Exome sequencing in families with severe mental illness identifies novel and rare variants in genes implicated in Mendelian neuropsychiatric syndromes.

AIM: Severe mental illnesses (SMI), such as bipolar disorder and schizophrenia, are highly heritable, and have a complex pattern of inheritance. Genome-wide association studies detect a part of the heritability, which can be attributed to common genetic variation. Examination of rare variants with next-generation sequencing may add to the understanding of the genetic architecture of SMI.
METHODS: We analyzed 32 ill subjects from eight multiplex families and 33 healthy individuals using whole-exome sequencing. Prioritized variants were selected by a three-step filtering process, which included: deleteriousness by five in silico algorithms; sharing within families by affected individuals; rarity in South Asian sample estimated using the Exome Aggregation Consortium data; and complete absence of these variants in control individuals from the same gene pool.
RESULTS: We identified 42 rare, non-synonymous deleterious variants (~5 per pedigree) in this study. None of the variants were shared across families, indicating a 'private' mutational profile. Twenty (47.6%) of the variant harboring genes were previously reported to contribute to the risk of diverse neuropsychiatric syndromes, nine (21.4%) of which were of Mendelian inheritance. These included genes carrying novel deleterious variants, such as the GRM1 gene implicated in spinocerebellar ataxia 44 and the NIPBL gene implicated in Cornelia de Lange syndrome.
CONCLUSION: Next-generation sequencing approaches in family-based studies are useful to identify novel and rare variants in genes for complex disorders like SMI. The findings of the study suggest a potential phenotypic burden of rare variants in Mendelian disease genes, indicating pleiotropic effects in the etiology of SMI.

Bipolar disorder (BD) and schizophrenia (SCZ) are severe mental illness (SMI) syndromes with a median lifetime prevalence of 2.4 and 3.3 per thousand persons, respectively,1, 2 and an estimated heritability of 70–90%.3, 4 Evidence from family and molecular genetic studies suggests shared, perhaps overlapping, risk factors across these syndromes.5, 6 The outcomes from large‐scale genome‐wide association studies (GWAS) exploring the common disease–common variant (CDCV) hypothesis detect a proportion of the estimated genetic risk.7 In this context, next‐generation sequencing (NGS) technology, by evaluating rare genetic variants, has enabled a deeper examination of complex traits using alternate models of risk, such as the ‘oligogenic quasi‐Mendelian model’8 and the ‘omnigenic models’9 of inheritance. Several recent studies in autism, SCZ, BD, and depression have detected rare variants using NGS in case–control or family‐based designs, across different genes implicated to play a key role in critical biological pathways.10, 11 Findings from such studies have shown that the majority of the rare variants identified are private to a family (Table S1),12, 13, 14 indicating the underlying heterogeneity in the genetic architecture of SMI. Multiplex families may provide valuable insights into the genetic correlates of these syndromes15, 16 when tested using high throughput sequencing.

A cross‐nosology approach has been quite informative in identifying potential disease‐relevant pathways in SCZ and BD.17, 18 Single‐nucleotide polymorphisms (SNP) associated with these two syndromes show a high mutual correlation, among combinations of neuropsychiatric syndromes.7 Such overlaps have also been observed across diverse neuropsychiatric syndromes, for both common and rare genetic variations, as well as in gene expression profiles in the cerebral cortex.19, 20 These findings indicate an underlying shared molecular pathology in the pathobiology of SMI.

As part of a longitudinal study, ‘Accelerator Program for Discovery of Brain Disorders Using Stem Cells’ (ADBS),21 aimed at understanding the developmental trajectories and basic biology of SMI, we describe in this study the results of a variant discovery analysis using whole‐exome sequencing (WES) in eight multiplex pedigrees with SCZ and BD phenotypes from a well characterized Indian cohort. Such studies have been predominantly conducted in large cohorts of European origin,16 and representation from other populations is perhaps necessary to validate earlier findings, and identify population‐specific signatures underlying SMI. In the current study, we aimed to identify rare, damaging, exonic variants that co‐segregate with SMI in multiplex families, and to examine their relevance to the disease.

Methods

Sample selection

The families were recruited as part of the ADBS longitudinal study, which has been approved by the ethics committee of the National Institute of Mental Health and Neurosciences, Bengaluru, India. The details of screening, informed consent, recruitment, and phenotyping have been previously published.21 Some of the families in the cohort have been on follow up for longer than 10 years. We have previously noted evidence of linkage in psychosis at chromosome 18p11.2,22 and the sex‐specific association to the DISC1 gene using a case–control study design23 in samples taken from this cohort. For the current study, eight families (A through H) with high loading of SMI (SCZ, BD, and psychosis in the context of these eight pedigrees; Fig. 1a, Fig. S1) were assessed in detail. From these, 32 individuals (‘cases’; 16 females) with SMI were available for blood sampling and were subjected to WES. Two senior psychiatrists evaluated all patients and unaffected relatives independently. Diagnoses were made with the ICD‐10 Classification for Mental and Behavioural Disorders and were verified in the longitudinal course of follow ups. From five of the eight families, we could also sample eight unaffected individuals who had crossed the age at risk and are defined as a ‘family‐specific control’ for the respective pedigree henceforth in this report. An independent set of 25 individuals without a history of SMI were further sampled as population‐matched controls. Together this group constituted a total of 33 asymptomatic ‘controls.’

Figure 1

(a) Two representative pedigrees analyzed with exome sequencing (Families A and B). (b) Cluster dendrogram created with a distance matrix based on the degree of variant sharing between pairs of cases and controls analyzed in the study. (c) ‘varPrio’ – variant prioritization pipeline with numbers indicating the reduction in the total number of variants in each prioritization step. (d) Ideogram representing the 42 genes that harbored variants prioritized by non‐synonymous damaging strict (NSD‐S) and disruptive definition generated with NCBI genome decoration page.

Exome sequencing and analysis

Sequencing was carried out on the Illumina Hiseq NGS platform with libraries prepared using Illumina exome kits. Reads were aligned with reference human genome GRCh37 using the Burrows–Wheeler algorithm tool.24 Variants were called from realigned BAM files using Varscan2 with the standard criteria (min coverage = 8, MAF ≥ 0.25, and P ≤ 0.001).25 Standard quality control protocols were employed at sequencing, alignment, and variant calling (Fig.S2). The resulting variant called files were annotated with ANNOVAR.26

Pedigree‐based analysis

All variant segregation analysis was performed at the level of individual pedigrees. To ascertain the degree of variance (between pedigrees) and relatedness within family structures, we generated a dendrogram with hierarchical clustering analysis using an allele‐sharing matrix of the exonic variants (Appendix S1).

Variant prioritization

Variants were prioritized if:

the variant was found to be shared by all affected individuals within the pedigree while allowing for one missing genotype, a method shown to be useful in an earlier study of familial BD12;

the variant fell into any of the following deleterious categories – the non‐synonymous damaging strict (NSD‐S) set predicted to be damaging by five prediction algorithms (SIFT,27 Polyphen‐2 HDIV,28 Mutation taster 2,29 Mutation assessor,30 and LRT31[Appendix S1]), the Disruptive set predicted to result in protein truncation (splice site, stop gain, or stop loss variants), or the non‐synonymous damaging broad (NSD‐B) set predicted to be damaging by one or more of the five prediction algorithms; and

the variant was rare <1% in Exome Aggregation Consortium – South Asian sample (ExAC‐SAS)32 and completely absent from a control cohort of 33 individuals from the same gene pool (http://indexdb.ncbs.res.in).

The above variant prioritization was carried out using an in‐house automated pipeline ‘varPrio’. Details of the pipeline and the resulting variant enrichment are summarized in Figure1c. To rule out any false positive calls at the final variant list, a representative set of prioritized variants (n=10) was independently confirmed by Sanger sequencing and we noted a 100% concordance.

Functional annotation

We adapted two approaches for evaluating functional impact to the prioritized variants:

We reviewed the literature on individual genes identified in the NSD‐S and the disruptive set carrying rare variants of highest priority (all five in silico predictors) for prior evidence of disease association in neuropsychiatric phenotype.

For the NSD‐B set carrying rare variants of plausible disease relevance (1–5 in silico predictors), we tested for enrichment of the aggregate list using DAVID functional annotation tool 6.8.33, 34 To test the enrichment on the categories of biological process, molecular function, protein domain, protein–protein interaction, and tissue expression we selected the sources as –‘GOTERM_BP_DIRECT’, ‘GOTERM_MF_DIRECT’, ‘INTERPRO’, ‘KEGG_PATHWAY’ and‘UP_TISSUE’ in this in silico approach. Modified Fisher's exact test with Benjamini–Hochberg correction built‐in to this algorithm was used to infer enrichment.

Results

Sample characteristics

Of the 32 cases sequenced in the study, 26 were diagnosed with BD, four with SCZ, and one each with SCZ‐like psychosis and schizoaffective disorder. They had been ill for a mean (SD) duration of 23.7 (11.1) years, and the mean (SD) age at onset was 23.1 (7.9) years. In most of the pedigrees, there was heterogeneity in the age of onset, illness severity, global outcomes, and segregation of suicidality and psychosis (in BD) with the primary phenotype. Substance use disorder was a common comorbidity, followed by hypothyroidism, seizure disorder, and dementia (Appendix S2, Table S2).

In the analysis of relatedness using the cluster dendrogram, ‘cases’ and ‘controls’ formed a single cluster possibly resulting from sharing of a large number of common and/or benign exonic variants. As expected, members from each pedigree clustered together due to the relatively larger magnitude of variant sharing (Fig. 1b, Appendix S1).

Rare deleterious variants in Mendelian genes segregate within SMI families

Familywise prioritization identified a total of 39 NSD‐S, three disruptive, and 248 NSD‐B variants. The NSD‐S and disruptive sets of variants (Table 1, Fig. 1d) spanning 42 genes were private to individual pedigrees (~5 variants per pedigree). Twelve of these were novel (not reported in dbSNP or other published databases) and the remaining were noted in very low frequencies (<1e −07 to 7.8e‐03) in ExAC‐SAS. None of the variants prioritized were present in 33 healthy Indian control samples (http://indexdb.ncbs.res.in). Nine (21.4%) of the 42 variants were found in genes that have been reported in Mendelian syndromes with early onset neurodevelopmental features, such as infantile epilepsy, intellectual disability, and structural brain abnormalities. Seven (16.67%) of these gene‐phenotype relationships were reported in the Online Mendelian Inheritance in Man (OMIM)35 and the remaining two were noted in the MedGen (NCBI) and ClinVar36 databases. This was significantly higher in comparison to a background list of 1310 out of 15 857 (8.26%) such genes (Appendix S1) listed in the OMIM database (P = 0.039, odds ratio [OR] = 2.423, confidence interval [CI] = 1.07–5.513, Fisher's exact test) while not accounting for potential gene length bias. Some of these variants were observed in close proximity to reported ‘pathogenic’ mutation of the relevant Mendelian syndrome and/or in highly conserved regions (Table 2(a)). Two of these nine variants, one each on the GRM1 gene (chr6:146351218, GRCh37) and NIPBL gene (chr5:37010263, GRCh37), were novel. Pathogenic mutations on the GRM1 gene, coding for metabotropic glutamate receptor 1 (mGluR1) result in autosomal dominant (type 44) (OMIM:617691) and recessive (type 13) (OMIM:614831) forms of spinocerebellar ataxia, both of which are characterized by early age of onset and associated intellectual disability. Missense variants have been identified spanning the entire exome of this gene in individuals and families with SCZ and other neuropsychiatric syndromes.37 Mutations in the NIPBL gene, coding for Cohesin Loading Factor involved cortical neuronal migration,38 cause Cornelia de Lange syndrome 1. The novel missense variant identified in the pedigree G (chr5:37010263), segregating with BD would result in substitution of polar amino acid glutamine by a hydrophobic amino acid proline. A non‐sense mutation at the same codon (rs797045760) is reported to be pathogenic of Cornelia de Lange syndrome 1 (ClinVarSCV000248215.1).

Table 1

List of novel or rare variants prioritized by non‐synonymous damaging strict and disruptive definition

Gene symbol	rsID/novel	chr:location	Transcript	Exon	Variant	Amino acid change	ExAC_SAS
LRRC8B	NOVEL	chr1:90049348	NM_015350	Exon5	c.A1139C	p.Y380S
GRM1	NOVEL	chr6:146351218	NM_001278064	Exon1	c.A565G	p.S189G
SETD6	NOVEL	chr16:58552094	NM_001160305	Exon6	c.C932G	p.A311G
SYF2	NOVEL	chr1:25555567	NM_015484	Exon3	c.A180C	p.K60N
RAB3IL1	NOVEL	chr11:61675047	NM_001271686	Exon3	c.G491A	p.S164N
BCDIN3D	NOVEL	chr12:50236792	NM_181708	Exon1	c.G79A	p.G27S
NDRG3	NOVEL	chr20:35317139	NM_022477	Exon3	c.G106T	p.G36C
PARP14	NOVEL	chr3:122423522	NM_017554	Exon8	c.G3467A	p.S1156N
NIPBL	NOVEL	chr5:37010263	NM_015384	Exon21	c.A4496C	p.Q1499P
SCUBE3	NOVEL	chr6:35211460	NM_001303136	Exon16	c.C1996T	p.L666F
NBPF11	NOVEL	chr1:147599423	Splicing
CKMT2	NOVEL	chr5:80550306	Splicing
KRT85	rs112554450	chr12:52758810	NM_002283	Exon2	c.G565A	p.D189N	6.000E‐03
NRG2	rs148371256	chr5:139231286	NM_001184935	Exon7	c.C1477T	p.R493W	5.000E‐04
MDN1	rs148868949	chr6:90397121	NM_014611	Exon68	c.C11392T	p.R3798W	3.000E‐03
MYO1A	rs151269703	chr12:57431355	NM_005379	Exon19	c.A2032T	p.I678F	5.500E‐03
EFHC1	rs1570624	chr6:52319050	NM_018100	Exon5	c.G881A	p.R294H	5.100E‐03
CNGB1	rs192843629	chr16:57950041	NM_001286130	Exon22	c.C2191T	p.R731C	7.000E‐04
PCCB	rs371155999	chr3:136002730	NM_000532	Exon6	c.C595T	p.P199S	7.100E‐03
TRMT44	rs373816157	chr4:8467199	NM_152544	Exon8	c.C1405T	p.R469W	0.000E+00
CLUAP1	rs531380218	chr16:3558347	NM_015041	Exon4	c.C278T	p.A93V	9.000E‐04
GOLM1	rs534059912	chr9:88661389	NM_016548	Exon5	c.G463A	p.D155N	3.700E‐03
LGALS12	rs534811017	chr11:63277314	NM_001142537	Exon3	c.C320T	p.T107M	2.200E‐03
ADPRH	rs547308034	chr3:119301144	NM_001291949	Exon2	c.T128C	p.L43S	7.800E‐03
FAM208B	rs548531206	chr10:5789582	NM_017782	Exon15	c.T4198C	p.S1400P	4.000E‐03
PM20D1	rs553380022	chr1:205809408	NM_152491	Exon10	c.G1088A	p.R363Q	2.700E‐03
CD1D	rs569233577	chr1:158152752	NM_001766	Exon5	c.C692G	p.P231R	1.400E‐03
PLXND1	rs569306898	chr3:129279222	NM_015103	Exon31	c.G5084C	p.R1695P	6.124E‐05
WDFY4	rs571808731	chr10:50030541	NM_020945	Exon35	c.C5941A	p.P1981T	3.100E‐03
CC2D2A	rs574421639	chr4:15559035	NM_001080522	Exon22	c.A2734G	p.R912G	1.300E‐03
ANLN	rs575071809	chr7:36435984	NM_001284301	Exon2	c.C128T	p.P43L	1.800E‐03
PARVB	rs575240566	chr22:44528830	NM_001243385	Exon6	c.C463A	p.H155N	1.000E‐04
DOCK5	rs61732769	chr8:25174610	NM_024940	Exon14	c.C1406T	p.T469M	4.300E‐03
KIF7	rs749711306	chr15:90176400	NM_198525	Exon13	c.G2690C	p.G897A	6.478E‐05
C20orf194	rs750188084	chr20:3251118	NM_001009984	Exon30	c.A2741G	p.N914S	6.063E‐05
TCEA3	rs753347636	chr1:23720470	NM_003196	Exon8	c.C721T	p.R241C	6.058E‐05
ARHGEF40	rs756016433	chr14:21553914	NM_001278529	Exon19	c.C1885T	p.R629W	0.000E+00
PLB1	rs760022335	chr2:28814039	Splicing				0.0004
SCN3A	rs775711350	chr2:166032822	NM_001081676	Exon3	c.G83A	p.R28H	0.000E+00
INPP5A	rs775793924	chr10:134521844	NM_005539	Exon7	c.C502T	p.R168W	6.083E‐05
DENND5A	rs779817963	chr11:9171664	NM_001243254	Exon15	c.A2699G	p.H900R	6.132E‐05
COL4A5	rs78972735	chrX:107865996	NM_000495	Exon33	c.G2858T	p.G953V	6.900E‐03

Chr:location (chromosomal location); ExAC_SAS (variant allele frequency in ExAC south Asian sample).

Table 2

Disease relevance of the genes harboring prioritized variants

(a) Genes implicated in a Mendelian syndrome
Gene symbol	Name	Mendelian disease	Selected gene functions
GRM1 †	Glutamate metabotropic receptor 1	Spinocerebellar ataxia AR 13 (MIM:617691) and SCA 44 (MIM:614831)	GO:0007216~G‐protein coupled glutamate receptor signaling pathway; GO:0007268~chemical synaptic transmission
EFHC1	EF‐hand domain containing 1	Myoclonic epilepsy, juvenile, susceptibility to, 1 (MIM:254770)	GO:0021795~cerebral cortex cell migration
DENND5A	DENN domain containing 5A	Epileptic encephalopathy, early infantile, 49 (MIM:617281)	GO:0043547~positive regulation of GTPase activity; GO:0070588~calcium ion transmembrane transport
KIF7	Kinesin family member 7	Acrocallosal syndrome, Joubert syndrome 12 (MIM:200990)	GO:0007018~microtubule‐based movement; GO:0045879~negative regulation of smoothened signaling pathway
SCN3A	Sodium voltage‐gated channel alpha subunit 3	Cryptogeneicpaediatric partial epilepsy (Medgen CN240377)	GO:0019228~neuronal action potential; GO:0060078~regulation of postsynaptic membrane potential
PCCB	Propionyl‐CoA carboxylase beta subunit	Propionicacidemia (MIM:606054)	GO:0006633~fatty acid biosynthetic process
NIPBL †	NIPBL, cohesin loading factor	Cornelia de Lange syndrome 1(MIM:122470)	GO:0007420~brain development; GO:0045995~regulation of embryonic development
CLUAP1	Clusterin‐associated protein 1	Oculoectodermal syndrome, Joubert syndrome (ClinVar)	GO:0001843~neural tube closure; GO:0021508~floor plate formation
CC2D2A	Coiled‐coil and C2 domain containing 2A	COACH syndrome (MIM:216360), Joubert syndrome 9 (MIM:612285), Meckel syndrome 6 (MIM:612284)	GO:1990403~embryonic brain development; GO:0001843~neural tube closure

Variant in close proximity to a pathogenic mutation for a Mendelian syndrome.

CNV, copy number variation; EWAS, epigenome wide association study; GO, gene ontology; GWAS, genome‐wide association studies; MIM, Mendelian Inheritance in Man.

Ten other genes that harbored prioritized variants have been implicated in neuropsychiatric syndromes. We identified a variant (rs148371256) in the NRG2 gene (Neuregulin 2) that was earlier reported to be associated with gamma band oscillations in SCZ with suggestive genome‐wide significance.39 The encoded protein neuregulin‐2 has been shown to be critical for the formation and maturation of GABAergic synapses40 and its ablation results in dopamine dysregulation.41 Another novel variant (chr3:122423522, GRCh37) was identified in the PARP14 gene (Poly ADP ribose polymerase 14), and the gene has been implicated in post‐traumatic stress disorder (PTSD), major depressive disorder (MDD), and attention deficit hyperactivity disorder (ADHD).42 We also noted a variant (rs534059912) in the GOLM1 gene (Golgi membrane protein 1), which was earlier reported in sporadic Alzheimer's dementia (AD) to influence the pre‐frontal cortical volume.43 A list of these 10 genes, evidence for disease association, and gene ontology descriptions are presented in Table 2(b).44, 45, 46, 47, 48, 49, 50, 51, 52

Of the remaining genes, there were several with a plausible role in the biology of SMI, but not thus far implicated in any disease phenotype. These genes, with the ontology descriptions and plausible biological implications, are provided in Table 2(c).53, 54, 55, 56, 57, 58

Enrichment of coding variants with plausible functional role in SMI

The NSD‐B set consisted of 248 variants; of these, except for rs570064523 in the PCSK1 gene, which was identified in cases from two families (G and H), no other overlap at the level of family was noted for the remaining 247 variants (Appendix S2, Table S3). In the ‘protein domains’ category tested using the Interpro database as the source, the term ‘epidermal growth factor like domain’ showed a nominally significant enrichment with P = 0.0013, Benjamini–Hochberg false discovery rate corrected P = 0.073. Twelve genes that were enriched for this domain included the NRG2 and SCUBE3 genes, which were also categorized in the NSD‐S set along with the NOTCH1, JAG1 and WIF1 genes, which form critical nodes in the notch signaling pathway implicated in neurodevelopment and embryogenesis (Appendix S2, Table S4).59 There was no statistically significant enrichment in any of the remaining categories tested with this in silico approach.

Discussion

The results of our study highlight the usefulness of WES in multiplex families with SMI to identify rare and novel variants that may contribute to the susceptibility to common polygenic syndromes. Many of these variants prioritized by NSD‐S, and presumed to be disruptive, map to genes that have been previously reported in GWAS, candidate gene association, post‐mortem expression, or animal model studies of SMI. In addition, consistent with the WES approach, we identify variants in genes hitherto not reported in the context of an SMI, but that could potentially contribute to disease biology.

The segregation of rare and deleterious variants in Mendelian disease genes with a neuropsychiatric phenotype is in keeping with some recent observations. Studies have shown that heterozygous carriers of Mendelian disease mutations are at increased risk for specific common diseases.60 While Mendelian forms of common, complex traits, such as Alzheimer's disease, hypertension, hypercholesterolemia, and hypertriglyceridemia, have long been attributed to rare causal variants in single genes, population‐based GWAS in these traits have often implicated genes that also cause single gene disorders.60 More recently, using electronic health record data, the disease‐relevant phenotypic burden of rare variants in Mendelian genes, thus far not characterized as ‘pathogenic,’ has been demonstrated across diverse phenotypes.61

We explored the clinical significance of nine variants in Mendelian genes in the ClinVar database, a publicly available archive of human phenotype‐variation relations.36 None of these variants was annotated as ‘pathogenic’ or ‘likely pathogenic’ in the database for the corresponding Mendelian phenotype. As a corollary, none of the families had any identified or suspected case of a severe neurodevelopmental syndrome. However, the predicted deleteriousness by in silico algorithms, a very low prevalence in the population, physical proximity to known pathogenic mutations, and the reported physiological gene function suggest a plausible role for these variants in the etiology of SMI. The impact of these variants in cellular and/or animal models needs to be examined to validate these observations and to establish their causal role in SMI. Interestingly, an earlier WES study in families with BD also reported variants in genes of monogenic syndromes: holoprosencephaly and progressive myoclonic epilepsy.13

We detected rare variants in 10 additional genes that have been noted in earlier studies to contribute to the risk for polygenic syndromes, such as SCZ, BD, autism, MDD, ADHD, PTSD, AD, and Parkinson's disease. This finding is congruent with the evolving concept of shared molecular neuropathology across SMI.19 These, along with other identified genes known to be involved in neurodevelopmental processes (e.g., PLXND1) or known to have manifold higher brain expression (e.g., ANLN, LRRC8B) are potential targets to be examined in future studies of SMI. Lastly, of the 12 genes encoding highly conserved epidermal growth factor‐like domains and showing nominally significant enrichment to this domain, many encode for proteins that play critical roles during embryogenesis and neurodevelopment.59

Certain limitations are to be considered while interpreting the results of this study. The relatively small control set sequenced in our study precluded statistical association testing at the level of a variant or a gene. It has been estimated that rare variant association testing at gene level using case–control samples would require sample sizes greater than 20 000 individuals.62 As an alternative, we considered the minor allele frequency of the variant in ExAC South Asian samples in the prioritization approach, and many of the identified variants were noted to be extremely rare. Second, although we sampled a nearly equal number of affected persons from each family, the relationships within pedigrees were not uniform, potentially adding heterogeneity to the number of identified variants. Thus, we prioritized variants with complete sharing allowing for one missing genotype. This resulted in identification of some variants that were not fully penetrant. Third, like the previous studies of WES in SCZ and BD, we have relied on in silico predictions to infer the deleteriousness of a variant and have considered those predicted by five algorithms as the primary variants of interest. Supporting this approach, a recent analysis noted that the strength of disease association for a non‐synonymous variant increased with the greater number of deleterious predictions in silico. 63 Fourth, inherent to the prioritization criteria of rarity, deleteriousness and segregation, the NSD‐S and disruptive variant set presented above would explain only a part of an individual's liability to disease. The results of this analysis represent the shared familial risk for SMI, private to each pedigree, determined by variants of possible major/moderate effect. Lastly, we have not been able to sample all affected individuals from each multiplex pedigree. Among the unaffected individuals, we have been able to sample one to two representative individuals from five of the pedigrees. Thus, the prioritized variants might represent only a part of the shared genetic risk within each pedigree.

Using WES data in multiplex families with SMI, we find evidence that suggests intersections in the molecular pathways leading to the expression of polygenic SMI and Mendelian neuropsychiatric syndromes. The patient‐derived neural stem cell lines being developed as part of the program21 will be useful to explore the functional significance of the identified variants accounting for ‘modifier genetic background’,64 and to characterize mechanisms that underlie the observed genotype–phenotype correlates.

Conclusions

NGS approaches in a family‐based study design are useful to identify novel and rare variants in genes potentially relevant to complex disorders, such as SMI. The study further provides an independent validation for the phenotypic burden of rare deleterious variants in Mendelian disease genes that segregate privately in multiplex pedigrees with SCZ and BD. Our findings support the role of heterogeneity and pleiotropy in the genetic architecture of SMI encompassing a spectrum of neurodevelopmental and degenerative phenotypes.

Disclosure statement

The authors declare that they have no conflicts of interest.

Author contributions

S.G. analyzed the data and wrote the manuscript; H.A.P. built the VarPrio algorithm and performed variant prioritization and secondary analysis; R.K.N. and M.S. performed the detailed clinical assessments of the study participants under the supervision of B.V.; R.P.M. performed whole‐exome sequence data mining; O.M. supervised sequencing data generation, analysis of the results, and manuscript preparation. S.J. and M.S. provided vital inputs to data analysis and manuscript preparation. The study was conceived by the ADBS Consortium. All authors took part in editing the manuscript and approved the final version.

Appendix S1. Supplemental material for Mendelian disease genes in familial SMI

Click here for additional data file.

Appendix S2. Microsoft Excel file containing supplementary tables

Click here for additional data file.