New perspective in diagnostics of mitochondrial disorders: two years’ experience with whole-exome sequencing at a national paediatric centre

Background Whole-exome sequencing (WES) has led to an exponential increase in identification of causative variants in mitochondrial disorders (MD). Methods We performed WES in 113 MD suspected patients from Polish paediatric reference centre, in whom routine testing failed to identify a molecular defect. WES was performed using TruSeqExome enrichment, followed by variant prioritization, validation by Sanger sequencing, and segregation with the disease phenotype in the family. Results Likely causative mutations were identified in 67 (59.3 %) patients; these included variants in mtDNA (6 patients) and nDNA: X-linked (9 patients), autosomal dominant (5 patients), and autosomal recessive (47 patients, 11 homozygotes). Novel variants accounted for 50.5 % (50/99) of all detected changes. In 47 patients, changes in 31 MD-related genes (ACAD9, ADCK3, AIFM1, CLPB, COX10, DLD, EARS2, FBXL4, MTATP6, MTFMT, MTND1, MTND3, MTND5, NAXE, NDUFS6, NDUFS7, NDUFV1, OPA1, PARS2, PC, PDHA1, POLG, RARS2, RRM2B, SCO2, SERAC1, SLC19A3, SLC25A12, TAZ, TMEM126B, VARS2) were identified. The ACAD9, CLPB, FBXL4, PDHA1 genes recurred more than twice suggesting higher general/ethnic prevalence. In 19 cases, variants in 18 non-MD related genes (ADAR, CACNA1A, CDKL5, CLN3, CPS1, DMD, DYSF, GBE1, GFAP, HSD17B4, MECP2, MYBPC3, PEX5, PGAP2, PIGN, PRF1, SBDS, SCN2A) were found. The percentage of positive WES results rose gradually with increasing probability of MD according to the Mitochondrial Disease Criteria (MDC) scale (from 36 to 90 % for low and high probability, respectively). The percentage of detected MD-related genes compared with non MD-related genes also grew with the increasing MD likelihood (from 20 to 97 %). Molecular diagnosis was established in 30/47 (63.8 %) neonates and in 17/28 (60.7 %) patients with basal ganglia involvement. Mutations in CLPB, SERAC1, TAZ genes were identified in neonates with 3-methylglutaconic aciduria (3-MGA) as a discriminative feature. New MD-related candidate gene (NDUFB8) is under verification. Conclusions We suggest WES rather than targeted NGS as the method of choice in diagnostics of MD in children, including neonates with 3-MGA aciduria, who died without determination of disease cause and with limited availability of laboratory data. There is a strong correlation between the degree of MD diagnosis by WES and MD likelihood expressed by the MDC scale. Electronic supplementary material The online version of this article (doi:10.1186/s12967-016-0930-9) contains supplementary material, which is available to authorized users.

A particular challenge is the diagnosis of MD in neonates below 3 months of age as these patients may account for up to 30 % of all MD cases [12,13]. However, so far, this group has not been specifically focused on in terms of diagnostic effectiveness of WES. The prevailing majority (96.5 %) of cases with a molecular diagnosis of MD established at our national reference centre until 2013 included children older than 3 months, indicating considerable under-diagnosis rates in the youngest infants in the Polish population. We have achieved some improvement in neonatal MD detection by performing targeted DNA sequencing (frequently post mortem) in cases of neonates with lactic aciduria (LA-uria) found in selective GC-MS screening, including over 90 % of SCO2 [14] and DGUOK [15] deficiencies, and ~ 50 % of SURF1 deficiency [16].
The purpose of our study was to evaluate WES as a tool for diagnosis of MD depending on the disease probability assessed according to mitochondrial disease criteria (MDC) [17]. We considered both patients with full-range mitochondrial diagnostics (Leigh syndrome features in MRI and/or muscle biopsy evaluation) and those in whom only fragmentary clinical data e.g. abnormal result of GC-MS screening indicating the presence LA-uria and/or 3-methylglutaconic aciduria (3-MGA-uria) were available.

Patients
WES was performed in patients with probable or possible MD, in whom a molecular defect had not been identified within the analysed period. In the retrospective subgroup (88/113 patients) the lag time was 2-25 years (mean 7.5 +/5.9 years). Since 2013 WES has been considered in consecutive patients (25/113). To undergo WES, a patient had to fulfil at least one of the following criteria: 1/neonatal onset; 2/basal ganglia involvement (Leigh syndrome-LS, nonspecific basal ganglia involvement); 3/increased 3-MGA in urine (patients recruited from a group of >250 cases of 3-MGA aciduria identified by national selective GC-MS screening for metabolic disorders since 2000), and 4/genetic counselling demands. Access to biological material and informed consent of parents were sine qua non conditions for participation in the study. Details of criteria for patient selection and their clinical characteristics are shown in Table 1 and Additional file 1: Table S1.
The study included cases with a high probability of MD and those in whom MD was considered possible. The level of probability was assessed according to the MDC score proposed by the Nijmegen mitochondrial team as follows: 2-4 points: MD possible; 5-8 points: MD probable [17]. The MDC scoring for this study did not include the results of muscle biopsy (panels A+B, without C). The mean MD score in the study group was 4.1 ± 1.5 (range 2-8). Muscle biopsy with subsequent OXPHOS evaluation was performed in 67 cases, and autopsy in 15 cases. The family history was positive in 26 cases and three couples were consanguineous.
In the retrospective group, DNA was isolated from fibroblast cultures or frozen tissue samples obtained by muscle/liver biopsy or by autopsy. Whenever possible, skeletal muscle was preferred. In the remaining cases, DNA was isolated from blood. Throughout the paper the genes were classified as MD-related if they had a connection with mitochondrial disorders documented in the literature [9] or non MD-related when this was not the case.
Parents of the patients gave informed consent for the WES analysis. The study protocol was in agreement with the Helsinki Convention and the study was approved by the Ethics Committee of The Children's Memorial Health Institute.

Whole-exome sequencing
WES was performed using TruSeqExome Enrichment Kits according to the manufacturer's instructions (Illumina). The samples were run on 1/4 of a lane on HiSeq 1500 using 2 × 100 bp paired-end reads. Bioinformatics analysis was performed as previously described [18]. Briefly, after initial processing with CASAVA, the sequencing reads were aligned to the hg19 reference availability of laboratory data. There is a strong correlation between the degree of MD diagnosis by WES and MD likelihood expressed by the MDC scale.
Keywords: Whole-exome sequencing, Mitochondrial disorders, Mitochondrial disease criteria scale, Neonates, Basal ganglia involvement, Leigh syndrome, 3-methylglutaconic aciduria, Novel mutation, Candidate gene genome with the Burrows-Wheeler Alignment Tool and further processed by Genome Analysis Toolkit [19]. Base quality score recalibration, indel realignment, duplicate removal, and SNP/INDEL calling were done as described [20]. The detected variants were annotated using Annovar and converted to MS Access format for final manual analyses. Alignments were viewed with Integrative Genomics Viewer [21,22]. The complete results of WES, including VCF and/or FASTQ files, are available on demand to qualified researchers. All samples were sequenced so that min. 80 % of target was covered 20× or more. The presence of the variants identified by WES was confirmed by Sanger sequencing.

Results
Among 67 probands, we found 99 variants in 49 different genes with a Known disease link ( Table 2). They were variants in mtDNA (6 patients) and nuclear DNA (nDNA): X-linked (9 patients), autosomal dominant (5 patients), and autosomal recessive (47 patients), including 11 homozygotes. In 50.5 % (50/99) the detected variants were novel (Table 3). Sixty-six of the variants found in the study group occurred in MD-related genes, whereas 31 were found in non MD-related loci. In addition, deleterious variants in a gene not previously linked to disease in humans were identified in one proband ( Table 2).
Mutations in MD-related genes were found in 47 probands. Identified pathogenic variants in 31 different genes included 27 located in nDNA and 4 in mtDNA (Table 2). Eleven genes were found defective more than once (PDHA1-4x, ACAD9, CLPB, and FBXL4-3x, COX10, EARS2, MTND1, MTND5, PC, RRM2B, SLC19A3-2x). The majority of these genes were not previously screened for in our mitochondrial diagnostic centre, with the exceptions of TAZ, PDHA1 [23], SCO2, and the genes encoding MTND and MTATP subunits. Below we present the results that were analysed according to selected phenotypic features (neonatal onset, basal ganglia involvement, 3-MGA) and MD likelihood.

Subgroup with 3-methylglutaonic aciduria
Positive WES results were obtained in seven of 16 patients with persisting 3-MGA (Fig. 1b). In two subjects [P28 and P37] we found mutations in TAZ and SERAC1 genes known to cause mitochondrial diseases with 3-MGA as a discriminative feature [24]. Ex post it was apparent that earlier some important clinical features, including hearing impairment in the patient with SERAC1 mutations and increased excretion of 3-MGA in the terminal stage in the boy with the TAZ mutation, had been overlooked.
In three unrelated 3-MGA neonates included in this study, we identified mutations in the CLPB gene, whose link to human disease was subsequently established [25]. Two of them [P5 and P27] have already been reported in the first disease description [25].
Additionally, in two 3-MGA patients [P13, P40] we found molecular variants in the CPS1, a non MD-related gene linked to urea cycle disorder. In remaining patients in whom the reason for inclusion in the study group was a single GC-MS assessment (ACAD9 and MYBPC3 patients [P15, P8]), increased excretion of 3-MGA has been apparently transient or it was within normal limits after quantitative verification (Additional file 1: Table  S1). Since traces of 3-MGA excretion were also found in a number of healthy siblings and parents of the patients the transient or mild increase in patients was most likely without a causal relationship.
In three patients with basal ganglia involvement one MD-related candidate (NDUFB8) and two known non MD-related genes (ADAR, CDKL5) were identified.  (Table 2; Additional file 1: Table S1).

New MD-related disorders
While our project was ongoing new candidate genes found by us including PARS2 [26] and CLPB have been described by other research teams [25]. The causal role of another two of our candidates has been recognized even more recently. The NAXE gene (APOA1BP according to old nomenclature), a susceptibility locus for migraine [27], in which likely pathogenic variants were found by us in two brothers with a fatal encephalitis-like disorder [P12], has been described in April 2016 as the cause of lethal infantile leukoencephalopathy in a large consanguineous family [28]. A homozygous variant in the TMEM126B gene encoding a subunit required for mitochondrial complex I assembly [29,30], found by us in a complex I deficient girl with extra-neurological presentation [P59], has been discovered and verified functionally as a cause of the disease in a subset of other patients (ESHG 2016, Alston et al.).
The interesting remaining candidate for a novel disease gene identified in our study is NDUFB8. Compound heterozygosity for two variants in NDUFB8 was found in a boy with a typical course of LS and complex I deficiency in muscle homogenate [P26] (Additional file 1: Table S1). NDUFB8 [31] encodes a known subunit of complex I, but, to the extent of our knowledge, its association with complex I deficiency and LS in humans has not been published so far.

Mitochondrial disease criteria score
In the studied cohort there were 40 patients with high probability of MD, i.e., with an MDC score above 4 (5-8, criteria A+B, without C). Positive WES results were obtained in 36 of them (90 %). In this group, pathogenic variants were found mainly in MD-related genes (CPS1 being the exception). WES failed in four patients [P49, P62, P77, P105] with an MDC score above 4. Some of them were found to carry a deleterious variant in one of the known MD-related genes only on one allele. The definite diagnosis still remains open in these cases. Bioinformatics tools for identification of structural variants using NGS have not been applied to our data so it is possible that in some cases the disease may be caused by large deletions/duplications. The complete lists of variants detected in the subjects without fully conclusive results and/or the respective FASTQ files are available on demand to qualified researchers.
Intermediate probability of MD (MDC = 4) was associated with the occurrence of variants in both MD-related and non MD-related genes, in ten (10/31) and six (6/31) patients, respectively. MD-related genes were repre- The percentage of positive results rose gradually as the likelihood of MD increased, as shown by the MDC score (Fig. 2). In the subset of high probability of MD (MDC above 4), the detection percentage reached 90 %. There was a broad range of MD-related genes ( Table 2). Only one non MD-related gene (CPS1) was found in a neonate with a MDC score of 5.
The participation of detected MD-related genes as compared with non MD-related genes also grew as the likelihood of MD probability increased (from 20 to 97 %, data not shown).

WES diagnostics of current cases vs. archival DNA samples
Characteristics of the patients stratified by the waiting period between disease onset and WES qualification into archival material and current diagnostics subset is shown in Table 4. WES efficacy assessed as percentage of molecularly confirmed diagnoses was comparable being higher than 50 % in both subsets. Contribution of MDrelated genes expressed by the ratio of MD-related/non MD-related genes was higher in the archival than current subset (3.4 vs. 1.0, respectively) indicating that this subset contained more patients with non-mitochondrial genetic disorders and that our current qualification for WES became less demanding.

Muscle biopsy findings
OXPHOS assessment available for 67 muscle homogenates showed isolated complex I deficiency in 16 cases, complex IV deficiency in 6 cases and combined OXPHOS defect in 10 cases. There were unspecific changes in 22 bioptates and normal OXPHOS activity in 10. The results were not conclusive in three cases due to technical problems (too small muscle specimen, low protein concentration, low citric synthase activity). Complex

Verification of mitochondrial genome variants
Interestingly, in six patients with typical MD phenotype the search for pathogenic variants in MD-related nuclear genes by WES was negative yet pathogenic variants were found in mtDNA. Each mtDNA variant identified by WES, was subsequently verified by Sanger sequencing using specific primers for mitochondrial genome. All detected changes are known and have been repeatedly reported. Examination of different tissues in probands and maternally related family members showed varying levels of heteroplasmy (Fig. 3).

Discussion
Our results confirm that the implementation of WES led to a significant breakthrough in the diagnostics of MD in children [32]. This is expressed by both the increased number of identified genes and faster establishment of final diagnosis. The total number of genes with likely causative defects found in the present work was 47, a very satisfactory diagnostic yield when compared with 8 genes identified by us by single-gene Sanger sequencing before the introduction of WES (203 such diagnoses per ~1200 patients studied in the period from 1996 to 2013).
In our study we observed a pronounced upward trend in the detection of the molecular background of mitochondrial diseases that was associated with increased MD probability (Fig. 2). According to the MDC scale that we used, a final genetic diagnosis was achieved in over 90 % of patients with the highest MDC scores (5-8 points). In all such cases (with one exception for a neonate with CPS1 mutation), variants were found exclusively in MD-related genes. The diagnostic yield was the lowest (36 %) in the patients with low MD suspicion (MDC score 2-3), and most of the variants in this group were present in non MD-related genes. A similar correlation between detection rate and the level of MD probability was described recently in a similar patient group studied by WES at the Nijmegen Mitochondrial Centre [10]. However, our results differed from that study in terms of the scope of detected defects. In our cohort, mutations in MTO1, TK2, C12orf65, COA6, TUFM, GFM1 were absent and the defects in nuclear encoded complex I subunits are different. This may be a result of random patient selection, but we should also take into account ethnic differences among European populations, e.g., the Slavonic vs. north-western European populations.
It is worth emphasizing that in some cases WES allowed for a diagnosis in statu nascendi, that is, at the time of the first publication of the new gene. This concerned, for example, mutations in CLPB [25,60], PARS2 [26], FBXL4 [61,62] 14:174 According to published literature, every third paediatric MD case (approximately 30 % of all MD diagnoses in this age group) manifests clinically shortly after birth [12,13]. The fatal outcome in such cases precludes transport to a reference centre and proper mitochondrial diagnostics. We have previously shown significantly reduced (up to ten times, about 3 % of all diagnoses) recognition of MD in this age group in Poland [16]. Therefore, neonates with suspected MD intentionally constituted a significant proportion of patients (47/113) undergoing WES in the present study.
Surprisingly, in the neonatal subgroup WES proved to be particularly useful, allowing identification of pathogenic variants in 24 various genes in 63.8 % of patients, including those without muscle biopsy or even autopsy. Our results extend the list recommended by Honzik [13] for neonatal MD diagnostics by at least 15 genes (MDrelated: RRM2B, CLPB, ACAD9, FBXL4, PC, AIFM1, SLC25A12, MTND5, NDUFS6 and non MD-related: CPS1, PGAP2 and more).
In the LS subgroup WES expanded the set of patients from our centre diagnosed with complex I deficiency by three known genes: NDUFS6 [63,64], NDUFV1 [65,66], NDUFS7 [67], a new candidate NDUFB8 [68] and five MTNDs mentioned above. Despite this, complex I deficiency continues to be underrepresented in our cohort in relation to complex IV deficiency because of the high carriage rate of SURF1 mutations in Poland [69]. In a number of cases with basal ganglia brain changes, WES failed to show mutations in known LS-associated genes. This was especially the case in patients without lactic acidaemia and MDC scores below 5 (MD possible but not likely). We speculate that other, still unknown, genes or non-genetic factors might influence the occurrence of LS-brain changes.
Taken together, our results indicate that WES rather than targeted NGS should be the method of choice for MD testing, at least until all MD-associated genes are identified. Furthermore, the rationale for choosing WES in MD-suspected neonates is the non-specificity of symptoms and overlapping results of biochemical tests with non-mitochondrial errors of metabolism.
In 50.5 % the molecular variants were novel (Table 3). However, a number of recurrent rare pathogenic variants found in some recently discovered MD genes (p.Arg22* in FBLX4, p.Arg518Cys in ACAD9, p.Arg417* in CLPB and c.1822_1828+10delinsACCAACAGG in SERAC1) may extend the ethnic specificity of MD in the Polish population reported earlier by us for variants p.Glu140Lys in SCO2 [14] and c.845_846delCT in SURF1 genes [69]. Confirmation of these findings could facilitate in-house diagnostics in selected suspected cases.

Conclusions
1. In a nationwide reference centre, WES provided positive results in >90 % of children with high likelihood of MD (MDC score above 4); 2. WES should be recommended for diagnostics of mitochondrial pathology considering remarkable representation of non MD-related genes among causal factors in patients with lower likelihood of MD, as well as a possibility to discover new mitochondrial genes; 3. WES significantly improves recognition of MD in newborns, even in the case of limited availability of appropriate diagnostic procedures; 4. Despite being a sine qua non for certain diagnoses 3-MGA is not a universal marker of mitochondrial dysfunction; 5. Recurrent variants recognized in some relatively new MD genes (FBLX4, ACAD9, and CLPB) may extend the known ethnic specificity of MD in the Polish population reported earlier for SCO2 and SURF1 variants.