Articles https://doi.org/10.1038/s41590-019-0569-9

Homeostatic regulation of STING at the resting state by stabilizer TOLLIP

Vladislav Pokatayev1,3, Kun Yang 1, Xintao Tu 1, Nicole Dobbs1, Jianjun Wu 1, Robert G. Kalb2 and Nan Yan 1*

STING (stimulator of interferon ) is an important innate immune protein, but its homeostatic regulation at the resting state is unknown. Here, we identified TOLLIP as a stabilizer of STING through direct interaction to prevent its degradation. Tollip deficiency results in reduced STING protein in nonhematopoietic cells and tissues, and renders STING protein unstable in immune cells, leading to severely dampened STING signaling capacity. The competing degradation mechanism of resting-state STING requires IRE1α and lysosomes. TOLLIP mediates clearance of Huntington’s disease-linked polyQ protein aggregates. Ectopically expressed polyQ in vitro or endogenous polyQ proteins in Huntington’s disease mouse striatum sequester TOLLIP away from STING, leading to reduced STING protein and dampened immune signaling. Tollip–/– also ameliorates STING- mediated autoimmune disease in Trex1 –/– mice. Together, our findings reveal that resting-state STING protein level is strictly regulated by a constant tug-of-war between ‘stabilizer’ TOLLIP and ‘degrader’ IRE1α-lysosome that together maintain tissue immune homeostasis.

he cGAS–STING (cyclic guanosine monophosphate-adenos- Results ine monophosphate synthase–stimulator of interferon genes) Identification of TOLLIP as a positive regulator of STING- pathway is an important innate immune signaling pathway mediated immune response. A CRISPR–Cas9 screen was per- T 1 that is responsible for detecting a variety of microbial pathogens . formed to look for regulators of the STING signaling pathway (data cGAS senses cytosolic DNA and produces a cyclic dinucleotide, 2′3′- not shown). The original screen was performed using a pooled cGAMP, which activates the (ER)-localized mouse CRISPR–Cas9 gRNA library (GeckoCRISPRv2; Addgene) protein STING (encoded by the Tmem173). STING activa- and wild-type mouse embryonic fibroblasts (MEFs) expressing tion requires translocation from the ER to ER–Golgi-intermediate a fluorescent reporter as a readout for STING activation. During compartments and then to the Golgi2. During translocation, STING this pilot screen, we identified several candidates, which we further activates IRF3 and nuclear factor-κB transcription factors that validated in a secondary screen using individual siRNA knockdown induce expression of type I interferon (IFN), IFN-stimulated genes and endogenous Ifnb expression as a readout (Fig. 1a). We further (ISGs) and inflammatory . selected candidates that specifically regulate the STING pathway. Constitutive activation of the cGAS–STING pathway leads TOLLIP was identified as the top candidate because pooled siRNA to autoinflammatory and autoimmune diseases3, of which the knockdown of Tollip in wild-type MEFs potently inhibited the best-studied example is Trex1-associated Aicardi–Goutières syn- intracellular cGAMP-induced, but not the double-stranded RNA- drome. TREX1, also known as DNase III, is a DNase in the cyto- induced, Ifnb expression (Fig. 1a,b). Other candidates are also of plasm, and Trex1 deficiency causes accumulation of self-DNA interest and will be pursued in the future. that activates cGAS–STING-mediated inflammation and sys- We next knocked down Tollip with four individual siRNA oli- temic disease in mice4,5. Autoimmunity in Trex1–/– mice originates gonucleotides, and we observed a strong correlation between Tollip from nonhematopoietic cells in the heart, and these mice suc- knockdown efficiency and cGAMP-induced IFN response (Fig. 1c). cumb to inflammatory myocarditis at around 3 months of age6,7. The reduced response to cGAMP was evident throughout a 24-h Either Tmem173+/– or Tmem173–/– can fully rescue Trex1–/– mouse time course and under a broad range of cGAMP concentrations survival, demonstrating a strong dependence of Trex1–/– disease (Fig. 1d,e). Tollip knockdown also substantially inhibited type I IFN, on STING6. inflammatory gene and ISG activation by another cell-permeable Given the essential function of STING, the STING signaling STING agonist 5,6-dimethylxanthenone-4-acetic acid (DMXAA) cascade is regulated by multiple checks and balances, most of (Fig. 1f). This reduced immune was restored when which act after STING activation, to enhance or limit downstream Tollip-knockdown cells were reconstituted with wild-type TOLLIP IFN signaling. For example, posttranslational modifications (siRNA-resistant; Fig. 1g), suggesting that TOLLIP is required for such as phosphorylation and ubiquitination play important roles STING-mediated signaling. Tollip–/– MEFs also showed reduced in regulating ligand-mediated STING activation8–12. However, response to DMXAA, but not to intracellular double-stranded RNA homeostatic regulation of STING protein at the resting state is (Fig. 1h,i). Tollip–/– bone marrow-derived macrophages (BMDMs) much less understood. Here, we identify TOLLIP as a STING sta- also showed reduced IFN response to cGAS or STING ligands bilizer at the resting state that is important for establishing tissue (Fig. 1j). These results collectively suggest that TOLLIP positively immune homeostasis. regulates the STING-mediated immune response.

1Department of Immunology, University of Texas Southwestern Medical Center, Dallas, TX, USA. 2Northwestern University Neurology, Feinberg School of Medicine, Chicago, IL, USA. 3Present address: The Broad Institute of Harvard and MIT, Cambridge, MA, USA. *e-mail: [email protected]

158 Nature Immunology | VOL 21 | February 2020 | 158–167 | www.nature.com/natureimmunology Nature Immunology Articles

a b 3 × 103 2.5 × 103

2 × 103 1 × 103

1 × 103 0.5 × 103 Ifnb1 mRNA (relative) 0 Ifnb1 mRNA (relative) 0 siRNA: siRNA: siCtrl siCtrl siCtrl siCtrl siPigg siKlc3 siKlc3 siPigg siJak1 siGjb2 siGjb2 siJak1 siEmr1 siPalm siTollip siPalm siTollip siEmr1 siIfi205 sirArl10 siPrmt7 siIfi205 siPole2 siPole2 sirArl10 siPrmt7 siKif21b siKif21b siOas1h siOas1h siBhlc23 siAtp6v0 siBhlc23 siRnf151 siAtp6v0 siRnf151 siTuba4a siTuba4a siTrappc8 siTrappc8 siAdam33 siAdam33 siZdhhc12 siZdhhc12 siRabgap1l siRabgap1l Mock + 2'3'-cGAMP Mock + intracellular poly(I:C)

cd15 e 15 104 200 R 2 = 0.9212 siCtrl 3 10 siTollip 10 #4 2 150 10 #3 10 101 100 5 0 5 #2 10 50 10–1 #1 Ifnb1 mRNA (relative) Ifnb1 mRNA (relative) Ifnb1 mRNA (relative) 0 0 10–2 Ifnb1 mRNA (relative) 0 020406080 100 0312 24 2'3'-cGAMP: –– #1 #2 #3 #4 #1-4 siCtrl siCtrl Tollip mRNA knockdown (%) Time (h) siTollip – + 2'3'-cGAMP Mock DMXAA f g NS 1,500 siCtrl 200 150 siTollip siCtrl siTollip siCtrl Fold siTollip *** 150 Ifit2 40 1,000 Usp18 100 siCtrl + Vector Ifit3 30 siTollip + Vector 100 Ifit1 siTollip + FLAG-TOLLIP 500 Oas3 20 50 50 Cxcl10

Il6 mRNA (relative ) Adar1

10 Ifnb1 mRNA (relative) Ifnb1 mRNA (relative) 0 Irf7 0 Mock DMXAA Mock DMXAA Rsad2 Mock DMXAA

h ij MEF BMDM 100 +/+ 200 50,000 Tollip Tollip+/+ 80 Tollip–/– 40,000 150 Tollip–/– 60 30,000 100 10,000 40 50 20 5,000 Ifnb1 mRNA (relative ) Ifnb1 mRNA (relative ) 0 0 Ifnb1 mRNA (relative) 0 DMXAA: –– Poly(I:C): – – Mock DMXAA dsDNA 2'3'-cGAMP

Fig. 1 | Tollip deficiency impairs STING-mediated IFN signaling. a,b, Arrayed siRNA screen of 21 different gene targets in MEFs. qRT–PCR expression of Ifnb mRNA after cGAMP stimulation (2 μg ml−1, transfected by Lipofectamine, same below) for 5 h (a) or intracellular poly(I:C) (1 μg ml−1, transfected by Lipofectamine, same below) for 5 h (b). A representative experiment with two biological replicates is shown (same below). Experiments were repeated at least three times (same below). c, qRT–PCR analysis of Ifnb mRNA after cGAMP stimulation for 5 h. Wild-type MEFs were treated with control siRNA (siCtrl) with four distinct siRNA sequences targeting Tollip for 48 h before cGAMP stimulation. Right panel shows that cells with greater Tollip knockdown had less Ifnb mRNA expression. d, qRT–PCR analysis of Ifnb mRNA in Tollip knockdown MEFs stimulated with cGAMP over a 24-h time course. e, qRT–PCR analysis of Ifnb mRNA in Tollip knockdown MEFs stimulated with an increasing amount of cGAMP (0, 2, 3 and 5 μg ml−1) for 5 h. f, qRT–PCR analysis of Ifnb, Il6 mRNA and several ISGs in Tollip knockdown MEFs stimulated with STING agonist DMXAA (50 μg ml−1) for 2 h. Heatmap summarizes qRT–PCR data on ISG expression. g, qRT–PCR analysis of Ifnb in Tollip knockdown cells stably expressing an empty vector or siRNA-resistant FLAG–TOLLIP. Cells were stimulated with DMXAA (50 μg ml−1) for 2 h. h,i, Tollip+/+ or Tollip–/– MEFs were stimulated with increasing amounts of DMXAA (0, 1, 10 and 100 μg ml−1) for 2 h (h) or intracellular poly(I:C) (0, 1.0, 5.0 and 10.0 ng ml−1) for 5 h (i). Ifnb mRNA was analyzed by qRT–PCR. j, BMDMs from Tollip+/+ or Tollip–/– mice were stimulated with dsDNA (1 μg ml−1), cGAMP (2 μg ml−1) or DMXAA (50 μg ml−1), and Ifnb mRNA was measured by qRT–PCR. g, ***P < 0.001. Error bars represent the s.e.m. Unpaired Student’s t-test (two-sided). Data are representative of at least three independent experiments. NS, not significant.

TOLLIP maintains resting-state STING protein level. Activated knockdown MEFs showed reduced IRF3 phosphorylation after STING recruits the kinase TBK1 to phosphorylate IRF3, a transcrip- DMXAA stimulation (Fig. 2a), consistent with reduced IFN signal- tion factor for type I IFN and ISG induction. To pinpoint which step ing activation. STING protein was substantially reduced in unstim- of the STING signaling cascade is affected by TOLLIP, we next ana- ulated Tollip-knockdown cells compared with control cells (at lyzed IRF3 phosphorylation and STING signaling kinetics. Tollip- time 0, Fig. 2a), suggesting that TOLLIP is required for stabilizing

Nature Immunology | VOL 21 | February 2020 | 158–167 | www.nature.com/natureimmunology 159 Articles Nature Immunology

a b Time after siCtrl siTollip DMXAA (min): 0153060 120 180 0153060 120 180 NS 1.00 100 siCtrl p-IRF3 siTollip 0.75 IRF3 50 0.50 STING (relative) 0.25 MAVS STING protein Tmem173 mRNA relative intensity (%) IRAK1 0 0.00 050 100 150 200 siCtrl HMGB1 Time (min) siTollip (Loading control)

cdNS 1.2 120 Tollip+/+ Tollip –/– Tollip+/+ 100 1.0 Tollip –/– STING 0.8 80 TREX1 0.6 60

CALNEXIN (relative) 0.4 40

TUBULIN 0.2 (% remaining) Tmem173 mRNA Tmem173 mRNA 20 (Loading MEF 0.0 control) +/+ –/– 0 p 024681012 TollipTolli Time after Act. D (h)

e FLAG-TOLLIP f g Vector +/+ +/+ –/– +/+ –/– Tollip Tollip Tollip Tollip Tollip –/– +/+ Tollip –/– 1.5 Tollip Tollip ** STING ** ** * ** ** ***** ** TUBULIN 1.0 Mock MG132 BafA1 (Loading Mock 3-MA control) MEF STING 1.5 TUBULIN 0.5 (Loading MEF control) ISG mRNA (relative) 1.0 0.0

Ifit1 Ifit3 Irf7 Oas3 Ifi44 Ifitm3 Stat1Stat2 0.5 Oas1a Usp18 Rsad2 Cxcl10 STING protein (AU ) 0.0 –/– –/– +/+ +/+ Tollip Tollip Tolli p Tollip

Vector TOLLIP

Fig. 2 | TOLLIP prevents lysosomal-mediated degradation of STING at the resting state. a, Immunoblot analysis of control or Tollip-knockdown MEFs stimulated with DMXAA (50 μg ml−1) for the indicated amount of time (top). Relative intensity of STING protein bands was quantified and presented on the right. Red arrowheads denote reduced STING protein in Tollip-knockdown cells at time 0. b, qRT–PCR analysis of basal Tmem173 mRNA in control or Tollip knockdown MEFs. n = 4 biological replicates. c, Immunoblot analysis of STING and other ER-associated proteins in Tollip+/+ and Tollip–/– MEFs. d, qRT–PCR analysis of Tmem173 mRNA in Tollip+/+ and Tollip–/– MEFs. Bar graph shows basal expression level. Line graph shows mRNA turnover after actinomycin D (Act. D) treatment. n = 3 biological replicates. e, Immunoblot analysis of Tollip+/+ and Tollip–/– MEFs stably expressing an empty vector or FLAG–TOLLIP. Relative protein band intensity from at least two independent experiments was quantified and presented on the bottom. A representative experiment with two biological replicates is shown. Experiments were repeated twice. f, Immunoblot analysis of Tollip+/+ and Tollip–/– MEFs treated overnight with 3-MA (5 mM), MG-132 (5 μM) and BafA1 (0.5 μM). g, qRT–PCR analysis of basal ISG mRNA in Tollip+/+ and Tollip–/– MEFs. n = 3 biological replicates. b,d,g, *P < 0.05; **P < 0.01; ***P < 0.005. Error bars represent the s.e.m. Unpaired Student’s t-test (two-sided). Data are representative of at least two independent experiments. AU, arbitrary units; NS, not significant.

resting-state STING protein. Tmem173 mRNA levels were the knockdown compared with control cells (Fig. 2a), suggesting that same in control and Tollip-knockdown cells (Fig. 2b), suggesting TOLLIP does not alter the kinetics of trafficking-mediated STING that TOLLIP regulates STING at the posttranscriptional level. As degradation. controls, we also analyzed MAVS (involved in RNA sensing) and We also analyzed Tollip–/– cells to substantiate the siRNA knock- IRAK1 (involved in TLR signaling and interacts with TOLLIP13) down findings. Tollip–/– MEFs contain reduced STING protein protein levels. Neither protein was altered in Tollip-knockdown cells compared with Tollip+/+ MEFs without affecting Tmem173 mRNA compared with control cells (Fig. 2a). Ligand-mediated STING acti- expression or stability (Fig. 2c,d). Other ER-associated proteins vation causes STING to translocate from the ER to vesicles where such as TREX1 and CALNEXIN were not affected in Tollip–/– cells, STING is rapidly degraded by lysosomes14. We observed a similar excluding the possibility of a general loss of ER mass in Tollip–/– rate of STING degradation after DMXAA stimulation in Tollip- MEFs (Fig. 2c). Stable expression of wild-type TOLLIP (using a

160 Nature Immunology | VOL 21 | February 2020 | 158–167 | www.nature.com/natureimmunology Nature Immunology Articles

–/– a b retroviral vector) in Tollip MEFs restored STING protein amount HA–STING + + + IP +/+ to that of Tollip cells (Fig. 2e). FLAG–TOLLIP ––+ IgG STING We next investigated which protein degradation pathway is FLAG–TOM1 ––+ responsible for the reduced STING levels in Tollip–/– cells. We DMXAA0 01(h) –/– TBK1 treated Tollip cells with inhibitors for autophagosome formation FLAG *IgL (3-MA), proteasome degradation (MG-132) and lysosome acidifi- IP STING cation (BafA1). Tollip–/– cells treated with BafA1 completely restored HA

IP HA–STING TOLLIP STING protein levels to that of the wild-type cells, and MG-132 *IgL treatment led to a slight increase, suggesting that STING protein is FLAG STING

degraded mostly by lysosomes (Fig. 2f). We also found that base- Lysate –/– line expression of several ISGs is reduced in Tollip MEFs, similar Lysate TOLLIP to what has been previously observed with Tmem173–/– MEFs15,16 HA (Fig. 2g). Together, these data suggest that TOLLIP prevents resting- c state STING protein degradation by a lysosome pathway. TOLLIP (endogenous) STING–GFP DAPI Merge TOLLIP interacts with STING. We next explored whether TOLLIP directly interacts with STING and potentially acts as a stabilizer at the resting state. Hemagglutinin (HA)-tagged STING co-immuno- precipitated (co-IP) with FLAG-tagged TOLLIP in HEK293T cells, but not with FLAG-tagged TOM1, a known binding partner of TOLLIP used as a negative control (Fig. 3a). Endogenous STING and TOLLIP also co-IP together (Fig. 3b). TOLLIP–STING interaction d e is not dependent on STING activation because we detected a similar TOLLIP-binding motif STING-binding motif amount of TOLLIP co-IP with endogenous STING pulldown before 1 153 340 379 154 180 229 274 STING TM1 TM2 TM3 TM4 CBD CTT TOLLIP C2 Cue or shortly after DMXAA stimulation (Fig. 3b). As a control, TBK1 154180 229 1 139 C2 interacted with STING only after DMXAA stimulation. Fluorescent TM1 TM2 TM3 TM4 microscopy also confirmed endogenous TOLLIP and STING–GFP 154180 140 153 340 379 C2 colocalization (Fig. 3c). We next performed truncation mapping to CBD CTT 54 180 229 274 determine which domains of TOLLIP and STING are crucial for C2 Cue this interaction. We found that the N-terminal transmembrane FLAG–STING 54 180 domain of STING (amino acids 1–139), but not the C terminus, was C2 required for interaction with full-length TOLLIP (Fig. 3d). Because FLAG–TOLLIP –379 1 1–139 TOLLIP does not have any transmembrane domains, it is likely 140–379 HA–TOLLIP ++++ that TOLLIP associates with short loops between the four trans- –180 –274 54–274 54–180 1 1 membrane domains in the STING N terminus that are exposed to 1–229 the cytosol. We also created TOLLIP truncations and identified a FLAG HA–STING ++++++ C-terminal motif between the C2 domain and the CUE domain that

IP HA is required for interaction with STING (Fig. 3e). These data suggest HA IP HA–TOLLI P

that TOLLIP interacts with the STING N terminus and potentially FLAG–TOLLIP stabilizes STING protein on the ER. FLAG FLAG Lysat e

PolyQ-rich protein sequesters TOLLIP and promotes STING Lysat e degradation. TOLLIP functions as a selective autophagy HA HA for proteins harboring polyQ-rich domains. Such proteins are asso- ciated with neurodegenerative diseases such as Huntington’s disease Fig. 3 | TOLLIP interacts with STING. a, Co-IP of STING and TOLLIP in 17 (HD) . We wondered what would happen to STING if proteins HEK293T cells. HEK293T cells were transfected with indicated plasmids with polyQ tracts sequester TOLLIP away from STING. We used (top), and 24 h later, anti-HA was used to pull down HA–STING. FLAG– the HUNTINGTIN-polyQ74 (HTTq74) as a model substrate for TOLLIP or FLAG–TOM1 (a negative control) co-IP was analyzed by 17 TOLLIP-mediated degradation as reported previously . Increasing immunoblot. b, Co-IP of endogenous STING and TOLLIP in MEFs. Wild- amounts of TOLLIP led to decreased HTTq74 protein in HEK293T type MEFs were untreated or stimulated with DMXAA for 1 h. Anti-STING cells, confirming TOLLIP’s known function in clearing polyQ-rich was used to pull down endogenous STING. Both IP and lysate were blotted proteins (Fig. 4a). We next expressed increasing amounts of HTTq74 for endogenous TBK1, STING and TOLLIP. c, Fluorescent microscopy in HEK293T cells, and we observed a dose-dependent decrease of analysis of TOLLIP and STING localization. Scale bar, 10 μm. d,e, Domain STING protein (Fig. 4b). We also treated HTTq74-expressing cells mapping studies. HEK293T cells were transfected with full-length HA- with BafA1 or MG-132, and we found that STING protein was TOLLIP plus various truncations of FLAG-STING (d) or full-length HA- restored after lysosome inhibitor BafA1, but not proteasome inhibi- STING plus various truncations of FLAG-TOLLIP (e). Co-IP analysis was tor MG-132, treatment (Fig. 4c). These data suggest that HTTq74 performed as in a. Diagrams of each construct used are shown on the top. accumulation results in STING protein degradation by lysosomes. Data are representative of at least three independent experiments. *IgL, Moreover, HTTq74 expression also inhibited STING-mediated IFN immunoglobulin light chain. response in a dose-dependent manner (Fig. 4d). To assess whether HTTq74 protein disrupts the TOLLIP– STING interaction, we expressed increasing amounts of HTTq74 co-IP with TOLLIP (Fig. 4e). Together, these data suggest that polyQ in HEK293T cells that are also expressing FLAG–TOLLIP and HA– protein sequesters TOLLIP away from STING, which mimics Tollip STING. We also treated cells with BafA1 to avoid STING degradation deficiency and results in STING degradation by lysosomes. caused by HTTq74, which would otherwise complicate our interpre- We next wondered whether naturally occurring endogenous tation of co-IP data. HTTq74 indeed very potently inhibited STING polyQ proteins associated with HD in vivo would also impact

Nature Immunology | VOL 21 | February 2020 | 158–167 | www.nature.com/natureimmunology 161 Articles Nature Immunology

a b e HTTq74: –+++ + FLAG–STING: –+++ FLAG–TOLLIP: –– HTTq74: –– HTTq74: –– FLAG–TOLLIP: +++++ HTTq74 FLAG–STING HA–STING: –++++ HTTq74 FLAG–TOLLIP HMGB1 HA–STING HMGB1 (Loading control) (Loading FLAG–TOLLIP control)

d 15 IP FLAG–TOLLIP c FLAG–STING: –+ ++ + HTTq74 HTTq74: –– ++ + 10 DMSO DMSO DMSO BafA1 MG132 HA–STING FLAG–STING Lysate 5 FLAG–TOLLIP IFN-Luc Activity

HTTq74 HMGB1 0 (Loading HMGB1 IFNβ-Luc: +++++ + control) (Loading cGAS: ––+++ + control) STING: –++++ + HTTq74: –––

f Striatum NS NS Striatum 1.5 3 Wild type zQ175/+ zQ175/zQ175 ** 123123 123 NS 1.0 2 STING 0.5 1 TUBULIN (Loading STING protein (AU ) 0 0 control) Tmem173 mRNA (relative) Cortex Cortex NS Wild type zQ175/+ zQ175/zQ175 1.5 NS 1.5 NS 123123 123 NS 1.0 1.0 STING

0.5 0.5 TUBULIN

(Loading STING protein (AU ) control) 0 0 Tmem173 mRNA (relative)

WT WT zQ175/+ zQ175/+

zQ175/zQ175 zQ175/zQ175

Fig. 4 | PolyQ protein disrupts TOLLIP–STING interaction and results in STING protein turnover in vitro and in vivo. a, Immunoblot analysis of HEK293T cells transfected with the same amount of HTTq74 plasmid and increasing amounts of FLAG–TOLLIP plasmids (as indicated on top). b, Immunoblot analysis of HEK293T cells transfected with the same amount of FLAG–STING plasmid and increasing amounts of HTTq74 plasmids (as indicated on top). c, Immunoblot analysis of STING and HTTq74 coexpression treated overnight with MG-132 (5 μM) and BafA1 (0.5 μM). d, IFNβ-Luciferase assay of cGAS–STING activation after dose titration of HTTq74 protein. A representative experiment with two biological replicates is shown. Experiments were repeated twice. e, Co-IP of STING and TOLLIP in HEK293T cells. HEK293T cells were transfected with indicated plasmids (top) in the presence of BafA1 (0.5 μM), and 24 h later, anti-FLAG was used to pull down FLAG–TOLLIP. HA–STING co-IP was analyzed by immunoblot. BafA1 was used to prevent STING degradation caused by HTTq74 polyQ protein. f, STING protein and mRNA analysis in zQ175 knock-in mouse brain tissues. Wild-type, zQ175/+ and zQ175/zQ175 mouse brain striatum and cortex were freshly dissected and processed for immunoblot and qRT–PCR analysis. n = 3 mice per group. STING protein blots are shown on the left, densitometry quantitation of STING/TUBULIN band ratio is shown in the middle and STING mRNA level (normalized to glyceraldehyde-3 phosphate dehydrogenase, GAPDH) is shown on the right. f, **P < 0.01. Error bars represent the s.e.m. Unpaired Student’s t-test (two-sided). NS, not significant. Data are representative of at least two independent experiments.

STING protein. We chose the zQ175 knock-in mouse that expresses the zQ175 knock-in mouse. Notably, the brain striatum is the main human HTT exon 1 sequence with an ~190 CAG repeat track, and component of the basal ganglia, which is the most affected region of these mice accumulate extensive protein aggregates in the striatum the brain by HD. Taken together, these results suggest that naturally and display neuropathological abnormalities at around 4–6 months occurring polyQ protein can lead to reduced STING protein in the of age18. We isolated cortex and striatum tissues from 10-month- striatum in vivo. old wild-type, zQ175 heterozygous and zQ175 homozygous mouse brain and analyzed tissue lysates by immunoblots. We found sig- TOLLIP specifically stabilizes STING protein in cells. We next nificantly reduced STING protein in the striatum, but not in the determined whether TOLLIP regulates STING protein stability. We cortex, of zQ175 homozygous mice and a decreasing trend in the treated cells with cycloheximide (CHX), and STING protein became heterozygous mice (Fig. 4f). STING mRNA level was not affected in rapidly degraded in Tollip–/– cells with a half-life of approximately

162 Nature Immunology | VOL 21 | February 2020 | 158–167 | www.nature.com/natureimmunology Nature Immunology Articles

120 +/+ a Tollip +/+ Tollip –/– STING in Tollip 100 STING in Tollip –/– CHX (h): 048120 4812 I B in Tollip +/+ 80 κ α –/– STING IκBα in Tollip 60

I B Protein κ α 40

HMGB1 (relative intensity %) 20 (Loading 0 control) 04812 Time (h) b Tollip +/+ Tollip –/– CHX CHX mock BafA BT Z mock BafA BT Z mock BafA BT Z mock BafA BT Z

STING

TUBULIN (Loading control)

Fig. 5 | TOLLIP stabilizes STING protein in cells. a, Immunoblot analysis of STING protein in Tollip+/+ and Tollip–/– MEFs treated with cycloheximide (CHX). IκBα is a control protein with a short half-life that is unaffected by TOLLIP. Densitometry quantitation of protein bands is shown on the right. b, Immunoblot analysis of STING protein in Tollip+/+ and Tollip–/– MEFs treated with CHX alone or in combination with proteasome inhibitor bortezomib (BTZ) or lysosome inhibitor BafA. Data are representative of at least two independent experiments.

6 h (Fig. 5a). In contrast, STING protein is stable in Tollip+/+ cells for IRE1α activation, was significantly upregulated in Tollip–/– cells with a half-life greater than 12 h. As a control, IκBα protein half-life compared with Tollip+/+ cells (Fig. 6a). Stable expression of wild- was not altered by Tollip–/–. The rapid degradation of STING after type TOLLIP suppressed Xbp1 mRNA splicing in Tollip–/– cells back CHX treatment in Tollip–/– cells was prevented by BafA1 treatment to wild-type levels (Fig. 6a). (Fig. 5b). These data suggest that TOLLIP stabilizes STING protein To substantiate these findings, we next assessed UPR proteins to prevent lysosome-mediated degradation. after treating cells with thapsigargin to induce ER stress. Tollip–/– To more rigorously assess whether Tollip–/– causes global protein cells showed increased total IRE1α at the resting state and after instability, we performed a chemical pulse-and-chase experiment fol- thapsigargin treatment (time 0 and 1 h, Fig. 6b). Consistent with lowed by proteomic analysis comparing Tollip+/+ and Tollip–/– MEFs. increased IRE1α, we also observed a remarkable increase in down- First, we treated cells with Click-iT AHA (l-azidohomoalanine) stream BIP protein, an IRE1α-regulated chaperone, in Tollip–/– com- for 4 h to label nascently translated proteins. Then, we IP AHA- pared with Tollip+/+ cells (Fig. 6b). In contrast, other UPR effectors labeled proteins at various times after labeling and analyzed them such as CHOP, phospho-eIF2α and ATF4 (downstream of PERK), by quantitative Tandem Mass Tag mass spectrometry (TMT–MS; and ATF6, were induced equally by thapsigargin in Tollip+/+ and Supplementary Fig. 1a). TMT–MS identified more than 1,000 pro- Tollip–/– cells. We also did not observe any difference in UPR pro- teins with high confidence in each sample (STING was not detected, teins comparing thapsigargin-treated Tmem173+/+ and Tmem173–/– likely because of relatively low expression in MEFs). Of the top pro- MEFs (Supplementary Fig. 2). These data suggest that Tollip defi- teins that show substantial turnover at 20 h, we observed a gradual ciency leads to hyperactivation of IRE1α, which may then facilitate decrease of protein amounts in both Tollip+/+ and Tollip–/– cells that STING turnover. are indistinguishable to each other (Supplementary Fig. 1b), validat- ing that our method can indeed detect protein turnover. When we IRE1α promotes resting-state STING turnover independently of compared the amount of each protein individually in Tollip+/+ versus the ER stress pathway XBP1. We next assessed Tollip–/– cells, very few proteins either increased or decreased more whether IRE1α promotes resting-state STING turnover. We found than twofold, and this pattern remains the same throughout the 20-h a substantially greater amount of STING protein in Ern1–/– (the chase period (Supplementary Fig. 1c). Collectively, these data sug- Ern1 gene encodes IRE1α) compared with Ern1+/+ cells (Fig. 6c). gest that TOLLIP specifically stabilizes STING protein in cells. Tmem173 mRNA levels were similar in both cells (Fig. 6d). Tollip knockdown led to reduced STING protein in Ern1+/+ but not Ern1–/– Tollip –/– hyperactivates the ER stress sensor IRE1α. We next inves- cells, suggesting that IRE1α is required for STING protein turnover tigated how STING protein is degraded in Tollip–/– cells. It was previ- in cells that lack TOLLIP (Fig. 6e). In a complementary experiment, ously shown that STING gain-of-function mutant induces ER stress we inhibited IRE1α activity with a small-molecule compound 4μ8C and the unfolded protein response (UPR)19. Although STING signal- in Tollip+/+ and Tollip–/– cells. Treatment of Tollip–/– cells with 4μ8C ing is not activated in Tollip–/– cells (Fig. 2g), we wondered whether restored STING protein to amounts observed in Tollip+/+ cells, fur- ER stress and the UPR could in turn regulate homeostasis of STING ther supporting that STING turnover in Tollip–/– cells is mediated at the resting state. Mammalian cells contain three major UPR effec- through IRE1α (Fig. 6f). tors, IRE1α, PERK and ATF6. To assess whether the UPR is acti- When IRE1α is activated, it cleaves Xbp1 mRNA to its spliced vated in Tollip–/– cells, we measured several UPR-dependent genes form to produce mature XBP1 protein that activates a variety of that are known to be activated through one of the three major UPR UPR genes. However, Xbp1–/– cells displayed reduced amounts of pathways. ATF6- or PERK-dependent genes such as Chop, Gadd34 STING protein similar to that of the Tollip–/– cells (Fig. 6g). Several and Xbp1 mRNA were not induced in Tollip–/– cells compared with studies have shown that IRE1α is chronically activated in Xbp1–/– Tollip+/+ cells (Fig. 6a). In contrast, spliced Xbp1 mRNA, a marker cells because of the loss of a regulatory feedback loop20–22. We thus

Nature Immunology | VOL 21 | February 2020 | 158–167 | www.nature.com/natureimmunology 163 Articles Nature Immunology

a IRE1α PERK ATF6

4 4 4 4

3 3 3 3

2 2 2 2 (relative)

1 1 1 1 Spliced Xbp1 mRNA Xbp1 mRNA (relative) Chop mRNA (relative) Gadd34 mRNA (relative)

0 0 0 0 Tollip MEF: +/+ –/– –/– +/+ –/– –/– +/+–/– –/– +/+–/– –/– Vector: ++– ++– ++– ++– FLAG–TOLLIP: ––+ ––+ ––+ ––+

b Time after Tollip +/+ Tollip –/– Tg (h): 0136 9120136912

IRE1α

BIP

CHOP

ATF4 p90 ATF6 p50

p-EIF2α

TUBULIN (Loading control)

c +/+ –/– d 1.5 Time after Ern1 Ern1 DMXAA (h): 013612 24 01 36 12 24 1.0 STING

IRE1α (relative) 0.5 Tmem173 mRNA

TUBULIN (Loading 0.0 control) +/+ –/– 1 1 Ern Ern

e +/+ –/– f Ern1 Ern1 DMSO 4µ8C Tollip +/+ –/– Tollip +/+ –/– siCtrl siTollip siCtrl siTollip Tollip Tollip STING STING TUBULIN IRE1α (Loading control) TOLLIP g DMSO 4µ8C TUBULIN Xbp1 +/+ Xbp1 –/– Xbp1 +/+ Xbp1 –/– (Loading control) STING

TUBULIN (Loading control)

Fig. 6 | Tollip deficiency selectively activates IRE1α, which facilitates resting-state STING protein turnover. a, qRT–PCR analysis of basal expression of UPR genes in Tollip+/+ and Tollip–/– MEFs stably expressing an empty vector or FLAG–TOLLIP. A representative experiment with two biological replicates is shown. Experiments were repeated three times. b, Immunoblot analysis of UPR proteins in Tollip+/+ and Tollip–/– MEFs treated with thapsigargin (Tg, 500 nM) for the indicated amount of time (top). Red arrowheads denote increased IRE1α in Tollip–/– MEFs. c, Immunoblot analysis of Ern1+/+ and Ern1–/– MEFs (Ern1 encodes IRE1α) stimulated with DMXAA (10 μg ml−1) for the indicated amount of time (top). d, qRT–PCR analysis of basal Tmem173 mRNA in Ern1+/+ and Ern1–/– MEFs. A representative experiment with two biological replicates is shown. Experiments were repeated three times. e, Immunoblot analysis of Ern1+/+ and Ern1–/– MEFs after treatment with control siRNA or siTollip. f, Immunoblot analysis of Tollip+/+ and Tollip–/– MEFs treated with DMSO or IRE1α inhibitor 4μ8C (100 μM) overnight. g, Immunoblot analysis of Xbp1+/+ and Xbp1–/– MEFs treated with DMSO or 4μ8C (100 μM) overnight. Data are representative of at least three independent experiments.

164 Nature Immunology | VOL 21 | February 2020 | 158–167 | www.nature.com/natureimmunology Nature Immunology Articles

a b ** –/– –/– –/– 30 WT Trex1 Trex1 Tollip Heart

20 40×

10 Body weight (g)

0 10× –/– –/– –/– WT Tollip Trex1 Tollip –/–

Trex1

–/– –/– –/– –/– c 150 NS WT Tollip Trex1 Trex1 Tollip **** 100

50 Spleen weight (mg)

0

–/– –/– –/– WT Tollip Tollip Trex1 –/–

Trex1

–/–

Tollip d –/– –/– –/– NS NS

WT Tollip Trex1 Trex1 Row 40 5 * Irf7 30 4 Usp18 Adar1 3 20 Oas3 2 Irf3 10 Ifit3 1 Ifit2 mRNA (relative) Oas3 mRNA (relative) Rsad2 0 0 Oas1a –/– –/– –/– –/– –/– –/– Ifit2 WT WT Tollip Trex1 Tollip Tollip Trex1 Tollip Ifitm3 –/– –/–

Row min Trex1 Trex1

e 30 200 80

) 150

) 60 –1 ) 20 –1 –1

(pg ml 100 40 (pg ml 10 IL-6 CCL5 (pg ml CXCL1 50 20

0 0 0 –/– –/– –/– –/– –/– –/– –/– –/– –/– WT WT WT Tollip Tollip Tollip Tollip Tollip Tollip Trex1–/– Trex1–/– Trex1–/–

Trex1 Trex1 Trex1

Fig. 7 | Tollip–/– ameliorates autoinflammatory disease of Trex1 –/– mice. a–c, Body weight (a), heart H&E staining (b) and spleen (c) from 16-week-old mice of indicated genotypes. n = 8 mice. Scale bars, 100 μm. d, A heatmap of expression of ISGs in BMDMs of indicated genotypes. Each ISG mRNA expression level was measured by qRT–PCR. Two representative ISGs, Oas3 and Ifit2, are shown on the right. A representative experiment with 3 biological replicates is shown. Experiments were repeated three times. e, Serum measurement. n = 2 mice. a,c,d, *P < 0.05; **P < 0.01; ****P < 0.001. Error bars represent the s.e.m. Unpaired Student’s t-test (two-sided). NS, not significant. treated Xbp1–/– cells with 4μ8C to inhibit IRE1α, which restored Tollip –/– ameliorates STING-mediated autoimmune disease in STING protein to the wild-type amounts. These data suggest that Trex1–/– mice. To assess whether TOLLIP regulates STING function Tollip–/– hyperactivates IRE1α; then IRE1α promotes resting-state in vivo, we crossed Tollip–/– mice to Trex1–/– mice, which develop STING protein turnover independently of XBP1. autoimmune and autoinflammatory disease in a STING-dependent

Nature Immunology | VOL 21 | February 2020 | 158–167 | www.nature.com/natureimmunology 165 Articles Nature Immunology

a WT Tollip –/– Tmem173 +/– Tmem173 –/– b WT Tollip–/– Tmem173 +/– Tmem173 –/–

STING DAPI

10 µm 10 µm 10 µm 10 µm

Isotype c H&E STING

10 µm 10 µm 10 µm 10 µm

d 4 × 106 **** 6 3 × 10 ****

2 × 106

(signal intensity ) 1 × 106 STING protein in heart

0

–/– +/– –/– WT

Tollip

Tmem173 Tmem173 Genotype

Fig. 8 | STING protein level is reduced in Tollip–/– mouse heart tissue. a, H&E staining of fixed mouse heart from indicated genotypes (top). Top images show the whole heart, and bottom images show enlarged view of cardiomyocytes. Scale bars, 500 μm (upper panels); 10 μm (lower panels). b, Fluorescent IHC staining with anti-STING (top, and enlarged view in the middle) or isotype (bottom). STING is in green and DAPI in blue. Scale bars, 60 μm (upper panels); 10 μm (middle and lower panels). c, A representative enlarged view of cardiomyocytes in wild-type heart stained for H&E (left) and anti-STING (right) using adjacent tissue slides cut from the same paraffin-embedded block. Dotted lines indicate the same three cardiomyocytes from both images. Scale bars, 10 μm. d, Quantification of STING protein fluorescence intensity from b. Four separate areas were chosen from each heart, and 25 ROIs were randomly chosen from each area (see Supplementary Fig. 3). STING fluorescent signal intensity from 100 ROIs are shown for each genotype. ****P < 0.001. Error bars represent the s.e.m. Unpaired Student’s t-test (two-sided). Experiments were repeated three times using a different mouse of each genotype. manner3. Genetic ablation of one or both copies of Tmem173 We first analyzed wild-type, Tollip–/–, and Tmem173+/– and rescues Trex1–/– mouse survival6. Tmem173–/– mouse heart (a key tissue that is relevant to Trex1–/– dis- We found that Tollip–/– substantially rescued Trex1–/– mouse dis- ease). Hematoxylin and eosin (H&E) staining revealed no abnor- ease phenotypes. Trex1–/–Tollip–/– mice are equal in size and weight mality in hearts from either genotype (Fig. 8a). Fluorescent IHC as Tollip+/+ or Tollip–/– mice, whereas Trex1–/– mice show significantly staining of STING protein revealed abundant STING protein signal reduced body weight (Fig. 7a). Trex1–/– mice develop splenomegaly, in cardiomyocytes in wild-type mouse heart, whereas both Tollip–/– inflammation in the heart and eventually succumb to inflamma- and Tmem173+/– mice showed significantly reduced STING protein tory myocarditis7. We found that Trex1–/–Tollip–/– heart lacked signs in cardiomyocytes (Fig. 8b–d and Supplementary Fig. 3). Very little of inflammatory cell infiltration and was much improved compared signal was detected in Tmem173–/– mice or with the isotype control with the Trex1–/– mouse heart (Fig. 7b). Trex1–/–Tollip–/– spleen was also antibody. STING protein staining pattern in cardiomyocytes resem- significantly smaller compared with Trex1–/– mice (Fig. 7c). Trex1– bles that of sarcoplasmic reticulum that surrounds muscle fibers, /–Tollip–/– BMDMs show significantly reduced expression of ISGs which is equivalent to the ER in nonmuscle cells (Fig. 8c). compared with Trex1–/– BMDMs (Fig. 7d). Inflammatory cytokines We also analyzed STING expression in T cells, B cells, primary den- such as interleukin-6, CXCL1 (also known as KC) and CCL5 (also dritic cells, BMDMs and bone marrow-derived dendritic cells from known as RANTES) were also significantly reduced in Trex1–/–Tollip–/– wild-type, Tollip–/–, Tmem173+/– and Tmem173–/– mice by immunoblot serum compared with Trex1–/– (Fig. 7e). These data demonstrate that and flow cytometry (Supplementary Fig. 4a–c). We observed ~50% Tollip deficiency dampens STING signaling and ameliorates STING- reduction of STING protein in Tmem173+/– compared with wild-type mediated autoimmune and autoinflammatory disease in vivo. cells. Tollip–/– immune cells did not show substantially reduced STING protein at the resting state; however, low-dose DMXAA stimulation TOLLIP stabilizes STING protein in tissues. Detecting endogenous (2.5 μg ml−1, mimics chronic activation) led to rapid STING degra- STING protein in tissues by fluorescent immunohistochemistry (IHC) dation in Tollip–/– BMDMs and bone marrow-derived dendritic cells, has been technically challenging in the past. We developed an IHC but not in wild-type cells, whereas high-dose DMXAA degraded procedure with a recently available anti-STING antibody that allowed STING quickly in both wild-type and Tollip–/– cells (Supplementary us to reliably detect STING protein in mouse tissues (see Methods). Fig. 4c). Together, these data suggest that Tollip–/– leads to substantially To ensure reliability and reproducibility, we included isotype anti- reduced resting-state STING protein level in nonhematopoietic cells body and Tmem173–/– mice as negative controls. We also included and tissues, and renders STING protein unstable in immune cells (see Tmem173+/– as a control for reduced STING protein amounts. a model in Supplementary Fig. 5).

166 Nature Immunology | VOL 21 | February 2020 | 158–167 | www.nature.com/natureimmunology Nature Immunology Articles

Discussion code availability are available at https://doi.org/10.1038/s41590- Homeostatic regulation is an area of STING biology that was 019-0569-9. incompletely understood. Much of the regulatory mechanisms described to date were focused on STING activation and down- Received: 21 March 2019; Accepted: 3 December 2019; stream signaling. Here, we identified TOLLIP as a critical cofactor Published online: 13 January 2020 that directly interacts with STING and stabilizes it on the ER at the resting state. We also determined a competing degradation mecha- References nism for STING at the resting state that involves IRE1 activation 1. Tan, X., Sun, L., Chen, J. & Chen, Z. J. Detection of microbial infections α through innate immune sensing of nucleic acids. Annu. Rev. Microbiol. 72, and lysosome function. Therefore, our study reveals that resting- 447–478 (2018). state STING protein is strictly regulated by a constant tug-of-war 2. Dobbs, N. et al. STING activation by translocation from the ER is associated between ‘stabilizer’ TOLLIP and ‘degrader’ IRE1α lysosome. Cells with infection and autoinfammatory disease. Cell Host Microbe 18, that lack either activity result in drastic swings of STING protein 157–168 (2015). levels to either low (Tollip–/–) or high (Ern1–/–) levels. This type of 3. Yan, N. Immune diseases associated with TREX1 and STING dysfunction. J. Interferon Cytokine Res. 37, 198–206 (2017). homeostatic regulation is fundamentally important for STING biol- 4. Gao, D. et al. Activation of cyclic GMP-AMP synthase by self-DNA causes –/– ogy in disease settings, because Tollip rescued STING-mediated autoimmune diseases. Proc. Natl Acad. Sci. USA 112, E5699–E5705 (2015). autoinflammatory disease in Trex1–/– mice. Chronic IFN signaling 5. Gray, E. E., Treuting, P. M., Woodward, J. J. & Stetson, D. B. Cutting edge: originates from nonhemopoietic cells in Trex1–/– mouse heart (by cGAS is required for lethal autoimmune disease in the Trex1-defcient mouse genetic lineage tracing experiments6), and Trex1–/– mice succumb model of Aicardi–Goutières syndrome. J. Immunol. 195, 1939–1943 (2015). 7 6. Gall, A. et al. Autoimmunity initiates in nonhematopoietic cells and to inflammatory myocarditis . Consistent with this, we confirmed progresses via lymphocytes in an interferon-dependent autoimmune disease. by fluorescent IHC that STING protein is substantially reduced in Immunity 36, 120–131 (2012). cardiomyocytes in Tollip–/– mouse hearts. 7. Morita, M. et al. Gene-targeted mice lacking the Trex1 (DNase III) 3′→5′ We also found that STING protein level is more substantially DNA exonuclease develop infammatory myocarditis. Mol. Cell Biol. 24, 6719–6727 (2004). reduced in nonhemopoietic tissues compared with hemopoietic 8. Luo, W.-W. et al. iRhom2 is essential for innate immunity to DNA viruses by –/– +/– cells in Tollip mice. As a comparison, Tmem173 mice show mediating trafcking and stability of the adaptor STING. Nat. Immunol. 17, ~50% STING protein in both. This suggests that the balance 1057–1066 (2016). between stabilization and turnover of STING may vary in different 9. Zhang, M. et al. USP18 recruits USP20 to promote innate antiviral response cell types; immune cells may shift the balance toward stabilization through deubiquitinating STING/MITA. Cell Res. 26, 1302–1319 (2016). 10. Zhong, B. et al. Te ubiquitin ligase RNF5 regulates antiviral responses by to ascertain STING-mediated antimicrobial capacity. We showed mediating degradation of the adaptor protein MITA. Immunity 30, that a low-dose DMXAA stimulation (that mimics chronic low- 397–407 (2009). level STING activation) can shift this balance back in immune cells, 11. Xing, J. et al. TRIM29 promotes DNA virus infections by inhibiting innate likely by boosting lysosomal degradation. This balance may also immune response. Nat. Commun. 8, 945 (2017). depend on expression levels of TOLLIP and IRE1α, as well as other 12. Wang, Y. et al. TRIM30α is a negative-feedback regulator of the intracellular DNA and DNA virus-triggered response by targeting STING. PLoS Pathog. unknown components on either side. Another potentially interest- 11, e1005012 (2015). ing aspect of this mechanism is that TOLLIP may communicate 13. Burns, K. et al. Tollip, a new component of the IL-1RI pathway, links IRAK with IRE1α to ‘set’ the homeostatic balance of STING protein level. to the IL-1 receptor. Nat. Cell Biol. 2, 346–351 (2000). For example, in yeast, TOLLIP is required for clearance of certain 14. Gonugunta, V. K. et al. Trafcking-mediated STING degradation requires protein aggregates, and Tollip loss-of-function results in depletion sorting to acidifed endolysosomes and can be targeted to enhance anti-tumor 23 response. Cell Rep. 21, 3234–3242 (2017). of ER protein chaperone and activation of ER stress . This could 15. Warner, J. D. et al. STING-associated vasculopathy develops independently of –/– explain how IRE1α is activated in Tollip cells. Whether this exact IRF3 in mice. J. Exp. Med. 214, 3279–3292 (2017). mechanism in yeast also applies in mammals requires further study. 16. Yang, K., Huang, R., Fujihira, H., Suzuki, T. & Yan, N. N-glycanase NGLY1 Another intriguing observation is diminished STING protein regulates mitochondrial homeostasis and infammation through NRF1. and its signaling capacity when polyQ protein accumulates in the J. Exp. Med. 215, 2600–2616 (2018). 17. Lu, K., Psakhye, I. & Jentsch, S. Autophagic clearance of polyQ proteins cell. TOLLIP has been shown to promote clearance of HD-linked mediated by ubiquitin-Atg8 adaptors of the conserved CUET protein family. 17 polyQ proteins . We showed here that polyQ protein disrupts Cell 158, 549–563 (2014). TOLLIP–STING interaction and results in STING degradation by 18. Menalled, L. B., Sison, J. D., Dragatsis, I., Zeitlin, S. & Chesselet, M.-F. Time lysosomes. Using a knock-in mouse model of HD that expresses course of early motor and neuropathological anomalies in a knock-in mouse polyQ protein zQ175, we found that STING protein, but not model of Huntington’s disease with 140 CAG repeats. J. Comp. Neurol. 465, 11–26 (2003). mRNA, is notably reduced in the striatum, which is the area of brain 19. Wu, J. et al. STING-mediated disruption of calcium homeostasis chronically most affected by HD. It remains unclear which cell type STING is activates ER stress and primes T cell death. J. Exp. Med. 216, 867–883 (2019). most affected in and whether reduced STING protein contributes 20. Fink, S. L. et al. IRE1α promotes viral infection by conferring resistance to to neurodegeneration. STING mediates multiple IFN-dependent apoptosis. Sci. Signal 10, eaai7814 (2017). and IFN-independent signaling pathways, such as inflammation, 21. Hur, K. Y. et al. IRE1α activation protects mice against acetaminophen- 19,24–27 induced hepatotoxicity. J. Exp. Med. 209, 307–318 (2012). cell death and autophagy . STING-mediated inflammation has 22. Aden, K. et al. ATG16L1 orchestrates interleukin-22 signaling in the intestinal been implicated in promoting microglia activation and neuronal epithelium via cGAS-STING. J. Exp. Med. 215, 2868–2886 (2018). death in a Parkinson’s disease mouse model25. Whether STING also 23. Duennwald, M. L. & Lindquist, S. Impaired ERAD and ER stress are early and plays a role in HD or other neurodegenerative disease will be an specifc events in polyglutamine toxicity. Genes Dev. 22, 3308–3319 (2008). 24. Gui, X. et al. Autophagy induction via STING trafcking is a primordial exciting area of future investigation. Together, our study uncovers a function of the cGAS pathway. Nature 567, 262–266 (2019). homeostatic regulatory mechanism of STING protein at the resting 25. Sliter, D. A. et al. Parkin and PINK1 mitigate STING-induced infammation. state by competing activities between TOLLIP and IRE1α lysosome Nature 561, 258–262 (2018). that together maintain tissue immune homeostasis. 26. Gulen, M. F. et al. Signalling strength determines proapoptotic functions of STING. Nat. Commun. 8, 427 (2017). 27. Larkin, B. et al. Cutting edge: activation of STING in T cells induces type I Online content IFN responses and cell death. J. Immunol. 199, 397–402 (2017). Any methods, additional references, Nature Research reporting summaries, source data, extended data, supplementary informa- Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in tion, acknowledgements, peer review information; details of author published maps and institutional affiliations. contributions and competing interests; and statements of data and © The Author(s), under exclusive license to Springer Nature America, Inc. 2020

Nature Immunology | VOL 21 | February 2020 | 158–167 | www.nature.com/natureimmunology 167 Articles Nature Immunology

Methods Laboratories) and an ABI-7500 Fast Real-Time PCR system (Applied Biosystems) Mice, cells and viruses. Tollip−/− mice were obtained from Dr Michel Maillard were used for quantitative RT–PCR (qRT–PCR) analysis. (Centre hospitalier universitaire vaudois, CHUV, Switzerland). Primary MEFs were / –/– isolated from embryos at either embryonic day 10 or 13.5. HEK293T cells were CHX chase assay. Tollip+ + and Tollip MEFs were seeded into six-well plates and 1 obtained from American Type Culture Collection. Tese cells were maintained incubated overnight. Cells were treated with CHX (50 μg ml− ) and chased for 4, in DMEM with 10% (vol/vol) heat-inactivated FCS, 2 mM l-glutamine, 10 mM 8 and 12 h. Cell lysates were collected at indicated time points and analyzed for HEPES and 1 mM sodium pyruvate (complete DMEM) with the addition of STING and IκBα proteins using immunoblot. 100 U ml−1 penicillin and 100 mg ml−1 streptomycin, and were cultured at 37 °C +/+ with 5% CO2. All tissue culture cells were tested for Mycoplasma on a monthly basis Quantification of protein degradation by AHA labeling/TMT–MS. Tollip and –/– and were confrmed to be negative. Experiments performed in BSL-2 conditions Tollip MEFs were seeded into 10-cm plates and incubated overnight. Cells were were approved by the Environmental Health and Safety Committee at University of washed once with warm PBS and refed with methionine-free medium (catalog Texas Southwestern Medical Center. Experiments involving mouse materials were (cat.) no. 21013; Gibco) for 1 h to deplete methionine reserve. Then cell culture approved by the Institutional Animal Care and Use Committees of the University medium was replaced with a methionine-free medium containing 50 μM AHA of Texas Southwestern Medical Center. (cat. no. C10102; Invitrogen) for 4 h to label nascent proteins. After labeling, cells were refed with normal medium and chased for an indicated time. At each time Reagents and . Herring testis DNA was used as dsDNA for immune point, cells were washed twice with ice-cold PBS and lysed in 1% SDS in 50 mM −1 stimulation (Sigma). 2′3′-cGAMP and DMXAA were used for STING agonists Tris–HCl (pH 8.0) containing 250 U ml Benzonase endonuclease. Cell lysate was (Invivogen). Polyinosinic:polycytidylic acid (poly(I:C)) was transfected as a MAVS centrifuged at 16,000g for 5 min to remove debris. AHA-labeled nascent proteins pathway agonist (Invivogen). Lipofectamine 2000 was used as transfection reagent in cell lysates were biotinylated using Click-iT Protein Reaction Buffer Kit (cat. for intracellular stimulations (Thermo Fisher). no. C10276; Invitrogen) with biotin-alkyne (cat. no. B10185; Invitrogen) per Antibodies used included ATF4 (D4B8; Cell Signaling Technology (CST)), manufacturers’ instructions. Biotinylated nascent proteins were further pulled ATF6 (D4Z8V; CST), BIP (C50B12; CST), CHOP (L63F7; CST), phospho-EIF2α down using streptavidin magnetic beads (cat. no. 88817; Pierce). Precipitated (D968; CST), HMGB1 (18256; Abcam), IRAK1 (D5167; CST), IRE1α (14C10; biotinylated nascent proteins were quantified using TMT–MS in the University CST), phospho-IRE1α (PA1-16927; Thermo Fisher), IRF3 (D83B9; CST), of Texas Southwestern Proteomics Core. Data were analyzed using Proteome phospho-IRF3 (4D4G; CST), MAVS (4983; CST), STING (D2P2F; CST), TOLLIP Discoverer 2.2 and were searched using the mouse database from UniProt. (ab187198; Abcam), goat anti–rabbit IgG-HRP conjugate (1706515; Bio-Rad Laboratories) and goat anti–mouse IgG-HRP conjugate (1706516; Bio- Fluorescent Microscopy. Cells were grown on glass coverslips and treated as Rad Laboratories). indicated. Cells were fixed with 4% paraformaldehyde, permeabilized with 0.1% Triton X-100 and blocked with 5% normal donkey serum. Slides were incubated Immunoblotting. Cells were lysed in radioimmunoprecipitation assay (RIPA) with primary antibody for 1 h at room temperature followed by fluorescent- buffer (150 mM NaCl, 5 mM EDTA, 50 mM Tris, 1.0% Nonidet-P40, 0.5% sodium conjugated secondary antibody incubation and mounted with VECTASHIELD deoxycholate, 0.1% SDS) supplemented with protease and phosphatase inhibitors Mounting Media containing DAPI (Vector Laboratories). and then centrifuged at 4 °C to obtain cellular lysate. Equal amounts of protein (10–50 μg) were loaded into a 12% SDS–PAGE gel. Semidry transfer onto Histology and Fluorescent IHC. Mice were euthanized and perfused with 1× nitrocellulose membrane was performed. Membrane was blocked in 5% milk in PBS followed by formalin. Tissues were collected and incubated for 24 h in 10% TBS-T buffer for 1 h at room temperature, followed by overnight incubation in buffered formalin and then transferred to 70% ethanol for 24–48 h. Tissues were 3% milk in TBS-T with primary antibodies. Membrane was washed with TBS-T embedded in paraffin embedding blocks, and 5-μM sections were prepared on buffer, incubated at room temperature with HRP-conjugated IgG secondary slides. H&E staining was performed by the University of Texas Southwestern antibody, washed with TBS-T, and then developed with SuperSignal West Pico Molecular Pathology Core. Chemiluminescent Substrate (Thermo Fisher) or SuperSignal West Femto The pathology slides were deparaffinized and rehydrated by washing Maximum Sensitivity Chemiluminescent Substrate (Thermo Fisher). sequentially in xylene three times for 3 min each, absolute (100%) ethanol three times for 3 min each, 95% ethanol three times for 3 min and the final wash with Co-IP. Wild-type MEFs or HEK293T cells transfected with indicated plasmids water two times for 5 min each. Antigen retrieval was performed by autoclaving were collected, washed once with PBS, lysed in IP lysis buffer (20 mM Tris–HCl, the slides in citric acid solution at 121 °C for 20 min. These samples were cooled pH 7.4, 150 mM NaCl, 0.5% Nonidet-P40 and 1× protease inhibitor mixture) to room temperature overnight. After antigen retrieval, slides were washed twice and centrifuged at 20,000g for 10 min at 4 °C. The supernatants were mixed with in water for 5 min, PBS for 5 min one time and PBS with 0.1% Tween 20 once for primary antibody and Dynabeads Protein G (Life Technology), and incubated 5 min. Slides were blocked for 1–4 h with 10% normal goat serum in 1% BSA and overnight at 4 °C. A small portion of supernatant was saved as input. The following PBS at room temperature. The blocking solution was poured off, but the slides day, the beads were washed once with IP buffer, then twice with high-salt IP buffer were not washed. The samples were incubated with primary Rabbit anti-STING 1 (500 mM NaCl) and finally once with low-salt IP buffer (50 mM). IP complex was (cat. no. 19851-1-AP, 0.33 mg ml− ; Proteintech) diluted to 1:200 in 1% BSA in 1 eluted in immunoblot sample buffer and boiled at 95 °C for 5 min. Samples were PBS or in isotype control Goat anti–Rabbit IgG (cat. no. R5506, 1 mg ml− ; Sigma- analyzed by immunoblot as described earlier. Aldrich) diluted at 1:600 in 1% BSA in PBS incubated overnight at 4 °C. The next day, slides were warmed to room temperature and washed three times for 3 min Compound inhibition. For protein degradation pathway inhibition, cells were each in PBS with 0.1% Tween. Slides were then incubated at room temperature seeded overnight and then treated with 5 mM 3-MA (tlrl-3ma; Invivogen), 0.5 μM for 1 h with secondary Goat anti–Rabbit IgG–Alexa Fluor 488 (Invitrogen). Slides Bafilomycin A1 (tlrl-baf1; Invivogen) and 5 μM MG-132 (133407-82-6; Sigma) for were washed three times for 5 min each with PBS with 0.1% Tween and then two 16 h. Cells were then collected and lysed in RIPA buffer for immunoblot analysis. times for 5 min each with PBS. To reduce the amount of autofluorescence in the For IRE1α inhibition, cells were seeded overnight and then treated with 100 μM background cardiac tissues, we used the Vector TrueVIEW Autofluorescence 4μ8C (SML0949; Sigma) for 24 h. Cells were then collected and lysed in RIPA Quenching Kit (cat. no. SP-8400) right before mounting per the manufacturer’s buffer for immunoblot analysis. instructions. Tissues were stained and mounted with ProLong Glass Antifade Mountant with NucBlue stain (cat. no. P36983; Thermo Fisher). siRNA knockdown. Predesigned siRNA oligomers were obtained from Sigma- Images of the H&E staining slides were acquired using a NanoZoomer 2.0-HT Aldrich and resuspended in water at 20 μM. A total of 105 MEFs were plated digital slide scanner (Hamamatsu). Immunofluorescence images were acquired and transfected with siRNA in Opti-Mem media (Thermo Fisher) for 48 h with using a Zeiss LSM 880 confocal microscope (Zeiss). Further deconvolution was Lipofectamine RNAiMAX (Thermo Fisher) and then validated for knockdown performed using AutoQuant X software (Media Cybernetics). For intensity efficiency with PCR with reverse transcription (RT–PCR) or immunoblotting. quantification, four different areas were selected within the specific tissue, and each Oligonucleotides with efficient knockdowns (>50% mRNA reduction) area was sampled by 25 smaller regions of interest (ROIs) using ImageJ. were used in subsequent experiments. The sense and antisense sequences (5′–3′) were as follows: siTollip#1, sense, CCAUCAAUCCUUGCUGCA Statistical methods. Data are presented as the mean ± standard error of the mean and antisense, UGCAGCAAGGAAUUGAUGG; siTollip#2, sense, (s.e.m.). Prism 6 (GraphPad) was used for statistical analysis. Statistical tests GCACUUACUUACAGGUUAU and antisense, AUAACCUGUAAGUAAGUGC; performed are indicated in the figure legends: *P < 0.05; **P < 0.01; ***P < 0.001; siTollip#3, sense, GAGUUCAUGUGCACUUACU and antisense, ****P < 0.0001. AGUAAGUGCACAUGAACUC; and siTollip#4, sense, CCAAGAACCCUCGCUGGAA and antisense, UUCCAGCGAGGGUUCUUGG. Reporting Summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article. RNA isolation and quantitative RT–PCR. Total RNA was isolated with TRI reagent according to the manufacturer’s protocol (Sigma-Aldrich), and Data availability complementary DNA (cDNA) was synthesized with iScript cDNA synthesis The data that support the findings of this study are available from the kit (Bio-Rad Laboratories). iTaq Universal SYBR Green Supermix (Bio-Rad corresponding author upon reasonable request.

Nature Immunology | www.nature.com/natureimmunology Nature Immunology Articles

Acknowledgements Competing interests We thank M. Maillard (Centre hospitalier universitaire vaudois) for Tollip–/– mice; M. The authors declare no competing interests. Lehrman (University of Texas Southwestern Medical Center) for Ern1+/+, Ern1–/–, Xbp1+/+ and Xbp1–/– MEFs; U. Deshmukh (Oklahoma Medical Research Foundation) for the initial IHC protocol; and members of the Yan laboratory for helpful discussion. This work was Additional information supported by the National Institutes of Health (grant nos. AR067135 and AI134877 to Supplementary information is available for this paper at https://doi.org/10.1038/ N.Y.; grant nos. NS0870778 and NS052325 to R.G.K.), the Cancer Prevention and Research s41590-019-0569-9. Institute of Texas (CPRIT, RP180288 to N.Y.) and the Burroughs Wellcome Fund (N.Y.). Correspondence and requests for materials should be addressed to N.Y. Author contributions Peer review information Zoltan Fehervari was the primary editor on this article and V.P. initiated the project and performed most of the experiments. K.Y. and J.W. helped with managed its editorial process and peer review in collaboration with the rest of the some experiments. X.T. and N.D. helped with mouse IHC staining. R.G.K. performed the editorial team. zQ175 mouse experiment. N.Y. and V.P. wrote the paper with input from all co-authors. Reprints and permissions information is available at www.nature.com/reprints.

Nature Immunology | www.nature.com/natureimmunology

α α α