Network-based modelling reveals cell-type enriched patterns of non-coding RNA regulation during human skeletal muscle remodelling

Abstract

Most human genes are non-protein-coding RNA (ncRNA). A handful of ncRNAs have characterised functions, including important epigenetic roles in development and disease. Neither ncRNA nor multinucleated muscle is ideally suited to sequencing technologies. We therefore used customised RNA profiling methods and quantitative network modelling to study cell-type specific ncRNA transcriptome responses during load-induced skeletal muscle hypertrophy. We completed five independent supervised exercise-training studies (n=144) and 61% of individuals accrued muscle mass beyond normal technical variation (lean mass responders, LMR). The remainder were defined as having no measurable lean mass response (NMLMR). Fifty ncRNA genes (FDR <1%) were differentially regulated in LMR, and in total we identified 110 ncRNAs for further study. A network model of the human muscle transcriptome was built (n=437 samples), assigning ncRNAs to protein coding modules representing functional pathways or single-cell types. We identified that the known hypertrophy-related ncRNA, CYTOR, was leukocyte-associated in vivo in humans (FDR = 4.9 x10-7; Fold Enrichment [FE] = 6.6). Other ncRNA modules included PPP1CB-DT, which was segregated with myofibril assembly genes (FDR = 8.15 x 10-8; FE = 47.5), while EEF1A1P24 and TMSB4XP8 were associated with vascular remodelling and angiogenesis genes (FDR = 2.77 x 10-5; FE = 3.6). MYREM was positively associated with hypertrophy, and we established its myonuclear expression pattern in vivo in humans using spatial transcriptomics probes. We show that single-cell type associations of ncRNA are identifiable from bulk transcriptomic data and that hypertrophy-linked ncRNA genes appear to mediate their association with muscle growth via multiple cell types.