Exome sequences versus sequential gene testing in the UK highly specialised Service for Limb Girdle Muscular Dystrophy

Patients selected for WES had undergone more extensive prior investigations than the cohort attending the UK-LGMD clinic (Additional file 1: Tables S1 and Additional file 2: Table S2): muscle biopsy was performed in at least one individual in 74/84 (88%) of UK LGMD-clinic families, compared to 74/75 (99%) of WES families; and the mean number of individual genes tested per family was 3 for those undergoing standard testing, compared to 8 genes previously tested in the cohort selected for WES (Fig. 1a).

We compared clinical characteristics which may impact on rate of diagnosis, including the inheritance pattern, serum creatine kinase (CK) levels and age of symptom onset, between the standard testing and WES cohort (Fig. 1b). Both cohorts mainly comprised patients with sporadically occurring disease; the WES cohort had a higher proportion of families with an autosomal recessive inheritance pattern (24% compared to 8% in the standard testing cohort). Similar proportions of patients had dominant or X-linked inheritance patterns, and rates of consanguinity were similar between groups (Fig. 1). The majority of patients in both the WES and standard testing cohorts had an elevated CK level (59% in standard testing and 65% in WES cohort). A greater proportion of patients in the standard testing cohort had adult onset of symptoms (59%, compared to 38% in the WES cohort). In the combined WES and standard testing cohorts the five most frequently tested genes were sequencing of ANO5, CAPN3, FKRP, LMNA and sequencing of TTN exons 293 and 308, which are mutation hotspots for Hereditary Myopathy with Early Respiratory Failure, a relatively frequent diagnosis in the North of England [29].

In the WES cohort a genetic diagnosis due to mutations in a known disease gene were made in 25/75 of families (Fig. 2), and disease due to mutation(s) in a novel disease genes (ZAK [30], and two currently unpublished genes) was made in a further 6 individuals from 3 families (3/75), making the overall rate of diagnosis 28/75 (37%). Using standard genetic testing a diagnosis was obtained for 28/84 (33%) families (Fig. 2). Fifteen panel tests were requested in 14 families in the standard testing cohort and one panel test was request in a family in the WES cohort (details of panel tests are included in Additional file 3). As a result of panel testing a single family was diagnosed with a peripheral neuropathy due to a MFN2 mutation (Additional file 1: Table S1).

A specific family structure was not required for an individual to be selected for WES, and the majority of patients included were single individuals with no family history or additional family members available for sequencing. The diagnostic rate in sporadically occurring disease was 32% (15/47 families), and where only index patients’ DNA was available was 34% (13/38 families). The highest diagnostic rate was obtained in families with an autosomal dominant history, where 6 out of 8 families were diagnosed (75%). In families with an autosomal recessive inheritance pattern 5 of 17 families were diagnosed (29%). In the nine families where trio WES data was obtained a diagnosis was obtained in two families (22%).

Diagnoses made by standard genetic testing are shown in Fig. 1a, and include LGMDs as well as diseases with overlapping phenotypes, such as collagen VI-related dystrophy (COL6-RD) and myofibrillar myopathies. Notably diagnoses in three families, of fascioscapulohumeral muscular dystrophy (FSHD) types 1 and 2 and Duchenne manifesting carrier due to an intragenic deletion, would have been missed by WES given the limitations of WES currently for detecting copy number changes or more complex genetic abnormalities as in FSHD.

Disease causing mutations in 17 previously known muscle disease genes were identified by WES (Fig. 2b). Additional clinical investigations to evaluate the plausibility of a diagnosis suggested by WES findings, so-called ‘reverse phenotyping’, were required in most instances to confirm the diagnosis indicated by WES (Table 1). The most frequent genetic diagnosis made by WES was disease due to mutations in TTN, as this large gene was not routinely sequenced in the pre-NGS era. Titinopathy phenotypes encompass a wide spectrum and in this study distal myopathy (family WES5), congenital myopathy (families WES18 and WES23), and LGMD (families WES19, WES29, WES61 – reported in detail elsewhere [31]).

Explanations for why the genetic diagnoses made by WES had been missed by standard diagnostic pathways are summarised in Table 2. In five instances mutations in known genes were identified in patients with typical phenotypes where this particular genetic test had not yet been requested due to the number of candidate diagnoses associated with a non-specific phenotype. At the time of study GMPPB and STIM1 were recently identified myopathy genes and therefore their testing was not yet part of routine diagnostic screening. Diagnosis in an MTM1 manifesting carrier (WES67) and a family with distal arthrogryposis due to PIEZO2 mutation (WES13) were missed due to the rarity of these disorders [32‐34], which meant that the necessary genes had not been selected for screening by standard testing.

Several patients had not been diagnosed due to the limitations of phenotype-driven testing. These include an individual with normal dysferlin expression on immunoanalysis who was compound heterozygous for two missense mutations in dysferlin, a rarely reported phenotype which may become more common as genetic testing may precede muscle biopsy [8, 35]. Similarly one patient with adult onset symptoms had compound heterozygous mutations in a congenital myopathy gene, MEGF10 (WES36), who would not have been diagnosed by phenotype driven testing as diagnosis in this individual expands the phenotype associated with mutations in this gene. In one family (WES30) the apparent inheritance pattern, with an affected male proband, affected maternal uncle and unaffected mother, had directed testing to X-linked genes, but in fact the phenotype was compatible with autosomal dominant disease due to MYH7 mutation as identified on WES with non-penetrance in the mother.

The second most frequent diagnosis made by WES and also frequently made in the standard testing cohort (Fig. 2) was collagen VI-related dystrophy (COL6-RD). Patient WES9 had a typical COL6-RD phenotype, and had had prior diagnostic Sanger sequencing of COL6A1, COL6A2 and COL6A3 which identified five variants of uncertain clinical significance (Fig. 3). We observed a discrepancy in detection of variants in these genes by Sanger sequencing and by WES (Fig. 3a). Following repeated Sanger sequencing of variants detected by WES we noted that a COL6A1 variant (c.957G > T p.Lys319Asn) detected by WES but not by initial Sanger sequencing was present at a low level in DNA from patient blood and cultured fibroblasts but not in DNA from either parent, consistent with a de novo mosaic mutation (Fig. 3d). One patient with this mutation has been previously reported [36] with a more severe phenotype (loss of ambulation aged 7 years, forced vital capacity (FVC) of 15% and non-invasively ventilated aged 11 years) than our patient who is now 25 years old and remains ambulant with an FVC of 56% at last assessment. There was evidence against the pathogenicity of the other identified variants in COL6A1, COL6A2 and COL6A3 (Fig. 3). Given that mosaicism for collagen VI genes is a recently recognised cause of disease [37] we feel that the de novo mosaic COL6A1 variant is likely to be the cause of disease in our patient. We observed asymmetry in the pattern of muscle involvement on muscle MRI (Fig. 3e) and on clinical assessment, which is atypical for COL6-RD and could be explained by somatic mosaicism.

Sequential gene by gene testing and gene panels as were routinely available within the National Health Service (NHS) during the time period examined are referred to as “standard testing”. Details of the gene panels used are included in Additional file 1: Tables S1 and Additional file 2: Table S2. There was no LGMD gene panel available for diagnostic use through the NHS during the period studied. Data, including genetic tests performed, serum CK level, age of symptom onset, inheritance pattern and whether a muscle biopsy was performed, was collected for all undiagnosed patients attending the UK-LGMD clinic from 1st April 2013 to 31st March 2015. Six patients attending clinic in this time period underwent WES as part of this research study and therefore their data have been excluded from the analysis of standard genetic testing. All investigations were performed as part of standard clinical care.

Patients were selected from a database of those attending the UK-LGMD clinic between 2002 to present day. Individuals presenting with limb girdle weakness of a probably genetic aetiology who had been assessed by our neuromuscular specialist clinic at any time since 2002, and in whom genetic testing to date was negative, were selected. Priority was given to family structures likely to be most informative, for example those consistent with autosomal recessive inheritance or consanguineous families; however such families were relatively uncommon in this cohort of patients.

Affected individuals underwent WES; in families with apparent autosomal recessive inheritance two affected siblings were sequenced, and in those with autosomal dominant inheritance two affected family members were sequenced. If this did not yield a diagnosis additional affected or unaffected family members were subsequently sent for WES as appropriate.

WES in 30 individuals (20 affected, 10 unaffected) from 9 families was performed at de CODE genetics (Iceland) using Illumina Nextera Rapid Capture exome kit (37 Mb) and sequenced 90 nt paired-end on Illumina HiSeq2000 sequencer. WES in a further 104 (84 affected, 20 unaffected) individuals from 65 additional families was performed at the Genomics Platform at the Broad Institute of Harvard and MIT (Cambridge, MA, USA), with Agilent Sure-Select Human All Exon v2.0 (44 Mb) on Illumina HiSeqXs platform (Illumina, Inc. San Diego, CA). Sequence alignment, variant calling and functional annotation were performed with Burrows Wheel Aligner, Genomes Analysis Tool Kit and Variant Effect Predictor (VEP).

Analysis of WES data was with proprietary software from deCODE genetics (Clinical Sequence Miner) and The Broad Institute (xBrowse) respectively. Variants were filtered to include those that were rare (MAF <0.01) in publically available reference datasets (ExAC and 1000 Genomes) and predicted to be moderately-severely damaging by VEP. Analysis was initially performed to identify variants in known muscle disease genes (www.​musclegenetable.​fr). Where a putatively pathogenic variant in a known gene was identified this was evaluated according to current guidelines including assessment of frequency, in silico predictions of pathogenicity, review of previous reports of patients with same variants, and published experimental work to review the effect of a variant on protein function [53, 54].

Putative genetic diagnoses were correlated with clinical findings including for example muscle MRI, clinical review of phenotype and assessment for extra-neuromuscular features as suggested by genotype. Where relevant and possible genetic diagnoses were correlated with histopathology phenotype, by identification of characteristic findings on muscle biopsy histopathology and immunoanalysis using antibodies typically using a standard dystrophy panel [55] as carried out by the diagnostic service provided by the Muscle Immunoanalysis Unit, which is part of the UK HSS LGMD referral service. Genetic findings, clinical features, muscle MRI and muscle biopsy results were then integrated at a multidisciplinary meeting to reach a consensus as to whether variants were likely to be disease causing.

In the absence of a positive diagnosis due to mutation in a known muscle disease gene additional affected or unaffected family members underwent WES and likely inheritance pattern and the scope of variant analysis was widened to include variants in novel genes.