A family-based study of genetic and epigenetic effects across multiple neurocognitive, motor, social-cognitive and social-behavioral functions

Ron Nudel; Richard Zetterberg; Nicoline Hemager; Camilla A. J. Christiani; Jessica Ohland; Birgitte K. Burton; Aja N. Greve; Katrine S. Spang; Ditte Ellersgaard; Ditte L. Gantriis; Jonas Bybjerg-Grauholm; Kerstin J. Plessen; Jens Richardt M. Jepsen; Anne A. E. Thorup; Thomas Werge; Ole Mors; Merete Nordentoft

doi:10.1186/s12993-022-00198-0

A family-based study of genetic and epigenetic effects across multiple neurocognitive, motor, social-cognitive and social-behavioral functions

Nudel, Ron; Zetterberg, Richard; Hemager, Nicoline; Christiani, Camilla A. J.; Ohland, Jessica; Burton, Birgitte K.; Greve, Aja N.; Spang, Katrine S.; Ellersgaard, Ditte; Gantriis, Ditte L.; Bybjerg-Grauholm, Jonas; Plessen, Kerstin J.; Jepsen, Jens Richardt M.; Thorup, Anne A. E.; Werge, Thomas; Mors, Ole; Nordentoft, Merete 2022-12-01 00:00:00 Many psychiatric and neurodevelopmental disorders are known to be heritable, but studies trying to elucidate the genetic architecture of such traits often lag behind studies of somatic traits and diseases. The reasons as to why relatively few genome‑ wide significant associations have been reported for such traits have to do with the sample sizes needed for the detection of small effects, the difficulty in defining and characterizing the phenotypes, partially due to overlaps in affected underlying domains (which is especially true for cognitive phenotypes), and the complex genetic architectures of the phenotypes, which are not wholly captured in traditional case–control GWAS designs. We aimed to tackle the last two issues by performing GWASs of eight quantitative neurocognitive, motor, social‑ cognitive and social‑behavioral traits, which may be considered endophenotypes for a variety of psychiatric and neurodevel‑ opmental conditions, and for which we employed models capturing both general genetic association and parent‑ of‑ origin effects, in a family‑based sample comprising 402 children and their parents (mostly family trios). We identified 48 genome‑ wide significant associations across several traits, of which 3 also survived our strict study‑ wide quality criteria. We additionally performed a functional annotation of implicated genes, as most of the 48 associations were with variants within protein‑ coding genes. In total, our study highlighted associations with five genes (TGM3, CACNB4, ANKS1B, CSMD1 and SYNE1) associated with measures of working memory, processing speed and social behavior. Our results thus identify novel associations, including previously unreported parent‑ of‑ origin associations with relevant genes, and our top results illustrate new potential gene → endophenotype → disorder pathways. Keywords: GWAS, Neurodevelopment, Cognitive functions, Endophenotype, Parent‑ of‑ origin effect Introduction As a species, humans are adept at using communica- tion (both verbal and nonverbal), mental facilities, social interaction abilities and fine motor skills in their every - *Correspondence: merete.nordentoft@regionh.dk day lives. These aptitudes mature during neurodevel - CORE ‑ Copenhagen Research Centre for Mental Health, Mental Health opment. Some individuals, however, have non-typical Centre Copenhagen, Copenhagen University Hospital, Copenhagen, neurodevelopment, which is associated with cognitive, Denmark Full list of author information is available at the end of the article motor, behavioral and/or social-cognitive impairments. © The Author(s) 2022. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. The Creative Commons Public Domain Dedication waiver (http:// creat iveco mmons. org/ publi cdoma in/ zero/1. 0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Nudel et al. Behavioral and Brain Functions (2022) 18:14 Page 2 of 20 Disorders characterized by these impairment are collec- as an endophenotype for schizophrenia [17]. In psychia- tively known as neurodevelopmental disorders, and they try in general, endophenotypes tend to be electrophysi- often exhibit high comorbidity [1]. Many of these dis- ological e.g., electroencephalogram (EEG), eye tracking orders have a strong genetic component, but they often or certain reflexes [18] or behavioral e.g. gaze direction exhibit both genetic and clinical heterogeneity [1–4]. towards specific facial features [19]. An example of a rela - Such a high degree of heterogeneity, in turn, encum- tively highly studied gene → behavioral endophenotype bers studies into the molecular underpinnings of these → disease pathway is that of the Calcium Voltage-Gated disorders. Channel Subunit Alpha1 C (CACNA1C) gene, which is a One strategy which has been proposed as a means to known susceptibility gene for several psychiatric disor- tackle this issue in psychiatric genetics is the use of endo- ders including schizophrenia [20]. A recent study showed phenotypes. Endophenotypes are heritable traits that are that deletions in that gene in mice led to behaviors asso- (typically) convenient to measure and exhibit an associa- ciated with psychiatric disorders [21]. tion with the psychiatric condition; more formally, they Even though many genetic studies of the aforemen- are said to be heritable traits that are associated with the tioned traits and disorders (and of complex traits and disease in the population, are primarily disease-state- diseases, in general) have been conducted, these stud- independent and co-segregate with the disease in families ies together have not identified enough associations to (an additional criterion for complex diseases is that endo- account for the heritabilities of the investigated traits phenotypes found in affected family members be found or diseases, a problem known as “the missing heritabil- in non-affected family members at a higher rate than in ity” [22]. As genome-wide association studies (GWAS) the general population) [5]. Endophenotypes can also become larger, more associations are identified at the be quantitative, in which case they should be “milder” conventional genome-wide significance threshold. How - in unaffected relatives of affected individuals and cor - ever, there are other reasons why some associations elude related with the severity of the disease, and, if this cor- the GWAS design, even as sample sizes grow larger: for relation is not due to disease progression or medication, example, there may be phenotypic heterogeneity not then it could suggest that the correlation with the disease only across individuals, but also in the sense that dif- is by way of disease liability [6]. Many traits that can be ferent studies may use different definitions for disor - measured using standardized tests meet these criteria. ders, different ascertainment criteria and/or different Pertinent to this study is the case of heritable quantitative assessment tools, and, at times, the studied phenotypes traits, which, in turn, may themselves be composites of themselves might reflect several overlapping underlying different measures. For example, it has long been known abilities. From the genetic perspective, an important rea- that general intelligence is heritable [7]. Although the son is that the common GWAS study design, i.e., using issue of what the intelligence quotient (IQ) itself meas- only unrelated individuals and modeling only specific ures is debated, as are the assumptions about the models types of effects, might not capture all the aspects of the estimating its heritability, the overall evidence from twin genetic architecture of a trait [22, 23]. Pertinent to this studies and other family-based studies suggests that a study is the case of the epigenetic phenomenon (i.e., large proportion of the variation in IQ between individ- a heritable phenomenon not caused by changes in the uals is due to additive genetic effects [8, 9]. Specifically, DNA sequence itself) known as parent-of-origin effect indices from subtests of the Wechsler Intelligence Scale (POE), whereby the effect of an allele is dependent on its for Children also have moderate to high heritabilities parental origin. Family-based genetic studies, where both [10]. Moreover, specific tests designed to measure vari - parental DNA and proband DNA are available, are ideal ous phenotypic expressions of autism spectrum disorder, for studying these effects. POEs have been implicated in namely the “Strange Stories” test, which can identify The - many studies of complex traits and diseases [24]. Studies ory of Mind impairments, and the Social Responsiveness have shown that, when these effects do operate but are Scale, which provides a quantitative measure of autistic not modeled, they can be missed in traditional GWAS behavioral traits, have both been shown to have modest designs [25, 26]. Moreover, the same allele may have (“Strange Stories”) to high (Social Responsiveness Scale) opposite effects when inherited paternally vs. maternally heritabilities [11, 12]. In fact, measures from the Social [25, 26]. Responsiveness Scale and from the Wechsler Intelligence Genomic imprinting is the epigenetic mechanism Scale for Children have been successfully used as endo- considered the primary underlying cause of POEs [27]. phenotypes in studies of autism spectrum disorder (ASD) Imprinted loci are loci at which the two parental alleles and attention deficit/hyperactivity disorder (ADHD) [13, are not functionally equivalent (and one of them may 14]. Lastly, both motor skill and motor learning are also even be silenced completely). One molecular mechanism heritable [15, 16], and motor deficits have been suggested that could lead to imprinting is methylation (the presence Nudel et al. Behavioral and Brain Functions (2022) 18:14 Page 3 of 20 of a methyl group on the DNA nucleotide). Allele-specific working memory [39], social responsiveness [40], and methylation in differentially methylated regions (DMRs), motor function [41]. Interestingly, these studies did not or, in this context, imprinting control regions, can lead to find similar differences between children of parents with differential gene expression depending on the parental no diagnosis of schizophrenia or bipolar disorder and origin of the allele [27]. Modification of histones (basic children who had at least one parent with a diagnosis of proteins around which DNA is wound to form nucle- bipolar disorder. osomes, a compact package of DNA which makes it pos- The main goal of our study is thus twofold: (i) to find sible for the DNA to fit within the nucleus of the cell) can specific genotype–phenotype associations for the quan - also result in altered gene expression; protein complexes titative phenotypes from across the aforementioned that modify histones covalently can lead to repression domains, and (ii) to model POEs in addition to general of transcription [28]. POEs may also result from mecha- association to identify associations that would not be nisms other than genomic imprinting, for example, bias captured in case–control GWAS designs. While we do in transmission of specific types of genetic, such as tri - not set out to show that the investigated traits are endo- nucleotide expansions variation, depending on the sex of phenotypes for specific disorders [as mentioned earlier, the parent [29]. Several disorders which involve genomic some of them have already been used as endophenotypes imprinting have strong behavioral and cognitive manifes- in previous studies, and they (or similar traits measured tations. Perhaps the most often-cited examples thereof by other tests) have been shown to be heritable)], they are are Prader-Willi syndrome and Angelman syndrome. The all inherently relevant to neurodevelopment in their own genes involved in both of these disorders map to chromo- right. Moreover, a recent article examining the history some 15q11q13, but different genes are involved in the of the use of endophenotypes in psychiatry proposed to two disorders, and they exhibit opposite POEs (paternal expand the definition to include transdiagnostic traits, for Prader-Willi syndrome and maternal for Angelman which are not necessarily associated with only one dis- syndrome); similarly, the cognitive and behavioral defi - order [42]. In this context, identifying genetic variants cits differ between the two disorders [30]. Most cases of influencing neurodevelopmental traits is an important these disorders are caused by a deletion of the parentally endeavor in its own right. To our knowledge, this is the expressed DNA, but some cases are the result of imprint- first study which examined these four neurodevelopmen - ing defects, leading to aberrant methylation patterns [31, tal domains in the same cohort, incorporating both gen- 32]. In the case of complex neurodevelopmental disor- eral GWAS models and POE models. ders, some notable examples for which POEs have been reported include specific language impairment [33, 34], Materials and methods dyslexia [35] and autism spectrum disorder [36]. A study Participants of 97 traits in mice, where the parent-of-origin of alleles The sample used in this study is part of the Danish High could be determined, found that most of them exhibited Risk and Resilience Study—VIA 7 (hereafter the VIA 7 POEs, to which a large component of their heritability study) [38]. The VIA 7 study recruited children aged 7 was attributable. Moreover, the study showed that non- and their biological parents. Families were recruited from imprinted loci could also exhibit POEs through interac- Danish registries on account of having at least one par- tion with imprinted loci [37]. These examples illustrate ent with a diagnosis of either schizophrenia spectrum the importance of considering POEs in studying behavio- psychosis or bipolar disorder (“high risk” families) or as ral and cognitive phenotypes. control families, in which neither parent had schizophre- Our study aimed to examine both general genetic asso- nia or bipolar disorder; however, these disorder were not ciation as well as parent-of-origin effects, in a deeply investigated directly in this study. Overall, of the 402 chil- phenotyped family-based cohort, in which families were dren with genetic data included in this study (after quality chosen based on the presence (in at least one parent) or control), 244 come from high risk families (schizophre- absence (in both parents) of a diagnosis of schizophrenia nia: 147; bipolar disorder: 97), and 158 come from control or bipolar disorder, and in which DNA from parents and families. The sample size varies per marker per analysis, children was collected, as well as data on a wide array of as the number of informative children depends on the quantitative neurocognitive, motor, social-cognitive and availability of trait data, marker (genotype) data, and, in social-behavioral traits [38]. In prior studies which used the parent-of-origin analyses, parental genotypes as well. this cohort, several of the investigated traits have been We therefore specify the number of informative children shown to differ significantly between children of parents (probands) for all significant results individually. Regard - with no diagnosis of schizophrenia or bipolar disorder ing parental data, only genetic data were used in the asso- and children who had at least one parent with a diagnosis ciation tests. After the quality control described below, of schizophrenia. These included processing speed and there were 261.117 trios, 88.0364 child-mother duos, Nudel et al. Behavioral and Brain Functions (2022) 18:14 Page 4 of 20 24.1713 child-father duos, 17.0366 children, 0.352642 Pearson’s correlation coefficients across the traits in the parents, 0.173135 mothers and 0.0495053 fathers (as well sample of children with genotypes used in this study, as 37.9879 parents without children in the dataset), on which are shown in Fig. 2. This was done using the Hmisc average per marker, as counted with PREMIM [43], with- package v.4.7-0 [51] for R, and the plots were generated out taking siblings into account. These numbers add up with the corrplot package v.0.92 for R [52]. Descriptive to ~ 391 (not counting parents without children in the statistics for the traits are found in Table 1, which also dataset), which is the number of independent children includes p-values from the Shapiro–Wilk normality test with genetic data in our sample (11 families included a as implemented in the shapiro.test function in R. sibling as well). All traits deviated from normality to some degree. However, as the effective sample for the majority of our Phenotypic data tests depends on the parental genotypes, it varies greatly We investigated eight traits derived from different tests across genetic markers (which are tested individually). selected from the comprehensive battery of the VIA 7 Therefore, different subsets of children were used for dif - study: MABC (total score from the Danish version of ferent markers, and it is not practical to try to transform Movement Assessment Battery for Children (Movement the scores so as to force them to have a normal distribu- ABC-2), 2nd edition [44]. N.B.: the norm sample for the tion, when each transformation will not necessarily work Danish version was from the UK, but it has cross-cultural for more than one marker. Moreover, transforming scores validity [45]); WISC Coding [score (total number cor- in this way would hinder the interpretation of the results, rect) from the Coding subtest of the Danish version of as the spaces between scores would have been changed the Wechsler Intelligence Scale for Children, 4th edition unevenly, and, therefore, the interpretation of the effect (WISC-IV) [46]]; WISC Symbol Search [score (total sizes would be problematic. We discuss this at length in number correct) from the Symbol Search subtest of the a previous paper, where we also examined the difference Danish version of the WISC-IV]; SSR (score from the normalization had made for our top result [53]. Lastly, as Strange Stories—Revised [47], based on the total number we explain below, we used variance components (which of correct answers to 8 mentalizing questions translated assume normality) to correct for relatedness among the into Danish); SRS (T-score from the Danish version of children within a given family. As we had only 11 fami- the Social Responsiveness Scale (SRS-2) [48], 2nd edi- lies with more than one child, we investigated the effect tion, completed by the child’s teacher); WISC Arithme- of removing a child from each family and not modeling tic [score (total correct responses) from the Arithmetic the variance components, and we saw that it had very subtest of the Danish version of the WISC-IV]; WISC little impact on our top result [53]; we therefore employ Letter-Number Sequencing [score (number of correct the phenotypic scores as detailed above, without an addi- trials) from the Letter-Number Sequencing subtest of tional transformation. the Danish version of the WISC-IV]; RIST Index [index score from the Danish version of the Reynold’s Intellec- Genetic data tual Screening Test (RIST) [49]]. We had DNA samples from a subset of the VIA 7 study The WISC Arithmetic, WISC Letter-Number Sequenc - sample, and these were genotyped on the Illumina Psy- ing, RIST Index and MABC scores were age-standardized chChip v1-1_15073391_C, which had a loci count of based on the norms from the manual of each respective 603,144 (according to the information lines in the Illu- test. Where norms were not available for some tests or mina manifest file for this array). The dataset has been subtests e.g., when we used the versions of the WISC- described in detail in our previous studies [53, 54]. IV Coding and Symbol Search subtests for children aged Briefly, the quality control (QC) steps for the samples 8 to 16, or when there were no norms (SSR scores), the and markers were as follows: initial QC on raw genetic raw scores were rescaled into Z-scores in SPSS v25.0.0.2 data: individuals with low call rates or discordant sex using the mean of the population control subset of VIA 7 information were removed in the first step, as were children, who were age-matched to the rest of the cohort. markers with a Gentrain score < 0.3. At this point 18 The SRS total T-score was not adjusted for age, as this individuals had been removed (including one possible score was not associated with age in children aged 7–15 duplicate sample), and there were 600,282 markers left [12]. More details about these tests can be found in pre- in the dataset. Subsequent QC was done with PLINK vious publications on the VIA 7 study [39–41]. The dis - [55] v.1.90b5.2: individuals and markers with > 1% tributions of the test scores for each trait are shown in Mendelian errors were removed (N = 10). Genotypes Fig. 1, which contains histograms and density plots for with remaining Mendelian errors below this threshold the traits and was generated in R [50] v3.6.3 using the hist were set to missing. Markers with > 5% missing data and density functions. We also calculated the pairwise were removed (at this point all remaining individuals Nudel et al. Behavioral and Brain Functions (2022) 18:14 Page 5 of 20 MABC WISC Coding WISC Symbol Search SSR 05 10 15 20 −4 −2 02 −3 −2 −1 012 3 −3 −2 −1 0123 Score Score Score Score SRS WISC Arithmetic WISC Letter−Number Sequencing RIST Index 40 60 80 100 2468 10 12 14 05 10 15 60 80 100 120 Score Score Score Score Fig. 1 Histograms and density plots for all phenotypes across all children with non‑missing phenotype values per trait. MABC, movement assessment battery for children; WISC Coding, coding subtest of the Wechsler Intelligence Scale for Children ( WISC); WISC Symbol Search, symbol search subtest of the WISC; SSR, strange stories—revised; SRS, social responsiveness scale; WISC Arithmetic, arithmetic subtest of the WISC; WISC Letter‑Number Sequencing, letter ‑number sequencing of the WISC; RIST Index, index score from the Reynold’s Intellectual Screening Test had < 5% missing data). Individuals with extreme het- as a minor allele frequency (MAF) threshold of 1% (in erozygosity rates (with a threshold of ± 3 SD from the founders). Markers with a significant HWE p-value sample mean) were removed (N = 21). Genetic ances- based on the above threshold or MAF below 1% were try was estimated in a principal component analysis excluded. We removed one marker per pair in case of (PCA). The threshold for the exclusion of samples was pairs of markers with identical positions included in 2 SD above or below the VIA 7 mean for either PC1 or the PsychChip, either using PLINK --list-duplicate- PC2, using the VIA 7 samples and the CEU, CHB, JPT vars suppress-first, if the allele codes matched, other - and YRI HapMap samples to create the PC space, as wise prioritizing markers with rsIDs. The number of described in a published QC protocol [56]. To reduce individuals removed during these steps was 64, includ- bias from population stratification, individuals of diver - ing 3 duplicate samples (note that some samples were gent ancestry were removed along with their relatives flagged in more than one step, in cases in which sev - (N = 36), while the rest of the sample clustered with eral checks were performed before the final exclusion the CEU individuals. Individuals who exhibited cryptic of samples, namely, after the Mendelian errors check). relatedness or who were less related to biological fam- In total, 1094 genotyped individuals and 299,604 auto- ily members than expected from pedigree information somal markers passed these QC steps, from which we were removed (N = 13) (the Pi-hat threshold for the further removed 125 indels, for a total of 299,479 auto- exclusion of individuals expected to be unrelated was somal markers. Only this dataset of genotyped markers 0.185). A Hardy–Weinberg Equilibrium (HWE) p-value was used in this study. Positions in the text and tables –6 threshold of 1 × 10 was employed for markers, as well are in genome build hg19. A minority of markers on the Density Density 0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.00 0.02 0.04 0.06 0.08 0.10 0.12 Density Density 0.00 0.05 0.10 0.15 0.20 0.00.1 0.20.3 0.4 Density Density 0.00 0.05 0.10 0.15 0.0 0.10.2 0.3 0.4 0.5 Density Density 0.00 0.01 0.02 0.03 0.04 0.0 0.1 0.20.3 0.40.5 0.6 Nudel et al. Behavioral and Brain Functions (2022) 18:14 Page 6 of 20 MABC 0.8 WISC Coding 0.6 0.4 WISC Symbol Search 0.2 SSR SRS −0.2 WISC Arithmetic −0.4 −0.6 WISC Letter−Number Sequencing −0.8 RIST Index −1 Fig. 2 Pearson correlations across the investigated traits. All correlation coefficients were different from zero with P < 0.05 array had positions in the Illumina manifest file which the position of interest (the direction depended on the differed slightly from the ones in dbSNP for the same strand), as expected. One marker among the top hits major build. For that reason we checked all the probes had an incorrect position in the manifest file, but the from the manifest file for our array for the top hits in probe mapped to the right place. this study, as for the top hit in our previous study [53]. This was done using the UCSC Genome Browser BLAT Statistical analyses—GWAS stage tool, and, where a marker in our top hits had an rsID, In the GWAS stage, we used QTDT (quantitative trans- we checked that the SNP was indeed 1 bp away from the mission-disequilibrium test) [57] v2.6.1 for the statistical probe. Otherwise we checked that the probe mapped to genetic analyses. MERLIN [58] v1.1.2 was used for esti- a position 1 bp away from the position in the manifest mating identity by descent (IBD) scores for each marker file. All probes mapped to a position 1 bp away from to be used by QTDT. Three tests were performed for MABC WISC Coding WISC Symbol Search SSR SRS WISC Arithmetic WISC Letter−Number Sequencing RIST Index Nudel et al. Behavioral and Brain Functions (2022) 18:14 Page 7 of 20 ‑ ‑ ‑ ‑ ‑ ‑ ‑ ‑ Table 1 Descriptive statistics for the investigated traits in the study sample. POE: parent of origin effect a b b Trait Domain The function Descriptive statistics for phenotypes in the full sampleGeneral test Paternal POE t est Maternal POE t est measured by Children Mean Median Standard Minimum Maximum Shapiro– Minimum Maximum Minimum Maximum Minimum Maximum the test with deviation value value Wilk test Number of Number of Number of Number of Number of Number of phenotype p-value probands probands probands probands probands probands data MABC Motor A combined 399 8.105 8 3.31 1 19 0.0001 337 399 196 382 196 382 score from tests of manual dex terity, aiming and catching, and balance WISC Cod Neuro Processing 401 − 0.134 − 0.059 0.992 − 3.61 3.219 0.0417 338 401 197 384 197 384 ing cognitive speed –5 WISC Neuro Processing 398 − 0.121 − 0.073 0.994 − 3.292 2.767 8.74 × 10 335 398 196 381 196 381 Symbol cognitive speed Search –6 SSR Social Theory of mind 400 − 0.1 0.046 0.999 − 3.227 2.909 1.69 × 10 337 400 197 383 197 383 cognitive –19 SRS Social Social respon 345 49.014 46 10.441 37 106 4.68 × 10 292 345 172 331 172 331 behav siveness ioral –10 WISC Neuro Verbal working 398 9.475 10 2.396 2 15 1.18 × 10 335 398 196 381 196 381 Arithmetic cognitive memory –13 WISC Neuro Verbal working 398 10.508 11 3.103 1 16 7.34 × 10 336 398 196 381 196 381 Letter cognitive memory Number Sequenc ing –10 RIST Index Neuro A combined 401 103.85 105 10.458 56 127 5.75 × 10 338 401 198 384 198 384 cognitive score from tests of non verbal intelligence (approximation of fluid intelli gence) and ver bal intelligence (a measure of crystallized intelligence) QTDT chooses informative probands from the pedigree file. Additionally, probands could have been excluded from the tests based on phenotype missingness, genotype missingness and/or problems with their IBD estimation with MERLIN In addition to the reasons for exclusion in the general test, probands could have also been excluded based on parental genotypes (see “Materials and methods”) Nudel et al. Behavioral and Brain Functions (2022) 18:14 Page 8 of 20 each trait-marker combination: a general (i.e., not a POE significantly different. It could be that a child effect is sig - test) total test of association using all family data (qtdt nificant and appears as such also when looking at pater - -at), a paternal parent-of-origin total test of association, nally inherited alleles or only at maternally inherited in which only paternally inherited alleles were used (-at alleles separately. Therefore, it is necessary to test for a -op) and a maternal parent-of-origin total test of associa- difference between these parental allele transmissions. tion, in which only maternally inherited alleles were used This can be achieved by controlling for risk parameters (-at -om). The total association model (as opposed to the other than the POE parameter by including them in both orthogonal model, which QTDT can also run) is not a the null and the full models. QTDT does not allow a TDT, and it was used because it is more powerful in the free choice of parameters in the null and the full mod- absence of population stratification [59]. In this model, a els, but it incorporates a test for the difference in the combined between/within family component X, or Xpat effects between the paternal and maternal allelic trans - and Xmat in the paternal and maternal tests, respectively, missions (qtdt -at -ot). In this test, the null model has X, denoting the between/within effect on the means, is and the full model has both X and Xmat. It is not pos- tested. X is the effect size reported for the QTDT analy - sible to include Xpat instead of Xmat in the full model ses in this paper. X is estimated from the data in the full (to test for a POE when a paternal POE is suspected); model and is fixed to zero in the null model. The likeli - therefore, as a precaution, we tested both parameteriza- hoods of these two models are then assessed through a tions for a known POE using a different program, EMIM likelihood ratio test, resulting in a χ statistic, which can [43], which allows to model both parental risk parame- be used to compute a one-sided p-value from the χ dis- ters (one at a time in this case) in addition to the child tribution. The tests in this study had one additional free risk parameter, and saw that the overall likelihood of the parameter in the full model as compared with the null full model was roughly the same in both cases. Thus, we model, and so the χ statistic was evaluated with 1 degree used this test to filter out associations that are significant of freedom. We included variance components in both in the GWASs when testing paternally inherited alleles models (-wega), incorporating an environmental compo- and maternally inherited alleles separately, but which do nent, a polygenic component and an additive major locus not show a significant difference from the other parent’s component. This allowed for the use of families with transmissions. Note, however, that these models do not multiple children, although only 11 families included a test for the simple parental effect at the locus of inter - sibling. Age was taken into account in the scoring of the est or for the type of POE (if the POE is real), and, for phenotypes, as explained earlier. For all traits, a covari- ate for sex was added to both the null model and the full 1 2 Consider a previously reported association [34, 54] with a paternal POE for model. The Manhattan plots and the QQ plots were specific language impairment: rs4280164 allele A having a paternal imprint - generated with the “qqman” R scripts by Stephen Turner ing parameter I (a multiplicative factor by which the probability of disease and Daniel Capurso (with the (major update) version is multiplied if the child receives a (paternal) copy of the effect allele from –8 their father) = 0.255 with P = 2.918 × 10 (1 degree of freedom). In this case, from April 19, 2011 for the former type of plot and the the null model has all risk parameters fixed to 1, and in the full model I is version from June 10, 2013 for the latter, available from: freely estimated; the two models are assessed using a likelihood ratio test. https:// github. com/ steph entur ner/ qqman/ bloqb/ v0.0. 0/ As the EMIM software (43) (used in the original study) allows many param- eterizations, it is also possible to include a free R parameter (the factor by qqman.r). Regional association plots were generated with which the disease risk is multiplied if the child has a single copy of the effect LocusZoom [60], after converting marker IDs to rsIDs allele, and assuming that the risk from the child’s having two risk alleles (where possible) using a key from the Illumina website. is R = R ) in both the null and full models and a free I parameter only in 2 1 p 2 –8 the full model. This results in a χ of 28.752840302, P = 8.22 × 10 (1 degree The QTDT output files were tabulated using an in-house of freedom), suggesting that the paternal POE is still significant even when program (included in the Additional file 1), but the sta- allowing for a child effect. Importantly, even when estimating the maternal tistics for the top hits in our study were also examined imprinting parameter, I (a multiplicative factor by which the probability of disease is multiplied if the child receives a (maternal) copy of the effect allele manually using the raw QTDT output, and they matched from their mother), instead of I , at the same locus (with R ), the likelihood p 1 the output of the program. of the full model here provides very similar evidence of association (χ of 28.752840295436), although the parameter estimates are different. Similarly, even if we do not make the assumption that R = R (i.e., we freely estimate Statistical analyses—post hoc tests for paternal 2 1 both R and R in the null and full models), we still obtain very similar likeli- 1 2 and maternal allelic transmission differences hoods, resulting in χ = 25.88917562832 when estimating I in the full model, When a POE is detected with one parent, it does and χ = 25.889147279256 when estimating I in the full model (in both cases including free R and R in both the null and full models), both obtaining not mean that the other parent’s transmissions are 1 2 –7 P = 3.62 × 10 (1 degree of freedom). We believe that QTDT is doing some- thing similar with its test of allelic difference, and that a test of either combi - nation can be used to determine the presence of a POE, even though QTDT N.B.: The Danish manual for the SRS we used did not include a sex-adjust - always models the maternal POE effect parameter as the additional one. ment for the T-score; hence, we used a covariate for sex for this trait as well. Nudel et al. Behavioral and Brain Functions (2022) 18:14 Page 9 of 20 that reason, we only use them to test for the presence of the MAF in founders), and N is the sample size (we used a POE and not for effect estimation; we always report the number of probands). We do this for the top results the effect of the allele from a test in which only Xmat or of our analyses. One further point needs to be taken into Xpat are included without X in the full model. Lastly, it consideration with regards to the current set of analyses: should be noted that a POE may be significant with both the power estimates from the literature are for the QTDT paternal and maternal transmissions separately and there orthogonal model. In the absence of population stratifi - may still be a significant difference between them, if the cation (as is the case in our study), the total association association trends are discordant across both paren- model can be used, and, all other things being equal, this tal transmissions (i.e., the same allele may increase the model has greater power than the orthogonal model [59]. score significantly when inherited from the mother but Regarding the sample sizes in the various tests, for the decrease it significantly when inherited from the father, general test, all children who had non-missing genotypes or vice versa). and IBD information for a given marker and non-missing phenotypes were used in the test for that marker. In the The power and effective sample size of a QTDT analysis POE tests, two groups of children are included: (i) chil- The power of a QTDT analysis depends on several fac - dren whose both parents are genotyped and where one tors, including: the marker allele frequencies, the effect parent is homozygous, or whose mother and father have size, the linkage disequilibrium between the marker and different genotypes (in addition, when paternal parent- the quantitative trait locus, the number of child geno- of-origin effects are tested, the father must be heterozy - types in the analysis and the parental genotypes. Stud- gous and, when maternal effects are tested, the mother ies which evaluated family-based association methods must be heterozygous), and (ii) all children with at least used simulations of models with the above parameters to one homozygous parent, even if the other parent has a estimate the power of those methods. For dichotomous missing genotype [65]. This may reduce the sample size traits, for example, 300 case-mother duos offered rea - based on parental genotypes, which is why we report the sonable power for detection of child genetic effects [61], number of informative probands (probands who meet all when the effects were R = 1.5 and R = 2.25 (see second the above criteria (for the general test, and, where appli- 1 2 footnote for an explanation of the parameters), the base- cable, the additional criteria for the POE tests) for each line risk was 0.1, the significance level was 0.05, and the association in the top results. risk allele frequency was 0.3. When strong POEs operate and are included in the model, some methods achieved Statistical analyses—correction for multiple testing power of ~ 90% with as few as 100 case-parents trios and quality measures for GWAS results [62], with I = 2.5 or I = 2.5, a significance level of 0.05, We employed the following strategy for correction for p m a baseline risk of 0.05 or 0.01 and a risk allele frequency multiple testing in this study: in the GWASs, we present of 0.3 or 0.1, for 20% and 80% of the population, respec- all the associations that met the following two criteria: tively. With regards to quantitative traits, as relevant to (i) they pass the conventional genome-wide signifi - –8 this study, we considered published reports of simula- cance threshold (P ≤ 5 × 10 ), (ii) for POE associations, tions estimating the power of various QTDT models. For they have P ≤ 0.0008 in the test of difference between example, in the original QTDT paper, assuming a maxi- paternal and maternal alleles, which was calculated as mum D’, h of 0.1, a risk allele frequency of 0.5, a signifi - the conventional threshold (0.05) Bonferroni-corrected cance level of 0.001 and including parental genotypes, a for the number of post hoc tests for POE associations sample of 480 children (families with a sibship of 1 and which met the first criterion (n = 63). We then prior- parental genotypes available) resulted in a power esti- itize associations that, in addition to meeting the above mate of 97.4% [57]. In another study, a power of 74% was two criteria, also meet the following criteria: (iii) they achieved with a sample size of 200, h of 0.1, and a risk have a p-value (in the GWAS) equal to or below the allele frequency of 0.3 [63]. We can translate the effects conventional significance threshold (0.05) Bonferroni- of an allele into proportion of variance explained (PVE) corrected for the actual number of tests performed –9 using the following formula, taken from the supplemen- across all GWASs (n = 299,479 × 24), i.e., P ≤ 7 × 10 ; tary note of a previous study [64]: and (iv) at least 30 children had the minor allele for 2 2 2 PVE = 2 × β × MAF × (1 − MAF) 2 × β × MAF × (1 − MAF) + SE(β) × 2 × N × MAF × (1 − MAF) , the associated marker (N.B.: this is not the same as the where β is the effect size, SE(β) is the standard error of β, MAF is the minor allele frequency of the marker (we used number of informative probands for QTDT, but rather Nudel et al. Behavioral and Brain Functions (2022) 18:14 Page 10 of 20 it means that at least 30 children in the sample had the the observed χ distribution from each GWAS divided minor allele for the marker in question; while this does by qchisq(0.5, 1). not guarantee that a specific number of children in a given test had the allele (as this also depended on the Functional annotation of variants and genes factors explained earlier), it could highlight associa- For functional annotation of variants, we used the eQTL- tions for which the effect size is less likely to be biased Gen [67] portal and the GTEx V8 portal [68] for finding due to one of the alleles being relatively rare). Associa- expression quantitative trait locus (eQTL) associations tions surviving all four criteria are discussed in more and PhenoScanner [69] for finding DNA methylation and detail in the Results section. For these associations, we histone modification associations for the associations also repeated the relevant association test while adding meeting our four study-wide criteria for significance. For a covariate for the high risk status (HRS) of the fam- gene-level annotation we used VarElect [70], which ranks ily (that is, a dummy variable (0/1) for whether the genes based on their association with free text keywords child is from a family with a parent with schizophrenia using the GeneCards [71] database. or bipolar disorder, or from a family in which neither parent has either of these diagnosis). Additionally, we Results used EMIM v3.22 [43], a program for multinomial fam- Across all 24 GWASs, 88 associations achieved genome- ily-based genetic association models, to test for asso- –8 wide significance (P ≤ 5 × 10 ), of which 25 were high- ciation between the top results (Table 2) and the HRS lighted in the general test and the rest were highlighted as a binary outcome. We used a model for child trend in the POE tests. Additional file 2: Fig. S1 shows Manhat- analysis [61] in which the factor by which the risk of tan plots for all 24 GWASs, and Additional file 3: Fig. S2 disease is multiplied when the child has two risk alleles shows the corresponding QQ plots. Across all analyses, is constrained to be the square of the risk from having the genomic inflation factor ranged from 0.967 to 1.077 one risk allele, or, using the aforementioned notation, (with a mean value of 1.008 and a standard deviation of R = R . In this analyses we used both case and control 2 1 0.024). Of the POE associations among the aforemen- family subsets, but we did not use controls without par- tioned 88 associations, only 23 were significant in the test ents, since EMIM does not distinguish between con- of difference between paternal and maternal alleles after trols and individuals with an unknown disease status correction for multiple testing (Methods), and the rest (which means that parents, who by definition have an were therefore excluded from downstream analyses. The unknown HRS, might be used as controls if the child 48 remaining associations are shown in Table 2. does not have genetic data for a given marker). The Of the 48 associations that were genome-wide signifi - p-values for this test are derived from the χ distribu- cant and, where applicable, showed a significant differ - tion with one degree of freedom (since only one risk ence between paternal and maternal alleles, only 3 met parameter was freely estimated in the full model), and our extra conditions pertaining to the study-wide sig- the test statistic comes from twice the difference in the nificance level and a minimum number of 30 probands log-likelihoods of a null model (in which the multipli- with the minor allele. Regional association plots for cative risk parameter is fixed to 1) and a full model, in these 3 markers are shown in Fig. 3. We employed these which it is estimated from the data. extra criteria to identify more robust associations, espe- QTDT does not output standard errors (SEs) for the cially because very rare alleles could lead to biased effect estimates it computes. In order to obtain SEs for the sizes. Of the 3 associations meeting all four criteria, 2 observed effect in the top associations in our results we were with the WISC Arithmetic score and were high- used the following approach: using the χ statistics lighted in the general test and the remaining association from the QTDT output, we calculated the error as was with the SRS score and showed a maternal POE. SE = X χ , where X is the effect size from QTDT. Two of these associations were with intragenic variants: This is an approximation of the SE, because it is calcu - rs214831 (general test, associated with WISC Arithme- lated from a Wald statistic, whereas QTDT uses a likeli- tic) in Transglutaminase 3 gene (TGM3) and rs7604835 hood ratio test for two nested models which differ by (maternal POE test, associated with SRS) in the Cal- the presence of the effect of the genetic variant, but cium Channel, Voltage-Dependent, Beta 4 Subunit gene these two methods are at least asymptotically equiva- (CACNB4). Marker rs214831 was strongly associated lent [66]. Lastly, the genomic inflation factor was calcu - with the expression of the gene it was located in, namely, 2 –34 lated for each GWAS in R using the χ statistics from TGM3, on eQTLGen (P = 5.72 × 10 ), whereby the A the QTDT output directly (as QTDT rounds the p-val- allele was associated with higher expression of the gene; ues themselves in the output) as follows: the median of in our study, the effect allele (G) was associated with a lower test score, suggesting that lower expression would Nudel et al. Behavioral and Brain Functions (2022) 18:14 Page 11 of 20 ff Table 2 Top results from the GWASs Test Trait Marker ID Chromosome Position MAF Gene Eect allele Other allele Probands Effect SE χ p-value (protein coding) –8 General WISC Arithmetic psy_rs7826548 8 2955158 0.4717 CSMD1 C T 389 − 0.199 0.036 31.12 2 × 10 –8 General WISC Arithmetic rs2554728 8 3842999 0.2731 CSMD1 T C 396 − 0.155 0.028 30.47 3 × 10 –8 General WISC Arithmetic rs11995240 8 3848025 0.1691 CSMD1 T C 396 − 0.016 0.003 29.84 5 × 10 –8 General WISC Arithmetic rs17068473 8 3854987 0.2854 CSMD1 G T 396 0.183 0.033 30.72 3 × 10 –8 General WISC Arithmetic rs4875262 8 3876265 0.3858 CSMD1 G A 396 0.262 0.046 31.95 2 × 10 –8 General WISC Arithmetic rs2740929 8 3879918 0.4877 CSMD1 C T 396 − 0.191 0.034 31.03 3 × 10 –9 General WISC Arithmetic rs2740878 8 3918118 0.3288 CSMD1 A G 396 − 0.309 0.054 33.05 9 × 10 –8 General WISC Arithmetic rs2552166 8 3920395 0.1734 CSMD1 G A 396 − 0.326 0.057 32.16 1 × 10 –8 General WISC Arithmetic rs6049002 20 296172 0.3721 C T 398 0.132 0.024 30.56 3 × 10 –9 General WISC Arithmetic rs6117457 20 733963 0.2883 T C 398 − 0.359 0.062 33.9 6 × 10 –8 General WISC Arithmetic psy_rs79359757 20 937853 0.07381 C T 398 − 0.053 0.01 30.01 4 × 10 –8 General WISC Arithmetic rs7261002 20 1002656 0.1835 G A 398 0.366 0.064 32.9 1 × 10 –8 General WISC Arithmetic rs6118727 20 1030235 0.4465 A G 398 0.071 0.013 30.15 4 × 10 –9 General WISC Arithmetic rs200896 20 1789409 0.4292 A C 398 − 0.326 0.056 33.38 8 × 10 –8 General WISC Arithmetic psy_rs200888 20 1796461 0.3931 G T 396 0.138 0.025 30.49 3 × 10 –8 General WISC Arithmetic psy_rs4813309 20 1880550 0.1331 SIRPA C T 397 − 0.444 0.078 32.7 1 × 10 –9 General WISC Arithmetic psy_rs6035018 20 1882954 0.19 SIRPA T C 398 − 0.392 0.068 33.07 9 × 10 –8 General WISC Arithmetic exm1519370 20 1896100 0.3994 SIRPA C T 388 − 0.128 0.023 30.41 4 × 10 –8 General WISC Arithmetic psy_rs73069290 20 1904515 0.06936 SIRPA G T 398 0.068 0.012 30.03 4 × 10 –8 General WISC Arithmetic rs6035139 20 1936275 0.06014 G A 398 − 0.905 0.163 30.94 3 × 10 –9 General WISC Arithmetic rs214831 20 2321363 0.3967 TGM3 G A 398 − 0.395 0.067 35.28 3 × 10 –8 General WISC Arithmetic rs6137776 20 2338454 0.2392 T C 398 0.281 0.05 31.82 2 × 10 –8 General WISC Symbol Search rs2673776 6 152522812 0.4429 SYNE1 C A 396 0.163 0.029 31.84 2 × 10 –8 General WISC Symbol Search exm2270431 6 152644111 0.4565 SYNE1 C T 396 0.146 0.026 31.44 2 × 10 –8 General WISC Symbol Search psy_rs9478324 6 152677815 0.05058 SYNE1 G A 396 0.287 0.052 30.51 3 × 10 –8 Maternal SRS rs1451197 2 2741446 0.4993 G A 213 − 4.888 0.872 31.42 2 × 10 –8 Maternal SRS rs2176347 2 45968233 0.4473 PRKCE T G 231 − 3.441 0.624 30.39 4 × 10 –9 Maternal SRS rs7604835 2 152881908 0.1315 CACNB4 G A 287 − 11.482 1.927 35.51 3 × 10 –9 Maternal SRS psy_rs77672109 4 35576468 0.01662 C A 304 − 58.859 10.144 33.67 7 × 10 –9 Maternal SRS rs11934637 4 92981939 0.0289 T G 304 − 34.434 5.853 34.61 4 × 10 –8 Maternal SRS psy_rs146416593 5 60662093 0.01592 ZSWIM6 G A 324 − 33.216 5.844 32.31 1 × 10 –10 Maternal SRS psy_rs79166730 5 175229558 0.01951 CPLX2 G T 321 − 37.303 5.817 41.13 1 × 10 Nudel et al. Behavioral and Brain Functions (2022) 18:14 Page 12 of 20 ff Table 2 (continued) Test Trait Marker ID Chromosome Position MAF Gene Eect allele Other allele Probands Effect SE χ p-value (protein coding) –9 Maternal SRS exm573219 6 116325108 0.01879 FRK G A 325 − 58.949 9.981 34.88 4 × 10 –9 Maternal SRS psy_rs78989171 8 49645477 0.01301 EFCAB1 G T 308 − 58.867 10.204 33.28 8 × 10 –10 Maternal SRS exm754630 9 74319677 0.01806 TMEM2 T C 322 − 26.113 4.085 40.87 2 × 10 –9 Maternal SRS rs1681993 12 63338414 0.01084 G A 324 − 58.96 10.14 33.81 6 × 10 –9 Maternal SRS rs7956933 12 63345018 0.01085 A G 323 − 58.964 10.151 33.74 6 × 10 –9 Maternal SRS psy_rs10860381 12 99309750 0.01375 ANKS1B G A 305 − 58.851 10.206 33.25 8 × 10 –8 Maternal WISC Arithmetic rs11784069 8 2119582 0.3736 T G 271 0.341 0.061 31.1 2 × 10 –8 Maternal WISC Arithmetic rs7261002 20 1002656 0.1835 G A 314 0.661 0.116 32.68 1 × 10 –8 Maternal WISC Symbol Search rs2256135 6 152464839 0.4566 SYNE1 G A 262 0.264 0.047 31.95 2 × 10 –9 Paternal RIST index exm693219 8 28929739 0.02746 KIF13B G A 371 42.012 7.191 34.13 5 × 10 –9 Paternal SRS exm109001 1 156314440 0.01517 TSACC T G 320 − 42.257 7.231 34.15 5 × 10 –17 Paternal SRS psy_rs191695175 8 122841477 0.01016 T C 326 − 47.854 5.665 71.36 3 × 10 –9 Paternal SRS rs16908233 11 21604897 0.01158 A G 317 − 34.271 5.86 34.2 5 × 10 –8 Paternal SRS psy_rs117476444 13 113442872 0.02168 ATP11A G A 320 − 33.185 5.847 32.21 1 × 10 –8 Paternal WISC Arithmetic rs11784069 8 2119582 0.3736 T G 271 − 0.273 0.049 30.71 3 × 10 –8 Paternal WISC Arithmetic rs2740939 8 3872513 0.4899 CSMD1 C A 253 − 0.244 0.044 30.73 3 × 10 Results meeting criteria (i) and (ii) (see “Materials and methods”) are shown. Results meeting criteria (iii) and (iv) are shown in boldface. MAF: minor allele frequency (in founders; does not necessarily correspond to the effect allele frequency); SE, standard error Nudel et al. Behavioral and Brain Functions (2022) 18:14 Page 13 of 20 Fig. 3 Regional association plots for associations surviving all four statistical quality criteria (Methods). a rs6117457 (general test, WISC Arithmetic); b rs6117457 (general test, WISC Arithmetic); c rs7604835 (maternal POE test, SRS) Nudel et al. Behavioral and Brain Functions (2022) 18:14 Page 14 of 20 be associated with a lower score. It was also associated study. We therefore tested whether these markers were with the expression of PTPRA in the basal ganglia on themselves associated with the high risk status as the GTEx (P = 0.000022), with allele G being associated with outcome; none of the markers in Table 2 were associated lower expression. This marker remained at least nomi - with it after Bonferroni correction for multiple testing, nally significant when adding a covariate for high risk and the top three markers were not nominally associated status (i.e. for whether the child comes from a high risk even before correction. Thus, if, for these markers, the family or a control family) to the model (P = 0.0312). high risk status of the family is not associated with the Marker rs7604835, which showed a maternal POE in our genetic exposure, then this eliminates both the potential study, was associated with multiple DNA methylation confounding and potential collider bias from the model, and histone modification sites on PhenoScanner (mini - even if HRS is not included as a covariate. Even though –45 mum P = 1.19 × 10 ), based on evidence from two dif- the high risk status refers to the parent and not the child’s ferent studies [72, 73]. This provides further support for phenotype, this lack of association could suggest that the the association with a POE at this locus. This marker traits highlighted in Table 2 might not be useful endophe- remained genome-wide significant when adding a covari - notypes for schizophrenia or bipolar disorder, but might –9 ate for high risk status (P = 2 × 10 ). The last association nonetheless be associated with other disorders. which met all four criteria, namely, between rs6117457 Most of the associated markers in Table 2 (29 out of 48) and WISC Arithmetic in the general test, did not impli- fall within protein-coding genes. In total, 15 unique genes cate any protein-coding gene, and we could not find any are implicated by at least one genome-wide significant relevant prior association with it in the literature or func- association [meeting criteria (i) and, where relevant, (ii)] tional databases. This marker did not remain significant with a variant within them: ANKS1B, ATP11A, CACNB4, when adding a covariate for high risk status (P = 0.0765). CPLX2, CSMD1, EFCAB1, FRK, KIF13B, PRKCE, SIRPA, Translating the effects of the top markers into PVEs, we SYNE1, TGM3, TMEM2, TSACC and ZSWIM6. Given get: 0.078, 0.08 and 0.11 for rs6117457, rs214831 and that the associations in Table 2 were with the Arithme- rs7604835, respectively. The associations adjusted for tic and Symbol Search subtests of the WISC, the SRS and HRS were in the same direction as before in all cases. It the RIST, we used the following terms together with the should be noted, however, that the interpretation of the gene names when running the VarElect analysis: autism models with the covariate for HRS can be difficult: both OR "working memory" OR behavior OR communication WISC Arithmetic and SRS are associated with the child’s OR intelligence OR "processing speed" OR "Wechsler schizophrenia family status in the VIA 7 study [39, 40]; Intelligence Scale for Children" OR "Reynolds Intellec- since the covariate in this case may imply some genetic tual Screening Test" OR "Social Responsiveness Scale" predisposition to schizophrenia, a disorder which is OR schizophrenia OR "bipolar disorder". The last two genetically correlated with cognitive traits [74], the same terms were added because they represent the disorders SNP could have some association with both the psychiat- based on which the VIA sample had been ascertained. ric disorder and the phenotype of interest. Furthermore, Fourteen out of the fifteen genes were directly associated both schizophrenia and bipolar disorder are complex, with at least one of the terms (i.e., the gene’s GeneCard meaning they have both genetic and environmental risk contained the term), with the average number of associ- factors [75, 76]. Thus, the high risk status of the family, ated terms per gene being 4.43 (± 2.41). Two genes were determined by the presence of a psychiatric diagnosis in associated with 8 terms, the highest number of terms any one of the parents, is influenced both by genetic factors one gene was associated with: Ankyrin Repeat and Sterile and environmental factors; the parental genetic factors Alpha Motif Domain-containing Protein 1B (ANKS1B), influence both the child’s genetics (the exposure) and and Synaptic Nuclear Envelope Protein 1 (SYNE1). Addi- the high risk status of the family (the parent’s illness and tional file 4: Table S1 lists all direct associations between potential covariate), which could influence the outcome the terms and the genes and a discussion of the scores. in the child (the investigated trait). Similarly, environ- The gene with the highest VarElect score was CUB And mental factors, which may be unmeasured (or external Sushi Multiple Domains 1 (CSMD1), and the gene with factors in general e.g., parental IQ), could influence both the highest average disease causing likelihood was the the high risk status of the family and the investigated trait aforementioned CACNB4. in the child. In this scenario, adjusting for the covariate may reduce bias from possible confounding but intro- Discussion duce collider bias. A further complication would be the Our study investigated eight neurocognitive, motor and fact that most of our tests were for POEs, which limit the social-cognitive and social-behavioral functions using genetic causal path but not the causal path of the high a family-based GWAS design, including a general asso- risk status on which families were ascertained in this ciation test as well as tests of parent-of-origin effect tests. Nudel et al. Behavioral and Brain Functions (2022) 18:14 Page 15 of 20 We have identified 48 genome-wide significant associa - with the largest number of terms: CSMD1, ANKS1B and tions, of which 3 met our study-wide significance thresh - SYNE1. CSMD1 is of particular interest because it has old. Our results highlighted several protein-coding genes, been implicated in schizophrenia [85–87]. Interestingly, some of which have been implicated in prior genetic in our study, this gene was implicated through markers analyses of relevant phenotypes. associated with a measure of working memory; a study Two genes were highlighted through associations of this gene reported that a schizophrenia risk variant which met all four of our significance criteria: TGM3 in CSMD1 was associated with spatial working memory and CACNB4. The association with TGM3 was further [88]. This could illustrate the effect of a genetic variant supported by the marker’s being an eQTL for the gene. on an endophenotype for schizophrenia. In this con- This gene is involved in terminal epidermal differentia - text it is also important to note a proposal to redefine tion and has been implicated in some cancers [77, 78]. the notion of endophenotype in psychiatry to allow it to In our study, the marker in this gene was associated with include transdiagnostic traits that may be shared across a measure of working memory. Interestingly, previous several disorders [42]. ANKS1B was implicated through studies have found relevant associations between the the association between the marker psy_rs10860381 gene and related phenotypes: a study of the RNA blood and social responsiveness in the maternal POE test. This transcriptome of patients with Alzheimer’s disease (AD), gene encodes an activity dependent postsynaptic effec - a disease which involves severe memory impairments, tor protein highly expressed in the brain, and it has been found that the largest expression fold change among dif- implicated in a wide array of neurodevelopmental phe- ferentially expressed genes across AD cases and controls notypes [89]. Importantly, haploinsufficiency of this gene was with TGM3 [79]. Genes of the same family have been in a mouse model resulted in impaired social interaction implicated in several neurodegenerative diseases [80]. and sensorimotor dysfunction, which are core features of Also of note, the associated marker in our study was also autism spectrum disorder [90]. Even more importantly, a brain eQTL for PTPRA, a gene which is important for this gene exhibits allelic expression imbalance in the hippocampal neuronal migration; mice deficient for the brain, which could be an outcome of genomic imprinting PTPRA protein exhibit impairments in learning and (which could result in a POE), although this is only one short-term memory [81]. The association between social possible explanation [89]. SYNE1 was implicated through responsiveness (SRS) and CACNB4 was with a mater- the associations between several markers and process- nal POE. This marker was also associated with methyla - ing speed (WISC Symbol Search) in the general test. The tion and histone modifications sites, providing further gene encodes a protein that is involved in anchoring spe- support for a POE. The gene encodes a member of the cialized myonuclei underneath neuromuscular junctions, beta subunit family of voltage-dependent calcium chan- but it is also expressed in the brain—predominantly in nels, and it belongs to a family of genes which has been the cerebellum [91]. It has been implicated in a recessive implicated in several psychiatric and neurodevelopmen- form of cerebellar ataxia, which may also include cogni- tal disorders, including autism spectrum disorder, across tive deficits [91]. Interestingly, individuals with SYNE1 many studies [82]. The subunit encoded by CACNB4, mutations exhibit processing speed deficits compared specifically, is highly expressed in the brain and is prom - with controls [92], which is in line with our result show- inent in the cerebellum [83]. A recent study found that ing association between this gene and processing speed a pathogenic missense variant in this gene resulted in a in the general test. Both SYNE1 and TGM3 have been severe neurodevelopmental impairment which included highlighted in a study of de novo mutations in autism intellectual disability, language impairment, movement spectrum disorder [93]. impairment and seizures [84]. When adding a covariate Some of the other associations in Table 2 are also for high risk status to the statistical models for the top of note. The paternal POE association between social associations, we observe that it either slightly improved responsiveness and rs191695175 was the most sig- the association (with SRS, maternal POE test) or reduced nificant association in our study. The minor allele it drastically (with WISC Arithmetic, general test). frequency for this marker was very low at ~ 0.01 (in Whether or not it is appropriate to include this covari- founders), which could lead to a biased effect size. ate in the model depends on the causal paths between However, this marker is found on chromosome 8 in the genetic variant, the trait, and the covariate, which chromosomal band 8q24.13, a locus which was part are complex and not known. Hence, the interpretation of of a suggestive linkage peak for the same trait, namely, these post hoc tests should be done with caution. SRS, in a genome-wide linkage study [13]. The same Among the other genes in Table 2, three genes were locus also showed linkage to SRS in addition to an anxi- highlighted in the functional annotation either as hav- ety score and a score for pragmatic language skills, in ing the highest VarElect score or as being associated another study [94]. These studies, however, did not Nudel et al. Behavioral and Brain Functions (2022) 18:14 Page 16 of 20 model POEs. Thus, even though we may not be able to domains. This does not mean that traits from the other trust the estimated effect size for this locus, the asso - domains would not make good endophenotypes; our ciation itself might be valid and supported by previ- study did not examine that, and the lack of genetic asso- ous studies, and it is possible that the POE, if it indeed ciation could result from lower heritability for those operates at this locus, contributed to the stronger signal traits and/or insufficient sample sizes. in this study as compared with previous studies. There have been other previous studies which included similar Limitations of our study phenotypes, such as social interaction and social com- Our results should be evaluated in the light of several munication (neither was measured with the SRS), but potential limitations. Firstly, our study sample was a they did not model POEs, and their significant results family-based sample, and, as such, not a very large one. do not overlap with ours [95, 96]. We also observe an While this has the advantage of our being able to have interesting association trend with marker rs11784069: a deeply phenotyped sample, it can be detrimental to allele T, when inherited from the mother, is associated genetic association studies. While, as shown in previous with a higher WISC Arithmetic score (better work- simulations studies of QTDT models, our sample should ing memory functions), but, when inherited from the be large enough to detect some effects, it is to expected father, it is associated with a lower score (Table 2). This that only strong effects could be detected in our sample, is an illustration of the phenomenon mentioned in the which can explain why the majority of our genome-wide introduction, namely, opposite POEs of different paren - significant associations were intragenic. It should also be tal types at the same locus, which has been observed emphasized that some of the effect sizes could be over - for other quantitative traits in humans. This marker is estimated due to confounding. As it is difficult to deter - a highly significant eQTL for MYOM2 on eQTLGen mine the appropriateness of the adjustment for high risk –310 –20 (P = 3.2717 × 10 ) and GTEx (P = 9.5 × 10 ) in status, it should be borne in mind that the effects for whole blood. Interestingly, the mouse ortholog of this some associations might not be accurate. However, since gene, Myom2, was significantly upregulated and had the GWASs were performed with the goal of discovering the fifth largest fold change among upregulated genes new genetic associations for downstream analyses and in the hippocampus of memory-enhanced mice in one not for estimating their effects, we adopted this approach study [97], which is relevant for the association in our rather than potentially over-adjust the models, as dis- study, as the WISC Arithmetic score is a measure of cussed earlier. Another limitation is that we did not have working memory. a suitable replication sample which included the same phenotypes and genetic data from children and parents. Although our candidate genes have been highlighted in The top results in the context of endophenotypes previous studies of related traits, providing more cred- and the investigated domains ibility to their association with our traits, the associations The traits implicated by the top results in our study, with specific variants need to be replicated in an inde - namely, SRS (social responsiveness) and WISC Arith- pendent sample. metic (working memory), had been proposed as endo- phenotypes for ASD and ADHD, respectively [13, 98–100]. However, these studies did not identify links Future perspectives between specific genes and these endophenotypes at It has been shown that the heritability of cognitive abil- a genome-wide significant level; they focused on link - ity increases from childhood to young adulthood [102]. age analyses or candidate genes, and, where association Interestingly, a similar trend (reaching its peak around was modeled, it was only suggestive. Thus, our study age of 13 for girls and 14 for boys) was observed for provides genetic evidence for the missing piece in the height [103]. When the proportion of phenotypic vari- pathway from gene to disorder through endophenotype, ance explained by genetics increases, the proportion of namely: TGM3 → working memory → ADHD and the variance explained by the environment decreases, and CACNB4 → social responsiveness → ASD, through the vice versa. In the case of height, this trend could reflect top genetic associations we identified. Similarly, memory the effect of early childhood living conditions and/or pre - impairments, including verbal working memory impair- natal environmental factors [103]. For cognitive ability, ment, are common feature of schizophrenia [101], sug- the authors theorize that this trend could be a result of gesting further pathways between TGM3, CSMD1 and genotype-environment correlation, whereby their genet- PTPRA and schizophrenia through the working memory ics influences children increasingly in selecting, modify - endophenotype. The highlighted associations in Table 2 ing and creating their own experiences as they grow up belong to the neurocognitive and social-behavioral [102]. From the statistical genetic perspective, a higher Nudel et al. Behavioral and Brain Functions (2022) 18:14 Page 17 of 20 Acknowledgements heritability means that more genetic associations could The authors would like to express their gratitude to the dedicated families be identified if the sample of children were studied when participating in the study. We would also like to thank Heather Cordell for very they are older; this could mean that repeating the analy- helpful discussions on the model parameterizations of parent‑ of‑ origin effects, and Oskar Hougaard Jefsen for a valuable discussion on covariate adjustments ses within the VIA sample with these functions measured and potential biases in the models. in early adulthood could result in further associations. Furthermore, functional studies of the genes highlighted Author contributions RN conceived the study, performed the QC of the genetic data, performed in our study could provide further insight into the molec- the genetic and statistical analyses, analyzed the results, wrote the paper; ular etiologies of the neurodevelopmental disorders RZ wrote the program that tabulated the QTDT output; JO performed data whose endophenotypes were investigated in this study. management for VIA 7, assisted with the QC of the pedigree information, per‑ formed the standardization of test scores which lacked norms; CAJC, NH, DE, KSS, BKB, ANG, DLG contributed to the VIA 7 data collection and/or pilot study; CAJC, NH, BKB were responsible for the non‑ genetic study of the investigated Conclusions traits in the VIA 7 cohort and provided information about the trait scores Our study identified several candidate genes for social- relevant to this study; J‑BG oversaw sample preparation and genotyping and behavioral and neurocognitive functions, implicated performed initial QC on the raw genetic data; TW designed and oversaw the genetic part of the VIA 7 study; KJP, AAET, JRMJ, OM, MN contributed to the either through a general test, or a test of POEs; asso- conception of the VIA 7 study and its design, coordination and funding appli‑ ciations in the latter test were also supported by exter- cations. All authors have read and approved the manuscript. nal studies which had identified methylation or histone Funding modification sites associated with the relevant marker. The VIA 7 project is supported by the Mental Health Services of the Capital Importantly, most of our genome-wide significant asso - Region of Denmark (Region Hovedstadens Psykiatri), the Lundbeck Founda‑ ciations were within protein-coding genes, and many of tion Initiative for Integrative Psychiatric Research (iPSYCH), Aarhus University, the Tryg Foundation and the Beatrice Surovell Haskell Fund for Child Mental these had previously been implicated in studies of related Health Research of Copenhagen. The funding bodies had no influence on the traits and disorders, although many of these previous study design, analysis, or interpretation or on the writing of the manuscript. associations were with rare and/or deleterious muta- Availability of data and materials tions. Our study provides further evidence to the effect Access to the dataset used in the current study is available from the corre‑ that common variants may influence related traits in sponding author upon reasonable request. The program used to tabulate the individuals not diagnosed with severe mental disorders, QTDT output is available in a additional file accompanying this article. and it further supports a role for the highlighted genes in the studied traits, which can be seen as a replication of Declarations those genes’ implications in the previous studies. We did Ethics approval and consent to participate not identify significant associations for traits in some of The study was approved by the Danish Data Protection Agency and follows the other functions/domains included in this study; this all laws concerning the processing of personal data. Permission to draw data from registers was granted by the Danish Ministry of Health. The study could be the result of the lower heritability of those traits, protocol was sent to the Danish Committee on Health Research Ethics, who as well as potentially smaller effects that could not be dis - decided that ethical approval was not needed due to the observational covered in the VIA sample. Our results also illustrate the nature of the study. The genetic part of the study obtained ethical approval from the outset of the study and The Danish High Risk and Resilience Study usefulness of modeling POEs in human genetic studies, –VIA 7 was later incorporated into the protocol (Arv og Miljø—genetics and and, while previous studies focused on an array of quan- environment) as an appendix, which has then been approved by the ethics titative non-social-cognitive, non-social-behavioral, and committee [ARV OG MILJØ: betydning for psykisk sygdom hos børn og unge (H‑B‑2009‑026)]. Written informed consent was obtained from all adult partici‑ non-neurocognitive traits, our study highlights the pres- pants and from the legal guardians of participating children. ence of potential POEs in several of these traits studied in a systematic way, thus providing further evidence for this Consent for publication Not applicable. phenomenon in humans. Competing interests Supplementary Information The authors have no competing interests to declare, but TW states that he has acted as a lecturer and scientific counselor to H. Lundbeck A/S. DE has been The online version contains supplementary material available at https:// doi. employed by H. Lundbeck A/S from March until August 2020. org/ 10. 1186/ s12993‑ 022‑ 00198‑0. Author details Additional file 1. Archive file containing std_qtdt, which can parse and 1 CORE ‑ Copenhagen Research Centre for Mental Health, Mental Health tabulate the QTDT output. Centre Copenhagen, Copenhagen University Hospital, Copenhagen, Den‑ mark. iPSYCH, The Lundbeck Foundation Initiative for Integrative Psychiatric Additional file 2: Figure S1. Manhattan plots for all 24 GWASs. Research, Aarhus, Denmark. Institute of Biological Psychiatry, Mental Health Additional file 3: Figure S2. QQ plots for all 24 GWASs. Centre Sct. Hans, Mental Health Services Copenhagen, Roskilde, Denmark. Additional file 4: Table S1. Direct associations from VarElect for genes in Mental Health Centre for Child and Adolescent Psychiatry ‑ Research unit, Table 1. Mental Health Services in the Capital Region of Denmark, Copenhagen, Denmark. Department of Clinical Medicine, Faculty of Health and Medi‑ cal Sciences, University of Copenhagen, Copenhagen, Denmark. Psychosis Research Unit, Aarhus University Hospital ‑ Psychiatry, Aarhus, Denmark. Nudel et al. Behavioral and Brain Functions (2022) 18:14 Page 18 of 20 Center for Neonatal Screening, Department for Congenital Disorders, Statens 19. Constantino JN, Kennon‑McGill S, Weichselbaum C, Marrus N, Haider Serum Institut, Copenhagen, Denmark. Division of Child and Adolescent A, Glowinski AL, et al. Infant viewing of social scenes is under genetic Psychiatry, Department of Psychiatry, Hospital University Lausanne, Lausanne control and is atypical in autism. Nature. 2017;547(7663):340–4 (Epub University, Lausanne, Switzerland. Center for Neuropsychiatric Schizophrenia 2017/07/13). Research and Center for Clinical Intervention and Neuropsychiatric Schizo‑ 20. Moon AL, Haan N, Wilkinson LS, Thomas KL, Hall J. CACNA1C: associa‑ phrenia Research, Mental Health Services in the Capital Region of Denmark, tion with psychiatric disorders, behavior, and neurogenesis. Schizophr Copenhagen, Denmark. Bull. 2018;44(5):958–65 (Epub 2018/07/10). 21. Dedic N, Pohlmann ML, Richter JS, Mehta D, Czamara D, Metzger MW, Received: 24 February 2022 Accepted: 9 September 2022 et al. Cross‑ disorder risk gene CACNA1C differentially modulates sus‑ ceptibility to psychiatric disorders during development and adulthood. Mol Psychiatry. 2018;23(3):533–43 (Epub 2017/07/12). 22. Eichler EE, Flint J, Gibson G, Kong A, Leal SM, Moore JH, et al. Missing heritability and strategies for finding the underlying causes of complex disease. Nat Rev Genet. 2010;11(6):446–50 (Epub 2010/05/19). References 23. Manolio TA, Collins FS, Cox NJ, Goldstein DB, Hindorff LA, Hunter DJ, 1. Parenti I, Rabaneda LG, Schoen H, Novarino G. Neurodevelopmental et al. Finding the missing heritability of complex diseases. Nature. disorders: from genetics to functional pathways. Trends Neurosci. 2009;461(7265):747–53 (Epub 2009/10/09). 2020;43(8):608–21 (Epub 2020/06/09). 24. Kong A, Steinthorsdottir V, Masson G, Thorleifsson G, Sulem P, Besen‑ 2. Lenartowicz A, Loo SK. Use of EEG to diagnose ADHD. Curr Psychiatry bacher S, et al. Parental origin of sequence variants associated with Rep. 2014;16(11):498 (Epub 2014/09/23). complex diseases. Nature. 2009;462(7275):868–74 (Epub 2009/12/18). 3. Masi A, DeMayo MM, Glozier N, Guastella AJ. An overview of autism 25. Benonisdottir S, Oddsson A, Helgason A, Kristjansson RP, Sveinbjorns‑ spectrum disorder, heterogeneity and treatment options. Neurosci Bull. son G, Oskarsdottir A, et al. Epigenetic and genetic components of 2017;33(2):183–93 (Epub 2017/02/19). height regulation. Nat Commun. 2016;7:13490 (Epub 2016/11/17). 4. Bishop DVM. What causes specific language impairment in children? 26. Mozaffari SV, DeCara JM, Shah SJ, Sidore C, Fiorillo E, Cucca F, et al. Curr Dir Psychol Sci. 2006;15(5):217–21 (Epub 2008/11/15). Parent‑ of‑ origin effects on quantitative phenotypes in a large Hutterite 5. Gottesman II, Gould TD. The endophenotype concept in psychiatry: pedigree. Communications biology. 2019;2:28 (Epub 2019/01/25). etymology and strategic intentions. Am J Psychiatry. 2003;160(4):636– 27. Lawson HA, Cheverud JM, Wolf JB. Genomic imprinting and parent‑ 45 (Epub 2003/04/02). of‑ origin effects on complex traits. Nat Rev Genet. 2013;14(9):609–17 6. Almasy L, Blangero J. Endophenotypes as quantitative risk factors (Epub 2013/08/07). for psychiatric disease: rationale and study design. Am J Med Genet. 28. Weaver JR, Bartolomei MS. Chromatin regulators of genomic imprint‑ 2001;105(1):42–4 (Epub 2001/06/27). ing. Biochem Biophys Acta. 2014;1839(3):169–77 (Epub 2013/12/19). 7. Blokland GAM, Mesholam‑ Gately RI, Toulopoulou T, Del Re EC, Lam 29. Pearson CE. Slipping while sleeping? Trinucleotide repeat expansions in M, DeLisi LE, et al. Heritability of neuropsychological measures in germ cells. Trends Mol Med. 2003;9(11):490–5 (Epub 2003/11/08). schizophrenia and nonpsychiatric populations: a systematic review and 30. Cassidy SB, Dykens E, Williams CA. Prader‑ Willi and Angelman syn‑ meta‑analysis. Schizophr Bull. 2017;43(4):788–800 (Epub 2016/11/23). dromes: sister imprinted disorders. Am J Med Genet. 2000;97(2):136–46 8. Plomin R, von Stumm S. The new genetics of intelligence. Nat Rev (Epub 2001/02/17). Genet. 2018;19(3):148–59 (Epub 2018/01/18). 31. Buiting K. Prader‑ Willi syndrome and Angelman syndrome. Am J Med 9. Visscher PM, Hill WG, Wray NR. Heritability in the genomics era—con‑ Genet C Semin Med Genet. 2010;154C(3):365–76 (Epub 2010/08/31). cepts and misconceptions. Nat Rev Genet. 2008;9(4):255–66 (Epub 32. Chamberlain SJ, Lalande M. Neurodevelopmental disorders involving 2008/03/06). genomic imprinting at human chromosome 15q11–q13. Neurobiol Dis. 10. van Soelen IL, Brouwer RM, van Leeuwen M, Kahn RS, Hulshoff Pol HE, 2010;39(1):13–20 (Epub 2010/03/23). Boomsma DI. Heritability of verbal and performance intelligence in a 33. Nudel R, Simpson NH, Baird G, O’Hare A, Conti‑Ramsden G, Bolton PF, pediatric longitudinal sample. Twin Res Hum Genet. 2011;14(2):119–28 et al. Associations of HLA alleles with specific language impairment. J (Epub 2011/03/24). Neurodevelop Disord. 2014;6(1):1 (Epub 2014/01/18). 11. Ronald A, Viding E, Happe F, Plomin R. Individual differences in theory 34. Nudel R, Simpson NH, Baird G, O’Hare A, Conti‑Ramsden G, Bolton PF, of mind ability in middle childhood and links with verbal ability and et al. Genome‑ wide association analyses of child genotype effects and autistic traits: a twin study. Soc Neurosci. 2006;1(3–4):412–25 (Epub parent‑ of‑ origin effects in specific language impairment. Genes Brain 2008/07/18). Behav. 2014;13(4):418–29 (Epub 2014/02/28). 12. Constantino JN, Todd RD. Autistic traits in the general population: a 35. Pettigrew KA, Frinton E, Nudel R, Chan MTM, Thompson P, Hayiou‑ twin study. Arch Gen Psychiatry. 2003;60(5):524–30 (Epub 2003/05/14). Thomas ME, et al. Further evidence for a parent‑ of‑ origin effect at 13. Lowe JK, Werling DM, Constantino JN, Cantor RM, Geschwind DH. the NOP9 locus on language‑related phenotypes. J Neurodev Disord. Social responsiveness, an autism endophenotype: genomewide 2016;8:24 (Epub 2016/06/17). significant linkage to two regions on chromosome 8. Am J Psychiatry. 36. Connolly S, Anney R, Gallagher L, Heron EA. A genome‑ wide investiga‑ 2015;172(3):266–75 (Epub 2015/03/03). tion into parent‑ of‑ origin effects in autism spectrum disorder identifies 14. Pineda DA, Lopera F, Puerta IC, Trujillo‑ Orrego N, Aguirre‑Acevedo DC, previously associated genes including SHANK3. Eur J Hum Genet. Hincapie‑Henao L, et al. Potential cognitive endophenotypes in multi‑ 2017;25(2):234–9 (Epub 2016/11/24). generational families: segregating ADHD from a genetic isolate. Atten 37. Mott R, Yuan W, Kaisaki P, Gan X, Cleak J, Edwards A, et al. The archi‑ Deficit Hyp Disord. 2011;3(3):291–9 (Epub 2011/07/23). tecture of parent‑ of‑ origin effects in mice. Cell. 2014;156(1–2):332–42 15. Missitzi J, Gentner R, Misitzi A, Geladas N, Politis P, Klissouras V, (Epub 2014/01/21). et al. Heritability of motor control and motor learning. Physiol Rep. 38. Thorup AA, Jepsen JR, Ellersgaard DV, Burton BK, Christiani CJ, Hemager 2013;1(7):e00188 (Epub 2014/04/20). N, et al. The Danish High Risk and Resilience Study—VIA 7—a cohort 16. Williams LR, Gross JB. Heritability of motor skill. Acta Genet Med study of 520 7‑ year‑ old children born of parents diagnosed with either Gemellol. 1980;29(2):127–36 (Epub 1980/01/01). schizophrenia, bipolar disorder or neither of these two mental disor‑ 17. Burton BK, Hjorthoj C, Jepsen JR, Thorup A, Nordentoft M, Plessen KJ. ders. BMC Psychiatry. 2015;15:233 (Epub 2015/10/04). Research review: do motor deficits during development represent an 39. Hemager N, Plessen KJ, Thorup A, Christiani C, Ellersgaard D, Spang KS, endophenotype for schizophrenia? A meta‑analysis. J Child Psychol et al. Assessment of neurocognitive functions in 7‑ year‑ old children at Psychiatry. 2016;57(4):446–56 (Epub 2015/11/19). familial high risk for schizophrenia or bipolar disorder: the danish high 18. Iacono WG, Vaidyanathan U, Vrieze SI, Malone SM. Knowns and risk and resilience study VIA 7. JAMA Psychiat. 2018;75(8):844–52 (Epub unknowns for psychophysiological endophenotypes: integration and 2018/06/22). response to commentaries. Psychophysiology. 2014;51(12):1339–47 40. Christiani CJ, Jepsen JRM, Thorup A, Hemager N, Ellersgaard D, Spang (Epub 2014/11/13). KS, et al. Social cognition, language, and social behavior in 7‑ year‑ old Nudel et al. Behavioral and Brain Functions (2022) 18:14 Page 19 of 20 children at familial high‑risk of developing schizophrenia or bipolar through maternal effects and that may be subject to parental imprint ‑ disorder: the Danish high risk and resilience study VIA 7—a popula‑ ing. Am J Hum Genet. 1998;62(4):969–78 (Epub 1998/06/13). tion‑based cohort study. Schizophr Bull. 2019;45(6):1218–30 (Epub 63. Lange C, DeMeo DL, Laird NM. Power and design considerations for a 2019/03/11). general class of family‑based association tests: quantitative traits. Am J 41. Burton BK, Thorup AAE, Jepsen JR, Poulsen G, Ellersgaard D, Spang KS, Hum Genet. 2002;71(6):1330–41 (Epub 2002/11/28). et al. Impairments of motor function among children with a familial 64. Shim H, Chasman DI, Smith JD, Mora S, Ridker PM, Nickerson DA, risk of schizophrenia or bipolar disorder at 7 years old in Denmark: an et al. A multivariate genome‑ wide association analysis of 10 LDL sub‑ observational cohort study. Lancet Psychiatry. 2017;4(5):400–8 (Epub fractions, and their response to statin treatment, in 1868 Caucasians. 2017/03/28). PLoS ONE. 2015;10(4):e0120758 (Epub 2015/04/22). 42. Beauchaine TP, Constantino JN. Redefining the endophenotype con‑ 65. QTDT—online reference. https:// csg. sph. umich. edu/ abeca sis/ QTDT/ cept to accommodate transdiagnostic vulnerabilities and etiological docs/ index. html. Accessed 30 March 2020. complexity. Biomark Med. 2017;11(9):769–80 (Epub 2017/09/12). 66. Engle RF. Chapter 13: wald, likelihood ratio, and Lagrange multiplier 43. Howey R, Cordell HJ. PREMIM and EMIM: tools for estimation of mater‑ tests in econometrics. In: Leamer E, Heckman JJ, editors. Handbook nal, imprinting and interaction effects using multinomial modelling. of econometrics. Amsterdam: Elsevier; 1984. p. 775–826. BMC Bioinformatics. 2012;13:149 (Epub 2012/06/29). 67. Võsa U, Claringbould A, Westra H‑ J, Bonder MJ, Deelen P, Zeng B, et al. 44. Henderson S, Sugden D, Barnett A. The movement assessment battery Unraveling the polygenic architecture of complex traits using blood for children. 2nd ed. London: The Psychological Corporation; 2007. eQTL metaanalysis. Nat Genet. 2018. https:// doi. org/ 10. 1101/ 447367. 45. Cools W, Martelaer KD, Samaey C, Andries C. Movement skill assessment 68. GTEx Consortium. The genotype‑tissue expression (GTEx) project. Nat of typically developing preschool children: a review of seven move‑ Genet. 2013;45(6):580–5 (Epub 2013/05/30). ment skill assessment tools. J Sports Sci Med. 2009;8(2):154–68 (Epub 69. Kamat MA, Blackshaw JA, Young R, Surendran P, Burgess S, Danesh J, 2009/01/01). et al. PhenoScanner V2: an expanded tool for searching human geno‑ 46. Wechsler D. Wechsler intelligence scale for children. 4th ed. San Anto‑ type‑phenotype associations. Bioinformatics. 2019;35(22):4851–3 nio: The Psychological Corporation; 2003. (Epub 2019/06/25). 47. White S, Hill E, Happe F, Frith U. Revisiting the strange stories: revealing 70. Stelzer G, Plaschkes I, Oz‑Levi D, Alkelai A, Olender T, Zimmerman mentalizing impairments in autism. Child Dev. 2009;80(4):1097–117 S, et al. VarElect: the phenotype‑based variation prioritizer of the (Epub 2009/07/28). GeneCards Suite. BMC Genomics. 2016;17(Suppl 2):444 (Epub 48. Constantino JN, Davis SA, Todd RD, Schindler MK, Gross MM, Brophy 2016/07/01). SL, et al. Validation of a brief quantitative measure of autistic traits: 71. Stelzer G, Rosen N, Plaschkes I, Zimmerman S, Twik M, Fishile‑ comparison of the social responsiveness scale with the autism diag‑ vich S, et al. The GeneCards Suite: from gene data mining to nostic interview‑revised. J Autism Dev Disord. 2003;33(4):427–33 (Epub disease genome sequence analyses. Curr Protocols Bioinform. 2003/09/10). 2016;54(1):1.30.1‑1.3. 49. Reynolds C, Kamphaus R. Reynolds intellectual assessment scales 72. Chen L, Ge B, Casale FP, Vasquez L, Kwan T, Garrido‑Martin D, et al. (RIAS). Lutz: Psychological Assessment Resources; 2003. Genetic drivers of epigenetic and transcriptional variation in human 50. R Core Team. R: a language and environment for statistical computing. immune cells. Cell. 2016;167(5):1398‑414 e24 (Epub 2016/11/20). Vienna: R Foundation for Statistical Computing; 2014. 73. Bonder MJ, Luijk R, Zhernakova DV, Moed M, Deelen P, Vermaat M, et al. 51. Harrell FEJ, Dupont C. Hmisc: Harrell miscellaneous. 4.5‑0 ed2021. Disease variants alter transcription factor levels and methylation of their 52. Wei TS, Viliam. R package ’corrplot’: visualization of a correlation. 0.90 binding sites. Nat Genet. 2017;49(1):131–8 (Epub 2016/12/06). ed2021. 74. Hubbard L, Tansey KE, Rai D, Jones P, Ripke S, Chambert KD, et al. 53. Nudel R, Christiani CAJ, Ohland J, Uddin MJ, Hemager N, Ellersgaard Evidence of common genetic overlap between schizophrenia and D, et al. Quantitative genome‑ wide association analyses of receptive cognition. Schizophr Bull. 2015;42(3):832–42. language in the Danish high risk and resilience study. BMC Neurosci. 75. Sawa A, Snyder SH. Schizophrenia: diverse approaches to a complex 2020;21(1):30 (Epub 2020/07/09). disease. Science. 2002;296(5568):692–5 (Epub 2002/04/27). 54. Nudel R, Christiani CAJ, Ohland J, Uddin MJ, Hemager N, Ellersgaard 76. Kerner B. Toward a deeper understanding of the genetics of bipolar DV, et al. Language deficits in specific language impairment, attention disorder. Front Psych. 2015;6:105 (Epub 2015/08/19). deficit/hyperactivity disorder, and autism spectrum disorder: an analysis 77. Stacey SN, Sulem P, Gudbjartsson DF, Jonasdottir A, Thorleifsson G, of polygenic risk. Autism Res. 2020;13(3):369–81 (Epub 2019/10/03). Gudjonsson SA, et al. Germline sequence variants in TGM3 and RGS22 55. Purcell S, Neale B, Todd‑Brown K, Thomas L, Ferreira MA, Bender D, et al. confer risk of basal cell carcinoma. Hum Mol Genet. 2014;23(11):3045– PLINK: a tool set for whole‑ genome association and population‑based 53 (Epub 2014/01/10). linkage analyses. Am J Hum Genet. 2007;81(3):559–75. 78. Wu X, Cao W, Wang X, Zhang J, Lv Z, Qin X, et al. TGM3, a candidate 56. Anderson CA, Pettersson FH, Clarke GM, Cardon LR, Morris AP, Zonder‑ tumor suppressor gene, contributes to human head and neck cancer. van KT. Data quality control in genetic case‑ control association studies. Mol Cancer. 2013;12(1):151 (Epub 2013/12/03). Nat Protoc. 2010;5(9):1564–73 (Epub 2010/11/19). 79. Bai Z, Stamova B, Xu H, Ander BP, Wang J, Jickling GC, et al. Distinctive 57. Abecasis GR, Cardon LR, Cookson WO. A general test of association for RNA expression profiles in blood associated with Alzheimer disease quantitative traits in nuclear families. Am J Hum Genet. 2000;66(1):279– after accounting for white matter hyperintensities. Alzheimer Dis Assoc 92 (Epub 2000/01/13). Disord. 2014;28(3):226–33 (Epub 2014/04/16). 58. Abecasis GR, Cherny SS, Cookson WO, Cardon LR. Merlin—rapid analy‑ 80. Martin A, De Vivo G, Gentile V. Possible role of the transglutaminases in sis of dense genetic maps using sparse gene flow trees. Nat Genet. the pathogenesis of Alzheimer’s disease and other neurodegenerative 2002;30(1):97–101 (Epub 2001/12/04). diseases. Int J Alzheimer’s Dis. 2011;2011: 865432 (Epub 2011/02/26). 59. van der Sluis S, Posthuma D. Single‑locus association models. In: Fer ‑ 81. Petrone A, Battaglia F, Wang C, Dusa A, Su J, Zagzag D, et al. Receptor reira MAR, Medland SE, Posthuma D, Neale BM, editors. Statistical genet‑ protein tyrosine phosphatase α is essential for hippocampal neuronal ics: gene mapping through linkage and association. London: Taylor & migration and long‑term potentiation. EMBO J. 2003;22(16):4121–31. Francis; 2008. 82. Heyes S, Pratt WS, Rees E, Dahimene S, Ferron L, Owen MJ, et al. Genetic 60. Pruim RJ, Welch RP, Sanna S, Teslovich TM, Chines PS, Gliedt TP, et al. disruption of voltage‑ gated calcium channels in psychiatric and neuro‑ LocusZoom: regional visualization of genome‑ wide association scan logical disorders. Prog Neurobiol. 2015;134:36–54 (Epub 2015/09/20). results. Bioinformatics. 2010;26(18):2336–7 (Epub 2010/07/17). 83. Dolphin AC. β subunits of voltage‑ gated calcium channels. J Bioenerg 61. Ainsworth HF, Unwin J, Jamison DL, Cordell HJ. Investigation of mater‑ Biomembr. 2003;35(6):599–620. nal effects, maternal‑fetal interactions and parent ‑ of‑ origin effects 84. de Bagneaux PC, von Elsner L, Bierhals T, Campiglio M, Johannsen (imprinting), using mothers and their offspring. Genet Epidemiol. J, Obermair GJ, et al. A homozygous missense variant in CACNB4 2011;35(1):19–45 (Epub 2010/12/25). encoding the auxiliary calcium channel beta4 subunit causes a severe 62. Weinberg CR, Wilcox AJ, Lie RT. A log‑linear approach to case ‑parent ‑ neurodevelopmental disorder and impairs channel and non‑ channel triad data: assessing effects of disease genes that act either directly or functions. PLoS Genet. 2020;16(3):e1008625. Nudel et al. Behavioral and Brain Functions (2022) 18:14 Page 20 of 20 85. Schizophrenia Psychiatric Genome‑ Wide Association Study (GWAS) Publisher’s Note Consortium. Genome‑ wide association study identifies five new schizo ‑ Springer Nature remains neutral with regard to jurisdictional claims in pub‑ phrenia loci. Nat Genet. 2011;43(10):969–76 (Epub 2011/09/20). lished maps and institutional affiliations. 86. Havik B, Le Hellard S, Rietschel M, Lybaek H, Djurovic S, Mattheisen M, et al. The complement control‑related genes CSMD1 and CSMD2 associate to schizophrenia. Biol Psychiat. 2011;70(1):35–42 (Epub 2011/03/29). 87. Cross‑Disorder Group of the Psychiatric Genomics Consortium. Identifi‑ cation of risk loci with shared effects on five major psychiatric disorders: a genome‑ wide analysis. Lancet. 2013;381(9875):1371–9. 88. Koiliari E, Roussos P, Pasparakis E, Lencz T, Malhotra A, Siever LJ, et al. The CSMD1 genome‑ wide associated schizophrenia risk variant rs10503253 affects general cognitive ability and executive function in healthy males. Schizophr Res. 2014;154(1–3):42–7 (Epub 2014/03/19). 89. Younis RM, Taylor RM, Beardsley PM, McClay JL. The ANKS1B gene and its associated phenotypes: focus on CNS drug response. Pharmacog‑ enomics. 2019;20(9):669–84. 90. Carbonell AU, Cho CH, Tindi JO, Counts PA, Bates JC, Erdjument‑ Bromage H, et al. Haploinsufficiency in the ANKS1B gene encoding AIDA‑1 leads to a neurodevelopmental syndrome. Nat Commun. 2019;10(1):3529. 91. Gros‑Louis F, Dupre N, Dion P, Fox MA, Laurent S, Verreault S, et al. Muta‑ tions in SYNE1 lead to a newly discovered form of autosomal recessive cerebellar ataxia. Nat Genet. 2007;39(1):80–5 (Epub 2006/12/13). 92. Gama MTD, Braga‑Neto P, Dutra LA, Alessi H, Maria LA, Gadelha AA, et al. Cognitive and psychiatric evaluation in SYNE1 ataxia. Cerebellum. 2019;18(4):731–7 (Epub 2019/05/03). 93. O’Roak BJ, Deriziotis P, Lee C, Vives L, Schwartz JJ, Girirajan S, et al. Exome sequencing in sporadic autism spectrum disorders identi‑ fies severe de novo mutations. Nat Genet. 2011;43(6):585–9 (Epub 2011/05/17). 94. Piven J, Vieland VJ, Parlier M, Thompson A, O’Conner I, Woodbury‑Smith M, et al. A molecular genetic study of autism and related phenotypes in extended pedigrees. J Neurodevelop Disord. 2013;5(1):30 (Epub 2013/10/08). 95. Yousaf A, Waltes R, Haslinger D, Klauck SM, Duketis E, Sachse M, et al. Quantitative genome‑ wide association study of six phenotypic subdo‑ mains identifies novel genome ‑ wide significant variants in autism spec‑ trum disorder. Transl Psychiatry. 2020;10(1):215 (Epub 2020/07/07). 96. St Pourcain B, Whitehouse AJ, Ang WQ, Warrington NM, Glessner JT, Wang K, et al. Common variation contributes to the genetic architec‑ ture of social communication traits. Mol Autism. 2013;4(1):34 (Epub 2013/09/21). 97. Li C, Dong S, Wang H, Hu Y. Microarray analysis of gene expression changes in the brains of NR2B‑induced memory‑ enhanced mice. Neuroscience. 2011;197:121–31 (Epub 2011/09/20). 98. Loo SK, Rich EC, Ishii J, McGough J, McCracken J, Nelson S, et al. Cogni‑ tive functioning in affected sibling pairs with ADHD: familial clustering and dopamine genes. J Child Psychol Psychiatry. 2008;49(9):950–7 (Epub 2008/07/31). 99. Duvall JA, Lu A, Cantor RM, Todd RD, Constantino JN, Geschwind DH. A quantitative trait locus analysis of social responsiveness in multiplex autism families. Am J Psychiatry. 2007;164(4):656–62 (Epub 2007/04/04). 100. Coon H, Villalobos ME, Robison RJ, Camp NJ, Cannon DS, Allen‑Brady K, et al. Genome‑ wide linkage using the social responsiveness scale in Utah autism pedigrees. Mol Autism. 2010;1(1):8 (Epub 2010/08/04). Re Read ady y to to submit y submit your our re researc search h ? Choose BMC and benefit fr ? Choose BMC and benefit from om: : 101. Conklin HM. Verbal working memory impairment in schizophrenia patients and their first ‑ degree relatives: evidence from the digit span fast, convenient online submission task. American J Psychiatry. 2000;157(2):275–277. https:// doi. org/ 10. thorough peer review by experienced researchers in your ﬁeld 1176/ appi. ajp. 157.2. 275. 102. Haworth CM, Wright MJ, Luciano M, Martin NG, de Geus EJ, van rapid publication on acceptance Beijsterveldt CE, et al. The heritability of general cognitive ability support for research data, including large and complex data types increases linearly from childhood to young adulthood. Mol Psychiatry. • gold Open Access which fosters wider collaboration and increased citations 2010;15(11):1112–20 (Epub 2009/06/03). 103. Jelenkovic A, Sund R, Hur YM, Yokoyama Y, Hjelmborg JV, Moller S, et al. maximum visibility for your research: over 100M website views per year Genetic and environmental influences on height from infancy to early adulthood: an individual‑based pooled analysis of 45 twin cohorts. Sci At BMC, research is always in progress. Rep. 2016;6:28496 (Epub 2016/06/24). Learn more biomedcentral.com/submissions http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Behavioral and Brain Functions Springer Journals http://www.deepdyve.com/lp/springer-journals/a-family-based-study-of-genetic-and-epigenetic-effects-across-multiple-xlVVbfKGuR

Loading next page...

Publisher: Springer Journals
Copyright: Copyright © The Author(s) 2022
eISSN: 1744-9081
DOI: 10.1186/s12993-022-00198-0
Publisher site: See Article on Publisher Site

Abstract

Many psychiatric and neurodevelopmental disorders are known to be heritable, but studies trying to elucidate the genetic architecture of such traits often lag behind studies of somatic traits and diseases. The reasons as to why relatively few genome‑ wide significant associations have been reported for such traits have to do with the sample sizes needed for the detection of small effects, the difficulty in defining and characterizing the phenotypes, partially due to overlaps in affected underlying domains (which is especially true for cognitive phenotypes), and the complex genetic architectures of the phenotypes, which are not wholly captured in traditional case–control GWAS designs. We aimed to tackle the last two issues by performing GWASs of eight quantitative neurocognitive, motor, social‑ cognitive and social‑behavioral traits, which may be considered endophenotypes for a variety of psychiatric and neurodevel‑ opmental conditions, and for which we employed models capturing both general genetic association and parent‑ of‑ origin effects, in a family‑based sample comprising 402 children and their parents (mostly family trios). We identified 48 genome‑ wide significant associations across several traits, of which 3 also survived our strict study‑ wide quality criteria. We additionally performed a functional annotation of implicated genes, as most of the 48 associations were with variants within protein‑ coding genes. In total, our study highlighted associations with five genes (TGM3, CACNB4, ANKS1B, CSMD1 and SYNE1) associated with measures of working memory, processing speed and social behavior. Our results thus identify novel associations, including previously unreported parent‑ of‑ origin associations with relevant genes, and our top results illustrate new potential gene → endophenotype → disorder pathways. Keywords: GWAS, Neurodevelopment, Cognitive functions, Endophenotype, Parent‑ of‑ origin effect Introduction As a species, humans are adept at using communica- tion (both verbal and nonverbal), mental facilities, social interaction abilities and fine motor skills in their every - *Correspondence: merete.nordentoft@regionh.dk day lives. These aptitudes mature during neurodevel - CORE ‑ Copenhagen Research Centre for Mental Health, Mental Health opment. Some individuals, however, have non-typical Centre Copenhagen, Copenhagen University Hospital, Copenhagen, neurodevelopment, which is associated with cognitive, Denmark Full list of author information is available at the end of the article motor, behavioral and/or social-cognitive impairments. © The Author(s) 2022. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. The Creative Commons Public Domain Dedication waiver (http:// creat iveco mmons. org/ publi cdoma in/ zero/1. 0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Nudel et al. Behavioral and Brain Functions (2022) 18:14 Page 2 of 20 Disorders characterized by these impairment are collec- as an endophenotype for schizophrenia [17]. In psychia- tively known as neurodevelopmental disorders, and they try in general, endophenotypes tend to be electrophysi- often exhibit high comorbidity [1]. Many of these dis- ological e.g., electroencephalogram (EEG), eye tracking orders have a strong genetic component, but they often or certain reflexes [18] or behavioral e.g. gaze direction exhibit both genetic and clinical heterogeneity [1–4]. towards specific facial features [19]. An example of a rela - Such a high degree of heterogeneity, in turn, encum- tively highly studied gene → behavioral endophenotype bers studies into the molecular underpinnings of these → disease pathway is that of the Calcium Voltage-Gated disorders. Channel Subunit Alpha1 C (CACNA1C) gene, which is a One strategy which has been proposed as a means to known susceptibility gene for several psychiatric disor- tackle this issue in psychiatric genetics is the use of endo- ders including schizophrenia [20]. A recent study showed phenotypes. Endophenotypes are heritable traits that are that deletions in that gene in mice led to behaviors asso- (typically) convenient to measure and exhibit an associa- ciated with psychiatric disorders [21]. tion with the psychiatric condition; more formally, they Even though many genetic studies of the aforemen- are said to be heritable traits that are associated with the tioned traits and disorders (and of complex traits and disease in the population, are primarily disease-state- diseases, in general) have been conducted, these stud- independent and co-segregate with the disease in families ies together have not identified enough associations to (an additional criterion for complex diseases is that endo- account for the heritabilities of the investigated traits phenotypes found in affected family members be found or diseases, a problem known as “the missing heritabil- in non-affected family members at a higher rate than in ity” [22]. As genome-wide association studies (GWAS) the general population) [5]. Endophenotypes can also become larger, more associations are identified at the be quantitative, in which case they should be “milder” conventional genome-wide significance threshold. How - in unaffected relatives of affected individuals and cor - ever, there are other reasons why some associations elude related with the severity of the disease, and, if this cor- the GWAS design, even as sample sizes grow larger: for relation is not due to disease progression or medication, example, there may be phenotypic heterogeneity not then it could suggest that the correlation with the disease only across individuals, but also in the sense that dif- is by way of disease liability [6]. Many traits that can be ferent studies may use different definitions for disor - measured using standardized tests meet these criteria. ders, different ascertainment criteria and/or different Pertinent to this study is the case of heritable quantitative assessment tools, and, at times, the studied phenotypes traits, which, in turn, may themselves be composites of themselves might reflect several overlapping underlying different measures. For example, it has long been known abilities. From the genetic perspective, an important rea- that general intelligence is heritable [7]. Although the son is that the common GWAS study design, i.e., using issue of what the intelligence quotient (IQ) itself meas- only unrelated individuals and modeling only specific ures is debated, as are the assumptions about the models types of effects, might not capture all the aspects of the estimating its heritability, the overall evidence from twin genetic architecture of a trait [22, 23]. Pertinent to this studies and other family-based studies suggests that a study is the case of the epigenetic phenomenon (i.e., large proportion of the variation in IQ between individ- a heritable phenomenon not caused by changes in the uals is due to additive genetic effects [8, 9]. Specifically, DNA sequence itself) known as parent-of-origin effect indices from subtests of the Wechsler Intelligence Scale (POE), whereby the effect of an allele is dependent on its for Children also have moderate to high heritabilities parental origin. Family-based genetic studies, where both [10]. Moreover, specific tests designed to measure vari - parental DNA and proband DNA are available, are ideal ous phenotypic expressions of autism spectrum disorder, for studying these effects. POEs have been implicated in namely the “Strange Stories” test, which can identify The - many studies of complex traits and diseases [24]. Studies ory of Mind impairments, and the Social Responsiveness have shown that, when these effects do operate but are Scale, which provides a quantitative measure of autistic not modeled, they can be missed in traditional GWAS behavioral traits, have both been shown to have modest designs [25, 26]. Moreover, the same allele may have (“Strange Stories”) to high (Social Responsiveness Scale) opposite effects when inherited paternally vs. maternally heritabilities [11, 12]. In fact, measures from the Social [25, 26]. Responsiveness Scale and from the Wechsler Intelligence Genomic imprinting is the epigenetic mechanism Scale for Children have been successfully used as endo- considered the primary underlying cause of POEs [27]. phenotypes in studies of autism spectrum disorder (ASD) Imprinted loci are loci at which the two parental alleles and attention deficit/hyperactivity disorder (ADHD) [13, are not functionally equivalent (and one of them may 14]. Lastly, both motor skill and motor learning are also even be silenced completely). One molecular mechanism heritable [15, 16], and motor deficits have been suggested that could lead to imprinting is methylation (the presence Nudel et al. Behavioral and Brain Functions (2022) 18:14 Page 3 of 20 of a methyl group on the DNA nucleotide). Allele-specific working memory [39], social responsiveness [40], and methylation in differentially methylated regions (DMRs), motor function [41]. Interestingly, these studies did not or, in this context, imprinting control regions, can lead to find similar differences between children of parents with differential gene expression depending on the parental no diagnosis of schizophrenia or bipolar disorder and origin of the allele [27]. Modification of histones (basic children who had at least one parent with a diagnosis of proteins around which DNA is wound to form nucle- bipolar disorder. osomes, a compact package of DNA which makes it pos- The main goal of our study is thus twofold: (i) to find sible for the DNA to fit within the nucleus of the cell) can specific genotype–phenotype associations for the quan - also result in altered gene expression; protein complexes titative phenotypes from across the aforementioned that modify histones covalently can lead to repression domains, and (ii) to model POEs in addition to general of transcription [28]. POEs may also result from mecha- association to identify associations that would not be nisms other than genomic imprinting, for example, bias captured in case–control GWAS designs. While we do in transmission of specific types of genetic, such as tri - not set out to show that the investigated traits are endo- nucleotide expansions variation, depending on the sex of phenotypes for specific disorders [as mentioned earlier, the parent [29]. Several disorders which involve genomic some of them have already been used as endophenotypes imprinting have strong behavioral and cognitive manifes- in previous studies, and they (or similar traits measured tations. Perhaps the most often-cited examples thereof by other tests) have been shown to be heritable)], they are are Prader-Willi syndrome and Angelman syndrome. The all inherently relevant to neurodevelopment in their own genes involved in both of these disorders map to chromo- right. Moreover, a recent article examining the history some 15q11q13, but different genes are involved in the of the use of endophenotypes in psychiatry proposed to two disorders, and they exhibit opposite POEs (paternal expand the definition to include transdiagnostic traits, for Prader-Willi syndrome and maternal for Angelman which are not necessarily associated with only one dis- syndrome); similarly, the cognitive and behavioral defi - order [42]. In this context, identifying genetic variants cits differ between the two disorders [30]. Most cases of influencing neurodevelopmental traits is an important these disorders are caused by a deletion of the parentally endeavor in its own right. To our knowledge, this is the expressed DNA, but some cases are the result of imprint- first study which examined these four neurodevelopmen - ing defects, leading to aberrant methylation patterns [31, tal domains in the same cohort, incorporating both gen- 32]. In the case of complex neurodevelopmental disor- eral GWAS models and POE models. ders, some notable examples for which POEs have been reported include specific language impairment [33, 34], Materials and methods dyslexia [35] and autism spectrum disorder [36]. A study Participants of 97 traits in mice, where the parent-of-origin of alleles The sample used in this study is part of the Danish High could be determined, found that most of them exhibited Risk and Resilience Study—VIA 7 (hereafter the VIA 7 POEs, to which a large component of their heritability study) [38]. The VIA 7 study recruited children aged 7 was attributable. Moreover, the study showed that non- and their biological parents. Families were recruited from imprinted loci could also exhibit POEs through interac- Danish registries on account of having at least one par- tion with imprinted loci [37]. These examples illustrate ent with a diagnosis of either schizophrenia spectrum the importance of considering POEs in studying behavio- psychosis or bipolar disorder (“high risk” families) or as ral and cognitive phenotypes. control families, in which neither parent had schizophre- Our study aimed to examine both general genetic asso- nia or bipolar disorder; however, these disorder were not ciation as well as parent-of-origin effects, in a deeply investigated directly in this study. Overall, of the 402 chil- phenotyped family-based cohort, in which families were dren with genetic data included in this study (after quality chosen based on the presence (in at least one parent) or control), 244 come from high risk families (schizophre- absence (in both parents) of a diagnosis of schizophrenia nia: 147; bipolar disorder: 97), and 158 come from control or bipolar disorder, and in which DNA from parents and families. The sample size varies per marker per analysis, children was collected, as well as data on a wide array of as the number of informative children depends on the quantitative neurocognitive, motor, social-cognitive and availability of trait data, marker (genotype) data, and, in social-behavioral traits [38]. In prior studies which used the parent-of-origin analyses, parental genotypes as well. this cohort, several of the investigated traits have been We therefore specify the number of informative children shown to differ significantly between children of parents (probands) for all significant results individually. Regard - with no diagnosis of schizophrenia or bipolar disorder ing parental data, only genetic data were used in the asso- and children who had at least one parent with a diagnosis ciation tests. After the quality control described below, of schizophrenia. These included processing speed and there were 261.117 trios, 88.0364 child-mother duos, Nudel et al. Behavioral and Brain Functions (2022) 18:14 Page 4 of 20 24.1713 child-father duos, 17.0366 children, 0.352642 Pearson’s correlation coefficients across the traits in the parents, 0.173135 mothers and 0.0495053 fathers (as well sample of children with genotypes used in this study, as 37.9879 parents without children in the dataset), on which are shown in Fig. 2. This was done using the Hmisc average per marker, as counted with PREMIM [43], with- package v.4.7-0 [51] for R, and the plots were generated out taking siblings into account. These numbers add up with the corrplot package v.0.92 for R [52]. Descriptive to ~ 391 (not counting parents without children in the statistics for the traits are found in Table 1, which also dataset), which is the number of independent children includes p-values from the Shapiro–Wilk normality test with genetic data in our sample (11 families included a as implemented in the shapiro.test function in R. sibling as well). All traits deviated from normality to some degree. However, as the effective sample for the majority of our Phenotypic data tests depends on the parental genotypes, it varies greatly We investigated eight traits derived from different tests across genetic markers (which are tested individually). selected from the comprehensive battery of the VIA 7 Therefore, different subsets of children were used for dif - study: MABC (total score from the Danish version of ferent markers, and it is not practical to try to transform Movement Assessment Battery for Children (Movement the scores so as to force them to have a normal distribu- ABC-2), 2nd edition [44]. N.B.: the norm sample for the tion, when each transformation will not necessarily work Danish version was from the UK, but it has cross-cultural for more than one marker. Moreover, transforming scores validity [45]); WISC Coding [score (total number cor- in this way would hinder the interpretation of the results, rect) from the Coding subtest of the Danish version of as the spaces between scores would have been changed the Wechsler Intelligence Scale for Children, 4th edition unevenly, and, therefore, the interpretation of the effect (WISC-IV) [46]]; WISC Symbol Search [score (total sizes would be problematic. We discuss this at length in number correct) from the Symbol Search subtest of the a previous paper, where we also examined the difference Danish version of the WISC-IV]; SSR (score from the normalization had made for our top result [53]. Lastly, as Strange Stories—Revised [47], based on the total number we explain below, we used variance components (which of correct answers to 8 mentalizing questions translated assume normality) to correct for relatedness among the into Danish); SRS (T-score from the Danish version of children within a given family. As we had only 11 fami- the Social Responsiveness Scale (SRS-2) [48], 2nd edi- lies with more than one child, we investigated the effect tion, completed by the child’s teacher); WISC Arithme- of removing a child from each family and not modeling tic [score (total correct responses) from the Arithmetic the variance components, and we saw that it had very subtest of the Danish version of the WISC-IV]; WISC little impact on our top result [53]; we therefore employ Letter-Number Sequencing [score (number of correct the phenotypic scores as detailed above, without an addi- trials) from the Letter-Number Sequencing subtest of tional transformation. the Danish version of the WISC-IV]; RIST Index [index score from the Danish version of the Reynold’s Intellec- Genetic data tual Screening Test (RIST) [49]]. We had DNA samples from a subset of the VIA 7 study The WISC Arithmetic, WISC Letter-Number Sequenc - sample, and these were genotyped on the Illumina Psy- ing, RIST Index and MABC scores were age-standardized chChip v1-1_15073391_C, which had a loci count of based on the norms from the manual of each respective 603,144 (according to the information lines in the Illu- test. Where norms were not available for some tests or mina manifest file for this array). The dataset has been subtests e.g., when we used the versions of the WISC- described in detail in our previous studies [53, 54]. IV Coding and Symbol Search subtests for children aged Briefly, the quality control (QC) steps for the samples 8 to 16, or when there were no norms (SSR scores), the and markers were as follows: initial QC on raw genetic raw scores were rescaled into Z-scores in SPSS v25.0.0.2 data: individuals with low call rates or discordant sex using the mean of the population control subset of VIA 7 information were removed in the first step, as were children, who were age-matched to the rest of the cohort. markers with a Gentrain score < 0.3. At this point 18 The SRS total T-score was not adjusted for age, as this individuals had been removed (including one possible score was not associated with age in children aged 7–15 duplicate sample), and there were 600,282 markers left [12]. More details about these tests can be found in pre- in the dataset. Subsequent QC was done with PLINK vious publications on the VIA 7 study [39–41]. The dis - [55] v.1.90b5.2: individuals and markers with > 1% tributions of the test scores for each trait are shown in Mendelian errors were removed (N = 10). Genotypes Fig. 1, which contains histograms and density plots for with remaining Mendelian errors below this threshold the traits and was generated in R [50] v3.6.3 using the hist were set to missing. Markers with > 5% missing data and density functions. We also calculated the pairwise were removed (at this point all remaining individuals Nudel et al. Behavioral and Brain Functions (2022) 18:14 Page 5 of 20 MABC WISC Coding WISC Symbol Search SSR 05 10 15 20 −4 −2 02 −3 −2 −1 012 3 −3 −2 −1 0123 Score Score Score Score SRS WISC Arithmetic WISC Letter−Number Sequencing RIST Index 40 60 80 100 2468 10 12 14 05 10 15 60 80 100 120 Score Score Score Score Fig. 1 Histograms and density plots for all phenotypes across all children with non‑missing phenotype values per trait. MABC, movement assessment battery for children; WISC Coding, coding subtest of the Wechsler Intelligence Scale for Children ( WISC); WISC Symbol Search, symbol search subtest of the WISC; SSR, strange stories—revised; SRS, social responsiveness scale; WISC Arithmetic, arithmetic subtest of the WISC; WISC Letter‑Number Sequencing, letter ‑number sequencing of the WISC; RIST Index, index score from the Reynold’s Intellectual Screening Test had < 5% missing data). Individuals with extreme het- as a minor allele frequency (MAF) threshold of 1% (in erozygosity rates (with a threshold of ± 3 SD from the founders). Markers with a significant HWE p-value sample mean) were removed (N = 21). Genetic ances- based on the above threshold or MAF below 1% were try was estimated in a principal component analysis excluded. We removed one marker per pair in case of (PCA). The threshold for the exclusion of samples was pairs of markers with identical positions included in 2 SD above or below the VIA 7 mean for either PC1 or the PsychChip, either using PLINK --list-duplicate- PC2, using the VIA 7 samples and the CEU, CHB, JPT vars suppress-first, if the allele codes matched, other - and YRI HapMap samples to create the PC space, as wise prioritizing markers with rsIDs. The number of described in a published QC protocol [56]. To reduce individuals removed during these steps was 64, includ- bias from population stratification, individuals of diver - ing 3 duplicate samples (note that some samples were gent ancestry were removed along with their relatives flagged in more than one step, in cases in which sev - (N = 36), while the rest of the sample clustered with eral checks were performed before the final exclusion the CEU individuals. Individuals who exhibited cryptic of samples, namely, after the Mendelian errors check). relatedness or who were less related to biological fam- In total, 1094 genotyped individuals and 299,604 auto- ily members than expected from pedigree information somal markers passed these QC steps, from which we were removed (N = 13) (the Pi-hat threshold for the further removed 125 indels, for a total of 299,479 auto- exclusion of individuals expected to be unrelated was somal markers. Only this dataset of genotyped markers 0.185). A Hardy–Weinberg Equilibrium (HWE) p-value was used in this study. Positions in the text and tables –6 threshold of 1 × 10 was employed for markers, as well are in genome build hg19. A minority of markers on the Density Density 0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.00 0.02 0.04 0.06 0.08 0.10 0.12 Density Density 0.00 0.05 0.10 0.15 0.20 0.00.1 0.20.3 0.4 Density Density 0.00 0.05 0.10 0.15 0.0 0.10.2 0.3 0.4 0.5 Density Density 0.00 0.01 0.02 0.03 0.04 0.0 0.1 0.20.3 0.40.5 0.6 Nudel et al. Behavioral and Brain Functions (2022) 18:14 Page 6 of 20 MABC 0.8 WISC Coding 0.6 0.4 WISC Symbol Search 0.2 SSR SRS −0.2 WISC Arithmetic −0.4 −0.6 WISC Letter−Number Sequencing −0.8 RIST Index −1 Fig. 2 Pearson correlations across the investigated traits. All correlation coefficients were different from zero with P < 0.05 array had positions in the Illumina manifest file which the position of interest (the direction depended on the differed slightly from the ones in dbSNP for the same strand), as expected. One marker among the top hits major build. For that reason we checked all the probes had an incorrect position in the manifest file, but the from the manifest file for our array for the top hits in probe mapped to the right place. this study, as for the top hit in our previous study [53]. This was done using the UCSC Genome Browser BLAT Statistical analyses—GWAS stage tool, and, where a marker in our top hits had an rsID, In the GWAS stage, we used QTDT (quantitative trans- we checked that the SNP was indeed 1 bp away from the mission-disequilibrium test) [57] v2.6.1 for the statistical probe. Otherwise we checked that the probe mapped to genetic analyses. MERLIN [58] v1.1.2 was used for esti- a position 1 bp away from the position in the manifest mating identity by descent (IBD) scores for each marker file. All probes mapped to a position 1 bp away from to be used by QTDT. Three tests were performed for MABC WISC Coding WISC Symbol Search SSR SRS WISC Arithmetic WISC Letter−Number Sequencing RIST Index Nudel et al. Behavioral and Brain Functions (2022) 18:14 Page 7 of 20 ‑ ‑ ‑ ‑ ‑ ‑ ‑ ‑ Table 1 Descriptive statistics for the investigated traits in the study sample. POE: parent of origin effect a b b Trait Domain The function Descriptive statistics for phenotypes in the full sampleGeneral test Paternal POE t est Maternal POE t est measured by Children Mean Median Standard Minimum Maximum Shapiro– Minimum Maximum Minimum Maximum Minimum Maximum the test with deviation value value Wilk test Number of Number of Number of Number of Number of Number of phenotype p-value probands probands probands probands probands probands data MABC Motor A combined 399 8.105 8 3.31 1 19 0.0001 337 399 196 382 196 382 score from tests of manual dex terity, aiming and catching, and balance WISC Cod Neuro Processing 401 − 0.134 − 0.059 0.992 − 3.61 3.219 0.0417 338 401 197 384 197 384 ing cognitive speed –5 WISC Neuro Processing 398 − 0.121 − 0.073 0.994 − 3.292 2.767 8.74 × 10 335 398 196 381 196 381 Symbol cognitive speed Search –6 SSR Social Theory of mind 400 − 0.1 0.046 0.999 − 3.227 2.909 1.69 × 10 337 400 197 383 197 383 cognitive –19 SRS Social Social respon 345 49.014 46 10.441 37 106 4.68 × 10 292 345 172 331 172 331 behav siveness ioral –10 WISC Neuro Verbal working 398 9.475 10 2.396 2 15 1.18 × 10 335 398 196 381 196 381 Arithmetic cognitive memory –13 WISC Neuro Verbal working 398 10.508 11 3.103 1 16 7.34 × 10 336 398 196 381 196 381 Letter cognitive memory Number Sequenc ing –10 RIST Index Neuro A combined 401 103.85 105 10.458 56 127 5.75 × 10 338 401 198 384 198 384 cognitive score from tests of non verbal intelligence (approximation of fluid intelli gence) and ver bal intelligence (a measure of crystallized intelligence) QTDT chooses informative probands from the pedigree file. Additionally, probands could have been excluded from the tests based on phenotype missingness, genotype missingness and/or problems with their IBD estimation with MERLIN In addition to the reasons for exclusion in the general test, probands could have also been excluded based on parental genotypes (see “Materials and methods”) Nudel et al. Behavioral and Brain Functions (2022) 18:14 Page 8 of 20 each trait-marker combination: a general (i.e., not a POE significantly different. It could be that a child effect is sig - test) total test of association using all family data (qtdt nificant and appears as such also when looking at pater - -at), a paternal parent-of-origin total test of association, nally inherited alleles or only at maternally inherited in which only paternally inherited alleles were used (-at alleles separately. Therefore, it is necessary to test for a -op) and a maternal parent-of-origin total test of associa- difference between these parental allele transmissions. tion, in which only maternally inherited alleles were used This can be achieved by controlling for risk parameters (-at -om). The total association model (as opposed to the other than the POE parameter by including them in both orthogonal model, which QTDT can also run) is not a the null and the full models. QTDT does not allow a TDT, and it was used because it is more powerful in the free choice of parameters in the null and the full mod- absence of population stratification [59]. In this model, a els, but it incorporates a test for the difference in the combined between/within family component X, or Xpat effects between the paternal and maternal allelic trans - and Xmat in the paternal and maternal tests, respectively, missions (qtdt -at -ot). In this test, the null model has X, denoting the between/within effect on the means, is and the full model has both X and Xmat. It is not pos- tested. X is the effect size reported for the QTDT analy - sible to include Xpat instead of Xmat in the full model ses in this paper. X is estimated from the data in the full (to test for a POE when a paternal POE is suspected); model and is fixed to zero in the null model. The likeli - therefore, as a precaution, we tested both parameteriza- hoods of these two models are then assessed through a tions for a known POE using a different program, EMIM likelihood ratio test, resulting in a χ statistic, which can [43], which allows to model both parental risk parame- be used to compute a one-sided p-value from the χ dis- ters (one at a time in this case) in addition to the child tribution. The tests in this study had one additional free risk parameter, and saw that the overall likelihood of the parameter in the full model as compared with the null full model was roughly the same in both cases. Thus, we model, and so the χ statistic was evaluated with 1 degree used this test to filter out associations that are significant of freedom. We included variance components in both in the GWASs when testing paternally inherited alleles models (-wega), incorporating an environmental compo- and maternally inherited alleles separately, but which do nent, a polygenic component and an additive major locus not show a significant difference from the other parent’s component. This allowed for the use of families with transmissions. Note, however, that these models do not multiple children, although only 11 families included a test for the simple parental effect at the locus of inter - sibling. Age was taken into account in the scoring of the est or for the type of POE (if the POE is real), and, for phenotypes, as explained earlier. For all traits, a covari- ate for sex was added to both the null model and the full 1 2 Consider a previously reported association [34, 54] with a paternal POE for model. The Manhattan plots and the QQ plots were specific language impairment: rs4280164 allele A having a paternal imprint - generated with the “qqman” R scripts by Stephen Turner ing parameter I (a multiplicative factor by which the probability of disease and Daniel Capurso (with the (major update) version is multiplied if the child receives a (paternal) copy of the effect allele from –8 their father) = 0.255 with P = 2.918 × 10 (1 degree of freedom). In this case, from April 19, 2011 for the former type of plot and the the null model has all risk parameters fixed to 1, and in the full model I is version from June 10, 2013 for the latter, available from: freely estimated; the two models are assessed using a likelihood ratio test. https:// github. com/ steph entur ner/ qqman/ bloqb/ v0.0. 0/ As the EMIM software (43) (used in the original study) allows many param- eterizations, it is also possible to include a free R parameter (the factor by qqman.r). Regional association plots were generated with which the disease risk is multiplied if the child has a single copy of the effect LocusZoom [60], after converting marker IDs to rsIDs allele, and assuming that the risk from the child’s having two risk alleles (where possible) using a key from the Illumina website. is R = R ) in both the null and full models and a free I parameter only in 2 1 p 2 –8 the full model. This results in a χ of 28.752840302, P = 8.22 × 10 (1 degree The QTDT output files were tabulated using an in-house of freedom), suggesting that the paternal POE is still significant even when program (included in the Additional file 1), but the sta- allowing for a child effect. Importantly, even when estimating the maternal tistics for the top hits in our study were also examined imprinting parameter, I (a multiplicative factor by which the probability of disease is multiplied if the child receives a (maternal) copy of the effect allele manually using the raw QTDT output, and they matched from their mother), instead of I , at the same locus (with R ), the likelihood p 1 the output of the program. of the full model here provides very similar evidence of association (χ of 28.752840295436), although the parameter estimates are different. Similarly, even if we do not make the assumption that R = R (i.e., we freely estimate Statistical analyses—post hoc tests for paternal 2 1 both R and R in the null and full models), we still obtain very similar likeli- 1 2 and maternal allelic transmission differences hoods, resulting in χ = 25.88917562832 when estimating I in the full model, When a POE is detected with one parent, it does and χ = 25.889147279256 when estimating I in the full model (in both cases including free R and R in both the null and full models), both obtaining not mean that the other parent’s transmissions are 1 2 –7 P = 3.62 × 10 (1 degree of freedom). We believe that QTDT is doing some- thing similar with its test of allelic difference, and that a test of either combi - nation can be used to determine the presence of a POE, even though QTDT N.B.: The Danish manual for the SRS we used did not include a sex-adjust - always models the maternal POE effect parameter as the additional one. ment for the T-score; hence, we used a covariate for sex for this trait as well. Nudel et al. Behavioral and Brain Functions (2022) 18:14 Page 9 of 20 that reason, we only use them to test for the presence of the MAF in founders), and N is the sample size (we used a POE and not for effect estimation; we always report the number of probands). We do this for the top results the effect of the allele from a test in which only Xmat or of our analyses. One further point needs to be taken into Xpat are included without X in the full model. Lastly, it consideration with regards to the current set of analyses: should be noted that a POE may be significant with both the power estimates from the literature are for the QTDT paternal and maternal transmissions separately and there orthogonal model. In the absence of population stratifi - may still be a significant difference between them, if the cation (as is the case in our study), the total association association trends are discordant across both paren- model can be used, and, all other things being equal, this tal transmissions (i.e., the same allele may increase the model has greater power than the orthogonal model [59]. score significantly when inherited from the mother but Regarding the sample sizes in the various tests, for the decrease it significantly when inherited from the father, general test, all children who had non-missing genotypes or vice versa). and IBD information for a given marker and non-missing phenotypes were used in the test for that marker. In the The power and effective sample size of a QTDT analysis POE tests, two groups of children are included: (i) chil- The power of a QTDT analysis depends on several fac - dren whose both parents are genotyped and where one tors, including: the marker allele frequencies, the effect parent is homozygous, or whose mother and father have size, the linkage disequilibrium between the marker and different genotypes (in addition, when paternal parent- the quantitative trait locus, the number of child geno- of-origin effects are tested, the father must be heterozy - types in the analysis and the parental genotypes. Stud- gous and, when maternal effects are tested, the mother ies which evaluated family-based association methods must be heterozygous), and (ii) all children with at least used simulations of models with the above parameters to one homozygous parent, even if the other parent has a estimate the power of those methods. For dichotomous missing genotype [65]. This may reduce the sample size traits, for example, 300 case-mother duos offered rea - based on parental genotypes, which is why we report the sonable power for detection of child genetic effects [61], number of informative probands (probands who meet all when the effects were R = 1.5 and R = 2.25 (see second the above criteria (for the general test, and, where appli- 1 2 footnote for an explanation of the parameters), the base- cable, the additional criteria for the POE tests) for each line risk was 0.1, the significance level was 0.05, and the association in the top results. risk allele frequency was 0.3. When strong POEs operate and are included in the model, some methods achieved Statistical analyses—correction for multiple testing power of ~ 90% with as few as 100 case-parents trios and quality measures for GWAS results [62], with I = 2.5 or I = 2.5, a significance level of 0.05, We employed the following strategy for correction for p m a baseline risk of 0.05 or 0.01 and a risk allele frequency multiple testing in this study: in the GWASs, we present of 0.3 or 0.1, for 20% and 80% of the population, respec- all the associations that met the following two criteria: tively. With regards to quantitative traits, as relevant to (i) they pass the conventional genome-wide signifi - –8 this study, we considered published reports of simula- cance threshold (P ≤ 5 × 10 ), (ii) for POE associations, tions estimating the power of various QTDT models. For they have P ≤ 0.0008 in the test of difference between example, in the original QTDT paper, assuming a maxi- paternal and maternal alleles, which was calculated as mum D’, h of 0.1, a risk allele frequency of 0.5, a signifi - the conventional threshold (0.05) Bonferroni-corrected cance level of 0.001 and including parental genotypes, a for the number of post hoc tests for POE associations sample of 480 children (families with a sibship of 1 and which met the first criterion (n = 63). We then prior- parental genotypes available) resulted in a power esti- itize associations that, in addition to meeting the above mate of 97.4% [57]. In another study, a power of 74% was two criteria, also meet the following criteria: (iii) they achieved with a sample size of 200, h of 0.1, and a risk have a p-value (in the GWAS) equal to or below the allele frequency of 0.3 [63]. We can translate the effects conventional significance threshold (0.05) Bonferroni- of an allele into proportion of variance explained (PVE) corrected for the actual number of tests performed –9 using the following formula, taken from the supplemen- across all GWASs (n = 299,479 × 24), i.e., P ≤ 7 × 10 ; tary note of a previous study [64]: and (iv) at least 30 children had the minor allele for 2 2 2 PVE = 2 × β × MAF × (1 − MAF) 2 × β × MAF × (1 − MAF) + SE(β) × 2 × N × MAF × (1 − MAF) , the associated marker (N.B.: this is not the same as the where β is the effect size, SE(β) is the standard error of β, MAF is the minor allele frequency of the marker (we used number of informative probands for QTDT, but rather Nudel et al. Behavioral and Brain Functions (2022) 18:14 Page 10 of 20 it means that at least 30 children in the sample had the the observed χ distribution from each GWAS divided minor allele for the marker in question; while this does by qchisq(0.5, 1). not guarantee that a specific number of children in a given test had the allele (as this also depended on the Functional annotation of variants and genes factors explained earlier), it could highlight associa- For functional annotation of variants, we used the eQTL- tions for which the effect size is less likely to be biased Gen [67] portal and the GTEx V8 portal [68] for finding due to one of the alleles being relatively rare). Associa- expression quantitative trait locus (eQTL) associations tions surviving all four criteria are discussed in more and PhenoScanner [69] for finding DNA methylation and detail in the Results section. For these associations, we histone modification associations for the associations also repeated the relevant association test while adding meeting our four study-wide criteria for significance. For a covariate for the high risk status (HRS) of the fam- gene-level annotation we used VarElect [70], which ranks ily (that is, a dummy variable (0/1) for whether the genes based on their association with free text keywords child is from a family with a parent with schizophrenia using the GeneCards [71] database. or bipolar disorder, or from a family in which neither parent has either of these diagnosis). Additionally, we Results used EMIM v3.22 [43], a program for multinomial fam- Across all 24 GWASs, 88 associations achieved genome- ily-based genetic association models, to test for asso- –8 wide significance (P ≤ 5 × 10 ), of which 25 were high- ciation between the top results (Table 2) and the HRS lighted in the general test and the rest were highlighted as a binary outcome. We used a model for child trend in the POE tests. Additional file 2: Fig. S1 shows Manhat- analysis [61] in which the factor by which the risk of tan plots for all 24 GWASs, and Additional file 3: Fig. S2 disease is multiplied when the child has two risk alleles shows the corresponding QQ plots. Across all analyses, is constrained to be the square of the risk from having the genomic inflation factor ranged from 0.967 to 1.077 one risk allele, or, using the aforementioned notation, (with a mean value of 1.008 and a standard deviation of R = R . In this analyses we used both case and control 2 1 0.024). Of the POE associations among the aforemen- family subsets, but we did not use controls without par- tioned 88 associations, only 23 were significant in the test ents, since EMIM does not distinguish between con- of difference between paternal and maternal alleles after trols and individuals with an unknown disease status correction for multiple testing (Methods), and the rest (which means that parents, who by definition have an were therefore excluded from downstream analyses. The unknown HRS, might be used as controls if the child 48 remaining associations are shown in Table 2. does not have genetic data for a given marker). The Of the 48 associations that were genome-wide signifi - p-values for this test are derived from the χ distribu- cant and, where applicable, showed a significant differ - tion with one degree of freedom (since only one risk ence between paternal and maternal alleles, only 3 met parameter was freely estimated in the full model), and our extra conditions pertaining to the study-wide sig- the test statistic comes from twice the difference in the nificance level and a minimum number of 30 probands log-likelihoods of a null model (in which the multipli- with the minor allele. Regional association plots for cative risk parameter is fixed to 1) and a full model, in these 3 markers are shown in Fig. 3. We employed these which it is estimated from the data. extra criteria to identify more robust associations, espe- QTDT does not output standard errors (SEs) for the cially because very rare alleles could lead to biased effect estimates it computes. In order to obtain SEs for the sizes. Of the 3 associations meeting all four criteria, 2 observed effect in the top associations in our results we were with the WISC Arithmetic score and were high- used the following approach: using the χ statistics lighted in the general test and the remaining association from the QTDT output, we calculated the error as was with the SRS score and showed a maternal POE. SE = X χ , where X is the effect size from QTDT. Two of these associations were with intragenic variants: This is an approximation of the SE, because it is calcu - rs214831 (general test, associated with WISC Arithme- lated from a Wald statistic, whereas QTDT uses a likeli- tic) in Transglutaminase 3 gene (TGM3) and rs7604835 hood ratio test for two nested models which differ by (maternal POE test, associated with SRS) in the Cal- the presence of the effect of the genetic variant, but cium Channel, Voltage-Dependent, Beta 4 Subunit gene these two methods are at least asymptotically equiva- (CACNB4). Marker rs214831 was strongly associated lent [66]. Lastly, the genomic inflation factor was calcu - with the expression of the gene it was located in, namely, 2 –34 lated for each GWAS in R using the χ statistics from TGM3, on eQTLGen (P = 5.72 × 10 ), whereby the A the QTDT output directly (as QTDT rounds the p-val- allele was associated with higher expression of the gene; ues themselves in the output) as follows: the median of in our study, the effect allele (G) was associated with a lower test score, suggesting that lower expression would Nudel et al. Behavioral and Brain Functions (2022) 18:14 Page 11 of 20 ff Table 2 Top results from the GWASs Test Trait Marker ID Chromosome Position MAF Gene Eect allele Other allele Probands Effect SE χ p-value (protein coding) –8 General WISC Arithmetic psy_rs7826548 8 2955158 0.4717 CSMD1 C T 389 − 0.199 0.036 31.12 2 × 10 –8 General WISC Arithmetic rs2554728 8 3842999 0.2731 CSMD1 T C 396 − 0.155 0.028 30.47 3 × 10 –8 General WISC Arithmetic rs11995240 8 3848025 0.1691 CSMD1 T C 396 − 0.016 0.003 29.84 5 × 10 –8 General WISC Arithmetic rs17068473 8 3854987 0.2854 CSMD1 G T 396 0.183 0.033 30.72 3 × 10 –8 General WISC Arithmetic rs4875262 8 3876265 0.3858 CSMD1 G A 396 0.262 0.046 31.95 2 × 10 –8 General WISC Arithmetic rs2740929 8 3879918 0.4877 CSMD1 C T 396 − 0.191 0.034 31.03 3 × 10 –9 General WISC Arithmetic rs2740878 8 3918118 0.3288 CSMD1 A G 396 − 0.309 0.054 33.05 9 × 10 –8 General WISC Arithmetic rs2552166 8 3920395 0.1734 CSMD1 G A 396 − 0.326 0.057 32.16 1 × 10 –8 General WISC Arithmetic rs6049002 20 296172 0.3721 C T 398 0.132 0.024 30.56 3 × 10 –9 General WISC Arithmetic rs6117457 20 733963 0.2883 T C 398 − 0.359 0.062 33.9 6 × 10 –8 General WISC Arithmetic psy_rs79359757 20 937853 0.07381 C T 398 − 0.053 0.01 30.01 4 × 10 –8 General WISC Arithmetic rs7261002 20 1002656 0.1835 G A 398 0.366 0.064 32.9 1 × 10 –8 General WISC Arithmetic rs6118727 20 1030235 0.4465 A G 398 0.071 0.013 30.15 4 × 10 –9 General WISC Arithmetic rs200896 20 1789409 0.4292 A C 398 − 0.326 0.056 33.38 8 × 10 –8 General WISC Arithmetic psy_rs200888 20 1796461 0.3931 G T 396 0.138 0.025 30.49 3 × 10 –8 General WISC Arithmetic psy_rs4813309 20 1880550 0.1331 SIRPA C T 397 − 0.444 0.078 32.7 1 × 10 –9 General WISC Arithmetic psy_rs6035018 20 1882954 0.19 SIRPA T C 398 − 0.392 0.068 33.07 9 × 10 –8 General WISC Arithmetic exm1519370 20 1896100 0.3994 SIRPA C T 388 − 0.128 0.023 30.41 4 × 10 –8 General WISC Arithmetic psy_rs73069290 20 1904515 0.06936 SIRPA G T 398 0.068 0.012 30.03 4 × 10 –8 General WISC Arithmetic rs6035139 20 1936275 0.06014 G A 398 − 0.905 0.163 30.94 3 × 10 –9 General WISC Arithmetic rs214831 20 2321363 0.3967 TGM3 G A 398 − 0.395 0.067 35.28 3 × 10 –8 General WISC Arithmetic rs6137776 20 2338454 0.2392 T C 398 0.281 0.05 31.82 2 × 10 –8 General WISC Symbol Search rs2673776 6 152522812 0.4429 SYNE1 C A 396 0.163 0.029 31.84 2 × 10 –8 General WISC Symbol Search exm2270431 6 152644111 0.4565 SYNE1 C T 396 0.146 0.026 31.44 2 × 10 –8 General WISC Symbol Search psy_rs9478324 6 152677815 0.05058 SYNE1 G A 396 0.287 0.052 30.51 3 × 10 –8 Maternal SRS rs1451197 2 2741446 0.4993 G A 213 − 4.888 0.872 31.42 2 × 10 –8 Maternal SRS rs2176347 2 45968233 0.4473 PRKCE T G 231 − 3.441 0.624 30.39 4 × 10 –9 Maternal SRS rs7604835 2 152881908 0.1315 CACNB4 G A 287 − 11.482 1.927 35.51 3 × 10 –9 Maternal SRS psy_rs77672109 4 35576468 0.01662 C A 304 − 58.859 10.144 33.67 7 × 10 –9 Maternal SRS rs11934637 4 92981939 0.0289 T G 304 − 34.434 5.853 34.61 4 × 10 –8 Maternal SRS psy_rs146416593 5 60662093 0.01592 ZSWIM6 G A 324 − 33.216 5.844 32.31 1 × 10 –10 Maternal SRS psy_rs79166730 5 175229558 0.01951 CPLX2 G T 321 − 37.303 5.817 41.13 1 × 10 Nudel et al. Behavioral and Brain Functions (2022) 18:14 Page 12 of 20 ff Table 2 (continued) Test Trait Marker ID Chromosome Position MAF Gene Eect allele Other allele Probands Effect SE χ p-value (protein coding) –9 Maternal SRS exm573219 6 116325108 0.01879 FRK G A 325 − 58.949 9.981 34.88 4 × 10 –9 Maternal SRS psy_rs78989171 8 49645477 0.01301 EFCAB1 G T 308 − 58.867 10.204 33.28 8 × 10 –10 Maternal SRS exm754630 9 74319677 0.01806 TMEM2 T C 322 − 26.113 4.085 40.87 2 × 10 –9 Maternal SRS rs1681993 12 63338414 0.01084 G A 324 − 58.96 10.14 33.81 6 × 10 –9 Maternal SRS rs7956933 12 63345018 0.01085 A G 323 − 58.964 10.151 33.74 6 × 10 –9 Maternal SRS psy_rs10860381 12 99309750 0.01375 ANKS1B G A 305 − 58.851 10.206 33.25 8 × 10 –8 Maternal WISC Arithmetic rs11784069 8 2119582 0.3736 T G 271 0.341 0.061 31.1 2 × 10 –8 Maternal WISC Arithmetic rs7261002 20 1002656 0.1835 G A 314 0.661 0.116 32.68 1 × 10 –8 Maternal WISC Symbol Search rs2256135 6 152464839 0.4566 SYNE1 G A 262 0.264 0.047 31.95 2 × 10 –9 Paternal RIST index exm693219 8 28929739 0.02746 KIF13B G A 371 42.012 7.191 34.13 5 × 10 –9 Paternal SRS exm109001 1 156314440 0.01517 TSACC T G 320 − 42.257 7.231 34.15 5 × 10 –17 Paternal SRS psy_rs191695175 8 122841477 0.01016 T C 326 − 47.854 5.665 71.36 3 × 10 –9 Paternal SRS rs16908233 11 21604897 0.01158 A G 317 − 34.271 5.86 34.2 5 × 10 –8 Paternal SRS psy_rs117476444 13 113442872 0.02168 ATP11A G A 320 − 33.185 5.847 32.21 1 × 10 –8 Paternal WISC Arithmetic rs11784069 8 2119582 0.3736 T G 271 − 0.273 0.049 30.71 3 × 10 –8 Paternal WISC Arithmetic rs2740939 8 3872513 0.4899 CSMD1 C A 253 − 0.244 0.044 30.73 3 × 10 Results meeting criteria (i) and (ii) (see “Materials and methods”) are shown. Results meeting criteria (iii) and (iv) are shown in boldface. MAF: minor allele frequency (in founders; does not necessarily correspond to the effect allele frequency); SE, standard error Nudel et al. Behavioral and Brain Functions (2022) 18:14 Page 13 of 20 Fig. 3 Regional association plots for associations surviving all four statistical quality criteria (Methods). a rs6117457 (general test, WISC Arithmetic); b rs6117457 (general test, WISC Arithmetic); c rs7604835 (maternal POE test, SRS) Nudel et al. Behavioral and Brain Functions (2022) 18:14 Page 14 of 20 be associated with a lower score. It was also associated study. We therefore tested whether these markers were with the expression of PTPRA in the basal ganglia on themselves associated with the high risk status as the GTEx (P = 0.000022), with allele G being associated with outcome; none of the markers in Table 2 were associated lower expression. This marker remained at least nomi - with it after Bonferroni correction for multiple testing, nally significant when adding a covariate for high risk and the top three markers were not nominally associated status (i.e. for whether the child comes from a high risk even before correction. Thus, if, for these markers, the family or a control family) to the model (P = 0.0312). high risk status of the family is not associated with the Marker rs7604835, which showed a maternal POE in our genetic exposure, then this eliminates both the potential study, was associated with multiple DNA methylation confounding and potential collider bias from the model, and histone modification sites on PhenoScanner (mini - even if HRS is not included as a covariate. Even though –45 mum P = 1.19 × 10 ), based on evidence from two dif- the high risk status refers to the parent and not the child’s ferent studies [72, 73]. This provides further support for phenotype, this lack of association could suggest that the the association with a POE at this locus. This marker traits highlighted in Table 2 might not be useful endophe- remained genome-wide significant when adding a covari - notypes for schizophrenia or bipolar disorder, but might –9 ate for high risk status (P = 2 × 10 ). The last association nonetheless be associated with other disorders. which met all four criteria, namely, between rs6117457 Most of the associated markers in Table 2 (29 out of 48) and WISC Arithmetic in the general test, did not impli- fall within protein-coding genes. In total, 15 unique genes cate any protein-coding gene, and we could not find any are implicated by at least one genome-wide significant relevant prior association with it in the literature or func- association [meeting criteria (i) and, where relevant, (ii)] tional databases. This marker did not remain significant with a variant within them: ANKS1B, ATP11A, CACNB4, when adding a covariate for high risk status (P = 0.0765). CPLX2, CSMD1, EFCAB1, FRK, KIF13B, PRKCE, SIRPA, Translating the effects of the top markers into PVEs, we SYNE1, TGM3, TMEM2, TSACC and ZSWIM6. Given get: 0.078, 0.08 and 0.11 for rs6117457, rs214831 and that the associations in Table 2 were with the Arithme- rs7604835, respectively. The associations adjusted for tic and Symbol Search subtests of the WISC, the SRS and HRS were in the same direction as before in all cases. It the RIST, we used the following terms together with the should be noted, however, that the interpretation of the gene names when running the VarElect analysis: autism models with the covariate for HRS can be difficult: both OR "working memory" OR behavior OR communication WISC Arithmetic and SRS are associated with the child’s OR intelligence OR "processing speed" OR "Wechsler schizophrenia family status in the VIA 7 study [39, 40]; Intelligence Scale for Children" OR "Reynolds Intellec- since the covariate in this case may imply some genetic tual Screening Test" OR "Social Responsiveness Scale" predisposition to schizophrenia, a disorder which is OR schizophrenia OR "bipolar disorder". The last two genetically correlated with cognitive traits [74], the same terms were added because they represent the disorders SNP could have some association with both the psychiat- based on which the VIA sample had been ascertained. ric disorder and the phenotype of interest. Furthermore, Fourteen out of the fifteen genes were directly associated both schizophrenia and bipolar disorder are complex, with at least one of the terms (i.e., the gene’s GeneCard meaning they have both genetic and environmental risk contained the term), with the average number of associ- factors [75, 76]. Thus, the high risk status of the family, ated terms per gene being 4.43 (± 2.41). Two genes were determined by the presence of a psychiatric diagnosis in associated with 8 terms, the highest number of terms any one of the parents, is influenced both by genetic factors one gene was associated with: Ankyrin Repeat and Sterile and environmental factors; the parental genetic factors Alpha Motif Domain-containing Protein 1B (ANKS1B), influence both the child’s genetics (the exposure) and and Synaptic Nuclear Envelope Protein 1 (SYNE1). Addi- the high risk status of the family (the parent’s illness and tional file 4: Table S1 lists all direct associations between potential covariate), which could influence the outcome the terms and the genes and a discussion of the scores. in the child (the investigated trait). Similarly, environ- The gene with the highest VarElect score was CUB And mental factors, which may be unmeasured (or external Sushi Multiple Domains 1 (CSMD1), and the gene with factors in general e.g., parental IQ), could influence both the highest average disease causing likelihood was the the high risk status of the family and the investigated trait aforementioned CACNB4. in the child. In this scenario, adjusting for the covariate may reduce bias from possible confounding but intro- Discussion duce collider bias. A further complication would be the Our study investigated eight neurocognitive, motor and fact that most of our tests were for POEs, which limit the social-cognitive and social-behavioral functions using genetic causal path but not the causal path of the high a family-based GWAS design, including a general asso- risk status on which families were ascertained in this ciation test as well as tests of parent-of-origin effect tests. Nudel et al. Behavioral and Brain Functions (2022) 18:14 Page 15 of 20 We have identified 48 genome-wide significant associa - with the largest number of terms: CSMD1, ANKS1B and tions, of which 3 met our study-wide significance thresh - SYNE1. CSMD1 is of particular interest because it has old. Our results highlighted several protein-coding genes, been implicated in schizophrenia [85–87]. Interestingly, some of which have been implicated in prior genetic in our study, this gene was implicated through markers analyses of relevant phenotypes. associated with a measure of working memory; a study Two genes were highlighted through associations of this gene reported that a schizophrenia risk variant which met all four of our significance criteria: TGM3 in CSMD1 was associated with spatial working memory and CACNB4. The association with TGM3 was further [88]. This could illustrate the effect of a genetic variant supported by the marker’s being an eQTL for the gene. on an endophenotype for schizophrenia. In this con- This gene is involved in terminal epidermal differentia - text it is also important to note a proposal to redefine tion and has been implicated in some cancers [77, 78]. the notion of endophenotype in psychiatry to allow it to In our study, the marker in this gene was associated with include transdiagnostic traits that may be shared across a measure of working memory. Interestingly, previous several disorders [42]. ANKS1B was implicated through studies have found relevant associations between the the association between the marker psy_rs10860381 gene and related phenotypes: a study of the RNA blood and social responsiveness in the maternal POE test. This transcriptome of patients with Alzheimer’s disease (AD), gene encodes an activity dependent postsynaptic effec - a disease which involves severe memory impairments, tor protein highly expressed in the brain, and it has been found that the largest expression fold change among dif- implicated in a wide array of neurodevelopmental phe- ferentially expressed genes across AD cases and controls notypes [89]. Importantly, haploinsufficiency of this gene was with TGM3 [79]. Genes of the same family have been in a mouse model resulted in impaired social interaction implicated in several neurodegenerative diseases [80]. and sensorimotor dysfunction, which are core features of Also of note, the associated marker in our study was also autism spectrum disorder [90]. Even more importantly, a brain eQTL for PTPRA, a gene which is important for this gene exhibits allelic expression imbalance in the hippocampal neuronal migration; mice deficient for the brain, which could be an outcome of genomic imprinting PTPRA protein exhibit impairments in learning and (which could result in a POE), although this is only one short-term memory [81]. The association between social possible explanation [89]. SYNE1 was implicated through responsiveness (SRS) and CACNB4 was with a mater- the associations between several markers and process- nal POE. This marker was also associated with methyla - ing speed (WISC Symbol Search) in the general test. The tion and histone modifications sites, providing further gene encodes a protein that is involved in anchoring spe- support for a POE. The gene encodes a member of the cialized myonuclei underneath neuromuscular junctions, beta subunit family of voltage-dependent calcium chan- but it is also expressed in the brain—predominantly in nels, and it belongs to a family of genes which has been the cerebellum [91]. It has been implicated in a recessive implicated in several psychiatric and neurodevelopmen- form of cerebellar ataxia, which may also include cogni- tal disorders, including autism spectrum disorder, across tive deficits [91]. Interestingly, individuals with SYNE1 many studies [82]. The subunit encoded by CACNB4, mutations exhibit processing speed deficits compared specifically, is highly expressed in the brain and is prom - with controls [92], which is in line with our result show- inent in the cerebellum [83]. A recent study found that ing association between this gene and processing speed a pathogenic missense variant in this gene resulted in a in the general test. Both SYNE1 and TGM3 have been severe neurodevelopmental impairment which included highlighted in a study of de novo mutations in autism intellectual disability, language impairment, movement spectrum disorder [93]. impairment and seizures [84]. When adding a covariate Some of the other associations in Table 2 are also for high risk status to the statistical models for the top of note. The paternal POE association between social associations, we observe that it either slightly improved responsiveness and rs191695175 was the most sig- the association (with SRS, maternal POE test) or reduced nificant association in our study. The minor allele it drastically (with WISC Arithmetic, general test). frequency for this marker was very low at ~ 0.01 (in Whether or not it is appropriate to include this covari- founders), which could lead to a biased effect size. ate in the model depends on the causal paths between However, this marker is found on chromosome 8 in the genetic variant, the trait, and the covariate, which chromosomal band 8q24.13, a locus which was part are complex and not known. Hence, the interpretation of of a suggestive linkage peak for the same trait, namely, these post hoc tests should be done with caution. SRS, in a genome-wide linkage study [13]. The same Among the other genes in Table 2, three genes were locus also showed linkage to SRS in addition to an anxi- highlighted in the functional annotation either as hav- ety score and a score for pragmatic language skills, in ing the highest VarElect score or as being associated another study [94]. These studies, however, did not Nudel et al. Behavioral and Brain Functions (2022) 18:14 Page 16 of 20 model POEs. Thus, even though we may not be able to domains. This does not mean that traits from the other trust the estimated effect size for this locus, the asso - domains would not make good endophenotypes; our ciation itself might be valid and supported by previ- study did not examine that, and the lack of genetic asso- ous studies, and it is possible that the POE, if it indeed ciation could result from lower heritability for those operates at this locus, contributed to the stronger signal traits and/or insufficient sample sizes. in this study as compared with previous studies. There have been other previous studies which included similar Limitations of our study phenotypes, such as social interaction and social com- Our results should be evaluated in the light of several munication (neither was measured with the SRS), but potential limitations. Firstly, our study sample was a they did not model POEs, and their significant results family-based sample, and, as such, not a very large one. do not overlap with ours [95, 96]. We also observe an While this has the advantage of our being able to have interesting association trend with marker rs11784069: a deeply phenotyped sample, it can be detrimental to allele T, when inherited from the mother, is associated genetic association studies. While, as shown in previous with a higher WISC Arithmetic score (better work- simulations studies of QTDT models, our sample should ing memory functions), but, when inherited from the be large enough to detect some effects, it is to expected father, it is associated with a lower score (Table 2). This that only strong effects could be detected in our sample, is an illustration of the phenomenon mentioned in the which can explain why the majority of our genome-wide introduction, namely, opposite POEs of different paren - significant associations were intragenic. It should also be tal types at the same locus, which has been observed emphasized that some of the effect sizes could be over - for other quantitative traits in humans. This marker is estimated due to confounding. As it is difficult to deter - a highly significant eQTL for MYOM2 on eQTLGen mine the appropriateness of the adjustment for high risk –310 –20 (P = 3.2717 × 10 ) and GTEx (P = 9.5 × 10 ) in status, it should be borne in mind that the effects for whole blood. Interestingly, the mouse ortholog of this some associations might not be accurate. However, since gene, Myom2, was significantly upregulated and had the GWASs were performed with the goal of discovering the fifth largest fold change among upregulated genes new genetic associations for downstream analyses and in the hippocampus of memory-enhanced mice in one not for estimating their effects, we adopted this approach study [97], which is relevant for the association in our rather than potentially over-adjust the models, as dis- study, as the WISC Arithmetic score is a measure of cussed earlier. Another limitation is that we did not have working memory. a suitable replication sample which included the same phenotypes and genetic data from children and parents. Although our candidate genes have been highlighted in The top results in the context of endophenotypes previous studies of related traits, providing more cred- and the investigated domains ibility to their association with our traits, the associations The traits implicated by the top results in our study, with specific variants need to be replicated in an inde - namely, SRS (social responsiveness) and WISC Arith- pendent sample. metic (working memory), had been proposed as endo- phenotypes for ASD and ADHD, respectively [13, 98–100]. However, these studies did not identify links Future perspectives between specific genes and these endophenotypes at It has been shown that the heritability of cognitive abil- a genome-wide significant level; they focused on link - ity increases from childhood to young adulthood [102]. age analyses or candidate genes, and, where association Interestingly, a similar trend (reaching its peak around was modeled, it was only suggestive. Thus, our study age of 13 for girls and 14 for boys) was observed for provides genetic evidence for the missing piece in the height [103]. When the proportion of phenotypic vari- pathway from gene to disorder through endophenotype, ance explained by genetics increases, the proportion of namely: TGM3 → working memory → ADHD and the variance explained by the environment decreases, and CACNB4 → social responsiveness → ASD, through the vice versa. In the case of height, this trend could reflect top genetic associations we identified. Similarly, memory the effect of early childhood living conditions and/or pre - impairments, including verbal working memory impair- natal environmental factors [103]. For cognitive ability, ment, are common feature of schizophrenia [101], sug- the authors theorize that this trend could be a result of gesting further pathways between TGM3, CSMD1 and genotype-environment correlation, whereby their genet- PTPRA and schizophrenia through the working memory ics influences children increasingly in selecting, modify - endophenotype. The highlighted associations in Table 2 ing and creating their own experiences as they grow up belong to the neurocognitive and social-behavioral [102]. From the statistical genetic perspective, a higher Nudel et al. Behavioral and Brain Functions (2022) 18:14 Page 17 of 20 Acknowledgements heritability means that more genetic associations could The authors would like to express their gratitude to the dedicated families be identified if the sample of children were studied when participating in the study. We would also like to thank Heather Cordell for very they are older; this could mean that repeating the analy- helpful discussions on the model parameterizations of parent‑ of‑ origin effects, and Oskar Hougaard Jefsen for a valuable discussion on covariate adjustments ses within the VIA sample with these functions measured and potential biases in the models. in early adulthood could result in further associations. Furthermore, functional studies of the genes highlighted Author contributions RN conceived the study, performed the QC of the genetic data, performed in our study could provide further insight into the molec- the genetic and statistical analyses, analyzed the results, wrote the paper; ular etiologies of the neurodevelopmental disorders RZ wrote the program that tabulated the QTDT output; JO performed data whose endophenotypes were investigated in this study. management for VIA 7, assisted with the QC of the pedigree information, per‑ formed the standardization of test scores which lacked norms; CAJC, NH, DE, KSS, BKB, ANG, DLG contributed to the VIA 7 data collection and/or pilot study; CAJC, NH, BKB were responsible for the non‑ genetic study of the investigated Conclusions traits in the VIA 7 cohort and provided information about the trait scores Our study identified several candidate genes for social- relevant to this study; J‑BG oversaw sample preparation and genotyping and behavioral and neurocognitive functions, implicated performed initial QC on the raw genetic data; TW designed and oversaw the genetic part of the VIA 7 study; KJP, AAET, JRMJ, OM, MN contributed to the either through a general test, or a test of POEs; asso- conception of the VIA 7 study and its design, coordination and funding appli‑ ciations in the latter test were also supported by exter- cations. All authors have read and approved the manuscript. nal studies which had identified methylation or histone Funding modification sites associated with the relevant marker. The VIA 7 project is supported by the Mental Health Services of the Capital Importantly, most of our genome-wide significant asso - Region of Denmark (Region Hovedstadens Psykiatri), the Lundbeck Founda‑ ciations were within protein-coding genes, and many of tion Initiative for Integrative Psychiatric Research (iPSYCH), Aarhus University, the Tryg Foundation and the Beatrice Surovell Haskell Fund for Child Mental these had previously been implicated in studies of related Health Research of Copenhagen. The funding bodies had no influence on the traits and disorders, although many of these previous study design, analysis, or interpretation or on the writing of the manuscript. associations were with rare and/or deleterious muta- Availability of data and materials tions. Our study provides further evidence to the effect Access to the dataset used in the current study is available from the corre‑ that common variants may influence related traits in sponding author upon reasonable request. The program used to tabulate the individuals not diagnosed with severe mental disorders, QTDT output is available in a additional file accompanying this article. and it further supports a role for the highlighted genes in the studied traits, which can be seen as a replication of Declarations those genes’ implications in the previous studies. We did Ethics approval and consent to participate not identify significant associations for traits in some of The study was approved by the Danish Data Protection Agency and follows the other functions/domains included in this study; this all laws concerning the processing of personal data. Permission to draw data from registers was granted by the Danish Ministry of Health. The study could be the result of the lower heritability of those traits, protocol was sent to the Danish Committee on Health Research Ethics, who as well as potentially smaller effects that could not be dis - decided that ethical approval was not needed due to the observational covered in the VIA sample. Our results also illustrate the nature of the study. The genetic part of the study obtained ethical approval from the outset of the study and The Danish High Risk and Resilience Study usefulness of modeling POEs in human genetic studies, –VIA 7 was later incorporated into the protocol (Arv og Miljø—genetics and and, while previous studies focused on an array of quan- environment) as an appendix, which has then been approved by the ethics titative non-social-cognitive, non-social-behavioral, and committee [ARV OG MILJØ: betydning for psykisk sygdom hos børn og unge (H‑B‑2009‑026)]. Written informed consent was obtained from all adult partici‑ non-neurocognitive traits, our study highlights the pres- pants and from the legal guardians of participating children. ence of potential POEs in several of these traits studied in a systematic way, thus providing further evidence for this Consent for publication Not applicable. phenomenon in humans. Competing interests Supplementary Information The authors have no competing interests to declare, but TW states that he has acted as a lecturer and scientific counselor to H. Lundbeck A/S. DE has been The online version contains supplementary material available at https:// doi. employed by H. Lundbeck A/S from March until August 2020. org/ 10. 1186/ s12993‑ 022‑ 00198‑0. Author details Additional file 1. Archive file containing std_qtdt, which can parse and 1 CORE ‑ Copenhagen Research Centre for Mental Health, Mental Health tabulate the QTDT output. Centre Copenhagen, Copenhagen University Hospital, Copenhagen, Den‑ mark. iPSYCH, The Lundbeck Foundation Initiative for Integrative Psychiatric Additional file 2: Figure S1. Manhattan plots for all 24 GWASs. Research, Aarhus, Denmark. Institute of Biological Psychiatry, Mental Health Additional file 3: Figure S2. QQ plots for all 24 GWASs. Centre Sct. Hans, Mental Health Services Copenhagen, Roskilde, Denmark. Additional file 4: Table S1. Direct associations from VarElect for genes in Mental Health Centre for Child and Adolescent Psychiatry ‑ Research unit, Table 1. Mental Health Services in the Capital Region of Denmark, Copenhagen, Denmark. Department of Clinical Medicine, Faculty of Health and Medi‑ cal Sciences, University of Copenhagen, Copenhagen, Denmark. Psychosis Research Unit, Aarhus University Hospital ‑ Psychiatry, Aarhus, Denmark. Nudel et al. Behavioral and Brain Functions (2022) 18:14 Page 18 of 20 Center for Neonatal Screening, Department for Congenital Disorders, Statens 19. Constantino JN, Kennon‑McGill S, Weichselbaum C, Marrus N, Haider Serum Institut, Copenhagen, Denmark. Division of Child and Adolescent A, Glowinski AL, et al. Infant viewing of social scenes is under genetic Psychiatry, Department of Psychiatry, Hospital University Lausanne, Lausanne control and is atypical in autism. Nature. 2017;547(7663):340–4 (Epub University, Lausanne, Switzerland. Center for Neuropsychiatric Schizophrenia 2017/07/13). Research and Center for Clinical Intervention and Neuropsychiatric Schizo‑ 20. Moon AL, Haan N, Wilkinson LS, Thomas KL, Hall J. CACNA1C: associa‑ phrenia Research, Mental Health Services in the Capital Region of Denmark, tion with psychiatric disorders, behavior, and neurogenesis. Schizophr Copenhagen, Denmark. Bull. 2018;44(5):958–65 (Epub 2018/07/10). 21. Dedic N, Pohlmann ML, Richter JS, Mehta D, Czamara D, Metzger MW, Received: 24 February 2022 Accepted: 9 September 2022 et al. Cross‑ disorder risk gene CACNA1C differentially modulates sus‑ ceptibility to psychiatric disorders during development and adulthood. Mol Psychiatry. 2018;23(3):533–43 (Epub 2017/07/12). 22. Eichler EE, Flint J, Gibson G, Kong A, Leal SM, Moore JH, et al. Missing heritability and strategies for finding the underlying causes of complex disease. Nat Rev Genet. 2010;11(6):446–50 (Epub 2010/05/19). References 23. Manolio TA, Collins FS, Cox NJ, Goldstein DB, Hindorff LA, Hunter DJ, 1. Parenti I, Rabaneda LG, Schoen H, Novarino G. Neurodevelopmental et al. Finding the missing heritability of complex diseases. Nature. disorders: from genetics to functional pathways. Trends Neurosci. 2009;461(7265):747–53 (Epub 2009/10/09). 2020;43(8):608–21 (Epub 2020/06/09). 24. Kong A, Steinthorsdottir V, Masson G, Thorleifsson G, Sulem P, Besen‑ 2. Lenartowicz A, Loo SK. Use of EEG to diagnose ADHD. Curr Psychiatry bacher S, et al. Parental origin of sequence variants associated with Rep. 2014;16(11):498 (Epub 2014/09/23). complex diseases. Nature. 2009;462(7275):868–74 (Epub 2009/12/18). 3. Masi A, DeMayo MM, Glozier N, Guastella AJ. An overview of autism 25. Benonisdottir S, Oddsson A, Helgason A, Kristjansson RP, Sveinbjorns‑ spectrum disorder, heterogeneity and treatment options. Neurosci Bull. son G, Oskarsdottir A, et al. Epigenetic and genetic components of 2017;33(2):183–93 (Epub 2017/02/19). height regulation. Nat Commun. 2016;7:13490 (Epub 2016/11/17). 4. Bishop DVM. What causes specific language impairment in children? 26. Mozaffari SV, DeCara JM, Shah SJ, Sidore C, Fiorillo E, Cucca F, et al. Curr Dir Psychol Sci. 2006;15(5):217–21 (Epub 2008/11/15). Parent‑ of‑ origin effects on quantitative phenotypes in a large Hutterite 5. Gottesman II, Gould TD. The endophenotype concept in psychiatry: pedigree. Communications biology. 2019;2:28 (Epub 2019/01/25). etymology and strategic intentions. Am J Psychiatry. 2003;160(4):636– 27. Lawson HA, Cheverud JM, Wolf JB. Genomic imprinting and parent‑ 45 (Epub 2003/04/02). of‑ origin effects on complex traits. Nat Rev Genet. 2013;14(9):609–17 6. Almasy L, Blangero J. Endophenotypes as quantitative risk factors (Epub 2013/08/07). for psychiatric disease: rationale and study design. Am J Med Genet. 28. Weaver JR, Bartolomei MS. Chromatin regulators of genomic imprint‑ 2001;105(1):42–4 (Epub 2001/06/27). ing. Biochem Biophys Acta. 2014;1839(3):169–77 (Epub 2013/12/19). 7. Blokland GAM, Mesholam‑ Gately RI, Toulopoulou T, Del Re EC, Lam 29. Pearson CE. Slipping while sleeping? Trinucleotide repeat expansions in M, DeLisi LE, et al. Heritability of neuropsychological measures in germ cells. Trends Mol Med. 2003;9(11):490–5 (Epub 2003/11/08). schizophrenia and nonpsychiatric populations: a systematic review and 30. Cassidy SB, Dykens E, Williams CA. Prader‑ Willi and Angelman syn‑ meta‑analysis. Schizophr Bull. 2017;43(4):788–800 (Epub 2016/11/23). dromes: sister imprinted disorders. Am J Med Genet. 2000;97(2):136–46 8. Plomin R, von Stumm S. The new genetics of intelligence. Nat Rev (Epub 2001/02/17). Genet. 2018;19(3):148–59 (Epub 2018/01/18). 31. Buiting K. Prader‑ Willi syndrome and Angelman syndrome. Am J Med 9. Visscher PM, Hill WG, Wray NR. Heritability in the genomics era—con‑ Genet C Semin Med Genet. 2010;154C(3):365–76 (Epub 2010/08/31). cepts and misconceptions. Nat Rev Genet. 2008;9(4):255–66 (Epub 32. Chamberlain SJ, Lalande M. Neurodevelopmental disorders involving 2008/03/06). genomic imprinting at human chromosome 15q11–q13. Neurobiol Dis. 10. van Soelen IL, Brouwer RM, van Leeuwen M, Kahn RS, Hulshoff Pol HE, 2010;39(1):13–20 (Epub 2010/03/23). Boomsma DI. Heritability of verbal and performance intelligence in a 33. Nudel R, Simpson NH, Baird G, O’Hare A, Conti‑Ramsden G, Bolton PF, pediatric longitudinal sample. Twin Res Hum Genet. 2011;14(2):119–28 et al. Associations of HLA alleles with specific language impairment. J (Epub 2011/03/24). Neurodevelop Disord. 2014;6(1):1 (Epub 2014/01/18). 11. Ronald A, Viding E, Happe F, Plomin R. Individual differences in theory 34. Nudel R, Simpson NH, Baird G, O’Hare A, Conti‑Ramsden G, Bolton PF, of mind ability in middle childhood and links with verbal ability and et al. Genome‑ wide association analyses of child genotype effects and autistic traits: a twin study. Soc Neurosci. 2006;1(3–4):412–25 (Epub parent‑ of‑ origin effects in specific language impairment. Genes Brain 2008/07/18). Behav. 2014;13(4):418–29 (Epub 2014/02/28). 12. Constantino JN, Todd RD. Autistic traits in the general population: a 35. Pettigrew KA, Frinton E, Nudel R, Chan MTM, Thompson P, Hayiou‑ twin study. Arch Gen Psychiatry. 2003;60(5):524–30 (Epub 2003/05/14). Thomas ME, et al. Further evidence for a parent‑ of‑ origin effect at 13. Lowe JK, Werling DM, Constantino JN, Cantor RM, Geschwind DH. the NOP9 locus on language‑related phenotypes. J Neurodev Disord. Social responsiveness, an autism endophenotype: genomewide 2016;8:24 (Epub 2016/06/17). significant linkage to two regions on chromosome 8. Am J Psychiatry. 36. Connolly S, Anney R, Gallagher L, Heron EA. A genome‑ wide investiga‑ 2015;172(3):266–75 (Epub 2015/03/03). tion into parent‑ of‑ origin effects in autism spectrum disorder identifies 14. Pineda DA, Lopera F, Puerta IC, Trujillo‑ Orrego N, Aguirre‑Acevedo DC, previously associated genes including SHANK3. Eur J Hum Genet. Hincapie‑Henao L, et al. Potential cognitive endophenotypes in multi‑ 2017;25(2):234–9 (Epub 2016/11/24). generational families: segregating ADHD from a genetic isolate. Atten 37. Mott R, Yuan W, Kaisaki P, Gan X, Cleak J, Edwards A, et al. The archi‑ Deficit Hyp Disord. 2011;3(3):291–9 (Epub 2011/07/23). tecture of parent‑ of‑ origin effects in mice. Cell. 2014;156(1–2):332–42 15. Missitzi J, Gentner R, Misitzi A, Geladas N, Politis P, Klissouras V, (Epub 2014/01/21). et al. Heritability of motor control and motor learning. Physiol Rep. 38. Thorup AA, Jepsen JR, Ellersgaard DV, Burton BK, Christiani CJ, Hemager 2013;1(7):e00188 (Epub 2014/04/20). N, et al. The Danish High Risk and Resilience Study—VIA 7—a cohort 16. Williams LR, Gross JB. Heritability of motor skill. Acta Genet Med study of 520 7‑ year‑ old children born of parents diagnosed with either Gemellol. 1980;29(2):127–36 (Epub 1980/01/01). schizophrenia, bipolar disorder or neither of these two mental disor‑ 17. Burton BK, Hjorthoj C, Jepsen JR, Thorup A, Nordentoft M, Plessen KJ. ders. BMC Psychiatry. 2015;15:233 (Epub 2015/10/04). Research review: do motor deficits during development represent an 39. Hemager N, Plessen KJ, Thorup A, Christiani C, Ellersgaard D, Spang KS, endophenotype for schizophrenia? A meta‑analysis. J Child Psychol et al. Assessment of neurocognitive functions in 7‑ year‑ old children at Psychiatry. 2016;57(4):446–56 (Epub 2015/11/19). familial high risk for schizophrenia or bipolar disorder: the danish high 18. Iacono WG, Vaidyanathan U, Vrieze SI, Malone SM. Knowns and risk and resilience study VIA 7. JAMA Psychiat. 2018;75(8):844–52 (Epub unknowns for psychophysiological endophenotypes: integration and 2018/06/22). response to commentaries. Psychophysiology. 2014;51(12):1339–47 40. Christiani CJ, Jepsen JRM, Thorup A, Hemager N, Ellersgaard D, Spang (Epub 2014/11/13). KS, et al. Social cognition, language, and social behavior in 7‑ year‑ old Nudel et al. Behavioral and Brain Functions (2022) 18:14 Page 19 of 20 children at familial high‑risk of developing schizophrenia or bipolar through maternal effects and that may be subject to parental imprint ‑ disorder: the Danish high risk and resilience study VIA 7—a popula‑ ing. Am J Hum Genet. 1998;62(4):969–78 (Epub 1998/06/13). tion‑based cohort study. Schizophr Bull. 2019;45(6):1218–30 (Epub 63. Lange C, DeMeo DL, Laird NM. Power and design considerations for a 2019/03/11). general class of family‑based association tests: quantitative traits. Am J 41. Burton BK, Thorup AAE, Jepsen JR, Poulsen G, Ellersgaard D, Spang KS, Hum Genet. 2002;71(6):1330–41 (Epub 2002/11/28). et al. Impairments of motor function among children with a familial 64. Shim H, Chasman DI, Smith JD, Mora S, Ridker PM, Nickerson DA, risk of schizophrenia or bipolar disorder at 7 years old in Denmark: an et al. A multivariate genome‑ wide association analysis of 10 LDL sub‑ observational cohort study. Lancet Psychiatry. 2017;4(5):400–8 (Epub fractions, and their response to statin treatment, in 1868 Caucasians. 2017/03/28). PLoS ONE. 2015;10(4):e0120758 (Epub 2015/04/22). 42. Beauchaine TP, Constantino JN. Redefining the endophenotype con‑ 65. QTDT—online reference. https:// csg. sph. umich. edu/ abeca sis/ QTDT/ cept to accommodate transdiagnostic vulnerabilities and etiological docs/ index. html. Accessed 30 March 2020. complexity. Biomark Med. 2017;11(9):769–80 (Epub 2017/09/12). 66. Engle RF. Chapter 13: wald, likelihood ratio, and Lagrange multiplier 43. Howey R, Cordell HJ. PREMIM and EMIM: tools for estimation of mater‑ tests in econometrics. In: Leamer E, Heckman JJ, editors. Handbook nal, imprinting and interaction effects using multinomial modelling. of econometrics. Amsterdam: Elsevier; 1984. p. 775–826. BMC Bioinformatics. 2012;13:149 (Epub 2012/06/29). 67. Võsa U, Claringbould A, Westra H‑ J, Bonder MJ, Deelen P, Zeng B, et al. 44. Henderson S, Sugden D, Barnett A. The movement assessment battery Unraveling the polygenic architecture of complex traits using blood for children. 2nd ed. London: The Psychological Corporation; 2007. eQTL metaanalysis. Nat Genet. 2018. https:// doi. org/ 10. 1101/ 447367. 45. Cools W, Martelaer KD, Samaey C, Andries C. Movement skill assessment 68. GTEx Consortium. The genotype‑tissue expression (GTEx) project. Nat of typically developing preschool children: a review of seven move‑ Genet. 2013;45(6):580–5 (Epub 2013/05/30). ment skill assessment tools. J Sports Sci Med. 2009;8(2):154–68 (Epub 69. Kamat MA, Blackshaw JA, Young R, Surendran P, Burgess S, Danesh J, 2009/01/01). et al. PhenoScanner V2: an expanded tool for searching human geno‑ 46. Wechsler D. Wechsler intelligence scale for children. 4th ed. San Anto‑ type‑phenotype associations. Bioinformatics. 2019;35(22):4851–3 nio: The Psychological Corporation; 2003. (Epub 2019/06/25). 47. White S, Hill E, Happe F, Frith U. Revisiting the strange stories: revealing 70. Stelzer G, Plaschkes I, Oz‑Levi D, Alkelai A, Olender T, Zimmerman mentalizing impairments in autism. Child Dev. 2009;80(4):1097–117 S, et al. VarElect: the phenotype‑based variation prioritizer of the (Epub 2009/07/28). GeneCards Suite. BMC Genomics. 2016;17(Suppl 2):444 (Epub 48. Constantino JN, Davis SA, Todd RD, Schindler MK, Gross MM, Brophy 2016/07/01). SL, et al. Validation of a brief quantitative measure of autistic traits: 71. Stelzer G, Rosen N, Plaschkes I, Zimmerman S, Twik M, Fishile‑ comparison of the social responsiveness scale with the autism diag‑ vich S, et al. The GeneCards Suite: from gene data mining to nostic interview‑revised. J Autism Dev Disord. 2003;33(4):427–33 (Epub disease genome sequence analyses. Curr Protocols Bioinform. 2003/09/10). 2016;54(1):1.30.1‑1.3. 49. Reynolds C, Kamphaus R. Reynolds intellectual assessment scales 72. Chen L, Ge B, Casale FP, Vasquez L, Kwan T, Garrido‑Martin D, et al. (RIAS). Lutz: Psychological Assessment Resources; 2003. Genetic drivers of epigenetic and transcriptional variation in human 50. R Core Team. R: a language and environment for statistical computing. immune cells. Cell. 2016;167(5):1398‑414 e24 (Epub 2016/11/20). Vienna: R Foundation for Statistical Computing; 2014. 73. Bonder MJ, Luijk R, Zhernakova DV, Moed M, Deelen P, Vermaat M, et al. 51. Harrell FEJ, Dupont C. Hmisc: Harrell miscellaneous. 4.5‑0 ed2021. Disease variants alter transcription factor levels and methylation of their 52. Wei TS, Viliam. R package ’corrplot’: visualization of a correlation. 0.90 binding sites. Nat Genet. 2017;49(1):131–8 (Epub 2016/12/06). ed2021. 74. Hubbard L, Tansey KE, Rai D, Jones P, Ripke S, Chambert KD, et al. 53. Nudel R, Christiani CAJ, Ohland J, Uddin MJ, Hemager N, Ellersgaard Evidence of common genetic overlap between schizophrenia and D, et al. Quantitative genome‑ wide association analyses of receptive cognition. Schizophr Bull. 2015;42(3):832–42. language in the Danish high risk and resilience study. BMC Neurosci. 75. Sawa A, Snyder SH. Schizophrenia: diverse approaches to a complex 2020;21(1):30 (Epub 2020/07/09). disease. Science. 2002;296(5568):692–5 (Epub 2002/04/27). 54. Nudel R, Christiani CAJ, Ohland J, Uddin MJ, Hemager N, Ellersgaard 76. Kerner B. Toward a deeper understanding of the genetics of bipolar DV, et al. Language deficits in specific language impairment, attention disorder. Front Psych. 2015;6:105 (Epub 2015/08/19). deficit/hyperactivity disorder, and autism spectrum disorder: an analysis 77. Stacey SN, Sulem P, Gudbjartsson DF, Jonasdottir A, Thorleifsson G, of polygenic risk. Autism Res. 2020;13(3):369–81 (Epub 2019/10/03). Gudjonsson SA, et al. Germline sequence variants in TGM3 and RGS22 55. Purcell S, Neale B, Todd‑Brown K, Thomas L, Ferreira MA, Bender D, et al. confer risk of basal cell carcinoma. Hum Mol Genet. 2014;23(11):3045– PLINK: a tool set for whole‑ genome association and population‑based 53 (Epub 2014/01/10). linkage analyses. Am J Hum Genet. 2007;81(3):559–75. 78. Wu X, Cao W, Wang X, Zhang J, Lv Z, Qin X, et al. TGM3, a candidate 56. Anderson CA, Pettersson FH, Clarke GM, Cardon LR, Morris AP, Zonder‑ tumor suppressor gene, contributes to human head and neck cancer. van KT. Data quality control in genetic case‑ control association studies. Mol Cancer. 2013;12(1):151 (Epub 2013/12/03). Nat Protoc. 2010;5(9):1564–73 (Epub 2010/11/19). 79. Bai Z, Stamova B, Xu H, Ander BP, Wang J, Jickling GC, et al. Distinctive 57. Abecasis GR, Cardon LR, Cookson WO. A general test of association for RNA expression profiles in blood associated with Alzheimer disease quantitative traits in nuclear families. Am J Hum Genet. 2000;66(1):279– after accounting for white matter hyperintensities. Alzheimer Dis Assoc 92 (Epub 2000/01/13). Disord. 2014;28(3):226–33 (Epub 2014/04/16). 58. Abecasis GR, Cherny SS, Cookson WO, Cardon LR. Merlin—rapid analy‑ 80. Martin A, De Vivo G, Gentile V. Possible role of the transglutaminases in sis of dense genetic maps using sparse gene flow trees. Nat Genet. the pathogenesis of Alzheimer’s disease and other neurodegenerative 2002;30(1):97–101 (Epub 2001/12/04). diseases. Int J Alzheimer’s Dis. 2011;2011: 865432 (Epub 2011/02/26). 59. van der Sluis S, Posthuma D. Single‑locus association models. In: Fer ‑ 81. Petrone A, Battaglia F, Wang C, Dusa A, Su J, Zagzag D, et al. Receptor reira MAR, Medland SE, Posthuma D, Neale BM, editors. Statistical genet‑ protein tyrosine phosphatase α is essential for hippocampal neuronal ics: gene mapping through linkage and association. London: Taylor & migration and long‑term potentiation. EMBO J. 2003;22(16):4121–31. Francis; 2008. 82. Heyes S, Pratt WS, Rees E, Dahimene S, Ferron L, Owen MJ, et al. Genetic 60. Pruim RJ, Welch RP, Sanna S, Teslovich TM, Chines PS, Gliedt TP, et al. disruption of voltage‑ gated calcium channels in psychiatric and neuro‑ LocusZoom: regional visualization of genome‑ wide association scan logical disorders. Prog Neurobiol. 2015;134:36–54 (Epub 2015/09/20). results. Bioinformatics. 2010;26(18):2336–7 (Epub 2010/07/17). 83. Dolphin AC. β subunits of voltage‑ gated calcium channels. J Bioenerg 61. Ainsworth HF, Unwin J, Jamison DL, Cordell HJ. Investigation of mater‑ Biomembr. 2003;35(6):599–620. nal effects, maternal‑fetal interactions and parent ‑ of‑ origin effects 84. de Bagneaux PC, von Elsner L, Bierhals T, Campiglio M, Johannsen (imprinting), using mothers and their offspring. Genet Epidemiol. J, Obermair GJ, et al. A homozygous missense variant in CACNB4 2011;35(1):19–45 (Epub 2010/12/25). encoding the auxiliary calcium channel beta4 subunit causes a severe 62. Weinberg CR, Wilcox AJ, Lie RT. A log‑linear approach to case ‑parent ‑ neurodevelopmental disorder and impairs channel and non‑ channel triad data: assessing effects of disease genes that act either directly or functions. PLoS Genet. 2020;16(3):e1008625. Nudel et al. Behavioral and Brain Functions (2022) 18:14 Page 20 of 20 85. Schizophrenia Psychiatric Genome‑ Wide Association Study (GWAS) Publisher’s Note Consortium. Genome‑ wide association study identifies five new schizo ‑ Springer Nature remains neutral with regard to jurisdictional claims in pub‑ phrenia loci. Nat Genet. 2011;43(10):969–76 (Epub 2011/09/20). lished maps and institutional affiliations. 86. Havik B, Le Hellard S, Rietschel M, Lybaek H, Djurovic S, Mattheisen M, et al. The complement control‑related genes CSMD1 and CSMD2 associate to schizophrenia. Biol Psychiat. 2011;70(1):35–42 (Epub 2011/03/29). 87. Cross‑Disorder Group of the Psychiatric Genomics Consortium. Identifi‑ cation of risk loci with shared effects on five major psychiatric disorders: a genome‑ wide analysis. Lancet. 2013;381(9875):1371–9. 88. Koiliari E, Roussos P, Pasparakis E, Lencz T, Malhotra A, Siever LJ, et al. The CSMD1 genome‑ wide associated schizophrenia risk variant rs10503253 affects general cognitive ability and executive function in healthy males. Schizophr Res. 2014;154(1–3):42–7 (Epub 2014/03/19). 89. Younis RM, Taylor RM, Beardsley PM, McClay JL. The ANKS1B gene and its associated phenotypes: focus on CNS drug response. Pharmacog‑ enomics. 2019;20(9):669–84. 90. Carbonell AU, Cho CH, Tindi JO, Counts PA, Bates JC, Erdjument‑ Bromage H, et al. Haploinsufficiency in the ANKS1B gene encoding AIDA‑1 leads to a neurodevelopmental syndrome. Nat Commun. 2019;10(1):3529. 91. Gros‑Louis F, Dupre N, Dion P, Fox MA, Laurent S, Verreault S, et al. Muta‑ tions in SYNE1 lead to a newly discovered form of autosomal recessive cerebellar ataxia. Nat Genet. 2007;39(1):80–5 (Epub 2006/12/13). 92. Gama MTD, Braga‑Neto P, Dutra LA, Alessi H, Maria LA, Gadelha AA, et al. Cognitive and psychiatric evaluation in SYNE1 ataxia. Cerebellum. 2019;18(4):731–7 (Epub 2019/05/03). 93. O’Roak BJ, Deriziotis P, Lee C, Vives L, Schwartz JJ, Girirajan S, et al. Exome sequencing in sporadic autism spectrum disorders identi‑ fies severe de novo mutations. Nat Genet. 2011;43(6):585–9 (Epub 2011/05/17). 94. Piven J, Vieland VJ, Parlier M, Thompson A, O’Conner I, Woodbury‑Smith M, et al. A molecular genetic study of autism and related phenotypes in extended pedigrees. J Neurodevelop Disord. 2013;5(1):30 (Epub 2013/10/08). 95. Yousaf A, Waltes R, Haslinger D, Klauck SM, Duketis E, Sachse M, et al. Quantitative genome‑ wide association study of six phenotypic subdo‑ mains identifies novel genome ‑ wide significant variants in autism spec‑ trum disorder. Transl Psychiatry. 2020;10(1):215 (Epub 2020/07/07). 96. St Pourcain B, Whitehouse AJ, Ang WQ, Warrington NM, Glessner JT, Wang K, et al. Common variation contributes to the genetic architec‑ ture of social communication traits. Mol Autism. 2013;4(1):34 (Epub 2013/09/21). 97. Li C, Dong S, Wang H, Hu Y. Microarray analysis of gene expression changes in the brains of NR2B‑induced memory‑ enhanced mice. Neuroscience. 2011;197:121–31 (Epub 2011/09/20). 98. Loo SK, Rich EC, Ishii J, McGough J, McCracken J, Nelson S, et al. Cogni‑ tive functioning in affected sibling pairs with ADHD: familial clustering and dopamine genes. J Child Psychol Psychiatry. 2008;49(9):950–7 (Epub 2008/07/31). 99. Duvall JA, Lu A, Cantor RM, Todd RD, Constantino JN, Geschwind DH. A quantitative trait locus analysis of social responsiveness in multiplex autism families. Am J Psychiatry. 2007;164(4):656–62 (Epub 2007/04/04). 100. Coon H, Villalobos ME, Robison RJ, Camp NJ, Cannon DS, Allen‑Brady K, et al. Genome‑ wide linkage using the social responsiveness scale in Utah autism pedigrees. Mol Autism. 2010;1(1):8 (Epub 2010/08/04). Re Read ady y to to submit y submit your our re researc search h ? Choose BMC and benefit fr ? Choose BMC and benefit from om: : 101. Conklin HM. Verbal working memory impairment in schizophrenia patients and their first ‑ degree relatives: evidence from the digit span fast, convenient online submission task. American J Psychiatry. 2000;157(2):275–277. https:// doi. org/ 10. thorough peer review by experienced researchers in your ﬁeld 1176/ appi. ajp. 157.2. 275. 102. Haworth CM, Wright MJ, Luciano M, Martin NG, de Geus EJ, van rapid publication on acceptance Beijsterveldt CE, et al. The heritability of general cognitive ability support for research data, including large and complex data types increases linearly from childhood to young adulthood. Mol Psychiatry. • gold Open Access which fosters wider collaboration and increased citations 2010;15(11):1112–20 (Epub 2009/06/03). 103. Jelenkovic A, Sund R, Hur YM, Yokoyama Y, Hjelmborg JV, Moller S, et al. maximum visibility for your research: over 100M website views per year Genetic and environmental influences on height from infancy to early adulthood: an individual‑based pooled analysis of 45 twin cohorts. Sci At BMC, research is always in progress. Rep. 2016;6:28496 (Epub 2016/06/24). Learn more biomedcentral.com/submissions

Journal

Behavioral and Brain Functions – Springer Journals

Published: Dec 1, 2022

Keywords: GWAS; Neurodevelopment; Cognitive functions; Endophenotype; Parent-of-origin effect

References

Neurodevelopmental disorders: from genetics to functional pathways

Parenti, I; Rabaneda, LG; Schoen, H; Novarino, G
Use of EEG to diagnose ADHD

Lenartowicz, A; Loo, SK
An overview of autism spectrum disorder, heterogeneity and treatment options

Masi, A; DeMayo, MM; Glozier, N; Guastella, AJ
What causes specific language impairment in children?

Bishop, DVM
The endophenotype concept in psychiatry: etymology and strategic intentions

Gottesman, II; Gould, TD
Endophenotypes as quantitative risk factors for psychiatric disease: rationale and study design

Almasy, L; Blangero, J
Heritability of neuropsychological measures in schizophrenia and nonpsychiatric populations: a systematic review and meta-analysis

Blokland, GAM; Mesholam-Gately, RI; Toulopoulou, T; Del Re, EC; Lam, M; DeLisi, LE
The new genetics of intelligence

Plomin, R; von Stumm, S
Heritability in the genomics era—concepts and misconceptions

Visscher, PM; Hill, WG; Wray, NR
Heritability of verbal and performance intelligence in a pediatric longitudinal sample

van Soelen, IL; Brouwer, RM; van Leeuwen, M; Kahn, RS; Hulshoff Pol, HE; Boomsma, DI
Individual differences in theory of mind ability in middle childhood and links with verbal ability and autistic traits: a twin study

Ronald, A; Viding, E; Happe, F; Plomin, R
Autistic traits in the general population: a twin study

Constantino, JN; Todd, RD
Social responsiveness, an autism endophenotype: genomewide significant linkage to two regions on chromosome 8

Lowe, JK; Werling, DM; Constantino, JN; Cantor, RM; Geschwind, DH
Potential cognitive endophenotypes in multigenerational families: segregating ADHD from a genetic isolate

Pineda, DA; Lopera, F; Puerta, IC; Trujillo-Orrego, N; Aguirre-Acevedo, DC; Hincapie-Henao, L
Heritability of motor control and motor learning

Missitzi, J; Gentner, R; Misitzi, A; Geladas, N; Politis, P; Klissouras, V
Heritability of motor skill

Williams, LR; Gross, JB
Research review: do motor deficits during development represent an endophenotype for schizophrenia? A meta-analysis

Burton, BK; Hjorthoj, C; Jepsen, JR; Thorup, A; Nordentoft, M; Plessen, KJ
Knowns and unknowns for psychophysiological endophenotypes: integration and response to commentaries

Iacono, WG; Vaidyanathan, U; Vrieze, SI; Malone, SM
Infant viewing of social scenes is under genetic control and is atypical in autism

Constantino, JN; Kennon-McGill, S; Weichselbaum, C; Marrus, N; Haider, A; Glowinski, AL
CACNA1C: association with psychiatric disorders, behavior, and neurogenesis

Moon, AL; Haan, N; Wilkinson, LS; Thomas, KL; Hall, J
Cross-disorder risk gene CACNA1C differentially modulates susceptibility to psychiatric disorders during development and adulthood

Dedic, N; Pohlmann, ML; Richter, JS; Mehta, D; Czamara, D; Metzger, MW
Missing heritability and strategies for finding the underlying causes of complex disease

Eichler, EE; Flint, J; Gibson, G; Kong, A; Leal, SM; Moore, JH
Finding the missing heritability of complex diseases

Manolio, TA; Collins, FS; Cox, NJ; Goldstein, DB; Hindorff, LA; Hunter, DJ
Parental origin of sequence variants associated with complex diseases

Kong, A; Steinthorsdottir, V; Masson, G; Thorleifsson, G; Sulem, P; Besenbacher, S
Epigenetic and genetic components of height regulation

Benonisdottir, S; Oddsson, A; Helgason, A; Kristjansson, RP; Sveinbjornsson, G; Oskarsdottir, A
Parent-of-origin effects on quantitative phenotypes in a large Hutterite pedigree

Mozaffari, SV; DeCara, JM; Shah, SJ; Sidore, C; Fiorillo, E; Cucca, F
Genomic imprinting and parent-of-origin effects on complex traits

Lawson, HA; Cheverud, JM; Wolf, JB
Chromatin regulators of genomic imprinting

Weaver, JR; Bartolomei, MS
Slipping while sleeping? Trinucleotide repeat expansions in germ cells

Pearson, CE
Prader-Willi and Angelman syndromes: sister imprinted disorders

Cassidy, SB; Dykens, E; Williams, CA
Prader-Willi syndrome and Angelman syndrome

Buiting, K
Neurodevelopmental disorders involving genomic imprinting at human chromosome 15q11–q13

Chamberlain, SJ; Lalande, M
Associations of HLA alleles with specific language impairment

Nudel, R; Simpson, NH; Baird, G; O'Hare, A; Conti-Ramsden, G; Bolton, PF
Genome-wide association analyses of child genotype effects and parent-of-origin effects in specific language impairment

Nudel, R; Simpson, NH; Baird, G; O'Hare, A; Conti-Ramsden, G; Bolton, PF
Further evidence for a parent-of-origin effect at the NOP9 locus on language-related phenotypes

Pettigrew, KA; Frinton, E; Nudel, R; Chan, MTM; Thompson, P; Hayiou-Thomas, ME
A genome-wide investigation into parent-of-origin effects in autism spectrum disorder identifies previously associated genes including SHANK3

Connolly, S; Anney, R; Gallagher, L; Heron, EA
The architecture of parent-of-origin effects in mice

Mott, R; Yuan, W; Kaisaki, P; Gan, X; Cleak, J; Edwards, A
The Danish High Risk and Resilience Study—VIA 7—a cohort study of 520 7-year-old children born of parents diagnosed with either schizophrenia, bipolar disorder or neither of these two mental disorders

Thorup, AA; Jepsen, JR; Ellersgaard, DV; Burton, BK; Christiani, CJ; Hemager, N
Assessment of neurocognitive functions in 7-year-old children at familial high risk for schizophrenia or bipolar disorder: the danish high risk and resilience study VIA 7

Hemager, N; Plessen, KJ; Thorup, A; Christiani, C; Ellersgaard, D; Spang, KS
Social cognition, language, and social behavior in 7-year-old children at familial high-risk of developing schizophrenia or bipolar disorder: the Danish high risk and resilience study VIA 7—a population-based cohort study

Christiani, CJ; Jepsen, JRM; Thorup, A; Hemager, N; Ellersgaard, D; Spang, KS
Impairments of motor function among children with a familial risk of schizophrenia or bipolar disorder at 7 years old in Denmark: an observational cohort study

Burton, BK; Thorup, AAE; Jepsen, JR; Poulsen, G; Ellersgaard, D; Spang, KS
Redefining the endophenotype concept to accommodate transdiagnostic vulnerabilities and etiological complexity

Beauchaine, TP; Constantino, JN
PREMIM and EMIM: tools for estimation of maternal, imprinting and interaction effects using multinomial modelling

Howey, R; Cordell, HJ
Movement skill assessment of typically developing preschool children: a review of seven movement skill assessment tools

Cools, W; Martelaer, KD; Samaey, C; Andries, C
Revisiting the strange stories: revealing mentalizing impairments in autism

White, S; Hill, E; Happe, F; Frith, U
Validation of a brief quantitative measure of autistic traits: comparison of the social responsiveness scale with the autism diagnostic interview-revised

Constantino, JN; Davis, SA; Todd, RD; Schindler, MK; Gross, MM; Brophy, SL
Quantitative genome-wide association analyses of receptive language in the Danish high risk and resilience study

Nudel, R; Christiani, CAJ; Ohland, J; Uddin, MJ; Hemager, N; Ellersgaard, D
Language deficits in specific language impairment, attention deficit/hyperactivity disorder, and autism spectrum disorder: an analysis of polygenic risk

Nudel, R; Christiani, CAJ; Ohland, J; Uddin, MJ; Hemager, N; Ellersgaard, DV
PLINK: a tool set for whole-genome association and population-based linkage analyses

Purcell, S; Neale, B; Todd-Brown, K; Thomas, L; Ferreira, MA; Bender, D
Data quality control in genetic case-control association studies

Anderson, CA; Pettersson, FH; Clarke, GM; Cardon, LR; Morris, AP; Zondervan, KT
A general test of association for quantitative traits in nuclear families

Abecasis, GR; Cardon, LR; Cookson, WO
Merlin—rapid analysis of dense genetic maps using sparse gene flow trees

Abecasis, GR; Cherny, SS; Cookson, WO; Cardon, LR
Single-locus association models

van der Sluis, S; Posthuma, D; Ferreira, MAR; Medland, SE; Posthuma, D; Neale, BM
LocusZoom: regional visualization of genome-wide association scan results

Pruim, RJ; Welch, RP; Sanna, S; Teslovich, TM; Chines, PS; Gliedt, TP
Investigation of maternal effects, maternal-fetal interactions and parent-of-origin effects (imprinting), using mothers and their offspring

Ainsworth, HF; Unwin, J; Jamison, DL; Cordell, HJ
A log-linear approach to case-parent-triad data: assessing effects of disease genes that act either directly or through maternal effects and that may be subject to parental imprinting

Weinberg, CR; Wilcox, AJ; Lie, RT
Power and design considerations for a general class of family-based association tests: quantitative traits

Lange, C; DeMeo, DL; Laird, NM
A multivariate genome-wide association analysis of 10 LDL subfractions, and their response to statin treatment, in 1868 Caucasians

Shim, H; Chasman, DI; Smith, JD; Mora, S; Ridker, PM; Nickerson, DA
Chapter 13: wald, likelihood ratio, and Lagrange multiplier tests in econometrics

Engle, RF; Leamer, E; Heckman, JJ
Unraveling the polygenic architecture of complex traits using blood eQTL metaanalysis

Võsa, U; Claringbould, A; Westra, H-J; Bonder, MJ; Deelen, P; Zeng, B
The genotype-tissue expression (GTEx) project
PhenoScanner V2: an expanded tool for searching human genotype-phenotype associations

Kamat, MA; Blackshaw, JA; Young, R; Surendran, P; Burgess, S; Danesh, J
VarElect: the phenotype-based variation prioritizer of the GeneCards Suite

Stelzer, G; Plaschkes, I; Oz-Levi, D; Alkelai, A; Olender, T; Zimmerman, S
The GeneCards Suite: from gene data mining to disease genome sequence analyses

Stelzer, G; Rosen, N; Plaschkes, I; Zimmerman, S; Twik, M; Fishilevich, S
Genetic drivers of epigenetic and transcriptional variation in human immune cells

Chen, L; Ge, B; Casale, FP; Vasquez, L; Kwan, T; Garrido-Martin, D
Disease variants alter transcription factor levels and methylation of their binding sites

Bonder, MJ; Luijk, R; Zhernakova, DV; Moed, M; Deelen, P; Vermaat, M
Evidence of common genetic overlap between schizophrenia and cognition

Hubbard, L; Tansey, KE; Rai, D; Jones, P; Ripke, S; Chambert, KD
Schizophrenia: diverse approaches to a complex disease

Sawa, A; Snyder, SH
Toward a deeper understanding of the genetics of bipolar disorder

Kerner, B
Germline sequence variants in TGM3 and RGS22 confer risk of basal cell carcinoma

Stacey, SN; Sulem, P; Gudbjartsson, DF; Jonasdottir, A; Thorleifsson, G; Gudjonsson, SA
TGM3, a candidate tumor suppressor gene, contributes to human head and neck cancer

Wu, X; Cao, W; Wang, X; Zhang, J; Lv, Z; Qin, X
Distinctive RNA expression profiles in blood associated with Alzheimer disease after accounting for white matter hyperintensities

Bai, Z; Stamova, B; Xu, H; Ander, BP; Wang, J; Jickling, GC
Possible role of the transglutaminases in the pathogenesis of Alzheimer's disease and other neurodegenerative diseases

Martin, A; De Vivo, G; Gentile, V
Receptor protein tyrosine phosphatase α is essential for hippocampal neuronal migration and long-term potentiation

Petrone, A; Battaglia, F; Wang, C; Dusa, A; Su, J; Zagzag, D
Genetic disruption of voltage-gated calcium channels in psychiatric and neurological disorders

Heyes, S; Pratt, WS; Rees, E; Dahimene, S; Ferron, L; Owen, MJ
β subunits of voltage-gated calcium channels

Dolphin, AC
A homozygous missense variant in CACNB4 encoding the auxiliary calcium channel beta4 subunit causes a severe neurodevelopmental disorder and impairs channel and non-channel functions

de Bagneaux, PC; von Elsner, L; Bierhals, T; Campiglio, M; Johannsen, J; Obermair, GJ
Genome-wide association study identifies five new schizophrenia loci
The complement control-related genes CSMD1 and CSMD2 associate to schizophrenia

Havik, B; Le Hellard, S; Rietschel, M; Lybaek, H; Djurovic, S; Mattheisen, M
Identification of risk loci with shared effects on five major psychiatric disorders: a genome-wide analysis
The CSMD1 genome-wide associated schizophrenia risk variant rs10503253 affects general cognitive ability and executive function in healthy males

Koiliari, E; Roussos, P; Pasparakis, E; Lencz, T; Malhotra, A; Siever, LJ
The ANKS1B gene and its associated phenotypes: focus on CNS drug response

Younis, RM; Taylor, RM; Beardsley, PM; McClay, JL
Haploinsufficiency in the ANKS1B gene encoding AIDA-1 leads to a neurodevelopmental syndrome

Carbonell, AU; Cho, CH; Tindi, JO; Counts, PA; Bates, JC; Erdjument-Bromage, H
Mutations in SYNE1 lead to a newly discovered form of autosomal recessive cerebellar ataxia

Gros-Louis, F; Dupre, N; Dion, P; Fox, MA; Laurent, S; Verreault, S
Cognitive and psychiatric evaluation in SYNE1 ataxia

Gama, MTD; Braga-Neto, P; Dutra, LA; Alessi, H; Maria, LA; Gadelha, AA
Exome sequencing in sporadic autism spectrum disorders identifies severe de novo mutations

O'Roak, BJ; Deriziotis, P; Lee, C; Vives, L; Schwartz, JJ; Girirajan, S
A molecular genetic study of autism and related phenotypes in extended pedigrees

Piven, J; Vieland, VJ; Parlier, M; Thompson, A; O'Conner, I; Woodbury-Smith, M
Quantitative genome-wide association study of six phenotypic subdomains identifies novel genome-wide significant variants in autism spectrum disorder

Yousaf, A; Waltes, R; Haslinger, D; Klauck, SM; Duketis, E; Sachse, M
Common variation contributes to the genetic architecture of social communication traits

St Pourcain, B; Whitehouse, AJ; Ang, WQ; Warrington, NM; Glessner, JT; Wang, K
Microarray analysis of gene expression changes in the brains of NR2B-induced memory-enhanced mice

Li, C; Dong, S; Wang, H; Hu, Y
Cognitive functioning in affected sibling pairs with ADHD: familial clustering and dopamine genes

Loo, SK; Rich, EC; Ishii, J; McGough, J; McCracken, J; Nelson, S
A quantitative trait locus analysis of social responsiveness in multiplex autism families

Duvall, JA; Lu, A; Cantor, RM; Todd, RD; Constantino, JN; Geschwind, DH
Genome-wide linkage using the social responsiveness scale in Utah autism pedigrees

Coon, H; Villalobos, ME; Robison, RJ; Camp, NJ; Cannon, DS; Allen-Brady, K
Verbal working memory impairment in schizophrenia patients and their first-degree relatives: evidence from the digit span task

Conklin, HM
The heritability of general cognitive ability increases linearly from childhood to young adulthood

Haworth, CM; Wright, MJ; Luciano, M; Martin, NG; de Geus, EJ; van Beijsterveldt, CE
Genetic and environmental influences on height from infancy to early adulthood: an individual-based pooled analysis of 45 twin cohorts

Jelenkovic, A; Sund, R; Hur, YM; Yokoyama, Y; Hjelmborg, JV; Moller, S

Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

A family-based study of genetic and epigenetic effects across multiple neurocognitive, motor, social-cognitive and social-behavioral functions

A family-based study of genetic and epigenetic effects across multiple neurocognitive, motor, social-cognitive and social-behavioral functions

Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

A family-based study of genetic and epigenetic effects across multiple neurocognitive, motor, social-cognitive and social-behavioral functions

A family-based study of genetic and epigenetic effects across multiple neurocognitive, motor, social-cognitive and social-behavioral functions

Abstract

Journal

Recommended Articles

References

Our policy towards the use of cookies