Abstract
The breadth and complexity of genetic testing in patients with suspected Mendelian neurodevelopmental disorders has rapidly expanded in the past two decades. However, in spite of advances in genomic technologies, genetic diagnosis remains elusive in more than half of these patients. Epigenomics, and in particular genomic DNA methylation profiles, are now known to be associated with the underpinning genetic defects in a growing number of Mendelian disorders. These often highly specific and sensitive molecular biomarkers have been used to screen these patient populations, resolve ambiguous clinical cases and interpret genetic variants of unknown clinical significance. Increasing the diagnostic yield beyond genomic sequencing technologies has rapidly propelled epigenomics to clinical utilization, with recent introduction of DNA methylation ‘EpiSign’ analysis in clinical diagnostic laboratories. This review provides an overview of the principles, applications and limitations of DNA methylation episignature analysis in patients with neurodevelopmental Mendelian disorders, and discusses clinical implications of this emerging diagnostic technology.
Introduction
Classically defined, mitotically heritable molecular mechanisms that impact gene expression without underlying changes in DNA sequence are referred to as epigenetics. However, it also stands that epigenetic profiles across genomic DNA are established, regulated and maintained by molecular mechanisms which are essentially genetic, and are influenced by the underpinning constitutive genotypes. Mammalian genomic DNA methylation patterns are established in early development by two waves of nearly global demethylation followed by targeted remethylation (1). Errors in methylation during these early time points, often as a result of an underlying genetic mutation, can lead to various developmental disorders, including neurodevelopmental syndromes (2,3). Since these genetic changes occur early in development they may be propagated across all cell lineages and can result in pathologies in multiple tissues. The consequent DNA methylation changes may also therefore be widespread across the genome. When these epigenetic changes are maintained throughout development and across cell types these genome-wide DNA methylation ‘episignatures’ can be used as biomarkers for the diagnosis of neurodevelopmental diseases using easily accessible tissues such as peripheral blood (4,5).
An expanding range of genetic conditions has been shown to exhibit DNA methylation episignatures. One group of such disorders involves genes that establish and maintain DNA methylation, the DNA methyltransferases. Unique episignatures have been identified in association with mutations in DNMT1 (6,7), DNMT3A (8,9) and DNMT3B (8). The majority of disorders with defined episignatures involve chromatin remodeling genes and chromatin regulators, with <30 such genes identified so far (8). Lastly, DNA methylation signatures are also being mapped in disorders whose genes’ primary functions are not associated with epigenetic and chromatin regulatory mechanisms, such as SMS and UBE2A (8).
Although the DNA sequence changes in genes associated with a particular Mendelian condition can vary widely, each syndrome that is currently known to exhibit a DNA methylation episignature typically demonstrates only one (or occasionally two (10)) methylation signature, which may or may not be associated with a unique clinical presentation. The practical impact of these findings is that we can now use DNA methylation signatures as biomarkers for screening and diagnosis of neurodevelopmental syndromes and to enable reclassification of variants with unknown/uncertain clinical significance (VUS). This technology which is referred to as ‘EpiSign’ forms the basis of the emerging diagnostic discipline clinical epigenomics, and which is now being applied in diagnostic testing of patients with rare disorders (https://www.ggc.org/episign; https://genomediagnostics.amsterdamumc.nl/epigenetic-test/).
Mapping of DNA Methylation Signatures
DNA methylation profiles show extensive variation across human tissues, however the majority of current DNA methylation episignature literature has focused on peripheral blood as the tissue of choice, primarily because this is the most commonly used tissue in clinical laboratory settings and one for which genomic DNA methylation reference databases are most abundant.
Developmental disorders with published peripheral blood DNA methylation episignatures used for patient diagnosis and/or VUS classification
| Disease/disorder | Gene(s) | References |
|---|---|---|
| Alpha thalassemia/mental retardation X-linked syndrome (ATR-X) | ATRX | (6,8,26) |
| Autism, susceptibility to, 18 (AUT18) | CHD8 | (8,27) |
| BAFopathies | ARID1B, ARID1A, SMARCB1,SMARCA4, SMARCA2, SOX11, DPF2, ARID2, SMARCE1 | (8,20) |
| Börjeson–Forssman–Lehmann syndrome (BFLS) | PHF6 | (8) |
| Cerebellar ataxia, deafness and narcolepsy, autosomal dominant (ADCADN) | DNMT1 | (6–8) |
| CHARGE syndrome | CHD7 | (6,8,18) |
| Chromosome 16p11.2 deletion syndrome | Chr16p.11 deletion | (27) |
| Cornelia de Lange syndrome (CdLS) | NIPBL, RAD21, SMC3, SMC1A | (8,12) |
| Down syndrome | Trisomy 21 | (8,28) |
| Epileptic encephalopathy, childhood-onset (EEOC) | CHD2 | (8) |
| Floating-Harbor syndrome (FLHS) | SRCAP | (6,8,17) |
| Genitopatellar syndrome (GTPTS) and Ohdo syndrome, SBBYSS variant (SBBYSS) | KAT6B | (6,8) |
| Helsmoortel-van der Aa syndrome (HVDAS)a | ADNP | (8,10) |
| Hunter McAlpine syndrome (HMA) | Chr5q35-qter duplication | (8) |
| Immunodeficiency-centromeric instability-facial anomalies syndrome (ICF)b | DNMT3B, CDCA7, ZBTB24, HELLS | (8) |
| Kabuki syndrome | KMT2D, KDM6A | (6,8,18,29) |
| Kleefstra syndrome | EHMT1 | (8) |
| Koolen de Vreis syndrome (KDVS) | KANSL1 | (8) |
| Mental retardation, autosomal dominant 51 (MRD51) | KMT5B | (8) |
| Mental retardation, X-linked 93 (MRD93) | BRWD3 | (8) |
| Mental retardation, X-linked 97 (MRD97) | ZNF711 | (8) |
| Mental retardation, X-linked syndromic, Nascimento-type (MRXSN) | UBE2A | (8) |
| Mental retardation, X-linked, Snyder–Robinson type (MRXSSR) | SMS | (8) |
| Mental retardation, X-linked, syndromic, Claes-Jensen syndrome (MRXSCJ) | KDM5C | (6,8,25) |
| Rahman syndrome (RMNS) | HIST1H1E | (8,30) |
| Rubinstein–Taybi syndrome (RSTS) | CREBBP, EP300 | (8) |
| SETD1B-related syndrome | SETD1B | (8,22) |
| Sotos syndrome | NSD1 | (6,8,16) |
| Tatton–Brown–Rahman syndrome (TBRS) | DNMT3A | (8) |
| Wiedemann–Steiner syndrome (WDSTS) | KMT2A | (8) |
| Williams–Beuren deletion syndrome (WBS) and Williams–Beuren regions duplication syndromec | 7q11.23 deletion/duplication | (8,31) |
| Weaver syndrome and Cohen–Gibson syndrome | EZH2 and EED | (8,32) |
| Disease/disorder | Gene(s) | References |
|---|---|---|
| Alpha thalassemia/mental retardation X-linked syndrome (ATR-X) | ATRX | (6,8,26) |
| Autism, susceptibility to, 18 (AUT18) | CHD8 | (8,27) |
| BAFopathies | ARID1B, ARID1A, SMARCB1,SMARCA4, SMARCA2, SOX11, DPF2, ARID2, SMARCE1 | (8,20) |
| Börjeson–Forssman–Lehmann syndrome (BFLS) | PHF6 | (8) |
| Cerebellar ataxia, deafness and narcolepsy, autosomal dominant (ADCADN) | DNMT1 | (6–8) |
| CHARGE syndrome | CHD7 | (6,8,18) |
| Chromosome 16p11.2 deletion syndrome | Chr16p.11 deletion | (27) |
| Cornelia de Lange syndrome (CdLS) | NIPBL, RAD21, SMC3, SMC1A | (8,12) |
| Down syndrome | Trisomy 21 | (8,28) |
| Epileptic encephalopathy, childhood-onset (EEOC) | CHD2 | (8) |
| Floating-Harbor syndrome (FLHS) | SRCAP | (6,8,17) |
| Genitopatellar syndrome (GTPTS) and Ohdo syndrome, SBBYSS variant (SBBYSS) | KAT6B | (6,8) |
| Helsmoortel-van der Aa syndrome (HVDAS)a | ADNP | (8,10) |
| Hunter McAlpine syndrome (HMA) | Chr5q35-qter duplication | (8) |
| Immunodeficiency-centromeric instability-facial anomalies syndrome (ICF)b | DNMT3B, CDCA7, ZBTB24, HELLS | (8) |
| Kabuki syndrome | KMT2D, KDM6A | (6,8,18,29) |
| Kleefstra syndrome | EHMT1 | (8) |
| Koolen de Vreis syndrome (KDVS) | KANSL1 | (8) |
| Mental retardation, autosomal dominant 51 (MRD51) | KMT5B | (8) |
| Mental retardation, X-linked 93 (MRD93) | BRWD3 | (8) |
| Mental retardation, X-linked 97 (MRD97) | ZNF711 | (8) |
| Mental retardation, X-linked syndromic, Nascimento-type (MRXSN) | UBE2A | (8) |
| Mental retardation, X-linked, Snyder–Robinson type (MRXSSR) | SMS | (8) |
| Mental retardation, X-linked, syndromic, Claes-Jensen syndrome (MRXSCJ) | KDM5C | (6,8,25) |
| Rahman syndrome (RMNS) | HIST1H1E | (8,30) |
| Rubinstein–Taybi syndrome (RSTS) | CREBBP, EP300 | (8) |
| SETD1B-related syndrome | SETD1B | (8,22) |
| Sotos syndrome | NSD1 | (6,8,16) |
| Tatton–Brown–Rahman syndrome (TBRS) | DNMT3A | (8) |
| Wiedemann–Steiner syndrome (WDSTS) | KMT2A | (8) |
| Williams–Beuren deletion syndrome (WBS) and Williams–Beuren regions duplication syndromec | 7q11.23 deletion/duplication | (8,31) |
| Weaver syndrome and Cohen–Gibson syndrome | EZH2 and EED | (8,32) |
aDNP has two distinct signatures depending on where in the gene the mutation occurs.
bICF1 exhibits one signature whereas ICF 2, 3 and 4 exhibit a separate, common signature.
cThese two deletion/duplication syndromes exhibit symmetrical increased/decreased DNA methylation signatures, respectively.
Developmental disorders with published peripheral blood DNA methylation episignatures used for patient diagnosis and/or VUS classification
| Disease/disorder | Gene(s) | References |
|---|---|---|
| Alpha thalassemia/mental retardation X-linked syndrome (ATR-X) | ATRX | (6,8,26) |
| Autism, susceptibility to, 18 (AUT18) | CHD8 | (8,27) |
| BAFopathies | ARID1B, ARID1A, SMARCB1,SMARCA4, SMARCA2, SOX11, DPF2, ARID2, SMARCE1 | (8,20) |
| Börjeson–Forssman–Lehmann syndrome (BFLS) | PHF6 | (8) |
| Cerebellar ataxia, deafness and narcolepsy, autosomal dominant (ADCADN) | DNMT1 | (6–8) |
| CHARGE syndrome | CHD7 | (6,8,18) |
| Chromosome 16p11.2 deletion syndrome | Chr16p.11 deletion | (27) |
| Cornelia de Lange syndrome (CdLS) | NIPBL, RAD21, SMC3, SMC1A | (8,12) |
| Down syndrome | Trisomy 21 | (8,28) |
| Epileptic encephalopathy, childhood-onset (EEOC) | CHD2 | (8) |
| Floating-Harbor syndrome (FLHS) | SRCAP | (6,8,17) |
| Genitopatellar syndrome (GTPTS) and Ohdo syndrome, SBBYSS variant (SBBYSS) | KAT6B | (6,8) |
| Helsmoortel-van der Aa syndrome (HVDAS)a | ADNP | (8,10) |
| Hunter McAlpine syndrome (HMA) | Chr5q35-qter duplication | (8) |
| Immunodeficiency-centromeric instability-facial anomalies syndrome (ICF)b | DNMT3B, CDCA7, ZBTB24, HELLS | (8) |
| Kabuki syndrome | KMT2D, KDM6A | (6,8,18,29) |
| Kleefstra syndrome | EHMT1 | (8) |
| Koolen de Vreis syndrome (KDVS) | KANSL1 | (8) |
| Mental retardation, autosomal dominant 51 (MRD51) | KMT5B | (8) |
| Mental retardation, X-linked 93 (MRD93) | BRWD3 | (8) |
| Mental retardation, X-linked 97 (MRD97) | ZNF711 | (8) |
| Mental retardation, X-linked syndromic, Nascimento-type (MRXSN) | UBE2A | (8) |
| Mental retardation, X-linked, Snyder–Robinson type (MRXSSR) | SMS | (8) |
| Mental retardation, X-linked, syndromic, Claes-Jensen syndrome (MRXSCJ) | KDM5C | (6,8,25) |
| Rahman syndrome (RMNS) | HIST1H1E | (8,30) |
| Rubinstein–Taybi syndrome (RSTS) | CREBBP, EP300 | (8) |
| SETD1B-related syndrome | SETD1B | (8,22) |
| Sotos syndrome | NSD1 | (6,8,16) |
| Tatton–Brown–Rahman syndrome (TBRS) | DNMT3A | (8) |
| Wiedemann–Steiner syndrome (WDSTS) | KMT2A | (8) |
| Williams–Beuren deletion syndrome (WBS) and Williams–Beuren regions duplication syndromec | 7q11.23 deletion/duplication | (8,31) |
| Weaver syndrome and Cohen–Gibson syndrome | EZH2 and EED | (8,32) |
| Disease/disorder | Gene(s) | References |
|---|---|---|
| Alpha thalassemia/mental retardation X-linked syndrome (ATR-X) | ATRX | (6,8,26) |
| Autism, susceptibility to, 18 (AUT18) | CHD8 | (8,27) |
| BAFopathies | ARID1B, ARID1A, SMARCB1,SMARCA4, SMARCA2, SOX11, DPF2, ARID2, SMARCE1 | (8,20) |
| Börjeson–Forssman–Lehmann syndrome (BFLS) | PHF6 | (8) |
| Cerebellar ataxia, deafness and narcolepsy, autosomal dominant (ADCADN) | DNMT1 | (6–8) |
| CHARGE syndrome | CHD7 | (6,8,18) |
| Chromosome 16p11.2 deletion syndrome | Chr16p.11 deletion | (27) |
| Cornelia de Lange syndrome (CdLS) | NIPBL, RAD21, SMC3, SMC1A | (8,12) |
| Down syndrome | Trisomy 21 | (8,28) |
| Epileptic encephalopathy, childhood-onset (EEOC) | CHD2 | (8) |
| Floating-Harbor syndrome (FLHS) | SRCAP | (6,8,17) |
| Genitopatellar syndrome (GTPTS) and Ohdo syndrome, SBBYSS variant (SBBYSS) | KAT6B | (6,8) |
| Helsmoortel-van der Aa syndrome (HVDAS)a | ADNP | (8,10) |
| Hunter McAlpine syndrome (HMA) | Chr5q35-qter duplication | (8) |
| Immunodeficiency-centromeric instability-facial anomalies syndrome (ICF)b | DNMT3B, CDCA7, ZBTB24, HELLS | (8) |
| Kabuki syndrome | KMT2D, KDM6A | (6,8,18,29) |
| Kleefstra syndrome | EHMT1 | (8) |
| Koolen de Vreis syndrome (KDVS) | KANSL1 | (8) |
| Mental retardation, autosomal dominant 51 (MRD51) | KMT5B | (8) |
| Mental retardation, X-linked 93 (MRD93) | BRWD3 | (8) |
| Mental retardation, X-linked 97 (MRD97) | ZNF711 | (8) |
| Mental retardation, X-linked syndromic, Nascimento-type (MRXSN) | UBE2A | (8) |
| Mental retardation, X-linked, Snyder–Robinson type (MRXSSR) | SMS | (8) |
| Mental retardation, X-linked, syndromic, Claes-Jensen syndrome (MRXSCJ) | KDM5C | (6,8,25) |
| Rahman syndrome (RMNS) | HIST1H1E | (8,30) |
| Rubinstein–Taybi syndrome (RSTS) | CREBBP, EP300 | (8) |
| SETD1B-related syndrome | SETD1B | (8,22) |
| Sotos syndrome | NSD1 | (6,8,16) |
| Tatton–Brown–Rahman syndrome (TBRS) | DNMT3A | (8) |
| Wiedemann–Steiner syndrome (WDSTS) | KMT2A | (8) |
| Williams–Beuren deletion syndrome (WBS) and Williams–Beuren regions duplication syndromec | 7q11.23 deletion/duplication | (8,31) |
| Weaver syndrome and Cohen–Gibson syndrome | EZH2 and EED | (8,32) |
aDNP has two distinct signatures depending on where in the gene the mutation occurs.
bICF1 exhibits one signature whereas ICF 2, 3 and 4 exhibit a separate, common signature.
cThese two deletion/duplication syndromes exhibit symmetrical increased/decreased DNA methylation signatures, respectively.
A decade ago, Down syndrome and Cornelia de Lange syndrome were shown to harbor genomic DNA methylation signatures (11,12). Since then, much of the research in this field has been focused on the study of syndromes resulting from mutations in chromatin regulatory genes (Table 1). With a few notable exceptions, the majority of syndromes studied were found to have detectable DNA methylation episignatures in peripheral blood. A recent report focused on syndromes presenting with overgrowth has shown robust DNA methylation signatures in those caused by histone modifying genes, but no evidence of episignatures in genes with no direct functional involvement in the epigenetic machinery (13). In contrast, several conditions resulting from genes linked to chromatin remodeling showed no evidence of DNA methylation signatures including Coffin-Lawry syndrome (RSK2), Rett syndrome (MeCP2), and Mental retardation syndrome, X-linked, Siderius type (PHF8) (6,8). Although future studies are likely going to expand the number and type of genetic syndromes exhibiting DNA methylation episignatures, it is also likely that increasing the resolution of genomic analyses (array based versus genomic bisulfite sequencing) and expansion of patient and control databases will help us discover episignatures in what are currently episignature-negative conditions.
One of the challenges in mapping episignatures is establishing reference cohorts of patient samples in what are commonly known as ‘rare disorders’. The ability to detect a DNA methylation signature is highly contingent upon the intensity (effect size) and extent (number of differentially methylated CpGs) in the signature, which can highly vary in different syndromes. Intuitively, conditions with large effect sizes require smaller number of samples. An extreme example of such disorder is the Nascimento-type X-linked mental retardation caused by mutations in UBE2A which was successfully mapped using DNA from only three mutation-positive patients (8). For conditions with milder signatures a larger sample size is needed, which in the context of rare disorders generally involves 10–20 samples.
The majority of genomic DNA methylation data that have been published in this field has utilized genomic microarrays (currently the Illumina Infinium Methylation EPIC array). Various statistical procedures have be implemented for prioritization of the informative CpG sites (6,8). Common considerations include the effect size, level of differentiation, pairwise correlations, and most significantly, the intersample variability in the overall combined DNA methylation pattern produced after probe selection. Prior to performing feature selection and analysis, data are adjusted for a number of confounders including biological factors associated with DNA methylation changes such as age, sex and blood cell type composition, as well as technical batch effects, and random data structures. Critically, replicability should be demonstrated by using distinct patient sample sets (cohorts) for signature detection and signature validation for each syndrome independently. Most syndromes published thus far exhibit a single, unique DNA methylation signature, even in situations where multiple clinical subtypes exist (e.g. Kabuki types 1 and 2, Cornelia de Lang types 1–4). This is also the case for syndromes caused by mutations in different domains of a single gene (Genitopatellar syndrome and Say–Barber–Biesecker–Young–Simpson syndrome, both caused by mutations in KAT6B) (6,8). An interesting outlier to this observation is that in Helsmoortel-van der Aa syndrome (ADNP syndrome) mutations within the central domain of the ADNP protein generate a strikingly different episignature relative to mutations outside of this domain, with no apparent clinical distinctions in patient subgroups currently identified (10,14). This data serve as a caution of the possibility of multiple episignatures existing in other syndromes for which mutation spectrums may not be fully represented in reference cohorts used to derive the original episignature. It is noted that episignature mapping is a dynamic process highly dependent on patient and referent cohort sizes, requiring regular reanalyses and updating of the list of most differentiating probes in these episignatures.
DNA methylation assessment (EpiSign analysis) of peripheral blood from a patient with clinical features of Kabuki syndrome presenting with a VUS in the KMT2D (c.15641G > T; p.R5214L). (A) Using a multiclass supervised classification algorithm (8), the sample was scored for 43 episignatures, including Kabuki syndrome. For all of the classes (X-axis) the sample received a score close to 0 (dots, Y-axis) except for the Kabuki syndrome category (score = 0.9). (B) Euclidean clustering analysis shows Kabuki syndrome patient DNA methylation signature clustering with Kabuki cases and not with healthy controls. Rows are ~150 informative CpG probes, columns are samples, and the color scale indicate the methylation levels ranging between 0 and 1. (C) Multidimensional scaling using the same set of CpGs on a larger group of confirmed cases and controls shows that the sample clusters with the Kabuki cases.
Use of the prioritized CpGs in disease classification should ideally be performed by using both unsupervised and supervised bioinformatics procedures. The current number of diseases with a positive episignature exceeds 40 (with >55 associated genes). There is a significant overlap in genomic regions forming the disease episignatures, particularly in disorders with closely related genetic etiology (similar genes/molecular pathways affected), or with overlapping phenotypes. Our previous work has shown that assessing a patient for one signature at a time using unsupervised methods (e.g. hierarchical clustering, multidimensional scaling, etc.) can lead to misclassifications or uncertain conclusions, particularly in syndromes with moderate episignatures (8). An alternate approach that uses multiclass supervised modeling and which ranks samples simultaneously for all of the known episignature conditions has proven much more robust and accurate. Another challenge is that the number and spectrum of episignatures is not known. The currently-mapped episignatures likely represent only a small subset of all the potentially identifiable episignatures. Some of the yet-unmapped conditions may bear significant overlap and similarities to the mapped signatures, which in turn can complicate interpretation of episignature analyses in individual cases.
Annotation and Classification of Variants of Unknown Significance
One of the major challenges in the field of genetic diagnostics is genetic variants of unknown/uncertain significance. DNA methylation episignatures are consequences of underpinning genetic defects, and as such can be used to interpret and clinically reclassify VUSs. Episignature analysis has enabled conclusive interpretation of VUSs in hundreds of families, and is currently the primary indication for EpiSign analysis in the clinical laboratories. Genomic DNA methylation profiles of a patient carrying a VUS in one of the EpiSign genes/syndromes is in contrast with the reference database (EpiSign Knowledge Database). The analysis involves comparison of patient’s DNA methylation profile to the particular syndrome being tested (based on the clinical presentation of gene locus of the VUS), or to all other known syndromes in the databases (in cases where screening analysis for all syndromes is requested). Figure 1 shows an example of data output and interpretation of a VUS in the KMT2D gene in a patient with clinical features consistent with Kabuki syndrome. In this case the VUS is clinically reclassified to likely pathogenic, ACMG category 2 (15), enabling a conclusive diagnosis for this patient.
An early studies to demonstrate utility of episignatures in variant reclassification was in Sotos syndrome, caused by mutations in the NSD1 gene (16). Authors used the NSD1 episignature to classify NSD1 VUSs from 16 individuals as either benign or pathogenic by determining whether the VUS samples clustered statistically with samples from confirmed Sotos patients or controls. It was found that these classifications based on methylation analysis were in agreement with blinded, expert assessment based on clinical presentation, but differed in some cases from in silico pathogenicity predication, suggesting that methylation analysis rather than genetic prediction algorithms may be a more reliable way to classify VUSs (16). Many studies demonstrating this principle have been published since including DNMT1 related Autosomal Dominant Cerebellar Ataxia, Deafness and Narcolepsy syndrome (7), SRCAP related Floating-Harbor syndrome (17), KMT2D associated Kabuki syndrome (18,19), CHD7 related Charge syndrome (18), BAFopathy related disorders (ARID1A, ARID1, SMARCA2) (20,21), SETD1B-related syndrome (22), and most recently in the largest study to date involving 42 Mendelian neurodevelopmental disorders (8), to highlight a few.
One major challenge in interpretation of VUSs based on DNA methylation episignatures relates to evidence of existence of multiple episignatures in individual genes. Helsmoortel-van der Aa syndrome is caused by mutations in the ADNP gene (23). The majority of reported mutations in this gene are truncating variants which result in DNA methylation disruption in these patients. However, depending on the location of these truncating mutations within the gene domains, they result in distinct and largely opposing DNA episignatures, and interestingly, no apparent difference in the consequent clinical phenotypes. (10). Hence, whenever possible, it is important to include a broad range and type of genetic variants in discovery cohorts in order to better capture the possible heterogeneity in the DNA methylation signature(s). Unfortunately, in many cases this is not feasible in the context of rare genetic disorders. Undoubtedly, expansion of the DNA methylation databases will uncover further complexity.
In a series of recent studies focused on describing episignatures in >50 genes, our group has described a method for concurrent classification of VUSs from multiple neurodevelopmental syndromes. One study described 93 patients with VUS findings, 37 of which confirmed positive for episignatures being reclassified likely pathogenic in CHARGE (CHD7), Kabuki (KMT2D) or Sotos (NSD1) syndrome, and the remaining VUSs being reclassified to likely benign (6). This and other studies have described patients where an episignature was found for a gene alternate to the gene where the VUS was identified, highlighting the complexity of clinical diagnosis of neurodevelopmental syndromes and furthering the utility of episignature analysis. A follow-up study included 44 patients with VUS findings, 17 of which exhibited specific episignatures for 14 different neurodevelopmental syndromes (24). In the most recent study, VUSs in 12 patient samples were successfully reclassified (8). In these studies, we also highlight a large number of patients with suspected neurodevelopmental genetic conditions without a prior genetic diagnosis where an episignature was the initial finding, which was later confirmed by a targeted genetic analysis. In a subset of the patients that where episignature-positive, in spite of exhibiting a matching clinical presentation, no detectable genetic changes in related genes were identified.
Adaption of DNA methylation analysis of patients with suspected genetic neurodevelopmental conditions in the clinical setting requires that the clinical community is aware of assumptions, limitations and exceptions associated with this technology, which may vary across syndromes and even in individual patients. Although the absence of a corresponding DNA methylation signature represents strong evidence against the VUS pathogenicity and diagnosis of the related genetic disorder, the possibility of the existence of an alternate, as of yet undefined DNA methylation ‘sub-profile’ associated with the variant cannot be ruled out. Not all pathogenic variants within a gene will necessarily produce the same DNA methylation signature (for example, as seen in ADNP syndrome), one gene may therefore be associated with multiple DNA methylation subprofiles, all of which have not yet been identified. Similarly, a positive episignature finding will not always be confirmed by genetic analysis (e.g.. ADNP syndrome). This may be because of our inability to clinically assess non-coding and regulatory sequences, or alternatively may be the result of genetic heterogeneity for a common episignature (e.g. BAFopathy complex genes). Similar to genetic testing, mosaicism may also limit the sensitivity of DNA methylation episignatures. Finally, heterozygous mutation carriers in some cases may exhibit DNA methylation signatures, even in absence of corresponding clinical features (25).
Conclusion
DNA methylation profiling in patients with Mendelian neurodevelopmental conditions has rapidly evolved from a scientific concept to a clinical diagnostic technology. The rapid adoption of DNA methylation episignatures as clinical biomarkers used in patient care is influenced by both an existing unmet need for conclusive genetic diagnosis in half of this patient population, as well as substantial accuracy, sensitivity and specificity of these biomarkers. Although technological and bioinformatics approaches for episignature analysis are well defined and scalable, the primary limitation for a broader adaption of this technology is related to the limited DNA methylation reference databases in these relatively rare genetic conditions. Although individually rare, cumulatively, <4000 known Mendelian conditions are a common cause of neurodevelopmental condition. It is unclear what proportion of these genetic disorders exhibit DNA methylation episignatures. Continued efforts in mapping of DNA methylation profiles in these disorders will ultimately broaden the utility of this technology. In addition to studies of specific patient cohorts, and further adaption of this technology in clinical testing, broader studies focused on the clinical impact of this technology on health systems are warranted. The first such national clinical trial recently launched and is titled ‘Beyond Genomics: Assessing the Improvement in Diagnosis of Rare Diseases using Clinical Epigenomics in Canada (EpiSign-CAN)’ (https://www.genomecanada.ca/en/beyond-genomics-assessing-improvement-diagnosis-rare-diseases-using-clinical-epigenomics-canada). This trial aims to validate the conditions for maximizing patient and health system impact and assess the evidence for first-visit and reflex scenarios for adoption of genome-wide DNA methylation testing within Canada. Although adaption of epigenomic DNA methylation analysis marks the first wave of postgenomic technologies for diagnosis of patients with Mendelian genetic conditions, it is likely that the ultimate goal of reaching an ‘omic’ diagnosis in all patients will require broader utilization of this and similar technologies in the future.
Acknowledgements
The authors have no financial or funding information to declare.
Conflict of Interest statement. None declared.
