Neurological disorder-associated genetic variants in individuals with psychogenic nonepileptic seizures

Psychogenic nonepileptic seizures (PNES) are diagnosed in approximately 30% of patients referred to tertiary care epilepsy centers. Little is known about the molecular pathology of PNES, much less about possible underlying genetic factors. We generated whole-exome sequencing and whole-genome genotyping data to identify rare, pathogenic (P) or likely pathogenic (LP) variants in 102 individuals with PNES and 448 individuals with focal (FE) or generalized (GE) epilepsy. Variants were classified for all individuals based on the ACMG-AMP 2015 guidelines. For research purposes only, we considered genes associated with neurological or psychiatric disorders as candidate genes for PNES. We observe in this first genetic investigation of PNES that six (5.88%) individuals with PNES without coexistent epilepsy carry P/LP variants (deletions at 10q11.22-q11.23, 10q23.1-q23.2, distal 16p11.2, and 17p13.3, and nonsynonymous variants in NSD1 and GABRA5). Notably, the burden of P/LP variants among the individuals with PNES was similar and not significantly different to the burden observed in the individuals with FE (3.05%) or GE (1.82%) (PNES vs. FE vs. GE (3 × 2 χ2), P = 0.30; PNES vs. epilepsy (2 × 2 χ2), P = 0.14). The presence of variants in genes associated with monogenic forms of neurological and psychiatric disorders in individuals with PNES shows that genetic factors are likely to play a role in PNES or its comorbidities in a subset of individuals. Future large-scale genetic research studies are needed to further corroborate these interesting findings in PNES.


Scientific RepoRtS
| (2020) 10:15205 | https://doi.org/10.1038/s41598-020-72101-8 www.nature.com/scientificreports/ phenotypic overlap 10 . Widespread pleiotropy for genes associated with Mendelian forms of neurological and psychiatric disorders suggests that these disorders are the response of a complex neurologic network, which is altered in several domains of function. For example, individuals affected by pathogenic variants in well-established epilepsy genes usually have additional neurological or psychiatric comorbidities 11 . While recent success in genetic studies has led to the identification of disease-associated genetic factors for virtually all disorders comorbid with PNES (i.e., anxiety, bipolar disorder, depression, epilepsy, intellectual disability, migraine, mood disorders, personality disorders, post-traumatic stress disorder, and schizophrenia), a genetic basis of PNES has been speculated 12 , but never formally investigated.
To characterize genetically the heterogeneous phenotypic spectrum of individuals referred to a tertiary care center-we genotyped and sequenced 550 individuals with PNES or epilepsy, from the Cleveland Clinic Epilepsy Center, while excluding individuals with both PNES and epilepsy. We then selected all copy-number (CNVs) and single nucleotide variants (SNVs) with strong computational support for a deleterious effect on the involved gene/s and classified them according to the American College of Medical Genetics and Genomics and the Association for Molecular Pathology (ACMG-AMP) 2015 guidelines 13 . The guidelines require an established gene to phenotype association as one criterion for pathogenicity prediction. Because the genetic basis of PNES is unknown and no established PNES genes exist, we applied the ACMG-AMP guidelines for research purposes only and considered genes associated with neurological or psychiatric disorders as candidate genes for PNES. Our overall study design is presented in Fig. 1.

Results
Cohort and analysis overview. We generated genotyping as well as exome data for 694 individuals ascertained through the Epilepsy Biorepository of the Cleveland Clinic Epilepsy Center. The data was used to perform genome-wide CNV and whole-exome SNV analyses. After data quality control, 550 individuals with epilepsy or PNES were included in the downstream analyses ( Table 1). The most prevalent diagnosis was focal epilepsy (FE, N = 393, 71.5%) followed by psychogenic nonepileptic seizures without epilepsy (PNES, N = 102, 18.5%), and generalized epilepsy (GE, N = 55, 10%).
Individuals in this study were 40 years old on average (standard deviation, SD = 15) and completed approximately 14 years of education (SD = 2). Fifty-three percent of the cohort was female and all of genetically defined European ancestry. The three study groups were well matched in terms of age and education. There were more females in the PNES group (75%) than in the epilepsy groups (FE: 47%, GE: 56%). The PNES group showed the highest proportion of individuals with comorbid psychiatric disorders or a history of chronic pain compared to the two epilepsy groups (Table 1). Individuals with FE were less likely to have a family history of epilepsy, defined as seizures in a first-or second-degree relative than those with GE (χ 2 = 7.70, P = 0.006) or PNES (χ 2 = 6.22, P = 0.013). However, the family history of seizures was similar between individuals with GE and PNES (χ 2 = 0.402, P = 0.526). The percentage of individuals with comorbidities and a family history of seizures was similar between females and males with PNES (Supplementary Table 4).
Individuals with PNES carry pathogenic and likely pathogenic variants similar to individuals with epilepsy. Rare genetic variants are well-established risk factors associated with epilepsy 14 . We first examined if individuals with PNES carry deleterious genetic variants, as classified by state-of-the-art in silico prediction methods. The results were compared to the control group of 448 individuals with epilepsy from the same epilepsy clinic. Identified potential deleterious insertion/deletion polymorphisms (indels, N = 56) did not survive our strict criteria for the visual inspection of the reads supporting each variant and were excluded from subsequent analyses. After data quality control and stringent variant filtering, we identified deleterious variants in 8 (Fig. 2). The proportions of individuals carrying only a specific variant type (single nucleotide or copy number variants) are displayed in Supplementary Figures 1 and 2. The deleterious variant burden across the three disorder types was not significantly different (PNES vs. FE vs. GE, two-tailed 3 × 2 χ 2 , P = 0.59; PNES vs. epilepsy, two-tailed 2 × 2 χ 2 , P = 0.38).
Next, we classified all deleterious variants according to the ACMG-AMP 2015 guidelines to determine if individuals with PNES carry ACMG-AMP-classified pathogenic (P) or likely pathogenic (LP) genetic variants in genes associated with neurological disorders, and/or psychiatric disorders, including epilepsy. We found P/ LP variants in 3.05% of all individuals with FE (N = 12, Supplementary  (Fig. 2). The P/ LP variant burden across the three disorder types was not significantly different (PNES vs. FE vs. GE, two-tailed 3 × 2 χ 2 , P = 0.30; PNES vs. epilepsy, two-tailed 2 × 2 χ 2 , P = 0.14). We also identified ACMG-AMG variants of uncertain significance (VUS) in 5 (Fig. 2). Out of all six P/LP variant carriers with PNES, one (16.7%) reported a family history of seizures, and two (33.3%) a family history of psychiatric disorders (anxiety, mood disorders, and post-traumatic stress disorder).   Table 3 and Supplementary Tables 5 and 6. Six individuals with PNES were affected by P/LP variants (three males and three females). These variants included four CNVs and two missense variants ( Table 2). Three out of the six P/LP variants were found to be known epilepsy-associated CNVs (deletions at 10q23.1-q23.2, distal 16p11.2, and 17p13.3). Deletions at 10q23.1-q23.2 are associated with complex genetic neurodevelopmental syndromes 15 . Our patient with PNES and the 10q23.1-q23.2 deletion had normal neurodevelopment, an onset of PNES at 33 years of age, a history of migraine, chronic pain and mood disorder, and no history of febrile seizures or family history of seizures.
Deletions and duplications at distal 16p11.2 locus are associated with epilepsy, autism spectrum disorder, schizophrenia, lower IQ, and subcortical brain abnormalities 16 . Our patient with PNES and the distal 16p11.2 deletion had normal neurodevelopment, an onset of PNES at 55 years of age, and no history of febrile seizures or a family history of seizures. The third epilepsy-associated variant in PNES was a PAFAH1B1 gene disrupting 17p13.3 deletion. PAFAH1B1 is a well-established gene for lissencephaly with seizures as the core symptom of this disorder 17 . The 17p13.3 deletion also disrupts the gene METTL16. However, METTL16 is not associated with any disease and is unlikely to play a role in PNES causation, because METTL16 is known to tolerate missense (missense Z-score = 2.35) and loss-of-function variants (pLI = 0.61) 18 . Our patient with PNES and deletion of PAFAH1B1 had normal neurodevelopment, an onset of PNES at 30 years of age, a history of anxiety and posttraumatic stress disorder (PTSD), and no history of febrile seizures or family history of seizures.
The next three P/LP variants (one P, two LP) were identified in two genes and at a chromosomal position, each implicated in neurological or psychiatric disorders with or without seizures (NSD1, 10q11.22-q11.23 deletion, and GABRA5; Table 2). Heterozygous mutations in the NSD1 gene are associated with Sotos syndrome  19 . Sotos syndrome is characterized by distinctive facial features, intellectual disability (ID), and overgrowth or macrocephaly, and seizures are reported in 9-50% of cases 20 . Our patient with PNES and an NSD1 mutation had normal neurodevelopment, an onset of PNES at 12 years of age, a history of mood disorder, and no history of febrile seizures or family history of seizures. Deletions and duplications at 10q11.22-q11.23 are associated with developmental delay or ID 21 . Our patient with PNES and the 10q11.22-q11.23 deletion had normal neurodevelopment, an onset of PNES at 50 years of age, a history of mood disorder and anxiety, no history of febrile seizures, and a suggestive family history of seizures (affected aunt). Finally, GABRA5 is a candidate gene for epilepsy and developmental delay 22 , which has not yet been statistically associated with epilepsy through an exome-wide cohort screen. Our patient with PNES and a GABRA5 mutation had normal neurodevelopment, an onset of PNES at 40 years of age, a history of mood disorder, anxiety, and PTSD, and no history of febrile seizures or family history of seizures. Out of the six identified VUS, four were identified in genes with a potential role in brain function (LHX9, MAPKAPK2, CAMKV, and MYH9). Expression of LHX9 was shown to be repressed by kainic acid-induced seizures 23 , while MAPKAPK2 expression was induced 24 . CAMKV encodes for a synaptic protein crucial for dendritic spine maintenance 25 . Heterozygous mutations in MYH9 are associated with a spectrum of autosomal dominant thrombocytopenias 26 . The most devastating consequence of thrombocytopenia is intracranial hemorrhage, and one case with recurrent seizures related to intracranial hemorrhage was recently described in the literature 27 .

Discussion
We generated whole-exome sequencing and whole-genome genotyping data to identify rare, P/LP variants in a cohort of 550 individuals with PNES or epilepsy (focal or generalized). This study represents the first genetic investigation of PNES. We used the ACMG-AMP 2015 guidelines 13 to classify variants from the perspective of Mendelian forms of neurological or psychiatric disorders, including epilepsy. We show that P/LP variants in genes implicated in a broad range of neurological and psychiatric disorders are found in individuals with PNES evaluated in a tertiary care epilepsy center. Interestingly, we did not observe a significant difference in the burden of P/LP variants, when comparing individuals with PNES without coexistent epilepsy to individuals with epilepsy. The observed variant burden is not surprising for the epilepsies, and elevated burdens of CNVs or SNVs have been observed in several large-scale studies that compared individuals with epilepsy against population controls 11,14,28,29 . In contrast, genetic factors have not been previously identified for PNES.
We executed a very stringent state-of-the-art variant filtering strategy, optimized for high specificity to identify deleterious SNVs. Following evidence from large-scale studies in epilepsy and other neurological disorders showing that most of the causal variants are ultra-rare in the general population 14,30 , we only selected and classified unique SNVs not seen in > 200 k population controls. We only considered CNVs if the locus had previously been associated with neurological or psychiatric disorders. The goal of this strategy was to prioritize genetic variants that have a high likelihood to be true and pathogenic for a robust comparison of the PNES and epilepsy groups. Future studies will have to balance sensitivity vs. specificity to prioritize variant discovery. www.nature.com/scientificreports/ Among the 12 deleterious variants identified in individuals with PNES without coexistent epilepsy, 50% (6/12) were classified according to the ACMG-AMP 2015 guidelines as pathogenic or likely pathogenic. The ACMG-AMP 2015 guidelines require an established gene to phenotype association as one criterion for pathogenicity prediction 13 . In a clinical setting, the same variants would not be classified as pathogenic since no single established PNES gene exists. Nevertheless, the detection of pathogenic variants that are likely to cause Mendelian forms of neurological or psychiatric disorders in individuals with PNES is in line with emerging evidence that neurological or psychiatric disorders share a broad range of pleiotropic acting genetic variation 8,11,31 . Our results are also in line with the observation that up to 48% of all individuals with PNES report a family history of epilepsy and 22% a family history of psychiatric disorders 6 . PNES could represent one of several clinically defined phenotypes associated with pleiotropic acting genetic variants, affecting genes essential for brain development and function. Surprisingly, none of the P/PL variant carriers with PNES had an abnormal bedside cognition/neurological examination or any major structural abnormality on MRI. However, formal neuropsychological testing was not performed; therefore, it is possible that individuals had more subtle forms of cognitive impairment. It is also possible that our results represent phenotype expansions for some of the genes affected by P/PL variants. Phenotype expansions are common as shown for SLC6A1, a known gene for autism spectrum disorders, epilepsy, and ID 32 , which was recently associated with schizophrenia without ID or other neurodevelopmental disorder 33 . Finally, despite an ACMG-AMP 2015 guidelines classification as VUS, we cannot exclude the involvement of the six identified VUS in PNES causation, as all variants identified in this study were predicted in silico as highly pathogenic, never seen in the general population, and affected highly variant intolerant genes. Future case/control studies are needed to identify genes that have not been implicated in other disorders as associated with PNES.
The four identified deletions in individuals with PNES in this study have been reported in the literature in epilepsy and other neurological disorders, with high phenotypic variability and incomplete penetrance 15,16,21,34 . This high degree of pleiotropy has been observed for the vast majority of all deletions that have been associated with epilepsy or other neurological disorders 35 . Most likely, these deletions impair neurodevelopmental processes  www.nature.com/scientificreports/ in a rather nonspecific manner and contribute to the genetic variance of a broad spectrum of neurological disorders. The specific disease phenotype is likely further specified by the interplay with genetic background effects and environmental influences following an oligo-/polygenic inheritance model with substantial genetic heterogeneity. Environmental factors which predispose to PNES are very well established 1 . Genetic vulnerability for a neurological or psychiatric condition could predispose to PNES when combined with environmental stressors. Individual differences in the genetic vulnerability to specific types of trauma and other environmentally relevant variables have been demonstrated, for example, for PTSD 36,37 . In our study, the P/LP variant carriers showed a similar rate of presence/absence of a history of trauma or abuse as individuals with PNES and no identified variants (P/LP carriers: 3/3 vs. no variants identified: 30/25). However, more research is needed to support this hypothesis and to explain our observation that individuals with PNES alone can carry pathogenic variants that affect genes linked to clinically severe phenotypes.
In conclusion, in this report, we provide the first evidence that genetic factors may play a role in the etiology of PNES. Future large-scale projects that employ comprehensive genetic testing, including polygenic risk scores, are needed for PNES and related genetically understudied disorders such as PTSD, and other conversion disorders. Potentially, such work could provide clues to the etiology and pathophysiology of the disorder, enable the identification of disease biomarkers, and set new directions for the development of new therapies for PNES.

Material and methods
Study participants. Data for this study were obtained from a Cleveland Clinic Institutional Review Boardapproved epilepsy biorepository. All methods were performed in accordance with relevant guidelines and regulations regarding research involving human subjects. All participants provided written, informed consent, and a blood or saliva sample for use in medical/genetic research. Participants were selected for study inclusion if they met the following criteria: (1) age 18 years or older, (2) video-EEG confirmed diagnosis of focal epilepsy (FE), generalized epilepsy (GE), or psychogenic nonepileptic seizures (PNES) without comorbid epileptic seizures, and (3) had blood or saliva DNA available for whole-exome sequencing (N = 694). Our study cohort is detailed in Fig. 1. Phenotyping procedures. Initial medical record review was conducted for all potential study participants by an epilepsy biorepository research coordinator trained in clinical epilepsy phenotyping (L.F.) and supervised by clinical epileptologists (J.F.B., L.J.). Diagnoses of epilepsy or PNES were established after review of the history and physical examination, scalp EEG video evaluation report, progress notes, social work, psychology and/or psychiatry report, and discharge summary as well as other documentation pertinent to the medical history and clinical diagnoses. Psychiatric diagnoses were established through review of the medical records and patient self-reports during the video-EEG admission. Individuals were classified as having FE, GE, or PNES based on video-EEG and concordant history. All individuals classified as PNES on initial review were re-reviewed by a clinical epileptologist (J.F.B.) to confirm the diagnosis. The diagnosis of PNES required video-EEG recording of a semiology typical for PNES, and an absence of epileptiform abnormalities on either interictal or ictal EEG 38 . Accordingly, all individuals with PNES included in this study had video-EEG diagnosed PNES without comorbid epilepsy. Individuals with episodes explained by another diagnosis such as physiologic nonepileptic events (e.g., syncope, migraine, sleep disorder) were excluded from the study as were those with comorbid PNES and epileptic seizures and those with episodes consisting of purely subjective symptoms without objective signs (to prevent the inclusion of individuals with epileptic auras, which often do not have EEG abnormalities).

Copy number calling.
A total of 688,032 single nucleotide polymorphisms (SNPs) were genotyped for all samples of this study using the Global Screening Array with Multi-disease drop-in (GSA-MD v1.0) (Illumina, San Diego, CA, USA). The SNP-data was used to detect CNVs in our dataset, using PennCNV's copy number variant (CNV) calling algorithm 39 with GC-wave adjustment. We generated a custom population B-allele frequency (BAF) file before calling CNVs. Adjacent CNVs were merged if the number of intervening markers between them was less than 20% of the total number of the whole segment encompassing both CNVs. CNV calling was followed by extensive quality control (QC) for both samples and CNVs, respectively. Samples with signal intensity log R Ratio (LRR) standard deviation < 0.23, variability of the average LRR values in sliding windows (waviness factor, WF) < 0.02, departure of the BAF from the expected values for two copies (BAF drift) < 0.003, total numbers of CNVs < 80, and European ancestry were included in the analysis. CNV calls were removed from the dataset if they spanned less than 20 markers, were less than 20 Kb in length, had a marker density (amount of markers/length of CNV) < 0.0001, overlapped by > 50% of their total length with regions known to generate artifacts 40 , or had a frequency > 1% in the study sample. CNVs that were spanning more than 20 markers over ≥ 1 Mb were included in the analysis, even if the marker density was < 0.0001. All CNVs of interest were examined visually by plotting the signal intensities using PennCNV 39 (Supplementary Figures 3-9).
Whole-exome sequencing. Whole-exome sequencing (WES) of all samples in this study was performed using Nextera Rapid Capture Exomes enrichment and paired-end reads (151 bp) Illumina sequencing on the Illumina HiSeq 4,000. Duplicate read removal, format conversion, and indexing of the reads, aligned to the GRCh37 human genome reference (RefSeq: GCF_000001405.13), were performed using Picard (https ://broad insti tute.githu b.io/picar d). All samples were jointly called using the Genome Analysis Toolkit (GATK) best practice pipeline 41  We also filtered out samples that exceeded three standard deviations (SD) from the mean of the entire study cohort on any of the following WES metrics: (i) low mean DP; (ii) low singleton count; (iii) low SNV count; (iv) low or high singleton/SNV ratio; (v) low transition/transversion ratio; (vi) low or high heterozygous/homozygous variant ratio; and (vii) low or high insertion/deletion ratio. Finally, we applied the following exclusion thresholds for variants in the samples that survived previous QC filtering: (i) genotype call rate < 0.95; (ii) minor allele frequency > 0.1; (iii) deviation from the Hardy-Weinberg equilibrium with P < 1 × 10 -20 . The supporting aligned reads of all variants that survived filtering for deleteriousness, detailed in the following paragraph, were visually inspected using the IGV browser 44 .
Variant deleteriousness assessment. We applied for the identified CNVs and SNVs two different strategies to assess the likelihood of a deleterious effect on disease-relevant loci or genes. CNVs: We only considered CNVs as deleterious if the locus had previously been associated with neurological or psychiatric disorders. SNVs: We used ANNOVAR 45 with custom databases to perform a state-of-the-art variant annotation for subsequent deleteriousness assessment. We applied two different filters, based on: i) variant type and frequency and (ii) predicted variant deleteriousness. The frequency-based filter was based on the following criteria: (i) variant not in a genomic duplication (> 1,000 bases of non-repeat masked sequence); (ii) not present in multiple variant databases, totaling > 200 k population controls (Supplementary Table 3). From the remaining variants, we selected only variants with a high confidence prediction to be deleterious using following criteria: (i) loss-of-function (LoF) variants ranked in the top 1% most deleterious variants in the human genome (scaled CADD ≥ 20) 46 found in known epilepsy genes or highly LoF-intolerant genes 18 (probability, pLI ≥ 0.95); (ii) missense variants ranked in the top 1% most deleterious variants in the human genome (scaled CADD ≥ 20) 46 found in missense-constrained regions (MPC ≥ 2 and MTR centile < 15%) 47,48 , of epilepsy genes or missense intolerant genes (missense Z-score > 3.09, corresponding to P < 10 -3 ) 18 .

ACMG-AMP 2015 classification of the deleterious CNVs and SNVs.
The pathogenicity of all deleterious SNVs was assessed according to the ACMG-AMP guidelines 13 using InterVar 49 . We adjusted the InterVar interpretations were needed for: (i) variant type in genes with a known mode of inheritance; (ii) location of the variant in protein; (iii) absence of variant in population controls; (iv) known gene function and association with disease. Deleterious SNVs expected to cause disorders that are not neurological or psychiatric (i.e., incidental findings), were considered as variants of uncertain significance. The pathogenicity of all deleterious CNVs was assessed directly, using the ACMG-AMP criterions PVS1, PS1, and PS4 13 , when supported by definite evidence.

Data availability
Data are available from the corresponding authors on reasonable request.