International consensus recommendations on the diagnostic work-up for malformations of cortical development

Malformations of cortical development (MCDs) are neurodevelopmental disorders that result from abnormal development of the cerebral cortex in utero. MCDs place a substantial burden on affected individuals, their families and societies worldwide, as these individuals can experience lifelong drug-resistant epilepsy, cerebral palsy, feeding difficulties, intellectual disability and other neurological and behavioural anomalies. The diagnostic pathway for MCDs is complex owing to wide variations in presentation and aetiology, thereby hampering timely and adequate management. In this article, the international MCD network Neuro-MIG provides consensus recommendations to aid both expert and non-expert clinicians in the diagnostic work-up of MCDs with the aim of improving patient management worldwide. We reviewed the literature on clinical presentation, aetiology and diagnostic approaches for the main MCD subtypes and collected data on current practices and recommendations from clinicians and diagnostic laboratories within Neuro-MIG. We reached consensus by 42 professionals from 20 countries, using expert discussions and a Delphi consensus process. We present a diagnostic workflow that can be applied to any individual with MCD and a comprehensive list of MCD-related genes with their associated phenotypes. The workflow is designed to maximize the diagnostic yield and increase the number of patients receiving personalized care and counselling on prognosis and recurrence risk.

Many MCDs are caused by an underlying genetic defect. Rapid advances in molecular genetics and neuro imaging techniques in recent years have substantially increased the number of recognized MCD forms and their associated genes, and have highlighted the con siderable genetic heterogeneity associated with these disorders 1 . Nextgeneration sequencing (NGS) of a selec tion of genes related to a phenotype (gene panel), the coding exons of the human genes (exome sequencing) International consensus recommendations on the diagnostic work-up for malformations of cortical development or the genome of an individual (genome sequencing) has enabled rapid sequencing of large numbers of genes.
Even following intensive diagnostic assessments, many individuals with an MCD remain without a molecular diagnosis [4][5][6] . The complex nature and high degree of clinical and genetic heterogeneity of MCDs demand highly specialized and multidisciplinary exper tise. However, MCD experts usually work individually or in small multidisciplinary teams. Currently, com prehensive guidelines for diagnosis and management are lacking, adding to the variability in the diagnostic approach between different centres. The disease course and longterm clinical outcome are often difficult to pre dict at an early stage, and medical management is rarely evidencebased. These challenges highlight the need for an expertdriven multidisciplinary effort to better understand these disorders. The availability of carefully curated MCD gene panels to the wider medical commu nity will enable accurate molecular diagnosis in a larger number of patients without long delays or unnecessary investigations.
We established the international multidisciplinary network NeuroMIG with the aim of disseminating knowledge to the broad medical community, improving the diagnosis and management of MCDs and accelerat ing research into MCDs 7 . In this article, we first review the clinical presentation and aetiology of the main MCD types. On the basis of a critical review of the literature, expert surveys and discussions, we then present a consen sus statement on the clinical and molecular investigations in patients with MCDs, including specific recommenda tions on clinical workup, molecular diagnostic methods and alternative strategies in undiagnosed patients.

Methods
This article represents a consensus document based on three facetoface expert meetings within the NeuroMIG network that were held in St Julians, Malta, from 21 to 23 February 2018, in Lisbon, Portugal, on 13 and 14 September 2018, and in Rehovot, Israel, on 17 March 2019. The meetings were funded by the European Cooperation in Science & Technology (COST Action CA16118). Two NeuroMIG working groups, WG1 and WG3, took the lead in preparing the draft, although a larger group within the network was invited to participate in the Delphi consensus procedure and comment on the second draft. The final version of the consensus docu ment was reviewed by the drafting team and circulated among all COST network members before submission.
From the MCD expert laboratories within the NeuroMIG network, headed by M.W., K.S., U.H., E.P. and N.D.D., we collected data regarding gene panels, enrichment strategies and diagnostic yield. Using the data obtained as described above, we compiled lists of genes associated with the various MCD subtypes and defined a diagnostic strategy for patients with MCDs. The gene list was curated -that is, checked, cor rected and completed -by all authors on the basis of longstanding personal experience gained through molecular diagnostics in patients with MCDs. The first draft was finalized before the second meeting. During the first round of voting, 21 of the authors voted on 101 recommendation statements. Agreement (>90% positive votes) was reached for 89 statements, and the remain ing 12 were revised according to the reasons provided for disagreement. The second round of voting involved 42 experts. At the end of the process, 94 recommenda tions found >90% consensus. In addition, five statements were agreed on by 80-90%, two statements by 75-80% and one statement by 70-75% of the participants (Supplementary Table 1). Recommendations with con sensus <80% were excluded from the recommendations section below. Unless specified otherwise, we report on recommendation statements with >90% consensus.

Clinical presentation of MCDs
MCDs can be isolated or associated with a wide variety of neurological and extraneurological features, includ ing other birth defects and facial dysmorphism. The age at clinical referral and the severity of neurological defi cits vary substantially between affected individuals. The most common presenting features are epilepsy, develop mental delay and/or motor abnormalities of tone, move ment and posture 1 . These features are listed in relation to the typical ages of presentation in Box 1.

Main MCD types
In this section, we provide an overview of the most common types of MCD and their aetiologies. Different descriptions have been introduced in the literature over the years depending on the study design and the med ical background of the research group (for example, neurologists, radiologists, geneticists or pathologists). TaBle 1 summarizes the consensus definitions that were agreed on by our working group. These defini tions are used throughout the text, and brain imaging examples are provided in Fig. 1. The descriptions are specific to each term and do not consider the presence of abnormalities of other brain structures, which often coexist with MCD. Each MCD type can be further clas sified on the basis of morphology, topography, severity gradient and involvement of other brain structures 1 . A detailed paper on the MCD neuroimaging features has been published separately by representatives from the NeuroMIG network 8 .

Microcephaly
Microcephaly is defined as a significant reduction in the occipitofrontal circumference (OFC) compared with controls matched for age and sex. Microcephaly is the most common MCD and is present in 15% of children referred for evaluation of developmental disabilities 9 . The relevant degree of reduction differs throughout the literature, being set at 2-3 s.d. below the mean 9-12 . Strictly speaking, microcephaly is a clinical finding rather than a disease; however, it provides a reliable estimation of the brain volume 10 . The final brain size is the result of a complex process of neural stem cell proliferation, migration, and ongoing organization, synaptogenesis and apoptosis 11 . Microcephaly is classed as congenital if present at birth (primary microcephaly) or postnatal if it develops after birth (secondary micro cephaly) 10,13,14 . These two groups also have different molecular aetiologies 11 . Microcephaly can present with a normal or simplified gyral pattern, or with additional, more complex brain abnormalities 11 . The clinical out come cannot be predicted by head size alone and largely depends on the underlying cause and the appearance of the brain on MRI.

Macrocephaly and megalencephaly
Macrocephaly is defined as an OFC ≥2 s.d. above the mean, whereas megalencephaly refers to an abnormally large brain size 1 . Macrocephaly has a wide variety of causes besides megalencephaly, including hydrocepha lus and increased skull thickness. Mild megalencephaly (2-3 s.d. above the mean) with an otherwise struc turally normal brain can be seen in typically devel oping children, often in the setting of benign familial macrocephaly 15 . However, megalencephaly can point to an underlying neurodevelopmental or generalized overgrowth disorder.

Periventricular nodular heterotopia
The term neuronal heterotopia refers to groups of neurons in an abnormal location, and periventricular nodular heterotopia (PVNH) describes nodular masses of grey matter located along the ventricular walls pro truding into the ventricle 1 . PVNH can occur in isola tion or together with other brain or body malformations and is not rare: in one study, PVNH was observed in 0.48% of the general paediatric population 16 . The nod ules can occur unilaterally or bilaterally, and should be further defined according to their number and location (for example, involving the frontal or temporal and/or occipital horns of the lateral ventricles).

Lissencephaly spectrum
The lissencephaly spectrum encompasses agyria, pachy gyria and subcortical band heterotopia (SBH) 17 . Agyria and pachygyria are characterized by an abnormal gyral pattern with absent gyri (agyria) or broad gyri (pachy gyria) in combination with an abnormally thick cortex 18 . SBH describes a band of grey matter separated from the cortex and lateral ventricles by zones of white matter 18 . In rare cases, pachygyria and SBH can cooccur in the same brain, with a typical pattern of frontal pachygyria and posterior SBH 19 . Microlissencephaly represents a separate subgroup and is defined as a combination of lissencephaly (usually in the form of agyria or pachy gyria) with severe congenital microcephaly (OFC at birth ≥3 s.d. below the mean) 20 .

Subcortical heterotopia
Subcortical heterotopia (SUBH) refers to brain malfor mations with clusters of neurons located within the white matter, between the cortex and lateral ventricles 21 . The wellrecognized and aforementioned PVNH and SBH have distinct imaging patterns and are classified sepa rately. Multiple terms have been used to describe this type of malformation, including giant, curvilinear, nodular, focal and massive heterotopias 21 . In 2019, a group within the NeuroMIG network provided the first framework for an imaging classification of SUBH that encompasses five groups further subdivided into specific entities 21 .

Cobblestone malformation
COB is recognized as an undersulcated, irregular and 'pebbled' cerebral surface, with a moderately thick cortex 22,23 . This malformation is caused by defects of the pial limiting membrane with resulting neuronal overmi gration from the cortical plate into the leptomeninges 3,24 . COB often cooccurs with eye, muscle and addi tional brain malformations within the spectrum of the αdystroglycanopathies, with Walker-Warburg syndrome at the most severe end 25 .
COB was originally described as lissencephaly type 2 but this term has now been abandoned 26 . In addition, COB is often confused with polymicrogyria 27 . The strict differentiation of COBrelated and polymicrogyria related genes in the literature remains difficult, as sev eral conditions characterized by COB were reported as polymicrogyriaassociated disorders (for example, GPR56associated frontoparietal 'polymicrogyria' and CHIME syndrome).

Polymicrogyria
Polymicrogyria is one of the most frequent types of MCD and is also one of the most heterogeneous in aetiology 1 . Polymicrogyria is defined as an excessive number of abnormally small cerebral gyri with cortical overfold ing, an irregular, pebbled cortical surface and a stippled grey-white matter boundary 28 . Megalencephaly specifically refers to a brain size that is ≥3 s.d. above the mean and is primarily a developmental brain disorder, whereas macrocephaly (defined as an OFC ≥3 s.d. above the mean) has a wide variety of causes besides megalencephaly, including ventriculomegaly, hydrocephalus and increased skull thickness.
volume 16 | November 2020 | 621 NATure revIeWS | NeuROlOGy As highlighted in the previous section, polymicro gyria can be difficult to differentiate from COB, and might also be confused with dysgyria or pachygyria. Highresolution imaging can aid the differentiation of these conditions, as it can show microgyri, microsulci and stippling of the grey-white matter junction -a spe cific feature of polymicrogyria that is not seen in other MCDs 1 Fig. 1 | MRI scans showing common malformations of cortical development. The brain was scanned in the axial plane unless otherwise stated. a | Normal brain on T1-weighted images. b | Normal brain on T2-weighted images. c | Primary microcephaly with a small brain. d | Abnormally large brain (megalencephaly) with abnormal appearance of the perisylvian cortex (arrows point to small gyri suggestive of polymicrogyria). e | Bilateral nodular heterotopia (arrows) situated along the ventricular walls. f | Lissencephaly spectrum with agyria-severe pachygyria (arrows). g | Lissencephaly spectrum with subcortical band heterotopia visible as a thick band isointense to the cortex (asterisks). h | Generalized thickened cortex with broad gyri and white matter abnormalities consistent with cobblestone complex (arrows). i | Bilateral frontoparietal polymicrogyria with abnormally small gyri and shallow sulci (arrows). j | Coronal scan showing schizencephaly, characterized by a cleft lined by grey matter extending from the cortex to the ventricle (arrow). k | Abnormally oriented sulci of varying depth with normal cortical thickness (arrows). l | Focal cortical dysplasia with blurring of the grey-white matter boundary and hyperintensity of the white matter on T2-weighted imaging (arrow).
viewed on sagittal imaging, should be closely scrutinized as polymicrogyria often affects these areas preferentially, with abnormal posterior extension and sulcal branching being observed 28 . Polymicrogyria is frequently seen in association with many other brain malformations and is sporadically described in various syndromic disorders. Polymicrogyria has been classified into six topographic patterns that are further divided into 13 morphological subtypes 28 . Moreover, at least six polymicrogyria syn dromes have been defined on the basis of radiological and clinical features 29 .

Dysgyria
Dysgyria translates as abnormal gyration and can there fore be applied to almost every type of MCD. However, this term was introduced to describe cortical malfor mations that do not meet classic features of any of the abovementioned wellestablished MCD types. Dysgyria describes a cortex of variable thickness and an abnormal gyral pattern characterized by abnormalities of sulcal depth or orientation (for example, obliquely oriented sulci directed radially towards the centre of the cere brum and narrow gyri separated by abnormally deep or shallow sulci) 30,31 . In the vast majority of cases, the term dysgyria describes an abnormal nonlissencephaly, nonpolymicrogyria cortex within the spectrum of tubulinopathies.

FCD and hemimegalencephaly
FCD is identified on brain imaging by focal irregular ities of cortical morphology and thickness, blurring of the grey-white matter boundary, and white matter T2 hyperintensity. Depending on the size of the lesion and the resolution of the brain imaging, FCD can be missed on MRI. Smaller lesions are often only identified on neuropathological studies after surgery for epilepsy. FCD type II is characterized by the presence of dysplas tic, megalocytic neurons, a feature that is also present in hemimegalencephaly. Balloon cells are also observed in FCD IIB and hemimegalencephaly 32 . The size of the lesion varies from submicroscopic involvement of one or several sulci (FCD) to a larger area involving a lobe (par tial hemimegalencephaly) or involvement of an entire cerebral hemisphere (classic hemimegalencephaly) 32 . In the latter condition, the affected hemisphere is visibly enlarged. In hemimegalencephaly, the lesion can extend to nonbrain tissue, and clinicians should look out for skin abnormalities and localized overgrowth of one or several body parts.

Molecular testing: current practice Chromosomal testing
MCDs have been linked to a wide range of CNVs, as detected by chromosomal microarray analysis (CMA) 1,33,34 . Several CNVs are consistently associated with MCD, the most common of which are the 22q11 and 1p36 deletions associated with polymicrogyria, the 17p13.3 deletion (encompassing LIS1 (also known as PAFAH1B1), YWHAE and other genes) that causes Miller-Dieker syndrome and isolated lissencephaly, and 6qter deletions associated with various brain mal formations including polymicrogyria and PVNH 33,35,36 .
A study published in 2019 reported a diagnostic yield of 36% when CMA was used in patients who had PVNH with or without other malformations, and 9% in a group with polymicrogyria only 37 . Another study did not show an increased burden of rare CNVs in people with polymicrogyria compared with healthy controls 38 . In patients with microcephaly, the yield was ~5-7% 13,39 . In a large cohort of patients with lissenceph aly (n = 811), Miller-Dieker syndrome was diagnosed in 9% of cases 40 . Several MCDrelated genes frequently har bour intragenic deletions or duplications, which might be identified by standard microarrays [41][42][43] .

Single gene testing
Single gene testing is being superseded by NGS gene panels, and we were only able to identify systematic studies for a small number of MCD types. The yield of single gene testing varies greatly depending on the MCD type and extension of the malformation. For SBH, the yield of molecular testing is high, with path ogenic variants in DCX or LIS1 being found in 79% of patients (123 of 155) 40 . Pathogenic variants in FLNA are important aetiological factors for PVNH. The highest frequency is found in women with bilateral frontocentral PVNH, especially in combination with cerebellar hypo plasia and/or mega cisterna magna, with a positive fam ily history of PVNH 44,45 . The yield varies from 80-100% in female familial cases to 9-26% in sporadic cases [44][45][46] . In a cohort of 113 patients with MCDs, a molecular diagnosis was established in 21 patients (19%) by tar geted testing of one or more genes selected on the basis of the phenotype 4 . In a more recent study consisting of an Argentinian cohort of 38 patients with lissencephaly, SBH or PVNH, pathogenic variants were identified in 36% of cases 46 .
Pathogenic variants of ASPM are the most common genetic cause of primary microcephaly, with a mutation rate of 10-40% depending on ethnicity and the presence or absence of consanguinity 47,48 . Among consanguineous families, alterations in ASPM and WDR62 accounted for >50% of cases of primary microcephaly 49,50 .
For COB, mutation detection rates vary considerably, depending on the age at diagnosis and clinical inclusion criteria. For the most severe prenatal manifestations, the detection rate was usually >60% when the six genes most commonly linked to dystroglycanopathy were analysed 25 .

Gene panels
Despite multiple publications reporting on the yield of gene panels in cohorts of patients with neurodevel opmental disorders [51][52][53] , similar studies for MCDs are scarce. The only study that we identified reported on testing of a small gene panel (ten genes) in 158 individ uals with brain malformations, including 30 individuals with SBH, 20 with megalencephaly, 61 with PVNH and 47 with pachygyria. Causal pathogenic variants were found in 27 individuals (17%, range 10-30% depending on the phenotype) 54 .
Several genes encoding components of the PI3K-AKT-mTOR pathway have been implicated in FCD, and targeted testing of PI3K-AKT-mTOR pathway volume 16 | November 2020 | 623 NATure revIeWS | NeuROlOGy genes, using highly sensitive sequencing methods that allowed detection of lowfrequency brain somatic var iants, produced diagnostic yields ranging from 12% to 40% [55][56][57] . In a different cohort, a targeted NGS panel that included the most commonly mutated PI3K-AKT-mTOR pathway genes uncovered PIK3CA pathogenic variants in 50 of 131 individuals (174 samples) with the megalencephaly-capillary malformation syndrome 58 .

Exome sequencing
One study investigated the yield of exome sequencing, combined with CMA, in 54 patients with various MCD types 5 . This approach yielded a definitive (9/16) or pre sumptive (7/16) molecular diagnosis in 16 of 54 enrolled individuals (30%). Another study of 62 patients with microcephaly followed a similar approach and identified causative variants in 48% of the individuals 39 .

Neuro-MIG laboratories
We have also analysed the yield from the diagnostic lab oratories within the NeuroMIG network. Targeted gene panels resulted in diagnostic yields of 15-37%, although wide variability was observed among the different clinical subtypes ( TaBle 2). The combination of expert evaluation of MRI scans followed by targeted analysis of the most plausible causative variants can considerably increase the diagnostic yield. Substantiating this point, the availability of MRI scans resulted in an improved mutation detection rate of 37% in a mixed cohort of 117 patients with MCDs, compared with only 18% in a cohort of 784 patients analysed without previous expert reevaluation of MRI scans at the Human Genetics Center Regensburg (U.H., unpublished work). In the former cohort, the testing strategy was selected by the laboratory depending on the MRI pattern, and the approaches included single gene, panel and exome sequencing. A similar trend was noted in the Department of Clinical Genetics, Erasmus MC University Medical Center, Rotterdam, where the diagnostic yields from inhouse requests accompa nied by expert MRI review by G.M.S.M. were almost double those from the tests ordered from other med ical specialists outside the university hospital (M.W., unpublished work).

In utero infections
Prenatal infections can cause extensive damage to the fetal brain, including the cerebral cortex [59][60][61] . Cytomegalovirus (CMV) is one of the most frequent nongenetic causes of MCDs and is specifically asso ciated with polymicrogyria, intracranial calcifications, white matter abnormalities and microcephaly 1 . In a cohort of 26 patients with bilateral polymicrogyria, six (31%) tested positive for CMV; however, it was unclear whether these patients were infected prenatally or postnatally 62 . In a larger group of 50 patients with poly microgyria, six (12%) tested positive on Guthrie cards (W.B.D., unpublished work).

New recommendations
The NeuroMIG network recommends that a con certed effort be made to reach an aetiological diagnosis in every individual with an MCD. The diagnosis serves several functions. First, it explains the cause of the mal formation, ends the diagnostic odyssey and prevents further unnecessary investigations. Second, it provides information on prognosis and recurrence risk for the patient and family members 4 . Third, it aids the predic tion of treatment outcomes; for example, the success rate for epilepsy surgery depends on the underlying genetic cause 75 . Fourth, it directs patient management (for example, antiviral treatment and screening for pro gressive hearing loss in infants with congenital CMV infection 76 , cardiovascular surveillance in FLNArelated and ARFGEF2related PNVH 77,78 or mTORC1 inhibition in patients with tuberous sclerosis complex (TSC)) 79 . Fifth, it enables natural history studies 80,81 and targeted research into personalized therapy and prevention 82,83 .
Imaging findings, such as generalized versus focal and bilateral versus unilateral malformations, cannot reliably distinguish genetic from nongenetic causes, and the diagnostic yield of targeted testing is determined to a large extent by the availability of a multidisciplinary expert evaluation. However, such an ideal setting can rarely be met in practice. Therefore, we have formu lated a general diagnostic workflow that can be applied in most clinics to any individual with an MCD (Fig. 2). Lists of currently known MCDassociated genes are pre sented in Supplementary Tables 2 and 3. These lists can assist variant interpretation and guide targeted testing if exome (or genome) sequencing is not available. These general recommendations should minimize the chance of missing a known causative variant. The workflow can be started when a person is first diagnosed with an MCD, although clinicians should check whether any of the investigations have already been performed. For some MCD subtypes, the most costeffective strategy would be targeted gene analysis, but the success of this approach depends greatly on accurate pattern rec ognition. The relevant subtypespecific patterns and aeti ologies are outlined in the section 'Phenotypespecific considerations' below.
The correct interpretation of genetic test results requires detailed phenotypic analysis, including reevaluation of the brain MRI, to confirm that the identified single nucleotide variant (SNV) or CNV fully explains the phenotype. In the case of a negative result, the reevaluation should help determine whether the malformation was correctly classified, whether additional diagnostic testing, such as deep sequencing or analysis of a different tissue, might be helpful, and whether a nongenetic cause is more likely.
We recommend that a final clinical interpretation is done by a qualified medical geneticist, preferably after an interdisciplinary discussion with a molecular genet icist, neuroradiologist and/or neurologist. Unusual cases can be presented at an expert review session. Selected case reports demonstrating the importance of phenotypeguided interpretation of the test results are summarized in Supplementary Box 1.

Strategy if no diagnosis is reached
If no diagnosis has been reached after the general workflow has been applied, several strategies can be considered. Patients with an MCD pattern that is known to be highly specific for one or a few genes could bene fit from visual inspection of NGS reads and/or alter native targeted sequencing methods such as Sanger sequencing complemented by deletion/duplication testing of genes of interest 84 , as outlined in the section 'Phenotypespecific considerations' below. Review of NGS data might reveal inadequate coverage of the genes of interest, or the filtering out of potentially relevant splice site or flanking intronic sequences.
If not performed previously, karyotype analysis should be considered in undiagnosed patients with MCDs (86% consensus from the NeuroMIG network). Balanced translocations and ring chromosome abnor malities are a rare cause of MCDs but have occasionally been described 35,85 .
Patients from consanguineous pedigrees and fami lies with multiple affected siblings might benefit from a single nucleotide polymorphism microarray analysis to identify regions of homozygosity. If a homozygous region contains a known MCDrelated gene that is com patible with the phenotype, special attention must be given to the known deep intronic variants [86][87][88][89] (listed in  Supplementary Table 3).
Metabolic investigations should be considered in patients with microcephaly, polymicrogyria or COB, as a broad range of metabolic diseases, including peroxi somal disorders, glutaric aciduria, fumarase deficiency and Dbifunctional protein deficiency, can manifest with cortical malformations resembling these MCD patterns 1 .
In patients with unexplained MCDs and muscle weak ness and/or elevated creatine kinase, a muscle biopsy might be considered to allow specific analysis for dys troglycanopathies and mitochondrial disorders. The results of muscle biopsy allied to characteristic brain imaging findings in the CNS may help to indicate the affected gene 90 .
Some patients might benefit from repeat brain imaging, especially if the first MRI scan was performed before completion of myelination (3 months to 2.5 years of age) or was of low quality (for example, low resolu tion, or inadequate exploration of the brain according to the axial, coronal and sagittal plan and/or inadequate sequences). Occasionally, brain MRI scans of the par ents can identify a previously unrecognized familial malformation syndrome 41,91,92 .
Autopsy represents an important final procedure in deceased patients with unexplained MCDs as it can pro vide additional information that cannot be obtained dur ing life 93 . Also, after brain surgery, DNA can be extracted from affected brain tissue to identify somatic patho genic variants. Specific protocols are recommended for the evaluation of perinatal and postnatal brain tissue, including both frozen and fixed tissue samples from key brain regions (that is, regions that are vulnerable to epilepsyrelated damage) to identify specific structural abnormalities and rule out other pathologies 94 .
Finally, patients without a diagnosis should be con sidered for triobased wholegenome sequencing and RNA sequencing, preferably within a large collabo rative research network to allow rapid discovery of novel causative variants, noncoding variants in regulatory elements and epigenetic variations [95][96][97] .

Recurrence risk and genetic counselling
Only when the cause of the MCD is known can an accu rate recurrence risk be provided to the patient and their family. When the cause is unknown, an attempt should be made to provide an empirical risk figure. This figure depends on the type of malformation, clinical presenta tion and the causes that have been reliably excluded (TaBle 3). We should point out that empirical risk coun selling requires very high confidence in correct MRI interpretation and recognition of the specific phenotype.

Phenotype-specific considerations
Microcephaly. The aetiology of microcephaly is hetero geneous and includes both genetic and nongenetic factors. Nongenetic causes, including intrauterine tera togen exposure (for example, alcohol or drugs), congen ital infections and perinatal and postnatal brain injuries (placental insufficiency, birth complications, postnatal infarcts and concussions), account for almost 30% of microcephaly cases. Recognized genetic causes include chromosomal aneuploidies, CNVs, some of which are submicroscopic, and a rapidly growing number of single gene disorders (reviewed by Pirozzi et al. 11 ). Accurate perinatal historytaking aids the identification of teratogen exposure and infections, although a negative history can never reliably rule out these causes. Brain scans should be scrutinized for signs of fetal injury, including gliosis, cysts and calcifications. Clinicians should be aware that cortical malformations, espe cially polymicrogyria, can also be caused by fetal injury (see also below). Recurrence in the family, dysmorphic features and congenital abnormalities outside the CNS can be indicative of a genetic cause.
Ophthalmological abnormalities are found in up to 48% of patients with microcephaly 98,99 , including chorioretinal lacunae in Aicardi syndrome, chorio retinopathy in KIF11related microcephaly, micro phthalmia and cataract in Warburg Micro syndrome and cerebrooculofacioskeletal syndrome, chorioret initis after in utero CMV or toxoplasmosis infection, and a wide spectrum of abnormalities of the macula, retina and optic nerve after in utero Zika virus infec tion. Therefore, a detailed eye examination should be routinely performed in every individual with micro cephaly so that appropriate support and diagnostics can be implemented.

Megalencephaly.
Examination of an individual with megalencephaly should include an assessment of whether the malformation is confined to the brain or whether it is associated with a generalized or seg mental overgrowth syndrome. Careful assessment of serial height, weight and OFC measurements is helpful, as is examining the body for any asymmetries and skin abnormalities. Overgrowth usually manifests within the first 2 years of life 100 . Currently, >20 generalized over growth syndromes are known (reviewed elsewhere 100,101 ). Distinctive facial features can also aid identification of the underlying syndrome.
Generalized overgrowth syndromes are most often caused by germline gene mutations or CNVs, which can be identified with the standardized workflow. By contrast, segmental overgrowth syndromes and some isolated megalencephaly syndromes are caused by somatic mutations that might elude detection by standard workflows. To increase the chance of iden tifying the diseasecausing variant, it might be neces sary to sequence DNA derived from affected tissue (for example, skin or brain specimens) instead of blood. Further details of this approach are provided in the section 'Detecting mosaic variants' below. Several over growth syndromes, as well as the PTEN hamartoma tumour syndrome, are associated with an increased risk of malignancies.
An increasing number of defects in genes involved in cell growth and proliferation pathways are being identified in megalencephaly. The affected pathways and molecules include the PI3K-AKT-mTOR and RAS-MAPK-ERK pathways, DNA methyltransferases, transcription initiation regulators and receptor tyrosine kinases 11,102,103 . In our experience, PI3K-AKT-mTOR pathwayassociated megalencephaly is often ≥3 s.d. above the mean. Mutations in this pathway can cause either isolated or syndromal megalencephaly, with other fea tures including somatic (body) overgrowth and/or other MCDs, including polymicrogyria 104,105 . Given the high prevalence of mosaicism in these disorders, a tailored approach is recommended (see below).
Lissencephaly spectrum. The lissencephaly imaging classification was updated in 2017 and now includes 21 patterns 17 . Lissencephaly is considered to be an exclu sively genetic disorder 40 , with 28 genes currently known to be associated with this condition (Supplementary Table 3). Four lissencephaly patterns are highly specific for pathogenic variants in one or two genes, with diag nostic yields >90% 40 . The first pattern is diffuse agyria with cortical thickness >10 mm, which is caused by LIS1 and DCX variants. The main cause in this group is a microdeletion at chromosome 17p13.3, the LIS1 locus, which can cause isolated lissencephaly, or Miller-Dieker syndrome in the case of a larger deletion 40 . The second specific pattern is occipital agyria combined with frontal pachygyria, which is primarily associated with deletions and pathogenic variants in LIS1, but also in rare cases with TUBG1 variants and TUBA1A variants affecting codon Arg402. The third pattern is pachygyria with a cortical thickness of 5-10 mm, most prominent over the temporal lobes, combined with complete agenesis of the corpus callosum and severe hypomyelination. This pattern is caused by ARX pathogenic variants. Note that pathogenic variants in DYNC1H1 have been linked to a similar lissencephaly pattern but without hypomyelination. The fourth pattern, diffuse SBH with a band thickness >5 mm, is a pathognomonic pattern strongly associated with pathogenic variants in DCX in both women and men 40 . Posteriorpredominant SBH is associated with mild or mosaic LIS1 mutations 40 .
No other genes have been associated with these pat terns. Therefore, a negative test result for those genes in a patient with a specific phenotype should prompt an offer to the family to participate in a research project focusing on gene discovery.

Periventricular nodular heterotopia.
PVNH is associ ated with numerous CNVs and single gene mutations and can be part of a complex syndromic disorder, such as van Maldergem syndrome, Donnai-Barrow syn drome, Au-Kline syndrome or Noonanlike syndrome with loose anagen hair 37 . Proteins encoded by the genes associated with PVNH are involved in several cellular and molecular mechanisms, including the formation of the radial glial scaffold, cell-cell adhesion and vesicle trafficking. In addition, dysregulation of PI3K-AKT-mTOR or SMAD2/3 signalling pathways, RNA process ing or transcriptional regulation has been reported in people with PVNH [106][107][108] . At least 20 genes have been associated with this condition (Supplementary Table 2).
FLNA mutations are an important monogenic cause of PVNH and, owing to a substantial risk of car diovascular and other organ complications, identifi cation of FLNArelated disorders is of great clinical importance 77,109 . Although no single feature is pathog nomonic, several features should raise suspicion of an FLNA mutation, including female sex, with or without a positive family history that follows an Xlinked dom inant pattern; absence of overt intellectual disability, although learning difficulties, dyslexia and/or psychi atric problems can be present 110,111 ; bilateral clusters of confluent nodules extending along the walls of the frontocentral lateral ventricles (classic PVNH) 44 ; and the presence of a retrocerebellar cyst or mega cisterna magna 44,110 . Less frequently, corpus callosum hypo plasia, inward rotated anterior ventricular horns, white matter abnormalities and/or focal cortical abnormali ties can be observed 77,110 . Systemic involvement is not an obligatory feature but can be present, leading to cardiovascular abnormalities such as patent ductus arteriosus, aortic aneurysm and cardiac valvular dystro phy; obstructive lung disease; constipation; coagulop athy; joint hypermobility; and other connective tissue abnormalities 77,109,110 .
In individuals with one or two single nodules, normal cognitive functioning and no other congenital abnor malities, the yield of genetic testing is low. However, these individuals can harbour mosaic FLNA mutations that might be passed on through the germline to their offspring 44 .
Posteriorpredominant PVNH is a common pat tern that is often associated with overlying poly microgyria and/or subcortical heterotopia, as well as abnormalities of the fossa posterior, corpus callosum and/or hippocampus 112 . This pattern can be caused by a microdeletion of chromosome 6q27, but has also been associated with fetal brain injury 36,113 .
Subcortical heterotopia. Several rare, mostly symmetri cal bilateral forms of SUBH have a genetic origin, usu ally with an autosomal recessive mode of inheritance. Extensive brain involvement is seen in the mesial paras agittal form associated with Chudley-McCullough syndrome, which results from biallelic variants in GPSM2, and ribbonlike heterotopia, in combination with agenesis of the corpus callosum and megalen cephaly, is observed in individuals with biallelic EML1 variants 114,115 . Another rare subtype affecting the peri trigonal regions has been observed in patients with variants in genes encoding a microtubule component (TUBB), a microtubulesevering protein that localizes to the centrosome and mitotic spindle during cell division (KATNB1), or a centrosomal protein with tubulindimer binding activity (CENPJ) 21 .
In parallel with the diverse morphology of SUBH, the aetiology of this condition is also very heterogene ous, and for certain subtypes is largely unknown. For example, no genetic cause has been identified for cur vilinear heterotopia, which is often asymmetric and can extend from the cortex to the ependyma 21,116 . However, a vascular disruptive cause has been suggested in several patients on the basis of a prenatal history of twinning, near miscarriage or trauma [117][118][119][120] , and some cases are hypothesized to result from postzygotic mutations 21 .
Polymicrogyria. The aetiology of polymicrogyria can be either genetic or disruptive 27 , and our new clinical workflow has been designed to make the physician aware of potential pitfalls. Despite extensive workup, including genomic testing, the underlying aetiology of polymicrogyria often remains unknown.
In a substantial proportion of patients, polymicro gyria has a genetic aetiology. Various CNVs, in par ticular, 22q11.2 and 1p36 deletions, have been linked to this condition, along with a rapidly growing number of monogenic causes, including several metabolic disorders (Supplementary Table 2). Dozens of genes implicated in different pathways or groups of related disorders, includ ing the mTORopathies (affecting the PI3K-AKT-mTOR pathway), the tubulinopathies and the RABopathies, have been associated with polymicrogyria 121 .
A common cause of polymicrogyria is a congenital CMV infection, which is thought to account for 12-30% of cases, or even more among patients with specific white matter changes 62,64 . Congenital CMV infection should be suspected if polymicrogyria is observed in the presence of clinical features such as microcephaly and congenital sensorineural hearing loss. Imaging features suggestive of congenital CMV, besides polymicrogyria, include white matter hyperintensities and intracranial calcifications 62,64,122 . Toxoplasmosis, syphilis, varicella zoster virus and Zika virus have also been associated with polymicrogyria 27,60 . Additional nongenetic causes include vascular disruptive events during pregnancy and, according to a few reports, maternal ergotamine use 123 . Twinning is also a risk factor for polymicrogyria, par ticularly in the case of death of a monozygotic cotwin, and in some cases of twintotwin transfusion syndrome, in which the donor twin is most commonly affected 124 . The association with twinning is proposed to be related to vascular disturbance and/or hypoperfusion 125 .
Dysmorphic features, multiple congenital abnormali ties, megalencephaly and microcephaly are all indicative of a genetic cause, although the latter condition can also be associated with congenital infection. Evaluation of head circumference is an essential part of the clinical workup and could assist with variant interpretation, as several genes are specifically associated with microceph aly or megalencephaly 121 . The bestknown gene associ ated with polymicrogyria and microcephaly is WDR62, and germline or somatic variants in genes encoding components of the mTOR pathway, such as PIK3CA and PIK3R2, are usually associated with megalenceph aly, often with other abnormalities such as vascular skin lesions and digital anomalies 121 . Calcifications on brain imaging are indicative of fetal brain injury (dystrophic calcification). However, COL4A1 and COL4A2 patho genic variants can genetically predispose to fetal vascu lar injuries, and the pseudoTORCH syndrome mimics congenital infection 126,127 .
Polymicrogyria can be associated with peroxisomal disorders such as Zellweger syndrome or Dbifunctional protein deficiency, and is reported in up to 65% of patients with the latter condition 128 . A peroxisomal dis order should be suspected if a child with polymicrogy ria is unusually sick for an individual with a static brain malformation, particularly in the neonatal period or early infancy. Additional abnormalities might be found, including dysmorphic features, hepatomegaly and pro found hypotonia. In addition to polymicrogyria, brain MRI will usually show severe leukoencephalopathy 129 . If a peroxisomal disorder is suspected, plasma levels of very long chain fatty acids (VLCFAs) should be checked, and further investigations such as skin fibroblast enzymatic analysis or genomic testing should be initiated.
The workup of a patient with polymicrogyria first requires astute clinical assessment and review of the brain MRI scan. If CMV is suspected, attempts should be made to retrieve the Guthrie neonatal blood spot for CMV PCR. VLCFA analysis should be requested if a peroxisomal disorder is suspected. CMA remains the first tier of genomic analysis. Although many genes have been associated with polymicrogyria, the yield of stand ard genomic testing is generally ~20% (unpublished work from NeuroMIG laboratories). Deep sequencing might be required to identify mosaic variants, especially in patients with megalencephaly. However, patients with mosaic PIK3R2 mutations and normal OFC have been reported.
Cobblestone malformation. All currently known COB syndromes are genetic and inherited in an autosomal recessive mode. A major group is the dystroglycan opathies, which are linked to various genes required for Oglycosylation of αdystroglycan (Supplementary  Table 1). Patients often have muscular dystrophy with markedly elevated serum creatine kinase levels. Moreover, eye involvement, such as severe myopia or structural malformations, is frequently observed.
Recurrent biallelic microdeletions at the ISPD locus are the most common cause of dystroglycanopathies. Other COB syndromes include laminopathies, congenital dis orders of glycosylation and basement membrane trans migration disorders (reviewed by Dobyns et al. 27 ). At the imaging level, COB can be difficult to distinguish from polymicrogyria 27 , but creatine kinase analysis and/or an ophthalmological examination can potentially guide the clinical diagnosis 25 .
Differentiation of COB syndromes from polymicro gyria might be especially challenging on lowresolution images and at a young age when myelination is still ongoing (from 3 months to 2 years of age). Useful distin guishing characteristics include the intracortical striations that appear at regular intervals vertical and perpendicular to the grey-white matter border in COB and that differ from the chaotic striations seen in polymicrogyria 27 . Other structural malformations that can cooccur with COB include hydrocephalus, brainstem hypoplasia and cerebellar cysts. The white matter might show an abnor mal MRI signal and small cysts. However, what clearly appears as polymicrogyria on MRI can present as typical neuronal overmigration on microscopic examination, suggesting that COB and polymicrogyria have a common pathogenesis 130 .
Tubulinopathies. Tubulinopathy is caused by heterozy gous missense variants in any one of six tubulinencoding genes, TUBA1A, TUBB2A, TUBB2B, TUBB3, TUBB and TUBG1. The variants probably exert dominantnegative effects on microtubule assembly and/or function. Although several pathogenic variants are recurrent, many patients harbour a unique variant, which can be difficult to confidently classify as pathogenic without functional studies 131 .
The tubulinopathies present with highly heterogene ous yet very recognizable patterns of brain malforma tions. The presence of a typical tubulinopathy pattern can be helpful in the interpretation of variants of uncer tain significance (VOUS) 131 . Abnormalities of the cor tex can be obvious or subtle, and the range encompasses microlissencephaly, pachygyria with a cortical thickness >10 mm, pachygyria with a 5-10 mm thick cortex (often more prominent in the perisylvian regions), polymicro gyria, dysgyria and a simplified gyral pattern 17,30,131,132 . The basal ganglia are usually dysmorphic, including an enlarged caudate and absent or diminutive anterior limb of the internal capsule (dividing the caudate from the putamen), resulting in a fused striatum that in turn gives the frontal horns of the lateral ventricles a characteristic 'hooked' appearance. Callosal abnormalities (partial or complete agenesis of the corpus callosum), ventricu lomegaly, vermian dysplasia with 'diagonal' folia (folia crossing the midline at an oblique angle), cerebellar hypoplasia and asymmetric hypoplasia of the brainstem might also be seen 30,31,131,133 . TUBB3 pathogenic variants can cause an ocular motility disorder, known as congen ital fibrosis of the extraocular muscles type 3, with or without MCD or axonal polyneuropathy 132 .
Pathogenic variants in DYNC1H1 and KIF2A, which encode microtubuleassociated motor proteins, also lead to a spectrum of MCDs, ranging from pachygyria volume 16 | November 2020 | 629 NATure revIeWS | NeuROlOGy to dysgyria. Similar to the tubulinopathy spectrum, most individuals demonstrate a large caudate and ver mian hypoplasia. DYNC1H1 variants can be associated with peripheral nerve disease ranging from fetal aki nesia to spinal muscular atrophy with lower extremity predominance 134 .
FCD and hemimegalencephaly. Somatic and/or germline variants in numerous PI3K-AKT-mTOR pathway genes, including TSC2, TSC1, MTOR, PIK3CA, AKT3, RHEB, DEPDC5, NPRL3 and NPRL2, are known to be associated with malformations within the FCDhemimegalencephaly spectrum 55,[135][136][137][138][139] . TSC encompasses a wide spectrum of severity and clinical presentation, including FCD, and the diagnosis has consequences for surveillance and treatment 79 . In people who present with FCD, the skin and MRI should be checked for manifes tations such as hypomelanotic macules, shagreen patch, additional FCD foci and subependymal nodules. If any of these features are present, a full diagnostic workup including TSC1/TSC2 testing is recommended 140 . Germline pathogenic variants in the GATOR1 complex genes DEPDC5, NPRL2 and NPRL3 are associated with focal onset seizures with or without FCD on imaging.
In families with epilepsy in particular, these genes should be carefully checked for SNVs and CNVs that segre gate in an autosomal dominant pattern with reduced penetrance [141][142][143] . Twohit models involving germline plus somatic variants in TSC2 and DEPDC5 have been proposed to explain the aetiology of TSCassociated FCD and isolated FCD type IIA 141,142,144 . In recent years, somatic mutations in SLC35A2, which encodes an enzyme involved in glycosylation, have been found in focal epilepsy specimens and seem to be specific to FCD type I 137,145,146 . Analysis of resected brain tissue using deep sequencing and singlecell techniques might be required for detection of somatic mutations.

Cerebrovascular disorders associated with MCDs.
Prenatal and postnatal cerebrovascular events can lead to ischaemic and disruptive brain malformations, including schizencephaly, polymicrogyria, intracranial calcifica tions, cysts and porencephaly. Disorders with a vascu lar and/or inflammatory basis, such as familial stroke, pseudoTORCH syndrome, Aicardi-Goutières syn drome, leukoencephalopathy with cortical cysts, and cer ebral microangiopathy syndromes with calcifications and cysts, can cause damage to the developing brain. A case series of 119 individuals with intracranial calcifications revealed a specific diagnosis in 50% of the cases 147 . Of these, 33 had Aicardi-Goutières syndrome, 6 had OCLN-related pseudoTORCH syndrome and 3 had a COL4A1-related disease. Pathogenic variants in USP18 have been associated with cerebral haemorrhage in utero, leading to polymicrogyria 148 . However, polymicrogyria is a rare feature in cerebrovascular disorders.
Several reports have shown porencephaly, schizen cephaly, polymicrogyria and PVNH associated with COL4A1 pathogenic variants, which cause imbal ance or structural distortion of the collagen IV triple helix 126,149,150 . Evidence for a link between COL4A2 and MCDs is weaker, although, considering the functional interactions between the two collagen IV proteins, COL4A1 and COL4A2 should be tested together 149 . Despite reports of EMX2 as a 'schizencephaly gene' , evi dence of a role for EMX2 mutations in schizencephaly is lacking 151,152 .
A list of genes that have been associated with earlyonset and often severe cerebrovascular phenotypes is provided in Supplementary Table 4.

Laboratory requirements
Chromosomal microarray analysis A survey within the NeuroMIG network, which was conducted in preparation for this Consensus Statement, indicated that multiple different microarray platforms can be used, with no specific technology showing a clear advantage.
When choosing CMA platforms for MCD diagnos tics, special attention should be paid to the exonlevel resolution of genes in which singleexon aberrations have been described (Supplementary Table 3). Single nucleotide polymorphism arrays have the advantage of detecting regions of homozygosity, thereby facilitating diagnostics in consanguineous families. Mosaic CNVs showing as little as 15-20% chromosomal mosaicism were successfully detected in patients with neurode velopmental disorders 153 . We anticipate that CMA will become redundant in the future as NGS costs further decrease and algorithms for CNV analysis from NGS data become more robust.

High-throughput sequencing
As MCDs constitute a genetically heterogeneous group of disorders and the number of known diseaseassociated genes is rapidly increasing, we strongly recommend genomewide testing approaches combined with tar geted evaluation of genes that are currently implicated in MCDs (the 'slice approach'). If the results of these tests are negative, the strategy can be expanded to a full trio exome analysis after appropriate genetic counselling. NeuroMIG network laboratories are applying various exome enrichment strategies with comparable efficiency across the platforms and compliance with published NGS guidelines 154,155 . Most current exome sequencing enrichment kits provide sufficient coverage to offer an MCD panel as a type A or type B test 154 . The terms type A and type B refer to the definitions from the cur rent guidelines for diagnostic NGS from the European Society of Human Genetics (ESHG), whereby the lab oratory guarantees >99% reliable reference or variant calls of the target regions (type A) or describes exactly which regions are sequenced at >99% reliable reference or variant calls (type B) 154 .

Variant calling and prioritization
Our experience shows that an average per base cover age of 100 reads with a minimum coverage of 30 reads is sufficient for reliable calls within coding and flank ing intronic regions. NeuroMIG network members preferentially use a variant calling threshold of 20% of the nonreference (alternative) reads and variant calling is performed within exons and 10 bp of the flanking intronic sequence (80% consensus). However, deep intronic variants affecting splicing have already been described in several MCDassociated genes (Supplementary Table 5). Such variants need to be con sidered in patients with highly suggestive phenotypes, but might require genome or targeted sequencing.
The described approach is applicable for the iden tification of constitutional (germline) and highgrade mosaic variants (>30% of cells). Special considerations regarding detection and validation of lowgrade mosaic variants are summarized in the section 'Detecting mosaic variants' below.
Supplementary Table 2 provides a curated list of the core MCDassociated genes, including information on the observed mutational spectrum and associated phenotypes. Supplementary Table 3 summarizes selected genes associated with syndromic, often postnatal micro cephaly. Microcephaly is a frequent accompanying fea ture of these conditions but is not a key manifestation. Genes associated with disorders that always present with microcephaly are listed in Supplementary Table 2. Taking into account the number of novel diseaseassociated genes that are emerging, we strongly suggest updating the gene lists according to the current literature every 6 months.
Variant interpretation follows the general rec ommendations of EuroGentest, the ESHG and the American College of Medical Genetics and Genomics (ACMG) 154,156 .
As all MCD entities are rare disorders, we recommend classifying a variant as benign if the allele frequency is >1% in the Genome Aggregation Database (gnomAD), which differs from the ACMG standalone evidence of benign impact with an allele frequency of >5% 157 . As the Neuro cohort of gnomAD includes individuals with neuropsychiatric disorders, which represent a rare manifestation of MCDs, one should consider excluding variants from this cohort when estimating gnomAD allele frequency, as pathogenic MCDassociated var iants might be present. The presence of a variant as a homozygous allele in multiple (at least five) individu als in gnomAD strongly suggests its benign impact and irrelevance for the phenotype. However, one should be careful to check that the variant is truly homozygous and not hemizygous, combined with a deletion of the second allele. The impact of a homozygous SNV might differ substantially from the impact of deletion of one allele and the same SNV on the remaining allele 158 .

Pitfalls in variant prioritization.
Inhouse variant data bases, which contain data from a single institution, are another important source to distinguish benign from potentially causative variants. However, some MCDrelevant genes, especially those encoding tubulin, which are prone to readalignment errors, might have high falsepositive inhouse frequencies. One TUBB2B pathogenic variant, Ala248Val 159 , was listed in gnomAD with an allele frequency of 3% but is currently flagged as failed -that is, probably an artefact -by random forest filters. However, when inhouse data are analysed, this variant might erroneously show up in control samples in up to 30% of the reads (K.S., unpublished work) and might, therefore, be filtered out as a 'frequent' inhouse variant, despite being pathogenic. On the basis of this example, we suggest that manual curation of inhouse variants in the tubulinencoding genes should include consideration of mapping quality and comparison of inhouse frequencies with the curated gnomAD data set. Sanger sequencing of TUBB2B could be consid ered in undiagnosed patients with an MCD pattern highly suggestive of a tubulinopathy. In the near future, such misalignment errors should be solved through highresolution mapping and application of longread DNA sequencing platforms 160 .
The presence of highly homologous pseudogenes also complicates accurate variant calling for a number of MCDrelevant target genes 161 (Supplementary Table 2).

Penetrance of MCD-associated variants.
With the exception of Xchromosomal genes such as ARX and DCX, and gene encoding components of the GATOR1 complex, variants in other MCDassociated genes seem to be fully penetrant, as carrier probands always show characteristic structural changes in the brain. However, individuals with these variants might be clinically asymptomatic and therefore never undergo brain MRI. In the case of inheritance of likely pathogenic variants from apparently unaffected parents, parental brain imag ing is essential for accurate variant interpretation 91,110 . Female carriers of the Xchromosomal variants might be clinically unaffected and have normal brain scans 43,162 . Incomplete and/or agerelated penetrance were reported for variants mainly associated with a seizure phenotype (for example, GATOR1 complex genes 163 ); therefore, variants inherited from unaffected parents might be considered causative.

Clinical laboratory report
The final laboratory report, including reporting of incidental findings, should follow the general require ments published by EuroGentest, the ESHG and the ACMG, as well as countryspecific guidelines for genetic laboratory reports.
If the review board includes a medical professional with sufficient expertise in MRI interpretation, we rec ommend that MRI scans should be presented together with the clinical information and relevant variants. Relevant clinical information and brain imaging are important for accurate interpretation of the variants and should be actively requested.
If parents and similarly affected siblings (if applica ble) were not analysed together with the index patient, segregation analysis must be strongly recommended in the final report.
All pathogenic and likely pathogenic variants (class 5 and class 4 variants, respectively) must be included in the final report. The final report should also contain all VOUS (class 3 variants) in MCDassociated genes. The laboratory should consider including proteinaltering de novo, homozygous or compound heterozygous rare variants in potentially relevant genes of uncertain sig nificance in the final report. The relevance for the MCD phenotype might be determined on the basis of the expression pattern of the gene or its potential importance for human brain evolution (humanspecific genes or volume 16 | November 2020 | 631 NATure revIeWS | NeuROlOGy transcripts). Despite the fact that most MCDassociated genes are evolutionarily conserved, primatespecific genes and isoforms should not be ignored as they can be linked to neurodevelopmental disorders 164 . The rele vance of such variants must be continually reevaluated over time.
Highresolution, singleexonlevel CNV analy sis is essential to complement the sequencing report. CNV analysis can be provided with different meth ods including CNV calling from NGS data if robustly established and validated, multiplex ligationdependent probe amplification, quantitative PCR or customized highresolution microarrays.
The final report must specify whether CNV analysis has been performed, including information about the genes analysed and methods used for the analysis. If no copy number analysis has been carried out, the report must contain information about the genes that require copy number tests.
If (likely) pathogenic variants or VOUS have been identified, patients and/or their families should be referred to a clinical geneticist for return of results and counselling on their clinical and prognostic implications.
We recommend sharing VOUS in the available data bases, such as ClinVar and the Leiden Open Variation Database. Depending on the local ethical and legal regu lations, some laboratories might choose to use different countryspecific databases.

Additional considerations Detecting mosaic variants
Mosaic (postzygotic somatic) mutations, including mutations in PI3K-AKT-mTOR pathway genes, as well as in DCX, LIS1, FLNA and TUBB2B, have been described in a wide range of MCDs 54 . Mosaic mutation variant detection requires dedicated deep sequencing and bioinformatics tools, as these variants are likely to be missed by standardcoverage exon sequencing, especially in bloodderived DNA 58 . When available, affected brain tissue is the recommended tissue for genetic testing. If this tissue is not available, the use of 'proxies' such as saliva or skinderived fibroblasts is recommended over lymphocytes when a mosaic dis order is suspected 58,165 . Ideally, multiple tissues from the same individual should be examined.
Reliable testing requires a targeted approach to spe cific loci, using a customized gene panel with ultradeep sequencing (for example, >1,000times coverage). A gene panel for PI3K-AKT-mTORrelated syndromes is provided in Supplementary Table 6. As a general rule, hybridizationbased assays offer superior performance over amplicon assays 166 . However, amplicon protocols with unique molecular identifiers during library prepa ration have also proved effective for detecting somatic mutations 167 . The variantcalling algorithm (percent age of nonreference allele reads) must be adapted for detecting lowgrade mosaic SNVs.
As falsepositive mosaic mutation calls can arise from many different sources, we strongly suggest confirma tion of every lowgrade mosaic variant using an orthog onal technology such as droplet digital PCR or a second independent round of ultradeep sequencing [168][169][170] .

Neuropathological work-up
Detailed neuropathological examination, biobanking and genetic testing are required after epilepsy surgery or autopsy in patients with MCDs, and also after sudden unexpected death in epilepsy, as individuals who die as a result of sudden unexpected death in epilepsy might have a previously undiagnosed MCD.
In 2016, the task force of neuropathology from the International League Against Epilepsy (ILAE) Commission on Diagnostic Methods published a con sensus standard operational procedure for collection and processing of cortical samples from patients with MCDs such as FCDs 94 . Whenever feasible, anatomically intact surgical neocortical samples should be obtained to allow systematic analysis to identify the affected area. Correct orientation of the cortical sample and determi nation of its relationship to neurophysiologically aber rant sites and MRI findings requires an interdisciplinary diagnostic approach with good communication between pathology, neurology, radiology and neurosurgical teams. Representative tissue should be apportioned for histol ogy and biobanking. Brain tissuederived DNA is often required for genetic diagnosis in FCD and hemimegal encephaly; thus, highly standardized tissue processing is recommended. A neuropathologist should be involved in the interpretation of the brain pathology, and molecular biologists (or pathologists) and geneticists should partici pate in the setup and analysis of the sequencing results 171 . A consensus protocol with details of how to best process resected brain specimens for somatic mutational analysis to detect mosaic variants for hemimegalencephaly, FCD types I and II, and other MCDs is under development by a task force of the ILAE (E.A., unpublished work).
Analysis of lowlevel mosaic mutations, such as those reported in FCDs 57,136,142,172 , requires careful selection of brain regions and cells to ensure enrichment of the mutated cells, followed by deep sequencing 136,138 . A study published in 2019 used resected brain tissue from a large cohort of patients after epilepsy surgery to explore the possibility of detecting lowlevel somatic mutations in unmatched formalinfixed paraffinembedded (FFPE) brain tissue samples (that is, brain samples without a blood sample from the same patient). FFPE samples often represent the most relevant samples in the standard neuropatho logical diagnostic approach to MCDs 146,173 . The research ers showed that deep sequencing, even when applied to unmatched FFPE brain tissues, can be used to accurately and efficiently detect lowlevel somatic mutations.

Conclusions
In this Consensus Statement, we present a diagnostic workup for individuals affected by brain malformations within the MCD spectrum, encompassing current best practices and recommendations based on the consen sus of a multidisciplinary group of international experts within the NeuroMIG network. With this approach, we aim to increase diagnostic yield, thereby improving patient care and management worldwide and facilitat ing the development of targeted therapeutic approaches in the long term.
Published online 7 September 2020