STING Nuclear Partners Contribute to Innate Immune Signalling Responses

STING and cGAS initiate innate immune responses (IIR) by recognizing cytoplasmic pathogen dsDNA and activating signaling cascades from the ER; however, another less investigated pool of STING resides in the nuclear envelope. We find that STING in the inner nuclear membrane increases mobility and changes localization upon IIR activation both from dsDNA and poly(I:C) stimuli. We next identified nuclear partners of STING from isolated nuclear envelopes. These include several known nuclear membrane proteins, bromodomain and epigenetic enzymes, and RNA- or DNA-binding proteins. Strikingly, 17 of these DNA and RNA-binding STING partners are known to bind direct partners of the IRF3/7 transcription factors that are central drivers of IIR. We find that several of these STING partners —SYNCRIP, Men1, Ddx5, snRNP70, RPS27a, Aatf— can contribute to IIR activation and SYNCRIP can moreover protect against influenza A virus infection. These data suggest that the many roles identified for STING likely reflect its interactions with multiple RNA and DNA-binding proteins that also function in IIR.


Introduction
The innate immune response (IIR) is the first line of defense against pathogens, recognizing molecular patterns in infected cells such as the presence of cytoplasmic DNA or dsRNA 1,2 . STING (STimulator of INterferon Genes) (also called MITA, ERIS, MPYS, NET23, and TMEM173) is the essential adaptor protein in innate immune signaling cascades triggered by cytosolic DNA. Cyclic GMP-AMP synthase (cGAS) senses cytoplasmic dsDNA and catalyzes the synthesis of a second messenger, cGAMP, which binds to and activates dimeric STING at the ER, activating IIR signaling cascades that stimulate IRF3/7 transcription factors to activate IIR genes such as type I interferons (IFN) 3,4,5,6,7,8,9 . The cGAS-STING pathway for IIR triggered by cytoplasmic dsDNA has been well characterized in molecular detail; however, STING clearly has other important functions. As well as restricting the proliferation of DNA viruses through type-I IFN induction, STING also restricts the proliferation of some RNA viruses 10,11,12 , although the mechanisms through which it does so remain to be fully elucidated. Although STING does not directly bind to RNA, the replication of multiple positive-and negative-sense RNA viruses is enhanced in the absence of STING 6,7,11,12,13,14,15,16,17,18,19,20,21 . Several reports have argued that STING is not involved in interferon activation in response to foreign RNA 7,13,22 ; so how it acts against RNA viruses is less clear than its counteraction of DNA viruses through the induction of type I IFN. Moreover, other IIR roles for STING have been identified in pro-apoptotic signaling with MHC II from the plasma membrane 23 , the induction of autophagy 24 , and in NF-κB activation downstream of DNA damage 25 . The many distinct functions and localizations reported for STING in IIR make it difficult to distinguish direct from downstream signaling effects.
Though originally introduced in a proteomics study of the nuclear envelope (NE) as NET23 26 , potential nuclear roles for STING have been largely unaddressed. A significant pool of STING was confirmed in the NE that was partly lost in the absence of lamin A 27 , suggesting it can reside in the inner nuclear membrane (INM); however, super resolution microscopy employed in that study suggested it was limited to the outer nuclear membrane (ONM). A subsequent study showing a role for STING in promoting chromatin compaction further suggested a function inside the nucleus 28 , though without demonstration of a pool inside the nucleus.
The potential importance to IIR of a pool of STING inside the nucleus has been enhanced greatly by recent reports of cGAS, its upstream partner in cytoplasmic dsDNA sensing, in the nucleus 29,30,31,32,33 . cGAS directly binds DNA and so a long-standing question in the field was how it was prevented from binding and being activated by chromosomal DNA. Recent studies have shown that in fact a large portion of cGAS is in the nucleus bound to chromosomes 3,29,31,34 , so it is critical to keep this pool from activating IIR. At the same time, this also raises the possibility that nuclear cGAS could sense pathogen nucleic acids inside the nucleus as well as in the cytoplasm and thus activate IIR from inside the nucleus in which case nuclear STING could be required for such signaling responses.
Here we confirm inner nuclear membrane residence for endogenous STING and show that tagged nuclear STING can redistribute from the nucleus to the ER upon treatment with dsDNA or, surprisingly, the dsRNA mimetic poly(I:C). We further show that STING dynamics in the nuclear envelope increase with both DNA and RNA triggered immune responses. Moreover, we identify NE partners for STING, which are enriched for RNA and DNA binding proteins, and testing several of these partners indicates that they can contribute to IIR activation and one of the partners identified, SYNCRIP, can moreover protect against influenza and RNA virus infection.

STING targets to the inner nuclear membrane.
An earlier attempt to determine if the NE pool of STING was in the outer (ONM) or inner (INM) nuclear membrane was inconclusive 27 . Therefore, we used structured illumination (OMX) super resolution microscopy that can distinguish INM proteins from ONM proteins by their being in the same plane with respectively nuclear basket or cytoplasmic filament proteins of the nuclear pore complexes (NPCs) that are separated by ~100 nm (Fig. 1a) 35 . Analyzing several cells on the same coverslip revealed STING-GFP to be in the ONM of some cells and the inner nuclear membrane (INM) of other cells (Fig. 1b). The finding of some cells with STING in the ONM and others in the INM was striking, as most NE membrane proteins tested with this method were unambiguously resident in either the INM or ONM 36,37 . This suggested STING may redistribute into different nuclear locations under certain cell-specific conditions.
As the super resolution approach used STING fused to GFP, we also analyzed endogenous STING within the NE using immunogold electron microscopy (Fig. 1c, left images). Similar numbers of gold particles were observed at the INM (142) as at the ONM (112). Specificity of immunogold labelling against endogenous STING was confirmed by the absence of gold particles in samples stained only with secondary antibodies ( Supplementary   Fig. S1a). To further determine whether the GFP fusion interfered with this distribution, a cell line stably expressing STING-GFP was also analyzed by immunogold electron microscopy ( Fig. 1c, right images). The distribution of particles between inner and outer nuclear membranes was not notably altered by the tag (Fig. 1d). A larger proportion of particles could not be clearly ascribed to either INM or ONM as they were detected in the lumen between the two membranes of the NE. This is likely a result of an increased distance between gold particle and protein due to the addition of the C-terminal GFP tag combined with sectioning resulting in STING-GFP residing within the INM or ONM appearing as luminal by immunogold EM.
Finally, we confirmed the inner nuclear membrane pool of STING by the ability of a C-terminally tagged STING-RFP construct to accept photons from lamin A-GFP through Förster resonance energy transfer by fluorescence lifetime imaging (FRET-FLIM). The lifetime of the activated lamin A-GFP fluorescence was reduced when cells also expressed STING-RFP as a photon acceptor (Fig. 1e). On average this INM pool of STING dropped the mean lifetime (τ) of lamin A-GFP from 2.281 to 2.142 ns. Expected targeting of STING-RFP to the ER/NE was confirmed by confocal microscopy (Supplementary Fig. S1b).
STING dynamics are altered upon stimulation of IIR. One possible explanation for the finding of STING-GFP in the inner nuclear membrane of some cells and not in others in Figure 1b is that its localization might be altered by activation of IIR, especially as the transient transfections used for most of these experiments could have stimulated IIR in a subpopulation of transfected cells due to sensing plasmid DNA. Although nuclear redistribution of STING during IIR has not been investigated, STING accumulates in perinuclear aggregates, with a visible decrease in nuclear localization, upon IIR activation with dsDNA but not dsRNA 7, 13 (see also Supplementary Fig. S2). This implies that, at least in response to DNA immune stimulation, STING at the INM must translocate to peri-nuclear foci, as for STING localized in the peripheral ER. Measurement of STING dynamics required using fluorescently tagged fusion proteins, therefore, we first compared the redistribution of STING in a stable STING-GFP cell line with that of endogenous STING using two different antibodies and fixations and stimulating IIR with either plasmid DNA or poly(I:C), finding a similar redistribution pattern for the dsDNA and a similar lack of visible redistribution for the dsRNA mimic, poly(I:C) (Supplemental Fig. S2a-d).
We hypothesized that if STING activation promotes shuttling between the nucleus and cytoplasm, its mobility measured by fluorescence recovery after photobleaching (FRAP) should change upon IIR induction. To avoid unintentional IIR induction due to transfected plasmid DNA, STING-GFP or a control nuclear envelope transmembrane protein (NET) fused to GFP (NET55) were stably expressed in HT1080 cells. Induction of IIR by infection with herpes simplex virus type 1 (HSV-1) visibly increased the speed of fluorescence recovery (Fig. 2a), reducing the t½ for STING in the NE by ~⅓ from 11.1 to 6.7 s (Fig. 2b,c).
In contrast, the t½ of control NET55 was unaffected. This most likely indicates an increase in STING shuttling upon IIR activation, because NE FRAP has been shown to principally measure translocation through the peripheral channels of the NPC 38 . Interestingly, STING was not observed to accumulate in peri-nuclear foci as occurs with dsDNA stimulation.
However, this is consistent with reports that HSV-1 inhibits STING activation and can prevent translocation of STING to the Golgi apparatus 39,40,41 . Suprisingly, STING mobility in the NE was also increased by stimulation with poly(I:C) (Fig. 2d), dropping the t½ from 13.3 s for the control in that experiment to 7.59 s (Fig. 2e). This was unexpected because the STING perinuclear accumulation is known to only occur in response to DNA stimuli and not to poly(I:C), suggesting this increased mobility of STING for both stimuli is not related to the canonical pathway.
We next turned to a different super resolution approach, Single-Molecule Fluorescence Recovery After Photobleaching (smFRAP) microscopy, that enables the tracking of individual NETs as they diffuse along the INM and ONM of the NE 42,43,44 . The nuclear pool of STING-GFP redistributed out of the nucleus upon stimulation of the cells with poly(I:C) or dsDNA (Fig. 2f). A similar number of STING-GFP molecules were in the inner versus outer nuclear membranes (as was observed by immunogold electron microscopy) in unstimulated cells, while the ratio of outer nuclear membrane to inner nuclear membrane signals more than doubled in the poly(I:C) and dsDNA ( Fig. 2f and h) stimulated cells.
Measuring the diffusion coefficient of this mobile STING revealed a near doubling of the speed of particles from 0.27 to 0.48 µm 2 /s in the poly(I:C) treated cells and to 0.49 µm 2 /s in dsDNA treated cells ( Fig. 2g and 's strong association with the INM and nuclear lamina, NEs   were first isolated from HEK293T cells transiently expressing STING-GFP and then treated   with a reversible cross-linker. Cross-linking chased STING-GFP into complexes between 130 and 300 kDa that could be reverted to the expected 70 kDa upon reversal of the cross-linking with DTT ( Fig. 3a and b). Cross-linked NEs were fragmented by sonication, immunoprecipitated (IP'd), cross-links reversed, and putative partners identified by tandem mass spectrometry (Table S1). The proteins that co-IP'd in the STING-crosslinked NEs were weighted for likely abundance based on spectral counts and plotted based on Gene Ontology (GO)-biological process terms. This revealed an enrichment in proteins with GO-terms for chromatin/chromosome organization and RNA/DNA binding compared to their representation among all proteins encoded by the genome (Fig. 3c). Plotting the normalized spectral abundance in the STING-GFP sample compared to mock transfected cells revealed the most abundant of the enriched co-IP proteins to be histone H1 variants (Fig. 3d) followed ( Fig. 3e) by a mixture of known NE proteins (e.g. Lamin A, LAP2), nucleotide-binding proteins (e.g. snRNP70, UBTF, RPS27a), and bromodomain proteins (e.g. Brd2, Brd3, Rbmx) that could mediate the reported STING function in chromatin compaction 28 and other epigenetic changes associated with IIR 45 (Table 1). Many proteins in all these categories bind DNA/RNA and nearly half of all STING partners identified are listed as nucleotide-binding proteins (Fig. 3f). Strikingly, although some known STING interactors were identified in the NE-STING proteome (DDX41 46 and CCDC47 47 ) many well-known interactors such as TBK1 and MAVS were not found, suggesting that the NE-STING proteome differs significantly from that of STING localized in the ER.
Downstream of STING ER/Golgi functions, IRF3/7 transcription factors induce IFN and other IIR genes in the nucleus. Therefore, we wondered if some of these STING NE co-IP partners have known interactions with IRF3/7 and may modulate immune signaling cascades. Accordingly we searched the HPRD interactome database 48 using Cytoscape 49 , finding that IRF3/7 had no known direct interactions with any of the putative STING partners. However, six known direct IRF3/7-binding partners interact directly with 17 of the proteins identified in the STING-NE co-IP (Fig. 3g). Of these, 12 are RNA-binding proteins (dark blue). The rest, as well as some of the RNA binding proteins, have also been reported to bind DNA. Although these proteins have not been previously shown to affect IRF3/7 transcriptional responses in IIR, several interact with viral proteins and affect viral replication, so may contribute to host cell IIR. For example, DDX5 is bound by the N(pro) protease of pestivirus 50,51 and may inhibit hepatitis C virus replication 52 and vesicular stomatitis virus triggered IFNβ induction 53 , although it appears to be a positive regulator of HIV-1 54 , and Japanese encephalitis virus (JEV) 55 among others 56 . Meanwhile, the hepatitis B virus HBx protein alters the intracellular distribution of RPS27a 57 , AATF is specifically targeted by an HIV-encoded miRNA 58 , and SYNCRIP is involved in hepatitis C virus replication 59 and mouse hepatitis virus RNA synthesis 60 . Therefore, we postulated that proteins identified in the NE STING co-IP experiment could contribute to IIR signaling and the potential links to IRF3/7 transcription factors suggested a signaling network through which STING might influence IIR from the NE. STING NE co-IP partners contribute to IIR. To test whether putative partner proteins identified in the NE STING co-IP experiment are involved in dsDNA triggered IIR, we used a dual-luciferase reporter system in combination with siRNA mediated knockdown of 7 partner proteins with links to IRF3/7 ( Fig. 4a) to test for effects on expression of an IFNβ promoterdriven reporter or a reporter activated by NFκB binding. The NFκB-and IFNβ-luciferase reporters are activated upon co-transfection of STING and cyclic GMP-AMP synthase (cGAS) (Fig. 4b). cGAS produces a second messenger (cGAMP) that is bound by STING during IIR 8 and HEK293FT cells were used because they do not express cGAS ( Supplementary Fig. S1e) so that the only source was the transfected plasmid. The cells were also co-transfected with a Renilla luciferase reporter under a thymidine kinase promoter to allow for normalization of transfection efficiency and cell number. Using this assay siRNA knockdown of MEN1, DDX5, snRNP70, and RPS27a all caused a statistically significant drop in IFNβ promoter driven luciferase expression (Fig. 4c), while SYNCRIP, MEN1, DDX5, snRNP70, RPS27a and AATF all exhibited a statistically significant drop in NFκB activated luciferase expression (Fig. 4d). This suggests that these putative STING partner proteins can themselves contribute to IIR. It is interesting that SYNCRIP and AATF were more restricted in only being able to affect luciferase expression from the NFκB driven reporter.
To further confirm the role of these STING NE co-IP partners in IIR, independent of the luciferase assay system we measured transcripts of IFNβ with the various knockdowns in  Supplementary Fig. S2a)). siRNA knockdown of STING partners SYNCRIP, MEN1, and SNRNP70 did not affect STING or cGAS protein levels as determined by Western blot (Fig. 5a and b). However, knockdown of DDX5 caused a modest reduction in STING protein levels, suggesting that either DDX5 is required for STING stability or an off target effect of the siRNA used. As expected, STING knockdown strongly reduced the amount of IFNβ induction upon plasmid DNA but not poly(I:C) stimulation of IIR ( Fig. 5c-e). SYNCRIP, MEN1 and snRNP70 knockdown all reduced IFNβ induction by more than 50% ( Fig. 5c and d) an effect that was more marked at 8h post-DNA transfection.
Surprisingly, DDX5 knockdown significantly enhanced IFNβ induction with both DNA and poly(I:C) immune stimulation. This effect is especially interesting given that DDX5 siRNA treatment caused a reduction in STING protein levels and suggests that DDX5 indirectly functions as a negative regulator of both DNA and RNA triggered IIR. Surprisingly, despite that several of these STING NE coIP partners are RNA-binding proteins, when IIR was induced with poly(I:C) the other proteins tested did not have a significant effect on IFNβ induction suggesting that they specifically modulate IFNβ induction during DNA triggered immune responses (Fig. 5e).
Given the potential links between NE STING partners and IRF3/7 and the effects on IFNβ induction we decided to look at whether their knockdown affected IRF3 activation as measured by IRF3 phosphorylation. IRF3 phosphorylation upon treatment with plasmid DNA was reduced when STING or its NE co-IP partners were knocked down compared to cells treated with control siRNA (Fig. 5f). In contrast, with poly(I:C) treatment there were no obvious effects on IRF3 phosphorylation in cells knocked down for STING and MEN1; however, it was significantly enhanced in cells knocked down for DDX5 and snRNP70, while also slightly increased in cells knocked down for SYNCRIP although not at a statistically significant level (Fig. 5f). As another measure of IRF3 activation, cells treated with siRNAs against STING and partner proteins were assayed for accumulation of IRF3 in the nucleus by microscopy. In agreement with the reduction in phosphorylated IRF3 seen in cells treated with siRNAs against STING and partners following immune stimulation with dsDNA, the percentage of cells positive for accumulation of IRF3 in the nucleus was reduced in all conditions compared to cells treated with a control siRNA ( Fig. 5g and Supplementary Fig.   S3b). Nuclear accumulation of NFκB (RelA) was also tested, revealing that in response to dsDNA treatment RelA is only weakly activated (phosphorylated RelA accumulates in the nucleus) in HT1080 cells ( Fig. 5g and h). In contrast treatment of knockdown cells with poly(I:C) led to a robust activation of IRF3 and NFκB, as determined by the accumulation of NFκB (p65) in the nucleus ( Supplementary Fig. S3c).

STING NE co-IP partner SYNCRIP is antiviral against influenza A virus.
Although none of the STING co-IP partners tested were found to have negative effects on poly(I:C) stimulated IFN expression (Fig. 5e), we decided to test whether they might play a role in IIR against an RNA virus, since STING may function in IIR triggered by RNA virus infection in a manner independent of IFN induction 7,13,22 . Following siRNA mediated knockdown of STING NE partners, HT1080 cells were infected with the nuclear replicating RNA virus, influenza A virus (IAV). Knockdown of SYNCRIP resulted in significantly higher viral titers as determined by plaque assays, both at low and high multiplicity of infection (MOI) ( Fig. 6a and b). This effect was stronger for a mutant IAV (PR8 -N81 61 ) which expresses an NS1 protein with a deletion of the effector domain and so is less able to antagonise host IIR (Fig.   6c). To determine whether SYNCRIP expression is altered during viral infection, cell lysates were harvested at multiple time points during infection and blotted for SYNCRIP, revealing no obvious difference in SYNCRIP protein levels during infection (confirmed by presence of viral proteins NP and NS1) compared to mock infected cells (Fig. 6d). STING knockdown has previously been shown to affect IAV replication 18, 62 and we replicated this here in HT1080 cells. Knockdown of STING resulted in significantly higher viral titers compared to cells treated with a control siRNA (Fig. 6e). Further, we have found that the phenotypic effects of SYNCRIP knockdown replicate those of STING knockdown.

Discussion
This study identified several proteins that co-IP with the NE pool of STING and found that several of these putative STING partners can themselves contribute to IIR.
Specifically, we identified SYNCRIP, MEN1, snRNP70, RPS27a, DDX5, and AATF as novel modulators of IIR. Furthermore, this study directly shows for the first time that part of the endogenous NE pool of STING is present in the INM and that this pool becomes more mobile and redistributes during IIR triggered by both dsDNA and dsRNA stimuli.
In the best characterized STING pathway, recognition of cytoplasmic dsDNA by cGAS triggers cGAMP association with STING to promote its activation. Activated STING dimers translocate from the ER to the Golgi where they accumulate in perinuclear aggregates 6,7,8,63,64,65 . From here STING dimers oligomerize, inducing TANK-binding kinase 1 (TBK1) activation which trans phosphorylates itself and neighboring STING dimers 15,66,67 , leading to the recruitment and activation of IRF3 and eventually the induction of type-I IFNs and pro-inflammatory cytokines. However, STING translocation to or from the nucleus in IIR was previously unknown. Here we have shown for the first time that endogenous STING is clearly present in the INM and that this NE STING pool increases mobility and translocates from the INM to ONM upon IIR activation with the dsRNA mimic poly(I:C) or dsDNA. In light of recent reports that most cGAS is in the nucleus 3, 29, 31, 34 , this raises the possibility that INM STING could be activated before ER STING following detection of nuclear-localized viral dsDNA. Indeed, our finding that STING mobility in the NE increases during infection with the dsDNA nuclear-replicating virus HSV-1 would support this notion. This finding also has implications for a recent report of cGASindependent activation of STING following detection of DNA damage, in which the authors propose a non-canonical signaling complex composed of STING, TRAF6, IFI16, and p53 that forms in response to DNA damage sensed by PARP1 and ATM, and initiates an NFκB dominated transcriptional response 25 . It is possible that such a signaling complex forms in the nucleus given that the authors of this study reported no redistribution of ER resident STING to perinuclear foci following the induction of DNA damage by etoposide. It is interesting in this regard that one of the more abundant STING NE coIP hits was the DNA damage response protein PARP1 (see Supplementary Table S1). Although this was not included as a top hit because there were only twice as many spectra in the NET23 sample as in the mock when our cutoff was 3-fold, this and other proteins identified in the proteomics further support a role of nuclear STING functioning to sense nuclear DNA damage to induce immune responses in cancer.
Furthermore, we find that the dsRNA mimic poly(I:C) increases STING mobility in the NE and similarly promotes its redistribution from the INM to the ONM. STING protection against RNA viruses seems to function in a different pathway from the wellcharacterized dsDNA route since STING does not redistribute from the ER to Golgi perinuclear aggregates with poly(I:C) treatment. STING also is reported to not directly bind RNA or poly(I:C) 68 . It was recently reported that STING restricts the replication of RNA viruses through a proposed mechanism dependent on the cytosolic dsRNA sensor RIG-I, and due to a general inhibition of translation independent of PKR and translocon functions 13 .
Several of the STING NE partners we identified here could potentially mediate STING effects on translation (e.g. RPS27a, SYNCRIP, snRNP70); however, it also is possible that these partners could provide specific recognition of different RNA viruses and thus serve to provide a variety of novel IIR nucleotide sensors or adaptors. Interestingly, despite having no effect on poly(I:C)-mediated interferon expression, we find SYNCRIP to play a role in antagonizing IAV. Whether this is through a general translation effect, or indirectly through the canonical cGAS-STING pathway stimulated by mitochondrial stress and DNA leakage as reported for other RNA viruses over the years 69,70 , remains to be determined.
That the INM STING pool can mobilize to translocate to the ONM through the peripheral channels of the NPC may reflect a backup mechanism to signal IIR responses using the peripheral NPC channels when viruses inhibit central channel transport. Viruses often target the central channel of the NPCs to either block transport or to usurp it so that virus transcripts are preferentially transported over host-directed transport 71,72,73,74 , but the peripheral channels are normally used for membrane protein transport 38,42,43,75 so that STING as a multi-spanning transmembrane protein could bypass this block to signal IIR. This does not preclude the well-established STING signaling cascades from the ER/Golgi compartment normally using the central channel of the NPC -indeed IRF3 is known to translocate through the NPC central channel 76,77 , but our findings of increased STING mobility and nucleo-cytoplasmic shuttling during IIR through the peripheral NPC channels suggest that STING may provide a backup system for activating IIR when central channel transport is disrupted. Moreover, the 17 STING NE co-immunoprecipitation partners identified here that are known to themselves interact with six IRF3/7 partners could potentially contribute to such IIR activation, enhancing STING functions through a multiply redundant backup system.
The identification of these putative STING partners through the reverse-crosslinking approach does not necessarily mean they directly bind STING and confirming such interactions may require finding particular conditions for each. Moreover, the increasing complexity of STING interactions and its multiple pathways for activating IIR make confirmation of STING's involvement in their contributions to IIR difficult. Nonetheless, these new putative STING partners clearly can contribute to IIR from the measures shown here; indeed, several reports in the literature further suggest these proteins play roles in IIR to different viruses when re-evalutated in light of our results. MEN1 binds and represses the activity of the AP1 transcription factor JunD 78 and the related cJun is an IIR activator 79 , possibly explaining its functioning in IIR. MEN1 was also recently found to affect promoter fidelity at the interferon-gamma inducible IRF1 gene 80 .
Interestingly, this function of MEN1 involves its functioning in a complex with the major histone K4 methyltransferase, MLL1, which was also identified as a STING NE co IP partner. STING interaction with this methyltransferase complex could also contribute to the other reported nuclear function for STING in chromatin compaction 28 . In addition, several bromodomain proteins identified as putative STING partners here (e.g. BRD2, BRD3) could also explain this chromatin compaction function or, more excitingly, chromatin remodeling reported to occur in IIR 45 . Furthermore MEN1 is a tumor suppressor 81 and as such could contribute to reported STING roles in DNA damage sensing in cancer 3,29,31,34,82,83 . A recent study showed that MEN1 depletion results in mis-regulation of the p53 pathway leading to increased levels of chromosomal instability and accumulation of DNA damage 83 .
The STING-NE co-IP partners that bind RNA also have several previously reported functions that would be consistent with their ability to support IIR indicated here. For example, DDX5 is targeted by the N(pro) protease of pestivirus, presumably to counter host antiviral defenses 50,51 . At the same time, DDX5 can be a negative regulator of IFN responses.
Our finding that DDX5 knockdown increases type-I IFN expression following DNA or poly(I:C) transfection and IRF3 phosphorylation following poly(I:C) transfection is consistent with a recent report that DDX5 suppressed IFN responses triggered by VSV infection 84 . This study also showed that DDX5 knockdown increased IRF3 phosphorylation, but without testing whether DDX5 knockdown influences IRF3 phosphorylation triggered by a DNA ligand as we do. Our findings here that DDX5 knockdown reduces dsDNA-induced IRF3 activation while elevating IFNβ expression might appear contradictory without the context of these other studies suggesting it can be both pro and anti IIR. Regardless, these results strongly suggest that DDX5 may contribute a regulatory function to IIR. While this study has focused on characterising potential IIR activity of the specific binding partners highlighted for having upstream effects on IRF3/7 transcription factors, it is notable that many more of the top STING NE coIP partners bind RNA and some have previously been shown to mediate IIR. One of these is DDX23 that is a dsRNA sensor recently reported to pair with TRIF or MAVS to mediate IIR 85 .
Other links to viral infection for these newly identified STING partners include the hepatitis B virus HBx protein that alters the intracellular distribution of RPS27a 57 . Also, STING-NE coIP partner AATF is specifically targeted by HIV to impair cellular responses to infection 58 . SYNCRIP/hnRNPQ interestingly facilitates hepatitis virus replication 59 , suggesting an alternate pathway where STING sequestration of this factor might provide another avenue towards host protection from the virus. SYNCRIP was separately reported to interact with the IAV NS1 protein 86 , a major antagonist of the host cell immune response, suggesting the virus may target SYNCRIP due to its positive immune functions. These many RNA-binding partners would provide a highly redundant backup system so that knockout of any single one would only moderately impact on IIR, consistent with the moderate but significant reduction in IIR signaling observed for SYNCRIP knockdown. Our data are consistent with the notion that STING plays a wider role in signaling than its initial description as an adaptor in cytosolic DNA sensing, initiating different responses based on diverse inputs from DNA damage 25 to RNA virus infection 13 . The wide range of STING nuclear partners combined with its ability to translocate out of the nucleus with treatments that activate IIR potentially provides a valuable redundancy and novel mechanism for STING functions that could better elucidate how it protects cells against both RNA and DNA viruses.  Image data was processed using Image-Pro Premier (Media Cybernetics Inc., MD, USA). Background and photobleach corrections were engaged using an algorithm written by D.A.K. according to 89 . A macro was written in VB.Net within Image Pro Premier whereby a region of interest (ROI) was applied to the bleach spot, background and non-bleached area of a nearby cell and corrected for movement automatically compared to the 5 pre-bleach images.

Methods
The t½s were calculated from the normalized fluorescence values.

Single-Molecule Fluorescence Recovery After Photobleaching (smFRAP) microscopy.
Imaging was performed using an Olympus IX81 equipped with a 1.

Mass spectrometry analysis. An LTQ-Orbitrap mass spectrometer (Thermofisher Scientific)
was coupled on-line to an Agilent 1100 binary nanopump and an HTC PAL autosampler (CTC). The peptides were separated using an analytical column with a self-assembled particle frit 94   Sigma Aldrich (UK) Dual-luciferase reporter assay. Following siRNA knockdown transfection in 6-well plates as described for HT1080 cells, 5x10 4

Reporting summary. Further information on research design is available in the Nature
Research Reporting Summary linked to this article.

Data availability
All mass spectrometry data will be deposited to ProteomeXchange Consortium and MassIVE databases upon acceptance of the manuscript.

Acknowledgements
We thank Ravi Badwe for assistance in nuclear envelope preparations from transfected