Tyrosine phosphorylation regulates the partitioning of STAT1 between different dimer conformations

Abstract

The activation/inactivation cycle of STAT transcription factors entails their transition between different dimer conformations. Unphosphorylated STATs can dimerize in an antiparallel conformation via extended interfaces of the globular N-domains, whereas STAT activation triggers a parallel dimer conformation with mutual phosphortyrosine:SH2 domain interactions, resulting in DNA-binding and nuclear retention. However, despite the crucial role of STAT tyrosine phosphorylation in cytokine signaling, it has not been determined how this modification affects the stability and the conformational flexibility of STAT dimers. Here, we use analytical ultracentrifugation and electrophoretic mobility shift assay (EMSA) to study the association of STAT1 in solution before and after tyrosine phosphorylation. It is revealed that STAT1 formed high-affinity dimers (K_d of ≈50 nM) with estimated half-lives of 20–40 min irrespective of the phosphorylation status. Our results demonstrate that parallel and antiparallel conformations of STAT1 were present simultaneously, supported by mutually exclusive interfaces; and the transition between conformations occurred through affinity-driven dissociation/association reactions. Therefore, tyrosine phosphorylation was dispensable for DNA binding, but the phosphorylation enforced preformed SH2 domain-mediated dimers, thus enhancing the DNA-binding activity of STAT1 >200-fold. Moreover, upon STAT1 activation the N-domains adopted an open conformation and engaged in interdimer interactions, as demonstrated by their participation in tetramerization instead of dimerization. Yet, homotypic N-domain interactions are not conserved in the STAT family, because the N-domain dissociation constants of STAT1, STAT3, and STAT4 differed by more than three orders of magnitude. In conclusion, STAT1 constantly oscillated between different dimer conformations, whereby the abundance of the dimerization interfaces was determined by tyrosine phosphorylation.

Interferons and other cytokines are extracellular signaling molecules that exert their effects after binding to specific cell surface expressed receptors (1). The binding of IFN-γ activates receptor-associated Jak kinases, which catalyze a series of tyrosine phosphorylation reactions leading to the activation of the transcription factor STAT1. The STATs, of which seven different genes are expressed in mammals, all consist of three independent structural subunits (2, 3). At the N terminus there is a proteolytically separable N-domain of 130 residues, followed by a large contiguous core fragment of almost 600 residues containing a coiled-coil domain, a DNA-binding domain, and an SH2 domain. A transferable transactivation domain of variable length and sequence that frequently is subject to alternative splicing is located at the C terminus (3). The STATs are best known for their cytokine-dependent transcriptional activities, and STAT activation was recognized as the central event in cytokine signaling. Structurally, STAT1 activation entails the phosphorylation of residue tyrosine⁷⁰¹ and the subsequent homo- or heterodimerization via reciprocal phosphor-Tyr:SH2 domain interactions (4). The dimerization event has important functional implications. It transforms the STATs into high-affinity DNA-binding transcriptional regulators and triggers their retention in the nucleus (3, 5, 6). Initially, the STATs were believed to be monomeric before activation. However, accumulating structural and functional evidence indicates the presence of dimers and higher-order aggregates for multiple unphosphorylated STATs (7, 8). Crystallographic analysis of STAT1 demonstrates the same monomer core structures before and after phosphorylation, yet an entirely different association of monomers (9, 10). Contrary to activated STAT1, the SH2 domains of unphosphorylated monomers do not participate in dimerization. Rather, the SH2 domains are situated at opposite ends of the dimers, a conformation called antiparallel, as opposed to the parallel conformation of the DNA bound STATs. The crystals of unphosphorylated STAT1 contain dimers formed by reciprocal interactions of the conserved N-terminal domains, as well as by interactions between the coiled-coil domain and the DNA-binding domain (10). Mutations that prevent N-domain interactions not only reduce dimerization of unphosphorylated STAT1 but also abolish cooperative DNA binding of activated STAT1 and STAT5 (11). Moreover, homologous mutations preclude activation of STAT4 (8), whereas dephosphorylation of STAT1 and STAT5 is diminished (11, 12). It was thus proposed that the N-domains assist in interrupting the phosphor-Tyr:SH2 domain interactions and promote rotation of the core fragments to facilitate dephosphorylation by exposing the phosphorylated tyrosine (13).

The aforementioned studies provide considerable insight into the structures that STAT molecules may adopt before and after activation. However, the structural flexibility that became apparent raises important questions concerning the molecular dynamics associated with STAT activation, which have not been addressed to date. These pertain to the influence of tyrosine phosphorylation on dimer or multimer stability, to the role of the various interfaces in promoting or inhibiting monomer association, and to the interchange between different dimer conformations. Here, we analyze the multimerization of STAT1 and focus on the roles of N-domain and tyrosine phosphorylation. Our data demonstrate that Tyr phosphorylation determines the abundance of different STAT1 dimer conformations that transmute via reversible dissociation/association reactions, and we propose that this dynamic process may serve as a paradigm for STAT activation and inactivation.

Results and Discussion

For these studies, we produced full-length human STAT1α and the alternative splice variant STAT1β, which lacks the C-terminal transcription activation domain of 38 residues. Moreover, two experimental deletion mutants were included that lack the N-domain either alone or in combination with the transactivation domain, termed STAT1ΔN or STAT1tc, respectively. Of the latter, additionally the three point mutants Phe¹⁷²Trp and Tyr⁷⁰¹Ala/Arg were studied. These proteins were well expressed in baculovirus-infected insect cells or bacteria and purified by virtue of a C-terminal Strep-tag. Proteins were alkylated to prevent cystine oxidation and denaturation, a procedure that does not adversely affect the SH2 domain and the DNA binding of STAT1 (9, 14). Proteins intended for analyses of the phosphorylated state were subjected to in vitro tyrosine phosphorylation with immunopurified EGF receptor, followed by heparin agarose chromatography to remove unphosphorylated material. This treatment consistently delivers highly pure STAT1 preparations exclusively phosphorylated on tyrosine⁷⁰¹ (14). The presence or absence of tyrosine phosphorylation was confirmed by Western blotting with a Tyr⁷⁰¹ phosphospecific antibody and mass spectrometry (data not shown). In addition, the isolated N-domains of human STAT1, STAT3, and STAT4 were purified similarly by Strep-affinity chromatography. Before use all proteins were further purified by gel filtration. A schematic overview of the proteins used in this study is shown in Fig. 1[see also supporting information (SI) Fig. S1].

Fig. 1.

Schematic representation of the proteins included in this study. Shown are the STAT1 domain structure (Upper) and the STAT variant proteins used herein (Lower). Given are the chemical molecular mass and the domain composition. Absence and presence of tyrosine⁷⁰¹ phosphorylation are indicated (± PY701). TAD, transactivation domain.

Unphosphorylated STAT1 Forms High-Affinity Dimers.

Sedimentation experiments were done for each protein at specific concentrations varying between 1 μg/ml and 3 mg/ml (10 nM to 80 μM). We first performed sedimentation velocity experiments to determine the distribution of sedimentation coefficients (data not shown). Next, we performed sedimentation equilibrium studies, and the results are summarized in Table S1. For the unphosphorylated STAT1α, STAT1β, and the N-domain deletion mutants, it was found that they exist in a monomer:dimer equilibrium (Fig. 2A). Dimerization was strongly dependent on the N-domain because its deletion significantly increased the dissociation constant ≈100-fold from ≈50 nM to 3–4 μM. These data are in agreement with a previous ultracentrifugation study that also found a crucial role for the N-domain in stabilizing unphosphorylated STAT1 dimers (10). To determine how the dissociation constant of the purified proteins relates to the concentration of endogenous STAT1, we calibrated a STAT1 antibody with varying concentrations of purified STAT1α to assess the STAT1 concentration of HeLa cell extracts. The cellular STAT1 concentration was determined to be 40 ± 7 nM, or ≈1 × 10⁵ molecules (Fig. S2A). Although caution must be exerted when translating biochemical equilibrium constants to cell activity, the affinities and concentrations are such that both monomers and dimers are likely to be present in the living cells.

Fig. 2.

Sedimentation equilibrium analyses reveal identical dimerization of STAT1 before and after tyrosine⁷⁰¹ phosphorylation, but different oligomeric states. Partial dimer concentrations are plotted against total concentrations. The K_d values were calculated with fixed molecular mass (see also Table S1). (A) The results for unphosphorylated STAT1 demonstrate that the dimerization of N-domain deletion mutants is strongly reduced. (B) Comparison of the phosphorylated and unphosphorylated states of STAT1α and STAT1β. Note that the dimerization is indistinguishable before and after phosphorylation for both variants. The dotted line denotes the partial tetramer concentration for phosphorylated STAT1α. (C) The results for phosphorylated STAT1 show that the N-domain is dispensable for dimerization, but it is required for tetramerization.

STAT1 Dimers Are Equally Stable Before and After Activation.

We then examined the effects of Tyr⁷⁰¹ phosphorylation on the dimerization and association of STAT1. Strikingly, the dimer stability remained unchanged (Fig. 2B). The activation of STAT1α and STAT1β resulted in dissociation constants of 30–40 nM, which constitutes a negligible and statistically insignificant change compared with the monomer association (K_d of 50 nM) before tyrosine phosphorylation. The dimerization constant of activated full-length STAT1 determined here agrees remarkably well with phosphopeptide:STAT1-binding constants previously determined by fluorescence polarization and surface plasmon resonance (15, 16), demonstrating that in solution the attraction between activated STAT1 monomers to a large extent results from docking of the phosphorylated Tyr⁷⁰¹ into the SH2 domain binding pocket. We determined that ≈35% of the STAT1 protein in HeLa cells is activated in response to IFN-γ (Fig. S2B). Given the unchanged dissociation constants before and after activation, cytokine stimulation of cells therefore is not expected to alter the concentration of STAT1 dimers. Nonetheless, the activation reaction markedly affected the conformation of STAT1 dimers. Foremost, and opposite to the unphosphorylated protein, the association of Tyr-phosphorylated STAT1 monomers was not adversely affected by deletion of the N-domains (Fig. 2C). This result agrees with crystallographic data obtained with phosphorylated STATs in the absence or presence of DNA, which revealed that the N-domains are too far apart to make contacts (9, 17). Thus, both in the crystals and in solution, phosphorylated STAT1 appeared to adopt a dimer conformation that depended solely on SH2 domain interactions—irrespective of the presence of DNA.

The spatial reorganization of dimers is pivotal for STAT1 activity (12). To relate the STAT1 self-association to the time course of cytokine activation, we estimated the kinetics of dimerization. Although the thermodynamic equilibrium constants reported here do not carry kinetic information, they can be used in conjunction with hydrodynamic modeling of the parallel and antiparallel conformations to assess kinetic rate constants (see SI Materials and Methods). Two different methods were used to estimate the kinetic parameters, which gave similar results, indicating that the half-life of both dimer conformations ranges between 20 and 40 min (Table S2). These values are consistent with the overall time course of cytokine activation, which usually lasts up to a few hours, and with the ≈20-min turnover time of activated STAT1 (3).

An important consequence of the switch to an SH2 domain-mediated dimer conformation was the release of N-domains from their intradimer bondage. This became obvious by the occurrence of high-affinity tetramers that required the interaction of N-domains (Fig. 2C). The sedimentation of activated STAT1α adhered best to a model that invokes the presence of monomers, dimers, and tetramers, whereby the monomer:dimer equilibrium was not different from the other STAT1 phosphodimer variants. Of note, the monomer:dimer and the dimer:tetramer dissociation constants are virtually identical. This indicated that the attraction between phosphor-Tyr⁷⁰¹ and SH2 domain in the dimer were comparable with the forces between dimers in the tetramer. It is long known that dimers of activated STAT1 can interact cooperatively on neighboring GAS sites via their N-domains, resulting in strongly reduced dissociation of the tetramers from DNA (14). Because of the translational and rotational offset between adjacent STAT1 dimers on DNA, it is impossible for a DNA-bound tetramer that all four N-domains dimerize simultaneously. This, however, allows the N-domain-mediated recruitment of more STAT dimers, as was previously observed (18). Here, we show that STAT1 formed very stable tetramers in the absence of DNA without evidence of further oligomerization, which indicated that the association capacity of the N-domains was saturated. Therefore, STAT1 tetramerization in the presence or absence of DNA is unlikely to be identical. Interestingly, tetramerization was more prominent for STAT1α than for the C-terminally truncated STAT1β, pointing out a contribution of the transactivation domain to the self-association of STATs. This outcome was confirmed by chemical cross-linking of highly dilute STAT1 protein (Fig. S3). Together, these results demonstrated that tyrosine phosphorylation did not affect the dimer stability. Rather, activated STAT1 associated in a different dimer conformation, in which the N-domains adopted an open conformation conducive to tetramerization.

Dimerization of N-Domains Is Not Conserved in the STAT Family.

Given the prominent role of homotypic N-domain interactions for dimerization and tetramerization, we explored the N-domain association of STAT1, STAT3, and STAT4. Their N-domains are structurally highly conserved (10, 19), they were shown to homodimerize (8), and they have all been demonstrated to promote tetramerization of the respective full-length molecules on DNA (14, 18, 20). Despite these similarities, our sedimentation equilibrium experiments with the isolated N-domains uncovered significant differences (Fig. 3). The N-domains both of STAT1 and STAT4 formed dimers in the low micromolar range, with a slightly enhanced dimerization constant for STAT4. It should be pointed out that these dimerization constants are several hundred-fold higher than for the full-length STAT1 (Fig. 2A), demonstrating considerable additional attracting forces between STAT1 monomers, as denoted by the crystal structure (10). Nevertheless, the N-domain dimers of STAT1 and STAT4 are of considerable affinity when compared with well known dimerization modules such as glutathione-S-transferase (K_d of ≈0.4 μM) (22). In clear contrast, appreciable association of the N-domains of STAT3 required millimolar amounts of protein, demonstrating a significant shift of the dissociation equilibrium to the monomeric state; a conclusion that was confirmed by analytical gel filtration (Fig. S4). These unexpected quantitative differences indicate that high-affinity homotypic interactions of the N-domains are not conserved in the STAT family. In addition, this considerable functional diversity among the N-domains possibly explains the proposed sequence-selective DNA recognition afforded by STAT multimerization (18). Moreover, weak N-domain interactions are also expected to compromise the ability of unphosphorylated STAT3 to form stable dimers. A previous ultracentrifugation study found no differences in this regard between STAT1 and STAT3 (22), suggesting that other regions of the STAT3 molecule may compensate for the low-affinity N-domain interactions. We note, however, that Braunstein et al. failed to detect monomers of unphosphorylated STAT1 and STAT3, which is not in accordance with results for STAT1 reported here and elsewhere (10).

Fig. 3.

Sedimentation equilibrium analyses demonstrate the differential dimerization of the isolated N-domains of STAT1, -3, and -4. Partial dimer concentrations are plotted against total concentrations. The listed K_d values were calculated with fixed molecular mass (see also Table S1). Note that STAT3 N-domains contain the least dimer.

Tyrosine Phosphorylation Regulates the Abundance of Different Dimer Conformations.

The ultracentrifugation results indicated that STAT1 was present in an antiparallel conformation before activation, whereas a parallel dimer was present thereafter. However, it remained unclear whether these conformations are linked exclusively to the respective activation states, or whether multiple conformations can coexist. To answer this question, we turned to DNA binding—one of the best studied consequences of STAT activation—as an indicator for dimers in the parallel conformation (9, 23). The specific DNA binding of unphosphorylated STAT1 was thus examined to explore whether STAT1 adopted a parallel conformation already before activation. Shown in Fig. 4A is a titration of increasing amounts of Tyr⁷⁰¹-phosphorylated or unphosphorylated STAT1α with a limiting concentration (below the K_d of 1 nM) of the M67 high-affinity STAT1-binding site (14). Phosphorylated STAT1 (lanes 1 and 2) formed the expected complexes of dimeric protein with DNA, as well as a minor fraction of slower migrating tetrameric protein:DNA complexes. Remarkably, the unphosphorylated STAT1 (lanes 3–9) formed well defined bands corresponding to activated STAT1 dimers with electrophoretic migrations that were identical at all protein concentrations. The DNA-binding activity of unphosphorylated STAT1 was very weak, however, because even a 200-fold elevated concentration of unphosphorylated protein did not match the binding activity of the phosphorylated material. Nonetheless, it appeared that DNA binding required dimeric STAT1, regardless of the phosphorylation status. Thus, the question arose of whether the prevalent conformation of unphosphorylated STAT1—antiparallel—contributed to the observed DNA-binding activity. This may occur either directly by providing an alternate DNA-binding surface or indirectly by facilitating the transition to the conventional parallel conformation of DNA-bound STAT1. Alternatively, the antiparallel dimer may be irrelevant for DNA binding, and the dimer conversion results from the spontaneous dissociation and subsequent re-association in the alternate orientation. According to the former model that DNA binding necessitates the pre-association of unphosphorylated STAT1, any weakening of the antiparallel conformation would be expected to impair DNA binding. To explore this possibility, we cancelled the two single most important interactions that afford the antiparallel dimer. At first, the N-domain was deleted; yet, the DNA-binding activity was not reduced (Fig. 4B, compare lanes 1–3 and 7–9), an outcome previously reported for activated STAT1, too (14). On the contrary, DNA binding of N-terminally truncated STAT1tc was increased by ≈4-fold. Moreover, further destabilization of the antiparallel dimer by the additional mutation of residue Phe¹⁷² to tryptophan (10) also was without adverse effects on DNA binding (compare lanes 1–3 and 4–6). We thus concluded that the antiparallel dimer conformation did not promote, but rather suppressed, the DNA binding of unphosphorylated STAT1. These data therefore do not support a role for the N-domain in facilitating the reversible transition between the antiparallel and parallel dimers.

Fig. 4.

DNA-binding properties indicate identical dimer conformations of STAT1 before and after Tyr⁷⁰¹ phosphorylation. Shown are EMSA analyses (representative of two to three experiments); the positions of dimeric (D) and tetrameric (T) STAT1/DNA complexes are indicated as well as the unbound probe (free). Radioactively labeled DNA contained a single high-affinity M67 STAT1 binding site and was used at a concentration of 0.1 nM. (A) Phosphorylated and unphosphorylated STAT1α were assayed at the indicated varying concentrations. (B) DNA binding of nonphosphorylated full-length STAT1 (1α), N- and C-terminally truncated STAT1 (1tc), and truncated STAT1 with mutation Phe¹⁷²Trp (1tcF172W). The different proteins were assayed at identical concentrations. (C) Inactivation of the STAT1 SH2 domain by a nonpeptidic small molecule. Phosphorylated and unphosphorylated STAT1tc were incubated with increasing concentrations of inhibitor (6-NBT), before DNA was added. Note that unphosphorylated STAT1 was used at 100-fold higher concentration. Quantitative data are means ± SEM (Lower). (D) Influence of residue Tyr⁷⁰¹ on the DNA binding of unphosphorylated STAT1. Shown are the results for wild type (WT), and the alanine⁷⁰¹- (Y701A) and arginine⁷⁰¹ mutants (Y701R) of truncated STAT1. (E) Competition binding assays for probing the DNA contacts of unphosphorylated (unp1tc, 200 nM) and phosphorylated (p1tc, 1 nM) STAT1 dimers. Comparable DNA-binding activities of STAT1tc were equilibrated with labeled M67 probe. Subsequently, a 100-fold molar excess of unlabeled M67 or nonspecific (n.s.) probe was added where denoted, before EMSA was performed. (F) Salt sensitivity of STAT1 DNA binding. Comparable DNA-binding activities of unphosphorylated STAT1tc (unp1tc, 200 nM) or activated STAT1tc (p1tc, 0.5 nM) were equilibrated with labeled probe, before KCl was added. Shown are the EMSA (Upper) and the quantification of the binding intensities (Lower).

Next, we investigated whether the parallel dimer conformation accounted for the DNA-binding activity of unphosphorylated STAT1. At first, we used 6-NBT, a nonpeptidic small molecule that impairs the function of the SH2 domains specifically of STAT1 and STAT3. The chemical inhibits the association of STATs with physiologically relevant tyrosine-phosphorylated peptide motifs (24). In stark contrast to the results obtained after destabilizing the antiparallel dimer, this inhibitor of the parallel dimer conformation strongly reduced the DNA binding of STAT1 (Fig. 4C). Importantly, the inhibitory effect was indistinguishable for the phosphorylated (lanes 5–8) and unphosphorylated (lanes 1–4) states. To corroborate these results, we targeted the parallel dimer conformation by mutating residue tyrosine⁷⁰¹—the SH2 domain counterpart—to alanine or arginine. Both mutants, as well as the aforementioned Phe¹⁷²Trp mutant, were expressed as stable proteins and did not differ from wild-type STAT1tc in their self-association, as determined by analytical gel filtration and dynamic light scattering (Fig. S5 and Table S3). Notwithstanding this similarity, the DNA-binding activity of the unphosphorylated Tyr⁷⁰¹-mutant proteins was significantly reduced, because an ≈6-fold increased protein concentration was required to match the DNA-binding activity of Tyr⁷⁰¹ wild type (Fig. 4D). To further compare activated and unphosphorylated STAT1 dimers, we examined their interactions with DNA by adding an excess of specific or nonspecific competitor DNA (Fig. 4E). Remarkably, when similar DNA-binding activities were compared, there was no difference between unphosphorylated and phosphorylated STAT1. Both phosphorylation states were insensitive to competition by nonspecific DNA (lanes 3 and 6), whereas both dimer forms were readily displaced by an excess of specific competitor DNA (lanes 2 and 5). Moreover, we also did not note differences in the kinetics of replacement between the two variants, indicating that their dimerization was not differently affected by the presence of DNA (data not shown). Collectively, these results demonstrated that STAT1 dimers engage in identical DNA contacts both before and after tyrosine phosphorylation; and they indicated that unphosphorylated STAT1 can adopt the parallel dimer conformation, albeit at a small proportion only. Despite similar conformations, differential dimer stability became apparent when we tested for the salt sensitivity of the protein:DNA complexes. Although phosphorylated STAT1 was relatively insensitive to increased ionic strength, with >50% of the dimers remaining even at 800 mM salt, ≈70% of the unphosphorylated dimers were disrupted already in the presence of 200 mM salt (Fig. 4F).

Based on the available data, we propose that STAT1 coexists in two dimer conformations both before and after activation. The transition from antiparallel to parallel conformation is not facilitated by the pre-association of dimers; it is a spontaneous dissociation/re-association reaction whereby the tyrosine phosphorylation reaction determines the conformation abundance (Fig. 5). In the unphosphorylated state, N-domain interactions characterize the prevalent dimer species, indicative of an antiparallel conformation. At the same time, a small proportion of unphosphorylated STAT1 adopts the parallel conformation of activated STAT1 (Fig. 5 A and B). Because its dissociation constant exceeds 10 μM, unphosphorylated STAT1 is unlikely to have physiological significance as a DNA-binding protein independent of cofactors. Upon STAT1 activation, the interaction of Tyr⁷⁰¹ with the SH2 domain is enforced, as demonstrated by N-domain-independent dimerization and the at least 200-fold increased DNA-binding activity. Consequently, the weak parallel conformation of unphosphorylated STAT1 is turned into a high affinity interaction with a K_d of 50 nM (Fig. 5C). This, in turn, allows the N-domains to adopt an open conformation, which results in high-affinity tetramerization. Based on the low tetramer dissociation rate of ≈100 nM and the absence of higher-order oligomers, we conclude that in solution all four N-domains contribute to the STAT1 tetramer (Fig. 5C, conformation I), whereas DNA-bound tetramers need to adopt conformation II. For STAT1 dimers, our data suggest that N-domain and SH2 domain interfaces are mutually exclusive and thus do not cooperate. Given the presence of unphosphorylated STAT1 in both antiparallel and parallel conformations, as well as their identical dimerization constants before and after activation, respectively, we therefore infer that for activated STAT1 both conformations are equally stable with K_d of ≈50 nM (Fig. 5D).

Fig. 5.

Model for STAT1 oligomerization before (A and B) and after (C and D) tyrosine phosphorylation. N-terminal domains are shown in purple, SH2 domains are shown in yellow, and phosphorylation of Tyr⁷⁰¹ is shown in red. Dissociation constants for full-length (A and C) and N-terminally truncated (B and D) STAT1 are given. I and II represent two possible tetramer conformations. K_d (underlined) was determined by AUC. a, K_d was estimated from EMSA experiments; b, K_d was adopted from the dimerization of unphosphorylated STAT1; c, K_d exceeds the concentration range used in the AUC experiments.

Our data indicate significant differences in the dimerization of STATs, such as the influence of the N-domains. It will therefore be important to study further members of this protein family to decide the extent to which differential self-association contributes to the functional differentiation of STAT transcription factors in cytokine signaling.

Materials and Methods

Recombinant Proteins.

The cDNAs of human STAT1α, 1β, 1ΔN (amino acid 132–750), 1tc (amino acid 132–712), STAT1-N (amino acid 1–130), and murine STAT3-N (amino acid 1–130), all furnished at their 3′ end with sequence encoding a 15-residue linker and the 8-aa Strep-Tag II (IBA), were cloned into the baculovirus transfer vector pFastBac1 (Invitrogen) and expressed in baculovirus-infected Sf9 insect cells (14). The cDNA of human STAT4-N (amino acid 2–124) and additionally 1tc were cloned into vector pASK-IBA3plus (IBA) and expressed with a C-terminal Strep-Tag II in Escherichia coli BL21. Site-directed point mutagenesis was performed with the QuikChange kit (Stratagene). Mutations were verified by DNA sequencing (Geneservice). Purification of the Strep-tagged proteins was done according to a protocol provided by IBA. Except for the N-domains, proteins were alkylated, Tyr⁷⁰¹-phosphorylated, and purified by Heparin-affinity chromatography as described in ref. 14. Before use the proteins were gel-filtrated in PBS (pH 7.4) on a Superose 12 column (GE Healthcare).

Sedimentation Analyses.

A Beckman Optima XL-I analytical ultracentrifuge at the Leibniz-Institut für Molekulare Pharmakologie was used for sedimentation experiments. All experiments were performed in PBS (pH 7.4). For details on sedimentation velocity runs, see SI Materials and Methods. For sedimentation equilibrium experiments, samples were analyzed at 4°C at multiple rotor speeds corresponding to sigma values between 1 and 4 [defined as σ = M(1 − vρ)ω²/2RT] in six-channel epon/charcoal centerpieces in the AN-60-TI rotor. Scans were collected at equilibrium at three appropriate wavelengths for loading concentrations ranging between 0.1 and 0.9 absorption units in radial step mode with 0.001-cm step-size setting and 15-point averages. Data analyses were performed with a global nonlinear least-squares method, using UltraScan 9.1 (25). The data were fit to multiple reversible and nonreversible association models, and the most appropriate model to describe the data was selected based on a minimized variance and by visual inspection of residual run patterns. Ninety-five-percent confidence limits for the best-fit parameters were determined by Monte Carlo analysis. Hydrodynamic corrections for buffer conditions and estimations of partial specific volumes of all proteins from their peptide sequence were done with UltraScan.

Electrophoretic Mobility Shift Assay (EMSA).

EMSAs were performed with purified STAT1 variant proteins essentially as described in ref. 14, using the following duplex oligonucleotides (complementary strands and the 5′-ACGT overhangs on both strands for radioactive labeling by the Klenow reaction are not listed; binding site is boldface and italic): M67, 5′-CGACATTTCCCGTAAATCTG; nonspecific, 5′-CGACAACGCGCGTTCTTCTG. For DNA-competition assays, the EMSA reaction was equilibrated at room temperature for 15 min before adding a 100-fold molar excess of the respective unlabeled DNA. The salt sensitivity was determined in a fixed volume by addition of a concentrated KCl solution to the equilibrated DNA-binding reaction. 6-Nitrobenzo[b]thiophene-1,1-dioxide (6-NBT) (M_r 211.2) was dissolved in ethanol and added to the binding reaction for 45 min at 37°C before the addition of DNA. The solvent was present in all reactions at the final concentration of 1.7% (vol/vol). Binding activity was detected and quantified by phosphor-imaging.