TBK1 Is a Synthetic Lethal Target in Cancer with VHL Loss

Estimated new cases and deaths from renal (renal cell and renal pelvis) cancer in the United States in 2014 were 63,920 and 13,860, respectively (1). Kidney cancer incidence has been increasing steadily for the past several decades, although the reasons for this are unclear (1). The von Hippel-Lindau (VHL) tumor suppressor gene was identified as a site of germline mutations in patients at risk for clear cell renal cell carcinoma (ccRCC), which accounts for approximately 85% of all kidney cancers (2). Inactivating VHL mutations also play major roles in sporadic kidney cell cancer (3). It is well established that the VHL-associated complex has E3 ubiquitin ligase activity and VHL loss leads to hypoxia-inducible factor α (HIFα, including HIF1α and HIF2α) and ZHX2 stabilization, which contributes substantially to the transforming phenotype of renal cancer (2, 4–7). Further research shows that VHL interacts with HIF1α through the hydroxylation of defined HIF1α proline residues (prolines 402 and 564) by members of the EGLN family of iron- and 2-oxoglutarate–dependent dioxygenases (EGLN1, EGLN2, and EGLN3; ref. 8). As a result of accumulation and translocation of HIFα factors into the nucleus, HIFαs dimerize with a constitutively expressed HIFβ subunit and transactivate genes that have hypoxia response elements (NCGTG) in promoters or enhancer regions, such as genes involved in angiogenesis (e.g., VEGF), glycolysis and glucose transport (e.g., GLUT1), and erythropoiesis (e.g., EPO; ref. 9). HIF signaling/activation is an important oncogenic signature for VHL-deficient ccRCC. However, it remains challenging to target HIF signaling in ccRCC. HIF2α stabilization, as a result of VHL loss, is sufficient and necessary for promoting kidney tumor growth (7). Recent reports showed that the specific HIF2α inhibitor PT2399 inhibits primary tumor growth and invasion of a subset of kidney cancer (10, 11). However, a significant portion of kidney cancers remain resistant to HIF2α inhibitor treatment (10, 11), highlighting the importance of identifying additional therapeutic vulnerabilities of VHL-deficient kidney cancer.

Tumor-specific genetic alterations (such as VHL loss) reveal not only the biological changes that drive tumor progression but also vulnerabilities that can be exploited therapeutically. Because 70% to 80% of kidney tumors harbor VHL functional loss, it remains very attractive to identify synthetic lethality partners for VHL loss in kidney cancer while sparing normal cells. Previous research has identified a handful of pharmacologic inhibitors, including the autophagy modulator STF-62247 (12), homoharringtonine (13), EZH inhibitors (14), GLUT1 inhibitors (15), and ROCK inhibitors (16), that displayed the selective killing of VHL-null ccRCC cells. In addition, CDK6, MET, and MAP2K1 were reported to be essential for ccRCC cell lines with VHL loss (17). Some of these pathways are known HIF signaling regulators whereas the mechanisms for other VHL synthetic lethality partners remain unknown.

Tank binding kinase 1 (TBK1) is a member of the atypical IκB kinase (IKK) family, which also features another highly related family member, IKKϵ. Upon DNA and RNA virus infection, stimulator of interferon genes (STING) binds with TBK1 and promotes its phosphorylation on Ser172 within the TBK1 activation loop, which is necessary for its kinase activity to induce STING phosphorylation on Ser366 and the type I IFN response by directing IRF3 phosphorylation (18, 19). As such, TBK1 is a required element of innate immune signaling in cells. In recent years, the role of TBK1 has been expanded into cancers (20, 21). Although previous research suggested that RALB–SEC5 effector or AXL signaling may act upstream of TBK1 signaling (22, 23), it is largely unclear regarding how TBK1 activity is dynamically regulated in cancers and whether this activation is connected to its canonical signaling in innate immunity. Here we identify a novel role of TBK1 signaling in cancer, distinct from its role in innate immune signaling, by serving as a synthetic lethal partner for VHL-null kidney cancer in a HIF-independent manner.

By using a pan-prolyl hydroxylation antibody to perform pulldown followed by mass spectrometry analysis in HeLa cell lysates, TBK1 was indicated to be hydroxylated (24). Because cells were not treated with MG132, many VHL degradation substrates may not be retrieved from the pulldown, including HIF1α and HIF2α (24). Among the list that was pulled down from the mass spectrometry, TBK1 is one of the handful of kinases that may be therapeutically targetable. Because hydroxylated protein may interact with and be potentially regulated by VHL, we set to determine whether TBK1 protein level or its canonical phosphorylation on Ser172, which governs its activity, may be regulated by VHL. To this end, we examined TBK1 or pTBK1 (Ser172) levels in ccRCC isogenic cell lines (786-O, UMRC2, RCC4, and UMRC6) that are either VHL-null [with empty vector (EV)] or with VHL restoration. Interestingly, whereas total TBK1 protein levels did not change upon VHL expression, phosphorylated TBK1 was significantly suppressed by VHL expression in all four cell lines examined (Fig. 1A and B). Similar regulation was also found in 293T cells overexpressing VHL (Supplementary Fig. S1A). To show whether the level of pTBK1 is a good predictor for TBK1 kinase activity, we performed an in vitro kinase assay using UMRC2 cell lysates as the source of the kinase and purified STING protein as a substrate. STING phosphorylation was detected by a STING pSer366 antibody. As a control, we depleted TBK1 in UMRC2 cells by specific single guide RNA (sgRNA) and found that TBK1 depletion led to decreased STING phosphorylation on Ser366 (Supplementary Fig. S1B), suggesting that this kinase assay can be used to reliably detect TBK1 activity. We found that lysates from UMRC2 EV cells displayed stronger STING phosphorylation than lysates from the isogenic cells restored with VHL (Fig. 1C). Conversely, we also depleted VHL by several independent sgRNAs with CRISPR/Cas9 in VHL-proficient cells (293T and Caki-1). Consistently, VHL depletion led to profound upregulation of pTBK1 while not affecting total TBK1 levels (Fig. 1D and E). It is worth mentioning that the protein levels and phosphorylation state of IKKϵ, a close family member of TBK1, was not affected by VHL status in these cells (Fig. 1E; Supplementary Fig. S1C), suggesting that VHL regulation of TBK1 phosphorylation is specific. IL1β and TNFα can activate TBK1 activity independent of adaptors such as STING (25). To exclude the potential effect of these cytokines on TBK1 signaling, we examined the conditional medium from EV and isogenic VHL cell culture. Although TNFα in most medium samples is lower than the detection limit, the IL1β level is comparable in EV and VHL medium (Supplementary Fig. S1D).

VHL loss creates a pseudohypoxia condition in ccRCC that is characterized by constitutive HIF stabilization (26). Hypoxia is also a characteristic of most solid cancers (27). We exposed VHL-proficient cells (293T, Caki-1, and HKC) and RCC4 cells expressing exogenous HA-VHL with hypoxia (1% O₂) and also found consistent pTBK1 upregulation (Fig. 1F; Supplementary Fig. S1E). As an orthogonal approach, we treated VHL-proficient Caki-1 cells with hypoxia mimetics [dimethyloxaloylglycine (DMOG), deferoxamine (DFO)] or VHL-restored RCC4, 786-O, or UMRC2 cells with FG4592 or DFO. In all cases, we found TBK1 phosphorylation was significantly upregulated (Supplementary Fig. S1F and S1G). To address the question of whether this upregulation was due to HIF upregulation, we depleted HIF2A expression by two independent sgRNAs and found that TBK1 phosphorylation level was not decreased by HIF2A depletion (Supplementary Fig. S1H). We also used two independent sgRNAs to deplete ARNT, encoding an essential binding partner of HIF2α for its transcriptional activity, in 786-O and UMRC2 cells. Consistently, we did not find that TBK1 phosphorylation was affected by ARNT depletion (Supplementary Fig. S1I). Altogether, our results suggest that VHL or hypoxia regulates TBK1 phosphorylation in a HIF-independent fashion.

To further determine whether VHL-mediated pTBK1 downregulation is mediated by prolyl hydroxylation, we treated HA-VHL–restored 786-O cells with prolyl hydroxylase inhibitors (DMOG, DFO) and found that these inhibitors upregulated pTBK1 level in VHL-restored cells to a similar level as in the vector cells (Supplementary Fig. S1J). To examine whether TBK1 and VHL directly interact, we performed GST–VHL pulldown with in vitro–translated FLAG–TBK1 and found a direct interaction between them (Fig. 1G). To determine whether the interaction of VHL and TBK1 mediates TBK1 phosphorylation regulation by VHL, we examined their binding in the absence or presence of prolyl hydroxylase inhibitors (DFO and DMOG) and found that prolyl hydroxylase inhibitors eliminated the binding between TBK1 and VHL, which corresponded with increased TBK1 phosphorylation (Fig. 1H). Consistently, DFO and DMOG treatment also led to increased TBK1 phosphorylation and activity in VHL-restored UMRC2 cells (Supplementary Fig. S1K). Therefore, our data suggest that prolyl hydroxylation may promote the interaction between TBK1 and VHL, therefore contributing to decreased TBK1 phosphorylation and activity.

To address the prolyl hydroxylase that may contribute to TBK1 hydroxylation and its regulation by VHL, we examined the interaction between TBK1 and three EGLN family members (EGLN1, 2, and 3) and found that EGLN1 is the primary prolyl hydroxylase that showed the most robust binding with TBK1 (Supplementary Fig. S2A). In addition, we showed that EGLN1 and TBK1 could bind endogenously in several cell lines including 293T, RCC4, and 786-O (Fig. 1I). Next, we depleted EGLN1 by two independent sgRNAs and found that EGLN1 depletion led to increased TBK1 phosphorylation (Fig. 1J). In addition, depletion of EGLN1 in HA-VHL–expressing RCC4 cells upregulated TBK1 phosphorylation to a similar level as in the RCC4 cells with vector control (Fig. 1K). As an orthogonal approach, we also showed that TBK1 phosphorylation was upregulated in EGLN1 knockout mouse embryonic fibroblasts (MEF), but not in EGLN2/3 knockout MEFs (Supplementary Fig. S2B). In addition, we also depleted EGLN1 with two independent shRNAs in 293T cells and consistently observed increased TBK1 phosphorylation (Supplementary Fig. S2C). In these cells with EglN1 depletion, DMOG treatment could not further upregulate TBK1 phosphorylation, suggesting that EGLN1 is the major hydroxylase enzyme that regulates TBK1 phosphorylation (Supplementary Fig. S2C). Conversely, we also overexpressed EGLN1 in Caki-1 cells and observed decreased TBK1 phosphorylation with wild-type (WT) but not a catalytically dead EGLN1 mutant (EGLN1-CD, H374A/D376A; Fig. 1L). This regulation is further confirmed by the upregulation of TBK1 phosphorylation in EGLN1-overexpressing cells treated with the prolyl hydroxylase inhibitors FG4592 or DFO (Supplementary Fig. S2D and S2E). To examine whether EGLN1 regulates the interaction between TBK1 and VHL, therefore contributing to TBK1 phosphorylation regulation, we first performed in vitro hydroxylation followed by VHL binding assay. In the presence of EglN1, the binding between TBK1 and VHL increased, suggesting that EGLN1-mediated hydroxylation on TBK1 promotes TBK1–VHL interaction (Supplementary Fig. S2F). We then depleted EGLN1 by two independent shRNAs and found that EGLN1 depletion led to decreased binding between TBK1 and VHL, which corresponded with increased TBK1 phosphorylation (Fig. 1M). Overall, our results suggest that EGLN1 hydroxylates TBK1 which leads to its binding with VHL and decreased TBK1 phosphorylation.

Next, we aimed to identify TBK1 prolyl hydroxylation sites that may be important for its regulation by EGLN1 and VHL. To this end, we overexpressed FLAG-tagged TBK1 in 293T cells followed by either DMSO or DMOG treatment and analyzed immunoprecipitated FLAG–TBK1 via mass spectrometry to identify potential hydroxylation sites. We were specifically interested in potential proline sites that were hydroxylated, but whose hydroxylation levels were diminished upon DMOG treatment. We identified two such prolyl hydroxylation sites: Proline 678 and 48 (Supplementary Fig. S3A and S3B). Next, we mutated TBK1 Proline 678 and 48 to alanines (P678A, P48A). Whereas TBK1 WT or P678A mutant binds VHL efficiently, P48A mutant displayed diminished binding with VHL confirmed by reciprocal immunoprecipitations (Fig. 2A), suggesting that Proline 48 is the major hydroxylation site that binds VHL. Subsequently, we synthesized TBK1 WT or P48-OH peptides and performed binding assay with VHL. Among these peptides, only Proline 48 hydroxylated peptide binds VHL, although considerably more weakly compared with the binding of hydroxylated HIF1α peptide with VHL (Fig. 2B). We also performed immunoprecipitation using a pan-hydroxyl proline antibody and found that WT but not catalytic dead EGLN1 or EGLN1 in combination with DMOG, promoted TBK1 hydroxylation (Fig. 2C). Consistent with the notion that Proline 48 is the major TBK1 hydroxylation site, pan-hydroxylation immunoprecipitation followed by FLAG immunoblot also confirmed that P48A mutation abrogated all hydroxylation signals present in WT TBK1 (Fig. 2D). To determine whether the Proline 48 site is the major site affecting TBK1 phosphorylation, we depleted endogenous TBK1 from UMRC2 cells and reintroduced either WT or P48A-mutant TBK1 into these cells (Fig. 2E). In WT TBK1–expressing cells, VHL overexpression led to decreased TBK1 phosphorylation that can be upregulated upon concurrent treatment with the prolyl hydroxylase inhibitor DMOG (Fig. 2F). On the other hand, cells expressing TBK1 P48A mutant displayed constitutive TBK1 phosphorylation that is resistant toward VHL expression (Fig. 2F). It is important to note that TBK1 P48A mutant also showed higher TBK1 phosphorylation at the basal level despite a lower total TBK1 level. Next, we also wanted to examine whether TBK1 hydroxylation by EGLN1 is mainly contributing to the phosphorylation regulation, and we found that TBK1 phosphorylation was upregulated by EGLN1 depletion by two independent EGLN1 siRNAs in WT TBK1–expressing cells (Fig. 2G). On the other hand, TBK1 P48A mutant displayed constitutive upregulation of phosphorylation that was resistant to EGLN1 depletion (Fig. 2G). According to crystal structure of TBK1 dimer, P48 residue sits on the periphery of TBK1 kinase domain that might be accessible by EGLN1 (Supplementary Fig. S3C). We then examined the binding between EGLN1 and TBK1 WT or P48A mutant and found that P48A could not bind with EGLN1, suggesting that the Proline 48 site is the major EGLN1 hydroxylation site (Fig. 2H). Finally, VHL-restored UMRC2 cells harboring TBK1 P48A mutant could form more colonies compared with cells harboring TBK1 WT in 3-D soft-agar growth assay (Fig. 2I). Although IKKϵ and TBK1 exhibit high similarity, the sequences surrounding Proline 48 are distinct in these two proteins (Supplementary Fig. S3D), suggesting that Proline 48–mediated regulation is specific for TBK1. Furthermore, TBK1 Proline 48 (LRP motif) is highly conserved in all vertebrates, suggesting its functional importance (Supplementary Fig. S3E). In summary, TBK1 is hydroxylated on Proline 48 residue that promotes VHL binding and its decreased phosphorylation.

The next question was how VHL binding to TBK1 may contribute to its decreased phosphorylation. VHL is often found as a component of an E3 ligase complex in which VHL binds its hydroxylated protein targets for ubiquitination and degradation. We examined the potential effect of VHL on TBK1 ubiquitination and did not find TBK1 ubiquitination induced by VHL (Supplementary Fig. S4A and S4B). PPM1B was reported to be a phosphatase for TBK1 (28). In addition, VHL was also shown to bind with PPM1B (29). We speculated that VHL might affect PPM1B binding with TBK1, therefore contributing to TBK1 dephosphorylation. To test this hypothesis, first we overexpressed VHL and found that VHL overexpression led to increased binding between TBK1 and PPM1B (Fig. 2J), which corresponded with decreased TBK1 phosphorylation. Next, we depleted VHL by two independent sgRNAs and found that VHL depletion led to decreased binding between TBK1 and PPM1B, which contributed to increased TBK1 phosphorylation (Fig. 2K). Cumulatively, our data suggest that EGLN1 can promote TBK1 hydroxylation, VHL binding, and recruitment of PPM1B, which contributes to the dephosphorylation of TBK1 and its decreased activity.

Because we showed that VHL loss led to TBK1 hyperactivation, we next aimed to see whether TBK1 depletion would cause synthetic lethality toward VHL-null ccRCC cells. To this end, we depleted TBK1 using several independent sgRNAs in several ccRCC cell lines that are either VHL-null (EV) or with VHL restoration. In UMRC6 cells, CRISPR/Cas9-mediated TBK1 depletion by sgRNAs led to efficient TBK1 depletion in the isogenic cell lines (Fig. 3A). It is important to note that VHL restoration in VHL-null ccRCC cells did not affect cell proliferation and colony formation in serum-rich medium, which is consistent with previously published literature (30). Remarkably, TBK1 depletion led to profound cell proliferation defects in VHL-null cells, whereas VHL-restored cells remained largely unaffected (Fig. 3B). We also confirmed this phenotype with 3-D soft-agar growth (Fig. 3C and D) and performed these experiments in UMRC2 and RCC4 cells and got similar results (Fig. 3E–H; Supplementary Fig. S5A and S5B). We then depleted TBK1 expression in VHL-intact cell lines such as 293T and HKC and found no robust growth defect could be observed (Supplementary Fig. S5C and S5D). To ensure the effect of TBK1 sgRNA was due to its on-target effect on TBK1, we also restored TBK1 expression with the TBK1 construct that carries the silent mutation on sgRNA recognition site and found that TBK1 restoration rescued cell-proliferation defect induced by TBK1 depletion (Fig. 3I and J), demonstrating the on-target effect of the TBK1 sgRNA.

To further determine whether the effect of TBK1 loss on these phenotypes was dependent on its enzymatic activity, we expressed TBK1 catalytically dead mutant K38A (31) and found that this mutant displayed cell-proliferation defects as compared with WT TBK1 (Supplementary Fig. S5E), suggesting that TBK1 enzymatic activity is important for cell proliferation. We also treated UMRC6 and UMRC2 EV or VHL-expressing cells with Compound 1 (CMPD1), a recently developed TBK1 inhibitor (32), and found that CMPD1 efficiently blocked soft-agar colony growth of EV cells but not VHL-restored cells (Fig. 3K-M; Supplementary Fig. S5F and S5G). In addition to CMPD1, we also utilized two previously published TBK1 inhibitors, BX795 and MRT67307 (33, 34). We also found that these TBK1 inhibitors preferentially inhibited ccRCC cell growth when VHL was lost and they only modestly affected these cells upon VHL restoration (Supplementary Fig. S5H). UMRC2 cells are resistant to PT2399, a recently reported HIF2α inhibitor (10). We wondered whether there is any cooperative effect between TBK1 inhibitor and PT2399. To answer this question, we treated UMRC2 cells with CMPD1 and PT2399. Compared with single-agent treatment, we did not find any cooperative effect between these two compounds (Supplementary Fig. S5I and S5J), which is consistent with the recent publication showing that CDK4/6 inhibitor did not enhance the HIF2α inhibitor efficacy in HIF2α inhibitor–resistant cell lines (35).

To eliminate the potential off-target effect associated with these inhibitors, we also generated a novel Cereblon-based TBK1 proteolysis targeting chimera (PROTAC) based on the previous literature (named UNC6587; Supplementary Fig. S6A) because previously reported TBK1 PROTACs facilitated TBK1 degradation via the recruitment of VHL E3 ligase (36, 37). Treatment with UNC6587 efficiently eliminated TBK1 protein levels in ccRCC cells, while not affecting its close family member IKKϵ (Supplementary Fig. S6B and S6C). More importantly, although TBK1 PROTAC efficiently degraded TBK1 in both VHL-null and VHL WT ccRCC cells in a similar fashion (Fig. 3N; Supplementary Fig. S6D), TBK1 degradation only caused cell growth defect on 3-D soft agar in VHL-null cells (Fig. 3O and P; Supplementary Fig. S6E and S6F), while leaving VHL WT cells unaffected. Therefore, our data suggest that TBK1 may be a therapeutic target in ccRCC with VHL loss.

In addition, we examined whether TBK1 was important for maintaining ccRCC tumor growth by introducing two inducible TBK1 shRNAs into UMRC2 cells. These hairpins efficiently depleted TBK1 levels upon doxycycline addition (Fig. 4A). TBK1 depletion led to decreased cell proliferation and reduced soft-agar growth upon doxycycline addition (Fig. 4B and C). Next, either control or TBK1 shRNA (#3185) cells were orthotopically injected into the renal capsules of NOD/SCID gamma (NSG) mice. Upon confirmation of tumor growth in vivo by consecutive weekly bioluminescence imaging, we fed mice doxycycline to induce TBK1 hairpin expression and imaged mice to monitor kidney tumor growth. Whereas cells expressing control hairpin grew readily 10 weeks after the addition of doxycycline, TBK1 hairpin–expressing cells failed to proliferate in vivo (Fig. 4D–F). Lung ex vivo imaging data showed that depletion of TBK1 also inhibited tumor cells' spontaneous lung metastasis (Fig. 4G and H). Our accumulative results suggest that TBK1 is important for ccRCC tumorigenesis both in vitro and in vivo.

To examine the physiologic relevance of TBK1 hyperactivation in ccRCC, we obtained 18 pairs of ccRCC tumors and adjacent normal tissues followed by staining with total TBK1 and pTBK1. VHL expression in these tumors was examined by both sequencing and IHC staining. Eight of 18 tumors were confirmed by sequencing that contain VHL mutations or splice variants (6). The other 10 tumors, although they did not show VHL mutation by sequencing, still displayed decreased VHL expression by IHC staining (Supplementary Fig. S7), which may be due to epigenetic or other mechanisms. Overall, a majority of ccRCC tumors (13 of 18 pairs) displayed higher TBK1 phosphorylation compared with normal tissues (Fig. 5A and B; Supplementary Table S1). To validate this finding in large clinical cohorts, we also obtained two sets of ccRCC tissue microarrays (TMA) and stained these TMAs with TBK1 and pTBK1 with IHC. By calculating the ratio between pTBK1 and TBK1, there was significant upregulation of pTBK1 in ccRCC tumors compared with normal tissue in both TMA sets, which was reflected by both representative IHC images and quantitative analyses (Fig. 5C–F; Supplementary Table S1). Annotation information suggested that all tumor samples, except one from TMA2, are ccRCC, which may contain up to 85% of VHL mutations (2). To further confirm VHL loss status, we stained TMA2 with VHL antibody and found 37 of 45 pairs show decreased VHL level in tumor than pairednormal tissue, 3 pairs show a comparable level between normal and tumor tissue, and the other 5 tumors have higher VHL than normal tissue. When all samples were divided on the basis of VHL status, we found that tumors expressing the highest pTBK1 levels are all from the VHL(N>T) group (Fig. 5G). On the basis of our data, a considerable number of tumors with VHL loss results in no effect on TBK1 phosphorylation. This may result from the complexity of kidney TMAs, including the possessing method, time of procurement, and the age of TMAs that might lead to loss of TBK1 phosphorylation in some tumors with VHL loss.

Next, to relate kidney cancer–associated TBK1 activity with its known innate immune function, we found that we could not detect canonical IRF3 phosphorylation despite detectable IRF3 protein levels (Supplementary Fig. S8A), indicating that there exists a novel mechanism by which TBK1 activation contributes to kidney tumorigenesis. In addition, depletion of STING, an important upstream regulator of TBK1 in innate immunity upon double-strand DNA infection, did not affect TBK1 protein level or its phosphorylation in multiple ccRCC cell lines (Supplementary Fig. S8B), further strengthening that TBK1 may exert its effect in ccRCC in an innate immunity-independent manner. Because AKT is another TBK1 substrate which has an oncogenic function (38, 39), we also examined AKT activity by detecting its phosphorylation on Ser473 and Thr308. Our result suggested phosphorylation of Akt Ser473 and Thr308 was not decreased upon TBK1 depletion (Supplementary Fig. S8C). To this end, we performed an unbiased phospho-proteomic screen using UMRC2 cells treated withthe specific TBK1 inhibitor CMPD1 for 0 minutes, 15 minutes, 1 hour, and 3 hours (Fig. 6A). We chose to use the TBK1 inhibitor for the short duration of treatment in the hope of identifying its direct substrates. We identified 3,320 localized phospho-peptides that are matched to 1,243 proteins (Supplementary Fig. S9A). For each treatment we performed triplicate analysis, and we achieved high correlation between replicates. Among differentially expressed phospho-peptides, we are especially interested in the phospho-peptides that were decreased at least at two consecutive time points (either 15 minutes/1 hour, 1 hour/3 hours or 15 minutes/1 hour/3 hours). By ANOVA comparison, we obtained a subset of phospho-peptides that showed a significant decrease, defined as P < 0.05 with log₂ fold change ≤ −0.5 (Fig. 6B). Among this list, we noticed that p62 has been reported to be an oncogene on chromosome 5q that is significantly amplified in kidney cancer (40). By comparing our list with the previous research on the potential p62 phosphorylation sites retrieved from mass spectrometry (41), we found that p62/SQSTM1 Ser366 was the only common site retrieved from both studies. It is interesting to point out that another phosphorylation site on p62 (Ser403) indicated previously in innate immune signaling was not recovered in our list (42), suggesting the context-dependent p62 phosphorylation in innate immunity versus certain cancer settings. Therefore, we decided to pursue the potential role of p62 as a novel TBK1 substrate and its phosphorylation (Ser366) in kidney cancer.

Next, we performed an in vitro kinase assay and found that recombinant TBK1, but not IKKϵ, phosphorylates p62 (Fig. 6C). In addition, CMPD1 inhibits TBK1-induced p62 phosphorylation in a dose-dependent manner (Fig. 6C). We then performed the nonradioactive in vitro kinase assay with p62 followed by mass spectrometry and retrieved that p62 Ser366 as the major site being phosphorylated (Supplementary Fig. S9B and S9C). Furthermore, we also confirmed that TBK1 can promote the phosphorylation of p62 on the Ser366 site following transfection (Fig. 6D), which can be inhibited by CMPD1 treatment in kidney cancer cells including 786-O and UMRC2 (Supplementary Fig. S9D). Interestingly, by transfecting cells with TBK1, we found that whereas WT p62 level was significantly elevated, p62 S366A mutant displayed diminished upregulation (Fig. 6D). p62 upregulation induced by TBK1 is dependent on TBK1 enzymatic activity because the TBK1 catalytic dead mutant K38A failed to induce p62 upregulation (Supplementary Fig. S9E). Conversely, TBK1 depletion in several ccRCC cell lines (786-O, UMRC2, and UMRC6) all led to decreased p62 protein levels (Fig. 6E). To exclude p62 protein level change as a secondary effect from TBK1 loss–induced cell death, we treated inducible TBK1 shRNA with doxycycline in a time-course manner (1–3 days). We found that cleaved caspase-3, the marker for cell apoptosis, did not appear apparently until the third day of doxycycline induction, which happened after the decreased TBK1 and p62 levels, suggesting that the regulation of TBK1 on p62 is direct rather than indirect resulting from cell death (Fig. 6F). Our results suggest that TBK1 may directly phosphorylate p62 on the Ser366 site and this is important for the stabilization of p62 protein. In accordance with this hypothesis, VHL-null ccRCC restored with exogenous VHL displayed diminished TBK1 phosphorylation and decreased p62 protein levels (Fig. 6G). To determine whether TBK1 depletion–induced p62 downregulation was due to proteasomal degradation or lysosomal degradation, we treated these cells with either the proteasomal inhibitor MG132 or lysosomal inhibitors including BA1 and NH₄Cl and found that the lysosomal inhibitors, but not the proteasomal inhibitors, significantly elevated p62 expression in TBK1-depleted cells or VHL-overexpressing cells (Fig. 6H and I; Supplementary Fig. S9F and S9G). Next, to examine whether p62 Ser366 phosphorylation is important for mediating TBK1 knockdown–induced phenotype, we overexpressed p62 S366A or S366D in TBK1 sgRNA–infected cells and found that p62 S366D could efficiently rescue TBK1 depletion–induced cell proliferation defect, but S366A could not (Fig. 6J–L). Therefore, TBK1-induced p62 Ser366 phosphorylation is critical to promote ccRCC tumor cell growth.

We also performed p62 IHC staining on an 18-pair tumor cohort as well as two TMAs. In both TMAs, p62 protein level increased significantly in tumors than in paired normal tissues (Supplementary Fig. S10A and S10B; Supplementary Table S1). We then performed correlation analysis but failed to see significant correlation between pTBK1/TBK1 ratio and p62 protein level. This may be explained by multiple mechanisms that contribute to p62 protein stability (43, 44). Especially, it was reported that p62 degradation could be promoted by hypoxia-induced autophagy (45), which may widely exist in cancer cells. Kidney cancer is characterized with chromosome 5q gain as previously published (46), which leads to increased levels of SQSTM1/p62 (40). Therefore, it is likely that tumors with 5q gain may not require increased pTBK1 because p62 is hyperactivated in this setting. This may partially explain the lack of correlation between pTBK1 level and p62 protein level as well. However, it does not rule out the possibility that TBK1 phosphorylation may also contribute to other pathway activation besides p62 that will coordinate with 5q gain to increase kidney tumorigenesis, which will be investigated in future research.

Taken together, our results provide novel mechanistic insight by which TBK1 hyperactivation promotes p62 phosphorylation on Ser366, which leads to upregulation of total p62 and phospho-p62, therefore contributing to ccRCC tumorigenesis.

In this study, we have uncovered a novel mechanism by which VHL loss or hypoxia promotes TBK1 phosphorylation and activity, therefore promoting its downstream p62 phosphorylation, protein stability, and ccRCC tumorigenesis. Under VHL restoration or normoxic condition, TBK1 undergoes EGLN1-mediated hydroxylation on Proline 48 residue, which will induce the VHL binding and its recruitment of PPM1B phosphatase to be associated with VHL and TBK1. As a result, PPM1B will dephosphorylate TBK1 that leads to decreased TBK1 activity. Under hypoxic condition, EGLN1 enzymatic activity is inhibited and cannot hydroxylate TBK1 on Proline 48. As a result, similar to VHL loss, TBK1 cannot associate with VHL and PPM1B, which will lead to constitutive TBK1 phosphorylation. As a result of TBK1 hyperactivation, TBK1 will phosphorylate p62 on Ser366 residue and promote renal tumorigenesis (Fig. 6M). Our study characterizes the autonomous function of TBK1 in kidney cancer cells, which is distinctive from its role in innate immune signaling.

VHL loss is a hallmark of ccRCC, which accounts for the majority of kidney cancers. HIF2α is a well-known oncogene in this setting that has led to clinical trials with the new HIF2α antagonist PT2399. However, only a subset of ccRCC cell lines will display growth inhibition with this HIF2α inhibitor (10, 11). In addition, only a subset of patients with ccRCC will respond to HIF2α inhibitors and some of them develop resistance to this inhibitor, which could be due to lack of targeting strategies for recently discovered VHL novel substrates including ZHX2 (6). Our study not only adds the TBK1 kinase as a novel synthetic lethality partner for VHL-null ccRCC, but also puts TBK1 as a potential therapeutic target for tumor cells that are hypoxic. However, although hypoxia is a characteristic of most solid cancers, it is not necessarily a feature universal to all cancer cells because tumor-wide pseudohypoxia is observed in a limited number of cancers, including VHL-deficient ccRCC. The role of TBK1 phosphorylation in other types of cancer cells under hypoxia awaits further detailed investigation.

p62 is an important mediator for tumor cell autophagy because it will facilitate the degradation of specific proteins by autophagy (47). Previous research also showed that VHL-null ccRCC displayed high basal levels of autophagy (48). It is interesting that the inhibitors that lead to increased autophagy caused lethality in VHL-null ccRCC cells (12). One important consideration is that p62 post-translational modification can modulate the phenotype in VHL-null ccRCC cells. Our study first demonstrates that p62 Ser366 phosphorylation, mediated by TBK1 kinase, plays an important role in maintaining ccRCC oncogenic phenotype. Our ongoing investigation focuses on whether p62 Ser366 phosphorylation will modulate the autophagy signaling in ccRCC.

Over the decade, research and interest in TBK1 have expanded along with the identification and development of small molecules targeting TBK1. Now, at least seven distinct small molecules are known to inhibit TBK1 activity including BX795, compound II, CYT387 (momelotinib), MRT67307, GSK2292978A, amlexanox, and CMPD1 (49). However, the problem of selectivity restricts application of these compounds. For example, BX-795 has been reported to inhibit other kinases including PDK1 (33, 50). MRT67307 may also inhibit ULK1/2 and block autophagy (51). Only momelotinib and amlexanox have been tested in clinical trials, but not for treating kidney cancer (49). The recently developed inhibitor CMPD1, which showed the better selectivity for TBK1 over IKKϵ (32), was used in our study and showed the selective killing of VHL-null ccRCC cells. We have shown in this study that a Cereblon-based TBK1 PROTAC may be efficacious in inhibiting VHL-null ccRCC soft-agar colony growth but will not affect VHL WT cells. Our ongoing research is to optimize TBK1 PROTACs to achieve higher efficacy and test their effects in tumor xenografts. Recent reports have shown that depletion of TBK1 can boost the efficacy of immune checkpoint inhibitors in cancers (32, 52). Therefore, it is likely that by targeting TBK1 kinase, we can not only inhibit tumor cell–autonomous signaling, but also boost immunotherapy efficacy in kidney cancer.

RCC4, 786-O, UMRC2, UMRC6, Caki-1, HKC, and 293T cells were cultured in DMEM containing 10% FBS plus 1% penicillin/streptomycin. The 786-O, Caki-1, and 293T cells were obtained from ATCC. HKC cells were obtained from Dr. W. Kimryn Rathmell (Vanderbilt University Medical Center; ref. 53). RCC4, UMCR2, and UMRC6 were obtained from Sigma as described previously (6). Cells were used for experiments within 10 to 20 passages from thawing. All cells were authenticated via short tandem repeat testing. All cells were tested to ensure they were Mycoplasma free.

Rabbit anti-TBK1 (3504), rabbit anti-TBK1 phospho-Ser172 (pTBK1, 5483), rabbit anti-IKKϵ (2905), rabbit anti-IKKϵ phospho-Ser172 (pIKKϵ, 8766), rabbit anti-STING (13647), rabbit anti-STING phosphor-Ser366 (19781), rabbit anti-VHL (68547), rabbit anti-EGLN1 (3293), rabbit anti-HIF1α (3716), rabbit anti-p62 (39749), rabbit anti-GST (2625), rabbit anti-HA tag (3724), rabbit anti-FLAG tag (14793), rabbit anti–cleaved caspase-3 (9664), mouse anti-α-Tubulin (3873) were from Cell Signaling Technology. Mouse anti-HIF2α (ab157249), mouse anti-TBK1 (ab12116), mouse anti-EGLN1 (ab103432), rabbit anti-PPM1B (ab70804) were from Abcam. Mouse anti-Ub (8017) was from Santa Cruz Biotechnology. Rabbit anti-p62 phosphor-Ser366 (AF7374) was from Affinity BioSciences. Peroxidase-conjugated goat anti-mouse secondary antibody (31430) and peroxidase-conjugated goat anti-rabbit secondary antibody (31460) were from Thermo Scientific. DMOG (D1070, 1g) was from Frontier Scientific; DFO (D9533, 1G), BX-795 (204001, 10 mg), MRT67307 (506306, 5 mg), and Bafilomycin-A1 (BA1, B1793) were from Millipore-Sigma; FG4592 (15294, 25 mg) was from Cayman Chemical. CMPD1 was synthesized by WuxiAPP Tech following the procedure described previously (32).

To prepare samples for mass spectrometry analysis, 293T cells were transfected with FLAG-TBK1 followed by DMSO or 1 mmol/L DMOG treatment. Cells were lysed in EBC buffer supplemented with protease inhibitor and phosphatase inhibitor (Roche Applied Bioscience). Then, cell lysates were incubated with FLAG M2 beads (Sigma) at 4°C overnight to capture FLAG-TBK1 protein. The next day, beads were washed three times by EBC buffer and the bond proteins were resolved by SDS-PAGE. The gel was stained by Coomassie Brilliant Blue and FLAG-TBK1 bands were excised for mass spectrometry analysis to identify potential hydroxylation site.

TBK1 peptide containing WT Pro48 (VFNNISFLRPVDVQMREFE) or hydroxylated Pro48 (VFNNISFLR(P-OH)VDVQMREFE) were synthesized by LifeTein. All peptides were labeled with N-terminal biotin group. To perform the pulldown assay, 10 μg WT or hydroxylated peptides were incubated with 10 μL NeutrAvidin beads (Thermo Fisher Scientific 29200) for 4 hours at 4°C in NETN buffer. After incubation, Avidin beads were washed by NETN buffer to remove free peptides and then incubated with VHL protein generated by in vitro translation system (Promega L1170) in NETN buffer. Then input and bond VHL was examined by Western blot analysis.

Purified GST-TBK1 protein (SignalChem T02-10G) was mixed with 10 μL EGLN1 (in vitro translated) and 75 μL hydroxylation buffer (0.5 mol/L HEPES pH7.4 containing 10 mmol/L FeSO₄, 1 mol/L ascorbic acid, 0.1 mol/L α-Ketoglutarate, and 1,500 U/mL Catalase C-100) to make a 100 μL reaction system. The reaction was carried out at 37°C for 1 hour. After reaction, 500 μL NETN buffer and 10 μL GST beads (GE Healthcare) were added in and incubated at 4°C for 4 hours to pull down GST-TBK1. Then, GST beads were washed by NETN buffer to remove unbound proteins and incubated with VHL protein (in vitro translated) overnight. The next day, after washing three times, all input and pulldown samples were examined by Western blot analysis.

For kinase assay using p62 as substrate, 6 μL recombinant human p62 protein (Enzo Life Sciences) was mixed with 22 μL kinase reaction buffer (20 mmol/L Tris-HCl pH 7.4, 500 mmol/L β-glycerol phosphate, 12 mmol/L magnesium acetate), 1 μL 10 mmol/L ATP, and 1 μL GST-TBK1 (SignalChem) to reach a 30 μL reaction system. Kinase reaction was carried out at 30°C, 500 rpm for 1 hour. After reaction, p62 proteins were resolved by SDS-PAGE and stained by Coomassie Brilliant Blue. The bands were cut out for MS analysis to identify phosphorylation site.

For kinase assay using STING as substrate, 5 μL recombinant human STING protein (Active Motif, 81182) was mixed with 33 μL kinase reaction buffer (the same as above), 2 μL 10 mmol/L ATP, and 10 μL UMRC2 cell lysis to reach a 50 μL reaction system. Kinase reaction was carried out at 30°C, 500 rpm for 1 hour. After reaction, total STING and phospho-STING (Ser366) were detected by Western blot analysis.

For phosphoproteomics, UMRC2 cells were treated with DMSO or CMPD1 for 15 minutes, 1 hour, or 3 hours (n = 3 biological replicates per time point). Cells were washed three times with ice-cold PBS, then lysed in 8 mol/L urea (Sigma, U4883), Tris-HCl (pH 7.6) with protease and phosphatase inhibitors (Roche Applied Bioscience). Lysates were reduced with 5 mmol/L DTT, alkylated with 15 mmol/L iodoacetamide, then subjected to digestion with trypsin (Promega) at a 1:50 enzyme:protein ratio. The resulting peptide samples were acidified and desalted using SepPak C18 SPE cartridges (Waters, 100 mg sorbent). Eluates were dried via vacuum centrifugation. Peptide concentration was determined using Pierce Quantitative Colorimetric Peptide Assay. A pooled sample was generated by combining an aliquot (85 μg) of each of the 12 samples. Five hundred micrograms of each sample (and 1 mg of pooled sample) was reconstituted with 200 mmol/L TEAB, then individually labeled with a TMT 10-plex reagent (Thermo Fisher Scientific) for 5 hours at room temperature. Labeling efficiency and sample ratios were evaluated by LC/MS-MS analysis of two test mixes. Samples were quenched with 50% hydroxylamine to a final concentration of 0.4%. Labeled peptide samples were mixed 1:1 to create two TMT 7-plex sets, dried via vacuum centrifugation, then desalted using SepPak C18 SPE cartridges (Waters, 500 mg sorbent). One hundred micrograms of each TMT 7-plex set was fractionated into six “global proteome” fractions using high pH reverse-phase spin columns (Pierce). The remaining mixed samples (3 mg of each TMT 7-plex) were enriched for phosphopeptides using MagReSyn Ti-MAC beads (ReSyn Biosciences), as described previously (54). The phosphopeptide-enriched samples were fractionated into three “phosphoproteome” fractions using high pH reverse-phase spin columns (Pierce). The proteome and phosphoproteome fractions were dried via vacuum centrifugation and stored at −80°C until further analysis.

The proteome and phosphoproteome fractions were analyzed by LC/MS-MS using an Easy nLC 1200 coupled to a QExactive HF mass spectrometer (Thermo Fisher Scientific). Samples were injected onto an Easy Spray PepMap C18 column (75 μm id × 25 cm, 2 μm particle size; Thermo Fisher Scientific) and separated over a 90-minute method. For the in vitro kinase reaction samples, the samples were separated over a 45-minute method. The gradient for separation consisted of 5% to 50% mobile phase B at a 250 nL/minute flow rate, where mobile phase A was 0.1% formic acid in water and mobile phase B consisted of 0.1% formic acid in 80% ACN. The QExactive HF was operated in data-dependent mode where the 15 most intense precursors were selected for subsequent HCD fragmentation. Resolution for the precursor scan (m/z 350–1,600) was set to 60,000 with a target value of 3 × 10⁶ ions and a maximum injection time of 100 ms. MS-MS scan resolution was set to 60,000 with a target value of 1 × 10⁵ ions and a maximum injection time of 100 ms. Fixed first mass was set to 110 m/z and the normalized collision energy was set to 32% for HCD. Dynamic exclusion was set to 30 seconds, peptide match was set to preferred, and precursors with unknown charge or a charge state of 1 and ≥ 8 were excluded.

For the proteome and phosphoproteome data, raw data files were processed using MaxQuant version 1.6.1.0., set to “reporter ion MS2” with “10plex TMT.” The isolation purity was set to >0.7. Peak lists were searched against a reviewed Uniprot human database (downloaded Feb 2018 containing 20,245 sequences), appended with a common contaminants database, using Andromeda within MaxQuant. All fractions were searched with up to two missed trypsin cleavage sites, fixed Cys carbamidomethylation modification, dynamic Met oxidation, and N-terminal acetylation modifications. Peptide false discovery rate was set to 1%. Data were further analyzed and visualized in Perseus, Microsoft Excel, and R. Only phosphopeptides identified in both TMT sets with a localization probability score > 0.7 were reported. To account for protein levels, log₂ intensities were median normalized on both the proteome and phosphopeptide level. The phosphopeptide log₂ intensities were normalized to the protein by calculating a ratio using the median normalized protein log₂ intensities. ANOVA multiple-sample test was performed in Perseus using a P value cutoff of 0.05.

For the p62 in vitro kinase assay samples, raw data were processed using Proteome Discoverer 2.1 (Thermo Fisher Scientific). Peak lists were searched against a reviewed Uniprot human database, appended with a common contaminants database, using Sequest. The following parameters were used to identify tryptic peptides for protein identification: 10 ppm precursor ion mass tolerance; 0.02 kDa product ion mass tolerance; up to two missed trypsin cleavage sites. Carbamidomethylation of Cys was set as a fixed modification and oxidation of Met and phosphorylation of Ser, Thr, and Tyr were set as variable modifications. The phosphoRS node was used to localize the sites of phosphorylation. Peptide false discovery rates (FDR) were calculated by the Percolator node using a decoy database search and data were filtered using a 1% FDR cutoff.

Six-week-old NSG mice (Jackson Laboratories) were used for xenograft studies. Approximately 5 × 10⁵ viable UMRC2 kidney cancer cells were resuspended in 20 μL fresh DMEM and injected orthotopically into the left kidney of each mouse as described previously (40). Bioluminescence imaging was performed as described previously (55, 56). For inducible TetOn-TBK1 shRNA, after injection and following consecutive weeks of bioluminescence imaging to make sure tumor was successfully implanted in kidney, mice were fed with purina rodent chow #5001 with doxycycline (Research Diets, Inc.). After mice were sacrificed, lung ex vivo imaging was performed immediately to examine tumor metastasis. The rough mass of tumors was presented as mean ± SEM and evaluated statistically using t test. All animal experiments were in compliance with NIH guidelines and were approved by the University of North Carolina at Chapel Hill Animal Care and Use Committee.

Unless indicated, the unpaired two-tailed Student t test was used for experiments comparing two sets of data. All graphs depict mean ± SEM unless otherwise indicated. Graphs were generated by GraphPad Prism. *, **, and *** denote P value of <0.05, 0.01, and 0.001, respectively; n.s., no significance.

No potential conflicts of interest were disclosed.

Conception and design: L. Hu, L.E. Herring, Q. Zhang

Development of methodology: L. Hu, Q. Zhang

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): L. Hu, F. Potjewyd, E.M. Wilkerson,L.E. Herring, L. Xie, X. Chen, J.C. Cabrera, K. Hong, A.S. Baldwin, K. Gong, Q. Zhang

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): L. Hu, H. Xie, F. Potjewyd, L.I. James, E.M. Wilkerson, L.E. Herring, J.C. Cabrera, X. Tan, K. Gong, Q. Zhang

Writing, review, and/or revision of the manuscript: L. Hu, H. Xie, L.I. James, C. Liao, A.S. Baldwin, Q. Zhang

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K. Hong, K. Gong, Q. Zhang

Study supervision: K. Gong, Q. Zhang

Other (hydroxylation assay): X. Liu

We thank the UNC LCCC Tissue Procurement Facility, UNC Animal Studies Core, and UNC Translational Pathology Laboratory for excellent help. We thank Prof. Dr. Michael Kracht, Dr. Samuel Bakhoum, and Dr. Xuewu Zhang for providing FLAG-TBK1, shSTING plasmids, and TBK1 structure information, respectively. We would like to thank the Kidney Cancer Research Program (KCRP) at UT Southwestern for providing us materials with testing. This work was supported in part by a Department of Defense Kidney Cancer Research Program Idea Development Award (W81XWH1910813, to Q. Zhang), Kidney Cancer Research Alliance (KCCure), Cancer Prevention and Research Institute of Texas (CPRIT; RP190058, to Q. Zhang), and the National Cancer Institute (R01CA211732 and R21CA223675, Q. Zhang; R35CA197684, to A.S. Baldwin). Q. Zhang is an American Cancer Society Research Scholar, V Scholar, Kimmel Scholar, Susan G. Komen Career Catalyst awardee, and Mary Kay Foundation awardee. This research is based in part upon work conducted using the UNC Proteomics Core Facility, which is supported in part by the P30 CA016086 Cancer Center Core Support Grant to the UNC Lineberger Comprehensive Cancer Center.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.