Introduction

Rolandic epilepsy (RE), or benign epilepsy with centrotemporal spikes (BECTS), is one of the most common epilepsy syndromes of childhood, comprising about 15% of epilepsies in children under the age of 16 years [1]. Age of onset ranges from years 3 to 13 and peaks between 8–9. Characteristic features are 1.) a somatosensory onset with unilateral paresthesias involving the tongue, lips, gums, and inner cheeks 2.) unilateral, tonic, clonic or tonic-clonic convulsions involving the face, lips, tongue as well as the pharyngeal and laryngeal muscles, causing 3.) speech arrest and drooling due to sialorrhea and saliva pooling. At this stage the seizure may end, or it may develop into a generalized tonic clonic seizure. Nocturnal seizures, the most frequent variant of this syndrome, frequently become generalized. The electroencephalographic hallmark, a prerequisite of diagnosis, are blunt high-voltage characteristically shaped centrotemporal spikes (CTS), often followed by slow waves [2], [3]. Based on a number of family studies it is generally assumed that, like in all other common idiopathic epilepsies a multifactorial mode of inheritance appears most likely [2], [4], [5]. The most recent family study on RE reported an increased rate of RE, febrile seizures, and an “epilepsy aphasia spectrum disorder” in relatives of children with RE [6]. To date, a number of loci or genes have been linked to various forms of idiopathic focal epilepsies or the EEG endophenotype of centrotemporal spikes (CTS). Linkage was reported to markers on 15q13.2 and to 16p12-11.2 [7], [8]. But no causative gene has been reported at either loci yet. In a small number of cases variants and mutations in KCNQ2 and KCNQ3 were found to be associated with RE and the respective EEG trait [9]. In addition, mutations in SRPX2 in two families with mental retardation, severe language dysfunction and rolandic seizures have been reported [10]. By genome-wide linkage analysis RE and the CTS trait have been associated with the elongator protein complex 4 [11]. Furthermore, in a girl with early-onset epileptic encephalopathy and CTS a de novo GRIN2A mutation was identified [12]. However, all these findings in single or few patients or small families still lack replication.

Rare copy number abnormalities of RBFOX1 have also been associated with mental retardation in comorbidity with and without seizures, attention deficit disorder and autism [13], [14], [15], [16]. Recently, we have shown in 1408 idiopathic generalized epilepsy (IGE) patients and 2256 population controls that deletions affecting 5′ located exons of RBFOX1 are significantly enriched in IGE cases compared to population matched controls [17]. The RBFOX genes (RBFOX1: chr16∶6069132-7763340, NM_001142333; RBFOX2:, chr22∶36134783-36424585, NM_001082578, and RBFOX3: chr17∶77085427-77512230; NM_001082575) encode neuron-specific splicing factors predicted to regulate neuronal splicing networks. Several epilepsy candidate genes are downstream targets of Rbfox proteins (FLNA, SLC1A3, DCX, GABRB3, GAD2, KCNQ2, SCN8A, SLC12A5, SV2B, SYN1) and their regulation of expression and splicing has been demonstrated [18], [19]. FOX family members regulate splicing of the other FOX members and autoregulate themselves [20], [21].

The present candidate gene analysis tested whether (i) deletions in RBFOX1, RBFOX2 and RBFOX3 might increase risk of RE and (ii) exonic mutations affecting the protein structure occur more frequently in RE compared to control subjects.

Methods

The institutional review board of the University Giessen, Germany specifically approved this study. Registration number: No 03/11. Written informed consent was obtained from all subjects or their legal guardians according to study protocols approved by the institutional ethical review board of the University Giessen, Germany under the title “genomic variation in patients with idiopathic epilepsy”.

Results

Detection of RBFOX1 and RBFOX3 Microdeletions in RE Patients

In total, 289 RE patients were screened for copy number variations in RBFOX1, RBFOX2, and RBFOX3 using the Infinium OmniExpressExome BeadChip® (Illumina Inc., San Diego, CA). We identified one RE patient among 289 (0.34%) with a hemizygous deletion in both RBFOX1 and RBFOX3. A deletion of 365 kb was found to be located in the genomic region of RBFOX1 affecting the untranslated 5′-terminal exons 3 and 1B (Fig. 1A, S1, exon annotation according to [17]) whereas a smaller deletion of 43 kb affected RBFOX3 by removing exon 3 of the known isoform NM_001082575 along with flanking intronic sequences (Fig.1B, S1). The RBFOX1 gene deletion affects the largest transcript variants 4–5 and 6 (NM_018723, NM_001142333, NM_001142334). No additional mutations were identified in the remaining undeleted RBFOX1 and RBFOX3 exonic sequences.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Overview of RBFOX1 and RBFOX3 affecting variants.

Hg19 genomic localization and overview of known transcript variants. Red bars represent microdeletion size and location for the patient. Red dashes indicate the genomic location of single nucleotide variants. (A) Overview RBFOX1 (B) RBFOX3.

https://doi.org/10.1371/journal.pone.0073323.g001

Detection of Rare Exonic Variants in RBFOX1and RBFOX3

Mutational screening of RBFOX1, RBFOX2, RBFOX3 in 242 RE patients did not reveal any mutation in RBFOX2, while a total of three rare mutations (1.2%) were identified in RBFOX1 and RBFOX3 (Fig. 1A,1B; Table 1). In contrast to the identified RBFOX1 and RBFOX3 microdeletions, all exonic variants were located near the 3′-terminal region of both genes. The identified exonic variants included two RBFOX1 variants in exon 11 (c.690_696delGTATCCAins(GTATCCA)2; p.A233Vfs*74, NM_001142333) and exon 13 (c.893_898delCTGCCG, p.A299_A300del, NM_001142333), as well as a nonsense mutation in exon 13 of the RBFOX3 gene (c.861C>A, p.Y287*, NM_001082575). The p.A299_A300del variant of patient E699 (Table 1, S1) deletes two out of three consecutive alanine residues which are conserved among mammals but not vertebrates (Fig. S2, S3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. RBFOX1and RBFOX3 variants and phenotype of index-patients.

https://doi.org/10.1371/journal.pone.0073323.t001

To our knowledge nonsense mutations in FOX genes have not been described in the literature and are not found in any of the available databanks. They are absent from 6503 individuals whose exomes were deposited in the EVS database. Furthermore, no deleterious mutations have been identified in >450 exomes of our in-house database (various non epilepsy projects; about 80% Caucasian ancestry). In the truncated proteins the C-terminal fragments of RBFOX1 and RBFOX3 are affected which are critical for cassette-exon activation and repression [29] and the nuclear localization of the FOX proteins [21], [30], [31].

Familial Segregation and Comorbidity Analysis

The segregation of RBFOX1 and RBFOX3 variants identified in the RE index-patients were tracked in four families (Fig. 2). Where testing was possible (n = 3), all variants identified were inherited, one maternally and two paternally. Five out of ten variant carriers were affected by RE, one by an encephalopathy with status epilepticus during sleep (ESES) and one by the RE-characteristic CTS EEG-trait only. The transmitting parents were all unaffected at the date of evaluation. Notably, we cannot rule out that the parents might have expressed the RE-characteristic CTS EEG-trait at younger age, considering its age-related expression with maximum manifestation during childhood. Six out of eight family members affected by either RE, CTS or ESES carried one exonic variant of RBFOX1 or RBFOX3 and one RE patient had a deletion in both genes (Fig. 2). The index patient in Family 1 had RE and exon-removing microdeletions in both RBFOX1 and RBFOX3 whereas her sister with the CTS EEG-trait alone carried only the RBFOX1 microdeletion. In Families 1, 2 and 4, all affected family members carried a RBFOX1 deletion or one of the truncating mutations. In family 3, the RBFOX1 p.A299_A300del deletion identified in the RE-index patient was not present in his RE-affected sister. The phenotypic RE features of the four variant carrying index cases did not differ from those RE patients lacking RBFOX1 and RBFOX3 variants (Table 1). The ESES phenotype variant is generally assumed to represent the most severe expression of the RE CTS EEG-trait. A mild to moderate developmental speech delay, that frequently resolves later, is a known feature in many RE patients [32].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. Segregation of RBFOX1 and RBFOX3 affecting variants.

For three mutations for which DNA samples of family members were available, segregation analyses could be performed. The respective RBFOX1 and RBFOX3 truncating mutations co-segregated with a variable phenotype of either seizures or pathologic EEG patterns in most family members. Only a few individuals carried the respective familial mutation but did not present any clinical features, indicating incomplete penetrance of the mutations. However, subclinical phenotypes (e.g. EEG patterns) have not been investigated in these individuals (indicated by question mark). In family 3 the variant (deletion of two consecutive alanine residues at position 299–300 of RBFOX1) did not segregate with the epilepsy phenotype. Abbreviations: n.a = DNA was not available for testing; RE = rolandic epilepsy; CTS = centrotemporal spikes; ESES = encephalopathy with status epilepticus during sleep.

https://doi.org/10.1371/journal.pone.0073323.g002

Discussion

A previous study of 1408 unrelated individuals with idiopathic generalized epilepsy revealed exon-removing RBFOX1 microdeletions in five patients, whereas none was found in 2256 ethnically matched controls [17]. Furthermore, rare copy number abnormalities of RBFOX1 have been reported for patients with neurological diseases like epilepsy, mental retardation and autism [13], [14], [15], [16] strongly indicating that partial RBFOX1 deletions are a recurrent risk factor of neurodevelopmental defects in human. The heterozygous Rbfox1 knockout mouse model shows deregulated splicing which impacts genes involved in synaptic transmission and membrane excitability, leading to an increased susceptibility for seizure events. Notably, homo- and heterozygous Rbfox1 knockouts display normal brain morphology [33]. Consistent with the splicing alterations in mice, RNA interference-mediated 50% knockdown of RBFOX1 transcripts in human neurons changes the alternative splicing pattern and expression of primarily neuronal genes involved in synapse formation and function [19].

Our identification of a RE patient carrying a rare microdeletion affecting the 5′ part of RBFOX1 replicates the association found in idiopathic epilepsy, but now in an entirely different, i.e. focal, epilepsy syndrome. Her sister, who also carried the deletion, exhibited the RE-specific EEG trait but did not suffer from overt seizures. In agreement with our previous IGE study, this indicates that RBFOX1 microdeletions act as a susceptibility factor but not as a highly penetrant variant. The frequency of the RBFOX1 deletion in the RE patient cohort (0.34%) is in the range of that observed for IGE (0.35%; [17]). The RBFOX1 microdeletion presented here is located in the untranslated 5′-terminal region of the gene, like the previously reported microdeletions and translocations [13]–[17]. A female with autism carrying a 5′-terminal microdeletion of RBFOX1 due to a de novo translocation t(15p;16p) displayed a significantly reduced RBFOX1 mRNA expression in lymphocytes [14]. A second FOX microdeletion also detected in patient E103 is affecting the paralogous RBFOX3 gene which also encodes the highly conserved RNA recognition motif (RRM) [20]. Interestingly, the sister of E103, who also carries the RBFOX1 deletion but lacks the RBFOX3 deletion, only expressed CTS but without having epileptic seizures. Due to the lack of further statistical and functional evidence, we can only hypothesize that both deletions affect neuronal splicing in Family 1 synergistically. Microdeletions in RBFOX3 are extremely rare. Only two other exon-removing microdeletions have been noted in the Database of Genomic Variants (DGV, accessed 2/2013) in the general population. Being present in control individuals might indicate a variable expressivity of RBFOX3 microdeletions or a lack of careful assessment of neuropsychiatric phenotypes in these probands. Both microdeletions do not delete exon 3 (NM_001082575) of RBFOX3 as in our RE patient. It has been demonstrated that Rbfox3 regulates alternative splicing and nonsense mediated decay of Rbfox2 mRNA [21]. This complex interplay of Fox family members has been further reported in the Rbfox1 knockout mouse model where the loss of Rbfox1 inhibits an upregulation of Rbfox2 [33]. Mutational screening did not reveal any exonic mutation in RBFOX2, while three rare mutations have been identified (1.2%) in RBFOX1 and RBFOX3 together. The C-terminal fragment is critical for cassette-exon activation and repression [29] in RBFOX1 as well as for nuclear localization for RBFOX1 [30], [31] and RBFOX3 [21]. These mechanisms are likely to be affected by the observed mutations. Furthermore, no frameshift nor nonsense mutations have been found in the exomes of 6503 control subjects reported in the ESV database. All individuals with an epileptic phenotype were carriers of the truncating mutations in family 1, 2 and 4. Only in family 3 the variant (deletion of two consecutive alanine residues at position 299–300 of RBFOX1) did not segregate with the epilepsy phenotype. This deleted sequence is conserved among mammals only, suggesting that most likely only truncating variants may be risk-conferring for RE. In vitro studies demonstrated that the C-terminal domain is critical for cellular localization in both FOX genes and also for targeted splicing in RBFOX1 [21], [30], [31]. As a model for haploinsufficiency in vitro knock down of RBFOX1 in primary human neural stem cells resulted in an altered expression and splicing of several epilepsy candidate genes (FLNA, SLC1A3, DCX, GABRB3, GAD2, KCNQ2, SLC12A5, SV2B, SYN1) [19]. Interestingly, variants and mutations in KCNQ2 were recently found associated with RE [9].

In summary, our results strengthen the association of partial RBFOX1 deletions in neurodevelopmental diseases and extend the RBFOX1-related phenotypic spectrum by RE. The present RBFOX3 mutations highlight this neuronal gene as a plausible novel genetic risk factor of RE, and suggest that besides genomic microdeletions affecting the 5′-terminal exons, truncating mutations of RBFOX1 and RBFOX3 increase the risk of RE.

Supporting Information

Acknowledgments

We thank all participants and their families for participating in this study. The authors would like to thank the NHLBI GO Exome Sequencing Project and its ongoing studies which produced and provided exome variant calls for comparison: the Lung GO Sequencing Project (HL-102923), the WHI Sequencing Project (HL-102924), the Broad GO Sequencing Project (HL-102925), the Seattle GO Sequencing Project (HL-102926) and the Heart GO Sequencing Project (HL-103010).