Implications of an antiparallel dimeric structure of nonphosphorylated STAT1 for the activation–inactivation cycle

Abstract

IFN-γ treatment of cells leads to tyrosine phosphorylation of signal transducer and activator of transcription (STAT) 1 followed by dimerization through a reciprocal Src homology 2–phosphotyrosine interaction near the –COOH end of each monomer, forming a parallel structure that accumulates in the nucleus to drive transcription. Prompt dephosphorylation and return to the cytoplasm completes the activation–inactivation cycle. Nonphosphorylated STATs dimerize, and a previously described interface between N-terminal domain (ND) dimers has been implicated in this dimerization. A new crystal structure of nonphosphorylated STAT1 containing the ND dimer has two possible configurations for the body of STAT1, one of which is antiparallel. In this antiparallel structure, the Src homology 2 domains are at opposite ends of the dimer, with the coiled:coil domain of one monomer interacting reciprocally with the DNA-binding domain of its partner. Here, we find that mutations in either the coiled:coil/DNA-binding domain interface or the ND dimer interface block dimerization of nonphosphorylated molecules and cause a resistance to dephosphorylation in vivo and resistance to a tyrosine phosphatase in vitro. We conclude that a parallel STAT1 phosphodimer not bound to DNA most likely undergoes a conformational rearrangement (parallel to antiparallel) to present the phosphotyrosine efficiently for dephosphorylation.

Signal transducers and activators of transcription (STATs) are latent cytoplasmic transcription factors that can be activated by a variety of tyrosine kinases in response to many different cytokine, growth factor, and peptide ligands binding to their respective cell surface receptors. Accumulation in the nucleus of tyrosine phosphorylated STAT dimers is followed by DNA binding, activation of target gene transcription, dephosphorylation, and return to the cytoplasm (1). The crystal structure of the phosphodimer core (amino acids 130–712), bound to DNA, showed a reciprocal phosphotyrosine (pY)–Src homology 2 (SH2) interaction at one end of a parallel dimer (2), with the DNA separating the monomers along the long axis of the monomers. Mao et al. (3) have now described a dimeric structure of nonphosphorylated STAT1, including the core (amino acids 130–712), as well as the core plus the N-terminal domain (ND) (amino acids 1–683). (The terminal 29 aa, 684–712, in the core were unstructured and omitted from study.) The body of each of the monomers in the nonphosphorylated structure is identical to the monomers in the previously reported phosphorylated dimer (2). However, the monomers in the nonphosphorylated core structure are arranged differently in space. In contrast to the dimeric, parallel structure of the phosphorylated STAT1, the SH2 domains in the nonphosphorylated core structure project from the opposite ends of an antiparallel dimer with a dimeric interface formed by reciprocal contacts between the coiled:coil (CC) and DNA-binding domains (DBD) of the monomers. In the crystal structure of nonphosphorylated STAT1 containing the ND, two types of dimers were observed [Mao et al. (3); see diagrams in Figs. 1B and 4]. One of these contained the antiparallel structure seen in the core fragment with a loose association between the DBD regions and a dimeric ND. The dimeric NDs in the nonphosphorylated structure have the same interface as was previously recognized from the crystallographic structure of isolated dimeric NDs (4, 5). The other possible nonphosphorylated STAT1 (1–683) structure had a parallel orientation of the bodies of the dimeric partners with the ND dimer lodged between.

Fig. 1.

Coimmunoprecipitation of nonphosphorylated differentially tagged STAT1 proteins. (A) Sites of domain boundaries, point mutations, and epitope tagging are indicated. (B) Diagram of known STAT1 structures (2, 3). The DNA (blue coil)-bound tyrosine-phosphorylated structure is labeled parallel; in this structure, the NDs cannot interact (2). The two antiparallel diagrams are from ref. 3 and are discussed in the Introduction. (C) 293T cells were transfected to express the indicated epitope-tagged proteins: F172W, single mutant; F77A/L78A, double mutant; F172W/F77A/L78A, triple mutant. Cell extracts were immunoprecipitated with anti-Flag antibody (IP:Flag) or anti-Myc antibody (IP:Myc). After precipitation and epitope release, SDS/PAGE and immunoblotting (IB) with anti-Myc or anti-Flag antibody was carried out.

Fig. 4.

Possible scenarios for STAT1 inactivation and predicted phenotypes for the F172W, F77A, and L78A mutants.

These new structural data provide a basis for the often reported occurrence of nonphosphorylated STAT dimers (including STAT1) in cell extracts and with pure protein in solution (6–10). Biophysical assays show that, at moderate concentration (2 mg/ml or less), the nonphosphorylated full-length STAT1 or STAT1 (1–683) used in the crystal structure study were mainly dimeric in solution (9). Mao et al. (3) suggest that the two dimeric structures (parallel and antiparallel) containing the NDs might be in equilibrium in solution. Murphy and colleagues (10) showed that mutations in the N-terminal interface (4) prevent dimerization of full-length nonphosphorylated STATs and that N-terminal STAT4 mutants are not tyrosine phosphorylated in response to cytokines, providing the first indication of any physiologic function of nonphosphorylated dimers.

From our examination of mutations in the ND or CC/DBD that disrupt either of the possible nonphosphorylated STAT1 intermolecular interactions seen in the crystal structures, we uncovered an unusual phenotype. Such mutations greatly decreased nonphosphorylated STAT1 dimer formation in solution, but the mutant protein became phosphorylated. More striking, the mutants exhibited a persistent phosphatase resistance both in vivo and in vitro. It is the phosphorylated, parallel wild-type molecule that accumulates in the nucleus to bind DNA and must be dephosphorylated to return to the cytoplasm. However, because of the persistent phosphorylation phenotype of the CC/DBD and ND mutants, we propose that, in the nucleus, a dimer rearrangement (parallel to antiparallel) is required for efficient presentation of STAT1 pY to phosphatases, allowing dephosphorylation to complete the activation–inactivation cycle.

Materials and Methods

Cell Culture, Antibodies, and Plasmids. U3A cells and derivatives and 293T cells were maintained in DMEM (GIBCO) supplemented with 10% bovine calf serum (Cosmic calf serum, Hy-Clone) and antimicrobials (GIBCO) at 37°C and 5% CO₂. Stable transfectants were maintained in 1 mg of G418 per ml. Transient transfections were performed with Lipofectamine reagent (Invitrogen) by using 2 μg of total DNA for each 60-mm plate. IFN-γ was at a final concentration of 5 ng/ml, in the case of transient transfections, usually 24 h after the introduction of DNA. Staurosporine (Sigma) was dissolved in DMSO and used at a final concentration of 500 nM.

STAT1 and Myc antibodies were from Santa Cruz Biotechnology. Activated STAT1 was detected with a polyclonal antibody raised to a peptide containing pY 701 (Cell Signaling Technology, Beverly, MA). The FLAG epitope was detected with M2 monoclonal antibody (Sigma).

All mammalian expression constructs were derived from pRc/CMV (Invitrogen). The point mutations (F172W, F77A, L78A, F77A/L78A, F172W/F77A/L78A, and N460A) were prepared by using a QuikChange site-directed mutagenesis kit (Stratagene) and verified by sequencing. The F77A/L78A/N460A mutation was derived by swapping the F77/L78A NcoI/BssSI fragment with the corresponding N460A. The Flag- and Myc-tags were added to the C-terminal of STAT1α by PCR.

EMSA: Cell Extracts. The same amount of fractionated- and whole-cell extracts were assayed for DNA binding activity with ³²P-labeled M67 probe (5′-GATCGATTTCCCGTAAATCATGATC-3′). GAS site is underlined, and italics indicates artifical sequence. DNA–protein complexes were then resolved on a 4% (29:1) polyacrylamide gel at 4°C (400 V, 0.25× Tris-borate-EDTA).

Cell Extracts, Immunoprecipitations, and SDS/PAGE. Cytoplasmic and nuclear extracts were prepared as described (11). After washing with cold PBS, cells were lysed at 4°C by incubating 5 min in hypotonic buffer (20 mM Hepes, pH 7.9/10 mM KCl/0.1 mM Na₃VO₄/1 mM EDTA/10% glycerol/0.5 mM PMSF/1 μg/ml aprotinin/1 μg/ml pepstatin/1 μg/ml leupeptin/1 mM DTT) with 0.2% Nonidet P-40. After centrifugation at 4°C (16,100 × g in microfuge) for 10 s, supernatants were collected as cytoplasmic extracts. Nuclear extracts were prepared by resuspension of the crude nuclei in high-salt buffer (hypotonic buffer with 20% glycerol and 420 mM NaCl) at 4°C for 30 min, and the supernatants were collected after centrifugation at 4°C (13,000 rpm) for 5 min. Whole-cell extracts were prepared in lysis buffer (150 mM Tris, pH 7.5/150 mM KCL/1 mM EDTA/0.1% Triton X-100/protease inhibitor cocktails).

Immunoprecipitations were carried out by adding 3 μg of anti-Flag or anti-Myc Ab to each extract and incubating overnight at 4°C followed by incubation at 4°C with protein A/G-Sepharose for 2 h. Samples were washed three times with whole-cell extract buffer and twice with PBS followed by resuspension in 2× Laemmli running buffer. Samples then were heated at 95°C for 4 min and subjected to SDS/PAGE on a 6% gel.

In Vitro Phosphatase Assays. Reactions were performed as described (11). Purified tyrosine phosphorylated wild-type and mutant STAT1 (50 ng) were incubated with various amounts of GST-TC45 (1, 3, 10, and 30 ng) for 60 min at 30°C.

Results

Mutants Directed by Crystallographic Results. Based on newly available crystallographic structure information (3), three mutations were designed to interrupt the apparent dimer interfaces of nonphosphorylated STAT1 (Fig. 1 A). The interface in the dimer between the CC and DBD (diagrammed in Fig. 1B) shows a reciprocal interaction between the F172 in the CC and a surface pocket in the DBD and therefore the F172W mutation was introduced. As we have described (4), the wild-type STAT1 ND is dimeric in solution whereas either F77A or L78A mutant protein is a monomer in solution. We introduced, individually or in combination, all three of these mutations into a truncated STAT1 core (130–683), a longer near full-length STAT1 version (1–683) that includes all of the SH2 domain, and also into full-length STAT1. To examine the intracellular properties of various STAT proteins, experiments have been carried out with purified full-length protein bearing these mutations and also with U3A cells that lack STAT1 but have had either wild-type or mutant STAT1 proteins introduced either by acute transfection or in permanently transfected lines.

Existence of, and Mutational Interruption of, Nonphosphorylated STAT1 Dimers in U3A Cells. To assay for interaction between nonphosphorylated STAT monomers, vectors encoding epitope tagged full-length STAT1 (either Flag or Myc) were introduced alone or together into 293T cells (Fig. 1C). In extracts of cells with a single tagged protein, each specific antibody precipitated only the expected tagged protein. However, in cells expressing both wild-type tagged proteins (in approximately equal amounts), antibody precipitation with either specific antibody precipitated the opposite epitope as well, providing evidence of interaction, most likely dimeric between nonphosphorylated monomers (Fig. 1C Left). When the coprecipitation assay was carried out with cells expressing the epitope-tagged mutant proteins F172W and F77A/L78A, or proteins with all three mutations (Fig. 1C Right), there was little or no coprecipitation. [Single mutations F77A or L78A also failed to show coprecipitation in this assay (data not shown).] Thus, mutations in either the CCD or the ND caused disruption of the nonphosphorylated dimers.

Time Course of Tyrosine Phosphorylation and Dephosphorylation in STAT1 and Mutant Proteins. Our first idea upon seeing the antiparallel, dimeric, nonphosphorylated core that presents SH2 groups 50–60 Å apart was that such a structure might facilitate binding of two STAT molecules to the dimeric IFN-γ receptor complex at the cell surface (12). This structure would provide an optimum chance for two interacting STAT1 molecules to be tyrosine phosphorylated simultaneously and then to interact to form the pY-SH2-mediated dimer. Thus, a mutation that decreased (or prevented) nonphosphorylated dimers might not be phosphorylated as readily as wild type. However, such was not the case for STAT1. IFN-γ treatment caused approximately equally rapid phosphorylation as assayed by DNA binding (EMSA) after 2, 4, 8, and 15 min in U3A cell lines complemented with wild-type or with any of the mutant proteins (Fig. 2A; similar results were found with the double mutant F77A/L78A and with the triple mutant F77A/L78A/F172W, data not shown).

Fig. 2.

Effects of various STAT1 mutations on IFN-γ activation and subsequent inactivation. (A) Time course (min) of IFN-γ activation of cells expressing STAT1α or mutants assayed by EMSA. (B) Time course (hr) of activation of STAT1 and mutants assayed by EMSA. (C) Persistent tyrosine phosphorylation of mutant STAT1. U3A cells were transiently transfected (or not) with the indicated expression vectors and treated (or not) with IFN-γ for 0.5 or 6 h. Western blotting was with anti-pY STAT1 antibody or STAT1 antibody. (D) Failure to bind DNA, accomplished with the N460A mutation, does not affect the prolonged phosphorylation phenotype. Reconstituted U3A cells were treated with IFN-γ for 0.5 or 5 h. A portion of the extracts was subjected to EMSA (Left) whereas the remainder was analyzed for tyrosine phosphorylation by Western blotting with anti-pY STAT1 antibody (Right). Anti-STAT1 Western blots show the total expressed protein levels.

Murphy and colleagues (10) recently reported that mutations in the ND that prevented formation of nonphosphorylated dimers of STAT4 completely prevented phosphorylation of STAT4 in response to either IL-12 or IFN-α. However, they also reported, in agreement with Fig. 2 A (supplemental figure 2 in ref. 10), that mutations designed to disrupt STAT1 N-terminal interactions do not block STAT1 tyrosine phosphorylation. Moreover, STAT1 lacking an ND can be phosphorylated (13) and in fact remains phosphorylated longer than wild-type protein (14). Therefore, there is a distinct difference in STAT1 and STAT4 in their interactions with, respectively, the IFN-γ and IL-12 or IFN-α receptors and associated kinases.

After these short-term phosphorylation experiments, we also assayed STAT1 activation in the nucleus over longer times, both by EMSA in cell extracts and by Western blotting with anti-pY antibodies (Fig. 2 B and C). In contrast to wild-type protein, all of the mutations (F172W, L78A, F77A, or multiple mutants) that affect nonphosphorylated dimer formation led to persistent DNA binding and to persistent tyrosine phosphorylation. Although activation of wild-type STAT1 by IFN-γ peaks within 30–60 min (Fig. 2B and ref. 15) and declines greatly by 2 h, the mutant proteins remained activated for up to 6 h (Fig. 2 B and C).

Persistent Phosphorylation of F172W Mutant Does Not Require DNA Binding. Because DNA binding might protect against dephosphorylation (16), we introduced mutations in a crucial DNA contact residue N460 (2, 17, 18) either by itself or with F77A/L78A to block DNA binding and introduced these into U3A cells. Like the wild-type protein, the N460A mutant was phosphorylated after 30 min of IFN-γ treatment, but the N460A protein did not bind DNA (Fig. 2D Left and Right, fifth and sixth lanes). The triple mutant F77A/L78A/N460A was also phosphorylated at 30 min and did not bind DNA. However, this latter non-DNA-binding mutant resisted dephosphorylation for 5 h (Fig. 2D Right, seventh and eighth lanes). Thus, the resistance to dephosphorylation brought about by the F77A/L78A mutation is independent of DNA binding.

Persistent Phosphorylation Is Based on Phosphatase Resistance in Vivo and in Vitro. The rates of in vivo STAT dephosphorylation have frequently been studied by stopping further phosphorylation by the kinase inhibitor staurosporine after brief IFN-γ treatment, and then after the disappearance of tyrosine phosphorylated STAT1 (11, 14, 15). This assay (Fig. 3A) showed that, in the absence of further phosphorylation caused by staurosporine, the dephosphorylation of wild-type STAT1 (tested by DNA binding) was completely gone in 30 min whereas the phosphorylation of F172W, L78A, and F77A mutants lasted for hours, supporting the conclusion that blocked dephosphorylation is the basis for the prolonged tyrosine phosphorylation with mutant protein.

Fig. 3.

Dephosphorylation resistance of the STAT1 mutants. (A) Effect of the tyrosine kinase inhibitor staurosporine. After a 30-min induction with IFN-γ, U3A cells containing indicated STAT1 protein were treated as indicated with staurosporine and extracts were subjected to EMSA or pY Western blotting. The F172W (B) and F77A/L78A (C) mutants resist in vitro dephosphorylation. Purified tyrosine-phosphorylated proteins were incubated with increasing amounts of GST-TC45, and tyrosine phosphorylation was assayed by Western blotting with anti-pY STAT1 antibody. The anti-STAT1 Western blots show the total expressed protein levels.

STAT1 is not dephosphorylated normally in cells lacking TC45, a tyrosine phosphatase (11), and this enzyme can dephosphorylate purified STAT1 in vitro (11, 16). We tested the sensitivity to TC45 dephosphorylation of purified phosphorylated wild-type and F172W and F77A/L78A mutant STAT1 proteins. There was little or no dephosphorylation of F172W or of the double mutant F77A/L78A compared with wild type in accord with the slower in vivo dephosphorylation of these mutants (Fig. 3 B and C). Thus, whatever the cause for the persistent phosphorylation phenotype, the mutant molecules resist dephosphorylation both in vivo and in vitro.

Discussion

The results described here, plus the new crystallographic structure of nonphosphorylated STAT1 (3), present two major findings that we wish to relate: (i) an antiparallel dimeric structure is evident in nonphosphorylated STAT1, and the dimeric association depends both on ND/ND and CCD/DBD interfaces; and (ii) mutations in either or both of these interfaces result in abnormally persistent phosphorylation in vivo and phosphatase resistance in vitro.

We earlier showed that the STAT1 phosphodimer by itself is very stable in solution (19) whereas dephosphorylation of wild-type STAT1 in vivo is quite rapid (e.g., see Fig. 2B and ref. 15). How the two reciprocal pY-SH2 bonds of phosphorylated STAT1 are broken to allow dephosphorylation is unknown. To relate the two central observations above, we propose that a molecular rearrangement occurs (parallel to antiparallel) that facilitates dephosphorylation.

As is the case with all transcription factors, dissociation from DNA occurs with STAT1 (20). Fig. 4 considers events that might occur to a tyrosine-phosphorylated STAT1 dimer after dissociation from DNA on the way to dephosphorylation in light of the persistent phosphorylation phenotype of the F172W, F77A, and L78A mutants. The simplest possibility would be that the parallel phosphodimer held together only by pY-SH2 interaction spontaneously dissociates and yields monomers as the target of a phosphatase. Because the STAT1 phosphodimer is very stable in vitro (19) and because F172W or F77A/L78A mutations should have no obvious effect on spontaneous dissociation and subsequent dephosphorylation, and yet these mutations prevent dephosphorylation, we discard this scenario. A second possibility is that the parallel phosphodimer not bound to DNA might be flexible [this is the case with the Dictyostelium STAT (21)], allowing a configuration where each ND on its flexible 24-aa tether (2) could come in contact and interact. [There is no interaction between the ND of one monomer and its partner when the phosphodimer is bound to DNA because of the angle and separation imposed by the DNA (Fig. 1B and ref. 2).] Such an ND/ND interaction in the phosphodimer off DNA seems very reasonable because (i) mutations show that the ND/ND interaction is required for dimeric, nonphosphorylated STAT1 molecules in solution and (ii) the nonphosphorylated crystal structure (3) includes the possibility of ND/ND interaction, with both a parallel and an antiparallel dimer. With the ND/ND dimer lodged between the bodies of the two monomers of a parallel phosphodimer, perhaps a stress develops that facilitates breakage of pY-SH2 interaction, leading to phosphatase access and dephosphorylation. In this scenario, the F77A/L78A mutation, disrupting the ND/ND interaction, would lead to persistent phosphorylation because pY/SH2 dissociation would no longer be facilitated, implying that this possibility is reasonable. But in a continuing parallel phosphoSTAT1 structure on the way to phosphorylation, there would be no effect of F172W on dephosphorylation because there is no CC/DBD interaction possible in the parallel alignment.

A final possibility, the most radical but a possibility that satisfies all of the data, begins with the unbound parallel phosphodimer off DNA allowing the ND/ND interaction. Whether the ND/ND interaction helps to force apart pY/SH2 or whether spontaneous pY/SH2 dissociation might occasionally occur, the ND/ND interaction [which has a K_D of 6.4 μM (3)] could hold the dimer together. Rotation of the bodies of the monomers from the parallel orientation to the antiparallel orientation would then allow the crystallographically confirmed, reciprocal CC/DBD interaction to occur. Such an ND/ND interaction followed by a rotation seems feasible because the NDs are connected to the body of the STAT molecule by the above mentioned, unstructured, flexible 24-aa tether. [Because of the configuration of the CC, this tether emerges near the center of STAT1 structure (diagrammed in Fig. 1B; see ref. 2.] Stabilized by two interfaces (ND/ND and CC/DBD), this antiparallel phosphodimer with exposed pY at each end would form an easy target for a phosphatase; and this mechanism is consistent with mutations in two different regions, CC/DBD and ND/ND, both inhibiting dimerization of nonphosphorylated monomers and both causing the persistent phosphorylation phenotype (in vivo and in vitro). The antiparallel, dephosphorylated STAT1 dimer might then exit to the cytoplasm. In the antiparallel configuration, the nuclear exit signal, residues 302–314 (22), would be exposed, and the nuclear localization signal (K410, K413) would be at least partially hidden (16, 23–27). In the proposed model, the two partners of a phosphodimer should remain together after dephosphorylation.

Could the mutations discussed simply result in blocking a phosphatase binding site(s) and then negate the necessity of the large conformational change we propose? Although such a possibility might exist, we regard it as quite unlikely; the mutations discussed are in distinctly different domains and are very distant from the phosphorylation site on this rigid molecule. We, of course, cannot rule out the participation of other proteins in the cell, but the dephosphorylation in vitro of wild-type protein but not mutant protein certainly allows for the possibility that only the STAT and phosphatase are involved.

We note as a general point that molecular details on the removal of a phosphate from tyrosine in which SH2 interaction has occurred is generally poorly understood. In the case of activation of src-like tyrosine kinase, removal of an inhibitory intramolecular pY-SH2, residue 527, requires a displacement by a pY on a phosphatase (28, 29) so that the inhibitory pY can be removed. Thus, the radical molecular reorientation suggested for STAT1 would accord with the need for structural changes to remove a pY phosphate, a necessary step in regulation of this transcription factor. The importance of regulated pY removal from STATs is emphasized by the oncogenic nature of persistently phosphorylated STAT3 (1, 30).