The oncometabolites D2HG and fumarate induce muscle atrophy
After consulting published articles 8–11, we listed cancer cachexia-related metabolites, identifying 157 cachexia-related metabolites (Table S1). A total of 66 common metabolites were selected after excluding metabolites that changed in only one project. The metabolites were matched with the Human Metabolome Database (www.hmdb.ca) and the function annotations were included. Based on muscle-related functional annotation and accessibility of these metabolites, nineteen candidate metabolites were selected for the in vitro experiment (Fig. 1A, Table S2). The concentrations of these metabolites were listed based on reference (Table S2).
Well-differentiated myotubes were treated with these metabolites for in vitro screening, and myotubes diameters were measured (Fig. 1B-C). Fumarate (Fum), lactate, D2HG, pyruvate, adenosine, inosine, carnosine, phenylacetate, 1-methylhistidine, 3-methylhistidine, 4-hydroxyproline, and creatine induced different degrees of myotubes atrophy. To confirm the effect of metabolites mediated muscle wasting, we used the widely used in vitro muscle atrophy system and used mRNA expression of Trim63 and Fbxo32 upregulation as the index of muscle proteolysis 21. When well-differentiated myotubes were treated with these 19 metabolites for 48 hours, succinate, fumarate, D2HG, 1-methylhistidine, 3-methylhistidine, and 4-hydroxyproline treatment resulted in significantly increase of mRNA expression of Trim63 and Fbox32 (Fig. 1D-E).
Genome instability and gene mutations of cancer cells drive the gain or loss of certain enzyme functions. On the other hand, they exert pro-oncogenic capabilities through metabolic reprogramming 22. We searched the oncogenes and common cancer-based genetic alterations in the database of CancerGenetics Web (http://www.cancerindex.org/geneweb/), Cancer Gene Census(https://cancer.sanger.ac.uk/census), oncogene database (http://ongene.bioinfo-minzhao.org/), and database of Mutational Signatures (https://cancer.sanger.ac.uk/cosmic). A total of 2739 genes were included in the database (Table S3). Genotype/phenotype/disease associations were input based on comprehensive biological context for OMICs data interpretation 23. To screen the metabolites-related genes, 84 genes were selected based on the association of gene annotation and metabolism (Table S4). Next, we searched the association between these genes and metabolic phenotype. It was found SDH and IDH1 were related with metabolites fumarate and D2HG. Thus, we further investigated the IDH1-mediated D2HG accumulation and SDH-mediated fumarate (Fum) accumulation (Fig. 1F). We treated the C2C12 myotube with D2HG and Fum for 72 h and confirmed morphological changes. The total RNA was then extracted for transcriptome sequencing. Heatmaps showed distinct transcriptional characteristics after D2HG treatment based on fragments per kilobase of exon model per million mapped fragments (FPKM) (Fig. 1G). The Fum treatment and NTC groups showed similar transcriptional features. These results confirmed that D2HG-induced myotube wasting occurs through proteolysis and with distinct transcriptional characteristics.
IDH1 mutation mediates high concentrations of D2HG in cancer patients with cachexia and in vivo cachexia mice model
D2HG is produced by mutant IDH1 15, an unique R132H/C/G mutation at rs121913500 (Fig S1A). We obtained the IDH1 mutation information from patients, and nineteen out of 149 cancer patients were confirmed as an IDH1 mutation with PCR sequencing (Table S5). Then serum D2HG levels were measured using high-performance liquid chromatography-tandem mass spectrometry. The serum D2HG levels were higher in IDH1-mutated patients compared to controls (Fig. 2A). This was consistent with previous studies that abnormally elevated levels of D2HG to millimolar per gram of tissue in patients with a single mutant copy of IDH1 6. We defined cancer cachexia as weight loss >5% in the past 6 months or weight loss >2% in the past 6 months and a body mass index <20 kg/m2 2. Among 19 IDH1-mutated patients, 8 had cancer cachexia. The serum levels of D2HG were higher in cancer cachexia patients compared to those stable-weight cancer patients (Fig. 2B).
Next, we evaluated the IDH1 genotype-related cancer phenotype based on the public biological information resources of the TCGA database. The average alteration frequency of IDH1 was 5% in all cancer types, and the most common alteration type was mutation (Table S6). Substitution of the arginine 132 by histidine (R132H) accounted for >80% of all IDH mutations. To evaluate the overall survival of IDH1 alteration on pan-cancer, a total of 10,802 patients from 32 studies with mutation data were included after excluding 10 overlapping patients. A total of 627 IDH1-altered patients showed poorer overall survival than the IDH1-unaltered cancer patients (Fig S1B). Though IDH mutations have a different prognostic value depending on the cancer type, we still noted the IDH1 mutation mediated poor survival in pan-cancer patients. Moreover, the transcripts per million in different cancers ranged from 20 to 200, and cancer patients were higher than control patients (Fig S1C). We thus tested the hypothesis that IDH1 mutation mediated D2DH accumulation contributed to cancer cachexia progression.
IDH1 mutation at the amino acid arginine 132 (R132) is unique since it is localized in the isozymes’ substrate-binding site 14. IDH1 catalyzes the oxidative decarboxylation of isocitrate to ketoglutarate with NADPH’s concomitant production. When IDH1 mutation occur in R132H, it hindered the hydrophilic interactions between the arginine and both α-carboxylate, and thus mutated IDH1 has gained a function that converts keto-glutarate and NADPH into D2HG and NADP. Based on the mutation frequency, we imported R132H into a cancer cell and used an in vivo experiment to evaluate whether IDH1 mutation mediated D2HG accumulation during cancer cachexia progression (Fig. 2C). BALB/c mice bearing CT26.wt colon adenocarcinoma cells are the most commonly used cancer cachexia model 24. We first cloned IDH1-R132H into the lentivirus plasmid pLV-EF1α-FLAG-IRES-Puro and produced pLV-EF1α-IDH1-R132H-FLAG-IRES-Puro lentiviruses, which were used to infect CT26 colon adenocarcinoma cells. The amplification and the protein expression of IDH1 R132H were confirmed by sequencing the cDNA and Western blot (Fig S2A, Fig S2B). No IDH1 protein changes were observed in the wild-type group, while there was a higher expression of IDH1-R132H protein in the IDH1-mutated group (Fig S2B). Consistent with protein expression, the mRNA expression of IDH1-R132H was found higher (Fig S2C). Then, 2 million CT26 cancer cells were subcutaneously transplanted into the right flanks of male mice. The survival of the mice bearing IDH1-mutated cancer was shorter than in control mice (Fig. 2D). Cachexia was defined as lean body weight (mice without transplanted tumor) loss of more than 5% from the lean body weight change curve. For the IDH1-R132H mutation cancer-bearing mice, cancer cachexia syndrome occurred at DPI 17, and the average lean body weight decreased by 5.4% (from 26.83+1.12 g to 25.45+0.98 g) (Fig. 2E). However, cachexia was observed at DPI 22 in the wild-type tumor group since there was a decrease in lean body weight, and the bodyweight was decreased by 5.0% (from 27.09±1.10 g to 25.81±1.23 g). These results implied that the mutation of IDH1 in CT26 cells could accelerate the growth of tumors and induce the wasting of body weight.
To reveal the contributor of lean body weight loss, we measured the mass of typical skeletal muscle. Compared with the control group, skeletal muscle gastrocnemius and tibialis anterior loss in mice bearing IDH1-mutated cancer was 26.1% and 16.3%, respectively (Fig. 2F). Furthermore, the loss of muscle gastrocnemius was 10.6% between the IDH1-mutated cancer group and IDH-wt group. Since skeletal muscles account for about 40% of total body weight, skeletal muscle weight loss was the main contributor to body weight loss. From the histopathology results of the muscle gastrocnemius (Fig. 2G), IDH1-mutated in tumor resulted in a smaller cross-sectional area in muscle gastrocnemius compared to the control groups and non-mutation CT26 tumor-bearing cancer cachexia mice (Fig. 2H).
Next, we extracted total RNA from the muscle gastrocnemius and measured the expression of the E3 ligases. Trim63 and Fbxo32 expression were increased by 2.3 and 1.4 times in CT26-bearing cachexia mice without IDH1 mutation (Fig. 2I). In contrast, they were dramatically increased 3.4 and 2.3 times in muscle gastrocnemius from the CT26-bearing cancer cachexia mice with IDH1 mutation compared with controls, respectively. These results implied that the degradation of skeletal muscle protein was enhanced via ubiquitinated protein system. To reveal the mediator of IDH1R132H mutation in cancer, we measured the levels of total D2HG in serum and tumor tissue. D2HG was enriched in IDH1 mutation tumor and serum (Fig. 2J), which was consistent with clinical data revealing that IDH1 mutation at R132 resulted in a high concentration of D2HG and high cachexia frequency in cancer patients. These results confirmed that muscle atrophy's deterioration was mediated by the high concentration of D2HG in CT26 bearing mice with an IDH1 mutation.
D2HG induces proteolysis via up-regulation of the ubiquitinated protein system and metabolic reprogramming
An ex vivo analysis of D2HG on differentiated multi-nucleus myotubes was designed (Fig. 3A). We treated well-differentiated multi-nucleus myotubes with 93 µM D2HG for 5 days, after which immunofluorescence was performed to detect myosin heavy chain (Fig S3A). The average myotube diameter of the D2HG treatment group was smaller than the NTC group (Fig S3B). To confirm the effect of protein degradation, we treated well-differentiated multi-nucleus myotubes with D2HG for 5 days and extracted total RNA. The mRNA expressions of E3 ligases of Trim63 and Fbxo32 were increased by about 11.7 and 20.4 folds, respectively (Fig S3C). Moreover, the E2 ubiquitin-conjugating enzyme Ube2d1 was also increased 2.8 folds (Fig S3D). These results indicated that the upregulated expression of UPP contributed to D2HG-induced muscle atrophy.
D2HG can regulate transcription and metabolic processes, such as glycolysis 22, lipogenesis, oxidative stress, and methylation of histone12. We analyzed the transcriptional change of D2HG induced muscle atrophy and found distinct transcriptional profiling (Fig. 3B). D2HG treatment resulted in 412 transcriptional alterations based on fold change at 2.0 and adjusted q at 0.05 (Table S7). To reveal the primary mechanism responsible for the catabolic pathway, over-representation analysis (ORA) was used to screen the different genes. ORA-based gene ontology (GO) enrichment showed altered molecular function, biological process, and cellular component based on up- or down-regulated genes (Fig. 3C). Extracellular matrix structural constituent, structural molecule activity, structural constituent of the cytoskeleton, oxidoreductase activity, acting on the CH-OH group of donors NAD or NADP as acceptor, glutathione transferase activity, and NAD binding were the top molecular functions. Moreover, the biological process and cellular components, including muscle cell differentiation, muscle tissue development, muscle system process, and muscle contraction, were also disturbed after D2HG treatment (Fig S4). We confirmed the correlation between muscle structure and metabolism resulting from D2HG treatment in differentiated multi-nucleus myotubes based on the GO enrichment analysis.
To reveal the distinct metabolic process responsible for the D2HG treatment, we extracted and measured the selected metabolites from well-differentiated myotubes using a targeted metabolomics strategy. Seventy-one metabolites were included based on the metabolic pathway of KEGG and previously reported metabolites [8–11]. Heatmap showed a distinct separation of the D2HG treatment group (Fig. 3D). A Joint Pathway Analysis was then used to integrate the changed metabolites and genes (Fig. 3E). The most characteristic pathways included glutathione metabolism, hypertrophic cardiomyopathy, aminoacyl-tRNA biosynthesis, citrate cycle synthesis, and degradation of ketone bodies. The changed metabolites based on the KEGG metabolic pathway were constructed (Fig. 3F) and we found D2HG exerted metabolic pathway reprogramming to disturb the maintenance of metabolic homeostasis.
D2hgdh mediated catabolism reverse D2HG induced proteolysis
We then analyzed the catabolic pathway of D2HG and explored the effect of D2HG catabolism on the reversal of proteolysis and muscle atrophy. D2HG is catabolized by D2hgdh, a mitochondrial enzyme that encodes D-2hydroxyglutarate dehydrogenase. Mutation of this gene in humans has been associated with developmental delay, epilepsy, hypotonia, and dysmorphic features 25. Since well-differentiated myotubes cannot catabolize D2HG, we cloned D2hgdh into C2C12 myoblasts and induced its differentiation to myotubes. Typical immunofluorescence of myosin heavy chain staining showed normal differentiation for D2hgdh overexpressed C2C12 myoblast induced with a differentiation medium. Yet, there were sloppy myotubes after D2HG treatment (Fig. 4A). The myotube diameter of the D2HG-treated group was shorter than normal myotubes, while D2hgdh overexpression reversed D2HG induced myotube atrophy (Fig. 4B). We then measured the relative levels of D2HG and keto-glutarate in the differentiated myotubes. D2HG treatment resulted in high levels of D2HG and low keto-glutarate (Fig. 4C). No differences were found in keto-glutarate, while the levels of D2HG were decreased, which implied that the overexpression of D2hgdh in myotubes catabolism D2HG. We also measured the mRNA expression of Ube2d1, Trim63, and Fbxo32 (Fig. 4D-E). Still, D2HG induced high expression of Ube2d1, Trim63, and Fbxo32, while D2hgdh over-expressing myotube inhibited the up-regulation of mRNA expression of these genes after D2HG treatment.
To confirm the expression of D2hgdh and related metabolic enzymes, we measured the protein expression by Western blot (Fig. 4F). D2hgdh overexpression showed high expression of D2hgdh, Hmgcr, Hsd17b7, Dhrs3, and Adh7, but low expression of Idh1. Hmgcr is a rate-limiting enzyme for cholesterol synthesis regulated via a negative feedback mechanism mediated by sterols. Anti-HMGCR antibody-positive patients often showed autoimmune myopathy and resemble limb-girdle muscular dystrophy 26. Hsd17b7 regulates fatty acid metabolism and testosterone synthesis. Testosterone has pronounced effect on muscle protein synthesis and muscle mass enlargement, especially during rapid muscle cell growth. A higher expression level of Hsd17b7 was observed in broilers, which might result in a higher testosis required for early embryonic patterning27. Dhrs3 protein is required for early embryonic patterning and upregulation of Dhrs3 was associated with osteogenic differentiation28. Single nucleotide polymorphism in Adh7 was associated with multiple system atrophy29. Interestingly, the D2hgdh over-expression myotubes reversed the effect of D2HG. These results indicated that D2hgdh over-expression could enhance the catabolism of metabolites D2HG and subtract its proteolysis effect.
D2hgdh reprogramming metabolism of D2HG involving in multiple processes
HPLC-MS-MS-based targeted metabolomics was used to cluster the samples and discrete metabolites (Fig. 4G). All four groups were categorized based on 60 metabolites, and 2HG levels were low in the NTC and D2hgdh groups. Myotubes overexpressing D2hgdh showed a similar metabolic profile as the NTC group. D2HG treatment groups showed distinct metabolic characteristics, which indicated that D2HG could indicate distinct metabolism change. Moreover, D2HG treated myotubes overexpression D2hgdh were clustered between the two groups, thus indicating that D2hgdh overexpression could resist D2HG induced proteolysis. To directly show the metabolic pathway, we drew the metabolic pathway and the relative concentration of metabolites (Fig S5). From the changed metabolites of two independent experiments, 12 were enriched, and 15 were simultaneously depleted (Fig. 4H and Table S8).
To further reveal the transcriptional features of the four groups, a heatmap was used to reveal the clustering profile as metabolomics (Fig. 4I). D2HG triggered distinct transcriptional characteristics since the heatmap was clustered far from the NTC group. D2hgdh overexpression could resist D2HG induced gene alterations. The overall differentially expressed genes profile of the four groups showed that D2HG upregulated 1340 genes and downregulated 786 genes, while myotube overexpressing D2hgdh showed 20 upregulated genes and 65 downregulated genes with fold change at 2 and padj at 0.05 (Fig. 4J). These implied that D2HG induced a wide range of gene transcriptional level changes, while myotube overexpressing D2hgdh showed subtle perturbation. This set of experiments also included D2HG-treated well-differentiated myotubes and control myotubes (Table S9). In addition, based on the repeatability assay, the Venn diagram revealed the distinct transcriptional alteration, showing that 37 genes were commonly upregulated, and 101 genes were downregulated after D2HG treatment (Fig. 4K and Table S9).
To reveal the distinct transcriptional and metabolic characteristics of D2hgdh and evaluate the reversal of D2HG-mediated metabolic reprogramming with D2ghdh over-expression, we used the paired comparison on D2hgdh overexpression myotubes. First, we compared D2ghdh overexpression in well-differentiated myotubes with control well-differentiated myotubes (Table S10). The volcano plot showed the changed genes resulting from D2ghdh overexpression in well-differentiated myotubes (Fig. 5A). The downstream metabolites 2HG were decreased, which was consistent with the theoretical metabolic trend (Fig. 5B). The joint pathway analysis revealed several pathways, such as CoA biosynthesis and glutathione metabolism (Fig. 5C). To reveal the direct effect of D2ghdh overexpression on well-differentiated myotubes, GO analysis was used to display the molecular function, biological process, and cellular component (Fig S6). Microtubule motor activity, microtubule-binding, tubulin binding, fibronectin-binding, ATPase activity, cell adhesion molecule binding, and activin receptor binding were the top molecular function.
Next, we compared the transcriptional and metabolic features between D2HG-treated myotubes over-expressing D2ghdh and NTC treated myotubes overexpressing D2ghdh to evaluate the D2hgdh over-expression on alleviation D2HG induced proteolysis (Table S11). It was found 241 genes were up-regulated, and 306 were down-regulated after D2HG treatment (Fig. 5D). NADH, lactate, 3-Methylhistidine, carnitine, and 2-hydroxyglutarate were enriched, while NADPH, NAD, and creatinine were depleted. GO analysis revealed the molecular function of oxidoreductase activity, acting on the CH−NH group of donors as top pathway enrichment (Fig S7). Moreover, D2ghdh induced the biology process alteration, such as sterol biosynthetic process, organic hydroxy compound metabolic process, and Acetyl-CoA metabolic process.
Furthermore, D2HG-treated D2ghdh over-expressing myotubes and D2HG-treated control myotubes were also compared (Table S12). A total of 1356 genes were depleted, and 640 genes were enriched (Fig. 5E); ATP-dependent serine/threonine kinase regulator activity was the top molecular function (Fig S8).
To interpret the gene expression data from D2hgdh over-expression and D2HG treatment, Gene Set Enrichment Analysis (GSEA) was used. D2HG-treated well-differentiated myotubes over-expressing D2hgdh showed enriched proteasome accessory complex and SUMO transferase activity, as well as depleted positive regulation of stem cell differentiation (Fig. 5F). These function annotations confirmed the D2HG mediated enhanced proteolysis and attenuated muscle differentiation.
To show the commonality relationships, the changed genes (Table S13) and Venn diagrams were used. Fifty-five common enriched genes and 52 common depleted genes were found (Fig. 5G). Metabolomics showed a series of common metabolic profiling, including depletion of 12 metabolites and enrichment of 8 metabolites after D2hgdh overexpression in well-differentiated myotubes, and depletion of 15 metabolites and enrichment of 12 metabolites after D2HG treated myotubes over-expressing D2hgdh when compared with D2HG treated control myotubes (Table S14). Venn diagram showed 1 commonly enriched metabolite and 5 commonly depleted metabolites (Fig. 5H). D2hgdh over-expression showed decreased 2HG, as well as isocitrate, carnitine, Fum, and NADH. D2hgdh could catalyze the conversion of D2HG to keto-glutarate with the driver of NAD to NADH. It was interesting that D2HG treatment resulted in decreasing NADPH/NADP ratio and increasing NADH/NAD ratio. However, when D2hgdh was over-expressed in myotubes, it could release the effect of D2HG and rescue the exhaustion of NADPH by cycling the redox balances and metabolic homeostasis30. Excessive D2HG could induce dysfunction of well-differentiation myotubes, while D2hghd over-expression myotube could catalyze D2HG and take advantage of metabolic reprogramming to drive the cycle of redox balances and metabolic homeostasis (Fig. 5I).
Ivosidenib relieves Idh1 mutation mediated exacerbated cancer cachexia
Ivosidenib is a selective inhibitor of IDH1 mutation that blocks the abnormal IDH1 protein and can reduce abnormal D2HG levels 31. We evaluated the effect of ivosidenib on relieving IDH1 mutation exacerbated cancer cachexia in mice bearing CT26 tumor (Fig. 6A). From DPI 9 of palpable tumor, mice bearing CT26 tumor and mice bearing CT26 tumor with IDH1 mutation were intravenously administrated with 50mg/kg ivosidenib or PBS as NTC control every day for the following experimental period 31. Mice bearing CT26 tumor with IDH1 mutation showed a worse survival compared to the control mice (Fig. 6B). After treatment with ivosidenib, the survival of mice bearing CT26 tumor did not significantly change, while it was prolonged the survival of mice bearing CT26 tumor with an IDH1 mutation. As for the tumor weight curve, there were no changes between the two groups that did not receive treatment. Yet, after treatment, the tumor weight of control mice did not significantly change, while it was decreased in CT26 tumor with an IDH1 mutation. Ivosidenib preserved lean body weight in the IDH1 mutation tumor and delayed the cancer cachexia progression since the occurrence of cancer cachexia was first found at DPI 25 (Fig. 6C). To verify that ivosetinib has an impact on cachexia directly, we compared the lean body weight with a similar tumor weight of 1.56g (tumor volume 3000mm3) and found there was a significant preserve of lean body weight in ivosidenib treated mice bearing CT26 tumor with IDH1-mutation when compared with NTC treated mice bearing CT26 tumor with IDH1-mutation (Fig. 6D). Next, we measured the cross-sectional area of the muscle gastrocnemius (Fig. 6E), and found the IDH1 mutation group was smaller than the control (Fig. 6F). Ivosidenib treatment prevented muscle atrophy. The IDH1 mutation also led to decreases in muscle gastrocnemius weight, while ivosidenib reversed the loss of muscle gastrocnemius (Fig. 6G). The mRNA expression of Trim63 and Fbxo32 was consistent with the histopathological results (Fig S9). IDH1 mutation in CT26 tumor-bearing mice resulted in increased expression of Ube2d1, Trim63, and Fbxo32, while ivosidenib treatment inhibited the upregulation of UPP-related enzymes. Serum D2HG concentration was increased after IDH1 mutation, while it was decreased after treatment with ivosidenib (Fig. 6H). These results indicated D2HG was a mediator of muscle wasting. Inhibition of the production of D2HG through IDH1 mutation inhibitor ivosidenib or catabolism of D2HG through over-expression of D2HGDH may reverse D2HG induced muscle proteolysis and slow down cancer cachexia procession.
Since IDH1 mutation was commonly found in glioma. We also established an orthotopic tumor model of in vivo bearing GL261 glioma cells with IDH1-mut. It was found the import of mutational IDH1 into GL261 glioma tumor resulted in significant improvement in survival (Fig. 6I). Ivosidenib could prolong the survival of mice bearing GL261 glioma tumor with mutational IDH1, however, not change the survival of mice bearing wild type GL261 glioma tumor. The loss of body weight was observed and loss of body weight over 5% occurred at DPI 15 for the mice bearing wild type glioma tumor and IDH1 mutational glioma tumor (Fig. 6J). After ivosidenib treatment, the cachexia syndrome occurred at DPI 17 for the mice bearing wild type glioma tumor, while at DPI 20 for the mice bearing IDH1 mutational glioma tumor. Though no significant change in muscle weight was found (Fig S10A), there was a significant decrease in muscle area (Fig. 10S B-C). Ivosidenib treatment resulted in significant improvement of muscle area. Moreover, the expression of E3 ligases Trim63 and Fbxo32 were increased in the muscle of mice bearing IDH1 mutation glioma tumor, while alleviated after ivosidenib treatment (Fig S10D). Moreover, serum D2HG concentration was increased after IDH1 mutation, while it was decreased after treatment with ivosidenib (Fig. 6K).