Glycolysis Diverges from Its Transcripts in Human Hypertrophic Cardiomyopathy

Human hypertrophic cardiomyopathy does not remodel cardiac metabolism through a single, uniform metabolic switch, according to a multi-omics analysis of eight public datasets. The study found that glycolytic gene-expression programs were broadly suppressed at the RNA level, while several glycolytic enzymes moved in the opposite direction at the protein level. Oxidative pathways behaved differently: genes and proteins associated with fatty-acid oxidation, the tricarboxylic acid cycle and oxidative phosphorylation generally declined together. The selective mismatch between transcripts and proteins points to post-transcriptional regulation that cannot be captured by RNA measurements alone.

Researchers combined four transcriptomic cohorts, three proteomic datasets and one phosphoproteomic dataset to examine metabolic remodeling in HCM myocardium. The material included single-nucleus RNA sequencing, microarray and bulk RNA sequencing, alongside tandem mass tag proteomics, liquid chromatography tandem mass spectrometry, SomaScan aptamer assays and phosphoproteomics. Some analyses used paired RNA and protein measurements from the same myectomy cohort, while others provided cross-cohort validation. Across more than 400 patients with HCM, glycolytic and hypoxia-related gene sets were predominantly negatively enriched at the transcript level.

The largest transcriptomic dataset, comprising 106 HCM and 39 control septal samples, showed significant suppression of both glycolysis and hypoxia programs, with normalized enrichment scores of −1.56 and −1.91, respectively, and false-discovery rates below 0.001. Single-cell data reproduced the pattern. In the Chaffin cohort, cardiomyocytes from hearts with preserved and reduced ejection fraction both showed negative enrichment of glycolysis and hypoxia pathways. Independent single-nucleus data from another cohort also indicated reduced glycolytic transcription, although its small comparison of three control and 10 HCM hearts did not significantly suppress the hypoxia set. The bulk RNA-seq dataset supplied further support, with hypoxia genes negatively enriched at a normalized score of −1.89.

Not every glycolytic transcript behaved identically. In the 106-sample microarray cohort, five of 12 core glycolytic enzymes were significantly downregulated, including LDHA, ENO1 and HK2, while LDHB and PFKM increased. Sparse detection in single-nucleus data complicated some individual estimates, particularly for ALDOA, which was detected in only 10 of 122,362 cardiomyocyte nuclei. But donor-level aggregation of the single-cell data produced estimates that correlated with the independent bulk RNA-seq results, and transcript detection was not systematically associated with larger discrepancies. Sensitivity analyses excluding ALDOA did not materially change the pathway findings, suggesting that sparsity did not explain the overall suppression pattern.

HIF1-associated transcription also ran counter to the expectation that advanced HCM would activate a hypoxia-driven glycolytic program. A curated set of established HIF1α targets, including SLC2A1, LDHA, PGK1 and ENO1, was strongly negatively enriched in the Garmany transcriptomic cohort, with a normalized score of −1.99 and an adjusted P value of 2.8 × 10−4. Two other HIF1α target sets were directionally negative but did not pass the false-discovery threshold. These results concern transcription attributed to HIF1α targets; they do not directly measure HIF1α protein abundance, stabilization, nuclear localization or activity.

The strongest evidence for transcript–protein discordance came from the paired Garmany myectomy dataset. Among 12 core glycolytic enzymes, five changed in opposite directions between RNA and protein. PFKM, LDHB and PGK1 had higher transcript levels but lower protein levels, whereas HK1 and PFKL showed the reverse pattern. The correlation between RNA and protein fold changes was weak and inverse, at r = −0.28 with P = 0.38 for 12 enzymes, but the researchers treated this correlation as descriptive because of the small sample. At pathway level, glycolysis had a negative transcriptomic enrichment score of −1.11 and a positive proteomic score of +1.40, although neither comparison met the false-discovery threshold. The hypoxia set showed a similar sign reversal, from significant negative transcriptomic enrichment to positive but nonsignificant protein enrichment.

HK1 emerged as the clearest example of decoupling. Its transcripts were reduced in all four transcriptomic cohorts, yet its protein was increased across all three independent proteomic platforms: tandem mass tag mass spectrometry, liquid chromatography tandem mass spectrometry and SomaScan. Because those technologies use different analytical principles, their agreement made HK1 less likely to be a platform-specific artifact. Other enzymes were less consistent. PGAM1, for example, increased in the LC-MS/MS dataset but decreased in the SomaScan dataset, while PFKM also showed platform-dependent behavior. The comparison between cardiomyocyte-specific single-cell RNA data and bulk-tissue proteomics provided supportive but less definitive evidence because the cohorts differed and the protein measurements could not be assigned to individual cell types.

Phosphoproteomic measurements added another layer of complexity. HK1 and PGAM1 showed increased phosphorylation, with log2 fold changes of +0.18 and +0.40, respectively. ALDOA, LDHB, PKM and PFKM instead had reduced phosphorylation, with adjusted P values of 0.02 or lower. The mixed pattern did not resemble a coordinated, pathway-wide increase in modification across glycolytic enzymes. The data also did not directly establish whether any enzyme was more or less active. Phosphorylation can have different effects depending on the modified residue, and the study’s phosphoproteomic evidence came from one dataset. It therefore identified candidate post-translational regulation without measuring metabolic flux or catalytic activity.

Integrated molecular and metabolite evidence suggested that the more reproducible defect involved impaired coupling between glycolysis and mitochondrial oxidation. In HCM myocardium with preserved ejection fraction, transcripts for oxidative phosphorylation, the TCA cycle, fatty-acid oxidation, glycolysis and hypoxia programs were all negatively enriched. Corresponding oxidative proteins and several TCA-associated enzymes were reduced. Metabolites were also depleted: acetyl-CoA fell by a log2 fold change of −1.69, NADH by −1.05 and lactate by −0.99, while medium- and long-chain acylcarnitines showed an average decrease of −1.43. Lower lactate alongside elevated amounts of selected glycolytic proteins is difficult to reconcile with a simple model of globally increased glycolytic flux. Nevertheless, steady-state metabolite concentrations are indirect indicators and do not by themselves determine pathway flux.

The findings have important boundaries. Most human myocardium came from patients with symptomatic, advanced, obstructive HCM undergoing septal myectomy, so the patterns may not represent earlier disease, non-obstructive HCM or every genotype. The cohorts also differed in severity, treatment exposure, cellular composition and tissue processing, while bulk proteomics cannot distinguish changes within cardiomyocytes from shifts caused by fibrosis or altered stromal and immune-cell abundance. Single-nucleus RNA sequencing can underestimate weakly detected transcripts, and the phosphoproteomic analysis was limited to one dataset. The results therefore support a selective transcript–protein disconnect in glycolysis rather than proving a mechanism for it or demonstrating a therapeutic target. Future work with spatial transcriptomics, single-cell proteomics and isotope tracing would be needed to resolve cell-specific regulation and test pathway activity directly.

Subject of Research: Transcript–protein discordance in glycolytic and oxidative metabolism in human hypertrophic cardiomyopathy

Article Title: Selective transcript–protein discordance of glycolytic remodeling in human hypertrophic cardiomyopathy

Article References: Chen, J., Hu, D., Yang, Y., Ouyang, M., & Chen, X. (2026). Selective transcript–protein discordance of glycolytic remodeling in human hypertrophic cardiomyopathy. Frontiers in Cardiovascular Medicine, 13, Article 1937619. https://doi.org/10.3389/fcvm.2026.1937619

Image Credits: AI Generated

DOI: 10.3389/fcvm.2026.1937619

Keywords: hypertrophic cardiomyopathy, glycolysis, multi-omics, transcriptomics, proteomics, HIF1α, phosphoproteomics, cardiac metabolism, HK1, mitochondrial oxidation, Selective, transcript