Follow-up screening of GABRG2 was done by Sanger sequencing in the 43 probands of the Italian cohort as previously described.13 Screening in the cohort of 64 probands from French families was performed using a targeted epilepsy gene panel (SeqCapEZ custom design, Roche NimbleGen) including GABRG2, GABRA1, STX1B, PRRT2, CHRNA4, CHRNA2, CHRNB2, LGI1, DEPDC5, KCNT1, and EFHC1. Quality of raw sequencing data was assessed with the FASTQC tool. Reads were trimmed with Trimmomatic. BWA was used to align reads on human genome (hg19). Finally, GATK best practices were applied to call variants. Variants with low quality (<20) were excluded. Variants with a read depth >30 and allele read frequency >0.30 were filtered. We selected heterozygous nonsynonymous variants with frequency <1% in the exome variant server, affecting the protein-coding sequence (nonsense, missense, frameshift, or essential splice-site) and predicted to be damaging by at least 1 of 2 in silico prediction tools (Sift, http://sift.jcvi.org/; PolyPhen-2, http://genetics.bwh.harvard.edu/pph2/). We followed the guidelines for assigning causality of sequence variants in human disease as outlined.14 Coverage of GABRG2 was satisfactory (>250×) except for exon 6, which was
