Methylation assay in KMT2B-related dystonia: a novel diagnostic validation tool

Abstract

Background/objectives

KMT2B-related dystonia (DYT28, OMIM #617284) is a progressive neurological condition characterized by early onset movement disorders with autosomal dominant inheritance. In this study, we describe the use of a genome methylation episignature methodology to functionally validate two variants of uncertain significance (VUS) in the KMT2B gene.

Methods

Genome-wide methylation status was assessed using the EPIC methylation assay in peripheral blood samples from two subjects with early onset movement disorder and missense variants of uncertain significance in the KMT2B gene (p.Leu1720Phe and p.Tyr2515Cys). After QC and normalization steps, we compared the M values for all 144 probes, previously described as an EpiSign for KMT2B-related dystonia, between the two subjects and 14 controls individuals.

Results

The individual harboring the p.Tyr2515Cys variant exhibited a hypermethylation profile compatible with pathogenic/likely pathogenic variants in KMT2B, allowing for variant reclassification, conclusive genetic counseling, and patient stratification for deep brain stimulation. In contrast, the individual harboring the p.Leu1720Phe variant had a methylation status similar to controls, practically ruling out KMT2B-related dystonia.

Conclusion

Investigation of methylation status can be a powerful tool to determine pathogenicity when facing KMT2B variants of uncertain significance. Methylation results may optimize genetic counseling and positively impact patient care.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13148-024-01780-1.

Introduction

The lysine-specific methyltransferase 2b gene (KMT2B, OMIM *606834) is involved in transcriptional regulation through posttranslational histones modification. The encoded protein is ubiquitously expressed, has multiple domains, and is responsible for methylation of histone H3, an important mechanism in regulating gene expression. Diseases associated with pathogenic variants in KMT2B have an autosomal dominant pattern of inheritance and include childhood-onset dystonia 28 (OMIM #617284, also known as DYT-KMT2B or DYT28) and autosomal dominant intellectual developmental disorder-68 (OMIM # 619934, also known as MRD68). Individuals with DYT-KMT2B typically have childhood-onset dystonia, often with secondary generalization affecting gait, upper limbs, neck, and oropharyngeal regions. Occasionally, affected individuals may present with myoclonus, facial dysmorphisms, early onset puberty, and variable degrees of intellectual disability and/or neuropsychiatric findings [1–4]. Imaging may demonstrate a hypointense streak in the lateral globus pallidus externa, more visible on susceptibility-weighted sequences [2]. In some series, KMT2B is the most frequently implicated gene in childhood-onset dystonia [5].

Pathogenic variants in KMT2B typically lead to loss of function of the mutated allele. The original disease description in 2017 [2] included several individuals with a microdeletion involving coding areas of the gene, and haploinsufficiency-leading variants (frameshift, stop codons and canonical-splice altering) predominate in ClinVar. However, missense variants have also been reported and can pose a challenge for pathogenicity interpretation. In the absence of segregation data or functional studies, previously unreported missense variants are frequently classified as variants of uncertain significance (VUS). Furthermore, even when a variant is inherited from an asymptomatic parent, this does not rule out completely pathogenicity since incomplete penetrance has been described [2].

The epigenetic regulation role ascribed to KMT2B poses an opportunity for functional validation of missense variants. Disease-causing variants are expected to disrupt the genome methylation pattern. Genome-wide DNA methylation analysis has increased the sensitivity in detecting epigenomic signatures (‘EpiSigns’) in individuals with complex clinical conditions presenting with variants of uncertain significance [6, 7].

Mirza-Schreiber et al. (2022) predicted that patients with loss-of-function variants in KMT2B would present with DNA methylation changes and described an EpiSign for patients with DYT-KTM2B following an epigenome-wide association study (EWAS) in blood DNA samples. Through regression analysis, it was possible to detect a set of 113 probes significantly hypermethylated in the case group. Thus, based on methylation data, a recent support vector machine (SVM) classification model was developed to distinguish variants associated with KMT2B dystonia [8, 9].

Here, we describe two subjects with early onset dystonia and missense VUSs in KMT2B, investigated with the EpiSign approach. Our results highlight the diagnostic value of methylation analysis in unsolved cases of DYT-KMT2B.

Materials and methods

Subjects

Subject 1 is a female patient with unremarkable antenatal and perinatal histories. She achieved all developmental milestones at the expected age, sitting independently at 6 months, walking at 12 months, and speaking her first words shortly after her first birthday. Her clinical history is free of delays or significant events. This subject presented at 22 months old with paroxysmal gait changes, best described as transient imbalance. Events were characterized by truncal hyperextension with possible appendicular dystonic posturing, occurring without a clear trigger and occasionally leading to injury. Events lasted several seconds to a minute, occurred with preserved consciousness, and were interpreted as paroxysmal nonkinesigenic dyskinesia (Supplementary video documentation 1). The events recurred with sporadic frequency until she was first seen in clinic at age 26 months old. Neurological examinations, EEG, and MRI were normal.

Whole-exome sequencing for subject 1 demonstrated the KMT2B variant ENST00000420124: c.5160G > C; p.Leu1720Phe. This variant is absent from population databases and in silico tools (REVEL, AlphaMissense, SpliceAI) offered conflicting interpretations about its potential pathogenicity (Table 1). Subsequent segregation analysis demonstrated that the L1720F variant was inherited from the subject’s asymptomatic father (Supplementary material 3). While these findings decreased the clinical suspicion that the KMT2B variant was a potential culprit, this was not completely ruled out. Paroxysmal dystonia and incomplete penetrance have been described in previous cases [2]. Thus, we opted to further investigate this patient due to the potential for an atypical presentation. In addition, no other significant variant was identified in any other gene previously associated with dystonia or paroxysmal dyskinesia, such as the PNKD, KCNMA1 and PRRT2 genes.

Table 1.

Stratification of ACMG criteria and in silico prediction tool scores

	Subject 1	Subject 2
Variant	p.Leu1720Phe	p.Tyr2515Cys
ACMG criteria	PM2 PP2	PM2 PP2 PP3
In silico prediction	CADD: 24.7 SIFT: 0.02 Revel: 0.46 AlphaMissense: 0.507	CADD: 34 SIFT: 0.007 Revel: 0.86 AlphaMissense: 0.772

Subject 2 is a female patient with unremarkable antenatal, perinatal, and developmental history. She first presented at age 9 with left leg dystonia. Over a period of months, dystonia progressed to involve the left and subsequently the right arm. When first seen in our clinic at age 13, there was dystonia affecting the upper more than the lower limbs, impacting handwriting and gait (Supplementary video documentation 2). The laboratory work-up yielded unremarkable results. The patient demonstrates age-appropriate adaptive skills and is currently enrolled in school, where she is performing at grade level. Brain MRI demonstrated slight bilateral pallidal hypointensities in SWAN sequences.

The whole-exome sequencing of subject 2 demonstrated the KMT2B variant ENST00000420124 c.7544A > G; p.Tyr2515Cys. This variant is absent from population databases and in silico tools predicted it to be deleterious (REVEL, AlphaMissense) (Table 1). Additionally, the SpliceAI tool predicted that this nucleotide change activates a cryptic splice donor site located 11 base pairs upstream from the canonical donor site, potentially leading to abnormal mRNA processing. RNA sequencing was not possible to confirm this hypothesis, and variant segregation was not feasible. Similar to subject 1, no other significant variants were identified in genes previously associated with dystonia.

Genome-wide methylation analysis

Based on previously described methodology, we used the Illumina Infinium MethylationEPIC Beadchip (v1) assay to investigate the methylation status of 144 probes from the two research subjects [8, 9] and compared with 14 control individuals. For the technical execution of the assay, the Illumina equipment (iScan Starter Kit) is required, taking about three to four days to obtain the experimental results, and with an already developed pipeline, only a few hours are needed for the analysis of these results.

Raw DNA methylation data were analyzed using a bioinformatics analysis pipeline in R 4.3.2 version. For normalization, we used the preprocessing function ‘preprocessQuantile’ implemented in minfi. To access M values, defined as M = logit(β) = log(Methylated/Unmethylated), we used getM function also implemented in minfi [10].

Then, with M values, we fitted an analysis of variance model and performed an ANOVA test. Subsequently, we conducted a multiple comparison test using Tukey test, which controls the overall experiment-wise error rate, to determine which M values were responsible for statistical differences between group means. Finally, we used a single-tailed t test to determine a possible hypermethylation in subject 2.

To validate the potential methylation findings in these two subjects, we employed the EpiSign analysis platform (v5) in addition to the SVM model previously described, provided courtesy of Oexle and colleagues. This tool uses analysis models that are even more sensitive than the statistical tests applied in this study.

Results

ANOVA revealed a significant difference between methylation status across groups (F = 2.165 p.val = 0.005). Using the Tukey test, we determined that subject 2 (harboring the Y2515C variant) had M values significantly different from control samples (mean adj.p-val = 0.04). In contrast, subject 1 (harboring the L1720F variant) did not show a statistically significant difference in M values (mean adj.p-val = 0.9). Using a single-tailed t test, we determined the presence of a hypermethylation status in subject 2 (t = 3.818 p = 8.249e-05), corresponding to the EpiSign described by Oexle and colleagues (2023) (Fig. 1).

Fig. 1.

Distribution of M values between all individuals in sample cohort. Subject 2 presents an upper deviation

Clustering analysis evidenced a clear distinction of subject 2 to the rest of cohort (Fig. 2).

Fig. 2.

Heatmap evidencing clustering results. Subject 2 differs substantially from control individuals, while subject 1 does not present distinction

Using the results from the SVM model applied by Oexle and colleagues, we calculated the standard deviation (SD) of the group. All subjects exhibited a mean of approximately ± 2 SDs, while subject 2 demonstrated more than 3 SDs above the mean.

Subsequently, after analyzing the methylation data of the two subjects with the EpiSign tool, we confirmed the findings definitively. This demonstrates that the episignature associated with this condition is sufficiently sensitive to be detected by conventional statistical methodologies, even in studies with small sample sizes.

Discussion

Widespread availability of next-generation sequencing methodologies has increased the diagnostic yield when investigating neurogenetic conditions. However, the amount of data generated poses a challenge in variant interpretation, particularly when dealing with variants of uncertain significance. Additionally, laboratories have differing policies and internal variant re-interpretation guidelines, which leads to heterogeneous reporting practices [11].

Clinicians face a conundrum when a VUS is identified in a test report. While its presence is insufficient to conclusively determine a diagnosis, at the same time it cannot be entirely disregarded. Functional validation of testing results is often too costly or unavailable, and in conditions with reduced penetrance such as KMT2B-dystonia, segregation analysis may not conclusively rule in or out a diagnosis. Often, this context leaves clinicians confused and/or frustrated.

A conclusive diagnosis ends the diagnostic odyssey and allows for proper genetic counseling. Additionally, in some conditions like KMT2B-dystonia, results have direct implications on patient management. These patients are typically poorly responsive to oral drugs, but often benefit from advanced neurosurgical therapies such as deep brain stimulation (DBS). The ability to functionally validate a variant of uncertain significance (VUS) in KMT2B enhances patient care by enabling DBS stratification with appropriate risk–benefit considerations, ultimately improving quality of life [4].

Our work supports the use of methylation analysis when interpreting VUSs in KMT2B. Subject 1 did not have a typical phenotype and it was possible to rule out a contributing role of the KTM2B variant identified on WES. Clinical follow-up supports this hypothesis, as she has not had any further episodes in over one year. Even considering the evidence for allelic series, suggesting moderate effects sizes and variation in the clinical phenotype [9], KMT2B-related dystonia does not appear to be a diagnostic possibility for this patient. However, a potential biological effect of the L1720F variant cannot yet be entirely ruled out in this case. Subject 2 had a more typical phenotype, and the functional analysis results allowed for more certainty in diagnosis, with genetic counseling and stratification for future DBS.

It is likely that methylation analysis studies will become more frequent in clinical practice. Genome-wide DNA methylation disruption has been observed in a significative number of congenital anomalies and neurodevelopmental disorders. As of current, about 70 different Mendelian and non-Mendelian disorders present an EpiSign (in addition to the established episignatures for tumor entities). This number tends to increase as new, larger, and more representative series are investigated [12–16].

A growing number of studies suggest that DNA methylation dosage assays can solve cases that could not be explained by traditional genetic tests [17, 18]. It is conceivable that for some clinical entities, identifying episignatures may become the first tier testing modality, before even the genomic variants investigate themselves [7].

Conclusion

This work supports the use of methylation analysis as means of functional validating variants in KTM2B, with direct implications for genetic counseling and clinical management. Unfortunately, the ACMG guidelines for variant analysis do not yet include a criterion that directly evaluates methylation data. However, our results indicate that the KMT2B Y2515C variant can be re-classified as likely pathogenic. To consider this change in classification, it is necessary to adapt the ACMG criteria for functional evidence, considering DNA methylation results as a functional assay. Regarding the evidence from allelic series for KMT2B variants, which suggest moderate effect sizes on the clinical phenotype and potential influences on the episignature, we were unable to propose a reclassification to likely benign for L1720F variant. Methylation status investigation is a powerful tool to accumulate variant pathogenicity evidence and improve diagnostic yield in patients with VUS in KMT2B. The episignature model is a promising study for a growing number of neurogenetic conditions.

Author contributions

G.F.S. Carvalho and C.M. Gusmão were involved in writing—review and editing, formal analysis, methodology; B.M. Wolff, L.L. Vieira, M.R. Costa, and R.S. Mendes helped in methodology—methylation assay, review; Y. Gasparini, M.A.A. Castro, M.T. Sakuma, and F. Kok contributed to review; B. Sadikovic helped in EpiSign analysis; L.D. Kulikowski took charge of project administration and funding acquisition.

Funding

This work was supported by the Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq-402951/2021-2 and CNPq-305313/2023-1).

Availability of data and materials

All raw files used in this article are available upon request from the corresponding author.

Declarations

Ethics approval and consent to participate

The ethics committee of the University of São Paulo approved this study (HC-FMUSP—CAPPesq 4 0809520.3.0000.0068/2021). The patients or parents of the patients signed the consent form for participation in the study.

Consent for publication

All authors have agreed to have this article published.

Competing interests

The authors declare no competing interests.