|Home | About | Journals | Submit | Contact Us | Français|
Deletions and reciprocal duplications of the chromosome 16p13.1 region have recently been reported in several cases of autism and mental retardation (MR). As genomic copy number variants found in these two disorders may also associate with schizophrenia, we examined 4345 schizophrenia patients and 35 079 controls from 8 European populations for duplications and deletions at the 16p13.1 locus, using microarray data. We found a threefold excess of duplications and deletions in schizophrenia cases compared with controls, with duplications present in 0.30% of cases versus 0.09% of controls (P = 0.007) and deletions in 0.12 % of cases and 0.04% of controls (P > 0.05). The region can be divided into three intervals defined by flanking low copy repeats. Duplications spanning intervals I and II showed the most significant (P = 0.00010) association with schizophrenia. The age of onset in duplication and deletion carriers among cases ranged from 12 to 35 years, and the majority were males with a family history of psychiatric disorders. In a single Icelandic family, a duplication spanning intervals I and II was present in two cases of schizophrenia, and individual cases of alcoholism, attention deficit hyperactivity disorder and dyslexia. Candidate genes in the region include NTAN1 and NDE1. We conclude that duplications and perhaps also deletions of chromosome 16p13.1, previously reported to be associated with autism and MR, also confer risk of schizophrenia.
Copy number variants (CNVs) are emerging as an important genomic cause of autism, mental retardation (MR) and schizophrenia.1–8 There is no simple pattern: several CNVs are associated with a broad range of neuropsychiatric phenotypes with no respect for traditional clinical diagnostic boundaries,4–6,8 and no agreed explanation is found for this clinical and genetic overlap.9–11 Despite high mutation rates,6 some recurrent CNVs seem to maintain themselves at low frequencies in human populations through the reduced fecundity and negative selection associated with severe neuropsychiatric disorders such as autism, MR and schizophrenia.
Ullmann et al.12 performed a high-resolution, tiling path BAC array comparative genomic hybridization on DNA from 70 autistic individuals from an Australian cohort. Identical submicroscopic 16p13.1 duplications were found in two unrelated patients. The duplication encompassed an interval of 1.5 Mb, ranging from chromosome 16p, 14.89–16.39 Mb (Human Genome Build 35). A third duplication was identified by quantitative PCR in a second Australian cohort of 112 patients. Two of the duplications were familial, and in one family, a severely autistic brother of the index case also carried the duplication. One of the brothers was continuously hyperactive, destructive and aggressive, whereas the younger brother was passive and easy to manage. Other carriers included a sister, who had learning difficulties and mother who had learning difficulties coupled with obsessive–compulsive disorder. They also reported two deletion patients with severe MR. The former was de novo; the latter had a mildly affected carrier mother. Hannes et al.13 in a study of 1027 patients with MR and/or multiple congenital anomalies found six deletions and seven duplications in the chromosome 16p13 region. Five deletions and five duplications corresponded roughly to the CNVs found by Ullmann et al.12 with distal breakpoints at 14.7–14.75 Mb and proximal breakpoints at 16.3–16.77 Mb (Human Genome Build 36). Two further duplications had distal breakpoints at 15.1–15.4 Mb and proximal breakpoints at 18.05–18.45 Mb, and one deletion had breakpoints at 16.3–16.77 and 18.3–18.4 Mb. Despite phenotypic variability, common features included three deletions with MR, microcephaly and epilepsy, and three duplications with pronounced behavioral problems in addition to MR and/or multiple congenital anomalies. However, when they examined the region in roughly 2000 controls, they found five duplications (no estimate of breakpoints provided) and no deletions, leading them to the conclusion that, although the deletion was probably pathogenic, the reciprocal duplication might be benign.13
The chromosome 16p13.1 duplication/deletion interval is located in a region with reported linkage to bipolar disorder14–17 and to puerperal psychosis.18 In addition, analysis of 458 Finnish schizophrenia families reported linkage to the DISC1 locus;19 however, when these families were later conditioned for a risk haplotype spanning intron 1 and exon 2 of the DISC1 gene, linkage was found in 16p13.1 region.20 The duplicated/deleted region contains the gene coding the DISC1-binding protein, NDE1, and the authors found significant allelic association between NDE1 and schizophrenia. However, this has not been confirmed in a recent Japanese study.21 Finally, another significant association was recently reported between NDE1 and schizophrenia, when schizophrenia cases and controls were conditioned for the presence of Cys residue at the non-synonymous polymorphism Ser704Cys encoded by the DISC1 gene.22
In the present study, we assessed association of CNVs in the 16p13.1 region with schizophrenia as part of a genome-wide scan using the Illumina microarrays HumanHap300, HumanHap550v3 and Human 610-Quad (Illumina, San Diego, CA, USA); and the Affymetrix SNP 6.0 microarray (Affymetrix, Santa Clara, CA, USA) in a sample of 4345 schizophrenia patients and 35 079 controls from 8 European populations, namely, Iceland, Finland, Germany, the Netherlands, Norway, Italy, Denmark and the United Kingdom. Preliminary details on the Scottish sample and part of the German sample have already been documented.7,23
A total of 4345 schizophrenia affected individuals and 35 079 screened controls from eight European populations were successfully examined at deCODE for CNVs at the locus studied here; they comprised 1435 schizophrenia patients and 28 554 control individuals from Iceland, Scotland, Germany, England, Italy and Finland (The SGENE sample; http://www.SGENE.eu); to these were added a further 866 patients and 856 controls from Aberdeen, Scotland and Munich, Germany; 491 patients and 881 controls from Bonn, Germany; 502 patients and 477 controls from Denmark; 806 Dutch cases and 4039 controls; and 245 patients and 272 controls from Norway. For a full description of samples see Supplementary Information, part one. Ethical approval was obtained from the local Ethics Committees. All participants gave a written informed consent.
The SGENE samples were typed on the HumanHap300 BeadArray (Illumina) at deCODE genetics. The additional samples from Aberdeen and Munich were typed at Duke University in collaboration with GlaxoSmithKline on HumanHap550v1 and HumanHap300 BeadArray (Illumina, respectively). The samples from Bonn were typed at Bonn University on the HumanHap550v3 BeadArray (Illumina). The Dutch samples from Utrecht University were genotyped at the University of California, Los Angeles, on HumanHap550v3 BeadArray (Illumina). The remaining Dutch samples were genotyped at deCODE genetics on HumanHap300 BeadArray (Illumina). The Danish samples were genotyped at deCODE genetics on the Human 610-Quad BeadArray (Illumina). The Norwegian samples were genotyped on Affymetrix Gene-Chip(r) GenomeWide SNP 6.0 array and analyzed using the Affymetrix Power Tools 1.8.0.
Dosage Miner software developed at deCODE genetics and QuantiSNP software developed at Wellcome Trust Centre for Human Genetics and the University of Oxford (http://www.well.ox.ac.uk/QuantiSNP/)24 were used to identify deletions and duplications within the region reported by Ullmann et al. and Hannes et al.12,13 Dosage Miner, described in detail elsewhere,6 uses the intensities from single nucleotide polymorphism (SNP) probes on the Illumina microarrays to estimate copy number of genomic regions and models factors, such as SNP effect, sample effect and GC-content the in the neighboring regions, to normalize the intensities. The software then automatically registers SNP loci in which intensities fall above or below an empirical threshold.
The QuantiSNP program relies on an Objective Bayes Hidden-Markov Model to estimate copy number variations. In this model, the hidden states denote the unknown copy number at the inspected SNPs. Genotype data was used to compute different states. The algorithm computes a Bayes factor that is used to calibrate the model to a fixed type I (false-positive) error rate. A Bayes factor threshold of 10 is considered as a promising value for the possible presence of a CNV. Usually, such values occur when 5–10 consecutive SNPs are deleted/duplicated. Differences in GC base pairs may result in biased hybridization behavior of SNP probes bearing the risk of miscalling genotypes. To normalize this, QuantiSNP assigns a locus-specific GC value to each probe. All potential CNVs that were detected by both softwares, and spanned at least 10 consecutive SNP probes, were subsequently visually inspected and confirmed.
A two-tailed Cochrane–Mantel–Haentzel analysis assuming common odds ratios was performed, stratifying samples by country of origin to take account of the possible effect of geographical variation on the results of the analysis.
We limited our search on 16p13.1 to the region between 14.66 and 18.70 Mb (Human Genome Build36). We subdivided the region into three single copy sequence intervals, which we called intervals I, II and III; each interval is flanked by sequences rich in low copy repeats with 99% sequence homology (Figure 1). The duplications and deletions previously reported12,13 are all contained within this region, with the most common breakpoints in the low copy repeat clusters distal to interval I and proximal to interval II (Figure 1).
Table 1 lists the duplications and deletions found in our sample, together with the population of origin. None were found in cases or controls from England (n = 104 and 95), Italy (n = 86 and 92), Finland (n = 191 and 200) and Norway (n = 245 and 272). Accordingly these samples are not included in the table or in the Cochran–Mantel–Haentzel statistical analysis.
Owing to varying geographical origin of the samples, we analyzed the data for association using Cochran–Mantel–Haentzel algorithm to correct for stratification. Within the total population examined, we found a threefold excess of duplications and deletions in cases compared with controls (see Table 2). We found six distinct CNV categories in the region, corresponding to deletions and reciprocal duplications of three overlapping regions (Figure 2). Duplications were present in 0.30% of schizophrenia cases versus 0.09% controls (P = 0.007). When analysis was restricted to duplications containing intervals I and II (Dup I + II, Figure 2), the significance increased further (P = 0.00010). The Dup I + II was present in four male and two female Scottish cases, two male Icelandic, one male Danish and three male Dutch cases, as well as in 12 female and six male Icelandic controls, in none of the Scottish controls and in two Dutch male controls. Common odds ratio was 8.79 (males) and 3.63 (females). The two Icelandic cases were independently ascertained and are included in the analysis as separate probands. However, when genealogical analysis was later performed, we found that the two were second-degree relatives. Other carriers in the family included single cases of alcoholism (under treatment), dyslexia and attention deficit hyperactivity disorder (See Supplementary Figure 1).
Deletions were present in 0.12% cases and 0.04% controls (P > 0.05). A deletion of interval I and II (Del I + II, Figure 2) was present in 3 German and 1 Danish schizophrenia cases, and 1 German and 10 Icelandic controls (P > 0.05), whereas a deletion of intervals II and III (Del II + III, Figure 2) was present in 1 Scottish schizophrenia case, and 2 German and 2 Icelandic controls (P > 0.05).
We also tested allelic association for all SNP markers on the Illumina microarrays that spanned the 16p13.1 region in 2687 schizophrenia cases and 13 484 controls. Although none of the markers in the interval were genome-wide significant, one marker, rs2283508, was associated (P = 1.5E-05) and remained significant after locus-wide correction P = 0.0043 (117 markers tested). This marker is located within an intron of the ABCC6 gene. No other markers spanning the region were significant after correcting for the 117 markers on the HumanHap300 chip in the interval spanning the duplication (see Supplementary Information, part 3).
In view of the report22 of significant associations between NDE1 and schizophrenia when schizophrenia cases were stratified by the presence of a Cys residue at Ser704Cys of the DISC1 gene, we also conditioned our schizophrenia cases. The DISC1 Ser704Cys SNP, rs821616, is not on the HumanHap300 chip. However, a SNP that is in complete linkage disequilibrium (r2 = 1) with rs821616 in the Hapmap CEU population, rs821596, was present. We therefore used rs821596 to divide the schizophrenia cases into Cys704-carrier and non-carrier groups, and then looked for allelic association with SNPs at the NDE1 locus in the two groups. None were significantly associated after correcting for the number of tested markers. The data regarding 51 SNPs in, or within, 200 Kb of NDE1 for the Cys704 carrier and non-carrier groups are given in Supplementary Information, part 3.
We also performed exploratory analyses. As the majority of the duplication cases were Scottish in origin, we examined the haplotype background of the duplicated regions. The CNV occurred on a different haplotype background in each individual. In addition, none of the non-Icelandic carriers, for whom we had genotype data, had a CNV on the same haplotype background as any of the Icelandic carriers. This suggests that there was no founder mutation, and each of the events is likely to have arisen independently. Within Iceland itself, for each of the CNV duplication and deletion subtypes (Figure 2) found in more than one individual, there was no founder mutation. There were also enough individuals in the Icelandic population with Dup I + II to look at clustering patterns. Clustering occurred at a rate of three- to fourfold less than expected if the duplications were selectively neutral. See Supplementary Information, part 4 for discussion of cluster analysis.
We have found a statistically significant (P = 0. 007) threefold overrepresentation of duplications in the chromosome 16p13.1 region in schizophrenia cases compared with controls. We found a similar threefold overrepresentation of deletions in cases, but this was not significant (P > 0.05). The great majority of duplications and deletions we found using Illumina microarrays are identical to those reported by two groups using array comparative genomic hybridization.12,13 They span the same 1.5 Mb region that includes intervals that we refer to as intervals I and II; we have also identified novel duplications and a single deletion involving interval II only.
The distinction between CNVs involving intervals I and II and interval II alone is dependent on a group of seven SNP probes at 15.0–15.1 Mb (Figure 1). However, we are confident that our classifications are correct; every CNV detected by the independent Dosage Miner and QuantiSNP algorithms was then inspected by eye, and in all cases, we could make a clear distinction between those involving both intervals I and II (Dup/Del I + II) and interval II only (Dup/Del II). In addition, genealogical analysis of the Icelandic data set, performed post hoc, showed a clear separation of CNV types. In total, 36 out of 54 carriers were clustered in 11 families (ancestral clustering, max depth three generations), including seven families with Dup I + II (nine, three and five times two carriers), two families with Del I + II (six and two carriers), one family with Dup II (four carriers) and one family with Dup II + III (two carriers). In no instance did we observe two different CNVs segregating within the same family.
The breakpoints for all three types of deletion/duplication are located in areas with high low copy repeat content, reflecting apparent genomic instability of the region.25 The repeats are in the same orientation, and non-allelic homologous recombination (NaHR) between these low copy repeats seems to be the most likely explanation for the recurrence of the rearrangements and for their identical size. Three inversion polymorphisms have previously been described in the 16p13.1 region.26,27 A large duplication in a patient with MR has also been reported,28 as has a smaller de novo duplication.29 A much larger duplication (8 Mb) of the region has also been reported in two unrelated patients with autistic features.30
Our most striking finding is the increased risk of schizophrenia associated with duplications at the 16p13.1 locus. Recurrent deletions at several loci have now been reported to be significantly associated with schizophrenia, but, to date, duplications associated with schizophrenia have mostly been isolated case reports. It is more difficult to decide whether the duplications are genuinely pathogenic or benign. The problem has already been encountered by Hannes et al.13 with duplications in MR at this locus. They found duplications in five out of 1027 cases, three of whom had striking behavioral problems similar to those reported by Ullmann et al.12 However, because the duplication was present in 5 out of 2000 controls, they could not decide whether it was genuinely pathogenic. The different sizes of duplications and deletions at the 16p13.1 locus also present difficulties when it comes to assessing statistical association. Statistically, we have used the straightforward approach of counting all duplications and then deletions as equivalent events, and only then tried to condition on those duplications or deletions that have the same breakpoints as those reported previously in autism and MR/multiple congenital anomalies.12,13
Although caution must be exercised when interpreting results from such a small number of cases, there are several grounds supporting that our findings are indeed genuine. First, given the rarity of the duplications, the overall association with schizophrenia is statistically (P = 0.0071) significant, and increases (P = 0.00010) when the Dup I + II is considered separately. In addition, identical duplications at the 16p13.1 locus have already been found associated with autism.12 Second, four of the schizophrenia duplication cases had an early onset of illness (12, 17, 19 and 19 years) and in this respect resembled the 16p13.1 deletion cases in which three of the five schizophrenia cases also had early onset of illness (15, 17 and 18 years, see Supplementary Information, part 2). With a mean age at onset of 24 years, this reflects an overrepresentation of early onset cases, although not reaching significance. Third, in an Icelandic family, the duplication co-segregates with neuropsychiatric disorders, including two cases of schizophrenia, and individual cases of attention deficit hyperactivity disorder, dyslexia and alcoholism. Owing to the small size of the family and the diversity of the neuropsychiatric phenotypes, calculating a logarithm of the odds (LOD) score did not seem appropriate. This range of phenotypes we observe in a single family is not unexpected, as an overlap of phenotypic features between autism and attention deficit hyperactivity disorder has been extensively reported, and individuals with attention deficit hyperactivity disorder are at increased risk of schizophrenia.31–33 Fourth, the duplications at this locus seem to be under negative selection. Cluster analysis of the Dup I + II events in the Icelandic population finds that the carriers cluster less than expected in families, that is, the genealogical clusters are smaller, and sporadic carriers more numerous than would be expected if the duplications were selectively neutral. The arguments and methods are laid out in Supplementary Information, part 4 and are also given in Stefansson et al.6 These clustering results, however, are informal, and so any conclusions must be provisional. Fifth, Kirov et al.34 have identified three chromosome 16p13.11 duplications in 471 schizophrenia cases and 6 out of 2792 controls. In addition, the International Schizophrenia Consortium finds overrepresentation of large 16p13.11 duplications (13 of 3391 cases versus 7 of 3181 controls have duplications encompassing interval II, see at http://pngu.mgh.harvard.edu/isc/isc-r1.cnv.bed),7 but this overrepresentation is mostly explained by the Aberdeen sample, which overlaps between our study and the International Schizophrenia Consortium study (although separately genotyped and analyzed), and therefore, does not provide further support to our finding.
Finally, with regard to the Dup I II duplications, although the Scottish sample seems+to be driving the association observed, as 6 out of 12 carriers among affected cases belong to that sample, the association of Dup I + II in this study remains significant even after removing the Scottish samples from the analysis (P = 0.0033, OR = 5.94).
The duplicated region contains two strong candidate genes (NTAN1 and NDE1), over- or underexpression of either or both of which at key stages of neurodevelopment could predispose to autism, MR and/or schizophrenia.
The NTAN1 gene is located in the small island of non-repeated sequence called interval one. It encodes an N-terminal asparagine amidase that has been implicated in social behavior and memory. Over-expression of NTAN1 leads to reduction in the MAP2 (microtubule-associated protein 2) expression through the ubiquitin proteasome pathway. Reduced expression of MAP2 may be a useful marker for diagnosis of schizophrenia and bipolar disorder in vivo35,36 and in vitro.37,38 Mice with a disrupted NTAN1 gene show less locomotion in an open field and impairment of several spatial memory tasks.39,40
NDE1 and NDEL are highly homologous genes involved in brain development, neuronal proliferation, migration and synapse formation. They encode proteins that biologically interact with DISC1 and LISI proteins, with NDE1 appearing to be interchangeable with its homolog NDEL, except that NDE1 is expressed earlier in development. nde1-null mice are viable and display microcephaly with thinning cortical layer and reduced numbers of neurons. Interestingly, two out of three reported autism cases with duplication had increased head circumference.12 The nde1-null mice display defects in neuronal proliferation and neuronal migration. The NDE1 protein directly interacts with the DISC1 protein at the C-terminal end that is distal to a truncating mutation that is reported to segregate with schizophrenia and other forms of major mental illness in a large Scottish DISC1 translocation family.41,42 As truncated DISC1 is known to alter NDE1/NDEL function, the duplications we report here may have a similar biological effect as the truncating mutation associated with schizophrenia in the Scottish family.
All duplications and deletions in our study involve interval II that harbors the NDE1 gene and this makes dysregulation of NDE1 expression the most parsimonious explanation for the increased risk of the phenotypes reported by Ullmann et al.12 and by the authors of this paper. On the other hand, the strongest association is with duplications that also involve interval I. It is possible that combined changes in expression of NTAN1 and NDE1 increase susceptibility over changes in expression of NDE1 alone. No clear-cut findings emerged from our examination of the 16p13.1 region for allelic association for schizophrenia. The findings are discussed along with association studies following conditioning for DISC1 Ser704Cys in Supplementary Information, part 3.
Further work is required before the clinical implications of our findings become clear. On the one hand, the data strongly suggest that recurrent duplications at 16p13.1 locus increase the risk of schizophrenia. They also strengthen the hypothesis that there are shared genetic risk factors between schizophrenia, MR and autism. However, the odds ratios, even for the Dup I + II, are substantially less than the increased risks we have observed for three out of four recurrent deletions on chromosomes 1, 15 and 22.6 Whether the smaller odds ratio we observe for duplications is a feature of the 16p13.1 locus itself, or it is part of a broader rule that recurrent duplications are generally less penetrant than recurrent deletions, remains to be determined. The 16p13.1 duplications we observe are rare, at a rate of about 3 or 4 per 1000 schizophrenia cases, and, estimating from the control population in the present study, about 0.08% in the general population. This makes it difficult to obtain precise measurements of overall risk or risk for individual neuropsychiatric disorders. Analysis of CNV data from sets of cases and controls considerably larger than the sets we report in this paper, which itself to date is one of the largest assembled, will be required. These and many other questions will need to be answered before the exciting findings that arise from CNV analysis can be used in clinical practice for diagnostics, disease classification or genetic testing.
Supplementary Information accompanies the paper on the Molecular Psychiatry website (http://www.nature.com/mp)
We thank the participating subjects and their relatives, and staff at the recruitment centres. We thank David Goldstein for permission to use the genotype data from the Scottish samples typed at Duke University. This work was sponsored by EU grant LSHM-CT-2006–037761 (Project SGENE). Genotyping of the Dutch samples was sponsored by NIMH funding, R01 MH078075 (to RAO).
GROUP investigators include: René S Kahn and Wiepke Cahn, Department of Psychiatry, Rudolf Magnus Institute of Neuroscience, University Medical Centre Utrecht, Utrecht, The Netherlands; Don Linszen and Lieuwe de Haan, Department of Psychiatry, Academic Medical Centre University of Amsterdam, Amsterdam, The Netherlands; Jim van Os, Lydia Krabbendam and Inez Myin-Germeys, Maastricht University Medical Centre, EURON South Limburg Mental Health Research and Teaching Network, Maastricht, The Netherlands; Durk Wiersma and Richard Bruggeman, Department of Psychiatry, University Medical Centre Groningen, University of Groningen, The Netherlands.
Conflict of interest: The authors declare no conflict of interest.