Gene Interactions of the INSIG1 and INSIG2 in Metabolic Syndrome in Schizophrenic Patients Treated with Atypical Antipsychotics

Gene–gene interactions of the INSIG1 and INSIG2 in metabolic syndrome in schizophrenic patients treated with atypical antipsychotics

Y-J Liou1,2,3,8, YM Bai1,2,8, The use of atypical antipsychotics (AAPs) is associated with increasing the risk 4,5 6 6 of the metabolic syndrome (MetS), which is an important risk factor for E Lin , J-Y Chen , T-T Chen , cardiovascular disease and diabetes. Two insulin-induced gene (INSIG) 1,2,7 1,2 C-J Hong and S-J Tsai isoforms, designated INSIG-1 and INSIG-2 encode two proteins that mediate feedback control of lipid metabolism. In this genetic case–control study, we 1Department of Psychiatry, Taipei Veterans General Hospital, Taipei, Taiwan; 2Division of investigated whether the common variants in INSIG1 and INSIG2 genes were Psychiatry, School of Medicine, National Yang- associated with MetS in schizophrenic patients treated with atypical Ming University, Taipei, Taiwan; 3Institute of antipsychctics. The study included 456 schizophrenia patients treated with Clinical Medicine, School of Medicine, National 4 clozapine (n ¼ 171), olanzapine (n ¼ 91) and risperidone (n ¼ 194), for an Yang-Ming University, Taipei, Taiwan; Vita ± Genomics, Taipei, Taiwan; 5Graduate Institute of average of 45.5 27.6 months. The prevalence of MetS among all subjects Clinical Medical Science, China Medical was 22.8% (104/456). Two single-nucleotide polymorphisms (SNPs) of the University, Taichung, Taiwan; 6Department of INSIG1 gene and seven SNPs of the INSIG2 gene were chosen as haplotype- Psychiatry, Yuli Veterans Hospital, Yuli, Hualien, tagging SNPs. In single-marker-based analysis, the INSIG2 rs11123469-C Taiwan and 7Institute of Brain Science, National Yang-Ming University, Taipei, Taiwan homozygous genotype was found to be more frequent in the patients with MetS than those without MetS (P ¼ 0.001). In addition, haplotype analysis Correspondence: showed that the C-C-C haplotype of rs11123469-rs10185316- rs1559509 Dr C-J Hong or Dr S-J Tsai, Department of of the INSIG2 gene significantly increased the risk of MetS (P ¼ 0.0023). Psychiatry, Taipei Veterans General Hospital, No significant associations were found between polymorphisms of INSIG1 No. 201, Shih-Pai Road, Sec. 2, Taipei 11217, Taiwan. gene and MetS, however, INSIG1 and INSIG2 interactions were found in the E-mail: [email protected] or significant 3-locus and 4-locus gene–gene interaction models (P ¼ 0.003 and [email protected] 0.012, respectively). The results suggest that the INSIG2 gene may be associated with MetS in patients treated with AAPs independently or in an interactive manner with INSIG1. The Pharmacogenomics Journal (2012) 12, 54–61; doi:10.1038/tpj.2010.74; published online 28 September 2010

Keywords: gene–gene interaction; insulin-induced gene 1; insulin-induced gene 2; antipsychotic; metabolic syndrome

Introduction

Metabolic syndrome (MetS) is a combination of risk factors, including weight gain, abdominal or visceral adiposity, dyslipidemia and elevated plasma glucose that increase the risk of developing cardiovascular disease and diabetes. 8 These authors contributed equally to this Individuals with schizophrenia are a high-risk group for MetS due to genetic work. predisposition, medication effects and lifestyle.1,2 Regarding the medication effects, the increased risk to develop MetS under atypical antipsychotic (AAP) Received 7 February 2010; revised 18 August 2010; accepted 26 August 2010; agents, such as clozapine or olanzapine, have been a major concern for the published online 28 September 2010 treatment of schizophrenia.3 In a study of MetS in first-episode patients with INSIGs and metabolic syndrome Y-J Liou et al 55

schizophrenia treated with typical or AAPs, it was found SCAP have a crucial role in feedback regulation of lipid that, at first episode, there was no difference in the metabolism.17 In addition of lipid metabolism, Herbert et prevalence of MetS between the historical and the current al., in a genome-wide association study, demonstrated a cohort.4 However, rates of MetS increased over time in both genetic variant (rs7566605) situated 10-kb upstream of the groups, but patients treated with AAPs had a three times transcription start site of INSIG2 was associated with obesity higher incidence rate of MetS (odds ratio 3.6). In another that was replicated by several replication studies, though study of 3-month follow-up of female patients with schizo- not by all.18 phrenia treated with AAPs, it was found, at baseline, 15% From the above findings, INSIG proteins may be impli- fulfilled criteria for MetS, and, after 3 months of treatment, cated in the pathogenesis of MetS in patients treated 27% fulfilled criteria for MetS.5 Recently, Correll et al. with AAPs. Skelly et al. conducted the first study on the investigated patients with bipolar disorder or schizophrenia association of INSIG2 genetic variation and BMI in 756 patients treated with AAPs.6 They found both bipolar schizophrenia patients before and after 18-month Clinical disorder and schizophrenia patients had comparable and Antipsychotics Trials of Intervention Effectiveness (CATIE) high rates of MetS (43.2 and 45.9%, respectively). The phase 1A trail consisting of four AAPs: olanzapine, risper- pathogenesis underlying AAP-related MetS is still unknown idone, quetiapine and ziprasidone.19 Most of the partici- but some risk factors such as patients older than 50 years, pants in CATIE study were European, African, Spanish, higher body mass index (BMI) at initiation of antipsychotic Hispanic and Latino origins. However, they could not treatment, BMI increase after initiation of antipsychotic observe an association between INSIG2-rs7566605 and the treatment, duration of antipsychotic drug exposure, and baseline BMI or BMI change across all CATIE phase 1A those taking clozapine or multiple antipsychotics have been treatment. A later study consisted of 160 German patients reported.7–9 with clozapine treatment for 12.0±1.2 weeks.20 Although During the past decade, dramatic advances in molecular they also did not observe the association of rs7566605, they biology have led to the modern era of pharmacogenetics, found an association between three SNPs (rs17587100, with variability in drug response, which includes both rs10490624 and rs17047764) localized with or near INSIG2 therapeutic and adverse effect, attributed to genetic factors gene and clozapine-related BMI changes.20 The two studies discovered in different populations. The identification of yielded divergent results. Furthermore there was no report the genetic variants that influence drug-related adverse regarding the association of INSIG1 or INSIG2 genetic effect may help pre-treatment selection and development variations with body weight changes or MetS in patients of drugs that are safe for individual patients on the basis treated with AAP in Oriental population. In the present of their idiosyncratic pharmacogenetic profile.10 Currently, study, we hypothesized that the INSIG1 or INSIG2 genes may there are three reports of pharmacogenetic studies of be associated with MetS in schizophrenic patients treated antipsychotic-related MetS. In 2007, Mulder et al. first with AAPs. Thus we conducted a case–control study in demonstrated that genetic variants in 5-hydroxytryptamine Chinese population to investigate the relationship between (serotonin) receptor 2C is associated with MetS in patients the genetic variants in INSIG1 or INSIG2 and risk of MetS. with schizophrenia.11 They replicated the finding in a recent We also determined whether significant gene–gene inter- report and further demonstrated that, the increased risk for actions exist between these two genes in affecting MetS, MetS, is particularly strong in carriers of the 5-hydroxy- using a novel algorithm called generalized multifactor tryptamine (serotonin) receptor 2C rs1414334 C allele using dimensionality reduction (GMDR) method, an extension clozapine or risperidone.12 Another study investigated of the MDR method.21 association between methylenetetrahydrofolate reductase genetic variants and MetS in schizophrenic patients receiving AAPs and demonstrated that the methylenetetrahydro- Materials and methods folate reductase 677T allele is associated with a 3.6-fold greater risk for developing AAP-associated MetS and the Subjects and measurements 677TT genotype may place individuals at greater risk for This study was approved by the Ethnic Review Committee of insulin resistance with greater central adiposity.13 Yuli Veterans Hospital and was carried out in accordance Two insulin-induced gene (INSIG) isoforms, designated with the principles of the Declaration of Helsinki. Informed INSIG-1 and INSIG-2, are closely related proteins of the consent was obtained from all subjects before commence- endoplasmic reticulum, encoded by individual gene ment. This study was conducted in the largest psychiatric (INSIG1 and INSIG2, respectively).14,15 In the mouse, an hospital in Taiwan, which has approximately 2500 inpati- integrating quantitative trait loci and high-density single- ents. As our previous report,8 patients selected for the study nucleotide polymorphism (SNP) analyses identified INSIG2 had used clozapine, olanzapine or risperidone for at least 3 as a strong susceptibility gene for plasma cholesterol months for treatment of schizophrenia, as diagnosed levels.16 INSIG proteins block the processing of sterol according to the Diagnostic and Statistical Manual of Mental regulatory element binding proteins (SREBPs) by sterol- Disorders (fourth edition) (DSM-IV).22 Consensus diagnoses dependent binding to SREBP cleavage-activating protein for each patient was made according to DSM-IV criteria for (SCAP), and thus prevent SCAP from escorting SREBPs to the schizophrenia by two certificated psychiatrists on the basis Golgi. Thus, the interaction of INSIG proteins, SREBPs and of an interview, clinical observation, medical record and

The Pharmacogenomics Journal INSIGs and metabolic syndrome Y-J Liou et al 56

past history. Patients with the following exclusion criteria Table 1 Basic and clinical characteristics of patients with or were not recruited: history of mood disorder, psychotic without metabolic syndrome, which is defined according to the disorder or mood disorder due to general medical condition, International Diabetes Federation Asia criteria dementia, mental retardation, substance use or neurological illness. All study subjects were continuously hospitalized, Metabolic Without metabolic P syndrome syndrome used the same antipsychotic drug and underwent routine (n ¼ 104) (n ¼ 352) monthly body weight assessment throughout the study period. Alcohol consumption was prohibited on all wards. Gender, male/female 70/34 299/123 0.671 These conditions ensured optimal control of drug treatment Age, year±s.d. 46.9±13.3 48.8±12.3 0.232 compliance and fasting status for assessment of metabolic AAP (clozapine/ 47/22/35 124/69/159 0.094 parameters. With the inclusion and exclusion, there were olanzapine/risperidone) initially 629 patients with schizophrenia treated with the Duration of AAPD use, 46.7±27.9 45.4±27.6 0.819 three AAPs for at least 3 months. Among them, 173 subjects month±s.d. did not have at least one of the information necessary Baseline BMI±s.d. 26.9±4.4 23.1±4.0 o0.001 for making a diagnosis of MetS. Therefore, there were Abbreviations: AAP, atypical antipsychotic; AAPD, atypical antipsychotic drug; 456 included for further study. BMI, body mass index; s.d., standard deviation. Retrospective reviews of the study subjects’ medical records were performed to obtain demographic data, including age at initiation of antipsychotic treatment, BMI gene and six tSNPs of the INSIG2 gene were chosen (Table 2). at initiation of antipsychotic treatment and change in BMI The eight tSNPs and rs7566605 were included for the study. after initiation of antipsychotic treatment. A cross-sectional Peripheral venous blood was taken from the study subjects assessment of anthropometric and metabolic parameters for genotyping of the INSIG1 and INSIG2 SNPs. Genomic was performed to determine the prevalence of MetS DNA was isolated using the PUREGENE DNA purification according to the 2005 International Diabetes Federation system (Gentra Systems, Minneapolis, MN, USA). For DNA Asia criteria: waist circumference 490 cm in males or quality examination, all the samples were genotyped for 480 cm in females, as the essential criterion for central eight unrelated SNPs. The samples were diluted onto 96-well obesity, plus two of the following four criteria: (a) fasting plates, and only the plate with an average successful serum triglyceride levels X150 mg/dl; (b) fasting high- genotyping rate greater than 95% for the eight SNPs were density lipoprotein cholesterol (HDL) cholesterol level used for further study. Briefly, primers and probes were o40 mg per 100 ml in men or o50 mg per 100 ml in designed with SpectroDESIGNER software (Sequenom, San women; (c) blood pressure X130/85 mm Hg; and (d) fasting Diego, CA, USA). The sequences of the primers are glucose level X100 mg per 100 ml. Overnight fasting blood summarized in the Supplementary Table 1. A multiplex samples were drawn between 0700 and 0800 h from all polymerase chain reaction was performed, and unincorpo- patients. Serum glucose, triglyceride and cholesterol levels rated double stranded nucleotide triphosphate bases were measured using a glucose oxidase autoanalyzer, a were dephosphorylated with shrimp alkaline phosphatase triglyceride enzyme autoanalyzer and a cholesterol oxidase (Hoffman-LaRoche, Basel, Switzerland) followed by primer autoanalyzer, respectively (Dimension RxL, DADE Behring extension. The purified primer extension reaction was Company, Newark, DE, USA). spotted on to a 384-element silicon chip (SpectroCHIP, The final study sample consisted of 456 patients Sequenom) and analyzed in the Bruker Biflex III matrix- treated with clozapine (n ¼ 171), olanzapine (n ¼ 91) and assisted laser desorption or ionization-time of flight Spectro- risperidone (n ¼ 194) with average antipsychotic treatment READER mass spectrometer (Sequenom). The resulting duration of 45.5±27.6 months in the total group. Table 1 spectra were processed with SpectroTYPER (Sequenom). shows the demographic data and treatment characteristics, The genotyping successful rate ranges from 95.6 to 98.2% metabolic parameters and prevalence of MetS among the with the exception of rs2161829 that has a successful rate of three groups. 77.8%. All the experiments were done by investigators who were blind to the phenotypes.

SNPs selection and genotyping Statistical analysis We selected candidate markers from the CHB (Han Chinese Statistical analysis was performed using SPSS 10.0. in Beijing, China) database in the International HapMap Differences in continuous variables were evaluated using Project (HapMap data Rel24/phase II, dbSNP 126, www. Student’s t-test. The categorical data were analyzed using the hapmap.org), except of rs7566605 in INSIG2, which has w2-test or the Fisher exact test if necessary. The criterion for been found to be associated with obesity and BMI.18 significance was set at Po0.05 for tests for demographic and We used the program Tagger implemented in Haploview clinical data. Logistic regression analyses were conducted v4.0 and the option ‘two-and-three aggressive tagging’ in single-marker analysis, and gender, age, baseline BMI, (r2 threshold) to selected tagSNPs (tSNPs) that could duration of current AAP exposure and types of AAPs represent the overall genetic variations of INSIG1 and were entered as covariates. For the analyses of difference INSIG2 genes. With the approach, two tSNPs of the INSIG1 in the mean of quantitative factors between the genotypes,

The Pharmacogenomics Journal INSIGs and metabolic syndrome Y-J Liou et al 57

Table 2 Single-marker analysis for the association of INSIG1 and INSIG2 genotypes with MetS

Gene SNP rs no. Location Genotype MetS, n (%) Non-MetS, n (%) Crude P-value (genotype/allele)

INSIG1 rs9767875 Intron 2 GG 44 (43.6) 163 (47.8) 0.582/0.319 GT 44 (43.6) 145 (42.5) TT 13 (12.9) 33 (9.7) rs9719268 Intron 7 GG 75 (75.0) 263 (76.7) 0.936/0.706 GT 22 (22.0) 71 (20.7) TT 3 (3.0) 9 (2.6) INSIG2 rs7566605 50-flanking region CC 15 (14.9) 60 (17.4) 0.239/0.129 CG 40 (39.6) 160 (46.4) GG 46 (45.5) 125 (36.2) rs11123469 50-flanking region TT 44 (43.6) 192 (56.3) 0.001/0.0008a CT 39 (38.6) 127 (37.2) CC 18 (17.8) 22 (6.5) rs10185316 50-flanking region CC 72 (71.3) 240 (69.2) 0.718/0.879 CG 24 (23.8) 94 (27.1) GG 5 (5.0) 13 (3.7) rs1559509 Intron 2 CC 79 (79.8) 264 (77.6) 0.190/1.000 CT 17 (17.2) 73 (21.5) TT 3 (3.0) 3 (0.9) rs2161829 Intron 2 TT 24 (27.6) 60 (22.4) 0.397/0.229 CT 52 (59.8) 160 (59.7) CC 11 (12.6) 48 (17.9) rs17047757 Intron 3 AA 59 (59.6) 224 (65.3) 0.311/0.146 AG 31 (31.3) 101 (29.4) GG 9 (9.1) 18 (5.2) rs9308762 Intron 3 TT 28 (29.2) 115 (33.8) 0.362/0.159 CT 40 (41.7) 149 (43.8) CC 28 (29.2) 76 (22.4)

Abbreviations: BMI, body mass index; CI, confidence interval; MetS, metabolic syndrome; OR, odds ratio; SNP, single-nucleotide polymorphism. aRisk for having metabolic syndrome, CC vs TT: adjusted OR ¼ 5.10 (95% CI ¼ 2.23 B11.70, P ¼ 0.00011). The adjusted OR was the crude OR adjusted for gender, age, baseline BMI, duration of current atypical antipsychotic exposure and type of atypical antipsychotic drug using binary logistic regression analysis.

we applied analysis of covariance with controlling for to controls (non-MetS) in that class. This is performed for all appropriate covariates, followed by post hoc Bonferroni possible combinations of SNPs, and the combination with procedure. The quantitative factors included those for the lowest misclassification error is selected. To remove the defining MetS, such as (triglycerides, HDL, systolic blood possibility of spurious associations due to missing geno- pressure, diastolic blood pressure, fasting blood sugar and types, we only included cases and controls for which waist circumstance, as well as baseline BMI and BMI changes genotypes were called for all SNPs. Moreover, we tested all after AAP treatment. For multi-marker analysis, we used possible two- to five-loci interactions using 10-fold cross- Haploview v4.0 to the status of pair-wise linkage disequili- validation in an exhaustive search, which considers all brium between the studied polymorphisms, and used possible SNP combinations. GMDR is a non-parametric data Unphased 3.0.1323 to evaluate the haplotypic association mining approach that can apply to both continuous and with MetS in a sliding-window fusion. The overall signifi- dichotomous phenotypes.21 In addition, GMDR permits cance in each window size was determined after 1000 adjustment for discrete and quantitative covariates in permutation procedures to correct for the multiple haplo- various population-based studies with unbalanced case– types tested. To further correct the inflation of significance control subjects.21 The GMDR software provides a number because of the multiple comparisons in genetic analysis, we of output parameters, including cross-validation consis- considered all single-marker and haplotype-based tests and tency, the testing balanced accuracy and the sign test, to assumed a false discovery rate less than 0.05 as significant assess each selected interaction.21 The cross-validation after correction.24 consistency score is a measure of the degree of consistency To investigate gene–gene interactions, we employed the with which the selected interaction is identified as the best GMDR method, which is described in detail elsewhere.21 model among all possibilities considered. Furthermore, the Briefly, the n-dimensional space formed by a given set of testing balanced accuracy is a measure of the degree to SNPs is reduced to a single dimension to analyze n-way which the interaction accurately predicts case–control interactions. And each class is classified as either high risk or status with scores between 0.50 (indicating that the model low risk according to the relative proportion of cases (MetS) predicts no better than chance) and 1.00 (indicating perfect

The Pharmacogenomics Journal INSIGs and metabolic syndrome Y-J Liou et al 58

prediction). Finally, we provided gender, age, baseline BMI, SNPs and the two INSIG1 SNPs did not show any association duration and type of current AAP exposure used as with MetS. covariates in our gene–gene interaction analyses. Permuta- We also analyzed the association of the nine SNPs on the tion testing obtains empirical P-values of prediction accu- quantitative serum factors of defining MetS (triglycerides, racy as a benchmark based on 1000 shuffles. In addition, we HDL, systolic blood pressure, diastolic blood pressure, performed logistic regression models to confirm the results fasting blood sugar and waist circumstance), and also from GMDR analyses, with gender, age, baseline BMI, type examined the association of baseline BMI and BMI changes and duration of current AAP exposure as the covariates. after AAP treatment with analysis of covariance, as shown The powers for the studied variants contributing to in Supplementary Table 4. The MetS-associated SNP various levels of odds ratio (OR) are summarized in rs11123469 had significant effect on serum HDL levels Supplementary Table 2. Briefly, for a risk allele of MetS with across AAP treatment: patients with rs11123469-TT had OR ¼ 1.2, the study has powers ranging from 0.13 B0.20, significantly higher HDL levels than those with rs11123469- 0.45 B0.70 at OR ¼ 1.5 and 0.89 B0.99 at OR ¼ 2.0. CC, post hoc Bonferroni P-value ¼ 0.005). With regard to other studied SNPs, there were no significant differences in Results the mean of the eight quantitative factors between different genotypes (Supplementary Table 4). Clinical characteristics This study sample consisted of 456 patients treated with Haplotype analysis clozapine, olanzapine or risperidone. The prevalence of The association of the genetic sequences with MetS was MetS was 22.8% (104/456) in the total group, and the rates further investigated by looking into the INSIG1 and INSIG2 for the subgroups were 27.5%, 24.2% and 18.0% for patients haplotypes. Pairwise linkage disequilibrium estimation of taking clozapine, olanzapine and risperidone, respectively the nine SNPs is showed in the Supplementary Figure 1 (P ¼ 0.094) (Table 1). The mean age, sex distribution and and 2. The results of the case–control haplotype analysis duration of AAPs used were similar between patients with between groups are presented in Table 3 and Supplementary and without MetS, however, patients with MetS had a larger Table 5. The overall haplotype distributions between groups baseline BMI than patients without MetS (Po0.001). were significantly different in the two, three and four- marker windows by permutation analysis and by correction Single polymorphism analysis for multiple testing (Supplementary Table 5, all permutation Two INSIG1 SNPs and seven INSIG2 SNPs were genotyped in Po0.05; false discovery rate for all the three permutation this study. The genotype distributions of the nine SNPs for P ¼ 0.043). Further analysis for the individual haplotype in the patient with and without MetS are shown in Table 2. each of the three windows identified several haplotypes Most of the genotype distribution of the nine SNPs in the significantly associated with MetS (Table 3). The most MetS and non-MetS were in Hardy–Weinberg equilibrium significant haplotype was the C-C-C type of rs11123469- with the exception of rs2161829 in the non-MetS that was rs10185316- rs1559509 that was more frequent in the deviated significantly from Hardy–Weinberg equilibrium patients with MetS (Table 3, P ¼ 0.0023, false discovery (Supplementary Table 3). The analysis for single locus effects rate ¼ 0.043). revealed that one INSIG2 SNP (rs11123469) had a statisti- cally significant association with MetS, when analyzed by Multi-loci interaction analysis genotype frequency that the INSIG2 rs11123469-C homo- The GMDR analysis was used to assess the impacts of zygous genotype was found to be more frequent in the combinations between the nine SNPs, including gender, age, patients with MetS than those without MetS (P ¼ 0.00011, baseline BMI, duration of current AAP exposure and type of false discovery rate ¼ 0.043) (Table 2). Comparing with the AAPs as covariates because these factors may affect the risk TT carriers, the OR for rs11123469-C homozygotes to have of MetS. Table 4 summarizes the results obtained from MetS was 5.10 (95% CI ¼ 2.23 B11.70). The other six INSIG2 GMDR analysis for 2-locus to 5-locus model with covariate

Table 3 INSIG2 haplotypes with significant different frequency in schizophrenia with or without metabolic syndrome (MetS)

SNP combination Haplotype MetS, % Non-MetS, % P s7566605-rs11123469 C-T 0.3149 0.4032 0.039 G-C 0.3351 0.241 0.011 rs11123469-rs10185316 T-C 0.4745 0.5765 0.011 C-C 0.3673 0.2515 0.0048 rs11123469-rs10185316-rs1559509 T-C-C 0.4579 0.5797 0.0026 C-C-C 0.3711 0.2455 0.0023 s7566605-rs11123469-rs10185316-rs1559509 G-C-C-C 0.3455 0.2391 0.0063

Abbreviations: MetS, metabolic syndrome; SNP, single-nucleotide polymorphism.

The Pharmacogenomics Journal INSIGs and metabolic syndrome Y-J Liou et al 59

Table 4 Best gene–gene interaction models identified by the GMDR method with adjustment for gender, age, baseline BMI, duration of current atypical antipsychotic exposure and type of atypical antipsychotic drugs

Locus Best combination Cross-validation Testing P* no. consistency accuracy (%)

2 INSIG1 rs9767875, INSIG2 rs7566605 10/10 55.58 0.125 3 INSIG1 rs9767875, INSIG2 rs1559509, INSIG2 rs2161829 10/10 62.29 0.003 4 INSIG1 rs9767875, INSIG2 rs7566605, INSIG2 rs1559509, INSIG2 rs2161829 10/10 60.89 0.012 5 INSIG1 rs9767875, INSIG2 rs7566605, INSIG2 rs1559509, INSIG2 rs2161829, 10/10 57.56 0.081 INSIG2 rs17047757

Abbreviations: BMI, body mass index; GMDR, generalized multifactor dimensionality reduction. *P-value based on 1000 permutations. adjustment. There was a significant 3-locus model involving microarray approach, they demonstrated that the anti- INSIG1 rs9767875 and INSIG2 (rs1559509 and rs2161829), psychotic drugs clozapine and haloperidol both activated indicating a potential gene–gene interaction between the SREBP system.25 In another study, using cell cultures of INSIG1 and INSIG2 (Table 4). Moreover, there was a signi- human liver cells, they found clozapine, which has been ficant 4-locus model involving INSIG1 rs9767875 and strongly associated with metabolic adverse effects, has a very INSIG2 (rs7566605, rs1559509 and rs2161829). Overall, the pronounced activation of SREBP.26 Ziprasidone, however, 3-locus model had the highest level of testing accuracy which has less metabolic adverse effects, did hardly (62.29%) and showed good cross-validation consistency stimulate the SREBP system. In a recent study using female (10/10). Therefore, we chose the 3-locus model as the best rats, they demonstrated that acute clozapine exposure GMDR model. Furthermore, the significant interactions among affected SREBP-regulated lipid biosynthesis, as well as other the above 3-locus and 4-locus models were confirmed by lipid homeostasis pathways, suggesting that such drug- logistic regression models (Po0.001 and o0.001, respectively). induced effects on lipid metabolism may be related to the metabolic adverse effects associated with clozapine.27 In 2007, Skelly et al. investigated the association between Discussion rs7566605 and 10 additional tagging SNPs near INSIG2 with BMI in 756 participants in CATIE study, a double-blinded The results of our study showed an overall prevalence of randomized clinical trial of typical and AAPs.19 They were MetS of 22.8% among the 456 Chinese schizophrenic unable to find the association of rs7566605 or other SNPs patients treated with AAPs. The aims of the present study investigated with BMI. However, recently, Steen and collea- was to examine not only the main effects but also epistatic gue reported a strong association (P ¼ 0.0003–0.00007) effects of the INSIG1 and INSIG2 genes in the risk of MetS in between three markers localized within or near the INSIG2 patients treated with AAPs. Among the nine SNPs tested, we gene (rs17587100, rs10490624 and rs17047764) and anti- found, for the rs11123469 CC carriers, 45.0% met MetS psychotic-related weight gain from the time before the criteria, compared with 18.6% in the TT genotype group, initiation of antipsychotic treatment (BMI-1) to the time giving an OR ¼ 5.10, (95% CI ¼ 2.23B11.70, P ¼ 0.00011). before initiation of clozapine treatment (BMI-2).20 The three Thus, for the CC homozygotes, the risk was five times SNPs (rs17587100, rs10490624 and rs17047764) were also greater to have MetS in patients treated with AAPs. Our found to be associated with BMI changes after 12-week finding might imply that the associated marker, rs11123469, clozapine treatment (BMI-3 versus BMI-2).20 In the present may have effects on INSIG2 functioning, and contribute to study, none of the studied INSIG2 SNPs, including the the underlying pathophysiology of MetS in patients treated obesity-associated SNP, rs7566605 were found to have with AAPs directly. For instance, the significant rs11123469 association with BMI changes (Supplementary Table 4). This was subjected to in silico prediction for its function with is consistent with the result of Skelly et al.19 supporting the FastSNP (http://fastsnp.ibms.sinica.edu.tw/), an online tool genetic variation in INSIG2 has no effect on body weight for functional analysis and prioritizing SNP, and found that change across AAP treatment, but partially conflicts with the the C-to-T substitution on the polymorphism creates a report of Le Hellard et al.20 The discrepancy might be multi- binding site for transcription factor octamer-binding factor factorial: (1) studied population with different antipsychotic 1. On the other hand, it is also possible that the association used across studies: olanzapine, risperidone, quetiapine of INSIG2 rs11123469 may not have real biological effects and ziprasidone in Skelly et al.,19 typical antipsychotics and it is most likely in linkage disequilibrium with the alone, AAP alone, combined typical and AAPs and clozapine responsible variants that harbor the biological effects on the in Le Hellard et al.20 and risperidone, olanzapine and reported metabolic phenotypes. clozapine in our study (Table 1); (2) various follow-up The pioneering work of Steen and colleagues has high- duration across antipsychotic treatment: average 18 months lighted the possible effect of the SREBP pathway in in Skelly et al.,19 average 2.7 years before and 12 weeks after antipsychotic-induced metabolic adverse effects. Using a clozapine treatment in Le Hellard et al.,20 and average

The Pharmacogenomics Journal INSIGs and metabolic syndrome Y-J Liou et al 60

45 months in our studies (Table 1), and (3) ethnic difference we only included nine genetic variations in this study. in the minor allele frequency of studied SNP: the minor Therefore, the contributions of other markers in the INSIG1 allele frequency of the three weight-change-associated SNPs and INSIG2 genes should be further examined in future in Le Hellard et al.20 are all less than 10% in Han population work. Finally, INSIG proteins interact with SREBPs and SCAP according to the information in the International HapMap to control the metabolism of cholesterol and fatty acids. Project. Despite of the discrepancy on body weight change Whether genetic variants of these proteins may interact to after antipsychotic treatment between studies, the observa- influence the MetS in patients treated with AAPs needs tion that patients with rs11123469-TT had significantly further exploration. higher HDL levels than those with rs11123469-CC (Supple- In conclusion, our study has tested the association mentary Table 4, post hoc Bonferroni P-value ¼ 0.005) between INSIG1and INSIG2 genetic polymorphisms and indirectly supports that Insig2 as an quantitative trait locus MetS in schizophrenic patients treated with AAPs. Our for mouse serum cholesterol levels.16 The positive associa- findings suggested that INSIG2 gene might contribute to the tion between INSIG2 genetic variants and MetS in patients risk of MetS independently or in an interactive manner with treated with AAPs in this study further supports the role of INSIG1. Independent replications in other population are SREBP pathway in AAP-associated metabolic adverse effects. needed to confirm the role of the INSIG polymorphisms in Another major finding of this study is that, by the GMDR MetS in patients treated with AAPs. analyses, we further established a two-gene interaction model between INSIG1 and INSIG2 genes in MetS. INSIG1 and INSIG2 interactions were found in the significant Conflict of interest 3-locus and 4-locus gene–gene interaction models. Further- more, the significant interactions among the above 3-locus The authors declare no conflict of interest. and 4-locus models were confirmed by logistic regression models (Po0.001 and o0.001, respectively), adjusting for Acknowledgments gender, age, baseline BMI, duration and type of current AAP exposure used as covariates. GMDR is a non-parametric data This work was supported by grant V98C1–061, V98C1-055 from mining approach that can apply to both continuous and Taipei Veterans General Hospital, Taiwan, and grants NSC 95-2314- dichotomous phenotypes. In addition, GMDR permits B-075 –011 and NSC 97-2314-B-075-001-MY3 from National Science Council Grant, Taiwan. The authors also thank Jer-Yuan Wu, the adjustment for discrete and quantitative covariates in National Genomic Center and the National Clinical Core at various population-based studies with unbalanced case– Academia Sinica, Taiwan, for genotyping support. control subjects. The results from the GMDR analyses confirmed our hypothesis that the genes from INSIG1 and References INSIG2 may have interactive effects on risk of MetS in patients treated with AAPs. Our findings also suggested that 1 DE Hert M, Schreurs V, Vancampfort D, VAN Winkel R. Metabolic the use of multiple statistical approaches could be a better syndrome in people with schizophrenia: a review. World Psychiatry strategy to elucidate complex gene interaction, such as drug- 2009; 8: 15–22. therapeutic effects. Beside the statistical significance, the 2 Meyer JM, Stahl SM. The metabolic syndrome and schizophrenia. Acta Psychiatr Scand 2009; 119:4–14. potential biological mechanism under this interaction 3 Newcomer JW. Antipsychotic medications: metabolic and cardio- model was our concern. The protein encoded by INSIG2 is vascular risk. J Clin Psychiatry 2007; 68(Suppl 4): 8–13. highly similar to the protein product encoded by gene 4 De Hert M, Schreurs V, Sweers K, Van Eyck D, Hanssens L, Sinko S et al. INSIG1. Both INSIG proteins are endoplasmic reticulum Typical and atypical antipsychotics differentially affect long-term incidence rates of the metabolic syndrome in first-episode patients proteins that block the processing of SREBPs by binding to with schizophrenia: a retrospective chart review. Schizophr Res 2008; SCAP, and thus prevent SCAP from escorting SREBPs to 101: 295–303. the Golgi. Data from knockout study showed that INSIG1 5 Medved V, Kuzman MR, Jovanovic N, Grubisin J, Kuzman T. Metabolic and INSIG2 mouse appear grossly normal, however, the syndrome in female patients with schizophrenia treated with second generation antipsychotics: a 3-month follow-up. J Psychopharmacol combined disruption of both genes was associated with 2009; 23: 915–922. over-accumulated cholesterol and triglycerides in liver.28 6 Correll CU, Frederickson AM, Kane JM, Manu P. Equally increased risk This finding suggested that both proteins may interact or for metabolic syndrome in patients with bipolar disorder and schizo- compensate with each other in the regulation of cholesterol phrenia treated with second-generation antipsychotics. Bipolar Disord 2008; 10: 788–797. and triglyceride synthesis which is in line with our findings. 7 Correll CU, Frederickson AM, Kane JM, Manu P. Does antipsychotic The notable strength of the present study was that, during polypharmacy increase the risk for metabolic syndrome? Schizophr Res the entire study period, these patients were hospitalized, 2007; 89: 91–100. remained on the same antipsychotic with optimal control of 8 Bai YM, Chen TT, Yang WS, Chi YC, Lin CC, Liou YJ et al. Association of adiponectin and metabolic syndrome among patients taking atypical drug compliance, and had the same diet and activity. At the antipsychotics for schizophrenia: a cohort study. Schizophr Res 2009; same time, there are several limitations of this study. First, 111: 1–8. the positive associations in this study may have simply been 9 Lamberti JS, Olson D, Crilly JF, Olivares T, Williams GC, Tu X et al. because of chance or a stratification effect in the sample Prevalence of the metabolic syndrome among patients receiving clozapine. Am J Psychiatry 2006; 163: 1273–1276. collection. The resolution of this suggestion requires 10 Tsai SJ, Hong CJ. Pharmacogenetic studies of psychotropic drug- replication in future studies with different samples. Second, induced adverse effects. Current Pharmacogenomics 2005; 3: 157–164.

The Pharmacogenomics Journal INSIGs and metabolic syndrome Y-J Liou et al 61

11 Mulder H, Franke B, van der-Beek van der AA, Arends J, Wilmink FW, (INSIG2) and weight gain in a German sample of antipsychotic-treated Scheffer H et al. The association between HTR2C gene polymorphisms schizophrenic patients: perturbation of SREBP-controlled lipogenesis and the metabolic syndrome in patients with schizophrenia. J Clin in drug-r-lated metabolic adverse effects? Mol Psychiatry 2009; 14: Psychopharmacol 2007; 27: 338–343. 308–317. 12 Mulder H, Cohen D, Scheffer H, Gispen-de Wied C, Arends J, Wilmink 21 Lou XY, Chen GB, Yan L, Ma MJZ, Zhu J, Elston RC et al. A g neralized FW et al. HTR2C gene polymorphisms and the metabolic syndrome in combinatorial approach for retecting gene-by-gene and gene- patients with schizophrenia: a replication study. J Clin Psychopharmacol by-environment i t-ractions withtapplication to nicotine d pendence. 2009; 29: 16–20. Am J Hum Genet 2007; 80: 1125–1137. 13 Ellingrod VL, Miller del D, Taylor SF, Moline J, Holman T, Kerr J. Metabolic 22 American Psychiatric Association 1994. Diagnostic and Statistical Manual syndrome and insulin resistance in schizophrenia patients receiving of Mental Disorders, 4th edn (DSM-IV). American Psychiatric Association: antipsychotics genotyped for the methylenetetrahydrofolate reductase Washington, DC. (MTHFR) 677C/T and 1298A/C variants. Schizophr Res 2008; 98: 47–54. 23 Dudbridge F. Likelihood-based association analysis for nuclear families 14 Yang T, Espenshade PJ, Wright ME, Yabe D, Gong Y, Aebersold R et al. and unrelated subjects with missing genotype data. Hum Hered 2008; Crucial step in cholesterol homeostasis: sterols promote binding of 66: 87–98. SCAP to INSIG-1, a membrane protein that facilitates retention of 24 Storey JD. A direct approach to false discovery rate. J R Statist Soc B SREBPs in ER. Cell 2002; 110: 489–500. 2002; 64: 479. 15 Yabe D, Brown MS, Goldstein JL. Insig-2, a second endoplasmic reticulum 25 Fernø J, Raeder MB, Vik-Mo AO, Skrede S, Glambek M, Tronstad KJ et al. protein that binds SCAP and blocks export of sterol regulatory element- Antipsychotic drugs activate SREBP-regulated expression of lipid binding proteins. Proc Natl Acad Sci USA 2002; 99: 12753–12758. biosynthetic genes in cultured human glioma cells: a novel mechanism 16 Cervino AC, Li G, Edwards S, Zhu J, Laurie C, Tokiwa G et al. Integrating of action? Pharmacogenomics J 2005; 5: 298–304. QTL and high-density SNP analyses in mice to identify Insig2 as a suscep- 26 Raeder MB, Fernø J, Vik-Mo AO, Steen VM. SREBP activation by tibility gene for plasma cholesterol levels. Genomics 2005; 86:505–517. antipsychotic- and antidepressant-drugs in cultured human liver 17 Liu JP. New functions of cholesterol binding proteins. Mol Cell Endocrinol cells: relevance for metabolic side-effects? Mol Cell Biochem 2006; 2009; 303: 1–6. 289: 167–173. 18 Herbert A, Gerry NP, McQueen MB, Heid IM, Pfeufer A, Illig T et al. 27 Fernø J, Vik-Mo AO, Jassim G, Ha˚vik B, Berge K, Skrede S et al. Acute A common genetic variant is associated with adult and childhood clozapine exposure in vivo induces lipid accumulation and marked obesity. Science 2006; 312: 279–283. sequential changes in the expression of SREBP, PPAR, and LXR target 19 Skelly T, Pinheiro AP, Lange LA, Sullivan PF. Is rs7566605, a SNP near genes in rat liver. Psychopharmacology 2009; 203: 73–84. INSIG2, associated with body mass in a randomized clinical trial of 28 Engelking LJ, Liang G, Hammer RE, Takaishi K, Kuriyama H, Evers BM antipsychotics in schizophrenia? Mol Psychiatry 2007; 12: 321–322. et al. Schoenheimer effect explained—feedback regulation of cholesterol 20 Le Hellard S, Theisen FM, Haberhausen M, Raeder MB, Fernø J, synthesis in mice mediated by Insig proteins. J Clin Invest 2005; 115: Gebhardt S et al. Association between the insulin-induced gene 2 2489–2498.

Supplementary Information accompanies the paper on the The Pharmacogenomics Journal website (http://www.nature.com/tpj)

The Pharmacogenomics Journal