“Selective Deficiency of Mitochondrial Respiratory Complex I Subunits Ndufs4/6 Causes Tumor Immunogenicity”

Abstract

It is appreciated that cancer cells rewire their metabolism to support their proliferation and evade immune surveillance. Here we provide evidence for a specific means of mitochondrial respiratory Complex I (CI) inhibition that improves tumor immunogenicity and sensitivity to immune checkpoint blockade (ICB). Targeted genetic deletion of either Ndufs4 or Ndufs6, but not other CI subunits, induces an immune-dependent growth attenuation in melanoma and breast cancer models. We show that deletion of Ndufs4 induces expression of the MHC class-I co-activator Nlrc5 and antigen presentation machinery components, most notably H2-K1. This induction of MHC-related genes is driven by a pyruvate dehydrogenase-dependent accumulation of mitochondrial acetyl-CoA, which leads to an increase in histone H3K27-acetylation within the Nlrc5 and H2-K1 promoters. Taken together, this work shows a novel modality by which selective CI inhibition restricts tumor growth, and that specific targeting of Ndufs4, or related CI subunits, increases T-cell surveillance and ICB responsiveness.

Introduction

The landscape of melanoma therapeutics has witnessed remarkable advancements over the past 15 years, transforming a previously untreatable disease into a more manageable one where affected patients have some factual hope of survival. Combinatorial oncogene-targeted BRAF- and MEK-inhibitor therapies have proven effective in patients harboring BRAF V600-missense mutations, although therapeutic resistance often emerges and curbs long-term clinical benefit^1,2. Additionally, immune checkpoint blockade (ICB) therapies such as anti-PD-1 and anti-CTLA-4 antibodies, have demonstrated remarkable efficacy in certain immunogenic cancers, including melanomas, and act by blocking inhibitory signals that limit the ability of CD8⁺ T-cells to surveil tumor cells^3–7. While some patients will see durable responses to ICB therapy, and a fraction will achieve cures, many patients will not meaningfully respond to ICB treatments due to poorly immunogenic tumors, thus highlighting the need to identify pathways and cellular components that could be leveraged to improve the immunogenicity of tumors⁸.

Targeting the mitochondria of tumor cells has been a promising area of therapeutic development because a large number of studies over the last decade have implicated mitochondria in cancer cell survival, proliferation, and metastasis. Since mitochondria are a biosynthetic and bioenergetic organelle, and many cancer cells generally proliferate quickly and have high energy demands, cancers often substantially rely on mitochondria to sustain growth and survival^9,10. In addition, several studies have found that metastatic melanoma and breast tumors in a variety of animal models have increased oxidative phosphorylation (OxPhos) signatures^11–13, suggesting that targeting OxPhos in these cancers may be of therapeutic benefit. OxPhos is driven by a series of five protein complexes made of the four electron transport chain complexes and the F-type proton ATPase, ATP-synthase, which are comprised of over 100 individual proteins that drive ATP and electron equivalent synthesis in mitochondria¹⁴. Due to the prominent role of OxPhos and mitochondria in tumor growth and metastasis, significant efforts have been made to develop OxPhos inhibitors, especially against CI, the NADH dehydrogenase complex. CI inhibitors have shown promising preclinical results across multiple cancer types^15–19. However, once tested in phase I clinical trials, all inhibitors have so far been found to display early dose-limiting and adverse toxicities^20,21. The success seen in preclinical models suggests that mitochondria, and in particular CI, are promising targets; however, different approaches to inhibit their activity need to be identified to increase the therapeutic window of this class of drugs.

In this work we adopted a granular methodology to selective targeting of CI subunits. Using a genetic approach, we used CRISPR/Cas9 to systematically target many nuclear encoded subunits in the different domains of CI to examine its role in mediating an anti-tumor response in a mouse melanoma model. We identify the CI subunits, Ndufs4 and Ndufs6, which when deleted using CRISPR/Cas9 technology, induces an immune-dependent growth attenuation of murine tumor cell models. Because previous attempts to inhibit CI focused on cell intrinsic growth inhibition, we prioritized understanding this cell extrinsic growth inhibition to introduce a different CI inhibition modality. Using this model, we show that selective CI inhibition through Ndufs4 deletion can induce expression of MHC class-I (MHC-I) antigen presentation and processing machinery through mitochondrial produced acetyl-CoA, drive CD8⁺ and NK1.1⁺ cell responses to tumors and sensitize tumors to murine anti-PD-1 ICB. These studies suggest that selectively affecting CI structure, rather than broadly interfering with CI catalytic activity, can promote anti-tumor immunity, providing support for further research into mitochondrial control of tumor immunogenicity.

Results

Deletion Ndufs4 and Ndufs6 causes an immune-dependent tumor growth attenuation

To study the effects of CI deficiency on in vivo tumor growth and immunogenicity, we took a genetic approach using CRISPR/Cas9 technology to knock-down several CI subunits from the three CI modules in B16-BL6 mouse melanoma cells. We then injected these cell lines into immune competent C57BL/6 and immunodeficient Fox1^nu (nude) mice, which lack functional T-cells, to determine which genetic perturbations caused tumor growth attenuations and whether this attenuation was dependent on the adaptive immune system.

For the assessed Q-module or P-module subunits within CI (Ndufs3, Ndufa9, Ndufa1 and Ndufb8), which are responsible for transferring electrons to ubiquinone and pumping protons respectively²², deletion of these subunits had similar effects on tumor growth in both C57BL/6 and nude mice. In the Q-module sgNdufs3 cells had no tumor growth phenotype while sgNdufa9 cells exhibited mild tumor growth attenuation in both C57BL/6 and nude mice (Fig. 1a-c, Extended Data Fig. 1e). For the P-module, sgNdufa1 and sgNdufb8 exhibited similar growth defects in both C57BL/6 and nude mice (Fig.1d-f, Extended Data Fig. 1e). These data pointed to an immune independent mechanism of tumor growth attenuation that is likely caused by disrupted bioenergetics and/or redox balance in these cells. This is further supported by blue-native polyacrylamide gel electrophoresis (BN-PAGE) analysis of these knockout cell lines that demonstrated a lack of CI activity and a lack of cell growth in galactose culture media (Extended Data Fig. 1a,d) and is in keeping with previous structural studies of CI²³.

Fig. 1:

Deletion of mitochondrial CI components leads to cell intrinsic and cell extrinsic tumor growth attenuation in B16-BL6 melanoma tumors. a, Structure of murine CI with the Q-module highlighted (PDB: 6G2J), matrix refers to mitochondrial matrix while IMS refers to the inner membrane space⁶⁸. b, Tumor growth curves of sgNT and sgNdufs3 tumors in C57BL/6 and nude mice. c, Tumor growth curves of sgNT and sgNdufa9 tumors in C57BL/6 and nude mice. d, Structure of murine CI with the P-module highlighted (PDB: 6G2J), matrix refers to mitochondrial matrix while IMS refers to the inner membrane space. e, Tumor growth curves of sgNT and sgNdufa1 tumors in C57BL/6 and nude mice. f, Tumor growth curves of sgNT and sgNdufb8 tumors in C57BL/6 and nude mice. g, Structure of murine CI with the N-module highlighted (PDB: 6G2J), matrix refers to mitochondrial matrix while IMS refers to the inner membrane space. h, Tumor growth curves of sgNT and sgNdufa12 tumors in C57BL/6 and nude mice. i, Tumor growth curves of sgNT and sgNdufv2 tumors in C57BL/6 and nude mice. j, Tumor growth curves of sgNT and sgNdufv1 tumors in C57BL/6 and nude mice. k, Tumor growth curves of sgNT and sgNdufs6 tumors in C57BL/6 and nude mice. l, Tumor growth curves of sgNT and sgNdufs4 tumors in C57BL/6 and nude mice. sgNdufs4-1 and sgNdufs4-2 denote different CRISPR guides against Ndufs4. P1 compares sgNT to sgNdufs4-1 and P2 compares sgNT to sgNdufs4-2. n= 5 mice per condition in b-c, e-f, and h-I, N≥2. P-values in b-c, e-f, and h-k were calculated using a Mann Whitney u-test, p-values in l were calculated using the Brown-Forsythe and Welch ANOVA test with Dunnett’s T3 multiple comparisons test. Error bars are mean ± s.e.m.

Most subunits in the catalytic NADH dehydrogenase N-module followed the same trend as the other modules. For example, the partial knockdown in sgNdufa12 cells have no growth phenotype in C57BL/6 or nude mice, while sgNdufv1 and sgNdufv2 cells had attenuated tumor growth in both C57BL/6 and nude mice (Fig. 1g-j, Extended Data Fig. 1e). Similar to the other modules, there was a strong trend for subunits that caused growth defects to display a lack of CI activity and sensitivity to galactose (Extended Data Fig. 1a,d). A notable exception was found following deletion of the Ndufs4 subunit which exhibited a stark tumor growth attenuation in C57BL/6 but not nude mice (Fig. 1l, Extended Data Fig. 1e). Similar data was also observed upon deletion of Ndufs6 (Fig. 1k). This unique phenotype indicated to us that there was a cell extrinsic, immune-dependent, tumor growth attenuation in these CI subunit deletion melanoma models.

To better understand what was unique about deletion of Ndufs4, we assayed the assembly and activity of CI. In keeping with previously published work²³, knocking out Ndufs4 resulted in partial, but not complete disassembly of CI. As read out by the CI subunit Ndufa9, OxPhos supercomplexes (co-assemblies of different stoichiometries of complex I, III and IV) in sgNdufs4 cells migrated slightly faster than those in sgNT or sgNdufa12 cells on BN-PAGE, which have previously been shown to not have an assembly defect, indicating a partial disassembly of the complex (Extended Data Fig. 1b)²³. Using a BN-PAGE in-gel activity assay, CI enzymatic activity could still be detected in sgNdufs4 cells, albeit at a lower molecular weight (Extended Data Fig. 1a,c). This indicates that the N-module is assembled in a much smaller complex, possibly migrating alone, and shows partial functionality. Additionally, when assaying CI driven respiration measure as oxygen consumption rate (Seahorse) assays and complex I + complex III activities in vitro as described²⁴, we found that sgNdufs4 and sgNdufs6 exhibited partial CI activity while sgNdufs3 showed almost none (Extended Data Fig. 1f-g). This partial functionality of CI sets sgNdufs4 and sgNdufs6 apart and likely underlies the normal tumor growth in nude mice and the ability to grow in galactose culture media (Fig. 1l, Extended Data Fig. 1d).

Ndufs4 deletion induces an anti-tumor response in multiple tumor models

To determine whether the immune-mediated anti-tumor effects of Ndufs4 deletion in tumors was more broadly applicable, we performed CRISPR/Cas9 mediated deletion of Ndufs4 in an addition murine melanoma cell line, YUMM1.7 and the murine breast cancer line EO771. Both cell lines displayed similar tumor growth phenotypes as B16-BL6, where Ndufs4 deletion caused tumor growth attenuation in immunocompetent C57BL/6 mice but not immunodeficient nude mice (Fig. 2a-b). In addition to the marked attenuation of primary subcutaneous tumor growth, strikingly sgNdufs4 cells exhibited almost undetectable signs of lung metastasis when injected into immunocompetent C57BL/6 mice via tail vein compared to controls. However, in nude mice, tail vein injection of sgNT and sgNdufs4 B16-BL6 cells equally formed extensive lung melanoma metastasis and growth with similar survival rates (Fig. 2c-d).

Fig. 2:

CRISPR/Cas9 mediated knockout of Ndufs4 causes primary tumor growth defects in multiple cell line and attenuated metastasis in B16-BL6 cells. a, Tumor growth curves of YUMM1.7 sgNT and sgNdufs4 tumors in C57BL/6 and nude mice and western blot of Ndufs4 expression. b, Tumor growth curves of E0771 sgNT and sgNdufs4 tumors in C57BL/6 and nude mice and western blot of Ndufs4 expression. c, Kaplan-Meier plot of C57BL/6 mice injected via tail vein with sgNT or sgNdfus4 B16-BL6 cells. d, Kaplan-Meier plot of nude mice injected via tail vein with sgNT or sgNdufs4 B16-BL6 cells. n=5 mice per condition in a-b, n=10 mice per condition in c-d N≥2. Representative pictures of lungs are shown in c-d. P-values in a-b were calculated using a Mann Whitney u-test and p-values in c-d using a Mantel-Cox log-rank test. Error bars are mean ± s.e.m.

To ensure that the observed effects on tumor growth were due to the loss of Ndufs4 and not an off-target effect of the CRISPR guides, we overexpressed Ndufs4 (or a control plasmid, V300) in sgNT and sgNdufs4 B16-BL6 cells. We observed a marked rescue in the tumor growth of sgNdufs4 cells re-expressing Ndufs4, while overexpression of Ndufs4 had no effect on the tumor growth of sgNT cells (Extended Data Fig. 2a). While the rescue of tumor growth was not complete, expression of Ndufs4 in the re-expression cell line was lower than in sgNT cells, and we could not detect a statistically significant difference between the tumor growth of the sgNT and the Ndufs4 re-expressing cells.

Mutations in CI in humans and mouse models have been shown to cause inflammation and cell death^25,26, with some of these effects having been attributed to altered NAD⁺/NADH ratios resulting from CI deficiencies²⁵. To test whether the observed immunogenic effects of sgNdufs4 cells could be caused by cellular redox state changes, we overexpressed a NADH oxidase from Lactobacillus brevis, LbNOX, which we have previously shown to rescue cell death in mitochondrial disease models²⁵. Overexpression of the cytosolic or mitochondrial targeted versions of the LbNOX protein did not alter the growth of sgNT or sgNdufs4 tumors; however, overexpression of the mitochondrial LbNOX did restore wild-type NAD⁺/NADH ratios in sgNdufs4 cells (Extended Data Fig. 2b-d). Based on these data, we conclude that the observed sgNdufs4 immunogenicity was not a consequence of NADH-dependent redox imbalance but rather a different cellular consequence of attenuated CI function.

sgNdufs4 tumors display increased expression of MHC-I genes

To identify cellular components and processes that cause immunogenicity of sgNdufs4 cells, we performed quantitative TMT-proteomics on tumors from C57BL/6 and nude mice and bulk RNA-seq from cells grown in vitro (Extended Data Fig. 3a-b). We performed an integrated analysis of these datasets, seeking to find proteins and their encoding genes that were consistently differentially regulated across the proteome and transcriptome data (Fig. 3a). We found that there were 10 upregulated and 29 downregulated proteins (log2FoldChange ≥ 1.0, p<0.05) and genes (log2FoldChange ≥ 0.7, p<0.05) across the three datasets (Fig. 3b). Enrichr/GSEA (gene set enrichment analysis) revealed that the upregulated proteins and genes (as human orthologs/paralogs) were associated with MHC-I antigen presentation and Cancer Immune Escape, while the downregulated proteins and genes were associated with arachidonic acid metabolism and lipogenesis (Fig. 3c-d). Given the extensive literature linking MHC-I expression with tumor immunogenicity, as well as response to immune checkpoint blockage treatments ^27,28, we chose to further pursue this pathway. The signature components driving the pathway enrichment were Nlrc5, the transcriptional co-activator that regulates MHC-I antigen presentation machinery²⁹, and the MHC-I allele H2-K1³⁰ (using either of the human paralogs HLA-A, HLA-B, or HLA-C for the Enrichr/GSEA analysis). As validation of our assumption that this pathway was indeed relevant, we used the TCGA-SKCM dataset for stage IV disease and could see a trend for the upregulated signature for predicting overall survival while the downregulated gene set did not (Extended Data Fig. 3c). Using a combined rank-based metric of NLRC5 and HLA-B, the MHC-I molecule that is the closest paralog by genomic location to mouse H2-K1, we found a trend towards association with survival (Mantel-Cox log rank test, p = 0.062) (Extended Data Fig. 3c), while NLRC5 expression levels alone significantly stratifies survival (Fig. 3e). These analyses show similar results to what has been shown in the literature regarding response to immune checkpoint treatment³¹. Furthermore, using a literature based gene signature MHC-I antigen processing and presentation pathway³², the bulk RNA-seq and proteomics from tumors in nude mice show a consistent upregulation across all steps of the antigen presentation pathway (Extended Data Fig. 3d). While we could see similar changes reflected in the C57BL/6 proteomics, there is more variability, possibly due to immune pressures to downregulate these proteins and attenuation of the signal from increased immune cell presence. These components, including NLRC5, the MHC-I molecules HLA-A and HLA-B, and the antigen processing machinery genes TAP1, PSMB8, PSMB9 and PSMB10 correlate with melanoma immunotherapy responsiveness in human patients (Extended Data Fig. 3e). We also used qPCR to validate that these components where upregulated in B16-BL6 and YUMM1.7 tumors grown in C57BL/6 mice (Extended Data Fig. 4a,b). Finally, we validated that expression of Nlrc5 and its target genes H2-K1, Psmb8, Psmb9, Psmb10 and Tap1 were upregulated in sgNdufs4 cells grown in culture to ensure that the upregulation we were seeing was not caused by cytokines in the tumor microenvironment in vivo (Fig. 3f). These data in aggregate suggest that deletion of Ndufs4 leads to upregulation of the MHC-I co-activator Nlrc5 and downstream regulated components of the antigen processing and presentation machinery, specifically H2-K1, in a cell intrinsic manner.

Fig. 3:

MHC-I and antigen processing and presentation components are increased in sgNdufs4 tumors and cells in a cell intrinsic manner. a, Bioinformatics strategy to identify differentially regulated genes/proteins in sgNdufs4 cells. b, Plot of log2FC of the proteomics performed in sgNT and sgNdufs4 tumors from nude and C57BL/6 mice colored by log2FC of the RNA-seq done with sgNT and sgNdufs4 cells in vitro. Labeled points are differentially expressed genes that meet the criteria set out in a. c, Significantly downregulated Elsevier Pathways Collection terms from proteins/gene that were downregulated in sgNdufs4 cells and tumors, FDR q-values < 0.05. d, Significantly upregulated Elsevier Pathways Collection terms from proteins/gene that were upregulated in sgNdufs4 cells and tumors, FDR q-values < 0.05. e, TCGA-SKCM stave IV survival stratification based on NLRC5 expression (n=68). f, qPCR mRNA levels of MHC-I and antigen processing genes of sgNT and sgNdufs4 cells cultured in vitro, N=3. P-values in f were calculated using a two-tailed, unpaired Student’s t-test. Error bars are mean ± s.e.m.

Nlrc5 and target gene expression is driven by altered mitochondrial metabolism

To better understand the mechanism by which Ndufs4 deletion drives increased Nlrc5 expression, we first tested several pathways that had previously been shown to affect Nlrc5 levels. First, Nlrc5 expression has been found to be regulated by DNA CpG methylation³³ and to this end we performed bisulfite conversion followed by sequencing of the Nlrc5 promoter region in sgNT and sgNdufs4 cells in vitro. We found that both sgNT and sgNdufs4 cells had presence of methylated CpG’s in the Nlrc5 promoter region, but there were no significant differences in the methylation patterns (Extended Data Fig. 5a). Second, given that NF-κB activation has been shown to induce Nlrc5 expression³⁴, we therefore assessed whether other downstream targets of NF-κB where likewise affected. A primary activator of NF-κB, especially in the case of mitochondrial disruption, is thought to be the cytosolic DNA sensing pathway cGAS-STING. At baseline, the cGAS-STING downstream target TBK1 was not found to be phosphorylated in either sgNT or sgNdufs4 cells (Extended Data Fig. 5b). Additionally, neither STING, nor the downstream transcription factor STAT1, were found to be necessary for the observed tumor immunogenicity (Extended Data Fig. 5c,d). Lastly, NF-κB protein levels were not shown to be changing in sgNdufs4 tumors from C57BL/6 or nude mice (Extended Data Fig. 3d).

We next searched for alternative mechanisms by which Ndufs4 deletion could induce Nlrc5 upregulation. Because many mitochondrial derived metabolites such as acetyl-CoA, NADH/NAD+, succinate and fumarate have previously been implicated in retrograde signaling from the mitochondria to the nucleus^35–37, we performed metabolomics on crude mitochondrial fractions from sgNT, sgNdufs3, sgNdufs4, sgNdufs6 melanoma cells to compare the mitochondrial metabolomes of immunogenic and non-immunogenic CI mutants (Fig. 4a). We found an accumulation of acetyl-CoA and S-adenosyl-L-methioninamine as well as a decrease in several tricarboxylic acid cycle (TCA) metabolites and amino acids in sgNdfus4 and sgNdfus6 but not sgNdufs3 mitochondria (Fig. 4b, Extended Data Fig. 6a-c), indicating these metabolites could be important in mediating the observed immunogenicity. Given the known role for acetyl-CoA as an epigenetic substrate, we confirmed the elevated acetyl-CoA levels in sgNdufs4 and sgNdufs6 in cultured cells, as well as sgNdufs4 tumors in vivo using a fluorometric acetyl-CoA detection kit (Fig. 4c), further supporting the differential regulation of acetyl-CoA in sgNdfus4 and sgNdufs6 cells.

Fig. 4:

Pdha derived acetyl-CoA is accumulated in immunogenic CI mutants and drives Nlrc5 and H2-K1 promoter H3K27ac and tumor immunogenicity. a, Metabolomic strategy for identifying mitochondrial metabolites responsible for observed gene expression changes. b, Volcano plots of mitochondrial enriched metabolomics from sgNdufs3, sgNdufs4 and sgNdufs6 cell lines (N=3). c, Acetyl-CoA levels in sgNT, sgNdufs3, sgNdufs4 and sgNdufs4+sgPdha in cells and sgNT and sgNdufs4 tumors in vivo measured using a fluorometric kit (N=5). d, Tumor growth curves of sgNdufs4 cells transfected with either sgNT or sgPdha CRISPR guides in C57BL/6 and nude mice (n=5). e, ChIP qPCR H3K27ac on the Nlrc5 and H2-K1 promoters in sgNT and sgNdufs4 cells (N=3). f, ChIP qPCR H3K27ac on the Nlrc5 and H2-K1 promoters in sgNdufs4 + sgNT and sgNdufs4 + sgPdha cells (N=3). P-values in c were calculated using one-way ANOVA followed by Tukey’s post-hoc test, while p-values in d were calculated using a Mann Whitney u-test.

To further test the hypothesis that mitochondrial metabolism could cause the observed immunogenicity of the sgNdufs4 tumors, we interrogated the role of acetyl-CoA in this system. Acetyl-CoA can be derived from multiple sources in mitochondria: pyruvate³⁸, acetyl-carnitine³⁹, and branched chain amino acid (BCAA) catabolism⁴⁰. However, because we do not observe accumulations in the intermediates of acetyl-carnitine catabolism, such as carnitine³⁹, or the BCAA catabolism, such as succinyl-CoA, propionyl-CoA, or acetoacetate⁴⁰, we hypothesized that pyruvate-derived acetyl-CoA was driving the phenotypic changes in sgNdufs4 cells. To this end, deletion of pyruvate dehydrogenase (Pdha), the mitochondrial enzyme that converts pyruvate to acetyl-CoA, rescues acetyl-CoA accumulation in sgNdufs4 cells, supporting our hypothesis (Fig. 4c). To test whether mitochondria generated acetyl-CoA could induce the observed gene expression changes, we inhibited acetyl-CoA formation from pyruvate through chemical inhibitors and genetic perturbations. First, treating cells in vitro with devimistat, an inhibitor of pyruvate dehydrogenase (PDH), potently reduced the expression of Nlrc5 and its target genes in both B16-BL6 and EO771 sgNdfus4 cells but had little effect on sgNT cells (Extended Data Fig. 7a-b). Second, CRISPR/Cas9 mediated deletion of Pdha, the gene encoding pyruvate dehydrogenase, partly rescued sgNdufs4 tumor growth in immune competent C57BL/6 mice but had no effect on tumor growth in nude mice (Fig. 4e, Extended Data Fig. 6e). Finally, deletion of Pdha also greatly reduced expression of Nlrc5, MHC-I antigen processing machinery, and immune activation markers like Gzmb and Ifng in the tumors (Extended Data Fig. 7c). These data suggest that accumulation of acetyl-CoA in the mitochondria drives these gene expression changes we observed in sgNdufs4 cells.

Because mitochondrial derived acetyl-CoA is a substrate for histone acetylation that affects gene expression⁴¹, we assessed whether histone acetylation was changing globally in sgNdfus4 cells. Using whole cell Western blot assays, we found that there were negligible differences in histone acetylation marks such as H3K27ac, H3K9ac, and H4K5ac (Extended Data Fig. 6d). However, ChIP q-PCR for H3K27ac at the Nlrc5 and H2-K1 promoters showed that sgNdufs4 cells had substantially increased H3K27ac (Fig. 4e). Importantly, Pdha deletion in the sgNdufs4 cells reduced the elevated H3K27ac acetylation levels to similar levels as sgNT cells (Fig. 4f). While the mechanism behind the selective increase in H3K27ac remains unknown, mitochondrial Pdha derived acetyl-CoA seems to drive MHC-I antigen presentation and processing machinery expression through alterations in H3K27ac marks.

Ndufs4 deletion leads to effector cell activation and improved anti-PD-1 responsiveness

To characterize what cells are involved in mounting an immune response against sgNdufs4 tumors, we performed flow cytometry analysis on the associated CD45⁺ cells within the tumor immune microenvironment (TIME). While there was no change in the total number of CD8⁺ or CD4⁺ T-cells (Fig. 5a, Extended Data Fig. 8a), there were significantly more CD44⁺CD69⁺ double positive and PD-1⁺TIM3⁺ double positive CD8⁺ T-cells (Fig. 5a, Extended Data Fig. 8a), indicating that the T-cells were at as an increased fraction activated and consistent with elevated MHC-I allele expression in sgNdufs4 tumor cells.

Fig. 5:

The immune response against sgNdufs4 tumors is dependent on both CD8b⁺ and NK1.1⁺ cells and synergizes with PD-1 therapy. a, Quantification of FACS based analysis of CD4⁺ and CD8⁺ T-cell infiltration and CD8⁺ T-cell activation in sgNT and sgNdufs4 tumors grown in C57BL/6 mice. N = 8 mice per condition, representative plots shown of two independent experiments. b, Quantification of FACS based analysis of NK cell infiltration and activation in sgNT and sgNdufs4 tumors grown in C57BL/6 mice. n = 8 mice per condition, representative plots shown of two independent experiments. c, Tumor growth curves of sgNT or sgNdufs4 tumor grown in C57BL/6 mice grown treated with either control IgG or anti-NK1.1 antibodies. P1 compares sgNT+anti-NK1.1 to sgNdufs4+anti-NK1.1 and p2 compares sgNT+IgG to sgNdufs4+IgG. d, Tumor growth curves of sgNdufs4 tumors grown in C57BL/6 mice treated either with control IgG, anti-CD8β, anti-NK1.1 or both anti-CD8β and anti-NK1.1 antibodies. P1 compares sgNdufs4+IgG to anti-CD8β+NK1.1, p2 compares sgNdufs4+IgG to anti-NK1.1, p3 compares sgNdufs4+IgG to anti-CD8b. e, Tumor growth curves of sgNdufs4 tumors transfected with H2-K1 or non-targeting CRISPR guides in C57BL/6 and nude mice, n=5 mice per condition. f, Tumor growth curves of sgNT tumors in C57BL/6 mice treated either with a control IgG or anti-PD-1, n=5 mice per condition. g, Tumor growth curves and corresponding survival plot of C57BL/6 mice bearing sgNdufs4 tumors and treated either with a control IgG or anti-PD-1 antibodies, n=5 mice per condition. P-values in a, b, e-g, were calculated using a Mann Whitney u-test, p-values in c and d were calculated using one-way ANOVA followed by Tukey’s post-hoc test. Error bars are mean ± s.e.m.

Akin to the T-cell phenotype, we found that there was no significant difference in the number of NK cells in the sgNdufs4 tumors (Fig. 5b, Extended Data Fig. 8b); however, tumor resident NK cells displayed increased levels of the activation marker CD69 (Fig. 5b, Extended Data Fig. 8b). Additionally, by qPCR, we found a significant increase in the mRNA levels of immune related genes such as Ifng, Prf1, and Gzmb in sgNdufs4 B16-BL6 and YUMM1.7 tumors from C57BL/6 mice (Extended Data Fig. 8c,d), further supporting a model of immune activation in the sgNdufs4 tumors. Taken together, these results support a model of an inflamed TIME that is leading to the enhanced activation of different immune cell populations and increased killing of sgNdufs4 tumor cells.

To test whether NK and/or CD8⁺ were required for the sgNdufs4 immune surveillance, we performed immunodepletion of these two cell types in C57BL/6 mice using anti-NK1.1 and anti-CD8β antibodies, respectively. Using single antibody treatments, we observed similar tumor growth rescue using each of anti-NK1.1 and anti-CD8β antibodies (Fig. 5c,d). Additionally, the combination of anti-NK1.1 and anti-CD8β antibodies displayed improved effectiveness, indicating that NK and CD8⁺ cells each contribute to surveillance of the sgNdufs4 tumors (Fig. 5d). Finally, when we delete H2-K1 in sgNdufs4 cells we observed a robust tumor growth rescue akin to anti-NK1.1 and anti-CD8β depletions, which suggests that a specific increase in H2-K1 protein levels causes tumor antigenicity and immune surveillance downstream of Ndufs4 deletion (Fig. 5e).

Because of the increased T-cell activation and MHC-I expression in sgNdufs4 tumors, we investigated whether sgNdufs4 tumors would be sensitive to murine ICB therapy. As expected, sgNT B16-BL6 tumors showed no response to anti-PD-1 antibody treatment (Fig. 5f), which is in line with previous work ⁴². However, mice bearing B16-BL6 sgNdufs4 tumors lived longer and had substantially smaller tumors on average when treated with anti-PD-1 therapy (Fig. 5g). Taken together these results indicates that a co-therapeutic modality using selective CI inhibition with ICB therapeutic improves ICB responses through increased tumor antigenicity.

Discussion

Together, this work has identified a previously uncharacterized modality of mitochondrial CI inhibition in tumors that leads to an immune-dependent attenuation of tumor growth. Our current model is that partial CI inhibition leads to accumulation of Pdh dependent acetyl-CoA, which in turn drive Nlrc5 and MHC-I antigen presentation, thus potently activating CD8⁺ T-cells and creating an inflamed tumor environment that synergizes with anti-PD-1 ICB (Fig. 6). This finding has potential implications in the field of immunotherapy. While many cancer patients, especially with those with baseline inflamed (immunologically hot) melanomas, respond to current generation ICB therapies, a large proportion of patients still have a limited or no response to current treatment modalities⁴³. Therefore, development of new therapies that could boost the effectiveness of current ICB therapies should be investigated for patient benefit and may be found within inhibitors that target specific components in the mitochondria, as illustrated herein. To this end, the highlighted mechanism of action of CI inhibitors as cancer therapeutics has previously focused on cell intrinsic apoptosis⁴⁴, this previously undescribed link between selective CI inhibition and induction of tumor immunogenicity opens up a new potential for co-therapeutic opportunities. Development of small molecules that prevents the binding of NDUFS4 to CI or low doses of CI inhibitors in combination with pyruvate dehydrogenase kinase inhibitor, an inhibitory kinase of pyruvate dehydrogenase, to increase mitochondrial acetyl-CoA could prove effective in increasing immunotherapy sensitivity.

Fig. 6:

The model of Ndufs4 mediated tumor immunogenicity. Partial inhibition of complex I through non-essential subunit deletion leads to increased mitochondrial acetyl-CoA that leads to the increased acetylation of Nlrc5 and H2-K1 promoters and increased expression of antigen presentation pathways. This increased antigen presentation pathway then leads to increased immune cell activation and killing.

The discovery of previously unknown mechanisms to regulate MHC-I expression in tumors is also of great interest to the field of immunotherapy. In order for ICB to be successful, patients need to be able to develop and sustain a T-cell based immune response against their cancer prior to ICB, as ICB can only stimulate a preexisting T-cell response⁴⁵. Therefore, the downregulation of MHC-I molecules by tumor cells to evade T-cell surveillance has been linked to both intrinsic and acquired ICB resistance in multiple systems^46,47. Thus, the development of therapies that can induce MHC-I molecules and antigen processing machinery is an active area of research. We present a mechanism by which selective and partial CI inhibition can induce MHC-I antigen presentation and processing machinery through mitochondrial nuclear metabolic signaling.

Mitochondria have long been seen as a metabolic signaling cellular organelle for bioenergetics, redox status, metabolism and inflammation^48–51. Here we show that alterations in mitochondrial -related metabolism can lead to changes in immunogenic-dependent tumor growth inhibition. However, unlike most previously published examples such as succinate, fumarate, aconitate and alpha-ketoglutarate⁵², the effect of the metabolite changes seems to be centered on the tumor cells versus immune cells. While mitochondria have been widely implicated in inflammatory signaling, less work has demonstrated the ability of mitochondrial metabolites to induce a pro-inflammatory tumor environment. Previous works have focused much more on mitochondria as a signaling platform for MAVS/RIG-1 signaling for RNA recognition, mtDNA recognition by cGAS-STING leading to NF-κB activation, or how mitochondrial dysfunction can activate the inflammasome⁴⁹. A recent publication has shown that inhibition of mitochondrial CII can promote MHC-I antigen processing transcription through a succinate driven epigenetic mechanism⁵³. However, they claimed that CI had no such role in regulating antigen presentation pathways. Here we show that selective CI inhibition can also promote similar transcriptional changes through a distinct mechanism centered upon accumulation of acetyl-CoA. Additionally, others have found that mitochondrial DNA mutations can induce sensitive to immunotherapies by shifting tumor metabolism towards glycolysis, which in turn remodels the tumor associated neutrophil populations, inducing tumor immunogenicity⁵⁴. To our knowledge these are the first examples of how altering mitochondrial electron transfer chain metabolism can drive anti-tumor immunity in a cell intrinsic manner in tumor cells. Taken together these works establish mitochondrial metabolism as a potential therapeutic target in cancers with low MHC-I expression.

While we have demonstrated that partial inhibition of CI leads to the accumulation of acetyl-CoA that seems to drive epigenetic changes in tumor cells that increase immune recognition, the exact mechanisms governing this phenomenon are still not understood. More investigation into how CI activity regulates TCA flux, acetyl-CoA levels, and how mitochondrial acetyl-CoA fluctuations regulate the epigenome is warranted. Because the observed increases in promoter acetylation are not global (Extended Data Fig. 6d), we hypothesize that there are additional signaling mechanisms involved in this mitochondrial-nuclear communication that have yet to be uncovered. Additionally, the mechanism of NK cell activation in this system remains unclear. Because nude mice have functional NK cells⁵⁵ and we see no anti-tumor activity towards sgNdufs4 and sgNdufs6 tumors, we hypothesize that the observed NK cell activation is T-cell dependent. However, the type of crosstalk that is occurring, whether through IFN-γ secretion by CD4⁺ and CD8⁺ cells^56,57, IL-2 secretion of CD4⁺ T-cells^58–60, IL-12 secretion by APCs⁶⁰, or some combination therein, remains unknown.

Methods

Further information on research design is available in the Nature Research Reporting Summary linked to this article. Research conducted in this manuscript complies with all relevant ethical regulations at Dana-Farber Cancer Institute and Harvard Medical school, and all animal experiments were performed according to Institutional Animal Care and Use Committee-approved protocols at Beth Israel Deaconess Medical Center.

Animals and cancer cell lines

Homozygous outbred nude mice (Jackson Lab: Foxn1 nu /Foxn1 nu #007850) and C57BL/6 (Jackson Lab: #000664) female mice were used for xenograft tumor experiments under the auspice of Beth Israel Deaconess Medical Center animal facility and IACUC approved protocols. Mice were between 8–10 weeks old and sex was not considered in experimental design. Mice were housed at 23̊C, 30% HR, and 12/12 h light/dark cycles with free access to chow diet food (Formulab Diet #5008) and water. YUMM1.7 was a kind gift from Marcus Bosenberg (Yale University School of Medicine). B16-BL6 was a kind gift from James Allison (MD Anderson Cancer Center). E0771 and HEK293T cells were purchased from ATCC (cat #CRL-3461, # CRL-3216) where they were kindly donated by Robin Anderson (Olivia Newton-John Cancer Research Institute). All cell lines were maintained in a humidified incubator at 37°C with 5% CO2, and if not otherwise indicated, in DMEM (Sigma-Aldrich) with 10% FBS, 100 U/ml penicillin, and 100 mg/ml streptomycin.

Western blot assays

For Western immunoblotting, cells were lysed in RIPA buffer at indicated time points, and the protein concentration was quantified using the DC protein concentration assay (Pierce) before being subjected to gradient 4–12% (30:1) SDS polyacrylamide electrophoresis and subsequently transferred to PVDF membranes. Specifically, MES SDS running buffer was used in gels to detect proteins whose molecular weight are below 20 kD, while MOPS SDS running buffer was used for all other Western blot assays. Membranes were blocked with 2% BSA in TBST, then probed with primary antibodies overnight, and subsequently using secondary antibodies for 1 h at room temperature. All antibodies used for western blots were diluted 1:1000.

Blue Native PAGE and CI in-gel activity assays

Cells at 80% confluent from around 4 15-cm-dishes were suspended in the ice-cold mitochondrial isolation buffer (0.32 M sucrose, 0.001M EDTA, 0.01 M Tris) after 3 times of wash with PBS. The suspension was then dounced 20 times with a Dounce homogenizer and then centrifuged at 1000g for 5 minutes at 4 °C. The supernatant was collected into a new falcon tube and then centrifuged at 1000g for 5 minutes again at 4 °C. The supernatant was then centrifuged at 9000 x g for 15 minutes at 4 °C. The mitochondrial pellet was resuspended into 1 ml mitochondrial isolation buffer⁶¹ and the protein concentration was measured with DC protein assay. For blue native assay, 200 ug mitochondria were suspended into an 80 µl buffer (20µl 4 x NativePAGE Sample Buffer, 16µl 10% digitonin, 44µl H2O) and kept on ice for 30 minutes. The solution was then centrifuged for 10 min at 12000 x g and the supernatant was collected and mixed with 8µl 5% Coomassie G-250. 15 µl of samples in loading buffer were loaded into the 3–12% and run with Native-PAGE gel. 1 x Native-PAGE running buffer and 1 x Native-PAGE cathode buffer. After 1 hour running, 1 x Native-PAGE cathode buffer was changed to 0.1 x Native-PAGE cathode buffer. For mitochondrial complex I in gel activity, the gel was washed in H2O for 10 min and then incubated in 10 ml reaction buffer (2mM Tris buffer pH 7.4, 0.1 mg/ml NADH and one tab of Nitro Blue Tetrazolium (N5514–10TAB, Sigma-Aldrich)). The transfer was performed at 100 V 4 °C for 3 hours with the regular transfer buffer. The membrane was washed with methanol for 10 minutes to destain Coomassie G-250.

Cell growth and galactose challenge assays

0.1 million cells were seeded into 6-well plates and cultured with DMEM no glucose medium plus with 10% FBS and glucose or galactose at a final concentration at 10 mM. The cell number was counted in the following days after trypsinization with a Countess II from Invitrogen using Trypan Blue (Thermofisher # T10282).

Poly I:C treatment

Polyinosinic-polycytidylic acid (Sigma-Aldrich #P0913) was resuspended in nuclease free water at a concentration of 1mg/mL. 2.5*10⁵ cells were plated in a 6 well plate and either treated with 10 µg/ml of poly I:C for 24 hours or a water control. Cells were then harvested and processed for western blots as described above.

Quantitative real-time PCR (qPCR)

RNA was isolated with Trizol (Invitrogen, 15596–026) and the Zymo-Spin Direct-zol RNA Kit (Zymo Research, R2050). 1 μg of RNA was used to generate cDNA with a High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems, 4368813) following the manufacturer’s protocol. For gene expression analysis, cDNA samples were mixed with Sybr Green quantitative PCR master mix (Applied Biosystems, 4309155) and were analyzed by a CFX 384 Real-Time system (Bio-Rad). Primer sequences are provided in the supplemental data.

ChIP-qPCR

Cells were cross-linked with 10% Neutral buffered formalin at room temperature for 2 min and then quenched with 125 mM glycine at room temperature for 5 min. Cells were washed twice with ice-cold 1X PBS and harvested by scraping in 1.5 ml ChIP lysis buffer (50 mM HEPES, pH7.5, 140 mM NaCl, 1mM EDTA, 0.5mM EGTA, 10% Glycerol, 0.5% NP40, 0.25%TX100, 1× proteinase inhibitor cocktail) and incubated on ice for 15 min with vertex several times. The nuclei pellet was spin down by centrifugation at 1500 RCF for 5 min at 4oC. The nuclei pellet in was resuspended 1ml RIPA ChIP buffer (50mM Tris-HCl (pH 8.0); 150mM NaCl; 5mM EDTA; 0.5mM EGTA; 1% Igepal CA-630; 0.1% SDS; 0.5% Na deoxycholate; 10mM NaF; 0.2mM sodium orthovanadate; 5μM trichostatin A; 5mM sodium butyrate; Protease inhibitor cocktail). The suspended pellet was sonicated with the Qsonica sonicator on ice bath for 8 cycles (~10 amplitude microns for 15sec sonication + 1 min interval) and then centrifuged for 10 min at 13,200 rpm at at 4oC. 5% of the lysate was taken as input and the remaining lysate was incubated with H3K27ac antibody and rabbit IgG control at 4 °C for overnight. 30ul of pre-washed Dynabeads Protein G were added to each sample and incubated at 4 °C for 1–2 h. The beads were washed with different buffers, once with ChIP wash buffer (100mM Tris-HCl at pH 8.5, 500 mM LiCL, 1mM EDTA, 1% NP-40, 1% Na deoxycholate), once with RIPA basic buffer and twice with Tris/EDTA buffer (50 mM Tris, pH8.0, 10 mM EDTA). After washing, 50ul of elution buffer (50mM Tris-HCl (pH8.0), 10mM EDTA with 1% SDS) was added to the washed protein-G beads and incubated at 65 °C for 10 min and 50µl Reverse Cross Link Buffer (10mM Tris-HCl (pH8.0), 1mM EDTA, 400mM NaCl) was added to the tube and incubated at 94̊C for 15 minutes. After incubation, 2.5µl Proteinase K (8mg/ml) was added at 65oC and incubated for 2 hours. 600µl PCR-A buffer (40ml PB buffer + 4ml 3M NaAc) was added to each sample. The samples were then purified by EconoSpin DNA column and dissolved into 300µl H2O. The obtained samples were then used as the templates for qPCR analysis. Primer sequences are provided in the supplemental data.

Complex I+III activity measurements

The measurement of Complex I+III activity has been described previously²⁴. Briefly, 6 µg of isolated mitochondria were incubated in 700 µl of distilled water for 2 min and then 100 µl of potassium phosphate buffer (0.5 M, pH 7.5), 20 µl of fatty acid–free BSA (50 mg/ml), 30 µl of KCN (10 mM) and 50 µl of oxidized cytochrome c (1 mM). were added and the final volume was adjusted to be 980 µl with distilled water. Samples plus 10 µl of 1 mM rotenone solution were also prepared as the control to determine Complex I specific activity. The baseline was read at 550 nm for 2 min. The reaction was started by adding 20 µl of 10 mM NADH, and the the increase of absorbance at 550 nm was measured continuously for 2 min. The Complex I+III activity (nmol /min/ mg) was calculated as (∆ Absorbance/min × 1,000)/ [(18.5/ mM/ cm× volume of sample used in ml) × (sample protein concentration in mg/ml)].

CI-driven respiration assays

Seahorse XF96 Cell Culture Microplate was coated with PEI (1:10000 dilution) in 1 X PBS for 5 min at room temperature and then washed with PBS twice. 2 × 104 cells per well in 100 µl growth medium were seeded into the coated plate and spined at 200 x g for 5 min to let the cells attach to the bottom of plate. After 2 h of incubation in cell culture incubator, 500 µl growth medium was added to the plate. The plate was then incubated at 37 °C overnight. The sensor cartridge was hydrated at 37 °C overnight. The following day, growth media was exchanged with Phenol Red-free DMEM without glutamine and supplied with 10 mM glucose and 1 mM pyruvate 1 h before seahorse assay. The OCR was measured following treatment with/without 1 μM rotenone, 2 μM oligomycin, 5 μM FCCP and 1 μM antimycin A. CI driven basal respiration and maximal respiration were calculated by the subtraction the respiration of cells treated with rotenone from the respiration of untreated cells.

Acetyl-CoA Assay

Acetyl-Coenzyme A in cells and in tumors was measured with an Acetyl-CoA Assay Kit (Sigma-Aldrich, MAK039). Briefly, 2×10⁶ cells were suspended in 500 µl Assay Buffer XXII/assay buffer on ice and homogenized. The supernatant was collected after spined at 13,000 x g for 10 min. Ice cold PCA was added to a final concentration of 1 M and incubated on ice for 5 min. Ice-cold 2 M KOH was added equals 34% of the supernatant and vortexed to precipitate PCA. The supernatant collected after centrifuge at 13,000 x g for 15 minutes was ready for test. In the test, reaction with Acetyl-CoA will interact with PicoProbe II/PicoProbe to generate fluorescence (Ex/Em = 535/587 nm). The amount of Acetyl-CoA was calculated according to the Acetyl-CoA standard curve.

Measurement of NAD⁺/NADH ratios

Cellular and tumor NAD⁺/NADH ratios were measured using a commercial assay kit (Abcam, ab65348) according to the manufacturer’s instructions. Briefly, 2 × 10^6 cells or 20 mg tumor tissues were homogenized in 400 μl of extraction buffer. The insoluble debris was removed by centrifuge and the protein was removed by filtering the samples through a 10 kD Spin Column. NAD⁺ was decomposed by incubation in 60 °C for 30 min. Samples with/without NAD⁺ decomposition were used for the following NADH assay by reading OD 450 nm after mixing with the reaction buffer. The amount of NADH was calculated according to the NADH standard curve. NAD⁺ level was calculated by subtracting the NADH in NAD⁺ decomposition samples from NADH in untreated samples.

Plasmid construction, lentiviral generation, and transduction

The plasmid for overexpression of Ndufs4 was purchased from VectorBuilder with the guide RNA targeted sequence synonymously altered. Primer sequences are provided in the supplemental data. CRISPR/Cas9 mediated gene knockout was performed using the GeCKO system, where pLentiCRISPRv2 (Addgene, 98290) was digested with BbsI enzyme and pre-annealed 5’-end phosphorylated sgDNA sequences inserted using Quick Ligase (New England Biolabs), and subsequently transformed into Stabl3 TM E. coli. Resulting plasmid inserts were verified by sequencing (Genewiz). For protein overexpression, replication-defective lentiviral supernatants were generated using transfection of 600ng psPAX2 (Addgene, #12260), 300ng pMD2.G (Addgene, #12259) and 900ng lentiviral plasmid backbone using PolyFect (Qiagen, 301105) into HEK293t cells in 6-well format according to the manufacturer’s instructions. Supernatants collected twice (at 48 and 72h post-transfection, and filtered through a 0.45-µm filter, then used to transduce cells in the presence of 8 μg/ml polybrene. The transduced cells were then selected with 2 μg/ml of puromycin or 8 μg/ml blasticidin for 4 days and then cultured without antibiotics for at least 7 days prior to use in experiments. For protein knockdown, the lentiviral backbone was transfected into target cells using Lipofectamine 3000 in a 6-well format according to the manufacturer’s protocol. The transfected cells were then selected with 5 μg/ml of puromycin or 8 μg/ml blasticidin for 2 days and then cultured without antibiotics for at least 7 days prior to use in experiments.

Tumor allograft studies using melanoma and breast cancer cell lines

Tumor cells (1×10^5 cells for B16-BL6 and Yumm1.7 melanoma and EO7071 breast cancer cells) were injected subcutaneously into the flanks of 7-week-ld female C57BL/6 mice or nude mice with a 30G needle. The tumor volumes were measured by a caliper and calculated as (Volume = length x widtĥ2 / 2). The experiments were when the tumor volumes reach around 2000 mm 3 or any sign of suffering of mouse, such as decreased activity, 20% body weight loss, back-arching, etc. The tumor samples at the final time points were collected for proteomics analyses and infiltrated immune cells analyses by flow cytometry. For the depletion of NK1.1⁺ and CD8⁺ cells in mouse, InVivoMAb anti-mouse NK1.1 (Bioxcell, Catalog# BE0036, Clone: PK136) and InVivoMAb anti-mouse CD8β (Lyt 3.2) (Bioxcell, Catalog# BE0223 Clone: 53–5.8) were administrated at 0.1 mg per mouse on one day before melanoma cell subcutaneous inoculation and then twice per week. The tumor volumes were closely monitored. Spleens were collected for the analysis of NK1.1⁺ and CD8⁺ cells to validate the efficacy of antibody treatment. For immunotherapy treatments, mice were treated with with anti-mouse PD-1 (Bioxcell, Catalog# BE0146, Clone: RMP1–14) or rat IgG2a isotype control (Bioxcell, Catalog# BE0089, Clone: 2A3) when tumors reached between 50–100mm³ with 0.1mg per mouse I.P. every other day for 3 treatments.

In vivo metastasis assay

5*10⁵ B16-BL6 cells with/without Ndufs4 deletion in cold DMEM medium were tail-vein injected into 7-week-old female C57BL/6 mice or nude mice. The mice were closely monitored for their health. The mice were sacrificed if any signs of suffering of mouse were observed, such as decreased activity, 20% body weight loss, back-arching and short-breathing. The sacrificed mice were dissected to check and photograph lung metastasis status. The last days of mice were recorded and used for the analysis of survivals.

Flow cytometry analyses

Tumors were harvested on day 16 following inoculation, weighed, and dissected into small pieces using sterile scalpels in serum-free RPMI 1640 media. Tumor fragments were dissociated in 160 U/mL Collagenase Type IV (Gibco), 80 U/mL DNase I (Sigma), 0.1 mg/ml Hyaluronidase Type V (Sigma) using GentleMACS C tubes on the GentleMACS™ Dissociator (Miltenyi Biotec #130–093-237) followed by incubation at 37°C for 30 min. Following enzymatic digestion, the reaction was quenched with FACS buffer and passed through 70µm cell strainers, and then centrifuged at 300 × g for 6 minutes to pellet the cells. CD45+ cells were enriched using CD45 TIL microbeads (Miltenyi Biotec) following the manufacturer’s protocol. For flow cytometry analysis, cells were stained for 15 minutes at RT with 100 μL of Zombie UV dye (BioLegend) diluted 1:500 in PBS. Cells were washed once with FACS buffer and pre-incubated with 5 μg/mL TruStain FcX anti-mouse CD16/32 antibody (clone 93, BioLegend) in 100 μL FACS buffer for 5 minutes on ice before immunostaining. For flow cytometry analysis of T-cells and NK cells were incubated with a combination of fluorochrome-conjugated antibodies to the following surface markers (from Biolegend unless otherwise indicated): CD3ε (17A2, Biolegend/ BD Bioscience), TCRβ (H57–597, Biolegend/ BD Bioscience), CD4 (RM4–4), CD8α (53–6.7), CD44 (1M7), CD45 (30-F11), CD62L (MEL14), CD69 (H1.2F3), NK1.1 (PK13), CD49b (HMα2), NKG2D (CX5), PD1 (29F.1A12), Tim3 (RMT3–23) at a dilution of 1:250. The cells were stained in 100µl FACS buffer for 15 minutes at RT, washed twice with FACS buffer, resuspended in 300µl of FACS buffer and analyzed using a LSRFortessa X-20 flow cytometer. Data were analyzed using FlowJo software version 10.8.0.

Mitochondrial metabolomics

sgNT and sgNdufs4 B16/BL6 cells were grown in four p-150 dishes to 80% confluent before being trypsinized and washed three times with cold PBS. The cell pellets were suspended in mitochondrial isolation buffer (0.32 M sucrose, 0.001 M EDTA and 0.01 M Tris with pH 7.2). The cell suspense was transferred into a dounce homogenizer and dounced 20 times and then subjected to centrifuge at 1000 x g for 5 min at 4 °C. The supernatant was transferred to a new tube and centrifuged at 1000 x g for 5 min at 4 °C again. The new obtained supernatant was the centrifuged at 9000 x g for 15 min at 4 °C. The pellet was resuspended into mitochondrial isolation buffer and for protein concentration test with DC protein assay (Bio-Rad, 5000111). The obtained mitochondria were aliquoted to 1.5 ml Eppendorf tube with 100 ug per tube and centrifuged at 9000 x g for 10 min. The obtained pellets were resuspended in 80% cold methanol and incubated in −80°C for one hour and were then centrifuged at maximum speed for 10 min at 4 °C. The top layer, which contains the polar metabolites, was removed, and transferred to a new tube and dried down overnight using a SpeedVac (Thermo Fisher Scientific). Lyophilized samples were resuspended in 20 μl HPLC quality water and subjected to metabolomics profiling using the AB/SCIEX 5500 QTRAP triple quadrupole instrument. Data analysis was performed using the GiTools software. Raw integrated peaks were sum normalized by sample for downstream analysis. KEGG pathway analysis was performed using MetaboAnalyst 5.0⁶².

Protein digestion and isobaric labelling for mass spectrometry analysis

Was performed as previously described⁶³. Briefly tumor samples from allograft study were lysed with 4 mL of SDS lysis buffer (2.0% SDS (w/v), 250 mM NaCl, 5 mM TCEP, EDTA-free protease inhibitor cocktail (Promega), and 100 mM HEPES, pH 8.5) using an Omni tissue homogenizer. Extracts were reduced at 57°C for 30 min and cysteine residues were alkylated with iodoacetamide (14 mM) in the dark at room temperature for 45 min. Extracts were purified by methanol–chloroform precipitation and pellets were washed with ice-cold methanol. Pellets were resuspended in 2 mL of 8 M urea (containing 50 mM HEPES, pH 8.5) and protein concentrations were measured by BCA assay (Thermo Fisher Scientific) before protease digestion. 100 µg of protein were diluted to 4 M urea and digested overnight with 4 µg LysC (Wako). Digests were diluted further to a 1.5 M urea concentration and 5 µg of trypsin (Promega) was added for 6 hours at 37°C. Digests were acidified with 50 µl of 20% formic acid and subsequently desalted by C18 solid-phase extraction (50 mg, Sep-Pak, Waters). Digested peptides were resuspended in 100 µl of 200 mM HEPES, pH 8.0. 10 μL of TMTpro reagents (Thermo Fisher Scientific) was added to each solution for 1 hour at room temperature (25°C). After incubating, the reaction was quenched by adding 4 μL of 5% (w/v) hydroxylamine. Labelled peptides were combined and subsequently desalted by C18 solid-phase extraction (50 mg, Sep-Pak, Waters) before basic pH reversed- phase separation.

Basic pH reverse-phase separation for mass spectrometry analysis

Was performed as previously described⁶³. Briefly, tandem mass tag (TMT)-labeled peptides were solubilized in 500 μL solution containing 5% acetonitrile in 10 mM ammonium bicarbonate, pH 8.0 and separated by an Agilent 300 Extend C18 column (5 mm particles, 4.6 mm inner diameter and 220 mm in length). An Agilent 1100 binary pump coupled with a photodiode array detector (Thermo Fisher Scientific) was used to separate the peptides. A 40-min linear gradient from 20% to 40% acetonitrile in 10 mM ammonium bicarbonate, pH 8 (flow rate of 0.6 ml/min) separated the peptide mixtures into a total of 96 fractions (30 sec). A total of 96 fractions was consolidated into 12 samples in a checkerboard fashion, acidified with 20 µl of 20% formic acid and vacuum dried to completion. Each sample was re-dissolved in 5% formic acid, 5% ACN, desalted via StageTips before liquid chromatograph–tandem mass spectrometry (LC–MS/MS) analysis. LC–MS/MS analysis. Data were collected using an Orbitrap Fusion Lumos mass spectrometer (Thermo Fisher Scientific, San Jose, CA) coupled with a Proxeon EASY-nLC 1200 LC pump (Thermo Fisher Scientific). Peptides were separated on a 100 μm inner diameter microcapillary column packed with 35 cm of Accucore C18 resin (2.6 μm, 100 Å, Thermo Fisher Scientific). Peptides were separated using a 3 hours gradient of 6–22% acetonitrile in 0.125% formic acid with a flow rate of ~400 nL/min. Each analysis used an MS3-based TMT method as described previously described⁶⁴. The data were acquired using a mass range of m/z 400 – 1400, resolution at 120,000, AGC target of 1 × 10⁶, a maximum injection time 100 ms, dynamic exclusion of 180 seconds for the peptide measurements in the Orbitrap. Data dependent MS 2 spectra were acquired in the ion trap with a normalized collision energy (NCE) set at 35%, AGC target set to 2.0 × 10⁵ and a maximum injection time of 120 ms. MS3 scans were acquired in the Orbitrap with an HCD collision energy set to 45%, AGC target set to 1.5 × 10⁵, maximum injection time of 200 ms, resolution at 50,000 and with a maximum synchronous precursor selection (SPS) precursors set to 10.

Mass spectrometry data processing and spectra assignment

Was performed as previously described⁶³. Briefly, in-house developed software was used to convert acquired mass spectrometric data from the .RAW file to the mzXML format. Erroneous assignments of peptide ion charge state and monoisotopic m/z were also corrected by the internal software. SEQUEST algorithm was used to assign MS2 spectra by searching the data against a protein sequence database including Mouse Uniprot Database (downloaded June 2018) and known contaminants such as mouse albumin and human keratins. A forward (target) database component was followed by a decoy component including all listed protein sequences. Searches were performed using a 20ppm precursor ion tolerance and requiring both peptide termini to be consistent with trypsin specificity. 16-plex TMT labels on lysine residues and peptide N termini (+304.2071 Da) were set as static modifications and oxidation of methionine residues (+15.99492 Da) as a variable modification. An MS2 spectra assignment false discovery rate (FDR) of less than 1% was implemented by applying the target-decoy database search strategy. Filtering was performed using a linear discrimination analysis method to create one combined filter parameter from the following peptide ion and MS2 spectra properties: XCorr and ΔCn, peptide ion mass accuracy, and peptide length. Linear discrimination scores were used to assign probabilities to each MS2 spectrum for being assigned correctly and these probabilities were further used to filter the data set with an MS2 spectra assignment FDR to obtain a protein identification FDR of less than 1%.

Determination of TMT reporter ion intensities

Was performed as previously described⁶³. Briefly, for reporter ion quantification, a 0.003 m/z window centered on the theoretical m/z value of each reporter ion was monitored for ions, and the maximum intensity of the signal to the theoretical m/z value was recorded. Reporter ion intensities were normalized by multiplication with the ion accumulation time for each MS2 or MS3 spectrum and adjusted based on the overlap of isotopic envelopes of all reporter ions. Following extraction of the reporter ion signal, the isotopic impurities of the TMT reagent were corrected using the values specified by the manufacturer’s specification. Total signal-to-noise values for all peptides were summed for each TMT channel and all values were adjusted to account for variance and a total minimum signal-to-noise value of 200 was implemented.

RNA-seq

RNA samples were prepared with TRIzol^® Reagent. Briefly, logarithmically growing cultures of sgNdufs4 and sgNT cells in 60 mm dish were incubated with 1 ml TRIzol^® Reagent for 5 minutes and then transferred into a 1.5 ml tube and mixed with 0.2 ml chloroform. After vigorously shaking for 15 seconds, samples were centrifuged at 12000 x g for 15 minutes at 4 °C. The aqueous phage was collected and mixed with an equal volume of 100% isopropanol. The mixture was then centrifuged at 12000 x g for 15 minutes at 4 °C to obtain RNA pellet. The supernatant was removed, and the obtained RNA pellet was washed with 75% ethanol 3 times. RNA pellet was dry and resuspended into RNase free water. The concentration of RNA was measured with a nanodrop spectrophotometer, and the quality of RNA was determined by agarose gel electrophoresis. RNA samples were sent to Genewiz for next generation sequencing. The resulting reads were mapped to GRCm38.9 using HiSat2.1 and then quantified using SeqMonk v48.1. CPM normalized counts were then exported and analyzed in R and Excel.

TCGA-SKCM data analyses

The TCGA-SKCM dataset was downloaded with clinical outcome and stage information. The FPKM level data was converted to TPM level, and we limited the analyses to only contain stage IV. We used ssGSEA⁶⁵ to rank the samples based on each of sgNdufs4 UP and DOWN signature. Mantel-Cox log rank calculations were used for statistical analyses of sample segregation on survival as endpoint.

Integrative analyses of proteome and transcriptome data

The quantitative proteomic data was curated to include protein levels that were represented by at least two peptides. Log two-fold (L2F) differences were calculated as a comparison sgNdufs4 vs sgNT across each of B6:wt and nude (four replicates each) with significance for each protein based on unpaired two-sided t-test (p < 0.050). We only considered proteins that were changed in each of B6:wt and nude and satisfied p<0.050 and with an average L2F=1 cut-off across each state to derive proteins with robust changes. For RNA data, log two-fold (L2F) differences were calculated as a comparison sgNdufs4 vs sgNT (three replicates each) with significance for each gene based on unpaired two-sided t-test (p < 0.050). Genes that were either increased/decreased by L2F≥ ±0.7 were used to integrate with proteomic data to find consistent changes. Resulting UP and DOWN signatures were analyzed by Enrichr (GSEA) to identify signatures with p-adj. < 0.050.

Bisulfite conversion and Nlrc5 promoter sequencing

Genomic DNA from B16-BL6 sgNdufs4 and sgNT cells grown in vitro was isolated (DNeasy Blood and Tissue kit, Qiagen), and 5ug DNA each used for bisulfite conversion and subsequently purified according to the manufacturer’s recommendation (Epimark Bisulfite Conversion kit, NEB). To amplify the Nlrc5 promoter region, PCR (HotStart Tq 2x Master mix, NEB) used 1/10^th of the converted DNA and the primers biseq-Nlrc5_F: TGG TTT TTT TTG TTT TTA TTT TAT AT and biseq-Nlrc5_R: AAC AAA TAA CTC ACC AAT TCA ACT C with 48°C annealing and 42 cycles, followed by TOPO-TA cloning (Invitrogen) and Sanger sequencing of resulting inserts. Sequencing was analyzed and graphics were made using CpG Viewer⁶⁶.

Statistical analysis and reproducibility

All measurements were biological replicate samples. In general, for two experimental comparisons in vivo, an unpaired two-tailed Mann-Whitney U-test was used. Otherwise, unpaired Student’s t-tests were used with Welch’s correction as necessary. As Mann-Whitney tests are non-parametric, there are no underlying assumptions about data distributions that can be violated. For data using Student’s t-tests, normality was tested using Shapiro–Wilk test and equal variance was tested using the F-test. No statistical method was used to predetermine sample size but are in line with other literature in the field. No data were excluded from the analyses. The experiments were not randomized. The Investigators were not blinded to allocation during experiments and outcome assessment. GraphPad Prism software (version 10.1.0) and R (version 4.2.1) were used to generate graphs and perform statistical analyses. Microsoft Excel was used for analysis of small-molecule screen data and proteomics. Immunotherapy response datasets were accessed and analyzed through TIGER⁶⁷.

Extended Data

Extended Data Fig. 1:

CI in gel activity assay and galactose sensitivity test for CI subunit knockouts. a, Blue native page gel showing CI activity from isolated crude mitochondria of CI subunit CRISPR knockouts. Purple bands indicate CI activity, blue bands indicate protein presence from Coomassie. Supercomplexes are labeled based on previous work⁶⁹. b, Blue native page gel showing CI activity in sgNdufs6, sgNdufa12, and sgNdufs4 cells and western blot showing NDUFA9 assembly into CI supercomplexes in sgNdufa12 and sgNdufs4 cells. Different numbers under sgNdufs6 represent different CRISPR guides. c, Blue native page gels showing CI activity from isolated crude mitochondria from sgNT and sgNdufs4 tumors from C57BL/6 mice. d, Cell growth curves measured using trypan blue in sgNT and CI mutant cell lines, N=3. e, Western blots showing knockouts for sgNdufa12, sgNdufa9, sgNdufs6 and sgNdufs4 cell lines. f, CI driven respiration in sgNT, sgNdufs4, sgNdufs6, and sgNdufs3 measured by Seahorse assay. Basal respiration p-values: p-(sgNT-sgNdufs4) = 1.3e-10, p-(sgNT-sgNdufs6) = 9.2e-10, p-(sgNT-sgNdufs3) = 9.7e-13, p-(sgNdufs4-sgNdufs3) = 1.7e-6, p-(sgNdufs6-sgNdufs3) = 1.2e-7. Maximal respiration p-values: p-(sgNT-sgNdufs4) = 6.9e-10, p-(sgNT-sgNdufs6) = 3.5e-9, p-(sgNT-sgNdufs3) = 2.87e-14, p-(sgNdufs4-sgNdufs3) = 2.2e-4, p-(sgNdufs6-sgNdufs3) = 2.1e-5. N≥2. g, CI and CIII driven respiration in sgNT, sgNdufs4, sgNdufs6, and sgNdufs3 measured by in vitro cytochrome c reduction. p-values in f and g were calculated using one way ANOVA with Šídák’s multiple comparisons test.

Extended Data Fig. 2:

Anti-tumor effects of sgNdufs4 cells are rescued by Ndufs4 overexpression and are not dependent on the redox state of the cell. a, Growth curves of sgNT and sgNdufs4 tumors transduced with either an empty or a Ndufs4 over expression vector in C57BL/6 mice. P1 compares sgNT-V300 to sgNT-Ndufs4^OE, P2 compares sgNT-Ndufs4^OE to sgNdfus4-Ndufs4^OE, P3 compares sgNT-V300 to sgNdufs4-V300. Western blot confirmation of Ndufs4 over expression. b, Growth curves of sgNT and sgNdufs4 tumors transduced with either an LbNox or mitochondrial target LbNox over expression vector in C57BL/6 mice. P1 compares sgNT-mitoLbNox^OE to sgNdufs4-mitoLbNox^OE and p2 compares sgNT-LbNox^OE to sgNdufs4-LbNox^OE. c, NAD+/NADH ratios in sgNT and sgNdufs4 cells overexpressing either LacZ or mitochondrial LbNox measuring using colorimetric assay (N=4). d, qPCR validation of LbNox overexpression in sgNT and sgNdfus4 cells. n = 5 mice per condition in a and b. P-values in a and b were calculated using Brown-Forsythe and Welch ANOVA test with Dunnett’s T3 multiple comparisons test. Error bars are mean ± s.e.m.

Extended Data Fig. 3:

sgNdufs4 RNA-seq and tumor proteomics reveal MHC-I signature that trends towards significant correlation with SKCM stage IV survival. a, Volcano plots of proteomics done in sgNT and sgNdufs4 tumors in C57BL/6 and nude mice (n=4 tumors per condition). NDUFS4 is labeled as a positive control. b, Volcano plots of RNA-seq done in sgNT and sgNdufs4 cells in culture (n=3 replicates per condition). NDUFS4 is labeled as a positive control. c, TCGA-SKCM Stage IV survival stratified on high or low expression of the gene sets found to be either upregulated or down regulated in sgNdufs4 cells in Fig. 3b, and NLRC5:HLA-B combined rank-based stratification of survival. d, Heatmap of MHC-I antigen presentation pathway in proteomic dataset of sgNT and sgNdufs4 proteomics and RNA-seq datasets. e, RNA expression values of human MHC-I and antigen processing machinery homologs that are upregulated in sgNdufs4 cells in melanoma patients segregated by patients response to immunotherapy (data from PMID: 30753825⁷⁰, accessed through TIGER⁶⁷). Statistics in c and e were calculated using Mantel-Cox log rank test. Box and whisker bars represent min to max with the box representing the interquartile range with the median plotted.

Extended Data Fig. 4:

mRNA levels of antigen processing and immune related genes from tumor grown in C57BL/6 mice. a, Expression levels of MHC-I and antigen processing genes from sgNT or sgNdufs4 B16-BL6 tumors from C57BL/6 mice (N=4 per condition). b, Expression levels of MHC-I and antigen processing genes from sgNT or sgNdufs4 YUMM1.7 tumors form C57BL/6 mice (N=3 per condition). P-values in a and b were calculated using a two-tailed, unpaired Student’s t-test. Error bars are mean ± s.e.m.

Extended Data Fig. 5:

Nlrc5 gene regulation isn’t associated with increased promoter methylation or regulation of NF-κB downstream of cGAS-STING or STAT1. a, CpG Viewer output of bisulfate sequencing of the Nlrc5 promoter region in sgNT and sgNdufs4 cells in vitro. b, Western blot analysis of TBK1 and phospho-TBK1 protein levels in sgNT and sgNdufs4 cells in vitro. Poly(I:C) treated cells were treated with 10µg/mL Poly(I:C) for 24 hours. c, Growth curves of sgNdufs4 and sgNdufs4-sgStat1 tumors along with western blot of protein levels. d, Growth curves of sgNdufs4 and sgNdufs4-sgSting1 tumors along with western blot of protein levels. N=5 mice per condition for b-c (N=2). P-values in b-c were calculated using a Mann Whitney u-test. Error bars are mean ± s.e.m.

Extended Data Fig. 6:

Elucidation of acetyl-CoA increase and effects on global acetylation levels. a, Venn diagram of consistently upregulated and down-regulated metabolites in sgNdufs3, sgNdufs4, and sgNdufs6 mitochondrial enriched metabolomics. b, KEGG metabolite pathway enrichment of commonly downregulated metabolites between sgNdufs4 and sgNdufs6 mitochondrial enriched metabolomics. c, Heatmap of identified TCA metabolites in sgNT, sgNdufs3, sgNdufs4, and sgNdufs6 cells. d, Western blot of histone acetylation marks in sgNT and sgNdufs4 cell lines grown in culture. e, Western blot showing knockout of Pdha in sgNdufs4 cells.

Extended Data Fig. 7:

PDH derived acetyl-CoA drives antigen process and presentations gene transcription. a, qPCR of Nlrc5, H2-K1, Psmb10 and H2-M3 in sgNT and sgNdufs4 B16-BL6 cells treated with DMSO or 100μM Devimistat (PDH inhibitor), N=3. b, qPCR of Nlrc5, H2-K1, Psmb9, Psmb10 and H2-M3 in sgNT and sgNdufs4 EO771 cells treated with DMSO or 100μM Devimistat (PDH inhibitor), (N=3). c, qPCR of Nlrc5, H2-K1, H2-M3, Gzmb, Psmb8, Psmb8, Prf1 and Ifng in B16-BL6 sgNdufs4 tumors transfected with sgNT or sgPdha CRISPR guides (N=3). For a-b shapes indicate averages of independent replicates, p-values are from Brown-Forsythe and Welch ANOVA test with Dunnett’s T3 multiple comparisons test. For c p-values are calculated using Welch’s corrected t-test.

Extended Data Fig. 8:

Gating strategy for T-cell and NK cells in sgNT and sgNdufs4 tumors. a, Density plots showing CD8 versus CD4 discrimination, expression of activation markers CD44 and CD69 and expression of exhaustion markers PD-1 and TIM-3 on CD8⁺ T-cells. b, Density plots showing NK cell discrimination using markers CD49b and NK1.1 and expression of the activation marker CD69 on NK cells. Plots are representative examples from experiments. c, Expression of immune related genes from B16-BL6 tumors grown in C57BL/6 mice (n=4 mice per condition). d, Expression of immune related genes from YUMM1.7 tumors grown in C57BL/6 mice (n=3 mice per condition).

Data Availability Statement

RNA-seq data were deposited into the Gene Expression Omnibus database under accession number GSE263533. Proteomics data were deposited to the Proteomics Identifications Database under accession number PXD051275. For proteomics analysis, the mouse proteome was downloaded from uniport (UP000000589) in 2018. Processed data for each of the datasets can be found in the supplemental materials.