Structural analysis of U-STAT1-U-STAT2 heterodimers (U-dimer) by EM
We purified U-STAT2 and U-STAT1 from mammalian cells stably expressing the FLAG-tagged proteins and imaged the chromatographic fraction corresponding to the heterodimer by electron microscopy (EM), following negative staining (Supplementary information, Fig. S1a). The EM map unambiguously demonstrates that the U-dimer has an anti-parallel conformation (Fig. 1a), and reveals that its conformation resembles the previously determined structure of the anti-parallel U-STAT1 homodimer.8,9 Based on the known structure of the U-STAT1 homodimer (PDB entry: 1YVL) and the three-dimensional reconstructed EM map,9 we modeled the conformation of the U-dimer (Fig. 1a). Our current EM map, despite in low resolution, enables us to observe three protein-protein interfaces that stabilize the anti-parallel conformation, including STAT1-ND (N-terminal domain)-STAT2-ND, STAT1-CCD (coiled-coil domain)-STAT2-DBD (DNA-binding domain), and STAT1-DBD-STAT2-CCD (Fig. 1b). To validate this structural model, we designed several truncated versions of STAT1 and STAT2 and used glutathione S-transferase (GST) pull-down assays to examine the formation of heterodimer, confirming that STAT2 interacts mainly through the DBD of STAT1, and vice versa (Supplementary information, Fig. S1b, c). Since the anti-parallel conformation represents an inhibitory state of this STAT complex, the question of how this inhibitory state is disfavored arises. Interestingly, a post-translational modification (PTM) site, T404, identified in our mass spectrometry analysis, resides in the core of the heterodimeric interfaces (Fig. 1b, c). We expect that the addition of a bulky, negatively-charged phosphate group would disrupt the stable STAT1-STAT2 heterodimer interfaces, providing a clear example of how PTMs might be critical regulators of IFN-I-dependent signaling.
Phosphorylation of STAT2 T404 disfavors the inactive U-dimer
To investigate the effect of T404 phosphorylation on the formation of the inactive U-dimer, we mutated this residue to alanine, preventing its phosphorylation, and to glutamate, mimicking the negative charge of phospho-threonine. Without IFN-I treatment, the binding of U-STAT1 to U-STAT2 was greatly enhanced by the T404A mutation of STAT2 (Fig. 1d) and severely inhibited by the T404E mutation. In IFN-treated cells, there was no effect of these mutations on the stability of the complex between tyrosine-phosphorylated STAT1 and STAT2. EM analysis revealed that the U-dimer with a T404A mutation of STAT2 has the same conformation as the U-dimer with WT (wild-type) STAT2 (Supplementary information, Fig. S1d), confirming that preventing the phosphorylation of T404 stabilizes the anti-parallel conformation of the U-dimer.
To further investigate how U-STAT1 and U-STAT2 interact in this anti-parallel conformation, we studied the effects of a series of mutations of STAT1 on U-dimer stability. F77A (an ND mutation), F172A (a CCD mutation), and Q340W (a DBD mutation) abolished U-dimer formation, while Y701F (a tyrosine phosphorylation-deficient mutation) had no effect (Fig. 1e). We also found that the expression of canonical ISGs (IFIT1 and OAS1) was enhanced by the F172A and Q340W mutations, which destabilize the anti-parallel conformation (Fig. 1f). Notably, because the tyrosine phosphorylation of STAT1 and STAT2 was delayed by the F77A mutation of STAT1 (Supplementary information, Fig. S1e), ISG induction was compromised upon IFN-I treatment. We conclude that, although the ND-domain and CCD-DBD interactions are both critical for stabilizing the U-dimer, only the mutations that affect the CCD-DBD interactions specifically disturb the anti-parallel, inactive U-dimer and thus enhance IFN-I responses.
We propose that U-STAT1 associates with U-STAT2 through F173 and F175 in the CCD and T404 in the DBD of U-STAT2. The F175A mutant of U-STAT2 fails to bind to U-STAT1 and IRF9, suggesting that U-STAT1 and IRF9 compete for binding to U-STAT2 through this residue.5 The T404A mutant of U-STAT2 has a stronger interaction with U-STAT1, but a weaker interaction with IRF9, while the effect of the T404E mutation of U-STAT2 on its interactions with STAT1 and IRF9 is completely opposite (Fig. 1g). In our model of the U-dimer, F173 of STAT2 is located in the CC domain near F175, but facing inward, suggesting that the inhibition of ISGF3 function seen with the F173A mutant might be due to a local conformational change. Consistent with this idea, the response to IFN-β is severely compromised in U6A cells expressing the mutants T404A or F175A (Fig. 1h). These results reveal that F175 of U-STAT2 is important for its interaction with both U-STAT1 and IRF9, and that T404 phosphorylation is the switch for a conformational change of U-dimer, which is a potent negative regulator of IFN-I-dependent signaling.
Phosphorylation of STAT2 on T404 enhances the affinity of ISGF3 for DNA
To further investigate the functional outcomes of T404 phosphorylation, we analyzed the expression of some typical ISGs in response to IFN-I in STAT2-null U6A cells in which STAT2 variants are expressed. The T404A mutation dramatically inhibited the induction of IFIT1 and OAS1 after treatment with either IFN-α or IFN-β (Fig. 2a). We also analyzed the kinetics of IFN-β-induced gene expression, finding that the induction of these ISGs was greatly inhibited in cells expressing T404A at all times, especially after 8 h (Supplementary information, Fig. S2a). Compared to the WT protein, T404E STAT2 mediated more ISG induction in response to IFN-β, contrasting with the effect of the T404A mutation, while T404D partially mimicked the effect of T404 phosphorylation (Supplementary information, Fig. S2b).
We used an Illumina Gene Expression Array to understand in more detail how T404 phosphorylation affects ISG induction. Of the 31,425 genes on the array, 181 were induced in cells containing WT STAT2 after 4 h, 210 after 8 h, and 227 after 24 h. Of the total of 358 genes that responded to treatment with IFN-β at all three time periods, 266 were induced less well in cells with the T404A mutant than in cells with WT STAT2. A gene ontology analysis revealed that genes induced less well in U6A-T404A-STAT2 cells fell into sub-groups whose products are involved in antiviral defense, immune responses, or the IFN-I signaling pathway (Fig. 2b). Of these genes, 55 (Supplementary information, Fig. S2c) were affected at all three time periods. Most of them are well-known ISGs (Supplementary information, Table S1), with canonical interferon-stimulated response elements (ISREs) in their promoters (Supplementary information, Fig. S2d). Since the expression of most ISGs is inhibited by the T404A mutation, we conclude that the phosphorylation of STAT2 on T404 has a very substantial positive effect on the transcriptional activity of ISGF3.
We investigated whether the phosphorylation of STAT2 on T404 stabilizes ISGF3 and enhances its ability to bind to DNA. IRF9 is primarily responsible for the binding of ISGF3 to ISREs. In a chromatin immunoprecipitation (ChIP) assay with anti-IRF9, we found that, following treatment with IFN-β, the recruitment of IRF9 to ISREs was not induced in control U6A cells, which lack STAT2 expression, as expected (Fig. 2c). IRF9 in company with WT STAT2 was well recruited to ISREs, but was recruited much more strongly with ISGF3 variants containing T404E, but not T404A STAT2. We performed a ChIP-seq analysis with the above samples, followed by next-generation sequencing. Shown in Fig. 2c are the top ranked binding motifs for each sample, determined by using the Multiple Em for Motif Elicitation (MEME) algorithm.10 We observed binding to a 5′-TTCTCAGAAA-3′ motif in IFN-β-treated cells expressing WT or T404E STAT2, but not T404A STAT2 (Fig. 2d). We also used electrophoretic mobility shift assays (EMSAs) with an ISRE probe to assess the ability of ISGF3 to bind to DNA. In U6A cells that express WT STAT2, ISGF3 was detected readily. The amount of ISGF3 capable of binding to an ISRE sequence is greatly enhanced by the T404E mutation of STAT2 and severely inhibited by the T404A mutation (Supplementary information, Fig. S2e). We conclude that the phosphorylation of T404, which lies in the DNA binding domain of STAT2, enhances the affinity of ISGF3 for ISREs. Previously, we found that the phosphorylation of STAT2 T387 has an effect opposite to that of the phosphorylation of T404 in IFN-I-dependent signaling, and in U-STAT1-STAT2 formation.6 However, the T-to-D mutation of T387 does not mimic the phosphorylation, indicating that the effect of T387 phosphorylation is not mediated by the negative charge, and suggesting that T387 phosphorylation functions through a specific mechanism, e.g., by recruiting one or more proteins that regulate the balance between U-STAT1-STAT2 and ISGF3.
Phosphorylation of STAT2 on T404 expedites the tyrosine phosphorylation of STAT1 and STAT2
We investigated whether the phosphorylation of STAT1 and STAT2 on specific tyrosine residues, required for IFN-I-dependent signaling, is affected by the status of the U-dimer, which is destabilized by the phosphorylation of T404. Phosphorylation of T404 enhanced the tyrosine phosphorylation of STAT1 and STAT2 at the early time of 30 min, and the level of tyrosine phosphorylation reached a plateau after 6 h (Supplementary information, Fig. S2f). A similar result was observed with HME (human mammary epithelial) cells expressing WT, T404A, or T404E STAT2 (Supplementary information, Fig. S2g). Since it has been shown that the anti-parallel U-STAT1 homodimer contributes to the de-phosphorylation of Y701 of STAT1,8 we investigated whether the differences in tyrosine phosphorylation of STAT1 and STAT2 might be due to effects on de-phosphorylation. Cells were treated with IFN-β for 2 h, followed by treatment with Staurosporine, a kinase inhibitor that rapidly blocks tyrosine phosphorylation. This procedure permits analysis of the rate of decay of the initially phosphorylated residues. The T404E mutation did not suppress the de-phosphorylation of tyrosine residues of either STAT1 or STAT2 (Supplementary information, Fig. S2h). In addition, an assay for nuclear translocation of the STATs showed that T404 phosphorylation did not affect this process either (Supplementary information, Fig. S2i). Taken together, our results reveal that the phosphorylation of STAT2 on T404 enhances the tyrosine phosphorylation of STAT1 in response to IFN-I, especially at early time.
The IFN-I receptor consists of two subunits, IFNAR1 and IFNAR2. Following ligand binding, IFNAR1 is phosphorylated on tyrosine 466, which recruits STAT2 (but not STAT1) via its SH2 domain,11 followed by the sequential phosphorylation, first of STAT2 and then of STAT1.12 We investigated how STAT2 T404 phosphorylation affects the tyrosine phosphorylation of STAT1 and STAT2, observing that the T404A mutation of STAT2 reduces and delays the interaction of STAT2 with phosphorylated IFNAR1, whereas the T404E mutation leads to a stronger interaction following treatment with IFN-I (Fig. 2e). These results indicate that T404 phosphorylation enhances and expedites the tyrosine phosphorylation of both STAT2 and STAT1, increasing the affinity of STAT2 for activated IFNAR1 by destabilizing the U-STAT1-U-STAT2 dimer (Fig. 2f).
Virus infection induces the T404 phosphorylation of STAT2 by IKK-ε
We generated a specific polyclonal antibody against a STAT2 peptide that includes phosphorylated T404 and used ELISA to demonstrate that this antibody does not cross-react with the corresponding un-phosphorylated peptide (Supplementary information, Fig. S3a). To identify kinases responsible for STAT2 T404 phosphorylation, we found that the amino acid sequence surrounding T404 in STAT2 is a good match with sequences that are preferentially phosphorylated by IKK-ε. Phosphorylation of STAT1 on S708 by IKK-ε is known to regulate the formation of STAT1 homodimers in response to IFN-γ, but not ISGF3 formation in response to IFN-I.13,14 Western analyses employing this antibody showed that the phosphorylation of STAT2 on T404 is induced by ectopic expression of IKK-ε, but not a kinase-dead mutant enzyme (Fig. 3a). We also used an in vitro kinase assay to demonstrate that IKK-ε phosphorylates T404 of STAT2 directly (Fig. 3b). Confirming this result, the phosphorylation of STAT2 was greatly impaired when T404 was mutated. We also evaluated TBK1, a homolog of IKK-ε, finding that it can also phosphorylate STAT2 on T404, although not as robustly as IKK-ε (Fig. 3b). To obtain more information on how these kinases phosphorylate STAT2 on T404, we performed in vitro pull-down assays with purified IKK-ε and TBK1. STAT2 binds directly to both IKK-ε (Fig. 3c) and TBK1 (Supplementary information, Fig. S3b). Using a set of GST-fusion proteins with different IKK-ε truncations, we detected a direct interaction between the kinase domain of IKK-ε and STAT2 (Supplementary information, Fig. S3c), and also found that the STAT2 DBD, which includes T404, binds directly to IKK-ε (Supplementary information, Fig. S3d). In addition, a co-immunoprecipitation experiment using HEK293T cells with ectopic expression of both STAT2 and IKK-ε revealed that the T404A mutation did not abrogate their interaction (Fig. 3d). These data indicate that IKK-ε binds directly to the DBD of STAT2 to phosphorylate T404 as an important component of a robust response to IFN-I, while TBK1 has a weaker ability to phosphorylate this residue.
Pretreatment of HME cells with Amlexanox, an IKK-ε inhibitor, dramatically inhibited the induction of ISGs by IFN-I (Fig. 3e). Furthermore, knocking down the expression of IKK-ε in the same cells with an shRNA (Fig. 3f) significantly suppressed ISG expression in response to IFN-β (Fig. 3g). We observed similar results when the same experiment was done with Hela cells (Supplementary information, Fig. S3e, f). We also noted that down-regulation of TBK1 expression, driven by an shRNA, also suppressed ISG expression (Supplementary information, Fig. S3g, h).
IKK-ε and TBK1 phosphorylate and activate IFN regulatory factor 3 (IRF3) in response to virus infection, further evoking cellular antiviral responses, including the production of IFN-I.15,16,17 We found that, in human primary fibroblast cells, VSV strongly induced the phosphorylation of STAT2 on T404, along with that of IRF3 (Fig. 3h), and the same observation was made in HME cells (Supplementary information, Fig. S3i). Poly I:C, which simulates virus infections, significantly induced T404 phosphorylation, and this activity was abolished by silencing IKK-ε (Fig. 3i). We conclude that virus infections induce T404 phosphorylation in an IKK-ε-dependent manner.
The phosphorylation of STAT2 on T404 is critical for IFN-dependent antivirus defenses in vitro and in vivo
The amino acid sequence surrounding T404 in human STAT2 is highly conserved in several other mammalian species (Supplementary information, Fig. S4a). We generated STAT2 T403A/T403A mice, using CRISPR/Cas9 technology (Supplementary information, Fig. S4b, c), in order to examine the effects of the T404 mutation on virus defense in vivo (the corresponding murine residue is T403). Signaling in response to IFN-I was assessed in murine embryonic fibroblasts (MEFs) and in bone marrow-derived dendritic cells (BMDCs) from STAT2 WT/WT or T403A/T403A mice. Four hours after treatment with murine IFN-β, the induction of ISGs was greatly reduced in cells derived from T403A/T403A mice (Supplementary information, Fig. S4d, e).
The expression of about 74% (266/358) of ISGs induced by IFN-I was reduced in U6A cells expressing exogenous T404A STAT2 (Supplementary information, Fig. S2c). To determine how T404 phosphorylation affects IFN-I signaling in cells expressing endogenous STAT2, especially immune cells, we used RNA-seq to performed transcriptome profiling in macrophages derived from WT/WT or T403A/T403A mice. Most of the 606 genes that are well-induced (fold change > 2) in WT cells after stimulation with IFN-I (Fig. 4a, left) are induced much less well in T403A cells (Fig. 4a, right). The top 40 genes are listed in Fig. 4b. A gene ontology enrichment analysis demonstrated that the genes with lower expression in T403A cells can be categorized into cytokine activity, adaptive immune response, and myeloid leukocyte activation (Supplementary information, Fig. S4f).
The protein products of IFN-induced ISGs are critical for host cell defense against virus infection. To determine whether the differences in gene expression between WT and T404A STAT2 cells translate into a compromised antiviral defense, MEFs from WT/WT or T403A/T403A mice were infected with VSV-GFP. The amount of VSV-G protein in T403A/T403A MEFs was substantially higher than in WT/WT MEFs (Fig. 4c; Supplementary information, Fig. S5a). We also monitored the kinetics of infection with WT VSV (Fig. 4d). The difference in the amount of VSV genomic RNA between MEFs derived from WT/WT and T403A/T403A mice became more dramatic with time. We repeated the same experiment in HME cells, in which the expression of STAT2 is constitutively low, stably expressing ectopic FLAG-tagged WT, T404A, or T404E STAT2 at comparable levels.6 The T404E cells were more resistant to VSV infection, while cells expressing T404A STAT2 were more susceptible than cells expressing WT STAT2 (Supplementary information, Fig. S5b). These results indicate that the phosphorylation of STAT2 on T404 is required for full antiviral activity in both human and murine cells.
To determine whether the phosphorylation of STAT2 on T403 is crucial for protection against infection in vivo, WT/WT, WT/T403A, and T403A/T403A mice were challenged with VSV, administered by tail vein injection. The T403A/T403A mice were highly susceptible to infection, while their WT siblings were resistant (Fig. 4e). T403A/WT mice, in which both forms of STAT2 are expressed, responded similarly to WT/WT mice, suggesting that STAT2 phosphorylated on T403 is vital in the defense against this infection. T403A/T403A mice are also much more sensitive to exposure of VSV in terms of weight change following infection (Fig. 4f). The body masses of T403A/T403A mice started to drop on day 4 post infection and continued thereafter, while the WT/WT mice started to recover at this time. We also observed that T403A/T403A mice developed conjunctivitis (Supplementary information, Fig. S5c). In addition to VSV, an RNA virus, we also challenged the mice with HSV, a DNA virus, and found that T403 phosphorylation of STAT2 is essential for antiviral defense in vivo (Fig. 4g).
Impaired virus clearance, due to a compromised IFN response, leads to severe viral encephalitis in T403A/T403A mice
Viral infection activates host immune responses and induces the expression of IFNs and other cytokines. When the mice were challenged with VSV for a short time, the amount of IFN-β induced was higher in most of the tissues from T403A/T403A mice (Supplementary information, Fig. S5d), showing that the susceptibility of these mice to VSV is not due to a deficiency in the production of IFN-I. We also determined how infection correlated with induction of the IFN response, by checking the expression of ISGs in spleens, as the lymphoid organ most critical for detection and response to infection. The level of VSV genomic RNA was dramatically higher in the tissues of T403A/T403A mice than in WT/WT mice (Fig. 4h). Interestingly, the expression of ISGs induced by infection was much lower in the former, especially at early time points (Fig. 4i). The above data is consistent with our previous finding (Fig. 2f), indicating that T404/403 phosphorylation expedites antiviral defense by enhancing responses to IFN-I.
Although VSV infects a variety of tissues, neurons are a major target. We monitored the levels of viral genomic RNA in different tissues six days after infection (Supplementary information, Fig. S5e). Very little viral RNA was detected in the brains of WT/WT mice, which had already recovered from the infection at this time (Supplementary information, Fig. S5e). Titers of VSV were about 1000-fold higher in the brains of the T403A/T403A mice (Fig. 5a). To further characterize the consequences of VSV infection, we stained different tissues with hematoxylin and eosin, revealing significant differences in the brains (Fig. 5b), but not in other tissues (Supplementary information, Fig. S5f). A dramatic lesion in the brains of the T403A/T403A mice was perivascular cuff, defined as an accumulation of lymphocytes or plasma cells in a dense mass around blood vessels. Cuffing is usually seen in the margins of plaques, where it serves as a marker of inflammation.18 We used flow cytometry to quantify the numbers of immune cells, as indications of inflammation,19 in the brains four days after infection. The percentages of CD45hiCD11bhi cells were substantially higher in infected T403A/T403A mice (Fig. 5c, left). Within this infected population, the percentages of cells expressing high levels of CD206, an anti-inflammatory marker, were much lower (Fig. 5c, middle), and the percentages of cells with high expression of Ly6C, a pro-inflammatory marker, were much higher in T403A/T403A than WT/WT mice (Fig. 5c, right). These results reveal that many myeloid cells had infiltrated into the brains of T403A/T403A mice, promoting local inflammation at a late stage of VSV infection, due to failure to remove the virus. In encephalitis, T cells are viewed primarily as detrimental.20 CD8+ cytotoxic T cells play an especially important role in viral encephalitis, helping to eliminate the virus along with infected cells. CD8+ T cells derived from T403A/T403A mice had much higher expression of CD69, a T cell activation marker, than T cells derived from WT/WT mice (Fig. 5d), while CD4+ T cells and CD19+ B cells remained similar (Supplementary information, Fig. S5g).
Along with massive inflammatory cell infiltration, elevated proinflammatory cytokine and chemokine expression, called a cytokine storm, result in deleterious encephalitis that is caused by rapid virus replication.18,21 We found significantly higher levels of CCL2 and CSF1, myeloid cell-derived chemokines, in the brain supernatants and plasma from T403A/T403A mice, along with robust production of IFN-β (Fig. 5e and Supplementary information, Fig. S5h). We also observed excessive expression of classical cytokine storm-associated cytokines and chemokines in the brain tissues (Fig. 5f). Together with our previous observation of a delayed IFN response (Fig. 4i), the mice with T403 phosphorylation deficiency failed to release U-STAT1 and U-STAT2 from the inactive anti-parallel conformation to facilitate a rapid antiviral response, which makes them vulnerable to viral encephalitis with an overactive immune response and an intractable cytokine storm.