Full-length Isoform Sequencing for Resolving the Molecular Basis of Charcot-Marie-Tooth 2A

Abstract

Objectives

Transcript sequencing of patient-derived samples has been shown to improve the diagnostic yield for solving cases of suspected Mendelian conditions, yet the added benefit of full-length long-read transcript sequencing is largely unexplored.

Methods

We applied short-read and full-length transcript sequencing and mitochondrial functional studies to a patient-derived fibroblast cell line from an individual with neuropathy that previously lacked a molecular diagnosis.

Results

We identified an intronic homozygous MFN2 c.600-31T>G variant that disrupts the branch point critical for intron 6 splicing. Full-length long-read isoform complementary DNA (cDNA) sequencing after treatment with a nonsense-mediated mRNA decay (NMD) inhibitor revealed that this variant creates 5 distinct altered splicing transcripts. All 5 altered splicing transcripts have disrupted open reading frames and are subject to NMD. Furthermore, a patient-derived fibroblast line demonstrated abnormal lipid droplet formation, consistent with MFN2 dysfunction. Although correctly spliced full-length MFN2 transcripts are still produced, this branch point variant results in deficient MFN2 levels and autosomal recessive Charcot-Marie-Tooth disease, axonal, type 2A (CMT2A).

Discussion

This case highlights the utility of full-length isoform sequencing for characterizing the molecular mechanism of undiagnosed rare diseases and expands our understanding of the genetic basis for CMT2A.

Introduction

Transcript sequencing is emerging as a powerful clinical tool and has been reported to increase diagnostic yield by 2%–24% vs DNA sequencing alone when evaluating cases of suspected Mendelian conditions.^1-5 These studies have used short-read sequencing to infer the identity of full-length transcripts.⁶ However, exon skipping and alternative splice site usage within multi-intronic genes can limit the ability of short-read sequencing to accurately predict the coding impact of aberrantly spliced full-length transcripts. Distinguishing among potential full-length transcript outcomes is important for appropriately evaluating conditions whereby distinct phenotypes and inheritance patterns are associated with dominant-negative or loss-of-function (LOF) variants in the same gene.

Charcot-Marie-Tooth 2A (CMT2A) is an axonal peripheral nerve disorder characterized by motor, sensory, or autonomic neuropathy. Approximately 90% of individuals with CMT2A have monoallelic variants associated with a dominant-negative mode of action in mitofusin 2 (MFN2)⁷ and a dominant mode of inheritance. By contrast, autosomal recessive inheritance is associated with biallelic LOF MFN2 variants, which typically do not result in a clinical phenotype in the heterozygous state. Splicing variants in MFN2 can cause both dominant and recessive forms of CMT2A,^8-10 indicating the need to accurately identify the effect of novel splicing variants. We report a patient found via short-read and full-length transcript sequencing to have a homozygous branch point variant in MFN2 that causes MFN2 deficiency via the creation of 5 altered transcripts, all subject to nonsense-mediated mRNA decay (NMD).

Methods

Exome Sequencing and Analysis

Quad exome sequencing (proband, unaffected mother, unaffected brother, unaffected paternal half brother) was performed through the Undiagnosed Diseases Network (UDN) (Baylor College of Medicine).

Short-read Transcript Sequencing and Analysis

RNA extraction, library preparation, and short-read sequencing were performed on cultured skin fibroblasts from the proband, as previously described.⁵ A control data set of short-read transcript/RNA sequencing from 236 skin fibroblast samples was used to identify RNA expression outliers and aberrant splicing products using OUTRIDER¹¹ and IRFinder,¹² respectively.

Full-length Transcript Sequencing and Analysis

Cultured fibroblasts were incubated with or without cycloheximide (100 μg/mL, Sigma-Aldrich) for 6 hours before RNA extraction (RNeasy Mini kit—Qiagen). cDNA synthesis was performed following the ISO-Seq protocol and sequenced using a Sequel II (PacBio, Menlo Park, CA). Iso-Seq data were processed using the Iso-Seq3 pipeline, mapped to GRCh38, and visualized using IGV.

Sanger Validation

cDNA was synthesized with random hexamers and SuperScript III reverse transcriptase (Invitrogen). The MFN2 region of interest was PCR amplified for 35 rounds with an exon 5 sense primer (5′-GCCATGAGGCCTTTCTCCTT) and an exon 8 antisense primer (5′-AGACGCTCACTCACCTTGTG). PCR products were separated using 7% polyacrylamide, and then DNA bands were excised and incubated in water overnight to elute DNA, which was amplified using the primers mentioned earlier and subjected to Sanger sequencing.

Lipid Droplet Analysis

Fibroblasts from the proband and a patient not harboring MFN2 variants were separately plated and grown on glass bottom dishes and then incubated with 0.1 μg/mL Mitotracker Red CMX Ros (Molecular Probes), 5 mM BODIPY 493/503 (Invitrogen), and 3 drops of NucBlue (Invitrogen) for 15 minutes at 37°C with 5% CO₂. Cells were moved into complete media for ≥45 minutes and then imaged using a Z-series step size of 0.3 μm on a Nikon Ti-E widefield microscope with a 63X NA 1.4 oil objective (Nikon), solid-state light source (Spectra X, Lumencor), and an sCMOS camera (Zyla 5.5 megapixel). Each line was imaged on 3 separate occasions (n > 100 cells/experiment). Images were deconvolved using 7 iterations of 3D Landweber deconvolution. For each image, 13–34 cells were captured and analyzed together. The number and fluorescence intensity of lipid droplets was quantified using the automated Spot Detection Analysis after removing background signal (Nikon Elements). Detection was set to detect “bright spots” of different sizes (typical diameter 0.54 μm) with a contrast of 10.1. Function “detect all objects” was selected. Using the ROI statistics function, the number of spots detected and pixel intensity were quantified. Maximum intensity projections were generated using ImageJ (NIH). All quantification was performed by an experimenter blinded to sample identity.

Standard Protocol Approvals, Registrations, and Patient Consents

This study was approved by the National Institutes of Health Institutional Review Board (IRB) (IRB # 15HG0130), and written informed consent was obtained from all participants in the study.

Data Availability

Sequencing data generated through the UDN are available at dbGaP Study Accession phs001232.v5.p2.

Results

Clinical Phenotype

We evaluated a 42-year-old woman who initially presented with abnormal “foot-slapping” gait at 1 year of age that progressed into distal leg weakness requiring a wheelchair for mobility by age 8 years (eTable 1, links.lww.com/NXG/A621). She underwent spinal fusion and Harrington rod placement for scoliosis in her teens and developed respiratory involvement in her thirties. She had normal cognitive development and no family history of neuromuscular disease. EMG and nerve conduction velocity studies at age 2 years revealed distal motor and sensory polyneuropathy, with positive waves and fibrillation. Nerve and muscle biopsy revealed marked denervation atrophy. Neurologic examination at age 42 years showed normal facial strength, hypophonia, severe muscle wasting of arms and legs, and 1–2/5 proximal and 0/5 distal motor strength. Sensation was present but reduced to all modalities distally, and reflexes were absent throughout.

Identification of a Deep Intronic MFN2 Variant

Initial genetic evaluation revealed paternal uniparental isodisomy of chromosome 1 (UPD[1]), while panel testing for neuromuscular disorder–associated genes was nondiagnostic. She was enrolled into the UDN, and initial exome analysis was nondiagnostic. Short-read transcript sequencing of patient-derived fibroblasts identified MFN2, located on chromosome 1, as an expression outlier with expression approximately half that of control fibroblasts (Z score −6.9) (Figure 1A) In addition, MFN2 exhibited increased retention of intron 6 (Z score 8.6) (Figure 1B). Reanalysis of exome data identified a homozygous MFN2 c.600-31T>G variant within intron 6 that is absent from population databases and is predicted to disrupt the U nucleotide in a yUnAy consensus branch point sequence¹³ (Figure 1C).

Figure 1. Identification of a Homozygous MFN2 Branch Point Variant that Disrupts MFN2 Splicing.

(A) Short-read RNA sequencing identified MFN2 as an expression outlier in this patient's sample, exhibiting 51% of the RNA expression level seen in control fibroblast samples. (B) Genetic testing of MFN2 identified a deep intronic homozygous variant in intron 6 of MFN2. Short-read RNA sequencing identified that the intron retention ratio of intron 6 of MFN2 was significantly abnormal compared with that in controls. (C) Long-read full-length transcript sequencing (ISO-Seq) of this patient's sample after treatment with the nonsense–mediated decay (NMD) inhibitor cycloheximide (CHX) identified 6 major MFN2 transcripts that differ in their splicing patterns for intron 6. The predicted protein impact and transcript count of each are indicated to the right. Inset below shows the alternative splice acceptor sites used for each transcript, as well as the sequence context of the patient's variant relative to the canonical branch point sequence. (D) Polyacrylamide gel electrophoresis showing altered spliced products with and without CHX treatment affecting exon 7 of MFN2 in cells from the patient and the normal splicing of MFN2 in cells from a control sample.

Characterizing the Transcript Impact of an MFN2 Branch Point Variant

Because branch point variants can induce complex splicing alterations,¹⁴ we performed full-length isoform sequencing (ISO-Seq) to determine the identity of all full-length MFN2 transcripts. ISO-Seq of patient-derived fibroblasts treated with the NMD inhibitor cycloheximide revealed 5 altered MFN2 transcripts that each use a distinct splice acceptor site in lieu of the canonical exon 7 splice acceptor (Figure 1C). Notably, all 5 altered transcripts have disrupted open reading frames that make them subject to NMD, and none of them are present within control fibroblasts (Figure 1D).

MFN2 Branch Point Variant Causes MFN2 Deficiency

To determine whether this branch point variant causes MFN2 deficiency, we analyzed patient-derived fibroblasts for hallmarks of MFN2 dysfunction. Pathogenic MFN2 variants are associated with diverse mitochondrial phenotypes, including increased lipid droplet formation.¹⁵ We found that patient-derived fibroblast cells had both increased number and intensity of lipid droplets compared with control cells (Figure 2), consistent with this branch point variant causing MFN2 deficiency.

Figure 2. MFN2 Branch Point Variant Results in Abnormal Lipid Droplet Formation.

(A) Representative images of control and proband fibroblast cells. Mitochondria were labeled with Mitotracker CMXRos, lipid droplets with Bodipy 493/503, and nuclei with NucBlue. Images represent maximum intensity projections. Scale bar = 5 μm. (B) Violin and swarm plots showing the number of lipid droplets per cell from 3 independent biological replicates where the number of lipid droplets per cell was quantified in 10 distinct fields each containing 13–34 cells. The median is indicated with a thick dashed line and quartiles with fine dashed lines. (C) Fold increase in lipid droplet fluorescence intensity and number in proband compared with that in control.

Discussion

We describe a pathogenic MFN2 intronic branch point variant that causes autosomal recessive CMT2A—expanding our understanding of the molecular basis of CMT2A. Full-length transcript sequencing revealed that all altered transcripts induced by this variant are subject to NMD, consistent with an LOF mechanism for disease. Notably, the proband's asymptomatic father is presumed to be heterozygous for this variant, consistent with prior observations that LOF MFN2 variants only cause disease in the recessive state.

Recent advances in long-read full-length transcript sequencing have the potential to transform clinical workflows for evaluating patients with suspected Mendelian conditions. This study provides a proof-of-concept for the utility of full-length transcriptome data to identify disease-associated variants and to characterize the mechanism by which these variants cause disease. Further studies are needed to fully evaluate the utility of full-length transcript data in clinical practice.

Glossary

CMT2A

Charcot-Marie-Tooth 2A

LOF

loss of function

MFN2

mitofusin 2

NMD

nonsense-mediated mRNA decay

UDN

Undiagnosed Diseases Network

UPD[1]

uniparental isodisomy of chromosome 1

Appendix 1. Authors

Appendix 2. Coinvestigators

Study Funding

A.B. Stergachis holds a Career Award for Medical Scientists from the Burroughs Wellcome Fund and is a Pew Biomedical Scholar. This study was supported by NIH grants 1U01HG010233, 1DP5OD029630, R01GM118509, and U01HG007703, in addition to funds from the Collagen Diagnostic Laboratory, University of Washington. Fibroblast data and analysis were partially supported by the California Center for Rare Diseases within the UCLA Institute of Precision Health. Sequence data analysis was supported by the University of Washington Center for Mendelian Genomics (UW-CMG), which was funded by NHGRI grant UM1 HG006493. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

Disclosure

The authors report no relevant disclosures. Go to Neurology.org/NG for full disclosures.