Abstract

   Myopathy refers to a large group of heterogeneous, rare muscle
   diseases. Bulk RNA-sequencing has been utilized for the diagnosis and
   research of these diseases for many years. However, the existing
   valuable sequencing data often lack integration and clinical
   interpretation. In this study, we integrated bulk RNA-sequencing data
   from 1221 human skeletal muscles (292 with myopathies, 929 controls)
   from both databases and our local samples. By applying a method similar
   to single-cell analysis, we revealed a general spectrum of muscle
   diseases, ranging from healthy to mild disease, moderate muscle
   wasting, and severe muscle disease. This spectrum was further partly
   validated in three specific myopathies (97 muscles) through clinical
   features including trinucleotide repeat expansion, magnetic resonance
   imaging fat fraction, pathology, and clinical severity scores. This
   spectrum helped us identify 234 genuinely healthy muscles as
   unprecedented controls, providing a new perspective for deciphering the
   hallmark genes and pathways among different myopathies. The newly
   identified featured genes of general myopathy, inclusion body myositis,
   and titinopathy were highly expressed in our local muscles, as
   validated by quantitative polymerase chain reaction.

   Subject terms: Neuromuscular disease, Molecular medicine
     __________________________________________________________________

   This study analysed bulk RNA-sequencing data from 1221 human skeletal
   muscles and validated the progressive spectrum of myopathy diseases.

Introduction

   Myopathy is a general term that refers to a large group of diseases
   primarily affecting the skeletal muscles. These conditions can be
   categorized into inherited or acquired forms, according to their
   aetiology. Myopathies exhibit heterogeneous phenotypes, including
   weakness, abnormal gait, muscle pain, difficulty swallowing,
   contractures, and systemic impairments, among others^[40]1–[41]3. One
   common feature of myopathies is a large spectrum of the degree of
   severity, similar to other progressive diseases, which includes
   asymptomatic, mild-to-moderate, and severe stages^[42]4–[43]7. However,
   this spectrum is usually more based on clinical observation than on
   well-established objective findings.

   RNA-sequencing for bulk skeletal muscles has been utilized in the
   diagnosis and research of muscle diseases (e.g., for ectopic splicing
   and molecular mechanism investigation)^[44]8–[45]10. It offers a
   sensitive perspective to understand the ongoing molecular activities in
   the muscle. Numerous studies have deposited their transcriptional data
   in various online databases, and this data is of significant value for
   integration, especially considering the rarity of most myopathies.
   However, these isolated datasets are somewhat prone to various biases
   from small sample size, selection of control materials, sequencing
   methods, different read length, etc^[46]11.

   Furthermore, with the emergence of more advanced and sophisticated
   technologies developed for myopathies in the past decade, deep
   phenotyping (including quantitative MRI, muscle biopsy evaluation, CTG
   expansion size in myotonic dystrophy) provides a multi-dimensional
   evaluation to depict the muscle deterioration process. Correlating
   these muscle-specific features with their genetic data can assist in
   characterizing and deciphering myopathies.

   In this study, we integrated transcriptional data from 1221 human
   skeletal muscles obtained from both online databases and our local
   patients, ultimately identifying a general spectrum separating normal
   and myopathy-affected muscles. In contrast to the traditional approach
   of focusing primarily on diseased samples, we reversed the perspective,
   emphasizing the control samples to characterize this spectrum validated
   using clinical features from different sources. We offered a novel
   perspective by using genuinely healthy muscles as an unprecedented
   control reference, aiming to identify both common pathways and specific
   features of the studied myopathies.

Methods

Data source and participant selection

   This is a retrospective integrative analysis (Fig. [47]1). The data
   sources include 803 muscles from the GTEx Consortium (dbGaP Accession
   phs000424.v8.p2)^[48]12, 291 muscles from the GEO database
   ([49]GSE115650^[50]13, [51]GSE140261^[52]14, [53]GSE175861^[54]15,
   [55]GSE184951^[56]16, [57]GSE201255^[58]17, [59]GSE202745^[60]18), and
   127 muscles from Helsinki (39 of which have also been reported as
   [61]GSE151757^[62]19). The ethics approval of using local muscles
   (195/13/03/00/11) was approved by HUS (Helsingin Uudenmaan
   Sairaanhoitopiiri) and informed consent was obtained from each subject.
   All ethical regulations relevant to human research participants were
   followed. The inclusion and exclusion criteria for participant
   selection were as follows: (1) only human skeletal muscle tissue was
   included (no cell lines or organoids); (2) bulk-RNA sequencing was
   performed using high-throughput techniques (no chip arrays or
   single-cell data); (3) datasets were preserved in raw count format
   (those shared in transformed count format were excluded).

Fig. 1. The workflow.

   [63]Fig. 1
   [64]Open in a new tab

   Human skeletal muscle bulk-RNA-seq data from three sources (GTEx
   database, GEO database, and Helsinki) were integrated into a combined
   dataset (1221 muscles × 9231 genes). A spectrum order can be observed
   in this integrated dataset: Healthy→Mild disease→Moderate muscle
   wasting →Severe muscle disease. Different clinical features were mapped
   to the transcriptional data to validate this spectrum order. Tissue
   deconvolution was performed using skeletal muscle single-cell datasets
   as references, allowing us to infer the cell type composition in