Long-read sequencing improves diagnostic rate in neuromuscular disorders

Abstract

Massive parallel sequencing methods, such as exome, genome, and targeted DNA sequencing, have aided molecular diagnosis of genetic diseases in the last 20 years. However, short-read sequencing methods still have several limitations, such inaccurate genome assembly, the inability to detect large structural variants, and variants located in hard-to-sequence regions like highly repetitive areas. The recently emerged PacBio single-molecule real-time (SMRT) and Oxford nanopore technology (ONT) long-read sequencing (LRS) methods have been shown to overcome most of these technical issues, leading to an increase in diagnostic rate. 

LRS methods are contributing to the detection of repeat expansions in novel disease-causing genes (e.g., ABCD3, NOTCH2NLC and RILPL1 causing an Oculopharyngodistal myopathy or PLIN4 causing a Myopathy with rimmed ubiquitin-positive autophagic vacuolation), of structural variants (e.g., in DMD), and of single nucleotide variants in repetitive regions (TTN and NEB). Moreover, these methods have simplified the characterization of the D4Z4 repeats in DUX4, facilitating the diagnosis of Facioscapulohumeral muscular dystrophy (FSHD).

We review recent studies that have used either ONT or PacBio SMRT sequencing methods and discuss different types of variants that have been detected using these approaches in individuals with neuromuscular disorders.

Diagnosis of neuromuscular disorders in the 21st century

Next-generation sequencing (NGS) or massive parallel sequencing (MPS) approaches were introduced in the early 21^st century and their clinical use has aided molecular diagnosis of genetic diseases ¹. Interestingly, large genes associated with muscle diseases, such as DMD and TTN, were among the first targets of NGS-approaches ^2,3. Since then, NGS-based genetic tests have been developed to make the diagnosis of rare diseases more effective and accurate ^4-7. Gene panels and whole exome sequencing (WES) have become the first-tier test to identify disease-causing variants in NMDs as well as in other genetic diseases ^8-17. Although these methods have helped identify several new genes and variants causing these diseases, many patients still remain without a molecular diagnosis ^8,18. Clinically, many variants of uncertain significance (VUS) are difficult to interpret ^8,10. Reanalysis of previously identified VUS only partly increases the diagnostic rate, demonstrating that technical aspects may limit the overall detection of DNA causative variants ¹⁹. In particular, the short-read length (usually 50-300 bp) is a clear limitation, especially in sequencing certain genomic regions such as highly repetitive areas or long homopolymers ²⁰. Similarly, localizing large structural variants in their entirety is extremely difficult using short reads ^21-25.

The development of Long-read sequencing

Long-read sequencing (LRS), also called third-generation sequencing, generates reads with a size from 1000 bases to several kilobases (kb) ^21,23. LRS can detect both small and large structural variants, repeat expansions, and even epigenetic modifications, such as DNA methylation, with up to 99.9% accuracy ^21,25,26. Moreover, LRS eliminates bias associated with amplification since isolated DNA can be used directly for sequencing, and instead of clusters, single DNA molecules are sequenced, resulting in an improved coverage. LRS also aids variant phasing ^26,27. LRS can increase the diagnostic rate of genetic diseases (including neuromuscular disorders) and reduce the time it takes to achieve a molecular diagnosis ^{21,24,25,28,29}. Moreover, the combined use of LRS and of a more complete human reference genome (T2T-CHM13), also increases the detection of de novo variants ³⁰.

In 2022, LRS was named the method of the year by Nature Methods due to the new opportunities and improvements it has given both individual labs and large-scale genomics projects. The Vertebrate Genomes Project (VGP) and the Telomere-to-Telomere Consortium (T2T) are two examples of the genomics initiatives that were made possible with the introduction of LRS ³¹.

Single-molecule real-time sequencing (SMRT) by Pacific Biosciences (PacBio) and Nanopore long-read sequencing by Oxford Nanopore Technologies (ONT) are the two leading LRS methods (Fig. 1) ^21,32. Both companies provide long, highly accurate reads in a short turnover time, although the processes work differently from each other. In PacBio SMRT sequencing, hairpin adapters are ligated to both ends of the target double-stranded DNA fragment, creating a template called the SMRTbell. Closed circular DNA molecules are formed, creating a library that is then loaded into flowcells containing nanoscale zero-mode waveguides (ZMWs). Fluorescently labeled polymerase at the base of the ZMWs excite fluorescent signals, which emit light that is captured in real-time, creating a circular consensus sequence (CCS) with high fidelity from multiple reads that cover the entire original DNA template.

Oxford nanopore sequencing involves a nanopore embedded into a synthetic bilayer. With the help of motor proteins, DNA or RNA is unwound and translocated through the nanopore. Exonucleases, meanwhile, cut off individual nucleotides, disrupting the electric current which is captured in real time ²³.

Long read sequencing for the identification and characterization of structural variants

In the field of NMDs research, LRS has been successfully used to detect novel rare structural variants, particularly in the DMD gene for individuals with Duchenne muscular dystrophy (DMD; MIM #٣١٠٢٠٠) ³³. Bruels and colleagues used ONT sequencing to evaluate a cohort of unsolved patients with suspected pathogenic variants in muscular dystrophy genes. They collected blood and saliva samples from ١٢ unsolved individuals belonging to ١٠ different families. The average read length obtained using DNA from saliva samples was ٧٠٠٠ bp, while DNA from blood samples averaged a read length of 8300 bp. In an unsolved Duchenne patient, they identified one 5.9 Mb structural variant (SV); an inversion that disrupts all exons except 1 and 2 in the DMD gene. In addition, a DMD in-frame duplication was identified in an unrelated asymptomatic patient but was interpreted as benign after it was confirmed to be in tandem through nanopore. In another patient, a 3.6 Mb duplication encompassing exon 30, predicted to cause a frameshift and a premature termination, was identified in LAMA٢ in a patient presenting with sporadic congenital muscular dystrophy. The same variant was confirmed in the proband’s mother. Finally, Nanopore also identified two novel splice-altering DMD variants in two other patients. These results were confirmed via PCR and Sanger, and in one case through a minigene splicing assay ³³.

In a study by Geng and colleagues, ONT was performed secondary to RNA sequencing, cDNA capture sequencing, and optical mapping in a single family, but as the primary method in a different family. The long-read sequencing assay was customized with probes to cover the entire DMD gene as well as ٢٠ kb upstream and downstream. In the first family, no pathogenic variants were found through regular genetic testing. Optical genome mapping (OGM) was later performed, and a pathogenic DMD variant was identified. The breakpoints of this inversion were confirmed through LRS. They revealed an 80 bp short interspersed nuclear element (SINE) and a 467 bp long terminal repeat (LTR). The inversion was predicted to affect exons 3-55. In the second family, LRS was directly performed, and an inversion was identified. In addition, several SINEs, LINEs, and LTRs were observed nearby ³⁴.

A novel complex structural variant was identified by Xie and colleagues through long-read whole-genome sequencing (WGS) in a dystrophinopathy patient who remained without a molecular diagnosis after conventional genetic testing. LRS of the DMD gene identified a large-scale inversion/deletion-insertion rearrangement mediated by long interspersed nuclear element-1 (LINE-1) retrotransposons. Long-read WGS confirmed the structural variant was a 982323 bp inversion flanked by a 3719 bp deletion insertion ³⁵. In another recent study, LRS contributed to the identification of two Alu-mediated deletions in the SMN1 gene and to correctly identify the breakpoints ³⁶.

The clinical utility of LR genome sequencing in a prenatal setting for accurate and rapid characterization of structural variants was demonstrated in a complex case with a duplication involving DMD ³⁷. Chin and colleagues used nanopore sequencing to quickly sequence the whole genome of a healthy pregnant individual with a duplication of uncertain significance encompassing a portion of DMD. Her healthy daughter, as well as her male fetus, were found to have inherited the variant. Comparable duplications in the same location had been reported with varying clinical significance from benign to pathogenic. LRS contributed to the identification of the precise breakpoints of the duplication and proved that the duplicated region did not disrupt DMD and was in the same orientation as DMD. The variant was interpreted as likely benign and, later on, its identification in a healthy maternal uncle further supported this interpretation ³⁷.

Repeat expansions and repeat sequences

In two studies by Yu and colleagues, DNA from patients with oculopharyngodistal myopathy (OPDM) types 3 (MIM #619473) and 4 (MIM #619790) was analyzed by long-read WGS using nanopore sequencing, after short-read WGS failed to identify a pathogenic variant. In the earlier study, LRS revealed a heterozygous GGC repeat in NOTCH2NLC in two patients and, after segregation studies, the gene was reported as the disease-causing gene ^38,39. In the later study, a heterozygous CCG repeat upstream of the RILP1 gene was identified. As in the earlier study, the identified CCG repeat was reported as a novel OPDM-causing variant, and this finding was supported by segregation in the family ⁴⁰.

Another recent study identified a CCG expansion, ranging from 118 to 694 repeats, in the 5´ UTR of ABCD3 in OPDM patients from several unrelated families. The expansion results in upregulation of ABCD3 expression ⁴¹.

Using ONT-sequencing, Yeetong and colleagues identified contracted D4Z4 repeats in seven individuals with Facioscapulohumeral muscular dystrophy (FSHD). The D4Z4 array was normal in the control groups, which included healthy individuals and unaffected parents of the FSHD patients ⁴².

Myotonic dystrophy type 1 (DM1; MIM: #160900) is caused by a CTG trinucleotide repeat expansion in DMPK. In some cases, the mutation can be up to 4000 triplets long ^43,44. Mangin and colleagues used SMRT LRS to sequence the DM1 locus in several patients, detecting de novo CCG interruptions and somatic mosaicism⁴⁴. Similarly, Rasmussen and colleagues identified the CTG expansion in a cohort of DM1 patients using a Cas9-enrichment, combined with nanopore sequencing ⁴⁵.

A study by Ruggieri and colleagues identified 99-mer repeat expansion in PLIN4 causing myopathy. They used a multi-omics approach, combining genomic and transcriptomic data. Nanopore LRS of cDNA from RNA extracted from the muscle of affected patients revealed a ٤٠x٩٩-repeat sequence in exon ٣ of PLIN4, compared to the normal 31x99-nucleotide sequence in unaffected individuals. This repeat expansion results in nearly 300 additional amino acids, which leads to an increased PLIN4 expression in some muscle tissues ⁴⁶.

Perrin and colleagues (2022) developed a strategy based on a long-range PCR combined with ONT sequencing to identify variants in the repeated sequences of TTN. The sequencing data helped to assign variants to specific exons and phase variants in the repeated regions ⁴⁷.

Long-read sequencing for haplotyping variants

Haplotyping, using LRS, has been carried out successfully in research of many different diseases, including NMDs, in particular in Spinal muscular atrophy (SMA; MIM IDs: SMA1 #253300, SMA2 #253550, SMA3 #253400, SMA4 #271150) ^48-50. Two recent studies adopted HiFi long-reads generated with SMRT sequencing for haplotyping and phasing variants in the highly homologous SMN1 and SMN2 genes associated with SMA ^48,50. For this purpose, Chen and colleagues developed a new bioinformatics tool called ‘Paraphase’. The pipeline identified full-length haplotypes in both genes, and the samples with more than 20x coverage were distributed into haplogroups through population-wide analysis. Co-segregation of the haplotypes was performed via pedigree-based analysis, which resulted in the identification of ten major haplogroups in SMN1 and nine in SMN2⁴⁸.

By performing a comprehensive SMA trio-analysis (CASMA-trio), Li and colleagues detected silent carriers (SMA 2+0) and ascertained the inheritance patterns of SMN1 and SMN2 haplotypes in most of their families. The CASMA assay identified full-length SMN haplotypes by combining PacBio SMRT LRS with long-range PCR ⁵⁰.

Conclusions

The benefits of using LRS for molecular diagnosis in rare diseases become evident by the increase in diagnostic rates observed in several studies resulting in the discovery of novel causative variants ^30,33,46.

Both PacBio SMRT and ONT offer several advantages over conventional short-read sequencing methods (Fig. 2). While ONT offers the longest reads, SMRT has very high accuracy. Combining these two technologies or coupling them with other strategies, e.g., OGM, could give the most comprehensive result.

LRS is a powerful tool that can identify novel isoforms and transcript variants of very large genes ^51,52. Similarly, LRS may contribute to the characterization of splice variants in genes with complex splice patterns, such as TTN ⁵³. Finally, LRS also shows promise in the detection of structural alterations in the mitochondrial genome ⁵⁴.

LRS has yet to become the standard tool for clinical applications due to its relatively high cost. However, we are observing a continued increase in quality and a decline in cost. Databases for LRS samples and standardized analytic pipelines are still needed. Several recent studies have already made efforts to facilitate the characterization of variants identified through LRS ^48,55,56.

We are at the beginning of a new era of sequencing and novel exciting findings are expected to emerge through LRS.

Acknowledgment

We are grateful to Johanna Ranta-aho for critically reading the manuscript.

Conflict of interest statement

Authors declare no conflict of interest.

Funding

M.S. is supported by Academy of Finland (grant #339437: “Improving the clinical interpretation of sequence variants”), and by Sigrid Jusélius Foundation.

Author contributions

RO: data collection, writing and finalizing the manuscript; MS: critical evaluation, proofreading, and supervision.

Figures and tables

Figure 1. Long-read sequencing technologies by Oxford Nanopore Technologies and Pacific Biosciences. (A) ONT sequencing is based on a nanopore attached to a bilayer that unwinds the DNA/RNA strand. It goes through the nanopore, and individual nucleotides are then cut off with exonucleases, disrupting the electric current. This electric current is measured in real-time and captured in the output. (B) In PacBio SMRT sequencing, an SMRTbell template is formed when hairpin adapters are attached to the double-stranded DNA fragment. The SMRTbell forms a circular consensus sequence (CCS) that is loaded into a flowcell containing ZMWs. A fluorescently labeled polymerase is attached to the CCS at the base of the ZMW, which emits a fluorescent signal that can be measured in real-time. Figure created with Biorender.com

Figure 2. Advantages of long-read sequencing. (1) Comprehensive genome assembly: Longer reads overlap each other, thereby eliminating sequence gaps and covering the entire genome. Larger are also easier to piece together, which makes assembly easier. (2) Sequencing of repetitive sequences: Long-read sequencing technologies allow the sequencing of highly repetitive regions, which is an issue with SRS. (3) Epigenetics: Long-read sequencing does not require amplification of extracted DNA, which allows the detection of epigenomic modifications such as methylation. (4) Structural variant detection: Large complex structural variants have been identified using LRS. The benefit of longer reads is that the entire variant can be covered with a single read ^21,25. (5) Haplotyping/phasing: Instead of combining the maternal and paternal copy of a chromosome, tools for LRS can be used to assemble them separately; thereby, you can determine if an allele is maternally/paternally inherited and if a variant is in cis or in trans ⁵⁷. Figure created with BioRender.com