A non-nucleotide agonist that binds covalently to cysteine residues of STING

Stimulator of interferon genes (STING) is an ER-localized transmembrane protein and the receptor for 2',3'-cyclic guanosine monophosphate–adenosine monophosphate (cGAMP), which is a second messenger produced by cGAMP synthase (cGAS), a cytosolic double-stranded DNA sensor. The cGAS-STING pathway plays a critical role in the innate immune response to infection of a variety of DNA pathogens through the induction of the type I interferons. Pharmacological activation of STING is a promising therapeutic strategy for cancer, thus the development of potent and selective STING agonists has been pursued. Here we report that mouse STING can be activated by phenylarsine oxide (PAO), a membrane permeable trivalent arsenic compound that preferentially reacts with thiol group of cysteine residue (Cys). The activation of STING with PAO does not require cGAS or cGAMP. Mass spectrometric analysis of the peptides generated by trypsin and chymotrypsin digestion of STING identifies several PAO adducts, suggesting that PAO covalently binds to STING. Screening of STING variants with single Cys to serine residues (Ser) reveals that Cys88 and Cys291 are critical to the response to PAO. STING activation with PAO, as with cGAMP, requires the ER-to-Golgi traffic and palmitoylation of STING. Our results identify a non-nucleotide STING agonist that does not target the cGAMP-binding pocket, and demonstrate that Cys of STING can be a novel target for the development of STING agonist.


Introduction
STING is an ER-localized transmembrane protein essential for control of infections of DNA viruses and tumor immune surveillance (Ishikawa and Barber, 2008;Barber, 2015).After its binding to cGAMP (Wu et al., 2013), which is generated by cGAS in the presence of cytosolic DNA (Sun et al., 2013), STING translocates to the trans-Golgi network (TGN), where STING recruits TBK1 from the cytosol and triggers the type I interferon and proinflammatory responses through the activation of interferon regulatory factor 3 (IRF3) and nuclear factor-kappa B (NF-κB) (Mukai et al., 2021).Activation of TBK1 by STING requires its palmitoylation and the Golgi lipid raft (Mukai et al., 2016;Takahashi et al., 2021), which is composed of cholesterol and sphingomyelin (Simons and Ikonen, 1997).Given that protein palmitoylation has been implicated in the clustering of a number of proteins (Linder and Deschenes, 2007) such as H-ras and Fas into lipid rafts, palmitoylated STING may cluster in the lipid rafts in the TGN, which then facilitates the phosphorylation of TBK1 and IRF3 by bringing them close to each other.
Phosphatidylinositol-4-phosphate (PI4P) is the most abundant phosphoinositide in cells (Balla, 2013).Initially identified as a key phosphoinositide that controls membrane trafficking at the Golgi and the actin cytoskeleton in yeast (Cleves et al., 1989), PI4P soon emerged as a critical regulator of membrane trafficking at the Golgi, also in mammals (Godi et al., 1999).PAO is a membrane permeable trivalent arsenic compound that preferentially reacts with thiol group of Cys and inhibits PI4P synthesis (Wiedemann et al., 1996;Balla et al., 2002).While we performed the experiments to investigate the role of PI4P in the STING trafficking and signalling at the Golgi, we unexpectedly found that PAO could affect STING at the ER and activated the STING signalling pathway.
In the present study, we show that PAO binds covalently to STING at the ER and induces the translocation of STING to the Golgi, where STING activates TBK1 and IRF3 in a STING palmitoylation-dependent manner.Our work identifies a novel STING agonist that does not target the cGAMP-binding pocket, and demonstrates that Cys of STING can be a novel target for the development of STING agonist.

PAO activates the STING-TBK1-IRF3 pathway
While we performed the experiments to examine the role of PI4P in the STING trafficking and signalling at the Golgi, we happened to find that PAO affected STING at the ER.PAO treatment (1 μM for 1 h) induced the phosphorylation of STING, TBK1, and IRF3 in mouse embryonic fibroblasts (MEFs), but not Sting -/ -(SKO) MEFs (Fig. 1a).To visualize STING, immortalized SKO MEFs were reconstituted with EGFP-tagged mouse STING.As cGAMP (Wu et al., 2013), PAO induced the translocation of STING from the ER to the Golgi (Fig. 1b).Most importantly, PAO induced the expression of the STING-downstream genes, such as Cxcl10, Ifnb, and Il6 in a STING-dependent manner (Fig. 1c).PAO exhibited its half-maximum activity at about 1 μM (Fig. 1d).Of note, 1 μM of PAO activated nuclear factor-erythroid 2-related factor 2 (Nrf2) through its Cys modifications (He and Ma, 2009).
Given the translocation of STING from the ER to the Golgi by PAO treatment, we reasoned that PAO acted directly on STING, or molecules upstream of STING, such as cGAS/cGAMP.cGAS/cGAMP are dispensable for the activation of the STING signalling with PAO PAO may activate cGAS, which produces the STING ligand cGAMP.To examine the role of cGAS in the STING activation with PAO, cGAS-knockout (CKO) MEFs (Mukai et al., 2021) were exploited.As shown (Fig. 2a), PAO induced the phosphorylation of STING, TBK1, and IRF3 in two cell lines (B8 and C2) of CKO MEFs, as wild-type (WT) MEFs.PAO also induced the expression of Cxcl10, Ifnb, and Il6 in CKO MEFs (Fig. 2b).These results suggested that cGAS was not required for STING activation with PAO.We further examined the contribution of cGAS/cGAMP using STING (Y239A) variant (Gao et al., 2013), which loses the affinity to cGAMP, and a mouse STING agonist DMXAA.As shown (Fig. 2c), PAO, not cGAMP or DMXAA, induced the phosphorylation of STING, TBK1, and IRF3 in SKO MEFs reconstituted with STING (Y239A).PAO also induced the translocation of STING (Y239A) from the ER to the Golgi (Fig. 2d).These results suggested that activation of the STING pathway with PAO was not via cGAS activation or cGAMP production.

PAO covalently binds to STING
As PAO makes a covalent bond with Cys of some proteins, PAO may also bind to Cys of STING to exert its effect.Indeed, the treatment of cells with a reducing agent N-acetylcysteine (NAC), suppressed PAO-induced phosphorylation of STING, TBK1, and IRF3 (Supplementary Fig. 1), suggesting that PAO activated the STING pathway through cysteine modifications.We particularly focused on Cys88 and Cys91, as these two residues are the specific targets of electrophilic compounds, such as nitrounsaturated fatty acids (Hansen et al., 2018), H-151 (Haag et al., 2018), and 4-hydroxylnonenal (Jia et al., 2020).We exploited the fact that human STING generated a 8-amino acid peptide (A 87 CLGCPLR 94 ) including Cys88 and Cys91 after trypsin digestion, which was successfully identified with mass spectrometry (Hansen et al., 2018;Jia et al., 2020).In contrast to human STING, mouse STING has histidine residue (His) at its position 94, preventing mouse STING from generating the 8amino acid peptide.Instead, trypsin digestion of mouse STING is expected to yield a longer peptide (A 87 CLGCPIHCMAMILLSSYF YFLQNTADIYLSWMFGLLVLYK 127 ), which appeared not suitable for mass spectrometric detection.We mutated His94 to Arg94 in mouse STING and used the resultant variant (STING H94R) for the following analyses.
Mouse STING H94R was confirmed to have the same reactivity as WT STING to PAO, regarding the generation of the phosphorylation of STING, TBK1, and IRF3 (Supplementary Fig. 2a) and the translocation from the ER to the Golgi (Supplementary Fig. 2b).
Cells expressing mouse STING H94R tagged with EGFP were treated with PAO for 1 h and lysed.EGFP-STING was immunoprecipitated with anti-GFP-nanobody and eluted from the beads.EGFP-STING was further purified by SDS-PAGE under non-reducing conditions and digested with trypsin and chymotrypsin in situ.The resulting peptides were subjected to liquid chromatography-mass spectrometry (LC-MS) analysis, and two PAO-peptide adducts were identified with a MASCOT score higher than 50 using MASCOT cross-linking analysis (Fig. 3a).conditions.No other bands, indicative of dimer or higher oligomers, were detected after PAO stimulation.In contrast to STING, a fraction of transferrin receptor was upshifted in nonreducing conditions, consistently with the fact that transferrin receptor exists as a dimer by disulfide bond (Jing and Trowbridge, 1987).These results suggested that the peptides crosslinked through PAO (Fig. 3a) were formed intramolecularly, not intermoleculary.

STING C88S and STING C291S variants are not activated with PAO
The mass-spectrometric analysis of the PAO-peptide adducts indicated that Cys205 and Cys291 were covalently modified with PAO.Cys88, Cys91, and/or Cys95 was (were) also indicated to be modified covalently with PAO.We thus examined if which Cys modification(s) with PAO was essential for the STING activation.
To address this, we generated STING variants with single Cys to Ser substitution.We found that STING C88S and STING C291S, but not the other variants (STING C91S, STING C95S, and STING C205S) did not translocate to the Golgi (Fig. 4a, b and Supplementary Fig. 3).These results suggested that the modification of Cys88 and Cys291 with PAO was essential for STING activation.STING C88S and STING C291S mostly lacked the activity to phosphorylate TBK1 and IRF3 (Fig. 4c) and to induce the expression of Cxcl10 (Fig. 4d), further supporting the (2 μg ml -1 ), DMXAA (25 μg ml -1 ), or PAO (1 μM) for 1 h.Cell lysates were prepared and analyzed by western blot.(d) Cells were treated with 2',3'-cGAMP (2 μg ml -1 ), DMXAA (25 μg ml -1 ), or PAO (1 μM) for 1 h, fixed, permeabilized, and stained for GM130 (a Golgi protein, magenta).Nuclei  Of note, we found that STING C147S and STING C205S appeared to lose the typical ER localization without stimulation and partially accumulated around the perinuclear area, with some co-localization with a Golgi protein GM130 (Fig. 4a, b and Supplementary Fig. 3).Thus, Cys147 and Cys205 may be essential for the steady-state ER localization of STING.In contrast, regarding Cys147, the previous report suggested that Cys148 of human STING (equivalent to mouse Cys147) was essential for the translocation of STING to the Golgi (Ergun et al., 2019).The different behavior of the two Cys147 variants may be due to the species difference and/or the different experimental system used, i.e., MEFs stably expressing STING vs. HEK293T cells overexpressing STING.

STING activation with PAO requires the ER-to-Golgi traffic and palmitoylation
The activation of the STING signalling pathway with cGAMP requires the exocytic membrane traffic from the ER to the Golgi, palmitoylation of STING at the Golgi, and the lipid-raft microdomains at the trans-Golgi network (TGN) (Mukai et al., 2016;Takahashi et al., 2021).The treatment of cells with BFA (a fungal macrocyclic lactone that blocks the ER-to-Golgi traffic) (Lippincott-Schwartz et al., 1990), or D-ceramide-C6 (an inhibitor of the formation of the lipid raft microdomains), suppressed the phosphorylation of STING and TBK1 with PAO (Fig. 5a, b).
These results suggested that the activation of the STING signaling pathway with PAO occurred at the Golgi, in particular at the TGN, but not the ER.This notion was further supported by the result that phosphorylation of TBK1, a hallmark of the STING activation, was confined to the TGN (Supplementary Fig. 5).The treatment with H-151 (Haag et al., 2018) (an inhibitor of palmitoylation of STING Cys91) or 2-BP (a panpalmitoyltransferase inhibitor) also suppressed the phosphor- ylation of STING and TBK1 with PAO (Fig. 5c).STING C91S had a reduced activity to induce the expression of Cxcl10 (Supplementary Fig. 6).These results suggested that STING activation by PAO required STING palmitoylation.
Surf4 is a protein that circulates between the ER and the Golgi (Mitrovic et al., 2008).We previously showed that STING was constantly retrieved back from the Golgi to the ER by the binding to Surf4, by which the steady-state localization of STING at the ER was warranted (Mukai et al., 2021).Intriguingly, the amount of Surf4 that bound to STING decreased after treatment with PAO (Fig. 5d).The weaker binding of Surf4 to PAO-conjugated STING may result in inefficient retrograde transport of STING to the ER, leading to the accumulation and activation of STING at the Golgi.

Discussion
Several studies have demonstrated that STING activity is tightly associated with cellular redox status.Treatment of cells with hydrogen peroxide generated intermolecular disulfide bonds between STING, leading to the formation of inactive STING oligomers (Jin et al., 2010;Tao et al., 2020).Cys64, Cys148, Cys292, and Cys309 of STING were suggested to be involved in the formation of disulfide bonds (Jin et al., 2010).High-resolution proteomics showed that Cys148 was constitutively oxidized, whereas Cys206 was oxidized by hypochlorous acid or by cGAMP (Zamorano Cuervo et al., 2021).Glutathione peroxidase 4 (GPX4) deficiency promotes lipid peroxidation (Ursini et al., 1982).4-Hydroxynonenal, one of the end products of lipid peroxidation, covalently bound Cys88 of STING and suppressed STING activation by inhibiting palmitoylation of Cys88 (Jia et al., 2020).In the present study, PAO was shown to bind covalently Cys88/Cys291 of STING and to activate STING in the absence of cGAMP.Thus, these studies, including the present one, suggest that cytoplasmic Cys of STING is susceptible to oxidation, which affects STING activity.The molecular basis underlying the specificity of different oxidants to different Cys of STING remains to be elucidated.
Translocation of STING from the ER to the Golgi is essential for the activation of the STING signalling (Ishikawa et al., 2009;Mukai et al., 2016;Taguchi and Mukai, 2019;Kemmoku et al., 2022).There are a number of molecules involved in the translocation process.The exit from the ER requires the coat protein complex-II (COP-II), a protein complex that is responsible for creating membrane vesicles (COP-II vesicles) that bud from the ER (Barlowe and Helenius, 2016).Sar1 (a small GTPase), Sec23/Sec24 (inner coat proteins), and Sec13/Sec31 (outer coat proteins), and several other proteins associated with COP-II vesicles were shown to be involved in the STING translocation (Ogawa et al., 2018;Sun et al., 2018;Ran et al., 2019;Zhang et al., 2020;Gui et al., 2019).The Ca 2+ sensor stromal interaction molecule 1 (STIM1), an ER-resident protein, was suggested to act as a tether of STING to the ER (Srikanth et al., 2019).Surf4 is a protein that circulates between the ER and the Golgi (Mitrovic et al., 2008).We and others have recently shown that STING was constantly retrieved back from the Golgi to the ER by the binding to Surf4 (Mukai et al., 2021;Deng et al., 2020;Steiner et al., 2022).The PAO binding may induce the conformational change of STING, which strengthens or weaken the affinity of STING to these regulators of the membrane trafficking.Indeed, PAOconjugated STING had a lower affinity to Surf4 (Fig. 5d).
As mentioned above, we observed PAO-crosslinked peptides, one of which was modified with palmitoyl group (Fig. 3b).Given the critical role of Cys88 in the activation of STING with PAO (Fig. 4c, d), we reasoned that Cys88 was crosslinked to Cys291 with PAO and that the other Cys in the beta peptide [A 87 CLGCPIR 94 ], i.e., Cys91, was palmitoylated.The activation of STING with PAO required the ER-to-Golgi traffic, palmitoylation of STING, and the raft-lipid microdomains at the TGN (Fig. 5a-c).
These results indicated that PAO-conjugated STING, as cGAMPbound STING, would cluster in the lipid raft at the TGN with the aid of its palmitoylation, leading to the activation of TBK1 and IRF3 (Mukai et al., 2016).
As shown (Supplementary Fig. 7), PAO could not activate human STING.Human STING also possesses Cys88 and Cys292 (equivalent to mouse Cys291), which can be the targets of PAO.One possibility to account for the inability of PAO to activate human STING is that the three-dimensional distance between Cys88 and Cys292 in human STING may differ from the one between Cys88 and Cys291 in mouse STING.If it is the case, towards the development of human STING agonists with the same mechanism of action as PAO, it will be essential to design the compound that has the appropriate molecular length to crosslink Cys88 and Cys292.
A number of STING agonists have been developed recently (Chin et al., 2022).Most of the STING agonists target the cGAMP-binding pocket, whereas two STING agonists, PC7A (Li et al., 2021) and C53 (Lu et al., 2022), do not.The present study provides another example of STING agonist that does not target the cGAMP-binding pocket, and demonstrates that cytoplasmic Cys of STING can be a novel target for the development of STING agonists, besides STING antagonists (Hansen et al., 2018;Haag et al., 2018;Decout et al., 2021).

Cell culture
MEFs were obtained from embryos of WT or Sting -/-mice at E13.5 and immortalized with SV40 Large T antigen.MEFs were cultured in DMEM supplemented with 10% fetal bovine serum (FBS) and penicillin/streptomycin/glutamine (PSG) in a 5% CO 2 incubator.MEFs that stably express tagged proteins were established using retrovirus.Plat-E cells were transfected with pMXs vectors, and the medium that contains the retrovirus was collected.MEFs were incubated with the medium and then selected with puromycin (2 μg mL -1 ).

PCR cloning
N-terminal EGFP-or FLAG-tagged mouse STING (NM_028261) was introduced into pMXs-IPuro.STING variants were generated by site-directed mutagenesis.

Immunocytochemistry
Cells were fixed with 4% paraformaldehyde (PFA) in PBS at room temperature for 15 min, permeabilized with 0.1% Triton X-100 in PBS at room temperature for 5 min.After blocking with 3% BSA in PBS, cells were incubated with primary antibodies.
After washing with PBS three times, cells were then incubated with the secondary antibody at room temperature for 60 min, washed, and mounted with ProLong™ Glass Antifade Mountant (P36982, Thermo Fisher Scientific).
Target gene expression was normalized on the basis of Gapdh content.

Western blotting
Proteins were separated in polyacrylamide gel and then transferred to polyvinylidene difluoride membranes (Millipore).These membranes were incubated with primary antibodies, followed by secondary antibodies conjugated to peroxidase.The proteins were visualized by enhanced chemiluminescence using Fusion SOLO.7S.EDGE (Vilber-Lourmat).

Mass spectrometry
Purified GFP nanobodies from bacteria were incubated with GST-Accept resin (09277-56, nacalai) at room temperature for 2 h.Cells were lysed with IP buffer (50 mM HEPES-NaOH (pH 7.2), 150 mM NaCl, 5 mM EDTA, 1% Triton X-100, protease inhibitors, and phosphatase inhibitors).The lysates were centrifugated at 15,000 rpm for 10 min at 4°C, and the resultant supernatants were incubated overnight at 4°C with the resin.The beads were washed six times with immunoprecipitation wash buffer (50 mM HEPES-NaOH (pH 7.2), 150 mM NaCl, 1% Triton X-100), and boiled at 95°C for 5 min with elution buffer (50 mM HEPES-NaOH (pH 7.2), 150 mM NaCl, 5 mM EDTA, 1% Triton X-100).Technos Co) at a flow rate of 300 nl min -1 .The separation was performed with 0.1% formic acid as solvent A and 100% acetonitrile containing 0.1% formic acid as solvent B, with a linear gradient of 40 min with solvent B rising from 0% to 80%.The separated peptides were analyzed using an online coupled Q Exactive Mass Spectrometer (Thermo Fisher Scientific) using the data-dependent Top 10 method.The acquired data were processed using MASCOT 2.8 (Matrix Science) and Proteome

Quantification of imaging data
For quantification of imaging data of multiple cells, individual cells were segmented using ROI.Pearson's correlation coefficient was then quantified by Coloc 2 in Fiji plugin with ROI data.

Statistical analyses
Error bars displayed in the bar plots throughout this study represent s.e.m. unless otherwise indicated and were calculated from triplicate samples.In box-and-whisker plots, the box bounds the interquartile range (IQR) divided by the median, and whiskers extend to a maximum of 1.5 × IQR beyond the box.The corresponding data points are overlayed on the plots.Statistical significance was determined with one-way ANOVA followed by Tukey-Kramer post hoc test.

K
. Matsumoto et al. 60 Cell Structure and Function 48: 59-70 (2023) https://doi.org/10.1247/csf.22085Cell Structure and Function of the peptides was acylated with one palmitoyl group and phosphorylated at threonine residue 293.MS/MS spectra of PAO cross-linked peptides are shown in Fig. 3b.The y-series ions of the alpha peptides of [F 290 CRTLEEILEDVPESR 305 ] and [F 201 PLDCGVPDNLSVVDPNIR 219 ] were observed in agreement with theoretical fragmentation patterns.Although no fragments of the cross-linking partner (beta peptides of [A 87 CLGCPIR 94 ] or [G 90 CPIRCMAMILL 101 ] were detected, these parent ions of cross-linked peptides agreed with the theoretical masses, suggesting that PAO covalently bounds cysteine residues of STING to form the cross-linking between peptide chains.To distinguish the intramolecular or intermolecular crosslink of Cys with PAO, EGFP-STING was analyzed by non-reducing SDS-PAGE.As shown (Fig. 3c: lane 2 and lane 4 from the left), the mobility of EGFP-STING was essentially similar both in reducing (with dithiothreitol (DTT))-and non-reducing (without DTT)

Fig. 2
Fig. 2 Activation of STING with PAO does not require cGAS/cGAMP (a) cGAS knockout MEFs (clone B8 and clone C2) were stimulated with PAO (1 μM) for 1 h.Cell lysates were prepared and analyzed by western blot.(b) Cells (WT and cGAS KO clone B8) were treated with PAO (1 μM) for 1 h, followed by 4 h incubation without PAO.The expression of Cxcl10, Ifnb, and Il6 was quantified by qRT-PCR.(c) Sting -/-MEFs expressing EGFP-STING (WT or Y239A) were stimulated with 2',3'-cGAMP Fig. 3 PAO covalently binds to STING (a) Sting -/-MEFs expressing EGFP-STING (H94R) were stimulated with PAO (1 μM) for 1 h and lysed.EGFP-STING in the lysates was immunoprecipitated with EGFP nanobody.EGFP-STING were separated with SDS-PAGE under non-reducing conditions and analyzed by mass spectrometry.PAO cross-linked peptides from STING identified by MASCOT cross-linking analysis.The cross-linked peptides consisting of an alpha peptide of 290-FCRTLEEILEDVPESR-305 and a beta peptide of 87-ACLGCPIR-94, an alpha peptide of 201-FPLDCGVPDNLSVVDPNI R-219, and a beta peptide of 90-GCPIRCMAMILL-101 were identified with MASCOT scores of 56 and 69 and expected value of 7.2e -06 and 1.7.e -07, respectively.(b) The MSMS spectra of PAO cross-linked peptides of STING.The peaks shown in red were consistent with theoretical values.The y-series ions of the α-peptide were detected, and the parent ions agreed well with the theoretical mass of the PAO cross-linked peptide.(c) Cells were stimulated with PAO (1 μM) for 1 h.Cell lysates were prepared without dithiothreitol (DTT) and analyzed by western blot.