VHL loss enhances antitumor immunity by activating the anti-viral DNA-sensing pathway

Summary

von Hippel-Lindau (VHL), known as a tumor suppressor gene, is frequently mutated in clear cell renal cell carcinoma (ccRCC). However, VHL mutation is not sufficient to promote tumor formation. In most cases other than ccRCC, VHL loss alters cellular homeostasis and causes cell stress and metabolic changes by stabilizing hypoxia-inducible factor (HIF) levels, resulting in a fitness disadvantage. In addition, the function of VHL in regulating immune response is still not well established. In this study, we demonstrate that VHL loss enhances the efficacy of anti-programmed death 1 (PD1) treatment in multiple murine tumor models in a T cell-dependent manner. Mechanistically, we discovered that upregulation of HIF1α/2α induced by VHL loss decreased mitochondrial outer membrane potential and caused the cytoplasmic leakage of mitochondrial DNA, which triggered cyclic GMP-AMP synthase-stimulator of interferon genes (cGAS-STING) activation and induced type I interferons. Our study thus provided mechanistic insights into the role of VHL gene loss in boosting antitumor immunity.

Graphical abstract

Highlights

• •VHL loss induced the cytoplasmic mitochondrial DNA leakage via HIF1/2-BNIP3 axis
• •VHL loss caused cGAS-STING activation and type I interferon production
• •cGAS-STING activation in VHL-deficient tumors facilitated ICB therapy

Molecular biology; Immunity; Cell biology; Cancer

Introduction

VHL (von Hippel-Lindau) is an E3 ligase responsible for regulating the cellular levels of hypoxia-inducible factors (HIFs),^1,2,3 which are critical transcriptional factors regulating the cellular response to low oxygen tension.^4,5 VHL is a recognized tumor suppressor, particularly in sporadic clear cell renal cell carcinomas (ccRCCs), where ∼91% of VHL mutations are found. Loss-of-function mutations can cause high HIF activities and increased vascular endothelial growth factor (VEGF) levels, which can promote renal cell carcinoma (RCC) tumor formation.^6,7,8,9 However, VHL inactivation alone is insufficient for RCC development.^10,11,12,13 In fact, VHL loss can lead to metabolic changes and create a fitness disadvantage in many cells.^14,15,16 For example, HIF1α impairs cellular respiration and mitochondrial biogenesis in VHL-deficient cells.¹⁴ Loss of VHL also increases the expression of cyclin kinase inhibitors p21 and p27 that cause cell arrest and/or senescence.^17,18

VHL plays a crucial role in regulating the immune system through its regulation of the stable-state level and activities of HIFs.^19,20,21 For instance, loss of VHL stabilizes and increases HIF activity, resulting in enhanced CD8⁺ T lymphocyte cytotoxicity.²⁰ In addition, VHL loss-mediated HIF activation leads to changes in gene expression that can significantly impact the tumor immune microenvironment (TIME) with regard to both the composition and the function of the immune cells.²² HIF activation can be a double-edged sword that may reduce T cell differentiation while also enhancing T cell proliferation and activation.^20,23 Moreover, the activity of HIFs affects cytokine production, which can reprogram immune responses. For example, HIF activation in T cells is associated with the upregulation of certain pro-inflammatory cytokines and cytolytic molecules, such as interferon-gamma (IFNγ), granzymes, and tumor necrosis factors, thereby stimulating cytotoxic T cell responses and impacting the effectiveness of immune checkpoint blockade (ICB) therapy.^20,24,25

ICB therapy has emerged as a potent tool in the fight against various types of cancer in recent years, including melanoma and ccRCC.²⁶ The existing paradigm for determining the efficacy of cancer immunotherapy in a given malignancy centers on multiple intra- and intercellular factors, such as gene mutations, antigen presentation, and the diversity of major histocompatibility complex (MHC) as well as T cell receptor (TCR) repertoires.²⁷ Tumor mutation burden (TMB) is a crucial parameter for predicting tumor responses to immunotherapy.^28,29,30 The main reasoning is that a higher TMB leads to higher numbers of neoantigens that are likely to be recognized by the host’s T cells, which is critical for the success of ICB therapy.³¹ However, it cannot explain the responsiveness of ccRCC to ICB therapy, where the TMB of ccRCC is only very moderate.²⁸ A previous study has implicated that endogenous retrovirus activation may be involved in ccRCC response to ICB.³² In view of the prevalence of VHL mutations in ccRCC, we focus on VHL gene to examine in detail its potential roles in modulating tumor responses to ICB therapy.

Results

Vhl deficiency increased susceptibility to ICB therapy in murine tumor models

To evaluate the relevance of Vhl gene loss in ICB therapy, we conducted anti-programmed death 1 (αPD1) therapy in several Vhl-deficient tumor lines following the schedule depicted in Figure S1D. In the murine RCC Renca tumor model, VHL loss alone suppressed tumor growth (Figures S1A and 1A). This result is consistent with a published study demonstrating that VHL deletion in Renca cells restrains tumor growth in vivo using mouse models.³³ Moreover, αPD1 treatment had minimal effects in control tumors but synergized with Vhl gene knockout (KO), significantly suppressing tumor growth (Figure 1A) and extending the survival of tumor-bearing mice (Figure 1B). Indeed, 2/5 mice in the αPD1-treated Vhl-KO group remained long-term survivors, indicating significant synergy between Vhl gene loss and ICB therapy. In the B16F10 melanoma line, Vhl gene loss (Figure S1B) also significantly attenuated tumor growth in immunocompetent C56BL/6J mice (Figure 1C). Moreover, αPD1 treatment significantly suppressed the growth of Vhl-KO tumors (Figures 1C and 1D), with 1/6 mice remaining tumor-free in the entire period of observation (Figure 1D). We further examined the efficacy of ICB therapy in Vhl-KO MC38 tumors (Figure S1C). Our data indicated that αPD1 treatment caused additional growth delay in Vhl-KO MC38 tumors, which were already substantially slower than control tumors (Figures 1E and 1F). Notably, 3/5 mice in the αPD1-treated Vhl-KO group were long-term survivors (Figure 1F). To rule out the possibility that the off-target effects of CRISPR-Cas9 were responsible for the behavior of the Vhl-KO tumors, we restored Vhl expression by ectopic gene transduction (Figure S1E). Our data showed that ectopic expression of wild-type (WT) mouse Vhl (mVhl) in Vhl-KO MC38 cells abrogated the delay in tumor formation from the latter (Figures S1F and S1G), thereby supporting an essential role for VHL in this process. Our results, therefore, suggest that VHL loss significantly attenuated tumor growth and made them more susceptible to ICB therapy in multiple murine tumor models.

Figure 1.

VHL loss enhances murine tumor responses to ICB therapy

(A and B) Tumor growth (A) and Kaplan-Meier analysis (B) of BALB/c mice subcutaneously inoculated with 1 × 10⁶ VC or Vhl-KO Renca cells (n = 5) and treated with isotype control or an αPD-1 antibody. About 100 μg isotype control or αPD-1 antibodies were injected intraperitoneally (i.p.) on days 5, 8, and 11 post tumor cell inoculation.

(C and D) Tumor growth (C) and Kaplan-Meier survival analysis (D) of C57BL/6J mice subcutaneously inoculated with 1 × 10⁵ VC or Vhl-KO B16F10 cells (n = 5) and treated with isotype control or an αPD-1 antibody. About 100 μg isotype control or αPD-1 antibodies were injected i.p. on days 7, 10, 13, and 16 post tumor cell inoculation.

(E and F) Tumor growth (E) and Kaplan-Meier analysis (F) of C57BL/6J mice subcutaneously inoculated with 5 × 10⁵ VC or Vhl-KO MC38 cells (n = 5) and treated with isotype control or αPD-1 antibodies. About 100 μg isotype control or αPD-1 antibodies were injected i.p. on days 6, 9, and 12 post tumor cell inoculation. Error bars represents mean ± SEM. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; ∗∗∗∗p < 0.0001; n.s. not significant.

VHL deficiency stimulates intratumoral lymphocyte infiltration

The substantial tumor growth delay caused by VHL gene loss and its synergy with αPD1 treatment led us to posit that VHL played a critical role in regulating the TIME. We used flow cytometry to analyze intratumoral infiltration of lymphocytes in control and Vhl-KO MC38 tumors to examine our hypothesis (Figure S2A). Our results show significantly more CD3⁺ T cells in VHL-KO than in control MC38 tumors (Figure 2A). The increase was observed in both CD4⁺ T and CD8⁺ T subsets (Figures 2B and 2C). Notably, both the numbers of GZMB⁺ CD8⁺ T cells and IFNγ⁺ CD8⁺ T cells were significantly increased (Figures 2D and 2E), indicating heightened activation of cytotoxic T lymphocytes in Vhl-KO tumors. Our analysis also showed high natural killer (NK⁺) cells in Vhl-KO tumors that did not reach statistical significance (Figure 2F). In contrast, the numbers of F4-80⁺ macrophages (Mφ), γσTCR⁺ T cells, and Foxp3⁺ Tregs were comparable between control and Vhl-KO tumors (Figures 2G–2I).

Figure 2.

VHL loss increases intratumoral lymphocyte infiltration

(A–I) Flow cytometry profiling of intratumoral lymphocytes in control and VHL-KO MC38 tumors grown in syngeneic C57BL/6 mice. Mice bearing VC or Vhl-KO MC38 tumors (n = 9) were euthanized on day 15 post tumor cell inoculation. The numbers of immune effector cells per mg of tumor were then obtained using flow cytometry. Error bars represents mean ± SEM.

(J–N) ImmunoSEQ analysis of TCRβ repertoire. Total productive TCRs (J), productive unique TCRs (K), productive clonality (L), and maximum productive frequency (M) were shown for control and Vhl-KO tumors. Heatmap (N) showing the top 10 (5%) most frequent productive TCR sequences in VC and Vhl-KO tumors.

(O and P) In vivo tumor growth and Kaplan-Meier survival analysis of C57BL/6J mice bearing VC and Vhl-KO MC38 tumors (n = 6) following depletion of CD8⁺ T cells using anti-CD8 antibodies. Each mouse received i.p. injection of 100 μg isotype or αCD8b antibodies on days 1, 4, and 7 post tumor cell inoculation. Error bars represents mean ± SEM. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; ∗∗∗∗p < 0.0001; n.s. not significant.

To further characterize the effect of Vhl-KO on intratumoral T cells, we analyzed the TCR repertoire of control and Vhl-KO MC38 tumors using the ImmunoSEQ approach.³⁴ Our analysis indicated that the numbers of both total productive TCRs (Figure 2J) and productive unique TCRs (Figure 2K) were significantly increased in Vhl-KO tumors, suggesting substantially more T cells with increased TCR diversity. Furthermore, productive clonality, an index representing the diversity of rearranged T cell clones, increased substantially in Vhl-KO MC38 tumors (Figure 2L). Since each distinct TCR clonal expansion can be regarded as a result of T cell proliferation and activation in response to the specific recognition of tumor antigen fragments, higher productive clonality in Vhl-KO MC38 tumors suggests the activation of different T cell clonotypes that are likely to target a higher number of different tumor-specific antigens. Moreover, the maximum productive frequency measuring the frequency of dominant TCR clones also trended higher in Vhl-KO MC38 clones, despite not reaching statistical significance (Figure 2M). Finally, the predominant T cell clones in Vhl-KO tumors had different TCR sequences than those in the control tumors (Figure 2N), suggesting activation of different T cell subsets targeting different tumor antigens between VHL-WT and Vhl-KO tumors.

To evaluate the relative contribution of different immune effector subsets involved in antitumor immunity, we used antibodies to deplete CD4⁺ T cells, CD8⁺ T cells, and NK cells in mice bearing control and Vhl-KO MC38 tumors and observed their growth kinetics. Administration of an αCD8 antibody significantly accelerated the growth of vector control (VC) and Vhl-KO tumors and completely abrogated the substantial growth delay of Vhl-KO tumors (Figures 2O and 2P). On the other hand, administration of αCD4 and αNK1.1 antibodies did not affect tumor growth compared to those receiving isotype controls (Figures S2B–S2E). Therefore, our data suggest a pivotal role for CD8⁺ T cells in mediating the growth delay in tumors with Vhl gene loss.

VHL deficiency causes constitutive activation of type I interferons

To unravel the molecular mechanism of how VHL loss stimulates the antitumor immune response, we analyzed a publicly available dataset (accession number: GSE108229) comparing the transcriptional differences between parental 786-O RCC cells that carry a homozygous nonsense mutation in the VHL gene and 786-O cells with ectopic human VHL (hVHL) expression.³⁵ Consistent with our findings from in vivo tumor growth delay studies, Gene Ontology analysis showed that VHL deficiency led to the activation of multiple immune-stimulating pathways (Figure 3A). Notably, we found that IFNα/β signaling is highly upregulated in VHL-deficient 786-O cells (Figure 3A). To further validate this finding, we generated VHL-KO Caki-1 cells, a human clear cell carcinoma cell line with WT VHL expression (Figures S3A and S3B). We then carried out RNA sequencing (RNA-seq) to identify transcriptomic changes in VHL-KO Caki-1 cells. Principal component analysis of the RNA-seq data demonstrated significant changes between VC and VHL-KO Caki-1 cells (Figure S3C). Furthermore, gene set enrichment analysis (GSEA) identified significant enrichment of gene signatures involving IFNα/β signaling and positive regulation of T cell-mediated cytotoxicity in VHL-KO Caki-1 cells (Figures 3B and S3D), consistent with our findings in 786-O cells and intratumoral lymphocytes infiltration in MC38 tumors. IFNα/β are type I interferons (IFNs) that can upregulate antigen presentation on the surface of tumor cells, which in turn stimulates the cross-presentation of tumor-specific antigens by professional antigen presentation cells such as macrophages and dendritic cells, which are essential for antitumor immunity.³⁶ Indeed, flow cytometry analysis showed a higher level of mouse H2K^b/H2D^b expression on the surface of Vhl-KO MC38 cells compared to control MC38 cells (Figures S3D and S3E). Furthermore, the mRNA levels of multiple human MHC class I (human leukocyte antigen) proteins were also substantially higher in VHL-KO Caki-1 cells (Figure S3F). These data thus provide strong evidence indicating a vital role for VHL deficiency in stimulating the type I IFNs.

Figure 3.

Type I interferon production is responsible for VHL-KO-induced tumor growth suppression

(A) Gene Ontology (GO) analysis shows top-10 pathways enriched in 786-O cells compared to 786-O cells with restored wild-type human VHL (hVHL) expression using publicly available data (accession number: GSE108229).

(B) GSEA score curve plot showing the IFNα/β signaling pathway enriched in VHL-KO Caki-1 cells. NES, normalized enrichment score.

(C) Quantitative reverse-transcription PCR (RT-qPCR) analysis of mRNA levels of Ifnα and Ifnβ in VC, Vhl-KO MC38 cells, and Vhl-KO MC38 cells with an exogenously expressed wild-type mouse VHL (mVhl) gene. We performed three independent experiments. Error bars represent mean ± SD.

(D and E) Tumor growth and Kaplan-Meier survival analysis of C57BL/6J mice bearing VC and Vhl-KO MC38 tumors (n = 6) following αIFNAR1 antibody treatments to suppress type I interferon signaling. About 100 μg isotype or αIFNAR1 antibodies per mouse were administered i.p. on days 5, 8, and 11 post tumor cell inoculation. Error bars represent mean ± SEM. p < 0.05 is considered statistically significant. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; ∗∗∗∗p < 0.0001; n.s. not significant.

To further characterize the involvement of IFNα/β signaling in suppressing the growth of VHL-KO murine tumors, we examined the expression of Ifnα and Ifnβ in VC and Vhl-KO MC38 cells, and Vhl-KO MC38 cells with restored mVhl gene expression. Ifnα/β mRNA levels increased by approximately 32-fold and 64-fold in Vhl-KO MC38 cells, respectively. Notably, expression levels of both Ifnα/β reverted to control levels with the restoration of mVhl expression in Vhl-KO MC38 cells (Figure 3C). To determine the functional importance of IFNα/β signaling in regulating antitumor immunity in vivo, we conducted a tumor growth delay experiment where we blocked the type I IFN signaling in vivo by injecting an αIFNAR1 antibody, which blocks the receptor for both IFNα/β. Strikingly, blocking IFNAR signaling abrogated the tumor growth delay in Vhl-KO MC38 cells. Moreover, tumor growth from Vhl-KO MC38 tumors in mice receiving the αIFNAR1 antibody was even faster than that of control tumors, similar to VC MC38 tumors treated with the αIFNAR1 antibody (Figures 3D and 3E). These results suggest that type I IFN signaling played an essential role in Vhl-KO-mediated antitumor immunity.

Constitutive activation of the cGAS-STING pathway responsible for VHL loss-induced type I IFN activation

To determine the mechanism of type I IFN induction in VHL-KO cells, we further examined RNA-seq data of VHL-KO Caki-1 cells using GSEA analysis. Our analysis demonstrated that VHL deficiency induced the expression of genes associated with the cyclic GMP-AMP synthase (cGAS)-stimulator of interferon genes (STING) pathway involved in sensing the cytoplasmic presence of double-stranded DNA (dsDNA) (Figures 4A and 4B). The cGAS-STING pathway activation can induce interferon-stimulated genes and type I IFN production.^37,38,39 Based on this analysis, we conducted immunoblot analysis of proteins associated with the VHL protein and the cGAS-STING pathway. As expected, we found that VHL loss in MC38 and Caki-1 cells caused the accumulation of HIF1 and 2α (Figures 4C and 4D). In addition, it upregulated the phosphorylated STING in both cell lines (Figures 4C and 4D), consistent with our RNA-seq analysis. Interestingly, we also found an elevation in the total protein level of STING in all VHL-deficient cells. The cGAS expression was increased in Vhl-KO MC38 cells but reduced in VHL-KO Caki-1 cells. This is likely due to the cell variation between mouse and human lines (Figures 4C and 4D). Therefore, VHL may also regulate cGAS/STING protein levels through a mechanism that is not fully understood at present. Furthermore, VHL loss activated several downstream effectors of the cGAS-STING axis, including the TANK-binding kinase 1 (TBK1), interferon regulatory factor 3 (IRF3), and their phosphorylated forms, consistent with the activation of the cGAS-STING pathway (Figures 4C and 4D). Furthermore, the observation of VHL gene loss-induced TBK1 activation was consistent with a recently published study.⁴⁰ In addition, we also observed a similar relationship between Vhl-KO and activation of the cGAS-STING pathway in B16F10 melanoma cells (Figure S4A). Moreover, to further establish the relationship between VHL loss and activation of the cGAS-STING axis, we re-expressed mVhl and hVHL in Vhl-KO mouse MC38 cells and VHL-deficient human 786-O cells, respectively (Figures 4C and 4E). Immunoblot analysis showed that the restoration of VHL abrogated the activation of STING, TBK1, and IRF3 in VHL-deficient cells (Figures 4C and 4E).

Figure 4.

Activation of cGAS-STING signaling in VHL-KO cells and ccRCC human tumor tissues

(A) GSEA analysis of control and VHL-KO Caki-1 cells indicating elevated expression of genes involved in sensing cytosolic DNA.

(B) Heatmap of top-10 differently expressed genes involved in regulating innate immune response to cytosolic DNA in control and VHL-KO Caki-1 cells.

(C and D) Western blot analysis examining the effect of Vhl/VHL gene loss on cGAS-STING signaling and its downstream effectors in MC38 cells (C) and Caki-1 cells (D). In (C), ectopic mVhl was re-expressed in Vhl-KO MC38 cells.

(E) Western blot analysis evaluating the effect of the ectopic expression of hVHL on cGAS-STING signaling in VHL-deficient 786-O cells.

(F and G) Immunoblot analysis of that STING and TBK1 activation in VHL/STING DKO Caki-1 (F) and MC38 (G) cells.

(H) Quantitative PCR determination of Ifnα/β levels in Vhl-KO, Sting-KO, and Vhl/Sting DKO MC38 cells.

(I and J) Tumor growth (I) and Kaplan-Meier analysis (J) of C57BL/6 mice subcutaneously inoculated with VC, Vhl-KO, or Vhl/Sting DKO MC38 cells and treated with isotype control or αPD-1 antibody (100 μg per mouse) on days 6, 9, and 12 post tumor cell inoculation. n = 5 per group.

(K–M) Comparisons of STING RNA expression levels between tumor tissues and adjacent normal tissues from TCGA_KIRC (clear cell renal cell carcinoma) (K), TCGA_KIRP (kidney renal papillary cell carcinoma) (L), and TCGA_KICH (kidney chromophobe renal cell carcinoma) (M) cohorts. STING expression is only significantly higher in ccRCC (KIRC) vs. adjacent tissues where VHL is mutated in most tumors. Error bars represent mean ± SD. p < 0.05 is considered statistically significant. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; ∗∗∗∗p < 0.0001; n.s. not significant.

To determine if the cGAS-STING pathway is essential for VHL gene loss-induced IFNα/β production, we generated VHL/STING double-KO (DKO) cells for Caki-1 and MC38 cells, respectively (Figures 4F and 4G). Immunoblot analysis showed STING-KO abrogated VHL loss-induced type I IFN response in Caki-1 (Figure 4F) and MC38 (Figure 4G) cells. In contrast, genetic KO of Mda5, a factor involved in sensing the cytoplasmic presence of double-stranded RNA, did not affect Vhl-KO-induced STING and TBK1 activation (Figure S4B). In addition, Sting-KO eliminated Vhl-KO-dependent Ifnα and Ifnβ mRNA upregulation (Figure 4H), suggesting the cGAS-STING-TBK1 axis is functionally responsible for VHL loss-induced type I IFN response. Notably, Vhl/Sting DKO completely reversed tumor growth delay and sensitization to anti-PD1 therapy caused by Vhl loss (Figures 4I and 4J). Furthermore, the increased intratumoral lymphocyte infiltration observed in Vhl-KO tumors, especially those of CD3⁺ T, CD8α⁺ T, and CD8⁺ GZMB⁺ T cells, was significantly reduced in Vhl/Sting DKO tumors (Figures S4C–S4H), further supporting the key role of cGAS-STING in Vhl-deficient tumor sensitivity to anti-PD1 blockades.

We also determined STING mRNA expression levels in three human The Cancer Genome Atlas (TCGA) RCC cohorts: kidney renal clear cell carcinoma (KIRC), kidney renal papillary cell carcinoma (KIRP), and kidney chromophobe renal cell carcinoma (KICH). Compared with STING expression levels in adjacent normal tissues, only KIRC, representing ccRCC, showed significantly higher expression levels (Figure 4K). In contrast, KIRP (Figure 4L) showed lower expression levels than normal tissues, while KICH showed no difference (Figure 4M). Because among three types of RCC, VHL function is lost in most KIRC but not in KICH or KIRP, the human RCC STING expression data were thus consistent with our preclinical observation.

Mitochondrial dysfunction and cytoplasmic leakage of mtDNA responsible for VHL deficiency-induced cGAS-STING activation

We next attempted to determine the source of the cytosolic dsDNA that triggers cGAS-STING activation. We first fractionated VC and VHL-KO Caki-1 cellular lysates into pellet and cytosolic fractions and verified the success of the fractionation by using HSP60 and HDAC1 as markers for mitochondrial and nuclear fractions, respectively^41,42,43 (Figure 5A). We then purified DNA from the cytosolic fractions and performed quantitative PCR analysis using primers that amplify nuclear (nucDNA) and mitochondrial (mtDNA)-encoded genes. Our results indicate no significant differences in the amount of cytosolic nucDNA between VC and VHL-KO Caki-1 cells (Figure 5B). In contrast, we consistently detected more mtDNA in the cytosolic fraction of VHL-KO than VC Caki-1 cells (Figure 5B), strongly suggesting mtDNA leakage into the cytoplasm as the trigger for cGAS-STING activation in VHL-KO cells. We also made similar observations in murine MC38 and B16F10 tumor cells (Figures S5A–S5D). We next attempted to detect the cytoplasmic presence of dsDNA by immunofluorescence staining in VC and VHL-KO Caki-1 cells. Using an antibody specific for dsDNA, we observed that most cytoplasmic DNA was co-localized with HSP60 in the mitochondria in control cells (Figure 5C). However, a substantial amount of dsDNA did not co-localize with HSP60 in VHL-KO cells (Figures 5C and S5E), suggesting extra-mitochondrial locations for the dsDNA. These data thus provide correlative evidence that mtDNA leaked into the cytoplasm in VHL-KO cells is responsible for the observed cGAS-STING activation. To determine if mtDNA is functionally required for cGAS-STING activation, we used an established protocol to deplete mtDNA by ethidium bromide (EtBr).^44,45,46 After a 21-day EtBr treatment, we could deplete most cytosolic mtDNA as confirmed by immunostaining using an anti-dsDNA antibody in Caki-1(Figure 5D) and MC38 (Figure S5F) cells. Our data suggest that EtBr treatment substantially suppressed cGAS-STING and/or type I IFN response VHL-KO Caki-1 (Figure 5E) and MC38 (Figures 5F and S5G) cells. These results, therefore, support the notion that mtDNA leakage into the cytoplasm by VHL-KO is responsible for activating the cGAS-STING pathway and inducing type I IFN production.

Figure 5.

Cytoplasmic leakage of mtDNA as the key driver of cGAS-STING activation in VHL-deficient cancer cells

(A) Immunoblot analysis validation of the absence of nuclear (HDAC1 as the marker) and mitochondrial (HSP60 as the marker) proteins. Wcl, whole cell lysate; Pel, pellet fraction; Cyto, cytosolic fraction.

(B) RT-qPCR measurement of mtDNAs in the cytosols of VC and VHL-KO Caki-1 cells (n = 3). Error bars represent mean ± SEM.

(C) Immunofluorescent localization of the mitochondria (as detected by HSP60) and dsDNA in the cytoplasm of VC and VHL-KO Caki-1 cells. Scale bar, 20 μm.

(D) Immunofluorescence detection of cytoplasmic dsDNA in VC and VHL-KO Caki-1 cells after a 21-day treatment in 100 ng/mL EtBr to deplete mtDNA. Scale bar, 20 μm.

(E) Immunoblot analysis of proteins involved in cGAS-SITNG signaling in VC and VHL-KO Caki-1 cells treated with vehicle or 100 ng/mL EtBr for 21 days.

(F) RT-qPCR analysis of the mRNA levels of IFNα and IFNβ in VC and VHL-KO Caki-1 cells treated with vehicle or 100 ng/mL EtBr for 21 days (n = 6). Error bars represent mean ± SD.

(G and H) Mitochondrial membrane potential (MtMP) characterization by JC-1 staining in VC and VHL-KO Caki-1 cells as detected by immunofluorescence microscopy (G) and flow cytometry (H). Scale bar, 20 μM. Error bars represent mean ± SD. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; ∗∗∗∗p < 0.0001; n.s. not significant.

(I) Immunoblot analysis of BNIP3 expression in VC and VHL-KO Caki-1 cells.

Since mtDNA leakage into the cytoplasm is most likely due to mitochondria damage, we sought to understand how VHL deficiency may disrupt mitochondrial functions. mtDNA leakage is usually associated with lowered mitochondrial membrane potential (MtMP), closely associated with the permeability changes in mitochondrial membranes.⁴⁷ To determine whether VHL deficiency influences mitochondrial membrane permeability, we assessed mitochondrial permeability by using the JC-1 fluorescent probe^48,49 in VC and VHL-KO cells. Flow cytometry analysis found a significantly higher green(monomer)/red(aggregate) fluorescence ratio of JC-1 in VHL-KO cells, indicating significantly lowered MtMP in VHL-deficient Caki-1 cells (Figures 5G, 5H, and S6A) and MC38 cells (Figures S6B–S6D). These data suggest that VHL deficiency causes a significant reduction in MtMP. How does VHL deficiency cause a decrease in MtMP? Previous studies have shown that hypoxia can induce the expression of BCL2 interacting protein 3 (BNIP3),^50,51 which can increase mitochondrial outer membrane permeability and decrease membrane potential.^52,53 Therefore, we hypothesized that VHL deletion was akin to subjecting cells to a condition of perpetual hypoxia by permanently upregulating HIFα expression, which could upregulate BNIP3 expression. Indeed, our data indicate that VHL KO significantly increased the level of BNIP3 (Figure 5I), consistent with our hypothesis.

Increased HIF1/2α and BNIP3 levels in VHL-deficient cells are responsible for mitochondrial dysfunction and mtDNA leakage

We next sought to understand the molecular mechanism of how VHL deficiency causes the reduction in MtMP. Because HIF1α and HIF2α are direct targets of VHL and are associated with hypoxia-induced mitochondrial dysfunction^2,3,14,54,55 and VHL-deficient cells had elevated levels of HIF1α and HIF2α (Figures 4C and 4D), we decided to test whether forced expression of exogenous HIF1α and HIF2α could activate the cGAS-STING pathway. Therefore, we introduced into Caki-1 cells mutant versions of hHIF1α or hHIF2α, both of which are resistant to prolyl hydroxylase-induced hydroxylation and thus resistant to VHL-mediated degradation. Ectopic expression of mutHIF1α significantly enhanced monomer/aggregates ratio in JC-1-stained Caki-1 cells compared with control cells (Figures 6A and S6E), indicating a dramatic reduction in MtMP. In comparison, expression of mutHIF2α only moderately reduced MtMP (Figures 6A and S6E). Furthermore, expression of mutHIF1α, but not mutHIF2α, induced elevated expression of STING and activation of TBK1 (as indicated by detection of p-TBK1) in Caki-1 cells (Figure 6B), similar to those observed in VHL-KO cells (Figure 4D). Furthermore, similar to VHL KO, mutHIF1α also induced the activation of BNIP3 in Caki-1 cells (Figure 6B), consistent with it being a direct transcriptional target of HIF1α. To test whether BNIP3, that is upregulated by VHL-KO-mediated HIF1α accumulation, is required to promote cGAS-STING signaling activation, we further generated a mixed cell population of BNIP3 knockdown and VHL/BNIP3 double knockdown (DKD) using CRISPR-Cas9 system (Figures 6C and S6G). VHL/BNIP3 DKD almost completely abrogated VHL-KO-induced activation of STING and TBK1 (Figures 6C and S6G). The MtMP was also restored with the depletion of BNIP3 (Figure 6D), indicating the change induced by VHL loss in MtMP is mediated by BNIP3. Therefore, elevated cGAS-STING signaling due to VHL loss is most likely regulated by the VHL-HIF1α-BNIP3 axis.

Figure 6.

Elevated HIF1α and HIF2α levels are essential and sufficient for VHL deficiency-induced mitochondrial dysfunction and cGAS-STING activation

(A) Quantitative JC-1-based MtMP measurements in Caki-1 cells expressing VC, degradation-resistant mutHIF1α (hHIF1α-p402A/p564A), or mutHIF2α (hHIF2α-p405A/p531A) by flow cytometry. Error bars represent mean ± SD.

(B) Immunoblot analysis of cGAS-STING signaling proteins and BNIP3 in Caki-1 cells expressing exogenous VC, mutHIF1α, or mutHIF2α genes.

(C) Immunoblot analysis of cGAS-STING signaling proteins in VC, BNIP3-KO, VHL-KO, and VHL/BNIP3 DKO Caki-1 cells.

(D) JC-1 assay for MtMP measurements in VC, VHL-KO, BNIP3-KO, and VHL/BNIP3 DKO Caki-1 cells. Error bars represent mean ± SEM.

(E) Immunoblot analysis of cGAS-STING signaling proteins and BNIP3 in VC, VHL-KO, VHL/HIF1α DKO, VHL/HIF2α DKO, and VHL/HIF1α/HIF2α TKO Caki-1 cells.

(F) JC-1-based MtMP measurements in VC, VHL-KO, VHL/HIF1α DKO, VHL/HIF2α DKO, and VHL/HIF1α/HIF2α TKO Caki-1 cells by flow cytometry. Error bars represent mean ± SD.

(G) Quantitative PCR measurements of cytosolic mtDNAs in VC, VHL-KO, VHL/HIF1α DKO, VHL/HIF2α DKO, and VHL/HIF1α/HIF2α TKO Caki-1 cells.

(H and I) Tumor growth (G) and Kaplan-Meier analysis (H) of C57BL/6 mice subcutaneously inoculated with VC, Vhl-KO, or Vhl/Bnip3 DKO MC38 cells and treated with isotype control or an αPD-1 antibody (100 μg per mouse).

(J and K) Tumor growth (I) and Kaplan-Meier analysis (J) of C57BL/6 mice implanted with VC, Vhl-KO, or Vhl/Hif1a/Hif2a TKO MC38 cells and treated with isotype control or an αPD-1 antibody (100 μg per mouse). (G and I) Treatments were given on days 6, 9, and 12 post tumor cell inoculation. n = 5 per group.

(L–O) Normalized mRNA expression levels of cGAS (K), STING (L), BNIP3 (M), and MAVS (N) in tumor tissues and normal tissues from the TCGA_KIRC (ccRCC) cohort. Tumor tissues have higher levels of expressions with the exception of MAVS, which is not part of the cGAS-STING dsDNA-sensing pathway. Shown are results from three independent experiments. Error bars represent mean ± SEM. Two-way ANOVA. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; ∗∗∗∗p < 0.0001; n.s. not significant.

We further investigated whether HIF1α and HIF2α were essential for reducing MtMP and causing cytosolic mtDNA leakage in VHL-deficient cells. For this purpose, we generated VHL/HIF1α DKO, VHL/HIF2α DKO, and VHL/HIF1α/HIF2α triple-KO (TKO) Caki-1 cells, respectively (Figure 6E). Notably, KO of either HIF1α or HIF2α or both significantly reduced VHL-KO-induced activation of BNIP3 and cGAS-STING signaling (Figure 6E). This result was different from those in Figure 6A, where only forced expression of a stabilized mutHIF1α but not the stabilized mutHIF2α reduced MtMP. We suspect this discrepancy was caused by the different VHL background where the functions of HIF1α and HIF2α are examined. In Figures 6A and 6B, the VHL was WT while in Figures 6E and 6F, VHL was knocked out. In order to examine the role of VHL background on the functions of mutHIF1α and mutHIF2α in disrupting MtMP, we overexpressed the mutHIF1α and mutHIF2α in VHL-KO Caki-1 cells, respectively. Our results showed that HIF1a overexpression in VHL-KO cells further boosted the already activated BNIP3, similar to what is seen in cells with WT VHL (Figure S6F). On the other hand, different from HIF2a overexpression in VHL-WT cells, exogenous HIF2α in VHL-KO cells elevated BNIP3 activation and TBK1 activation compared to VHL-KO control cells (Figure S6F), suggesting differential abilities of HIF2α in VHL-WT and VHL-deficient cells.

Therefore, we conclude that, in VHL-KO cells, both HIF1α and HIF2α stabilization are required to activate cGAS-STING signaling. In agreement with this finding, VHL-KO-mediated reduction of MtMP was largely rescued in VHL/HIF1α DKO, VHL/HIF2α DKO, and VHL/HIF1α/HIF2α TKO cells, suggesting that both HIF1α and HIF2α are required for the VHL deficiency-induced mitochondrial membrane permeabilization (Figures 6F and S6H). Consistent with this finding, individual or combined HIF1/2α deletions in VHL-KO cells significantly reduced cytosolic mtDNA (Figure 6G), while changes of cytosolic nucDNA were insignificant (Figure S6I). Importantly, further KO of Bnip3 and Hif1α/2α abrogated tumor growth delay observed in Vhl-KO MC38 tumors. In addition, loss of Bnip3 or Hif1α/Hif2α in Vhl-deficient MC38 tumors also attenuated their sensitivity to ICB treatments significantly (Figures 6H–6K). Taken together, our results suggest that HIF1α and HIF2α are the critical effectors downstream of VHL responsible for reducing MtMP and increasing membrane permeability via BNIP3 to activate cGAS-STING signaling.

Many human tumors are resistant to immunotherapy because they downregulate STING signaling by various mechanisms,^56,57 a fact that testifies to the importance of the cGAS-STING pathway in mediating tumor response to ICB therapy. In order to see if the cGAS-STING pathway is indeed upregulated in ccRCC as has been shown in our experiments, we analyzed the mRNA expression levels of cGAS and STING in a cohort of patients with ccRCC from the TCGA Pan-Cancer cohort. Our analysis indicated that both cGAS (Figure 6L) and STING (Figure 6M) were expressed at higher levels in tumor vs. adjacent normal tissues, thereby confirming the activation of the cGAS-STING pathway. Furthermore, BNIP3 was similarly expressed at higher levels in tumor tissues (Figure 6N). Therefore, data from patients with ccRCC are consistent with the activation of the mtDNA-cGAS-STING pathway. In contrast, expression of mitochondrial antiviral signaling protein (MAVS), a vital component of the dsRNA-sensing pathway, showed no difference between tumor and adjacent normal tissues (Figure 6O). Taken together, our analysis of the TCGA human ccRCC data is consistent with the activation of the mtDNA-cGAS-STING pathway in these patients, the majority of whom have VHL deficiencies.

Discussion

Our finding of VHL gene loss causing constitutive activation of the cGAS-STING signaling and increased intratumoral lymphocyte infiltration (Graphic Abstract) provides critical insights into how ccRCC with moderate TMB unexpectedly responds well to ICB therapy,^58,59 despite the discrepancy that several published clinical studies did not show an improved response to ICB treatment in patients with VHL-mut ccRCC. We surmise that there could be several factors. One confounding factor might be that patients from most of the published studies also received different co-treatments. For example, in Javelin Renal 101,⁶⁰ ImMotion150,⁶¹ and ImMotion 151⁶² RCC cohorts, patients receiving anti-PD-L1 treatment were also treated with VEGF inhibitors or anti-VEGF treatments concurrently. The other factor undermining the use of VHL genomic mutation as a biomarker for ICB therapy may be that VHL genomic mutations alone may not fully reflect the actual functional status of the VHL gene. In fact, even those patients with ccRCC with no VHL genomic mutations may have compromised VHL functions. For example, methylation-mediated VHL gene inactivation or loss of heterozygosity, which may occur in up to 98% of patients with ccRCC,^35,63 may also disrupt VHL function. Therefore, even though we believe that compromised VHL function underlies the overall ccRCC responsiveness to ICB therapy, genomic mutations may not be a suitable biomarker for predicting individual ccRCC response to ICB therapy because of the prevalence of compromised VHL functions in ccRCC.

How do we reconcile the paradoxical roles of VHL loss, which promote tumor development in certain cancers, including ccRCC, and also predispose better response of tumors to ICB therapy? It may have to do with the dual roles of the cGAS-STING pathway, which acts downstream of VHL. Indeed, chronic cGAS-STING has been reported to promote tumor invasion and metastasis.⁶⁴ In fact, the paradoxical roles of VHL mutations in promoting tumor development while making them susceptible to ICB therapy are not unique. Similar examples include mutations in the ataxia telangiectasia mutated gene and mismatch repair genes. Both are well-established tumor suppressor genes whose gene deficiency promotes tumorigenesis but makes them more susceptible to immunotherapy.^44,65

Understanding the function of VHL in the immune response is complex. While its significance in tumor growth and angiogenesis is widely acknowledged, its effect on the immune system is just as significant and interrelated to these processes. Researchers are currently exploring how VHL and HIF signaling can influence immune responses. For instance, VHL and subsequent HIF1α stabilization can not only promote the expression of pro-inflammatory cytokines but also influence the fate and function of immune cells.^66,67 However, VHL loss or mutations may also contribute to chronic inflammation and the development of inflammatory diseases.^68,69,70 Therefore, further investigation regarding the intricate balance of VHL in inflammation is crucial for developing therapeutic strategies that leverage its benefits while mitigating its potential contributions to chronic inflammation and related diseases. Our efforts to uncover the precise mechanisms of VHL behind cancer immunotherapy could potentially lead to new therapeutic applications.

In addition, the constitutive cGAS-STING activation in VHL-deficient tumors can potentially create additional therapeutic opportunities beyond ICB immunotherapy. For example, VHL-deficient tumors may be more susceptible to chimeric antigen receptor T (CAR-T) therapy due to elevated levels of type I IFNs in the tumor microenvironment. Indeed, recent studies have shown that NKG2D CAR-engineered T cells can synergize with STING agonists in suppressing the growth of pancreatic tumors.⁷¹ By the same token, adoptively transferred tumor-infiltrating lymphocytes may have a better chance of penetrating the tumor mass and staying active due to a more hospitable TIME created by constitutive cGAS-STING activation.^72,73 Another possibility is using STING agonists in combination with ICB therapy in RCC. The agonists may “supercharge” the constitutively high STING levels and create a more immune-stimulating tumor microenvironment that further enhances ICB therapy.^74,75,76

In conclusion, our study provides insights into the VHL deficiency-induced cGAS-STING activation, providing a mechanistic underpinning of why tumors respond to ICB therapy. Finally, it also suggests potential treatment strategies for targeting VHL-deficient tumors based on constitutive cGAS-STING and type I IFN activation.

Limitations of the study

Our conclusions are mostly drawn from experiments conducted in murine tumor models. Therefore, whether there is a similar role for VHL mutation in ccRCC ICB therapy still awaits further studies in human patients. Furthermore, human studies in ccRCC have to resolve confounding factors such as high VHL mutation rates in patients with ccRCC, co-mutations with VHL, and frequent combination treatments such as kinase inhibitors. The status of cGAS-STING activation in patients with ccRCC also needs to be confirmed in human patients with cancer.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Trustain FcXTM anti-mouse CD16/32	BioLegend	Cat #101320; RRID:AB_1574975
FITC anti–mouse CD45	BioLegend	30-F11; Cat #103108; RRID:AB_312973
Pacific blue anti–mouse CD3	BioLegend	145-2c11; Cat #100334; RRID:AB_2028475
Alexa Fluor 647 anti–mouse CD4	BioLegend	GK1.5; Cat #100424; RRID:AB_389324
APC/FireTM 750 anti–mouse CD8a	BioLegend	53–6.7; Cat #100766; RRID:AB_2572113
phycoerythrin (PE) anti–mouse NK1.1	BioLegend	PK136; Cat #108707; RRID:AB_313394
APC anti–γ/δTCR	BioLegend	GL3; Cat #118116; RRID:AB_1731813
PE anti–mouse F4/80	BioLegend	BM8; Cat #123110; RRID:AB_893486
phycoerythrin (PE) anti–human/mouse Granzyme B	BioLegend	QA16A02; Cat #372207; RRID:AB_2687031
Alexa Fluor 647 anti–mouse IFN-γ	BioLegend	XMG1.2; Cat #505816; RRID:AB_493315
PE anti–mouse FOXP3	BioLegend	MF-14; Cat #126404; RRID:AB_1089117
PE anti-mouse H2-Kb/H2Db antibody	BioLegend	28-8-6, cat#114607; RRID:AB_313598
PE anti-mouse H-2Kb/H2Db antibody	BioLegend	Cat #114607; RRID:AB_313598
rat IgG2 isotype	BioXCell	Cat #BE0089; RRID:AB_1107769
rat anti-mouse αPD-1	BioXCell	Cat #BE0146; RRID:AB_10949053
mouse IgG1 isotype	BioXCell	Cat #EB0083; RRID:AB_1107784
anti-IFNAR1 antibody	BioXCell	Cat #BE0241; RRID:AB_2687723
anti-CD4 antibody	BioXCell	Cat #BE0003-1; RRID:AB_1107636
anti-CD8b antibody	BioXCell	Cat #BE0223; RRID:AB_2687706
anti-NK1.1 antibody	BioXCell	Cat #BE0036; RRID:AB_1107737
anti-cGAS	Cell Signaling Technology	D3080; cat# 31659; RRID:AB_2799008
anti-cGAS	Cell Signaling Technology	D1D3G; Cat #15102; RRID:AB_2732795
anti-STING	Cell Signaling Technology	2P2F; Cat# 13647; RRID:AB_2732796
anti-p-STING (Ser365)	Cell Signaling Technology	D8F4W; Cat #72971; RRID:AB_2799831
anti-p-STING (Ser366)	Cell Signaling Technology	D7C3S; Cat #19781; RRID:AB_2737062
anti-TBK1/Nak	Cell Signaling Technology	D1B4, Cat #3504; RRID:AB_2255663
anti–p-TBK1/p-Nak (Ser172)	Cell Signaling Technology	D52C2, Cat #5483; RRID:AB_10693472
anti-IRF-3	Cell Signaling Technology	D83B9; Cat #4302; RRID:AB_1904036
anti-p-IRF3 (Ser396)	Cell Signaling Technology	D601M; Cat #29047; RRID:AB_2773013
anti-HA-Tag	Cell Signaling Technology	C29F4; Cat #3724; RRID:AB_1549585
anti-HSP60	Cell Signaling Technology	D307; Cat #4870; RRID:AB_2295614
anti-HDAC1	Cell Signaling Technology	Cat #2062; RRID:AB_2118523
anti-BNIP3	Cell Signaling Technology	D7U1T; Cat #44060; RRID:AB_2799259
anti-MDA5	Cell Signaling Technology	D74E4, Cat#5321; RRID:AB_10694490
anti-dsDNA	Millipore Sigma	AC-30-10; Cat #CBL186; RRID:AB_11213573
anti-HIF-1 alpha	Novus Biologicals	H1alpha67; Cat #NB100-105; RRID:AB_10001154
anti-HIF-2 alpha/EPAS1	Novus Biologicals	Cat #NB100-122; RRID:AB_10002593
anti-GAPDH	Proteintech	cat #60004-1-Ig; RRID:AB_2107436
Anti-VHL	Santa Cruz Biotechnology	VHL40; Cat #sc-135657; RRID:AB_2215955
anti-actin	Thermo Fisher Scientific	ACTN05-C4; Cat #MA5-11869; RRID:AB_11004139
Alexa Fluor® 488 goat anti-mouse IgG	Thermo Fisher Scientific	Cat #A28175; RRID:AB_2536161
Alexa Fluor® 555 goat anti-rabbit IgG	Thermo Fisher Scientific	Cat #A27039; RRID:AB_2536100
Bacterial and virus strains
lentiCRISPRv2	Addgene	#52961; RRID:Addgene_52961
psPAX2	Addgene	#12260; RRID:Addgene_12260
pMD2.G	Addgene	#12259; RRID:Addgene_12259
lentiCRISPRv2 neo vector	Addgene	#98292; RRID:Addgene_98292
px330-mcherry	Addgene	#98750; RRID:Addgene_98750
pSpCas9(BB)-2A-GFP(PX458)	Addgene	#48138; RRID:Addgene_48138
HA-VHL-pRC/CMV	Addgene	#19999; RRID:Addgene_19999
pcDNA3-HA-HIF1α(P402A/P564A)	Addgene	#18955; RRID:Addgene_18955
pcDNA3-HA-HIF2α(P405A/P531A)	Addgene	#18956; RRID:Addgene_18956
HA-tagged mouse VHL ORF Clone	Origene	Cat #MR201630
LentiORF pLEX vector	Thermo Scientific	Cat #OHS4735
Biological samples
Xenograft MC38 tumors	This paper
Chemicals, peptides, and recombinant proteins
Trizol® Reagent	Ambion by life technologies	Cat #15596018
SDS	Bio-Rad	Cat #1610302
30% Acrylamide/Bis Soln, 37.5:1, 500 mL	Bio-Rad	Cat # 1610158
Resolving Gel Buffer, 1 L	Bio-Rad	Cat # 1610798
Stacking Gel Buffer 1L	Bio-Rad	Cat # 1610799
4x Laemmli Sample Buffer	Bio-Rad	Cat # 1610747
fixation buffer	BioLegend	Cat #420801
Ammonium persulphate	GE healthcare	Cat # 17-1311-01
RPMI-1640	Gibco	Cat #11875-119
Geneticin™ Selective Antibiotic (G418 Sulfate)	Gibco	Cat #10131035
DMEM medium	Gibco	Cat #11995073
McCoy’s 5A (Modified) Medium	Gibco	Cat # 16600082
Sodium pyruvate 100mM	Gibco	Cat #11360070
OPTI-MEM® I	Gibco	Cat # 11058021
HBSS	Gibco	Cat #14025092
Trypsin-EDTA (0.25%)	Gibco	Cat #25200114
Trypan Blue Solution, 0.4%	Gibco	Cat #15250061
penicillin-Streptomycin	Gibco	Cat #15140122)
FBS	Hyclone	Cat # SH30396.03
1× intracellular staining perm wash buffer	Invitrogen	Cat #00-8333-56
TURBOTM DNase	Invitrogen	Cat #AM2238
SuperScript II Reverse Transcriptase	Invitrogen	Cat #18064014
Paraformaldehyde	J.T. BakerTM	Cat #S898-07
DNase I	Millipore Sigma	SKU #10104159001
digitonin	Millipore Sigma	SKU #D141-100MG
BsmBI	NEB	Cat #R0580
BbsI	NEB	Cat #R0539S
Phusion® High Fidelity DNA Polymerase	NEB	Cat #M0530S
T4 DNA ligase	NEB	Cat #0202L
T4 Polynucleotide Kinase	NEB	Cat #0201S
1× red blood cell lysis buffer	Roche	Cat #11814389001
Puromycin	Sigma	Cat #8833
Ampicillin	Sigma	Cat #A0166
D-(+)-GLUCOSE SOLUTION 45% IN H2O	Sigma	Cat #G8769-100ML
1× RIPA buffer	Sigma	Cat #R0278
1× protease inhibitors	Sigma	Cat #P8340
Collagenase type IV	Sigma	Cat #C5138
ethidium Bromide (EtBr)	Sigma	Cat #E1510
chloroform	Sigma	Cat #C2432
bovine serum albumin	Sigma	Cat #A3983
Triton X-100	Sigma	Cat #T8787
LipofectamineTM 2000	Thermo Fisher Scientific	Cat #11668019
PBS (10X), pH 7.4	Thermo Fisher Scientific	Cat # 70011044
SuperSignalTM West Pico PLUS Chemiluminescent Substrate	Thermo Fisher Scientific	Cat #34580
LIVE/DEADTM fixable dead cell staining	Thermo Fisher Scientific	Cat #L23105
VECTASHIELD® Antifade Mounting Medium with DAPI	VECTOR LABORATORIES	Cat #H-1200-10
Critical commercial assays
JC-1(tetraethylbenzimidazolylcarbocyanine iodide) Mitochondrial Membrane Potential Assay Kit	Abcam	Cat #113850
Universal Mycoplasma Detection Kit	ATCC	Cat #30-1012K
qPCRBIO SyGreen Blue Mix Hi-ROX	Genesee Scientific	Cat #17-506C
QIAGEN® DNeasy Blood & Tissue kit	QIAGEN	Cat # 69506
Gene JET Gel extraction kit	Thermo Fisher Scientific	Cat #K0692
DNA Clean & Concentrator-5 Kit	ZYMO RESEARCH	Cat #D4004
ZR Plasmid Miniprep Classic Kit	ZYMO RESEARCH	Cat #D4016
Deposited data
Caki-1 RNAseq data	This paper	GEO: GSE196509
786-O RNAseq data (Published)	GEO Database	GEO: GSE108229
Experimental models: Cell lines
HEK293T	ATCC	CRL3216™; RRID:CVCL_0063
Renca	ATCC	CRL2947™; RRID:CVCL_2174
B16F10	ATCC	CRL6475™; RRID:CVCL_A4CJ
Caki-1	ATCC	HTB-46™; RRID:CVCL_0234
786-O	ATCC	CRL1932™; RRID:CVCL_1051
MC38	Kerafast	ENH204-FP; RRID:CVCL_B288
Experimental models: Organisms/strains
C57BL/6J	Jackson Laboratory	Strain #:000664; RRID:IMSR_JAX:000664
Balb/C	Jackson Laboratory	Strain #:000651; RRID:IMSR_JAX:000651
Oligonucleotides
random hexamer primers	Invitrogen	Cat #SO142
sgRNA for CRISPR/Cas9 mediated gene knockout	IDT (Integrated DNA Technology)	Table S1
Primers for ectopic gene expression	IDT (Integrated DNA Technology)	Table S2
Primers for quantitative RT-PCR and quantitative PCR	IDT (Integrated DNA Technology)	Table S3
Software and algorithms
ImmunoSEQ® mouse TCR-β CDR3 survey sequencing and ImmunoSEQ® Analyzer 3.0	Adaptive Technologies	NA
FACS Canto II Flow Cytometer	BD	NA
Astrios Sorter	Duke Cancer Institute flow cytometry core facility	NA
Prism	GraphPad	8.2.0
Illumina NovaSeq 6000	Illumina	NA
Leica TCS SP5 laser scanning confocal microscope	Leica	NA
Odyssey® Fc imaging system	LI-COR® Biosciences	NA
R	The R foundation	https://www.r-project.org/
Applied Biosystems® ViiATM 7 Real-Time PCR System with 384-well Block	Thermo Fisher Scientific	Cat #4453536

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact (chuan.li@duke.edu).

Materials availability

This study did not generate new unique reagents. Plasmids and cell lines generated in this study are available upon request. All the other materials in this study are commercially available. Any additional analysis information for this work is available by request to the lead contact.

Data and code availability

Raw RNAseq and metadata are deposited in the Gene Expression Omnibus database and are publicly available as of the date of publication. The accession number is listed in the key resources table. The code for RNAseq analysis is provided in detail in the supplemental information. Any additional information required to reanalyze the data reported in this paper is available upon request. Source data for other figures will also be provided upon request from the lead contact.

Experimental model and study participant details

Clinical samples and public datasets

All human data we obtained are publicly available. To compare the expression of cGAS, STING, BNIP3, and MAVS in RCC tumor versus normal tissues, we accessed three cohorts of studies from the open-access online tool cBioPortal (http://www.cbioportal.org).^77,78 These studies include kidney renal clear cell carcinoma (KIRC), kidney renal papillary cell carcinoma (KIRP), and kidney chromophobe carcinoma (KICH) patients from the Cancer Genome Atlas (TCGA) Pan-Cancer studies.^79,80,81

We used a published dataset (GSE108229) to perform gene ontology (GO) analysis. Top-ranked pathways were plotted with ‘ggplot2’ R package.

Cell lines and cell culture

We purchased the Renca mouse renal adenocarcinoma cells (ATCC CRL-2947), B16F10 mouse melanoma cells (ATCC CRL-6475), Caki-1 human clear cell carcinoma cells (ATCC HTB-46), and 786-O human renal cell adenocarcinoma (ATCC CRL-1932) from the Cell Culture Facility of Duke University School of Medicine. In addition, we obtained the MC38 mouse colon adenocarcinoma cells from Kerafast (Boston, MA). All cells were cultured and maintained following the manufacturer’s instructions. All cells undergo periodic mycoplasma testing using the Universal Mycoplasma Detection Kit from ATCC (Cat #30-1012K) to ensure they are mycoplasma-free.

In vivo tumor growth studies in mice

We purchased six-week-old C57BL/6J and Balb/C female mice from the Jackson Laboratory. Duke University Institutional Animal Use and Care Committee (IACUC) approved all mouse experiments in this study. For in vivo tumor growth experiments, we inoculated about 1∗10⁶ Renca cells, 1∗10⁵ B16F10 cells, and 5∗10⁵ MC38 cells (suspended in 50 μL 1× PBS) subcutaneously into the right flanks of syngeneic mice, respectively. The treatments were given via intraperitoneal injection according to the corresponding schedule. We measured tumor sizes by measuring the longest (length) and shortest (width) dimensions of the tumors every 2–3 days using a digital caliper and calculated tumor volumes using the following formula: (length) × (width)²/2. We euthanized mice when their volumes reached 2000 mm³.

In conducting immunotherapy in mice, we injected the antibodies intraperitoneally (i.p.), either with 100 μg of rat IgG2 isotype control (clone 2A3; Bio X Cell; Cat #BE0089), or 100μg of rat anti-mouse αPD-1 antibody in 150 μL 1×PBS per mouse on Day 5, 8, and 11 for Renca cells, on Day 7, 10, 13, and 16 for B16F10 cells, and on Day 6, 9, 12 for MC38 cells after tumor inoculation, respectively.

In experiments involving αIFNAR1 antibody treatments, we injected mice (i.p.) bearing MC38 tumors with 200 μg mouse IgG1 isotype control (clone MOPC-21; BioXCell; Cat #EB0083) or 200 μg mouse αIFNAR1 antibody (clone MAR1-5A3; BioXCell; Cat #BE0241) in 150 μL PBS per mouse on Day 5, 8, and 11, respectively.

In studies involving in vivo lymphocyte depletion, we injected 100 μg αCD4 antibody (Clone GK1.5; BioXCell; Cat #BE0003-1), or αCD8b antibody (Clone 53-5.8; BioXCell; Cat #BE0223), or 100 μg αNK1.1 antibody (Clone PK136; BioXCell; Cat #BE0036) into each MC38 tumor-bearing mouse on Day 1, 4, and 7 to deplete CD4⁺ T cells, CD8⁺ T cells, and NK cells, respectively. We also administered an equal amount of IgG isotype antibodies to control mice.

Method details

CRISPR/Cas9-mediated gene knockout

We used the CRISPR/Cas9 system to generate gene-specific knockout cells. We designed single guided RNAs (sgRNAs) using a public domain web-based CRISPR sgRNA design tool CHOPCHOP (https://chopchop.cbu.uib.no). Table S1 lists individual sgRNA sequences targeting different genes. When generating Human VHL-KO, human STING-KO, human HIF1α-KO, human HIF2α-KO, mouse Sting-KO, and mouse Mda5-KO human and murine cancer cells, we used the lentiCRISPRv2 vector (Addgene #52961) following a published protocol from the Zhang lab.⁸² We digested LentiCRISPRv2 with BsmBI (NEB; Cat #R0580) and gel purified it using the Gene JET Gel extraction kit (Thermo Fisher Scientific; Cat #K0692). Oligos encoding sgRNA sequences were phosphorylated, annealed, and subsequently ligated into digested LentiCRISPRv2.

To produce sgRNA-encoding lentivirus vectors, we used HEK293T cells co-transfected with lentiviral constructs encoding the target sgRNAs and second-generation packaging plasmids psPAX2 (Addgene #12260) and pMD2.G (Addgene #12259) following the instruction from the Trono Lab (https://www.epfl.ch/labs/tronolab/laboratory-of-virology-and-genetics/lentivectors-toolbox/).

To generate clonal knockout cell lines, we infected target cells with sgRNA-encoding CRISPR/Cas9 lentivirus and cultured them in a complete cell growth medium, selected with puromycin (1 μg/ml for Caki-1) for 7–10 days. Cells were then collected to test the expression of target genes by immunoblot.

To generate clonal VHL-KO Caki-1 cells, cells were seeded in 96-well plates after a 7-day puromycin selection and screened for pure knockout clones verified by immunoblot analysis. To generate VHL/STING double knockout (DKO) cells, including VHL/HIF1α DKO, VHL/HIF2α DKO, or VHL/HIF1α/HIF2α triple knockout (TKO) Caki-1 cells, we cloned sgRNAs sequences encoding HIF1α, and/or HIF2α into digested lentiCRISPRv2 neo vector (Addgene #98292), respectively. We then infected VHL-KO Caki-1 cells with the sgRNA-encoding lentivirus and selected the cells with neomycin (2 mg/ml) for 10–14 days. Lentivirus prepared from lentiCRISPRv2 neo vector was used to generate control cells. We then detected protein levels of STING, HIF1α, and/or HIF2α knockdown (KD) by western blot. To generate VHL/STING DKO and VHL/MDA5 DKO Caki-1 and MC38 cells, we infected VHL-KO Caki-1 cells with sgRNA-encoding lentivirus and cultured them in complete DMEM medium, and selected with puromycin (5 μg/ml for MC38 cells and 1 μg/ml for Caki-1 cells) for 7 days. We then seeded the cells in 96-well plates for single clone selection. We then verified the cells’ VHL/STING DKO and VHL/MDA5 DKO status by immunoblot analysis.

To generate Vhl knockout in Renca, B16, or MC38 cells, we digested the px330-mcherry (Addgene #98750) and pSpCas9(BB)-2A-GFP(PX458) (Addgene #48138) with BbsI (NEB; Cat #R0539S), and then gel-purified, and ligated the vectors with annealed oligos encoding sgRNA_1 or sgRNA_2 targeting Vhl. We then transfected the vectors in the murine tumor cells with sgRNA-encoding constructs using Lipofectamine 2000 (Thermo Fisher Scientific; Cat #11668019). Cells transiently transfected with px330-mCherry and PX458 vectors alone were as controls (VC). We then cultured the cells using a complete DMEM medium [DMEM with 10% FBS and 100U/ml penicillin-Streptomycin (Gibco by Life Technologies; Cat #15140122)] for 6 days and subjected them to FACS sorting (Duke Cancer Institute flow cytometry core facility). We then seeded GFP⁺ mCherry⁺ cells in 96 well plates for single clone selection. We screened for VHL-KO single clones by immunoblot analysis. Table S1 shows the sgRNA primer sequences used for targeting various genes in this study.

Exogenous gene expression

We modified a commercially available LentiORF pLEX vector (Thermo Scientific; Cat #OHS4735) with EF-1α and used it to generate constructs for exogenous gene expression. Gene of interest was amplified by PCR using the Phusion High Fidelity DNA Polymerase. Specifically, we used HA-VHL-pRC/CMV (Addgene #19999) and HA-tagged mouse VHL ORF Clone (Origene; Cat #MR201630) as templates to amplify human and mouse VHL, respectively. In addition, we amplified mutant human HIF1/2α using the pcDNA3-HA-HIF1α(P402A/P564A) (Addgene #18955) plasmid and the pcDNA3-HA-HIF2α(P405A/P531A) plasmid (Addgene #18956). Modified pLEX vector was used for lentiviral production as controls. Table S2 lists the primer sequences for PCR reactions.

Immunoblot analysis

To prepare cellular lysates, we washed the cells quickly with ice-cold 1X PBS buffer twice and immediately lysed them in 1× RIPA buffer (Sigma; Cat #R0278) supplemented with 1× protease inhibitors (Sigma; Cat #P8340) on ice using a cell scraper. We then collected the lysates in a 1.5 mL tube and incubated them on ice for 10 min, then centrifuged them at 15,000 rpm for 15 min at 4°C. We then transferred the supernatants to a new 1.5 mL tube and boiled them for 5 min. We then loaded equal amounts of lysates into SDS-PAGE gels. After electrophoresis, we transferred the proteins to PVDF membranes, incubated them with primary antibodies overnight at 4°C, and then incubated them with HRP-conjugated secondary antibodies at room temperature for 1 h. Afterward, we incubated the membranes with the SuperSignal West Pico PLUS Chemiluminescent Substrate (Thermo Fisher Scientific; Cat #34580) and detected target signal strength with the Odyssey Fc imaging system (LI-COR Biosciences).

Flow cytometry

To analyze tumor-infiltrating lymphocytes, we implanted about 5∗10⁵ vector control or Vhl-KO, or Vhl/Sting DKO MC38 cells subcutaneously into the right flank of C57BL/6J mice. We euthanized tumor-bearing mice on day 15 post-inoculation. We removed and weighed the tumors and cut them into pieces as small as possible. We prepared 1× digestive enzyme solution with 0.2U/ml Collagenase type IV (Sigma; Cat #C5138) and 50 mg/mL DNase I (Millipore Sigma; SKU #10104159001) in HBSS. We then added the triple enzyme solution to the homogenates and incubated for 1 h at 37°C to ensure full digestion. We then passed the dissociated cells through a 70 μm cell strainer (BD Falcon; Cat #35–2350) and rinsed them with HBSS three times. The pellets were collected, lysed in 1× red blood cell lysis buffer (Roche; Cat #11814389001) on ice for 5 min, and rinsed with 1X wash buffer (2% FBS in 1X PBS). About 1∗10⁶ live cells were then blocked with TruStain FcX anti-mouse CD16/32 antibody (Clone 93; Biolegend; Cat #101320) on ice for 10 min, followed by LIVE/DEAD fixable dead cell staining (Thermo Fisher Scientific; Cat #L23105), and cell surface staining using perspective antibodies (see Antibodies and Reagents Section) on ice for 20 min. We then fixed the cells with fixation buffer (BioLegend; Cat #420801), permeabilized them with 1× intracellular staining perm wash buffer (Invitrogen; Cat #00-8333-56), and subjected them to intracellular staining using indicated antibodies (See Antibodies and Reagents Section). Samples were analyzed using BD FACS Canto II Flow Cytometer (Flow Cytometry Shared Facility, Duke University School of Medicine).

To evaluate the surface levels of MHC class I in VC and Vhl-KO MC38 cells, cells were detached by dissociation buffer (2% EDTA in 1× PBS), followed by blocking with TruStain FcX anti-mouse CD16/32 antibody (Clone 93; Biolegend; Cat #101320) on ice for 10 min. Cells were then stained with LIVE/DEAD fixable dead cell staining (Thermo Fisher Scientific; Cat #L23105) and PE anti-mouse H-2K^b/H2D^b antibody (Clone 28-8-6; Biolegend; Cat #114607) on ice for 20 min in the dark, washed twice with 1× FACs buffer (2% FBS in 1×PBS with 2mM EDTA), and fixed with 1% PFA at room temperature for 15 min in the dark. Samples were washed twice, resuspended in 1× FACs buffer, and analyzed using BD FACS Canto II Flow Cytometer (Flow Cytometry Shared Facility, Duke University School of Medicine).

TCR sequencing

We implanted control and Vhl-KO MC38 cells into mice as described above. At 13 days post-inoculation, we euthanized the mice and collected tumor tissues, and extracted genomic DNA (gDNA) using QIAGEN DNeasy Blood & Tissue kit according to manufacturer’s instruction (QIAGEN; Cat # 69506). We sent about 3 μg gDNA (60 ng/μL in AE buffer) to Adaptive Technologies (Seattle, WA) for ImmunoSEQ mouse TCR-β CDR3 survey sequencing. Upon data acquisition, we analyzed them using ImmunoSEQ Analyzer 3.0, an Adaptive Biotechnologies online analysis platform.

Subcellular fractionation of cellular extracts

We followed a previously published procedure for subcellular fractionation of cellular extracts.^44,83 Briefly, we seeded cells in 100 mm dishes two days before the experiment. On the day of the experiment, we divided 8∗10⁶ cells into two equal aliquots. We then suspended one aliquot in 500 μL of 50 mM NaOH, boiled the lysates for 30 min and used it to serve as total mtDNA control from whole cell lysate (WCL). Next, we resuspended the other aliquot in 500 μL of solution with 25 μg/mL of digitonin (Millipore Sigma; SKU #D141-100MG) in a buffer with 150mM of NaCl, 50mM HEPES at pH 7.4, and homogenized the cells by vigorous pipetting, and followed by a 10-min incubation period on ice to allow selective permeabilization of the plasma membrane. We then centrifuged the homogenates at 980 ×g three times for 3 min each at 4°C to pellet the intact cells. After the centrifugation, we rinsed the pellet with 1× PBS and used it as the pellet (Pel) fraction for immunoblot. Next, the supernatant was transferred to a new tube and centrifuged at 17,000 ×g for 10 min at 4°C to spin down the remaining cellular debris. The supernatant from this centrifugation was transferred to a new tube and used as the cytosolic (Cyto) fraction free of nuclear, mitochondrial, and endoplasmic reticulum contamination. We then boiled the WCL, Pel, and Cyto lysates at 95°C for 5 min and subjected them to western blot analysis to ensure the cytosolic preparation was free of contamination. Finally, we purified and concentrated both WCL_DNA and Cyto_DNA using DNA Clean & Concentrator-5 Kit (ZYMO RESEARCH; Cat #D4004) and used the DNA for quantitative PCR (qPCR) tests.

Depletion of cellular mtDNA

We depleted mtDNA using ethidium Bromide (EtBr) according to a previously published method.^84,85 We treated VC and VHL-KO cells with 100 ng/mL of EtBr for 21 days with routine medium change every 2–3 days. We then validated the depletion of mtDNA by staining cells with anti-dsDNA and anti-HSP60 antibodies before we harvested the cells for immunoblot and qPCR.

Total RNA extraction

We plated about 6 ∗10⁵ VC and VHL-KO Caki-1 cells in 60 mm dishes two days before total RNA extraction. We then extracted total RNA using Trizol Reagent (Ambion by life technologies; Cat #15596018) according to a protocol from StarrLab (https://sites.google.com/a/umn.edu/starrlab/protocols/rna/rna-isolation-using-trizol). Briefly, we first rinsed the cells twice with ice-cold 1× PBS. We then added 1mL Trizol to lyse the cells. Next, we scraped the lysate off from the Petri dish and transferred it to a 1.5 mL Eppendorf tube, incubated the lysate at room temperature for 5 min, and added 200 μL chloroform (Sigma; Cat #C2432) with vigorous vortexing for 15 s. We then incubated the mixture at room temperature for 10 min and centrifuged it at 12,000 ×g for 15 min at 4°C. Next, we transferred the transparent top layer to a clean 1.5 mL tube, adding 500 μL isopropanol and incubating at room temperature for 10 min. The mixture was then centrifuged at 12,000 ×g for 10 min at 4°C to get a white pellet at the bottom. After removing the supernatant, the pellet was washed with 75% ethanol and centrifuged at 7,500 ×g for 5 min at 4°C. The pellet was then allowed to air dry and dissolved in 85 μL RNase-free H₂O. Subsequently, we added 5 μL of TURBO DNase (Invitrogen by Thermo Fisher Scientific; Cat #AM2238) and 10 μL of 10× Turbo DNase buffer to degrade the remaining DNA at 37°C for 30 min. We then added 200 μL chloroform to stop the reaction. Afterward, we repeated the RNA extraction procedures as described above. Finally, we dissolved total RNA free of DNA contamination in RNase-free H₂O and used it for quantitative PCR and bulk RNA sequencing analysis.

Quantitative RT-PCR (qRT-PCR) and quantitative PCR

To quantify RNA expression levels, we used the Trizol-extracted total RNA (described above) as the template for cDNA synthesis using random hexamer primers (Invitrogen by Thermo Fisher Scientific; Cat #SO142) and SuperScript II Reverse Transcriptase (Invitrogen Thermo Fisher Scientific; Cat #18064014) following the manufacturer’s instructions. Afterward, we performed qRT-PCR of the cDNA using qPCRBIO SyGreen Blue Mix Hi-ROX (Genesee Scientific; Cat #17-506C) and the Applied Biosystems ViiA 7 Real-Time PCR System with 384-well Block (Thermo Fisher Scientific; Cat #4453536). We used the comparative Ct (ΔΔCt) method to compare the relative changes in gene expression among different genes.

To quantify cytosolic DNA levels, we obtained cytosolic extracts as described above. We then conducted qPCR analysis using the cytosolic extracts as templates. To quantify cytosolic mitochondria DNA (mtDNA) and nuclear DNA (nucDNA) levels for any individual gene, we set the ratio between the cytosolic DNA level and the whole cell lysate as in the control cells as 1 and used it to obtain the relative levels of cytosolic DNA in other cells.

Table S3 lists the primers used for q-RT-PCR and qPCR analysis of different target genes.

JC-1 mitochondrial membrane potential assay

We measured mitochondrial membrane potential (MtMP) using the JC-1(tetraethylbenzimidazolylcarbocyanine iodide) mitochondrial membrane potential assay kit (Abcam; Cat #113850) following the manufacturer’s instruction. JC-1 is a lipophilic cationic carbocyanine dye that accumulates in mitochondria. In healthy mitochondria, the mitochondrial membrane is less permeable. The high mitochondrial membrane potential that results from the proton pump causes the aggregation of JC-1, which shows a red to orange color under fluorescence. However, damaged mitochondria are associated with highly permeable and depolarized mitochondrial membrane and reduced MtMP. Under such conditions, JC-1 predominantly forms monomers and produces a green fluorescence. We used FCCP (2-[2-[4-(trifluoromethoxy)phenyl]hydrazinylidene]-propanedinitrile) as a positive control for membrane depolarization as it is an uncoupler of oxidative phosphorylation in the mitochondria.⁸⁶ We assessed the red fluorescence (excitation 535 nm)/emission 590 nm) and green fluorescence (excitation 475 nm/emission 530 nm) using immunofluorescence microscopy and flow cytometry. To quantify green/red fluorescence ratios in flow cytometry, the following formula was used: Ratio = [Red⁺Green⁺ (Q2, quadrant 2) + Red⁻Green⁺ (Q3)]/Red⁺Green⁻ (Q1).

Immunofluorescence microscopy

We seeded the cells in 35 mm glass-bottomed poly-D-lysine-coated dishes (MatTek Life Sciences; Cat #P35G-1.5-10-C) two days before experiments. We used 4% Paraformaldehyde (PFA) to fix cells at room temperature for 15 min and permeabilized the cells using 0.5% Triton X-100 in PBS at room temperature for 10 min. We then washed the cells three times with PBS and blocked them with 5% bovine serum albumin (BSA; Sigma; Cat #A3983) at room temperature for 1 h. Next, we added primary antibodies and incubated the cells at 4°C overnight, followed by adding fluorophore-conjugated secondary antibodies after washing with PBS three times. Next, we incubated the cells at room temperature for 1 h in the dark and washed them three times. Finally, we added VECTASHIELD Antifade Mounting Medium with DAPI (VECTOR LABORATORIES; Cat #H-1200-10) to the glass bottom of the dish before analysis. We took fluorescence images using the Leica TCS SP5 laser scanning confocal microscope in the Light Microscopy Core Facility of Duke University School of Medicine.

Bulk RNA sequencing

To perform genome-wide transcriptome analysis of VC and VHL-KO Caki-1 cells, we prepared total RNAs from the cells using Trizol Reagent as described above. We then submitted our RNA samples to the Duke Center for Genomic and Computational Biology for sequencing, which QC’ed the samples and prepared cDNA libraries for analysis using Illumina NovaSeq 6000. We processed the RNA-seq data using the TrimGalore toolkit, which employs Cutadapt to trim low-quality bases and Illumina sequencing adapters from the 3′ end of the reads.^87,88 Only reads that were 20nt or longer after trimming were kept for further analysis. Reads were mapped to the GRCh38.p13 of the human genome and transcriptome using the STAR RNA-seq alignment tool.^89,90 Reads were kept for subsequent analysis if they mapped to a single genomic location using the SAMtools.⁹¹ Gene counts were compiled using the HTSeq tool.⁹² Only genes that had at least 10 reads in any given library were used in subsequent analysis. Normalization and differential expression were carried out using the DESeq2 Bioconductor package with the R statistical programming environment. Software for gene set enrichment analysis (GSEA; version 4.1.0) was used to identify differentially regulated pathways. The source RNA-seq data are deposited in the NCBI’s Gene Expression Omnibus (GEO) database.

Analysis of published RNAseq data

To assess enriched pathways from VHL overexpression in 786-O cells, a previously published dataset (GSE108229) was used to perform gene ontology (GO) analysis.⁹³ Top-ranked pathways were plotted with ‘ggplot2’ R package. Scripts are available in the supplemental information.

Quantification and statistical analysis

Statistical analysis was conducted using GraphPad Prism 8.2.0 software. Two-sided Student’s t test was used for comparing two experimental groups. One-way ANOVA was applied to compare gene expression levels among multiple groups. two-way ANOVA was applied to compare in vivo tumor growth rates within two or more experimental groups. mtDNAs and nucDNAs levels in VC, VHL-KO, VHL/HIF1α DKO, VHL/HIF2α DKO, and VHL/HIF1α/HIF2α TKO Caki-1 cells were also analyzed using two-way ANOVA. Log rank (Mantel-Cox) test was used for mouse and human patient survival analysis. ∗p < 0.05 was considered statistically significant.

Published: June 15, 2024