Translating Muscle RNAseq Into the Clinic for the Diagnosis of Muscle Diseases

Abstract

Objective: Approximately half of patients with hereditary myopathies remain without a definitive genetic diagnosis after DNA next-generation sequencing (NGS). Here, we implemented transcriptome analysis of muscle biopsies as a complementary diagnostic tool for patients with muscle disease but no definitive genetic diagnosis after exome sequencing.

Methods: In total, 70 undiagnosed cases with suspected genetic muscular dystrophies or congenital myopathies were included in the study. Muscle RNAseq comprised the analysis of aberrant splicing, aberrant expression, and monoallelic expression. In addition, existing NGS data or variant calling from RNAseq were reanalyzed, and genome sequencing was performed in selected cases. Four aberrant splicing open-source tools were compared and assessed.

Results: RNAseq established a diagnosis in 10/70 patients (14.3%) by identifying aberrant transcripts produced by single nucleotide variants (7/10) or copy number variants (3/10). Reanalysis of NGS data allowed the diagnosis in 9/70 individuals (12.9%). Based on this cohort, FRASER was the tool that reported more splicing outlier events per sample while showing the highest accuracy (81.26%).

Conclusions: We demonstrate the utility of RNAseq in identifying causative variants in muscle diseases. Evaluation of four aberrant splicing tools allowed efficient identification of most pathogenic splicing events, obtaining a manageable number of candidate events for manual inspection, demonstrating feasibility for translation into a clinical setting. We also show how the integration of omic technologies reduces the turnaround time to identify causative variants.

Keywords: RNA sequencing; alternative splicing; congenital myopathy; genetic diagnosis; muscular dystrophy; neuromuscular diseases; transcriptomics.

FIGURE 1

Overview of the diagnostic pipeline using RNAseq. (A) Filtered RNAseq BAM files with NMD genes (Muscle Gene Table, version 17/01/2024) were used as input for RNAseq VC and aberrant splicing tools. Splicing outlier events were annotated with nearby SNVs (from DNAseq or RNA VC) and the z‐score obtained in OUTRIDER. Candidate events were visually inspected in IGV and, if relevant, validated. For patients with available DNAseq VCF files, genes with MAE were annotated with candidate splice events and SNVs. OUTRIDER was used to identify samples with gene expression outliers: results were sorted by lowest z‐scores, and the first five genes with the lowest z‐scores for each category were visually inspected. Asterisks indicate analyses that included NMD genes, OMIM genes, highly expressed muscle genes, and all genes. (B) The global diagnostic rate in the cohort of 70 undiagnosed cases after ES/gene panel sequencing was 27.2%. In the RNAseq‐solved cases (10/70, 14.3%), RNAseq findings facilitated reaching the genetic diagnosis. Patients included in the reanalysis category (9/70, 12.9%) also included patients whose variant was identified by RNAseq VC. AS, aberrant splicing; CNV, copy number variant; ES, exome sequencing; IGV, integrative genomics viewer; LRS, long read sequencing; MAE, monoallelic expression; NMD, neuromuscular disease; RNAseq, RNA sequencing; SNV, single nucleotide variant; VC, variant calling; VUS, variant of uncertain significance.

FIGURE 2

Alternative splicing tools to detect aberrant splicing events from RNAseq data. (A) Boxplot showing statistically significant splicing events per sample identified by FRASER for a |Δ PSI| > 0.1 cutoff, FRASER for a |Δ PSI| > 0.3 cutoff, FRASER2, LeafCutterMD, and rMATS‐turbo. (B) Aberrant splicing junctions identified by each splicing tool and the overlapping events identified among them. (C) Pathogenic or Likely Pathogenic splicing events (detailed in Table 2) correctly detected by each tool. (D) Three examples of aberrant splicing events that can be detected with RNAseq. Upper panel: Intron retention and multiple low‐abundant transcripts produced due to the CAPN3(NM_000070.3):c.1746‐20C>G variant. Middle panel: Intron retention, caused by the missense variant TTN(NM_001267550.2):c.57262G>C, p.(Val19088Leu) in Case 28. Lower panel: Pseudoexon inclusion in the DMD gene due to the deep intronic variant DMD(NM_004006.3):c.10554‐2996T>G. PSI, percentage spliced in; RNAseq, RNA sequencing.

FIGURE 3

RYR1 alternative splicing in Case 22 due to a synonymous variant. (A–F) HE staining on muscle biopsy showing marked variability in fiber size, nuclear internalization and centralization with increased endomysial connective tissue (A). Several muscle fibers contain dark blue‐reddish irregular areas (arrows in B), devoid of ATPase activity (D). Some fibers contain well‐defined cores (single arrows in E), and others have irregular areas of myofibrillar disorganization (double arrows in E). In addition, some fibers contain collections of small nemaline bodies (arrows in C). There is type 1 fiber uniformity (D). (G) Sashimi plot of exons 48 and 49 showing alternative (139 junction counts) and wild‐type transcripts (813 junction counts) caused by the RYR1(NM_000540.3):c.7833C>T, p.(Cys2611=) variant. As represented in the lower panel, the aberrant transcript (dashed line) contains a 4‐bp exon deletion leading to a premature termination codon (p.(Ala2612Thrfs*132)). (H) OUTRIDER expression rank plot of RYR1 gene showing reduced expression in Case 22. COX, cytochrome c oxidase; HE, hematoxylin and eosin; NADH, nicotinamide adenine dinucleotide.

FIGURE 4

Single exon deletion mosaicism in COL6A1. (A) Immunofluorescence of Collagen‐VI from the proband's fibroblast (upper panel) showing decreased Collagen‐VI intensity compared to control and a diffuse microfibrillar pattern. Permeabilized cells (lower panel) show a marked Collagen‐VI intracellular retention. In the father (Case 10), Collagen‐VI staining and microfibrillar patterns are similar to controls and no intracellular Collagen‐VI retention occurs in permeabilized cells. (B) Sashimi plot of the father's muscle RNAseq revealed 15% of transcripts with exon 9 skipping, not present in any other sample in the cohort. (C) Exon 9 skipping was validated in the father's and proband's fibroblasts through RT‐PCR, evident in 12.3% and 47% of the transcripts in the father and proband, respectively. (D) IGV screenshot of GS from peripheral blood for the father and proband, showing a deletion encompassing exon 9 (split‐reads) in heterozygosity in the proband and mosaicism in the father. GS, genome sequencing; IGV, Integrative Genomics Viewer; RNAseq, RNA sequencing; RT‐PCR, reverse transcription polymerase chain reaction.