Targeted exome sequencing for mitochondrial disorders reveals high genetic heterogeneity

Patients with mitochondrial diseases present with widely variable phenotypes that can affect any organ, at any age. This clinical variability and a lack of reliable diagnostic tests can make diagnosing these disorders very challenging. Currently, the diagnosis relies upon the enzymatic analysis of respiratory chain complexes (RCC) in muscle biopsy tissues [1, 2]. Unfortunately, clinical laboratories have considerable differences in their RCC assay protocols and methods of interpretation, leading to inaccurate diagnoses that can affect the quality of patient care [3–5].

Recently, targeted next-generation sequencing (NGS) has been utilized to diagnose patients with several types of disorders [6–11]. For mitochondrial disease, NGS can be used not only to examine genes that have been classically considered to cause disease, such as RCC subunits and assembly factors, but also to analyze genes that are suspected to cause mitochondrial dysfunction, or genes causing conditions that mimic mitochondrial disease [12]. Over 200 nuclear genes have been identified that cause mitochondrial disease, and the list continues to grow as over 1500 genes have been identified controlling mitochondrial structure and function [13, 14]. We recently explored a panel of ~900 known and candidate disease genes for 26 patients with known or suspected mitochondrial disorders [15]. This study was successful in demonstrating both the analytical sensitivity and the clinical utility of NGS by diagnosing several of the patients included. Here, we utilize a smaller gene panel (447 genes) which includes 364 known genes causing mitochondrial disease or similar disease, and 83 candidate genes expected to be involved in critical mitochondrial functions, such as RCC subunits and tRNA synthetases. This gene panel was offered to physicians as a commercially available clinical diagnostic assay through Transgenomic, Inc. The samples were processed, analyzed, and interpreted at a CLIA-approved laboratory at Seattle Children’s Research Institute, and the final report was provided through Transgenomic, Inc. Here, we present data from the first 148 patients.

Diagnosis of mitochondrial disease by NGS of targeted gene panels has been explored on a research basis over the past years [2, 10, 11, 15]. This work has proven to be highly effective in diagnosing patients, and identifying new disease genes [2, 11, 15, 22, 38–41]. This technology is providing insights into the genetic basis of diseases that often remain undiagnosed even after extensive clinical workup [42]. The reliability of our approach has been previously demonstrated from our earlier research studies [10, 15], and is similar to the approaches used for the diagnosis of mitochondrial disorders and other inborn errors of metabolism in other laboratories [2, 40, 43, 44]. This approach has very high analytical sensitivity and specificity as demonstrated by our quality control metrics and use of internal and external controls. The clinical sensitivity and specificity of the approach are less easily defined. Our classification approach utilizes conservative minor allele frequency cut-offs to insure that disease causing variants are not removed from consideration. While this does allow for many rare, benign polymorphisms to be reported as variants of unknown significance, we must rely on the patient’s physician to determine if these variants fit the patient’s clinical presentation.

In this clinical study, patient samples were submitted by physicians, most of whom were experts in the field of mitochondrial medicine and ordered this test because other diagnostic avenues had been unsuccessful and mitochondrial disease was still suspected. As a result, this patient cohort is more representative of the wide variety of patients that are considered to have suspected mitochondrial disease in comparison to the research studies of the past.

Our data highlights the great underlying genetic heterogeneity of suspected mitochondrial disease, but also the difficulty that even experienced physicians endure when attempting to define mitochondrial disorders. Of note, possible disease-causing variants, confirmed by parental samples analysis, were found in only one gene that encodes an oxidative phosphorylation subunit of complex I (NDUFS2) and one gene for the assembly of the complex I (NDUFAF5). In a total of 148 patients we analyzed, there was only one additional gene SCO2 in the category of primary mitochondrial disease causative genes, with variants that we classified as of unknown significance as we did not receive parental samples, but most likely associated with disease.

Our results indicate that the majority of patients that cannot obtain a molecular diagnosis by traditional methods (RCC enzyme assay), have disease that is probably not the result of direct impairment of energy production as we observed in our previous study [15]. Instead, their disease may be caused by secondary inhibition of proper mitochondrial function [30, 45, 46]. This observation is consistent with other studies that utilized targeted sequencing on patients with biochemically confirmed defects [2, 15, 23]. Other NGS studies have also indicated that suspected “mitochondrial patients” were actually affected by conditions caused by mutations in genes that don’t impair mitochondria function [47, 48]. While we can’t entirely exclude that some patients may have mutations in mtDNA, nuclear genes are primarily causative for pediatric cases, which are the majority of the patients studied here [1]. The requisition forms for the patients with variants listed in the Tables indicated that 35 of them had mtDNA sequenced with either normal or inconclusive results.

From our observation, it appears that the clinical differential diagnosis for mitochondrial disease could be even more challenging, as the diagnostic spectrum is much wider than previously thought. Some of our patients could have been diagnosed by standard biochemical or genetic studies, without NGS sequencing. Unfortunately, the lack of clinical phenotypes that are specific to these diseases causes physicians to have a difficult time identifying which biochemical studies should be ordered for each case. A good example is represented by the patient affected by CLN6, a lysosomal storage disorder in which the clinical presentations can easily mislead to focus on mitochondria, leading to muscle biopsy and RCC assay.

While it is undoubtedly important to perform family segregation analysis for the accurate and appropriate interpretation of novel variants, we could not obtain more parental samples in this study as some of parents were not involved in patient’s care. We can nevertheless consider that even when definitive diagnoses cannot be made in the absence of family segregation supporting data, the results can still guide physical examinations or additional laboratory tests to look for signs specific to the diseases for which possibly pathogenic variants were found. Some examples include GAA (Pompe disease), HLCS (Holocarboxylase synthetase deficiency), ABCD1 (X-linked adrenoleukodystrophy), PC (Pyruvate carboxylase deficiency), UBE3A (Angelman syndrome), ASL (Argininosuccinic aciduria), or ZFYVE26 (Spastic paraplegia 15) (Additional file 3: Table S2). Presumptive diagnosis can prompt recommended interventions, including diet modification for certain patients such as for ornithine transcarbamylase deficiency, one of the urea cycle disorders. It is unfortunate that we cannot provide further details on any follow-up biochemical tests that were recommended to the referring physicians. The nature of clinical testing does not allow for us to regularly follow-up on patients tested and how our results affected their care.

Our results highlight the need to have better means to define the significance of variants found. For instance, the variants of unknown significance in genes with AD inheritance were also found in majority of one of presumably healthy parents. While this cannot entirely rule out the possibility of pathogenic involvement, only functional studies can truly assess the significance of these variants. This problem is an even larger issue when whole exome or whole genome sequencing data are interpreted.

Of interest, we have observed that many patients had possibly disease-causing variants in multiple genes. This suggests that some patients with suspected mitochondrial disease may be caused by oligogenic factors, where many mildly dysfunctional proteins may be insufficient to cause disease alone, but when in combination can cause disease in a patient. If this is possible, it even further complicates the molecular diagnosis of these patients and extended studies will be necessary to explore this possibility. We also observed a substantial number of VUS for Charcot-Marie-Tooth (CMT) disease, spinocerebellar ataxia or spastic paraplegia related genes. While these variants still require parental samples analysis to determine the clinical significance, these conditions may be an emerging group of mitochondrial disorders as supported by recent publications with evidence of mitochondrial dysfunction [49, 50].

Our study revealed that a broader diagnostic spectrum including neurodegenerative disorders or myopathy need to be considered during the clinical work-up of suspected mitochondrial patients and in developing future diagnostic tests. It appears that the current standard diagnostic algorithm along with the definition of mitochondrial disorders may require substantial revision to ultimately improve patient quality care. The clinical sensitivity of NGS assays for mitochondrial disorders will be further improved by targeting more genes and ultimately by whole exome sequencing with adequate coverage in the near future.