Association of Neuregulin 1 with schizophrenia and bipolar disorder in a second cohort from the Scottish population

Abstract

Neuregulin 1 (NRG1) is a strong candidate for involvement in the aetiology of schizophrenia. A haplotype, initially identified as showing association in the Icelandic and Scottish populations, has shown a consistent effect size in multiple European populations. Additionally, NRG1 has been implicated in susceptibility to bipolar disorder. In this first study to select markers systematically on the basis of linkage disequilibrium across the entire NRG1 gene, we used haplotype-tagging single-nucleotide polymorphisms to identify single markers and haplotypes associated with schizophrenia and bipolar disorder in an independently ascertained Scottish population. Haplotypes in two regions met an experiment-wide significance threshold of P=0.0016 (Nyholt's SpD) and were permuted to correct for multiple testing. Region A overlaps with the Icelandic haplotype and shows nominal association with schizophrenia (P=0.00032), bipolar disorder (P=0.0011), and the combined case group (P=0.0017). This region includes the 5′ exon of the NRG1 GGF2 isoform and overlaps the expressed sequence tag (EST) cluster Hs.97362. However, no haplotype in Region A remains significant after permutation analysis (P>0.05). Region B contains a haplotype associated with both schizophrenia (P=0.00014), and the combined case group (P=0.000062), although it does not meet Nyholt's threshold in bipolar disorder alone (P=0.0022). This haplotype remained significant after permutation analysis in both the schizophrenia and combined case groups (P=0.024 and P=0.016, respectively). It spans a 136 kb region that includes the coding sequence of the sensory and motor neuron derived factor (SMDF) isoform and 3′ exons of all other known NRG1 isoforms. Our study identifies a new of NRG1 region involved in schizophrenia and bipolar disorder in the Scottish population.

Introduction

Neuregulin (NRG1) on chromosome 8p12 is one of four neuregulin genes (NRG1, NRG2, NRG3 and NRG4). NRG1 has been implicated in the aetiology of breast cancer, heart disease, multiple sclerosis and schizophrenia.1 It is a large multiexon gene spanning 1.1 Mb with multiple transcription start sites and alternative splice sites that give rise to a complex series of isoforms. The isoforms have been classified into three types (I, II, III) on the basis on their N-terminal sequences and functional domains. However, recently, Steinthorsdottir et al.2 reported the identification of additional transcription initiation sites and internal exons and have proposed the addition of isoform types IV–VI, further increasing the isoform complexity. NRG1 is a ligand for ErbB tyrosine kinase receptors and is known to mediate cell–cell interactions in the nervous system, heart, breast, muscle and other organs (reviewed by Falls1, 3). NRG1-mediated signalling results in stimulation or inhibition of cell proliferation, apoptosis, migration, differentiation and adhesion (reviewed by Yarden and Sliwkowski4). The differences in structure and expression patterns of the isoforms are thought to modify the protein to suit different signalling strategies and requirements (reviewed by Falls1, 3).

Mice heterozygous for two different Nrg1 mutations, which are in domains that are present in all isoforms, display phenotypes thought to be related to schizophrenia. These include hyperactivity in behavioural tests, impaired prepulse inhibition (PPI) and reduced numbers of functional NMDA receptors in the prefrontal cortex.5, 6, 7 Hyperactivity, but not the PPI defect, is reversed by treatment with the antipsychotic drug clozapine.7 Mice heterozygous for a mutation in the Ig-like domain (type I and II) of Nrg1 show clozapine suppression of open-field and running wheel activity and impaired latent inhibition, but no hyperactivity.8 Recent experiments involving rat NRG type 1β suggest that it is involved in the reduction of NMDA receptor activity, through ErbB-mediated internalization of NMDA receptors, which is likely to affect synaptic plasticity, long-term potentiation and thus memory.9, 10 Rat NRG1 has also been shown to downregulate GABA-A receptors in the hippocampus.11 Thus, while the precise functions of the neuregulin isoforms have yet to be resolved, their involvement in dopaminergic neurotransmission, synaptic plasticity and neuronal migration, make NRG1 a promising functional candidate for schizophrenia.

Multiple genome linkage scans and meta-analyses have suggested chromosome 8p as the location of a susceptibility gene for schizophrenia in several different populations.7, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22 Fine mapping and haplotype analysis of the linkage peak in an Icelandic sample identified a haplotype, shared by seven out of the thirty-three families tested, implicating an 600 kb region.7 A core haplotype from this region (HapICE) spanning 290 kb and consisting of five single-nucleotide polymorphisms (SNPs) and two microsatellite markers was found to be over-represented in affected individuals (P=0.000087, one-tailed; Figure 1). This core haplotype spans the NRG1 promoter region, the first 5′ exon, which encodes the N-terminus of GGF2, and part of the first intron. A subsequent case–control study of schizophrenia patients and controls from North East Scotland gave one-tailed P-values of P=0.00031 for the same seven-marker HapICE haplotype, and P=0.000032 for the haplotype containing the five SNPs alone (Figure 1).23 Three further groups have studied UK/Irish and Portuguese samples and found no significant differences in the HapICE frequencies,24, 25, 26 although the HapICE haplotype was more frequent in the case versus the control group in all three studies. A fourth family-based study found no significant differences in either the individual alleles of the HapICE markers or the haplotypes thereof investigated, but did not formally test the HapICE haplotype.27 Petryshen et al. (2005)26 undertook a meta-analysis combining the Portuguese results with those of previous European NRG1 association studies.7, 23, 25, 28 The estimated combined odds ratios (OR) for the seven-marker HapICE was 1.5 (95% CI: 1.44–1.56).26 This result remained significant when the original Icelandic sample was excluded, OR=1.37 (95% CI: 1.31–1.44),26 suggesting a small, but consistent, effect size of this haplotype in schizophrenia. However, as noted by Stefansson et al.,7 this magnitude of effect is unlikely to be sufficient on its own to account for the linkage signals in this region. Corvin et al.25 reported that an alternate two-marker microsatellite haplotype is in excess both in their Irish sample and the original Scottish sample used by Stefansson (Figure 1). This HapBIRE includes one of the HapICE microsatellites and is positioned close to Hs.97362, an expressed sequence tag (EST) cluster of unknown function, within intron 1 of NRG1.25 Petryshen et al.26 also reported weak association in their Portuguese sample of two haplotypes at the exon-dense 3′ region of the gene, as well as the association of an alternative 5′ haplotype containing two HapICE SNPs (Figure 1). These reports suggest that there may be additional haplotypes associated with schizophrenia in NRG1 in European populations.

Figure 1
figure1

Most significant individual haplotype P-values from previous association studies of NRG1 and schizophrenia or bipolar disorder. The P-values of the most significant haplotypes are given for each study reporting significant association between NRG1 and schizophrenia7, 23, 24, 25, 26, 29, 30, 31, 32, 33 or bipolar disorder.24 Williams et al.28 is not included as this sample was not significant when reanalysed with an increased control sample.24 Yang et al.30 and Hall et al.29 reported P-values as less than a defined threshold, in these cases the threshold value is used.

Genetic evidence for the association of NRG1 with schizophrenia is not restricted to European populations. A South African Caucasian Afrikaner study29 and five studies of the Han Chinese reported significant association of haplotypes with schizophrenia, both for HapICE markers and for other markers located in the region of HapICE and elsewhere (Figure 1).30, 31, 32, 33 However, one report on the Han Chinese, one from the USA and a study from Japan failed to detect association.29, 34, 35

To date only one study has examined the effect of NRG1 in bipolar disorder. Green et al.24 found association in a Caucasian UK sample of bipolar disorder with a haplotype bearing the same alleles as seen on the HapICE haplotype (Figure 1). The odds ratio and 95% confidence limit was similar to that seen for schizophrenia, OR=1.37, 95% CI: 1.03–1.80.

We have performed case–control association studies of both schizophrenia and bipolar disorder testing the whole of NRG1, including the linkage disequilibrium (LD) blocks up- and downstream of the coding region. Data from the International HapMap Project (http://www.hapmap.org/) was downloaded into Haploview and haplotype tagging SNPs (htSNPs) were selected to represent the genetic variation across the gene. Our aim was to carry out a comprehensive survey of NRG1. We considered this important given the increasing evidence of alternate NRG1 haplotypes predisposing to schizophrenia and the recent report of association with bipolar disorder. In order to compare our results in the 5′ region of the gene with those of others, we subsequently genotyped four of the HapICE SNPs and determined their relationship to the haplotypes associated in our sample.

Materials and methods

The study was approved by the Scottish Multicentre Research Ethics Committee. All patients and controls gave informed written consent.

Case–control sample

Subjects were inpatients or outpatients of hospitals in South East or South Central Scotland. Subjects were interviewed by an experienced psychiatrist and a venous blood sample was given for DNA extraction. Diagnoses were made according to Diagnostic and Statistical Manual (DSM)-IV criteria36 based on case note review and personal interview using The Schedule for Affective Disorders and Schizophrenia – lifetime version (SADS-L).37 Final diagnoses were reached by consensus between two experienced psychiatrists (DB and WM).

Control subjects were drawn from the same population in South East and South Central Scotland. The majority was recruited from donors for the Scottish National Blood Transfusion service (395). Although the blood donors were not screened by interview for personal or family history of psychiatric illness, donors are accepted only if they are taking no medication and are not currently unwell with physical or psychiatric illness. The remaining controls (83) were recruited from the local population or from hospital staff. These controls were briefly screened by interview to exclude anyone currently on psychotropic medication or with a history of treatment for psychiatric illness.

Genomic DNA was extracted from venous blood samples using standard protocols.

SNP selection and genotyping

SNPs were selected to tag haplotypes of greater than 10% frequency in the HapMap data release #7 using the data from the CEPH trios of Utah residents of northern and western European ancestry (CEU) for the region chr8: 31 520 000–32 854 000 (NCBI Build 34). Data for SNPs with minor allele frequencies greater than 10% were downloaded into Haploview version 2.05,38 and LD blocks identified using Solid Spine of LD (D′>0.8). Blocks were joined where the multiallelic D′ (MAD38, 39) between blocks was >0.85. Haploview was then used to select htSNPs with a frequency of greater than 10% within each block, selecting at least one SNP per block. SNPs that fell outside or between blocks of LD, as defined by the above parameters, were also genotyped.

Tagging SNPs were either genotyped by Illumina, San Diego, using Bead Array technology, or by the Wellcome Trust CRF Genetics Core Facility using TaqMan assays on an ABI7900. The HapICE SNPs, SNP8NRG221132; SNP8NRG221533; SNP8NRG241930; SNP8NRG243177 (rs6994992), were genotyped by the Wellcome Trust CRF Genetics Core Facility using TaqMan assay-by-design assays. For information on HapICE SNPs and microsatellite see deCODE Genetics, http://www.decode.com/nrg1/markers.

Statistical analysis

All markers were tested for Hardy–Weinberg equilibrium (HWE) using a χ2 goodness-of-fit test.

Case–control association analyses were performed using the Unphased suite of programmes.40 Haplotypes were estimated in Unphased using an EM algorithm. Single-marker and sliding-window analyses of two- to four- SNP haplotypes were performed using the program Cocaphase (Unphased suite of programs), clumping together rare haplotypes of a frequency less than 1% in both the cases and controls. P-values were calculated for the differences in global haplotype frequencies (pg) and individual haplotypes frequencies (pi) for SCZ and BP both separately and combined. In addition, a seven SNP haplotype window, consisting of the four HapICE SNPs and the three most significant SNPs from Region A in this study, was also analysed using Cocaphase.

Multiple-testing correction

The nominal P-values given are two-tailed and uncorrected for multiple testing. An experiment-wide significance threshold was calculated using the method by Nyholt (http://genepi.qimr.edu.au/general/daleN/SNPSpD).41 P-values meeting this threshold were subsequently subjected to permutation analysis (1000 permutations) using Cocaphase. It should be noted that while this multiple-testing correction adjusts for the number of markers and sliding windows tested, it does so only for the size of the significant sliding window and not for other window sizes. All sliding windows covering a given region are likely to be highly correlated with each other and are therefore unlikely to greatly inflate the number of independent tests undertaken in this study.

Results

Thirty-six htSNPs were selected for case–control association studies of NRG1 from the International HapMap project CEU data. These htSNPs discriminated between haplotypes of frequency greater than 10% within haplotype blocks in a 1.3 Mb region spanning the NRG1 gene. This resulted in an average density of one SNP per 37 kb across 14 LD blocks. The SNPs were genotyped in 455 control (237 male, 218 female), 386 SCZ (276 male, 110 female) and 368 BP (160 male, 208 female) subjects. All markers were in Hardy–Weinberg equilibrium in our sample (P>0.01).

We selected htSNPs on a block-by-block basis to capture efficiently the genetic variation across the entire gene, but not to impose a definite block structure on the region and the subsequent analysis. Haplotype blocks are, to a certain extent, dependent on the algorithms used to construct them as well as to SNP density and population structure (see for example Nothnagel and Rohde42). Additionally, there is often substantial multiallelic D′ (MAD) between adjacent haplotype blocks, indicating that SNPs in neighbouring blocks may provide additional valuable information.43 For these reasons, and because of the wider range of haplotype frequencies and thus causal variant frequencies covered, rather than apply a strict block-by-block analysis, we followed a single-marker and sliding-window approach in order to increase our power to detect the causal variant (See Wray44). This ensured that our analysis was robust to block definitions that did not represent the true biological boundaries.

Sliding-window analysis identified clusters of single markers and individual haplotypes that show association in both SCZ and BP subjects (Table 1, Figure 2). The most significant haplotypes in each cluster (underlined in Table 1) define two regions, Region A that includes LD blocks 1–4 and Region B that includes LD blocks 9 and 10 (Figures 2 and 3). P-values are given for both global tests (pg) and individual tests (pi) of significance (Supplementary Information, Table 1).

Table 1 Individual association results for single-, two-, three- and four-marker haplotypes
Figure 2
figure2

LD map of the NRG1 region. (a) Schematic representation of chromosome 8 31 520 000–32 854 000 (NCBI build 34), with the position of the tagging SNPs (red), HapICE SNPs (pale blue), NRG1 reference sequences (dark blue) and ESTs that have been spliced (black) taken from the UCSC Human Genome Project genome browser (http://genome.ucsc.edu/). The lower panel shows the position of the 14 LD blocks (black triangles) relative to these reference sequences. The LD map is generated from 409 SNPs of >10% minor allele frequency genotyped by the International HapMap Project on 30 CEPH trios from Utah with western or northern European ancestry (CEU, Data Release #16c, June 2005). LD was calculated by Haploview version 2.5 using solid spine of LD>0.8, blocks were joined where the multiallelic D′>0.8 and shown using the standard Haploview color scheme in greyscale (see HTML for color version and description). (b) Table showing the htSNPs genotyped in each block. The blocks that overlap the most significant haplotypes associated with either schizophrenia or bipolar disorder in this study are indicated (Y). Blocks 11 and 12 were not tested (NT) as they did not appear in release #7 data used in SNP selection. The numbers of previous studies also reporting significant haplotypes in each block are also shown.

Figure 3
figure3

Individual haplotypes P<0.01 for schizophrenia, bipolar disorder and the combined group. The P-values are shown for the most significant individual haplotypes in each region in the groups: schizophrenia (squares), bipolar disorder (triangles) and the combined cases (circles). Regions A and B are defined by the boundaries of LD blocks containing the most significant individual haplotype in each region.

Our study presents a considerable number of tests and indeed our samples are used in other association studies. However, a Bonferroni correction for all tests performed would be overly conservative as many of the tests are correlated. We therefore used the method of Nyholt (http://genepi.qimr.edu.au/general/daleN/SNPSpD), which accommodates LD between the SNPs, to determine the effective number of independent SNPs and a preliminary significance threshold.41 Using this method the effective number of independent SNPs was 32 and an experiment-wide significance threshold of P=0.0016 is needed if the type I error rate is to be kept at 5%. We have used this threshold to determine which P-values to subject to permutation analysis using Cocaphase (corrected P-values). However, it should also be noted that the previous reports of genetic association with both SCZ and BP, the biological studies of mutations in Nrg1, and the function of the gene increase the prior probability of NRG1 being involved in susceptibility to both SCZ and BP. We have therefore also reported P-values that are only nominally significant (P<0.05) where they may be of interest, for example, with haplotypes involving the HapICE markers.

Region A

Region A (Table 1, Figure 3) contains a 3-SNP haplotype that is nominally significant with SCZ alone (SNPs 9–11, haplotype t-t-t, pg=0.0043, pi=0.00032), although it does not withstand permutation analysis (corrected P=0.057) This haplotype is not significant in the BP sample (pg=0.40, pi=0.084) and did not meet Nyholt's threshold in the combined sample of SCZ and BP (pg=0.0080, pi=0.0017).

The t-t-t haplotype has an estimated haplotype frequency of 1.3% in SCZ, but is unlikely to occur in BP or controls individuals (Supplementary Information, Table 1). This haplotype spans LD blocks 2 and 3 (Figure 2) and overlaps with the blocks containing the HapICE seven-marker haplotype. We therefore investigated the relationship between this three SNP haplotype and the HapICE SNPs. Genotyping of four of the five HapICE SNPs (SNP8NRG221132, SNP8NRG221533, SNP8NRG241930, SNP8NRG243177) showed no significant difference between SCZ and controls for the HapICE haplotype (g-c-g-t, 34.5% cases, 34.8% controls, pg=0.79, pi=0.92). However, analysis revealed that 77% of the SCZ t-t-t haplotype carriers coinherited the g-c-g-t alleles of the HapICE haplotype, with the remaining 23% coinheriting an alternate g-t-g-g haplotype at these SNPs. Association analysis of the seven-SNP haplotype g-c-g-t-t-t-t showed nominally significant association with SCZ (1.1% cases, 0.0% controls, pg=0.037, pi=0.0025), while the alternate seven-SNP haplotype containing g-t-g-g-t-t-t was not significant (pi=0.059). The results from the seven-SNP haplotypes may be inflated due to both the low estimated frequency of the haplotype in the control group and the level of accuracy of EM estimation of a seven-SNP haplotype. These results must therefore be viewed with caution.

A second haplotype in Region A is nominally associated with BP alone and spans LD blocks 3 and 4 (Table 1, SNPs 10–12, pg=0.026, pi=0.0011). Neither of these blocks were associated with either BP or SCZ in previous studies (Figure 2). This haplotype shares two out of three SNPs with the nominally associated schizophrenia haplotype described above, but the only allele identical between the two groups is that of SNP 10. This SNP is nominally significant in all three groups (SCZ P=0.029; BP P=0.029; combined case group P=0.011, 10.4% cases, 7.4% controls). SNP 10 is in LD block 3 adjacent to those associated in previous studies (Figure 2), potentially extending the region containing the putative functional variant(s) further into intron 1.

In the BP sample, analysis of the sliding windows that included the HapICE SNPs led to the identification of a four-SNP haplotype that was nominally significant at the individual level (SNP8NRG221533-rs4298458-SNP8NRG241930-SNP8NRG243177, pg=0.011, pi=0.0045, 3.8% cases 1.2% controls). This is the only nominally significant haplotype identified in the sliding-window analysis that contains HapICE SNPs (Table 1). Analysis of the four HapICE SNPs alone also identified one nominally significant haplotype g-t-g-t in the BP sample (pi=0.027), but it did not contain the same alleles as the HapICE core haplotype (g-c-g-t). This haplotype is within LD block 1 (Figure 2).

Region A overlaps with the HapICE haplotype, and LD blocks associated with schizophrenia in other populations (Blocks 1 and 2, Figure 3).25, 26, 31, 32, 33

Region B

In Region B, the most significant haplotype in SCZ (P=0.00014, corrected P=0.024) was also nominally significant in the BP sample (P=0.0022), with the most significant P-value resulting from the analysis of the combined group (SNPs 34–36, pg=0.0034, pi=0.000062, corrected P=0.016; OR 1.54, 95%CI 1.27–1.86; Table 1, Figure 3). This haplotype is at the 3′ end of the gene, and is present at high frequency in both the cases and the controls (32.6% cases, 24.0% controls). This 3-SNP haplotype contains SNPs from LD blocks 9 and 10 including two out of the three markers are nominally significant in the combined sample (SNP 34 P=0.011, SNP 35 P=0.0058). The only additional nominally significant marker in this sample, SNP 33, is also in LD block 9 (SNP 33 P=0.0060, Table 1). LD blocks 9 and 10 span the sensory and motor neuron derived factor, SMDF, isoform and 3′ regions of all other NRG1 isoforms (Figure 2). This was the only haplotype to withstand permutation analysis in this study, identifying a new region of association with both schizophrenia and the combined case group in the Scottish population (corrected P=0.024 and P=0.016, respectively).

Discussion

Here, we report significant association between schizophrenia and bipolar disorder and the NRG1 gene. NRG1 is one of the best replicated gene associations in psychiatric illness with more than 10 studies having described its involvement in schizophrenia. One previous study has also reported an association between NRG1 and bipolar disorder.24 However, to date only three studies have examined the 3′ region of the gene,26, 30, 32 and no other study has employed a systematic LD-based marker selection procedure across the entire genic region. Therefore, our study of NRG1 is the first to meet the criterion for a systematic study set by David Goldstein.45 Petryshen et al.26 did select SNPs to tag common haplotypes, for a region spanning exon 1 and introns 1 and 2, but not for the remainder of the gene. Using the publicly available information from the International HapMap Project release #7, we selected 36 htSNPs38 that tagged haplotypes with frequencies greater than 10% within the haplotype blocks covering the NRG1 gene and additional blocks that cover 35 kb 5′ and 170 kb 3′ of the known gene (NCBI build 34). We performed case–control association studies of schizophrenia and bipolar disorder and analysed the results as single markers and sliding windows of two- to four-marker haplotypes. This analysis method is robust to misdefinition of blocks, reducing errors due to the relatively small sample size tested in HapMap and to misspecification of block boundaries.

International HapMap Project release #16c includes two additional LD blocks not seen in the data set used to select SNPs. These two blocks lie in the region between two of the original blocks and were not directly examined in this study, however, both these blocks are in strong multiallelic D′ with blocks in release #7 (MAD>0.78) indicating that we should have good power to detect any effects in these regions.

Two regions showed nominally significant association with both bipolar and schizophrenia at both the single-marker and haplotype level. Region A was most significant in our schizophrenic sample (P=0.00032), and overlaps with the HapICE region and those of other studies which used SNPs from this haplotype (Figure 4). This region spans the HapIRE region defined by Corvin et al.,25 which is located slightly 3′ of the HapICE haplotype overlapping with Unigene cluster Hs.97362 (Figure 2a), and is consistent with the results of Zhao et al.33 Analysis of four out of five HapICE SNPs suggests that the majority of individuals carrying the significant haplotype (t-t-t) in our schizophrenic population coinherit the HapICE haplotype. However, unlike the independent Scottish cohort studied by Steffansson et al.,23 no association was seen with any single HapICE SNP or with haplotypes of these SNPs in our schizophrenic or combined samples (P>0.01). Furthermore, the haplotype nominally associated with schizophrenia did not withstand permutation analysis (corrected P=0.057). Therefore, our results do not replicate the original findings of Steffanson et al. at the marker/haplotype level. A second haplotype in Region A was nominally associated with bipolar disorder only (P=0.0011). This haplotype overlaps with our schizophrenia-associated haplotype. When schizophrenia and bipolar disorder are combined the haplotype associated with schizophrenia in Region A is only nominally significant (P=0.0017), reflecting the difference in the individual haplotypes nominally associated in schizophrenia and bipolar disorder.

Figure 4
figure4

Individual haplotypes P<0.01 from this and previous studies of schizophrenia, bipolar disorder. The P-values are shown for the most significant individual haplotypes in this study, and previous studies. Regions A and B are defined by the boundaries of LD blocks containing the most significant individual haplotype in each region.

A single haplotype containing the HapICE SNPs was nominally significant in the bipolar sample (SNPs 3–6, 3.8% cases, 1.2% controls, pi=0.0045). However, the alleles of the HapICE SNPs were not consistent with those on the original Icelandic core haplotype, and this result did not meet the Nyholt's corrected significance threshold. Association between markers of HapICE and bipolar disorder has been reported previously in a sample from the UK population.24 Green et al. genotyped a reduced core Icelandic haplotype using SNP8NRG221533 and the two microsatellites 478B14-848 and 420M9-1395. They reported pg=0.003 for the three-marker haplotypes, with pi=0.04 for the core HapICE haplotype. Several less frequent haplotypes (frequencies greater than 1%) also contributed to the global P-value; however, we do not know if our haplotype is one of these.

In Region B, the most significant haplotype in the schizophrenic sample was also nominally significant in the bipolar disorder sample (P=0.0022), resulting in the most significant P-value being that of a three-SNP haplotype in the combined sample (pi=0.000062, corrected P=0.016). This supports the involvement of this region in both disorders. This haplotype extends across two LD blocks that total 136 kb, spanning the SMDF isoform and 3′ exons of all other known isoforms (chr8:32 546 019–32 682 055 NCBI build 34). Three other studies have examined this region of the NRG1 gene (Figure 4).26, 30, 32 Yang et al.30 reported association in the Han Chinese population with three SNPs showing association in transmission/disequilibrium test (TDT) analysis of schizophrenia after Bonferroni correction (P=0.0078, P=0.00093, P=0.013). The global haplotype analysis of the three SNPs identified highly significant global haplotype association (pg<0.000001) in their study. However, when one of the three, a nonsynonymous SNP, was examined in an independent study of Han Chinese from Taiwan, a trend towards significant overtransmission of the opposite allele was observed (P=0.052).34 A second study in the Han Chinese population32 detected a four-marker microsatellite haplotype in the same 3′ region that was significant when tested in patients (23.8%) against non-transmitted parental chromosome controls (13.7%, P=0.000042), but not in the case-control comparison with unrelated controls (P=0.29). The only European study to examine this region26 reported nominal association of two nonoverlapping haplotypes in case–control association studies of this region, Hap7 P=0.044 and Hap9 P=0.031 (Figure 4). Interestingly, our most strongly associated single-marker, rs6988339 (combined case group P=0.0058), was also individually associated in this study (P=0.030). The association of this second region may help to explain the disparity between the strength of the linkage peak identified in previous studies of schizophrenia and the estimated effect size of the Icelandic haplotype, suggesting that multiple haplotypes in NRG1 affect susceptibility to schizophrenia and also to bipolar disorder.

Two studies have examined the expression of NRG isoforms in cells from schizophrenic individuals. Hashimoto et al.46 examined mRNA levels of type I, II and III isoforms by RT-PCR in the dorsolateral prefrontal cortex (DLPFC) of 20 schizophrenic post-mortem brains and 19 matched controls. They reported a small but significant increase in the expression of type I isoforms in schizophrenia, however, this was also correlated with antipsychotic medication dosage, and a significant decrease in typeII/I and type II/III ratios, which is consistent with the relative underexpression of type II in the DLPFC of schizophrenic patients. There was no evidence of change in the expression levels of type III. No correlation with genotype at SNP8NRG221533 or SNP8NRG243177 was detected. More recently Petryshen et al.26 reported 3.8-fold increased expression of SMDF, a type III isoform, in the peripheral leucocytes of schizophrenics relative to their unaffected siblings (P=0.013). None of the other isoforms studied in this report showed significant variation in gene expression between the two groups, although trends towards an effect of diagnosis on type 1 isoform HRG-beta2 (P=0.093) and overall NRG1 expression were noted (P=0.13). No association was detected between SMDF expression level and the genotypes of SNPs that uniquely mark protective haplotypes, including rs6988339, the most significant single-marker in our study. However, given the small sample size in the above association studies, the observed negative results are not unexpected. As our 3′ haplotype spans the coding region of the SMDF isoform, it is possible that the use of the common haplotype defined in our study would yield more significant results.

Genetic linkage and association studies have identified some chromosomal regions and genes that show association in both schizophrenia and bipolar disorder and others that are apparently specific to one but not both of these disorders. This evidence, along with the results of family and twin studies, suggests a partial overlap in the aetiology of these diseases (reviewed by Berrettini47). Splitting samples by diagnosis does raise multiple-testing concerns, but a prior study has suggested that there may be shared genetic predisposition across diagnostic boundaries in the NRG1 region.24 Interestingly, this study supports the candidacy of NRG1 as a possible gene for not only schizophrenia but also for bipolar disorder. We have shown that identical haplotypes are nominally associated with both diseases in a region at the 3′ end of the gene. However, it is noteworthy that in the 5′ region of the gene the haplotypes nominally associated with the two disorders are distinct. It is therefore interesting to note that Green et al.24 showed that subgroups of the two disorders, bipolar disorder with predominantly mood-incongruent psychotic features and schizophrenic cases who had experienced mania, showed a stronger association in their sample. Similarly, Bakker et al.48 reported that, while schizophrenia was not associated with the two HapICE markers examined, a subgroup of nondeficit schizophrenia was associated, and Kampman et al.49 reported association of the single HapICE SNP tested in only schizophrenic individuals who do not respond to conventional antipsychotics (P=0.013). These results may indicate that subgroups of schizophrenia and bipolar disorder are associated with NRG1, and further that some regions of the gene may confer risk to both bipolar and schizophrenia and others to only one of the diagnoses.

We have recently published the association of haplotypes in DISC1 with bipolar disorder and schizophrenia using the same sample set.50 It is interesting that two of the most promising candidate genes for susceptibility to psychiatric illness are associated with these disorders within the same sample set. In the future, we plan to test for interactions between these and other genes in an independent replication set.

The functional variant(s) in NRG1 remain(s) elusive, but our results suggest that it may be possible to define subgroups of both diseases where NRG1 is a risk factor and thus amenable to an NRG1 pathway specific intervention.

References

  1. 1

    Falls DL . Neuregulins: functions, forms, and signaling strategies. Exp Cell Res 2003; 284: 14–30.

  2. 2

    Steinthorsdottir V, Stefansson H, Ghosh S, Birgisdottir B, Bjornsdottir S, Fasquel AC et al. Multiple novel transcription initiation sites for NRG1. Gene 2004; 342: 97–105.

  3. 3

    Falls DL . Neuregulins and the neuromuscular system: 10 years of answers and questions. J Neurocytol 2003; 32: 619–647.

  4. 4

    Yarden Y, Sliwkowski MX . Untangling the ErbB signalling network. Nat Rev Mol Cell Biol 2001; 2: 127–137.

  5. 5

    Gassmann M, Casagranda F, Orioli D, Simon H, Lai C, Klein R et al. Aberrant neural and cardiac development in mice lacking the ErbB4 neuregulin receptor. Nature 1995; 378: 390–394.

  6. 6

    Gerlai R, Pisacane P, Erickson S . Heregulin, but not ErbB2 or ErbB3, heterozygous mutant mice exhibit hyperactivity in multiple behavioral tasks. Behav Brain Res 2000; 109: 219–227.

  7. 7

    Stefansson H, Sigurdsson E, Steinthorsdottir V, Bjornsdottir S, Sigmundsson T, Ghosh S et al. Neuregulin 1 and susceptibility to schizophrenia. Am J Hum Genet 2002; 71: 877–892.

  8. 8

    Rimer M, Barrett DW, Maldonado MA, Vock VM, Gonzalez-Lima F . Neuregulin-1 immunoglobulin-like domain mutant mice: clozapine sensitivity and impaired latent inhibition. Neuroreport 2005; 16: 271–275.

  9. 9

    Gu Z, Jiang Q, Fu AK, Ip NY, Yan Z . Regulation of NMDA receptors by neuregulin signaling in prefrontal cortex. J Neurosci 2005; 25: 4974–4984.

  10. 10

    Schillo S, Pejovic V, Hunzinger C, Hansen T, Poznanovic S, Kriegsmann J et al. Integrative proteomics: functional and molecular characterization of a particular glutamate-related neuregulin isoform. J Proteome Res 2005; 4: 900–908.

  11. 11

    Okada M, Corfas G . Neuregulin1 downregulates postsynaptic GABAA receptors at the hippocampal inhibitory synapse. Hippocampus 2004; 14: 337–344.

  12. 12

    Pulver AE, Lasseter VK, Kasch L, Wolyniec P, Nestadt G, Blouin JL et al. Schizophrenia: a genome scan targets chromosomes 3p and 8p as potential sites of susceptibility genes. Am J Med Genet 1995; 60: 252–260.

  13. 13

    Kendler KS, MacLean CJ, O'Neill FA, Burke J, Murphy B, Duke F et al. Evidence for a schizophrenia vulnerability locus on chromosome 8p in the Irish Study of High-Density Schizophrenia Families. Am J Psychiatry 1996; 153: 1534–1540.

  14. 14

    Levinson DF, Wildenauer DB, Schwab SG, Albus M, Hallmayer J, Lerer B et al. Additional support for schizophrenia linkage on chromosomes 6 and 8: a multicenter study. Schizophrenia Linkage Collaborative Group for Chromosomes 3, 6 and 8. Am J Med Genet 1996; 67: 580–594.

  15. 15

    Blouin JL, Dombroski BA, Nath SK, Lasseter VK, Wolyniec PS, Nestadt G et al. Schizophrenia susceptibility loci on chromosomes 13q32 and 8p21. Nat Genet 1998; 20: 70–73.

  16. 16

    Kaufmann CA, Suarez B, Malaspina D, Pepple J, Svrakic D, Markel PD et al. NIMH genetics initiative millennium schizophrenia consortium: linkage analysis of African-American pedigrees. Am J Med Genet 1998; 81: 282–289.

  17. 17

    Shaw SH, Kelly M, Smith AB, Shields G, Hopkins PJ, Loftus J et al. A genome-wide search for schizophrenia susceptibility genes. Am J Med Genet 1998; 81: 364–376.

  18. 18

    Brzustowicz LM, Honer WG, Chow EW, Little D, Hogan J, Hodgkinson K et al. Linkage of familial schizophrenia to chromosome 13q32. Am J Hum Genet 1999; 65: 1096–1103.

  19. 19

    Gurling HM, Kalsi G, Brynjolfson J, Sigmundsson T, Sherrington R, Mankoo BS et al. Genomewide genetic linkage analysis confirms the presence of susceptibility loci for schizophrenia, on chromosomes 1q32.2, 5q33.2, and 8p21–22 and provides support for linkage to schizophrenia, on chromosomes 11q23.3–24 and 20q12.1–11.23. Am J Hum Genet 2001; 68: 661–673.

  20. 20

    Lewis CM, Levinson DF, Wise LH, DeLisi LE, Straub RE, Hovatta I et al. Genome scan meta-analysis of schizophrenia and bipolar disorder, part II: Schizophrenia. Am J Hum Genet 2003; 73: 34–48.

  21. 21

    Liu CM, Hwu HG, Fann CS, Lin CY, Liu YL, Ou-Yang WC et al. Linkage evidence of schizophrenia to loci near neuregulin 1 gene on chromosome 8p21 in Taiwanese families. Am J Med Genet B Neuropsychiatr Genet 2005; 134: 79–83.

  22. 22

    Takahashi S, Faraone SV, Lasky-Su J, Tsuang MT . Genome-wide scan of homogeneous subtypes of NIMH genetics initiative schizophrenia families. Psychiatry Res 2005; 133: 111–122.

  23. 23

    Stefansson H, Sarginson J, Kong A, Yates P, Steinthorsdottir V, Gudfinnsson E et al. Association of neuregulin 1 with schizophrenia confirmed in a Scottish population. Am J Hum Genet 2003; 72: 83–87.

  24. 24

    Green EK, Raybould R, Macgregor S, Gordon-Smith K, Heron J, Hyde S et al. Operation of the schizophrenia susceptibility gene, neuregulin 1, across traditional diagnostic boundaries to increase risk for bipolar disorder. Arch Gen Psychiatry 2005; 62: 642–648.

  25. 25

    Corvin AP, Morris DW, McGhee K, Schwaiger S, Scully P, Quinn J et al. Confirmation and refinement of an ‘at-risk’ haplotype for schizophrenia suggests the EST cluster, Hs.97362, as a potential susceptibility gene at the Neuregulin-1 locus. Mol Psychiatry 2004; 9: 208–213.

  26. 26

    Petryshen TL, Middleton FA, Kirby A, Aldinger KA, Purcell S, Tahl AR et al. Support for involvement of neuregulin 1 in schizophrenia pathophysiology. Mol Psychiatry 2005; 10: 366–374, 328.

  27. 27

    Thiselton DL, Webb BT, Neale BM, Ribble RC, O'Neill FA, Walsh D et al. No evidence for linkage or association of neuregulin-1 (NRG1) with disease in the Irish study of high-density schizophrenia families (ISHDSF). Mol Psychiatry 2004; 9: 777–783; image 729.

  28. 28

    Williams NM, Preece A, Spurlock G, Norton N, Williams HJ, Zammit S et al. Support for genetic variation in neuregulin 1 and susceptibility to schizophrenia. Mol Psychiatry 2003; 8: 485–487.

  29. 29

    Hall D, Gogos JA, Karayiorgou M . The contribution of three strong candidate schizophrenia susceptibility genes in demographically distinct populations. Genes Brain Behav 2004; 3: 240–248.

  30. 30

    Yang JZ, Si TM, Ruan Y, Ling YS, Han YH, Wang XL et al. Association study of neuregulin 1 gene with schizophrenia. Mol Psychiatry 2003; 8: 706–709.

  31. 31

    Tang JX, Chen WY, He G, Zhou J, Gu NF, Feng GY et al. Polymorphisms within 5′ end of the Neuregulin 1 gene are genetically associated with schizophrenia in the Chinese population. Mol Psychiatry 2004; 9: 11–12.

  32. 32

    Li T, Stefansson H, Gudfinnsson E, Cai G, Liu X, Murray RM et al. Identification of a novel neuregulin 1 at-risk haplotype in Han schizophrenia Chinese patients, but no association with the Icelandic/Scottish risk haplotype. Mol Psychiatry 2004; 9: 698–704.

  33. 33

    Zhao X, Shi Y, Tang J, Tang R, Yu L, Gu N et al. A case control and family based association study of the neuregulin1 gene and schizophrenia. J Med Genet 2004; 41: 31–34.

  34. 34

    Hong CJ, Huo SJ, Liao DL, Lee K, Wu JY, Tsai SJ . Case-control and family-based association studies between the neuregulin 1 (Arg38Gln) polymorphism and schizophrenia. Neurosci Lett 2004; 366: 158–161.

  35. 35

    Iwata N, Suzuki T, Ikeda M, Kitajima T, Yamanouchi Y, Inada T et al. No association with the neuregulin 1 haplotype to Japanese schizophrenia. Mol Psychiatry 2004; 9: 126–127.

  36. 36

    American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders, 4th edn. Washington DC, 1994.

  37. 37

    Endicott J, Spitzer RL . A diagnostic interview: the schedule for affective disorders and schizophrenia. Arch Gen Psychiatry 1978; 35: 837–844.

  38. 38

    Barrett JC, Fry B, Maller J, Daly MJ . Haploview: analysis and visualization of LD and haplotype maps. Bioinformatics 2005; 21: 263–265.

  39. 39

    Hedrick PW . Gametic disequilibrium measures: proceed with caution. Genetics 1987; 117: 331–341.

  40. 40

    Dudbridge F . Pedigree disequilibrium tests for multilocus haplotypes. Genet Epidemiol 2003; 25: 115–121.

  41. 41

    Nyholt DR . A simple correction for multiple testing for single-nucleotide polymorphisms in linkage disequilibrium with each other. Am J Hum Genet 2004; 74: 765–769.

  42. 42

    Nothnagel M, Rohde K . The effect of single-nucleotide polymorphism marker selection on patterns of haplotype blocks and haplotype frequency estimates. Am J Hum Genet 2005; 77: 988–998.

  43. 43

    Halldorsson BV, Istrail S, De La Vega FM . Optimal selection of SNP markers for disease association studies. Hum Hered 2004; 58: 190–202.

  44. 44

    Wray NR . Allele frequencies and the r2 measure of linkage disequilibrium: impact on design and interpretation of association studies. Twin Res Hum Genet 2005; 8: 87–94.

  45. 45

    Goldstein DB . Pharmacogenetics in the laboratory and the clinic. N Engl J Med 2003; 348: 553–556.

  46. 46

    Hashimoto R, Straub RE, Weickert CS, Hyde TM, Kleinman JE, Weinberger DR . Expression analysis of neuregulin-1 in the dorsolateral prefrontal cortex in schizophrenia. Mol Psychiatry 2004; 9: 299–307.

  47. 47

    Berrettini W . Evidence for shared susceptibility in bipolar disorder and schizophrenia. Am J Med Genet C Semin Med Genet 2003; 123: 59–64.

  48. 48

    Bakker SC, Hoogendoorn ML, Selten JP, Verduijn W, Pearson PL, Sinke RJ et al. Neuregulin 1: genetic support for schizophrenia subtypes. Mol Psychiatry 2004; 9: 1061–1063.

  49. 49

    Kampman O, Anttila S, Illi A, Saarela M, Rontu R, Mattila KM et al. Neuregulin genotype and medication response in Finnish patients with schizophrenia. Neuroreport 2004; 15: 2517–2520.

  50. 50

    Thomson PA, Wray NR, Millar JK, Evans KL, Hellard SL, Condie A et al. Association between the TRAX/DISC locus and both bipolar disorder and schizophrenia in the Scottish population. Mol Psychiatry 2005; 10: 657–668, 616.

Download references

Acknowledgements

We are indebted to patients, their families and volunteers for their participation in this study. We would also like to thank Maura Walker and Margaret Van Beck for the collation of patient data. We thank Kirsty Millar, Helen Torrance, Susan Anderson, Alison Condie, John Beekman, Pat Malloy, Alan MacLean, Rosalind Launchbury, Sebastienne Buchanan and the Wellcome Trust CRF Genetics Core for their help in the preparation of the samples. We thank Illumina, San Diego, for SNP genotyping our samples and to Simon T Cooper for help in the preparation of this manuscript. This work was supported from grants from the Chief Scientists Office, the Scottish Executive; the Health Foundation, London; the Medical Research Council, UK and the Wellcome Trust.

Author information

Correspondence to P A Thomson.

Additional information

Supplementary Information accompanies the paper on the Molecular Psychiatry website (http://www.nature.com/mp)

Supplementary information

Supplementary Information (DOC 315 kb)

Rights and permissions

Reprints and Permissions

About this article

Keywords

  • NRG1
  • schizophrenia
  • bipolar disorder
  • association
  • neurodevelopment

Further reading