GFP-Stat5B VVVI co-immunoprecipitates, suggesting differential phosphorylation of the wild-type and mutant fusion proteins. These data indicate that dimer formation is not sufficient for tyrosyl-phosphorylated mutant GFP-Stat5B VVVI to localize to the nucleus and define the DNA binding domain as a novel determinant of Stat5B nuclear accumulation. DISCUSSION STATs play important roles in immune system function and hormonal signaling. Numerous elegant studies have provided much insight into how the STATs fulfill these roles (reviewed in Refs. 1-3). Yet, one key aspect of their regulation is poorly understood: we do not know the details of how the cellular localization of STATs is regulated. To begin to probe the mechanisms that control the cellular distribution of STATs, we constructed and characterized a GFP-Stat5B fusion protein.
We find that GFP-Stat5B is regulated by GH in a manner indistinguishable from untagged Stat5B.
GFP-Stat5B is distributed equally in the cytosol and nucleus of COS-7, 2C4, and NIH-3T3 cells prior to GH stimulation. This distribution is unlikely to result from constitutive activation of Stat5B since it was present even when GHR was not co-ex- pressed (Fig. 4C), and GFP-Stat5B in the nuclear fraction of unstimulated cells was not detectably tyrosyl-phosphorylated (Fig. 9A, upper and lower panels, lane G). Also, this distribution is unlikely to be an artifact of overexpression since the same distribution was seen even in weakly fluorescent cells. Furthermore, we find endogenous Stat5B in CHO-GHR cells present in both the nucleus and cytoplasm prior to ligand stimulation. A similar distribution has been reported for endogenous Stat5B in INS-1 insulinoma cells (9). However, Stat5B is restricted to the cytoplasm prior to activation in the liver-derived CWSV-1 cell line (41) and BRL-GHR hepatoma cells. 2 Significant cell type-specific variability in the localization of other STATs has also been documented. For example, Stat1 prior to activation is mostly cytoplasmic in some cell types (7,14,42,43) but distributed throughout the nucleus and cytoplasm in other cells (5,6,8). Thus, individual STATs may possess localization properties that are regulated in a cell typedependent manner.
We selected three regions of Stat5B to mutate and assay for GH-dependent nuclear localization. Only mutation of residues 466 VVVI 469 in the DNA binding domain disrupted GH-depend-ent nuclear accumulation of Stat5B. Luo and Yu-Lee (28) have shown previously that mutating these four amino acids abrogates binding to the GAS sequence of the IRF-1 promoter in response to PRL. We report here that these same amino acids are required for GFP-Stat5B to bind to a ␤-casein promoter probe. We find that the mutant protein, GFP-Stat5B VVVI , retains early signaling events; namely it is tyrosyl-phosphorylated and forms dimers in response to GH. The failure of mutant GFP-Stat5B VVVI to localize to the nucleus in response to GH despite its ability to form dimers suggests that dimerization is not sufficient for nuclear accumulation of Stat5B. To our knowledge, this is the first identification of a region required for STAT nuclear accumulation distinct from those regions needed for tyrosyl phosphorylation and dimerization. Interestingly, Stat5B VVVI is active as an inhibitor of PRL-induced IRF-1 transcription (28). The inability of Stat5B VVVI to localize to the nucleus suggests that this inhibition occurs in the cytoplasm.
In addition to Stat5B, Stats 1, 3, and 5A possess residues VVVI in their DNA binding domain. Evidence for important functional differences in the DNA binding domain of individual STATs is emerging. Mutation of VVVI abrogates Stat5B nuclear localization in response to ligand stimulation (Fig. 7), yet mutation of VVV in STAT3 does not (44,45). It will be important to determine if the isoleucine within VVVI of Stat3 is needed for nuclear localization. Also, while conserved residues EE and VVV in the DNA binding domain of Stat3 are required for binding to DNA (44), residues EE in Stat5B are dispensable for DNA binding (28). Hence, individual residues within the DNA binding domain may have discrete functions that are specific for a given STAT.
How does mutation of 466 VVVI 469 prevent Stat5B nuclear localization? Several possibilities exist. One model consistent with our results is that in addition to being a DNA-binding determinant, VVVI may also function as a protein-protein in-2 J. Herrington and C. Carter-Su, unpublished observations.  (lanes A, D, G, and J) or with 500 ng/ml GH for 10 (lanes B, E, H, and K) or 60 min (lanes C, F, I, and L). Cytosol and nuclear extracts were immunoprecipitated (IP) with ␣Stat5B, blotted with ␣PY, and reprobed with ␣Stat5B. The molecular weight of prestained standards (ϫ10 Ϫ3 ) are indicated. B, EMSAs were performed using the cytosol and nuclear extracts from the experiment in A and a GAS-like PRL response element from the ␤-casein promoter.  A and B), wild-type GFP-Stat5B (GFP-5B, 5 g) (lanes C and D), or DNA-binding mutant GFP-Stat5B VVVI (GFP-5B VVVI , 15 g) (lanes E and F). Cells were treated without (Ϫ) or with (ϩ) 500 ng/ml GH for 30 min. Cellular proteins were immunoprecipitated (IP) with ␣Stat5A, blotted with ␣Stat5B, and reprobed with ␣PY and ␣Stat5A. The molecular weight of prestained standards (ϫ10 Ϫ3 ) and the migration wildtype GFP-Stat5B, mutant GFP-Stat5B VVVI , and Stat5A are indicated. teraction domain. In this model, VVVI would mediate the association of Stat5B with nuclear import proteins or other proteins that would serve to shuttle Stat5B into the nucleus. In the case of Stats 1 and 3, the VVVI region does not make direct DNA contact but contributes to a buttress that supports the DNA binding loop (37,38). Alternatively, mutation of VVVI of Stat5B might disrupt the structure of the protein at a more distant location, preventing the interaction of Stat5B with proteins necessary for nuclear import. Several proteins known to associate with Stat5B move between the cytosol and nucleus, including glucocorticoid receptor (46), GHR (47,48), and SH2 domain-containing protein tyrosine phosphatase-1 (41). Interaction with these or other signaling proteins may be instrumental in regulating the cellular distribution of Stat5B.
Alternatively, mutation of these residues may abrogate nuclear localization by preventing binding to nuclear components, including DNA and possibly the nuclear matrix. In this model, in cells in which Stat5B is distributed throughout the cell prior to activation, non-activated Stat5B moves into and out of the nucleus by yet unidentified transport pathways, with export being faster than import. This constitutive transport combined with the interaction of non-activated Stat5B with nuclear components yields a near equal distribution between cytosol and nucleus. Once Stat5B is tyrosyl-phosphorylated and dimerized, the significantly higher affinity binding to nuclear components, including DNA, shifts the distribution of Stat5B to being mostly nuclear. Mutation of residues VVVI prevents both nonactivated and activated Stat5B from binding to nuclear components. Now the faster export dominates, shifting the equilibrium distribution of Stat5B VVVI to being mostly cytoplasmic.
Finally, it is possible that mutation of residues VVVI prevents a phosphorylation event required for nuclear accumulation. Based on studies of rat liver Stat5B (34), we suspect that the differential phosphorylation of GFP-Stat5B VVVI compared with wild-type GFP-Stat5B arises from GH-dependent serine/ threonine phosphorylation. Modulation of nucleocytoplasmic transport of transcription factors by serine/threonine phosphorylation has considerable precedence (11). Furthermore, phosphorylation of retinoblastoma protein regulates its interaction with the nuclear matrix (49). Identification of the serine/threonine phosphorylation sites of Stat5B may provide further insight into the mechanisms regulating nuclear localization.
At present, any of the above models seem plausible. Our challenge now is to define the role of the DNA binding domain in Stat5B nuclear localization. Doing so will shed light on how the STATs transduce cell-surface signals into transcriptional regulation of diverse physiological processes.