Short-chain acyl-CoA dehydrogenase initiates mtDNA demethylation and leakage to fuel antitumor immunity in colorectal cancer

Abstract

Reprogramming of lipid metabolism and cyclic GMP‒AMP synthase (cGAS)-stimulator of interferon genes (STING) signaling is associated with cancer development. However, whether and how fatty acid metabolism regulates the cGAS‒STING pathway in colorectal cancer (CRC) remains to be elucidated. In this study, we found that short-chain acyl-CoA dehydrogenase (ACADS) is aberrantly deficient in CRC cells and is associated with cancer progression in human patients. We further revealed that ablation of ACADS promoted CRC progression by orchestrating the cGAS‒STING signaling-dependent immunosuppressive tumor microenvironment (TME) in mouse xenografts and AOM/DSS-induced CRC models. Mechanistically, ACADS deficiency suppressed cGAS‒STING signaling by inhibiting mtDNA leakage in a nonmetabolic manner. ACADS binds to and inhibits mitochondrial DNMT1 (mito-DNMT1)-dependent mtDNA methylation, thereby stabilizing mtDNA and inhibiting its leakage. Genetic and pharmacological modulation of mito-DNMT1 restored ACADS-regulated mtDNA leakage, cGAS‒STING signaling, and CRC progression. Importantly, strong correlations between ACADS, mito-DNMT1, and STING signaling and the immune TME were found in patients with CRC. Furthermore, we screened and identified an old drug, hypericin, as an ACADS-binding compound that upregulates ACADS expression. Hypericin treatment can mimic ACADS overexpression-regulated pathways, ultimately improving the immune TME and suppressing CRC growth. These findings highlight a previously undiscovered ACADS/mito-DNMT1 complex that links fatty acid metabolism reprogramming to mtDNA methylation and cGAS‒STING signaling-dependent antitumor immunity.

Introduction

Colorectal cancer (CRC) is among the most prevalent cancers worldwide, with an estimated more than 1.92 million new cases and 903 thousand deaths occurring in 2022, ranking third in incidence and second in mortality.¹ Distant metastasis, involving the lymph nodes, liver, lung, and peritoneum, is the leading cause of death in most patients with CRC, with a 5-year survival rate of less than 20%.² Currently, prevalent strategies for the treatment of CRC encompass multimodal-combined approaches that include surgery, radiotherapy, and chemotherapy; however, the rates of recurrence and metastasis remain notably high.^3,4 In recent years, immunotherapy has demonstrated great potential for cancer treatment, whereas only a minor number of strong immune-activated CRC patients with microsatellite instability-high or deficient mismatch repair are benefit for it.⁴ Therefore, elucidating the underlying mechanisms by which CRC cells inhibit the immune microenvironment and evade immune surveillance, along with developing these novel target-based strategies, represents a critical focus for enhancing progression-free survival in patients with various types of CRC.

Metabolic reprogramming is an important feature and material basis of solid tumors such as CRC. This reprogramming involves altered expression and activity of metabolic enzymes, leading to dysregulated metabolite levels.⁵ Many studies suggest that metabolic enzymes can also regulate tumor initiation and progression through nonmetabolic pathways.⁶ Notably, remodeling of the lipid metabolism has emerged as a functional feature in CRC pathogenesis.^7,8 Our previous work demonstrated that lipid metabolic reprogramming, particularly of triglyceride,^9,10 cholesterol,¹¹ and phospholipid metabolism,¹² plays a critical role in CRC malignancy. Specifically, modulation of lipid metabolism-associated signaling pathways, including upregulation of fatty acid oxidation¹³ and de novo lipogenesis,¹⁴ alongside downregulation of key enzymes governing triglyceride hydrolysis¹⁵ promotes malignant phenotypes such as enhanced proliferation, migration, and metastasis in CRC. Fatty acids constitute the most abundant class of lipids and serve as precursors for diverse lipid species, such as triglycerides, phospholipids, and cholesterols. These molecules play essential and multifaceted roles in maintaining cellular homeostasis in normal tissues. Consequently, dysregulated fatty acid metabolism is mechanistically linked to oncogenic processes, including aberrant cell secretion, migration, and epithelial–mesenchymal transition.¹⁶ Actually, the expression levels of fatty acid synthase were increased in 86% of abnormal crypt foci in patients with sporadic CRC or familial adenomatous polyps, indicating that perturbations in fatty acid metabolism occur in the early stage of CRC development. Moreover, accumulating evidence revealed that the pathways involved in the de novo synthesis of fatty acids and others regulating fatty acid metabolism exert a unique role in promoting cancer growth under different conditions.¹⁷ Consistent with this, pharmacological inhibitors targeting key proteins in fatty acid metabolism have demonstrated robust antitumor efficacy in preclinical CRC models.¹⁸ Collectively, these findings underscore the therapeutic promise of targeting fatty acid metabolic pathways for CRC prevention and intervention. Nevertheless, the functional contributions of fatty acid metabolism-related enzymes in CRC remain incompletely understood.

The malignant progression of CRC is driven by an immunosuppressive tumor microenvironment (TME), characterized by increased infiltration of tumor-associated macrophages (TAMs), myeloid-derived suppressor cells (MDSCs), and regulatory T cells (Tregs), alongside diminished or functionally exhausted CD4⁺ and CD8⁺ T cells.¹⁹ The homeostasis of the TME is dynamically regulated by crosstalk between tumor cells and immune cells. Notably, dysregulated lipid metabolism in CRC cells serves as a key modulator of the TME,²⁰ and intervening lipid metabolism-related signals through dietary or chemical means can impede the progression of CRC by remodeling the TME, such as activating adipose tissue macrophages to differentiate into M1-like phenotype to enhance their phagocytosis,²¹ furtherly underscoring the importance of development new strategies based on lipid metabolism in CRC treatment. However, whether reshaping fatty acid metabolism in CRC cells influences the immune landscape of CRC and the underlying mechanisms remains to be elucidated.

The cyclic GMP‒AMP synthase (cGAS)-stimulator of interferon genes (STING) signaling pathway represents a pivotal innate immune mechanism that plays fundamental roles in microbial defense, antitumor immunity, and various disease processes.²² As a key surveillance system, this pathway detects cytosolic double-stranded DNA (dsDNA) originating from both the nuclear and the mitochondrial compartments. Following dsDNA recognition, pathway activation initiates downstream signaling cascades through the TBK1-IRF3 and NF-κB axes, leading to robust transcriptional induction of type I interferons (notably IFN-β), interferon-stimulated genes (ISGs, including CCL5, OASL, ISG15, ISG56, and CXCL10), and proinflammatory cytokines (such as IL-6).^22,23 These molecular events ultimately potentiate T-cell-mediated antitumor immunity, highlighting the crucial role of this pathway in immune surveillance. In fact, during tumor development, cGAS-STING signaling is often in an inhibitory state, and its regulatory mechanisms have not been fully elucidated. Using mouse tumor models and clinical CRC samples, we screened and identified a protein complex, ACADS/DNMT1, that links fatty acid metabolism reprogramming to mtDNA methylation, mtDNA leakage, and cGAS‒STING signaling. Furthermore, we identified an old drug, hypericin, as an ACADS-binding compound that upregulates ACADS expression, which can mimic ACADS overexpression-regulated signaling pathways, ultimately improving the immune TME and suppressing CRC growth. These findings provide potential biomarkers and therapeutic targets for the management of CRC.

Results

Deficiency of ACADS promotes CRC progression

To determine the key regulators of fatty acid metabolism involved in CRC progression, four GEO datasets from azoxymethane- and dextran sodium sulfate (AOM/DSS)-induced mouse CRC were used to analyze common differentially expressed genes (DEGs) (Fig. 1a, Supplementary Fig. 1a), which were further intersected with the fatty acid metabolism gene set (human gene set: HALLMARK_FATTY_ACID_METABOLISM) and DEGs from the TCGA-COAD dataset (Fig. 1b). Four fatty acid metabolism-associated DEGs, ACADS, carbonic anhydrase 4 (CA4), 15-hydroxyprostaglandin dehydrogenase (HPGD) and UDP-glucose 6-dehydrogenase (UGDH), were present in all the groups (Fig. 1c). Interestingly, according to data from TCGA, only ACADS was significantly decreased in transcription levels in CRC (Fig. 1d) and was associated with patient survival (Supplementary Fig. 1b‒e) and clinical stage (Supplementary Fig. 1f). Decreased ACADS expression in CRC tissues was also confirmed in freshly collected clinical samples (Fig. 1e‒g). To further clarify the function of ACADS in CRC progression, we modulated the expression of Acads via knockdown or overexpression in MC-38 cells and constructed subcutaneous tumor xenografts and peritoneal carcinomatosis models in C57BL/6 mice. The results revealed that CRC tumors derived from Acads-knockdown cells grew faster than those derived from control cells in both xenograft models (Supplementary Fig. 1g‒k). Conversely, tumors derived from Acads-overexpressing cells exhibited slower growth (Supplementary Fig. 1l‒p). These results indicate that Acads may play a tumor-suppressive role in CRC. To further validate the role of Acads in CRC, intestinal epithelial cell-specific Acads conditional knockout mice (Acads^flox/flox-Vil1^Cre+) were generated (Fig. 1h), and an AOM/DSS-induced CRC model was established (Fig. 1i). Compared with the control group (Acads^flox/flox), Acads^flox/flox-Vil1^Cre+ mice presented a significant reduction in body weight, along with increased tumor number and size, whereas colon length was unaffected (Fig. 1j‒m, Supplementary Fig. 1q), suggesting that Acads knockout promoted CRC progression in a colitis-associated CRC model, which was confirmed by H&E staining and Ki67 expression in colon tissues (Fig. 1n‒q). Notably, in contrast to that in control mice, Acads abundance in AOM/DSS-treated mice tended to decrease (Fig. 1n), suggesting that the loss of Acads may be a crucial step in CRC induction. Taken together, these findings indicate that ACADS deficiency promotes CRC progression.

Fig. 1.

Deficiency of ACADS promotes CRC progression. a Venn diagram of the intersection of significantly differentially expressed genes (DEGs) in four gene sets (GSE231709, GSE86299, GSE44988, and GSE155777). b After the 104 DEGs common in (a) were mapped to their corresponding human gene names, a Venn diagram was constructed to depict the intersection of these genes with the DEGs in TCGA-COAD and the genes within the HALLMARK_FATTY_ACID_METABOLISM gene set. c Heatmap of the expression changes in the four common DEGs in (b) across each dataset. d TCGA-COAD database showing the expression levels of ACADS in CRC. e Real-time quantitative PCR showing the expression levels of ACADS in paired samples of normal tissues and CRC tissues from 17 patients. f Representative Western blot results showing ACADS expression levels in paired normal tissues (N) and CRC tissues (C). g Statistical graphic of ACADS relative protein levels in normal and tumor tissues collected from 17 patients. h Protein lysates were prepared from the intestinal tissues of Acads conditional knockout mice (Acads^fl/fl-Vil1^Cre, fl/fl-Cre) or the control group (Acads^fl/fl, fl/fl) and subjected to Western blot analysis with an anti‑Acads antibody to validate the knockout efficiency. i Diagram of the AOM/DSS model used to induce colitis-associated colon cancer. j Body weight percentage changes in fl/fl and fl/fl-Cre mice during AOM/DSS treatment (n = 9 per group). k Representative image of colons from AOM/DSS-treated fl/fl and fl/fl-Cre mice. Tumor number per colon (l, n = 9 per group) and tumor number in different sizes per colon (m, n = 5 per group) of AOM/DSS-treated fl/fl and fl/fl-Cre mice. n‒q H&E and immunofluorescence or immunohistochemical staining for Acads (n, scale bar 20 μm) and Ki67 (p) in colon tumors from AOM/DSS-treated fl/fl and fl/fl-Cre mice, with corresponding quantification [scale bar, 1 mm (top), 200 μm (middle) μm, or 20 μm (bottom)]. Representative images are shown in (n, p); statistical data in (o, q) (n = 6 per group). The data are shown as the means ± SEMs; *p < 0.05, **p < 0.01

ACADS deficiency promotes CRC progression through orchestrating an immunosuppressive TME

To elucidate the mechanism of ACADS deficiency-triggered CRC progression, we first examined the biological behavior of CRC cells in vitro and found that cell viability and proliferation were markedly decreased (Supplementary Fig. 2a, b), whereas apoptosis was not affected by Acads knockdown in MC-38 cells (Supplementary Fig. 2c, d). However, the regulatory effect of Acads on CRC disappeared in the subcutaneous tumor xenograft and peritoneal carcinomatosis models in nude mice (Supplementary Fig. 2e‒l), indicating that the TME plays a decisive role in Acads knockdown-triggered tumor progression. Moreover, we screened ten pairs of colons exhibiting relatively high or low ACADS expression in tumor cells via single-cell sequencing data from human colon cancer patient samples (GSE178341). We first compared and verified the expression of ACADS in tumor cells across the two sample groups (Supplementary Fig. 3a). After dimensionality reduction and cell clustering (clustering results and associated markers are shown in Supplementary Fig. 3b, c), we statistically analyzed changes in the proportion of cells within each group under the two conditions. The results revealed that the proportions of myeloid and tumor cells increased, whereas the proportions of NK and T cells in the microenvironment significantly decreased in the low-ACADS-expressing group (Fig. 2a, b). Further analysis of myeloid, NK, and T cells revealed a significant increase in the number of lipid-associated macrophages (LA-TAMs, cluster 2 myeloid cells), proliferating tumor-associated macrophages (prolif-TAMs, cluster 9 myeloid cells) (Fig. 2c‒e), and immunosuppressive Treg cells (Fig. 2f‒h) in the low ACADS expression group. These findings suggest that low ACADS expression may be detrimental to the tumor immune microenvironment. Furthermore, multiple immunohistochemical staining results confirmed that the positive cell density of MDSCs, macrophages, and Tregs, but not total T cells, was significantly greater in AOM/DSS-induced CRC tissues from Acads^flox/flox-Vil1^Cre+ mice than in those from control mice (Fig. 2i‒m). In addition, the ratios of MDSCs and macrophages were increased, whereas the ratios of T cells and IFNγ⁺ T cells were decreased in Acads-deficient subcutaneous tumors in C57BL/6 mice (Supplementary Fig. 3d‒j). These results were reversed in the Acads-overexpressing setting (Supplementary Fig. 3k‒t). Notably, the alterations in MDSCs in the subcutaneous tumor models were similar to those in the peritoneal carcinomatosis models (Supplementary Fig. 3u‒z). Collectively, these data across multiple tumor models indicate that ACADS deficiency promotes CRC progression by orchestrating an immunosuppressive TME.

Fig. 2.

ACADS deficiency establishes an immunosuppressive TME. a t-SNE plot showing the main cell types in ten colon cancer samples (UAMP of 47499 cells colored by cluster) with high ACADS expression in tumor cells (left panel), contrasting ten colon cancer samples exhibiting low ACADS expression in tumor cells (right panel). b The proportion of each cluster in samples with high and low ACADS expression. c t-SNE plot displaying 11 subtypes of myeloid cells across these twenty colon cancer samples (UAMP of 3511 cells colored by cluster). Macro macrophage, Mono monocyte, Granulo granulocyte, DC dendritic cell. d The proportion of each subcluster of myeloid cells in samples with high and low ACADS expression. e Bubble plot showing markers across myeloid cell subclusters. f t-SNE plot showing subtypes of NK and T cells across these twenty colon cancer samples (UAMP of 11217 cells colored by cluster). g Proportion of each subcluster of NK and T cells in samples with high and low ACADS expression. h Bubble plot showing markers across subclusters of NK and T cells. i Representative multiple immunohistochemical staining of T cells (CD3), MDSCs (GR1), macrophages (F4/80), and Tregs (FOXP3) in tumors collected from AOM/DSS-treated Acads^flox/flox (fl/fl) and Acads^flox/flox-Vil1^Cre (fl/fl-Cre) mice (n = 3 per group; scale bar, 20 μm in the right image). j Statistical image of the positive cell density of T cells (CD3), MDSCs (GR1), macrophages (F4/80), and Tregs (FOXP3) within tumors. Representative multiple immunohistochemical staining of MDSC cells (k, scale bar 20 μm in the right image) and Treg cells (l, scale bar 20 μm in the right image) and the percentage of colocalized positive cells (m) within tumors collected from AOM/DSS-treated fl/fl and fl/fl-Cre mice (n = 4 per group). The data are shown as the means ± SEMs; n.s stands for not significant; *p < 0.05; ***p < 0.001

ACADS deficiency orchestrates an immunosuppressive TME by suppressing the Cgas-Sting signaling pathway

To investigate how the TME is regulated by Acads, RNA-Seq was performed in Acads-knockdown MC-38 cells. KEGG pathway enrichment analysis revealed that the cytosolic DNA-sensing pathway was one of the top downstream signaling pathways of Acads, in addition to fatty acid metabolism (Fig. 3a). Considering the expression of identified cytosolic DNA sensors in CRC cells (Fig. 3b), we assessed the activity of the Cgas-Sting signaling pathway. We found that the protein levels of Cgas, p-Sting, p-Tbk1, and p-Irf3 were markedly decreased in Acads-knockdown cells (Fig. 3c). Conversely, these proteins were increased in Acads-overexpressing cells (Fig. 3c). Notably, the downstream target type I IFNs and ISGs (Ifn-β, Isg15, and Ccl5) displayed a similar trend to that of the Cgas-Sting signaling pathway (Fig. 3d, e), indicating that Acads knockdown triggered the immunosuppressive TME, which likely resulted from a decrease in the Cgas-Sting signaling pathway and associated targets. Compared with those in wild-type mice, the protein levels of Sting were consistently lower in Acads-knockout mice treated with AOM/DSS (Fig. 3f). Furthermore, tumor number and progression were greatly reduced by the Sting agonist DMXAA in the AOM/DSS-induced CRC model in Acads-knockout mice (Fig. 3g‒j), accompanied by an improvement in the immunosuppressive TME (Fig. 3k‒m). Additionally, the suppression of subcutaneous MC-38 tumor growth by Acads overexpression was largely blocked by treatment with C-176, a Sting signaling inhibitor (Fig. 3n, o). These findings suggest that the Acads deficiency-triggered immunosuppressive TME and tumor progression are mediated by the suppression of the Cgas-Sting signaling pathway.

Fig. 3.

ACADS deficiency promotes CRC development through the Cgas-Sting signaling pathway. a KEGG pathway enrichment results showing the major pathways involved in Acads knockdown (#2 vs. NC). b Schematic diagram illustrating the regulation of Acads on type I IFNs and inflammatory response-related genes via the Cgas-Sting signaling pathway. c Western blot analysis of protein levels in the Cgas-Sting pathway, comparing Acads-knockdown (sh-Acads, left) and Acads-overexpressing (Lv-Acads, right) MC-38 cells to their respective negative controls (NCs). Real-time quantitative PCR showing the expression levels of Ifn-β, Isg15, and Ccl5 in MC-38 cells with Acads knockdown (d) or overexpression (e) (n = 6 per sample and treatment group). The data are shown as the means ± SDs. f Representative immunohistochemistry (IHC) staining of Sting in colons collected from AOM/DSS-treated Acads^flox/flox (fl/fl) and Acads^flox/flox-Vil1^Cre (fl/fl-Cre) mice (n = 6 per group, scale bar 50 μm), with the quantification of the percentage of the Sting-positive area shown on the right. g Schematic illustration of intraperitoneal DMXAA administration (15 mg/kg) in the AOM/DSS-induced colitis-associated colon cancer model. h Representative image of colons from DMXAA-treated or untreated AOM/DSS model fl/fl and fl/fl-Cre mice (n = 8 per group). Tumor number per colon (i) and H&E staining (j, scale bar 1 mm) of colons from DMXAA-treated or untreated AOM/DSS model fl/fl and fl/fl-Cre mice (n = 8 per group). Representative multiple IHC staining of MDSCs and Tregs (k, scale bar 20 μm) and their positive cell density (l, m) within tumors collected from DMXAA-treated or untreated AOM/DSS model fl/fl and fl/fl-Cre mice (n = 3 per group). Representative image (n) and histogram of tumor weight (o) in subcutaneous tumor models established with MC-38 cells with or without Acads overexpression. C-176 (16.7 mg/kg) was administered starting on day 2 postvaccination, with doses repeated at 48 h intervals for a total of 5 administrations (n = 5 per group). The data are shown as the means ± SEMs; n.s stands for not significant; *p < 0.05; **p < 0.01; ***p < 0.001

ACADS deficiency suppresses the Cgas-Sting signaling pathway via mtDNA leakage

Leakage of mtDNA from mitochondria into the cytoplasm is a crucial event that activates the Cgas-Sting signaling pathway.²² Given that Acads is located in mitochondria and is responsible for short-chain fatty acid catabolism, we first detected changes in mtDNA in Acads-knockdown MC-38 cells. Superresolution microscopy revealed that the amount of cytoplasmic mtDNA decreased in Acads-knockdown cells (Fig. 4a, b, and Supplementary Fig. 4a) and accumulated in Acads-overexpressing cells (Fig. 4d, e), as confirmed by qPCR (Fig. 4c, f). Notably, the total levels of mtDNA and nuclear DNA were not affected by changes in Acads expression (Supplementary Fig. 4b‒e). Moreover, EtBr, a reagent that depletes cytosolic mtDNA, blocked the Acads overexpression-induced Cgas-Sting signal (Fig. 4g, h) and its downstream targets (Supplementary Fig. 4f, g). These data indicate that the suppression of the Cgas-Sting signaling pathway induced by Acads knockdown is mediated by a decrease in mtDNA leakage.

Fig. 4.

ACADS deficiency inhibits mtDNA leakage-mediated Cgas-Sting signaling in a mtDNA channel-dependent manner. a Representative immunofluorescence image showing cytosolic mtDNA foci in MC-38 cells following Acads knockdown (sh-Acads #2) and in control cells (sh-NC) (scale bar, 5 μm). b Quantification of the mean number of cytosolic mtDNA foci per cell, corresponding to the images shown in (a) (n = 6 per group). c Cytosolic and whole-cell genomic DNA were separately isolated from Acads knockdown (#1, #2) and NC cells. The relative levels of cytosolic mtDNA (targeting the ND1 and D-loop regions) were then quantified via qPCR, with the nuclear gene Tert from the whole-cell genomic DNA used as the internal reference (n = 4 per group). d Representative immunofluorescence image showing cytosolic mtDNA foci in MC-38 cells following Acads overexpression (OE-Acads) and in control cells (OE-NC). e Quantification of the mean number of cytosolic mtDNA foci per cell, corresponding to the images shown in (d) (n = 6 per group). f Cytosolic and whole-cell genomic DNA were separately isolated from OE-Acads and NC cells. The relative levels of cytosolic mtDNA (targeting the ND1 and D-loop regions) were then quantified via qPCR, with the nuclear gene Tert from the whole-cell genomic DNA used as the internal reference. g qPCR showing the relative levels of cytosolic mtDNA, ND1, in OE-Acads and NC cells following 72 h of treatment with ethidium bromide (EtBr, 2 μM) or vehicle (DMSO) (n = 4 per group). β-actin was used as the internal reference. h Western blotting showing the protein levels of Cgas-Sting signaling pathway proteins (Cgas, Sting, Tbk1, and Irf3) in Acads-overexpressing (Lv-Acads, OE) and NC cells following 72 h of treatment with EtBr (2 μM) and DMSO. qPCR showing the relative levels of cytosolic mtDNA, ND1, in MC-38 cells with or without Acads intervention following 24 h of treatment with 200 μM MitoQ (i, n = 6 per group), 5 μM VBIT4 (j, n = 4 per group), or 5 μM cyclosporin A (CsA, k, n = 6 per group). Western blotting showing the protein levels of the Cgas-Sting pathway in MC-38 cells with or without Acads overexpression following 24 h of treatment with 5 μM VBIT4 (l) or 5 μM CsA (m). Representative image (n) and histogram of tumor weight (o) illustrating the impact of VBIT4 on the progression of subcutaneous tumors established by MC-38 cells with or without Acads overexpression. VBIT4 (20 mg/kg) was administered daily starting on day 5 postinoculation, with a total duration of 5 consecutive days (n = 5 per group). p, q Western blotting showing the protein levels of Bax, Vdac1, Vdac3, Hsp60, and Phb1 in whole cells (total) and mitochondria (Mito.) in Acads-knockdown (sh-Acads) or Acads-overexpressing MC-38 cells (Lv-Acads). r Representative images of flow cytometry analysis (left) and histogram (right) demonstrating the mean fluorescence intensity of TMRM in MC-38 cells with or without Acads knockdown (n = 3 per group). The data are shown as the means ± SDs; n.s stands for not significant; *p < 0.05; **p < 0.01; ***p < 0.001

ACADS deficiency inhibits mtDNA leakage in a mtDNA channel-dependent manner

Generally, leakage of mtDNA from mitochondria into the cytoplasm is determined by mitochondrial function and mtDNA channels.²⁴ To clarify how Acads affects mtDNA release from mitochondria, we first examined key mitochondrial factors involved in regulating mtDNA release, including reactive oxygen species (ROS) generation,²⁵ the calcium ion concentration,²⁶ and the ATP content.²⁷ We found that Acads suppressed ROS generation and promoted ATP production (Supplementary Fig. 4h‒k), whereas calcium ion levels were unaffected (Supplementary Fig. 4l, m). However, elimination of ROS with NAC (cellular) and MitoQ (mitochondrial) or ATP replenishment did not rescue the reduction in the cytosolic mtDNA content in Acads-knockdown cells (Fig. 4i, Supplementary Fig. 4n, o). Next, we examined biomarkers of mitochondrial fission (Dnm1l and Fis1), fusion (Mfn1, Mfn2, and Opa1), and biogenesis (Pgc1α, Tfam, and Nrf1). No significant changes in the expression of these genes were observed in Acads-knockdown or Acads-overexpressing cells compared with their respective controls (Supplementary Fig. 4p‒s). We also investigated the effect of sodium butyrate (NaBu), the principal substrate of Acads, on mtDNA leakage. Although NaBu stimulated mtDNA leakage in vitro, it had no obvious effect on subcutaneous or peritoneal xenografts established with MC-38 or CT-26 cells (Supplementary Fig. 4t‒x). These findings suggest that changes in mitochondrial function are not the primary regulators of Acads-mediated mtDNA leakage, which may instead be attributed to its nonenzymatic functions. Therefore, we examined the role of mtDNA channels in mtDNA leakage and found that Acads overexpression-enhanced mtDNA leakage was diminished by the voltage-dependent anion channel 1 (VDAC) oligomerization inhibitor VBIT4 and the mitochondrial permeability transition pore (mPTP)-opening inhibitor cyclosporin A (CsA) (Fig. 4j, k). This was further confirmed by changes in Cgas-Sting signaling (Fig. 4l, m). Moreover, inhibition of VDAC1 oligomerization disrupted the tumor-suppressive role of Acads in subcutaneous tumor xenograft models (Fig. 4n, o). Intriguingly, the protein levels of typical mtDNA channel molecules, including BCL2 antagonist/killer and BCL2-associated X apoptosis regulator (BAK/BAX) macropores,²⁸ VDAC oligomers,²⁹ and mPTP components such as VDAC1 and prohibitin 1 (PHB1),³⁰ did not change dramatically in Acads-knockdown or Acads-overexpressing cells (Fig. 4p, q). Additionally, FCM analysis revealed that the mitochondrial membrane potential (MMP) associated with mPTP opening was only slightly decreased following Acads knockdown (Fig. 4r). These results indicate that Acads deficiency inhibits mtDNA leakage in a mtDNA channel-dependent manner but that the components of mitochondrial channels are not affected by Acads modulation.

ACADS deficiency promotes mitochondrial DNA methyltransferase 1 (mito-Dnmt1)-dependent mtDNA methylation and leakage suppression

If mtDNA channels function normally, the regulation of mtDNA leakage is likely intrinsic to the mtDNA itself. Oxidized mtDNA has been reported to influence mtDNA leakage and Cgas-Sting signaling.³¹ Epigenetic modifications of mtDNA may alter its conformation and subsequently affect its translocation through mtDNA channels. Thus, we speculated that Acads may regulate specific mtDNA. To verify this hypothesis, we explored Acads-regulated epigenetic pathways. Transcriptomic analysis revealed that Acads knockdown did not affect epigenetic signaling pathways (Fig. 3a). We further analyzed the Acads-interacting proteome via mass spectrometry. A well-known epigenetic enzyme, Dnmt1, was identified and captured our attention (Fig. 5a). Protein‒protein interactions between Acads and Dnmt1 were further confirmed by coimmunoprecipitation (Fig. 5b, Supplementary Fig. 5a‒c), proximity ligation assay (Fig. 5c), immunofluorescence (Fig. 5d, Supplementary Fig. 5d), and high-resolution microscopy (Supplementary Fig. 5e). However, we were unable to detect any specific enrichment of Dnmt1 in the GST-Acads pull-down assays (Supplementary Fig. 5f, g). Considering that Acads is a fatty acid catabolism enzyme located in mitochondria, we extracted mitochondria and found that the protein level of mito-Dnmt1 was increased after Acads knockdown (Fig. 5e), which was also observed in CRC tumor samples from Acads conditional knockout mice compared with those from control mice (Supplementary Fig. 5h). However, the mRNA levels of Dnmt1 remained unchanged following Acads knockdown (Fig. 5f), suggesting that Acads deficiency may affect the posttranscriptional regulation of Dnmt1. We demonstrated that mito-Dnmt1 accumulated after treatment with MG132, a proteasome inhibitor (Fig. 5g). Furthermore, the coimmunoprecipitation results revealed that Dnmt1 ubiquitination was reduced after Acads was knocked down (Fig. 5h). As expected, immunofluorescence and methylation sequencing revealed that DNA methylation was greater in Acads-knockdown cells than in control cells (Fig. 5i‒k). Furthermore, ELISA (Fig. 5l), bisulfite sequencing (BSP; Supplementary Fig. 6a‒d), and third-generation nanopore sequencing (Supplementary Fig. 6e) revealed that 5-mC methylation of mtDNA was elevated following Acads knockdown. Consistently, almost all the mRNA levels of the mitochondrial genome-encoded genes decreased after Acads were knocked down (Supplementary Fig. 7a‒c). These data suggest that Acads knockdown promotes mtDNA methylation by suppressing mito-Dnmt1 degradation in a proteasome-dependent manner.

Fig. 5.

ACADS deficiency promotes mitochondrial DNA methyltransferase 1 (mito-Dnmt1)-dependent mtDNA methylation and leakage suppression. a LC-MS/MS results showing the interaction of Acads and Dnmt1 in MC-38 cells. b Lysates from MC-38 cells prepared in IP lysis buffer were incubated for 8 h with magnetic beads conjugated to an anti-Acads antibody or an IgG control. After washing, the proteins bound to the beads were eluted with 2× loading buffer and detected via Western blotting with antibodies against Dnmt1 and Acads. c MC-38 cells grown on coverslips were divided into four groups. After fixation, permeabilization, and blocking, the cells were incubated with different primary antibodies (or controls): BSA (control, Con.), rabbit anti-Acads, mouse anti-Dnmt1, or a combination of both antibodies. Following washing, the samples were incubated with corresponding secondary antibodies conjugated to fluorescent probes. Signal amplification was then performed via a DNA polymerase-mediated method. Nuclei were counterstained with DAPI (blue), and images were acquired. The figure shows representative immunofluorescence images from the four treatment groups (scale bar, 20 μm). d Immunofluorescence staining was performed on the two indicated groups (control and the combination of rabbit anti-Acads with mouse anti-Dnmt1), followed by imaging via confocal microscopy (scale bar, 5 μm). Representative images of the negative control group are presented in Supplementary Fig. 5a. e Western blotting showing the expression of Dnmt1 in whole cells (Total), mitochondria (Mito.), and cytoplasm (Cyto.) of MC-38 cells, with or without Acads knockdown. f qPCR showing the mRNA levels of Dnmt1 in MC-38 cells with or without Acads knockdown (n = 3 per group). g Western blot analysis showing the abundance of Dnmt1 in mitochondria (mito.) and whole-cell lysates (WCL) of MC-38 cells following 24 h of treatment with MG132 (5 μM). h Western blot analysis demonstrating the level of Dnmt1 ubiquitination in immunoprecipitated MC-38 cells via an anti-Dnmt1 antibody. Immunofluorescence assay demonstrating the levels of 5-mC in MC-38 cells with or without Acads knockdown (i, scale bar 20 μm), with the quantification of the mean fluorescence intensity of 5-mC shown in (j) (n = 4 per group). k Whole-cell DNA methylation sequencing showing the levels of whole-genome methylation in MC-38 cells, with or without Acads knockdown. l ELISA results showing 5-mC methylation in mtDNA following its extraction from the purified mitochondria of Acads knockout and control MC-38 cells (n = 3 per group). m qPCR results showing the relative levels of cytosolic mtDNA (ND1) in MC-38 cells with or without mitochondrial Dnmt1 overexpression (Lv-mito-Dnmt1, OE) (n = 9 per group). n Western blotting showing the protein levels of the Cgas-Sting pathway in MC-38 cells with or without mitochondrial Dnmt1 overexpression. o qPCR results showing the relative mRNA levels of Ifn-β and Oasl1 in MC-38 cells with or without mitochondrial Dnmt1 overexpression (n = 6 per group). p Histogram of tumor weight and representative image illustrating the progression of tumors in subcutaneous models established with MC-38 cells with or without mitochondrial Dnmt1 overexpression (n = 4 per group). q qPCR results showing the relative levels of cytosolic mtDNA in MC-38 cells with or without Acads knocked down following 48 h of treatment with 5 μM decitabine (n = 6 per group). Representative image (r) and histogram of tumor weight (s) illustrating the impact of decitabine on the progression of tumors in subcutaneous models established with Acads-knockdown MC-38 cells. During this period, decitabine (1.1 mg/kg) was administered via intraperitoneal injection every two days after the indicated cells were seeded (n = 6 per group). (t) qPCR showing the relative levels of cytosolic mtDNA in MC-38 cells with or without CRISPR/Cas9-based mito-Dnmt1 knockout. No. 51 and 58 were screened from two mito-Dnmt1 knockout cell lines (n = 4 per group). Representative image (u) and histogram of tumor weight (v) illustrating the progression of tumors in subcutaneous models established with MC-38 cells with or without specific knockout of mito-Dnmt1 (n = 5 per group). The data are shown as the means ± SDs; n.s stands for not significant; *p < 0.05; **p < 0.01; ***p < 0.001

DNMT1 is an enzyme that transfers methyl groups to cytosine nucleotides in genomic DNA in the nucleus.³² However, two groups reported that there is a 186 bp upstream open reading frame (uORF) sequence encoding a mitochondrial localization sequence (MLS) located immediately to the 5′ flanking of full-length DNMT1, which can transfer DNMT1 into the mitochondria to perform 5-mC and 5-hmC methylation.^33,34 To verify the role of mito-Dnmt1 in mtDNA leakage and CRC progression, we established mito-Dnmt1-overexpressing MC-38 cells containing MLS and full-length Dnmt1 (MLS-Dnmt1) via a lentivirus (Supplementary Fig. 7d) and found that some mitochondrial function parameters within Acads-knockdown cells, such as the MMP and mitoROS, rather than the ATP content, were restored in CRC cells overexpressing mito-Dnmt1 (Supplementary Fig. 7e‒g). Importantly, the cytoplasmic mtDNA content was markedly decreased in mito-Dnmt1-overexpressing cells (Fig. 5m), and several downstream targets of Cgas-Sting signaling were significantly decreased compared with those in control CRC cells (Fig. 5n, o, and Supplementary Fig. 7h). A tumor-promoting role was observed in subcutaneous tumor xenografts and peritoneal carcinomatosis models constructed with mito-Dnmt1-overexpressing CRC cells (Fig. 5p, Supplementary Fig. 7i). Furthermore, the reduction in cytosolic mtDNA and the Cgas-Sting downstream gene Ccl5 caused by Acads deficiency was reversed by the Dnmt1 inhibitor decitabine (Fig. 5q, Supplementary Fig. 7j). Accordingly, the CRC progression induced by Acads silencing was blocked by decitabine treatment (Fig. 5r, s). To further verify the function of mito-Dnmt1 in mtDNA release and CRC progression, we established mito-Dnmt1-specific knockout MC-38 cells via CRISPR/Cas9, which targets its uORF sequence, and obtained two clones, numbered 51 and 58. Analysis of the genomic DNA sequences revealed that clone 51 carried a 20 bp deletion (^-120TCTCTTGCCCTGTGTGGTAC^-101) within the uORF, whereas clone 58 harbored a single nucleotide insertion (T) upstream of −109 bp (Supplementary Fig. 8a). Both clones showed downregulation of the mito‑Dnmt1 protein (Supplementary Fig. 8b). Notably, in clone 51, total Dnmt1 protein and mRNA levels were also significantly reduced (Supplementary Fig. 8b, c). As expected, specific knockout of mito-Dnmt1 significantly increased mtDNA release (Fig. 5t) and inhibited CRC progression in two xenograft models (Fig. 5u, v, and Supplementary Fig. 8d). Interestingly, the two Cas9‑mito‑Dnmt1 clones exhibited divergent tumor growth, which may be attributed to their differential mito‑Dnmt1 abundance or to other mechanisms triggered by the overall reduction in cellular Dnmt1 levels.

To determine whether the enzymatic activity of mito-Dnmt1 contributes to CRC progression, we transfected clone 58 with either wild-type (WT) mito-Dnmt1 (mito-Dnmt1^WT-Flag) or a catalytic- dead mutant (mito-Dnmt1^Mut-Flag, R1576A/N1580A) plasmid. Compared with the WT group, the catalytically dead group presented increased mtDNA leakage (Supplementary Fig. 8e), accompanied by significant accumulation of Cgas, indicating increased activation of the Cgas-Sting pathway (Supplementary Fig. 8f). Moreover, a statistically nonsignificant decrease in tumor growth was observed in the catalytic-dead group relative to the wild-type group in the subcutaneous xenograft model, which may be partly attributable to the absence of stably overexpressing cell lines in the study (Supplementary Fig. 8g). Nevertheless, this trend supports the role of mito-Dnmt1 catalytic activity in promoting CRC progression. Collectively, our data suggest that Acads deficiency enhances mito-Dnmt1-mediated mtDNA methylation, which suppresses mtDNA leakage, thereby limiting Cgas-Sting signaling and ultimately accelerating CRC progression.

DNA methylation is a common type of DNA damage that can trigger DNA repair and maintain DNA stability.³⁵ Therefore, to determine whether changes in the methylation of mtDNA affect its ability to promote DNA repair, we assessed the protein levels of genes related to mtDNA replication and repair-related proteins, including DNA polymerase γ (Polg)³⁶ and Twinkle,³⁷ and found that these two proteins were increased upon Acads knockdown and decreased upon Acads overexpression in MC-38 cells (Supplementary Fig. 8h, i). A similar trend was observed in cells overexpressing mito-Dnmt1 (Supplementary Fig. 8j). These findings, together with the observation that total mtDNA levels remained largely unchanged despite Acads dysregulation (Supplementary Fig. 4b, c), suggest that the accumulation of mtDNA methylation may enhance DNA repair and stability, thereby contributing to a reduction in mtDNA leakage.

The ACADS-DNMT1-STING axis is correlated with the immune TME and CRC progression in human patients

To determine whether ACADS deficiency promotes CRC progression in humans, we collected 77 pairs of cancer and paracancer samples from patients with CRC and analyzed the expression and prognostic correlation of ACADS. The results revealed that the expression level of ACADS in tumor tissues was lower than that in paracancerous tissues, and this reduced expression was significantly associated with poor prognosis and tumor stage (Fig. 6a‒d, Supplementary Fig. 9a). Furthermore, we compared the changes in ACADS and DNMT1 expression via immunofluorescence staining and found that DNMT1 levels increased as ACADS levels decreased in CRC tissues (Fig. 6e). Additionally, we observed a significant increase in the expression of mito-DNMT1 in CRC tissues by assessing the positive area ratio and the proportion of cells positive for the target protein to the mitochondrial marker HSP60 (Fig. 6f, g). Notably, the expression of ACADS and DNMT1 in the mitochondria was significantly negatively correlated (Fig. 6h). Moreover, we analyzed colon cancer datasets from CPTAC and TCGA-COAD and found that ACADS and DNMT1 were significantly negatively correlated with both protein and mRNA expression levels in CRC (Fig. 6i, j). Moreover, we employed multiple algorithms to conduct immune infiltration analysis on the expression data from TCGA-COAD, examining the correlation between the expression levels of ACADS and STING signals in tumor tissues and various types of immune cells (Fig. 6k, l). Our findings revealed a significant positive correlation between the expression levels of ACADS and STING signaling (STING and IRF3) (Fig. 6l, m). The consistent trend in ACADS and STING expression was also observed in clinical samples (Supplementary Fig. 9a, b). Furthermore, the expression of both ACADS and STING was positively correlated with the proportions of CD4⁺ and CD8⁺ effector T cells in tumor tissues but negatively correlated with the proportions of monocytes, neutrophils, and Tregs in the same tissues (Fig. 6k‒m, Supplementary Fig. 9c). Additionally, we analyzed the immunophenoscore (IPS) of COAD patients who received anti-PD-1/CTLA-4 immunotherapy in the TCIA database and reported that low ACADS expression was accompanied by a poor response to immunotherapy in COAD patients (Fig. 6n). These results suggest that the ACADS-DNMT1-STING axis is correlated with the immune TME and CRC progression in human patients.

Fig. 6.

The ACADS-DNMT1-STING axis is correlated with the TME and CRC progression in human patients. a Representative H&E and immunohistochemistry images of ACADS in paracancerous and colon cancer tissues from 77 CRC patients (scale bar, 50 μm). b Histogram showing the H-scores of ACADS in 77 CRC and paracancerous tissues. c Kaplan‒Meier analysis of the survival curves of CRC patients with high and low ACADS expression from the microarray. d Histogram showing the H-scores of ACADS in different stages of CRC measured via a tissue microarray (paracancer from 77 CRC patients, stage 1 from 13 patients, stage 2 from 37 patients, and stage 3/4 from 27 patients). e Representative multiple immunohistochemistry images of DNMT1, ACADS, and HSP60 in CRC and paracancerous tissues from the microarray (scale bar, 50 μm). Histogram showing the percentage of mito-DNMT1-positive areas (f) and mito-DNMT1-positive cells (g) in CRC and paracancerous tissues from the microarray (n = 77 patients). h Correlations between mito-DNMT1 and ACADS levels in CRC tissues measured via a tissue microarray. i Correlation of DNMT1 and ACADS protein levels in CRC tissues obtained from the CPTAC database (n = 197). j The correlation of DNMT1 and ACADS mRNA levels in CRC tissues obtained from the TCGA-COAD database (n = 434). k The correlation heatmap illustrates the relationships among ACADS, DNMT1, STING signaling, and various immune cells derived from the immune infiltration analysis results obtained via the xCell algorithm for these genes. Dot plots illustrating the correlation in expression between ACADS and either STING (l) (n = 512) or IRF3 (m) (n = 512), utilizing expression data from the TCGA-COAD cohort. n Low expression of ACADS was significantly correlated with poor immunotherapy benefit. The IPS score (ctla4: neg; pd1: pos) represents the IPS score when receiving only anti-PD1 therapy, the IPS score (ctla4: pos; pd1: neg) represents the IPS score when receiving only anti-CTLA-4 therapy, and the IPS score (ctla4: pos; pd1: pos) represents the IPS score when receiving both anti-CTLA-4 therapy and anti-PD1 therapy. The data are shown as the means ± SEMs, *p < 0.05; **p < 0.01; ***p < 0.001

Activating the ACADS-DNMT1-mediated cGAS-STING signaling pathway restricts CRC growth

To investigate the potential therapeutic role of ACADS in CRC, we performed structure-based virtual screening and selected eight candidate compounds predicted to directly bind to ACADS (Fig. 7a, Supplementary Fig. 10a). Among these compounds, only hypericin, a natural polycyclic aromatic naphthodianthrone extracted from Hypericum, significantly increased ACADS protein levels (Fig. 7b, c, and Supplementary Fig. 10b, c), and surface plasmon resonance (SPR) assays demonstrated high affinity binding between ACADS and hypericin (Kd = 4.25 μM) (Fig. 7d). Importantly, treatment with hypericin decreased the protein levels of mito-DNMT1 (Fig. 7e), which is consistent with the results of Acads overexpression in CRC cells (Supplementary Fig. 10d). Furthermore, hypericin activated mtDNA leakage (Fig. 7f) and the Cgas-Sting signaling pathway (Fig. 7h), increasing the expression of its downstream factors (Ifn-β, Isg15, and Ccl5) (Supplementary Fig. 10e), accompanied by a marked decrease in nuclear DNA in the cytoplasm (Fig. 7g) and an obvious decrease in the levels of Polg and Twinkle (Supplementary Fig. 10f). Importantly, the ability of hypericin to activate the Cgas‑Sting signaling pathway was abolished in Acads‑knockdown cells (Fig. 7i, Supplementary Fig. 10g‒i). In vivo experiments further demonstrated that hypericin significantly inhibited CRC growth (Fig. 7j‒l, Supplementary Fig. 10j, k), whereas this inhibitory effect was lost in Acads-knockdown cells (Supplementary Fig. 10l, m). Flow cytometry analysis revealed that, compared with vehicle treatment, hypericin treatment increased the proportions of total T cells, CD4⁺ T cells, IFNγ⁺ T cells, and M1-like macrophages while reducing the numbers of MDSCs and Tregs in tumor tissues (Fig. 7m). Similar trends in the proportions of total T cells, IFNγ⁺ T cells, and Tregs were observed in human CRC tissue-derived cells following treatment with hypericin (Fig. 7n, o, and Supplementary Fig. 10n). Collectively, these findings suggest that hypericin activates the ACADS-DNMT1-cGAS-STING signaling pathway in an Acads-dependent manner, thereby suppressing CRC growth through the modulation of an immunosuppressive TME.

Fig. 7.

Hypericin targets ACADS for immune activation and CRC therapy. a Schematic diagram of structure-based virtual screening for Acads-specific binders. b 2D (upper) and 3D (below) views of hypericin with Acads. The light blue cartoon represents Acads. c Three cell lines (MC-38, CT-26, and HCT-116) were treated with varying concentrations of hypericin (0, 5, and 10 µM) for 24 h. The protein level of Acads was then analyzed by Western blotting, with β-actin serving as the loading control. d SPR assay showing responses measured in the resonance unit (RU) of the ACADS protein (chip-coupled) to hypericin. e Western blotting showing the expression of Dnmt1 and Acads in whole cells (total) and mitochondria (Mito.) of MC-38 cells after 24 h of treatment with hypericin (10 µM). f, g qPCR showing the relative levels of cytosolic mtDNA (ND1 and D-loop) in MC-38 cells after 24 h of treatment with hypericin (10 µM, n = 4 per group). h Western blotting showing the protein levels of Cgas-Sting signaling pathway proteins (Cgas, Sting, Tbk1, and Irf3) in MC-38 cells following 24 h of treatment with hypericin (10 μM). i MC-38 cells with or without Acads knockdown were treated with hypericin (10 µM) or vehicle control for 24 h. Whole‑cell lysates were then collected and analyzed by qPCR to quantify the expression levels of the downstream Cgas-Sting effector Ifn-β, with β-actin serving as the loading control. j Experimental scheme of hypericin administration in MC-38 tumor‑bearing mice. k, l Representative image and histogram of tumor volume and weight (l) illustrating the impact of hypericin (10 mg/kg) on the progression of tumors in a subcutaneous mouse model established with MC-38 cells (n = 5 per group). During this period, hypericin was first intraperitoneally injected (i.p.) on the 6th day after tumor inoculation. m Flow cytometry analysis showing the changes in MDSCs, macrophages, total T cells, CD8⁺ T cells, CD4⁺ T cells, IFNγ⁺ T cells, and Tregs in subcutaneous tumors established by MC-38 cells, with or without 24 h of treatment with hypericin (10 mg/kg) (n = 6 per group). n, o Flow cytometry analysis of the changes in total T cells (CD3⁺) and IFNγ⁺ T cells (CD3⁺IFNγ⁺) in a mixture of single cells digested from CRC patients after 24 h of treatment with vehicle or hypericin (10 μM) (n = 3 per group). p Schematic illustration of the mechanism by which ACADS deficiency promotes the progression of CRC. The data are shown as the means ± SDs; n.s stands for not significant; *p < 0.05; **p < 0.01; ***p < 0.001

Discussion

CRC remains a significant threat to human health, with its mortality rate being among the highest of all cancers despite the development of multiple therapeutic strategies.¹ Therefore, it is crucial to identify novel therapeutic targets to broaden the treatment landscape for CRC. Our study highlighted the role of fatty acid metabolism in CRC, showing that ACADS deficiency promotes tumor progression by establishing an immunosuppressive TME via suppression of the cGAS-STING signaling pathway. Importantly, we revealed a novel mechanism through which ACADS interacts with and promotes the degradation of mito-DNMT1, thereby regulating mtDNA methylation-dependent leakage and subsequent activation of the cGAS‒STING signaling pathway. Notably, this study identified an ACADS activator, hypericin, as a potent antitumor agent that promotes mtDNA leakage and cGAS-STING signaling.

As a hallmark of tumors, lipid metabolic reprogramming has demonstrated considerable potential in both diagnosis and therapy, particularly in the early stages of CRC, where fatty acid metabolism plays a crucial role.¹⁶ To find out the key fatty acid metabolism-related genes involved in the occurrence of CRC, we first analysis four groups of transcriptome sequencing data from AOM/DSS-induced mouse CRC samples, together with the TCGA-COAD database, and found that ACADS, the enzyme related to short-chain fatty acid metabolism, significantly decreased in CRC in all groups and were intimately associated with the survival of CRC patients. This tumor-suppressive role of ACADS was validated in three different mouse models, including subcutaneous tumor xenografts, peritoneal carcinomatosis models, and AOM/DSS-induced CRC models. These findings, combined with the clinical data, strongly suggested that ACADS serves as a tumor suppressor in CRC progression. Furthermore, we established subcutaneous tumor xenografts and peritoneal carcinomatosis models in nude mice, and found that the tumor-promoting role of Acads-deficiency had disappeared, suggesting Acads deficiency could promote CRC progression through reshaping the immune TME. This conclusion was further strengthened by the results that immunosuppressive cells, such as MDSCs and Tregs, were highly accumulated in Acads-silenced, and reduced in Acads-overexpressed CRC tumors.

Considering that Acads is a mitochondria-resident protein, the mitochondria might be involved in the regulation of immunosuppressive TME triggered by Acads deficiency. It has been revealed that mitochondria within immune cells or tumor cells can modulate TME through regulating AKT1-PGC1α, BNIP3-BNIP3L, cGAS-STING, NLRP3-Caspase 1, MAPK, and NF-κB signaling pathways.³⁸ To uncover the mechanism by which Acads-deficiency induced immune suppression, we performed transcriptome sequencing and found that the cytosolic DNA-sensing pathway was significantly changed, indicating that the cGAS-STING signal may be a crucial pathway mediating Acads-regulated tumor immunity. In fact, the cGAS-STING pathway constitutes an integral component of the innate immune system, functioning to detect DNA damage and viral infections, thereby initiating an immune response.²² Activation of this pathway results in the production of immune mediators, including type I IFN and an inflammatory response, which subsequently activate immune cells to target and eliminate cancer cells. Interestingly, this immune response has also been observed in tumor cells.²² In this study, we found that the type I IFN and inflammatory response-related factors were decreased in Acads-deficient CRC cells, and increased in Acads-overexpressed cells, compared to their respective controls. Notably, activation of the STING signal with the agonist can significantly reduce the infiltration of MDSCs and Tregs within TME, thereby inhibiting CRC progression. Moreover, eliminating the content of cytosolic mtDNA disrupted the cGAS-STING signal activation in Acads-overexpressed cells. These results suggest that Acads can regulate cGAS-STING-mediated inflammatory signaling by regulating mtDNA leakage.

A crucial finding in this study was that ACADS potentiates mtDNA leakage, which has not been previously reported. The translocation of mtDNA from the mitochondria to the cytoplasm is associated with multiple factors, including mitochondrial apoptosis,³⁹ channel activity,^28–30 ROS generation,²⁵ the calcium ion concentration,²⁶ and the ATP content.²⁷ However, our results demonstrated that none of these factors are involved in ACADS-regulated mtDNA leakage, suggesting the existence of a nonclassical mechanism. In particular, we found that ACADS initiated mtDNA leakage by inhibiting mito-DNMT1-dependent mtDNA methylation. The accumulation of DNA methylation triggers DNA repair and maintains DNA stability.³⁵ We observed that the expression of mtDNA repair-related proteins (Polg and Twinkle) was suppressed by ACADS and hypericin. These findings indicate that ACADS might regulate mtDNA leakage by regulating mtDNA methylation-associated DNA repair and enhancing its stability.

To explore the mechanism linking ACADS to mito-DNMT1 expression, we ruled out the involvement of metabolic pathways by evaluating the effect of butyric acid, a substrate of ACADS. Notably, we discovered that Acads can form a complex with mito-Dnmt1 through an indirect binding mechanism and suppress mito-Dnmt1 protein expression via the ubiquitin–proteasome system. This finding is intriguing because mitochondria lack a ubiquitin‒proteasome system.⁴⁰ Given that mitochondrial protein degradation typically occurs via ubiquitin-mediated processes at the outer mitochondrial membrane in the cytoplasm, our findings suggest that Acads likely form a complex with mito-Dnmt1 and facilitate its degradation in proximity to the mitochondrial membrane.

Notably, previous studies have demonstrated that abnormal ACADS expression is closely associated with the progression of several cancers, including liver cancer,⁴¹ pancreatic ductal cancer,⁴² and gastric cancer.⁴³ In this study, we identified robust correlations among ACADS, mito-DNMT1, STING signaling, and CRC progression in clinical samples. In particular, ACADS expression was found to be associated with the response of patients with CRC to immunotherapy. Furthermore, structure-based virtual screening and SPR assays identified hypericin as a natural compound that directly binds to and upregulates ACADS. This activation leads to the suppression of CRC growth by establishing an immunosuppressive TME via the mito-DNMT1-mediated cGAS-STING signaling pathway. Hypericin, a photosensitizer, has successfully completed phase III clinical trials for the treatment of mycosis fungoides-cutaneous T-cell lymphoma and has exhibited a favorable safety profile.⁴⁴ In particular, we found that treatment with hypericin improved the immune TME in ex vivo human CRC samples. Therefore, hypericin represents a promising therapeutic approach for patients with CRC.

In summary, we identified and confirmed for the first time that ACADS plays a tumor-suppressive role in CRC. ACADS deficiency contributes to the establishment of an immunosuppressive TME, thereby promoting CRC development. Activation of ACADS with hypericin can impede CRC growth. Mechanistically, we demonstrated that ACADS deficiency inhibited the mtDNA leakage-mediated cGAS‒STING signaling pathway by increasing mito-DNMT1-dependent mtDNA methylation (Fig. 7p).

Materials and methods

The mouse experiments were approved and performed in accordance with the guidelines of the Institutional Animal Care and Use Committee of the Third Military Medical University (TMMU, Chongqing, China) or Jinfeng Laboratory (Chongqing, China). All the mice were housed in a pathogen-free facility with a 12 h light/dark cycle at the TMMU or Jinfeng laboratory. All the mice were provided food and purified water ad libitum. Human CRC tissue samples were collected in accordance with the guidelines of the Second Affiliated Hospital of Army Medical University (Protocol No. 2019-YANDI 103-01) and the Cancer Hospital of the Chinese Academy of Medical Sciences (Protocol No. KY2023166).

Chemicals

Dimethyl sulfoxide (DMSO; #D8371) and sodium butyrate (NaBu; #156-54-7) were purchased from Solarbio (Beijing, China). DMXAA (#117570-53-3), C-176 (#314054-00-7), Cyclosporin A (CsA; #59865-13-3), mitoquinone mesylate (MitoQ; #845959-50-4), ethidium bromide (EtBr; #1239-45-8), decitabine (#2353-33-5), PEG300 (#25322-68-3), Tween 80 (#9005-65-6), MG132 (#133407-82-6), riboflavin phosphate (sodium), (#HY-B0964), Herbacetin 3,8-O-diglucoside, (HY-N10042), Diosmin (HY-N0178), rutin (HY-N0148), DAPTA (HY-P1034), adenosine 5′-diphosphoribose (sodium) (HY-100973A), Bimosiamose (HY-106139), and hypericin (HY-N0453) were obtained from MedChemExpress (MCE, Shanghai, China). VBIT4 (#2086257-77-2) and N-acetylcysteine (NAC; #616-91-1) were acquired from TargetMol (BOS, MA, USA).

Cell culture

The CRC cell lines MC-38, CT-26, and HCT-116 were maintained in our laboratory.^10,45,46 All the cells were authenticated and tested for mycoplasma. Primary mouse macrophages and cell lines were cultured with Dulbecco’s modified Eagle’s medium (DMEM) and high-glucose medium (#C3113, VivaCell, Shanghai, China) supplemented with 10% fetal bovine serum (FBS, #164210-50, Procell, Wuhan, China) at 37 °C in a humidified 5% CO₂ atmosphere.

Mice

Wild-type C57BL/6 and BALB/c mice and BALB/c nude mice were obtained from GemPharmatech (Jiangsu, China). C57BL/6 J mice harboring a loxP-flanked (fl) allele of exon 2 of Acads (Acads^fl/-) were obtained from Cyagen Biosciences (Suzhou, Jiangsu, China). Acads^fl/fl mice were crossed with Vil1-Cre mice (Cyagen Biosciences, Suzhou, Jiangsu, China) via a three-step backcrossing procedure to obtain Acads intestinal-conditional knockout mice (Acads^fl/fl-Vil1^Cre).

Mouse models of subcutaneous tumors

A mouse model was constructed as described in our previous work.¹⁰ Briefly, 100 μL of PBS containing either MC-38 or CT-26 CRC cells (2 × 10⁶) or without cells was subcutaneously injected into the groin of 6-week-old male C57BL/6 or BALB/c mice. Approximately three weeks post-injection, the mice were euthanized, and the subcutaneous tumor nodules were harvested, imaged, and weighed.

Mouse models of peritoneal carcinomatosis

A mouse model was generated as described in our previous studies.⁴⁵ Briefly, 100 μL of PBS containing MC-38 or CT-26 CRC cells (2 × 10⁶) or without cells was intraperitoneally injected into 6-week-old male C57BL/6 or BALB/c mice. Approximately two weeks post-injection, the mice were euthanized, and the peritoneal tumor nodules were harvested, imaged, and weighed.

Establishment of a colitis-associated cancer model

Acads^fl/fl and Acads^fl/fl-Vil1^Cre mice (8–12 weeks) were intraperitoneally injected with 10 mg/kg azoxymethane (AOM; #A5486, Sigma). Five days post-injection, the mice underwent three cycles of 2.5% dextran sodium sulfate (DSS; #160110, MP Biomedicals) treatment. Each cycle consisted of administering DSS in their drinking water for 5 consecutive days, followed by a 2-week period of normal water. The body weights of the mice were monitored every two days. After completing the three cycles, the mice were maintained on a regular diet until day 92 post-AOM injection, at which point the colon was excised for analysis.

Generation of Acads-knockdown and Acads-overexpressing MC-38 cells

Lentiviral vectors for Acads knockdown (targeting sequences: #1: 5′-CGCATCACTGAGATCTACGAA-3′; #2: 5′-CCTGGATTGTGCTGTGAAGTA-3′) with puromycin resistance and for the overexpression of C-terminal HA-tagged Acads with hygromycin B resistance were obtained from Genechem (Shanghai, China). MC-38 cells were infected with the respective lentiviruses following the manufacturer’s instructions. After infection, the cells were treated with the corresponding antibiotics for several days until all noninfected cells were eliminated. The successfully transduced cells were then harvested and validated for their Acad expression levels via qPCR and Western blot analysis.

Generation of mito-Dnmt1 knockout and overexpression in MC-38 cells

Mitochondrial Dnmt1-specific knockout MC-38 cells were generated via the pWST-lenti-CRISPR/Cas9 system (Chongqing Western Biomedical Technology Co., Ltd., Chongqing, China). A guide RNA targeting the uORF sequence (5′-CCCCACTCTCTTGCCCTGTG-3′) was designed to disrupt the MLS. The generated CRISPR/Cas9 lentivirus was used to infect MC-38 cells at 50% confluence. Thirty-six hours post-transfection, puromycin was added to the culture medium to select successfully transduced cells. Positive cells were then dissociated and plated into 96-well plates by limiting dilution to obtain single-cell clones. Once colonies formed, genomic DNA was extracted, and the targeted region was amplified via PCR for cloning and subsequent sequencing validation.

For mitochondrial Dnmt1 overexpression, a lentiviral vector containing the full-length Dnmt1 sequence with its uORF encoding the MLS at the N-terminus and a blasticidin resistance marker was packaged by Sangon Biotech (Shanghai, China). MC-38 cells were infected with the mito-Dnmt1 overexpression lentivirus for 48 h, followed by blasticidin selection for several days until all noninfected cells were eliminated. The successfully transduced cells were harvested and validated for mito-Dnmt1 expression via qPCR and Western blot analysis.

Cell viability assay

Acads-knockdown MC-38 cells and their respective negative control cells (5 × 10³) were seeded in 96-well plates. After the cells adhered to the bottom of the plates, this time point was designated 0 h. The cell counting kit-8 (CCK-8) working solution (#K009-500, Zetalife, Menlo Park, CA, USA) was added to the plates at predetermined time points (0, 12, 24, 36, and 48 h), followed by an additional incubation for 1 h. Subsequently, the optical density (OD) of each well was measured at 450 nm via a Synergy H1 microplate reader (BioTek, Winooski, VT, USA).

Cell proliferation detection

Acads-knockdown MC-38 cells and their respective negative control cells (5 × 10⁴) were planted in 6-cm petri dishes and cultured in 4 mL of complete medium for 7 days. After the incubation period, the cells were trypsinized and resuspended in an equal volume of PBS. The cell density was subsequently quantified via a hemocytometer.

Apoptosis analysis

Acads-knockdown MC-38 cells and their respective negative control cells (1 × 10⁵) were seeded in 12-well plates. Upon reaching 80% confluence, both the suspended cells in the culture medium and the adherent cells were collected and combined for staining via the Annexin V-FITC/PI Apoptosis Detection Kit (#556547, BD Biosciences, San Jose, CA, USA) according to the manufacturer’s protocol. Apoptosis analysis was then performed via a CytoFLEX flow cytometer (Beckman Coulter, Ltd., Miami, FL, USA). Data were acquired and analyzed via CytExpert software (version 2.4.0.28; Beckman Coulter, Inc.; Brea, CA, USA).

Quantitative real-time PCR (qPCR)

mRNAs from experimental cells were extracted with an Eastep^® Super Total RNA Extraction Kit (#LS1040, Promega, Beijing, China). One microgram of total RNA was reverse transcribed into cDNA using PrimeScript™ RT Master Mix (#RR036, TaKaRa, Beijing, China). qPCR was performed in a 20 µL reaction volume containing TB Green premix (#RR820a, TaKaRa, Beijing, China) via the CFX Connect™ Optics Module system (Bio-Rad, Hercules, CA, USA) following the instructions of the kit. Relative gene expression levels were calculated via the 2^−∆∆CT method, with β-actin serving as the housekeeping gene for normalization. The primer sequences are provided in Supplementary Tables 1 and 2.

Western blotting

The CRC cell lines MC-38 or CT-26 were plated in 6-well plates. Upon reaching 70%-80% confluence, the cells were washed with precooled PBS and lysed on ice in cell lysis/IP buffer (#P0013, Beyotime, Shanghai, China) supplemented with protease inhibitors (#HY-K0011, MedChemExpress, Shanghai, China). The protein concentration of the lysates was detected via an enhanced BCA protein assay kit (#P0010, Beyotime, Shanghai, China). All samples were adjusted to an equal concentration and denatured with 5× loading buffer at 100 °C for 10 min. Equal amounts of protein from each sample were then loaded into an 8% or 10% sodium dodecyl sulfate‒polyacrylamide gel (SDS‒PAGE) for electrophoresis. The resolved proteins were transferred onto PVDF membranes, which were subsequently blocked with 5% nonfat milk in Tris-buffered saline containing 0.1% Tween 20 (TBST) for 1 h at room temperature. The membranes were washed with TBST and incubated overnight at 4 °C with primary antibodies, including ACADS (#AP75024, Abcepta, Jiangsu, China), DNMT1 (#NB100-56519AF594, NOVUS), cGAS (Cat#31659, Cell Signaling Technology), p-STING (#PA5-105674, Invitrogen), STING (#A3575, ABclonal), p-TBK1 (#AP1026, ABclonal), TBK1 (#AF8103, Beyotime), p-IRF3 (#4947, Cell Signaling Technology), IRF3 (#4302, Cell Signaling Technology), BAX (#A0207, ABclonal), VDAC1 (#A19707, ABclonal), VDAC3 (#55260-1-AP, Proteintech), HSP60 (#AF0186, Beyotime), PHB1 (#10787-1-AP, Proteintech), Lamin B1 (#ab16048, Abcam), HA (#51064-2-AP, Proteintech), TFAM (#A13552, ABclonal), NRF1 (#12482-1-AP, Proteintech), PGC1α (#66369-1-Ig, Proteintech), Twinkle (#18793-1-AP, Proteintech), POLG (#AP14948B, Abcepta), and β-actin (#66009-1-Ig, Proteintech). Following incubation with primary antibody, the membranes were washed 3 times with TBST and then hybridized with the corresponding secondary antibody at room temperature for 2 h. In certain instances, the membrane was subjected to washing with primary and secondary antibody removal solution (#P0025, Beyotime), followed by reincubation with either the indicated antibody or a loading control β-actin antibody, along with the corresponding secondary antibodies. The chemiluminescence signals of the target proteins were collected via a chemiluminescence instrument after the addition of the enhanced chemiluminescence substrate. Protein band density was quantified via ImageJ software (version 1.54g; https://imagej.net/ij/).

Cytoplasmic mtDNA analysis

Experimental cells (5 × 10⁵) were seeded in 6-well plates. When the cells reached 70%-80% confluence, they were trypsinized, collected by centrifugation at 400 × g for 5 min, and then resuspended in PBS. The cell suspension was equally divided into two aliquots. The first aliquot was lysed on ice using 1% IGEPAL CA-630 (#ST2045, Beyotime, Shanghai, China) for 15 min, followed by centrifugation at 16,000 × g for 15 min at 4 °C. The resulting supernatant, representing the cytoplasmic fraction, was transferred to a new tube for subsequent genomic DNA extraction via a commercial kit (TIANGEN, #DP304-03). The second aliquot was processed for total genomic DNA extraction following the manufacturer’s protocol. To quantify the mtDNA content in the cytoplasmic fraction, qPCR was performed using primers specific for the mitochondrial ND1 and D-loop, while the nuclear Tert gene from total genomic DNA served as an internal reference for normalization. The primer sequences are shown in Supplementary Table 1.

Mitochondrial ROS detection

Mitochondrial ROS levels were assessed via the MitoSOX Red mitochondrial superoxide indicator (#1003197-00-9; MedChemExpress, Shanghai, China) according to the manufacturer’s protocol. In brief, experimental cells (1 × 10⁵) were seeded in 12-well plates and allowed to reach 70–80% confluence. The cells were then trypsinized, harvested by centrifugation at 600 × g for 5 min, and resuspended in 1 mL of culture medium containing 100 nM MitoSOX Red working solution. Following a 30 min incubation at room temperature in the dark, the cells were centrifuged at 400 × g for 5 min and washed twice with PBS. The cell suspension was subsequently filtered through a 200-mesh filter and subjected to flow cytometry analysis via appropriate excitation/emission settings (510/580 nm) for MitoSOX Red detection. The fluorescence intensity was quantified via CytExpert software (version 2.4.0.28; Beckman Coulter, Inc.; Brea, CA, USA).

Mitochondrial membrane potential (MMP) assay

The MMP was measured via the fluorescent probe tetramethylrhodamine methyl ester (TMRM) (#115532-49-5; MedChemExpress, Shanghai, China) following the manufacturer’s protocol. Briefly, experimental cells were seeded in 12-well plates and cultured until 80% confluence. The cells were then trypsinized and collected by centrifugation at 500 × g for 5 min at room temperature. The cell pellet was resuspended in 1 mL of serum-free medium containing 20 μM TMRM working solution and incubated at 37 °C in a 5% CO₂ atmosphere for 30 min. After incubation, the cells were washed with PBS following centrifugation at 500 × g for 5 min and finally resuspended in 200 μL of PBS. Flow cytometry analysis was performed using a 488 nm excitation laser with fluorescence emission collected at 570 nm (±10 nm) through the appropriate bandpass filter. Data acquisition was performed via a CytoFLEX flow cytometer (Beckman Coulter, Ltd., Miami, FL, USA), whereas data analysis was conducted via CytExpert software (version 2.4.0.28, Beckman Coulter, Inc., Brea, CA, USA).

ATP content assay

The intracellular ATP levels were quantified via an enhanced ATP assay kit (#S0027; Beyotime, Shanghai, China) following the manufacturer’s instructions. In brief, experimental cells cultured in 6-well plates were washed twice with PBS upon reaching 80% confluence, followed by the addition of lysis buffer. The cell lysates were collected via centrifugation at 12,000 × g for 5 min at 4 °C. The resulting supernatants were then mixed with the ATP detection working solution at a 1:1 ratio (v/v). Luminescence intensity was measured via a multimode microplate reader (SpectraMax iD5, Molecular Devices, USA). ATP concentrations were determined via interpolation from a standard curve generated with known ATP concentrations (0.1–10 μM), and the values were normalized to the total protein content determined via a BCA assay.

Mitochondrial calcium level analysis

The experimental cells were seeded in 6-well plates and cultured until they reached 80% confluence. The cells were then trypsinized and harvested via centrifugation at 500 × g for 5 min at room temperature. Following washing with PBS, the cell pellets were resuspended in PBS containing 10 μM Rhod-2 AM (#145037-81-6, MedChemExpress, Shanghai, China) and incubated at room temperature for 30 min in the dark. After incubation, the cells were collected by centrifugation at 500 × g for 5 min and washed twice with PBS. Finally, the cells were resuspended in PBS for fluorescence intensity analysis via flow cytometry. The data were further analyzed via FlowJo (v10.8.1).

Liquid chromatography‒tandem mass spectrometry (LC‒MS/MS) and coimmunoprecipitation assay

LC‒MS/MS and coimmunoprecipitation assays were performed following the protocol established in our previous study.⁴⁷ In brief, 4 μg of specific antibodies against Acads (#16623-1-AP, Proteintech) or Dnmt1 (#NB100-56519AF594, Novus Biologicals) and their respective IgGs were preincubated with magnetic beads (Sera-Mag SpeedBead Protein A/G, 17152104010150, GE) from a 40 μL suspension at 4 °C overnight. After incubation, the bead‒antibody complex was washed three times with cell lysis/IP buffer (#P0013, Beyotime, Shanghai, China). Subsequently, 500 μg of the protein lysate from MC-38 cells was added to the mixture, which was subsequently incubated overnight at 4 °C with gentle rotation. The protein bound to the beads was then eluted via 2× protein loading buffer at 100 °C for 10 min, and the eluate was subjected to SDS‒PAGE and stained with Coomassie brilliant blue G250 (#P0003S, Beyotime, Shanghai, China). The gels containing all the proteins from the IgG, Acads, and Dnmt1 groups were subsequently sent to Shanghai Applied Protein Technology Co., Ltd. (Shanghai, China) for LC‒MS/MS analysis. Additionally, the eluate was further separated and analyzed by Western blotting using antibodies against Acads (#16623-1-AP, Proteintech), Dnmt1 (#NB100-56519AF594, Novus Biologicals), or ubiquitin (#3933, Cell Signaling Technology) to perform the coimmunoprecipitation assay.

Protein expression, purification, and in vitro protein binding assay

The coding sequences of Acads and Dnmt1 were cloned and inserted into the glutathione S-transferase (GST)-containing pGEX-4T-3 plasmid. The recombinant plasmids, along with the empty vector, were transformed into E. coli BL21 cells. The transformed cells were cultured in 2× YT medium at 37 °C with shaking until they reached the logarithmic growth phase (OD₆₀₀ = 0.4–0.6). Protein expression was induced by adding isopropyl β-D-1-thiogalactopyranoside (IPTG) to a final concentration of 1 μM, followed by incubation at 20 °C for 8 h with shaking at 200 rpm. After induction, the bacterial cells were resuspended in PBS supplemented with protease inhibitor cocktails and subjected to ultrasonic lysis on ice to disrupt the cell membranes. The cell lysates were collected via centrifugation at 12,000 × g for 5 min at 4 °C. The supernatant was then incubated with glutathione Sepharose 4B beads (#17-0756-01, GE Healthcare, Sweden) at 4 °C overnight under gentle rotation. Following incubation, the supernatant was discarded, and the agarose beads were thoroughly washed five times with cold PBS to remove nonspecifically bound proteins. For subsequent analysis, the washed agarose beads were used for a second round of immunoprecipitation. The eluate was incubated overnight at 4 °C with gentle rotation using lysates from HEK293T cells transfected with either a Dnmt1-Flag plasmid or an empty control plasmid. After incubation, the beads were collected and washed five times with cold lysis buffer. Finally, the beads were eluted with glutathione elution buffer and mixed with 2× loading buffer for Western blot detection.

Immunofluorescence and superresolution imaging

The immunofluorescence assay was performed following the protocol established in our previous study.⁴⁷ Briefly, MC-38 cells were seeded onto glass coverslips placed in 12-well plates. Upon reaching 70% confluence, the cells were gently rinsed with PBS and fixed with 4% paraformaldehyde at room temperature for 15 min. Following fixation, the cells were treated with 5% bovine serum albumin (BSA) prepared in 0.3% Triton X-100 solution to block nonspecific antigen binding sites, and this blocking step was carried out at 37 °C for 30 min. The cells were then incubated overnight at 4 °C with primary antibodies targeting Hsp60 (#A85438, Antibodies), Acads (#16623-1-AP, Proteintech), Dnmt1 (#NB100-56519AF594, Novus Biologicals), 5-Methylcytosine (5-mC; #39649, Proteintech), or DNA (#CBL186, Merck Millipore). After thorough washing with PBS, the cells were exposed to a fluorophore-conjugated secondary antibody for 2 h at room temperature in a dark and humidified chamber. Nuclei were counterstained with Antifade Mounting Medium with DAPI (#P0131, Beyotime, Shanghai, China) in the dark for 20 min. Finally, the stained samples were imaged and captured via either a laser-scanning confocal microscope (IXplore SpinSR, Olympus) or a Multi SIM X superresolution microscope (NanoInsights, Beijing, China) for superresolution imaging.

Proximity ligation assay (PLA)

MC-38 cells were seeded onto glass coverslips placed in 12-well plates. Upon reaching 70% confluence, the cells were gently rinsed with PBS and fixed with 4% paraformaldehyde at room temperature for 15 min. After fixation, the cells were washed three times with PBS, and the PLA was performed following the protocol provided in the Duolink In Situ Red Starter Kit Mouse/Rabbit Kit (#DUO92101, Sigma‒Aldrich). Next, a blocking solution was applied to prevent nonspecific binding of the cellular antigens by incubation at 37 °C for 1 h. Antibodies against Acads (#16623-1-AP, Proteintech) and Dnmt1 (#NB100-56519AF594, Novus Biologicals), either individually or in combination, were prepared in Probemaker PLA Probe Diluent and incubated with the cells overnight at 4 °C. Next, the PLA probes were ligated to their respective antibodies and amplified in the appropriate buffer at 37 °C for 1 h. The nuclei were then stained with Duolink In Situ Mounting Medium containing DAPI and incubated for 20 min at room temperature in the dark. Finally, images were captured via a confocal microscope (IXplore SpinSR, Olympus).

Mitochondrial isolation

Mitochondria were isolated via a mitochondria isolation kit (#89874; Thermo Fisher Scientific) following the manufacturer’s protocol. Briefly, experimental cells were cultured in 15-cm petri dishes. Upon reaching 80% confluence, the cells were gently washed with PBS, trypsinized, and collected by centrifugation at 850 × g for 2 min. The cell pellet was resuspended in ice-cold Reagent A, followed by the addition of other isolation reagents and thorough mixing. The lysate was then centrifuged at 700 × g for 10 min to remove nuclei and cellular debris. The supernatant, containing mitochondria, was further centrifuged at 3000 × g for 15 min to pellet the mitochondria. The resulting pellet represented the mitochondrial fraction, while the supernatant contained the cytoplasmic components.

mtDNA methylation analysis

The analysis of mtDNA methylation was performed via the MethylFlash^TM Global DNA Methylation (5-mC) ELISA Easy Kit (Colorimetric) (#P-1030, EpigenTek), which strictly adhered to the manufacturer’s protocol. In brief, mtDNA was extracted from isolated mitochondria via a genomic DNA extraction kit (TIANGEN, #DP304-03). An equal amount of mtDNA (100 ng) was diluted with binding solution and then transferred to strip wells with high DNA affinity, followed by incubation at 37 °C for 1 h. Subsequently, each well was washed with washing buffer, and a 5-mC detection complex solution was added and incubated at room temperature for 50 min. After another wash with diluted washing buffer, the developer solution was added to each well. Following a 4 min reaction, the stop solution was added to halt the enzymatic reaction, and the absorbance was measured at 450 nm via a microplate reader (SpectraMax iD5, Molecular Devices, USA). The percentage of 5-mC was determined on the basis of a standard curve.

Hematoxylin and eosin (H&E) staining

H&E staining of the tissues was performed following the protocol established in our previous study.¹⁰ Briefly, the colorectum of each mouse was rinsed with PBS and prepared via the Swiss-roll technique. The tissue was then fixed in 4% paraformaldehyde and embedded in paraffin. The paraffin-embedded blocks were subsequently sectioned into slices, which underwent a series of processing steps, including dewaxing, dehydration with graded ethanol solutions, and rehydration with distilled water. Following pretreatment with a prestaining solution, the tissue sections were stained with H&E solution (#C0105, Beyotime, Beijing, China). After staining, the slides were dehydrated and mounted with a coverslip. The stained slides were finally examined and imaged via a light microscope.

Immunohistochemistry and tyramide signal amplification (TSA)-based multiple immunohistochemistry

The rehydrated slides were subjected to high-pressure-induced epitope retrieval in citrate buffer (pH 6.0). After cooling to room temperature, the slides were washed three times with PBS and treated with 3% H₂O₂ in the dark for 25 min to block endogenous peroxidase activity. The slides were subsequently incubated with 10% rabbit serum or 3% BSA at room temperature for 30 min to block nonspecific antigen binding sites, followed by overnight incubation with the primary antibody at 4 °C in a humidified chamber. After being washed three times with PBS, the slides were incubated with an HRP-labeled secondary antibody for 1 h at room temperature. For conventional immunohistochemistry, the slides were washed three times with PBS and treated with DAB substrate to develop the chromogenic reaction. For triple or quadruple immunohistochemistry, the slides were incubated with fluorophore-labeled TSA at room temperature for 10 min. After incubation, the slides were washed three times with TBST and subjected to microwave-induced epitope retrieval in citrate buffer. This was followed by two or three additional cycles, each comprising blocking nonspecific binding sites, incubation with primary and secondary antibodies, addition of TSA with a distinct fluorophore, and microwave-induced epitope retrieval to stain two or three additional target proteins. Upon completion of the staining process, the slides were treated with DAPI to stain the nuclei, followed by the application of an autofluorescence quencher and mounting with a coverslip. Finally, the results were visualized and captured via a fluorescence microscope (CLIPSE C1, Nikon) and a slide scanner (Pannoramic MIDI, 3DHISTECH). The antibodies used for immunohistochemistry were as follows: Ki67 (#GB11612, Servicebio), STING (#GB111415, Servicebio), GR1 (#GB11229, Servicebio), CD3 (#GB12014, Servicebio), F4/80 (#GB113373, Servicebio), FOXP3 (#GB112325, Servicebio), LY6C (#GB115601, Servicebio), LY6G (#GB11224, Servicebio), CD25 (#GB11612, Servicebio), ACADS (#TA800429S, Thermo Fisher Scientific), DNMT1 (#GB11328, Servicebio), and HSP60 (#GB15243, Servicebio). T-cell populations were defined on the basis of specific surface marker profiles: total T cells as CD3⁺; CD4⁺ T cells as CD3⁺CD8^-CD4⁺; and CD8⁺ T cells as CD3⁺ CD4^-CD8⁺. Macrophage populations were categorized as follows: total macrophages, F4/80⁺; M1-like cells, F4/80⁺CD206^-iNOS⁺; and M2-like cells, F4/80⁺iNOS^-CD206⁺. MDSCs were classified as total MDSCs as Gr1⁺; polymorphonuclear MDSCs (PMN-MDSCs) as CD11b⁺LYG⁺LY6C^-; and monocytic MDSCs (M-MDSCs) as CD11b⁺LYG^-LY6C⁺. Regulatory T cells (Tregs) were identified as CD3⁺CD25⁺FOXP3⁺ cells. In addition, the positive cell density of different immune cells and the percentage of colocalized positive cells within tumors were analyzed via the HALO software platform (Indica Labs, USA).

Fluorescence-activated cell sorting (FACS) of human CRC, mouse tumor, and adipose tissues

The human CRC samples and subcutaneous tumors from the mice were minced into 1 mm³ pieces and then digested in 20 mL of RPMI-1640 medium or DMEM supplemented with 1% FBS, 1 mg/mL collagenase IV (#C5138, Sigma), 0.1 mg/mL hyaluronidase (#BS171-1 g, Biosharp Life Science), or 0.1 mg/mL DNase I (#10104159001, Merck Millipore) at 37 °C for 45 min. For adipose tissue processing, epididymal fat pads were harvested, minced, and digested in 10 mL of DMEM containing 1% FBS and 1 mg/mL collagenase IV (#C5138, Sigma) at 37 °C for 45 min. Following complete digestion, the cell suspensions were filtered through a 70 μm strainer and centrifuged at 500 × g for 10 min. The pelleted cells were resuspended in 40% Percoll solution (#17089109, Cytiva) and carefully layered onto 70% Percoll solution, followed by centrifugation at 800 × g for 20 min. The intermediate layer containing the target cells was collected and treated with red blood cell lysis buffer (R1010, Solarbio, China) to remove erythrocytes. After lysis, the cells were centrifuged at 500 × g for 5 min, resuspended in PBS, and incubated with fluorochrome-conjugated antibodies targeting cell surface markers for 45 min at 4 °C. The cells were then washed twice with PBS at room temperature, subsequently resuspended, and filtered through a cell strainer. FACS analysis was then performed using a FACSVerse C6 flow cytometer (BD LSRFortessa). For intracellular marker staining, the cells were fixed and permeabilized before antibody incubation. The antibodies used for FACS analysis were as follows: PerCP/Cy5.5 anti-mouse CD45 (#103132, BioLegend), FITC anti-mouse CD90.2 antibody (#105338, BioLegend), APC/Cy7 anti-mouse F4/80 antibody (#123118, BioLegend), PE anti-mouse CD11c isotype control antibody (#117308, BioLegend), and APC anti-mouse CD206 isotype control antibody (#141708, BioLegend); APC anti-mouse CD11b antibody (#101212, BioLegend), Percp5.5 anti-mouse Gr1 antibody (#108428, BioLegend), and FITC anti-mouse LY6C antibody (#128006, BioLegend); Alexa Fluor 700 anti-mouse CD45 antibody (#103128, BioLegend), PE anti-mouse CD3 antibody (# 100206, BioLegend), APC/Cy7 anti-mouse CD4 antibody (#116020, BioLegend), Percp5.5 anti-mouse CD8a (#100734, BioLegend), and APC anti-mouse IFNγ (#505810, BioLegend); and APC anti-mouse CD25 antibody (#102012, BioLegend), FITC anti-mouse Foxp3-AF488 antibody (#126406, BioLegend). Brilliant Violet 421™ anti-human CD45 antibody (#368521, Biolegend), FITC anti-human CD11B antibody (301329, Biolegend), APC anti-human CD68 antibody (#333813, Biolegend), Brilliant Violet 510™ anti-human CD3 antibody (#317331, Biolegend), PE/Cyanine7 anti-human IFN-γ antibody (#502527, Biolegend), APC/Fire™ 810 anti-human CD25 antibody (#356149, Biolegend), and PerCP-Cy5.5 anti-human FOXP3 antibody (#320212, Biolegend) were used. T-cell populations were identified via the following surface marker profiles: total T cells, CD45⁺CD3⁺; CD4⁺ T cells, CD45⁺CD3⁺CD8^-CD4⁺; and CD8⁺ T cells, CD45 + CD3 + CD4^-CD8⁺. Macrophage populations were characterized as follows: total macrophages, CD90.2^-CD45⁺F4/80⁺; M1-like cells, CD90.2^-CD45⁺F4/80⁺CD206^-CD11c⁺; and M2-like cells, CD90.2^-CD45⁺F4/80⁺CD11c^-CD206⁺. The populations of MDSCs were classified as total MDSCs as CD11b⁺Gr1⁺; PMN-MDSCs as CD11b⁺LYG⁺LY6C^-; and M-MDSCs as CD11b⁺LYG^-LY6C⁺. Treg cells were identified as CD45⁺CD3⁺CD4⁺CD25⁺FOXP3⁺ cells.

RNA sequencing and DNA methylation sequencing

The experimental cells were seeded into 6-well plates and cultured until they reached 80% confluence. Next, the cells were gently washed twice with ice-cold PBS. Subsequently, 500 µL of TRNzol Universal reagent (#DP424, TIANGEN, Beijing, China) was added to lyse the cells. The resulting cell lysates were then collected, immediately placed on dry ice, and promptly transported to TSINGKE (Beijing, China) for transcriptome sequencing. Concurrently, MC-38 cells with Acads knockdown or mito‑Dnmt1 overexpression, along with their corresponding controls, were harvested. Genomic DNA was then extracted in parallel from two sources: (1) whole‑cell genomic DNA was sent to the BGI Center (Shenzhen, Guangdong, China) and TSINGKE (Beijing, China) for DNA methylome sequencing and BSP, respectively; and (2) mtDNA isolated from purified mitochondria was sent to TSINGKE (Beijing, China) for methylation analysis via BSP and third‑generation nanopore sequencing.

Transcriptomic data analysis and visualization

Transcriptomic sequencing data from four distinct datasets (GSE231709, GSE86299, GSE44988, and GSE155777) pertaining to AOM/DSS-induced mouse spontaneous colon cancer models were retrieved from the GEO database. These datasets were subsequently analyzed via the GEO2R online tool to identify differentially expressed genes (DEGs) by comparing the treatment groups to their respective controls. DEGs were defined as those with an absolute log2-fold change (|log2FC|) > 1 and an adjusted p-value < 0.05. Next, an online intersection analysis (http://bioinformatics.psb.ugent.be/webtools/Venn/) was carried out to compare these DEGs with those identified from The Cancer Genome Atlas (TCGA)-Colon Adenocarcinoma (COAD) database, the latter of which were processed via the DESeq2 package (version 1.40.2) in R software (version 4.4.1). Additionally, genes associated with fatty acid metabolism from the HALLMARK_FATTY_ACID_METABOLISM gene set in the MsigDB database were included in the subsequent analysis. The common DEGs identified through this process were then mapped, and their expression changes were visualized via a pheatmap generated via the heatmap package (version 1.0.12) in R software (version 4.4.1).

Single-cell RNA-Seq data analysis and visualization

Single-cell RNA sequencing data (GSE178341)⁴⁸ from human CRC patient samples were retrieved from the GEO database. The data were subjected to quality control, standardization, and dimensionality reduction clustering analysis via the Seurat package (version 5.1.0) in R software. Tumor cells were subsequently isolated to assess ACADS expression levels. All analyses were performed via default parameters unless otherwise specified. Cell clustering was achieved via the ‘FindClusters’ function in Seurat, with cell types and subtypes identified through nonlinear dimensional reduction (t-SNE) and annotated on the basis of established cell type-specific markers.^48,49 From the dataset, we identified 10 pairs of CRC tissue samples with distinct ACADS expression profiles in tumor cells. The low ACADS expression group comprised the following samples: C103_T_1_1_0_c1_v2, C114_T_1_1_0_c1_v2, C119_T_0_2_0_c1_v2, C125_T_1_1_0_c1_v2, C134_T_0_3_0_c2_v2, C136_T_1_1_0_c1_v2, C145_T_1_1_0_c1_v2, C159_T_1_1_0_c1_v3, C165_T_0_0_0_c1_v3, and C166_T_0_0_0_c1_v3. The high ACADS expression groups included C106_T_1_1_0_c1_v2, C122_T_1_1_0_c1_v2, C130_TA_1_1_0_c1_v2, C130_TB_1_1_0_c1_v2, C132_T_1_1_0_c1_v2, C137_T_1_1_0_c1_v2, C139_T_1_1_0_c1_v2, C139_T_1_1_0_c2_v2, C146_T_1_1_0_c1_v2, and C164_T_1_1_0_c1_v3. These 10 sample pairs were subsequently integrated and analyzed via the Seurat package, with batch effects corrected via the Harmony package (version 1.2.1). Following confirmation of differential ACADS expression in tumor cells between the groups, we examined changes in cell proportions within each group on the basis of the established clustering and grouping criteria, followed by statistical analysis.

Immune infiltration and immune epigenetic scoring analysis

Immune infiltration analysis was conducted with the xCell tool (version 1.1.0) in R software according to the expression data of the ACADS, DNMT1, and STING signaling pathways from the TCGA-COAD cohort. The genes associated with immune infiltration patterns were analyzed via multiple algorithms, including CIBERSORT, TIDE, EPIC, TIMER, and XCELL, via the TIMER2.0 online platform (http://timer.cistrome.org/). Additionally, the immune epigenetic score of ACADS patients was evaluated by utilizing the IPS of COAD patients who underwent anti-PD-1/CTLA-4 immunotherapy, as provided in the TCIA database (https://tcia.at/home).

Analysis of tissue microarrays

CRC and adjacent paracancerous tissues from the Cancer Hospital of the Chinese Academy of Medical Sciences (Protocol No. KY2023166) were processed into tissue microarrays and immunohistochemically stained with antibodies against ACADS (#TA800429S, Thermo Fisher Scientific), DNMT1 (#GB11328, Servicebio), and HSP60 (#GB15243, Servicebio). The staining results were analyzed via the HALO software platform (Indica Labs, USA) to quantify key parameters, including positive cell counts, positive areas, and colocalization. These metrics were subsequently used to calculate the ratio of positive cells to the positive area. Additionally, the H score for ACADS expression across different stages of CRC tissues was determined via the following formula: H score = ∑(pi × i) = (percentage of weakly stained cells × 1) + (percentage of moderately stained cells × 2) + (percentage of strongly stained cells × 3). Here, i represents the intensity grading of the positive area: negative (no staining, scored as 0), weakly positive (light yellow, scored as 1), moderately positive (brownish yellow, scored as 2), and strongly positive (dark brown, scored as 3). The term pi denotes the percentage of the positive area corresponding to each intensity grade.

Structure-based virtual screening

Virtual screening for Acads-specific binders was performed by MedChemExpress (Shanghai, China). Briefly, the 3D structure of Acads (PDB ID: 2VIG) was retrieved from the Protein Data Bank, followed by structure optimization, binding site determination, and grid construction. Moreover, 2D structures of small molecules from the MedChemExpress library were converted into 3D conformations. Then, molecular docking was performed on the basis of the geometric and energetic complementarity. The top 15% of the compounds identified in the primary screening were subjected to secondary screening in standard-precision mode. The top 15% of hits from this secondary screening were subsequently further evaluated in a tertiary screening via the extraprecision mode. Finally, the binding interactions and structural stability between Acads and the candidate compounds were manually validated.

Surface plasmon resonance (SPR)

The CM5 sensor chip (Cytiva, BR-1005-30) was immobilized with recombinant ACADS protein (amino acids 25–412) according to standard amine coupling protocols. The binding kinetics between ACADS and hypericin were evaluated via SPR in manual injection mode. A serial dilution of hypericin was prepared for kinetic analysis. The running buffer was delivered at a flow rate of 30 μL/min, with both the association and dissociation phases lasting 150 s each. Multicycle kinetic analysis was conducted, and sensorgrams were plotted with time (seconds) on the x-axis and response units (RUs) on the y-axis. Kinetic parameters were calculated via BIAcore Insight evaluation software (Cytiva, Marlborough, MA, UK) via the steady state affinity model, yielding the equilibrium dissociation constant (KD, M), association rate constant (Ka, 1/Ms), and dissociation rate constant (Kd, 1/s).

Statistical analysis

Statistical analyses were conducted via GraphPad Prism 8. The data from the animal experiments and clinical studies are presented as the means ± SEMs, whereas the other results are expressed as the means ± SDs. Comparisons between two groups were performed via Student’s t tests, and comparisons among multiple groups were analyzed via one-way analysis of variance (ANOVA). Survival curves were evaluated via the log-rank test. The relationship between ACADS and mito-DNMT1 in terms of mean fluorescence intensity (MIF) scores was assessed via Pearson’s correlation analysis. All the experiments were performed with biological replicates and were repeated at least three times. The mice were randomly assigned to experimental groups. A p-value of <0.05 was considered statistically significant.

Author contributions

H.M. and Y.X. conceived the project and were responsible for all phases of the research. F.Y., M.W., and S.H. conducted the majority of the experiments and data analyses. X.G. and W.X. performed the clinical sample collection. K.Z. and Y.Z. constructed several cell lines. L.L., Y.G., S.Z., N.L., T.Z., and H.Y. assisted with data collection and interpretation. The manuscript was drafted by Y.X., F.Y., M.W., and H.M. and was revised and approved by all the authors. All the authors have read and approved the article.

Data availability

All data supporting the findings of this study are available within the main manuscript, the Supplementary Materials, or from the designated public repository. Specifically, the RNA sequencing data generated in this study are provided in the Supplementary Materials, while the whole-genome DNA methylome sequencing data have been submitted to the China National Center for Bioinformation under reference number GRA030170.

Competing interests

Hongming Miao is the editorial board member of Signal Transduction and Targeted Therapy, but he has not been involved in the process of manuscript handling. The authors declare that they have no conflicts of interest.

Supplementary information

The online version contains Supplementary Material available at 10.1038/s41392-026-02675-8.