Diverse genetic causes of polymicrogyria with epilepsy

1 |. INTRODUCTION

Polymicrogyria is a developmental brain malformation characterized by the radiographic appearance of excessive and small folds of the cerebral cortex and pathological evidence of abnormal neuronal organization. Patients present with a range of clinical findings, including epilepsy, focal weakness, and oromotor apraxia. Polymicrogyria may affect the entire cerebral cortex or distinct brain regions, either bilaterally or unilaterally.^1,2 Although polymicrogyria can be due to nongenetic factors, including in utero ischemia and congenital cytomegaloviral infections,^3–5 some cases have been suspected to have a genetic basis, as polymicrogyria has been observed in families^6–8 and because it often presents with bilateral, regionally specified abnormalities, suggesting the involvement of genes expressed in specific distributions. Among the earliest polymicrogyria genes identified were GPR56 in individuals with bilateral frontoparietal polymicrogyria,⁹ TUBA1A with perisylvian polymicrogyria,¹⁰ and LAMC3 with occipital polymicrogyria.¹¹ The early identification of polymicrogyria as part of genetic syndromes associated with copy number variations (CNVs), such as the 22q11 deletion syndrome,¹² was followed by studies identifying a number of recurrent CNVs associated with polymicrogyria.¹³ More recently, polymicrogyria has been noted in patients with overgrowth syndromes that include megalencephaly and dysmorphic features.^14,15

The genes were discovered through the study of rare large multiplex families, small cohorts of patients with polymicrogyria in the setting of multisystem syndromes, case reports of whole exome sequencing (WES) findings, and targeted sequencing of candidate genes in polymicrogyria cohorts. To our knowledge, there have been no prior unbiased genome-wide sequencing studies of large polymicrogyria cohorts to evaluate the contribution of known and novel genes.

In this study, we aimed to perform trio-based WES in a multicenter, international cohort of patients with polymicrogyria and epilepsy recruited and extensively phenotyped through the Epilepsy Phenome/Genome Project (EPGP)^2,16 and sequenced through the Epi4K Consortium.¹⁷

2 |. MATERIALS AND METHODS

Patients with epilepsy and polymicrogyria on brain magnetic resonance imaging (MRI) were enrolled through the EPGP, and blood samples were collected from probands and their parents¹⁶; the research presented was approved by the institutional review boards at all 26 participating enrollment sites. All MRIs were reviewed by the EPGP MRI Core for eligibility. The cohort has been previously described²; additional phenotypic data were reviewed for this report. Blood samples were collected from patients and parents; DNA was obtained from the Coriell Institute DNA Bank (Table S1). Written informed consent was provided by patients, or their parents or legal guardians if the patient was underage.

We performed WES on 86 trios with either SeqCap EZ Exome v3 or SureSelect Human All Exon–65MB exome capture kits at the Institute for Genomic Medicine (IGM), Columbia University. De novo variants were called within trios using GATK multisample calling in the protein-coding regions (CCDS, release 14, GRCh37.p13) as described previously.^18,19 De novo variants were either Sanger validated in the probands and confirmed to be absent in parents in primary DNA, or predicted with high confidence using a validated machine-learning model using variant-level, individual-level, and genomic features to identify de novo variants (Table S2).¹⁹ Controls (n = 2711) used for single nucleotide variant (SNV) and indel calling and gene-level collapsing included individuals sequenced at the IGM for other studies.

In addition to de novo variants, we compiled rare inherited, hemizygous, homozygous, and compound heterozygous variants. We first listed all rare (minor allele frequency [MAF] < 1% among internal controls or any population in Exome Variant Server, Exome Aggregate Consortium [release 0.3], and gnomAD browser [v2.1.1]), putatively protein-a ltering (missense [PolyPhen-2 probably/possibly damaging], nonsense, or indels) variants residing in protein-coding regions. Genotypes were excluded if they had a quality score less than 30 or genotype quality score less than 20 in the proband. We required a minimum of 10-fold coverage at a variant site to call a homozygous genotype, and the same coverage in addition to at least five reads supporting the alternate allele to call a heterozygous variant. Inherited heterozygous variants were required to have a variant allele frequency (VAF) between 40% and 60%. For rare inherited variants, we included only those absent in the aforementioned control databases; newly hemizygous (present in proband, absent in mother), homozygous, and compound heterozygous variants were required to have a homozygote frequency of zero in all control databases listed above. Candidate variants are listed in Tables S3 and S4.

We evaluated for de novo or inherited CNVs using the CoNIFER pipeline as previously described.^20,21

Variants were assessed for pathogenicity using the principles outlined by the American College of Medical Genetics.²²

To assess whether inherited rare variation either augments the risk conferred by de novo alleles or confers risk independently, we performed a gene-level collapsing analysis comparing, for each protein-coding gene, the fraction of cases versus the fraction of controls with a rare, functional variant. We performed three separate case–control analyses comparing (1) ultrarare (absent from control databases), functional (missense and loss-of-function) variants; (2) ultrarare, loss-of-function variants only; and (3) rare (MAF < .1%) biallelic variants. We used a previously published method with a modification to allow for mosaic variants in the ultrarare comparisons.¹⁹

3 |. RESULTS

Among 86 individuals (39 male, 47 female) with polymicrogyria and epilepsy, 10 probands (12%) had a pathogenic or likely pathogenic de novo variant identified by individual case analysis (Table 1). Seven patients harbored heterozygous de novo variants in PIK3R2: five with the recurrent pathogenic PIK3R2 variant Gly373Arg, one with PIK3R2 Lys376Glu (pathogenic), and one with Asp557Tyr (likely pathogenic). Three of the seven had VAFs between 13% and 16%, which deviates from the 50% expected for a heterozygous variant (binomial probability distribution), a finding consistent with mosaicism due to the mutational event occurring during early postzygotic development.

All seven PIK3R2 patients had bilateral polymicrogyria, most prominent in the perisylvian region (Figure 1, Table 1). Six patients presented with focal epilepsy, with onset between 15 months and 9 years, and the other one presented with infantile spasms at 3 months of age. There was a wide developmental range among them; some had normal milestones or mild delays, whereas others were severely impaired. Given the association between this gene and overgrowth syndromes involving brain malformations (hemimegalencephaly, megalencephaly–capillary malformation syndrome, and megalencephaly–polydactyly–polymicrogyria–hydrocephalus),^15,23 we specifically evaluated for other features of these syndromes. All patients displayed macrocephaly with occipitofrontal circumference measuring from +3.1 to +7.1 standard deviation (SD) greater than the mean for age for those with available measurements (Table 1). None had capillary malformations of the skin, digital abnormalities, or other syndromic features. Interestingly, there was no correlation of VAF with the extent of brain malformation or other features, meaning that the patients with mosaic PIK3R2 pathogenic variants were not more mildly affected.

We observed de novo variants in one case each of CCND2, associated with bilateral perisylvian polymicrogyria (BPP) and enlarged ventricles on MRI,²⁴ and DYNC1H1, associated with BPP.²⁵ We identified a single case with biallelic variants in the gene WDR62, previously associated with microcephaly and a range of brain malformations²⁶; our case also presented with microcephaly. The phenotypic features of these three individuals are consistent with the reported literature for the corresponding genes (Table 1).

Among 86 trios, we identified 99 de novo variants (average = 1.15 per trio; Table S2). Three genes were found to harbor de novo variants in multiple probands: PIK3R2 (n = 7, described above), SCN2A (n = 2), and AASDH (n = 2, variants predicted to be benign). Only PIK3R2 had more than expected to occur by chance given the size of the cohort, gene size, and mutability, and correcting for the ~18K protein- coding genes evaluated in this study (FitDNM,²⁷ p = 3.6 × 10⁻²⁴). Beyond the enrichment of rare variants in PIK3R2 in polymicrogyria cases in the ultrarare functional collapsing analysis (Fisher exact test p = 9 × 10⁻¹⁰), no other gene was significantly enriched in cases versus controls across any of the models (Figure 2, Table S5).

To estimate aggregate genetic signal attributed to de novo variation, we performed “hot-zone” analysis as described previously,^18,19 to compare the fraction of cases versus controls harboring a highly likely damaging de novo variant in genes intolerant to functional variation, as estimated using the Residual Variation Intolerance Score method.²⁸ We observed that 26% of polymicrogyria probands had a highly likely damaging de novo variant, compared to 9.6% of previously published control trios (p = .04).²⁹ The increased de novo variant burden among polymicrogyria cases is largely attributed to the known or likely pathogenic de novo variants above, as the percentage dropped to 15% (p = .47) following the removal of the 10 genetically explained individuals.

Despite the majority of aggregate de novo variant burden being attributed to variants in PIK3R2, CCND2, and DYNC1H1, a small number of additional likely de novo candidates were also identified (Table S2). We identified two de novo likely pathogenic missense variants in SCN2A, both of whom had neonatal epilepsy presenting at 1 month of age; one had focal seizures, and the other had tonic seizures and burst suppression on electroencephalogram, consistent with Ohtahara syndrome. We identified a de novo variant in CACNA1C, a gene reported in the context of Timothy syndrome, features of which were not reported in our case. Similarly, we identified a de novo variant in GRIA3, a gene associated with intellectual disability, in a normocephalic child with BPP, normal early developmental milestones, and school difficulty recognized only at age 6 years. The CACNA1C and GRIA3 variants are predicted to be likely pathogenic in their known disease contexts, but because these genes are not associated with polymicrogyria and our cases did not display the recognized gene- associated phenotypes, we classified the variants as variants of uncertain significance.

In addition to SNVs, we identified a rare de novo 597-kb deletion on chromosome 15q25.2 encompassing nine genes in one individual with polymicrogyria, focal seizures with onset at 18 years, and normal development (Figure 3). This locus has been reported as a rare susceptibility allele for neurodevelopmental disorders,^30,31 although no reported individuals with a 15q25.2 deletion have polymicrogyria. We classified the 15q25.2 deletion as a variant of unknown significance (VUS).

Rare inherited, hemizygous, homozygous, and compound heterozygous variants identified across the cohort are presented in Tables S3 and S4.

4 |. DISCUSSION

We report the genetic findings of a large cohort of patients with polymicrogyria and epilepsy analyzed with a trio-based, genome-wide approach and analyze the phenotypes of the patients with notable genetic findings. Ten of the 86 individuals studied had variants in the known polymicrogyria-related genes PIK3R2, CCND2, DYNC1H1, and WDR62, comprising 12% of our cohort. In cohort-wide analyses, including gene-level assessments of genome-wide enrichment of de novo variants and overall rare variant enrichment, only PIK3R2, a gene previously implicated in polymicrogyria,^15,32–34 was significantly associated with polymicrogyria. The seven individuals with rare germline or early mosaic de novo variants in PIK3R2 had bilateral, mostly perisylvian, polymicrogyria and macrocephaly, consistent with the phenotype associated with genes in the PI3K- AKT-mTOR pathway.^{15,32– 34} Note that the onset and epilepsy phenotype varied, with one patient presenting with infantile spasms at 3 months and the other six with focal epilepsy between 15 months and 9 years. Earlier seizure onset did not correlate with the degree of developmental delay or intellectual disability.

Although the genes CCND2, DYNC1H1, and WDR62 did not reach statistical significance for an association with polymicrogyria in our cohort-wide analysis, we considered the variants in these three genes pathogenic or likely pathogenic given in silico predictions, phenotype, and the reported association with polymicrogyria in the literature. Specifically, the presence of macrocephaly in the patient with the CCND2 variant and microcephaly in the patient with the WDR62 biallelic variants are consistent with these genes’ described phenotypes.

An additional five individuals with polymicrogyria harbored variants in genes or loci that are currently associated with epilepsy or other neurodevelopmental conditions but not definitively with brain malformations. Two had de novo likely pathogenic variants in SCN2A, the gene encoding the voltage-gated sodium channel alpha 2 subunit, which is associated with epilepsy, intellectual disability, and autism spectrum disorder.^35,36 The two variants in SCN2A were predicted to be likely pathogenic, consistent with the patients’ neonatal epilepsy presentations, as SCN2A is a known cause of Ohtahara syndrome. Although SCN2A’s connection to polymicrogyria is emerging,^35,37 we could not conclude definitively that the SCN2A variants contributed to the brain malformation in these patients. One case had a de novo variant in CACNA1C, and another case had a de novo variant in GRIA3, each predicted to be likely pathogenic. As these genes are associated with clinical phenotypes unrelated to polymicrogyria, we considered the variants to be VUSs. Our findings add to emerging reports implicating ion channel genes in the pathogenesis of polymicrogyria (e.g., SCN3A and GRIN2D^38–41). Finally, we identified one patient with polymicrogyria and epilepsy with a deletion at 15q25.2. These genes and this CNV locus do not have a definitive prior association in the literature with polymicrogyria, so we consider their variants, even when compelling in other phenotypic contexts, as VUSs with respect to polymicrogyria.

Although polymicrogyria can occur in isolation or in conjunction with other features, we note that our study required that individuals have polymicrogyria and epilepsy as well as both parents available for DNA samples. It may be informative to compare our results with those in cohorts with different ascertainment strategies, for example, those studies of polymicrogyria not requiring that patients have epilepsy or including individuals with more syndromic presentations.

In summary, exome sequencing in a large cohort with polymicrogyria and epilepsy provided a likely causative explanation for 12% of the patients and a possible genetic cause for an additional 6%. Our findings support a role for genetic testing in the clinical setting for patients with polymicrogyria, particularly those with epilepsy. Specifically, polymicrogyria with macrocephaly suggests that a germline or mosaic variant in PIK3R2, or a related PI3K-AKT-mTOR pathway gene, may be involved. As treatment trials emerge for neurodevelopmental disorders related to this pathway, a genetic diagnosis may provide the opportunity for patients with polymicrogyria to participate in clinical trials. Other candidate polymicrogyria genes will require sequencing of additional cohorts to validate their role in polymicrogyria.

Supplementary Material