Main

Schizophrenia is a neuropsychiatric disorder characterized by delusions and hallucinations, paranoia, apathy, anhedonia, social withdrawal, and extensive cognitive impairments.1 The disorder is relatively common, with a prevalence of 1.1% of the population older than 18 years (National Institute of Mental Health) and a worldwide incidence of approximately 1 in 4000.

The etiology of schizophrenia is not well understood, but it has been established that genetic predisposition is a major determinant of susceptibility to schizophrenia; adoption and twin studies suggest heritability estimates as high as 80%.2 However, understanding the genetic architecture of a complex disorder such as schizophrenia has been challenging. Conventional genome-wide linkage scans on sib pairs and affected pedigrees have provided limited evidence for strong susceptibility loci, often with limited reproducibility between the various studies.3,4 More recently, large-scale genotyping of thousands of cases and controls with hundreds of thousands of tagged single-nucleotide polymorphisms throughout the genome has allowed for the examination of common genetic variability associated with complex disease traits. Although such genome-wide association studies (GWAS) have identified common variants associated with complex diseases such as macular degeneration, rheumatoid arthritis, and colon cancer, the results from GWAS of schizophrenia have been less promising.5–9 Although multiple single-nucleotide polymorphisms have been associated with schizophrenia, none have reached genome-wide significance or could be conclusively implicated in the disease; furthermore, replication of the results from different studies has been limited.5–9 These data suggest that the heretofore accepted “common-disease/common-variant” model, in which common alleles with low to moderate disease risk may have a cumulative effect,10 may not be sufficient to explain the genetic factors that cause schizophrenia.

The inconclusive results of GWAS on schizophrenia, combined with the widespread use of higher resolution molecular cytogenetic assays such as microarray-based comparative genomic hybridization (aCGH) for the characterization of individuals with intellectual and physical disabilities, have renewed interest in the search for novel copy number variants (CNVs) associated with schizophrenia. Cytogenetic aberrations have long been associated with schizophrenia.11 The most well-characterized CNV associated with schizophrenia is the 22q11 microdeletion that results in velocardiofacial syndrome/DiGeorge syndrome (DGS); most individuals with DGS have cognitive deficits of varying severity, and approximately 30% have behavioral abnormalities, including schizophrenia, bipolar disorder, and autism.12–15 A second reported chromosome abnormality is alteration of the gene DISC1 on 1q42.2, which was first identified by a balanced translocation t(1;11)(q42;q14) that cosegregated in a large Scottish family with neuropsychiatric diseases including schizophrenia16–18 and has also recently been implicated in large-scale GWAS.19 These studies have shown that the genome-wide burden of rare CNVs is significantly greater in individuals with schizophrenia than in healthy controls.20–22 More recently, recurrent microdeletions at 1q21.1,23 15q13.3,21,24 and 15q11.2,22,24 microduplications at 16p11.2,25 and CNVs at other genomic loci26–28 have been associated with schizophrenia in large cohorts by CNV analyses and other molecular studies. The identification of these rare variants supports a “common-disease/rare-variant” (CD-RV) model, which suggests that the heterogeneity of schizophrenia comes from multiple rare variants with high penetrance.29,30 Some of these CNVs have been reported in a very small number of individuals, and large pedigrees are an exception.

To delineate the clinical significance of CNVs previously suggested to be schizophrenia candidate susceptibility loci, we searched our database for specific CNVs previously associated with schizophrenia in all our patients regardless of the indication for study (IFS). In addition, we examined the genomic alterations detected by microarray in individuals with schizophrenia as an IFS to determine whether there was concordance with the CNVs identified in the first group. Our results suggest that CNVs previously associated with schizophrenia are associated with a diverse spectrum of neurologic deficits.

MATERIALS AND METHODS

Patient ascertainment

Individuals in this study for whom additional clinical information was obtained provided written informed consent using an Institutional Review Board Spokane-approved consent form.

Patient identification—abnormalities encompassing previously reported schizophrenia candidate susceptibility loci

Between March 2004 and February 2010, our laboratory performed microarray testing on 38,779 probands who were referred for unexplained physical and/or intellectual disabilities (IDs) with or without dysmorphic features. All individuals with abnormalities of the same or reciprocal copy number state (i.e., gain or loss) overlapping previously reported schizophrenia candidate susceptibility loci were identified in our database (Tables 1 and 2). These included six previously published recurrent loci and 20 “rare” loci >100 kb that have been identified in large-population studies of individuals with schizophrenia6,21,24 and for which adequate probe coverage was present on our microarrays. The analysis for the latter set of 20 “rare” loci was restricted to include only the most relevant from multiple population-based studies and adequate sensitivity and specificity of our array platform for those genomic intervals. Abnormalities that were substantially larger than the size of the CNV reported in the literature were excluded from further analysis to eliminate cases in which additional genes outside the candidate locus were disrupted (e.g., whole-arm gains/losses). Clinicians were asked to supply clinical information for all cases with overlapping abnormalities. Parental and prenatal sample analyses were not included.

Table 1 Gains and losses identified by our laboratory of schizophrenia candidate loci
Full size table
Table 2 Summary of recurrent CNVs associated with schizophrenia identified by our laboratory
Full size table

Patient identification—schizophrenia as an IFS

We searched our database of 38,779 probands for individuals with a CNV and an IFS that included “schizophrenia,” either as a primary indication or as part of a family history. We identified six such individuals. Information about how a diagnosis of schizophrenia was made was not obtained (Table 3). We then identified all individuals in our database with overlapping abnormalities of the same or reciprocal copy number state (i.e., gain or loss) using the same criteria as earlier.

Table 3 Summary of microarray results for six individuals referred for schizophrenia or family history of schizophrenia and 37 individuals with overlapping abnormalities
Full size table

Microarray analysis

Array comparative genomic hybridization

Microarray analyses were performed between March 2004 and February 2010 with an evolving series of genomic microarrays. The successive seven versions of microarrays have increasing coverage of the genome. The version of the array used on a particular patient depended on the date of sample receipt. Version 1.0 of the bacterial artificial chromosome (BAC)-based SignatureChip® (Signature Genomic Laboratories, Spokane, WA) was used from March 2004 until October 2004, version 2.0 until October 2005, version 3.0 until May 2006, version 4.0 until November 2007, and Whole Genome until December 2009. In addition, oligonucleotide-based microarray analysis was performed on some of the individuals reported in this article using a 105K-feature whole-genome microarray (SignatureChip Oligo Solution™, designed by Signature Genomic Laboratories and made by Agilent Technologies), which was in use from November 2007 to April 2009. The SignatureChipOS V2, a 135K-feature oligonucleotide array designed by Signature Genomic Laboratories and made by Roche-Nimblegen (Madison, WI), is currently in use in our laboratory. A comparison of the genomic coverage and content of each microarray platform can be found at http://www.signaturegenomics.com/clone_list.html. Microarray analysis was performed as previously described for BAC31 and oligo-based32 arrays. All results were visualized using our laboratory-developed computer software program Genoglyphix (http://www.signaturegenomics.com/genoglyphix.html). Referral to our laboratory is most commonly based on the clinical presentation of developmental delay, dysmorphic features, developmental disabilities such as autism or epilepsy, and/or multiple congenital anomalies. Copy number alterations detected by aCGH were reported according to previously described criteria.33 Briefly, reported abnormalities included those that were associated with established genomic disorders, were large and affected a significant gene or genes within the critical interval, or were part of a complex rearrangement such as an unbalanced translocation. Smaller abnormalities were reported if they impacted gene content likely to contribute to the patient's phenotype or when the size of the abnormality could not be well defined if run on a BAC array. For oligonucleotide arrays, based on accepted criteria and extensive validation of the proprietary Genoglyphix software, average intensity thresholds of ±0.300 for intervals containing a minimum of five probes, ±0.200 for intervals containing 200 probes, and ±0.100 for intervals containing 500 probes were used to identify aberrant regions.

Fluorescence in situ hybridization analysis

Abnormalities detected by aCGH were visualized by metaphase or interphase fluorescence in situ hybridization using one or more BAC clones determined to be in the abnormal region as described previously.34,35

RESULTS

Individuals with CNVs overlapping recurrent and rare schizophrenia susceptibility loci

We analyzed a subset of 20 important and previously implicated schizophrenia susceptibility loci listed in Tables 1 and 2.6,20,24 Of these loci, 14 are nonrecurrent, and six are recurrent. We identified 78 individuals harboring CNVs across genomic regions/genes implicated in rare instances (Table 1) and 1035 individuals with CNVs of one of six recurrent identified regions—1q21.1, 15q11.2, 15q13.3, 16p11.2, 16p13.11, and 22q11.2 (Table 2).

Table 1 presents inheritance, age at diagnosis, and copy number state of CNV for the individuals in each of the rare loci analyzed in this cohort. Examination of the IFS for these individuals showed a spectrum of neurologic deficits, including developmental and speech delays, behavioral problems, autism/autism spectrum disorder (ASD), and seizure disorders. None of the individuals had an IFS of schizophrenia, although medical records were not obtained to completely exclude the diagnosis. Of these individuals, the inheritance was de novo in seven, maternal in 14, paternal in 12, and unknown in 45 because parents were not tested. The age range for this group was from 1 month to 23 years; the majority (n = 60) were at or below 19 years of age.

Thirteen individuals had a second potentially significant CNV (Table 1). Notably, two individuals with CNVs at the NRXN1 locus (2p16.3) had an additional loss: GC26449 had a 1.34 Mb loss at 13q12.12 and GC45066 had a 2.9 Mb loss at 3p12.3. One individual (GC43660) with a deletion overlapping the CNTNAP2 locus at 7q35q36.1 had an additional abnormality, a 322 kb loss at 16p13.3 that encompassed CREBBP, associated with Rubinstein-Taybi syndrome. One individual (GC25195) with a gain of material from 7q36.1 also had a 2.5 Mb deletion at the 22q11.2 (DGS-velocardiofacial syndrome) region. Three individuals with CNVs at 11q14.1 (DLG2 locus) had an additional abnormality: one (GC33254) had a 1.3 Mb duplication at the 15q13.2q13.3 microdeletion syndrome locus, which includes CHRNA7, one (GC43330) had a 1.7 Mb loss at 2q13 that includes at least five OMIM genes, and the third (GC46017) had an additional 160 kb loss at 2p16.3 that overlaps the NRXN1 gene. Two individuals with CNVs at 18p11.31p11.23 (LAMA1, ARHGAP28, and PTPRM) had additional significant aberrations; the first case (GC30619) had a 1.73 Mb loss at 5q14.3 that encompassed MEF2C, deletions of which have been associated with mental retardation and epilepsy36–38; the second (GC44238) had a 468 kb loss at 10q23.1 that overlapped NRG3 and a 520 kb gain at 18p11.31 that overlapped DLGAP1.

We identified 1035 CNVs encompassing one of the six recurrent loci (Table 2). For all six recurrent schizophrenia susceptibility loci we analyzed, the spectrum of phenotypes remained diverse and heterogeneous and distinct from schizophrenia (Table 2). For at least two of these loci (16p11.2 and 22q11.2), there were substantially more cases with copy number losses than with copy number gains, reflecting a potentially benign outcome in some carriers of the duplication. Parental testing was limited except for 22q11.2 and 16p11.2 deletion cases.

Individuals with schizophrenia as IFS

Between March 2004 and February 2010, we analyzed 38,779 individuals by aCGH. Of these, six with a likely clinically relevant CNV were referred with an IFS of schizophrenia (n = 5) or family history of schizophrenia (n = 1) (Table 3). One abnormality, a 3.62 Mb duplication at 4q21.22q21.23, was present in a 33-year-old man with schizophrenia and his child with developmental delay. We also identified 37 individuals with an overlapping gain/loss of a CNV identified in an individual with schizophrenia. Table 3 lists inheritance, age at diagnosis, and copy number state of CNV for the six individuals with an IFS of schizophrenia and the 37 individuals with overlapping CNVs. Examination of the IFS for individuals with CNVs overlapping those identified in individuals referred for schizophrenia showed a spectrum of neurologic deficits, including developmental and speech delays, behavioral problems, autism/ASD, and seizure disorders. Although none of the individuals had an IFS of schizophrenia, medical records were not obtained to completely exclude the diagnosis. Of these individuals, the inheritance was de novo in one, maternal in 15, paternal in four, and unknown in 17 because parents were not tested. The age range for this group was from 1 month to 20 years; all but one individual was at or below 19 years of age. None of the CNVs identified in the individuals with an IFS of schizophrenia overlapped known or putative schizophrenia susceptibility loci.

DISCUSSION

Elucidating the genetic architecture of complex disorders is confounded by the large number, low frequency, and variable effect of predisposition loci. Genomic CNVs have been established as a major source of human genetic variation that underlie many neurologic and neurodevelopmental syndromes including schizophrenia. Although a number of large-scale studies have now revealed an increased load of copy number variation in patients with schizophrenia compared with the normal population,21,22,24–28 no large-scale analysis has been performed to determine the phenotypic spectrum of the CNVs that have been implicated in schizophrenia. Although previous studies have characterized rare CNVs in large populations of individuals with schizophrenia, this study identified CNVs previously associated with schizophrenia in individuals referred for genetic testing for a broad spectrum of physical and developmental deficits. Our results further refine the clinical spectrum associated with schizophrenia susceptibility loci.

CNVs present in individuals with schizophrenia are associated with a diverse spectrum of neurodevelopmental deficits and are not unique to schizophrenia. For the nonrecurrent and recurrent loci, the phenotypic spectrum was diverse and included developmental delay, ID, autism spectrum, and multiple congenital anomalies. The 22q11.2 microdeletions are the most well-characterized example of heterogeneity; the deletion can be associated with multiple neuropsychiatric disorders including schizophrenia, attention deficit/hyperactivity disorder, bipolar disorder, and ASD.10,39,40 We have now identified 280 cases of CNVs at 22q11.2 (deletions [n = 186] and duplications [n = 94]) in individuals with diverse neurologic deficits including ID, DD, and seizures (Table 2). Evidence from other disorders also suggests pleiotropic effects. CNVs of the 1q21.1 region, for example, are associated with a spectrum of neurodevelopmental deficits including autism41 and schizophrenia.23 Although phenotypic heterogeneity has been established for many of the recurrent CNVs, our results broaden the neurodevelopmental spectrum of sporadic CNVs that had previously been identified in a few individuals in large-population studies. For example, we identified 19 individuals with CNVs of 2p16.3 that encompass the NRXN1 gene, which is a neurodevelopmentally important gene. The spectrum of neurologic deficits in our cases ranged from mild MR to seizure disorders and ASD.

The frequency in our patient and parent populations of some of the CNVs associated with neurodevelopmental conditions suggests that the CD-RV model may not sufficiently explain the role of these CNVs in schizophrenia. For all loci, we identified de novo, maternally and paternally inherited aberrations, although the origin of the CNV in many probands was unknown because one or both parents were unavailable for testing. In some cases, the CNV was inherited from a carrier parent with milder neurologic deficits than the proband. For example, the 177 kb microdeletion at 9p24.2 identified in patient GC15750 (Table 1) was inherited from his carrier father, who had a history of anxiety and depression. The inheritance of a CNV from a parent who displays a milder form of the IDs identified in the proband has been reported for some CNVs.42 In addition, although recent evidence suggests CNVs of 15q11.2, which is within the larger Prader-Willi/Angelman syndrome deletion region between breakpoints I and II, may be enriched in patients with idiopathic generalized epilepsy, autism, and other neurocognitive phenotypes,23,43 the 15q11.2 CNV is relatively common in the normal population and has rarely been reported to be de novo.24,44 Although the CD-RV model hypothesizes that one or a few rare, highly penetrant CNVs contribute to schizophrenia, the presence of these CNVs in a high proportion of reportedly normal carrier parents suggests they may indeed not be highly penetrant. These results support other studies in which the penetrance of recurrent schizophrenia candidate loci is between 2 and 7.4%.45

To determine the relative frequencies of the recurrent CNVs among our study population, we compared the identification rates of each recurrent CNV from November 2007 to February 2010, during which time our laboratory performed testing using whole-genome platforms that had dense coverage for all six recurrent CNV regions. This comparison allowed us to retain consistency in size and breakpoints for the recurrent deletions and duplications. During this interval, we tested 23,250 individuals. The frequency of the recurrent CNVs ranged from 0.14% for 16p13.11 microdeletions to 0.70% for 15q11.2 microdeletions (Table 4). The incidence of these CNVs in the general population can be estimated by comparing their detection rate in our laboratory with that of a genomic disorder with a well-established incidence, such as Smith-Magenis syndrome (SMS), which has a frequency in the general population of approximately 1/15,000.48 During this same period, we identified 27 SMS deletions (0.12% of patient population). By comparison, 1q21.1 microduplications were identified in 113 individuals (0.49% of our study population). Thus, because they were identified more than four times more frequently in our study populations than SMS deletions, the incidence of 1q21.1 microduplications in the general population may be inferred to be >1/3700. These may be overestimates of frequency because some cases of SMS are diagnosed by other methods, and therefore, not all individuals with these syndromes will have aCGH, whereas the abnormalities associated with neurodevelopmental disorders do not have clearly recognizable constellations of clinical features and would not be expected to be diagnosed by other methods. Comprehensive compilation of data from our clinical collection and that from focused study populations provide a comparison of the frequencies between the groups. Taken in concert with the frequency of these recurrent CNVs in apparently normal carrier parents, our data indicate that some of the recurrent CNVs associated with schizophrenia may not fit the CD-RV model. Rather, such CNVs may lie on a spectrum between rare highly penetrant mutations and common low-risk CNVs that, in aggregate, contribute to complex disorders. These CNVs may act as modifiers or vulnerability factors for neurologic deficits rather than directly influencing the specific disease outcome.42,45 Additionally, incomplete penetrance and variable expressivity may be valid explanations for assessing the outcome of many of these inherited CNVs. Most of these recurrent CNVs—with the notable exception of the 15q13.3 microduplication—seem to be enriched in our study population compared with a previously published population of normal control individuals (Table 4); the statistical significance of the frequency of these 11 recurrent CNVs compared with their frequency in the control population (χ2 with Yates' correction) revealed significant P values (<0.0001) for microdeletions at 1q21.1, 15q11.2, 15q13.3, 16p11.2, and 22q11.2 and microduplications at 1q21.1 and 16p13.11, respectively. The significance for the remaining loci was less profound (Table 4). These data further suggest that the odds of most of these abnormalities to be associated with a variable phenotype are significant or approach significance, with the exception of duplication 15q13.3. Further studies of the frequencies of these CNVs in normal control populations are necessary to establish their degree of enrichment in individuals with neurodevelopmental impairments.

Table 4 Estimates of frequency for recurrent CNVs associated with schizophrenia in our patient population and a previously published control population
Full size table

For some of the rare rearrangements that have been previously reported in only one case, our data suggest that these abnormalities are indeed clinically relevant and not rare benign polymorphisms. For example, we identified eight CNVs (seven deletions and one duplication) encompassing an approximately 440 kb interval on 9p24.2 that includes the RFX3 gene, deletion of which was identified in one individual with schizophrenia in the cohort described by Walsh et al.20 The deletion sizes in these individuals ranged from 177 kb to 3.5 Mb, and the clinical features in these individuals included autism, behavioral problems, developmental delay, and ID. The deletion was de novo in two individuals, paternally inherited in two, maternally inherited in one, and of unknown origin in three. Thus, although these CNVs may not be unique to individuals with schizophrenia, our results suggest rare, nonrecurrent CNVs that have only been identified in single cases in previous large-population studies6,20–22,24 may be clinically significant.

Because the symptoms of schizophrenia typically do not appear until early adulthood (with the exception of the more severe childhood-onset schizophrenia), we cannot rule out the future manifestation of schizophrenia in the young patients in our study. However, because the prevalence of schizophrenia in individuals with IDs and developmental delay is reportedly three times that of the normal population,49 we would expect that schizophrenia would be enriched in patients older than 16 years identified by our laboratory. Of the 1035 individuals with recurrent CNVs, 123 were older than 16 years at diagnosis. However, none of these individuals had an IFS of schizophrenia, and many had an IFS of a distinct neurologic deficit (e.g., ASD). It has been suggested that based on structural studies of the brains of individuals with comorbid learning disability and schizophrenia and those with only schizophrenia that the presence of schizophrenia (whether diagnosed or destined to develop) predisposes some individuals to develop IDs.50,51 This raises the possibility that schizophrenia may be at one end of a phenotypic continuum, starting with early childhood-onset developmental/ID that eventually culminates in schizophrenia in adulthood in a fraction of cases. Thus, one might conclude that the cooccurrence of similar genetic aberrations in individuals with MR/ID and individuals with schizophrenia is a result of comorbidity between schizophrenia and MR. However, the presence of heterogeneous neurocognitive deficits in our study, including some quite distinct from schizophrenia—such as attention deficit/hyperactivity disorder—and the absence of schizophrenia in any of our patients with MR suggest not that schizophrenia predisposes to MR/DD but that a common genetic insult predisposes to a myriad of neurologic problems. Long-term follow-up of our patients to determine and document any evolving phenotype will be required to address these issues. The other, more acceptable explanation may be that the genetic defects are truly pleiotropic and the outcome in an individual harboring a CNV can be very different and potentially influenced by environmental factors and modifier effects of other loci. These data also imply the existence of shared biologic pathways among multiple neurodevelopmental conditions, and the suggestion that any specific locus is unequivocally associated with only schizophrenia or any other specific neurodevelopmental conditions should be made with caution.