Rare Structural Variants Disrupt Multiple Genes in Neurodevelopmental Pathways in Schizophrenia

See allHide authors and affiliations

Science  25 Apr 2008:
Vol. 320, Issue 5875, pp. 539-543
DOI: 10.1126/science.1155174


Schizophrenia is a devastating neurodevelopmental disorder whose genetic influences remain elusive. We hypothesize that individually rare structural variants contribute to the illness. Microdeletions and microduplications >100 kilobases were identified by microarray comparative genomic hybridization of genomic DNA from 150 individuals with schizophrenia and 268 ancestry-matched controls. All variants were validated by high-resolution platforms. Novel deletions and duplications of genes were present in 5% of controls versus 15% of cases and 20% of young-onset cases, both highly significant differences. The association was independently replicated in patients with childhood-onset schizophrenia as compared with their parents. Mutations in cases disrupted genes disproportionately from signaling networks controlling neurodevelopment, including neuregulin and glutamate pathways. These results suggest that multiple, individually rare mutations altering genes in neurodevelopmental pathways contribute to schizophrenia.

Schizophrenia is a debilitating neuropsychiatric illness with severe individual, family, and societal burdens. The illness typically arises in late adolescence or early adulthood, with prevalence ∼1% worldwide. The phenotype is heterogeneous and complex, with multiple genes and environmental exposures likely involved. Family, twin, and adoption studies all support a strong genetic influence, although patterns of inheritance are variable and not consistent with a single monogenic Mendelian trait (1).

The present working hypothesis for genetic influences on schizophrenia is the “common disease–common allele” model (2), in which the illness is caused by combinations of common alleles, each contributing a modest effect. We propose an alternative model: that some mutations predisposing to schizophrenia are highly penetrant, individually rare, and of recent origin, even specific to single cases or families (3).

Genomic microduplications and microdeletions, also known as structural variants or copy-number variants (CNVs), underlie many serious illnesses, including neurological and neurodevelopmental syndromes (4). Previously, we found a significantly increased frequency of de novo deletions and duplications in children with autism spectrum disorders (5), suggesting that rare structural variants collectively account for a larger proportion of these disorders than previously recognized. Direct methods for detection of microdeletions and microduplications can be used for gene discovery in other psychiatric disorders as well.

The hypothesis that chromosomal mutations may cause schizophrenia is not new. Karyotypic abnormalities have been detected in affected individuals or families (6), including events that led to the discovery of potential disease-causing mutations in DISC1 (7), PDE4B (7), and NPAS3 (8). Deletions at 22q11.2 are also associated with schizophrenia (9). However, the ability to detect structural mutations associated with schizophrenia was historically limited to large events identifiable by karyotype. The development of microarray-based methods now enables the detection of much smaller events. These methods have revealed structural mutations involving genes in neurodevelopmental pathways in individuals with schizophrenia, including a deletion disrupting NRXN1 in affected siblings (10), a de novo duplication involving APBA2 in one patient (10), and deletions containing CNTNAP2 in two unrelated patients (11).

If severe mutations leading to schizophrenia are individually rare, then each individual mutation will be responsible for the disorder's occurrence in only one or a few patients. The overall involvement of rare mutations in the illness can be tested statistically by comparing the collective frequency of rare variants in cases with their collective frequency among controls. We examine whether rare structural mutations throughout the genome are more frequent among persons with schizophrenia than among unaffected individuals. We then examine whether genes disrupted by structural mutations in cases are more likely than genes disrupted in controls to encode proteins critical to brain development.

Genome-wide scans for structural variants greater than 100 kb were carried out on genomic DNA of 418 individuals, including 150 persons with schizophrenia or schizoaffective disorder meeting DSM-IV criteria and 268 healthy individuals (controls) aged 35 years or older who were free of signs of neurological or psychiatric illness. Detailed clinical information regarding the cases is provided in the supporting online material [(12) and table S1]. Distributions of racial ancestries were the same in cases and controls. Representational oligonucleotide microarray analysis 85K probe microarrays were used for event discovery. Illumina 550K microarrays and NimbleGen 2.1M feature HD2 microarrays were used to validate events and refine genomic breakpoints (12). Stringent quality-control criteria were applied to microarray data to obtain reliable measures of intensity, to exclude false positives, and to ensure that ascertainment of structural variants was consistent between cases and controls (12). First, frequencies in cases versus controls were compared for the 115 common variants (i.e., with frequency >1%) that passed our quality-control criteria and were verified by an independent method. Frequencies of these common variants did not differ between cases and controls (fig. S1). Second, detection sensitivity of rare variants was measured by simulating structural variants of various sizes at many different genomic locales and screening for them in blind tests. Detection sensitivity varied by event size and locale but was the same in cases and controls (fig. S2). Variants that were ultimately included in this study passed all these quality-control tests [see also (12)] and met our criteria for rare variants, defined as those not previously described on the Database of Genomic Variants, version 3, 29 November 2007 update (13). In the 418 cases and controls, we identified and validated 53 previously unreported microdeletions and microduplications, of sizes ranging from 100 kb to 15 MB.

Individuals with schizophrenia were more than three times as likely as controls to harbor rare structural variants that deleted or duplicated one or more genes (Table 1, P = 0.0008). Cases with onset of psychotic symptoms at age 18 years or younger were more than four times as likely as controls to harbor such variants (Table 1, P = 0.0001). In contrast, there was no significant difference in the proportions of cases versus controls carrying rare mutations that did not alter any genes (table S2).

Table 1.

Individuals with schizophrenia and controls with novel structural variants (SVs) of size >100 kb. Obs, observations; OR, odds ratio; CI, confidence interval.

View this table:

We tested the same hypothesis in an independent series of cases and controls, using a rigorously characterized cohort of youth with childhood-onset schizophrenia (COS). COS is a rare and typically more severe form of the illness (14). Of the 92 patients in this cohort, 9 patients known to harbor major chromosomal abnormalities (1517) were excluded from this analysis, leaving 83 individuals. Given the heterogeneous ancestries of the COS patients, we evaluated the nontransmitted chromosomes of available parents (n = 154) as controls. Parental relationships were validated by genotyping multiple polymorphic markers. Because each parent contributed one haploid nontransmitted genome to analysis, the effective diploid sample size of our control group was 154/2 or 77. Genome-wide scans for structural variants >100 kb and ≥20 single-nucleotide polymorphisms (SNPs) were carried out on genomic DNA from COS cases and their parents by Affymetrix 500K SNP arrays (Nsp I and Sty I chips). All events were validated by Agilent 185K or 244K array comparative genomic hybridization and by custom bacterial artificial chromosome arrays targeting genomic regions of segmental duplications (18). Rare variants were defined by the same criteria as for those in the original series. Insofar as possible, genotypes from parents were used to determine whether variants were de novo or inherited.

In the COS cases, we identified and validated 27 previously unreported deletions and duplications in 23 individuals. A 2.5-MB deletion on chromosome 2q31.2-q31.3 and a 183-kb duplication on Yq11.221 were not inherited from confirmed parents and were therefore presumptively de novo; other events were either inherited or of unknown origin. A 115-kb deletion on chromosome 2p16.3 disrupting NRXN1 was found in identical twins concordant for COS. Deletions in NRXN1 have been reported in other patients with schizophrenia (10), mental retardation (19), and autism (20). In addition, two COS cases harbored the recurrent microduplication of 500-kb on chromosome 16p11.2 that was recently associated with autism and also was detected in two individuals with bipolar disorder (21, 22). As found with the 16p11.2 duplication in autism (21, 22), some duplications and deletions in COS patients were inherited from unaffected parents.

Among the 83 individuals with COS, 23 (28%) harbored one or more rare structural variants that deleted or duplicated genes; 4 individuals carried more than one such event (table S3). Among the 77 controls constructed from nontransmitted chromosomes of the COS cases' parents, 10 (13%) carried such variants. As for the original series, individuals with schizophrenia were significantly more likely than controls to harbor rare structural variants that altered genes (Table 1, P = 0.03). The proportion of controls harboring a rare deletion or duplication was higher in the COS series than in the original series. Schizophrenia of very young onset is associated with more severe family history as compared with schizophrenia of older onset (14), so parents of COS cases are likely to harbor more genomic lesions related to mental illness as compared with parents of persons from the general population. The use of nontransmitted parental chromosomes as controls is therefore likely to be very conservative. Deletions and duplications in COS cases and their parents were screened with different platforms from those used in the original series of cases and controls, but within each series, screening was the same for cases and controls. Definitions and sizes of CNVs are not yet identical across platforms (13).

Structural mutations that disrupt genes (rather than deleting or duplicating entire genes) may be especially likely to have biological consequences. Gene disruptions include premature truncation, internal deletion, gene fusion, or disruption of regulatory or promoter elements (23). Therefore, in our original series, we determined the breakpoints of genomic deletions and duplications in cases and controls by high-density microarray analysis and genomic sequencing (12). Genes disrupted by the breakpoints of each event were identified (Table 2). Individuals with schizophrenia were significantly more likely than controls to harbor rare structural variants that directly disrupted one or more genes (Table 1).

Table 2.

Novel SVs in genomic DNA that delete (del) or duplicate (dup) genes in schizophrenia cases and controls. Chr, chromosome; hg18, human genome assembly 18 (March 2006). Dashes indicate that no gene was disrupted by SV breakpoints.

View this table:

Virtually every rare structural mutation detected in our original series was different. Some deletions and duplications were multiple megabases in size; other variants altered only one or a few genes. Some events in unrelated patients altered the same genes, but with subtly different genomic breakpoints. For example, similar structural variants on chromosome 18p11 disrupted genes LAMA1 and PTPRM in a case in the original series (Table 2) and in a case with COS (table S3).

Next, we assessed whether the functions of the genes disrupted in cases or controls might be related to schizophrenia. From our original series, for which disrupted genes were fully identified, we compared genes disrupted in cases versus those disrupted in controls using PANTHER (24) and Ingenuity Pathways Analysis (IPA) (25) classification systems. These programs enable one to determine whether an experimentally derived set of genes is overrepresented, as compared with all known genes (26), in one or more functionally defined pathways. The analyses are undirected: that is, no a priori selection is imposed on the pathways to be queried.

Genes disrupted by structural variants in our series of cases were significantly overrepresented in pathways important for brain development, including neuregulin signaling, extracellular signal–regulated kinase/mitogen-activated protein kinase (ERK/MAPK) signaling, synaptic long-term potentiation, axonal guidance signaling, integrin signaling, and glutamate receptor signaling (Table 3). Of the 24 genes disrupted by microdeletions or microduplications in cases, 11 genes—ERBB4, MAGI2, DLG2, PRKCD, PRKAG2, PTK2, CAV1, GRM7, SLC1A3, PTPRM, and LAMA1—were components of one or more of these pathways. Genes disrupted in controls were not overrepresented in any pathway. It is therefore important that this association was found only for cases and not for controls, that this association was not the result of the larger number of genes disrupted by case events [(12) and table S4], and that the signaling pathways implicated were specifically involved with brain development.

Table 3.

Pathways and processes overrepresented by genes disrupted in schizophrenia cases. No pathways were overrepresented by genes disrupted in controls.

View this table:

Mutations in two genes, ERBB4 and SLC1A3, are predicted to have particularly deleterious effects on function. For each of these genes, we experimentally characterized mutant transcripts in the affected individuals. The ERBB4 gene, which encodes a type I transmembrane tyrosine kinase receptor for neuregulins, is disrupted by a 399-kb deletion in one case (Fig. 1A). Using 3′ rapid amplification of cDNA ends, we determined that the mutant allele produces an alternate transcript in which exons 1 to 19 are spliced to 198 base pairs (bp) of the intergenic sequence located downstream of ERBB4 and the deleted genomic region (Fig. 1B). The alternate transcript is predicted to encode amino acids 1 to 767 of ERBB4 followed by a novel extension of 15 amino acids, which would result in a protein lacking most of the intracellular kinase domain (Fig. 1C). The predicted mutant protein is likely to function similarly to an engineered dominant-negative ERBB4 (DN-ErbB4) (Fig. 1D), which causes defects in neuronal migration (27) and synaptic neurotransmission (28, 29).

Fig. 1.

Structural mutations in genomic DNA from schizophrenia patients leading to expression of truncated ERBB4 and to expression of an in-frame fusion of SKP2 and SLC1A3. (A) Genomic structure of the 399-kb deletion in ERBB4 in a schizophrenia case. (B) cDNA sequence of the mutant ERBB4 transcript detected in the patient's lymphoblasts, in which the sequence from exon 19 is spliced to the downstream intergenic sequence, with encoded amino acids indicated above (47). (C) Domain structure of the ERBB4 protein with the site of mutation indicated (*). TM, transmembrane domain. (D) Amino acid sequence of the wild-type ERBB4 kinase domain as compared with that of mutant ERBB4 and of an engineered dominant-negative ERBB4 construct (DN construct) that disrupts signaling (27). The predicted mutant protein has a truncated kinase domain and a novel 15 amino acid extension followed by a stop. Kinase domain amino acids are highlighted in yellow. (E) Chimeric SKP2-SLC1A3 gene formed by a 503-kb deletion in a schizophrenia case. The indicated fusion transcript, including exons 1 and 2 of SKP2 (yellow) and exons 4 to 10 of SLC1A3 (blue), was detected in the patient's lymphoblasts. The cDNA sequence and encoded amino acids surrounding the point of fusion are shown. Arrows indicate locations of primers used to amplify the fusion region by reverse transcription polymerase chain reaction.

SLC1A3, a glutamate transporter widely expressed in glial cells, regulates neurotransmitter concentrations at excitatory glutaminergic synapses (30). Elevated levels of SLC1A3 message have been described in postmortem thalamus tissue of individuals with schizophrenia (31). The 503-kb deletion in another case disrupts both SLC1A3 and SKP2 and leads to a chimeric transcript expressed in patient lymphoblasts (Fig. 1E). The predicted fusion protein is likely to differ in expression and function from that of wild-type SLC1A3.

These genes and nine other genes disrupted by structural variants in cases contribute to cellular signaling networks critical to neuronal cell growth, migration, proliferation, differentiation, apoptosis, and synapse formation (32). ERBB4 is a receptor for neuregulin (NRG1) and interacts with MAGI2 at neuronal synapses (33) and with DLG2 (34). The NRG1-ERBB4 complex helps regulate neuronal migration and differentiation, neurotransmitter receptor expression, glial proliferation, and synaptic plasticity (35) and is critical to the development of glutamate networks (28, 29). GRM7 [a G protein–coupled metabotropic glutamate receptor (36)] and SLC1A3 [a glutamate transporter (30)] are also involved with glutamate signaling. Glutamate is the primary excitatory neurotransmitter (37) in the human brain. Glutamate pathways coregulate intracellular signaling cascades, including ERK/MAPK (38), nitric oxide (39), and integrin (40) signaling pathways: all of which play key roles in the orchestration of neurogenesis and neural circuitry (4143). Aberrations of NRG1-ERBB4 pathways and of glutamate neurocircuitry have been implicated previously in schizophrenia. NRG1 is among the best-supported susceptibility genes in association studies of schizophrenia (1). Variants of ERBB4 and interactions among genes in the NRG1-ERBB signaling pathway have also been associated with the illness (4446).

In two independent cohorts, we found structural variants altering genes to be more common among individuals with schizophrenia than among controls. Our design does not prove the involvement with the illness of any specific variant or even the involvement of any specific gene. Rather, these results suggest that schizophrenia can be caused by rare mutations that disrupt genes in pathways of neuronal development and regulation. This model suggests a new approach to gene discovery for schizophrenia and likely for other psychiatric disorders. Neurodevelopmental pathways involve hundreds of genes. A severe mutation in any one of these genes may lead to a psychopathological phenotype. Furthermore, independent mutations in the same gene or genomic region, such as 16p11.2 or NRXN1, may lead to expression of different neurodevelopmental phenotypes, ranging from schizophrenia to autism to mental retardation (10, 1921) or to no clinical manifestation at all. Variable expression may arise by different genetic or epigenetic “second hits” or simply by chance.

Gene discovery efforts should focus on methods that allow detection of structural mutations genome-wide in affected individuals. Any gene harboring a deleterious structural mutation becomes a candidate gene to be screened by other methods for additional, presumably smaller, deleterious mutations in unrelated individuals. This second analytic step would entail complete analysis of variation of a limited number of candidate genes in a very large number of cases and controls. The underlying hypothesis is that a gene harboring one mutation for an illness is likely to harbor more than one mutation. Although each mutation may be individually rare, the total number of disease-causing variants in a gene relevant to the disorder may collectively explain a substantial number of cases. The long-term goals are to identify all genes with mutations leading to the illness and to develop treatment and prevention strategies tailored toward the remediation of the altered pathways.

Supporting Online Material

Materials and Methods

Figs. S1 to S3

Tables S1 to S4


References and Notes

Stay Connected to Science

Navigate This Article