Inherited variants in CHD3 demonstrate variable expressivity in Snijders Blok-Campeau syndrome

Interpretation of next-generation sequencing data of individuals with an apparent sporadic neurodevelopmental disorder (NDD) often focusses on pathogenic variants in genes associated with NDD, assuming full clinical penetrance with limited variable expressivity. Consequently, inherited variants in genes associated with dominant disorders may be overlooked when the transmitting parent is clinically unaffected. While de novo variants explain a substantial proportion of cases with NDDs, a significant number remains undiagnosed possibly explained by coding variants associated with reduced penetrance and variable expressivity. We characterized twenty families with inherited heterozygous missense or protein-truncating variants (PTVs) in CHD3, a gene in which de novo variants cause Snijders Blok-Campeau syndrome, characterized by intellectual disability, speech delay and recognizable facial features (SNIBCPS). Notably, the majority of the inherited CHD3 variants were maternally transmitted. Computational facial and human phenotype ontology-based comparisons demonstrated that the phenotypic features of probands with inherited CHD3 variants overlap with the phenotype previously associated with de novo variants in the gene, while carrier parents are mildly or not affected, suggesting variable expressivity. Additionally, similarly reduced expression levels of CHD3 protein in cells of an affected proband and of related healthy carriers with a CHD3 PTV, suggested that compensation of expression from the wildtype allele is unlikely to be an underlying mechanism. Our results point to a significant role of inherited variation in SNIBCPS, a finding that is critical for correct variant interpretation and genetic counseling and warrants further investigation towards understanding the broader contributions of such variation to the landscape of human disease.


Summary (250 words)
Interpretation of next-generation sequencing data of individuals with an apparent sporadic neurodevelopmental disorder (NDD) often focusses on pathogenic variants in genes associated with NDD, assuming full clinical penetrance with limited variable expressivity. Consequently, inherited variants in genes associated with dominant disorders may be overlooked when the transmitting parent is clinically unaffected. While de novo variants explain a substantial proportion of cases with NDDs, a significant number remains undiagnosed possibly explained by coding variants associated with reduced penetrance and variable expressivity. We characterized twenty families with inherited heterozygous missense or protein-truncating variants (PTVs) in CHD3, a gene in which de novo variants cause Snijders Blok-Campeau syndrome, characterized by intellectual disability, speech delay and recognizable facial features (SNIBCPS). Notably, the majority of the inherited CHD3 variants were maternally transmitted. Computational facial and human phenotype ontology-based comparisons demonstrated that the phenotypic features of probands with inherited CHD3 variants overlap with the phenotype previously associated with de novo variants in the gene, while carrier parents are mildly or not affected, suggesting variable expressivity. Additionally, similarly reduced expression levels of CHD3 protein in cells of an affected proband and of related healthy carriers with a CHD3 PTV, suggested that compensation of expression from the wildtype allele is unlikely to be an underlying mechanism. Our results point to a significant role of inherited variation in SNIBCPS, a finding that is critical for correct variant interpretation and genetic counseling and warrants further investigation towards understanding the broader contributions of such variation to the landscape of human disease.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Main text (2297 words)
The availability of whole exome sequencing (WES) in clinical practice has greatly improved the yield of genetic diagnostics for individuals with neurodevelopmental disorders (NDDs). In particular, sequencing of proband-parent trios, followed by filtering for de novo [1][2][3] or bi-allelic variants 4; 5 , has proven a powerful tool to identify causal variants in individuals with sporadic dominant and recessive NDDs. However, while de novo and bi-allelic variants explain a substantial proportion of cases with NDDs 1; 4; 5 , the majority remains undiagnosed 6 . Various factors may explain the difficulties to diagnose these individuals, including variation in genes not yet associated to disease, polygenic inheritance or variation in non-coding regions 7 . Also coding variants associated with reduced penetrance and variable expressivity may underlie unexplained NDD cases 6; 8 . Common diagnostic strategies to analyze next-generation sequencing data are not optimized to identify the contributions of these factors to disease. While penetrance indicates the proportion of carriers of a particular variant with a phenotype, expressivity describes the variability in severity of the phenotype between carriers of this variant 9 . Variable expressivity can cause highly variable symptoms, even in severe disorders that are caused by variants with a large effect 9; 10 .
In the present study we show variable expressivity for variation in CHD3. CHD3 is an ATPdependent chromatin remodeling protein that serves as core member of the NuRD complex 11 .
Heterozygous variants in CHD3 have recently been shown to cause a neurodevelopmental syndrome with a variable phenotype, ranging from mildly to more severely affected cases (MIM #618205, Snijders Blok-Campeau syndrome: SNIBCPS) 12; 13 . CHD3 is extremely intolerant for both loss-of-function (LoF) and missense variation (pLI = 1, o/e = 0.09 (0.05 -0.15); Z = 6.15, o/e = 0.5 (0.46 -0.53)), suggesting haploinsufficiency as a possible disease mechanism. However, the large majority of cases diagnosed with SNIBCPS carry confirmed de novo missense variants or single amino acid in-frame deletion variants (51/55, 93% of cases) [12][13][14] , clustering in the ATPase-Helicase domain of the encoded protein, and affecting its ATPase activity and/or . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 5, 2021. ; https://doi.org/10.1101/2021.10.04.21264162 doi: medRxiv preprint chromatin remodeling functions, which could be consistent with a dominant-negative mechanism 12 .
Computational facial analysis also confirmed the presence of a SNIBCPS facial gestalt in probands ( Figure S3; Table S2), and composite images showed similarities in facial features between probands with de novo and inherited CHD3 variants (squared face, deep set eyes, pointed chin; Figure 2B).
All of the carrier parents for whom phenotypic information was available about at least development and dysmorphisms (n = 6) had at least one feature of SNIBCPS (Table S2), although in five parents (31%) this was limited to only one (family 1 and 10) or two phenotypic features (family 9, 15 and 20; Table S2). Whereas the majority of carrier parents (15/16, 94%) presented with a single (e.g. prominent forehead or deep-set eyes) or several facial features known in SNIBCPS and 53% had macrocephaly (8/15) (Table 1 and S2), the parents had either mild/borderline intellectual disability (n = 4, 22%) or no history of intellectual disability (n = 14) ( Table 1, S1 and S2; Supplemental Notes 1). Taken together, these observations suggest a combination of both variable expressivity and reduced penetrance for these rare genetic variations in CHD3.
We more objectively compared the phenotypes of probands with de novo and inherited CHD3 variants based on Human Phenotype Ontology (HPO) terminology 15 , using a Partitioning Around Medoids clustering algorithm 16 . While this computational analysis did not identify a phenotypic difference between probands with de novo and inherited variants (31 of 54 individuals clustered correctly, p = 0.31; Figure 2C and S4), it confirmed a phenotypic difference between . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 5, 2021. ; https://doi.org/10.1101/2021.10.04.21264162 doi: medRxiv preprint probands with inherited CHD3 variants and their carrier parents (32 of 38 individuals clustered correctly, p = 0.00003; Figure 2C and S4). It is important to note that, although younger generations seem more severely affected than previous generations this may be due to ascertainment bias 17 . Our finding of variable expressivity of inherited CHD3 variants is consistent with the lack of a correlation between mutation location and phenotype 13 , and the variable severity of phenotypes already described for (recurrent) de novo CHD3 variants 12; 13 . Additional evidence for variable expressivity for CHD3 variation is provided by the recently identified association of 19 rare CHD3 missense variants with Chiari I malformations in individuals without features of SNIBCPS 18 .
We noticed that the majority of variants in our cohort were maternally inherited (14/20, 70%, p = 0.0577; Figure 1B, 1C and S5A). For single nucleotide variants with a LoF effect, 6/7 (86%, p = 0.0625) variants were maternally inherited ( Figure 1B and S5A). Notably, the only father transmitting a LoF single nucleotide variant was affected (mild intellectual disability). This observation could hint at a female-protective effect for genetic variation in CHD3. Previous studies have repeatedly demonstrated a male bias in NDDs, a higher pathogenic variant burden in females and a maternal transmission bias in rare inherited variants 6; 19-26 , suggesting that female gender protects against genetic variation in disease. This phenomenon might contribute to the variable expressivity observed for the inherited CHD3 variants. However, we did not observe a sex-bias in the affected probands (12/20 female, p > 0.9999), or more severe intellectual disability in male compared to female de novo or inherited cases 12; 13 . To further explore the hypothesis of a female protective effect at population level, we analyzed all CHD3 LoF variants in gnomAD (n = 15 in 198,800 individuals) and found that females had a significantly higher carrier rate than males (12/15, p = 0.0173; Figure S5B).
Few cases with SNIBCPS have been described with confirmed de novo CHD3 PTVs (4/55, 7.3% cases) 12; 13 including one which is predicted to escape nonsense-mediated decay (NMD; NP_001005273.1:p.Phe1935GlufsTer108). However, in our study we identified seven . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 5, 2021. ; https://doi.org/10.1101/2021.10.04.21264162 doi: medRxiv preprint families with inherited single nucleotide PTVs and one with an intragenic deletion with a predicted LoF effect (8/20, 40%; Figure 1A). None of the inherited PTVs were predicted to escape NMD.
We functionally confirmed this in family 1 ( Figure 3A), for which we treated lymphoblastoid cell lines from the proband (individual III-2), carrier mother (II-2) and grandmother (I-2) and non-carrier healthy sibling of the proband (III-1) with cycloheximide to inhibit NMD, followed by direct amplification and Sanger sequencing of the CHD3 transcript. We found that treatment with cycloheximide increased the expression of mutant allele, showing that the NM_001005273.2:c.3473G>A variant was targeted by NMD in all samples, as expected ( Figure   3B).
An explanation for variable expressivity of PTVs could be compensation of expression by the wildtype allele to maintain normal expression levels 27 . To test if such compensation plays a role in variable expressivity of CHD3 PTVs, we evaluated the expression of the CHD3 variant in family 1 (c.3473G>A, p.W1158*) on a transcript and protein level. We found that this variant resulted in lower levels of CHD3 transcript and CHD3 protein levels, when compared to lymphoblastoid cells from the non-carrier healthy sibling ( Figure 3C and 3D). These findings confirm the LoF effect of the stop-gain variant in this family, and makes compensation by the wildtype allele as an underlying mechanism for the milder phenotype in the carrier mother and grandmother unlikely. It remains, however, to be determined, whether such LoF variance can have a tissue-specific, temporal expression specific, and/or transcript specific effect. It is unclear whether results from blood-derived cells can be extrapolated to neuronal cell types, which would be more relevant considering the NDD phenotypes in our cohort, especially given that neuronspecific alternative splicing has previously been described for CHD3 28 . Other explanations for the clinical variable expressivity of inherited CHD3 variants include the presence of a second-hit on the other allele by either rare or common variation, a genome wide higher mutational burden of high-penetrant variants, or common variants in promoter/enhancers regions or in other genes, inherited from the non-carrier parent 9; 29; 30 . Such a compound inheritance mechanism, has, for . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 5, 2021. ; https://doi.org/10.1101/2021.10.04.21264162 doi: medRxiv preprint example, been described for thrombocytopenia with absent radii (TAR) syndrome, where the inheritance of a rare null allele together with one of two low-frequency SNPs in regulatory regions causes disease 31 . In four probands with an inherited CHD3 variant a copy number variant (CNV) was also reported, including one 22q11.2 duplication, which has been described with highly variable features (MIM # 608363) and three CNVs of unknown significance (Table S1). Proband 7 had other (likely) pathogenic variants contributing to the phenotype (Table S1; Supplemental Notes 1). A comparison with the prevalence of additional genetics finding in individuals with de novo CHD3 variants could not be made due to lack of reporting on additional genetic findings 12; 13 .
In addition to the seven inherited CHD3 single nucleotide PTVs and the intragenic deletion, we identified 12 families with inherited missense variants. One of the identified inherited missense variants also present in an unaffected carrier parent was identical to a variant previously reported as a de novo variant in an individual with SNIBCPS (p.R1342Q; individual 32 in 12 ).
Based on the phenotypes observed in the probands with inherited CHD3 missense variants, the conservation of affected positions ( Figure S1), and in silico predictions of pathogenicity ( Figure   S6; Table S1), we considered these inherited CHD3 missense variants as likely pathogenic with variable expressivity in the parents. Clinically, probands carrying a CHD3 missense variant did not seem to be more severely affected than individuals with PTVs (Table S3). The individuals with de novo missense variants published to date were mostly (although not entirely) localized to the ATPase-Helicase domain 12; 13 . No clustering to the ATPase-Helicase domain or elsewhere was observed among the inherited missense variants of our cohort ( Figure 1A). It has been speculated that the de novo missense variants clustering in the ATPase-Helicase domain are unlikely to lead to a sole LoF effect 12 , and may potentially act in a dominant-negative way. The identification of eight families with an inherited LoF variant and the lack of clustering of the inherited missense variants may suggest a LoF effect as the main mechanism for inherited cases, which may underlie the variable expressivity. However, our cell-based analyses of chromatin binding (for p.S477F) . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted October 5, 2021. ; https://doi.org/10.1101/2021.10.04.21264162 doi: medRxiv preprint and GATAD2B-binding (for p.R1342Q, p.E1837K and p.Q1888R) did not find evidence of LoF for the protein functions that we tested ( Figure S7). This does not exclude that these variants have an effect on other biological functions of CHD3. Based on 3D-protein modeling, the prior published de novo missense variants within the ATPase-Helicase domain localize more closely to the ATPbinding site than the inherited missense variants of our cohort (Supplemental Notes 2).
Interestingly, the p.I983V (family 13) variant was found to be closer to published de novo variants (Supplemental Notes 2) and the carrier parent with this missense variant did have a neurodevelopmental phenotype which was more pronounced than in other carrier parents ( Figure   1C and 2; Table S2; Supplemental Notes 1).
The presence of rare, likely pathogenic CHD3 variants in healthy individuals prompted us to study possible effects of variation in this gene at a population level, using data from the UK Biobank resource 32-37 . For a detailed description of these analyses, see Supplemental Notes 3.
We found no associations between rare missense variation at mutation-intolerant locations in CHD3 (minor allele frequency ≤ 1%, located in functional domains, damaging in PolyPhen or SIFT and with a CADD-PHRED score > 25) and fluid intelligence (N = 77,998), educational qualification (N = 120,596) or intracranial volume (N = 18,254). However, we identified a group difference in intracranial volume between rare CHD3 putative LoF variant carriers and non-carriers, with carriers having an increased volume compared to controls (n = 4, t = 2.37, p = 0.018). This seems consistent with the observation of macrocephaly in 44-53% of probands with a (likely) pathogenic CHD3 variant and in 53% of carrier parents (Table 1), and the link of rare CHD variants with abnormal brain growth 18 . To test possible relationships between CHD3 common genetic variation and head circumference and/or intracranial volume, we performed gene-level analyses using previously published SNP-wise association summary statistics for these traits 38; 39 , but no tests survived multiple testing correction (Supplemental Notes 3).
With the identification and characterization of inherited CHD3 variants with variable expressivity in twenty families, we showed that, in addition to highly penetrant de novo variants, . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted October 5, 2021. Clinically, we recommend that it can be helpful to evaluate the parents of children with CHD3 variants for subtle SNIBCPS features. In particular macrocephaly and facial dysmorphisms including a prominent forehead and pointed chin could be recognized in a substantial number of carrier parents (53% and 94% respectively; Figure 2A; Table S2). Taken together, our results illustrate the continuum of causality for NDDs with genetic origins 17; 42 and significantly underline the hypothesis that variable expressivity and reduced penetrance likely explain a large portion of as yet unexplained NDD cases. Overall, we show that even for genes already known to be implicated in a NDD inherited variation and variable expressivity can play a major role and are thus important to consider in genetic counseling.

Data availability
Supplementary Notes 1, Table S1, facial photographs ( Fig. 2A and Fig. S2) and all datasets generated and analyzed during the current study are available from the corresponding author on request.

Acknowledgements
We are extremely grateful to all families participating in this study. In addition, we wish to thank the members of the Cell culture facility, Department of Human Genetics, RadboudUMC, Nijmegen for culture of cell lines. This work was financially supported by the Dutch Research Council grant . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Individuals and consent
The cohort presented in this study was assembled from hospitals and laboratories across the Netherlands, Germany, United States of America, Slovenia, Australia and Canada. Informed consent for the use and publication of medical data and biological material was obtained from all . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted October 5, 2021. ; https://doi.org/10.1101/2021.10.04.21264162 doi: medRxiv preprint patients or their legal representative by the involved clinician. Consent for publication of photographs was obtained separately. Genetic testing and research were performed in accordance with protocols approved by the local Institutional Review Boards.

Next-generation-sequencing
CHD3 variants in all probands were identified using whole exome sequencing (WES) or whole genome sequencing (WGS; family 4 and 12), with filtering as previously described 3; 43-52 . Due to inheritance from seemingly healthy/mildly affected parents the CHD3 variants were initially classified as variants of unknown significance. Inheritance of variants was confirmed either as part of trio WES or using targeted Sanger sequencing after identification in singleton exon analysis. Similarly, if applicable, other family members were tested using targeted sequencing.

Facial analysis
We established a 2D hybrid facial model which combines the analysis of the 'Clinical Face Phenotype Space' pipeline with the facial recognition system of the 'OpenFace' pipeline 56; 57 .
First, we generated a 468-dimensional feature vector of the facial features of 30 individuals with de novo CHD3 variants. After extraction of the hybrid features for each of the individuals, we calculated whether the individuals with de novo CHD3 variants cluster together when compared to a group of matched controls based on the nearest neighbor principle (Euclidean distance)these matched controls were individuals with ID and are age-, ethnicity-and gender matched.
The Mann-Withney U test was used to determine whether the clustering of individuals with de novo CHD3 variants was significantly higher than expected based on random chance. A p-value smaller than 0.05 was considered significant.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.  Figure S3).

Construction of composite face
For 12 individuals with an inherited CHD3 variant and 30 with a de novo CHD3 variant, facial 2Dphotographs were available for generating a composite face. As previously described, average faces were generated while allowing for asymmetry preservation and equal representation by individuals 58 .

Human Phenotype Ontology (HPO)-based phenotype clustering analysis
HPO-based clustering analysis was performed as described elsewhere 59 . We included 35 individuals with de novo CHD3 variants 12 , 19 of 20 probands with an inherited CHD3 variant, and 19 of 20 carrier parents in the analysis: the proband and carrier mother of family 6 were excluded because no clinical data were available, and the mother is mosaic for the CHD3 variant (~37%).
The Wang score (a measure of semantic similarity) between all terms was calculated using the HPO Sim package 60; 61 . The terms were divided in groups, based on the similarity score: a new featurethe sum of the terms in the group -was created as a replacement for the terms in that specific group ( Figure S4A; Table S4). HPO terms that could not be added to a group feature were added as a separate term. To quantify and visualize possible differences in our cohort, we used Partitioning Around Medoids (PAM) clustering on these grouped features. We compared probands with a de novo and inherited variant and probands with inherited variants and their carrier parents in a second analysis. To assess statistical significance, a permutations test (n = . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted October 5, 2021. ; https://doi.org/10.1101/2021.10.04.21264162 doi: medRxiv preprint 100,000) was used with relabeling based on variant types, while keeping the original distribution of variant types into account.

Three-dimensional protein modeling
We modeled the protein structure of the ATPase-Helicase domain of CHD3 in interaction with the DNA using the homology modeling script in the WHAT IF 62 & YASARA 63 Twinset with standard parameters. As a template, we used PDB file 6RYR which contains the human Nucleosome-CHD4 complex structure of a single copy of CHD4 64 . The PHD2 variant (p.S477F) was modeled in the PHD2 domain of CHD4 (PDB 2L75, 89% sequence identity with CHD3) 65 .

Cell culture
Lymphoblastoid cell lines (LCLs) were established by Epstein-Barr virus transformation of peripheral lymphocytes from blood samples collected in heparin tubes, and maintained in RPMI medium (Sigma) supplemented with 15% fetal bovine serum and 5% HEPES (both Invitrogen).
HEK293T/17 cells (CRL-11268, ATCC) were grown in DMEM supplemented with 10% fetal . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Testing for nonsense mediated decay of truncating variants
LCLs of members of family 1 and controls were grown overnight with 100 µg/ml cycloheximide (Sigma) to block NMD. After treatment, cell pellets were collected, and RNA and protein were extracted using the RNeasy Mini Kit (Qiagen) or with 1x RIPA buffer supplemented with 1% PMSF and 1x PIC, respectively. RT-PCR was performed using SuperScript III Reverse Transcriptase (ThermoFisher) with random primers, and regions of interest were amplified from cDNA using primers listed in Table S5. Sanger trace peak sizes of the wildtype and variant allele were measured using the 'Area' option in ImageJ and proportion of the variant allele was calculated: peak area variant allele / (peak area variant + wildtype allele).

Direct fluorescent imaging
HEK293T/17 cells were grown on coverslips coated with poly-D-lysine (Sigma). Forty-eight hours after transfection with the YFP-tagged C-terminal CHD3 construct and HisV5-tagged GATAD2B, cells were fixed with 4% paraformaldehyde (PFA, Electron Microscopy Sciences). Nuclei were stained with Hoechst 33342 (Invitrogen). Fluorescence images were acquired with a Zeiss LSM880 confocal microscope and Airyscan unit using ZEN Image Software (Zeiss).

FRAP assays
HEK293T/17 cells were transfected in clear-bottomed black 96-well plates with YFP-tagged fulllength CHD3 or p.S477F. After 48 h, medium was replaced with phenol red-free DMEM supplemented with 10% fetal bovine serum (both Invitrogen), and cells were moved to a temperature-controlled incubation chamber at 37°C. Fluorescent recordings were acquired using a Zeiss LSM880 and Zen Black Image Software, with an alpha Plan-Apochromat 100x/1.46 Oil . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Co-immunoprecipitation
HEK293T/17 cells were transfected with the YFP-tagged C-terminal region of CHD3 and Rluctagged GATAD2B. After 48h, whole-cell lysates were collected in Pierce IP Lysis Buffer (25 mM Tris-HCl pH 7.4, 150 mM NaCl, 1 mM EDTA, 1% NP-40 and 5% glycerol; ThermoFisher) . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Population-based analysis of the association of CHD3 variation with intelligence, educational qualification and intracranial volume/head circumference
Using WES data of 200,000 individuals from the UKB Exome Sequencing Consortium ( 33; 34 , and https://www.ukbiobank.ac.uk/media/cfulxh52/uk-biobank-exome-release-faq_v9-december-2020.pdf) we studied the association of CHD3 missense and putative LoF variants with 'Fluid intelligence score' (data field ID 3533), 'Qualifications' (data field ID 6138) and 'Volume of EstimatedTotalIntraCranial' (data field ID 7054). Additionally, we used genome-wide association meta-analysis summary statistics of head circumference (N ≤ 18,881), and HC combined with intracranial volume (N ≤ 45,458) in childand adulthood 38 , and infant head circumference (N ≤ 10,768) 39 to calculate gene-level p-values reflecting the common variant associations of CHD3 with these traits using MAGMA 69 . For detailed description of the methods, see Supplemental Notes 3.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted October 5, 2021. ; https://doi.org/10.1101/2021.10.04.21264162 doi: medRxiv preprint CNS: central nervous system . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Figures
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted October 5, 2021.    . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.   . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted October 5, 2021.  . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted October 5, 2021. ; https://doi.org/10.1101/2021.10.04.21264162 doi: medRxiv preprint . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted October 5, 2021. ; https://doi.org/10.1101/2021.10.04.21264162 doi: medRxiv preprint . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 5, 2021. ; https://doi.org/10.1101/2021.10.04.21264162 doi: medRxiv preprint . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 5, 2021. ; https://doi.org/10.1101/2021.10.04.21264162 doi: medRxiv preprint