Molecular Psychiatry (2007) 12, 273–282 & 2007 Nature Publishing Group All rights reserved 1359-4184/07 $30.00 www.nature.com/mp ORIGINAL ARTICLE Interleukin 3 and schizophrenia: the impact of sex and family history X Chen1, X Wang1, S Hossain1, FA O’Neill2, D Walsh3, E van den Oord1, A Fanous4 and KS Kendler1 1Department of Psychiatry and Virginia Institute for Psychiatric and Behavior Genetics, Virginia Commonwealth University, Richmond, VA, USA; 2The Department of Psychiatry, The Queens University, Belfast, UK; 3The Health Research Board, Dublin, UK and 4Washington VA Medical Center-Georgetown University Medical Center, Schizophrenia Research Program, Washington, DC, USA

Chromosome 5q21–33 has been implicated in harboring risk for schizophrenia. In this paper, we report evidence that multiple single nucleotide polymorphisms in and around interleukin 3 (IL3) are associated with the disease in the Irish Study of High-Density Schizophrenia Families (ISHDSF), the Irish Case–Control Study of Schizophrenia (ICCSS) and the Irish Trio Study of Schizophrenia (ITRIO). The associations are sex-specific and depend on the family history (FH) of schizophrenia. In all three samples, rs31400 shows female-specific and FH-dependent associations (P = 0.0062, 0.0647 and 0.0284 for the ISHDSF, ICCSS and ITRIO, respectively). Several markers have similar associations in one or two of the three samples. In haplotype analyses, identical risk and protective haplotypes are identified in the ISHDSF and ITRIO samples in several multimarker combinations. For ICCSS, the same haplotypes are implicated; however, the risk haplotypes observed in the family samples become protective. Several significant markers, rs440970, rs31400 and rs2069803, are located in and around known estrogen response elements, promoter and enhancer of the IL3 . They may explain the sex-specific associations and be functional for the expression of IL3 gene. Molecular Psychiatry (2007) 12, 273–282. doi:10.1038/sj.mp.4001932; published online 19 December 2006 Keywords: linkage disequilibrium test; association; interleukin 3; sex; family history; schizo- phrenia

Introduction Immediately telomeric to the ACSL6 gene is inter- leukin 3 (IL3) gene. The ACSL6 and IL3 genes are Schizophrenia is a complex disorder in which both arranged in a head-to-head orientation, sharing a genetic and environmental factors are involved. potential regulatory region of about 46 kb. In our fine- Multiple linkage studies have implicated chromo- mapping, some single nucleotide polymorphisms some 5q21–33 as a region likely to harbor risk (SNPs) in this 46 kb interval showed signs of associa- genes,1,2 including the study of Irish Study of High tion. From the literature, multiple regulatory ele- Density Schizophrenia Families (ISHDSF).2,3 For the ments, including 4 estrogen response elements last few years, significant progress has been made in (EREs), have been identified in this 46 kb interval.7,8 the field, such as the identification of the dysbindin In an effort to define the boundary of association gene in 6p,4 the neuregulin gene in 1 in 8p5 and the signals for the ACSL6 gene, we extended our search to TRACE4 gene in 10p.6 include this 46 kb regulatory region and IL3 gene, We started fine-mapping of the 5q22–31 linkage hoping to exclude IL3 from the list of genes associated peak in the ISHDSF several years ago and found a with the disease in our broad linkage region. To our broad region showing consistent associations with surprise, we found that several SNPs in the well- schizophrenia. This region covers several genes and studied and well-defined IL3 promoter and enhancer spans about one million base pairs of genomic were associated with schizophrenia. These poten- sequence. One of the genes showing associations is tially functional associations in IL3 promoter and ACSL6 (acyl-CoA synthetase long-chain member 6). enhancer prompted us to study more carefully of the IL3 gene and its involvement in schizophrenia. For Correspondence: Dr X Chen, Department of Psychiatry and this reason, we used two more independent samples Virginia Institute for Psychiatric and Behavior Genetics, Virginia to verify the results obtained in our fine-mapping Commonwealth University, 800 E. Leigh Street, Richmond, VA sample. In this article, we report the evidence of 23298, USA. E-mail: [email protected] sex-specific, family-history-dependent association in Received 13 March 2006; revised 24 October 2006; accepted 29 this promoter region and its potential connection to October 2006; published online 19 December 2006 the IL3 gene. Interleukin 3 and schizophrenia X Chen et al 274 IL3 is one of the cytokines that has multiple was confirmed by a blind expert diagnostic review. biological functions.9 In addition to its roles in Controls, selected from several sources, including hematopoiesis, IL3 is involved in neural develop- blood donation centers, were included if they denied ment.10–12 Of particular interest is that IL3 can a lifetime history of schizophrenia. Both cases and stimulate cell growth and suppress apoptosis.13 controls were included only if they reported all four Dysfunction of apoptosis may be involved in the grandparents as being born in Ireland or the United pathophysiology of schizophrenia.14 Abnormal activ- Kingdom. Family history (FH), based on the family- ities of several interleukins,15,16 including IL3,17 have history research diagnostic criteria,20 was assessed by been observed in schizophrenia patients. All these clinical interview of probands and their relatives. lines of evidence suggest that IL3 is a plausible Subjects having a first or second-degree relative functional candidate for schizophrenia. diagnosed with schizophrenia were classified as FH positive. In the ICCSS sample, there were 117 subjects (71 males and 46 females) classified with positive FH Materials and methods of schizophrenia. The ISHDSF sample The ISHDSF was collected in Northern Ireland, The Irish Trio Study of Schizophrenia sample United Kingdom and the Republic of Ireland. Pheno- The Irish trios (Irish Trio Study of Schizophrenia types were assessed using estrogen response element (ITRIO)) were collected in the same period as the (DSM)-III-R. The diagnoses were formed into a ICCSS sample and the inclusion criteria and symptom hierarchy of 10 categories reflecting the probable ascertainments were the same as that of ICCSS genetic relationship of these syndromes to classic sample. The only difference was that for the trios, schizophrenia. This hierarchy consisted of three both parents of the proband were included. In ITRIO, definitions of affection: (i) narrow – categories D1 there were 187 families. Of these families, 26 families and D2, or ‘core schizophrenia’ – schizophrenia, poor- had two affected subjects, and two families had three outcome schizoaffective disorder and simple schizo- affected subjects, the remainder having a single phrenia; (ii) intermediate – categories D1–D5, or a affected subject. In addition, there were 29 subjects narrow definition of the schizophrenia spectrum, who had another first or second-degree relative adding to the narrow definition schizotypal person- diagnosed of schizophrenia. The total number of ality disorder, schizophreniform disorder, delusional probands with positive FH was therefore 87 (57 disorder, atypical psychosis and good-outcome schi- families, 65 males, 22 females). zoaffective disorder; (iii) broad – categories D1–D8, including all disorders that significantly aggregated in Marker selection and genotyping relatives of schizophrenic probands in the Roscom- We used the HapMap data and the available assays mon Family Study18 and adding to the intermediate developed by Applied BioSystems (Foster City, CA, definition mood incongruent and mood congruent USA) to assist our selection of markers. We elected psychotic affective illness, and paranoid, avoidant to use SNPs that tag major haplotypes (with fre- and schizoid personality disorder. The final inclusion quency > 1%) in the region. Rs3846726 and rs3916441 criteria for pedigrees in the ISHDSF sample required were included to give even spacing of markers. We two or more first, second or third-degree relatives found a non-synonymous SNP (rs40401) in the IL3 with a diagnosis of D1–D5, one or more of whom had gene, so we included it as well. Of the 12 markers a D1–D2 diagnosis. The sample contained 273 used, three (rs399714, rs77938 and 440970) were in pedigrees and about 1350 subjects had DNA sample the ACSL6 gene, five (rs246999, rs3914025, for genotyping. Of them, 515 were diagnosed with the rs3846726, rs3916441 and rs31400) were located narrow definition (351 males and 164 females), 634 between the coding sequences of ACSL6 and IL3, were diagnosed with the intermediate definition and four (rs31480, rs40401, rs31481 and rs2069803) and 686 were diagnosed with the broad definition. were in the IL3 gene (Figure 1a). Detail descriptions of the sample were published We used two techniques for SNP genotyping. previously.19 A majority of SNPs were typed by the TaqMan assay.21 SNPs typed by this method were either The Irish Case–Control Study of Schizophrenia sample validated assays or custom-designed assays devel- The Irish Case–Control Study of Schizophrenia oped by Applied BioSystems Incorporated (Foster (ICCSS) sample was collected in the same geographic City, CA, USA). We also used the FP-TDI protocol 22,23 regions as that of the ISHDSF sample. In this study, to type some SNPs. For the FP-TDI procedures, we used a subset of 657 affected subjects (436 males DNA sequences of SNPs obtained from dbSNP and 221 females) and 411 controls (233 males and 178 (http://www.ncbi.nlm.nih.gov/SNP/index.html) were females) from this study. The affected subjects were masked by the RepeatMasker program24 and poly- selected from in-patient and out-patient psychiatric merase chain reaction (PCR) and template-directed facilities in the Republic of Ireland and Northern dye incorporation assay with fluorescence polariza- Ireland. Subjects were eligible for inclusion if they tion detection (FP-TDI) extension primers were had a diagnosis of schizophrenia or poor-outcome designed by the PRIMER 3 program. We used the schizoaffective disorder by DSM-III-R criteria, which FP-TDI genotyping products from the Perkin Elmer

Molecular Psychiatry Interleukin 3 and schizophrenia X Chen et al 275

Figure 1 Genomic structure in the region and the comparison of pairwise LD between the ISHDSF and ICCSS samples. The LDs were computed by the HAPLOVIEW program. (a) Genomic structure of the interval between the ACSL6 and IL3 genes. There are one segmental duplication (WSSD 3284) and four EREs identified in this interval. The two genes share a regulatory interval of 46 kb and are transcribed in opposite directions. (b) Pairwise LD of the ISHDSF sample. (c) Pairwise LD of the ICCSS sample. (d) Pairwise LD of the affected female subjects with FH from the ICCSS sample. (e) Pairwise LD of the unaffected female subjects from the ICCSS sample.

Corporation (Boston, MA, USA) and followed the based on weighing all families equally (the ave option recommended procedures for PCR and single base in the program). In multilocus haplotype analyses, we extension. All markers typed were checked for used 10 restarts for the expectation–maximization deviation from the Hardy–Weinberg equilibrium algorithm28 and used 1% as the cutoff for minor (HWE) and Mendelian errors by the PEDSTATS haplotypes. For the case–control sample, w2 tests were program.25 performed for each SNP for allelic associations. For haplotype analyses, the COCAPHASE module of the Statistical analyses UNPHASED program27 was used to analyze multi- We used the pedigree disequilibrium test (PDT)26 as marker haplotype associations. As in the family implemented in the UNPHASED27 program sample, haplotypes with frequencies less than 1% (PDTPHASE module) to analyze the ISHDSF sample. were aggregated. For all tests, global and individual For each typed SNP, the PDT program was run for the haplotype tests were performed simultaneously and narrow, intermediate and broad disease definitions. In P-values obtained were w2 distributions. For the these analyses, both vertical and horizontal transmis- ITRIO, we used TDTPHASE module of the UN- sions were included. The P-values reported were PHASED program as there were no unaffected sibling

Molecular Psychiatry Interleukin 3 and schizophrenia X Chen et al 276 in the families. The P-values reported for the ISHDSF Rs 40401, which changes the serine to proline at the sample were obtained with permutations (1000 27th position of IL3 protein, was not significant. Of all permutations for the individual markers and 5000 markers typed, only rs31400 showed a significant permutations for multi-marker haplotype analyses) deviation from HWE (P = 0.0099), with heterozygotes and those for the ICCSS and ITRIO samples were observed less than expected. uncorrected. We report the permutated P-values for We carried out sliding window multimaker ana- the ISHDSF because almost all markers used in this lyses. There were two regions (markers 2–5 and 7–11) study had strong linkage disequilibrium (LD) with that produced significant results in the ISHDSF each other (see Figure 1b) and all phenotypes tested sample. In the region covered by markers 2–5, the were hierarchically organized, that is, the broad best associations were observed for marker combina- disease definition included the intermediate and the tions spanning LD blocks 1 and 2 (see Figure 1b). For narrow definitions, and the intermediate definition example, the P-values for the window for combina- included the narrow definition. In sex-conditioned tion 3–4 were 0.0028, 0.0036 and 0.0154 for the analyses, only male or female subjects (ICCSS) or narrow, intermediate and broad diagnoses, respec- affected offspring (ISHDSF and ITRIO) were used. For tively. Here, the significant haplotype was protective. FH-conditioned analyses, all subjects without posi- The major haplotype 1–1 or A-G was undertrans- tive FH were set to unknown affection status. We used mitted to affected subjects. For the region covered by the HAPLOVIEW program29 to estimate pairwise LD markers 7–11, the strongest associations were ob- and to illustrate haplotype blocks. The haplotype served for marker combinations spanning LD blocks 2 blocks were partitioned by the confidence interval and 3. In this region, significant risk haplotypes were algorithm.30 observed. Marker combination 8–9 (rs31400–rs31480) produced a global P-value of 0.0004 for the inter- mediate definition and the haplotype overtransmitted Results to the affected subjects was 2–1, or T-C, with P-values Association analyses of the ISHDSF sample of 0.0088, 0.0060 and 0.0162 for the narrow, inter- In single-marker association analyses, six of the 12 mediate and broad definitions, respectively. By markers listed in Table 1 reached 5% significance combining the information obtained from the sliding level with at least one diagnostic definition. Of the six window analyses, the risk haplotype appeared to significant markers, four (rs3914025, rs3846726, extend from markers 2 to 11. Other marker combina- rs3916441 and rs31400) are located between ACSL6 tions that included markers 8 and 9 (rs31400 and and IL3 genes. One marker, rs399714, is located in the rs31480) all produced significant P-values (see Table ACSL6 gene and the other marker, rs2069803, is 2). Protective haplotypes were always more signifi- located on the 30 end of the IL3 gene. In addition, cant than the corresponding risk haplotypes. rs440970 approached significance at the 5% level. (Of Inspired by the identification of EREs – a signature note, the theoretical w2 distribution P-values obtained of estrogen regulated genes – in the region,7 we for all single-marker tests were comparable with carried out sex-conditioned analyses. In the single- the permuted P-values, data not shown.) Of these marker analyses, we found that the significant markers, three (rs31400, rs31480 and rs2069803) markers were the same markers as those identified are located in known IL3 promoters and enhancers.8 in the standard tests (compare Tables 1 and 3). The

Table 1 Marker characteristics and single marker association (P-values) in the ISHDSF

Abbreviations: ISHDSF, Irish Study of High-Density Schizophrenia Families; PDT, pedigree disequilibrium test. Significant P-values are shaded.

Molecular Psychiatry Interleukin 3 and schizophrenia X Chen et al 277 Table 2 Haplotype analyses of the ISHDSF sample (PDT ave)

Abbreviations: ISHDSF, Irish Study of High-Density Schizophrenia Families; PDT, pedigree disequilibrium test. Significant P-values are shaded.

Table 3 Sex-specific association (P-values) in the ISHDSF, ICCSS and ITRIO samples

Abbreviations: FH, family history; ICCSS, Irish Case–Control Study of Schizophrenia; ISHDSF, Irish Study of High-Density Schizophrenia Families; ITRIO, Irish Case–Control Study of Schizophrenia; PDT, pedigree disequilibrium test. Significant P-values are shaded. differences were that the results from sex-conditioned Association analyses of the ICCSS and ITRIO samples analyses were more significant. These results sug- To verify the findings from the ISHDSF, we typed the gested that almost all association signals were derived same 12 SNPs in the ICCSS and ITRIO samples. For from female offspring. Sex-conditioned analyses with all markers typed, no HWE deviations were observed. the intermediate and broad definitions produced In standard analyses with all subjects, we found no similar results, but the overall signals were weaker significant results. As the most important difference (data not shown). We also performed sex-conditioned between the ISHDSF and the ICCSS and ITRIO was analyses for multimaker haplotypes. As was observed that ISHDSF had multiple affected individuals in in the single-marker analyses, the association signals each family, or FH with schizophrenia, we selected were largely from the females as well (see Table 4). only those subjects with FH to perform the compar- One exception was combination 5–6–7, where both ison. The results for single-marker analyses were males and females were significant. With this combi- summarized in Table 3. Despite our power was only nation, the risk haplotype in the males was protective moderate owing to reduced sample sizes (there were in the females and vice versa. This explained why we 87 affected subjects with FH in ITRIO and 117 did not see any signals for this combination when subjects with FH in the ICCSS), multiple markers both males and females were analyzed together reached or approached nominal significance in both (compare Tables 2 and 4). the ICCSS and ITRIO samples. Like what was

Molecular Psychiatry Interleukin 3 and schizophrenia X Chen et al 278 Table 4 Haplotype analyses of the ISHDSF, ICCSS and ITRIO samples

Abbreviations: FH, family history; ICCSS, Irish Case–Control Study of Schizophrenia; ISHDSF, Irish Study of High-Density Schizophrenia Families; ITRIO, Irish Case–Control Study of Schizophrenia; OR, odds ratio; PDT, pedigree disequilibrium test. aThe counts of the vertical (trios) and horizontal (sib pairs) transmission from the PDT tests. Significant P-values are shaded.

observed in the ISHDSF, the associations were largely comparable between the samples. For the ICCSS from females. Thus, we were able to confirm the sex- sample, although the haplotypes implicated were specific associations in all three samples. However, the same as that observed in the ISHDSF and ITRIO, there were some discrepancies between the samples. the risk haplotypes in the ICCSS were undertrans- We found that the transmitted alleles in ISHDSF and mitted in both the ISHDSF and ITRIO. ITRIO were the same for all significant markers, but the transmitted alleles in the ISHDSF and ITRIO were LD structure analyses underrepresented in the affected individuals in the To understand the association signals, we analyzed ICCSS. the haplotype structure and computed pairwise LD We also carried out haplotype analyses for those for the three samples. Figure 1 and Table 5 showed FH-positive subjects for both the ICCSS and ITRIO the pairwise LD (D 0 and r2, respectively) of the samples. In both the ICCSS and ITRIO, the associa- ISHDSF and ICCSS. The pairwise LD was very tions were observed only in the females (see Table 4). similar, and the LD block partitions were exactly The ITRIO produced exactly the same risk and same when all subjects were included. The ITRIO protective haplotypes as that observed in the ISHDSF. showed similar pairwise LD and same LD block While the individual P-values were much more partitions (data not shown). We also compared the significant in the ISHDSF than that in the ITRIO, subjects with and without FH from the ICCSS sample the odds ratios of all implicated haplotypes were and again, the LD patterns were similar and LD blocks

Molecular Psychiatry Interleukin 3 and schizophrenia X Chen et al 279 Table 5 Pairwise LD (r2) of the ISHDSF and ICCSS samples

Abbreviations: ICCSS, Irish Case–Control Study of Schizophrenia; ISHDSF, Irish Study of High-Density Schizophrenia Families; LD, linkage disequilibrium. ISHDSF, above the diagonal; ICCSS, below the diagonal. r2X0.5 are shaded. were the same (data not shown). Differences were associations in those subjects with FH of schizophre- also observed between the affected and unaffected nia in both samples. In LD and haplotype analyses, subjects for both sexes when only subjects with FH the three samples used in this study showed similar were analyzed (Figure 1d–e). Some pairwise LD patterns. For those subjects with FH of schizophrenia, (i.e. 2–3, 3–4, 3–11, 5–11, etc. in Figure 1d) were some differences of LD patterns were observed not statistically significant. We noticed that those between the affected and unaffected females, and markers showing different LD patterns in Figure 1d the markers involved were right next to the EREs. were the markers closest to the EREs (see Figure 1a). From the genomic structure, we know that the We reconstructed haplotypes for the 12 markers for interval between rs246999 and rs31480 can serve as all three samples using the MERLIN/FUGUE pro- promoters for both the ACSL6 and IL3 genes. This grams31 and estimated their frequencies. In all three interval contains multiple transcription factor bind- samples, the top six haplotypes were exactly the same ing sites – AP-1, SP1 and GATA sites – that has been and they accounted for 89, 91 and 94% of the shown to regulate the expression of IL3 gene.8 There haplotypes observed in the ISHDSF, ICCSS and ITRIO are also multiple EREs identified recently in this samples, respectively. region.7 However, the regulating target of these EREs has not yet been identified. The physical positions of these EREs imply that they may be involved in the Discussion and conclusion regulation of ACSL6 gene. There is also ample 5q21–33 is a region likely to harbor risk evidence that estrogen receptors have multiple path- genes for schizophrenia. We started fine-mapping this ways to regulate gene expression. These include the region several years ago and we found that a genomic direct binding to EREs to activate transcription and interval from 130.5 to 131.5 million base pairs indirect binding to AP-1, SP1 and GATA binding showed consistent associations in our ISHDSF sam- proteins to coactivate the expression of genes.32,33 In ple. This interval contains several genes, including other words, estrogen can directly and indirectly the ACSL6 and IL3. ACSL6 and IL3 are arranged in a regulate the expression of ACSL6 and IL3 genes. In head-to-head orientation. In the effort to exclude one report, sex-specific expression of IL3 was found the IL3 from the list of candidate genes, we were in breast cancer patients.34 In another report, sexually surprised to find that several potentially functional dimorphic expression of both ACSL6 and IL3 was SNPs in the promoter and enhancer of IL3 were reported in a mouse model.35 In addition, the key associated with the disease (rs3914025, rs3846726, SNPs defining all risk haplotypes observed in all rs3916441, rs31400 and rs2069803) in the ISHDSF. three samples are potentially functional. Rs31480 is Two- and three-marker haplotypes involving key only 16 bp upstream of the start codon and is within SNPs (rs31400, rs31480 and rs2069803) were signifi- the 500 bp essential promoter sequence of the IL3 cant for all three disease definitions. When sex- gene (for details of IL3 transcription regulation, see a conditioned analyses were performed, we found that recent review by Cockerill).8 Rs31400 is located the associations – in both single markers and about 6.8 kb upstream of the first exon, which is haplotypes – were largely derived from the female between the two well-defined and well-studied IL3 subjects. We followed up these findings with an enhancers located at 4.5 and14 kb upstream of the independent case–control sample and a proband– gene.8 Another marker, rs2069803, which is also parents trio sample. We found the same sex-specific significantly associated with the disease in the family

Molecular Psychiatry Interleukin 3 and schizophrenia X Chen et al 280 samples, is located in another well-defined, DNase level, might inflate type I error as indicated by a hypersensitive region downstream of IL3 that has recent report.51 For these reasons, further study with been shown to play an important role in the independent samples is necessary to clarify this issue. expression of IL3.8 These three markers provide a In this study, we find that sex plays an important direct connection to the IL3 gene. Although it is not role in the association. In a recent, extensive review clear at this time whether variants of rs31400, rs31480 of worldwide incidence rate, it was found that and rs2069803 change the expression of IL3, this males have significantly higher incidence rate than scenario is likely, as demonstrated by promoter females.52 The males have younger age of onset than variants of other genes.36,37 Despite all these informa- the females.53 Sex-specific associations with schizo- tion, we cannot exclude the possibility that these phrenia have been reported for the catechol o- regulatory elements also regulate the expression of methyltransferase gene 38 and several other genes.54,55 ACSL6 gene. Functional and expression studies are It was also argued recently that sex may play a more necessary to clarify this issue. important role in brain structure and psychiatric As seen in several other studies of schizophrenia disorders than commonly acknowledged.56–58 Our candidate genes, different samples tend to have finding is consistent with this converging evidence. different risk haplotypes,38–44 including flipping of Furthermore, our finding of sex-specific associations the risk and protective haplotypes of DTNBP1 gene in this region has strong genomic and molecular basis, between the ISHDSF45 and German nuclear families46 that is, the existence of EREs and other coordinating and other samples.47,48 In this study, the two family transcription factor binding sites in this region and samples identified the same risk and protective the report of sex-specific expression of IL3 and ACSL6 haplotypes, but in the case–control sample, the risk genes.34,35 We also find that FH is a contributing haplotype in the family samples become protective factor for the associations. Consistent associations (underrepresented in the affected individuals). This are observed only in those subjects with a positive raises a question of replication, that is, what consti- FH of schizophrenia. This is not the first time that FH tutes a replication? In recent literature, it has been has impacted so clearly on the association of a proposed that a replication can be made at the levels candidate gene. At least one prior study had shown of individual markers, haplotypes and the gene.49 In a a different pattern of association when FH was meta-analysis of NRG1 gene, the authors seem to considered.59 suggest that the replication at the level of gene is more The immune system has long been hypothesized to robust.50 Although both the ITRIO and ICCSS samples be involved in schizophrenia.60 IL3 is a major provide supporting evidence at the levels of indivi- component of this system and plays a critical role in dual markers, haplotypes and the gene as judged by the development and differentiation of all hemato- the criteria proposed by Neale and Sham,49 the poietic cell types. Recently, it has been shown to have conflict effect of alleles and haplotypes between the neurotropic activities in the brain. Transgenic mice ICCSS and the family samples is puzzling. There may with low levels of IL3 expression in the central be several possible explanations to these conflict nervous system demonstrate demyelination and results. The obvious difference between ICCSS and motor disease.10 Expression of antisense IL3 RNA in ISHDSF and ITRIO is that ICCSS is a case–control microglia causes progressive neurologic dysfunctions, design whereas the other two samples are family- including ataxia, bradykinesia and paralysis.11 IL3 based designs. It is possible, in theory, that some could potentially be involved in the pathophysiology unknown stratifications are responsible for these of schizophrenia in several ways. First, IL3 is the results as the case–control study is more susceptible upstream signal that activates the Akt1/Gsk3b apop- to this influence. Second, it is possible that the totic pathway.61 Imbalance of apoptosis in the early conflicting risk and protective haplotypes observed stages of development of the central nervous system might be due to sampling fluctuation, given the has been suggested as a cause of schizophrenia.14,62 relative small sample sizes for FH-positive subjects Indeed, a higher Bax/Bcl-2 ratio, an index of apoptotic and the nominal significance in both the ITRIO and activation, is observed in schizophrenia patients.63 In ICCSS samples. As the two family samples produce recent years, key components in this pathway, like the consistent results (i.e. the same alleles and haplotypes tumor necrosis factor,64 Akt165 and Gsk3b,66 have been are overtransmitted to affected subjects), this is implicated in schizophrenia. Second, IL3 is one of the equivalent to one-tailed tests for the ITRIO sample. major factors that activate the microglia in the brain. Based on this rationale, we incline to think that the Activated microglia lead to inflammatory responses signals from the ITRIO are more reliable. We should in the brain,10,12,67 and some researchers have sug- add that we performed 1000 permutation tests for all gested that such inflammatory processes contribute to individual markers for both the ITRIO and ICCSS, the etiology of schizophrenia.68 For this reason, other the results were very similar to the theoretical w2 inflammatory cytokines have been studied in schizo- distribution (data not shown). Ascertainment varia- phrenia.69,70 Clinical studies show that IL3 activity tion might be another factor potentially contributing may be lower in schizophrenia patients.17,71 The non- to the conflicting risk haplotypes even when the synonymous marker, rs40401, is reported to be probands were ascertained with same criteria. In associated with rheumatoid arthritis72 that is nega- addition, genotype errors, even at a relatively low tively correlated with schizophrenia.73

Molecular Psychiatry Interleukin 3 and schizophrenia X Chen et al 281 In summary, we present evidence that variants in antisense-IL-3 transgenic mice. J Neuropathol Exp Neurol 1999; and around IL3 are associated with schizophrenia in 58: 480–488. the ISHDSF, ICCSS and ITRIO samples. Specifically, 12 Giralt M, Carrasco J, Penkowa M, Morcillo MA, Santamaria J, Campbell IL et al. Astrocyte-targeted expression of interleukin-3 the associations are sex-specific and depend on FH and interferon-alpha causes region-specific changes in metal- for schizophrenia. The sex-specific associations have lothionein expression in the brain. Exp Neurol 2001; 168: a genomic and molecular basis. In all three samples, 334–346. key SNPs in known regulatory elements of the IL3 13 Reddy EP, Korapati A, Chaturvedi P, Rane S. IL-3 signaling and the gene underlie the risk haplotypes. They have the role of Src kinases, JAKs and STATs: a covert liaison unveiled. Oncogene 2000; 19: 2532–2547. potential to directly change the expression of IL3. But 14 Jarskog LF, Glantz LA, Gilmore JH, Lieberman JA. Apoptotic we cannot rule out that the regulatory elements, mechanisms in the pathophysiology of schizophrenia. Prog especially the EREs in the shared promoter region Neuropsychopharmacol Biol Psychiatry 2005; 29: 846–858. also control the expression of ACSL6. In light of these 15 Zhang XY, Zhou DF, Zhang PY, Wu GY, Cao LY, Shen YC. Elevated findings, further study of these elements and SNPs interleukin-2, interleukin-6 and interleukin-8 serum levels in neuroleptic-free schizophrenia: association with psychopathology. and their effects on schizophrenia are warranted. Schizophr Res 2002; 57: 247–258. 16 Brown AS, Hooton J, Schaefer CA, Zhang H, Petkova E, Babulas V et al. Elevated maternal interleukin-8 levels and risk of Acknowledgments schizophrenia in adult offspring. Am J Psychiatry 2004; 161: 889–895. This study is supported by a research grant 17 Sirota P, Schild K, Elizur A, Djaldetti M, Fishman P. Increased (RO1MH41953) to KSK from the National Institute interleukin-1 and interleukin-3 like activity in schizophrenic of Mental Health and by a young investigator award to patients. Prog Neuropsychopharmacol Biol Psychiatry 1995; 19: 75–83. XC from the National Alliance for Research on 18 Kendler KS, McGuire M, Gruenberg AM, O’Hare A, Spellman M, Schizophrenia and Depression. We thank the patients Walsh D. The roscommon family study. III. Schizophrenia-related and their families for participating in this study. personality disorders in relatives. Arch Gen Psychiatry 1993; 50: 781–788. 19 Kendler KS, Myers JM, O’Neill FA, Martin R, Murphy B, MacLean References CJ et al. Clinical features of schizophrenia and linkage to 5q, 6p, 8p, and 10p in the Irish Study of High- 1 Schwab SG, Eckstein GN, Hallmayer J, Lerer B, Albus M, Density Schizophrenia Families. Am J Psychiatry 2000; 157: Borrmann M et al. Evidence suggestive of a locus on chromosome 402–408. 5q31 contributing to susceptibility for schizophrenia in German 20 Endicott J, Andreasen N, Spitzer RL. Family History Research and Israeli families by multipoint affected sib-pair linkage Diagnostic Criteria. New York State Psychiatric Institute: analysis. Mol Psychiatry 1997; 2: 156–160. New York, 1978. 2 Straub RE, MacLean CJ, O’Neill FA, Walsh D, Kendler KS. Support 21 Livak KJ. Allelic discrimination using fluorogenic probes and the for a possible schizophrenia vulnerability locus in region 5q22–31 50 nuclease assay. Genet Anal 1999; 14: 143–149. in Irish families. Mol Psychiatry 1997; 2: 148–155. 22 Chen X, Levine L, Kwok PY. Fluorescence polarization in 3 Straub RE, MacLean CJ, Ma Y, Webb BT, Myakishev MV, homogeneous nucleic acid analysis. Genome Res 1999; 9: Harris-Kerr C et al. Genome-wide scans of three independent 492–498. sets of 90 Irish multiplex schizophrenia families and follow-up 23 van den Oord EJ, Jiang Y, Riley BP, Kendler KS, Chen X. FP-TDI of selected regions in all families provides evidence for multiple SNP scoring by manual and statistical procedures: a study of error susceptibility genes. Mol Psychiatry 2002; 7: 542–559. rates and types. Biotechniques 2003; 34: 610–620, 622. 4 Straub RE, Jiang Y, MacLean CJ, Ma Y, Webb BT, Myakishev MV et 24 Bedell JA, Korf I, Gish W. MaskerAid: a performance enhancement al. Genetic variation in the 6p22.3 gene DTNBP1, the human to RepeatMasker. Bioinformatics 2000; 16: 1040–1041. ortholog of the mouse dysbindin gene, is associated with 25 Wigginton JE, Cutler DJ, Abecasis GR. A note on exact tests schizophrenia. Am J Hum Genet 2002; 71: 337–348. of Hardy–Weinberg equilibrium. Am J Hum Genet 2005; 76: 5 Stefansson H, Sigurdsson E, Steinthorsdottir V, Bjornsdottir S, 887–893. Sigmundsson T, Ghosh S et al. Neuregulin 1 and susceptibility to 26 Martin ER, Monks SA, Warren LL, Kaplan NL. A test for linkage schizophrenia. Am J Hum Genet 2002; 71: 877–892. and association in general pedigrees: the pedigree disequilibrium 6 Duan J, Martinez M, Sanders AR, Hou C, Saitou N, Kitano T et al. test. Am J Hum Genet 2000; 67: 146–154. Polymorphisms in the trace amine receptor 4 (TRAR4) gene 27 Dudbridge F. Pedigree disequilibrium tests for multilocus haplo- on chromosome 6q23.2 are associated with susceptibility to types. Genet Epidemiol 2003; 25: 115–121. schizophrenia. Am J Hum Genet 2004; 75: 624–638. 28 Excoffier L, Slatkin M. Maximum-likelihood estimation of mole- 7 Bourdeau V, Deschenes J, Metivier R, Nagai Y, Nguyen D, cular haplotype frequencies in a diploid population. Mol Biol Evol Bretschneider N et al. Genome-wide identification of high-affinity 1995; 12: 921–927. estrogen response elements in human and mouse. Mol Endocrinol 29 Barrett JC, Fry B, Maller J, Daly MJ. Haploview: analysis and 2004; 18: 1411–1427. visualization of LD and haplotype maps. Bioinformatics 2005; 21: 8 Cockerill PN. Mechanisms of transcriptional regulation of the 263–265. human IL-3/GM-CSF locus by inducible tissue-specific promoters 30 Gabriel SB, Schaffner SF, Nguyen H, Moore JM, Roy J, Blumenstiel and enhancers. Crit Rev Immunol 2004; 24: 385–408. B et al. The structure of haplotype blocks in the . 9 Burdach S, Nishinakamura R, Dirksen U, Murray R. The Science 2002; 296: 2225–2229. physiologic role of interleukin-3, interleukin-5, granulocyte- 31 Abecasis GR, Cherny SS, Cookson WO, Cardon LR. Merlin – rapid macrophage colony-stimulating factor, and the beta c receptor analysis of dense genetic maps using sparse gene flow trees. Nat system. Curr Opin Hematol 1998; 5: 177–180. Genet 2002; 30: 97–101. 10 Powell HC, Garrett RS, Brett FM, Chiang CS, Chen E, Masliah E et 32 O’Lone R, Frith MC, Karlsson EK, Hansen U. Genomic targets of al. Response of glia, mast cells and the blood brain barrier, in nuclear estrogen receptors. Mol Endocrinol 2004; 18: 1859–1875. transgenic mice expressing interleukin-3 in astrocytes, an experi- 33 Bjornstrom L, Sjoberg M. Mechanisms of estrogen receptor mental model for CNS demyelination. Brain Pathol 1999; 9: signaling: convergence of genomic and nongenomic actions on 219–235. target genes. Mol Endocrinol 2005; 19: 833–842. 11 Sugita Y, Zhao B, Shankar P, Dunbar CE, Doren S, Young HA et al. 34 Speirs V, Birch MA, Boyle-Walsh E, Green AR, Gallagher JA, White CNS interleukin-3 (IL-3) expression and neurological syndrome in MC. Interleukin-3: a putative protective factor against breast

Molecular Psychiatry Interleukin 3 and schizophrenia X Chen et al 282 cancer which is secreted by male but not female breast fibroblasts. 53 Halbreich U, Kahn LS. Hormonal aspects of schizophrenias: an Int J Cancer 1995; 61: 416–419. overview. Psychoneuroendocrinology 2003; 28(Suppl 2): 1–16. 35 Yang X, Schadt EE, Wang S, Wang H, Arnold AP, Ingram-Drake L 54 Tan EC, Chong SA, Wang H, Chew-Ping LE, Teo YY. Gender- et al. Tissue-specific expression and regulation of sexually specific association of insertion/deletion polymorphisms in the dimorphic genes in mice. Genome Res 2006; 16: 995–1004. nogo gene and chronic schizophrenia. Brain Res Mol Brain Res 36 Chang HS, Kim JS, Lee JH, Cho JI, Rhim TY, Uh ST et al. A single 2005; 139: 212–216. nucleotide polymorphism on the promoter of eotaxin1 associates 55 Thomson PA, Wray NR, Thomson AM, Dunbar DR, Grassie MA, with its mRNA expression and asthma phenotypes. J Immunol Condie A et al. Sex-specific association between bipolar affective 2005; 174: 1525–1531. disorder in women and GPR50, an X-linked orphan G protein- 37 D’Adamo M, Perego L, Cardellini M, Marini MA, Frontoni S, coupled receptor. Mol Psychiatry 2005; 10: 470–478. Andreozzi F et al. The -866A/A genotype in the promoter of the 56 Salem JE, Kring AM. The role of gender differences in the human uncoupling protein 2 gene is associated with insulin reduction of etiologic heterogeneity in schizophrenia. Clin Psychol resistance and increased risk of type 2 diabetes. Diabetes 2004; 53: Rev 1998; 18: 795–819. 1905–1910. 57 Czlonkowska A, Ciesielska A, Gromadzka G, Kurkowska-Jastr- 38 Shifman S, Bronstein M, Sternfeld M, Pisante-Shalom A, zebska I. Estrogen and cytokines production – the possible cause Lev-Lehman E, Weizman A et al. A highly significant association of gender differences in neurological diseases. Curr Pharm Des between a COMT haplotype and schizophrenia. Am J Hum Genet 2005; 11: 1017–1030. 2002; 71: 1296–1302. 58 Cahill L. Why sex matters for neuroscience. Nat Rev Neurosci 39 Chen X, Wang X, O’Neill AF, Walsh D, Kendler KS. Variants in the 2006; 7: 477–484. catechol-o-methyltransferase (COMT) gene are associated with 59 Van Den BA, Schumacher J, Schulze TG, Otte AC, Ohlraun S, schizophrenia in Irish high-density families. Mol Psychiatry 2004; Kovalenko S et al. The DTNBP1 (dysbindin) gene contributes to 9: 962–967. schizophrenia, depending on family history of the disease. Am J 40 Chowdari KV, Mirnics K, Semwal P, Wood J, Lawrence E, Hum Genet 2003; 73: 1438–1443. Bhatia T et al. Association and linkage analyses of RGS4 60 Jones AL, Mowry BJ, Pender MP, Greer JM. Immune dysregulation polymorphisms in schizophrenia. Hum Mol Genet 2002; 11: and self-reactivity in schizophrenia: do some cases of schizo- 1373–1380. phrenia have an autoimmune basis? Immunol Cell Biol 2005; 83: 41 Stefansson H, Sarginson J, Kong A, Yates P, Steinthorsdottir V, 9–17. Gudfinnsson E et al. Association of neuregulin 1 with schizo- 61 Mitsiades CS, Mitsiades N, Koutsilieris M. The Akt pathway: phrenia confirmed in a Scottish population. Am J Hum Genet molecular targets for anti-cancer drug development. Curr Cancer 2003; 72: 83–87. Drug Targets 2004; 4: 235–256. 42 Li T, Stefansson H, Gudfinnsson E, Cai G, Liu X, Murray RM et al. 62 Wierzba-Bobrowicz T, Lewandowska E, Kosno-Kruszewska E, Identification of a novel neuregulin 1 at-risk haplotype in Lechowicz W, Pasennik E, Schmidt-Sidor B. Degeneration of Han schizophrenia Chinese patients, but no association with microglial cells in frontal and temporal lobes of chronic schizo- the Icelandic/Scottish risk haplotype. Mol Psychiatry 2004; 9: phrenics. Folia Neuropathol 2004; 42: 157–165. 698–704. 63 Jarskog LF, Selinger ES, Lieberman JA, Gilmore JH. Apoptotic 43 Funke B, Finn CT, Plocik AM, Lake S, DeRosse P, Kane JM et al. proteins in the temporal cortex in schizophrenia: high Bax/Bcl-2 ratio Association of the DTNBP1 locus with schizophrenia in a U.S. without caspase-3 activation. Am J Psychiatry 2004; 161: 109–115. population. Am J Hum Genet 2004; 75: 891–898. 64 Hanninen K, Katila H, Rontu R, Mattila KM, Hurme M, Lehtimaki 44 Chen X, Dunham C, Kendler S, Wang X, Oneill F, Walsh D et al. T. Tumor necrosis factor-alpha – G308A polymorphism in Regulator of G-protein signaling 4 (RGS4) gene is associated with schizophrenia in a Finnish population. Neurosci Lett 2005; 385: schizophrenia in Irish high density families. Am J Med Genet 76–81. 2004; 129B: 23–26. 65 Emamian ES, Hall D, Birnbaum MJ, Karayiorgou M, Gogos JA. 45 van den Oord EJ, Sullivan PF, Jiang Y, Walsh D, Neill FA, Kendler Convergent evidence for impaired AKT1-GSK3beta signaling in KS et al. Identification of a high-risk haplotype for the dystro- schizophrenia. Nat Genet 2004; 36: 131–137. brevin binding protein 1 (DTNBP1) gene in the Irish study of 66 Nadri C, Dean B, Scarr E, Agam G. GSK-3 parameters in high-density schizophrenia families. Mol Psychiatry 2003; 8: postmortem frontal cortex and hippocampus of schizophrenic 499–510. patients. Schizophr Res 2004; 71: 377–382. 46 Schwab SG, Knapp M, Mondabon S, Hallmayer J, Borrmann- 67 Natarajan C, Sriram S, Muthian G, Bright JJ. Signaling through Hassenbach M, Albus M et al. Support for association of JAK2-STAT5 pathway is essential for IL-3-induced activation of schizophrenia with genetic variation in the 6p22.3 gene, dysbin- microglia. Glia 2004; 45: 188–196. din, in sib-pair families with linkage and in an additional sample 68 Hanson DR, Gottesman II. Theories of schizophrenia: a genetic- of triad families. Am J Hum Genet 2003; 72: 185–190. inflammatory-vascular synthesis. BMC Med Genet 2005; 6:7. 47 Kirov G, Ivanov D, Williams NM, Preece A, Nikolov I, Milev R 69 Yu L, Yang MS, Zhao J, Shi YY, Zhao XZ, Yang JD et al. An et al. Strong evidence for association between the dystrobrevin association between polymorphisms of the interleukin-10 gene binding protein 1 gene (DTNBP1) and schizophrenia in 488 promoter and schizophrenia in the Chinese population. Schizophr parent-offspring trios from Bulgaria. Biol Psychiatry 2004; 55: Res 2004; 71: 179–183. 971–975. 70 Rosa A, Peralta V, Papiol S, Cuesta MJ, Serrano F, Martinez-Larrea 48 Williams NM, Preece A, Morris DW, Spurlock G, Bray NJ, A et al. Interleukin-1beta (IL-1beta) gene and increased risk for the Stephens M et al. Identification in 2 independent samples of a depressive symptom-dimension in schizophrenia spectrum novel schizophrenia risk haplotype of the dystrobrevin disorders. Am J Med Genet B Neuropsychiatr Genet 2004; 124: binding protein gene (DTNBP1). Arch Gen Psychiatry 2004; 61: 10–14. 336–344. 71 Bessler H, Levental Z, Karp L, Modai I, Djaldetti M, Weizman A. 49 Neale BM, Sham PC. The future of association studies: gene-based Cytokine production in drug-free and neuroleptic-treated schizo- analysis and replication. Am J Hum Genet 2004; 75: 353–362. phrenic patients. Biol Psychiatry 1995; 38: 297–302. 50 Munafo MR, Thiselton DL, Clark TG, Flint J. Association of the 72 Yamada R, Tanaka T, Unoki M, Nagai T, Sawada T, Ohnishi Y et al. NRG1 gene and schizophrenia: a meta-analysis. Mol Psychiatry Association between a single-nucleotide polymorphism in the 2006; 11: 539–546. promoter of the human interleukin-3 gene and rheumatoid 51 Souverein OW, Zwinderman AH, Tanck MW. Multiple imputation arthritis in Japanese patients, and maximum-likelihood estimation of missing genotype data for unrelated individuals. Ann Hum of combinatorial effect that two genetic loci have on susceptibility Genet 2006; 70: 372–381. to the disease. Am J Hum Genet 2001; 68: 674–685. 52 McGrath J, Saha S, Welham J, El Saadi O, MacCauley C, Chant D. A 73 Gorwood P, Pouchot J, Vinceneux P, Puechal X, Flipo RM, De systematic review of the incidence of schizophrenia: the distribu- Bandt M et al. Rheumatoid arthritis and schizophrenia: a negative tion of rates and the influence of sex, urbanicity, migrant status association at a dimensional level. Schizophr Res 2004; 66: and methodology. BMC Med 2004; 2: 13. 21–29.

Molecular Psychiatry