Germline and somatic mutations in cortical malformations: Molecular defects in Argentinean patients with neuronal migration disorders

Neuronal migration disorders are a clinically and genetically heterogeneous group of malformations of cortical development, frequently responsible for severe disability. Despite the increasing knowledge of the molecular mechanisms underlying this group of diseases, their genetic diagnosis remains unattainable in a high proportion of cases. Here, we present the results of 38 patients with lissencephaly, periventricular heterotopia and subcortical band heterotopia from Argentina. We performed Sanger and Next Generation Sequencing (NGS) of DCX, FLNA and ARX and searched for copy number variations by MLPA in PAFAH1B1, DCX, POMT1, and POMGNT1. Additionally, somatic mosaicism at 5% or higher was investigated by means of targeted high coverage NGS of DCX, ARX, and PAFAH1B1. Our approach had a diagnostic yield of 36%. Pathogenic or likely pathogenic variants were identified in 14 patients, including 10 germline (five novel) and 4 somatic mutations in FLNA, DCX, ARX and PAFAH1B1 genes. This study represents the largest series of patients comprehensively characterized in our population. Our findings reinforce the importance of somatic mutations in the pathophysiology and diagnosis of neuronal migration disorders and contribute to expand their phenotype-genotype correlations.


Introduction
Malformations of cortical development (MCD) are a major cause of intellectual disability and severe epilepsy. Among them, neuronal migration disorders (NMD), resulting from a disruption in the normal movement of neurons from their original birth site to their final location at early brain developmental phases [1] can be considered a sub-group with syndromic defining imaging features including periventricular nodular heterotopia (PNH), subcortical band heterotopia (SBH), also called double cortex syndrome (DC), and lissencephaly [2].
Despite recent increase in our knowledge of NMD genetics, the identification of diseasecausing mutations in individual patients remains a challenge on clinical grounds, even in the Next Generation Sequencing (NGS) era. A high proportion of cases from NMD cohorts persists as genetically undefined [3]. Moreover, somatic mosaicism could be the mechanism underlying the aetiology of a third of these patients [4]. Therefore, their molecular diagnosis should be addressed to look not only for germline, but for somatic variants as well.
We aimed to test the utility, feasibility and diagnostic yield of implementing a comprehensive molecular diagnostic strategy for NMD, by assessing the presence of germline mutations, copy number variants and somatic mutations in six of the main NMD causing genes in a cohort of MCD patients from an academic centre in a developing country.

Patients
As part of our on-going MCD registry (beginning in 2007) patients with NMD were referred to our Neurogenetic Clinic settled in a tertiary academic centre in Argentina. Among them, patients with MRI features characteristic of periventricular nodular heterotopia (PNH), subcortical band heterotopia (SBH) and lissencephaly (LIS) syndromes were selected. Detailed information regarding family history (pedigree analysis), pre-and perinatal events, epilepsy, psychomotor development, cognitive function, associated systemic malformations, and neurological examination was collected.

Neuroimages
Magnetic resonance images (MRI) of 1.5 T were available for all patients and were analyzed by the same specialist in neuroimages for the number and localization of nodular heterotopias in PNH, the extension, localization and thickness of the band in SBH (as previously described in [5]) and gradient of lissencephaly. The gradient comprises the severity of LIS along the anterior-posterior axis of the brain, which could be more severe in the frontal region (anterior greater than posterior, or a >p), equal (anterior similar to posterior, or a = p), or more severe in the parieto-occipital region (posterior greater than anterior, or p >a). Similarly, the presence of other anomalies affecting CNS was reviewed (i.e. agenesis of the corpus callosum). Patients with PNH were further classified according to the radiological findings in typical PNH and atypical PNH. The criteria for typical PNH included bilaterally grey matter lining the lateral ventricles with otherwise normal appearing white and cortical grey matter. Asymmetric heterotopia and associated findings of cerebellar hypoplasia, enlarged cisterna magna and thinning or agenesis of the corpus callosum were also accepted in typical PNH patients. Patients with subcortical heterotopia, disorders of the overlying cortex and nodules localized predominantly along the atria and temporal horns of the lateral ventricles were considered atypical PNH.

Diagnostic approach. Genetic assays
We performed Sanger sequencing of DCX, ARX, and FLNA in patients with SBH, lissencephaly with ambiguous genitalia and/or corpus callosum agenesis and typical PNH, respectively. Patients with atypical PNH, particularly those with bilateral posterior PNH were excluded from the FLNA mutational analysis as, according to previous evidence, they were presumed to be distinct entities (FLNA negative) [6,7]. In addition, we applied multiplex ligation-dependent probe amplification (MLPA) to detect copy number variants (CNV) disrupting PAFAH1B1, DCX and POMT1, POMGNT1 in all individuals with agyria-pachygyria spectrum (lissencephaly and SBH). Those patients with SBH and a mutation involving DCX detected during DCX sequencing were excluded from CNV analysis. Finally, we applied targeted highcoverage next generation sequencing for the detection of somatic mutations in DCX, PAFAH1B1, and ARX when previous assays were negative (Fig 1).
Sanger and NGS of NMD causing genes. All coding regions and exon-intron boundaries of DCX and ARX genes were amplified by polymerase chain reaction (primers and conditions are available upon request) and sequenced through Sanger sequencing by capillary electrophoresis (ABI BigDye terminator).
Depending on the date of inclusion of each patient, the coding sequence of FLNA was interrogated by means of Sanger sequencing, 454-based NGS or Illumina MiSeq-based NGS. In both NGS assays, the first step of FLNA exonic enrichment was based on a long-range PCR (LR-PCR) approach as previously described by us for mitochondrial disorders [8] CNV detection. MLPA. MLPA test was performed with the P061 kit, lot B1 (MRC HOL-LAND), following the procedure specified by the developer (http://www.mrc-holland.com). Case-control samples were processed under the same conditions and in the same experiment. The PCR products were quantified by capillary electrophoresis on an ABI 31xl analyzer. Electropherograms were analyzed using the software Coffalyser V 1.0 (MRC Holland).
Somatic mutations. Targeted high coverage NGS. Exons of the DCX, PAFAH1B1 and ARX genes were amplified by LR-PCR using a Q5 1 High-Fidelity DNA Polymerase (New England BioLabs Inc.). All amplified fragments for the same patient were mixed in the same tube and purified of the PCR reagents. Each pool was prepared for sequencing following the "Nextera XT DNA Sample Preparation Workflow" protocol and sequenced on Illumina MiSeq for high depth sequencing (> 1000X). The resulting FASTQ files were filtered with PRINSEQ (mean quality score > 10) and mapped to the human reference genome (GRCh37.75) with BWA [9]. The resulting SAM was sorted and converted to BAM with PICCARD and the alignment was recalibrated following the Broad Institute recommended best practices [10,11]Variants were called using Haplotype Caller (GATK) [12] and Platypus [13]. Structural variants (big deletion or insertions among the amplified fragments) were called with Delly2 [14]. The resulting VCF files were annotated with structural information with SnpEff [15] to detect variations with moderate or high impact. Allelic frequency was calculated considering the number of individual reads evidencing the mutations in relation to the total number of sequences covering that region. All pathogenic variants detected by NGS were validated. Variants for which the alternate-allele read frequency was 40% or lower were evaluated for mosaicism by subcloning. Original DNA was re-amplified by means of a polymerase chain reaction (PCR) assay, subcloned into a linearized vector (pGEM(R)-T Easy Vector System II PROMEGA) and transformed into JM109 Competent Cells (Promega). Multiple individual transformants were reisolated, and Sanger sequencing was performed to confirm the presence or absence of the predicted variant and to quantify the degree of mosaicism.
Reference sequences used:

Statements
This study was approved by our Institutional Ethics Committee (CODEI, Buenos Aires, Argentina). All patients and parents provided written informed consent for genetic analyses and use of their anonymized data. All experiments and methods were carried out in accordance with the relevant guidelines and regulations of CODEI. All clinical investigations have been conducted according to the principles expressed in the Declaration of Helsinki.

Results
From a cohort of 70 MCD patients, we selected 38 patients with NMD. The most frequent type of malformation was PNH (N = 19), followed by SBH (N = 12) and lissencephaly (N = 7) (S1 Table). A conclusive genetic diagnosis was achieved in 14 patients. All variants reached ACMG [16] criteria to be classified as pathogenic or likely pathogenic (for allele frequency and in silico analysis see S2 and S3 Tables). Thus, the diagnostic yield of our comprehensive approach was 36%. Clinical and genetic findings are summarized in Tables 1, 2 and 3. Further details and illustrative cases are presented below.

SBH
Seven patients with SBH were female (58%), thus representing a similar female/male ratio compared to other series [5]. All of them were sporadic cases, except MDC1045, a woman with SBH with a son affected by lissencephaly (SBH/LIS x-linked pedigree). The heterotopic band was bilateral and diffuse in all cases. A severe form with band thickness grade 3-4 (grade 3 = 8 to12 mm; grade 4 > 12 mm) was present in 50% of the patients and a milder form with band thickness grade 1-2 (grade 1 < 4 mm; grade 2 = 4 to 7 mm) in the other 50%.
We identified a pathogenic mutation in 7/12 patients. Remarkably, 4 mutations (57%) were somatic, including a PAFAH1B1 mutation in a girl with a p>a gradient (see clinical reports section bellow, MDC1070). The phenotypes in these 4 cases with a mosaic state were milder than expected for a germinal mutation: two males with DCX mutations (MDC1034 and MDC1092) and a woman with a PAFAH1B1 mutation (MDC1070), presenting with HBS instead of lissencephaly; and a woman with a somatic DCX mutation with a very thin predominantly anterior band of less than 4 mm without intellectual disability.

PNH
Nineteen patients with PNH were included. Heterotopic nodules were bilateral (n = 18), or unilateral (n = 1). We found a FLNA mutation in 4 individuals, all of them with at least one clinical or radiological feature supportive of a FLNA mutation: MDC1002, typical PNH and vermis hypoplasia; MDC1005, typical PNH and cardiac (ventricular septal defect); MDC 1019, typical PNH and family history of recurrent spontaneous abortions; MDC 1020, male foetus stillbirth.

Lissencephaly
We studied 7 patients with lissencephaly by means of MLPA and /or gene sequencing. One of them had frontal pachygyria (a >p gradient), and two had frontal pachygiria with posterior agyria (p >a gradient) A de novo PAFAH1B1 deletion of exon 1 and was found in MDC1075, one of the patients with a p >a gradient. MLPA and PAFAH1B1 sequencing was normal in the rest. The other four patients had lissencephaly in association with Corpus Callosum agenesia and abnormal genitalia (XLAG). Regarding this last subgroup, two mutations in ARX were found in two cases.

Genetic findings
Germline mutations. We successfully sequenced coding regions of DCX, FLNA, PAFAH1B1 or ARX in 24 patients and found 8 different germline mutations in 9 patients: 3 in FLNA, 3 in DCX and 2 in ARX. Among them, 3 were nonsense (2 in FLNA, 1 in ARX), 4 were missense (1 in FLNA, 2 in DCX, 1 in ARX), whereas one was a 1 bp insertion that results in a frameshift of DCX translation that predicts the premature introduction of a stop signal. All variants detected by NGS were successfully validated by Sanger. All of the identified variants can be classified as pathogenic or likely pathogenic. Three of them were previously reported in individuals with a similar phenotype, while the other five were considered to be novel ( Table 2). From this last group, two were null alleles, meaning there was a point mutation that translated into a gain of a stop codon (ARX:p.F476X and DCX:p.P288SfsX22), and three were predicted to result in a nonconservative missense substitution of an amino acid residue (FLNA:p.G1387S, DCX:p.R102P and ARX: pR345P), being absent in population genomic databases. These three variants represent drastic changes in the amino acid sequence, from a small flexible amino acid as glycine to a polar one capable of doing sulfured bonds as cysteine in the FLNA gene, or from the positive charged arginine to the rigid proline in both DCX and ARX genes. The positions involved in these changes are conserved in the phylogeny and are predicted to be pathogenic by several prediction tools (S3 Table). Addiionally, ARX: pR345P is located in the ARX homeodomain a DNA binding domain strongly associated with XLAG phenotypes. Copy number variants. Patients with lissencephaly or SBH, but negative for DCX mutations, were subsequently studied for copy number variants. Successful MLPA results were obtained for 9 out of 9 SBH/lissencephaly samples. A single copy loss of PAFAH1B1 exon 1 and 2 was detected in one sample with lissencephaly. The MLPA probe sets around this region covered all exons of PAFAH1B1 and the close genes METTL16 and H1C1 (17p13.3 probes). The upstream breakpoint was estimated to be localized between H1C1 and METTL16 genes and the downstream breakpoint to be between exon 2 and 3 of PAFAH1B1. MLPA performed in the parents showed two copies of exon 1 and 2, suggesting a de novo deletion.
Somatic mutations. We applied targeted high-coverage NGS to 12 patients. We obtained a mean coverage of about 4000x. Pathogenic or likely pathogenic variants were found in 4 subjects, including 2 variants (in PAFAH1B1 and DCX genes) with an alternate-allele read frequency lower than 15% (12-15%) and 2 DCX variants with and alternate-allele read frequency higher than 40%. These latter two were identified in two males with SBH (MDC1034 and MDC1092). Since DCX is localized on the X chromosome, and both individuals had a normal karyotype (46 XY), we concluded that this pseudo heterozygosity represented a mosaic state. Further validation with Sanger, confirmed these findings. Noteworthy, one of them consisted of a novel intraexonic 257 bp deletion and 2 pb insertion of DCX (S1 Fig) maintaining the coding frame but severely truncating the protein sequence. Since the alternate-allele read frequency is near 50%, these postzygotic mutations must have probably occurred in early embryonic stages of subjects MDC1034 and MDC1092.
On the other hand, Sanger sequencing would not have been sensitive for those two variants having alternate-allele read frequencies of less than 15%: DCX:p.A251V, previously reported as pathogenic and PAFAH1B1: p.A210P, a novel mutation (Table 3). Thus, we validated them by means of subcloning assays, followed by sequencing of individual colonies (S3 Fig). The latter novel mutation (PAFAH1B1: p.A210P) comprises the substitution of a highly conserved alanine to a proline, a rigider amino acid. This change is predicted to be damaging by several pathogenicity predictors. In addition, we studied the stability of the protein using a modelled peptid after the 1VYH structure in PDB, and evaluated the free energy difference between the wild and mutated protein using FoldX, resulting in a highly destabilizing change (> 4,06 Kcal/ mol).
All phenotypes caused by mosaic mutations were milder than expected for germline mutations. MDC1093 had a diffuse SBH without cognitive delay and MDC1070 was a female with a mosaic mutation in PAFAH1B1 that led to SBH instead of lissencephaly. Interestingly, the SBH had a posterior predominance, as seen in PAFAH1B1 associated lissencephaly (see clinical report below).

Phenotype-genotype correlations. Clinical reports
Here, we present a further description of some illustrative cases.
Periventricular nodular heterotopia-FLNA missense mutation with wide phenotypic spectrum. MDC1019 is a 19-year-old woman with normal early development who presented focal seizures since she was 6 years old. The MRI revealed bilateral and diffuse nodular heterotopía. Although her sister, MDC1020, was totally asymptomatic regarding epilepsy and showed a normal neurological development, she reported the antecedent of one male foetus stillbirth (Fig 2-A1). MRI performed to MDC1020 revealed the presence of a single heterotopic nodule adjacent to the lateral ventricle (Fig 2-A2). Although MRI could not be performed to The mutation is present at the X chromosome of the patient (male) but absent in the mother. Please consider that the Sanger sequencing was performed on the coding strand (reverse of reference).
A novel missense mutation (c.4159G>A; p.G1387S) in the FLNA gene was found. Different tools predicted deleterious impact for this variant (SIFT-Polyphen-Mutation Taster) (Fig 2-A3). This mutation was later confirmed in her sister PC.
Agyria-pachygyria spectrum: Posterior SBH in a girl mosaic for a PAFAH1B1 mutation. MDC1070 was a female born at term. Her mother was asymptomatic, as well as two of her siblings. Conversely, two other brothers were dead before the first 2 months of life (Fig 2-B1). MDC1070 presented with development delay and during childhood, she developed drugresistant epilepsy.
We decided to perform DCX sequencing first since 88% of females with SBH carry a DCX mutation. However, we were not able to find any DCX mutation by Sanger sequencing. DCX deletions and duplications were also discarded by MLPA. Subsequently, we did high coverage NGS searching for somatic mutations. Considering the posterior gradient of the cortical malformation (Fig 2-B2) we included also PAFAH1B1 in the analysis. Again, no mutation in DCX could be identified but PAFAH1B1 sequencing showed a mosaic missense mutation in PAFAH1B1 with an alternate allele frequency of 14.9% (Fig 2-B3) SBH in males-Somatic DCX mutation. MDC1034 is a male with no relevant family history that presented drug resistant epilepsy and developmental delay. His sister, a dizygotic twin, was asymptomatic (Fig 2-C1). No other family members were affected. MRI, in this case, showed a diffuse band of subcortical heterotopia (thickness> 12 mm) with a a = p gradient (Fig 2-C2).
DCX sequencing by NGS (genomic DNA extracted from peripheral blood leukocytes) revealed the mutation c.176G> A; p.R59H with an alternate-allele read frequency of 48.7% (Fig 2-C3). Given that DCX gene is located on the X chromosome and that our patient had a normal karyotype (46XY), this results can be considered a mosaic state in pseudo heterozygosity. These findings were subsequently confirmed by Sanger sequencing and subcloning followed by sequencing of individual colonies.
Finally, to explore the existence of mosaicism in other tissues, we studied genomic DNA extracted from epithelial cells obtained by buccal smear. Again, the c.176G> A substitution was revealed [15]. Mutational analysis in his mother showed a wild-type genotype.

Discussion
We presented clinical, neuroimaging and genetic data of a large cohort of NMD patients followed in an academic neurogenetic tertiary center in Argentina. Our results expand the mutation spectrum causing this group of disorders and give new insights about the evolving role of somatic mutations in neurological diseases, especially for diagnostic purposes. A phenotypedriven comprehensive strategy looking for germline and mosaic sequence variants could identify a disease-causing defect in 36% of our patients. These results highlight the genetic heterogeneity of NMD, along with their complex phenotypes, that often lead to difficult diagnostic scenarios.
Most of the previous reports were retrospective, included selected homogeneous cohorts and, many times, they were restricted to the analysis of a single gene [5,17]. Among them, the diagnostic yield was variable and dependent on selected phenotypes [18]. As an example, the chance of finding a mutation in FLNA in PNH patients could be as low as 4% within subjects with a non-classical phenotype, whereas it approximates 100% when only familial cases with X-linked PNH are considered [19]. On the contrary, we aimed to perform a prospective and comprehensive mutational analysis including typical and atypical phenotypes in our cohort. It is plausible that this heterogeneity resulted in a lower number of positive cases, but we believe that it comes closer to a real-life scenario. To our knowledge, this is the largest NMD series in Latin America reporting a comprehensive mutation spectrum.
In addition, our approach investigated not only germline variants but also mosaicism. Mosaic mutations might play a role in the phenotypic heterogeneity observed in NMD. Recent works demonstrated that early embryonic cell-doubling events contribute asymmetrically to adult tissues [20] because of stochastic allocation of embryonic cells in the inner cell mass. Thus, it is plausible that somatic mutations involving NMD genes may lead to diverse degrees of adult tissue affectation (and phenotypic severity) even when they occur during early embryonic stages. Germline DCX mutations in males and PAFAH1B1 in either sex result in lissencephaly. However, when mutations are mosaic a milder severity could be expected, such as the phenotypes observed in patients MDC1070 and MDC1034 [21].
Also, somatic mosaicism could be behind the missing genetic cause of several NMD cases. De novo mutations are frequently found in this group of malformations [5]. Furthermore, early embryonic mutation rate seems to be similar to, or even slightly higher than, germline mutational rates [20]. Thus, we hypothesized that postzygotic mosaic mutations, could be detected more often in sporadic NMD, if they were properly investigated. Supporting this, a substantial fraction (29%) of the 14 mutations found in our study were postzygotic mosaic mutations. Jamuar et al performed the most extensive investigation of somatic mutations in brain cortical malformations in 2014. [4] In their study, they investigated 158 persons with brain malformations and found a causal mutation in 27, 30% of which were somatic. Thus, our findings are in concordance with those obtained by Jamuar et al. Furthermore, in their series as well as in ours, mosaic mutations were mainly in individuals with Double-Cortex syndrome (SBH), Also, Zillhart et al. [22] have recently communicated that parental germline mosaicism accounts for about 15% of recurrent forms of MCD.
Overall, this evidence suggest that somatic mosaicism is not uncommon in MCD, specially in SBH, and call for a reappraisal of the most appropriate diagnostic strategies in clinical practice, In this aspect, our results, in addition to those of Jamuar et al. support the utility of targeted high coverage NGS for this purpose.
Considering our results, we propose a candidate gene targeted strategy only for patients presenting with a classical phenotype (eg. classical PNH in females), which in our opinion, should be implemented through NGS. This technique is more cost-effective for large genes such as FLNA [23] and it is more sensitive to detect mosaic mutations [24] In atypical cases, a more comprehensive approach that includes the scrutiny of multiple genes by high-coverage NGS based panels and structural variants detection by means of MLPA or microarray would be warranted [25]. Although, we did not perform whole exome (WES) or genome sequencing (WGS) in our work, they would certainly have been the next step in negative cases [26].
Delineating the several molecular mechanisms that disrupt neuronal migration might facilitate therapeutics discoveries while providing an improved understanding of normal brain development. However, since most of the neuronal migration disorders remain still genetically unexplained, we must remember that the molecular diagnosis of these malformations is a real challenge even in this current and promising genomic era.
Supporting information S1