A family-wide assessment of latent STAT transcription factor interactions reveals divergent dimer repertoires

The conversion of signal transducer and activator of transcription (STAT) proteins from latent to active transcription factors is central to cytokine signaling. Triggered by their signal-induced tyrosine phosphorylation, it is the assembly of a range of cytokine-specific STAT homo- and heterodimers that marks a key step in the transition of hitherto latent proteins to transcription activators. In contrast, the constitutive self-assembly of latent STATs and how it relates to the functioning of activated STATs is understood less well. To provide a more complete picture, we developed a co-localization-based assay and tested all 28 possible combinations of the seven unphosphorylated STAT (U-STAT) proteins in living cells. We identified five U-STAT homodimers—STAT1, STAT3, STAT4, STAT5A, and STAT5B—and two heterodimers—STAT1:STAT2 and STAT5A:STAT5B—and performed semi-quantitative assessments of the forces and characterizations of binding interfaces that support them. One STAT protein—STAT6—was found to be monomeric. This comprehensive analysis of latent STAT self-assembly lays bare considerable structural and functional diversity in the ways that link STAT dimerization before and after activation. The conversion of signal transducer and activator of transcription (STAT) proteins from latent to active transcription factors is central to cytokine signaling. Triggered by their signal-induced tyrosine phosphorylation, it is the assembly of a range of cytokine-specific STAT homo- and heterodimers that marks a key step in the transition of hitherto latent proteins to transcription activators. In contrast, the constitutive self-assembly of latent STATs and how it relates to the functioning of activated STATs is understood less well. To provide a more complete picture, we developed a co-localization-based assay and tested all 28 possible combinations of the seven unphosphorylated STAT (U-STAT) proteins in living cells. We identified five U-STAT homodimers—STAT1, STAT3, STAT4, STAT5A, and STAT5B—and two heterodimers—STAT1:STAT2 and STAT5A:STAT5B—and performed semi-quantitative assessments of the forces and characterizations of binding interfaces that support them. One STAT protein—STAT6—was found to be monomeric. This comprehensive analysis of latent STAT self-assembly lays bare considerable structural and functional diversity in the ways that link STAT dimerization before and after activation. The signal transducer and activator of transcription (STAT) proteins are an evolutionarily conserved family of seven transcription factors in mammals, namely, STAT1, STAT2, STAT3, STAT4, STAT5A, STAT5B, and STAT6 (1Levy D.E. Darnell Jr., J.E. STATs: transcriptional control and biological impact.Nat. Rev. Mol. Cell Biol. 2002; 3: 651-662Crossref PubMed Scopus (2526) Google Scholar). These proteins are the substrate of both receptor and non-receptor tyrosine kinases such as the Janus kinases (JAK), which catalyze their phosphorylation of a single C-terminal tyrosine residue after ligand binding to cell surface receptors (2Morris R. Kershaw N.J. Babon J.J. The molecular details of cytokine signaling via the JAK/STAT pathway.Protein Sci. 2018; 27: 1984-2009Crossref PubMed Scopus (324) Google Scholar). About 50 different growth factors and cytokines including interferons, interleukins, and growth hormones are known to signal via STAT proteins (3O'Shea J.J. Schwartz D.M. Villarino A.V. Gadina M. McInnes I.B. Laurence A. The JAK-STAT pathway: impact on human disease and therapeutic intervention.Annu. Rev. Med. 2015; 66: 311-328Crossref PubMed Scopus (882) Google Scholar). The activated STATs accumulate in the nucleus and participate in the transcription of hundreds of genes (4Philips R.L. Wang Y. Cheon H. Kanno Y. Gadina M. Sartorelli V. et al.The JAK-STAT pathway at 30: much learned, much more to do.Cell. 2022; 185: 3857-3876Abstract Full Text Full Text PDF PubMed Scopus (20) Google Scholar). This sequence of events, usually referred to as canonical JAK-STAT signaling, entails the dimerization of STAT proteins through mutual phosphotyrosine:SH2 domain interactions. Such dimers can bind short palindromic stretches of DNA called GAS elements in the promoter region of target genes (5Schindler C. Levy D.E. Decker T. JAK-STAT signaling: from interferons to cytokines.J. Biol. Chem. 2007; 282: 20059-20063Abstract Full Text Full Text PDF PubMed Scopus (973) Google Scholar). In addition to homodimers, activated STATs can assemble heterodimers with the other family members (6Delgoffe G.M. Vignali D.A. STAT heterodimers in immunity: a mixed message or a unique signal?.JAKSTAT. 2013; 2e23060PubMed Google Scholar). While the phosphotyrosine-mediated functions of STATs were linked to their dimerization early on (7Shuai K. Horvath C.M. Huang L.H. Qureshi S.A. Cowburn D. Darnell Jr., J.E. Interferon activation of the transcription factor Stat91 involves dimerization through SH2-phosphotyrosyl peptide interactions.Cell. 1994; 76: 821-828Abstract Full Text PDF PubMed Scopus (687) Google Scholar), much less is known to date regarding the constitutive self-assembly of latent, that is, unphosphorylated STATs (U-STATs). Initially, U-STATs were believed to be monomeric (7Shuai K. Horvath C.M. Huang L.H. Qureshi S.A. Cowburn D. Darnell Jr., J.E. Interferon activation of the transcription factor Stat91 involves dimerization through SH2-phosphotyrosyl peptide interactions.Cell. 1994; 76: 821-828Abstract Full Text PDF PubMed Scopus (687) Google Scholar), but further studies showed that STATs can assemble high-molecular-weight complexes already before tyrosine phosphorylation (8Droescher M. Vinkemeier U. Self-association of STAT proteins from monomers to paracrystals.in: Decker T. Müller M. Jak-Stat Signaling: From Basics to Disease. Springer, Vienna2012: 47-63Crossref Scopus (3) Google Scholar, 9Sehgal P.B. Paradigm shifts in the cell biology of STAT signaling.Semin. Cell Dev. Biol. 2008; 19: 329-340Crossref PubMed Scopus (106) Google Scholar). A well-documented example is STAT1, which forms equally strong dimers before and after activation as shown by analytical ultracentrifugation (10Wenta N. Strauss H. Meyer S. Vinkemeier U. Tyrosine phosphorylation regulates the partitioning of STAT1 between different dimer conformations.Proc. Natl. Acad. Sci. U. S. A. 2008; 105: 9238-9243Crossref PubMed Scopus (116) Google Scholar). Importantly, crystallographic evidence demonstrates that unphosphorylated dimers of STAT1, STAT3, and STAT5 are stabilized not by interactions of the carboxy-terminal SH2 domains but by interactions between amino-terminal regions (11La Sala G. Michiels C. Kükenshöner T. Brandstoetter T. Maurer B. Koide A. et al.Selective inhibition of STAT3 signaling using monobodies targeting the coiled-coil and N-terminal domains.Nat. Commun. 2020; 11: 4115Crossref PubMed Scopus (21) Google Scholar, 12Mao X. Ren Z. Parker G.N. Sondermann H. Pastorello M.A. Wang W. et al.Structural bases of unphosphorylated STAT1 association and receptor binding.Mol. Cell. 2005; 17: 761-777Abstract Full Text Full Text PDF PubMed Scopus (225) Google Scholar, 13Neculai D. Neculai A.M. Verrier S. Straub K. Klumpp K. Pfitzner E. et al.Structure of the unphosphorylated STAT5a dimer.J. Biol. Chem. 2005; 280: 40782-40787Abstract Full Text Full Text PDF PubMed Scopus (98) Google Scholar), resulting in antiparallel orientation of monomers as opposed to their parallel orientation in the phosphodimers (14Zhong M. Henriksen M.A. Takeuchi K. Schaefer O. Liu B. ten Hoeve J. et al.Implications of an antiparallel dimeric structure of nonphosphorylated STAT1 for the activation-inactivation cycle.Proc. Natl. Acad. Sci. U. S. A. 2005; 102: 3966-3971Crossref PubMed Scopus (124) Google Scholar). It has been recognized that unphosphorylated STATs are involved in many biological events in both normal and pathological situations (15Chatterjee-Kishore M. Wright K.L. Ting J.P. Stark G.R. How Stat1 mediates constitutive gene expression: a complex of unphosphorylated Stat1 and IRF1 supports transcription of the LMP2 gene.EMBO J. 2000; 19: 4111-4122Crossref PubMed Scopus (278) Google Scholar, 16Cui X. Zhang L. Luo J. Rajasekaran A. Hazra S. Cacalano N. et al.Unphosphorylated STAT6 contributes to constitutive cyclooxygenase-2 expression in human non-small cell lung cancer.Oncogene. 2007; 26: 4253-4260Crossref PubMed Scopus (57) Google Scholar, 17Majoros A. Platanitis E. Szappanos D. Cheon H. Vogl C. Shukla P. et al.Response to interferons and antibacterial innate immunity in the absence of tyrosine-phosphorylated STAT1.EMBO Rep. 2016; 17: 367-382Crossref PubMed Scopus (35) Google Scholar, 18Park H.J. Li J. Hannah R. Biddie S. Leal-Cervantes A.I. Kirschner K. et al.Cytokine-induced megakaryocytic differentiation is regulated by genome-wide loss of a uSTAT transcriptional program.EMBO J. 2016; 35: 580-594Crossref PubMed Scopus (53) Google Scholar, 19Stark G.R. Cheon H. Wang Y. Responses to cytokines and interferons that depend upon JAKs and STATs.Cold Spring Harb. Perspect. Biol. 2018; 10a02855Crossref PubMed Scopus (67) Google Scholar). In some instances, the self-assembly of U-STATs has been linked to tyrosine phosphorylation and cytokine signaling. For example, the homodimerization of U-STAT4 is a prerequisite for cytokine-induced STAT4 activation (20Ota N. Brett T.J. Murphy T.L. Fremont D.H. Murphy K.M. N-domain-dependent nonphosphorylated STAT4 dimers required for cytokine-driven activation.Nat. Immunol. 2004; 5: 208-215Crossref PubMed Scopus (85) Google Scholar), and heterodimerization of U-STAT1 with U-STAT2 has been shown to have positive or negative consequences for type 1 and type 2 interferon signaling, respectively (21Ho J. Pelzel C. Begitt A. Mee M. Elsheikha H.M. Scott D.J. et al.STAT2 is a pervasive cytokine regulator due to its inhibition of STAT1 in multiple signaling pathways.PLoS Biol. 2016; 14e2000117Crossref PubMed Scopus (42) Google Scholar, 22Wang Y. Song Q. Huang W. Lin Y. Wang X. Wang C. et al.A virus-induced conformational switch of STAT1-STAT2 dimers boosts antiviral defenses.Cell Res. 2021; 31: 206-218Crossref PubMed Scopus (23) Google Scholar). These and other findings indicate that U-STAT dimers fulfill roles critical for cytokine functioning, but the knowledge of dimerization before cytokine-induced phosphorylation remains incomplete. This is particularly true for heterotypic interactions across the STAT family and the in vivo situation generally. To fill in knowledge gaps, we devised an assay to systematically explore the repertoires of homo- and heterotypic interactions among U-STATs in living cells. The assay is based on U-STATs being nucleocytoplasmic shuttling proteins that can freely cross the soft diffusion barrier posed by the nuclear pore, presumably via direct contact with nuclear pore proteins (23Vinkemeier U. Getting the message across, STAT! design principles of a molecular signaling circuit.J. Cell Biol. 2004; 167: 197-201Crossref PubMed Scopus (91) Google Scholar, 24Popken P. Ghavami A. Onck P.R. Poolman B. Veenhoff L.M. Size-dependent leak of soluble and membrane proteins through the yeast nuclear pore complex.Mol. Biol. Cell. 2015; 26: 1386-1394Crossref PubMed Scopus (83) Google Scholar). It results in generally pancellular distributions of U-STATs (25Meyer T. Gavenis K. Vinkemeier U. Cell type-specific and tyrosine phosphorylation-independent nuclear presence of STAT1 and STAT3.Exp. Cell Res. 2002; 272: 45-55Crossref PubMed Scopus (74) Google Scholar), which for STAT1 and STAT2 can be shifted to nuclear or cytoplasmic accumulation by tagging them with transferable carrier-dependent nuclear localization (NLS) or nuclear export (NES) signals (26Frahm T. Hauser H. Köster M. IFN-type-I-mediated signaling is regulated by modulation of STAT2 nuclear export.J. Cell Sci. 2006; 119: 1092-1104Crossref PubMed Scopus (22) Google Scholar). We reasoned that such signal-tagged variants might function as baits that attract co-expressed test proteins if binding interactions occurred. This would be evident by their co-localization in the bait protein's compartment. After rigorous testing and verification, this approach was used to probe homo- and heterotypic binding interactions within cells across the entire STAT family for the first time. Latent STATs are nucleocytoplasmic shuttling proteins that can freely cross the nuclear envelope by directly contacting nuclear pore proteins and additional carrier-mediated mechanisms (23Vinkemeier U. Getting the message across, STAT! design principles of a molecular signaling circuit.J. Cell Biol. 2004; 167: 197-201Crossref PubMed Scopus (91) Google Scholar). Accordingly, they display near pancellular distributions in cells before cytokine treatment (with the exception of STAT2, see below), which is preserved upon C-terminal fusion of fluorescent marker proteins such as mEGFP or mCherry (Fig. S1A, panels 1,12,16,20,24,28). A Western blot demonstrating the expression of full-length STAT-fluorophore fusion proteins is shown in Fig. S1B. As mentioned earlier, STAT2 deviates from the near-pancellular distribution of the other STAT proteins; it accumulates in the cytoplasm due to potent nuclear export activity in its C-terminal transactivation domain (Fig. S1A, panel 6). Consistent with previous observations (26Frahm T. Hauser H. Köster M. IFN-type-I-mediated signaling is regulated by modulation of STAT2 nuclear export.J. Cell Sci. 2006; 119: 1092-1104Crossref PubMed Scopus (22) Google Scholar, 27Banninger G. Reich N.C. STAT2 nuclear trafficking.J. Biol. Chem. 2004; 279: 39199-39206Abstract Full Text Full Text PDF PubMed Scopus (103) Google Scholar), STAT2 variants with truncated transactivation domain, referred to as STAT2ΔC, showed strongly reduced cytoplasmic accumulation, thus adopting a distribution more like the other STAT family members (Fig. S1A, panel 9). The addition of transferable heterologous nuclear export (NES) or nuclear import signals (NLS) to wild-type STATs (or the C-terminally truncated STAT2) allowed us to direct bait proteins to the cytoplasm or the nucleus, respectively (Fig. S1A, panels 3,4,10,11,14,15,18,19,22,23,26,27,29). We used two well-characterized and highly active signals, namely, an NES derived from protein kinase A inhibitor and the NLS from simian virus 40 large T-antigen (28Fu S.C. Fung H.Y.J. Cağatay T. Baumhardt J. Chook Y.M. Correlation of CRM1-NES affinity with nuclear export activity.Mol. Biol. Cell. 2018; 29: 2037-2044Crossref PubMed Scopus (24) Google Scholar, 29Hodel A.E. Harreman M.T. Pulliam K.F. Harben M.E. Holmes J.S. Hodel M.R. et al.Nuclear localization signal receptor affinity correlates with in vivo localization in Saccharomyces cerevisiae.J. Biol. Chem. 2006; 281: 23545-23556Abstract Full Text Full Text PDF PubMed Scopus (83) Google Scholar). These nuclear translocation-related features potentially allow STATs to alter the localization of proteins they interact with. Schematics of the assay design are shown in Figure 1A. Another important aspect of the assay we considered was the strength of binding interactions, as this posed a main constraint on whether they are detectable as co-localization. Current data on U-STAT dimer assembly is largely qualitative; however, equilibrium sedimentation studies indicate that unphosphorylated STAT1 and STAT3 form high-affinity homodimers, both with a Kd in the low nanomolar range (10Wenta N. Strauss H. Meyer S. Vinkemeier U. Tyrosine phosphorylation regulates the partitioning of STAT1 between different dimer conformations.Proc. Natl. Acad. Sci. U. S. A. 2008; 105: 9238-9243Crossref PubMed Scopus (116) Google Scholar, 30Braunstein J. Brutsaert S. Olson R. Schindler C. STATs dimerize in the absence of phosphorylation.J. Biol. Chem. 2003; 278: 34133-34140Abstract Full Text Full Text PDF PubMed Scopus (163) Google Scholar). Moreover, structural and mutational studies of the dimer interfaces have identified the conserved N-domain and key anchoring residues therein as critical for unphosphorylated dimer assembly. Unphosphorylated STATs with N-domain truncation or mutated dimerization hot spot residues retain nucleocytoplasmic shuttling and their subcellular distributions differ little from their wild-type counterparts (31Domoszlai T. Martincuks A. Fahrenkamp D. Schmitz-Van de Leur H. Küster A. Müller-Newen G. Consequences of the disease-related L78R mutation for dimerization and activity of STAT3.J. Cell Sci. 2014; 127: 1899-1910Crossref PubMed Scopus (27) Google Scholar, 32Meyer T. Hendry L. Begitt A. John S. Vinkemeier U. A single residue modulates tyrosine dephosphorylation, oligomerization, and nuclear accumulation of stat transcription factors.J. Biol. Chem. 2004; 279: 18998-19007Abstract Full Text Full Text PDF PubMed Scopus (62) Google Scholar) (see also Fig. S1A, panels 2,8,13,17,21,25). Nonetheless, these mutations render them essentially monomeric, as quantitative analyses of N-domain-deleted or residue F77-mutated STAT1 showed a >100-fold drop in binding affinities (10Wenta N. Strauss H. Meyer S. Vinkemeier U. Tyrosine phosphorylation regulates the partitioning of STAT1 between different dimer conformations.Proc. Natl. Acad. Sci. U. S. A. 2008; 105: 9238-9243Crossref PubMed Scopus (116) Google Scholar, 12Mao X. Ren Z. Parker G.N. Sondermann H. Pastorello M.A. Wang W. et al.Structural bases of unphosphorylated STAT1 association and receptor binding.Mol. Cell. 2005; 17: 761-777Abstract Full Text Full Text PDF PubMed Scopus (225) Google Scholar). For STAT3, mutation of residue L78 similarly results in unphosphorylated dimer dissociation in living cells as demonstrated by Förster resonance energy transfer (31Domoszlai T. Martincuks A. Fahrenkamp D. Schmitz-Van de Leur H. Küster A. Müller-Newen G. Consequences of the disease-related L78R mutation for dimerization and activity of STAT3.J. Cell Sci. 2014; 127: 1899-1910Crossref PubMed Scopus (27) Google Scholar). These established results allowed us to use wild-type and mutant STAT1 and STAT3 as positive and negative controls to test if the co-localization of STAT proteins could be a reliable indicator for their dimerization inside cells. As shown in Figure 1B, co-expression of NES-fused STAT3 with wild-type STAT3 indeed resulted in their co-localization in the cytoplasm. Importantly, co-localization was lost if the monomeric STAT3-L78R mutant was used, which displayed pancellular distribution irrespective of the presence of STAT3-NES. This was the expected behavior of non-interacting, monomeric STAT3, in accord with the aforementioned in vitro and cellular studies. We then examined the homodimerization of STAT1 to validate further the co-localization assay. In the same way that STAT3 interacted with STAT3-NES, wild-type STAT1 co-localized with STAT1-NES in the cytoplasm, suggestive of homodimerization (Fig. 1B). To examine whether dimerization can also occur in the cell nucleus, we co-expressed STAT1 and nuclear-accumulated STAT1-NLS. As shown in Figure 1B, this likewise resulted in the co-localization of the two STAT1 variant proteins, albeit in the nucleus, demonstrating that dimerization was not limited to the cytoplasmic compartment. To test the consequences of a dimer-disrupting mutation for STAT1 homodimers, we used the mutant STAT1-F77A. In contrast to wild-type STAT1, the F77A mutant did not accumulate in the nucleus with co-expressed STAT1-NLS but retained pancellular distribution, similar to STAT3-L78R, thus again presenting the expected behavior of monomeric STATs. Finally, it was important to ascertain that the failure of a test protein to co-localize with bait was correctly attributed to a lack of dimer formation, rather than a shortfall of bait protein expression. Therefore, the relative expression of test and bait proteins was determined using quantitative fluorescence imaging as described in Experimental procedures, and cells were disregarded for inclusion in co-localization analyses if the expression of bait variants was below equimolar levels or exceeded the test protein's concentration by more than 4-fold. In summary, co-localization in nuclear or cytoplasmic compartments was a reliable indicator to assess the homodimerization of unphosphorylated STAT1 and STAT3. We, therefore, expanded this approach to assess the homodimer formation of all seven STATs as well as their ability to heterodimerize. To similarly examine the homodimerization of the other five STATs, we co-expressed STAT4, STAT5A, STAT5B, and STAT6 with their respective NES-tagged counterparts. As shown in Figure 2A, STAT6 failed to accumulate in the cytoplasm, in stark contrast to STAT4, STAT5A, and STAT5B, which co-localized with their NES-tagged equivalents. We quantified the extent of co-localization for each experiment by calculating Pearson's correlation coefficients (rP) for 10 to 30 cells. Numerical values for homodimers are shown in Figure 2C and summarized in Figure 3. STAT1, STAT3, STAT4, STAT5A, and STAT5B have Pearson's correlation coefficients of 0.94 or higher, indicative of near-complete co-localization due to stable homodimerization. STAT6 diverged strongly as it showed no cytoplasmic co-localization, with an accordingly significantly lower rP value of 0.41. To assess the homodimerization of STAT2, the C-terminally truncated variant STAT2ΔC with strongly reduced intrinsic nuclear export activity was used, as described earlier (Fig. S1A, panel 9). Although the C-terminus is needed for efficient constitutive nuclear export of STAT2, this region generally appears to be dispensable for the dimerization of U-STATs (10Wenta N. Strauss H. Meyer S. Vinkemeier U. Tyrosine phosphorylation regulates the partitioning of STAT1 between different dimer conformations.Proc. Natl. Acad. Sci. U. S. A. 2008; 105: 9238-9243Crossref PubMed Scopus (116) Google Scholar, 33Lim C.P. Cao X. Structure, function, and regulation of STAT proteins.Mol. Biosyst. 2006; 2: 536-550Crossref PubMed Scopus (247) Google Scholar), including the heterodimerization of U-STAT2ΔC and U-STAT1, which was demonstrated using STAT1-NES and STAT2ΔC-NLS as baits (Fig. S2). The C-terminally truncated STAT2 mutant was therefore co-expressed with wild-type STAT2 to probe U-STAT2 homodimerization. We observed incomplete cytoplasmic co-localization (Fig. 2A) and a correlation coefficient rP = 0.71 that was significantly reduced compared to stable dimers of other STAT family members yet higher than for non-interacting monomeric STAT6 (Fig. 2C). This could signify genuine homodimer assembly, albeit with lower affinity, or merely constitute an artifact reflecting co-localization due to residual cytoplasmic accumulation of STAT2ΔC. To distinguish between these possibilities, STAT2ΔC was directed to the nucleus through its tagging with an NLS (Fig. S1A, panel 11) to examine if this resulted in the nuclear localization of co-expressed STAT2ΔC. This was not the case, however, as the slight cytoplasmic accumulation of STAT2ΔC appeared unaltered, with no indication of increased nuclear translocation in the presence of the nuclear accumulated STAT2 bait protein. The correlation coefficient accordingly dropped sharply into the negative, rP = −0.45 (Fig. 2C) which indicated opposite distributions and hence a lack of co-localization of the two STAT2 variants. We concluded that unphosphorylated STAT2, like STAT6, did not homodimerize, in contrast to STAT1, STAT3, STAT4, STAT5A, and STAT5B, which formed stable homodimers in living cells. Next, we probed the heterodimer assembly of the seven unphosphorylated STATs by co-expressing NES-tagged STAT proteins as baits (except STAT2, where wild-type was used) and untagged wild-type STATs as the test proteins. Of the 21 possible heterotypic pairings, only two showed co-localization of bait and test proteins, namely, U-STAT1:STAT2 and U-STAT5A:STAT5B (Fig. 2, B and C), with rP values of .97 in both cases. All other combinations, including STAT1:STAT3 or STAT3:STAT4 (Fig. 2B), which readily assemble heterodimers upon their cytokine-induced tyrosine-phosphorylation (6Delgoffe G.M. Vignali D.A. STAT heterodimers in immunity: a mixed message or a unique signal?.JAKSTAT. 2013; 2e23060PubMed Google Scholar), did not appear to heterodimerize in the absence of cytokine stimulation, and the corresponding rP values were accordingly low (Fig. 3). Thus, unphosphorylated STATs were generally present as stable dimers, predominantly as homodimers. Heterodimers were formed only between STAT1 and STAT2 and the two very closely related STAT5 proteins. STAT6 was the only family member devoid of detectable dimerization activity.Figure 3Homo- and heterotypic dimerization of the unphosphorylated STAT proteins. Summary of Pearson correlation coefficients obtained with the translocation assay for the 28 possible pairings of unphosphorylated STAT proteins. Data are obtained with NES fusion proteins and wild-type STAT2 as the baits. STAT2 homodimer data are for U-STAT2:STAT2ΔC (†) and U-STAT2ΔC-NLS:STAT2ΔC (††). Given are means ± standard deviation. Light and dark green highlighting marks stable homo- and heterodimers, respectively. Number of cells analyzed in each experiment are shown in brackets. See data availability section for source data.View Large Image Figure ViewerDownload Hi-res image Download (PPT) Several lines of experimental inquiry indicate that unphosphorylated STAT dimers adopt an antiparallel conformation that is dependent on N-domain interactions (8Droescher M. Vinkemeier U. Self-association of STAT proteins from monomers to paracrystals.in: Decker T. Müller M. Jak-Stat Signaling: From Basics to Disease. Springer, Vienna2012: 47-63Crossref Scopus (3) Google Scholar). Deletion of the N-domain, and even single N-domain point mutations, can therefore result in the dissociation of dimers. For STAT1, N-domain residues phenylalanine 77 and leucine 78 have been shown to be critical for the assembly of unphosphorylated dimers (12Mao X. Ren Z. Parker G.N. Sondermann H. Pastorello M.A. Wang W. et al.Structural bases of unphosphorylated STAT1 association and receptor binding.Mol. Cell. 2005; 17: 761-777Abstract Full Text Full Text PDF PubMed Scopus (225) Google Scholar), and likewise, the leucine residues 78 for STAT3 (31Domoszlai T. Martincuks A. Fahrenkamp D. Schmitz-Van de Leur H. Küster A. Müller-Newen G. Consequences of the disease-related L78R mutation for dimerization and activity of STAT3.J. Cell Sci. 2014; 127: 1899-1910Crossref PubMed Scopus (27) Google Scholar, 33Lim C.P. Cao X. Structure, function, and regulation of STAT proteins.Mol. Biosyst. 2006; 2: 536-550Crossref PubMed Scopus (247) Google Scholar, 34Sgrignani J. Olsson S. Ekonomiuk D. Genini D. Krause R. Catapano C.V. et al.Molecular determinants for unphosphorylated STAT3 dimerization determined by integrative modeling.Biochemistry. 2015; 54: 5489-5501Crossref PubMed Scopus (14) Google Scholar) and STAT4 (20Ota N. Brett T.J. Murphy T.L. Fremont D.H. Murphy K.M. N-domain-dependent nonphosphorylated STAT4 dimers required for cytokine-driven activation.Nat. Immunol. 2004; 5: 208-215Crossref PubMed Scopus (85) Google Scholar), see also Figure 1. We, therefore, mutated the homologous N-domain residues of STAT2 (L82A), STAT4 (L78S), STAT5A (L82A), and STAT5B (L82A) to examine which of the unphosphorylated STAT dimers shared the dependence on N-domain interactions. As is shown in Figure 4A, all seven homo- and heterodimers were destabilized upon mutation of the same single homologous side chain in the N-domain. We concluded that the U-STAT dimers adopted similar N-domain-mediated conformations. However, the dissociating effect on the U-STAT1:STAT2 heterodimer was comparatively weak as indicated by the relatively small albeit statistically significant reduction in its rP value (Fig. 4B). This could be an indication that this STAT dimer adopted an exceptionally stable conformation, which was tested next. To compare the relative binding strengths of U-STAT dimers, we co-expressed STATs that harbored opposed localization signals, namely, PKI NES or SV40 NLS. The nuclear export activity conferred by the PKI NES was determined to dominate over import activity associated with the NLS of SV40 since proteins such as GFP or GST accumulate in the cytoplasm when both signals are appended simultaneously (35Knauer S.K. Moodt S. Berg T. Liebel U. Pepperkok R. Stauber R.H. Translocation biosensors to study signal-specific nucleo-cytoplasmic transport, protease activity and protein-protein interactions.Traffic. 2005; 6: 594-606Crossref PubMed Scopus (42) Google Scholar). Of note, the same outcome, namely, accumulation in the cytoplasm, was observed if STAT1 was used as the acceptor of the two opposed signals (Fig. S1A, panel 5). We reasoned that subjecting the different dimers to these same antipodal translocation activities might reveal differences in the forces driving U-STAT association (Fig. 5A). As shown in Figure 5, B and D, the five U-STAT homodimers and the U-STAT5A:STAT5B heterodimer showed the same behavior, that is, the dimer subunits localized to the nucleus or the cytoplasm in accordance with their respective localization signals. We inferred that the opposed translocation forces dissociated these dimers. In notable difference, STAT2 and the nuclear-targeted STAT1-NLS variant sustained their cytoplasmic co-localization (Fig. 5, C and D), suggesting that U-STAT1:STAT2 heterodimers uniquely resisted dissociation. However, the co-localization of STAT2 and STAT1-NLS was lost upon the alanine mutation of STAT2 hot spot interface residue L82 (Fig. 5, C and D). The weakening of U-STAT1:STAT2 interactions caused by this mutation (see Fig. 4, A and B) evidently sufficed to reduce the binding affinity below the threshold required for cont