Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Interactions among genes in the ErbB-Neuregulin signalling network are associated with increased susceptibility to schizophrenia

Interactions among genes in the ErbB-Neuregulin signalling network are associated with increased... Background: Evidence of genetic association between the NRG1 (Neuregulin-1) gene and schizophrenia is now well-documented. Furthermore, several recent reports suggest association between schizophrenia and single- nucleotide polymorphisms (SNPs) in ERBB4, one of the receptors for Neuregulin-1. In this study, we have extended the previously published associations by investigating the involvement of all eight genes from the ERBB and NRG families for association with schizophrenia. Methods: Eight genes from the ERBB and NRG families were tested for association to schizophrenia using a collection of 396 cases and 1,342 blood bank controls ascertained from Aberdeen, UK. A total of 365 SNPs were tested. Association testing of both alleles and genotypes was carried out using the fast Fisher's Exact Test (FET). To understand better the nature of the associations, all pairs of SNPs separated by ≥ 0.5 cM with at least nominal evidence of association (P < 0.10) were tested for evidence of pairwise interaction by logistic regression analysis. Results: 42 out of 365 tested SNPs in the eight genes from the ERBB and NRG gene families were significantly associated with schizophrenia (P < 0.05). Associated SNPs were located in ERBB4 and NRG1, confirming earlier reports. However, novel associations were also seen in NRG2, NRG3 and EGFR. In pairwise interaction tests, clear evidence of gene-gene interaction was detected for NRG1-NRG2, NRG1-NRG3 and EGFR-NRG2, and suggestive evidence was also seen for ERBB4-NRG1, ERBB4-NRG2, ERBB4-NRG3 and ERBB4-ERBB2. Evidence of intragenic interaction was seen for SNPs in ERBB4. Conclusion: These new findings suggest that observed associations between NRG1 and schizophrenia may be mediated through functional interaction not just with ERBB4, but with other members of the NRG and ERBB families. There is evidence that genetic interaction among these loci may increase susceptibility to schizophrenia. Page 1 of 11 (page number not for citation purposes) Behavioral and Brain Functions 2007, 3:31 http://www.behavioralandbrainfunctions.com/content/3/1/31 that had originally been identified as EGFR ligands (Fig. Background Schizophrenia is a complex psychiatric disorder which 1). Furthermore, NRG1 and NRG2 can also bind to ErbB3. affects 0.5–1% of the world wide adult population. A In contrast, none of the neuregulins bind to EGFR which number of putative schizophrenia susceptibility genes instead binds unrelated ligands such as EGF, amphiregu- have been identified recently [1,2]. Genomewide linkage lin and others (see Fig. 1). analysis on large Icelandic pedigrees showed evidence for linkage with an initial LOD score of 3 on chromosome Ligand binding to the extracellular domain of ErbB family 8p13, which led to the identification of the Neuregulin-1 members leads to receptor homo-and heterodimerisation gene (NRG1) as a potential genetic risk factor (O.R. = 2) and activation of various intracellular signalling pathways for schizophrenia [3]. The association was confirmed in a such as the Ras-Raf-MAPK and the PI-3 Kinase pathways large Scottish case control data set [4], and both studies (reviewed in [12]). indicated a core haplotype (HAP ) as the DNA variation ICE carrying most of the risk for schizophrenia. Subsequently, In order to investigate the involvement of neuregulin 15 published studies in schizophrenia data sets of various pathway genes in schizophrenia, beyond the previously Asian and Caucasian ethnic backgrounds have detected published association with NRG1 and ERBB4, we have association in NRG1 SNPs or haplotypes, while only four tested all eight genes from the ERBB and NRG families for studies were not able to replicate the association (see [5] association with schizophrenia. We have investigated and [6] for a comprehensive review). The association both single genes and gene-gene interactions, using a col- studies reported thus far strongly suggest a true effect of lection of schizophrenia cases and blood bank controls NRG1 as a susceptibility gene for schizophrenia. How- from Aberdeen, Scotland. ever, some caution should be exercised before calling NRG1 a true "schizophrenia gene" due to the lack of a Methods Subjects known functional effect of the identified NRG1 variants, the allelic heterogeneity reported across several studies 396 Caucasian cases were ascertained in Aberdeen, Scot- and the multiple SNPs and haplotypes analysed in differ- land. All have a basic diagnosis of schizophrenia or ent studies. Nonetheless, several recent publications schizoaffective disorder according to the Operational Cri- report an association between schizophrenia and ErbB4, one of the receptors for NRG1 [7-9]. This provides indirect evidence supporting NRG1 association with schizophre- nia and a role for aberrant neuregulin signalling in this disorder. The Neuregulin-ErbB signalling network is involved in a multitude of processes in the developing and adult brain. Neuregulins for example promote neuronal migration and differentiation, regulate the expression of neurotrans- mitter receptors, influence glial proliferation, survival and differentiation and play a role in synaptic plasticity. The neuregulins (NRG) are cell-cell signalling proteins that are ligands for receptor tyrosine kinases of the ErbB family. Whereas NRG1 is known to play essential roles in nervous system and heart development, as well as in the Binding Figure 1 specificities of ErbB receptor family ligands adult brain (see above), less is known about the other Binding specificities of ErbB receptor family ligands. members of the NRG gene family, NRG2, -3 and -4 EGF (epidermal growth factor), AREG (amphiregulin) and [10,11]. TGFA (transforming growth factor-α) bind EGFR (ErbB1) only. BTC (Betacellulin), HBEGF (heparin-binding EGF-like The ErbB family comprises four homologous typeI recep- growth factor) and EREG (Epiregulin) bind EGFR and ErbB4. tor tyrosine kinases (RTKs). The EGFR (epidermal growth NRG1 and NRG2 bind ErbB3 and ErbB4. NRG3 and NRG4 factor receptor, HUGO name for ErbB1), ErbB3 and are specific for ErbB4. Receptors form homo- and het- ErbB4 receptors can bind ligands, whereas ErbB2 lacks a erodimers after ligand-binding (not shown). As indicated (X), ligand binding domain and functions as a preferred and ErbB2 has no ligand-binding capacity and ErbB3 has no active very potent co-receptor. ErbB3 is devoid of an active kinase domain. Only NRG1-4 and ERBB1-4 genes were investigated in this study. (Modified from [46]). kinase domain. ErbB4 is the only ErbB family member that binds all four neuregulins, as well as several proteins Page 2 of 11 (page number not for citation purposes) Behavioral and Brain Functions 2007, 3:31 http://www.behavioralandbrainfunctions.com/content/3/1/31 teria Checklist (OPKRIT). 285 of the cases were male; 111 Statistical analysis were female. The mean recruitment age was 44 yr with a All SNPs were tested for departure from Hardy Weinberg standard deviation of 14 yr. equilibrium (HWE) in controls using a chi squared test with one degree of freedom. This test compares the 1342 Controls were ascertained as a series of anonymous expected genotype frequencies calculated from the allele blood donors from Aberdeen, Scotland. Due to subject frequencies, with observed genotype frequencies. A depar- anonymity, characteristics such as schizophrenia- status, ture from HWE in controls may indicate errors in the data. age and ethnic group could not be determined. However at the time of sample collection, the Aberdeen Blood Association testing of both alleles and genotypes was car- Transfusion Service made a concerted effort to exclude any ried out using the 'PROC FREQ' fast Fisher's Exact Test non-Caucasian samples. At the time of this study, non- (FET) procedure in the statistical software package Caucasians constituted around 2% of the Aberdeen popu- SASv8.2 (SAS Institute Inc., Cary, NC, USA). The Fast lation but a lower percentage of Aberdeen blood donors. Fisher's Exact Test computes exact P-values for general 724 control subjects were male; 618 were female. contingency tables using the network algorithm devel- Informed consent was obtained for all subjects involved oped by [16]. The network algorithm provides substantial in this study. advantages over direct enumeration and is rapid and accu- rate. The exact test is used in preference to the asymptotic As calculated by the power equations of Risch and Teng chi-square P-value as it is more accurate at loci where the [13], under a multiplicative model of association, a sam- minor allele frequency is low. ple of 400 cases and 1300 controls has 80% power to detect disease association at an α = 0.05 significance level In order to model and test for interactions between SNPs for SNPs having a genotypic relative risk of 1.5 [14] and in different candidate genes, standard logistic regression susceptibility allele frequencies between 0.07 and 0.9. analysis was performed using the statistical software pack- age R (The R Project for Statistical Computing, [17]). Fol- Polymorphism selection and genotyping lowing Cordell [18] the model tested had the following SNPs from each gene were selected at an approximate form: spacing of 1 SNP every 10 kb. Six of these SNPs are located in potentially functional regions of the genes, five in log(p/(1-p)) = μ + a x + d z + a x + d z + i x x + i x 1 1 1 1 2 2 2 2 aa 1 2 ad 1 exons and one in a promoter region (Table 1). Among the z + i z x + i z z 2 da 1 2 dd 1 2 SNPs that are in exons, two are synonymous, but the other three produce amino acid changes. The size of these genes where x , x , z and z are variables determined by the gen- 1 2 1 2 varied greatly, from 23 kb to 1110 kb, and there was a cor- otype under consideration, and μ, a , a , d , d , i , i , i 1 2 1 2 aa ad da responding variation in the number of SNPs analysed, and i are parameters to be estimated. Here, genotypes at dd from 4 to 109 (for further details see Table 2, in the two SNPs are being modelled and, for each of the 9 possi- Results section). ble genotype combinations at the two SNPs, p is the pro- portion of subjects with the genotype who are cases and Genotyping was performed using a modification of the (1-p) is the proportion of subjects with the genotype who single base chain extension assay previously described are controls. The first five terms in the equation represent [15], at GlaxoSmithKline, UK. A SNP passed QC if over the effects of the two SNPs on log(p/(1-p)) if there is no 80% of genotype calls were successful. statistical interactions between the two SNPs. A non-zero value for one or more of the remaining four coefficients (i , i , i , i ) indicates the presence of interaction. Statis- aa ad da dd Table 1: SNPs in potentially functional gene regions included in the present study. 1 2 Gene Polymorphism Chromosome Position Region Base change Amino acid change EGFR rs2072454 7 54988557 Exon 4 C/T Asn/Asn ERBB2 rs1801200 17 35133114 Exon 20 A/G Val/Ile ERBB2 rs1058808 17 35137563 Exon 30 C/G Pro/Ala ERBB4 rs3748962 2 212077370 Exon 27 C/T Val/Val NRG1 rs6994992 8 31615123 Promoter C/T - NRG1 rs3924999 8 32572900 Exon 2 C/T Gln/Arg Polymorphism names are taken from the dbSNP website, [47]. From map NCBI 35. Page 3 of 11 (page number not for citation purposes) Behavioral and Brain Functions 2007, 3:31 http://www.behavioralandbrainfunctions.com/content/3/1/31 Table 2: Polymorphism in genes in the families NRG and ErbB investigated in this study. 1 1 Gene Chromosome Size (kb) No. of analysable SNPs No. of SNPs with p < 0.10 No. of SNPs with p < 0.05 NRG1 8 1120 94 22 (23%) 11 (12%) NRG2 5 196 16 2 (13%) 1 (6%) NRG3 10 1110 105 21 (20%) 10 (10%) NRG4 15 70 9 0 0 EGFR 7 188 22 7 (32%) 6 (27%) ERBB2 17 29 4 1 (25%) 0 ERBB3 12 23 6 0 0 ERBB4 2 1160 109 26 (24%) 14 (13%) Either genotypic test or allelic test of association. tical significance was assessed by a likelihood ratio test, Of 374 SNPs submitted for genotyping, 3 were dropped comparing the full model above with a reduced model in from analysis due to monomorphism in cases and con- which the last four terms are absent (i.e. in which the coef- trols, and another 6 were dropped due to departure from ficients i , i , i and i are all set to zero). Hardy-Weinberg Equilibrium (P < 0.001). Among the aa ad da dd remaining SNPs, the proportion of individuals success- A significant interaction effect in the logistic regression fully genotyped ranged from 0.804 to 0.99, with a median analysis indicates that the combined effect of the two of 0.97. A subset of the individuals studied was genotyped SNPs is significantly different from the sum of their indi- in duplicate, giving a total of 14,323 duplicate genotypes. vidual effects. Pair-wise interactions may be of three types, Of these, only 2 (= 0.014%) were inconsistent. The 365 namely 'and', 'or' and 'exclusive or', defined as follows. In SNPs retained, located in 8 genes, were subjected to sin- an 'and' relationship, the effect of having the risk-enhanc- gle-point analysis. The number of SNPs per gene is given ing genotypes at both SNP loci is larger than the sum of in Table 2. The median frequency of the minor allele the single-locus effects. In an 'or' relationship, the effect of among the SNPs analysed was 22.9 %, but there were 35 having the risk-enhancing genotype at both loci is smaller SNPs (9.6% of those analysed) for which the minor allele than the sum of the single-locus effects. In the final type frequency was less than 5%. These SNPs, selected prior to of interaction, named 'exclusive or', the effect of having the availability of the HapMap [20], are less informative the risk-enhancing genotype at both loci is smaller than than the others; however, as they do provide some addi- that of having a risk-enhancing genotype at only one tional information they were retained in the analysis. locus. For each SNP pair, each of these configurations gives rise to a contingency table, classified by genotype Single-point results group (risk-enhancing or risk-reducing) and phenotype Both allelic and genotypic tests of association were con- (case or control). These contingency tables were evalu- ducted. Detailed results are provided in Additional file 1. ated, and the configuration with the lowest P-value for Significant results (P < 0.05) were observed in five genes, association with the phenotype was identified. This con- namely NRG1, NRG2, NRG3, EGFR and ERBB4. These figuration indicated the type of interaction between the confirmed previous findings on NRG1 and ERBB4 and two loci in question. suggest possible involvement also of NRG2, NRG3 and EGFR. Borderline evidence of association (P < 0.10) was observed for one SNP in ERBB2. Haplotypic tests of asso- Results Data checking ciation (not shown) provided similar results to the single- Although controls were selected from the same geo- point results described here. If none of the 365 single graphic area as the cases, a test was conducted to deter- SNPs were associated with the disease status (the null mine the presence of any latent population stratification. hypothesis), we would expect 18 SNPs to be significant at The lambda statistic of Devlin et al. [19], calculated on the 0.05 level and 37 SNPs to be significant at the 0.10 4,250 SNPs distributed over the whole genome, showed level on an allelic test. For the allelic test there were 34 no evidence of stratification (population heterogeneity, SNPs significant at the 0.05 level (9.3%) and 59 SNPs sig- overdispersion: Genotypic exact lambda = 1.061, Allelic nificant at the 0.10 level (16.2%). For the genotypic test exact lambda = 1.000), indicating that no genomic-con- there were 24 SNPs significant at the 0.05 level (6.6%) trol adjustment is required when performing tests of gen- and 62 SNPs significant at the 0.10 level (17.0%). This otypic and allelic association on these subjects. excess of significantly associated SNPs, beyond the num- Page 4 of 11 (page number not for citation purposes) Behavioral and Brain Functions 2007, 3:31 http://www.behavioralandbrainfunctions.com/content/3/1/31 bers expected by chance when accounting for the multiple were significantly associated in NRG3, including one tests, provides evidence that some of these associations which was highly significantly associated (P < 0.01) by the are real. allelic test (RS3924461: map position 084423371 on Chr10, OR = 1.283, CI 1.077–1.529). The pattern of LD in The distribution of evidence for single-point association this gene (results not shown) is highly localised, similar to with schizophrenia, and linkage disequilibrium (LD), that in NRG1, so the observation of association from spa- over gene NRG1 are presented in Fig. 2. In this figure, the tially separated SNPs may have arisen by chance or may magnitude of association of each SNP with schizophrenia suggest the presence of different functional variants is indicated by the log-transformed P-value. For the sake within the same gene. The spatial pattern of association is of simplicity, only values arising from the genotypic exact not related to the distribution of the exons in this gene. Six test are shown; allelic test results followed a very similar SNPs in EGFR were significantly associated with schizo- pattern. Across both tests, eleven SNPs in this gene phrenia, however, all fell in the range 0.01 <P < 0.05. Only showed significant genotypic association with schizo- four SNPs were studied in ERBB2, and none of these was phrenia (P < 0.05), and in three instances the association found to be significantly associated with schizophrenia. was highly significant (P < 0.01) (see Additional file 1: SNP RS776401: map position 031836504 on Chr 8, risk Interaction analysis results genotype odds ratio (OR) = 1.461, 95% confidence inter- As summarised in Table 2, 79 SNPs were identified which val (CI) 1.160–1.840; SNP RS1383887: map position had a P-value from an allelic or genotypic test that was less 031872698, OR = 1.589, CI 1.213–2.081; SNP RS- than 0.10 (details are given in Additional file 2). These GSK8116812: map position 032741837, OR = 1.651, CI SNPs were contained in the 6 genes considered above, 1.242–2.196). A putatively functional SNP in Exon 2, cod- namely NRG1, NRG2, NRG3, EGFR, ERBB2, and ERBB4. ing a non-synonymous (Gly/Arg) change, showed allelic All pairs of these SNPs that are at least 0.5 cM apart on an association to schizophrenia with P = 0.0279 (SNP interpolated Decode genetic map [21] were tested for RS3924999: map position 032572900, OR = 1.302, CI interaction effects in a logistic regression interaction anal- 1.028–1.648). It is possible that the cluster of associations ysis. observed at the 3' end of the gene, to the right of the fig- ure, has arisen through LD with this non-synonymous A total of 2640 statistical tests for pair-wise interactions change or another underlying functional variant. Further were performed (details of results are given in Additional evidence of association is seen at the 5' end of the gene, file 2). Of these, 212 were significant at the 5% level: that closer to Exon 1. The latter associations fall beyond the is 8.03% of all tests were significant compared to 132 (= apparent reach of LD. 5% of 2640) expected by chance if the null hypothesis of no interaction were true for all pairs of SNPs tested. If all The corresponding information for gene ERBB4 is pre- the tests performed were mutually independent the prob- sented in Fig. 3. In this gene, fourteen SNPs were signifi- ability that 212 or more would be significant by chance is -11 cantly associated with schizophrenia by either the allelic 2.575 × 10 (P(R ≥ 212) where R~binomial(N = 2640, p or genotypic test, including one allelic test result which = 0.05)). The true probability of such an outcome is cer- was highly significant (P < 0.01: SNP RS6754674: map tain to be larger, due to non-independence among the position 213033762 on Chr 2, OR = 1.387, CI 1.056– tests: nevertheless, this result provides convincing evi- 1.822). Again, the graph of genotypic test results shows dence that some of the observed interactions are real. The strong evidence of association occurring most commonly significance levels presented here are not adjusted for at the two ends of the gene, particularly at the 5'-end, to multiple testing: a Bonferroni correction would be inap- the right of the figure, although the pattern is not as clear propriate due to non-independence among the tests, and as in the case of NRG1. This region of the gene represents the multiple-testing issue is addressed by the comparison a 0.15 Mb region of moderate LD, so it is possible that a of observed and expected numbers of significant tests pre- single underlying variant, in LD with all of the SNPs that sented above. None of the significant results obtained show significant association, accounts for the observed would survive the Bonferroni correction. association in this region. Table 3 gives the percentage of significant interactions for The corresponding information for genes NRG2, NRG3, each gene combination; the corresponding counts are EGFR and ERBB2 is presented in Fig. 4, except that the LD given in brackets. There are four gene combinations in maps are omitted for brevity. A single SNP in NRG2 which the percentage of significant interactions is sub- showed significant genotypic association with schizo- stantially higher than the 5% expected by chance, viz. the phrenia (SNP RS2936651: map position 139401129 on two-gene combinations NRG1-NRG2, NRG1-NRG3 and Chr 5, OR = 2.081, CI 1.156–3.745). This finding was cor- NRG2-EGFR, and the single large gene ERBB4. The per- roborated by a significant allelic test result also. Ten SNPs centages of significant interactions for combinations Page 5 of 11 (page number not for citation purposes) Behavioral and Brain Functions 2007, 3:31 http://www.behavioralandbrainfunctions.com/content/3/1/31 RefSeq Genes NRG1 31600000 31800000 32000000 32200000 32400000 32600000 Chromosome Position (bp) Genotypic single-point a Figure 2 ssociation test results and LD over NRG1 Genotypic single-point association test results and LD over NRG1. In the plot of -log (P) v. map position, horizontal dashed lines are at P = 0.05, 0.01 and 0.001. Vertical dashed lines are at 20 kb intervals. The region covered by the gene is indi- cated in the centre of the figure by a horizontal line, with an arrow head indicating the direction of transcription. A crossing vertical line indicates the position of each exon. At the base of the figure, LD between SNP loci is indicated by shading. White: 2 2 r = 0; black: r = 1. NRG1-NRG2 and NRG2-EGFR are based on a small ple of significant interactions. This also might be number of tests and should therefore be treated with cau- expected, as it corresponds to a complex biological mech- tion, but the percentages for NRG1-NRG3 and ERBB4 are anism and to a rather specific pattern of results that is rel- based on substantial numbers. atively unlikely to occur by chance. The next step towards the identification of a mechanism Discussion to account for the statistical interactions is the identifica- In this manuscript we present confirmation of earlier pub- tion of the type of each significant interaction. This was lished associations between schizophrenia and the genes performed by pair-wise combination analysis (see Meth- NRG1 and ERBB4. We build a new, compelling picture of ods). Fig. 5 illustrates the distribution of the three types of pathway involvement by showing that associations exist interaction over the four gene combinations with a large with additional gene family-members, namely EGFR, percentage of significant interactions. The frequencies of NRG2, and NRG3. Lastly, we present statistical evidence and-type and or-type interactions are fairly similar. This of gene epistasis, consistent with a highly complex biolog- might be expected, as the patterns of results that corre- ical pathway of schizophrenia susceptibility. This is the spond to these two types of interaction are complemen- first time that the roles and inter-relationships of the ERBB tary, and equally likely to occur by chance. Interactions of and NRG gene families as a whole have been investigated the xor-type are rare: indeed, they are observed only in the in relation to schizophrenia and our approach sets a prec- two gene combinations that gave rise to a fairly large sam- edent in the investigation of common complex diseases. Page 6 of 11 (page number not for citation purposes) − log (P) 0.0 1.0 2.0 3.0 Behavioral and Brain Functions 2007, 3:31 http://www.behavioralandbrainfunctions.com/content/3/1/31 RefSeq Genes ERBB4 212200000 212400000 212600000 212800000 213000000 213200000 Chromosome Position (bp) Genotypic single-point a Figure 3 ssociation test results and LD over ERBB4 Genotypic single-point association test results and LD over ERBB4. In the plot of -log (P) v. map position, horizontal dashed lines are at P = 0.05, 0.01 and 0.001. Vertical dashed lines are at 20 kb intervals. The region covered by the gene is indi- cated in the centre of the figure by a horizontal line, with an arrow head indicating the direction of transcription. A crossing vertical line indicates the position of each exon. At the base of the figure, LD between SNP loci is indicated by shading. White: 2 2 r = 0; black: r = 1. Clear evidence for gene-gene interaction was detected for NRG1 overlap with those found by Stefansson et al. [3,4], gene combinations NRG1-NRG2, NRG1-NRG3 and Corvin et al. [23], Li et al. [24], Zhao et al. [25] and Petry- EGFR-NRG2, and suggestive evidence was also seen for shen et al. [26]. Similarly, the associations we observed at ERBB4-NRG1, ERBB4-NRG2, ERBB4-NRG3 and ERBB4- the 3' end of the gene overlap with those found by Yang et ERBB2. Evidence of intragenic interaction was seen for al. [27], Li et al. [23], Petryshen et al. [25] and Lachmann SNPs in ERBB4. Of the 2640 tests for pair-wise interac- et al. [28]. Two polymorphisms in our study directly rep- tion, 212 were significant at the 5% level, which is sub- licate these associations: the putatively functional SNP stantially more than the 131 expected by chance. This rs3924999 was found to be associated with schizophrenia large number of low P-values may be partly explained by in a Han Chinese family study conducted by Yang et al. non-independence among tests. However, it may also [26]. This SNP was also tested in a Caucasian study of 136 indicate that some of the 79 SNPs studied interact to mod- families by Duan et al. [29], but was not significant. ulate susceptibility to schizophrenia. Rs2466049 was significant as part of a haplotype found in a Caucasian family by Petryshen et al. [25]. Numerous studies have investigated the association between schizophrenia and NRG1 (reviewed in [22]). A For the ERBB4 gene, four polymorphisms (rs3748962, comparison of the associations generated in the current rs839523, rs2054617 and rs6435711) screened in this study with those previously published shows good agree- study were also analysed by Silberberg et al. [8]. However, ment. Associations that we observed at the 5' region of none of the individual SNP associations observed in the Page 7 of 11 (page number not for citation purposes) − log (P) 0.0 1.0 2.0 3.0 Behavioral and Brain Functions 2007, 3:31 http://www.behavioralandbrainfunctions.com/content/3/1/31 Genotypic single-point a Figure 4 ssociation test results and LD over NRG2, NRG3, EGFR and ERBB2 Genotypic single-point association test results and LD over NRG2, NRG3, EGFR and ERBB2. In the plot of - log (P) v. map position, horizontal dashed lines are at P = 0.05, 0.01 and 0.001. Vertical dashed lines are at 20 kb intervals. The region covered by the gene is indicated in the centre of the figure by a horizontal line, with an arrow head indicating the direc- tion of transcription. A crossing vertical line indicates the position of each exon. previous study was replicated here. As no underlying func- from the NRG and ERBB gene families, i.e. NRG2, NRG3 tional variant has been identified for ERBB4, this lack of and EGFR (ERBB1). It is interesting to note that we only association at the SNP level is unsurprising and is consist- detected evidence for association of NRG genes which are ent with a heterogeneous haplotypic background across expressed in the central nervous system, namely NRG1, studies. It is worth noting also that non-coding polymor- NRG2 and the nervous-system-specific NRG3 [31], phisms might lead to dysregulation of ErbB4 expression whereas NRG4, which is mainly expressed in pancreas levels, a hypothesis corroborated by the Silberberg et al. and absent from the brain [32] was not found to be asso- [8] study, which found enhanced expression of the ERBB4 ciated to schizophrenia. Jm-a CYT-1 isoform (one of four splice variants, see [30]) in schizophrenic patients. The altered balance of ErbB4 To our knowledge, an association of EGFR to schizophre- isoforms could lead to variations in homo- and het- nia has not previously been investigated. Association of a erodimer formation and thus in downstream signalling polymorphism in EGF (one of the six EGFR ligands, Fig. pathways. 1) with schizophrenia was reported in Finnish men [33], but not replicated in a Japanese population [34]. Further In addition to providing further evidence of NRG1 and support for a possible role of EGF and EGFR in the disease ERBB4 associations to schizophrenia, our study is the first comes from findings of abnormal expression of these to suggest possible involvement of three additional genes genes in the forebrain and serum of schizophrenic Page 8 of 11 (page number not for citation purposes) Behavioral and Brain Functions 2007, 3:31 http://www.behavioralandbrainfunctions.com/content/3/1/31 Table 3: Percentages of pair-wise interactions that were significant (P < 0.05), classified by the gene(s) in which the interacting SNPs lie. Gene combinations in which there are substantially more significant interactions than the 5% expected by chance are indicated in bold. The counts that contribute to each percentage are also shown in parentheses as a fraction. The presence of "-" in the table indicates that no pairs of SNPs in this gene are 0.5 cM apart. NRG1 NRG2 NRG3 EGFR ERBB2 ERBB4 NRG1 5.8% (6/103) 13.6% (6/44) 11.7% (54/462) 4.5% (7/154) 0.0% (0/22) 7.9% (45/572) NRG2 - 4.8% (2/42) 14.3% (2/14) 0.0% (0/2) 7.7% (4/52) NRG3 5.7% (6/105) 2.0% (3/147) 0.0% (0/21) 8.4% (46/546) EGFR - 0.0% (0/7) 5.5% (10/182) ERBB2 -7.7% (2/26) ERBB4 13.7% (19/139) patients [35]. EGFR is usually omitted in reviews on NRG- receptor. The NRG2-EGFR interaction is superficially ERBB signalling, although it is well-established that ErbB4 more difficult to explain since NRG2 is a ligand for ErbB3 can interact with and cross-phosphorylate EGFR upon and ErbB4 and does not usually recognize EGFR directly neuregulin-binding [36-39]. Furthermore, EGFR and or with high affinity [42-44]. However, a recent study ERBB4 show partially overlapping temporal and spatial reports that a single amino acid change in the EGFR lig- mRNA expression in several brain areas, such as cortex, and-binding domain dramatically increases the receptor striatum, cerebellum and hippocampus, and are coex- affinity for NRG2β which can then induce potent stimula- pressed in GABAergic interneurons [40,41]. Therefore, it is tion of EGFR signalling [45]. The interactions within the very likely that EGFR/ErbB4 heterodimers mediate impor- ERBB4 gene suggest that combinations of particular geno- tant functions in the developing and adult central nervous types may result in modified receptor activity, leading to system in vivo. Whilst we limited this study to the NRG altered risk of schizophrenia. and ERBB gene families, it would be interesting to investi- gate association of the additional EGFR and ErbB4 ligands Such suggestions concerning possible mechanisms of (see Fig. 1). interaction remain speculative until the reality of these interactions has been established biologically and the The significant gene-gene interactions reported in this effect of each two-locus genotype on the phenotype is study support our hypothesis of pathway involvement in understood. A follow-up study currently underway, using schizophrenia. These genes have related functions and, for independent cases and controls, will clarify which of these example, the NRG1-NRG2 and NRG1-NRG3 interactions interactions can be replicated and may generate further may suggest competition between these ligands for a hypotheses regarding the nature of the interactions. It will be interesting also to investigate whether observed associ- ations or their interactions correspond to particular phe- notypically defined patient sub-groups. The strong evidence of genetic associations and interactions reported here provides ample justification for such an enquiry. Conclusion This study confirms previously published associations between schizophrenia and NRG1 and ERBB4, and points to three other members of the NRG and ERBB gene fami- lies (EGFR, NRG2 and NRG3) as potential schizophrenia susceptibility genes. Furthermore, the study presents evi- dence for inter- and intragenic pair-wise interaction between loci in genes of both families (NRG1-NRG2, NRG2-EGFR, NRG1-NRG3 and ERBB4-ERBB4). These Distribution combinatio tions Figure 5 ns wi of the types of th a large perc pair-wise interaction in gene entage of significant interac- extensive genetic inter-relationships observed among lig- Distribution of the types of pair-wise interaction in ands and receptors of the ERBB and NRG gene families gene combinations with a large percentage of signifi- reflect the complexity of their interactions at the molecu- cant interactions. The total in each column corresponds lar level. The influence of so many closely related genetic to the number of significant interactions presented in Table factors on the potential development of disease patho- 3, i.e. the numerator of the fraction presented in that table for the gene combination in question. physiology has rarely, if ever, been demonstrated before and may partially explain the difficulty in arriving at an Page 9 of 11 (page number not for citation purposes) Behavioral and Brain Functions 2007, 3:31 http://www.behavioralandbrainfunctions.com/content/3/1/31 overarching genetic hypothesis of the etiology of schizo- Acknowledgements The authors wish to acknowledge the patients and blood donors who con- phrenia. In future studies, it will be important to replicate tributed to this study. Thanks are also due to members of the GSK Genetic our new findings in an independent cohort and to under- Sample Management and Genotyping groups; also Matt Nelson, Pierandrea stand the underlying biological mechanisms of the single- Muglia, Paul Skelding and Nicola Foote for their helpful contributions in the locus associations as well as the interactions among differ- writing of this paper. ent loci. References Competing interests 1. Owen MJ, Craddock N, O'Donovan MC: Schizophrenia: genes at last? [Review]. Trends in Genetics 2005, 21:518-525. The author(s) declare that they have no competing inter- 2. Harrison PJ, Weinberger DR: Schizophrenia genes, gene expres- ests. sion, and neuropathology: on the matter of their conver- gence [Review]. Molecular Psychiatry 2005, 10:40-68. 3. Stefansson H, Sigurdsson E, Steinthorsdottir V, Bjornsdottir S, Sig- Authors' contributions mundsson T, Ghosh S, Brynjolfsson J, Gunnarsdottir S, Ivarsson O, IB helped to formulate the study and identified the genes Chou TT, Hjaltason O, Birgisdottir B, Jonsson H, Gudnadottir VG, of interest; she led the writing of the manuscript and pro- Gudmundsdottir E, Bjornsson A, Ingvarsson B, Ingason A, Sigfusson S, Hardardottir H, Harvey RP, Lai D, Zhou M, Brunner D, Mutel V, vided the biological interpretation and context to our Gonzalo A, Lemke G, Sainz J, Johannesson G, Andresson T, Gudb- results. AB led the statistical analysis, and produced the jartsson D, Manolescu A, Frigge ML, Gurney ME, Kong A, Gulcher JR, Petursson H, Stefansson K: Neuregulin 1 and susceptibility to first draft of the manuscript. BB performed the single- schizophrenia. Am J Hum Genet 2002, 71:877-892. point statistical analyses and the pairwise analysis, and 4. Stefansson H, Sarginson J, Kong A, Yates P, Steinthorsdottir V, Gud- provided interpretation of his results. NG provided inter- finnsson E, Gunnarsdottir S, Walker N, Petursson H, Crombie C, Ingason A, Gulcher JR, Stefansson K, St Clair D: Association of neu- pretation of the linkage disequilibrium analysis and sub- regulin 1 with schizophrenia confirmed in a Scottish popula- stantially contributed to the interpretation of the tion. Am J Hum Genet 2003, 72:83-87. 5. Tosato S, Dazzan P, Collier D: Association between the neureg- statistical analyses in the manuscript. PM was instrumen- ulin 1 gene and schizophrenia: a systematic review. Schizophr tal in progressing the study, revised the manuscript and Bull 2005, 31:613-617. added important intellectual content concerning the biol- 6. Harrison PJ, Law AJ: Neuregulin 1 and Schizophrenia: Genetics, Gene Expression, and Neurobiology. Biol Psychiatry 2006, ogy of schizophrenia. RM designed the statistical analysis 60:132-140. plan and provided schizophrenia genetics expertise to the 7. Norton N, Moskvina V, Morris DW, Bray NJ, Zammit S, Williams writing of the manuscript. DS designed the laboratory NM, Williams HJ, Preece AC, Dwyer S, Wilkinson JC, Spurlock G, Kirov G, Buckland P, Waddington JL, Gill M, Corvin AP, Owen MJ, aspects of the experiment and oversaw their execution. O'Donovan MC: Evidence that interaction between neuregu- DSC recruited, worked up clinically and supplied DNA lin 1 and its receptor erbB4 increases susceptibility to schiz- ophrenia. Am J Med Genet B Neuropsychiatr Genet 2006, 141:96-101. from the Aberdeen schizophrenia case collection of 396 8. Silberberg G, Darvasi A, Pinkas-Kramarski R, Navon R: The involve- cases. PY provided the control collection of 1342 anony- ment of ErbB4 with schizophrenia: association and expres- mous blood donors from Aberdeen. IP helped to formu- sion studies. Am J Med Genet B Neuropsychiatr Genet 2006, 141:142-148. late the original idea for the study and revised the 9. Nicodemus KK, Luna A, Vakkalanka R, Goldberg T, Egan M, Straub manuscript critically, providing important intellectual RE, Weinberger DR: Further evidence for association between ErbB4 and schizophrenia and influence on cognitive inter- content. All authors read, edited and approved the final mediate phenotypes in healthy controls. Molecular Psychiatry manuscript. 2006, 11:1062-1065. 10. Falls DL: Neuregulins: functions, forms, and signaling strate- gies. Exp Cell Res 2003, 284:14-30. Additional material 11. Talmage DA, Role LW: Multiple personalities of neuregulin gene family members. J Comp Neurol 2004, 472:134-139. 12. Yarden Y, Sliwkowski MX: Untangling the ErbB signalling net- Additional file 1 work. Nat Rev Mol Cell Biol 2001, 2:127-137. Detailed results from allelic and genotypic tests of single-point association 13. Risch N, Teng J: The relative power of family-based and case- control designs for linkage disequilibrium studies of complex of SNP loci with schizophrenia. human diseases - I. DNA pooling. Genome Research 1998, Click here for file 8:1273-1288. [http://www.biomedcentral.com/content/supplementary/1744- 14. Risch N, Merikangas K: The future of genetic studies of complex 9081-3-31-S1.xls] human diseases. Science 1996, 273:1516-1517. 15. Taylor JD, Briley D, Nguyen Q, Long K, Iannone MA, Li MS, Ye F, Afshari A, Lai E, Wagner M, Chen J, Weiner MP: Flow cytometric Additional file 2 platform for high-throughput single nucleotide polymor- Detailed results from allelic and genotypic tests of association between phism analysis. Biotechniques 2001, 30:661-669. SNP-pairs and susceptibility to schizophrenia. 16. CR M, NR P: A network algorithm for performing Fisher's Click here for file Exact Test in r ×c contingency tables. J Am Stat Assoc 1983, 78:427-434. [http://www.biomedcentral.com/content/supplementary/1744- 17. The R Project for Statistical Computing 2003 [http://www.r- 9081-3-31-S2.xls] project.org/]. 18. Cordell HJ: Epistasis: what it means, what it doesn't mean, and statistical methods to detect it in humans. Hum Mol Genet 2002, 11:2463-2468. Page 10 of 11 (page number not for citation purposes) Behavioral and Brain Functions 2007, 3:31 http://www.behavioralandbrainfunctions.com/content/3/1/31 19. Devlin B, Roeder K, Wasserman L: Genomic control, a new 38. Riese DJ, van Raaij TM, Plowman GD, Andrews GC, Stern DF: The approach to genetic-based association studies. Theoretical Pop- cellular response to neuregulins is governed by complex ulation Biology 2001, 60:155-166. interactions of the erbB receptor family. Mol Cell Biol 1995, 20. Altshuler D, Brooks LD, Chakravarti A, Collins FS, Daly MJ, Donnelly 15:5770-5776. P: A haplotype map of the human genome. Nature 2005, 39. Zhang K, Sun J, Liu N, Wen D, Chang D, Thomason A, Yoshinaga SK: 437:1299-1320. Transformation of NIH 3T3 cells by HER3 or HER4 recep- 21. Kong A, Gudbjartsson DF, Sainz J, Jonsdottir GM, Gudjonsson SA, tors requires the presence of HER1 or HER2. J Biol Chem 1996, Richardsson B, Sigurdardottir S, Barnard J, Hallbeck B, Masson G, Shl- 271:3884-3890. ien A, Palsson ST, Frigge ML, Thorgeirsson TE, Gulcher JR, Stefansson 40. Fox IJ, Kornblum HI: Developmental profile of ErbB receptors K: A high-resolution recombination map of the human in murine central nervous system: implications for functional genome. Nat Genet 2002, 31:241-247. interactions. J Neurosci Res 2005, 79:584-597. 22. Munafo MR, Thiselton DL, Clark TG, Flint J: Association of the 41. Kornblum HI, Gall CM, Seroogy KB, Lauterborn JC: A subpopula- NRG1 gene and schizophrenia: a meta-analysis. Mol Psychiatry tion of striatal gabaergic neurons expresses the epidermal 2006, 11:613. growth factor receptor. Neuroscience 1995, 69:1025-1029. 23. Corvin AP, Morris DW, McGhee K, Schwaiger S, Scully P, Quinn J, 42. Carpenter G: ErbB-4: mechanism of action and biology. Exp Meagher D, Clair DS, Waddington JL, Gill M: Confirmation and Cell Res 2003, 284:66-77. refinement of an 'at-risk' haplotype for schizophrenia sug- 43. Hobbs SS, Coffing SL, Le AT, Cameron EM, Williams EE, Andrew M, gests the EST cluster, Hs.97362, as a potential susceptibility Blommel EN, Hammer RP, Chang H, Riese DJ: Neuregulin iso- gene at the Neuregulin-1 locus. Mol Psychiatry 2004, 9:208-213. forms exhibit distinct patterns of ErbB family receptor acti- 24. Li T, Stefansson H, Gudfinnsson E, Cai G, Liu X, Murray RM, vation. Oncogene 2002, 21:8442-8452. Steinthorsdottir V, Januel D, Gudnadottir VG, Petursson H, Ingason 44. Pinkas-Kramarski R, Shelly M, Guarino BC, Wang LM, Lyass L, Alroy A, Gulcher JR, Stefansson K, Collier DA: Identification of a novel I, Alimandi M, Kuo A, Moyer JD, Lavi S, Eisenstein M, Ratzkin BJ, Seger neuregulin 1 at-risk haplotype in Han schizophrenia Chinese R, Bacus SS, Pierce JH, Andrews GC, Yarden Y: ErbB tyrosine patients, but no association with the Icelandic/Scottish risk kinases and the two neuregulin families constitute a ligand- haplotype. Mol Psychiatry 2004, 9:698-704. receptor network. Mol Cell Biol 1998, 18:6090-6101. 25. Zhao X, Shi Y, Tang J, Tang R, Yu L, Gu N, Feng G, Zhu S, Liu H, Xing 45. Gilmore JL, Gallo RM, Riese DJ: The epidermal growth factor Y, Zhao S, Sang H, Guan Y, St Clair D, He L: A case control and receptor (EGFR)-S442F mutant displays increased affinity family based association study of the neuregulin1 gene and for neuregulin-2beta and agonist-independent coupling with schizophrenia. J Med Genet 2004, 41:31-34. downstream signalling events. Biochem J 2006, 396:79-88. 26. Petryshen TL, Middleton FA, Kirby A, Aldinger KA, Purcell S, Tahl AR, 46. Olayioye MA, Neve RM, Lane HA, Hynes NE: The ErbB signaling Morley CP, McGann L, Gentile KL, Rockwell GN, Medeiros HM, Car- network: receptor heterodimerization in development and valho C, Macedo A, Dourado A, Valente J, Ferreira CP, Patterson NJ, cancer. EMBO J 2000, 19:3159-3167. Azevedo MH, Daly MJ, Pato CN, Pato MT, Sklar P: Support for 47. NCBI dbSNP website 2004 [http://www.ncbi.nlm.nih.gov/SNP/]. involvement of neuregulin 1 in schizophrenia pathophysiol- ogy. Mol Psychiatry 2005, 10:366-74, 328. 27. Yang JZ, Si TM, Ruan Y, Ling YS, Han YH, Wang XL, Zhou M, Zhang HY, Kong QM, Liu C, Zhang DR, Yu YQ, Liu SZ, Ju GZ, Shu L, Ma DL, Zhang D: Association study of neuregulin 1 gene with schizo- phrenia. Mol Psychiatry 2003, 8:706-709. 28. Lachman HM, Pedrosa E, Nolan KA, Glass M, Ye K, Saito T: Analysis of polymorphisms in AT-rich domains of neuregulin 1 gene in schizophrenia. Am J Med Genet B Neuropsychiatr Genet 2006, 141:102-109. 29. Duan J, Martinez M, Sanders AR, Hou C, Krasner AJ, Schwartz DB, Gejman PV: Neuregulin 1 (NRG1 ) and schizophrenia: analysis of a US family sample and the evidence in the balance. Psychol Med 2005, 35:1599-1610. 30. Junttila TT, Sundvall M, Maatta JA, Elenius K: Erbb4 and its iso- forms: selective regulation of growth factor responses by naturally occurring receptor variants. Trends Cardiovasc Med 2000, 10:304-310. 31. Zhang D, Sliwkowski MX, Mark M, Frantz G, Akita R, Sun Y, Hillan K, Crowley C, Brush J, Godowski PJ: Neuregulin-3 (NRG3): a novel neural tissue-enriched protein that binds and activates ErbB4. Proc Natl Acad Sci U S A 1997, 94:9562-9567. 32. Harari D, Tzahar E, Romano J, Shelly M, Pierce JH, Andrews GC, Yarden Y: Neuregulin-4: a novel growth factor that acts through the ErbB-4 receptor tyrosine kinase. Oncogene 1999, 18:2681-2689. 33. Anttila S, Illi A, Kampman O, Mattila KM, Lehtimaki T, Leinonen E: Association of EGF polymorphism with schizophrenia in Publish with Bio Med Central and every Finnish men. Neuroreport 2004, 15:1215-1218. scientist can read your work free of charge 34. Watanabe Y, Fukui N, Muratake T, Kaneko N, Someya T: No asso- ciation of EGF polymorphism with schizophrenia in a Japa- "BioMed Central will be the most significant development for nese population. Neuroreport 2005, 16:403-405. disseminating the results of biomedical researc h in our lifetime." 35. Futamura T, Toyooka K, Iritani S, Niizato K, Nakamura R, Tsuchiya K, Sir Paul Nurse, Cancer Research UK Someya T, Kakita A, Takahashi H, Nawa H: Abnormal expression of epidermal growth factor and its receptor in the forebrain Your research papers will be: and serum of schizophrenic patients. Mol Psychiatry 2002, available free of charge to the entire biomedical community 7:673-682. 36. Cohen BD, Green JM, Foy L, Fell HP: HER4-mediated biological peer reviewed and published immediately upon acceptance and biochemical properties in NIH 3T3 cells. Evidence for cited in PubMed and archived on PubMed Central HER1-HER4 heterodimers. J Biol Chem 1996, 271:4813-4818. 37. Hatakeyama M, Yumoto N, Yu X, Shirouzu M, Yokoyama S, Konagaya yours — you keep the copyright A: Transformation potency of ErbB heterodimer signaling is BioMedcentral determined by B-Raf kinase. Oncogene 2004, 23:5023-5031. Submit your manuscript here: http://www.biomedcentral.com/info/publishing_adv.asp Page 11 of 11 (page number not for citation purposes) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Behavioral and Brain Functions Springer Journals

Interactions among genes in the ErbB-Neuregulin signalling network are associated with increased susceptibility to schizophrenia

Download PDF
11 pages

Loading next page...
 
/lp/springer-journals/interactions-among-genes-in-the-erbb-neuregulin-signalling-network-are-OykckQZOOM

References (51)

Publisher
Springer Journals
Copyright
Copyright © 2007 by Benzel et al; licensee BioMed Central Ltd.
Subject
Biomedicine; Neurosciences; Neurology; Behavioral Therapy; Psychiatry
eISSN
1744-9081
DOI
10.1186/1744-9081-3-31
pmid
17598910
Publisher site
See Article on Publisher Site

Abstract

Background: Evidence of genetic association between the NRG1 (Neuregulin-1) gene and schizophrenia is now well-documented. Furthermore, several recent reports suggest association between schizophrenia and single- nucleotide polymorphisms (SNPs) in ERBB4, one of the receptors for Neuregulin-1. In this study, we have extended the previously published associations by investigating the involvement of all eight genes from the ERBB and NRG families for association with schizophrenia. Methods: Eight genes from the ERBB and NRG families were tested for association to schizophrenia using a collection of 396 cases and 1,342 blood bank controls ascertained from Aberdeen, UK. A total of 365 SNPs were tested. Association testing of both alleles and genotypes was carried out using the fast Fisher's Exact Test (FET). To understand better the nature of the associations, all pairs of SNPs separated by ≥ 0.5 cM with at least nominal evidence of association (P < 0.10) were tested for evidence of pairwise interaction by logistic regression analysis. Results: 42 out of 365 tested SNPs in the eight genes from the ERBB and NRG gene families were significantly associated with schizophrenia (P < 0.05). Associated SNPs were located in ERBB4 and NRG1, confirming earlier reports. However, novel associations were also seen in NRG2, NRG3 and EGFR. In pairwise interaction tests, clear evidence of gene-gene interaction was detected for NRG1-NRG2, NRG1-NRG3 and EGFR-NRG2, and suggestive evidence was also seen for ERBB4-NRG1, ERBB4-NRG2, ERBB4-NRG3 and ERBB4-ERBB2. Evidence of intragenic interaction was seen for SNPs in ERBB4. Conclusion: These new findings suggest that observed associations between NRG1 and schizophrenia may be mediated through functional interaction not just with ERBB4, but with other members of the NRG and ERBB families. There is evidence that genetic interaction among these loci may increase susceptibility to schizophrenia. Page 1 of 11 (page number not for citation purposes) Behavioral and Brain Functions 2007, 3:31 http://www.behavioralandbrainfunctions.com/content/3/1/31 that had originally been identified as EGFR ligands (Fig. Background Schizophrenia is a complex psychiatric disorder which 1). Furthermore, NRG1 and NRG2 can also bind to ErbB3. affects 0.5–1% of the world wide adult population. A In contrast, none of the neuregulins bind to EGFR which number of putative schizophrenia susceptibility genes instead binds unrelated ligands such as EGF, amphiregu- have been identified recently [1,2]. Genomewide linkage lin and others (see Fig. 1). analysis on large Icelandic pedigrees showed evidence for linkage with an initial LOD score of 3 on chromosome Ligand binding to the extracellular domain of ErbB family 8p13, which led to the identification of the Neuregulin-1 members leads to receptor homo-and heterodimerisation gene (NRG1) as a potential genetic risk factor (O.R. = 2) and activation of various intracellular signalling pathways for schizophrenia [3]. The association was confirmed in a such as the Ras-Raf-MAPK and the PI-3 Kinase pathways large Scottish case control data set [4], and both studies (reviewed in [12]). indicated a core haplotype (HAP ) as the DNA variation ICE carrying most of the risk for schizophrenia. Subsequently, In order to investigate the involvement of neuregulin 15 published studies in schizophrenia data sets of various pathway genes in schizophrenia, beyond the previously Asian and Caucasian ethnic backgrounds have detected published association with NRG1 and ERBB4, we have association in NRG1 SNPs or haplotypes, while only four tested all eight genes from the ERBB and NRG families for studies were not able to replicate the association (see [5] association with schizophrenia. We have investigated and [6] for a comprehensive review). The association both single genes and gene-gene interactions, using a col- studies reported thus far strongly suggest a true effect of lection of schizophrenia cases and blood bank controls NRG1 as a susceptibility gene for schizophrenia. How- from Aberdeen, Scotland. ever, some caution should be exercised before calling NRG1 a true "schizophrenia gene" due to the lack of a Methods Subjects known functional effect of the identified NRG1 variants, the allelic heterogeneity reported across several studies 396 Caucasian cases were ascertained in Aberdeen, Scot- and the multiple SNPs and haplotypes analysed in differ- land. All have a basic diagnosis of schizophrenia or ent studies. Nonetheless, several recent publications schizoaffective disorder according to the Operational Cri- report an association between schizophrenia and ErbB4, one of the receptors for NRG1 [7-9]. This provides indirect evidence supporting NRG1 association with schizophre- nia and a role for aberrant neuregulin signalling in this disorder. The Neuregulin-ErbB signalling network is involved in a multitude of processes in the developing and adult brain. Neuregulins for example promote neuronal migration and differentiation, regulate the expression of neurotrans- mitter receptors, influence glial proliferation, survival and differentiation and play a role in synaptic plasticity. The neuregulins (NRG) are cell-cell signalling proteins that are ligands for receptor tyrosine kinases of the ErbB family. Whereas NRG1 is known to play essential roles in nervous system and heart development, as well as in the Binding Figure 1 specificities of ErbB receptor family ligands adult brain (see above), less is known about the other Binding specificities of ErbB receptor family ligands. members of the NRG gene family, NRG2, -3 and -4 EGF (epidermal growth factor), AREG (amphiregulin) and [10,11]. TGFA (transforming growth factor-α) bind EGFR (ErbB1) only. BTC (Betacellulin), HBEGF (heparin-binding EGF-like The ErbB family comprises four homologous typeI recep- growth factor) and EREG (Epiregulin) bind EGFR and ErbB4. tor tyrosine kinases (RTKs). The EGFR (epidermal growth NRG1 and NRG2 bind ErbB3 and ErbB4. NRG3 and NRG4 factor receptor, HUGO name for ErbB1), ErbB3 and are specific for ErbB4. Receptors form homo- and het- ErbB4 receptors can bind ligands, whereas ErbB2 lacks a erodimers after ligand-binding (not shown). As indicated (X), ligand binding domain and functions as a preferred and ErbB2 has no ligand-binding capacity and ErbB3 has no active very potent co-receptor. ErbB3 is devoid of an active kinase domain. Only NRG1-4 and ERBB1-4 genes were investigated in this study. (Modified from [46]). kinase domain. ErbB4 is the only ErbB family member that binds all four neuregulins, as well as several proteins Page 2 of 11 (page number not for citation purposes) Behavioral and Brain Functions 2007, 3:31 http://www.behavioralandbrainfunctions.com/content/3/1/31 teria Checklist (OPKRIT). 285 of the cases were male; 111 Statistical analysis were female. The mean recruitment age was 44 yr with a All SNPs were tested for departure from Hardy Weinberg standard deviation of 14 yr. equilibrium (HWE) in controls using a chi squared test with one degree of freedom. This test compares the 1342 Controls were ascertained as a series of anonymous expected genotype frequencies calculated from the allele blood donors from Aberdeen, Scotland. Due to subject frequencies, with observed genotype frequencies. A depar- anonymity, characteristics such as schizophrenia- status, ture from HWE in controls may indicate errors in the data. age and ethnic group could not be determined. However at the time of sample collection, the Aberdeen Blood Association testing of both alleles and genotypes was car- Transfusion Service made a concerted effort to exclude any ried out using the 'PROC FREQ' fast Fisher's Exact Test non-Caucasian samples. At the time of this study, non- (FET) procedure in the statistical software package Caucasians constituted around 2% of the Aberdeen popu- SASv8.2 (SAS Institute Inc., Cary, NC, USA). The Fast lation but a lower percentage of Aberdeen blood donors. Fisher's Exact Test computes exact P-values for general 724 control subjects were male; 618 were female. contingency tables using the network algorithm devel- Informed consent was obtained for all subjects involved oped by [16]. The network algorithm provides substantial in this study. advantages over direct enumeration and is rapid and accu- rate. The exact test is used in preference to the asymptotic As calculated by the power equations of Risch and Teng chi-square P-value as it is more accurate at loci where the [13], under a multiplicative model of association, a sam- minor allele frequency is low. ple of 400 cases and 1300 controls has 80% power to detect disease association at an α = 0.05 significance level In order to model and test for interactions between SNPs for SNPs having a genotypic relative risk of 1.5 [14] and in different candidate genes, standard logistic regression susceptibility allele frequencies between 0.07 and 0.9. analysis was performed using the statistical software pack- age R (The R Project for Statistical Computing, [17]). Fol- Polymorphism selection and genotyping lowing Cordell [18] the model tested had the following SNPs from each gene were selected at an approximate form: spacing of 1 SNP every 10 kb. Six of these SNPs are located in potentially functional regions of the genes, five in log(p/(1-p)) = μ + a x + d z + a x + d z + i x x + i x 1 1 1 1 2 2 2 2 aa 1 2 ad 1 exons and one in a promoter region (Table 1). Among the z + i z x + i z z 2 da 1 2 dd 1 2 SNPs that are in exons, two are synonymous, but the other three produce amino acid changes. The size of these genes where x , x , z and z are variables determined by the gen- 1 2 1 2 varied greatly, from 23 kb to 1110 kb, and there was a cor- otype under consideration, and μ, a , a , d , d , i , i , i 1 2 1 2 aa ad da responding variation in the number of SNPs analysed, and i are parameters to be estimated. Here, genotypes at dd from 4 to 109 (for further details see Table 2, in the two SNPs are being modelled and, for each of the 9 possi- Results section). ble genotype combinations at the two SNPs, p is the pro- portion of subjects with the genotype who are cases and Genotyping was performed using a modification of the (1-p) is the proportion of subjects with the genotype who single base chain extension assay previously described are controls. The first five terms in the equation represent [15], at GlaxoSmithKline, UK. A SNP passed QC if over the effects of the two SNPs on log(p/(1-p)) if there is no 80% of genotype calls were successful. statistical interactions between the two SNPs. A non-zero value for one or more of the remaining four coefficients (i , i , i , i ) indicates the presence of interaction. Statis- aa ad da dd Table 1: SNPs in potentially functional gene regions included in the present study. 1 2 Gene Polymorphism Chromosome Position Region Base change Amino acid change EGFR rs2072454 7 54988557 Exon 4 C/T Asn/Asn ERBB2 rs1801200 17 35133114 Exon 20 A/G Val/Ile ERBB2 rs1058808 17 35137563 Exon 30 C/G Pro/Ala ERBB4 rs3748962 2 212077370 Exon 27 C/T Val/Val NRG1 rs6994992 8 31615123 Promoter C/T - NRG1 rs3924999 8 32572900 Exon 2 C/T Gln/Arg Polymorphism names are taken from the dbSNP website, [47]. From map NCBI 35. Page 3 of 11 (page number not for citation purposes) Behavioral and Brain Functions 2007, 3:31 http://www.behavioralandbrainfunctions.com/content/3/1/31 Table 2: Polymorphism in genes in the families NRG and ErbB investigated in this study. 1 1 Gene Chromosome Size (kb) No. of analysable SNPs No. of SNPs with p < 0.10 No. of SNPs with p < 0.05 NRG1 8 1120 94 22 (23%) 11 (12%) NRG2 5 196 16 2 (13%) 1 (6%) NRG3 10 1110 105 21 (20%) 10 (10%) NRG4 15 70 9 0 0 EGFR 7 188 22 7 (32%) 6 (27%) ERBB2 17 29 4 1 (25%) 0 ERBB3 12 23 6 0 0 ERBB4 2 1160 109 26 (24%) 14 (13%) Either genotypic test or allelic test of association. tical significance was assessed by a likelihood ratio test, Of 374 SNPs submitted for genotyping, 3 were dropped comparing the full model above with a reduced model in from analysis due to monomorphism in cases and con- which the last four terms are absent (i.e. in which the coef- trols, and another 6 were dropped due to departure from ficients i , i , i and i are all set to zero). Hardy-Weinberg Equilibrium (P < 0.001). Among the aa ad da dd remaining SNPs, the proportion of individuals success- A significant interaction effect in the logistic regression fully genotyped ranged from 0.804 to 0.99, with a median analysis indicates that the combined effect of the two of 0.97. A subset of the individuals studied was genotyped SNPs is significantly different from the sum of their indi- in duplicate, giving a total of 14,323 duplicate genotypes. vidual effects. Pair-wise interactions may be of three types, Of these, only 2 (= 0.014%) were inconsistent. The 365 namely 'and', 'or' and 'exclusive or', defined as follows. In SNPs retained, located in 8 genes, were subjected to sin- an 'and' relationship, the effect of having the risk-enhanc- gle-point analysis. The number of SNPs per gene is given ing genotypes at both SNP loci is larger than the sum of in Table 2. The median frequency of the minor allele the single-locus effects. In an 'or' relationship, the effect of among the SNPs analysed was 22.9 %, but there were 35 having the risk-enhancing genotype at both loci is smaller SNPs (9.6% of those analysed) for which the minor allele than the sum of the single-locus effects. In the final type frequency was less than 5%. These SNPs, selected prior to of interaction, named 'exclusive or', the effect of having the availability of the HapMap [20], are less informative the risk-enhancing genotype at both loci is smaller than than the others; however, as they do provide some addi- that of having a risk-enhancing genotype at only one tional information they were retained in the analysis. locus. For each SNP pair, each of these configurations gives rise to a contingency table, classified by genotype Single-point results group (risk-enhancing or risk-reducing) and phenotype Both allelic and genotypic tests of association were con- (case or control). These contingency tables were evalu- ducted. Detailed results are provided in Additional file 1. ated, and the configuration with the lowest P-value for Significant results (P < 0.05) were observed in five genes, association with the phenotype was identified. This con- namely NRG1, NRG2, NRG3, EGFR and ERBB4. These figuration indicated the type of interaction between the confirmed previous findings on NRG1 and ERBB4 and two loci in question. suggest possible involvement also of NRG2, NRG3 and EGFR. Borderline evidence of association (P < 0.10) was observed for one SNP in ERBB2. Haplotypic tests of asso- Results Data checking ciation (not shown) provided similar results to the single- Although controls were selected from the same geo- point results described here. If none of the 365 single graphic area as the cases, a test was conducted to deter- SNPs were associated with the disease status (the null mine the presence of any latent population stratification. hypothesis), we would expect 18 SNPs to be significant at The lambda statistic of Devlin et al. [19], calculated on the 0.05 level and 37 SNPs to be significant at the 0.10 4,250 SNPs distributed over the whole genome, showed level on an allelic test. For the allelic test there were 34 no evidence of stratification (population heterogeneity, SNPs significant at the 0.05 level (9.3%) and 59 SNPs sig- overdispersion: Genotypic exact lambda = 1.061, Allelic nificant at the 0.10 level (16.2%). For the genotypic test exact lambda = 1.000), indicating that no genomic-con- there were 24 SNPs significant at the 0.05 level (6.6%) trol adjustment is required when performing tests of gen- and 62 SNPs significant at the 0.10 level (17.0%). This otypic and allelic association on these subjects. excess of significantly associated SNPs, beyond the num- Page 4 of 11 (page number not for citation purposes) Behavioral and Brain Functions 2007, 3:31 http://www.behavioralandbrainfunctions.com/content/3/1/31 bers expected by chance when accounting for the multiple were significantly associated in NRG3, including one tests, provides evidence that some of these associations which was highly significantly associated (P < 0.01) by the are real. allelic test (RS3924461: map position 084423371 on Chr10, OR = 1.283, CI 1.077–1.529). The pattern of LD in The distribution of evidence for single-point association this gene (results not shown) is highly localised, similar to with schizophrenia, and linkage disequilibrium (LD), that in NRG1, so the observation of association from spa- over gene NRG1 are presented in Fig. 2. In this figure, the tially separated SNPs may have arisen by chance or may magnitude of association of each SNP with schizophrenia suggest the presence of different functional variants is indicated by the log-transformed P-value. For the sake within the same gene. The spatial pattern of association is of simplicity, only values arising from the genotypic exact not related to the distribution of the exons in this gene. Six test are shown; allelic test results followed a very similar SNPs in EGFR were significantly associated with schizo- pattern. Across both tests, eleven SNPs in this gene phrenia, however, all fell in the range 0.01 <P < 0.05. Only showed significant genotypic association with schizo- four SNPs were studied in ERBB2, and none of these was phrenia (P < 0.05), and in three instances the association found to be significantly associated with schizophrenia. was highly significant (P < 0.01) (see Additional file 1: SNP RS776401: map position 031836504 on Chr 8, risk Interaction analysis results genotype odds ratio (OR) = 1.461, 95% confidence inter- As summarised in Table 2, 79 SNPs were identified which val (CI) 1.160–1.840; SNP RS1383887: map position had a P-value from an allelic or genotypic test that was less 031872698, OR = 1.589, CI 1.213–2.081; SNP RS- than 0.10 (details are given in Additional file 2). These GSK8116812: map position 032741837, OR = 1.651, CI SNPs were contained in the 6 genes considered above, 1.242–2.196). A putatively functional SNP in Exon 2, cod- namely NRG1, NRG2, NRG3, EGFR, ERBB2, and ERBB4. ing a non-synonymous (Gly/Arg) change, showed allelic All pairs of these SNPs that are at least 0.5 cM apart on an association to schizophrenia with P = 0.0279 (SNP interpolated Decode genetic map [21] were tested for RS3924999: map position 032572900, OR = 1.302, CI interaction effects in a logistic regression interaction anal- 1.028–1.648). It is possible that the cluster of associations ysis. observed at the 3' end of the gene, to the right of the fig- ure, has arisen through LD with this non-synonymous A total of 2640 statistical tests for pair-wise interactions change or another underlying functional variant. Further were performed (details of results are given in Additional evidence of association is seen at the 5' end of the gene, file 2). Of these, 212 were significant at the 5% level: that closer to Exon 1. The latter associations fall beyond the is 8.03% of all tests were significant compared to 132 (= apparent reach of LD. 5% of 2640) expected by chance if the null hypothesis of no interaction were true for all pairs of SNPs tested. If all The corresponding information for gene ERBB4 is pre- the tests performed were mutually independent the prob- sented in Fig. 3. In this gene, fourteen SNPs were signifi- ability that 212 or more would be significant by chance is -11 cantly associated with schizophrenia by either the allelic 2.575 × 10 (P(R ≥ 212) where R~binomial(N = 2640, p or genotypic test, including one allelic test result which = 0.05)). The true probability of such an outcome is cer- was highly significant (P < 0.01: SNP RS6754674: map tain to be larger, due to non-independence among the position 213033762 on Chr 2, OR = 1.387, CI 1.056– tests: nevertheless, this result provides convincing evi- 1.822). Again, the graph of genotypic test results shows dence that some of the observed interactions are real. The strong evidence of association occurring most commonly significance levels presented here are not adjusted for at the two ends of the gene, particularly at the 5'-end, to multiple testing: a Bonferroni correction would be inap- the right of the figure, although the pattern is not as clear propriate due to non-independence among the tests, and as in the case of NRG1. This region of the gene represents the multiple-testing issue is addressed by the comparison a 0.15 Mb region of moderate LD, so it is possible that a of observed and expected numbers of significant tests pre- single underlying variant, in LD with all of the SNPs that sented above. None of the significant results obtained show significant association, accounts for the observed would survive the Bonferroni correction. association in this region. Table 3 gives the percentage of significant interactions for The corresponding information for genes NRG2, NRG3, each gene combination; the corresponding counts are EGFR and ERBB2 is presented in Fig. 4, except that the LD given in brackets. There are four gene combinations in maps are omitted for brevity. A single SNP in NRG2 which the percentage of significant interactions is sub- showed significant genotypic association with schizo- stantially higher than the 5% expected by chance, viz. the phrenia (SNP RS2936651: map position 139401129 on two-gene combinations NRG1-NRG2, NRG1-NRG3 and Chr 5, OR = 2.081, CI 1.156–3.745). This finding was cor- NRG2-EGFR, and the single large gene ERBB4. The per- roborated by a significant allelic test result also. Ten SNPs centages of significant interactions for combinations Page 5 of 11 (page number not for citation purposes) Behavioral and Brain Functions 2007, 3:31 http://www.behavioralandbrainfunctions.com/content/3/1/31 RefSeq Genes NRG1 31600000 31800000 32000000 32200000 32400000 32600000 Chromosome Position (bp) Genotypic single-point a Figure 2 ssociation test results and LD over NRG1 Genotypic single-point association test results and LD over NRG1. In the plot of -log (P) v. map position, horizontal dashed lines are at P = 0.05, 0.01 and 0.001. Vertical dashed lines are at 20 kb intervals. The region covered by the gene is indi- cated in the centre of the figure by a horizontal line, with an arrow head indicating the direction of transcription. A crossing vertical line indicates the position of each exon. At the base of the figure, LD between SNP loci is indicated by shading. White: 2 2 r = 0; black: r = 1. NRG1-NRG2 and NRG2-EGFR are based on a small ple of significant interactions. This also might be number of tests and should therefore be treated with cau- expected, as it corresponds to a complex biological mech- tion, but the percentages for NRG1-NRG3 and ERBB4 are anism and to a rather specific pattern of results that is rel- based on substantial numbers. atively unlikely to occur by chance. The next step towards the identification of a mechanism Discussion to account for the statistical interactions is the identifica- In this manuscript we present confirmation of earlier pub- tion of the type of each significant interaction. This was lished associations between schizophrenia and the genes performed by pair-wise combination analysis (see Meth- NRG1 and ERBB4. We build a new, compelling picture of ods). Fig. 5 illustrates the distribution of the three types of pathway involvement by showing that associations exist interaction over the four gene combinations with a large with additional gene family-members, namely EGFR, percentage of significant interactions. The frequencies of NRG2, and NRG3. Lastly, we present statistical evidence and-type and or-type interactions are fairly similar. This of gene epistasis, consistent with a highly complex biolog- might be expected, as the patterns of results that corre- ical pathway of schizophrenia susceptibility. This is the spond to these two types of interaction are complemen- first time that the roles and inter-relationships of the ERBB tary, and equally likely to occur by chance. Interactions of and NRG gene families as a whole have been investigated the xor-type are rare: indeed, they are observed only in the in relation to schizophrenia and our approach sets a prec- two gene combinations that gave rise to a fairly large sam- edent in the investigation of common complex diseases. Page 6 of 11 (page number not for citation purposes) − log (P) 0.0 1.0 2.0 3.0 Behavioral and Brain Functions 2007, 3:31 http://www.behavioralandbrainfunctions.com/content/3/1/31 RefSeq Genes ERBB4 212200000 212400000 212600000 212800000 213000000 213200000 Chromosome Position (bp) Genotypic single-point a Figure 3 ssociation test results and LD over ERBB4 Genotypic single-point association test results and LD over ERBB4. In the plot of -log (P) v. map position, horizontal dashed lines are at P = 0.05, 0.01 and 0.001. Vertical dashed lines are at 20 kb intervals. The region covered by the gene is indi- cated in the centre of the figure by a horizontal line, with an arrow head indicating the direction of transcription. A crossing vertical line indicates the position of each exon. At the base of the figure, LD between SNP loci is indicated by shading. White: 2 2 r = 0; black: r = 1. Clear evidence for gene-gene interaction was detected for NRG1 overlap with those found by Stefansson et al. [3,4], gene combinations NRG1-NRG2, NRG1-NRG3 and Corvin et al. [23], Li et al. [24], Zhao et al. [25] and Petry- EGFR-NRG2, and suggestive evidence was also seen for shen et al. [26]. Similarly, the associations we observed at ERBB4-NRG1, ERBB4-NRG2, ERBB4-NRG3 and ERBB4- the 3' end of the gene overlap with those found by Yang et ERBB2. Evidence of intragenic interaction was seen for al. [27], Li et al. [23], Petryshen et al. [25] and Lachmann SNPs in ERBB4. Of the 2640 tests for pair-wise interac- et al. [28]. Two polymorphisms in our study directly rep- tion, 212 were significant at the 5% level, which is sub- licate these associations: the putatively functional SNP stantially more than the 131 expected by chance. This rs3924999 was found to be associated with schizophrenia large number of low P-values may be partly explained by in a Han Chinese family study conducted by Yang et al. non-independence among tests. However, it may also [26]. This SNP was also tested in a Caucasian study of 136 indicate that some of the 79 SNPs studied interact to mod- families by Duan et al. [29], but was not significant. ulate susceptibility to schizophrenia. Rs2466049 was significant as part of a haplotype found in a Caucasian family by Petryshen et al. [25]. Numerous studies have investigated the association between schizophrenia and NRG1 (reviewed in [22]). A For the ERBB4 gene, four polymorphisms (rs3748962, comparison of the associations generated in the current rs839523, rs2054617 and rs6435711) screened in this study with those previously published shows good agree- study were also analysed by Silberberg et al. [8]. However, ment. Associations that we observed at the 5' region of none of the individual SNP associations observed in the Page 7 of 11 (page number not for citation purposes) − log (P) 0.0 1.0 2.0 3.0 Behavioral and Brain Functions 2007, 3:31 http://www.behavioralandbrainfunctions.com/content/3/1/31 Genotypic single-point a Figure 4 ssociation test results and LD over NRG2, NRG3, EGFR and ERBB2 Genotypic single-point association test results and LD over NRG2, NRG3, EGFR and ERBB2. In the plot of - log (P) v. map position, horizontal dashed lines are at P = 0.05, 0.01 and 0.001. Vertical dashed lines are at 20 kb intervals. The region covered by the gene is indicated in the centre of the figure by a horizontal line, with an arrow head indicating the direc- tion of transcription. A crossing vertical line indicates the position of each exon. previous study was replicated here. As no underlying func- from the NRG and ERBB gene families, i.e. NRG2, NRG3 tional variant has been identified for ERBB4, this lack of and EGFR (ERBB1). It is interesting to note that we only association at the SNP level is unsurprising and is consist- detected evidence for association of NRG genes which are ent with a heterogeneous haplotypic background across expressed in the central nervous system, namely NRG1, studies. It is worth noting also that non-coding polymor- NRG2 and the nervous-system-specific NRG3 [31], phisms might lead to dysregulation of ErbB4 expression whereas NRG4, which is mainly expressed in pancreas levels, a hypothesis corroborated by the Silberberg et al. and absent from the brain [32] was not found to be asso- [8] study, which found enhanced expression of the ERBB4 ciated to schizophrenia. Jm-a CYT-1 isoform (one of four splice variants, see [30]) in schizophrenic patients. The altered balance of ErbB4 To our knowledge, an association of EGFR to schizophre- isoforms could lead to variations in homo- and het- nia has not previously been investigated. Association of a erodimer formation and thus in downstream signalling polymorphism in EGF (one of the six EGFR ligands, Fig. pathways. 1) with schizophrenia was reported in Finnish men [33], but not replicated in a Japanese population [34]. Further In addition to providing further evidence of NRG1 and support for a possible role of EGF and EGFR in the disease ERBB4 associations to schizophrenia, our study is the first comes from findings of abnormal expression of these to suggest possible involvement of three additional genes genes in the forebrain and serum of schizophrenic Page 8 of 11 (page number not for citation purposes) Behavioral and Brain Functions 2007, 3:31 http://www.behavioralandbrainfunctions.com/content/3/1/31 Table 3: Percentages of pair-wise interactions that were significant (P < 0.05), classified by the gene(s) in which the interacting SNPs lie. Gene combinations in which there are substantially more significant interactions than the 5% expected by chance are indicated in bold. The counts that contribute to each percentage are also shown in parentheses as a fraction. The presence of "-" in the table indicates that no pairs of SNPs in this gene are 0.5 cM apart. NRG1 NRG2 NRG3 EGFR ERBB2 ERBB4 NRG1 5.8% (6/103) 13.6% (6/44) 11.7% (54/462) 4.5% (7/154) 0.0% (0/22) 7.9% (45/572) NRG2 - 4.8% (2/42) 14.3% (2/14) 0.0% (0/2) 7.7% (4/52) NRG3 5.7% (6/105) 2.0% (3/147) 0.0% (0/21) 8.4% (46/546) EGFR - 0.0% (0/7) 5.5% (10/182) ERBB2 -7.7% (2/26) ERBB4 13.7% (19/139) patients [35]. EGFR is usually omitted in reviews on NRG- receptor. The NRG2-EGFR interaction is superficially ERBB signalling, although it is well-established that ErbB4 more difficult to explain since NRG2 is a ligand for ErbB3 can interact with and cross-phosphorylate EGFR upon and ErbB4 and does not usually recognize EGFR directly neuregulin-binding [36-39]. Furthermore, EGFR and or with high affinity [42-44]. However, a recent study ERBB4 show partially overlapping temporal and spatial reports that a single amino acid change in the EGFR lig- mRNA expression in several brain areas, such as cortex, and-binding domain dramatically increases the receptor striatum, cerebellum and hippocampus, and are coex- affinity for NRG2β which can then induce potent stimula- pressed in GABAergic interneurons [40,41]. Therefore, it is tion of EGFR signalling [45]. The interactions within the very likely that EGFR/ErbB4 heterodimers mediate impor- ERBB4 gene suggest that combinations of particular geno- tant functions in the developing and adult central nervous types may result in modified receptor activity, leading to system in vivo. Whilst we limited this study to the NRG altered risk of schizophrenia. and ERBB gene families, it would be interesting to investi- gate association of the additional EGFR and ErbB4 ligands Such suggestions concerning possible mechanisms of (see Fig. 1). interaction remain speculative until the reality of these interactions has been established biologically and the The significant gene-gene interactions reported in this effect of each two-locus genotype on the phenotype is study support our hypothesis of pathway involvement in understood. A follow-up study currently underway, using schizophrenia. These genes have related functions and, for independent cases and controls, will clarify which of these example, the NRG1-NRG2 and NRG1-NRG3 interactions interactions can be replicated and may generate further may suggest competition between these ligands for a hypotheses regarding the nature of the interactions. It will be interesting also to investigate whether observed associ- ations or their interactions correspond to particular phe- notypically defined patient sub-groups. The strong evidence of genetic associations and interactions reported here provides ample justification for such an enquiry. Conclusion This study confirms previously published associations between schizophrenia and NRG1 and ERBB4, and points to three other members of the NRG and ERBB gene fami- lies (EGFR, NRG2 and NRG3) as potential schizophrenia susceptibility genes. Furthermore, the study presents evi- dence for inter- and intragenic pair-wise interaction between loci in genes of both families (NRG1-NRG2, NRG2-EGFR, NRG1-NRG3 and ERBB4-ERBB4). These Distribution combinatio tions Figure 5 ns wi of the types of th a large perc pair-wise interaction in gene entage of significant interac- extensive genetic inter-relationships observed among lig- Distribution of the types of pair-wise interaction in ands and receptors of the ERBB and NRG gene families gene combinations with a large percentage of signifi- reflect the complexity of their interactions at the molecu- cant interactions. The total in each column corresponds lar level. The influence of so many closely related genetic to the number of significant interactions presented in Table factors on the potential development of disease patho- 3, i.e. the numerator of the fraction presented in that table for the gene combination in question. physiology has rarely, if ever, been demonstrated before and may partially explain the difficulty in arriving at an Page 9 of 11 (page number not for citation purposes) Behavioral and Brain Functions 2007, 3:31 http://www.behavioralandbrainfunctions.com/content/3/1/31 overarching genetic hypothesis of the etiology of schizo- Acknowledgements The authors wish to acknowledge the patients and blood donors who con- phrenia. In future studies, it will be important to replicate tributed to this study. Thanks are also due to members of the GSK Genetic our new findings in an independent cohort and to under- Sample Management and Genotyping groups; also Matt Nelson, Pierandrea stand the underlying biological mechanisms of the single- Muglia, Paul Skelding and Nicola Foote for their helpful contributions in the locus associations as well as the interactions among differ- writing of this paper. ent loci. References Competing interests 1. Owen MJ, Craddock N, O'Donovan MC: Schizophrenia: genes at last? [Review]. Trends in Genetics 2005, 21:518-525. The author(s) declare that they have no competing inter- 2. Harrison PJ, Weinberger DR: Schizophrenia genes, gene expres- ests. sion, and neuropathology: on the matter of their conver- gence [Review]. Molecular Psychiatry 2005, 10:40-68. 3. Stefansson H, Sigurdsson E, Steinthorsdottir V, Bjornsdottir S, Sig- Authors' contributions mundsson T, Ghosh S, Brynjolfsson J, Gunnarsdottir S, Ivarsson O, IB helped to formulate the study and identified the genes Chou TT, Hjaltason O, Birgisdottir B, Jonsson H, Gudnadottir VG, of interest; she led the writing of the manuscript and pro- Gudmundsdottir E, Bjornsson A, Ingvarsson B, Ingason A, Sigfusson S, Hardardottir H, Harvey RP, Lai D, Zhou M, Brunner D, Mutel V, vided the biological interpretation and context to our Gonzalo A, Lemke G, Sainz J, Johannesson G, Andresson T, Gudb- results. AB led the statistical analysis, and produced the jartsson D, Manolescu A, Frigge ML, Gurney ME, Kong A, Gulcher JR, Petursson H, Stefansson K: Neuregulin 1 and susceptibility to first draft of the manuscript. BB performed the single- schizophrenia. Am J Hum Genet 2002, 71:877-892. point statistical analyses and the pairwise analysis, and 4. Stefansson H, Sarginson J, Kong A, Yates P, Steinthorsdottir V, Gud- provided interpretation of his results. NG provided inter- finnsson E, Gunnarsdottir S, Walker N, Petursson H, Crombie C, Ingason A, Gulcher JR, Stefansson K, St Clair D: Association of neu- pretation of the linkage disequilibrium analysis and sub- regulin 1 with schizophrenia confirmed in a Scottish popula- stantially contributed to the interpretation of the tion. Am J Hum Genet 2003, 72:83-87. 5. Tosato S, Dazzan P, Collier D: Association between the neureg- statistical analyses in the manuscript. PM was instrumen- ulin 1 gene and schizophrenia: a systematic review. Schizophr tal in progressing the study, revised the manuscript and Bull 2005, 31:613-617. added important intellectual content concerning the biol- 6. Harrison PJ, Law AJ: Neuregulin 1 and Schizophrenia: Genetics, Gene Expression, and Neurobiology. Biol Psychiatry 2006, ogy of schizophrenia. RM designed the statistical analysis 60:132-140. plan and provided schizophrenia genetics expertise to the 7. Norton N, Moskvina V, Morris DW, Bray NJ, Zammit S, Williams writing of the manuscript. DS designed the laboratory NM, Williams HJ, Preece AC, Dwyer S, Wilkinson JC, Spurlock G, Kirov G, Buckland P, Waddington JL, Gill M, Corvin AP, Owen MJ, aspects of the experiment and oversaw their execution. O'Donovan MC: Evidence that interaction between neuregu- DSC recruited, worked up clinically and supplied DNA lin 1 and its receptor erbB4 increases susceptibility to schiz- ophrenia. Am J Med Genet B Neuropsychiatr Genet 2006, 141:96-101. from the Aberdeen schizophrenia case collection of 396 8. Silberberg G, Darvasi A, Pinkas-Kramarski R, Navon R: The involve- cases. PY provided the control collection of 1342 anony- ment of ErbB4 with schizophrenia: association and expres- mous blood donors from Aberdeen. IP helped to formu- sion studies. Am J Med Genet B Neuropsychiatr Genet 2006, 141:142-148. late the original idea for the study and revised the 9. Nicodemus KK, Luna A, Vakkalanka R, Goldberg T, Egan M, Straub manuscript critically, providing important intellectual RE, Weinberger DR: Further evidence for association between ErbB4 and schizophrenia and influence on cognitive inter- content. All authors read, edited and approved the final mediate phenotypes in healthy controls. Molecular Psychiatry manuscript. 2006, 11:1062-1065. 10. Falls DL: Neuregulins: functions, forms, and signaling strate- gies. Exp Cell Res 2003, 284:14-30. Additional material 11. Talmage DA, Role LW: Multiple personalities of neuregulin gene family members. J Comp Neurol 2004, 472:134-139. 12. Yarden Y, Sliwkowski MX: Untangling the ErbB signalling net- Additional file 1 work. Nat Rev Mol Cell Biol 2001, 2:127-137. Detailed results from allelic and genotypic tests of single-point association 13. Risch N, Teng J: The relative power of family-based and case- control designs for linkage disequilibrium studies of complex of SNP loci with schizophrenia. human diseases - I. DNA pooling. Genome Research 1998, Click here for file 8:1273-1288. [http://www.biomedcentral.com/content/supplementary/1744- 14. Risch N, Merikangas K: The future of genetic studies of complex 9081-3-31-S1.xls] human diseases. Science 1996, 273:1516-1517. 15. Taylor JD, Briley D, Nguyen Q, Long K, Iannone MA, Li MS, Ye F, Afshari A, Lai E, Wagner M, Chen J, Weiner MP: Flow cytometric Additional file 2 platform for high-throughput single nucleotide polymor- Detailed results from allelic and genotypic tests of association between phism analysis. Biotechniques 2001, 30:661-669. SNP-pairs and susceptibility to schizophrenia. 16. CR M, NR P: A network algorithm for performing Fisher's Click here for file Exact Test in r ×c contingency tables. J Am Stat Assoc 1983, 78:427-434. [http://www.biomedcentral.com/content/supplementary/1744- 17. The R Project for Statistical Computing 2003 [http://www.r- 9081-3-31-S2.xls] project.org/]. 18. Cordell HJ: Epistasis: what it means, what it doesn't mean, and statistical methods to detect it in humans. Hum Mol Genet 2002, 11:2463-2468. Page 10 of 11 (page number not for citation purposes) Behavioral and Brain Functions 2007, 3:31 http://www.behavioralandbrainfunctions.com/content/3/1/31 19. Devlin B, Roeder K, Wasserman L: Genomic control, a new 38. Riese DJ, van Raaij TM, Plowman GD, Andrews GC, Stern DF: The approach to genetic-based association studies. Theoretical Pop- cellular response to neuregulins is governed by complex ulation Biology 2001, 60:155-166. interactions of the erbB receptor family. Mol Cell Biol 1995, 20. Altshuler D, Brooks LD, Chakravarti A, Collins FS, Daly MJ, Donnelly 15:5770-5776. P: A haplotype map of the human genome. Nature 2005, 39. Zhang K, Sun J, Liu N, Wen D, Chang D, Thomason A, Yoshinaga SK: 437:1299-1320. Transformation of NIH 3T3 cells by HER3 or HER4 recep- 21. Kong A, Gudbjartsson DF, Sainz J, Jonsdottir GM, Gudjonsson SA, tors requires the presence of HER1 or HER2. J Biol Chem 1996, Richardsson B, Sigurdardottir S, Barnard J, Hallbeck B, Masson G, Shl- 271:3884-3890. ien A, Palsson ST, Frigge ML, Thorgeirsson TE, Gulcher JR, Stefansson 40. Fox IJ, Kornblum HI: Developmental profile of ErbB receptors K: A high-resolution recombination map of the human in murine central nervous system: implications for functional genome. Nat Genet 2002, 31:241-247. interactions. J Neurosci Res 2005, 79:584-597. 22. Munafo MR, Thiselton DL, Clark TG, Flint J: Association of the 41. Kornblum HI, Gall CM, Seroogy KB, Lauterborn JC: A subpopula- NRG1 gene and schizophrenia: a meta-analysis. Mol Psychiatry tion of striatal gabaergic neurons expresses the epidermal 2006, 11:613. growth factor receptor. Neuroscience 1995, 69:1025-1029. 23. Corvin AP, Morris DW, McGhee K, Schwaiger S, Scully P, Quinn J, 42. Carpenter G: ErbB-4: mechanism of action and biology. Exp Meagher D, Clair DS, Waddington JL, Gill M: Confirmation and Cell Res 2003, 284:66-77. refinement of an 'at-risk' haplotype for schizophrenia sug- 43. Hobbs SS, Coffing SL, Le AT, Cameron EM, Williams EE, Andrew M, gests the EST cluster, Hs.97362, as a potential susceptibility Blommel EN, Hammer RP, Chang H, Riese DJ: Neuregulin iso- gene at the Neuregulin-1 locus. Mol Psychiatry 2004, 9:208-213. forms exhibit distinct patterns of ErbB family receptor acti- 24. Li T, Stefansson H, Gudfinnsson E, Cai G, Liu X, Murray RM, vation. Oncogene 2002, 21:8442-8452. Steinthorsdottir V, Januel D, Gudnadottir VG, Petursson H, Ingason 44. Pinkas-Kramarski R, Shelly M, Guarino BC, Wang LM, Lyass L, Alroy A, Gulcher JR, Stefansson K, Collier DA: Identification of a novel I, Alimandi M, Kuo A, Moyer JD, Lavi S, Eisenstein M, Ratzkin BJ, Seger neuregulin 1 at-risk haplotype in Han schizophrenia Chinese R, Bacus SS, Pierce JH, Andrews GC, Yarden Y: ErbB tyrosine patients, but no association with the Icelandic/Scottish risk kinases and the two neuregulin families constitute a ligand- haplotype. Mol Psychiatry 2004, 9:698-704. receptor network. Mol Cell Biol 1998, 18:6090-6101. 25. Zhao X, Shi Y, Tang J, Tang R, Yu L, Gu N, Feng G, Zhu S, Liu H, Xing 45. Gilmore JL, Gallo RM, Riese DJ: The epidermal growth factor Y, Zhao S, Sang H, Guan Y, St Clair D, He L: A case control and receptor (EGFR)-S442F mutant displays increased affinity family based association study of the neuregulin1 gene and for neuregulin-2beta and agonist-independent coupling with schizophrenia. J Med Genet 2004, 41:31-34. downstream signalling events. Biochem J 2006, 396:79-88. 26. Petryshen TL, Middleton FA, Kirby A, Aldinger KA, Purcell S, Tahl AR, 46. Olayioye MA, Neve RM, Lane HA, Hynes NE: The ErbB signaling Morley CP, McGann L, Gentile KL, Rockwell GN, Medeiros HM, Car- network: receptor heterodimerization in development and valho C, Macedo A, Dourado A, Valente J, Ferreira CP, Patterson NJ, cancer. EMBO J 2000, 19:3159-3167. Azevedo MH, Daly MJ, Pato CN, Pato MT, Sklar P: Support for 47. NCBI dbSNP website 2004 [http://www.ncbi.nlm.nih.gov/SNP/]. involvement of neuregulin 1 in schizophrenia pathophysiol- ogy. Mol Psychiatry 2005, 10:366-74, 328. 27. Yang JZ, Si TM, Ruan Y, Ling YS, Han YH, Wang XL, Zhou M, Zhang HY, Kong QM, Liu C, Zhang DR, Yu YQ, Liu SZ, Ju GZ, Shu L, Ma DL, Zhang D: Association study of neuregulin 1 gene with schizo- phrenia. Mol Psychiatry 2003, 8:706-709. 28. Lachman HM, Pedrosa E, Nolan KA, Glass M, Ye K, Saito T: Analysis of polymorphisms in AT-rich domains of neuregulin 1 gene in schizophrenia. Am J Med Genet B Neuropsychiatr Genet 2006, 141:102-109. 29. Duan J, Martinez M, Sanders AR, Hou C, Krasner AJ, Schwartz DB, Gejman PV: Neuregulin 1 (NRG1 ) and schizophrenia: analysis of a US family sample and the evidence in the balance. Psychol Med 2005, 35:1599-1610. 30. Junttila TT, Sundvall M, Maatta JA, Elenius K: Erbb4 and its iso- forms: selective regulation of growth factor responses by naturally occurring receptor variants. Trends Cardiovasc Med 2000, 10:304-310. 31. Zhang D, Sliwkowski MX, Mark M, Frantz G, Akita R, Sun Y, Hillan K, Crowley C, Brush J, Godowski PJ: Neuregulin-3 (NRG3): a novel neural tissue-enriched protein that binds and activates ErbB4. Proc Natl Acad Sci U S A 1997, 94:9562-9567. 32. Harari D, Tzahar E, Romano J, Shelly M, Pierce JH, Andrews GC, Yarden Y: Neuregulin-4: a novel growth factor that acts through the ErbB-4 receptor tyrosine kinase. Oncogene 1999, 18:2681-2689. 33. Anttila S, Illi A, Kampman O, Mattila KM, Lehtimaki T, Leinonen E: Association of EGF polymorphism with schizophrenia in Publish with Bio Med Central and every Finnish men. Neuroreport 2004, 15:1215-1218. scientist can read your work free of charge 34. Watanabe Y, Fukui N, Muratake T, Kaneko N, Someya T: No asso- ciation of EGF polymorphism with schizophrenia in a Japa- "BioMed Central will be the most significant development for nese population. Neuroreport 2005, 16:403-405. disseminating the results of biomedical researc h in our lifetime." 35. Futamura T, Toyooka K, Iritani S, Niizato K, Nakamura R, Tsuchiya K, Sir Paul Nurse, Cancer Research UK Someya T, Kakita A, Takahashi H, Nawa H: Abnormal expression of epidermal growth factor and its receptor in the forebrain Your research papers will be: and serum of schizophrenic patients. Mol Psychiatry 2002, available free of charge to the entire biomedical community 7:673-682. 36. Cohen BD, Green JM, Foy L, Fell HP: HER4-mediated biological peer reviewed and published immediately upon acceptance and biochemical properties in NIH 3T3 cells. Evidence for cited in PubMed and archived on PubMed Central HER1-HER4 heterodimers. J Biol Chem 1996, 271:4813-4818. 37. Hatakeyama M, Yumoto N, Yu X, Shirouzu M, Yokoyama S, Konagaya yours — you keep the copyright A: Transformation potency of ErbB heterodimer signaling is BioMedcentral determined by B-Raf kinase. Oncogene 2004, 23:5023-5031. Submit your manuscript here: http://www.biomedcentral.com/info/publishing_adv.asp Page 11 of 11 (page number not for citation purposes)

Journal

Behavioral and Brain Functions – Springer Journals

Published: Jun 28, 2007

There are no references for this article.