EGFR‐mediated tyrosine phosphorylation of STING determines its trafficking route and cellular innate immunity functions

Abstract

STING (STimulator of INterferon Genes) mediates protective cellular response to microbial infection and tissue damage, but its aberrant activation can lead to autoinflammatory diseases. Upon ligand stimulation, the endoplasmic reticulum (ER) protein STING translocates to endosomes for induction of interferon production, while an alternate trafficking route delivers it directly to the autophagosomes. Here, we report that phosphorylation of a specific tyrosine residue in STING by the epidermal growth factor receptor (EGFR) is required for directing STING to endosomes, where it interacts with its downstream effector IRF3. In the absence of EGFR‐mediated phosphorylation, STING rapidly transits into autophagosomes, and IRF3 activation, interferon production, and antiviral activity are compromised in cell cultures and mice, while autophagic activity is enhanced. Our observations illuminate a new connection between the tyrosine kinase activity of EGFR and innate immune functions of STING and suggest new experimental and therapeutic approaches for selective regulation of STING functions.

A phospho‐switch governs trafficking of the interferon response stimulator STING from the ER to endosomes instead of autophagosomes, allowing it to bind and activate its effector IRF3.

Introduction

Microbial infections are detected by a variety of cellular pattern recognition receptors (PRRs) which trigger inflammatory responses to protect the infected cells as well as their neighbors. The same receptors that recognize pathogen‐associated molecular patterns (PAMPs) also recognize cellular damage‐associated molecular patterns (DAMPs) and respond to defects in cell metabolism and cell death (Takeuchi & Akira, 2010). PRRs include Toll‐like receptors (TLRs), Nod‐like receptors (NLRs), RIG‐I‐like receptors (RLRs), and DNA‐recognizing receptors, such as cGAS, DAI, IFI16, and DDX41 (Jønsson et al, 2017; Liu et al, 2019). Ligand‐induced activation of PRRs leads to their recruitment of specific adaptor proteins, which in turn recruit protein kinases, ubiquitin ligases, and transcription factors to assemble signaling complexes (Krachler et al, 2011; Liu et al, 2013, 2019; Hu & Sun, 2016; Prabakaran et al, 2018). Transcription factors, such as NF‐κB, IRF3, or AP1, activated by their post‐translational modifications by the enzymes of the signaling complexes, translocate to the nucleus to induce transcription of genes encoding inflammatory cytokines and type I interferons (IFNs) (Liu et al, 2016).

STING (also known as MITA, ERIS, or MPYS) is the nodal point of signaling in response to a variety of cellular stresses. It recognizes cyclic dinucleotides, such as c‐di‐GMP or c‐di‐AMP, produced by intracellular bacteria; it also recognizes cGAMP, synthesized by the cytoplasmic enzyme, cGAS, in response to cytoplasmic cellular or microbial DNA (Chen et al, 2016). Active nuclear cGAS, enriched on centromeres and LINE DNA repeats, is also found (Gentili et al, 2019). STING is needed for signaling by other DNA‐recognizing receptors, such as IFI16 and DDX41, as well (Dunphy et al, 2018). Physiologically, STING provides protection against many bacterial and viral infections by triggering the synthesis of IFN and other cytokines and autophagy (Ishikawa et al, 2009). For example, STING^−/− cells and mice are more susceptible to infection by herpes simplex virus 1 or Listeria monocytogenes (Ishikawa et al, 2009; Archer et al, 2014). The signaling duo cGAS/STING also respond to cytoplasmic DNA leaked from nuclei or mitochondria due to defects in DNA damage repair or organelle integrity (Li & Chen, 2018; Aarreberg et al, 2019). STING‐mediated inflammation has been connected to skin cancer development in 7, 12‐dimethylbenzanthracene (DMBA)‐treated mice (Ahn et al, 2014). On the other hand, STING provides protection against colorectal tumorigenesis (Zhu et al, 2014); it also promotes ionizing radiation‐induced tumor regression and T‐cell‐mediated control of tumor growth (Deng et al, 2014; Sivick et al, 2018). Therefore, STING signaling plays a critical role in combating both microbial diseases and cancer.

STING spans the membrane of the endoplasmic reticulum (ER) with a long cytoplasmic region that binds activating ligands and signals downstream to induce IFNs and other cytokines (Ishikawa & Barber, 2008). The adaptor protein, TRIF, is bound to STING constitutively and required for STING signaling (Wang et al, 2016). Ligand binding causes conformational change and multimerization of STING followed by several post‐translational modifications of the protein that affect its signaling functions. For example, K63‐linked ubiquitination of Lys224 is essential for STING signaling (Ni et al, 2017) and palmitoylation of Cys88 and Cys91 occurs, when activated STING translocates to the Golgi (Mukai et al, 2016). The recruitment of TBK1 leads to the phosphorylation of Ser366 of STING and its interaction with IRF3 which is essential for interferon induction (Tanaka & Chen, 2012). The different cellular activities of STING, including transcription factor activation, gene induction, and triggering of autophagy, require its translocation to different intracellular membrane compartments. Activated STING translocates from the ER to the ERGIC, the compartments from which the NF‐κB signaling occurs. Then, it follows two trafficking routes. The first one takes STING from the ERGIC to the Golgi network, late endosomes, and lysosomes which fuse with autophagosomes to form autolysosomes. In the second route, ERGIC serves as the membrane source for WIPI2 recruitment by STING and LC3 lipidation, leading to the formation of autophagosomes. Gui et al (2019) speculated that, when located in the endosomal compartment, STING binds to IRF3 and activates it through phosphorylation by the STING‐bound TBK1.

Similar to the cGAS/STING system, several TLRs, namely TLR3, TLR7/8, and TLR9, recognize intracellular nucleic acids to trigger cytokine synthesis (Kawai & Akira, 2010; Pandey et al, 2014). These TLRs reside on the endosomal, not ER, membrane. Recent studies have shown that many of these TLRs require ligand‐dependent phosphorylation of specific Tyr residues in their cytoplasmic domains in order to recruit the adaptors which assemble the signaling complexes (Chattopadhyay & Sen, 2014). TLR3 signaling requires phosphorylation of two specific Tyr residues by the protein Tyr kinases (PTK), EGFR and Src, which bind to TLR3 only after its ligand stimulation. Similarly, TLR9 phosphorylation by EGFR is required for its ability to recruit the adaptor, MyD88, but unlike TLR3, EGFR is constitutively bound to TLR9 on the endosomal membrane (Veleeparambil et al, 2018). Although EGFR's primary function, as the EGF receptor, requires its presence on the plasma membrane, the protein is abundant in the membranes of many intracellular organelles, including the ER (Wang & Hung, 2012). Moreover, EGF binding is not necessarily needed for EGFR activation; it can also be activated by conformational change due to interaction with another protein (Chakraborty et al, 2014; Nozaki et al, 2019).

Here, we report that the ER‐bound EGFR was associated with ligand‐activated STING and phosphorylated Tyr245 of STING. In the absence of this modification, activated STING translocated rapidly to the autophagosomes. More importantly, it did not traffic to the late endosomes at all and IRF3 was not recruited and activated by STING, even though its TBK1 binding and Ser366 phosphorylation were unaffected. Consequently, EGFR was essential for the induction of IRF3‐driven interferon synthesis by STING and its antiviral action.

Results

EGFR kinase is required for STING‐mediated gene induction

We used a variety of cell lines to inquire whether protein tyrosine kinases, especially EGFR, are required for gene induction by STING signaling. For this purpose, the STING ligand, cGAMP, was transfected into cells and the induction of IFN and other mRNAs was measured by qRT–PCR. IFN‐β mRNA was strongly induced by STING signaling in the mouse monocytic cell line, Raw 264.7, but the induction was inhibited by genistein, a general inhibitor of PTKs, indicating that protein tyrosine phosphorylation is required for STING signaling (Fig EV1A). To inquire whether EGFR was one of the PTKs, we used the EGFR‐specific inhibitor, gefitinib, or cells in which EGFR expression had been ablated. Gefitinib inhibited IFN‐β mRNA induction in RAW cells (Fig 1A). Similarly, IFN‐α mRNA induction in MEFs was inhibited by gefitinib; gefitinib dose–response analysis in the latter cells indicated an IC50 of below 5 μM (Fig EV1B). This value is higher than that needed to inhibit the kinase activity of EGFR on the cell surface, but the available concentration of the inhibitor could be much lower at the intracellular ER–cytoplasm interface, where EGFR acts on STING, as elaborated below. To ensure that the observed effect of the inhibitor on STING signaling was mediated by inhibiting EGFR, and not another PTK, we tested cells in which EGFR expression had been genetically ablated. Indeed, when Raw 264.7 cells, in which EGFR expression had been ablated by shRNA (Fig EV1C), were used for detecting STING signaling, STING‐mediated IFN‐β mRNA induction was inhibited (Fig 1B). To test whether the EGFR requirement was true in primary cells, we used bone marrow‐derived macrophages from WT mice and EGFRfl/fl, LysMCre mice, which carried a deletion of the EGFR gene only in the myeloid cells (Fig EV1D); induction of IFN‐α mRNA was strongly impaired in the absence of EGFR (Fig 1C). To test whether the phenomenon was true in human cells as well, we used a HeLa cell line, from which the EGFR gene had been deleted by CRISPR/Cas9; IFN‐β mRNA was not induced in that line (Fig 1D), demonstrating that the need of EGFR for IFN induction by STING signaling was not cell type‐ or species‐specific. When we rescued the HeLa EGFR^−/− cells by transfecting EGFR‐expressing vectors, EGFR WT rescued IFN‐β mRNA induction, not the EGFR kinase‐dead mutant (Fig EV1E). Next, we inquired whether the induction of all genes by STING required EGFR activity. For this purpose, RNA‐seq analysis was performed using RNAs isolated from gefitinib‐treated or gefitinib‐untreated and cGAMP‐stimulated or cGAMP‐unstimulated MEFs (Table EV1). The analysis showed that the induction of some genes, but not others, required EGFR activity (Fig 1E). Of the 259 genes induced by cGAMP, the induction of 86 genes was inhibited by twofold or more by gefitinib (Fig 1F). The RNA‐seq results were confirmed by qRT–PCR analysis of a sensitive mRNA, IFN‐α, and an insensitive mRNA, Ccl20 (Fig 1G). STING‐dependent gene induction is primarily mediated by two transcription factors, IRF3 and NF‐κB. It appeared that EGFR was required only for induction of the IRF3‐driven genes, a conclusion that was confirmed by transcription factor activation analysis (see below). The above results demonstrated that EGFR was needed for the induction of a subset of genes by STING, indicating that the function of EGFR, in this context, is signaling pathway‐specific.

Figure EV1. (related to Fig 1): Inhibition of IFN mRNA induction by PTK and EGFR inhibitors.

1. Raw 264.7 cells were pretreated with the general tyrosine kinase inhibitor, genistein (100 μM), or equal volume of solvent (DMSO) for 1 h, and then transfected with cGAMP. After 5 h, total RNA was extracted and IFN‐β mRNA was measured by qRT–PCR (n = 4). Bars are the mean ± SD of indicated (n) independent experiments. ***P < 0.001 when compared to WT by two‐tailed Student's t‐test.
2. MEF cells were pretreated with DMSO (0 μM) or different concentration of gefitinib as indicated for 1 h, and then transfected with cGAMP. After 5 h, IFN‐α mRNA induction was measured by qRT–PCR (n = 4). Bars are the mean ± SD of indicated (n) independent experiments.
3. EGFR expression in WT and EGFR knockdown RAW264.7 cells (Fig 1B) was analyzed by Western blot.
4. EGFR expression in WT and EGFR^−/− BMDMs (Fig 1C) was analyzed by Western blot.
5. WT, but not kinase‐dead EGFR, could rescue signaling in EGFR^−/− HeLa cells. EGFR^−/− HeLa cells were reconstituted by expressing WT or kinase‐dead EGFR. WT cells, EGFR^−/− cells, and EGFR^−/− cells reconstituted with WT or mutant EGFR were stimulated with cGAMP or left untreated. After 5 h, total RNA was extracted and IFN‐β mRNA was measured by qRT–PCR (n = 3). Bars are the mean ± SD of indicated (n) independent experiments.
6. Poly I:C signaling did not require EGFR. cGAMP or Poly (I:C) was transfected into WT and EGFR^−/− HeLa cells. After 5 h, total RNA was extracted and IFN‐β mRNA was measured by qRT–PCR (n = 3). Bars are the mean ± SD of indicated (n) independent experiments. ***P < 0.001 when compared to WT by two‐tailed Student's t‐test.
7. The chemical structures of gefitinib and AZD9291.
8. MEF cells were pretreated with DMSO or the EGFR inhibitor AZD9291 (10 μM) before cGAMP transfection; after 5 h, IFN‐α mRNA and Ccl20 mRNA were measured by qRT–PCR (n = 3). Bars are the mean ± SD of indicated (n) independent experiments. ***P < 0.001 when compared to AZD9291‐pretreated group by two‐tailed Student's t‐test.
9. HT1080 cells were pretreated with DMSO or the EGFR inhibitor AZD9291 (10 μM) or gefitinib (10 μM) before cGAMP transfection; after 5 h, IFN‐β mRNA and Ccl20 mRNA were measured by qRT–PCR (n = 3). Bars are the mean ± SD of indicated (n) independent experiments. ***P < 0.001 when compared to AZD9291‐pretreated group by two‐tailed Student's t‐test.

Figure 1. STING‐mediated induction of a subset of genes needs EGFR kinase activity.

• A
  Raw 264.7 cells were pretreated with DMSO or the EGFR inhibitor gefitinib (10 μM) before cGAMP transfection; after 5 h, IFN‐β mRNA was measured by qRT–PCR (n = 3). Bars are the mean ± SD of indicated (n) independent experiments. ***P < 0.001 when compared to gefitinib‐pretreated group by two‐tailed Student's t‐test.
• B
  WT and EGFR knockdown Raw 264.7 cells were transfected with cGAMP; 4.5 h post‐treatment, IFN‐β mRNA was measured (n = 3). Bars are the mean ± SD of indicated (n) independent experiments. ***P < 0.001 when compared to gefitinib‐treated group by two‐tailed Student's t‐test.
• C
  WT and EGFR^−/− BMDMs were transfected with cGAMP; 5 h post‐transfection, IFN‐α mRNA induction was measured (n = 4). Bars are the mean ± SD of indicated (n) independent experiments. ***P < 0.001 when compared to EGFR^−/− cells by two‐tailed Student's t‐test.
• D
  WT and EGFR^−/− HeLa cells were transfected with cGAMP; after 5 h, IFN‐β mRNA was measured (n = 3).
• E–G
  MEF cells were pretreated with gefitinib for 1 h and then transfected with cGAMP; 4.5 h post‐treatment, total RNA was harvested for analysis. (E) RNA‐seq analysis (n = 2). For RNA‐seq analysis, poly‐T-enriched cDNA library was prepared and 30 M single‐end 50‐bp reads were analyzed. The heat map for some of the genes is shown. (F) 259 mRNAs were induced by twofold or more upon cGAMP treatment; the induction of 86 of those genes was inhibited by gefitinib by twofold or more. (G) The RNA‐seq analysis results were confirmed by qRT–PCR analysis of a gefitinib‐sensitive and an gefitinib‐insensitive mRNA (n = 3). Bars are the mean ± SD of indicated (n) independent experiments. ***P < 0.001 when compared to gefitinib‐treated group by two‐tailed Student's t‐test.
• H
  MEF cells were pretreated with gefitinib for 1 h and followed with cGAMP or Poly (I:C) transfection; 4.5 h post‐treatment, total RNA was harvested and IFN‐α mRNA induction was measured (n = 3). Bars are the mean ± SD of indicated (n) independent experiments. ***P < 0.001 when compared to gefitinib‐pretreated group by two‐tailed Student's t‐test; NS means no significance.

Next, we evaluated the specificity of EGFR action. Gefitinib could not inhibit IFN‐α mRNA induction, in MEFs, by transfected Poly I:C (Fig 1H); similarly, poly I:C could induce IFN‐β mRNA in WT and EGFR^−/− HeLa cells equally strongly (Fig EV1F), indicating that EGFR was not required for the cytoplasmic RLR signaling pathway, which induces the same genes that are induced by STING signaling. The specificity of EGFR action was also verified using another pharmacologic inhibitor of EGFR, AZD9291, which has a structure quite distinct from the structure of gefitinib (Fig EV1G). Like gefitinib, AZD9291 inhibited the induction of IFN‐α mRNA, but not Ccl20 mRNA, in MEFs (Fig EV1H) and in HT1080 cells (Fig EV1I).

STING‐mediated innate immune response to HSV1 infection requires EGFR

Once we established the need of EGFR in STING signaling in an experimental setting, namely, after cGAMP transfection, we wanted to determine the same need in a natural setting. HSV1 infection is known to activate STING signaling leading to IFN induction which attenuates virus replication and resultant pathogenesis (Ishikawa et al, 2009; Parker et al, 2015; Reinert et al, 2016; Yang et al, 2018). As expected, in MEF, HSV1 infection induced IFN mRNA and this induction was inhibited by gefitinib (Fig 2A), and consequently, virus DNA replication was enhanced (Fig 2B). Similarly, in EGFR^−/− HeLa cells IFN was poorly induced by HSV1, compared to its induction in WT cells (Fig 2C). Next, we extended these observations in cultured cells to HSV1‐infected mice. With the chosen dose of virus, HSV1 infection did not kill any mouse and administration of gefitinib alone did not affect mouse survival. But when in the infected mice, EGFR activity was inhibited by gefitinib, 60% of mice died (Fig 2D). Similar death was observed in infected mice that had the EGFR gene deleted in their myeloid cells, indicating that those cells were the major producers of IFN (Fig 2E). As expected, the virus replicated more robustly in the brains of those mice (Fig 2F). These results demonstrated that EGFR activity is essential for STING‐mediated innate immune response to HSV1 infection in vitro and in vivo.

Figure 2. STING‐mediated IFN induction and attenuation of HSV1 replication and pathogenesis require EGFR .

1. MEF cells were pretreated with gefitinib or DMSO for 1 h and then infected with HSV1 at m.o.i. of 1. Four hours post‐infection (h.p.i.), IFN‐β mRNA induction was monitored by qRT–PCR (n = 3). Uninfected cells were used as negative control (U). Bars are the mean ± SD of indicated (n) independent experiments. ***P < 0.001 when compared to gefitinib‐pretreated group by two‐tailed Student's t‐test.
2. DNA was extracted from gefitinib‐treated or gefitinib‐untreated infected MEFs 0 and 16 h.p.i., and viral DNA was measured by qPCR (n = 3). Bars are the mean ± SD of indicated (n) independent experiments. **P < 0.01 when compared to DMSO group by two‐tailed Student's t‐test.
3. WT and EGFR KO HeLa cells were infected with HSV1, RNA was extracted at 4 h.p.i., and IFN‐α mRNA induction was measured (n = 3). Bars are the mean ± SD of indicated (n) independent experiments. **P < 0.01 when compared to EGFR KO group by two‐tailed Student's t‐test.
4. Survival of uninfected mice treated with gefitinib and HSV1‐infected mice treated or untreated with gefitinib.
5. Survival of WT and EGFR^fl+/− LysM^Cre+/+ mice infected with HSV1.
6. The infection was as in (E); HSV1 titers in the brains of the live mice at day 9 were determined. Bars are the mean ± SD of indicated (n) independent experiments. ***P < 0.001 when compared to WT by two‐tailed Student's t‐test.

IRF3 activation is impaired in the absence of EGFR activity

STING signaling proceeds through several discreet steps, and we investigated which step required EGFR activity. cGAMP‐induced dimerization of STING was unaffected by inhibiting EGFR (Fig 3A), indicating that the need of EGFR is at a step further downstream. As expected from the above conclusion, ligand‐independent gene induction by a constitutively dimeric STING mutant (V155M) required the presence of EGFR (Fig EV2A). The transcription factor IRF3 as well as TBK1, its activating kinase, was phosphorylated in response to cGAMP stimulation; gefitinib inhibited IRF3 phosphorylation, but TBK1 phosphorylation was unaffected (Fig 3B). To extend these analyses, we monitored nuclear translocation of the two transcription factors, IRF3 and NF‐κB, upon cGAMP stimulation of STING. Nuclei were isolated, and total cell extracts (T) and nuclear extracts (N) were analyzed by Western blot; both extracts contained RNA polymerase II, a nuclear protein, but only the total cell extracts contained GAPDH, a cytoplasmic protein, thus demonstrating the purity of the nuclear extracts. In the nuclei, we detected the presence of IRF3 after cGAMP stimulation and this nuclear import was impaired by gefitinib treatment of the cells. In contrast, the nuclear import of NF‐κB (p65) was not impaired by gefitinib (Fig 3C). This was because EGFR had a differential effect on activation of the two transcription factors: IRF3 activation, as monitored by its phosphorylation, was totally blocked by gefitinib, but NF‐κB activation, as judged by the phosphorylation and degradation of IkB, was unaffected (Fig EV2B). These results showed that IRF3 activation, but not NF‐κB activation, by STING signaling required the action of EGFR.

Figure 3. EGFR kinase activity is required for IRF3 activation.

1. EGFR was not required for ligand‐induced STING dimerization: MEFs were transfected with HA‐tagged STING and treated with gefitinib or DMSO before cGAMP transfection, 4 h later, STING was immunoprecipitated with anti‐HA antibodies and analyzed by Western blot.
2. EGFR was required for the activation of IRF3, but not TBK1: MEFs were pretreated with gefitinib for 1 h and then treated with cGAMP for 2 h; cell lysate was subjected to Western blot analysis using the indicated antibodies.
3. Nuclear translocation of IRF3, but not NF‐κB p65, required EGFR: MEFs were pretreated with gefitinib (+) or DMSO (−) for 1 h before transfecting cGAMP. Cells were harvested, and nuclei were isolated at the indicated time points; total cell extracts (T) and nuclear extracts (N) were analyzed by Western blot.
4. IRF3, but not TBK1, binding to STING required EGFR: MEFs were treated as in (B); cell lysates were immunoprecipitated with anti‐STING antibody and analyzed by Western blot.
5. STING‐IRF3 interaction in human cells required EGFR: WT and EGFR^−/− (KO) HeLa cells were transfected with cGAMP; 2 h later, cells were lysed, immunoprecipitated with anti‐STING antibody, and analyzed by Western blot.
6. STING and IRF3 co‐localized in the endosomes of stimulated cells: STING^−/− MEF cells were transfected with STING‐HA for 24 h and then treated with cGAMP for 1 h. The cells were then stained with the HA (purple), CD63 (red), and IRF3 (green) antibodies. Imaging data were analyzed by the Fuji software to reveal co‐localization as white dots. Scale bars represent 10 μm.
7. EGFR was not required for ligand‐induced phosphorylation of STING Ser366: EGFR^−/− HeLa cells or gefitinib‐treated and gefitinib‐untreated WT HeLa cells were treated with cGAMP for the indicated time, and cell lysates were analyzed by Western blot.
8. Kinetics of STING and IRF3 co‐localization in the endosomes of stimulated cells: STING^−/− MEF cells were transfected with STING‐HA for 24 h and treated with cGAMP for the indicated time. The cells were then stained with the HA (purple), CD63 (red), and IRF3 (green) antibodies and DAPI. Imaging data were analyzed by the Fuji software to reveal co‐localization as white dots. Scale bars represent 10 μm.

Source data are available online for this figure.

Figure EV2. (related to Fig 3): No involvement of ULK1/2 in mediating EGFR action on IRF3‐driven gene induction.

1. Gene induction by the constitutively active V155M STING mutant required EGFR. WT and EGFR KO HeLa cells were transfected with wild‐type STING (Wt) or STING mutant (V155M). 16 h post‐transfection, IFN‐β mRNA levels were estimated by qRT–PCR (n = 3). Bars are the mean ± SD of indicated (n) independent experiments. ***P < 0.001 when compared to EGFR KO group by two‐tailed Student's t‐test.
2. IRF3, but not IkB, phosphorylation required EGFR. Raw 264.7 cells were pretreated with gefitinib or DMSO for 1 h and then transfected with cGAMP. Total protein was harvested at the indicated time points and submitted to Western blot with different antibodies.
3. In L929 cells, ULK1 expression was knocked down with shRNA; cell clones expressing Sh‐ULK1 RNA or a non‐target control shRNA were analyzed for ULK1 expression.
4. ULK2 expression was also knocked down by shRNA in L929‐Sh-ULK1 cells.
5. Effects of gefitinib on IFN‐β mRNA induction in Wt, ULK1 KD, and ULK1/ULK2 KD L929 cells, respectively (n = 3). Bars are the mean ± SD of indicated (n) independent experiments.
6. The individual panels of STING, CD63 (late endosome marker, red), IRF3, and IRF3‐CD63 co‐localized in the endosomes of stimulated cells (related to Fig 3H). Imaging data were analyzed by the Fuji software to reveal co‐localization as white dots. Scale bars represent 10 μm.

IRF3 activation by STING requires interaction between the two proteins along with TBK1, the IRF3 kinase. IRF3 co‐immunoprecipitated with STING after cGAMP stimulation, but this interaction did not occur after gefitinib treatment (Fig 3D) or in EGFR^−/− cells (Fig 3E). STING‐IRF3 interaction was also demonstrated by confocal microscopy (Fig 3F). As speculated before (Gui et al, 2019), this interaction occurred in the endosomal compartment. After stimulation, diffused STING (purple) and IRF3 (green) relocated to punctate spots in late endosome (CD63 marker, red). Unbiased co‐localization analysis of the different markers was aided by the Fiji software program; co‐localized proteins appeared as white dots. The STING‐IRF3 interaction in the endosomes peaked at 1 h of cGAMP treatment. After activation, IRF3 probably leaves STING and moves to the nucleus (Figs 3H and EV2F). The above analyses led to the conclusion that the recruitment of IRF3 by STING and IRF3 activation takes place in late endosomes and they require EGFR kinase activity. Tanaka and Chen (2012) recently reported that STING phosphorylation on Ser366 (Ser365 in mouse) by TBK1 is essential for STING‐IRF3 interaction and IRF3 activation. Therefore, we examined whether the observed need of EGFR for STING‐IRF3 interaction was mediated through the regulation of Ser366 phosphorylation. However, STING Ser366 phosphorylation was unaffected in the absence of EGFR activity; in WT Hela cells, EGFR^−/− HeLa cells, and gefitinib‐treated Hela cells, the kinetics of Ser366 phosphorylation were very similar (Fig 3G), indicating that EGFR was not acting by promoting Ser366 phosphorylation. In an earlier report, Konno et al (2013) claimed that the protein kinase, ULK, phosphorylates Ser366 to terminate STING signaling. To examine a putative role of the ULK family of kinases in mediating the effect of EGFR, we knocked down the expression of ULK1 (Fig EV2C) or both ULK1 and ULK2 (Fig EV2D). Although IFN mRNA induction was stronger in the absence of ULK1, indicating a negative role of this kinase, it was still impaired by gefitinib (Fig EV2E). The above results demonstrated that neither ULK1 nor ULK2 mediated the effect of EGFR on STING signaling.

EGFR binds to STING and mediates its Tyr phosphorylation

To acquire further insights into the mechanistic basis for EGFR involvement in STING signaling, we inquired whether STING was Tyr‐phosphorylated. Indeed, STING was strongly phosphorylated upon ligand stimulation (Fig 4A); the Tyr phosphorylation was detectable as early as 1 h after cGAMP transfection and gefitinib treatment blocked it (Fig 4B). Similarly, STING phosphorylation was strongly impaired in EGFR knockdown cells (Fig 4C). Moreover, EGFR co‐immunoprecipitated with STING, but only after ligand stimulation (Fig 4D). Confocal microscopy confirmed EGFR‐STING co‐localization in the ER compartment of the stimulated cells (Fig 4E); diffused STING relocated to punctate spots after cGAMP stimulation (green). EGFR (red) was present in these spots, visualized as yellow spots in the merged image or as white spots after co‐localization analysis of the images with the Fiji software program. The two signaling proteins also co‐localized with calnexin, an ER marker. EGFR binding to STING was detectable after 30 min. However, the interaction was transient, disappearing after 2 h (Fig 4F). Interestingly, EGFR activity was not required for STING binding, but in the presence of gefitinib, EGFR stayed bound to STING even 2 h after cGAMP treatment (Fig 4G). Phosphorylation of Tyr1068 of EGFR is required for its kinase activity. Active EGFR was detected in cGAMP, not Poly I:C‐treated, cells, but only in the presence of STING (Fig 4H). cGAMP‐induced EGFR activation peaked around 3 h (Fig 4I). The STING‐bound EGFR was phosphorylated on Tyr1068, indicating that it was enzymatically active (Fig 4J). An analysis of the kinetics of the phosphorylations, upon cGAMP treatment, showed that the peak of phosphorylation of STING (Tyr and Ser), TBK1, EGFR, and IRF3 all happened around 3 h (Fig 4K). Neither Tyr and Ser366 phosphorylation of STING nor Tyr1068 phosphorylation of EGFR was induced by RLR activation by Poly I:C (Fig 4L). The above results indicated that STING could be a substrate of EGFR in cGAMP‐stimulated cells.

Figure 4. Upon cGAMP stimulation, EGFR binds STING and mediates its Tyr phosphorylation.

1. STING was Tyr‐phosphorylated in cGAMP‐treated cells. Raw 264.7 cells were transfected with cGAMP; after 60 min, cell lysates were immunoprecipitated with anti‐pTyr antibodies and then subjected to anti‐STING Western blot.
2. STING Tyr phosphorylation was inhibited by gefitinib. MEF cells were pretreated with gefitinib before cGAMP stimulation for 60 min. Cell lysates were immunoprecipitated with anti‐pTyr antibody and then analyzed by anti‐STING Western blot. NTC: no treatment control.
3. STING Tyr phosphorylation required EGFR. Raw 264.7 cells expressing EGFR shRNA (KD) or non‐target shRNA (CO) were transfected with cGAMP for 2 h; cell lysates were analyzed by Western blot before or after immunoprecipitation with anti‐pTyr antibodies.
4. cGAMP stimulation triggered EGFR binding to STING. Raw 264.7 cells were transfected with cGAMP for 1 h; cell lysates were immunoprecipitated with anti‐STING antibody and analyzed by Western blot.
5. STING co‐localized with EGFR after cGAMP stimulation. STING^−/− MEF cells were transfected with STING‐GFP (green) for 24 h and stimulated with cGAMP for 1 h; cells were fixed and labeled with EGFR (red) and calnexin (purple) antibodies. Co‐localization was analyzed by Fuji, and co‐localized dots are shown in white. Scale bars represent 10 μm.
6. STING‐EGFR interaction was transient. Raw 264.7 cell lysates were prepared at indicated time points after cGAMP transfection, immunoprecipitated with anti‐EGFR antibodies, and subjected to Western blot.
7. Gefitinib did not block, but prolonged STING‐EGFR interaction. Raw 264.7 cells were pretreated with gefitinib or DMSO for 1 h before cGAMP stimulation. Cell lysates were prepared at indicated time points, immunoprecipitated with anti‐STING antibody, and subjected to Western blot.
8. Poly I:C did not activate EGFR. STING^−/− MEF cells were transfected with cGAMP or Poly (I:C); after 3 h, p‐EGFR was detected by Western blot.
9. Kinetics of EGFR activation. Raw 264.7 cells were treated with cGAMP for the indicated time. Cell lysates were analyzed for p‐EGFR via Western blot.
10. STING‐bound EGFR was phosphorylated. Raw 264.7 cells were stimulated for 1 h; cell lysates were immunoprecipitated with anti‐STING antibody and subjected to Western blot.
11. Kinetics of phosphorylation. Raw 264.7 cells were treated with cGAMP for the indicated time. Cell lysates were subjected to Western blot using the indicated antibodies.
12. Poly I:C did not trigger Tyr or Ser phosphorylation of STING. HeLa cells were transfected with cGAMP or Poly (I:C); after 3 h, cell lysates were analyzed by Western blot using the indicated antibodies.

Source data are available online for this figure.

EGFR is needed for STING trafficking to endosomes

After stimulation, STING travels from the ER to other membrane compartments and different cellular functions of STING are elicited from different compartments. We wanted to determine whether the absence of EGFR affects the trafficking route of STING and, consequently, its signaling pattern. In unstimulated cells, STING‐GFP was diffusely distributed but cGAMP stimulation caused it to concentrate to punctate dots; however, in the presence of gefitinib, the dots were replaced by clusters (Fig EV3A). About 80% of gefitinib‐treated stimulated cells, as against 20% of untreated cells, displayed STING in clusters (Fig EV3B), indicating that EGFR affects STING trafficking pattern. To gain further insight, we performed a detailed analysis of the presence of STING in different subcellular compartments. Confocal microscopy in MEFs was used to determine the co‐localization of STING‐GFP and the respective marker proteins of ER (calreticulin), ERGIC (P58), autophagosomes (LC3), and endosomes (CD63), the compartments to which STING is known to travel. As expected, in unstimulated cells, STING was in the ER and gefitinib did not alter its location (Fig 5A, upper two panels). After 2.5 h of cGAMP treatment, in the absence of gefitinib, punctate dots of STING were still present in the ER, but in the gefitinib‐treated cells, it was almost absent (Fig 5A, lower two panels), suggesting that in the absence of EGFR activity, stimulated STING leaves the ER much more rapidly. From the ER, activated STING translocates to the ERGIC, where it was present in both gefitinib‐treated and gefitinib‐untreated cells (Fig 5B). On the other hand, there was plenty of STING in the autophagosomes of only cGAMP and gefitinib‐treated cells (Fig 5C). A kinetic analysis showed that STING was undetectable even 3 h after stimulation in the autophagosomes of untreated cells (Fig 5D), but abundant clusters of STING appeared in the autophagosomes of gefitinib‐treated cells as early as 0.5 h after stimulation (Fig 5E); the correlation coefficient between STING and LC3 was even higher than that in cells 3 h after cGAMP treatment without gefitinib (Fig EV3C). An alternate route of trafficking takes STING from the ERGIC to the endosomes. Indeed, STING was detected in the endosomes of gefitinib‐untreated cells 1 h after stimulation (Fig 5F); the correlation coefficient between STING and CD63 after 1 h of stimulation was quite high (Fig EV3D). But in gefitinib‐treated cells, it was not present in the endosomes even 2.5 h after stimulation (Fig 5F). Similar results were obtained by comparing STING trafficking in WT and EGFR^−/− HeLa cells. STING appeared in the autophagosomes of EGFR^−/− cells within 0.5 h, but it was not detectable in the endosomes even after 3 h (Fig EV3E–H). The above analysis demonstrated that EGFR activity was needed for appropriate trafficking of STING to the different membrane compartments from which it signals. Without active EGFR, stimulated STING quickly exits from the ER, travels through ERGIC, and accumulates in the autophagosomes. More importantly, EGFR activity was needed for STING to transit to the endosomes, where it recruits and activates IRF3 through its phosphorylation by TBK1.

Figure EV3. (related to Fig 5): Effects of EGFR on STING trafficking.

1. STING^−/− MEFs expressing GFP‐STING (Green) were incubated with cGAMP for 2.5 h; cells were fixed and observed under confocal microscope. In the images from cGAMP‐treated cells, the STING‐enriched regions were expanded and shown in the right panels (red border). Scale bars represent 10 μm.
2. STING translocation in cells was quantitated by ImageJ. At least 40 cells were randomly picked. The green dots/clusters were highlighted, and the area of each dots/clusters and the total area were calculated. The top 5 individuals in each cell were picked, and the sum of their total area was calculated as Area10. In case, Area10 was bigger than 30% of the area of all green dots/clusters in the cell; the cell was considered as cells with clusters. Otherwise, the cell would be considered as cells with dots. The percentage of cells with clusters or punctate dots in total of cells with translocated STING were calculated, and the number of cells in each part is labeled in each column. All values are means ± SD, Student's t‐test (unpaired); ***P < 0.001.
3. Pearson's co‐localization coefficient between STING and LC3 from Fig 5D and E. At least 25 cells were quantified from each independent experiment, which was repeated for three times with similar results. Values are presented as mean ± SD, ***P < 0.001, two‐tailed Student's t‐test (unpaired).
4. Pearson's co‐localization coefficient between STING and CD63 from Fig 5F. At least 25 cells were quantified from each independent experiment, which was repeated for three times with similar results. Values are presented as mean ± SD, ***P < 0.001, two‐tailed Student's t‐test (unpaired).
5. Hela cells were transfected with STING‐GFP (Green) for 24 h and then transfected with cGAMP for the indicated time; cells were fixed and labeled with LC3 (autophagosome marker, red) antibody. Co‐localization (Col) is indicated by white dots. Scale bars represent 10 μm.
6. EGFR^−/− Hela cells were transfected with STING‐GFP (Green) for 24 h and transfected with cGAMP for the indicated time; cells were fixed and labeled with LC3 (autophagosome marker, red) antibody. Co‐localization (Col) is indicated by white dots. Scale bars represent 10 μm.
7. Same treatments as in (E), but labeled with CD63 (late endosome marker, red) antibody. Scale bars represent 10 μm.
8. Same treatments as in (F), but labeled with CD63 antibody (red). Scale bars represent 10 μm.

Figure 5. In the absence of EGFR, cGAMP stimulation causes STING to translocate to autophagosomes but not to late endosomes.

1. In the absence of EGFR, after cGAMP stimulation, STING translocated out of the ER. STING^−/− MEF cells were transfected with STING‐GFP (green) for 24 h, pretreated with gefitinib for 1 h, and transfected with cGAMP for 2.5 h; cells were fixed and labeled with calreticulin (ER marker, red) antibody. Co‐localization (Col) is indicated by white dots. Scale bars represent 10 μm.
2. STING was present in the ERGIC of stimulated cells in the presence or the absence of gefitinib. With the same treatment as in (A), the cells were labeled with P58 (ERGIC marker, red) antibody. Scale bars represent 10 μm.
3. STING translocated to the autophagosomes in the absence of EGFR. With the same treatment in (A), the cells were labeled with LC3 (autophagosome marker, red) antibody. Scale bars represent 10 μm.
4. Less STING was detected in the autophagosomes after a 3 h stimulation if EGFR was not inhibited. The procedures were as in (C). Scale bars represent 10 μm.
5. In gefitinib‐treated cells, STING was detected in the autophagosomes as early as 0.5 h. The procedures were as in (C). Scale bars represent 10 μm.
6. STING translocation to late endosomes required EGFR activity. STING^−/− MEF cells were transfected with STING‐GFP (green) for 24 h, pretreated with gefitinib for 1 h, and transfected with cGAMP for the indicated time periods; cells were fixed and labeled with CD63 (late endosome marker, red) antibody. Scale bars represent 10 μm.

Tyr245 of STING is the direct target of phosphorylation by EGFR

To determine whether EGFR is the direct kinase for STING, we performed cell‐free kinase assays. STING‐GFP was expressed in HEK293XL cells, which do not express endogenous STING, and purified by affinity chromatography (Fig EV4A and B). Purified STING‐GFP was first treated with a phosphatase to remove any putative phosphate modifications (Fig 6A, lower panel) and then used as a substrate for phosphorylation by commercially available pure EGFR (Fig EV4A and B). The phosphorylated protein, if any, was detected by phos‐tag gel analysis (Nagy et al, 2018). The single slower migrating band indicated that STING was phosphorylated by EGFR at one Tyr residue (Fig 6A, upper panel). To identify the specific phosphorylated Tyr residue in the cytoplasmic domain of STING, we focused on the Tyr residues that are on its outer surface (Shang et al, 2012). Four Tyr residues were individually mutated to Phe, and the mutant proteins were expressed and purified to use them as substrates of EGFR. Only the Y245F mutant failed to be phosphorylated, indicating that Y245 is the direct target of EGFR in vitro (Fig 6B). To verify that Y245 was phosphorylated in vivo as well, we purified STING‐GFP from cGAMP‐stimulated cells and subjected it to LC‐MS/MS analysis. A doubly charged peptide had an observed m/z of 932.9266 Da and was within −0.96 ppm of the expected mass. The mass difference between the y8 and y9 ions was consistent with phosphorylation at Y245 which demonstrated the existence of phosphorylated Y245 site (Fig 6C). The degree of modification was determined by plotting chromatograms for both the unmodified and modified forms of the Y245 peptides (Fig EV4C). The sequence around this phosphorylation site is conserved from Bos taurus to Homo sapiens; in mouse STING, the corresponding phosphorylation site is Y244 (Fig 6D). EGFR targeted the phosphorylation of Tyr245, as evidenced by its absence in either gefitinib‐treated cells (Fig 6E) or EGFR^−/− HEK293XL cells (Fig 6F) even after cGAMP stimulation.

Figure EV4. (related to Fig 6): STING Tyr phosphorylation, signaling protein interaction, and phosphorylation.

1. Silver staining of EGFR and recombinant STING used in the experiments of Fig 6A.
2. Silver staining of EGFR and recombinant STING mutants shown in Fig 6B.
3. The degree of modification was determined by plotting chromatograms for both the unmodified and modified forms of the Y245 peptides. Chromatograms for the VYSNSIpYELLENGQR peptides are shown.
4. STING Y244F‐Myc and STING WT‐Myc recombinant stable cell lines in primary STING^−/− MEF were treated with cGAMP for 2 h; then, the cell lysates were subjected to Myc pull‐down by Myc trap beads. The pull‐down products were analyzed by Western blot with the indicated antibodies.
5. The same cell lines as in (D) were treated with cGAMP for the indicated time; the cell lysates were used for Western blot with p‐S365 STING antibody.
6. WT and EGFR^−/− HeLa cells were treated with Torin (10 μM) and CQ (Chloroquine, 20 μM) for 1 h and then stimulated with cGAMP for 1 h, where indicated. Cell lysates were subjected to Western blot analysis via LC3II antibody.

Figure 6. EGFR phosphorylates Tyr245 of STING .

1. EGFR phosphorylates STING in vitro. Purified STING‐GFP was incubated with increasing amounts of EGFR for in vitro kinase assays; the products were analyzed on phos‐tag gel by Western blot using GFP antibody (upper panel). The substrate was analyzed, before the assay incubations, by the same method (lower panel).
2. Y245F STING mutant is not a substrate of EGFR. WT and the indicated mutants of STING‐GFP were purified and used as substrates for EGFR and analyzed as in (A).
3. LC‐MS/MS analysis of STING identifies phosphorylated Y245. STING‐GFP was purified from cGAMP‐stimulated cells and subjected to protease digestion and LC-MS/MS analysis. Phosphopeptides corresponding to 239VYSNSIpYELLENGQR253 containing pY245 were identified. The MS/MS spectra for the phosphorylated peptide VYSNSIpYELLENGQR are shown. This doubly charged peptide has an observed m/z of 932.9266 Da and is within −0.96 ppm of the expected mass. The mass difference between the y8 and y9 ions is consistent with phosphorylation at Y245.
4. Evolutionary conservation of the STING Y245.
5. Gefitinib pretreatment inhibits phosphorylation of Y245. STING‐GFP-293XL cells were pretreated with gefitinib and then treated with cGAMP. LC-MS/MS analysis was done, as above, for detecting pY245 (n = 3). Bars are the mean ± SD of indicated (n) independent experiments. ***P < 0.001 when compared to gefitinib‐pretreated group by two‐tailed Student's t‐test.
6. EGFR is required for STING Tyr phosphorylation. STING‐GFP was expressed in 293XL cells in which EGFR expression had been knocked down. STING phosphorylation was analyzed by LC‐MS/MS as described above (n = 3). Bars are the mean ± SD of indicated (n) independent experiments. ***P < 0.001 when compared to EGFR KD group by two‐tailed Student's t‐test.
7. Y245F STING mutant cannot trigger IFN‐β mRNA induction. RNA was isolated from cells, expressing WT or Y245F STING, 3 h after stimulation and subjected to qRT–PCR analysis (n = 3). Bars are the mean ± SD of indicated (n) independent experiments. ***P < 0.001 when compared to WT by two‐tailed Student's t‐test.
8. Y245F STING is phosphorylated on S366 after stimulation of cells. Phosphorylation of S366 was detected in cells expressing the WT or the mutant STING after stimulation for the indicated time.
9. TBK1, but not IRF3, can bind to Y245F STING and be phosphorylated in cGAMP‐stimulated cells. Lysates from cells expressing the WT or the mutant STING were analyzed by Western blot after (upper panel) or before (lower panel) GFP pull‐down.
10. Y245F mutant induced more LC3 lipidation after 1 h cGAMP treatment of 293XL cells expressing WT or mutant STING.
11. Inhibition of EGFR by gefitinib induced more LC3II after 1.5 h cGAMP treatment of L929 cells.

Source data are available online for this figure.

Y245F STING has properties similar to those that WT STING has in the absence of EGFR

We determined various properties of the Y245F mutant STING to determine whether they were similar to the properties of WT STING, in the absence of EGFR. Stimulation of cells expressing WT STING‐GFP, but not the Y245F mutant, led to strong induction of IFN‐β mRNA, indicating that the Y245 residue of STING was needed for its ability to signal (Fig 6G). The kinetics of Ser366 phosphorylation of STING‐GFP and STING Y245F‐GFP were very similar (Fig 6H). The mutant STING did not bind IRF3, and IRF3 was not phosphorylated even in stimulated cells, but TBK1 bound to both WT and mutant STING and was phosphorylated (Fig 6I). The corresponding mouse STING mutant, Y244F, behaved similarly, when expressed in STING^−/− MEFs. There was little difference in S365 phosphorylation between WT and mutant STING (Fig EV4E), and TBK1 binding and its phosphorylation were unaltered in cells expressing the mutant; however, IRF3 was not phosphorylated nor did it bind to Y244F STING (Fig EV4D). Enhanced autophagy can be monitored biochemically by measuring the cellular content of LC3II, the lapidated form of LC3; as expected, gefitinib‐treated cGAMP‐stimulated cells had more LC3II (Fig 6K). Similarly, cGAMP‐treated EGFR^−/− cells had more LC3II compared to WT cells; however, the presence or absence of EGFR did not make any difference in the level of LC3II when an unrelated autophagy‐inducing agent, Torin, was used (Fig EV4F). The Y245F mutant also induced more LC3 lipidation upon cGAMP treatment (Fig 6J). The Y244F mutant of mouse STING, when expressed in STING^−/− MEF, mimicked the trafficking pattern of WT STING expressed in EGFR^−/− cells. After stimulation, it quickly relocated to autophagosomes (Fig 7A) and it never transited to the endosomes (Fig 7B). The corresponding human STING mutant, Y245F, when expressed in HeLa cells, showed similar trafficking pattern after stimulation (Fig EV5A and B). The above results demonstrated that the Y245F (Y244F in mouse) mutant of STING, expressed in WT cells, shared the same properties as WT STING expressed in cells lacking EGFR activity.

Figure 7. Y244F mSTING mutant translocates to autophagosomes but not late endosomes.

1. STING^−/− MEF cells were transfected with mSTING WT‐GFP and Y244F‐GFP (green) for 24 h, pretreated with gefitinib for 1 h, and transfected with cGAMP for the indicated time; cells were fixed and labeled with LC3 (autophagosome marker, red) antibody. Co‐localization (Col) is indicated by white dots. Scale bars represent 10 μm.
2. The cells with the same treatment as A and labeled with CD63 (late endosome marker).
3. A model for STING's trafficking routes in stimulated cells in the presence or the absence of EGFR.

Figure EV5. (related to Fig 7): Trafficking of the Y245F STING mutant.

1. HeLa cells expressing Y245F STING‐GFP (green) were treated with cGAMP for the indicated time, fixed and stained with LC3 antibody (red). Co‐localization (Col) is indicated by white dots. Scale bars represent 10 μm.
2. Same as in (A), but stained with CD63 antibody (red). Scale bars represent 10 μm.

Discussion

The cGAS/STING signaling pathway is emerging to be the critical pathway used by cells to respond to cytosolic DNA, often produced by invading microbes. In addition, it responds to cellular DNA that is leaked to the cytoplasm from the nucleus or the mitochondria in response to genotoxic stresses (Li & Chen, 2018). Although the DNA detected by cGAS is cytoplasmic, according to a recent report, the enzyme is tethered to the plasma membrane by its ability to bind phosphoinositides (Barnett et al, 2019). On the other hand, STING is an ER protein with a long cytoplasmic tail (Ishikawa & Barber, 2008). To activate the endosomal TLRs, which recognize intracellular nucleic acids, the ligands bind to the luminal side of the proteins, but unlike the endosomal TLRs, both ligand binding and signaling are carried out by the cytoplasmic domain of STING (Barber, 2015). However, for both classes of receptors, ligand‐induced conformational change is the triggering event for activating the signaling cascades by recruiting appropriate adaptors, kinases, and transcription factors to the receptor (Takeuchi & Akira, 2010). It is known that ligand‐induced ubiquitination, palmitoylation, and serine phosphorylation of specific residues of STING are required for its ability to signal (Tanaka & Chen, 2012; Mukai et al, 2016; Ni et al, 2017). Phosphorylation of STING Tyr245 has been reported before (Xia et al, 2019); our study confirmed that observation and determined the functional consequences of this phosphorylation. Moreover, EGFR, which mediates Tyr phosphorylation of TLR3 and TLR9 (Yamashita et al, 2012; Veleeparambil et al, 2018), was identified as the critical PTK for STING signaling as well. The need for EGFR‐mediated Tyr phosphorylation of STING was found to be true in a variety of human and mouse cells, indicating that the phenomenon described here is not cell type‐ or species‐specific. Moreover, the EGFR requirement for STING‐mediated IFN induction was true for HSV1 infection, a physiological activator of the STING pathway, not only in cultured cells but also in mice. It is interesting to note that eliminating myeloid cell‐specific EGFR expression in the mouse and global inhibition of EGFR in the mouse by gefitinib had similar effects on HSV 1‐induced pathogenesis (Fig 2D and E), indicating that myeloid cells are the major producers of IFN in HSV1‐infected mice. A STING mutant that does not induce IFN but triggers autophagy can inhibit HSV 1 replication in cell culture but only after cGAMP stimulation (Gui et al, 2019); in contrast, our virus experiments did not include cGAMP treatment, indicating that virally induced IFN was the primary restriction factor for HSV 1 replication both in vitro and in vivo.

Our experimental results, along with the existing literature, led us to formulate a working model for the STING trafficking steps and IRF3 activation by EGFR‐mediated STING signaling (Fig 7C). Ligand binding by STING leads to its EGFR binding in the ER, causing EGFR autophosphorylation at Tyr1068 and its activation. Enzymatically active EGFR phosphorylates Tyr245 of STING, the two proteins dissociate, and Tyr‐phosphorylated STING translocates slowly to the ERGIC compartment. This phosphorylation is not needed for ligand‐induced oligomerization of STING, but in its absence (EGFR^−/− cells, gefitinib‐treated cells, or cells expressing the Y245F STING mutant), STING moves from the ER to the ERGIC rapidly. In the ERGIC, both Tyr‐phosphorylated and unphosphorylated STING bind TBK1, which is activated by its autophosphorylation at Thr172; active TBK1 phosphorylates Ser366 of STING, a modification necessary for IRF3 recruitment. Because NF‐κB is activated by STING, when the protein is located in the ER or ERGIC, both Tyr‐phosphorylated and unphosphorylated STING can activate NF‐κB. The trafficking routes of Tyr245‐phosphorylated and unphosphorylated STING diverge sharply from this point on. The Tyr‐phosphorylated form moves to late endosomes where IRF3 is recruited and phosphorylated on Ser396 by STING‐bound TBK1 and activated IRF3 transits to the nucleus to induce transcription of IFN and other genes; eventually, STING moves to the lysosomal compartment and autolysosomes. In contrast, STING that is not phosphorylated at Tyr245 rapidly transits to autophagosomes and autolysosomes by a direct route avoiding the endosomes entirely; thus, even with phosphorylated Ser366, STING does not encounter IRF3 to activate it and drive IFN induction. The absence of EGFR activity itself might have aided the autophagy route of STING trafficking, because even without cGAMP stimulation there was a low level of autophagy in EGFR^−/− cells (Fig EV4F).

Unlike the global need of EGFR for all signaling by TLR3 and TLR9, the observations presented here clearly showed that the IRF3 branch, but not the NF‐κB branch, of STING signaling requires EGFR. Such dichotomy between the two branches of STING signaling has been observed before. Ni et al (2017) reported that K63‐linked ubiquitination of Lys224 of STING is required for its ability to activate IRF3, but not NF‐κB. Stempel et al (2019) observed that MCMV m152 protein binds STING and selectively blocks IRF3 activation. They also observed that the K288R mutant of STING can activate NF‐κB, but not IRF3. It remains to be seen whether EGFR‐mediated Tyr phosphorylation and modification of specific Lys residues of STING are causally connected. There are conflicting reports on the effects of Ser366 (S365 in mouse) phosphorylation on STING signaling. Tanaka and Chen (2012) reported that phosphorylation of this residue by TBK1 is required for IRF3 binding and activation. However, an earlier report (Konno et al) claimed that S366 phosphorylation by ULK1 selectively blocks IRF3 activation (Konno et al, 2013). In any event, our data indicate that the need of EGFR for IRF3 activation is not related to STING S366 phosphorylation. Even in the absence of EGFR activity, S366 (365) was phosphorylated in response to cGAMP stimulation, and in cells devoid of ULK1 and ULK2, the need for EGFR persisted.

Tyr phosphorylation of different components of the cGAS/STING signaling pathways has been recently reported. cGAS, TBK1, IRF3, and STING have been shown to be Tyr‐phosphorylated after cellular stimulation (Liu et al, 2017, 2018; Hu et al, 2019). Consistent with our findings, Xia et al (2019) observed that Y245 of STING is phosphorylated after HSV1 infection. However, their report (Hu et al, 2019) suggests that Src is the PTK responsible for STING phosphorylation and the Y245F mutation of STING reduced Src‐mediated phosphorylation. Upon transient transfection, Src co‐immunoprecipitated with STING and its overexpression caused STING Tyr phosphorylation. However, overexpression of transfected PTKs often leads to erroneous conclusions. Moreover, a key difference between our study and that by Hu et al is in the method of delivery of cGAMP to the cells. We transfected 8 μg/ml of the ligand, which delivers it directly to the cytoplasm, whereas the other group added 200 μg/ml cGAMP to the culture medium from where it was probably endocytosed. In any event, both EGFR and Src may mediate STING phosphorylation, which is reminiscent of the same requirement for TLR3 signaling, in which case EGFR binding precedes Src binding (Yamashita et al, 2012) and the two kinases target two different Tyr residues. Because the actions of the two kinases are interdependent, mutation of the first target Tyr residue eliminates phosphorylation of the other target Tyr as well. To avoid these complexities, we resorted to in vitro phosphorylation and mass spectrometric analyses of STING, which clearly showed that EGFR targets Y245 directly. Notwithstanding the issue of whether EGFR alone or EGFR and Src together mediate STING Tyr phosphorylation, the major finding by Xia et al is consistent with our finding of the functional consequences of Y245 phosphorylation. They observed that its de‐phosphorylation by the PTPN1/2 phosphatases leads to a diminution of IFN induction and enhanced STING degradation. Our report provides the mechanistic basis of their observations. In another relevant paper (Wu et al, 2019), Wu et al reported the role of HER2, a member of the EGFR family, in negatively regulating STING signaling by preventing STING‐TBK1 interaction. Similar to our observation, they did not observe STING‐EGFR interaction in unstimulated cells; however, they did not measure it after cGAMP stimulation. It remains to be seen how the opposing actions of EGFR (ErbB‐1) and HER2 (ErbB‐2) physiologically affect STING signaling; however, it is worth noting that Wu et al observed that a constitutively active EGFR enhances STING signaling regardless of the cellular levels of HER2.

It is unclear how Tyr245 phosphorylation affects activated STING trafficking. Unlike several constitutively active mutants of STING, which are not retained in the ER at all (Dobbs et al, 2015), the Y245F mutant needed ligand binding to transit to the ERGIC, but the migration was very rapid. Several vesicular proteins are known to interact with STING and affect its trafficking out of the ER and beyond. It is possible that Tyr245 phosphorylation of STING changes the avidity of its interaction with one or more of these proteins. For example, a weaker interaction of unphosphorylated STING with STIM1, which retains STING in the ER, will hasten its exit (Srikanth et al, 2019). Conversely, a stronger interaction with iRhom2, which facilitates trafficking, may have the same result (Luo et al, 2016). The preference of the direct route to autophagosomes may be caused by a stronger affinity of the unphosphorylated protein for p62/SQSTM1, which directs STING to that path and away from the IFN‐inducing route (Prabakaran et al, 2018). Similarly, unphosphorylated STING may enhance WIP12 and ATG5‐dependent LC3 lipidation and autophagosome formation (Gui et al, 2019). Regardless of the underlying mechanism, our results clearly demonstrate that Tyr245 phosphorylation profoundly affects STING trafficking routes and, consequently, its functional activities. Moreover, our observation provides a means for blocking selective actions of STING for experimental or therapeutic purposes. For example, EGFR inhibitors, which are widely used in treating cancer patients, can now be used to block IFN induction by STING in autoimmune and autoinflammatory diseases.

Materials and Methods

Reagents and antibodies

2′3′‐cGAMP (tlrl‐nacga23‐02) was purchased from InvivoGen. EGFR inhibitors gefitinib (S1025) and AZD9291 (S7297) were obtained from Selleckchem. Polyethyleneimine (PEI, 195444) was purchased from MP Bio. Poly (I:C) was purchased from GE Healthcare (27473201). Genistein was purchased from Calbiochem (345834). Lipofectamine 2000 (11668019) was obtained from Invitrogen. A comprehensive list of reagents and antibodies utilized is detailed in Table EV2.

Cell culture and transfection

Macrophages (BMDMs) were isolated and differentiated as described before (Yamashita et al, 2012). RAW264.7 cells (TIB‐71), HT1080 cells (CCL‐121), HEK293T cells (CRL‐11268), HeLa cells (CRM‐CCL‐2), MEF cells (CRL‐2907), and L929 cells (CCL‐1) were purchased from ATCC (Manassas, VA, USA), and primary STING^−/− MEF cells were a kind gift from Glen N. Barber (Ni et al, 2017). HeLa EGFR^−/− and mutant‐rescued cells (Chen et al, 2019) were a kind gift from Xiaoxia Li. The cells were maintained in DMEM containing 10% FBS and penicillin–streptomycin; 293XL cells were from InvivoGen (Veleeparambil et al, 2018), hSTING‐GFP and hSTING‐GFP mutant 293XL cells were generated by transfection, puromycin selection, and picking clones and maintained in DMEM containing 10% FBS, penicillin–streptomycin, blasticidin, and normocin. EGFR knockdown 293XL cells were generated by lentiviral transduction of human EGFR‐specific shRNA (Sigma, #TRCN0000121202) and selected by puromycin and picking the clones and a non‐targeting shRNA (#SHC002) as control. Lipofectamine 2000 was used for plasmid and 2′3′‐cGAMP transfection as per manufacturer's protocol. For transient transfection to 293XL cells, 2′3′‐cGAMP was transfected into related cells using polyethyleneimine. Approximately 1 μg of 2′3′‐cGAMP (final concentration for transfection: 8 μg/ml)/Poly (I:C) (final concentration for transfection: 10 μg/ml) and 2.25 μl of PEI (1 mg/ml, pH 7.0) were separately incubated in 250 μl of serum‐free DMEM for 5 min and then mixed together for 15 min and dropwise into the cell dish for the specific time.

RNAi‐mediated knockdown

ShRNA plasmid pLKO.1‐puro, which targets human EGFR (#TRCN0000121202), and non‐targeting control plasmid (SHC002) were purchased from Sigma (St. Louis, MO, USA). Target cells were transduced lentivirally in the presence of polybrene (8 μg/ml) following the manufacturer's instructions. Sixteen hours post‐infection, the media were removed, and cells were allowed to recover in complete growth media for 36–48 h before using the selection media containing puromycin (1 μg/ml) or hygromycin (100 ng/ml). Lentiviral shRNAs of pGIPZ were constructed via an inverse PCR method. ShULK1 sequence is ATTCATCAAAGTCCATGCG, and the pGIPZ ShULK1 vector contains puromycin selection marker. ShULK2 sequence is the same as previously reported (Ro et al, 2013), and the selection marker for ShULK2 is hygromycin. The pGIPZ and pLKO.1 shRNA vectors encoding shRNAs were co‐transfected with lentiviral packaging vectors pSPAX2 and pMD2.G into HEK293T cells.

Western blot

Cells were lysed in lysis buffer of 50 mM Tris–HCl, pH 7.4, 150 mM NaCl, 5 mM EDTA, 1 mM DTT, 1 mM PMSF, and cocktail (Roche) at 4 degrees for 15 min and centrifuged at 21,130 g at 4°C for 10 min; the supernatant was retained. The cell lysates were resolved by SDS–PAGE or used for IP. The proteins were electrophoresed on SDS–PAGE gels and then transferred to PVDF membrane (Bio‐Rad). Then, the membranes were kept in 5% skim milk in TBST buffer (150 mM NaCl; Tris, pH 7.4; and 0.1% Tween‐20) for 60 min at room temperature and then incubated with the primary antibody at cold room overnight. Western blot experiments were performed via the indicated antibodies and visualized by SuperSignal West Pico Chemiluminescent Substrate (Pierce Chemical) and film (Thermo Fisher, 34090). The Western blot image data were made by Adobe Illustrator.

Immunoprecipitation

For immunoprecipitations, cells were lysed in buffer containing 20 mM HEPES, pH 7.5, 150 mM NaCl, 10 mM NaF, 1.5 mM MgCl₂, 10 mM β‐glycerophosphate, 2 mM EGTA, 1 mM Na3VO4, 1% (v/v) Triton X‐100, 0.2% NP‐40, and protease inhibitors (Roche Applied Science, Indianapolis, IN, USA). Whole‐cell lysate was centrifuged for 10 min at 4°C and the supernatant used for immunoprecipitation via Protein A/G PLUS‐Agarose (Santa Cruz) and incubated with 3 μg of mouse or rabbit monoclonal antibodies overnight at 4°C overnight. The related beads were washed six times with IP buffer and boiled in SDS–PAGE buffer for 10 min at 95°C. The samples were separated by SDS–PAGE Gel and transferred onto PVDF membranes (Bio‐Rad). Western blot experiments were performed via the indicated antibodies and visualized by SuperSignal West Pico Chemiluminescent Substrate (Pierce Chemical).

In vitro EGFR kinase assay and Phos‐tag gel analysis

The EGFR kinase buffer contains 50 mM HEPES, 0.01% Tween‐20, 10 mM MnCl₂, 10 mM MgCl₂, 1 mM EGTA, 2.5 mM DTT, and 0.1 mM ATP, pH 7.4. Recombinant EGFR protein was purified from Sf9 insect cells (Boster, Catalog Number: PROTP00533‐1). 0.1 or 0.4 μg of EGFR and approximately 2 μg of GFP C‐terminal‐tagged recombinant STING and mutants which purified from 293XL cells via GFP trap (Chromotek, gta‐10) were added to the reaction buffer for 60 min at 30°C. All reactions were stopped by adding 5×SDS loading buffer and boiled for 10 min at 95°C for Phos‐tag gel‐related WB analysis. For analysis of phosphorylation of STING‐GFP, Phos‐tag (Wako) and MnCl₂ were added to regular 10% SDS–PAGE gels at the levels recommended by the manufacturer which is needed to slower migration of phosphorylated STING. Western blotting was performed normally after soaking the gels in 1 mM EDTA for 10 min to remove Mn²⁺. All other steps in this analysis were identical to normal SDS–PAGE and immunoblotting protocols.

RNA‐seq analysis and RT–PCR

For quantitative profiling of mRNA level via microarray, MEF cells were pretreated with DMSO or gefitinib (10 μM) for 1 h and stimulated with 2′3′‐cGAMP (8 μg/ml) for 6 h along with DMSO or gefitinib. Total RNA was isolated and treated with DNase I, and the RNA was further purified by using the RNeasy Kit (Qiagen). cDNA libraries were constructed by using TruSeq SBS Kit v3 (FC‐401‐3001; Illumina, San Diego, CA) with their protocol. Next‐generation sequencing was performed using Illumina HiSeq 4000 with 100‐bp single‐end reads at Genomics Core University of Chicago. RNA‐seq data analysis was done with the help of https://bioinfo.ca. We selected the level of cDNAs, which were induced at least twofold by 2′3′‐cGAMP, and quantified the inhibition index. The RNA‐seq data from this publication have been submitted to the BioProject database (http://www.ncbi.nlm.nih.gov/bioproject/609334) and assigned the BioProject ID PRJNA609334. For RT–PCR, RNA was isolated using Roche RNA Isolation Kit (Roche). cDNA was extracted by using ImProm‐II Reverse Transcription Kit (Promega), and 0.5 ng of cDNA was applied to 384‐well plate for real‐time PCR using Applied Biosystem's Power SYBR Green PCR Mix in Roche LightCycler 480 II. The levels of the induced mRNAs were normalized to 18S rRNA. The specificity was confirmed by analysis of the melting curves of the PCR products. The primer sequences used are described in Table EV2.

Mass spectrometry

STING protein was purified from 293XL‐hSTING‐GFP cells by using GFP trap beads (Chromotek, gta‐20). For protein digestions, the bands were cut from the gels as closely as possible and washed and destained in 50% ethanol and 5% acetic acid. The gel pieces were then dehydrated in acetonitrile, dried in a SpeedVac, and digested by adding 5 μl 10 ng/μl of trypsin or chymotrypsin, in 50 mM ammonium bicarbonate, followed by incubation overnight. The peptides were extracted into two portions of 30 μl each 50% acetonitrile and 5% formic acid. The combined extracts were evaporated to < 10 μl in a SpeedVac and then resuspended in 1% acetic acid to make up a final volume of ~ 30 μl for LC‐MS analysis. The LC‐MS system was a Thermo Scientific Fusion Lumos Tribrid Mass Spectrometry System. The HPLC column was a Dionex 15 cm × 75 μm id Acclaim PepMap C18, 2 μm, 100 Å reversed‐phase capillary chromatography column. Five microliters of volume of the extract was injected, and the peptides, eluted from the column in an acetonitrile, 0.1% formic acid gradient at a flow rate of 0.25 μl/min, were introduced into the source of the mass spectrometer online. The micro‐electrospray ion source was operated at 2.5 kV. The digest was analyzed in both a survey manner and a targeted manner. The survey experiments were performed using the data‐dependent multitask capability of the instrument, acquiring full scan mass spectra to determine peptide molecular weights and product ion spectra to determine amino acid sequences in successive instrument scans. The LC‐MS/MS data were searched with the programs Mascot and Sequest against both the full human UniProtKB database and specifically against the sequence of STING. The parameters used in this search include a peptide mass accuracy of 10 ppm, fragment ion mass accuracy of 0.6 Da, carbamidomethylated cysteines as a constant modification, and oxidized methionine and phosphorylation at Y as a dynamic modification. The results were filtered based on Mascot ion scores and SEQUEST XCorr scores. The positively identified phosphopeptide around Y245 was manually validated. The targeted experiments involve the analysis of specific STING peptide around Y245. The chromatograms for these peptides were plotted based on known fragmentation patterns, and the peak areas of these chromatograms were used to determine the extent of phosphorylation (Willard et al, 2003; Waitkus et al, 2014).

Immunofluorescence staining and confocal microscopy

Cells were cultured in 4‐well chamber slides and transfected with the indicated plasmids via Lipo2000 for 24 h. The cells were then fixed in 4% paraformaldehyde overnight in cold room and then permeabilized by PBS with 0.1% Triton X‐100 for 10 min at room temperature. The cells were washed four times with PBS and then blocked with 10% BSA in PBS for 1.5 h at RT. Then, cells were incubated with related antibody overnight at cold room. After washing three times with PBS for 10 min, respectively, the cells were incubated with goat anti‐rabbit Alexa Fluor 594 (Invitrogen), anti‐mouse Alexa Fluor 594 (Invitrogen), or anti‐rabbit Alexa Fluor 647 (Invitrogen). Objects were mounted using VECTASHIELD/DAPI, and images were taken by confocal laser scanning microscopy (Leica TCS SP8). Images were processed with Leica LCS software. Fiji (ImageJ) software co‐localization plug‐in was used for determining co‐localization of two or three proteins where the white dots represent co‐localization; for each independent figure, we set higher parameters to the plug‐in software to get clear results and the control and negative groups do not appear as white spots. Images were photographed at random positions for each condition. Pearson's co‐localization coefficient was also measured by Fiji (ImageJ).

Mouse experiments

EGFR^fl/− LysM^Cre+/+ and WT mice (all in C57Bl/6 genetic background) were used for mouse experiments. WT mice were provided by the Jackson Laboratory. The EGFR^fl/− LysM^Cre+/+ mice were generated as previously described (Veleeparambil et al, 2018). HSV‐1 (KOS) propagation and infection were performed as previously described (Blaho et al, 2005; Wang et al, 2016). Virus titers were determined by TCID50 assay. Male mice were injected with 5 × 107 pfu of HSV‐1 intraperitoneally. All mice procedures were approved by IACUC.

Quantification and statistical analysis

Error bars for qPCR analysis represent the standard deviation data which were collected from three or more independent biological experiments. For Western blot, co‐IP, and confocal experiments, unless indicated otherwise, results are representative of at least two independent experiments. The differences for statistical significance between two groups were tested via two‐tailed t‐test using Prism 5 software (GraphPad). All values are means ± SD of the indicated independent experiments. ^NS P > 0.05; *P < 0.05; **P < 0.01, ***P < 0.001.

Author contributions

CW designed, performed, and interpreted the experiments and participated in writing the manuscript. XW designed, performed, and interpreted the experiments. MV supervised experimental protocol development and edited the manuscript. PMK generated critical reagents and assisted in experiments. BW performed and interpreted the mass spectrometric analyses. SC provided critical conceptual input and important reagents. GCS designed and supervised the project, secured funding for its execution, and wrote and edited the manuscript.

Conflict of interest

The authors declare that they have no conflict of interest.

Supporting information

Expanded View Figures PDF

Table EV1

Table EV2

Review Process File

Source Data for Figure 3

Source Data for Figure 4

Source Data for Figure 6

The EMBO Journal (2020) 39: e104106

Data availability

Deposited data and new reagents will be made available readily upon request. The RNA‐seq data from this publication have been submitted to the BioProject database and assigned the BioProject ID PRJNA609334; https://bioinfo.ca.