Whole genome sequencing study of identical twins discordant for psychosis

The study investigates whether rare post-zygotic genetic variants contribute to diagnostic discordance for psychotic disorders within monozygotic (MZ) twin pairs. Schizophrenia and related psychotic disorders are highly heritable, yet much heritability remains unexplained by common-variant GWAS. Rare variants detectable by sequencing may account for part of this gap. Given that MZ twins share nearly all inherited genetic background and early-life environments, discordant diagnoses suggest roles for non-shared factors, including post-zygotic mutations. The authors hypothesize that affected twins may harbor rare, deleterious variants not present in their co-twins that increase disease risk. The purpose is to leverage whole-genome sequencing (WGS) in discordant MZ pairs to identify discordant single nucleotide variants (SNVs), indels, and copy number variants (CNVs), including regulatory variants and repeat expansions, that could underlie psychosis risk.

Prior work shows substantial heritability for schizophrenia (60–80%) and genetic overlap with schizoaffective disorder and bipolar disorder. GWAS have implicated many common loci but explain limited heritability; sequencing studies (WES/WGS) have identified rare coding variants of larger effect in psychotic disorders and shared rare-variant architecture between schizophrenia and bipolar disorder. Twin studies report higher heritability than GWAS-based estimates and ~50% concordance for schizophrenia in MZ twins, indicating roles for rare variants and non-shared environment. Recent studies demonstrate that a fraction of de novo variants arise post-zygotically, with some occurring after twinning. WES/WGS analyses in discordant MZ twins for various neuropsychiatric and neurodevelopmental disorders have identified post-zygotic variants, motivating similar analyses in psychotic disorders.

Design and cohort: From the Swedish STAR twin cohort, the authors selected all available monozygotic (MZ) twin pairs discordant for schizophrenia, schizoaffective disorder, or bipolar disorder; co-twins with major depressive disorder without psychosis were considered unaffected. Initial N=19 pairs were sequenced; after QC, 17 pairs remained. Diagnoses were established via SCID-I and evaluation of hospital records; ethical approvals and informed consent were obtained. DNA and sequencing: Peripheral blood DNA was extracted using a salting-out method (Puregene). WGS was performed at Edinburgh Genomics (HiSeqX) with ~30x mean coverage. FastQC and samtools were used for initial QC. Reads were aligned to GRCh38 using BWA-MEM; GATK Best Practices were followed (mark duplicates, BQSR, local realignment, HaplotypeCaller v3.8), with joint genotyping and VQSR for SNVs and indels. Sample QC: Peddy was used to check relatedness, sex, coverage, and ancestry (1000 Genomes PCA). One pair (T04) was excluded for relatedness error; another pair (T19) failed DNA QC pre-sequencing. Genotype concordance with SNP arrays (where available) and within-pair zygosity checks confirmed identities and monozygosity. SNV/indel processing: Variants with QUAL < 100 were removed; per-sample genotypes with GQ < 20 or DP < 10 were set to missing. Discordant variants were defined pairwise where one twin had exactly one additional copy of the alternative allele relative to the co-twin. VEP annotated consequences (sequence ontology), SIFT, PolyPhen-2, CADD, and population frequencies (1000 Genomes, gnomAD). Prioritization targeted rare, coding, putatively pathogenic SNVs meeting: (i) RefSeq protein-coding, (ii) VEP impact MODERATE or HIGH, (iii) SIFT deleterious or PolyPhen damaging, (iv) allele frequency <1% or absent in relevant populations in gnomAD/1000G, and (v) not observed in any other twin pair in the cohort. Multiallelic sites were split to biallelic. For regulatory variants, additional filters required CADD Phred >20 and RegulomeDB score 1 or 2; brain/ENCODE regulatory annotations were assessed; Wilcoxon signed-rank tests compared counts in affected vs unaffected co-twins across eight regulatory features. CNV calling: Germline CNVs were called by a family-based consensus from four algorithms (details in Supplementary Methods), merging calls within pairs to regions of interest (ROIs). ROIs detected by only one caller in a twin were removed. Common CNVs (≥1% frequency) overlapping ≥50% reciprocally with entries in gnomAD-SV, DDD, or DGV were filtered out (hg19 liftover where needed). ROIs with ≥50% reciprocal overlap with ClinGen pathogenic CNVs were retained regardless of frequency (acknowledging platform/reference differences). Somatic CNVs were assessed using MoChA (depth deviations from phased SNVs) with manual curation. Repeat expansions: ExpansionHunter was used to genotype 16 known repeat expansion loci; thresholds followed disorder-specific criteria. Statistical/ancillary analyses: Genome-wide discordant variant counts were summarized; ancestry confirmed by PCA; regulatory feature overlap counts compared within pairs using Wilcoxon signed-rank tests.

- Cohort and QC: 17 MZ twin pairs (post-QC) discordant for psychosis (schizophrenia, schizoaffective disorder, or bipolar disorder); all European ancestry except one East Asian pair. Zygosity and sample identities confirmed. - Discordant variant burden: After filtering, each sample had on average 44,306 discordant SNVs/indels (SD ~1,640; range 41,134–48,050), about ~1% of total detected variants per genome, exceeding expected sequencing error (~0.1%). - Protein-coding SNVs: Six rare, deleterious missense SNVs were identified across four genes, each private to an affected twin: NUTM2G (T13_A1; SCZ), FOXN1 (three proximal variants likely due to local realignment; T07_A; BD), FLOT2 (T09_A; BD), and KRTAP10-6 (T10_A; SCZ). FOXN1 and FLOT2 variants would be categorized as damaging in recent schizophrenia/bipolar exome analyses. None of the exact SNVs were in SCHEMA; FOXN1 S291R observed in BipEx in one case but in a missense class not enriched overall. - Regulatory variants: Applying stringent regulatory filters (CADD >20; RegulomeDB 1–2), one discordant regulatory variant was found but in an unaffected twin. No significant within-pair differences in counts of discordant variants overlapping eight regulatory features (Wilcoxon tests, all non-significant). - CNVs: Four rare discordant genic CNVs were identified in affected twins only: a 3q29 duplication (T17_A; BD) overlapping genes including DLG1 and BDH1; small deletions at chr10:92,847,856–92,849,207 (overlapping EXOC6; T02_A; SAD), chr12:120,201,498–120,204,299 (T16_A; BD), and chr19:11,261,852–11,262,999 (T01_A1; SCZ). Screening against 23 known schizophrenia-implicated CNVs found a chr13q12.11 duplication in both members of pair T09 (previously nominally protective). Extended ClinGen screen identified 14 clinically impactful CNVs across samples, but only the 3q29 duplication was exclusive to affected individuals. - Somatic CNVs and repeat expansions: No convincing discordant somatic CNVs after visual curation. No repeat expansions above pathogenic thresholds in any twin; some within-pair discordances were below thresholds. - Notable observation: In T09, the affected twin carried the rare FLOT2 missense variant alongside the protective 13q12.11 duplication present in both twins.

Findings support the hypothesis that rare post-zygotic variants can arise after twinning and may contribute to psychosis risk in an affected twin. The identification of six rare, deleterious missense SNVs private to affected individuals across four genes (including FOXN1 and FLOT2 categorized as damaging in large exome analyses) and four rare discordant genic CNVs in affected twins (notably a 3q29 duplication, a region implicated in neurodevelopmental disorders) underscores the potential etiologic role of such variants. The lack of significant differences in regulatory-annotation overlaps and absence of pathogenic repeat expansions suggest that, within the constraints of blood-derived 30x WGS, coding variation and selected CNVs provided the most compelling discordant signals. Some findings, such as the protective 13q12.11 duplication observed in both members of a pair and a deleterious FLOT2 variant in the affected co-twin, illustrate complex genetic architectures where multiple variants with opposing effects may co-occur. Overall, the results advance evidence that post-zygotic genetic variation is a plausible component of discordant psychotic phenotypes in MZ twins, complementing shared inherited risk and non-shared environmental influences.

This study presents the largest WGS analysis to date of MZ twin pairs discordant for psychotic disorders, identifying rare, deleterious coding SNVs and rare genic CNVs present only in affected co-twins, including a notable 3q29 duplication. The work provides candidate variants and regions for further study and demonstrates a framework integrating SNVs, CNVs, regulatory annotations, and repeat expansions in a co-twin design. Future research should include larger cohorts for statistical burden testing, parental sequencing to confirm de novo status and shared de novo events, higher-depth sequencing or targeted deep sequencing to better detect somatic mosaicism, analyses of brain tissue where feasible to capture tissue-specific mosaicism, integration of epigenomic data (e.g., methylation), and family-based follow-up (including offspring) to assess transmission and segregation of candidate variants.

- Sample size limits statistical power to assess burden of rare discordant coding variation. - Lack of parental genomes prevents confirmation of de novo status and evaluation of shared de novo variants with variable penetrance. - WGS depth (~30x) may be insufficient to reliably detect low-level somatic mosaicism; manual review also did not support somatic CNVs. - Blood-derived DNA may miss brain-specific post-zygotic variants. - Diagnostic status of unaffected co-twins could change over time; some were assessed after typical age of onset but longitudinal changes remain possible.