Polymicrogyria is associated with pathogenic variants in PTEN

Abstract

Objective:

Congenital structural brain malformations have been described in patients with pathogenic PTEN variants, but the frequency of cortical malformations in PTEN variants and their impact on clinical phenotype are not well understood. Our goal was to systematically characterize brain malformations in patients with PTEN variants and assess the relevance of their brain malformations to clinical presentation.

Methods:

We systematically searched a local radiology database for patients with pathogenic PTEN variants who had available brain magnetic resonance imaging (MRI). MRIs were reviewed systematically for cortical abnormalities. We reviewed EEG data and evaluated the electronic medical record to for evidence of epilepsy and developmental delay.

Results:

In total we identified 22 patients with PTEN pathogenic variants for which brain MRIs were available (age range 0.4 years – 17 years). Twelve among these 22 (54%) had polymicrogyria (PMG). Variants associated with PMG or atypical gyration encoded regions of the phosphatase or C2 domains of PTEN. Interestingly, epilepsy was present in only 2 of the 12 patients with PMG. We found a trend toward higher rates of GDD, ID, and motor delay in individuals with cortical abnormalities, though cohort size limited statistical significance.

Interpretation:

Malformations of cortical development, PMG in particular, represent an underrecognized phenotype associated with PTEN pathogenic variants and may have an association with cognitive and motor delays. Epilepsy was infrequent compared to the high risk of epilepsy in patients with PMG reported previously.

Introduction

Phosphatase and tensin homologue (PTEN) pathogenic variants have been identified in association with several clinical syndromes that are distinct yet have overlapping features of aberrant growth leading to macrocephaly or macrosomia and a susceptibility to tumor formation. Examples include Bannayan-Riley-Ruvalcaba (BRR) syndrome (OMIM 153480), characterized by macrocephaly, developmental delay, vascular malformations including hemangiomas, and hyperpigmented macules of the glans penis¹; Cowden syndrome (OMIM 158350), characterized by macrocephaly, hamartomas, and increased risk of breast, thyroid, and endometrial cancer; and macrocephaly/autism syndrome (OMIM 605309), characterized by macrocephaly, abnormal facial features, and delayed psychomotor development². There is clinical overlap between these syndromes, which fall under the general classification of PTEN Hamartoma Syndrome (PTHS).

Patients with PTEN variants have been shown to harbor specific brain imaging characteristics reflecting abnormalities in cortical development. Prior studies have noted enlarged perivascular spaces and periventricular white matter abnormalities^3,4. A separate study evaluating characteristics of PTEN-associated disorders reported that 2 of 14 patients undergoing MRI had Chiari Type I malformation but no other associated structural abnormalities, though more than half had associated systemic vascular anomalies⁵. A more recent study used quantitative evaluation of brain MRI in individuals with PTHS to demonstrate the frequent finding of megacorpus callosum, present in 9 of 12 patients (75%) and the malformation of cortical development polymicrogyria (PMG) in 4/12 (33%)⁶. Malformations such as PMG have clinical implications, such as epilepsy, that have not been explored specifically in the context of PTEN germline variants.

PMG is a characteristic brain folding pattern that has been associated with numerous genetic alterations as well as non-genetic or acquired causes. Genes that encode proteins of the PI3K phosphoinositide-3-kinase (PI3K) pathway, such as AKT3 and PIK3CA, have been implicated^7–9. PMG has also been associated with disorders linked to tubulin genes, transcriptional regulators, and numerous other genetic causes of abnormal neuronal migration⁷. Clinically, seizures have been reported in up to 78% of patients with PMG of any cause⁷.

PTEN encodes a tumor suppressor on chromosome 10q23.31, which catalyzes the degradation of phosphatidylinositol-(3,4,5)-triphosphate generated by PI3K, inhibiting downstream activation of PI3K pathway targets. When PTEN is mutated, this PI3K inhibition is reduced, leading to downstream activation of cellular growth pathways, angiogenesis, and cell differentiation¹⁰. Despite the known regulation of PI3K by PTEN and the known role of PI3K in PMG, there have been no systematic studies of the association between PTEN variants and PMG or other cortical abnormalities.

We performed a systematic review of 22 individuals who harbor variants in PTEN at a single large pediatric center. We report the genetic characterization, standardized review of MRI characteristics, review of electroencephalography (EEG), and neurodevelopmental characteristics of this cohort. We specifically assessed the rates of brain malformations, and in those with brain malformations, the presence of epilepsy and developmental disabilities.

Methods

This study received prior approval by the Boston Children’s Hospital (BCH) Institutional Review Board.

Patient ascertainment

We queried the BCH Radiology database (Nuance mPower, Nuance Communications, Inc.) for dictations including the term “PTEN.” Each identified case was manually reviewed for the availability of brain MRI and genetic confirmation of a PTEN variant. Individuals for whom genetic results could not be confirmed were removed from the study. Five individuals (3 who overlapped with radiological ascertainment) were identified through the BCH Brain Development and Genetics (BrDG) Clinic: patient numbers 2, 3, 4, 10, and 12. A total of 22 individuals with PTEN variants were included in this study.

MRI evaluation

Brain MRIs were systematically reviewed for the presence of PMG by an experienced, board-certified pediatric neuroradiologist (EY). We annotated individuals based upon their MRI characteristics as those having PMG (Patients 1–12) versus those without PMG, or with atypical gyration (Patients 13–22). Cases with increased gyral frequency and distortion of surface anatomy were scored as positive for polymicrogyria if these findings were observed in multiple planes, and as atypical gyration if suggestive of PMG in only one imaging plane. PMG was further classified by location (frontal F, parietal P, perisylvian Ps). Other malformations of cortical malformation and other structural abnormalities were also noted if present (e.g., callosal dysgenesis, developmental venous anomalies, white matter changes). Motion-free studies with definition of the cortical ribbon comparable to our best quality studies (usually 3T MRIs) were designated high quality studies; studies which had deficiencies in resolution, motion, or signal to noise that degraded diagnostic confidence were designated as lower quality.

Clinical presentation

Developmental history, epilepsy history, and physical examination were assessed directly during clinical encounters for the 5 patients seen by study authors. For the other patients, assessment was based on chart review. In particular, developmental assessments were limited to descriptions provided by review in developmental medicine, genetics, or neurology clinic. EEG or sleep study data were reviewed by a trained epileptologist for all individuals who had undergone these studies. To assess intellectual disability and/or global developmental delay (ID/GDD), chart data was reviewed by a neurodevelopmental specialist. ID was defined as full scale IQ score <70. For individuals for whom objective neuropsychological data were not available, a developmental quotient was estimated for clinical determination of ID/GDD.

Classification of PTEN variants

PTEN variants were classified regarding pathogenicity based on consensus recommendations of the American College of Medical Genetics and Genomics (ACMG)¹¹. References used for classification are detailed in Table S1.

Targeted Sequencing

Targeted sequencing was undertaken for 3 individuals. Gene capture was performed with molecular inversion probes (MIP) spanning across exomes of 41 genes previously associated with PMG (See Table S2 for full list of genes). For the MIP design, custom scripts incorporating the MipGen¹ tool were used for dense tiling of >98% of all targeted bases with an average of at least 2 unique MIPs. MIP pool was amplified with low cycles (17 cycles), high-fidelity polymerase and custom common primers. Sequencing libraries were generated by hybridization of MIPs with 250ng of DNA for 24 hours. Hybridized MIPs were then filled in and ligated, and linear DNA was removed. Captured products were amplified using 15 cycles of PCR with custom 8nt indexing primers and sequenced on the Illumina HiSeq platform with 2 × 150bp paired end reads. The paired end reads were mapped to the hg19 human genome build using default settings in BWA-mem. All mapped BAMs were processed for germline mutations using default settings within GATK 3.7 Haplotype caller. The resulting joint-called VCF file was annotated using custom Annovar scripts and databases for population allele frequency filtration and missense prediction databases.

Exome Sequencing

For 2 individuals, whole exome sequencing (WES) and data processing were performed by the Genomics Platform at the Broad Institute of MIT and Harvard. Libraries from DNA samples (>250 ng of DNA, at >2 ng/ul) were created with an Illumina Nextera or Twist exome capture (~38 Mb target) and sequenced (150 bp paired reads) to cover >80% of targets at 20x and a mean target coverage of >100x. Sample identity quality assurance checks were performed on each sample. The exome sequencing data was de-multiplexed and each sample’s sequence data were aggregated into a single Picard BAM file. Exome sequencing data was processed through a pipeline based on Picard, using base quality score recalibration and local realignment at known indels. The BWA aligner was used for mapping reads to the human genome build 38. Single nucleotide variants (SNVs) and insertions/deletions (indels) were jointly called across all samples using Genome Analysis Toolkit (GATK) HaplotypeCaller package version 3.5. Default filters were applied to SNV and indel calls using the GATK Variant Quality Score Recalibration (VQSR) approach. Annotation was performed using Variant Effect Predictor (VEP). Lastly, the variant call set was uploaded for collaborative analysis between the Broad Institute Center for Mendelian Genomics and investigator. Similar variant calling approaches were used in prior Broad Institute publications¹².

Results

Patient cohort and genetics

In total, between ascertainment from the BCH Radiology database and direct referral to our BrDG Clinic, we identified a total of 22 patients harboring PTEN variants (Table 1; Figure 1A). By searching MRI requisition forms or radiologist dictation for the search term “PTEN”, we identified a total of 31 individuals out of 60,086 MRIs in the database, of whom 20 individuals had confirmed PTEN variants based on clinical sequencing data or clinical documentation. In addition, 5 clinically ascertained individuals with PTEN variants and brain malformations were included in the study; 3 of these 5 individuals had also been identified by the Radiology database search. Of the total 22 individuals that comprise our PTEN cohort, 14 (63.6%) were male, 8 (36.4%) were female, and ages ranged from 2.8 years to age 17 years.

Table 1.

Complete list patients included in this study and salient genetic and MRI findings

Patient	Age at evaluation	Sex	OFC SD (age years)	Coding variant	Predicted functional change	ACMG classification	MRI finding
1	17 years	M	+6.9 (17)	c.209+5 G>A	Splice site	Pathogenic	FPs PMG
2	14 years	M	+4.6 (14)	c.380G>C	p.Gly127Ala	VUS	FP PMG
3	2.8 years	M	+5.6 (2.8)	c.388C>T	p.Arg130Ter	pathogenic	FPs PMG
4	4.75 years	F	+5.0 (3.75)	c.388C>T	p.Arg130Ter	Pathogenic	DMEG (+PMG)
5	12.5 years	M	+6.4 (12)	c.389G>A**	p.Arg130Gln	Pathogenic	FPs PMG
6	8 years	F	4.3 (8)	c.406T>C	p.Cys136Arg	Pathogenic	PMG (vs FCD)
7	0.4 years	M	n.d.	c.464A>C	p.Tyr155Ser	Pathogenic	FPPs PMG
8	6 years	F	+6.4 (6)	c.521A>G	p.Tyr174Cys	VUS	FP PMG
9	8 years	M	+4.1 (3.6)	c.611delC	p. Pro204Glnfs*17	Likely Pathogenic	FPs PMG
10	16 years	M	+6.9 (16)	c.737C>T	p.Pro246Leu	Pathogenic	FPPs PMG
11	16 years	M	+6.6 (16)	c.955insA	p.Thr319Asnfs*6	Pathogenic	FPPs PMG
12	16 years	M	+7.0 (6)	c.1027delG	p.Val343Ter	Likely pathogenic	FPPs PMG
13	21 years	M	+5.5 (19)	c.−1034–1030dupGCCCT	Promotor	VUS	No cortical abnormality
14	11 years	F	+4.7 (10)	c.27delT	p.Ser10Alafs*14	Pathogenic	No cortical abnormality
15	12 years	F	+9.0 (11)	c.164+1G>A	Splice Site	Pathogenic	Atypical gyration
16	9 years	M	+5.4 (8)	c.323T>C*	p.Leu108Pro	Pathogenic	Atypical gyration
17	13 years	F	+2.9 (10)	c.686C>A	p.Ser229Ter	Pathogenic	No cortical abnormality
18	5 years	F	+4.5 (5)	c.737C>T	p.Pro246Leu	Pathogenic	SEGMH
19	8 years	M	+4.8 (6)	c.1027–1G>A	Splice site	Pathogenic	No cortical abnormality
20	24 years	F	+3.8 (20)	c.1110–1111ins ATAGT	p.Asp371Ilefs*47	Likely Pathogenic	No cortical abnormality
21	14 years	M	n.d.	c.1176delT	p.Phe392Leufs*24	Likely Pathogenic	No cortical abnormality
22	11 years	M	+4.7 (11)	deletion chr10q23	Deletion	Pathogenic	Atypical gyration

n.d. = data not available, OFC=occipitofrontal circumference, SEGMH = subependymal grey matter heterotopia, DMEG = dysplastic megalencephaly, PMG = polymicrogyria, F=frontal, P=parietal, Ps=perisylvian.

^**cDNA variant inferred from amino acid change documented in the electronic medical record

FIGURE 1:

PTEN patient cohort. (A) Schematic of patient ascertainment and association with MRI characteristics. (B) Mutational spectrum of PTEN associated with cortical abnormalities. Promoter region variant and chromosomal deletion including entirety of PTEN are not represented. MRI = magnetic resonance imaging; PMG = polymicrogyria; PTEN = phosphatase and tensin homologue.

We annotated individuals as those having PMG (patients 1–12) and those without PMG (patients 13–22). All 22 ascertained patients harbored heterozygous PTEN variants. Patient 20 harbors an additional benign PTEN variant in the 5’UTR in addition to the pathogenic variant. 14 variants were SNVs that altered the coding sequence or splice sites, 7 variants were small indels, and 1 patient (22) harbored a small chromosomal deletion that included the PTEN gene. All individuals harbored pathogenic or likely pathogenic variants by ACMG criteria¹¹, except for patient 2, patient 8, and patient 13 who harbored variants of uncertain significance (VUS). Despite classification as VUS, these individuals exhibited features consistent with PTHS on review of clinical symptoms, and their associated variants were not observed in control population databases^13,14 (Table S1). Thus, these individuals were included in our study based on our determination that these individuals’ disorders were likely related to PTEN.

Due to the known incidence of multiple molecular diagnoses¹⁵, particularly in patients that have autosomal dominant variants, we performed additional DNA sequencing in 4 patients who could be consented for additional sequencing and ruled out additional gene variants that may be contributing to brain malformations. Targeted sequencing was performed for 41 genes typically associated with PMG in patients 3, 4, and 10 (Table S2). In addition, whole exome sequencing (WES) was performed on patients 4 and patient 10. No additional variants other than those in PTEN were identified that were likely to represent a cause of polymicrogyria for these individuals.

Retrospective MRI review reveals frequent cortical abnormalities in patients with PTEN pathogenic variants

Standardized review of radiologic features of all 22 patients with PTEN variants in our cohort revealed a significant association of PTEN pathogenic variants with the radiographic appearance of PMG on MRI in comparison to 6 individuals for whom PTEN sequencing was normal (Fig. 1; Table 1; p=0.002, Fisher’s exact test). The individuals with normal PTEN status were also identified through radiological database query for search term “PTEN”, likely due to initial clinical concern for PTHS, and thus it is less likely that our findings of PMG in our PTEN cohort are simply due to radiological ascertainment bias. Just over half (12/22, 54%) of the individuals we assessed with PTEN variants had PMG on MRI, and an additional 13.6% (3/22) showed atypical gyral patterns that were suggestive of PMG but that could not be confirmed in multiple radiographic planes. We classified Patient 4 as having PMG, though we note that this patient’s MRI would best be described as dysplastic megalencephaly (DMEG) due to appearance of overgrowth, transmantle/subcortical grey matter heterotopia, and cortical dysplasia, in addition to radiologic appearance of polymicrogyria (Table S3).

Among the 12 individuals with PTEN variants and PMG (patients 1–12), the pattern of PMG involved the frontal and parietal convexities predominantly at the depth of the sulci, e.g., inferior and superior frontal sulci (Fig. 2A–B). In addition, 8 individuals had PMG involving the perisylvian regions, and when present, the perisylvian region was the most conspicuous site of involvement. Some areas of PMG were subtle, resulting in fine areas of increased gyration and only localized disturbance of surface anatomy (Fig. 2C). In 3 individuals classified as having an atypical gyral pattern (Figure 2D), abnormal folding could not be convincingly demonstrated in multiple planes as required for PMG according to our classification. The cortical areas exhibiting atypical gyration were also perisylvian (1 patient) or frontal (2 patients).

FIGURE 2:

Cortical abnormalities associated with PTEN variants. Sagittal (A) and axial (B) views of affected individuals with structural brain abnormalities. (C) 3D reconstructed images of selected cortical abnormalities. (D) Sagittal and axial MRI images from individuals with atypical gyration (appearance of PMG observed only in 1 radiological plane) or with SEGMH. Upper left, patient identifier. Dotted regions highlight affected brain areas. DMEG = dysplastic megalencephaly; MRI = magnetic resonance imaging; PMG = polymicrogyria; PTEN = phosphatase and tensin homologue; SEGMH = subependymal grey matter heterotopia.

PTEN variants associated with brain abnormalities tended to encode regions of the PTEN phosphatase domain (Figure 1B). Of the 10 individuals who harbored variants in the phosphatase domain, 8 had PMG (including Patient 4 with PMG/DMEG) and 2 had an atypical gyral pattern. In comparison, of the 7 individuals with variants in the PTEN C2 domain, 4 had PMG, 1 had subependymal grey matter heterotopia (SEGMH), and 2 had no brain malformations. None of the 5 remaining variants, which were located in the promoter region, N- or C- terminal domains, or gene deletion, were associated with brain abnormalities.

The quality of the MRI study performed affected radiologic determination of PMG. 9 of the 12 studies which were called as PMG were high-quality studies, performed using 3T MRI without significant motion degradation or artifact. In contrast, only 2 of the 8 studies that we reported as showing atypical gyral patterns or no abnormalities were of high quality. Only 6 of 12 patients with PMG were noted to have PMG on the initial clinical MRI report, whereas only one case that was dictated as having PMG was reclassified as atypical gyration in our study.

The full table of MRI findings in each patient can be found in Table S3. Except for Patient 4 who had dysplastic megalencephaly and subcortical heterotopia, none of the other individuals with polymicrogyria had heterotopia or hamartoma. Patient 18, who did not have polymicrogyria, had a single focus of periventricular nodular heterotopia. Additional imaging findings in the PTEN cohort included white matter abnormalities and vascular abnormalities. Vascular abnormalities included a scalp venous malformation in one individual, an arteriovenous malformation and a sphenoid wing venous malformation. All patients with PTEN variants had macrocephaly with head circumference measurements ranging between +2.9SD up to +9SD at last measurement. The mean was nominally higher in individuals with PMG (+5.8SD) compared to those without (+5.0SD).

Patients with PMG in the setting of PTEN variants do not have a high rate of epilepsy

Because the presence of PMG has previously been associated with high risk for epilepsy, we reviewed the EEG characteristics and risk of epilepsy in our patient cohort (Table 2). We reviewed clinical records and EEG data to determine whether PMG was associated with epilepsy or electrographic brain abnormalities in this cohort.

Table 2.

Electrographic findings of patients with PTEN variants who underwent electroencephalography.

Patient	PMG	Reason for EEG/ Age performed	Results
1	Yes	Done for inattention/ 16y	Normal
3	Yes	Developmental delay and ASD/ 2y	Intermittent slowing and midline spikes
10	Yes	Done as a routine/14y	Normal awake and asleep
12	Yes	Events concerning for apparent GTCs/ 6y and 12 y	At 6: continuous generalized slowing, right posterior and midline spikes. At 12: normal awake and asleep EEG
14	No	Staring spells/13y	Normal EEG, events captured, not seizures
15	No	Single nocturnal episode of crying and urinary incontinence/9y	Right and left sleep activated centrotemporal spikes, otherwise normal background
18	No	Developmental delay and ASD/3 and 4y	Normal background with right parietal spikes
19	No	Sleep study	Normal
22	No	Sleep study	Normal

Amongst 12 patients with PTEN variants and PMG, two individuals (16.7%) had epilepsy. Patient 4 had DMEG on MRI and multiple reported seizures during infancy including subclinical seizures and infantile spasms. Diagnosis of epilepsy was determined from chart review as EEG data were not available at our institution. Patient 12 had epilepsy diagnosed at the age of 6, with seizures several times per year on lamotrigine monotherapy, and EEG showed continuous generalized slowing with right posterior and midline spikes on review.

Of the patients who had PMG but did not have epilepsy, we identified 1 patient with abnormal EEG amongst 3 who had EEG performed for any reason. Patient 3 had EEG performed as part of the work-up for developmental delay, which showed intermittent generalized slowing and midline spikes. Patient 1 and patient 10 had normal EEGs.

Of the 10 individuals with PTEN variants without PMG, none had documented seizures based on chart review. Four patients had EEGs, and two had limited EEG data from sleep studies (Table 2). Patient 15 had an abnormal EEG with sleep-activated right and left centrotemporal spikes with tangential dipole. This EEG was performed for evaluation of nocturnal crying and urinary incontinence, and the episodes were ultimately diagnosed as night terrors. The other 4 EEGs (includes 2 EEGs from sleep studies) were normal. In summary, most EEGs revealed normal background with no evidence of encephalopathy in the group without definite brain abnormality, and spikes were seen only in 1 individual (1 abnormal of 5 available EEGs) with a pattern that is seen commonly in pediatric epilepsy but was present in this case in a child without clinical epilepsy.

The number of individuals with epilepsy was not significantly different between patients with vs. without cortical abnormalities (p-value = 0.48, Fisher’s exact test). It is notable that epilepsy was generally infrequent in our cohort of individuals harboring PTEN variants, including in those patients with PMG.

Neurodevelopmental differences associated with cortical abnormalities in patients with PTEN variants

Rates of cognitive disability were found to be higher in the group of patients with PMG, though our small sample size precluded statistical significance (Fig. 3A, Table 3). We utilized data for from either full-scale IQ, developmental quotient as calculated from formal neuropsychiatric assessment, or neurology clinical assessment. We assessed whether or not ID/GDD were present in each case. We excluded individuals for which a clear determination could not be made based on available data, though the cognitive/developmental data for each case including the excluded individuals are noted in Table 3. Five of 9 (55%) individuals with PMG for whom cognitive status could be ascertained were noted to have GDD/ID, compared to 2 of 8 (25%) of individuals without clear abnormalities. Although this difference is not statistically significant (p-value = 0.33), some of the delays in patients with cortical abnormalities were quite severe, including one patient who spoke only 15 words by age 7 years and another who could only communicate with 20–25 signs by age 4 years 9 months.

FIGURE 3:

Comparison of (A) percentage of individuals with global developmental delay or intellectual disability (GDD/ID), (B) percentage of individuals with delayed motor development, and (C) percentage of individuals with ASD, in individuals with PTEN variants with or without polymicrogyria. ASD = autism spectrum disorder; GDD = global developmental delay; ID = intellectual disability; PMG = polymicrogyria; PTEN = phosphatase and tensin homologue. [Color figure can be viewed at www.annalsofneurology.org]

Table 3.

Cognitive, Motor, and Behavioral Development of patients with PTEN variants.

Patient ID	ID/GDD	Motor Delay	ASD	Assessment Method for ID/GDD
1	No	None	No	Full scale IQ 92
2	Yes	Walk at 3–4 years	No	Full scale IQ 57
3	Yes	Sitting at 18 months	Yes	Bayley III at 39 months showed problem solving DQ 36% and language DQ 29%
4	Yes	Sitting at 18 months	No	20–25 communicative signs at 4y9mo
5	No	Unknown, Receives OT	Yes	Scales of independent behavior 103
6	No	Walk at 19 months	No	Age appropriate functioning reported by neurologist
7	Unknown	Unknown	Unknown	n/a
8	Uncertain	Sitting at 10 months	No	No objective assessment, DQ at least 67% based on review of function reported in neurology clinic visit
9	Yes	Unknown	Yes	Diagnosis in electronic medical record
10	No	Walk at 20 months	No	Full scale IQ 114
11	Uncertain	Sit at 8 months	No	No objective assessment
12	Yes	Walk at 3 years	Yes	15 words at age 7 years, impaired self-help
13	No	Walk at 1.5 years	No	Age appropriate comprehension on Gray Oral Reading Paragraph 4
14	Yes	Walk at 16 months	Yes	Mild intellectual disability reported in electronic medical record
15	Uncertain	Walk at 2 years	No	IEP in school only supports math
16	No	Walk by 15 months	No	Delayed 1 year in elementary school
17	No	None	No	No cognitive concerns by clinicians
18	No	Delayed per clinician	No	At 6y9mo: KABC-II 79, Kauffman EVT-II 79
19	No	Walk at 21 months	No	Functioning at grade level with 504 plan
20	Unknown	Unknown	Unknown	n/a
21	Uncertain	Walk at 20 months	No	Separate classroom special education classes
22	Yes	Walk at 3 years	No	At 9 years: reported functioning at kindergarten level by special needs school (development quotient 55%)

In general, a high rate of motor delay was noted amongst all individuals with PTEN variants, and a nominally higher rate in those with PMG (Fig. 3B, Table 3). We determined motor delay based upon clinically documented parental reports of acquiring the ability to walk independently greater than 18 months, or clinician assessment of delay. In summary, 6 of 9 (66.7%) patients with PMG for whom motor developmental data was available were noted to be delayed. In patients who did not have PMG, 4 of 8 (50%) patients for whom developmental data were available had motor delays. Speech delay was not assessed as there was inconsistent reporting in the electronic medical record, and rates of primary language delay is confounded by rates of cognitive disability.

Finally, we found that rates of autism spectrum disorder (ASD; previously, pervasive developmental disorder (PDD)) were higher in the PTEN cohort with PMG. 4 out of 12 individuals (33.3%) carried a diagnosis of ASD (diagnosed between the ages of 2 to 5 years), with one additional individual described as having difficulty with social nuances but not meeting diagnostic criteria for ASD. In comparison, only 1 out of 10 individuals (10%) without PMG carried a diagnosis of ASD (Fig. 3C).

Discussion

Here we expand the phenotype of PTEN-associated disorders to include a strong association with cortical malformations, in particular PMG, always in the setting of macrocephaly. Additional brain malformations identified include SEGMH and DMEG. We note that individuals with PMG in our cohort also had higher rates of ASD, motor delay, and ID, though larger cohorts are necessary to confirm the strength of this association.

Variants associated with PMG or atypical gyration were frequently located in the phosphatase domain of PTEN, suggestive of a genotype-phenotype relationship. Targeted sequencing of additional known genes associated with PMG did not identify alternative causes of PMG. However, we cannot rule out the possibility of mosaic variants. Somatic variation has been indeed established as a cause of focal structural brain abnormalities¹⁶.

Rates of PMG reported in patients with PTHS have varied widely ranging from 0% to 33%^5,6,17. In our study, we find a high rate with just over half of our PTEN cohort showing MRI evidence of PMG (54.5%). While ascertainment bias can be a concern due to identification of individuals through a radiology database, we note that the individuals found via this radiology search who had normal PTEN status despite clinically suspected PTHS did not have cortical malformations. Furthermore, only 2 individuals were clinically ascertained that were not identified by radiology search, and even if they were excluded, the rate of PMG would be 50%.

Our data suggest two potential explanations for the wide range of reported PMG prevalence in prior PTEN studies: the localized high frequency nature of the PMG (fineness) in PTEN patients and the general requirement for high quality brain imaging (i.e. 3T MRI). Specifically, PMG can be difficult to detect when thick, non-isotropic brain MRI sequences are obtained or when image noise obscures the increased gyral frequency characteristic of this cortical malformation. Therefore, it is logical that higher quality studies are preferred for detection of cortical abnormalities. Consistent with this concept, the prior study which reported a rate of PMG of 33% used only 3T MRI images⁶. In our institution, patients with macrocephaly and developmental delay are generally scanned at 3T with a high-density multi-channel head coil (32 or 64-channel), high resolution imaging technique, and at least one isotropic sequence. However, several of the patients in our cohort were imaged for other reasons or had lower resolution outside imaging which was not repeated. We showed that improved scan resolution increased radiologist ability to call PMG. Therefore, it is possible that our cohort may actually underrepresent the true proportion of individuals that have cortical malformations (i.e. atypical gyration or normal brain MRIs may be false negatives for PMG).

Forty-one percent of individuals in our cohort exhibit cortical white matter abnormalities, which have been described previously in association with PTEN³ (Table S3). Although the prevalence of white matter changes in our cohort is lower than the prior report, this difference is likely related to the fact that patients from that study were recruited predominantly from patients referred to leukodystrophy centers for unclassified white matter disorders. Interestingly, it has separately been found that individuals with PTEN variants and white matter abnormalities may still have normal intelligence, thus clouding the picture of whether the neurodevelopmental phenotype of PTEN is related primarily to dysfunction of the white matter¹⁸.

The radiographic findings in our PTEN cohort are important for both diagnosis and patient counseling. Developmental delay and macrocephaly are common indications for brain MRIs in children. Our data would suggest that the finding of PMG should prompt consideration of PTEN hamartoma syndrome in addition to other syndromes such as MPPH/MCAP, a differential that can change genetic testing and can also lead to changes in approach to clinical evaluation (e.g. dermatologic evaluation). Future prospective studies of patients who present with macrocephaly and PMG identified on MRI would be helpful to clarify how many of the individuals fitting this phenotype have pathogenic variants in PTEN.

Conversely, the finding of a cortical malformation need not cast into doubt the diagnosis of PTEN-associated disorder. In our clinically ascertained cohort, multiple patients were referred to the BrDG clinic due to the noted PMG despite having a known PTEN variant. One family was concerned for a progressive brain abnormality given that a prior MRI had been reported to them as normal. Another family was fearful of epilepsy reported to be associated with PMG. This anxiety may be mitigated by the knowledge that these brain abnormalities are in fact a frequent finding associated with PTEN variants and that epilepsy is not universal—in fact, it was not common—in patients with PTEN-related PMG.

Patients with PMG in our study of patients with PTEN variants have a much lower incidence of epilepsy (16%) when compared more generally to reported rates of epilepsy up to 87% associated with PMG¹⁹. Furthermore, epilepsy for patients with PMG has been reported to be difficult to control^19,20, again in contrast to the two patients with epilepsy in our cohort. Reasons for this difference may include ascertainment bias, as the incidence of epilepsy in patients with PMG is likely to be higher in patients seen in epilepsy clinics versus other clinics. Another consideration is the improved detection of PMG given advancements in MRI technology and its increased use in patients with developmental disorders. Bilateral perisylvian PMG (present in most of our patients) is significantly correlated with lower age of seizure onset as well (12 months)²⁰, thus the current ages of individuals in our cohort is not likely the reason for the relatively low epilepsy prevalence in our population.

One intriguing possibility may be that PTEN pathogenic variants either modulate this epileptogenic effect or drive a structural pattern of PMG that is distinct from other causes of PMG. Patients who have PMG as part of megalencephaly-capillary malformation (MCAP) and megalencephaly-polydactyly-polymicrogyria-hydrocephalus syndrome (MPPH), caused by variants in the PI3K which is regulated by PTEN, have been also reported to have a relatively low (38%) incidence of epilepsy⁹, though when epilepsy is present in this setting it can be very severe and difficult to control. This difference in epilepsy prevalence for patients with PMG may reflect subtle cytoarchitectural differences as radiographically identified PMG has been shown to represent a spectrum of cortical abnormality at a histological level^21,22. Regardless of any underlying histologic differences, our findings suggest that risk of seizures and likelihood of developing drug-resistant epilepsy are both lower than might be expected based on imaging findings and analogy to other forms of megalencephalic PMG.

The individuals with cortical abnormalities in our cohort also had higher rates of GDD/ID, motor delay, and ASD, though our cohort size is insufficiently powered to determine the significance of these developmental findings. Consistent with prior reports in association with PTEN-associated syndromes^5,23, we note a high incidence of motor and speech delays regardless of the presence of cortical abnormalities. Thus, patients with PTEN variants, particularly those presenting with PMG, should be assessed and closely monitored for features of ASD and developmental delay, particularly when cortical abnormalities are present.