Schizophrenia, autism spectrum disorders and developmental disorders share specific disruptive coding mutations

Schizophrenia is a severe, highly heritable, polygenic psychiatric disorder. Prior work has established contributions from common variants, rare copy number variants (CNVs), and ultra-rare coding variants with elevated de novo mutation rates. Notably, CNVs conferring schizophrenia risk are largely pleiotropic with neurodevelopmental disorders (NDDs), and sequencing implicates overlap between schizophrenia and NDD genes, including SETD1A. Yet, two key questions remained underpowered previously: within shared genes, do the same functional classes of mutation (e.g., protein-truncating variants [PTVs] vs missense) confer risk across schizophrenia and NDDs, and do the same specific rare coding alleles occur in both disorders (allelic pleiotropy)? This study tests whether genic pleiotropy is functionally congruent across disorders and whether specific NDD de novo variants are enriched in schizophrenia, thereby clarifying the extent to which equivalent gene function changes underlie both schizophrenia and NDDs.

Prior literature shows: (1) common variants explain roughly one-third of schizophrenia’s genetic liability, (2) multiple rare CNVs increase risk for schizophrenia and also for NDDs, indicating pleiotropy at the CNV level, (3) ultra-rare coding variants contribute to schizophrenia risk, with enrichment in genes under strong selection against PTVs, and (4) SETD1A rare variants are associated with both schizophrenia and developmental disorders. However, most CNVs are multigenic, leaving uncertainty about genic pleiotropy. Previous exome-based studies found enrichment of de novo variants in schizophrenia within NDD-implicated genes, but lacked power to test whether the same classes of variants (PTV vs missense) were shared across disorders or whether specific alleles overlapped, questions critical for establishing allelic pleiotropy and inferring equivalent functional perturbations across conditions.

Design: Two complementary analyses were conducted: (A) Genic pleiotropy—testing whether schizophrenia de novo variants are enriched in genes implicated in NDDs by specific functional class (PTV or missense). (B) Allelic pleiotropy—testing whether the same specific NDD de novo single-nucleotide variants (SNVs) are enriched among de novo variants in schizophrenia, and replication in case-control exome sequencing.

Datasets: Schizophrenia trios: 3444 proband-parent trios (2121 male, 1323 female) from 11 published studies with DSM-IV/ICD-10 schizophrenia or schizoaffective disorder. NDD trios: 37,488 developmental disorder (DD) trios from Kaplanis et al. and 31,058 ASD trios from Satterstrom et al.; combined for allelic analyses (total unique de novo SNVs = 46,772). Case-control replication: 4070 schizophrenia cases and 5712 controls from the Swedish exome sequencing dataset.

Variant processing: Schizophrenia de novo variants were re-annotated using Ensembl VEP (v96). PTVs included stop-gain, frameshift, and essential splice donor/acceptor. Missense variants were assigned MPC scores; missense with MPC ≥ 2 were prioritized as likely deleterious.

Genic pleiotropy analysis: NDD-associated genes were taken from the DDD study at exome-wide significance (P < 2.5×10^-6): 180 PTV, 156 missense, of which 53 were significant for both; stratified into three independent sets: PTV-specific (127), missense-specific (103), and PTV+missense (53). For each set, expected schizophrenia de novo counts were computed using published per-gene mutation rates (coverage-adjusted where possible). Enrichment was tested using two-sample Poisson rate ratio tests for PTVs and for missense (MPC ≥ 2). Differences between schizophrenia enrichment for PTVs vs missense within each gene set were assessed via Poisson regression. A multivariable Poisson regression across 13,015 genes assessed whether NDD PTV and NDD missense gene-wide statistics independently predict schizophrenia de novo enrichment, including models adjusted for brain expression (BrainSeq DLPFC) and gene constraint (gnomAD observed/expected) as covariates.

Allelic pleiotropy analysis (de novo overlap): All unique de novo SNVs from ASD and DD studies were compiled (46,772 unique). Variants were split into: Primary NDD set—PTVs in LoF-intolerant genes (pLi ≥ 0.9) and missense with MPC ≥ 2; Comparator NDD set—PTVs in genes with pLi < 0.9, missense with MPC < 2, and all synonymous variants. Tri-nucleotide context mutation rates were used to estimate per-allele de novo expectations in 3444 schizophrenia trios. Overlap of schizophrenia de novo variants with each NDD variant set was tested against expectation using two-tailed Poisson exact tests. Additional two-sample Poisson tests compared enrichment in the primary set to the background schizophrenia enrichment for the same mutation classes (PTVs in LoF-intolerant genes and missense MPC ≥ 2) to ensure the signal exceeded general constraint-driven enrichment.

Case-control replication: In Swedish exomes, ultra-rare variants (gnomAD allele count ≤ 5) were intersected with the NDD primary and comparator sets (including indels). One-tailed Firth penalized logistic regression tested burden differences between cases and controls, adjusting for 10 ancestry PCs, synonymous burden, platform, and sex.

Phenotype follow-up: Clinical features were collated for schizophrenia probands carrying primary-set NDD variants (onset age, premorbid cognition, comorbid NDD features) and compared to clinical annotations in ClinVar and DECIPHER for the same variants where available.

Genic pleiotropy: In 3444 schizophrenia trios, de novo PTVs were significantly enriched in genes associated with NDD by PTVs (127 genes): 20 observed vs 3.41 expected; P = 2.14×10^-8; rate ratio (RR) = 4.89 (95% CI 2.95–7.67). No enrichment was found for schizophrenia missense (MPC > 2) in these PTV-enriched genes (6/4.57; P = 0.47; RR = 1.32). The excess of PTVs relative to missense in these genes was significant (P = 0.006; RR = 3.70).

• In NDD missense-enriched genes (103 genes), schizophrenia missense variants were enriched (14/7.85; P = 0.04; RR = 1.86), while PTVs were not (4/2.94; P = 0.79; RR = 1.09). The difference in enrichment between missense and PTVs was not significant (P = 0.35).
• In genes enriched for both classes in NDD (53 genes), both schizophrenia PTVs (7/1.64; P = 0.0051; RR = 3.45) and missense (9/3.77; P = 0.014; RR = 2.47) were enriched.
• Genome-wide multivariable Poisson regression showed schizophrenia PTV enrichment tracks NDD PTV gene-wide statistics (beta = 0.15; P = 3.3×10^-11), not NDD missense (beta = -0.018; P = 0.67). Conversely, schizophrenia missense enrichment tracks NDD missense statistics (beta = 0.049; P = 0.0047), not NDD PTV (beta = 0.012; P = 0.38). Results were robust to including brain expression and constraint covariates.

Allelic pleiotropy (de novo overlap): Of 46,772 unique NDD de novo SNVs, 17 were also observed de novo in schizophrenia trios. There was significant enrichment for overlaps in the NDD primary variant set (5863 variants): 9 observed vs 1.20 expected; P = 5.0×10^-6; RR = 7.48 (3.42–14.20). No enrichment in the comparator set (40,909 variants): 8 vs 9.86 expected; P = 0.75; RR = 0.81 (0.35–1.60). The enrichment for specific NDD variants exceeded the general schizophrenia enrichment of constrained de novo variants (PTVs in LoF-intolerant genes and missense MPC ≥ 2): P = 1.01×10^-5; RR = 6.91 (3.11–13.38). By class, enrichment was significant for NDD PTVs (P = 0.01; RR = 6.77 [1.40–19.80]) and for NDD missense (P = 0.00014; RR = 7.90 [2.90–17.19]).

Case-control replication: In Swedish exomes, variants from the NDD primary set were more frequent in cases than controls: case rate 0.0044 (18 variants) vs control rate 0.0023 (13 variants); P = 0.036; OR = 1.90 (0.94–3.95). No difference for the comparator set (P = 0.42; OR = 1.01 [0.89–1.15]).

Specific overlapping variants: Nine NDD primary-set variants observed as de novo in schizophrenia implicate KMT2D, NF1, AUTS2 (PTVs) and GRIA3, RUNX3, SLC6A1, CSNK2A1, KLHL20, SCN2A (missense; MPC ≥ 2). Several are recorded as pathogenic/likely pathogenic for syndromic NDDs in ClinVar/DECIPHER (e.g., Okur-Chung syndrome—CSNK2A1; Kabuki syndrome—KMT2D; Neurofibromatosis type 1—NF1; AUTS2 likely pathogenic in Decipher). Clinical review showed variable neurodevelopmental features among schizophrenia carriers, with some lacking intellectual disability or ASD.

The study addresses whether shared genetic risk between schizophrenia and NDDs reflects the same functional variant classes within genes and whether identical rare coding alleles confer risk to both disorders. Results show that genes enriched for PTVs (or missense) in NDDs are also enriched for the same variant class in schizophrenia, supporting functionally congruent genic pleiotropy. Moreover, specific NDD de novo variants are significantly enriched among schizophrenia de novo mutations, providing evidence for allelic pleiotropy. These findings suggest that equivalent disruptions in gene function can lead to diverse clinical outcomes spanning developmental disorders, ASD, and schizophrenia, pointing to shared molecular etiology and overlapping pathophysiology. Contrary to the initial hypothesis that more severe mutation classes would predominate in NDDs versus schizophrenia, the data indicate that outcome severity likely depends on additional genetic background (e.g., common variant liability), environmental, or stochastic modifiers, rather than mutation class alone. Clinically, some schizophrenia patients without evident neurodevelopmental comorbidities harbor variants known to cause severe NDDs, underscoring the need to consider comorbidity and cognitive assessment within schizophrenia care and to recognize schizophrenia as part of a neurodevelopmental continuum.

This study demonstrates that schizophrenia and neurodevelopmental disorders share disruptive coding mutations at both the genic and allelic levels. Genes implicated by PTVs or missense variants in NDDs show congruent enrichment of the same variant classes in schizophrenia, and specific NDD de novo variants are significantly overrepresented among schizophrenia de novo mutations, with supportive replication in a case-control exome dataset. The findings support a shared molecular etiology and the view that at least a subset of schizophrenia lies on a neurodevelopmental continuum with ASD and developmental disorders. Future research should leverage larger trio and case-control sequencing cohorts to definitively implicate individual genes/variants, dissect modifier effects (common variant background, environment), and systematically evaluate quantitative cognitive and biomarker impacts of NDD variants in schizophrenia.

• Power and sample size: Despite significant gene-set and allele-set signals, definitive association of individual genes or specific variants with schizophrenia requires larger datasets. The number of overlapping de novo variants is small.
• Variant scope: Allelic pleiotropy analyses considered only single-nucleotide variants due to lack of empirically established indel mutation rates; thus, indel contributions to allelic pleiotropy in de novo data were not assessed.
• Gene set definitions: NDD gene sets were derived using an exome-wide significance threshold from DDD; ASD gene-specific PTV vs missense statistics were unavailable in the largest ASD study, limiting inclusion in genic pleiotropy analyses.
• Statistical considerations: Many P-values are uncorrected for multiple testing at the table level; effect estimates rely on mutation rate models and coverage adjustments that introduce uncertainty.
• Phenotypic characterization: Limited clinical data for schizophrenia probands constrain inferences about penetrance, expressivity, and detailed neurodevelopmental features among variant carriers.