A single residue modulates tyrosine dephosphorylation, oligomerization, and nuclear accumulation of stat transcription factors.

The NH(2) terminus of Stat proteins forms a versatile protein interaction domain that is believed to use discrete surfaces to mediate oligomerization and tyrosine dephosphorylation of Stat dimers. Here we show for Stat1 and Stat5a/b that these interfaces overlap and need to be reassigned to an unrelated region of the N-domain. Unexpectedly, our study showed for Stat1 that defective oligomerization of DNA-bound dimers was associated with prolonged interferon-induced nuclear accumulation. This uncoupling of DNA binding and nuclear retention was explained by the concomitant dephosphorylation deficiency that both Stat1 and Stat5a/b have in common and that for Stat1 was due to defective dephosphorylation by the phosphatase TC45. Furthermore, diminished N-domain-mediated oligomerization affected transcriptional activation by both Stat1 and Stat5a/b in a promoter-specific manner. DNA binding analysis indicated that oligomerization of Stats on DNA may be common, irrespective of the presence of multiple canonical binding sites. Accordingly, also transcription from promoters with only a single discernable gamma-activated sequence site was negatively effected by reduced tetramerization. Thus, these results indicate that defective oligomerization cannot generally be compensated for by enhanced tyrosine phosphorylation and prolonged nuclear accumulation. In addition, these data clarify the role of DNA binding in nuclear retention of Stat1.

The signal transducer and activator of transcription (Stat) 1 proteins comprise an evolutionary conserved family of transcription factors (1). Canonical signaling through the Janus tyrosine kinase (Jak)/Stat pathway begins at the cell membrane with the engagement of cytokines with their cognate receptor (2). This triggers the autophosphorylation on tyrosines of non-covalently attached Jak kinases, which also phosphorylate signature tyrosine residues in the intracellular receptor tails thus providing phosphotyrosine docking sites for the Stat SH2 domain (3). The bound Stat monomers detach from their receptor docking sites after phosphorylation of a single tyrosine residue at their COOH terminus and form high avidity reciprocal homo-or heterodimers (4 -6). This sequence of events is commonly referred to as "Stat activation" and within minutes triggers the accumulation of Stat dimers in the nucleus due to their inability to leave this compartment (7). Here they can bind to palindromic DNA recognition sites (GAS) and directly induce transcription (8).
The products of seven different Stat genes and various splice variants thereof can be found in mammalian cells. Gene inactivation experiments using homologous recombination have revealed that the most striking phenotypes are associated with signal transduction by particular cytokines. Stat1 is required for interferon-␣ (IFN␣) and IFN␥ signaling, and mice and humans lacking Stat1 are extremely sensitive to viral and microbial infection (9 -11). The two highly related Stat5a and Stat5b proteins (ϳ96% sequence identity) are activated by a wide range of different cytokines such as interleukin 2 (IL-2), IL-7, prolactin, and growth factors. Accordingly, gene ablation experiments uncovered mammary gland defects, growth abnormalities, and an absent IL-2 response (12)(13)(14).
Stat proteins consist of a large core domain including the DNA binding module (15) and the SH2 domain responsible for dimer formation (5). The core is connected through a proteolytically cleavable peptide to the smaller N-domain, which engages in a multitude of protein-protein interactions. The Ndomain mediates cooperative DNA binding of Stat dimers (16 -20) and controls their tyrosine dephosphorylation (21,22). The structure of the isolated N-domain reveals an all-helical hooklike appearance (23). In solution the N-domain forms stable dimers; however, two alternate binding surfaces are formed in the crystals. Recent reevaluation of the crystal-packing has indicated that the initial assignment of one of these surfaces as being responsible for cooperative DNA binding is probably incorrect (24). An unrelated region of the N-domain has previously been identified by mutagenesis to be important for the dephosphorylation of Stat1, as mutation of an invariant residue caused severe hyperphosphorylation (21). However, this mutation affects the salt bridges that interlink the helices that make up this domain and thus may impair the structural integrity of this domain. Therefore, at present the physiological interfaces that modulate cooperative DNA binding and tyro-sine dephosphorylation of the Stats remain poorly defined. Here, we investigate these issues and find for Stat1 and Stat5a/b that the two interfaces overlap. Our data reveal that DNA binding, but not dimerization, is dispensable for nuclear retention of Stat proteins and demonstrate a promoter-specific but widespread need for cooperative DNA binding in gene activation.
Plasmids, Transfections, and Luciferase Assays-pCIStat5a, pCIStat5b, pStat1-GFP, and pStat1-DNA minus have been described previously (7,17,25). Site-specific mutagenesis of these plasmids according to manufacturer's instructions (QuikChange, Stratagene; Morph sitespecific mutagenesis, 5 Prime 3 3 Prime, Inc., Boulder, CO) generated the respective N-domain mutants (F77A for Stat1, F81A for Stat5). The mutations were confirmed by DNA sequencing. Additionally for use in reporter gene assays, the cDNAs of Stat1 and Stat1F77A were cloned into pcDNA3. For IL-2 receptor reconstitution transfection assays, 4 ϫ 10 6 293T cells were transfected with the following DNAs: 2 g of IL-2 receptor ␤, 0.5 g of ␥ c , 0.25 g of Jak3, and 0.5 g (for electrophoretic mobility shift assays) or 10 ng (for luciferase assay) of each of the WT Stat5 or F81A mutant expression plasmids. Additionally for the luciferase assay, 1 g of positive regulatory region III (PRRIII)-E1b-or M10-E1b-luciferase reporter constructs (17) and 0.5 ng of the transfection control plasmid, pRLTK-luciferase, was added. The total amount of DNA transfected was kept at a constant amount of 10 g with empty pCI vector (Promega). Twenty-four hours after transfection, where appropriate, cells were split into two, and 48 h after transfection one set was left untreated and the other was treated with 2 nM IL-2 for 30 min (for nuclear extracts) or overnight (for luciferase assays). Dual luciferase assays were performed according to manufacturer's protocols (Promega). Reporter gene assays with transfected U3A cells and IFN␥ stimulation were done as described previously (25). The reporter genes pIC339 and pIC1352 (a kind gift from Dr. van der Saag) contain 339 or 1352 bp (relative to the transcription start site) from the 5Ј-region of the human intercellular adhesion molecule 1 (ICAM-1) gene (27). Transfections for luciferase assays were performed in triplicate and the data presented are the means Ϯ S.D. values for the three independent experiments.
Nuclear Extracts and EMSAs-Following transfections, cytoplasmic and nuclear extracts were prepared as described previously (26). Total protein was determined by the Bradford method with a bovine serum albumin standard (Bio-Rad). All buffers used in the cell extract preparations and EMSAs contained protease inhibitor mixture-1 (Calbiochem), 1 mM sodium orthovanadate, and 10 mM sodium fluoride. Ten g of nuclear extracts was incubated with 40,000 cpm of 32 P-labeled DNA probe, labeled by end-filling reaction using Klenow fragment (New England Biolabs). The WT PRRIII and MGF probes have been previously described (28). EMSA reactions containing Stat1 were done with in vitro phosphorylated recombinant protein or cytoplasmic extracts from IFN␥-treated reconstituted U3A cells (25), and the following duplex oligonucleotides (4-bp overhangs at each 5Ј-end are not included; the respective antisense oligonucleotides are not listed): M67, 5Ј-CGA-CATTTCCCGTAAATCTG; human IRF-1, 5Ј-CGACATTTCCCCGAAA-TCTG; 2ϫ-GAS, 5Ј-CGTTTCCCCGAAATTGACGGATTTCCCCGAAC; GASϩGAS-like, 5Ј-CGTTTCCCGGAAATAGAAGATTATTATCATTAT; GASϩnon-GAS, 5Ј-CGTTTCCCGGAAATAGAAGAACCTCGTTGT-CAC; 2ϫ-non-GAS, 5Ј-CGAGGTACAGGTAAAGAAGAACCTCGTTGT-CAC; GAS-like sites are underlined. For competition assays the EMSA reaction was equilibrated for 15 min at room temperature before adding a 750-fold molar excess of the respective unlabeled DNA. Thereafter, the reaction was incubated at room temperature for another 30 min (or as indicated). In supershift assays 20 ng of a Stat1-specific antibody was added and incubated with the EMSA reaction for 15 min at room temperature.
Dephosphorylation Assays and Western Blot Analysis-In vivo dephosphorylation was determined as follows. Forty-eight hours after transfection, cells were treated with IFN␥ or IL-2 for 30 min, and then the cytokine-containing medium was replaced with either fresh me-dium or medium containing Me 2 SO, MG132, staurosporine, or MG132 ϩ staurosporine as indicated. Samples were harvested immediately (zero time point) or at the indicated times after removal of cytokine and replacement with new medium containing appropriate chemicals. Cells were washed once with ice-cold phosphate-buffered saline and whole cell lysates were prepared by boiling the cell pellet in 200 l of SDS sample buffer for 5 min. Ten l of each sample was loaded on 8% Tris-glycine-SDS gels. Western blotting, blot development, and reprobing were done as described previously (25) using a phosphotyrosinespecific Stat5 antibody (Tyr 694 , ␣Stat5-P) or a pan-Stat5 antibody (both from Transduction Laboratories) or a Tyr 701 phosphospecific Stat1-P (Cell Signaling) or Stat1 antibody (Santa Cruz Biotechnology). In vitro dephosphorylation reactions were performed for 60 min with 0.5 nM Stat1 dimers and 1.5 units of truncated TC45 (Sigma) and the indicated amount of unphosphorylated Stat1 as described previously (7). Unphosphorylated Stat1 was added at a concentration of 0.5, 5, or 20 nM. The mixture was resolved by SDS-PAGE after boiling in SDS sample buffer.
Expression and Purification of Recombinant Proteins-Recombinant Stat1␣ and His-tagged Stat5a proteins were produced in baculovirus infected insect cells and purified as described previously (7,28). Stat5a was tyrosine-phosphorylated by an endogenous kinase during expression in the insect cells (28). In vitro tyrosine phosphorylation with epidermal growth factor receptor and isolation of phosphorylated Stat1 were done as described previously (7). For production of Stat1F77A, the correctly mutated cDNA was excised (NotI) from pcDNA-Stat1F77A and cloned into the baculovirus transfer vector pFastBac (Invitrogen). For production of recombinant Stat5aF81A the wild type Stat5a cDNA sequence, which was cloned into the baculovirus transfer vector pBacPAK8 (Clontech) was mutated by site-directed mutagenesis at the appropriate residue and confirmed by DNA sequence analysis. Both Stat5a and Stat5aF81A were confirmed to be tyrosine-phosphorylated by Western blot analysis using anti-phospho-Stat5 antibody. Truncated TC45 was produced in bacteria and purified as described previously (7).
Fluorescence Microscopy and Fluorescence Recovery after Photobleaching (FRAP)-Fluorescence microscopy and FRAP were performed as described previously (7).

RESULTS
The Alternate N-domain Interface Mediates Cooperative DNA Binding of Stat1 and Stat5a/b-Studies of Stat oligomerization from our and other labs have indicated that a single invariant residue (Trp 37 of Stat1) that is located in the crystallographic interface of N-domain dimers affects tetramerization of fulllength Stats (17,20,23). However, in vitro studies with the isolated N-domain demonstrated that mutation of Trp 37 causes destabilization and degradation of the N-domain (24), and the Stat5W37A mutants were previously found to be expressed less efficiently compared with their WT counterparts in either a baculovirus expression system or in IL-2 receptor-reconstituted 293T cells (17,28). The crystal packing of the isolated Ndomain also revealed another mode of N-domain interactions not involving Trp 37 (24). Therefore, we mutated in the fulllength Stat molecule a conserved hydrophobic residue on the surface of the N-domain that was demonstrated to affect dimerization of N-domains (Phe 77 3 Ala in Stat1 and Phe 81 3 Ala in Stat5a/b, respectively). As is shown in Fig. 1, A and B, the resulting Stat1 and Stat5b mutants were expressed well and no indication of structural instabilities was detected by SDS-PAGE and Western blotting. However, expression of the Stat5aF81A mutant was reduced in 293T cells compared with WT (Fig. 1B). Therefore, we produced recombinant Stat5a protein in baculovirus infected insect cells (Fig. 1C). In these cells, Stat5aF81A was expressed even better than WT. Phosphoryl-

FIG. 1. Expression and DNA binding of N-domain mutants of Stat1 and Stat5a/b. A, HeLa cells transiently expressing GFP-tagged
Stat1WT or F77A mutant were treated with or without IFN␥ for 30 min and harvested. Shown are Western blots of whole cell extracts developed with a phospho-Stat1-specific antibody (␣Stat1-P) and a Stat1-specific antibody. B, Western blot of nuclear extracts (2 g of protein/lane) from IL-2 receptor-reconstituted 293T cells. The expression of transfected Stat5a or Stat5b was probed with a pan-Stat5 antibody. C, increasing amounts of purified recombinant Stat5aWT or F81A-mutant (125, 250, or 500 ng) were Western blotted and probed with a pan-Stat5 and phospho-Stat5 antibody. D, competition gel-shift assay with cytoplasmic extracts from IFN␥-stimulated reconstituted U3A cells and a single GAS site. Before loading on the gel, the reaction mixture was incubated for the indicated times in the presence of a 750-fold molar excess of unlabeled oligo; alternatively the reaction included a Stat1-specific antibody (S). The positions of dimeric Stat1 (D) and an unspecific band (*) are indicated. E, gel-shift analysis of wild type or mutant Stat5a and Stat5b with a single GAS element (MGF). Phosphorylated Stat5a (125, 250, or 500 ng) purified from insect cells and nuclear extracts (10 g protein/lane) from untreated or IL-2 treated reconstituted 293T cells expressing Stat5b were used. F, gel-shift analysis of wild type and mutant Stat5a and Stat5b. Left panel, nuclear extracts (10 g of protein/lane) from IL-2 receptor-reconstituted 293T cells expressing the indicated constructs were incubated with a single high affinity MGF site or the PRRIII probe containing tandem ation by an endogenous kinase yielded correctly tyrosine-phosphorylated Stat5a protein ( Fig. 1C and Ref. 17), and this material was therefore used in subsequent DNA binding experiments.
The Phe 77 and Phe 81 mutants of Stat1 or Stat5a/b, respectively, were tyrosine phosphorylated in response to cytokine stimulation of cells and bound to a single optimal GAS site (M67 and MGF, respectively) similar to wild type, as did recombinant Stat5aF81A (Fig. 1, D, E, and left panel of F). Next, we tested the ability to form tetramers on tandem GAS sites by EMSA analysis. For Stat5 we used an oligonucleotide containing tandemly linked GAS motifs, derived from the IL-2 response element that is contained in both the murine and human genes coding for the ␣-subunit of the IL-2 receptor, termed PRRIII. We and others (17,28,29) have previously shown that binding of endogenous Stat5 to isolated single GAS sites from PRRIII is not discernable, but an oligonucleotide comprising full-length PRRIII binds Stat5a and Stat5b as tetramers (Fig.  1F). Consistent with previous observations (28), tetramerization of Stat5b on PRRIII was weak compared with Stat5a (Fig.  1F, left panel). In contrast to wild type Stat5a/b, mutation of residue Phe 81 to Ala caused a loss of tetramer formation as indicated by the inability to bind to the tandemly linked GAS motifs in PRRIII (Fig. 1F, left panel). In addition to nuclear extracts from reconstituted 293T cells, we used purified recombinant Stat5a proteins to explore binding to the PRRIII probe, which confirmed the tetramerization defect of the Stat5aF81A mutant (Fig. 1F, right panel). EMSA analysis with tandem binding sites was also performed with Stat1WT and Stat1F77A (Fig. 1G). An artificial construct containing two high affinity GAS sites from the human IRF-1 promoter (5Ј-TTCCCCGAA) was used in these experiments. Stat1 can bind independently to either site, resulting in both a fast migrating Stat1/DNAcomplex containing a single Stat1 dimer as well as a slow migrating complex with two Stat1 dimers. When such complexes were challenged for 30 min with a 750-fold molar excess of unlabeled oligonucleotide, the dimeric complexes of both wild type and mutant were displaced. However, the tetrameric complex resisted displacement only with Stat1WT, whereas the respective complex of the F77A mutant was dispelled (Fig. 1G), indicating that adjacent dimers of Stat1F77A interacted only weakly via stabilizing contacts.
To further explore DNA binding and tetramerization of Stat1, we included in our analysis another three 41-mer oligonucleotides. Two of them each contained a single canonical GAS site (5Ј-TTCCCGGAA), and a second site 11 bp away that was either unrelated (non-GAS; 5Ј-TCGTTGTCA) or bore little resemblance to GAS sites (GAS-like; 5Ј-TTATCATTA). Additionally, a 41-mer without sequences resembling GAS elements was used (see "Experimental Procedures"). We used in vitro tyrosine-phosphorylated Stat1 at a concentration of 1 nM for these experiments (which is also the apparent equilibrium constant of the Stat1 DNA-binding reaction; see Ref. 18). The results are shown in Fig. 1, H and I. Expectedly, a fast migrating band of dimeric Stat1 bound to canonical GAS sites was seen (Fig. 1H, lanes 2-5). Strikingly, Stat1 tetramers formed not only with the tandem canonical GAS sites (Fig. 1H, lane 2) but also with the 41-mer oligonucleotides that harbored only a single GAS site (Fig. 1H, lanes 4 and 5). DNA binding of Stat1 was observed even in the complete absence of any GAS resemblance of the DNA, but binding was confined to tetramers (Fig.  1H, lane 6). Of note, tetramerization required the presence of a DNA stretch, as no band indicative of tetramer formation was seen with a short 28-mer oligonucleotide containing a single binding site (Fig. 1H, lane 3). Importantly, the mutant F77A did not form tetramers on DNA where a single high affinity GAS site was adjacent to a GAS-like site (5Ј-TTATCATTA) (Fig. 1I, compare with Fig. 1G, where oligonucleotides containing tandem canonical GAS sites were used). Thus, N-domain interactions of DNA-bound Stat molecules made tetramerization possible even in the presence of only a single canonical GAS site. In summary, these data revealed that the phenylalanine residues in position 81 of Stat5 and 77 of Stat1 maintain N-domain interactions. We thus conclude that the recently identified alternate interface is physiologically relevant.
Tetramerization Deficiency Is Associated with Prolonged Nuclear Accumulation-The molecular mechanisms underlying nuclear accumulation of Stat1 were elucidated very recently (7). It was shown that Stat1 can leave the nucleus only after its tyrosine dephosphorylation, and DNA binding was identified as the critical regulator of the conversion. Importantly, impaired DNA binding resulted in shortened nuclear retention. Thus, we investigated nuclear accumulation of the tetramerization-deficient Stat1F77A mutant, expecting diminished nuclear retention due to the negative effect that the mutation exerted on DNA binding. Surprisingly, our expectations were not met, as we observed a markedly prolonged accumulation phase in IFN␥-stimulated cells ( Fig. 2A). Moreover, we observed a distinct insensitivity to pharmacological kinase inactivation, since nuclear accumulation of the mutant persisted also in the presence of staurosporine (Fig. 2B). This is in stark contrast to wild type Stat1, where staurosporine caused a rapid collapse of nuclear accumulation (Fig. 2B). We then employed an antibody microinjection assay to explore the rate of nucleocytoplasmic shuttling of GFP-tagged Stat1 during nuclear accumulation (7,26). It was previously demonstrated that antibody microinjection can cause precipitation of the targeted molecule. In the case of a shuttling protein this lead to the depletion of the respective molecule from the non-injected compartment (26). This is shown in Fig. 2C for Stat1WT during its nuclear accumulation in interferon-␥ stimulated cells. Strikingly, the nuclear retention time during IFN␥-induced accumulation was considerably extended for mutant Stat1, as it was impossible to achieve its clearance from the nucleus after cytoplasmic microinjection of a GFP antibody (Fig. 2C).
According to our previous results the intranuclear mobility of Stat1 is inversely correlated with the strength of DNA binding (7). We employed FRAP analysis to examine this also for Stat1F77A (Table I). Expectedly, both the mutant and wild type Stat1 were highly mobile during nuclear accumulation, GAS sites. Twenty micrograms of nuclear extract were loaded in lanes 9 and 10 to account for the weaker tetramerization of Stat5b. Right panel, and the mobility was reduced only modestly by the inactivation of tyrosine phosphatases. Notably, though, the mutant was even less sensitive to phosphatase inhibition, as its mobility remained significantly higher than wild type (Table I). These results indicated that the prolonged nuclear accumulation period of the mutant Stat1F77A was uncoupled both from DNA binding capacity and reduced intranuclear mobility. The mutant thus differed from the nuclear retention mutants described to date (7). We therefore tested whether the loss of nuclear accumulation that was previously observed with the DNA binding mutant Stat1-DNA minus , which has lost both sequence-specific and -unspecific DNA binding (7), could be reversed by the additional mutation of Phe 77 . As is shown in Fig. 2D, this was indeed the case, as the F77A mutant of Stat1-DNA minus displayed completely restored nuclear accumulation after 60 min of IFN␥ stimulation. These observations can be reconciled with the model of Stat nucleocytoplasmic shuttling if we propose a dephosphorylation defect to be part of

FIG. 2. Nuclear accumulation and nucleocytoplasmic mobility of mutant Stat1. A, HeLa cells transiently expressing GFP fusions with
Stat1WT or the F77A mutant were pretreated without (w/o) or with IFN␥ for 60 min, before incubation was continued in IFN␥-free medium. After the indicated times the cells were fixed, and the GFP fluorescence was recorded. In addition the location of the nuclei is indicated (Hoechst stain). B, same as A, but the IFN␥-free medium was supplemented with staurosporine (500 nM) to block kinase activity. C, collapse of nuclear accumulation of wild type Stat1, but not of the F77A mutant, after cytoplasmic microinjection of Stat1-precipitating antibodies. HeLa cells expressing GFP fusions of Stat1WT or F77A mutant were treated with IFN␥ for 60 min to induce nuclear accumulation, before cytoplasmic microinjection of GFP antibody. Thereafter, the cells were further incubated in the absence of IFN␥ for another 60 min (Stat1WT) or 90 min (F77A) followed by fixation and fluorescence microscopy. Note the reduced nuclear GFP fluorescence in the injected cell expressing Stat1WT. Precipitated Stat1-GFP appears as dots in the cytoplasm. D, HeLa cells expressing GFP fusion proteins with Stat1-DNA minus or Stat1F77A-DNA minus were stimulated with IFN␥ for 60 min before fixation and fluorescence microscopy. the mutant phenotype. This was investigated next.
The Interfaces for Dimer-Dimer and Dimer-Phosphatase Interactions Are Overlapping-At first, the kinetics of tyrosine dephosphorylation was investigated in living cells. U3A cells, which do not express endogenous Stat1 (30), or 293T cells that were reconstituted with IL-2 receptor components, were transfected with plasmids expressing Stat1 and Stat5, respectively. Subsequently, the cells were stimulated for 30 min with IFN␥ to activate Stat1 or IL-2 to activate Stat5. After this initial stimulation period the cytokines were removed, and the cells were further incubated in the absence or presence of the kinase inhibitor staurosporine. As is shown in Fig. 3A, Stat1F77A displayed prolonged tyrosine phosphorylation and insensitivity toward staurosporine, which is indicative of a reduced dephosphorylation rate. A similar outcome was observed also for Stat5a and Stat5b (Fig. 3B). Interestingly, in the IL-2 receptorreconstituted 293T system the kinetics of dephosphorylation of wild type Stat5 proteins is longer (2-4 h) than that observed in lymphoid cells, where it occurs within 1-2 h 2 (Fig. 3B). Nevertheless, mutation of residue 81 also prolonged tyrosine phosphorylation even further, as the phosphotyrosine signal now remained detectable for more than 6 h. The addition of staurosporine to the cells terminated the phosphotyrosine signal of wild type Stat5a/b within 1 h after cytokine removal (Fig. 3B), while the mutants were less sensitive to kinase inhibition and showed resistance to dephosphorylation, particularly Stat5bF81A, where the signal was detectable for at least 2 h.
As was reported earlier, tyrosine-phosphorylated Stat5 is stabilized by proteasome inhibitors such as MG132 (Fig. 3B) (31). Addition of the kinase inhibitor staurosporine at the same time as the proteasome inhibitor cancelled the effect of MG132 and caused rapid dephosphorylation (Fig. 3B). Importantly, the N-domain mutants of Stat5a/b were again refractory to the action of staurosporine and stayed phosphorylated much longer than wild type Stat5. To exclude the trivial explanation that MG132 is simply inactivated by the presence of staurosporine, we examined the well known MG132-sensitive degradation of IB␣ after TNF␣ stimulation of cells (32). Clearly, MG132 was a potent inhibitor of the TNF␣-induced degradation of IB␣ both in the absence and presence of staurosporine (data not shown). We therefore conclude that the observed proteasomedependent down-regulation of phosphorylated Stat5 is not the result of direct Stat5 targeting by the proteasome. Rather, during MG132 treatment a constituent of the IL-2 pathway is preserved that ultimately enhances kinase activity. In summary, these findings show that Stat1F77A and Stat5F81A are not only tetramerization-deficient but also display defective tyrosine dephosphorylation in vivo.
To examine whether the N-domain mutation directly affected the interaction between Stat dimer and phosphatase, we performed in vitro dephosphorylation assays with purified tyrosine-phosphorylated recombinant Stat1F77A and TC45, a ubiquitously expressed Stat1-specific nuclear tyrosine phos-phatase (33). The dephosphorylation rate of the Stat1 variant protein was strongly reduced (Fig. 3C), indicating that this mutant is defective in direct Stat-phosphatase interactions. Thus, the N-domain surface responsible for cooperative DNA binding of Stat1 dimers is also governing interactions with the phosphatase TC45. When we added up to 40-fold molar excess of unphosphorylated Stat1 to the in vitro dephosphorylation reaction, we did not note an influence on the interaction of phosphatase and its substrate (Fig. 3D), which indicates that the dephosphorylation reaction is not subject to product inhibition.
Mutation of the Tetramerization Interface Affects Transcription in a Promoter-specific Manner-Finally the impact of the N-domain point mutation on gene transcription was investigated. Reporter gene assays were used to assess the consequences of decreased tetramer stability and enhanced tyrosine phosphorylation on gene induction. As shown in Fig. 4, the observed effects depended on the promoter construct used. Reporter genes containing optimal GAS sites were transactivated as good as or even better by the mutant Stats than by the respective wild type counterparts (Fig. 4, A-D). For Stat1 a synthetic promoter with three strong GAS sites separated by 10 bp was used (Fig. 4A). However, when natural promoter fragments were tested, transactivation was lower with the mutants as compared with the respective wild type Stat protein. This was exemplified for Stat1 by transient transfection of a luciferase reporter containing the full-length or truncated promoter of the human ICAM-1 gene (Fig. 4B). Besides binding sites for NFB, AP-2, and Sp1, this promoter harbors a single canonical GAS site (5Ј-TTCCGGGAA) (27, 34).
Analysis of Stat5-dependent transcription was examined in reporter gene assays with WT PRRIII and a PRRIII derivative (M10), in which the natural weak tandem GAS sites were transformed into strong Stat5 binding sites (17). IL-2 inducibility of the wild type PRRIII promoter was lost for both Stat5a and Stat5b mutants ( Fig. 4C; note the considerable reporter gene activation by IL-2 even in the absence of Stat5). However, both the Stat5F81A mutants were better than their wild type counterparts at activating transcription from the high affinity M10 promoter (Fig. 4D). Thus, together these results demonstrate that only natural GAS sites were sensitive to weakened N-domain interactions.
We also tested expression of endogenous IFN␥-responsive genes by semiquantitative RT-PCR (Fig. 4E). Of the seven genes tested, only LMP-2 expression was unaffected by the N-domain mutation of Stat1, whereas the others showed a strongly diminished IFN␥-responsiveness in cells expressing Stat1F77A. Strong effects were seen with the MIG-1, IRF-1, IRF-9, and GBP-1 genes, which had lost their IFN␥ responsiveness in cells expressing the Stat1 mutant protein. The importance of tetramer formation for gene induction in a natural setting was further evaluated by scoring gene induction in response to IFN␣. Contrary to IFN␥, which signals via a Stat1 homodimer, stimulation of cells with IFN␣ induces the additional formation of a ternary complex called ISGF-3, consisting of a Stat1/Stat2 heterodimer and IRF-9, which binds to an unrelated recognition site termed interferon-stimulated response element (35). Two genes were examined by RT-PCR. Expression of the ISG-15 gene was highly induced by IFN␣, and mutation of Phe 77 reduced its IFN␣ response by one half, whereas the weak induction of ISG-56 was not adversely affected by the N-domain mutation (Fig. 4E). DISCUSSION The formation of Stat tetramers and higher order oligomers on DNA results from cooperative DNA binding of Stat dimers that appears to be conserved throughout the Stat family, and 2 S. John, unpublished observation. an invariant Trp residue in position 37 of the N-domain has been demonstrated to influence this process (17,18,20). However, recent in vitro evidence indicated that mutation of residue 37 might cause global perturbations of the domain structure, and an alternate surface was proposed to mediate Stat oligomerization (24). This study was initiated to distinguish between these possibilities. Our data demonstrate that mutation of a conserved residue located in the alternate surface does not compromise protein stability but causes a severe decrease in tetramer formation of Stat1, Stat5a, and Stat5b without overtly affecting DNA binding of the dimer. During preparation of this manuscript, the same conclusion was reached also for Stat4 (36). Thus, based on these results we propose the alternate dimerization surface of N-domains as physiologically relevant for oligomerization of Stat proteins. Previous data generated with Trp 37 mutants most likely reflect the consequences of nonspecific N-domain destabilization rather than specific interface perturbations. Nevertheless, conclusions concerning cooperative DNA binding remain correct, but data generated in vivo, e.g. transcription analyses, should be reconsidered for the same reason. It remains to be seen whether the original interface of N-domain dimers with Trp 37 at its center is required for functions other than oligomerization, such as interactions between unphosphorylated Stats. Our study shows that the interface that organizes cooperative DNA binding is also required for tyrosine dephosphorylation of Stat1 and Stat5, as the oligomerization mutants were refractory to tyrosine dephosphorylation in vivo. By using in vitro dephosphorylation assays with purified proteins, for Stat1 this could be attributed to a loss of dephosphorylation by the phosphatase TC45. Moreover, progression of the Stat1 dephos-phorylation reaction was not inhibited by its reaction product, stressing the very different biochemical properties of the Ndomain of phosphorylated and unphosphorylated Stat1. Currently, the role of the N-domain in the inactivation of Stat5 is controversial, since the carboxyl-terminal transactivation domain has been implicated in the proteolytic turnover of tyrosine-phosphorylated Stat5 (31,37). Here, it is demonstrated that stabilization of activated Stat5 by the proteasome inhibitor MG132 is not due to protection from proteolytic degradation or phosphatase attack, as simultaneous treatment with proteasome and kinase inhibitors resulted in rapid dephosphorylation of wild type Stat5 (Fig. 3B). Importantly, the N-domain mutants both resisted dephosphorylation under these conditions. This indicates that the prolonged tyrosine phosphorylation observed after removal of the COOH terminus and in the presence of MG132 is in fact the result of Stat5 hyperphosphorylation and not of defective dephosphorylation. This conclusion is supported by reports showing that proteasome inhibition prevents Jak kinase inactivation and receptor-mediated endocytosis of the interleukin-2 receptor complex (38,39). The identity of the Stat5 phosphatase is less clear as different studies have indicated different phosphatases, such as SHP-2, PTP-1B, and TC45, to be involved in the dephosphorylation of Stat5 (33, 40 -42).
The cytokine-induced nuclear accumulation of Stats requires the presence of the N-domain (22). Mutation of residue 77 in Stat1 and 81 in Stat5a/b did not affect nuclear import after stimulation with cytokines. On the contrary, mutation of phenylalanine 77 of Stat1 even prolonged nuclear accumulation ( Fig. 2A). Both its dephosphorylation defect as well as its tetramerization deficiency measurably affected the nuclear accu- Cells were stimulated without (gray bars) or with IFN␥ for 6 h (black). C and D, IL-2-reconstituted 293T cells were co-transfected with wild type or mutant Stat5a or Stat5b and a luciferase reporter gene. Stimulation without (gray) or with IL-2 was overnight (black). Luciferase expression was activated from a natural PRRIII promoter (C) or a synthetic M10 element (D). E, left, RT-PCR analysis of endogenous IFN␣-and IFN␥-responsive genes. The assay was performed with U3A cells stably expressing Stat1WT or F77A mutant. Cells were grown for 15 h in Dulbecco's modified Eagle's medium with reduced serum (1%), followed by the addition of serum to 10% and treatment with or without IFN␣ or IFN␥ for 4 h. RNA was extracted and reverse transcribed followed by PCR. Expressions of Stat1 and GAPDH are shown as controls. Shown are the ethidium bromide-stained PCR products (no RT, no reverse transcription before PCR). Right, quantification of the ethidium bromide staining on the left. Absolute intensities were scored in arbitrary units, and the ratio between induced and uninduced signal is depicted for each gene. mulation phenotype of the mutant F77A. Most striking was the loss of nucleocytoplasmic shuttling during nuclear accumulation (Fig. 2C), which was also demonstrated during treatment with the kinase inhibitor staurosporine (Fig. 2B). Importantly, the phenotype of F77A reveals new details about the role that DNA binding plays in the nuclear accumulation of Stat1. It was hypothesized that access to a nuclear export signal within the DNA binding domain was regulated by DNA binding (43). According to this model, DNA binding would mask the nuclear export signal and thus preclude nuclear export. Consequently, loss of DNA binding would inadvertently result in loss of nuclear accumulation. However, this is not the case. Nuclear retention of DNA binding mutants can be restored by treatment of cells with phosphatase inhibitors (7), or as is shown in Fig. 2D, by additional mutation of the interface of Stat1 that specifically controls its dephosphorylation. These results confirm the finding that tyrosine-phosphorylated Stat1 is retained in the nucleus (7), which makes dephosphorylation the crucial step in the control of nuclear export. For this reason it is dimerization, and not subsequent DNA binding, that masks the region required for Stat1 nuclear export. Whether this region is indeed localized in the DNA binding domain awaits further study. Thus, DNA binding is not necessary, but sufficient, for nuclear retention of Stat1, because the DNA bound molecule is refractory to phosphatase attack.
At the same time also the concomitant tetramerization defect of the mutant F77A influenced nuclear accumulation, which became apparent when we examined the intranuclear mobility during cytokine stimulation. Interestingly, statistically significant mobility differences between wild type and mutant Stat1 were found only during vanadate treatment. The addition of vanadate, which inhibits tyrosine phosphatase activity, results in increased nuclear concentration of Stat1 dimers. It is reasonable to assume that oligomerization of Stat1 is favored in this situation. However, due to its tetramerization defect the mutant will form oligomers less frequently and thus maintain a comparatively high mobility even in the face of its intrinsic dephosphorylation defect. Nevertheless, despite its high intranuclear mobility, a feature that discriminates this protein from Stat1-DNA plus , which also displays prolonged nuclear retention (7), the concurrent dephosphorylation defect of mutant F77A prevented its nuclear export. Thus, because loss of dephosphorylation was associated with diminished oligomerization, the mutant F77A displayed a unique phenotype where efficient nuclear trapping was combined with elevated intranuclear mobility.
Our transcription analyses have indicated that mutation of Phe 77 influences gene induction in a promoter-specific manner. Transcription from transiently transfected "strong" synthetic promoters was elevated with the tetramerization-deficient Stat1 and Stat5, whereas reduced transcription was seen when natural promoter fragments were used in this kind of assay. These results clearly indicated that the mutant Stats are functional transcription factors that retained cytokine responsiveness. RT-PCR analysis of natural IFN␥-or IFN␣-dependent target genes in their native chromatin setting demonstrated reduced transcriptional activity of the Stat1 mutant protein. At present the molecular defects underlying these phenotypes are not clear. Together, the reporter gene results with synthetic promoters argue for an influence of tetramerization deficiency, since optimal GAS sites are characterized by a markedly lower DNA off rate as compared with many physiological sites. Therefore effective binding to such sites is less dependent upon cooperativity (18). The multimerization of optimal sites may mimic oligomerization, while the natural promoter fragments that were used contained "weak" sites and thus required Stat oligomerization for transcriptional activity. We note that the ICAM-1 enhancer not only contains a single canonical GAS site but additionally a non-canonical GAS site (5Ј-TTCCGGAGG) just 14 bp upstream. 3 Somewhat surprisingly, also an IFN␥responsive promoter with only a single discernable GAS site, such as IRF-1 (44), was suceptible to the mutation of Phe 77 , despite the fact that binding of the mutant F77A to single (not shown) or tandem IRF-1 sites (Fig. 1G) was indiscriminable from wild type. This may indicate that Stat tetramers are mandatory for full transcriptional activity at many promoters that contain only a single canonical GAS site. This interpretation is supported by the DNA binding data shown in Fig. 1, H and I. Clearly, the ability to form tetramers (or higher order polymers) on DNA did not require multiple GAS elements. Therefore, it is likely that tetramers of Stat1 are also bound at promoters that contain only a single canonical GAS site. This situation may not be limited to Stat1. In vitro binding site selection revealed that tetrameric binding of Stat5 also can be seen with a wide range of non-consensus motifs, which in many cases did not allow Stat5a binding as a dimer (28). Additionally, it was shown before that the N-domain is dispensable for high affinity DNA binding (18,19), and Fig. 1, D and E, show that mutation of Phe 77 or Phe 81 in Stat1 or Stat5, respectively, did not diminish interactions of the dimers with DNA.
Although mutation of residue 77 of Stat1 or 81 of Stat5 was without adverse effects on gene induction from promoters containing multiple strong GAS elements, an alternative explanation for the observed reduced transcription from natural promoters must be considered, since the N-domain has a role also in the recruitment of transcriptional co-activators (45). At present we cannot exclude the possibility that the mutation of residue Phe 77 modulates also recruitment of unrelated transcription factors. Further analysis is required to decide this point.
In summary, the interface centered around the critical phenylalanine fulfills distinct functions that are subject to control by the tyrosine phosphorylation and DNA binding status of the Stat molecule. The N-domains of unphosphorylated Stats appeared not to interact with the phosphatase (Fig. 3D). In the activated molecule, however, Phe 77 participates in recruitment of the phosphatase (Fig. 3C), whereas during DNA binding the Stat1 protein is protected from dephosphorylation (7). At present no structural data with regard to the position of the Ndomain in phosphorylated and unphosphorylated Stats are available. It is therefore not possible to describe the various states in structural terms. From this and other work (22,36,45) it becomes clear that the surface of the Stat N-domain influences a remarkable diversity of crucially important functions, and despite high structural conservation the degree of functional variability appears to be considerable.