Clinical Significance of Rare Copy Number Variations in Epilepsy:  A Case-Control Survey Using Microarray-Based Comparative Genomic Hybridization FREE

Quiz Ref IDIn approximately 30% to 40% of patients with epilepsy, there is an underlying cause of seizure recurrence, but it is not identifiable in most subjects.¹ Although causative mutations have been identified for several epilepsy syndromes, mendelian epilepsies account for only 1% of the cases.² By contrast, common epilepsies have a complex pattern of inheritance, likely involving combinations of variants in different genes that have proven challenging to discover.² Linkage and genome-wide association studies have failed to identify robust or unambiguous associations in large series of patients.³

Quiz Ref IDCopy number variations (CNVs) are increasingly recognized as a source of phenotypic variation among humans, and the implementation of genome-wide technologies such as microarray-based comparative genomic hybridization (array-CGH) and single-nucleotide polymorphism genotyping now allows their efficient identification.^4- 5 A number of studies have recently highlighted the role of CNVs in the etiology of various neuropsychiatric disorders.^6- 7 Targeted and genome-wide surveys of CNVs have been also performed in individuals with epilepsy, revealing recurrent deletions at 15q11.2, 15q13.3, and 16p13.11 in a small proportion of patients with different epilepsy syndromes.^8- 11 However, the role of nonrecurrent CNVs in the etiology of epilepsy has not yet been evaluated in detail.^2,5 We performed an array-CGH–based survey of CNVs in subjects with epilepsy of unknown cause.

We prospectively collected records of patients with unexplained epilepsy referred to different centers between September 1, 2007, and August 31, 2009. Clinical evaluation included general and neurological examination, familial and personal history, electroclinical findings, 1.5-T brain magnetic resonance imaging, treatment schedule, and outcome. Seizures and epilepsy type were defined according to international recommendations.¹² The patients' IQs were defined according to Diagnostic and Statistical Manual of Mental Disorders, fourth edition criteria (borderline intellectual functioning, IQ 71-84; mild mental retardation [MR], IQ 60-70; moderate MR, IQ 50-59; severe MR, IQ <50). We excluded patients with the following: (1) lesional or metabolic cause of epilepsy; (2) signs and symptoms of epileptic encephalopathy; (3) major congenital malformations or pathognomonic dysmorphisms; and (4) severe MR or major neuropsychiatric diseases (eg, schizophrenia, tic disorders, severe mood disorder). The ethics committee of each center approved the study, and informed consent was signed by participants.

Control populations comprised subjects with nonsyndromic MR but no epilepsy (control subjects with MR) and healthy blood donors (healthy control subjects). Subjects with MR only were selected from a consecutive series of cases referred to the Laboratory of Genetics, Galliera Hospital, Genova, Italy, between January 1, 2008, and December 31, 2010, and classified according to Diagnostic and Statistical Manual of Mental Disorders, fourth edition criteria. Healthy control subjects were selected among adults aged 40 to 60 years through a modified version of a structured questionnaire to exclude any seizure disorder.¹³

Genomic DNA of patients and control subjects was extracted from blood lymphocytes. Samples from patients and control subjects were screened by the Human Genome CGH Microarray 44B kit (Agilent Technologies, Inc, Santa Clara, California). Identified CNVs were checked in the Database of Genomic Variants (http://projects.tcag.ca/variation/) and in a data set of 790 control subjects with neurological disease who were of European descent.¹⁴ We adopted a quite conservative approach and did not follow up rearrangements overlapping more than 50% with CNVs reported in control subjects. Validation and parental origin of CNVs were obtained by fluorescence in situ hybridization, higher-density array-CGH slides (Microarray 244K kit; Agilent Technologies, Inc), or quantitative polymerase chain reaction.

We used array-CGH containing about 44 000 60-mer oligoprobes covering the nonredundant genome at an average distance of approximately 43 kilobases (kb) and allowing the genome-wide survey of CNVs. Labeling and hybridization were performed following the manufacturer's protocol. The arrays were scanned (Microarray Scanner G2565CA; Agilent Technologies, Inc) and analyzed by Feature Extraction version 9.5.0 software and Genomic Workbench version 5.0.14 software (Agilent Technologies, Inc). Assays showing a derivative log ratio spread score higher than 0.3 were excluded. Detection of gains and losses was performed using the Aberration Detection Method 2 algorithm with a moving average of 500 kb and a threshold of 6.0. In addition, we set the minimum number of consecutive probes to call CNVs at 8. The same protocol was applied for the higher-density Microarray 244K kit (Agilent Technologies, Inc).

Bacterial artificial chromosome clones were selected according to the UCSC Genome Browser hg18 assembly (http://genome.ucsc.edu; Genome Bioinformatics Group, Center for Biomolecular Science and Engineering, University of California, Santa Cruz). Bacterial artificial chromosome DNA was labeled by nick translation. Slides were counterstained with antifade Vectashield Mounting Medium with 4’,6-diamidino-2-phenylindole (Vector Laboratories, Inc, Burlingame, California). Signals were visualized on a Nikon E1000 microscope (Nikon Corp, Tokyo, Japan) equipped with a charge-coupled device camera and Genikon image analysis software (Genikon, Plano, Texas).

Quantitative polymerase chain reaction was performed using the TaqMan technology on an ABI 7500 instrument (Applied Biosystems, Inc, Foster City, California). The copy number of the unknown samples was normalized to an endogenous reference (RNase P gene). Locus-specific TaqMan probes were selected from the Applied Biosystems database.

Sample power estimates were calculated assuming an α error of .05 and a β error of 0.2 (ie, a power of 80%). Two-tailed unpaired t test and Fisher exact tests were used for categorical sample and phenotype correlations; the nonparametric Mann-Whitney U test was used to evaluate numerical data. The BioMart tool (http://www.ensembl.org/biomart/martview/f3cb7e173fa0ea9eeb69a890e659fa4a) was used to retrieve the lists of RefSeq deleted and duplicated genes in patients and control subjects. In addition, we searched for Gene Ontology (GO) annotations that are overrepresented in the group of rearranged genes using GeneCodis version 2.0 software (http://genecodis.dacya.ucm.es).¹⁵ This tool allows the classification of genes according to their putative biological function by screening the GO annotations included in the GO Slim database and identifying significant enrichment of specific GO annotations among any set of genes by normalizing for the total number of human genes belonging to each GO group using the hypergeometric distribution and correcting resulting P values for multiple hypothesis testing using the false discovery rate correction.¹⁵ The DECIPHER database of genomic unbalances was used for genotype-phenotype correlations (https://decipher.sanger.ac.uk/).

Records from 279 patients with epilepsy (152 males), 265 control subjects with MR (117 males), and 246 healthy control subjects (127 males) were collected. In the patients group, epilepsy type was generalized in 188 patients and focal in 91 patients. Eighty-one subjects (29.0%) showed slight to moderate MR or additional neuropsychiatric features (Table 1). The initial survey led to the identification of 182 CNVs (58 in patients with epilepsy, 74 in control subjects with MR, and 50 in healthy control subjects) (data available on request). Among these, 92 CNVs (ie, 28 in 26 patients with epilepsy, 48 in 45 control subjects with MR, and 16 in 16 healthy control subjects) were absent in the Database of Genomic Variants and were further followed up (Table 2). The overall rate of rare CNVs in patients with epilepsy (9.3%) was similar to that in healthy control subjects (6.5%) (P = .26) and significantly lower than that in control subjects with MR (17.0%) (P = .01). Power estimates, calculated on the frequency of rare CNVs in the control population, indicate that the available sample can detect at least a 2.1-fold increase in the frequency of rare CNVs among patients.

A total of 28 rare CNVs were identified in 26 patients with epilepsy (9.3%; 16 males) (Figure and Table 2). No recurrent CNVs were found except for the 22q13.32qter deletion (cases C012 and B004). Locus 19q13.43 was deleted in 1 case (case A022) and duplicated as part of a larger rearrangement in another one (case A211). Twenty-four CNVs (92.3%) involved a single chromosomal segment (12 deletions, 12 duplications), whereas the other 2 involved various segments in the same (case A096) or different (case A164) chromosomes. The CNVs varied in size from 0.12 to 13 megabases (Mb). Likewise, the number of rearranged genes varied from 1 to more than 100 (Table 2). Two duplications composed a single gene (cases C106 and A049). The parental origin of the rearrangements was assessed in 21 patients: 10 occurred de novo, 9 were inherited from a healthy parent (5 paternal, 4 maternal), and 2 resulted from unbalanced segregation of parental balanced rearrangements. Altogether, de novo CNVs were larger (P = .03) and showed higher gene content (P = .006) compared with those inherited, whereas the type of CNV (deletions vs duplications) did not differ between the 2 groups (P = .17).

The clinical features of individuals carrying these CNVs are shown in Table 3. Age at epilepsy onset ranged from 3 months to 18 years (mean [SD], 7.1 [5.0] years). Epilepsy was generalized in 16 subjects and focal in 10. Fourteen patients (53.8%) had MR, and 10 of these had MR associated with autistic tracts or attention-deficit/hyperactivity disorder. Nine patients (34.6%) showed rearrangements that have been already associated with microdeletion or microduplication syndromes (Figure). These rearrangements compared with the novel ones were more frequently deletions (7 of 8 vs 5 of 17, respectively; P = .03) and larger in size (mean, 5.2 vs 1.9 Mb, respectively; P = .04), whereas they did not differ in gene content (mean, 45.8 vs 20.0 genes, respectively; P = .13) or parental origin (de novo vs inherited, respectively; P = .07). No newly identified CNVs matched those reported in the DECIPHER database.

Among 246 ethnically matched healthy control subjects, we identified 16 unreported CNVs (3 deletions, 13 duplications) (Table 2). Largely overlapping 6q22.31 deletion (case 02957M) and 9p21.2 duplication (case 01503M) were identified in patients (cases I091 and E011), indicating 2 novel, recurrent, polymorphic CNVs.

The CNVs identified in patients compared with healthy control subjects were larger (mean, 3.04 vs 0.79 Mb, respectively; P = .03) and showed higher gene content (mean, 31.3 vs 7.1 genes, respectively; P = .02), whereas the type of unbalance (deletions vs duplications, respectively) did not differ between the 2 groups (P = .06). The CNVs larger than 1 Mb (P = .002) and including more than 10 genes (P = .005) occurred more frequently in patients than in healthy control subjects. Conversely, the 2 groups did not differ for CNVs smaller than 1 Mb (P = .05) and involving 10 or fewer genes (P = .55).

Genotype-phenotype correlations indicated that having MR or neuropsychiatric features was significantly associated with the occurrence of CNVs compared with not having them (14 of 81 cases [17.2%] vs 16 of 246 cases [6.5%], respectively; P = .004), whereas epilepsy type (focal vs generalized; P = .52) was not. In addition, CNVs in patients with epilepsy who have MR or neuropsychiatric features were larger (P = .02) and showed higher gene content (P < .001) compared with healthy control subjects, whereas the size and gene content of CNVs identified in patients without MR or neuropsychiatric features did not differ from those in healthy control subjects.

The incidence of CNVs in patients with epilepsy who have MR or neuropsychiatric features and in control subjects with MR is similar (45 of 265 cases [17.0%] vs 14 of 81 cases [17.2%]; P > .99). Likewise, the 2 groups did not differ in CNV size (mean, 4.51 vs 3.78 Mb, respectively; P = .54) or gene content (mean, 37.8 vs 29.7 genes, respectively; P = .43). Among CNVs identified in control subjects with MR, 3 were also observed in the epilepsy group. The 6q22.31 duplication has been also detected in 2 control subjects with MR. The 15q24.1 deletion was identified in a patient with epilepsy showing different types of generalized seizures plus mild MR with hyperactivity and in a control individual with moderate nonsyndromic MR. The 16p13.11 deletion was detected in a mentally healthy patient with idiopathic generalized epilepsy (case I104) and a control subject with MR showing moderate MR and slight dysmorphisms.

A set of 17 rearrangements identified in patients with epilepsy were validated by an independent assay, ie, the 244K array-CGH (cases C012, B004, C106, and A049), fluorescence in situ hybridization (cases C049, A096, C045, A018, A164, A139, A066, A211, and A022), or quantitative polymerase chain reaction (cases A181 and A002 and controls 14F and 18F). Validation assays confirmed the occurrence of CNVs in all of the cases (data available on request).

A total of 791 genes were involved in rearrangements identified in patients (data available on request). We classified the rearranged genes according to the putative biological role attributed by the GO database, and we searched for a significant enrichment of specific GO category using GeneCodis version 2.0. A total of 593 genes were associated with at least 1 GO annotation (Table 4). Genes encoding ion channels (P = .01) were the most significantly enriched among rearranged genes. In addition, significant associations were detected for genes involved in posttranslational modification of proteins (P = .01), cell differentiation (P = .02), and metabolic processes (P = .02) (Table 4). GeneCodis analysis performed on 1327 genes (1116 with GO identification) rearranged in control subjects with MR revealed a significant enrichment of genes involved in development (P = .004), signal transduction (P = .005), metabolic processes (P = .008), and cell differentiation (P = .008) (Table 4). By contrast, we did not find any overrepresented GO category in healthy control subjects.

Recent studies have shown that recurrent rearrangements such as 15q11.2, 15q13.3, and 16p13.11 microdeletions represent a risk factor for a wide spectrum of epilepsies.^8- 11 However, the role of nonrecurrent genomic rearrangements in patients with epilepsy of unknown cause is still unclear and the diagnostic impact of high-resolution screening of CNVs has not yet been assessed.

Quiz Ref IDWe showed that the cumulative incidence of rare CNVs is not significantly higher in patients with epilepsy of unknown cause compared with a healthy control population. However, CNVs identified in patients are larger and include more genes on average; in addition, rearrangements larger than 1 Mb and containing 10 or more genes are significantly overrepresented among patients with epilepsy. These data indicate that only a subgroup of rare CNVs play a role in the etiology of epilepsy.

It should be emphasized that CNV detection is largely dependent on the type of platform (array-CGH vs single-nucleotide polymorphism genotyping), the density and design of the array, and the setting of analytic parameters.¹⁶ Most population-based CNV surveys have been performed by single-nucleotide polymorphism arrays, whereas array-CGH has become the elective routine diagnostic tool to identify genomic rearrangements in patients.¹⁷ We identified 2 previously unreported CNVs in both patients and control subjects, indicating that recurrent polymorphic CNVs may have been missed by large-scale population single-nucleotide polymorphism–based studies. Accordingly, CNV screening should include appropriate internal controls, and heterogeneous data from a public database such as the Database of Genomic Variants should be considered with caution within a diagnostic context.

Our cohort includes patients referred to third-level centers for different epilepsy types of variable severity in the absence of severe MR, major neuropsychiatric disease, or pathognomonic dysmorphisms. These patients represent a common heterogeneous phenotypic group, raising complex diagnostic and management issues.

Genotype-phenotype correlations revealed that the frequency of rare CNVs as well as their size and gene content are significantly higher in patients showing mild to moderate MR or other minor neuropsychiatric features than in healthy control subjects. On the other hand, the incidence, size, and gene content of rare CNVs in patients with epilepsy who have healthy or borderline mental status are not increased compared with our healthy control population. Notably, comparative analysis of CNVs identified in patients with epilepsy and control subjects with MR revealed a minor overlap, suggesting the involvement of specific pathogenetic mechanisms for epilepsy. Although we cannot exclude that this lack of overlap is due to a sampling issue, analysis of putative biological processes related to rearranged genes pointed out the specific role of ion channel genes in the epilepsy group. In this scenario, the analysis of genomic rearrangements emerges as a first-line diagnostic test for patients with epilepsy showing MR and/or neuropsychiatric features, even if mild.

About one-third of the patients showed rearrangements that can be unambiguously considered pathogenetic. In these cases, epilepsy was the predominant phenotypic expression of emerging microdeletion or microduplication syndromes, such as 6q terminal deletion (case C049),^18- 19 invdupdel(8p) (case A096),²⁰ 9q34qter deletion (case A133),²¹ 15q11-q13 duplication (case C070),²² 15q24 deletion (case C045),²³ 16p13.11 deletion (case I104),²⁴ 22q13.3 deletion (cases C012 and B004),²⁵ and Xp22.31 deletion (case A040).²⁶ These syndromes are usually characterized by complex syndromic phenotypes including dysmorphisms, brain structural abnormalities, or multiple congenital defects. Conversely, our patients showed a mild clinical expression featuring mild to moderate MR, no coarse facial features, and no structural abnormalities in the brain or other organs and could not be diagnosed on clinical grounds. The heterogeneous clinical expressivity of these syndromes may be due to several factors. Phenotypic differences among affected individuals may be explained by the parental origin of the rearrangement (15q11-q13 duplication)²² or differences in the size of the rearranged region (15q24 deletion).²³ Variable phenotypic expressivity may also be due to other factors such as unmasking of recessive mutations, somatic mosaicism, and epigenetic or environmental modifiers.^17,27

The assessment of the pathogenetic effect of novel, unreported rearrangements remains an open issue. The de novo origin of a rearrangement is considered an important predictor.^17,27 Most of the novel rearrangements identified in this study were transmitted from unaffected parents. However, population studies indicate that the de novo origin of a CNV does not necessarily imply a pathogenetic effect.^17,27 Consistently, in our series, known pathogenetic and novel CNVs do not differ for the parental origin. Moreover, rearrangements inherited from a healthy parent cannot always be considered a benign variation. Indeed, it was recently demonstrated that CNVs may represent susceptibility alleles for a number of neuropsychiatric disorders, including idiopathic epilepsy.^8- 11

Quiz Ref IDThe size, the type (deletions vs duplications), and the number of genes involved in CNVs represent key factors underlying pathogenesis.²⁷ In our series, novel CNVs are more frequently duplications and are smaller compared with known pathogenetic rearrangements, whereas the gene content is not significantly different. Moreover, we detected a tight link between the mode of transmission and the size and gene content of CNVs. Indeed, inherited rearrangements were smaller and showed lower gene content compared with de novo rearrangements. In this scenario, mode of transmission, type, size, and gene content of a novel CNV should be evaluated together to assess its pathogenetic relevance.

A comprehensive bioinformatic analysis on CNVs identified in our cohort showed significant enrichment of ion transport genes, confirming that the impairment of neuronal ion currents represents a common pathophysiological mechanism in epilepsy.^1- 2,5 In addition, our analysis revealed a significant enrichment of genes involved in cell differentiation and metabolism. This finding is consistent with emerging evidence pinpointing the role of early maturation processes,²⁸ subtle cytoarchitectural defects,^29- 30 and metabolic impairment³¹ in the etiology of epilepsies lacking detectable brain abnormalities. The implication of developmental and metabolic processes in control subjects with MR provides further evidence of a pathogenetic link between epilepsy and MR. On the other hand, the implication in patients with epilepsy of pathways related to posttranscriptional modification of proteins in epilepsy cannot be readily interpreted.

Quiz Ref IDPatients with epilepsy have a significantly increased burden of large, rare, gene-rich CNVs, particularly when associated with MR or other neuropsychiatric features. Accordingly, we suggest that CNV screening should be performed for diagnostic purposes preferentially in such patients. Rearrangements smaller than 100 to 200 kb could have been missed in our study. However, small rearrangements involving few genes were usually inherited from healthy parents and have an uncertain clinical relevance. Therefore, the implementation of very high-resolution screening in clinical practice deserves caution.