Article Text

PDF

Diagnosing childhood-onset inborn errors of metabolism by next-generation sequencing
  1. Arunabha Ghosh1,2,
  2. Helene Schlecht1,
  3. Lesley E Heptinstall1,
  4. John K Bassett1,
  5. Eleanor Cartwright1,
  6. Sanjeev S Bhaskar1,
  7. Jill Urquhart1,
  8. Alexander Broomfield1,
  9. Andrew AM Morris1,
  10. Elisabeth Jameson1,
  11. Bernd C Schwahn1,
  12. John H Walter1,
  13. Sofia Douzgou1,2,
  14. Helen Murphy1,
  15. Chris Hendriksz3,4,
  16. Reena Sharma3,
  17. Gisela Wilcox3,
  18. Ellen Crushell5,
  19. Ardeshir A Monavari5,
  20. Richard Martin6,
  21. Anne Doolan7,
  22. Senthil Senniappan8,
  23. Simon C Ramsden1,
  24. Simon A Jones1,2,
  25. Siddharth Banka1,2
  1. 1 Manchester Centre for Genomic Medicine, St. Mary’s Hospital, Central Manchester NHS Trust, Manchester Academic Health Science Centre, Manchester, UK
  2. 2 Divison of Evolution and Genomic Sciences, School of Biology, Faculty of Biology, Medicine and Health, University of Manchester, Manchester, UK
  3. 3 Adult Inherited Metabolic Disorders, The Mark Holland Metabolic Unit, Salford Royal NHS Foundation Trust, Salford, UK
  4. 4 Department of Paediatrics and Child Health, Steve Biko Academic Unit, University of Pretoria, Pretoria, South Africa
  5. 5 National Centre for Inherited Metabolic Disorders, Temple Street Children’s University Hospital, Dublin, Ireland
  6. 6 Institute of Human Genetics, The International Centre For Life, Newcastle, UK
  7. 7 Cork University Maternity Hospital, Cork, Ireland
  8. 8 Department of Endocrinology, Alder Hey Children’s Hospital, Liverpool, UK
  1. Correspondence to Dr Siddharth Banka, Manchester Centre for Genomic Medicine, Manchester Academic Health Science Centre, Manchester M13 9WL, UK; siddharth.banka{at}manchester.ac.uk

Abstract

Background Inborn errors of metabolism (IEMs) underlie a substantial proportion of paediatric disease burden but their genetic diagnosis can be challenging using the traditional approaches.

Methods We designed and validated a next-generation sequencing (NGS) panel of 226 IEM genes, created six overlapping phenotype-based subpanels and tested 102 individuals, who presented clinically with suspected childhood-onset IEMs.

Results In 51/102 individuals, NGS fully or partially established the molecular cause or identified other actionable diagnoses. Causal mutations were identified significantly more frequently when the biochemical phenotype suggested a specific IEM or a group of IEMs (p<0.0001), demonstrating the pivotal role of prior biochemical testing in guiding NGS analysis. The NGS panel helped to avoid further invasive, hazardous, lengthy or expensive investigations in 69% individuals (p<0.0001). Additional functional testing due to novel or unexpected findings had to be undertaken in only 3% of subjects, demonstrating that the use of NGS does not significantly increase the burden of subsequent follow-up testing. Even where a molecular diagnosis could not be achieved, NGS-based approach assisted in the management and counselling by reducing the likelihood of a high-penetrant genetic cause.

Conclusion NGS has significant clinical utility for the diagnosis of IEMs. Biochemical testing and NGS analysis play complementary roles in the diagnosis of IEMs. Incorporating NGS into the diagnostic algorithm of IEMs can improve the accuracy of diagnosis.

  • Inborn Errors of Metabolism
  • Metabolic disorders
  • Next Generation Sequencing

Statistics from Altmetric.com

What is already known on this topic?

  • Next-generation sequencing (NGS) has changed the diagnostic paradigm for a number of rare genetic disorders.

  • Genetic testing in inborn errors of metabolism (IEMs) is important for accuracy of diagnosis to provide information on prognosis for cascade carrier testing and for prenatal or preimplantation diagnosis.

  • Although research-based NGS testing for IEM has been developed, the place of NGS in a clinical diagnostic setting has not been established.

What this study adds?

  • NGS-based testing can improve the accuracy of diagnosis of IEM and potentially avoid the need for invasive or expensive functional testing.

  • Biochemical testing and NGS are complementary in the diagnosis of IEM.

  • NGS testing in patients with suspected IEM does not significantly increase the burden of subsequent follow-up testing.

Introduction

Inborn errors of metabolism (IEMs) result in a range of childhood-onset or adult-onset phenotypes and are caused by the disruption of biochemical pathways. IEMs occur in 1 in 800–2500 births.1 2 Metabolic investigations are an integral part of their diagnostic process3 4 and are essential for enabling management that may prevent disability, limit disease or be curative.5–9 IEM investigations can be broadly divided in two overlapping categories—(1) screening tests (newborn screening or phenotype directed) to detect abnormal biochemical markers (eg, acylcarnitine profile, organic acid profile in urine or amino acid profiles in plasma/urine) and (2) specific assays to detect precise deficiencies (eg, specific pathognomonic biomarkers, cellular complementation or enzyme activity assays).10 However, some IEMs lack reliable biochemical markers or if present, the relevant investigations can be invasive, prohibitively expensive, time consuming or offered only on an ad hoc basis by research laboratories. Even for IEMs with accessible biochemical markers, genetic testing is often needed to improve the accuracy of diagnosis, determine prognosis, for cascade carrier testing and prenatal or preimplantation diagnosis. In the era of personalised medicine, genetic testing will facilitate emerging genotype-specific therapies.11

Next-generation sequencing (NGS) has changed the diagnostic paradigm of rare genetic disorders.12–17 Diagnostic yield is a key parameter of the clinical utility of NGS.18 Systematic evaluation of diagnostic yields of NGS across different disease groups will help in the rational integration of NGS into clinical practice. IEMs are potentially excellent candidates for an NGS-based approach due to their extreme genetic heterogeneity and complexities of the traditional diagnostic approach. However, the role of NGS in the diagnostic algorithm of IEMs and its broader clinical impact remain to be established. Here, we report 102 individuals who underwent NGS for suspected childhood-onset IEM in a clinical diagnostic setting.

Methods and results

Panel design and sequencing

A list of over 500 known IEM genes was prepared in April 201419 (www.omim.org). From this list, we excluded mitochondrial disorders and congenital disorders of glycosylation because their genetic heterogeneity remains substantially unresolved (online supplementary table 1). This makes whole exome (WES) or whole genome sequencing (WGS) more appropriate for these two groups. Finally, we selected 226 IEM genes with predominantly neurodevelopmental phenotypes that are genetically heterogeneous or are diagnosed through specialist investigations which are invasive, expensive or have poor availability. We created overlapping subpanels based on biochemical and clinical phenotypes to optimise the bioinformatic analysis and minimise secondary findings. The subpanels were: (1) amino acid and neurotransmitter defects, (2) organic acidaemias and vitamins-related disorders, (3) disorders of fatty acid oxidation or ketone metabolism and hyperammonaemia, (4) carbohydrate metabolism defects, (5) lysosomal disorders and neuronal ceroid lipofuscinoses and (6) peroxisomal disorders. A full list of the selected genes and the subpanels is provided in online supplementary table 2.

Supplementary Material

Supplementary file

We designed a targeted enrichment to sequence all the exons (+/−50 base pairs) of the canonical transcripts of the selected 226 genes. In addition, we included all the published pathogenic intronic variants for the selected genes (information taken from http://www.hgmd.cf.ac.uk). Using manufacturers’ protocols, DNA samples from peripheral blood were enriched by an Agilent SureSelect Custom Design target-enrichment kit (Agilent, Santa Clara, California, USA) and sequenced with the Illumina HiSeq 2500 (Illumina, San Diego, California, USA). Subsequent sequence alignment, variant calling and annotation and filtering were performed as reported previously.20 If a heterozygous variant in a clinically relevant recessive disorder-linked gene was identified, then ExomeDepth programme21 was used to identify call copy number variants (CNVs) in the relevant gene.

Sensitivity of the assay was determined by comparing calls from four cell lines, with published genotypes, derived from The International HapMap Project and the 1000 Genomes Project (NA19194, HG01970, NA19005, NA18907; Corriell Institute, 403 Haddon Avenue Camden, New Jersey, USA). In total, 1216 known single nucleotide variants, within the region of interest, were cross-referenced with the locally generated NGS data. This analysis did not identify any false-positive or false-negative calls. We undertook NGS runs on four individuals with previously known mutations in at least one of the 226 selected genes and confirmed the ability of the assay to identify the expected mutations.

A repeat independent enrichment and sequencing run on one of the four cell lines was undertaken and results compared with test assay repeatability, which demonstrated 100% match with original run along with comparable coverage for the region of interest. In summary, we designed and validated a robust targeted NGS analysis assay for a comprehensive range of IEMs that has a minimum 50× coverage for >97% bases and a specificity of 100% and sensitivity of 100% within exons+/−5 bases. No known areas in the 226 selected genes consistently failed to achieve this coverage.

Clinical data collection, reporting and statistical analysis

This work is a retrospective evaluation of a clinical diagnostic service and ethical approval was not required. Samples along with completed clinical proformas (supplementary information) were accepted from clinicians for analysis of either a single subpanel or any combination of subpanels. Clinical and biochemical profiling of most patients was undertaken prior to the referral for NGS analysis by IEM specialists or geneticists from tertiary centres (table 1, online supplementary table 3). NGS panel analysis was performed for 102 individuals with suspected IEMs of childhood onset.

Table 1

Primary pathogenic variants from IEM NGS panel analysis for individuals with childhood-onset disease

A multidisciplinary team (MDT) comprising clinical scientists, clinical geneticists, IEM specialists and genetic counsellors was constituted for clinical correlation of results. Findings were reported under three categories (1) primary—likely clinically significant pathogenic variants, (2) secondary—deleterious but likely incidental carrier findings and (3) variants of unknown clinical significance (VUS)—variants with insufficient population frequency data to evaluate their deleteriousness. All pathogenic or deleterious point mutations were confirmed with bidirectional Sanger sequencing and CNVs were confirmed by dosage assays. Where relevant and possible, parental segregation studies were performed.

Primary pathogenic variants were reported in 51/102 individuals (table 1). In 51/102, no primary pathogenic variants were detected (online supplementary table 3). In 18/102 (8 with and 10 without primary pathogenic findings), secondary pathogenic carrier variants were reported (online supplementary table 4). In total, 89 pathogenic variants were identified across 58/226 genes in 61 individuals. Of these, 27/89 (30%) were novel and the remainder have been previously reported in the literature as pathogenic variants.

For overall analysis of distribution of diagnostic yields and carrier frequencies in this study, the data were grouped according to the complexity of the presenting phenotype and number of subpanels requested. X2 tests were performed using IBM SPSS Statistics V.24. Results of these analyses are summarised in figure 1.

Figure 1

Frequency of diagnosis by subpanels and biochemical markers. (A) Proportion of individuals in whom primary pathogenic variants and incidental carrier findings were identified according to whether specific gene defects or subgroups of IEMs were suspected and subpanels requested. The frequency of a confirmed genetic diagnosis was significantly higher when a single specific gene defect or a specific subgroup of IEM was suspected, in comparison with suspected IEM tested by requests for combination or all subpanels (p<0.0001). (B) Proportion of individuals with primary pathogenic variants according to abnormal biochemical markers detected in screening investigations. The diagnostic yield was much higher for metabolic markers, such as MMA, abnormal VLCFA and specific enzyme deficiencies. It was lower for hypoglycaemia and hyperammonaemia, especially if no additional marker was identified in the previous biochemical testing. def, deficiency; MMA, methylmalonic aciduria; VLCFA, very long chain fatty acids.

Discussion

The relative merits and limitations of WGS, WES and custom targeted design have been discussed in the literature.12 14–17 20 22–24 We chose bespoke targeted enrichment to guarantee a uniformly high coverage of the regions of interest. Our approach allowed for the inclusion of known intronic mutations that would escape detection via exome sequencing. Notably, we identified a deep-intronic PTS c.84-323A>T mutation in subject 30 (table 1) with tetrahydrobiopterin-deficient hyperphenylalaninaemia (OMIM 261640).

Our cohort of 102 individuals includes five adults, three individuals with known onset of disease in childhood and two parents (subjects 11 and 12) of a sibship of three deceased children with suspected GM1 gangliosidosis (OMIM 230500). In 42/51 individuals, the reported primary pathogenic variant was fully consistent with their biochemical and clinical features (table 1). In a further 7/51 individuals, single heterozygous pathogenic variants were detected in relevant recessive disorder genes (subjects 3, 6, 8, 15, 27, 42, 50). These may be incidental carrier findings or partial diagnoses where a second variant escaped detection. In 2/51 individuals the genetic diagnosis did not fully explain the presenting phenotype (Tyrosinaemia type I, OMIM 276700 in subject 47 and peroxisomal acyl-CoA oxidase deficiency, OMIM 264470 in subject 51). Incidental additional diagnoses with clear management implications were made in two individuals with other confirmed or possible diagnoses (Fabry disease, OMIM 301500 in subject 11 and galactokinase deficiency, OMIM 230200 in subject 51). Overall, the panel provided clinically useful and actionable information in half (51/102) of the individuals tested.

Identification of causal mutations had direct clinical impact on a number of families and individuals (table 1). We diagnosed an ultra-rare condition in subject 41 (HMG-CoA synthase-2 deficiency, OMIM 605911) that could not have been identified just on clinical or biochemical features. Diagnosis of glycogen storage disorder 0 (OMIM 240600) in subject 43 led to a change in management strategy from frequent bolus feeds to regular overnight feeds, necessitating gastrostomy insertion. Accurate genetic diagnosis enabled targeted cascade genetic testing for a number of relatives and subsequently led to diagnostic confirmation in three siblings. Five prenatal tests were offered following NGS results and one foetus was confirmed to be affected. Although we have not undertaken a formal cost-benefit analysis, the cost of NGS was comparable with the traditional diagnostic approach in mutation positive cases (table 1) (online supplementary table 7).

When clinical or biochemical features suggested a specific single gene defect or a particular subgroup of IEMs, NGS analysis led to a confirmed diagnosis in 39/66 (59%) individuals, whereas when an IEM was suspected but clinical and biochemical features were non-specific, genetic diagnoses were made in only 3/36 by this approach (8%) (p<0.0001). This underlines the role of biochemical investigations and phenotyping in directing NGS data analysis.

The likelihood of confirming a genetic diagnosis in individuals with a suspected specific single gene defect was high (24/34, 71%). Admittedly, a panel-based approach in such cases appears counter-intuitive. However, targeted clinical sequencing is not available in the UK for 16/34 suspected single-gene defects. The NGS approach thus improves access to genetic testing for IEMs.

Certain biochemical markers, such as very long chain fatty acids (VLCFA; indicative of peroxisomal disorders), methylmalonic aciduria (MMA; indicative of methylmalonic acidaemia or cobalamin metabolism defects) or specific enzyme deficiencies led to relatively better diagnostic yields (figure 1B). Our results suggest that for certain groups of disorders, NGS can be reliable, cost-effective, quicker and less invasive than follow-up biochemical testing. For example, confirmatory biochemical diagnosis of specific peroxisomal disorders, MMA or cobalamin metabolism defects usually requires functional testing in cultured fibroblasts, necessitating a skin biopsy followed by lengthy highly specialist complementation or enzyme studies.25 26

The diagnostic yield of NGS for hypoglycaemia and hyperammonaemia was comparatively lower (figure 1B) but consistent with the known positive predictive value of these markers.27 28 For hypoglycaemia, the NGS diagnostic yield (27%) is comparable to that of diagnostic fasting (22%).27 Fasting requires hospital admission and is potentially hazardous, though it may still have a role in optimising management.29 For hyperammonaemia, the NGS diagnostic yield (45%) is comparable to that of standard confirmatory tests (54%).28 For both hypoglycaemia and hyperammonaemia, the yield improved in the presence of an additional biochemical marker (figure 1B).

There were 23 incidences of carrier findings across 21 genes in 18/102 (17%) individuals (online supplementary table 4). By dividing the assay into subpanels, we minimised incidental carrier findings. VUS were detected in 85/102 (83%) of individuals, with a median of three per individual (range 0–24) (online supplementary table 5). These results reflect the importance of pre-test counselling and consent for NGS-based tests.

Scientific analysis of variants to determine the pathogenicity is challenging and compounded by the large number of variants identified in multigene NGS analyses. We identified 89 pathogenic mutations across 58 genes, thus expanding the mutational spectrum for these genes. If there was insufficient prior evidence to confirm the significance of the detected genetic variant, we undertook reverse phenotyping using clinical, biochemical or in-silico approaches (eg, subject 41 with HMG-CoA synthase-2 deficiency, OMIM 605911).30 Interestingly, even though 27/89 pathogenic variants were novel, the burden of additional functional testing due to NGS results was minimal and had to be considered in only 3% of subjects (table 1). In most cases, review of previously performed biochemical tests was sufficient for correlation of unexpected NGS findings. In comparison, the NGS approach helped to avoid invasive, hazardous, lengthy or expensive investigations in 69% (table 1 and online supplementary table 2). This demonstrates the robustness of available variant and mutation databases and in-silico tools for this group of disorders and underscores the clinical utility of this approach.

A proportion of the 51/102 individuals (online supplementary table 3) in whom we failed to achieve a genetic diagnosis may have mutations in genes that are not part of our panel. In addition, certain classes of mutations may have escaped detection due to technical limitations of the assay. Additionally, certain biochemical phenotypes such as recurrent hypoglycaemia or hyperammonaemia may be multifactorial in origin in some patients, making it challenging to achieve precise genetic diagnoses.

Notably, using NGS is clinically useful even in cases where a molecular diagnosis is not achieved because a ‘negative’ NGS result can substantially reduce the likelihood of a high-penetrant single-gene disorder and thus informing the management of the patient and counselling of the families. Furthermore, NGS was the only practical approach in several cases or helped to avoid inconvenient or invasive investigations. Even in these ‘negative’ cases, the cost of NGS was comparable to the investigations that would otherwise have been performed (online supplementary table 6). These individuals could form a focused cohort for future WES or WGS research studies.

Although the potential for NGS in a few selected groups of IEMs has been previously reported, the use of NGS for multiple groups of IEMs has not been studied extensively15 31–39 (summarised in online supplementary figure 1). The results presented here establish the utility of NGS for diagnosis of multiple groups of IEMs in clinical practice. Figure 2 illustrates how the use of NGS may be incorporated into the diagnostic algorithm of IEMs. We have demonstrated that biochemical tests are pivotal for the diagnosis of IEMs as they help focus genetic testing and facilitate NGS analysis. In addition, we have demonstrated that NGS can help to avoid invasive, expensive or hazardous biochemical confirmatory studies in specific situations, can improve access to genetic testing and may be a comparable approach in cost terms to alternative investigations. Overall, incorporating NGS into the diagnostic algorithm of IEMs can improve the accuracy of diagnosis. Further work is required to formally assess the cost-effectiveness of NGS and explore the optimal approach to the timing of NGS in the diagnosis of IEMs.

Figure 2

(A) ‘Traditional’ approach to diagnosis of IEMs. (B) Incorporation of NGS techniques into diagnostic algorithm of IEMs. CSF, cerebro-spinal fluid; IEM, inborn errors of metabolism; NGS, next-generation sequencing; WES whole exome sequencing, WGS whole genome sequencing.

The future of genomic techniques in the diagnosis of IEMs is likely to involve WES and WGS, the latter conferring particular advantages including the ability to provide equal coverage across the genome, detect CNVs, indels, intronic and regulatory variants. However, the pivotal role of biochemical testing as we demonstrate here and the potential ability of genomic testing to reduce the need for additional studies will remain relevant to WES and WGS.

Acknowledgments

This work was supported by the Manchester Biomedical Research Centre, the British Inherited Metabolic Disease Group 2015 Studentship scheme and the Central Manchester NHS Foundation Trust Newly Appointed Consultants Leadership Programme 2014. We thank the members of the Willink Biochemical laboratory, Ms Teresa Wu, Ms Karen Tylee, Dr Heather Church and Dr Mick He nderson for help in providing the biochemical data. We thank Juan Pié, University of Zaragoza, Spain for providing HMG-CoA synthase enzyme activity data. We acknowledge Ms Georgina Hall and Ms Nasaim Khan for their contribution to the MDT meetings. We express our gratitude to Dr Ashique Ahamed, Dr Ben Grey, Dr Simon Hellings, Dr Lamiya Mohiyiddeen and Ms Liz Oldfield-Beechey for useful discussions on the project.

References

View Abstract

Footnotes

  • Contributors Study design was conceived by SB and SAJ. AG, SB and SAJ designed the panel. AB, AM, EJ, BS, JW, SD, HM, CH, RS, GW, EC, AAM, RM, AD, SS, SAJ and SB provided clinical data and analysed the impact of NGS for patients. Sequencing was performed by HS, LH, JU and SR. Bioinformatic analysis was performed by SSB. NGS panel validation and design was performed by JU and SR. Data collection was performed by AG, EC, JKB, HS and LH. Data analysis was performed by AG, EC, JKB, SSB, HS, LH, SR and SB. The manuscript was written by AG, SB and SR. All authors reviewed and approved the final version of the manuscript.

  • Competing interests Arunabha Ghosh reports travel grants from Shire Plc, Biomarin Pharmaceutical, outside the submitted work. Alexander Broomfield reports consulting fees from Genzyme, Synageva, outside the submitted work. Christian Hendriksz is director of FYMCA Ltd and reports consulting fees and travel grants from Actelion, Alexion, Amicus, Biomarin, Inventiva, Sanofi Genzyme and Shire and research grants from Actelion, Amicus, Biomarin, Sanofi Genzyme and Shire, outside the submitted work. Gisela Wilcox reports travel grants from Biomarin, Genzyme, Shire, Vitaflo and Actelion, outside the submitted work. Simon Jones reports travel grants, research grants and consultancy fees from Genzyme, Shire, Biomarin, Ultragenyx, Alexion, and PTC, outside the submitted work. Helene Schlecht, Lesley Heptinstall, John Bassett, Eleanor Cartwright, Sanjeev Bhaskar, Jill Urquhart, Andrew Morris, Elisabeth Jameson, Bernd Schwahn, John Walter, Sofia Douzgou, Helen Murphy, Reena Sharma, Ellen Crushell, Ardeshir Monavari, Richard Martin, Anne Doolan, Senthil Senniappan, Simon Ramsden and Siddharth Banka have nothing to disclose.

  • Patient consent This work is a retrospective evaluation of a clinical diagnostic service. Informed consent for NGS panel testing was obtained for all individuals in the study.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Linked Articles