UBXN9 governs GLUT4-mediated spatial confinement of RIG-I-like receptors and signaling

The cytoplasmic RIG-I-like receptors (RLRs) recognize viral RNA and initiate innate antiviral immunity. RLR signaling also triggers glycolytic reprogramming through glucose transporters (GLUTs), whose role in antiviral immunity is elusive. Here, we unveil that insulin-responsive GLUT4 inhibits RLR signaling independently of glucose uptake in adipose and muscle tissues. At steady state, GLUT4 is docked at the Golgi matrix by ubiquitin regulatory X domain 9 (UBXN9, TUG). Following RNA virus infection, GLUT4 is released and translocated to the cell surface where it spatially segregates a significant pool of cytosolic RLRs, preventing them from activating IFN-β responses. UBXN9 deletion prompts constitutive GLUT4 trafficking, sequestration of RLRs, and attenuation of antiviral immunity, whereas GLUT4 deletion heightens RLR signaling. Notably, reduced GLUT4 expression is uniquely associated with human inflammatory myopathies characterized by hyperactive interferon responses. Overall, our results demonstrate a noncanonical UBXN9-GLUT4 axis that controls antiviral immunity via plasma membrane tethering of cytosolic RLRs.


INTRODUCTION
The retinoic acid inducible gene I (RIG-I)-like receptors (RLRs), comprised of RIG-I, melanoma differentiation-associated protein 5 (MDA5) and Laboratory of Genetics and Physiology 2 (LGP2), are a family of cytosolic pattern recognition receptors (PRRs) that sense viral RNA and initiate an early antiviral immune response to numerous RNA viruses.Following the sensing of distinct viral RNA structures, RIG-I and MDA5 utilize the conserved mitochondrial antiviral-signaling protein (MAVS) to initiate a signaling cascade leading to the activation of transcription factors IRF3 and NF-κB and subsequent transcription of type I/III interferon (IFN-I/III) and proin ammatory cytokine genes 1 .
However, abnormal activation of RLRs is associated with autoimmune diseases and interferonopathies such as Type-I diabetes 2,3 , Aicardi-Goutières syndrome and dermatomyositis 4,5 , emphasizing that RLRs must be strictly regulated for discrimination between pathogenic RNA and unnecessary sterile responses.To date, post-translational modi cations (PTMs) of the receptors themselves or the downstream signaling proteins are the predominant mechanisms by which RLRs are controlled.Whether additional regulatory strategies exist upstream of RLR activation has not been thoroughly explored.
In addition to cytokine production, activation of PRRs leads to metabolic reprogramming characterized by enhanced glucose in ux, glycolysis, disruption of the tricarboxylic acid (TCA) cycle and accumulation of metabolites 6 .Similarly, RNA viruses (sensed by RLRs) also promote glucose uptake 7,8 , where metabolites feedback from glycolysis 9 , the pentose phosphate pathway 10 , the hexosamine biosynthesis pathway 11 and the TCA cycle 12 to intersect at MAVS to suppress RLR signaling.This rewiring of cellular metabolism hinges on glucose transporters (GLUTs), which belong to the solute carrier (SLC) family of nutrient transporters 13 .The relationship between GLUTs and immune function have been best described for GLUT1, which facilitates extensive glycolytic reprogramming in innate immune cells following Toll-like receptor (TLR) activation 14,15 .However, despite their importance to prompting glucose metabolism, the role of other GLUTs in innate immune responses is entirely unknown.Further, whether metabolites or glycolytic proteins regulate RLRs directly remains an intensive area of study.GLUT4 is the predominant, insulin/muscle contraction-regulated glucose transporter in adipose, skeletal, and cardiac muscle tissues, and is essential for the maintenance of whole-body glucose homeostasis 16 .In the basal state, GLUT4 is primarily sequestered within intracellular GLUT4 storage vesicles (GSVs), which are tethered to the Golgi matrix by a direct interaction with the Golgi-associated ubiquitin regulatory X (UBX) domain-containing protein UBXN9 (also known as TUG) 17,18 .Insulin triggers sitespeci c endoproteolytic cleavage of UBXN9 to release GSVs for transport to the plasma membrane; thus, deletion of UBXN9 increases constitutive tra cking and surface residence of GLUT4, resembling the acute insulin effect on GLUT4 tra cking 19 .Therefore, UBXN9-mediated GLUT4 tethering is the dominant mechanism for ensuring appropriate insulin signaling.Intriguingly, whether GLUT4 has any other biological function is entirely unknown.
Here, we uncover a new role of the UBXN9-GLUT4 axis in the regulation of RLR signaling, independent of GLUT4's canonical function as a glucose transporter.We nd that, in contrast to the current dogma stating PRR effector responses are supported by GLUTs, GLUT4 speci cally sequesters RLRs to the plasma membrane, thereby spatially segregating them from downstream signaling components and curtailing the primary IFN response.On the other hand, UBXN9 tethers GLUT4 to the Golgi matrix, thus maintaining normal RLR signaling.This study reveals GLUT4 as a negative regulator of RLR signaling and establishes the plasma membrane-mediated partitioning of RLRs as a strategy to ensure signaling delity of cytoplasmic RNA sensors.Critically, this novel mechanism could operate in autoimmune diseases, where GLUT4 is reduced in patients with autoin ammatory myopathies, coinciding with overexpression of RLRs and interferon stimulated genes (ISGs).Hence, new therapies for autoin ammatory myopathies targeting the UBXN9-GLUT4-RLR axis could prevent damaging in ammation.

UBXN9 promotes RLR signaling and inhibits RNA virus infection
To understand the function of the UBXN9-GLUT4 axis in regulating RLR signaling, we utilized the mouse myoblast cell line, C2C12, which recapitulates many aspects of primary mouse myotube biology 20 .
To address whether UBXN9 protects against infection with muscle tropic RNA viruses sensed by RLRs, we utilized encephalomyocarditis virus (EMCV, which speci cally stimulates MDA5) and arthritogenic O'nyong nyong virus (ONNV, which activates both RIG-I and MDA5) 22,23,24 .Consistent with compromised RLR signaling, UBXN9-de cient myocytes had higher EMCV titers and produced less IFN-β protein than their WT counterparts (Fig. 1f-h).We observed similar results with ONNV infection (Supplemental Fig. 1e, f).To validate the above-described in vitro studies, we utilized a mouse model of EMCV infection since EMCV has a tropism for the heart-which contains GLUT4 + cardiomyocytes-and is sensitive to type I IFNs 23 (Fig. 1i).Our recent work has demonstrated EMCV replication was ~ 100-fold higher in the heart than that in other tissues on day 4 post infection (p.i.) 25 .Critically, Ubxn9 −/− mice were more susceptible to EMCV-induced lethality, accompanied by higher viral loads in the blood and heart relative to Ubxn9 +/+ mice (Fig. 1j-m).Accordingly, knockout mice exhibited lower serum IFN-β concentrations (Fig. 1n).In an ONNV infection model, viral load in the skeletal muscle was relatively higher in Ubxn9 −/− mice compared to WT mice (Supplemental Fig. 1g, h).Collectively, these results argue that UBXN9 positively regulates RLR sensing of diverse RNA viruses in vitro and in mice.
In addition to skeletal muscle, GLUT4 is also the dominant GLUT expressed in cardiac tissue 27 .To investigate the conservation of GLUT4-mediated RLR suppression in cardiac muscle cells, we differentiated cardiomyocytes from human induced pluripotent stem cells (hiPSCs-CMs) based on the modulation of Wnt signaling 28 , silenced GLUT4 using siRNA and treated them with 3p-hpRNA or EMCV to activate RLRs (Supplemental Fig. 5a).hiPSC-CMs expressed the cardiac troponin T differentiation marker and GLUT4 mRNA was signi cantly reduced as compared to nontargeting control siRNA (Fig. 2g, Supplemental Fig. 5b).Notably, the knockdown of GLUT4 signi cantly decreased EMCV replication and augmented the IFN-β response when compared to control-treated hiPSC-CMs (Fig. 2h, i).Further, GLUT4 also suppressed RIG-I signaling as the stimulation of siSLC2A4 hiPSC-CMs with 3p-hpRNA led to an elevated IFN-β response compared to siCtrl cells (Fig. 2j).Finally, we tested whether RLR responses are blunted by GLUT4 in primary myocytes derived from Slc2a4 / MCK Cre+ mice where GLUT4 expression is ablated by muscle creatine kinase (MCK) synthesis in skeletal and cardiac muscle 29,30 (Fig. 2k).This permitted the direct assessment of GLUT4's putative proviral role in muscle cells without the systemic metabolic disturbances accompanied by GLUT4 deletion in vivo 30,31 .Consistent with C2C12 cells and iPSC-CMs, GLUT4 deletion reduced EMCV replication and heightened IFN-β responses at 48 hours postinfection (Fig. 2l, m).These data suggest that GLUT4 suppression of RLRs is a highly conserved mechanism in mouse and human muscle cells.
GLUT4-mediated RLR suppression is uncoupled from glucose in ux and lactate accumulation Until recently, glycolysis and RLR signaling pathways were thought to occur independently.It has now been reported that lactate-one of the chief by-products of glycolysis-is an inhibitor of RLR signaling by disrupting MAVS aggregation 9, , and MAVS itself coordinates glucose ux through the pentosephosphate pathway (PPP) and the hexosamine biosynthesis pathway (HBP) 10 .Based on the premise that UBXN9 tethers GLUT4 intracellularly, we reasoned that upon UBXN9 de ciency, GLUT4 would constitutively transport glucose resulting in increased intracellular/circulating levels of lactate, which then inhibits RLR signaling (Fig. 3a).In support of this possibility, deletion of lactate dehydrogenase A (Ldha −/− )-the end-stage glycolytic enzyme that catalyzes the conversion of pyruvate to lactateincreased IFN-β expression and reduced lactate release (Supplemental Fig. 6a-d).Moreover, global Ubxn9 −/− mice exhibited reduced glucose concentrations during fasting and after intraperitoneal injection of glucose, compared to Ubxn9 +/+ counterparts, consistent with the muscle-speci c Ubxn9 deletion mouse model 19 (Fig. 3b, Supplemental Fig. 6e, f).
To rst address this hypothesis, Ubxn9 +/+ and Ubxn9 −/− mice were fasted before EMCV infection, and lactate levels monitored throughout the course of acute disease.Despite higher lactate levels in Ubxn9 −/ − mice before infection, these levels equilibrated between WT and KO animals by 3 days post infection (Fig. 3c).Comparable trends in lactate were also noted in skeletal muscle cells (Fig. 3d).To rule out a direct role for lactate in our system, we overexpressed GLUT4 in Ldha +/+ and Ldha −/− cells and treated them with 3p-hpRNA.Albeit signi cantly lower lactate, GLUT4 suppressed RLR responses equally in Ldha −/− cells as those in WT control cells (Fig. 3e).Aligning with these results, silencing UBXN9, which effectively redistributes GLUT4 to the plasma membrane, markedly diminished IFN-β release from Ldha +/+ and Ldha −/− cells, compared with siCtrl cells (Fig. 3f).While we consistently observed LDHA indirectly contributes (e.g., catabolism of pyruvate) to suppression of IFN-β responses, our knockdown data argues that GLUT4 is upstream of lactate inhibition in the hierarchy of RLR regulation.
As EMCV infection promoted glucose in ux in Ubxn9 +/+ myocytes to levels equivalent in Ubxn9 −/− cells before and after infection (Fig. 3g), we postulated that EMCV activates UBXN9-controlled GLUT4 tra cking and glucose uptake.Confocal microscopy revealed GLUT4 translocation to the plasma membrane increased following EMCV infection in WT cells, whereas it remained constitutively elevated on the surface in Ubxn9 −/− myocytes and unaffected by EMCV infection, mirroring the kinetics of glucose uptake in these cells (Fig. 3g-i).The localization of GLUT4 on the surface was con rmed by using insulin treatment (positive control, Supplemental Fig. 7a, d), Slc2a4 −/− myocytes (negative control, Fig. 3h, i, Supplemental Fig. 7c, d) and the plasma membrane marker ZO-1 (Supplemental Fig. 7e).Consistent with these ndings, siRNA knockdown of Slc2a4 (GLUT4) in Ubxn9 −/− cells fully rescued IFN-β expression to WT levels after 3p-hpRNA treatment (Fig. 3j).Thus, these data reinforced the hypothesis that UBXN9 positively regulates RLR signaling through GLUT4 (Fig. 3a, f, j) and the activation of RLRs (via 3p-hpRNA or VSV/EMCV infection) can trigger GLUT4 exocytosis (Supplemental Fig. 7f).Notably, GLUT4-de cient myocytes continuously produced more IFN-β protein than their Slc2a4 +/+ counterparts in all sugar conditions albeit lactate levels remaining comparable between the two cell types (Fig. 3k, l).Lastly, blockage of glucose uptake with Fasentin 32 failed to rescue the Ifnb1 levels in EMCV-infected Ubxn9 −/− cells (Supplemental Fig. 6g, h).In total, these ndings propose that the activation of RLRs induces GLUT4 translocation and UBXN9 strengthens the IFN-β response in a GLUT4-dependent manner that is uncoupled from glucose in ux and lactate production.

GLUT4 tethers RLRs to the plasma membrane
We demonstrated that TBK1 and IRF3 phosphorylation was affected in Ubxn9 −/− or Slc2a4 −/− cells (Fig. 1d, e, Supplemental Fig. 4e, f), suggesting the UBXN9-GLUT4 axis targets a step at or upstream of TBK1.To pinpoint the mechanism of GLUT4-regulated RLR signaling, we assessed the direct RLR-MAVS interaction as well as MAVS lament formation-the two immediate upstream events that are necessary for TBK1 activation-from WT and Ubxn9 −/− myocytes after 3p-hpRNA stimulation.Compared to WT control cells, Ubxn9 −/− myocytes had reduced binding of RIG-I to MAVS, which compromised MAVS oligomerization following 3p-hpRNA treatment (Fig. 3m, n).These data suggested that UBXN9-GLUT4 targets the earliest step of RLR pathways.
As RLR signaling was suppressed when UBXN9 was deleted, we hypothesized that relocation of GLUT4 to the plasma membrane could negatively impact RLR signaling independent of glycolysis.To overcome the limitation of intracellular cytoskeleton staining with endogenous GLUT4 antibodies (Fig. 3h, Supplemental Fig. 7a-e) 33 , we utilized the 3T3-L1 broblast/adipocyte cell line with a stably expressed myc-GLUT4-GFP reporter 34 that recapitulate the tra cking kinetics of endogenous GLUT4 in response to insulin (Fig. 4a).Further, these cells i) model adipocytes as important sources of IFNs during dysglycemia 35 , and ii) are sensitive to RLR ligands 36 .At steady state, GLUT4 was concentrated near the nucleus-consistent with its sequestration in GSVs at the Golgi/ERGIC 34 -but acute insulin treatment (t = 10min) rapidly mobilized GLUT4 to the cell membrane, which suppressed the 3p-hpRNA-induced IFN-β response when compared with vehicle-treated cells (Fig. 4a, b).Moreover, blocking glucose uptake with Fasentin 32 did not rescue the insulin-mediated suppression (Fig. 4b).Although this data supports the postulate that GLUT4 suppresses RLR signaling independently of glucose transport (Fig. 3k, Supplemental Fig. 6g, h), it is unclear if GLUT4 physically impedes RIG-I.
As UBXN9-GLUT4 regulates RLR signaling likely at the level of receptors (Fig. 3m, n), we sought to visualize the spatial distribution of RIG-I when GLUT4 was mobilized to the cell surface.To focus on the primary RLR response, we chose an early timepoint (6hr after RLR activation) before RLR expression is ampli ed by the JAK-STAT1/2 pathway.Physiologically, RLR signaling is indispensable for the control of RNA viruses at the early stage of infection.Insulin or 3p-hpRNA (to promote GLUT4 translocation) induced signi cantly more RIG-I into the plasma membrane compartment (containing caveolin, a speci c marker of the plasma membrane) than cells treated with vehicle alone (Fig. 4c, (Supplemental Fig. 7a-c,  f).Accordingly, the pool of RIG-I and GLUT4 associated with the plasma membrane was enhanced in Slc2a4 +/+ cells after 3p-hpRNA treatment, whereas RIG-I was nearly undetectable in this fraction from Slc2a4 −/− myocytes, suggesting the relocation of RIG-I to the plasma membrane is dependent on GLUT4 (Fig. 4d).
RIG-I expression in skeletal muscle is very low 37 .To address if GLUT4 could hinder the rapidity of the primary IFN response through RLR tethering, we fractionated the crude mitochondria (marked by MAVS) and cytosolic fractions (marked by tubulin) from myocytes and assessed the spatial residency of RIG-I before and after 3p-hpRNA treatment 38 (Fig. 4e).Indeed, the cytosolic and crude mitochondrial pools of RIG-I were increased in Slc2a4 −/− cells (lane 7-8) as compared to their WT counterparts, which had signi cantly lower abundance before and after RLR stimulation (Fig. 4f, red star, lanes 1-2, 5-6).In the absence of treatment, deletion of GLUT4 also augmented the quantity of RIG-I in the mitochondrial fraction (Fig. 4f), consistent with previous reports of low, but detectable RIG-I in the crude mitochondria at steady state 38 .These data suggest that GLUT4 abundancy in myocytes is su cient to signi cantly deplete the cytosolic RIG-I pool accessible to MAVS before and after activation.These ndings were further recapitulated at early timepoints in skeletal muscle cells and 3T3-L1 myc-GLUT4-GFP adipocytes (Supplemental Fig. 7h).Aligning with these fractionation results, the RIG-I-MAVS interaction and MAVS lament formation were enhanced in Slc2a4 −/− cells only after 3p-hpRNA treatment relative to control myocytes (Fig. 4g, h).Confocal microscopy further showed RIG-I juxtaposed plasma membrane signals in Slc2a4 +/+ cells, whereas bright, segregated RIG-I puncta were observed in the perinuclear area of Slc2a4 −/− myocytes, consistent with active signaling and a magni ed pool of cytosolic RIG-I that was untethered by GLUT4 39 (Fig. 4f, i, Supplemental Fig. 7g, h).Measuring the distance of RIG-I to the plasma membrane indicated that deletion of GLUT4 signi cantly dissociated these two uorescent signals from a near overlap in Slc2a4 +/+ (mean = ~ 1.8µm) (Fig. 4i-k).In total, these ndings, in conjunction with the induction of IFN-β (Fig. 2), demonstrate that the GLUT4 governance of RIG-I redistribution to the plasma membrane profoundly in uences the primary RLR signaling program.
The magnitude of GLUT4 mobilization is directly proportional to the total UBXN9-GLUT4 complexes that are dissociated during insulin treatment 16 .The results in Fig. 3 indicated deletion of UBXN9 accelerates GLUT4 tra cking (Fig. 3h, Supplemental Fig. 7b, d).We further tested the speci city of RIG-I relocation by tracking the cellular localization and distribution of GLUT4 and RLRs in tandem during virus infection.Despite RIG-I being the primary sensor of VSV, both RIG-I and MDA5 increased simultaneously with GLUT4 in the plasma membrane following VSV infection in WT cells, indicating viral RNA-induced conformational changes in these sensors is not a prerequisite for GLUT4-mediated tethering (Fig. 4l).In contrast, RLRs tra cked in tandem with the constitutively translocated pool of GLUT4 in Ubxn9 −/− cells, which was virtually eliminated from the plasma membrane when GLUT4 was deleted (Fig. 4l).The synchronized translocation of RLRs with GLUT4 mirrored the kinetics of glucose uptake in WT, Ubxn9 −/− and Slc2a4 −/− cells before and after insulin treatment (Supplemental Fig. 7i, j).Overall, these results support our hypothesis that, within the hierarchy of RLR regulation, GLUT4-mediated redistribution of RIG-I to the plasma membrane is a dominant mechanism in skeletal muscle and adipocytes.

GLUT4 binds RLRs and translocates them to the plasma membrane to curb RLR signaling
The above ndings suggest that tra cking of GLUT4 to the plasma membrane suppresses RLR signaling by repositioning RIG-I away from MAVS.Mobilization of GSVs involves the vesicular tra cking machinery as well as numerous accessory proteins in close association with GLUT4 16,40 .Considering other protein(s) of anterograde transport could be responsible for RLR tethering, we rst examined the interaction between GLUT4 and RLRs under the same conditions known to promote GLUT4 translocation (Figs. 3 and 4, Supplemental Fig. 7).Using 3T3-L1 myc-GLUT4-GFP adipocytes, line tracing demonstrated a shift in GLUT4 and RIG-I signals from a uniform distribution at steady state (Vehicle), to the cell periphery after insulin treatment (Fig. 5a, b).Moreover, the GLUT4-RIG-I interaction strengthened over time upon stimulation with insulin (in adipocytes and C2C12 myocytes), 3p-hpRNA and virus infection in WT myocytes (Fig. 5c-e).Intriguingly, even in the face of constitutive GLUT4 tra cking and RLR tethering (Fig. 4l, Fig. 3h, Supplemental Fig. 7b, i, j), 3p-hpRNA and virus infection continued to induce stronger binding between GLUT4 and RLRs in Ubxn9 −/− cells (Fig. 5d, e).Confocal microscopy con rmed extensive colocalization of endogenous GLUT4 with RIG-I on the plasma membrane after VSV infection in Ubxn9 −/− as compared to WT cells (Fig. 5f, g, inset images).Such subcellular relocation of RIG-I to the membrane was dependent on GLUT4 as the brightest RIG-I signals were concentrated near the perinuclear region and failed to colocalize with GLUT4 or the plasma membrane in Slc2a4 −/− cells (Fig. 5f, g).Lastly, we detected equivalent RLR tethering by ectopic GLUT4 in Ldha +/+ and Ldha −/− cells after 3p-hpRNA treatment, indicating this mechanism operates upstream of lactate inhibition (Supplemental Fig. 8a, b).Thus, GLUT4 directly sequesters RLRs during translocation to the plasma membrane, induced by either exogenous stimuli or after UBXN9 deletion All GLUT proteins possess 12 transmembrane segments which facilitate their incorporation into the plasma membrane 13 (Fig. 5h).GLUT4 has a 24 amino acid (aa) long cytoplasmic N-terminus, a large cytoplasmic loop (Loop 6, ~ 65 aa) that is bound by UBXN9 to sequester it at the Golgi/ ERGIC 34 , and a 42 aa-long cytoplasmic C terminus.The remaining residues are in short, 8-12 aa intracellular loops (Loop 2, 4, 8 and 10).Therefore, given RLRs reside in the cytoplasm, we reasoned that the N-, C-terminus or Loop 6 of GLUT4 might mediate its interaction with RLRs.We generated three FLAG-tagged deletion mutant constructs of GLUT4: ΔN24, ΔC42 and ΔL6, and included a point mutation, Arg169Ala (R169A), that disrupts glucose uptake by > 70% 41 (Fig. 5h).Although the R169A mutant failed to rescue glucose uptake relative to Slc2a4 −/− cells reconstituted with WT GLUT4, it phenocopied the RIG-I binding potential of the WT protein (Supplemental Fig. 8d, e).These data reinforced the hypothesis that the transport activity of GLUT4 is decoupled from its tethering of RLRs to the plasma membrane.All the GLUT4 mutants were expressed except for ΔN24, which was likely unstable due to misfolding, among many other possibilities (Fig. 5i).Importantly, the ΔC42 and ΔL6 mutants were unable to bind RIG-I compared to FL and R169A GLUT4 (Fig. 5i).Further co-immunoprecipitation from these constructs revealed deletion of loop 6 compromised the UBXN9 interaction (Fig. 5i), which is the competitive binding site between RIG-I and UBXN9 (Supplemental Fig. 8f).Functionally, reconstitution of Slc2a4 −/− cells with the ΔC42 and ΔL6 mutants failed to inhibit RLR signaling to the levels in Slc2a4 +/+ cells, whereas FL WT and R169A GLUT4 inhibited 3p-hpRNA-induced IFN-β expression (Fig. 5j).These ndings indicated that the Cterminus and Loop 6 of GLUT4 are necessary for binding and suppressing RLRs.
Finally, to address where GLUT4 bound to RIG-I, we generated four domain mutants of RIG-I (FLAGtagged): caspase activation and recruitment domain (CARD), helicase, ΔCARD and the C-terminal domain (CTD) and assessed their reciprocal binding e ciencies to GLUT4 (Myc-tagged) (Fig. 5k).The FL and RIG-I CARD domain alone strongly interacted with GLUT4, whereas the helicase and the CTD failed to do so (Fig. 5l, Supplemental Fig. 8c).Similarly, the MDA5 FL protein and CARD domain was su cient for GLUT4 binding (Fig. 5m).Importantly, failure of the ΔCARD mutants to bind GLUT4 established this domain as the putative binding site between RLRs and GLUT4 (Fig. 5l, m).In summary, GLUT4 directly sequesters RIG-I and MDA5 to the plasma membrane during tra cking and blunts the primary activation of RLRs by interacting with the CARD domain.

Promotion of UBXN9 cleavage and GSV release underlie RLR sequestration during virus infection
Virus-induced metabolic reprogramming can be attributed to, in part, manipulation of upstream enzymes in various glycolytic pathways 7,8,42 .Our results suggest that in addition to RNA virus infection (Fig. 3h,  4l), acute insulin signaling can promote RLR tethering (Fig. 4c, 5a-d).Therefore, we focused on the conserved upstream nodes in the PI3K-AKT, AMPK and c-Cbl-TC10a GLUT4 tra cking pathways that may contribute to RLR sequestration (Fig. 6a).We detected an acute increase in AKT phosphorylation at 3-6 hpi that returned to the baseline shortly thereafter (Fig. 6b).On the other hand, AMPK activation was gradual and sustained throughout the course of RNA virus infection (Fig. 6b).Importantly, microtubulebased movement of GSVs was activated re ected by the phosphorylation of AS160 (Tbc1D4) (Fig. 6b).The c-Cbl pathway culminates in cleavage of UBXN9 to liberate GSVs from the Golgi matrix-accordingly, and consistent with endoproteolytic processing of UBXN9 after insulin treatment 19,43 , virus infection led to a noticeable increase in c-Cbl phosphorylation and the production of UBXN9 C-terminal cleavage products (Fig. 6c).Thus, virus infection coordinates GLUT4 translocation through the activation of vesicle tra cking and cleavage of UBXN9.
We further tested whether the activation of GLUT4 pathways was synchronized with the relocation of RLRs to the surface during virus infection.Indeed, GLUT4 and RLRs were transported to the plasma membrane in unison, although the speed at which GLUT4 tra cked differed depending on the RNA virus (Fig. 6d).Agreeing with our fractionation results (Fig. 4f), the peak of RLR sequestration (lane 4, VSV; lane 3, EMCV) was preceded by the rapid relocation of steady state RIG-I and MDA5 (lane 1-3, VSV; lane 1-2, EMCV) (Fig. 6d).As the infection progressed, even in the presence of more GLUT4, a signi cant drop in RLR surface abundance was evident at later timepoints (lane 5, VSV; lane 4, EMCV), A similar phenomena was also noted in 3T3-L1 myc-GLUT4-GFP reporter cells as early as 60 minutes after insulin stimulation (Supplemental Fig. 9a), suggesting con nement is a transient event and RLRs are eventually relinquished from GLUT4-mediated tethering.
As AKT regulates GLUT4 tra cking, we hypothesized that the relocation of RLRs during virus infection is dependent on AKT.We next genetically deleted the dominant Akt isoform in skeletal muscle, Akt2 44 , and evaluated GLUT4 tra cking after RNA virus infection (Supplemental Fig. 9b).In line with previous reports 44 , Akt2 −/− cells had comparable translocation of GLUT4 to the surface after VSV and EMCV infections, relative to Akt2 +/+ controls (Fig. 6e, Supplemental Fig. 9c).Loss of Akt2 may have been compensated for by the sustained activation of AMPK (Fig. 6b), which phosphorylates similar AS160 residues as AKT 45 .
Despite equivalent GLUT4 tra cking, Akt2 −/− cells were unable to mobilize RIG-I and MDA5 to the plasma membrane fraction, while WT cells showed a consistent increase in RLR abundance after virus infection (Fig. 6e).Moreover, Akt2 −/− had elevated RLR expression before and after VSV infection relative to those in the WT cells; these heightened ISGs in Akt2 −/− likely afforded better protection from VSV infection (Fig. 6e).In total, these data revealed AKT2 is dispensable for GLUT4 translocation yet necessary for sequestration of RLRs to the plasma membrane.

Myopathic diseases are associated with decreased GLUT4 expressions and elevated interferon signatures in skeletal muscle
The experiments above revealed a non-canonical function of GLUT4 in the relocation of RLRs, which perturbs signal transduction and antiviral immune responses.To test the biomedical impact of the putative tethering mechanism of GLUT4, we analyzed two datasets of skeletal muscle tissue biopsied and sequenced directly from patients with critical illness myopathy (CIM) (PRJNA491748) or dermatomyositis (DM) (GSE143323) (Fig. 7a, Supplemental Fig. 10a) 46,47 .Idiopathic in ammatory myopathies are autoimmune conditions characterized by muscle in ammation and weakness 48 , enriched in ammatory gene signatures, and an upregulation of RIG-I that is now considered a biomarker of DM 49 .To rst investigate the pertinence of GLUT4 in these autoimmune disorders, we analyzed the pathways enriched in DM patients compared to healthy skeletal muscle using QIAGEN ingenuity pathway analysis (IPA) 46 .Of the 382 pathways upregulated in DM muscle, IFN-related pathways were among the most enriched including ISGylation Signaling, Interferon Signaling and RIG-I Receptors in Antiviral Immunity (Fig. 7b).Consistent with previous reports, we also observed Pathogen Induced Cytokine Storm Signaling-comprised of in ammatory and interferon genes-was the most signi cantly activated pathway in myopathic patients by p-value and Z-score 50 .Similar in ammatory pathways were also enriched in the muscle of CIM patients compared to controls (Supplementary Fig. 10b; red dots).Of note, Oxidative Phosphorylation, Glycolysis, TCA cycle, and AMPK Signaling pathways were downregulated in DM patients, indicating a general dysregulation of muscle metabolism (Fig. 7b).Further RNA-seq analysis of the same muscle comparing the log 2 fold change in genes (FDR < 0.05) revealed similar patterns of exaggerated interferon/in ammatory transcripts in the primary (IFIH1, IRF3) and secondary (STAT1, MX1, ISG15) RLR response in myopathic patients (Fig. 7c).These results are consistent with previously published datasets reporting heightened interferon pathways in myopathic skeletal muscle 47,51,52 .Unexpectedly, SLC2A4 (encoding GLUT4), and several genes involved in GLUT4 tra cking 16 , were downregulated in DM muscle relative to controls (Fig. 7c, Supplemental Fig. 10d).
Since several GLUTs could contribute to the dysregulation of metabolic pathways in myopathic muscle, we next compared the expression of all the detected SLC2A genes in DM patients.Notably, only SLC2A4 transcripts were signi cantly decreased in myopathic muscle (Fig. 7d, dotted box), whereas several other SLC2A transporters and UBXN9 were increased, potentially re ecting a compensation in glucose homeostasis in DM (Fig. 7d, e).Further, SLC2A4 represented the only DEG of the SLC2A transporter family in CIM patients (Supplemental Fig. 10c).Linear regression analysis of relative GLUT4 expression with ISGs on a patient-by-patient basis revealed that GLUT4 inversely correlated with IFIH1, IFITM1, DDX58 and OAS1A (Fig. 7f, Supplemental Fig. 10e).Pearson's correlation analysis indicated GLUT4 was the only transporter that negatively correlated with a common set of IFN genes upregulated in both datasets (Fig. 7g).Importantly, neither GLUT1 expression from DM patients (the other major glucose transporter in muscle) nor GLUT4 from healthy muscle demonstrated such a relationship (Fig. 7f, g, Supplemental Fig. 10e).Collectively, these data highlight a key relationship between lower GLUT4 levels and heightened IFN signatures in patients with various myopathic diseases.

DISCUSSION
GLUT4 is the dominant, insulin/muscle contraction-regulated glucose transporter in adipose and muscle tissues, and indispensable for the maintenance of whole-body glucose homeostasis.Despite this, and the previous literature that GLUTs support immune activation, GLUT4 has not been thoroughly investigated in the context of innate immunity.Here, we demonstrate that GLUT4 inhibits the major cytosolic viral RNA-sensing pathways, RIG-I and MDA5, by con ning RLRs to the plasma membrane, displacing them from their downstream adaptor MAVS and attenuating the primary antiviral type I IFN response.This conclusion is further supported by studies with UBXN9, an important tether of GLUT4.
Deletion of UBXN9 leads to continuous tra cking of GLUT4 to the plasma membrane (comparable to the effect of acute insulin stimulation) 17,18 , together with redistribution of the resident pool of RLRs to the surface, and suppression of antiviral immune responses during RNA virus infection.Mechanistically, GLUT4-mediated tethering of RLRs was mediated by binding of the large cytoplasmic loop (L6) and Cterminus of GLUT4 to the CARD domain of RLRs.The ΔC42 and ΔL6 mutants also failed to restore normal signaling to Slc2a4 −/− cells, suggesting both domains contribute to the sequestration and inhibition of RLRs.These results are concordant with the idea that only UBXN9-free GLUT4 on the plasma membrane is able to tether RLRs as UBXN9 occupies the Loop 6 of GLUT4 at the Golgi matrix 18 .Thus, the compartmentalization of RNA sensors by translocated GLUT4 represents an important event upstream of RLR activation and an additional regulatory mechanism of innate immune signaling.
A growing body of evidence has highlighted that SLCs can have, beyond their established roles as transporters, functions related to the tra cking and signaling of innate immune receptors 53 , now termed "transceptors" (transporter-receptor).For example, the histidine-peptide cotransporter SLC15A4 controls TLR7,9-induced cytokine responses independent of its ligand-binding potential 54 .Moreover, the transporter activity of GLUT3 is dispensable for the coordination of IL-4/STAT6 signaling in M2 macrophage polarization 55 .Our work aligns with this transceptor model, suggesting a transporterindependent role of GLUT4 in RLR signaling.This proposal is substantiated by four pieces of evidence: (1) mutation of the glucose-binding residue (R169A) interacted equally with RLRs as compared to WT GLUT4, (2) pharmacological inhibition of glucose uptake failed to rescue the GLUT4-mediated suppression of IFN production in adipocytes and skeletal muscle, (3) glucose-free culture conditions did not alter the Slc2a4 −/− phenotype during RLR signaling, and (4) deletion of GLUT4 did not abolish steady state glucose uptake, but effectively depleted the insulin-and virus-responsive pool critical for relocating RLRs to the surface.Although glucose uptake by GLUT4 is not involved in RLR signaling, we and others 9,10 have demonstrated RNA virus infection can promote glucose in ux, providing a rapid source of ATP and raw materials for virus replication 56 .With these data in mind, our study cannot exclude the role of other immunomodulating metabolites, such as itaconate (and its derivatives), which feedback on STAT1 to downregulate IFN-β responses 57 .This metabolic change is generated during the shift to glycolysis and therefore warrants future investigation during RLR activation.
Investigation into virus hijacking of GLUT4 tra cking revealed the activation of AKT, AMPK and c-Cbl, aligning with numerous reports that viruses exploit upstream enzymes in glucose sensing pathways to support replication 8, 58 42 .However, our work highlights the apparent dichotomy between GLUT4 translocation and RLR con nement.While the proper delivery of GLUT4 to the surface is organized by mechanisms unique to adipocytes and skeletal muscle cells-including the phosphorylation of AS160 by AKT and AMPK 45,59 -RLR tethering was nearly abrogated in Akt2 −/− cells despite intact GLUT4 translocation.These data argue that GLUT4 is necessary but not su cient for RLR sequestration, and AKT2 may directly catalyze this interaction.Evidence for this postulate further stems from our studies in Ubxn9 −/− cells whereby constitutive GLUT4 tra cking continued to strengthen its interaction with RLRs following subsequent AKT activation (e.g., insulin, virus infection).In fact, the eventual untethering of RLRs from GLUT4 was synchronized with the return of AKT to its pre-infection phosphorylation state.
Given the con icting reports of AKT2's role for regulating GLUT4 translocation in vivo 44 , and its suppression of 14-3-3 chaperones necessary for IFN-β responses 60,61,62 , future studies aim to address the relative contribution of the two proposed functions for AKT2, i.e., conventional GLUT4 tra cking and the control of RLR signaling transduction.
The transient tethering of RLRs to the plasma membrane further coincided with cessation of IFN signaling as re ected by an overall decrease in RLR expression.Akin to canonical regulation of RLRs in the cytosol, post-translational modi cations-such as ubiquitination-also terminate signaling of surface receptors by degrading signalosome components or mislocalizing sorting adapters from their subcellular site of signal transduction 63 .Indeed, c-Cbl, which was phosphorylated after RNA virus infection herein, conjugated K48-linked polyubiquitination of RIG-I near the plasma membrane 58 .As degradative polyubiquitin chains are linked at later stages of infection 64 , sequestered RLRs may present an opportunity to couple compartmentalization mechanisms with degradative pathways.
Stringent regulation of PRRs is necessary to avoid recognition of self-ligands that elicit autoin ammatory conditions 4 5 .One such mechanism is compartmentalization, which is best exempli ed by shielding the ligand binding domains of endosomal-bound TLRs from cytoplasmic contents 65 .Of note, compartmentalization is also applicable to the cytosolic DNA sensor, cGAS 66,67 .Our data offers further support of this idea, as GLUT4 attenuated the primary stage of signaling by sequestering the pool of preexisting (i.e., steady state and 3-6hpi) and de novo cytosolic RLRs (i.e., JAK/STAT-induced ISG) away from MAVS.These results align with recent reports that RLRs are proximally regulated by the actin cytoskeleton 68,69 .Notably, tethering RLRs is reliant on the cleavage of UBXN9 from GLUT4 and subsequent tra cking of GSVs, which is constitutive at a low level but induced following virus infection.
As GSVs contain approximately ve GLUT4 molecules per vesicle 43,70 , viral-induced UBXN9 cleavage further increases the likelihood of a GLUT4-RLR interaction.Furthermore, generation of this UBXN9 Cterminal cleavage product stimulates thermogenesis 19 and promotes a feed-forward circuit that could account for increased body temperature (i.e., fever) when the IFN-β response is attenuated by GLUT4.Beyond an immune evasion strategy, it is tempting to speculate upon the potential protection that is afforded by compartmentalizing RNA sensors to the plasma membrane.Under such a model is where our results may be the most relevant, as GLUT4 and its associated tra cking genes (e.g., TBC1D4/AS160) are signi cantly downregulated in the muscle of dermatomyositis patients presenting with exaggerated interferon signatures.Overall, our data indicates that RLRs are poised for activation by virus infection when GLUT4 is largely sequestered by UBXN9.Progression of the response leads to the release of GLUT4 from UBXN9, which redistributes RLRs to the surface and attenuates RLR signaling and antiviral immunity (Fig. 8).
In sum, our study identi es spatial con nement as a dominant strategy to control RLR signal transduction in skeletal muscle and adipocytes and unveils a previously unknown glucose-independent function of GLUT4 in the regulation of innate immunity.Future studies are necessary to elucidate the molecular details of the GLUT4-RLR interaction at the plasma membrane, speci cally how AKT2 promotes the sequestration of RLRs, when GLUT4 tra cking intervenes with signal transduction and the role of RLR signaling in metabolic and autoin ammatory diseases.

Mouse models
All animal procedures were approved by the Institutional Animal Care and Use Committee at UConn Health adhering to federal and state laws.Mice with the exon 5 of Ubxn9 anked by loxP sites (Ubxn9 ox/ ox ) were generated via homologous recombination.The homozygous Ubxn9 ox/ ox were then crossed with homozygous tamoxifen-inducible Cre recombinase-estrogen receptor T2 mice (The Jackson Laboratory, Stock # 008463; Rosa26 Cre-ERT2 ) to generate male and female Cre +/-Ubxn9 littermates, which were mated to produce Cre +/-Ubxn9b / , Ubxn9b / , and Cre +/-.To induce global Ubxn9 knockout in ≥ 5-week-old mice (Ubxn9 -/-), 1mg of tamoxifen was administered (dissolved in corn oil) to each mouse every other day for a total of 5 injections.ERT2-Cre +/-Ubxn9 / treated with corn oil served as the wild-type control (Ubxn9 +/+ ).Tamoxifen was allowed to be cleared 10 days after the last injection before experimentation.Both male and female mice were used between 7-18 weeks of age.Genotyping was performed with genomic DNA and Choice Taq Blue Mastermix (Denville Scienti c, Cat# CB4065-8) using the following PCR protocol: 95°C for 1 s, 34 cycles of 94°C for 1 min, 60°C for 30 s, 72°C for 30 s, and then 72°C for 7 min, 4°C to stop.The genotyping primers for the loxP sites in exon 5 were: Ubxn9 F 5'GCTTCTCTCAAAGCTGGAGAGTCAC; R 5' CAAGGCACTGGGCCAGGGAG.The PCR reaction resulted in a product of 226bp (WT) and/or 276 bp (loxP).The Cre primers were common: WT F 5'-AAGGGAGCTGCAGTGGAGTA; WT R 5'-CCGAAAATCTGTGGGAAGTC; mutant R 5'-CGGTTATTCAACTTGCACCA.The PCR reaction resulted in a product of 297 bp (WT) and/or 450bp (Cre).
The GLUT4 LoxP mice (Slc2a4 ox/ ox ) were generated as previously described 71 and provided by Carol A. Witczak (Indiana University).Slc2a4 / mice were then bread with muscle creatine kinase Cre recombinase transgenic mice (The Jackson Laboratory, strain # 006475; MCK-Cre + ).Heterozygote mice (Slc2a4 /+ -MCK Cre ) were used for breeding to generate the following genotypes: wild-type (Slc2a4 +/+ ), GLUT4 LoxP HET (Slc2a4 /+ ), LoxP homozygous (Slc2a4 / ), MCK-Cre+ (control), muscle-speci c GLUT4 heterozygous (Slc2a4 /+ MCK Cre ) and muscle-speci c GLUT4 knockout (KO; Slc2a4 / MCK Cre ).For primary myocyte studies, Slc2a4 / MCK Cre (KO) and MCK-Cre+ (WT) mice were used at 12-weeks of age.For genotyping, the PCR protocol for GLUT4 LoxP was 95 o C for 3min, 35   Isolation and culture of primary skeletal myoblasts and bone-marrow derived macrophages (BMDMs) Primary skeletal myoblasts were isolated from Ubxn9 +/+ and Ubxn9 −/− mouse quadricep muscle tissue, puri ed and differentiated as previously described with minor adjustments 73 .Quadriceps (vastus medialis and vastus lateralis) were excised from two hindlegs of mice and placed in 1X phosphate-buffered saline (PBS) for washing.Muscle was then nely minced into small pieces, transferred to a 15ml conical tube, and spun down at 21,130 x g for 30s at RT to collect muscle pieces.Muscle was digested in Collagenase II (Cat #: 17101015, Gibco™) digestion media (400 U/ml in DMEM -FBS, sterile ltered) on a shaker at 100 rpm for 1.5 hours with vortexing halfway through digestion time.After shaking, vortex for 15s and muscle mix should appear cloudy.Tubes were then spun down at 1,400 x g for 5 min at RT, at which time supernatant was removed and muscle was resuspended in complete DMEM to neutralize collagenase II.To release myoblasts from muscle, tissue was pipetted up and down 30X using a sterile 10ml pipette and then strained over a 70µm strainer on a 50 ml tube.Cells were collected and strained through a 40µm lter to remove macrophages.Tubes were spun down at 1,400 x g for 5 min at RT and myoblast pellet resuspended in myoblast growth medium (MGM: DMEM/F12 + 1% p/s, 20% FBS, 10ng/ml basic broblast growth factor).Cells were seeded on plates precoated with 10% Matrigel (Cat #: 354234, Corning™).

Reagents and antibodies
For puri cation of a pure myoblast culture, digested muscle/cells were allowed to attach to Matrigelcoated plates for 72h without media removal.Three-days later, small, droplet-shaped myoblasts and larger broblasts should be visualized.Here, cells are ready for rst pre-plating step.MGM medium was removed, and cells washed 2X with MGM to remove lingering cell debris and trypsinized to detach cells.
Cells were centrifuged at 800 x g for 5 min and plated in 6-well plates that were not coated with Matrigel for 1h.This step is critical to remove contaminating broblasts as these will attach to the uncoated surface and myoblasts will remain in suspension.After 1h, broblasts were visualized on plate surfaces and supernatant containing myoblasts was removed and added to a new 10% Matrigel-coated plate.
These puri cation steps were performed every 48h until > 98% myoblasts were obtained (typically ~ 3 pre-plate steps are necessary).After a pure culture is achieved, cells are ready for experimentation, or alternatively, they were differentiated into myotubes using differentiation medium (DM: DMEM + 2% horse serum) once cells reached ~ 80-90% con uency.DM medium was changed every 48h until myoblasts align and fusion begins (~ 120h).At this point, cells were ready for experimentation.Skeletal muscle myoblasts were isolated from Slc2a4 +/+ and Slc2a4 −/− using the same protocol as above.

Human iPSC culture and cardiomyocyte differentiation
Human iPSCs were generated as previously described 75 .Brie y, the broblasts were derived from discarded female neonatal skin tissue under Yale Institutional Review Board approval.CytoTune TM -iPSC 2.0 Sendai Reprogramming Kit (ThermoFisher Scienti c, A16517) was used to reprogram broblast cells into iPSCs (Y6-iPSC).Twenty-four hours after viral transduction, infected cells were harvested and plated onto mitotically arrested MEF feeder layer with human iPSC medium (20% KSOR in DMEM/F12 medium supplemented with 10ng/ml bFGF), 1% non-essential amino acid (v/v), 2 mM L-Glutamine, 0.44 µM beta-Mercaptoethaol, and 1% Pen/Strep (v/v) (all from ThermoFisher Scienti c, USA).The medium was changed every other day for three weeks.Colonies of iPSCs were then picked and expanded on MEF feeder layers for cardiac differentiation.

Mouse infection and disease monitoring
Ubxn9 +/+ and Ubxn9 −/− mice were intraperitoneally injected with 100 PFU of EMCV and morbidity and mortality were monitored twice a day for survival studies.Prior to infection, mice were fasted overnight to collect baseline IFN-β, glucose, lactate, and viral RNA levels (considered cutoff for the level of detection by qPCR).Mice were then allowed to feed ad libitum following infection.IFN-β in the plasma was analyzed at 24 h post-infection (hpi) from mice injected with 1000 PFU EMCV.Viral titers in plasma and heart tissue at 3 days post-infection (dpi) and 4dpi were assessed by qPCR and plaque assay, respectively.Lactate concentrations were quanti ed from the same samples used for qPCR of viral RNA and quanti ed using a Lactate-Glo Assay (Promega, Madison, WI, Cat # J5021).

Cell culture and viruses
EMCV (Cat # VR129-B) and VSV (Indiana Strain, Cat # VR-1238) were purchased from American Type Culture Collection (ATCC) (Manassas, VA) and the multiplicity of infection was speci ed in each gure legend.ONNV UgMP30 strain (NR-51661) was provided by BEI Resources.Green uorescence protein (GFP)-VSV was made by inserting a VSV-G/GFP fusion sequence between the VSV G and L genes and propagated in our lab for use in several studies 25,74,76 .These viruses were propagated in Vero cells and titrated by a plaque forming assay.
Differentiation was induced on the rst day with DMEM containing 0.25 µM dexamethasone, 160 nM insulin, and 500 µM methylisobutylxanthine, followed by feeding cells with fresh medium containing 160nM insulin for two more days.On day four, cells were maintained in DMEM + 10% FBS until 8 days.
Experiments were then performed on differentiated adipocytes ~ 8-10 days after process was initiated.MD, Cat # SR416343).In brief, three unique targeting siRNA's for Slc2a4 ( nal concentration: 50nM) were mixed with TransIT-X2 in 1X Opti-MEM for 20 min before transfecting mixture dropwise onto cells.After 48h of incubation to achieve knockdown, cells were then treated with 3p-hpRNA as described above.A universal, non-speci c scramble siRNA was used as a non-targeting control.The same protocol was adopted for knockdown of Ubxn9 (Origene, Rockville, MD, Cat # SR417176).For iPSC-CMs, knockdown of human GLUT4 using two unique siRNA's ( nal concentration: 50nM) were prepared based on the user manual (Origene, Rockville, MD, Cat # SR304402) and transfected as above.
For viral infection, viruses were diluted in DMEM without FBS and allowed to attach and infect cells for 2h; the cells were then washed with 1X phosphate-buffered saline (PBS) once and incubated with fresh medium.The MOI and infection time were denoted in each experimental gure legend.

Plaque-forming assay
Quanti cation of infectious EMCV (Cat# VR-129B) viral particles in heart tissue/cell culture supernatant was performed on Vero cell monolayer with minor modi cations 74 .Brie y, heart tissue was weighed and 15mg of heart was digested in PBS using a Bio-Gen PRO200 Homogenizer (Pro Scienti c, Oxford, CT, Cat # 01-01200).A total of 30-100 µg (total proteins) of tissue lysate or cell supernatant was serially diluted in DMEM (-) FBS and applied to con uent Vero cells (12-well plate) at 37°C for 2h.The inoculum was then removed and replaced with 2 ml of DMEM complete medium with 1% SeaPlaque agarose (Lonza, Cat# 50100).Plates were inverted at 37°C, 5% CO 2 and plaques visualized using Neutral red (Sigma-Aldrich) after 24h of incubation.Viral titers were expressed as plaque forming units (PFU) / mL or gram of tissue.

Plasmid construction and molecular cloning
The pB retrovirus vector containing GLUT4myc7-GFP (herein referred to as pGLUT4-GFP) was used in our studies previously 77 .This plasmid was overexpressed in C2C12 using similar transfection methods described previously.The human full length (FL) and R169A point mutant GLUT4 were provided by Chuangye Yan at Tsinghua University 41 .The FL (NCBI accession: NM_001042) and R169A variant were cloned into a new, custom pcDNA3.1-FLAGvector 74 for expression as an N-terminal FLAG-fusion protein using standard PCR ampli cation and cloning techniques.The pcDNA3.1-FLAGvector was also used to generate FLAG-UBXN9 76 .To generate GLUT4 truncations on the same plasmid backbone, the FL GLUT4

Generation of gene knockout cell lines with CRISPR-Cas9 technology
Pre-designed, gene speci c guide (g) RNAs (Integrated DNA Technologies, Coralville, IA) were subcloned into a lentiCRISPR-v2 vector 78 and correct insertion was con rmed by sequencing.To generate lentiviral particles, each gRNA vector was transfected into HEK293T cells with the packaging plasmids pCMV-VSV-G and psPAX2 (Didier Trono lab via Addgene, Watertown, MA, Cat # 12259).After 24h, half of the cell culture medium was replaced with DMEM, and viral particles were collected at 48-72h after transfection.The viral supernatant was cleared by brief centrifugation at 2000 rpm for 5min.C2C12 target cells were then seeded at ~ 50% and transduced the next day with each individual lentivirus.The WT control was lentiCRISPRv2 vector only.C2C12 cells were then selected with 0.8 µg/mL of puromycin for 10-20 days, changing puromycin media every other day.Successful knockout clones were con rmed by western blotting.Guides targeting mouse UBXN9 (F-CACCGATGAAGTGCTACGACCCCGT; R-5' AAACACGGGGTCGTAGCACTTCATC) GLUT4 (F-5' CACCGAGGCACCCTCACTACGCTCT; R-5' AAACAGAGCGTAGTGAGGGTGCCTC), LDHA (F-5' CACCGTGTTCACGTTTCGCTGGACC; R-5' AAACGGTCCAGCGAAACGTGAACAC), and AKT2 (F-5' CACCGTCACAAAGCATAGGCGGTCA; R-5' AAACTGACCGCCTATGCTTTGTGAC) were used in this study.
Puri cation of total cellular RNA, reverse transcription and quantitative (q) PCR Approximately 15mg of mouse tissue, ~ 20 µL of whole blood and up to 1x 10 6 culture cells were collected in 300 µL of lysis buffer (RNApure Tissue & Cell Kit, CoWin Biosciences, Cambridge MA, United States).Heart tissue was homogenized as described, mixed with lysis buffer, and extracted according to the product manual.Isolated RNA was quanti ed using a spectrometer feature on BioTek Cytation 1 imaging reader (Agilent, Santa Clara, CA) and RNA concentration was normalized according to the lowest concentration across all samples with RNase-free water.RNA samples were normalized and converted into cDNA using the PrimeScript™ Reverse Transcription (RT) reagent Kit (TaKaRa Bio, Inc, Cat# RR037A).Quantitative PCR (qPCR) was performed with gene-speci c primers and iTaq Universal SYBR Green Supermix (BioRad, Cat# 1725124).Results were calculated using the 2 -ΔΔCt method from the C T values for each sample and gene of interest.A housekeeping gene (Actb, Polr2b) was used as an internal control.The qPCR primers have been used and validated in our previous studies 74,79 .

Mouse glucose tolerance test (GTT)
Ubxn9 +/+ and Ubxn9 -/-mice were fasted overnight (~ 12h) followed by being administered with 1 mg glucose/kg body weight (Sigma, St. Louis, MO, Lot # 82H0725) via intraperitoneal (i.p.) injection.Tail vein blood at baseline and indicated time points after injection was collected and measured for glucose levels by using a handheld glucometer (Germaine Laboratories, San Antonio, TX).Glycemia was reported as mg/dl.

2-deoxyglucose (2-DG) and lactate assays
Cells were plated in 96-well clear bottom black plates one day before assay to allow equilibration in metabolism.To quantify 2-NBDG uptake following insulin treatment, protocol was followed according to manufacturer's instructions, with minor modi cations (Cayman Chemical, Ann Arbor, MI, Item # 600470).
For acute glucose uptake, cells were serum starved in no glucose (-FBS) DMEM medium for 3h before 50 µg/ml 2-NBDG was added to no glucose DMEM containing FBS and 200nM human insulin (Millipore Sigma, Darmstadt, Germany, Cat # 91077C) for 15-20 min.Supernatant was removed and cells were washed 2X with assay buffer to remove non-speci c binding.After second wash, cells were placed in assay buffer and intracellular 2-NBDG taken up by cells was detected with uorescent lters (485nm/535nm) on BioTek Cytation 1 imaging reader (Agilent, Santa Clara, CA).To account for background signal, no insulin treated cells were used as "steady state" glucose consumption and 2-NBDG in DMEM was added to empty wells without cells.Both controls were subtracted from insulin treated groups to calculate relative glucose uptake.For experiments of longer duration (3p-hpRNA, virus, IFN-β), 2-NBDG was added to DMEM without glucose with or without speci ed treatment and incubated with cells until timepoint indicated in gure legends.To account for cell growth/background, "Mock" cells were incubated with 2-NBDG in DMEM alone and assayed in tandem with treated cells.Cells were washed and assayed as described above for insulin treatment.For testing inhibition of glucose uptake, cells were pretreated for 1h with 50µM of Fasentin (R&D Systems, Minneapolis, MN, Cat # 6100/10) diluted in DMEM without glucose (or DMSO control).Media was then replaced containing fresh inhibitor with 100 µg/ml 2-NBDG and cells were incubated for 1h before assessing glucose uptake as described previously.
Protocol was followed exactly as suggested in product manual as each type of sample required unique processing.For intracellular lactate, cells were plated in a white 96-well clear bottom plates and treated as described in gure legends.After speci ed timepoints, supernatant was removed, and lysate used for determining intracellular lactate concentration.For mock groups, cells were incubated in media side-byside with experimental wells and processed simultaneously.Mouse plasma was diluted 100-fold and processed as for cell culture supernatant samples.Plates were measured using a BioTek Cytation 1 imaging reader (Agilent, Santa Clara, CA) with luminescent lters.DMEM + 10% FBS was used as background control for cell culture supernatant and plasma; 1X PBS used for intracellular lactate background.
Enzyme-linked immunosorbent assay (ELISA) and immunoblotting IFN-β in cell supernatants, homogenized tissue and plasma were assessed using LumiKine™ Xpress mIFN-β 2.0 ELISA detection kit according to manufacturer's instructions (InvivoGen, San Diego, CA, Cat #luex-mifnbv2).Human IFN-β in cell supernatants of iPSCs were assessed using DuoSet® ELISA kit according to the manufacturer's instructions (R&D Systems, Minneapolis, MN, Cat #DY814-05).The IFNβ data are presented as pg/ml.For western blotting analysis, cells were lysed in RIPA (Alfa Aesar, Tewksbury, MA, Cat # J63306) or for overexpression/detection of GLUT4, 2% n-Dodecyl-β-D-Maltopyranoside (Anatrace, Maumee, OH, Cat # D3101GM).For detection of GLUT4, samples were not boiled and lysed on a rotator for 2h at 4 o C before centrifuging at 21,000 x g for 10min at 4 o C. Samples were run on standard dodecyl sulfate-polyacrylamide gels and transferred onto nitrocellulose membranes using Mini Trans-Blot® Cell transfer systems (Biorad, Hercules, CA, Cat #1703930).Blots were then blocked with 5% milk in TBS-T and probed with appropriate primary and secondary antibodies.

Co-immunoprecipitation
HEK293T cells were transfected with GLUT4 expression plasmids or RIG-I/MDA5 mutant plasmids using the TransIT-X2 system as described above.Total cell lysates were prepared from transfected cells in lysis buffer containing [ 2% n-Dodecyl-β-D-Maltopyranoside (DDM), 25mM HEPES, 150mM NaCl and 1X protease inhibitors] by gentle scrapping and pipetting before transferring to a 1.5ml centrifuge tube.
Lysate was rotated at 4 o C for 2h and then centrifuged at 21,000 x g for 10min.Supernatant was removed and again rotated overnight with 20µl of pre-washed (25mM HEPES, 150mM NaCl) anti-FLAG magnetic beads at 4 o C. Co-immunoprecipitation was then preformed according to manufacturer's instructions (Sigma Aldrich, St. Louis, MO, Cat # M8823).IP elution was mixed with 4X SDS loading buffer.Samples were not boiled to maintain GLUT4 transmembrane conformation.For harder to transfect cells such as C2C12, GLUT4 overexpression was carried out in cell suspension.Brie y, cells were dislodged by trypsin digestion and pelleted by brief centrifugation at 800 x g for 5min.The cell pellet was then resuspended in transfection mix (DNA + TransIT-X2 + 1X Opti MEM) prepared as described above for 7-10 min with intermittent agitation at 37 o C. Prewarmed DMEM was then added and plated for further culture.FLAG immunoprecipitation then proceeded as above.
To maintain GLUT4 conformation, lysate was kept on ice for 30min with intermittent vortexing every 5min.Lysate was then centrifuged at 10,000 x g for 10min Target complexes were eluted in 100µl SDS-PAGE sample lysis buffer on a rotator for 20min.Beads were collected with a magnetic stand and remaining elution was boiled for 10min.To promote RIG-I: MAVS interactions, 3p-hpRNA was transfected into cells as described in "Ligand treatments" section for indicated times detailed in gure legends; reconstitution water diluted to the same ratio in transfection mixture was used for mock groups.

Endogenous MAVS aggregation
To detect MAVS oligomers, cells were seeded in 6-well plates at a density of ~ 1 x 10 6 per well in duplicate and treated as denoted in gure legends.After treatment, cells were washed with 500 µL cold PBS before being lysed in 2% DDM lysis buffer (n-Dodecyl-β-D-Maltopyranoside, 25mM HEPES, 150mM NaCl and 1X protease inhibitors] by gentle scrapping and pipetting.Lysate was then rotated at 4 o C for 2h and centrifuged at 21,000 x g for 10min to remove cell debris.Next, supernatants were transferred and mixed with 20 µL Native Sample Buffer (Bio-Rad #1610738) that was then loaded into a 4-16% NativePAGE™ (Invitrogen™ #BN1004BOX).The samples were then run in an Invitrogen NuPage® Novex® Gel System at 120 V for 1 h, 180 V for 30 min, 240 V for 30 min and then 300 V for 1 h in sequence (to maintain at 9 mA electricity).Gels were then transferred to nitrocellulose membranes using Mini Trans-Blot® Cell transfer systems (Biorad, Hercules, CA, Cat #1703930).Membranes were then blocked with 5% milk in TBS-T, probed with appropriate primary overnight and proteins detected with HRP-conjugated secondary antibodies.Protein bands were visualized with enhanced chemiluminescent (ECL) substrate (Lumigen, South eld, MI, Cat # TMA-100) on a ChemiDoc™ MP Imaging system (Biorad, Hercules, CA, Cat # 12003154).

Plasma membrane fractionation
The membrane fraction was isolated from GLUT4 + C2C12 myocytes and 3T3-L1 adipocytes as previously described, with modi cations according to experimental goals 77,80

Subcellular fractionation
Subcellular fractionation by differential centrifugation was performed as previously described 38 , with slight modi cations.Brie y, myotubes were washed 1X with PBS before adding sucrose homogenization buffer (250mM sucrose, 50mM Tris-HCl, 5mM MgCl 2 and 1X protease inhibitor cocktail) and scrapping cells off plates.Cellular fractions were kept on ice or 4 o C for the remaining steps.Cells were broken up with a Dounce homogenizer for 5min (or until cells were observed to be > 90% broken) and then the nuclei/unbroken cells were vortexed for 15sec and centrifuged at 800 x g for 15min.The pellet discarded and the supernatant was again centrifuged at 800 x g for 10min.The supernatant was retained and centrifuged at 11,000 x g for 10min to separate the cytosol fraction (supernatant) and crude mitochondria (pellet).Crude mitochondria pellets were resuspended in 200µl sucrose buffer, centrifuged again at 11,000 x g for 10min, and pellets resuspended in lysis buffer (50mM Tris-HCl, 1mM EDTA, 0.5% Triton-X100 and 1X protease inhibitor cocktail).Mitochondrial lysis was sonicated on ice 3X in 5 sec increments with 30sec pauses.The lysis buffer was labeled "mitochondrial fraction' and ready for SDS-PAGE.The cytosol fraction (containing cytosol and microsomes) were centrifuged at 100,000 x g for 1hr in an ultra-centrifuge.The cytosol in the supernatant was concentrated with 100% ice-cold acetone in -20 o C for 1hr while the microsome pellet was discarded.The cytosolic fraction was again centrifuged at 12,000 x for 5min, and the pellet containing precipitated cytosolic proteins was resuspended in lysis buffer containing SDS, 10% β-mercaptoethanol and 8M urea.

Immuno uorescence microscopy
Cells were cultured as experiment requested and treated as described in gure legends.All microscopic images were rst processed in the Zen 2.3 microscopy software (Zeiss Group, Jena, Germany) before further analysis in ImageJ (NIH).The same microscope instrument settings were used for all samples, and all images were analyzed using the same settings.A vector (white arrow) was drawn from the edge of the nucleus (DAPI) to the plasma membrane (CellBrite 555); the midpoint of the indexed arrow was then considered the perinuclear and peri-plasma membrane regions, respectively.The relocalization of RIG-I intensity (AF488) was calculated by the ratio of uorescence in the perinuclear to the peri-plasma membrane regions.Fluorescence intensities of all three markers were then traced along the vector and distance (in µm) of RIG-I to the plasma membrane was calculated.Colocalization of membrane and RIG-I uorescence was measured on a per-cell basis and Pearson's coe cient calculated using the JACoP Plugin in ImageJ.Mean uorescence intensities (MFI) for surface GLUT4 were calculated on a per-image basis and identical settings used for each image.In brief, images were adjusted using "Threshold" setting to x a cutoff value as to highlight regions with uorescence above the background intensity.Final pmGLUT4 MFI = MFI of cell-background MFI.Because GLUT4 antibody recognizes only exofacial GLUT4 and the cells were not permeabilized or xed before staining, detected uorescence was considered on the plasma membrane.
For immunostaining of iPSC-derived cardiomyocytes (iPSC-CMs), cells were xed with 4% paraformaldehyde in PBS for 12 min at RT, washed with PBS and permeabilized with 0.01% Triton X-100 for 20 min at RT. iPSC-CMs were then blocked with 10% goat serum in PBS for 30 min at RT and stained for cardiac troponin T (cTnT) and nuclei with DAPI (SouthernBiotech, Cat #0100 − 20).Confocal uorescent microscope (Leica; Multiphoton Microscope TCS SP8 MP) was used to image immunostained cardiomyocytes.
RNA sequencing analysis from patient datasets RNA sequencing datasets were obtained from PRJNA491748 (CIM study) and GSE143323 (DM study).
Myopathic diseases are associated with decreased GLUT4 expression and elevated interferon signatures in skeletal muscle.a, schematic overview of dermatomyositis (DM) study (GSE143323) from which the data in (b-f) are derived.N=39 DM and 20 healthy controls.b, Ingenuity Pathway Analysis (IPA) for pathways activated or suppressed in DM patients compared to healthy controls.The Log P value and Z-score represent signi cance and activity of pathway enrichments (z-score > 0, activated; z-score < 0, inhibited), respectively.The dot size represents the number of genes found enriched/suppressed compared to the whole gene set in that pathway (ratio).Highly upregulated pathways related to interferon signaling and highly downregulated pathways associated with metabolism are denoted in boxes and lled with their respective signature colors.c,volcano plot depicting differentially expressed genes (DEGs) (FDR <0.05 and Log2FC >0.5) between healthy controls and DM patients.DEGs with a fold change >0.5 are indicated in red; DEGs with a fold change <0.5 are in blue.Nonsigni cant DEGs are indicated in grey.FDR, false discovery rate.d, e, relative FPKM of SLC2 genes (GLUTs) (d) and UBXN9 (e) in DM patients compared to controls.Dotted red box highlights expression differences for SLC2A4 (GLUT4).f, simple linear regression analysis of SLC2A4 (GLUT4) or SLC2A1(GLUT1) FPKMs with IFIH1 and IFITM1 FPKMs from DM patients.g, representative heatmap demonstrating correlations between all SLC2A(GLUTs) and ISGs identi ed to be upregulated in both DM and critical illness myopathy (CIM) study subjects (Supplemental cycles of 95 o C for 20sec, 64 o C for 30sec, 72 o C for 50sec, 72 o C for 2min, 10 o C to stop.The genotyping primers for LoxP were F: 5'GGCTGTGCCATCTTGATGACC 3'; R: 5'ACCCATGCCGACAATGAAGTTAC 3'.This PCR resulted in a WT band (~ 752 bp) and a GLUT4 LoxP + band (~ 900bp).The Cre primers for MCK-Cre were F: 5'TAAGTCTGAACCCGGTCTGC 3'; R: 5'GTGAAACAGCATTGCTGTCACTT 3' and resulted in Cre + band (450bp).An internal control primer for DNA quality was also included F: 5'CAAATGTTGCTT GTCTGGTG 3'; R: 5'GTCAGTCGAGTGCACAGTTT 3' and resulted in a band ~ 200bp.The Cre PCR protocol was 94 o C for 2min, stepdown protocol of 0.5 o C/cycle at 94 o C for 1min, 65 o C 30sec, 68 o C 50sec, then 28 cycles of 94 o C 1min, 60 o C for 30sec, 72 o C for 50sec, then 72 o C for 7min, 10 o C hold (The Jackson Lab strain # 006475; MCK-Cre + , Cre primer protocol).
plasmid served as a template to clone GLUT4 ΔN24, ΔL6, ΔC42 mutants (PCR primers from Integrated DNA Technologies, Coralville, IA) and these were subsequently cloned using the same protocol.The pcDNA-FLAG-GLUT4 plasmids were transformed into E. coli DH5α (Thermo Fisher Scienti c, Cat # 18265017) bacteria by electroporation at 42 o C for 45sec.E. coli were plated on LB Agar containing 100µg/mL ampicillin and inverted overnight at 37 o C. Antibiotic-resistant colonies were picked and grown in LB media at 37 o C overnight by shaking (100rpm).Plasmid DNA was extracted using PureLink™ Quick Plasmid Miniprep Kit (Thermo Fisher Scienti c, Cat # K210011).Sanger sequencing was used to verify GLUT4 insertion into plasmids.RIG-I mutant plasmids were similarly cloned as above where the human FL RIG-I inserted into the pcDNA3.1-FLAGvector74 served as the template to clone RIG-I CARD, Helicase, ΔCARD and CTD.MDA5 FL was also cloned into the pcDNA3.1-FLAGvector and served as the template to clone MDA5 CARD and ΔCARD plasmids.

Figure 2 The
Figure 2

Figure 5 GLUT4
Figure 5 h, schematic diagram of human GLUT4 membrane topology (left panel) and deletion/point mutant constructs (right panel).Yellow star (169, red text) denotes point mutation of glucose binding residue.i, co-IP of the truncated forms of GLUT4 (h) with endogenous RIG-I and UBXN9 from HEK293T cells.Quanti cation of protein band ratios of RIG-I to GLUT4 in IP elution samples (right panel).Vector, pcDNA3.1-FLAG.j, IFN-β protein concentrations from C2C12 cells transfected with WT or mutant GLUT4 plasmids for 24 h and then either mock treated or stimulated with 3p-hpRNA for 12 hr.k, schematic representation of the domain structure of RIG-I and MDA5 and its truncation mutants used for co-IP experiments in (l, m).FL, full length; CARD, caspase activation and recruitment domain; ∆CARD, RIG-I with deletion of CARD domain; CTD, C-terminal domain.l, m, co-IP of the WT and truncated mutants of RIG-I (l) or MDA5 (m) with GLUT4 in HEK293T cells.The results are representative of 2-3 independent experiments.Bar: mean ± SEM. *p<0.05,**p<0.01,***p<0.001by one-way ANOVA (d, j) or unpaired Student's t-test (g).

Fig. calculated
Fig. calculated by Pearson r correlation.Red box highlights a signi cant negative correlation of GLUT4 with all ISGs.Bar: mean ± SEM. *p<0.05,**p<0.01,by Mann-Whitney test in (d, e) or Pearson r correlation (g).

Figure 8 GLUT4
Figure 8 . All cell lines were grown in Dulbecco's modi ed Eagle medium (DMEM, Life Technologies, Grand Island, NY) supplemented with 10% FBS fetal bovine serum (FBS) and antibiotics/antimycotics.These cell lines are not listed in the database of commonly misidenti ed cell lines maintained by ICLAC and have not been authenticated in our hands.They are routinely treated with MycoZAP (Lonza) and tested for mycoplasma contamination in our hands.Cells were maintained in between experiments using stock DMEM supplied by Life Technologies (Grand Island, NY).During experiments with virus, ligands, etc., cells were cultured in base DMEM without glucose, sodium pyruvate or HEPES (Thermo Fisher Scienti c, Cat # 11966025) supplemented with 10% FBS.Glucose was then complemented back at the same concentration (25mM) normally present in DMEM.This was to account for sodium pyruvate that feeds lactate and would skew glucose uptake/glycolysis measurements.For bioassay, 2fTGH-ISRE reporter cells were used to determine the concentration of IFN-β in the cell supernatant of A549 cells 76 .
. Supernatant was removed and mixed with 20µl of pre-washed (25mM Tris, 150mM NaCl and 0.05% Tween-20) Pierce Anti-Myc Magnetic Beads and rotated overnight at 4 o C. Co-immunoprecipitation was then preformed according to manufacturer's instructions (Thermo Fisher Scienti c, Waltham, MA, Cat # 88842).IP elution was mixed with 4X SDS loading buffer.Samples were not boiled to maintain GLUT4 transmembrane conformation.
. Cells were grown on 6well plates to ~ 2-3 x 10 6 and serum starved for 3h before 20min of insulin treatment or transfected with 3p-hpRNA for 6h to stimulate innate immune receptors.Culture medium was removed, cells washed with cold 1X PBS twice and then lysed in 180µl Buffer A (50mM Tris HCl, 0.5mM DTT, 0.1% NP-40, 1X protease inhibitors).Cells were scrapped off plates and transferred to a 1.5ml centrifuge tube on wet ice before homogenized with a handheld electric pestle for 15sec (Kimble Chase LLC, United States).The homogenized lysate was passed 3X through a 23G needle attached to a 1mL syringe to shear DNA/nucleus and liberate intracellular proteins (small aliquot removed for total cell lysate and subsequently lysed in RIPA).Current and subsequent centrifugation steps were all done at 4 o C to maintain plasma membrane protein structures and prevent denaturing.The remainder of the sheared 1% NP-40) and incubated on ice for 1h with occasional mixing to solubilize membrane proteins.After 1h, proteins were centrifuged for 20min at 12,000 x g and supernatant containing pure plasma membrane (PM) proteins was removed and mixed with SDS-PAGE sample lysis buffer.The backup PPM fraction was incubated on ice for 1h with occasional mixing and centrifuged at 12,000 x g and mixed with SDS-PAGE sample lysis buffer.Total cell lysate (TCL) was incubated with RIPA for 60 min and centrifuged at the same conditions with PM and PPM fractions.Caveolin was used as loading control for the PM fraction and GAPDH for TCL.
goat serum containing 0.2% saponin for 15 min at RT. Cells were subsequently blocked intracellularly with 10% goat serum for 30 min on ice and then incubated with primary antibody diluted to 5µg/mL in 1% BSA, 0.3M glycine and 0.1% saponin.Intracellular staining was allowed to proceed overnight at 4 o C in the dark.After overnight incubation, cells were washed thrice with 1X PBS on ice and secondary antibody staining proceeded as before.At the secondary antibody step, phalloidin was used to mark the plasma membrane.
o C in the dark.Cells were washed 3X with cold 1X PBS for 5 min per wash.A mix containing a variety of Alexa Flour ™ -conjugated secondary antibodies diluted 1:200 (see "Reagent and antibodies" section for speci c species reactivity and uorophore) in antibody buffer was subsequently added to cells for 1h in the dark at RT. Cells were then washed 2X with cold 1X PBS and incubated with DAPI counterstain (Thermo Fisher Scienti c, Cat # D1306) diluted 1:1000 in antibody buffer for 10 min in the dark at RT. DAPI was removed and cells were washed once for 5 min in 1X PBS.before the addition of mixture containing secondary antibody diluted 1:200 (goat anti-human Alexa Fluor™ 488 Cat # A11013) and Hoechst 33342 (Thermo Fisher Scienti c, Cat # 62249) diluted 1:10,000 in 10% goat serum for 1h at RT. Secondary antibody was removed and cells were washed three times with 1X PBS.For co-staining of extracellular GLUT4 and intracellular RIG-I, cells were rst processed and stained with primary anti-GLUT4 as described above.Then, cells were xed as before and permeabilized with 10%