Constitutively Bound EGFR–Mediated Tyrosine Phosphorylation of TLR9 Is Required for Its Ability To Signal Free

Production of inflammatory cytokines and IFNs by innate immune cells is triggered by TLRs, which are pattern recognition receptors for pathogen-associated molecular patterns and damage-associated molecular patterns (1–4). Upon activation, TLRs recruit specific cytoplasmic adapters that nucleate the assembly of signaling proteins, protein kinases, and transcription factors; the activated transcription factors translocate to the nucleus and drive induced transcription of the target genes. TLRs are transmembrane proteins; their ectodomains interact with the ligands, whereas their short C-terminal regions are in the cytoplasm where they signal. Some TLRs are located on the plasma membrane, and others are on the endosomal membrane (5). TLR9 is an endosomal DNA-recognizing receptor that is primarily expressed in myeloid cells and activated by microbial pathogen-associated molecular patterns and mammalian damage-associated molecular patterns; it can provide protection against pathogenesis caused by selected bacteria and viruses (6, 7). In human cancers, TLR9 can be beneficial or detrimental. In lung cancer, TLR9 expression in tumor-infiltrating mononuclear cells is associated with a worse outcome (8, 9); in contrast, high TLR9 expression in triple-negative breast cancer is associated with better survival (10).

Experimentally, unmethylated CpG-rich oligodeoxynucleotide (ODN) is used to stimulate TLR9; it is endocytosed and binds to the ectodomain of TLR9 in the endosomal lumen. TLR9 dimerizes in endoplasmic reticulum and translocates to endolysosomes, a process regulated by UNC93B1, a membrane protein. In endolysosomes of mouse cells, TLR9 is proteolytically cleaved and undergoes ligand-induced conformational changes, leading to posttranslational modifications of amino acid residues in its cytoplasmic domain to generate a functional receptor (11–15). TLR9-mediated induction of proinflammatory cytokines and IFN may originate from two membrane compartments (16). To signal, TLR9 uses MyD88 as the adapter protein and IKKs and IRAKs as the protein kinases to activate transcription factors. Genome-wide RNA interference screening identified many proteins that affect TLR9 trafficking or signaling (17). Recent observations indicate that, to signal, many TLRs require ligand-dependent phosphorylation of specific tyrosine (Tyr) residues in their cytoplasmic domains. The phosphotyrosine residues of the TLR cytoplasmic domains can potentially serve as docking sites for signaling proteins that contain SH2 domains and, thereby, expand the repertoire of signaling pathways. Ligand-dependent Tyr phosphorylation of TLR9 has been reported; however, the specific phosphorylated Tyr residue and the relevant protein Tyr kinase (PTK) have not been identified (4, 18). A Tyr-containing motif, in the cytoplasmic domain of TLR9, is important for its appropriate intracellular localization and ability to signal; Tyr⁸⁸⁸, a part of this motif, is required for TLR9 Tyr phosphorylation, although this residue itself is not phosphorylated (19). CpG ODN treatment of cells causes rapid activation of Src family PTKs, such as Hck and Lyn, without any involvement of TLR9; this leads to the activation of another PTK, Syk, which binds to TLR9 (20).

Our investigation of the biochemistry of TLR3 signaling led to identification of the epidermal growth factor receptor (EGFR) as the critical PTK required for TLR3 signaling. EGFR binds TLR3 only after ligand stimulation, recruits another PTK, Src, and the two kinases phosphorylate two specific Tyr residues in the cytoplasmic domain of TLR3. This chain of events is required for the recruitment of the adapter, TRIF, and subsequent signaling (21). For TLR4 signaling in mouse myeloid cells, only the IRF3-mediated branch of endosomal signaling requires EGFR; however, no physical interaction between EGFR and TLR4 has been noted (22). In this article, we report that EGFR kinase activity is required for all TLR9 signaling in vitro and in vivo. However, unlike TLR3, endosomal TLR9 binds EGFR constitutively.

CpG oligonucleotides (CpG-A and CpG-B) were purchased from Integrated DNA Technologies. The EGFR inhibitors gefitinib (Gf) and erlotinib were obtained from Selleckchem (Santa Cruz, CA), and AG1478 was from Calbiochem. Phosphotyrosine Ab was obtained from Millipore; Abs against EGFR, MyD88, p-AKT, p-IκBα, p-ERK, p-JNK, p-TBK1, AKT, IκBα, TBK1, and actin were from Cell Signaling Technology; hemagglutinin (HA) Ab was purchased from Abcam; and Ab against murine Ifit proteins were raised in our laboratory (23). TLR4 Ab was obtained from Santa Cruz Biotechnology. TLR9 Ab (clone ABM1C51), which recognizes the N-terminal regions of murine and human TLR9, was from Abeomics. Cetuximab (Erbitux; Bristol-Myers Squibb) was obtained from the Cleveland Clinic pharmacy. d-galactosamine (GalN) was purchased from Sigma-Aldrich. Polyinosinic-polycytidylic acid [poly(I:C)] was obtained from GE Healthcare, and Lipofectamine 2000 (LF2000) was obtained from Invitrogen.

Primary bone marrow–derived dendritic cells (BMDCs), bone marrow–derived macrophages (BMDMs), and plasmacytoid dendritic cells (pDCs) were isolated and differentiated as described previously (24). RAW264.7 cells (American Type Culture Collection) were maintained in DMEM containing 10% FBS and penicillin-streptomycin. 293XL, 293XL–human TLR9–HA cells, and 293-CD14.MD2.TLR4 cells were from InvivoGen and were maintained in DMEM containing 10% FBS, penicillin-streptomycin, Normocin, and blasticidin. EGFR-knockdown 293XL–human TLR9–HA cells were generated by lentiviral transduction of human EGFR–specific short hairpin RNA (shRNA) (product number TRCN0000121202; Sigma-Aldrich), followed by selection under puromycin; a nontargeting shRNA (product number SHC002) was used as control. HT1080–human TLR9 (with double epitope tags, YFP and Flag) cells were generated by lentiviral transduction.

Four micrograms of poly(I:C) with 10 μl of LF2000 were incubated in 250 μl of Opti-MEM (10 μl of LF2000 was incubated in 250 μl of Opti-MEM as control) for 30 min and then added to the cell culture medium containing 10% FBS.

Wild-type (WT), EGFR^fl/−, and EGFR^fl/− LysMCre^+/+ mice (all on the C57BL/6 genetic background) were used for the experiments. WT mice were obtained from The Jackson Laboratory. EGFR^fl/fl mice, originally generated by Dr. D. Threadgill’s laboratory at the University of North Carolina (25), were a gift from Dr. X. Li’s laboratory (Cleveland Clinic). EGFR^fl/− mice were generated in the laboratory by crossing EGFR^fl/fl mice with CMVCre⁺ mice (The Jackson Laboratory). EGFR^fl/− LysMCre^+/+ mice were generated by crossing EGFR^fl/− mice with LysMCre^+/+ mice (The Jackson Laboratory). To induce septic shock, 8–10-wk-old WT, EGFR^fl/−, or EGFR^fl/− LysCre^+/+ mice were injected i.p. with CpG (30 μg per mouse) along with GalN (20 mg per mouse), and their survival was monitored for 5 d. To investigate the effect of EGFR inhibitor on CpG-induced septic shock, the mice were orally gavaged with Gf (2.4 mg per mouse) or vehicle (DMSO), followed by i.p. injection of CpG and GalN (26, 27). Blood samples were obtained by cardiac puncture 1 h later, and serum was isolated. The livers were harvested from the animals at 8 h post-CpG/GalN treatment for histopathological analyses. All mice procedures were approved by the Institutional Animal Care and Use Committee.

RNA was isolated using an RNA Isolation Kit (Roche). cDNA was prepared using an ImProm-II Reverse Transcription System (Promega), and 0.5 ng of cDNA was applied to a 384-well plate for real-time PCR using Power SYBR Green PCR Master Mix (Applied Biosystems) and a Roche LightCycler 480 Instrument II. The expression levels of the induced mRNAs were normalized to 18S rRNA or RPL32 mRNA.

RAW264.7 cells (in biological triplicate) were treated with DMSO or Gf (10 μM) for 1 h and stimulated with CpG (10 μg/ml) for 6 h, together with DMSO or Gf. Total RNA was isolated and treated with DNase I, and RNA was further purified using an RNeasy Kit (QIAGEN). Purified RNA was analyzed in a MouseRef-8 gene array, and data analysis was carried out using GenomeStudio v2011.1 (both from Illumina). We selected the mRNAs that were induced ≥2-fold by CpG and quantified the inhibition index (E/F) (Supplemental Table I). All datasets have been submitted to the National Center for Biotechnology Information/Gene Expression Omnibus under accession number GSE97366 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE97366).

Culture supernatants from CpG-treated cells were used for quantification of secreted TNF-α and IFN-β by ELISA, using the manufacturer’s instructions. Blood was collected by cardiac puncture from anesthetized mice into serum separator tubes (BD Biosciences), and sera were isolated following the manufacturer’s protocol for measuring the secreted cytokines. TNF-α levels were assessed in serum samples using a mouse TNF-α ELISA kit (eBioscience).

A Duolink In Situ Detection kit (product number DUO92008; Sigma-Aldrich) was used to perform a proximity ligation assay, following the manufacturer’s instructions. In brief, cells were grown on glass coverslips. Cells were incubated with wheat germ agglutinin 633 (Thermo Fisher Scientific) for 10 min at room temperature to label the plasma membrane, fixed with 4% paraformaldehyde, and permeabilized with 0.5% Triton X-100, followed by blocking with 2.5% goat serum. Cells were incubated overnight with primary Abs, rabbit Ab targeting Flag (Cell Signaling Technology) and mouse Ab targeting HA (Sigma-Aldrich), in 2.5% goat serum. Proximity ligation assay probes corresponding to the primary Abs, Duolink In Situ PLA Probe Anti-Mouse MINUS (product number DUO92004; Sigma-Aldrich) and Duolink In Situ PLA Probe Anti-Rabbit MINUS (product number DUO92005; Sigma-Aldrich), in 2.5% goat serum were used for further incubation for 1 h at 37°C, followed by incubation with a DNA ligase diluted in ligation buffer for 30 min at 37°C. Finally, cells were incubated with a DNA polymerase diluted in amplification buffer for 100 min at 37°C and mounted using VECTASHIELD/DAPI. Confocal images were acquired using a Leica laser scanning confocal microscope.

Human HT1080–TLR9 cells were grown on glass coverslips. The cells were fixed with 4% paraformaldehyde for 20 min and permeabilized with 0.2% Triton X-100 for 20 min. The cells were blocked with 5% normal goat serum for 1 h and then labeled overnight with anti-EGFR (C74B9; Cell Signaling Technology) and anti-EEA1 (BD Transduction Laboratories) to stain early endosomes. Goat anti-Rabbit Alexa Fluor 594 and goat anti-Mouse Alexa Fluor 647 (both from Invitrogen) (1 h) were used as secondary Abs, respectively. Objects were mounted using VECTASHIELD/DAPI, and images were acquired by confocal laser scanning microscopy (TCS SP8; Leica). Images were processed with Leica LCS software. The ImageJ Colocalization plugin was used for determining colocalization of two proteins, where the white dots represents colocalization.

Livers were harvested, fixed in formalin (Sigma-Aldrich), and embedded in paraffin. The 5-μm sections were cut and stained with H&E to assess liver pathology. Immunohistochemistry staining for infiltrated macrophages was performed using MAC-2 Ab. In brief, Ag retrieval was performed using a Tris/borate/EDTA buffer (DISCOVERY CC1; Ventana Medical Systems) (pH 8–8.5) for 32 min at 95°C. Slides were incubated with MAC-2 at 1:1800 dilution (CEDARLANE) for 40 min at room temperature. The Ab was visualized using a biotinylated rabbit anti-rat secondary Ab at 1:200 dilution (Vector Laboratories) and a DAB Map Detection Kit (Ventana Medical Systems). Lastly, the slides were counterstained with hematoxylin and bluing agent.

All statistical analyses were performed using GraphPad Prism 5.02 software. The p values were calculated using a two-tailed unpaired Student t test. The p values for survival curves were calculated using a log-rank test.

In view of our observation that TLR3 and TLR4 signaling requires EGFR activity, we inquired whether the same is true for TLR9 signaling. TLR9 activation by CpG ODN treatment of myeloid cells induced the transcription of many cytokine mRNAs, including Tnf-α and IFN-β secretion; such induction in primary mouse macrophages was completely blocked by Gf, an inhibitor of EGFR kinase (Fig. 1A). Testing of the Gf dose response indicated that 50% inhibition was achieved with an inhibitor concentration of 1–5 μM (Supplemental Fig. 1A); the high dose may have been needed because the inhibitor had to reach intracellular EGFR, which activated endosomal TLR9. To exclude any off-target effect of Gf, we pursued a genetic approach; TLR9 signaling was totally blocked in primary macrophages obtained from mice in which the EGFR gene had been selectively knocked out in myeloid cells (Fig. 1B). Similar to macrophages, in BMDCs, Gf inhibited the induction of Tnf mRNA and Ifit2 (an IFN-induced gene) mRNA (Fig. 1C). In contrast, in the same cells, as reported previously (22), TLR4-mediated Tnf mRNA induction by LPS was not inhibited by Gf (Supplemental Fig. 1B). A different inhibitor of EGFR, AG1478, had similar inhibitory effects on Il6 and Ifit2 induction by TLR9 (Supplemental Fig. 1B). In pDCs, Gf inhibited Tnf induction, measured at the mRNA and protein levels (Fig. 1D); inhibition was also observed for three other mRNAs (Supplemental Fig. 1C). Similarly, in the monocytic RAW264.7 cell line, induction of Tnf, Il6, and Ifn-β mRNAs by CpG ODN was strongly inhibited by Gf (Fig. 2A, Supplemental Fig. 2A); two other inhibitors of EGFR kinase, erlotinib and AG1478, had similar effects (Supplemental Fig. 2B). To ascertain whether the inhibitory effects of Gf on gene induction by TLR9 were global, we compared gene-expression profiles, by microarray analyses, using RNAs isolated from untreated, CpG ODN–treated, Gf-treated, and CpG ODN plus Gf–treated cells. It was clear that CpG ODN could not induce any gene in Gf-treated cells (Supplemental Table I). None of the 170 mRNAs that were induced >2-fold by CpG ODN were induced if the cells were also treated with Gf. The microarray results were verified by quantitative RT-PCR (qRT-PCR) analyses of four induced mRNAs that had not been examined before (Fig. 2B). The above results demonstrated that EGFR kinase activity was globally required for gene induction by TLR9 signaling.

Gene induction by TLR9 requires the activation of transcription factors, such as NF-κB and IRF, which, in turn, requires the activation of specific signaling kinases; we inquired whether TLR9 stimulation could activate these kinases if EGFR activity was blocked. NF-κB is activated by its release from IκB, which is degraded upon its phosphorylation. As expected, TLR9 stimulation caused IκBα phosphorylation and degradation, but both processes were blocked by Gf (Fig. 3A). Similarly, activation of ERK1, ERK2, JNK1, and JNK2, as measured by their phosphorylation, was impaired by Gf treatment (Fig. 3B). TBK1, which phosphorylates and activates IRF, also was not activated in the presence of Gf (Fig. 3C). The same was true for AKT, which is activated by PI3K and required for the full transcriptional activity of NF-κB and IRF (Fig. 3D).

To facilitate our investigation of the biochemical basis of the observed need for EGFR for TLR9 signaling, we used 293XL cells expressing human TLR9 with an HA-epitope tag at its C terminus. Unlike murine TLR9, no cleavage of human TLR9 could be detected in this cell line, even after ligand stimulation; an Ab that recognizes the N-terminal regions of murine and human TLR9s readily detected the full-length human TLR9 (FL; upper arrow, Fig. 4A, upper panel) but not a putative N-terminal fragment (N; lower arrow, Fig. 4A, upper panel). Similarly, HA Ab detected the full-length protein (FL; upper arrow, Fig. 4A, lower panel) but not a putative C-terminal fragment (C; lower arrow, Fig. 4A, lower panel). The absence of N-terminal and C-terminal fragments having a m.w. similar to their murine counterparts prompted us to ensure that the signaling properties of the TLR9-expressing 293XL cells were similar to those of murine myeloid cells. CpG ODN robustly induced TNF and IFN-β mRNAs in these cells, and the induction was totally abolished by Gf treatment (Fig. 4B).

Because EGFR has several isoforms with kinase activity, we determined which isoform is functionally required for TLR9 signaling in our system. For this purpose, expression of different EGFR isoforms was knocked down by respective shRNAs, clonal cell lines were established, and TLR9 signaling was assessed. In one such clone (clone 3), the expression of an shRNA for EGFR1 (erbB1) strongly impaired the expression of the EGFR1 protein without affecting the expression of TLR9 (Fig. 4C); neither TNF mRNA nor IFN-β mRNA was induced by CpG ODN treatment of these cells (Fig. 4D). In contrast, as expected, TNF mRNA induction in the same cells, by poly(I:C) transfection (retinoic acid inducible gene I–like receptor signaling), was unimpaired (Fig. 4E). The above results demonstrated that our experimental cell line 293XL-TLR9 shares the properties of primary myeloid cells; moreover, EGFR1 is the relevant isoform that is required for TLR9 signaling.

To understand the basis of the requirement for EGFR Tyr kinase activity for TLR9 signaling, we examined whether TLR9 itself is Tyr phosphorylated; we observed ligand-dependent and transient Tyr phosphorylation of TLR9 (Fig. 5A), which required EGFR kinase activity (Fig. 5B). Similar results were obtained with EKD cells in which EGFR expression had been knocked down (Fig. 5C). The lack of the total elimination of TLR9 phosphorylation (Fig. 5B, 5C) could be due to the residual EGFR in EKD cells (Fig. 4C, lane 3) or the action of another protein Tyr kinase. Assembly of the TLR9 signaling complex requires interaction of TLR9 with the adapter protein MyD88. We detected such an interaction, which, as expected, was ligand dependent and transient (Fig. 5D); note that MyD88 interacted with full-length TLR9 to form the signaling complex. More importantly, this interaction also was EGFR kinase dependent (Fig. 5E). The above results suggested that ligand-induced EGFR-mediated Tyr phosphorylation of TLR9 leads to MyD88 recruitment by TLR9 and triggering of the consequent signaling pathways.

Because EGFR was required for ligand-induced Tyr phosphorylation of TLR9, we wondered whether the two proteins physically interact. We used two experimental approaches to measure EGFR–TLR9 interaction. We could readily demonstrate their interaction by coimmunoprecipitation assays (Fig. 6A, left panel); their interaction was not affected by Gf and, more surprisingly, it did not require ligand stimulation. In contrast, we did not detect any interaction between EGFR and TLR4, even after LPS stimulation (Fig. 6A, right panel). The above conclusion was confirmed by a cell-based imaging assay; in the proximal ligation Duolink assay, a strong signal is generated if two proteins are in close proximity in a cell (28). As shown in Fig. 6B, cells scored positive for TLR9–EGFR interaction before and after CpG ODN treatment. In contrast, as reported before, TLR3–EGFR interaction required ligand stimulation (21). The above results demonstrated that EGFR was bound to TLR9 constitutively, but ligand stimulation was required for TLR9 phosphorylation and signaling.

Because cytoplasmic signaling is elicited by endosomal membrane-bound TLR9, we postulated that TLR9–EGFR interaction occurs, not on the plasma membrane, but on the internal membranes of the cell. In cells exposed to EGFR ligands, EGFR is found not only on the plasma membrane but also on the endosomal membrane (21). However, in cells cultured in serum-free medium, which does not contain EGF or other EGFR ligands, EGFR is localized almost exclusively on the plasma membrane. We observed that, in serum-starved (SS) cells, there was no interaction between TLR9 and EGFR, even after CpG ODN stimulation (Fig. 7A). Consequently, there was no TLR9 phosphorylation (Fig. 7B) or gene induction (Fig. 7C) in SS cells, indicating that TLR9 interaction with intracellular EGFR was required for its ability to signal. This conclusion was supported by the observation that cetuximab, an EGFR Ab that blocked cell surface EGFR signaling (Fig. 7D), could not block gene induction by CpG ODN (Fig. 7E). Unlike Gf, cetuximab, could not enter the cell and, hence, could not inhibit the function of endosomal EGFR. Confocal microscopy confirmed colocalization of TLR9 and EGFR on the early endosomal membrane. Colocalization of an early endosomal marker (EEA1), EGFR, and TLR9-YFP, visualized by three colors (Fig. 7F, three left panels), was confirmed by image analysis, using a software that produces white dots if two colors are colocalized (Fig. 7F, three right panels).

To establish the requirement for EGFR for TLR9 signaling in vivo, we used a liver injury model in mice. When CpG ODN is injected into GalN-sensitized mice, TLR9-induced TNF-α causes rapid liver damage and death (26). To test the need for EGFR in this process, mice were treated with Gf before CpG ODN challenge. TNF-α was strongly induced in the serum of CpG ODN–injected mice, but not if they were pretreated with Gf (Fig. 8A). The liver architecture of untreated and CpG ODN plus Gf–treated mice was very similar; in contrast, massive hemorrhage and cell apoptosis were present in the livers of mice injected with CpG ODN (Fig. 8B). There was pronounced macrophage infiltration in the livers of CpG ODN–treated mice, but the absence of EGFR or its kinase activity had no effect on the magnitude of the infiltration (Fig. 8C).

The observed liver damage caused the death of all CpG-injected mice within 12 h, whereas 80% of Gf-treated mice survived (Fig. 9A). We used this mouse-survival assay to ascertain the shortest time period needed for Gf administration, prior to CpG ODN treatment, to protect the mice. Sixteen hours of pretreatment (Supplemental Fig. 3A) and 1 h of pretreatment (Supplemental Fig. 3B) were effective; surprisingly, even coadministration of CpG ODN and Gf was protective (Supplemental Fig. 3C). These results clearly demonstrated that TLR9 signaling in vivo required EGFR kinase activity. To strengthen the conclusion drawn from pharmacological intervention of EGFR activity in vivo, we tested mice in which the EGFR gene has been manipulated. The need for EGFR for pathogenesis was supported by the observation that mice harboring only one allele of the EGFR gene were less susceptible to TLR9-induced pathogenesis (Fig. 9B). To ascertain the cell type in which EGFR expression was required to elicit the phenotype, we selectively deleted both alleles of the EGFR gene in myeloid cells; the corresponding mice were quite resistant to pathogenesis (Fig. 9B). These results indicated that EGFR-dependent TLR9 signaling in myeloid cells was essential for the deleterious effects of CpG ODN administration to mice.

This study provides support for the paradigm that endosomal TLRs require ligand-induced EGFR-mediated Tyr phosphorylation to trigger signaling; our results demonstrated an absolute need for EGFR kinase activity for TLR9 to induce gene transcription. This is reminiscent of a similar need for TLR3 signaling, but not for TLR4 signaling, in which only the IRF3 branch requires EGFR (21, 22). Among the different ligands of TLR9, CpG A ODN is a better inducer of IFN than CpG B ODN, which was used for all of our experiments. EGFR kinase activity was required for IFN induction by CpG A ODN as well (data not shown). Consistent with the need for EGFR for all gene induction by TLR9, its PTK activity was also needed to activate the signaling kinases.

Several highly specific pharmacological inhibitors of EGFR, Gf, erlotinib, and AG1478, inhibited gene induction by TLR9; however, for effective inhibition, the required dose of Gf was high compared with what is needed to inhibit biochemically purified EGFR in vitro (29). Because our assays were cell based, inefficient cellular uptake and intracellular degradation might have contributed to inefficient inhibition. Moreover, in contrast to inhibiting cell surface EGFR, to inhibit TLR9, Gf had to reach endosomal EGFR, which binds to TLR9. Nonetheless, to rule out off-target effects of high doses of Gf, we sought genetic evidence for the requirement for EGFR in TLR9 signaling; knocking down the expression of EGFR (Fig. 4D) or knocking out the EGFR gene itself (Fig. 1B) confirmed the conclusion that EGFR was necessary for TLR9 to signal.

To facilitate the analysis of EGFR’s involvement in TLR9 signaling, we took advantage of the cell line 293XL-TLR9, which expresses epitope-tagged human TLR9. Unlike murine myeloid cells, TLR9 was not cleaved in human 293 cells, and neither its N-terminal fragment nor its C-terminal fragment was detected (Fig. 4A); similar observations had been made by Latz et al. (15). We confirmed that full-length TLR9 was the signaling receptor by demonstrating its ligand-dependent Tyr phosphorylation and consequent recruitment of MyD88. TLR9 signaling in 293XL-TLR9 cells had characteristics similar to those in mouse myeloid cells; most importantly, these cells also required EGFR kinase activity for CpG ODN–induced gene induction, and like TLR3, TLR9 used the erbB1 isoform of EGFR for this purpose. In these cells, we observed transient Tyr phosphorylation of TLR9 in response to CpG ODN stimulation, a process that required EGFR kinase activity, which was also required for the recruitment of MyD88 to TLR9. From these observations, we concluded that EGFR directly, or indirectly through the acquisition of another PTK, phosphorylates TLR9, which leads to the binding of MyD88, triggering the complete signaling pathway. It is worth noting that a basal low level of Tyr phosphorylation of TLR9 was detectable in unstimulated cells (Fig. 5A–C), even in the presence of Gf (Fig. 5B). The role and the nature of the basal phosphorylation of TLR9 remain to be investigated. There are six Tyr residues in the cytoplasmic region of TLR9, and it is possible that basal and CpG ODN–induced phosphorylation of multiple Tyr residues is required for the receptor to signal. A specific Tyr, Tyr⁸⁸⁸, has been shown to be part of a structural motif required for correct intracellular localization of TLR9; although this Tyr is not phosphorylated, it is required for TLR9 Tyr phosphorylation and signaling (19).

The functional requirement for EGFR for ligand-induced activation of TLR9 signaling was consistent with the observed physical interaction between the two proteins. Their interaction was constitutive; it did not require ligand stimulation or EGFR kinase activity. This is in contrast to the nature of the EGFR interaction with TLR3, which depends on ligand-induced conformational change of the receptor, allowing accessibility of EGFR to its binding site on TLR3 (21). In the case of TLR9, most probably, ligand-induced conformational change allows prebound EGFR to access its target Tyr moiety on TLR9 and phosphorylate it. It is likely that the two proteins encounter each other on the endosomal membrane. In SS cells, without exogenous ligands, EGFR resides almost exclusively on the plasma membrane; we observed no interaction with TLR9 and no Tyr phosphorylation of TLR9 in those cells (Fig. 7). However, in cells cultured in media containing growth factors, EGFR is abundant in internal membranes, including endosomal membrane where TLR9 binds to it. Consequently, inhibiting EGFR kinase activity on the cell surface by a neutralizing Ab had no effect on TLR9 signaling, whereas cell-permeable small chemical inhibitors, such as Gf, could inhibit the action of endosomal EGFR and block TLR9 signaling. It is surprising that EGFR was not identified as a TLR9 partner in a previous study (17); however, because genome-wide RNA interference screening was used in that study, it is possible that the target cells were not viable in the absence of EGFR.

To verify the need for EGFR for TLR9 signaling in vivo, we resorted to a convenient model of CpG ODN–mediated liver injury in mice (26). In this model, administration of CpG stimulates TLR9 to induce TNF production, which promotes massive apoptotic death of hepatocytes in GalN-sensitized mice, causing fulminant liver failure and consequent death. We tested whether pretreatment of the mice with Gf would block TLR9 signaling and its deleterious effects in this model. Indeed, pretreatment, or even cotreatment (Supplemental Fig. 3), with Gf was very effective in preventing death caused by CpG ODN administration. As expected, the protected mice suffered lesser liver damage, and the level of TNF in their circulation was much lower. These pharmacological data were supported by results obtained from genetic manipulation of EGFR expression. Mice harboring only one functional allele of the EGFR gene were less susceptible than WT mice. Because TNF is primarily produced by the macrophages of CpG-treated mice (26), when we deleted the remaining allele of the EGFR gene only in those cells, the mice were highly protected from CpG-induced death. The pharmacological and genetic data suggest that TLR9 expressed in myeloid cells is activated by CpG ODN to produce TNF, a process that requires the physical and functional presence of EGFR in those cells. Thus, the in vivo results are in complete agreement with the mechanistic conclusions drawn from our in vitro experiments.

The requirement for EGFR kinase activity for signaling and gene induction by endosomal TLRs can be potentially exploited for clinical benefits. Especially in situations in which TLR signaling is harmful, such as autoimmunity, blocking EGFR therapeutically may be beneficial. Many EGFR chemical inhibitors and Abs are widely used clinically to treat specific cohorts of cancer patients; thus, their safety and efficacies are well documented. These drugs can be repurposed for treating patients suffering from hyperinflammation due to potent TLR signaling. On the flip side, in cancer patients receiving EGFR blockers on a long-term basis, signaling by the endosomal TLRs may be impaired. These TLRs are often activated by endocytosis of cellular DNA or RNA produced by apoptotic cancer cells; the consequent cytokine-mediated inflammation may promote disease progression, depending on the specific type of cancer. In such cases, the beneficial effects of EGFR inhibitors may be mediated by impairing cancer cell proliferation, as well as by blocking inflammatory cytokine production.

We thank Xiaoxia Li for providing EGFR^fl/fl mice, Volker Fensterl for helpful suggestions, and Judy Drazba for help with confocal microscopy (Leica SP8 confocal microscope).

The authors have no financial conflicts of interest.