Frequent SLC35A2 brain mosaicism in mild malformation of cortical development with oligodendroglial hyperplasia in epilepsy (MOGHE)

Focal malformations of cortical development (MCD) are linked to somatic brain mutations occurring during neurodevelopment. Mild malformation of cortical development with oligodendroglial hyperplasia in epilepsy (MOGHE) is a newly recognized clinico-pathological entity associated with pediatric drug-resistant focal epilepsy, and amenable to neurosurgical treatment. MOGHE is histopathologically characterized by clusters of increased oligodendroglial cell densities, patchy zones of hypomyelination, and heterotopic neurons in the white matter. The molecular etiology of MOGHE remained unknown so far. We hypothesized a contribution of mosaic brain variants and performed deep targeted gene sequencing on 20 surgical MOGHE brain samples from a single-center cohort of pediatric patients. We identified somatic pathogenic SLC35A2 variants in 9/20 (45%) patients with mosaic rates ranging from 7 to 52%. SLC35A2 encodes a UDP-galactose transporter, previously implicated in other malformations of cortical development (MCD) and a rare type of congenital disorder of glycosylation. To further clarify the histological features of SLC35A2-brain tissues, we then collected 17 samples with pathogenic SLC35A2 variants from a multicenter cohort of MCD cases. Histopathological reassessment including anti-Olig2 staining confirmed a MOGHE diagnosis in all cases. Analysis by droplet digital PCR of pools of microdissected cells from one MOGHE tissue revealed a variant enrichment in clustered oligodendroglial cells and heterotopic neurons. Through an international consortium, we assembled an unprecedented series of 26 SLC35A2-MOGHE cases providing evidence that mosaic SLC35A2 variants, likely occurred in a neuroglial progenitor cell during brain development, are a genetic marker for MOGHE. Electronic supplementary material The online version of this article (10.1186/s40478-020-01085-3) contains supplementary material, which is available to authorized users.


Introduction
Malformations of cortical development (MCD) represent an important cause of pediatric epilepsy resistant to antiseizure medication [13,37]. When focal, surgical resection of the malformed brain area currently represents the main treatment option for seizure control [3,19].
Focal MCD comprise distinct histopathological entities, including Focal Cortical Dysplasia (FCD) type 1, 2 and 3 [4], mild MCD (mMCD), and a newly recognized clinicopathological entity termed mild malformation of cortical development with oligodendroglial hyperplasia in epilepsy (MOGHE). MOGHE is histopathologically characterized by clusters of increased oligodendroglial cell density in the white matter and deep cortical layers, an augmented oligodendroglial proliferation activity, patchy areas with hypomyelination and an excess of heterotopic neurons in the white matter, which occurs most often in the frontal lobe of children with early seizure onset [35]. Magnetic resonance imaging (MRI) features of MOGHE are age-related and comprise increased laminar T2 and fluid attenuated inversion recovery (FLAIR) signal at the corticomedullary junction in young children (subtype I) and change to reduced corticomedullary differentiation due to increased signal of the adjacent white matter in older patients (subtype II) [16].
In recent years, somatic mosaicism has been shown to be a significant cause of MCD [23]. Brain mosaic variants in genes belonging to the mechanistic Target of Rapamycin (mTOR) pathway have been discovered as underlying cause of FCD type 2 [2,8,21,26,39]. Recently, mosaic variants in the SLC35A2 gene, encoding the major Golgilocalized UDP-galactose transporter required for protein and sphingolipid glycosylation, have been identified in various forms of non-lesional focal epilepsies, as well as FCD type 1 and mMCD [2,25,38,39,41]. Moreover, germline de novo variants of SLC35A2 cause a rare X-linked dominant form of congenital disorder of glycosylation (CDG) type IIm (MIM #300896), primarily presenting with epileptic encephalopathy, seizures, severe psychomotor developmental delay and delayed myelination [30,40]. Loss of SLC35A2 protein function abolishes transport of UDP-galactose from the cytosol into the Golgi apparatus, resulting in the synthesis of truncated glycans lacking galactose residues [30,38]; yet, how this relates to the clinical phenotype is still unknown.
In this study, we hypothesized that somatic brain variants in SLC35A2 may contribute to the pathogenesis of MOGHE and used surgically resected brain tissues obtained from the epileptogenic zone to test this hypothesis. We found that mosaic SLC35A2 variants are important factors in the etiology of MOGHE, and likely occurred in neuroglial progenitors during brain development.

Cohort recruitment
We enrolled a cohort of 20 pediatric patients with refractory focal epilepsy subjected to surgery between 2012 and 2019 in the center for pediatric neurology, neurorehabilitation and epileptology, Schoen Klinik, Vogtareuth, Germany, and with a histopathological diagnosis of MOGHE. Eight out of twenty cases were previously reported [16]. All patients underwent detailed clinical examination and review of the medical files, including high-resolution magnetic resonance imaging (MRI), long-term video-EEG-monitoring (EEG) and fluoro-deoxy-glucose (FDG) positron emission tomography (PET) investigations when performed. Written informed consent was obtained from all participants or their parents on their behalf. The study was approved by the University of Erlangen ethical review board (EEBB 160_12B).
We next assembled an international cohort of 43 additional mMCD, FCD, unclassified FCD or inconclusive cases operated between 2013 and 2019 from three epilepsy surgery centers: the neurosurgery department from the Rothschild Foundation Hospital (n = 20, Paris, France), Severance Children's Hospital (n = 17, Seoul, Republic of Korea) and Amsterdam UMC (n = 6, Amsterdam, the Netherlands). These studies received approval by the ethical committee CPP Ile de France II (No. ID-RCB/EUDRACT-2015-A00671-48), the Severance Hospital and KAIST Institutional Review Board and Committee on Human Research, and Amsterdam UMC (W13_85;W15_262).

Neuropathological evaluation
The histopathology diagnosis of MOGHE was given according to the previously published catalogue of microscopic measures [35], i.e. clusters of oligodendroglial cells with high density at the grey-white matter junction using H&E staining and confirmed by additional immunohistochemistry for Olig2 and Mib1. Reassessment of the 43 multicentric mMCD cases was performed by an expert neuropathologist (IB) blind to the genetics results and the initial neuropathological diagnosis. Digitally scanned slides of H&E staining, NeuN and Olig2 immunostainings in all selected cases were made accessible for online microscopic evaluation (Leica Aperio, Germany).

Gene panel sequencing
Genomic DNA from formalin-fixed paraffin-embedded (FFPE) brain tissues was extracted with QIAamp DNA Micro Kit (Qiagen) or Maxwell RSC DNA FFPE kit (Promega), following standard protocols. We designed a hybrid capture sequencing (Twist Bioscience), targeting coding exons and exon-flanking junctions (10 bp) of 59 genes including genes already involved in focal MCD, mTOR pathway-related genes and SLC35A2 (the full list of genes included in the panel is reported in Supplemental data). Libraries were prepared according to a standard protocol for FFPE DNA samples and sequenced on an Illumina NextSeq 500 sequencer (2 × 75 bp) at the iGenSeq sequencing facility of ICM (Paris, France). Bioinformatic analysis including quality control, processing and calling of the variants was conducted by GenoSplice as previously described [2,27]. FASTQ sequence files were mapped using hg19 reference with bwa-0.7.12 (mem) and converted to bam with samtools-1-1 (mpileup) to detect low frequency variants (i.e. < 10%). The pileup was parsed with pileup2base which provides the count of reads for each strand and nucleotides by coordinates. A home-made Perl script was developed to filter variants according to following criteria: (1) variants not retrieved by GATK; (2) ≥ 10 reads with the alternate allele; (3) ≥ 1% of allelic frequency; and (4) maximum strand bias value of 0.5. Candidate variants were validated by standard Sanger sequencing or deep site-specific amplicon sequencing. Variant pathogenicity for missense single nucleotide variant (SNV) was assessed using Mendelian Clinically Applicable Pathogenicity (M-CAP v1.4) [18] and Combined Annotation-Dependent Depletion (CADD v1.6) pathogenicity classifier website [7].

Laser capture microdissection and droplet digital PCR
Laser capture microdissection (LCM) was performed on the frozen resected tissue of one MOGHE case (FR-2) subjected to surgery at the Rothschild Foundation Hospital (Paris, France). LCM was performed using a Leica LMD7000 system on 20 μm frozen brain sections mounted on PEN-membrane slides after rapid (30 s) cresyl-violet staining. Cell populations were selected as follows: (1) neurons: morphologically normal appearing neurons, oval shape, without Nissl substance cytoplasmic aggregates and longest cell body diameter d = 10-25 μm; (2) glial cells: small round-shaped cells, and d = 5-7 μm. We collected pools of 4 × 250 cortical normal neurons, 3 × 250 glial cells, and 3 × 70 heterotopic neurons from the subcortical white matter. To assess variant enrichment specifically in the oligodendroglial cell lineage, we performed a rapid immunostaining on a 20 μm frozen brain section mounted on a PEN-membrane slide using an anti-Olig2 primary antibody (clone EP112, Epitomics, 1:25) adapted from a previously published protocol [31]. We collected one pool of 250 and one pool of 500 Olig2-positive cells located in the regions of higher density within the white matter. Microdissected cells were collected in AdhesiveCap 500 Opaque tubes (Zeiss) for DNA extraction, as previously described [9]. We then performed droplet digital PCR (ddPCR) (QX200 system, Bio-Rad Laboratories), a highly sensitive technique based on sample partitioning following a Poisson distribution and allowing DNA absolute quantification, to detect specifically the variant in the different pools of cells. All reactions were prepared using the ddPCR Supermix for probes (no dUTPs) according the manufacturer's protocol. Specific ddPCR Mutation Detection Assay (FAM + HEX) for the detection of SLC35A2:p. Ser212Leufs*9 variant was purchased from Bio-Rad. Data were analyzed using the QX200 droplet reader and Quantasoft Analysis Pro software version 1.0.596. The entire DNA extraction volume per pool of cells was used in each reaction.

Statistical analysis
Genotype-phenotype correlations were conducted using the computing environment R provided by R Core Team (2019). Statistical analyses were performed using Chi-2 test to compare percentages from categorical variable and Wilcoxon-Mann-Whitney test to compare quantitative variable. Values of p < 0.05 were considered statistically significant.

Neuropathological description of the MOGHE cohort
We enrolled 20 sporadic patients with refractory focal epilepsy subjected to neurosurgery (Schoen Klinik, Germany) and postoperatively diagnosed with MOGHE. Amongst subjects, 9/20 had preoperative MRI characteristics of MOGHE as previously described [16]. Clinical and imaging features are summarized in Table 1. The neuropathological diagnosis of MOGHE was in accordance with previously reported MOGHE criteria [35], consisting in the presence of: (1) clusters of increased Olig2-immunoreactive cell density as compared to neighboring regions, close to the grey-white matter junction and deep white matter (Fig. 1a, b); (2) increased densities of heterotopic neurons in the deep white matter according to Mühlebner et al. [28] (Fig. 1c); (3) patchy areas of hypomyelination in the white matter identified by immunohistochemistry for myelin proteins, i.e. CNPase (Fig. 1d).

Genetic findings in the MOGHE cohort
We performed targeted gene panel sequencing in 20 FFPE MOGHE tissues with a mean sequencing coverage of 1161X. We identified somatic SLC35A2 variants in 9/20 patients (5 males and 4 females), representing a genetic burden of 45% (Table 2). SLC35A2 variant allele frequencies (VAF) ranged from 7 to 52% (mean VAF: 26.7%). All variants were absent from the gnomAD database [12] (gnomAD v3.1 accessed November 2020). Six loss-offunction variants (frameshift insertions/deletions, nonsense) were considered pathogenic. The three missense variants (p.Thr69Ile, p.Ser302Phe and p.Ser312Phe) were considered likely pathogenic according to ACMG guidelines [34]: presence at a mosaic state (PS2); absence from gnomAD database (PM2); predicted deleterious using two in silico methods, M-CAP, and CADD (all variants scored > 23) and affecting highly conserved amino acids within (Thr69) or in close proximity (Ser302, Ser312) of a helical transmembrane domain (PP3). No matched blood samples were available to confirm the brain specificity of the variants. We did not identify any causative variants in the other panel genes in the 9 SLC35A2-positive cases, nor in the 11 SLC35A2-negative MOGHE tissue samples.

Neuropathological reassessment of SLC35A2-mutated MCD cases
Since MOGHE is a neuropathological entity newly recognized in 2017 [35] and not yet included in the current ILAE classification [4], we asked whether previously identified SLC35A2-related MCD cases may be reclassified as MOGHE. To this aim, we assembled an international cohort of 43 MCD cases (including 19 mMCD, 7 FCD1, 9 FCD2, 8 inconclusive diagnosis; see Additional file: Table 1 for histological and genetic cohort description). The cohort included 18 SLC35A2-positive cases carrying brain pathogenic variants with an initial diagnosis of mMCD (n = 11), FCD (n = 1) or inconclusive diagnosis (n = 6). A neuropathological reexamination, blind to the initial histological diagnosis and genetic finding, was performed by reviewing all H&E slides, together with anti-NeuN (to identify heterotopic neurons) and anti-Olig2 (to identify clusters of oligodendroglial cells) immunostainings. In 17 out of 18 cases with an SLC35A2-variant, we confirmed the microscopic signature of MOGHE ( Table 3) with clusters of increased Olig2-positive cells at the grey-white matter junction and deep white matter, and increased densities of heterotopic neurons in the white matter. In one SLC35A2 case, the brain specimen was too fragmented to allow a meaningful interpretation. No MOGHE diagnosis was established for non-SLC35A2-mutated  Table 4.

Clinical and genetic landscape of SLC35A2-mutated MOGHE
Altogether, we report a series of 26 sporadic MOGHE subjects with a brain somatic SLC35A2 variant (n = 9 from the German cohort described in this study and n = 17 histologically reassessed cases from France, Republic of Korea and the Netherlands). All patients had refractory focal epilepsy and were subjected to surgery. The whole cohort comprised 16 males and 10 females representing a sex-ratio M/F of 1.6. Age at epilepsy onset ranged from 3 months to 7 years In total, we reported 24 distinct brain somatic variants (Fig. 2). Among the 26 patients, five had a recurrent somatic pathogenic SLC35A2 variant. Loss-of-function variants were the most frequent with 37% frameshift insertions/deletions, 25% nonsense SNV and 4% splice site SNV. Missense variants (30%) and in-frame insertions/ deletions (4%) were considered as likely pathogenic based on in silico pathogenicity prediction, protein domain localization, amino acid conservation among species, absence from the gnomAD database and their presence at mosaic state in the lesional tissue. The variants p.Ser212Leufs*9, p.Leu120Hisfs*7 and p.Gln185* were recurrent somatic variants; p.Ser212Leufs*9 was previously reported in two were previously reported and were identified as P8 and P3 respectively in [16]. No blood sample was available to confirm the brain specificity  other sporadic cases with focal epilepsy related to MCD [25,41]. SLC35A2 variant allele frequencies (VAFs) ranged from 1.4 to 52% among the cases.
To determine possible genotype-phenotype correlations, we compared clinical features and surgery outcome between MOGHE SLC35A2-mutated cases (n = 26) and panel-negative cases (n = 11) (Additional file: Table 2). We did not observe significant differences in age at seizure onset, duration of epilepsy before surgery, neurodevelopmental status and post-surgical outcome. The only feature distinguishing SLC35A2 cases from panel-negative cases was a higher proportion of infantile spasms in SLC35A2-MOGHE cases (77% vs. 36%, p = 0.01).

Oligodendroglial cells and heterotopic neurons carry SLC35A2 variants
We next asked which are the specific cell types carrying SLC35A2 variants. To address this question, by laser capture microdissection we isolated glial cells located in the white matter, heterotopic neurons at the blurred grey-white matter junction and morphologically normal neurons within the grey matter, based on their morphology on cresyl-violet stained frozen sections from one male MOGHE case (FR-2, VAF in the bulk brain tissue of 22.6%). Variant allele frequency was determined in each pool by droplet digital PCR. Since SLC35A2 Table 3 Neuropathological reassessment of SLC35A2-mutated case *Patients with an initial histological diagnosis suggestive of MOGHE. Patient FR-2 underwent three surgeries, and we confirmed the presence of a mosaic gradient, with a 2% VAF identified in the tissue from the first surgery and a 22.6% VAF in the tissue from the second surgery. Tissue from the third surgery was not tested for the genetic variant SLC35A2 variants were absent from blood-derived DNA when available (all except KR-10, NL-1, NL-2) is an X-linked gene, present in one copy in males, the VAF directly reflects the percentage of cells carrying the variant. We found a VAF enrichment in both pools of glial cells (36.5 ± 6.2%) and heterotopic neurons (28.6 ± 4.2%) compared to morphologically normal neurons (8.7 ± 3.1%) (Fig. 3). These results indicate that both glial cells and heterotopic neurons, rather than morphologically normal cortical neurons are the main carriers of SLC35A2 variants in MOGHE. We then hypothesized that clusters of high Olig2-positive cells density have a higher VAF as a consequence of the SLC35A2 variant. Analysis of microdissected pools of Olig2-positive cells located in clusters within the white matter showed that nearly half of these Olig2-positive cells carried the variant (mean VAF = 50.5 ± 18%). Altogether, these results provide the proof of principle demonstration that SLC35A2 variants are enriched in pathological cell types featuring MOGHE.

Discussion
In this study we report the so far largest cohort of patients with MOGHE, a newly recognized clinico-pathological entity belonging to the spectrum of epileptogenic mild malformations of cortical development (mMCD). Here we took advantage of advanced deep sequencing technology and bioinformatic tools to detect brain somatic variants. We used unmatched FFPE tissues, the most common archival brain tissue source and highlighted the importance of genetic diagnosis for comprehensive genotype-phenotype studies. Our findings demonstrate that somatic brain-only variants in the UDP-galactose transporter gene SCL35A2 are a frequent cause in MOGHE.
Most SCL35A2 variants are nonsense or frameshift variants leading to loss-of-function of the protein in the mutated cells (Fig. 4a). Pathogenicity of three SLC35A2 variants reported in this study has been demonstrated in previous studies: (1) Sim et al. showed aberrant pattern of glycosylation in tissues carrying the SLC35A2 brain somatic variants p.Glu254* and p.Gln197* (patients KR-4 and KR-5 in this study respectively) [38]; (2) Ng et al. performed a biochemical assay on a SLC35A2-CDG patient fibroblasts carrying the p.Gln168* variant (same variant here described in patient KR-6) to confirm that SLC35A2-dependent UDP-galactose transport into the Golgi apparatus is altered [30]. Diagnosis of MCD remains challenging in everyday clinical practice [32] and white matter lesions in the borderland of FCD are lacking a comprehensive definition and description of clinical phenotypes, imaging features or specific molecular biomarkers [29]. MOGHE is currently not included in the 2011 ILAE classification of FCD [4], since it was first described histopathologically in 2017 [35] and is therefore likely underdiagnosed in current clinical practice. By reevaluating previously diagnosed SLC35A2 cases including anti-Olig2 immunostainings, we confirmed they all belong to the MOGHE entity. We next asked whether histopathological hallmarks of MOGHE comprising clusters of increased Olig2-immunoreactive cells, patchy areas of hypomyelination and foci of heterotopic neurons in the white matter were a direct consequence of the SLC35A2 variants. By means of a single cell approach performed in a MOGHE tissue, we identified that the SLC35A2 variant was enriched in both Olig2-positive cell clusters and heterotopic neurons (Fig. 4b), suggesting that the mutational event occurred during brain development in a neuroglial progenitor cell, likely in radial glia cells, before the differentiation into neuronal and glial cell lineages.
Protein N-glycosylation is a key post-translational process, and the majority of all secretory glycoproteins appear to be N-glycosylated. Altered glycan structures may affect folding, stability, interaction with other molecules, and cell-surface expression of numerous proteins. Studies of CDG which manifest with severe neurological symptoms have highlighted the importance of N-glycosylation on brain development [11,33]. There is evidence that N-glycans contribute to the regulation of neural transmission and excitability of neural circuits [36], and in neuronal adhesion and neuronal migration [24] due to altered extracellular matrix proteoglycans [1]. Glycosylation can also affect the myelination process due to abnormal development of myelin sheaths constituted by glycolipids or abnormal connections between oligodendrocytes and their myelin sheath [6,15,17]. This hypothesis is further sustained by the report of myelination defects in other congenital disorders of glycosylation [11,33]. It remains to be further investigated how a primary defect in glycosylation in neuroglial progenitors leads to the histopathological features described in MOGHE. Patients with SLC35A2-related MOGHE have intractable focal epilepsy; however the underlying pathomechanisms remain unclear. In particular, while heterotopic neurons interconnected within neuronal networks may drive the epileptogenicity, to which extent the oligodendroglial hyperplasia and/or hypomyelination contribute to epileptogenesis is unknown. To further clarify these mechanisms, SLC35A2-deficient mouse models will be needed to dissect the spatio-temporal contribution of each cell type. Accordingly, a recent study reported a correlation between interictal EEG spike density and the SLC35A2 variant allele burden from different brain areas of a patient who underwent multilobar resection [25]. In FCD type 2 and hemimegalencephaly linked to mosaic variants in genes of the mTOR pathway, a similar correlation between the size of the lesion or the epileptogenic activity and the mosaic rate has also been reported confirming the proepileptogenic role of mutated cells [2,20,21].
The finding of SLC35A2 variants as a cause of MOGHE may also promote the development of precision therapies. Two reports showed a clinical improvement in response to oral D-galactose supplementation in SLC35A2-CDG patients, resulting in reduction of seizure frequency [10,42]. Galactose supplementation may partially increase cytosolic UDP-galactose, thus facilitating galactosylation through alternative UDP-galactose transport into the Golgi [14,22]. Evaluating this therapeutic approach in patients with brain mosaic loss of SLC35A2, and for whom surgery is not an option or failed to reduce seizures, could offer a personalized treatment strategy. Regarding the 55% (11/20) SLC35A2-negative MOGHE cases of the cohort, the causal gene still needs to be identified, either by targeted sequencing of genes belonging to other glycosylation pathways or by deep whole-exome/genome sequencing.

Conclusion
Our findings highlight the important contribution of brain mosaicism in the etiology of focal epilepsies associated with MCDs. FCD type 2 are well recognized as mosaic mTORopathies, due to brain somatic variants in the PI3K-AKT3-mTOR pathway [2,8,26,39 ]. Here we confirm that MOGHE is a brain mosaic disorder affecting the N-glycosylation pathway. Our work further emphasizes the importance of genetic diagnosis in neuropathologically characterized human brain tissues to provide guidance toward precision therapeutic targets in these refractory focal epilepsies.