X‐linked neuronal migration disorders: Gender differences and insights for genetic screening

Cortical development depends on neuronal migration of both excitatory and inhibitory interneurons. Neuronal migration disorders (NMDs) are conditions characterised by anatomical cortical defects leading to varying degrees of neurocognitive impairment, developmental delay and seizures. Refractory epilepsy affects 15 million people worldwide, and it is thought that cortical developmental disorders are responsible for 25% of childhood cases. However, little is known about the epidemiology of these disorders, nor are their aetiologies fully understood, though many are associated with sporadic genetic mutations.


| INTRODUCTION
Neuronal migration is a developmental process responsible for the formation of the elaborate architecture of the human brain, particularly that of the cortex.During gestation, excitatory neurons migrate from the dorsal subventricular zone (SVZ) along radial glial processes, allowing for the formation of the six laminar cortical layers (Gleeson & Walsh, 2000).Meanwhile, inhibitory GABAergic interneurons migrate tangentially initially from the ventral telencephalon then radially to the cortex (Moffat et al., 2015).Dysregulated neuronal differentiation and migration give rise to a subset of cortical malformations, often termed neuronal migration disorders (NMDs) (Liu, 2011).This review is limited to the NMDs lissencephaly, subcortical band heterotopia (SBH), periventricular nodular heterotopia (PNH) and polymicrogyria (PMG) and excludes schizencephaly, focal cortical dysplasia and marginal glioneural heterotopia.
NMDs, usually detected postnatally using brain imaging, are permanent malformations which significantly decrease quality of life, with refractory epilepsy and intellectual disability (ID) amongst the most common symptoms.These conditions are usually diagnosed in children, and are a burden carried throughout life, hence why medical advancement in this field should be made a priority.
Many X-linked NMDs occur because of genetic mutations which are often familial.Due to the severity of phenotype ensued from NMDs, female probands outnumber males as they are afforded the opportunity of random X-inactivation.In females, approximately 50% of neurons express the faulty gene, and thus attenuate the severity of the disease, compared with males.Additionally, females may also experience skewed X-inactivation, where an imbalanced proportion of the paternal or maternal X chromosome is expressed, leading to further phenotypic heterogeneity in females.These mechanisms lead to functional mosaicism in females, whereas males, because they only have one X chromosome, express the mutation in 100% of neurons in familial cases.Therefore, females with mild symptoms may pass on X-linked NMD mutations to male offspring who then present with a more severe form of the disorder.Due to this, diagnosis and genetic screening is of utmost importance in prevention of NMDs in future generations.
In this review we will discuss NMDs associated with X-linked aetiology; consider the evidence for genetic causality; the involvement of the molecular/cellular function of the associated genes as pathogenic mechanisms for NMDs; and offer approaches for future research to build on current evidence.Awareness of specific mutations will allow prenatal screening efforts and open avenues for potential therapeutic approaches.Additionally, understanding of gene function at a cellular level may help explain the range of phenotypes that accompany NMDs, and pinpoint the developmental defect.X-linked genetic mutations engender complex clinical presentations which are only partially understood, requiring greater awareness from clinicians and scientists alike.
This review presents an overview of the 10 most pertinent X-linked mutations associated with NMDs we identified.

| LISSENCEPHALY
Lissencephaly refers to a spectrum of NMDs where the surface of the brain appears smooth.Such manifestations can range between absent gyri (agyria), broad gyri (pachygyria), which constitute classical lissencephaly, and heterotopic neurons inappropriately located in white matter (SBH) (Fry et al., 2014).Classical lissencephaly was found in 1.17 children per 100,000 births in the largest epidemiological study to date (de Rijk-van Andel et al., 1991).
Three subtypes of lissencephaly exhibit X-linked aetiology: (a) Classical lissencephaly and (b) SBH are both associated with mutations on the Doublecortin (DCX) gene ( Xq23) and (c) X-linked lissencephaly with abnormal genitalia (XLAG) is associated with mutations on the Aristaless-related homeobox (ARX) gene (Xp21.3).A summary of the condition can be found in Table 1.
Symptoms of lissencephaly include refractory epilepsy, severe developmental delay and intellectual disability, all of which vary depending on the severity of the anatomical defect.Prenatal ultrasound and magnetic resonance imaging (MRI) reveal abnormal cortical development after 20 weeks, by which time the sulci normally develop, although because of the differences in severity, the extent of disability may be difficult to identify prenatally (Gha et al., 2006).Long-term follow up of a 1991 study of births in the Netherlands, which remains the largest epidemiological study, found that phenotypic severity and life expectancy could be predicted by severity of prenatal neuroimaging, which is useful for prenatal counselling (de Wit et al., 2011).

| Classical Lissencephaly
Many genes are implicated in lissencephaly, but autosomal dominant mutations on PAFAH1B1 (also known as LIS1) and on the X-linked DCX account for up to 76% of cases.
DCX mutations give a characteristic appearance on brain MRI with strong genotype-phenotype correlation, showing a gradient of cortical malformation severity from rostral-to-caudal with particularly severe dysregulation in the orbitofrontal region (Dobyns, Truwit, et al., 1999).Although this is promising from a diagnostic point of view, genetic testing is still vital in these patients to explore familial risk and guide prognosis.Variation between sexes is observed with these DCX mutations, where males present with X-linked lissencephaly as discussed here and females present with SBH (Pilz et al., 1998).
DCX mutations include missense and splice mutations as well as deletions resulting in frame shifts (Bahi-Buisson et al., 2013).The latter of these were found to be more commonly de novo, producing more severe phenotypes; whereas, the point mutations resulting in amino acid substitutions inferred milder developmental and cognitive defects (Gleeson et al., 1998;Portes et al., 1998).Further research has identified that some de novo cases may in fact have been inherited from asymptomatic mothers with functional mosaicism, resulting in non-penetrance (Aigner et al., 2003;Demelas et al., 2001;Gleeson, Minnerath, et al., 2000).However, further investigation is required to reveal the true nature of the inheritance in mosaic cases.

| SBH
SBH is a form of lissencephaly, with less severe clinical manifestation (Dobyns, 2010).Heterotopia is characterised by morphologically intact tissue in a non-physiological position.In SBH, heterotopic grey matter fails to migrate and forms a characteristic smooth, symmetrical band within the subcortical white matter; hence, it is commonly referred to as 'double cortex syndrome'.The clinical severity of SBH differs widely, and patients experience ID of a varying degree, although 65% have refractory epilepsy.Other manifestations include cerebral palsy, language impairment, exotropia and congenital adrenal hyperplasia, but some females have SBH without these symptoms (Hung et al., 2016).
DCX mutations in males are typically associated with lissencephaly; whereas, females are usually seen with SBH (Bahi-Buisson et al., 2013).This is likely due to X-inactivation in females, which leads to reduced expression of the faulty gene compared with males, leading to a milder phenotype.Occasionally, however, SBH is seen in a minority of males with DCX mutations (D'Agostino et al., 2002).The occurrence of less severe SBH instead of lissencephaly in these males can be explained in some cases by (a) mutational differences (such as those affecting non-coding gene regions) and (b) somatic mosaicism of the DCX mutation, that is, the mutation occurred in a neuronal stem cell, and therefore, only a subset of clonally derived neurons carries the mutation (D'Agostino et al., 2002;Kato et al., 2001;Poolos et al., 2002).However, the development of SBH in males is not completely understood, and many cases are left unexplained; therefore, it remains an area of focus for future research.
In females, sporadic DCX mutations tend to cause a more severe SBH phenotype and symptom burden compared with inherited mutations (though not as severe as lissencephaly).In familial cases, all cells in the body carry the DCX mutation.Often, this leads to selection for cells with inactivation of the mutant X chromosome, particularly in tissues that DCX functions, such as neural stem cells (Bahi-Buisson et al., 2013;Gonz alez-Mor on et al., 2017).This contrasts with sporadic mutations, as they tend to occur later in development, and therefore, T A B L E 1 A table outlining the key features of each lissencephaly subtype and their X-linked genetic origins.effective skewed X-inactivation of the mutant chromosome does not take place, so a higher proportion of neurons express the mutation and experience disrupted neuronal migration (closer to the 50% seen with random X-inactivation).Overall, skewed X-inactivation in familial SBH is the mechanism that is involved in producing less severe and asymptomatic cases in females.
Familial cases account for one third of SBH patients (Bahi-Buisson et al., 2013); however, this could be an underestimate because of the underdiagnosis of subclinical phenotypes.Additionally, fewer may appear in the literature because of publication bias towards more severe cases.Bahi-Buisson et al. (2013) identified 14 cases of asymptomatic SBH with DCX mutation in females, drawing attention to the risk of inheritance from an asymptomatic mother to male offspring, who could then develop lissencephaly (Takeshita et al., 2015).These findings highlight the importance of performing genetic screening on family members following every SBH diagnosis, to ensure subclinical cases are detected and effective genetic counselling can be carried out.
As we have explored, clinical severity of the SBH phenotype can vary widely; however, most cases can be traced back to a DCX mutation.Differences in the style (e.g., missense, nonsense and deletion) and location of the mutation within the DCX gene contribute to its heterogeneity, as they impact the functionality of DCX in varying degrees (Bahi-Buisson et al., 2013;Matsumoto et al., 2001).Additionally, differences in expression of the mutated DCX gene between genders contribute to its broad clinical manifestation.Therefore, despite its diverse appearance, SBH aetiology can be predominantly traced back to a single X-linked gene, justifying the need for future research into DCX.

| DCX
The DCX gene encodes for a protein also known as DCX, which acts as a microtubule-associated protein involved in the polymerisation and stabilisation of neurons (Gleeson, Lin, et al., 1999;Horesh et al., 1999).Microtubules have been shown to drive the process of neuronal migration by forming the cytoskeleton of leading processes of neurons and providing tracks to enable directional movement of intracellular cargos within processes (Sakakibara et al., 2013).
A wide range of DCX mutation types are seen in lissencephaly and SBH such as missense, nonsense, frameshift with insertion or deletion, splice site and deletion of exon (Bahi-Buisson et al., 2013).
The high prevalence of DCX mutations in lissencephaly (23% of sporadic cases in males, 100% of XLIS) and SBH (67% of all cases, 100% of family pedigrees), strongly suggests that DCX mutations can cause these conditions.To strengthen this notion, the prevalence of DCX mutations in those with lissencephaly and SBH should be compared with a control group in a candidate gene casecontrol study.If DCX mutations are significantly more common in those with lissencephaly/SBH, this implies that the mutation is likely involved with the disease process.
Aborted human foetal brain studies show that DCX is expressed in the brain during development, predominantly in the ventricular zone and cortical plate (Meyer et al., 2002).More specifically, DCX has been seen to be expressed by migrating neurons in organoids derived from human induced pluripotent stem cells (iPSCs) (Bershteyn et al., 2017).Human neural iPSCs in vitro with reduced protein expression exhibit delayed differentiation, impaired migration and deficient neurite formation (Shahsavani et al., 2018).This provides evidence that DCX plays a role in both neuronal migration and axon growth.
In rats, reduced DCX function results in an SBH-like phenotype (Ramos et al., 2006); whereas, gene knockout in mice produces phenotypes associated with defective neuronal migration.However, lamination defects, abnormal positioning of neurons and susceptibility to epilepsy are not seen in mice.
DCX loss of function studies, which comprise the bulk of research in animal models, do not reflect the same effects of DCX mutations in humans.Mutations found on DCX may produce modified doublecortin, which has a dominant negative effect by interfering with the assembly of microtubules; whereas, loss of DCX may allow the assembly, but not into a functional form.In vitro studies show that different types of DCX mutations exhibit opposing effects on the association of doublecortin with microtubules, depending on their effect on doublecortin phosphorylation.Therefore, replacing DCX in mice with human mutant DCX may produce phenotypes that are more similar to those seen in humans.
Together, the above experiments provide compelling evidence that DCX mutations impair neuronal migration, leading to the development of lissencephaly and SBH.However, they also highlight their limitations and the need for experiments using human iPSC-cells.iPSC derived brain-organoids from patients with DCX mutations would give us more detailed and accurate knowledge of the effect of DCX mutation on neuronal migration.Organoids also provide an ethically preferred alternative to studies in primates.Such experiments would further verify our understanding of the effect of patient specific DCX mutations and enable DCX gene therapy in babies.

| XLAG
XLAG is a rare condition with less than 20 cases discussed in the literature, making wider prevalence and prognosis difficult to calculate.XLAG is attributed to an Xp21.3 defective gene ARX, which encodes a protein known also as ARX protein, and is implicated in several cognitive defect disorders because of its role as a transcription factor (Kato et al., 2004).
ARX is expressed at high levels throughout the telencephalon, with reduced levels expressed in the SVZ, hippocampus, cortical plate, ventral thalamus and basal ganglia (Kato et al., 2004).ARX mutations exhibit defects involving tangential migration and differentiation of GABAergic interneurons in the cortex.These produce a continuous spectrum of phenotypes, which range from non-syndromic ID, developmental and epileptic encephalopathies, Partington and Proud syndromes to XLAG as described here (Katsarou et al., 2017;Kitamura et al., 2002;Traversa et al., 2020).
The effects of XLAG on the nervous system include corpus callosum agenesis, neonatal refractory epilepsy, hormone dysregulation, temperature instability and lissencephaly, which is more severe caudally (Dobyns, Berry-Kravis, et al., 1999).In most cases, survival is less than 4 years (Spinosa et al., 2006).Males are primarily affected by XLAG, with females displaying a milder phenotype, although corpus callosum agenesis is usually seen in both sexes (Mattiske et al., 2017).

| ARX
Mouse models have managed to recapitulate many of these presentations, particularly the knocked-in ARX P353R mutation (ARXP353R/Y) and the ARX deficient (ARX-/Y) models, which caused a severe failure (<30%) of the tangential migration of GABAergic progenitor neurons as well as a thinner cortical plate (Kitamura et al., 2009).This increased failure of tangential migration could explain the severity of the XLAG presentation, as well as the early mortality observed in both humans and mouse models, compared with other ARX-associated disorders.Analysis of ARX in XLAG patients found strong evidence of association.Seventeen of the 24 XLAG subjects, who were all genotypically male, had a premature termination leading to loss-of-function of the critical domains of the transcription factor product, resulting in a more severe phenotype (Kato et al., 2004).Missense mutations were also identified in six of the 24 patients, producing a less severe XLAG presentation.Interestingly, one male patient was diagnosed with XLAG, with a mutation on ARX gene found also in healthy controls, suggesting that the mutation is a polymorphism (Katsarou et al., 2017).This patient presents an interesting potential for the identification of interacting ARX targets, which may contribute to the phenotype of this polymorphic mutation.However, there is also the possibility that this patient had a different aetiology that causes XLAG or that the condition was inappropriately classified as XLAG.This case raises the question of polygenic inheritance, so a wider range of genes must be studied in these patients.Nevertheless, the correlation between ARX mutation and XLAG is very strong, but further insights into ARX function would strengthen its causative role in XLAG.

| PNH
PNH is the most common type of heterotopia and can be characterised by the failure of radial neuronal migration during corticogenesis (Parrini, 2006).It results in an ectopic accumulation of neurons forming nodules along the surface of the lateral ventricles in the sub-ependymal layer of the cortex, suggesting failure of a subset of neurons to migrate away from the proliferating zone (Gonzaĺez et al., 2013).Little epidemiological data is available regarding the incidence and demographic distribution of PNH, but it is estimated that 25% of refractory epilepsy can be attributed to some type of malformation of cortical development, of which 13%-20% is associated with nodular heterotopias (Barkovich et al., 2015).
The exact biomolecular pathogenesis of PNH has also not been confirmed, limited by human brain samples but also by disease heterogeneity.The cardinal symptom of PNH is childhood refractory epilepsy, similar to most other NMDs; however, unlike other NMDs, PNH does not always present with ID, although some cases of reading and spelling impairment have been reported in the literature (Chang et al., 2005;Srour et al., 2011).Coupled with a delayed onset of seizures, usually presenting in the teenage years, diagnosis is often delayed, resulting in an average age of diagnosis of 28.5 years (Lange et al., 2015).
The Human Phenotype Ontology database lists various extra-neuronal manifestations that are 'very frequently' observed with PNH.Ehlers-Danlos syndrome, pyloric stenosis, gastroesophageal reflux, bleeding diathesis, scoliosis and hernia in addition to cortical malformations have all been cited.Given that the pathology in PNH is confined to the rostral neocortex, an area void of neurocrest cells that are precursors of head skeletomuscular tissue, craniofacial defects are seldom observed in these patients.
The most common phenotypic variant of PNH, thought to account for approximately 60% cases, is a primarily X-linked variant known as classical bilateral PNH (BPNH) (Barkovich et al., 2015;Parrini, 2006), which is associated with mutations in the Filamin A (FLNA) gene.BPNH almost always has X-linked in aetiology, even more so than PNH generally.It is estimated that 80%-100% of familial cases of BPNH are associated with an X-linked mutation (Chen & Walsh, 2002;Sheen et al., 2001).The remaining phenotypic variants of PNH, known as atypical PNH, have generally unknown aetiologies (Parrini, 2006).The vast majority of X-linked BPNH is thought to be caused by mutations in the FLNA gene, whereas a small minority have been reported to be caused by other X-linked mutations such as the Fragile X mental retardation 1 (FMR1) and L1 Cell Adhesion Molecule (L1CAM) genes (Moffat et al., 2015).A summary of PNH and its X-linked genetic origins can be found in Table 2.
As expected with a primarily X-linked disorder, most cases of BPNH are observed in females because of skewed X-inactivation.This also explains why females are likely to survive to an age where they are diagnosed and genetically screened, thus making up most of the population of BPNH cases.Despite many articles claiming certain perinatal death in males with X-linked BPNH, there have been some confirmed cases of males cited in the literature (Lange et al., 2015).However, somatic mosaicism in sporadic postzygotic mutations creating two neuronal populations, one of which maintains normal migratory function, could provide an explanation for the viability in male case (Guerrini et al., 2004).

| FLNA
Forty-nine percent of all BPNH cases and 77% of all female BPNH cases are associated with FLNA mutation (Parrini, 2006).FLNA is a gene located at Xq28, which encodes the structural protein FLNA that binds to F-actin and forms a three-dimensional orthogonal network (Figure 1), connecting integrins and other  membrane proteins to the cytoskeleton (Cannaerts et al., 2018;Hart et al., 2006).FLNA is highly expressed in human neurons, particularly in the leading process of the neuron, which is the integral component involved in migration (Sheen et al., 2002).Mutations in human FLNA resulting in the deletion of the actin-binding domain have been shown to inhibit radial migration of excitatory neurons because of an inability of the neuron to adopt the bipolar conformation required for locomotion (Fox et al., 1998).Additionally, inhibition of neuronal migration has been demonstrated in murine models where FLNA levels were reduced through RNA interference (RNAi) and indirectly through the expression of a dominant negative FLNA protein, which neutralised the wild-type FLNA protein (Carabalona et al., 2012).
Together, the strong genotype-phenotype correlation in patients, the function of the FLNA protein in the cytoskeleton and migration, its localised expression within the leading end in neurons, and the neuronal migration defects generated by genetic manipulation of FLNA function in animal models support the causative link between FLNA mutations and BPNH.
Many different genetic mutations in the 48-exon FLNA gene have been discovered in patients with familial and de novo BPNH, in both females and males (Parrini, 2006).In familial cases of BPNH (all female heterozygotes), all identified mutations involved truncations of the FLNA protein, rendering it dysfunctional (Fox et al., 1998;Sheen et al., 2001).In the male-cases of BPNH, de-novo missense mutations are largely observed because the single amino acid substitution only mildly affects FLNA function (Fox et al., 1998).This contrasts with the familial truncating mutation, which causes a complete loss of function, hence why it is only seen in females (Fox et al., 1998;Sheen et al., 2001).A significant shortcoming in the majority of FLNA studies to date is that they only focused on analysing coding regions of the FLNA gene, leaving large parts of the non-coding genetics unexamined and potentially harbouring mutations that affect splicing and transcription (Parrini, 2006).

| FMR1
Mutations in the Fragile X Mental Retardation-1 gene (FMR1), located at Xq23.1, is a less frequent PNH associated locus (Liu et al., 2018).FMR1 codes for a ubiquitous cytoplasmic protein called Fragile X Mental Retardation Protein (FMRP), which is present in the adult and embryonic brain with mRNA-translation suppressing functions (Laggerbauer et al., 2001).Major mRNA targets of FMRP are microtubule-associated protein 1B (MAP 1B) and semaphorin 3F (SEMA3F), both known to be involved in neuronal migration, axon guidance and synaptogenesis (Li et al., 2009).The well-defined mutation on FMR1 consists of an expansion of >200 CGG repeats, which are associated with hypermethylation of the gene leading most likely to transcriptional silencing and loss of function (Willemsen et al., 2011).This leads to deregulated translation of the aforementioned mRNA targets.The physiological function of FMRP provides compelling evidence that subsequent loss of function would in turn cause a pathology similar to that of PNH.However, to date, just two unrelated PNH probands with FMR1 mutation have been reported (Moro et al., 2006).Both cases were males who presented in their adolescence with a history of ID but an absence of seizures, which is surprising as seizures are the predominant symptom in PNH patients with FLNA mutations.Because both patients tested negative for all other PNH-causing mutations, either X-linked or otherwise, it suggests that FMR1 mutation is the most probable cause of their heterotopia.However, the prevalence of FMR1 mutation in the male population is estimated to be one in 4000, which weakens the causative evidence between FMR1 mutation and PNH, as more than just two cases would be expected if such a causative link existed (Hagerman et al., 2001).Given that somatic mosaicism relies on sporadic mutations, it is plausible that the extent of CGG expansion on FMR1 in the neuronal population varies, resulting in survival in the two male patients identified so far.

| L1CAM
L1CAM, located at Xq28 (Djabali et al., 1990) encodes cell adhesion molecule L1 with various functions including neuronal migration and differentiation, which is expressed in many tissues including throughout the nervous system, small intestine, colon and kidneys (GeneCards, n.d.-b).CRASH syndrome (corpus callosum hypoplasia, mental retardation, adducted thumbs, spastic paraplegia and hydrocephalus) has been termed because of the defining symptoms found in variant L1 genomes and have been shown to have a range of severity in phenotype, with more severe phenotypes resulting in death before 2 years old (Fransen et al., 1995).However, this phenotype is primarily shown in males, with affected females heterozygous for the defective gene having milder presentation if any presentation is shown (Kaepernick et al., 1994).The severity of phenotype correlates with the genotypic mutations, with extracellular domain mutations tending to cause a milder phenotype than cytoplasmic domain mutations, as well as truncation or absence of L1 (Fransen et al., 1998).
Only one case of PNH with overlying PMG associated with the gene L1CAM has been identified in the literature (Shieh et al., 2015).This case also presented with the far more common phenotype of L1 syndrome (corpus callosum hypoplasia, ID, thumb adduction and spasticity), which has been well-reported in L1CAM mutations.As in many L1 syndrome-associated mutations, a missense mutation, in this case substituting glycine-587 for arginine that has been identified as a key residue because of its interactions with lysine-606, was identified (Bateman et al., 1996).The phenotype shown is thought to be because of the loss of heterophilic integrin adhesion inhibiting cell migration pathways, as well as loss of L1-L1 homophilic binding, which facilitates axon fasciculation, extension and guidance (De Angelis et al., 1999).In mouse models, Lindner et al. (1983) showed that monoclonal and polyclonal L1 antibodies inhibited proper migration of neurons.Itoh et al. (2004) also demonstrated that L1-knockdown resulted in reduced cortical neuron radial migration.
The genetic aetiology of PNH remains unknown in most cases (Guerrini & Filippi, 2005); however, FLNA mutations are by far the most identified genetic cause, especially when focusing on X-linked aetiology.Genetic screening can reveal additional cases and provide counselling to patients with PNH.

| PMG
PMG refers to a spectrum of disorders clinically defined by characteristic small and crowded gyri (Leventer et al., 2010).Heterotopias are seen in most cases, which indicate disruption to the process of neuronal migration in these patients (Jansen et al., 2016;Judkins et al., 2011).X-linked PMG is commonly represented within wider syndromes detailed later in this review.PMG itself can be subdivided depending on where the defect lies anatomically.Amongst the 24 cases of X-linked PMG in the literature, symmetrical and bilateral forms are most common, particularly bilateral perisylvian PMG (BPP) and bilateral frontal PMG (BFP).Asymmetrical forms including generalised, frontal and posterior perisylvian PMG have also been associated with X-linked genes.A summary of the condition and its X-linked genetic origins can be found in Table 3.
Clinical manifestations are greatly variable; however, epilepsy is seen in 78% of cases, and ID with motor defects is also common.Although there are no apparent rostral-caudal severity gradients within PMG, clinical severity is closely related to the extent of pathologygeneralised and bilateral PMG confers worse prognosis and more severe clinical manifestations than focal or unilateral forms (Leventer et al., 2010).Micro-or macrocephaly and extra-cortical defects, which are often seen with PMG, contribute to the severity of the associated symptoms.All the X-linked PMG associated genes reviewed below are broadly expressed and regulate several cellular functions.Consequently, mutations within these genes affect several tissues and organs, leading to a broad spectrum of multi-system defects.
Over half of the cases in one of the largest PMG histopathological series to date were attributed to 'presumed genetic' origin (Jansen et al., 2016), and of these, a large proportion belonged to families displaying X-linked inheritance patterns.Here, we review the DEAD-Box Helicase 3 X-Linked (DDX3X) gene associated with the most PMG cases together with Sushi Repeat Containing Protein X-linked 2 (SRPX2), NAD(P) Dependent Steroid Dehydrogenase-Like (NSDHL), Cullin-4B (CUL4B) and Oral-facial-digital 1 (OFD1) gene mutations associated with fewer cases.The exact mechanisms by which mutations in these genes lead to PMG is often unclear.Strong evidence exists for the causality of DDX3X mutants; however, more epidemiological and molecular-functional evidence is required to attribute a causative role for other gene mutations often seen with PMG.Furthermore, new candidate genes are yet to be uncovered.Candidate loci at Xq28 and Xq27.2-q27.3 were attributed to multiple cases of PMG within several families, but they have not yet been matched to corresponding candidate genes (Santos et al., 2008;Villard et al., 2002).This is also true for Aicardi syndrome, an X-linked disorder in which PMG has been frequently noted (in all 23 female cases in one study) (Aicardi, 2005;Hopkins et al., 2008).
To date, all PMG cases with DDX3X mutations were females with heterozygous de novo variants (Blok et al., 2015;Lennox et al., 2020;Moresco et al., 2021;Scala et al., 2019;Tang et al., 2021).Lack of de novo variants in males suggests that most DDX3X mutations are not compatible with life in males (Nicola et al., 2019).Conversely, PMG associated with familial mutations of SRPX2 and NSDHL were only seen in males.This is in line with the protection conferred to females by X-inactivation, as affected females could display subclinical phenotypes that remain undetected.For example, one male case of BPP was linked with an SRPX2 mutation, and family screening detected four heterozygous female carriers of the mutation (two unaffected, two with mild ID) who did not display BPP on MRI (Roll et al., 2006).

| DDX3X
DDX3X, located at Xp11.4 (Park et al., 1998), encodes 'ATP-dependent RNA helicase DDX3X', a protein broadly expressed across multiple tissues throughout the body (GeneCards, n.d.-a).This protein is thought to play a role in transcriptional regulation, the assembly of messenger ribonucleoprotein (mRNP) granules involved in splicing, mRNA nuclear export and translation in the cytoplasm (GeneCards, n.d.-a).
DDX3X variants have been identified in a clinically diverse range of developmental defects grouped under 'DDX3X neurodevelopmental disorders', which involve PMG in 7%-12% of cases (Blok et al., 2015;Lennox et al., 2020;Moresco et al., 2021;Scala et al., 2019;Tang et al., 2021).Most cases of PMG with DDX3X mutations were missense (15/19); however, one duplication and three deletion mutations have also been reported (Blok et al., 2015;Lennox et al., 2020;Scala et al., 2019).The missense variants appear to impair the helicase unwinding of RNA.Lenox et al (2020) show that DDX3X controls cortical development by regulating neuron generation mainly in the cortex.Mutations in DDX3X disrupt both neuronal differentiation (Lennox et al., 2020) and migration, possibly through dysregulation of the Wnt signalling pathway.Through direct influence of the canonical Wnt pathway (Blok et al., 2015), DDX3X exerts an influence on processes involved in neuronal migration such as radial glial attachment and differentiation of cortical progenitor cells (Harrison-Uy & Pleasure, 2012).
T A B L E 3 A table outlining the key features of PMG and its X-linked genetic origins.Interestingly, missense DDX3X mutations are associated with more severe clinical phenotypes than loss of function mutations (Scala et al., 2019).This could be due to additional disruption of DDX3X helicase activity with a missense mutation, as opposed to simply reduced activity with a loss of function mutation.Clinical and radiological signs are highly diverse both between variants and individual patients with the same variants.Because of DDX3X broad expression, PMG cases often have concurrent neuro-anatomical deficits such as enlarged temporal ventricles, microcephaly and varying extents of corpus callosum (Blok et al., 2015;Lennox et al., 2020).Additionally, extra-cortical manifestations are also common (Lennox et al., 2020;Scala et al., 2019;Tang et al., 2021) as summarised in Table 3.
Experiments in animal models have demonstrated the role of DDX3X in neuronal migration.DDX3X depletion in mice embryonic brains significantly impaired neuronal migration in radial glial cells (neuronal progenitors) (Lennox et al., 2020).However, no replication of the pattern of anatomical defects typical in DDX3X-related PMG has been demonstrated in-vivo.This could be accounted for by the differences between mice and human brains; however, further studies in animals are required to support the notion that DDX3X mutation is a direct cause of PMG.
The above evidence suggests that DDX3X mutation plays a significant role in the development of PMG.DDX3X regulates several cellular proteins, which are different in varying cell contexts.Therefore, investigating specific downstream targets in different human neurons and neuronal precursors may reveal more relevant information on mechanisms of pathogenesis.

| SRPX2
SRPX2 maps to Xq22.1 (Roll et al., 2006) and encodes a protein responsible for a wide range of physiological functions, largely through binding of signalling receptors.Limited available evidence suggests SRPX2 is involved with glutamatergic synaptogenicity (Sia et al., 2013) and preservation of synapses against pathological pruning by microglial and complement-processes (Cong et al., 2020), which affects language processes (Roll et al., 2010;Royer-Zemmour et al., 2008).
SRPX2 mutation has been found in a small number of patientsone male with BPP and four of his female relatives who were either unaffected or had mild ID (Roll et al., 2006).The missense mutation seen in these cases occurred in the functional 'first sushi domain' of SRPX2 and results in change to protein conformation (Roll et al., 2006).This mutation is thought to affect formation the of urokinase-type plasminogen activator receptor (uPAR), which is involved in radial migration and axonal guidance and facilitates neuronal migration (Royer-Zemmour et al., 2008;Shmakova et al., 2021).However, the causative role of the gene in PMG is yet to be confirmed.
Animal studies show that in utero silencing of SRPX2 ortholog influences neuronal migration in the developing rat cerebral cortex by a mechanism that involves alpha-tubulin acetylation (Salmi et al., 2013).Postnatal silencing in utero caused spontaneous seizure activity, but the seizures and migration defects were prevented by administration of a tubulin deacetylase inhibitor to the mothers (Salmi et al., 2013).This evidence suggests that disruption to tubulin is involved with the pathogenesis of PMG, which is supported by the fact that tubulin related mutations are common in autosomal forms of PMG (Romaniello et al., 2014).It should be considered whether maternal drug therapies, such as tubulin deacetylase inhibitors, could prevent developmental defects in a foetus with a known SRPX2 mutation.
Mutations in NSDHL have been linked to CK syndrome, a disorder first noted in a case series considering a 5-generation family with ID, seizures, microcephaly, 'thin body habitus' and cerebral cortical malformations, which included two males with widespread, irregular BPP defects anatomically distinct from those found in most BPP (du Souich et al., 2009).However, PMG is not ubiquitous in families or individuals with CK syndrome (Preiksaitiene et al., 2015;Sun et al., 2018).
It is uncertain whether deficiency of cholesterol as an end-product or build-up of sterol precursors relate to pathogenesis of CK syndrome.Although no observable abnormalities in the sterol profile in blood plasma of patients with CK syndrome have been observed to date (du Souich et al., 2009;McLarren et al., 2010), abnormal sterol profiles (including accumulation of 4-methylsterol intermediates) in cells in-vitro expressing mutant NSDHL (McLarren et al., 2010) has led to a speculation of methylsterol toxicity as the pathogenic agent.Although this is compelling, it is unclear how this accumulation may impair neural development.However, it is known that the sonic hedgehog signalling (SHH) pathway is cholesterol-dependent (Porter et al., 1996), and it plays pivotal roles in neuronal development, including ventral brain patterning, where the cortical interneurons are generated (Xu et al., 2010).
Therefore, NSDHL mutations may impair SHH signalling leading to PMG, and this has been proposed from studies on a conditional knockout of the NSHDL (Xlinked ortholog) in mice (Cunningham et al., 2015).In this model, pups develop ataxia with progressive loss of cortical and hippocampal neurons, cerebellar precursor proliferation and migration deficits, and subsequent apoptosis of the cerebellar cortex.The lateral defect was shown to result from defective SHH signalling (Cunningham et al., 2015), supporting the notion that this pathway connects NSHDL mutation with the formation of PMG.However, the polymicrogyric phenotype was not replicated in these animal models, and the high lethality of the mutation (resulting in fatality by 2 weeks of age) poorly reflects CK syndrome in humans, which is usually not diagnosed until many years into childhood.Furthermore, abnormalities seen in humans with defective SHH signalling are not similar to those seen in CK patients (McLarren et al., 2010).Further functional studies using in-vivo or in-vitro iPSC/organoid models more closely reflecting human tissue may yield further insight into the pathogenesis and help determine whether NSHDL mutations play a role in the development of PMG.

| CUL4B
The CUL4B gene, located at Xq24, encodes CUL-4B, a member of the cullin protein family (Sarikas et al., 2011).Cullin proteins form complexes, which have E3 ubiquitin ligase activity and are involved in nucleotide excision repair, regulating gene expression and controlling DNA replication (Kerzendorfer et al., 2010).Mutations in the CUL4B gene have been linked to cortical malformations and a range of clinical manifestations including ID, craniofacial dysmorphism and seizures (Badura-Stronka et al., 2010;Isidor et al., 2010;Tarpey et al., 2007).
A recent study by Vulto-van Silfhout et al. (2015) sequenced the genome of a cohort of patients with neurodevelopmental disorders and identified several CUL4B missense mutations that were absent in control individuals.Many of these patients had PMG and a simplified gyral pattern, suggesting this gene plays a role in neuronal migration.
CUL4B gene knockout studies in animal models are limited, as the null mutation is typically lethal in mice.A study by Chun-Yu Chen et al. (2012) overcame this by crossing female mice with Sox2-Cre male mice to escape the critical time window of CUL4B-dependent embryonic development.The results showed impaired spatial learning and memory, and epileptic susceptibility, consistent with some clinical manifestations in humans.However, this method may not fully reveal the effect of CUL4B mutation on neuronal migration, as it rescues a key stage of embryonic development.Nonetheless, it would be useful to look for evidence of abnormal neuronal migration in these mice using imaging.Tripathi et al. (2007) showed that RNAi-mediated knock-down of CUL4B in a mouse cell line results in increased beta-catenin levels, which is a key transducer of the Wnt signalling pathway.Canonical Wnt/betacatenin signalling is known to regulate rostral-caudal and medial-lateral patterning in the developing cortex, as well as the regulation of neural precursor self-renewal and laminar fate determination (Chenn, 2008).This suggests that CUL4B mutation could impair neuronal migration through the regulation of beta-catenin levels; however, further studies in vitro and in vivo are required to confirm this.
A study, which knocked down cullin 5 (a protein encoded by the CUL4B gene) in migrating neurons, showed an accumulation of active Dab1 protein and a cortical layering defect characterised by excess migration and buildup of neurons at the top of the cortical plate (Feng et al., 2007).The results could implicate cullin 5 in the down-regulation of Dab1 in vivo and suggest that cullin 5 plays a role in regulating migrating neurons possibly by opposing a promigratory effect of Dab1.
Overall, these studies suggest that CUL4B plays a role in regulating neuronal migration and that mutations in CUL4B contribute to a wide range of developmental abnormalities.Further studies using animal models and iPSCs are required to delineate the impact of CUL4B mutation on neuronal migration, and further understanding of the molecular and cellular pathways involved could ultimately lead to the development of targeted therapies.

| OFD1
OFD1, located at Xp22.2, is another X-linked gene thought to be a rare cause of PMG (Stutterd et al., 1993).OFD1 encodes a cytoskeletal protein, which can be localised within centrioles and basal bodies of primary cilia, hence why OFD1 mutations are often interchangeably referred to as ciliopathy (Alfieri et al., 2020).Ciliary genes such as OFD1 have been shown to regulate various cell signalling pathways such as Wnt, which is integral to neuronal migration.Mouse derived embryonic stem cells with silenced OFD1 show aberrant Wnt signalling and, in turn, neuronal migration (Del Giudice et al., 2014).This reliance between OFD1 and signalling pathways, which are necessary for neuronal migration, provides compelling evidence to support the causative link between OFD1 and PMG.
However, the incidence of OFD1 mutations is approximately one in 100,000 live births, while PMG has only been reported in two patients with OFD1 mutation.Both of these patients exhibited a substitution of valine for aspartate at amino acid residue 307 (p.Val307Asp) (Dehghan Tezerjani et al., 2016;Giordano et al., 2009).These two cases showed radiological evidence of PMG in the caudal brain, which provides explanation as to why a multitude of craniofacial defects are also observed in patients with OFD1 mutations including frontal bossing, macrocephaly and wide nasal bridge.The rare incidence of PMG relative to the prevalence of OFD1 mutations, in addition to the lack of variety in OFD1 mutations causing PMG, provides weak support for the causative link between OFD1 and PMG.Instead, given that ID and craniofacial deformities are common symptoms of OFD1 mutation, it suggests that OFD1 does play some role in neuronal circuit formation, but that this role is not necessarily specific to the pathogenesis of PMG.For instance, neurons, which have been engineered from OFD1 mutant human iPS cells, have been shown to fail to differentiate, particularly into interneurons, which are normally generated in the ventral telencephalon and rely on tangential migration to reach their final destination, the cortex, in the dorsal telencephalon (Hunkapiller et al., 2011).This suggests that OFD1-associated PMG is a disorder of neuronal differentiation which precedes neuronal migration.However, it is still possible that OFD1 plays a role in both.

| CONCLUSION AND FUTURE PROSPECTS
Twenty-five percent of children with refractory epilepsy will have a cortical development disorder, making NMDs a critical differential which should be considered in these patients.Family history of ID or developmental disorders in the child should also elicit clinical suspicion (Hsu et al., 2020).PNH is the exception, as these patients may only develop seizures in later life, though cognitive difficulties can emerge in childhood too, albeit less commonly (Felker et al., 2011).As brain development occurs predominantly in utero, NMDs usually present in children and adolescents through debilitating symptoms such as seizures and ID.These cortical malformations are permanent and current medical treatments typically only offer mild symptom relief, highlighting the importance of a preventative approach to these disorders.
Currently, NMD diagnosis is mostly reliant on MRI imaging, which often results in a vague classification of what is largely a heterogeneous spectrum of disorders.Genetic testing enables more specific diagnoses, which can be managed more effectively through personalised pharmacotherapy and gene therapy (Walker et al., 2015).Using antenatal testing, gene-carriers can be identified, allowing for genetic counselling of the risks of inheritance.For example, after the identification of DCX mutations as a cause of multiple NMDs, prenatal detection of DCX has become possible through direct sequencing of chorionic villi or amniocentesis (Leventer et al., 2000).
NMDs possess a large degree of clinical and phenotypic heterogeneity, with often unclear molecular pathogenic mechanisms, which can make it difficult to determine their genetic aetiology at times.Early studies restricted genetic analysis to exome sequencing, which required previous knowledge of which exon contained the mutation in order to search for it (Perenthaler et al., 2019).Fortunately, recent advances in genome wide association studies (GWAS) have allowed researchers to screen the entire genome, including regions of transcriptional regulation, mRNA stability and splicing.With this technology, many more mutations can be identified without any prior knowledge of where the mutation may be situated.
Because of the potential severity of phenotype associated with NMDs, compatibility with life is incumbent on genetic mechanisms that reduce expressivity, such as X-inactivation.This explains why X-linked NMDs can be milder than autosomal NMDs and why these patients are more likely to survive to reproductive age (Germain, 2006).Using this knowledge, it is suggested that female-targeted screening programmes for those who possess mild symptoms of NMDs, such as ID, will allow more families with these mutations to be identified.Furthermore, genetic screening gives females with known mutations an opportunity to explore preconception reproductive options, such as preimplantation genetic diagnosis, which allows genetic screening of embryos in vitro and selection of unaffected embryos for implantation.This offers an alternative to prenatal diagnosis of NMD, which carries an inevitably difficult decision regarding termination should an NMD be detected.Preimplantation diagnosis provides a better option for many people; however, ethical challenges, such as whether heterozygous female embryos should be transferred, and practical issues due to the low success rate (18%-20%), need to be considered (Steffann et al., 2018).
In addition to detection, it may also be possible to actively modify these genes pre-emptively in order to prevent disease.The field of genome editing is developing with UK scientists successfully using somatic gene therapy in 2015 to treat a one-year-old girl with leukaemia (Reardon, 2015).Neuronal migration occurs within the first weeks of zygotic development; therefore, somatic gene therapy after this stage would be ineffectual.Germline therapies would instead be necessary to prevent the development of NMDs; however, there are ethical and safety concerns, as germline changes would affect every following generation (Gonçalves & Paiva, 2017).The NIH currently does not fund research into editing of human embryos; however, this remains a future possibility with improving technology (Gene Editing-Digital Media KitjNational Institutes of Health [NIH]).Preimplantation genetic diagnosis provides a safer alternative because it does not involve the risk of potential errors in genome editing.However, the birth rate following preimplantation genetic diagnosis is low because of the vast number of embryos that are discarded (Steffann et al., 2018).With an improvement in genome editing technology, these embryos could be treated to remove their pathogenic genes, rather than being discarded.Some may argue that this is a more ethical solution to preventing NMD births in the population.
Overall, with better recognition of NMDs and their causative genes, preventative efforts can be improved.Through the continuation of research into the genetic aetiologies of these life-altering conditions, we can deepen our understanding and strengthen our arsenal against them.

F
I G U R E 1 Diagrammatic representation of Filamin A (FLNA) and F-actin forming orthogonal three-dimensional cytoskeletal network.
T A B L E 2 A table outlining the key features of PNH and its X-linked genetic origins.
Abbreviations: FLNA, Filamin A; FLNB, Filamin B; FMR1, Fragile X mental retardation 1; ID, intellectual disability; L1CAM, L1 Cell Adhesion Molecule; PNH, periventricular nodular heterotopia.a 88% of PNH patients with FLNA mutation PNH present with epilepsy.b Intellectual ability is typically normal or borderline but problems in behaviour and reading proficiency have been cited.