Molecular Psychiatry (2009) 14, 308–317 & 2009 Nature Publishing Group All rights reserved 1359-4184/09 $32.00 www.nature.com/mp ORIGINAL ARTICLE Association between the -induced 2(INSIG2) and weight gain in a German sample of antipsychotic-treated schizophrenic patients: perturbation of SREBP-controlled lipogenesis in drug-related metabolic adverse effects? S Le Hellard1,2, FM Theisen3, M Haberhausen3, MB Raeder1,2, J Fernø1,2, S Gebhardt4 , A Hinney5, H Remschmidt3, JC Krieg4, C Mehler-Wex6,MMNo¨then7, J Hebebrand5 and VM Steen1,2 1Dr Einar Martens’ Research Group for Biological Psychiatry, Center for Medical Genetics and Molecular Medicine, Haukeland University Hospital, Bergen, Norway; 2Department of Clinical Medicine, Bergen Mental Health Research Center, University of Bergen, Bergen, Norway; 3Department of Child and Adolescent Psychiatry and Psychotherapy, Philipps University, Marburg, Germany; 4Department of Psychiatry and Psychotherapy, Philipps University, Marburg, Germany; 5Department of Child and Adolescent Psychiatry, University of Duisburg-Essen, Essen, Germany; 6Department of Child and Adolescent Psychiatry and Psychotherapy, University of Ulm, Ulm, Germany and 7Department of Genomics, Life & Brain Center, University of Bonn, Bonn, Germany

Atypical antipsychotics are nowadays the most widely used drugs to treat schizophrenia and other psychosis. Unfortunately, some of them can cause major metabolic adverse effects, such as weight gain, dyslipidemia and type 2 diabetes. The underlying lipogenic mechanisms of the antipsychotic drugs are not known, but several studies have focused on a central effect in the hypothalamic control of appetite regulation and energy expenditure. In a functional convergent genomic approach we recently used a cellular model and demonstrated that orexigenic antipsychotics that induce weight gain activate the expression of lipid biosynthesis controlled by the sterol regulatory element-binding (SREBP) transcription factors. We therefore hypothesized that the major genes involved in the SREBP activation of fatty acids and cholesterol production (SREBF1, SREBF2, SCAP, INSIG1 and INSIG2) would be strong candidate genes for interindividual variation in drug-induced weight gain. We genotyped a total of 44 HapMap-selected tagging single nucleotide polymorphisms in a sample of 160 German patients with schizophrenia that had been monitored with respect to changes in body mass index during antipsychotic drug treatment. We found a strong association (P = 0.0003–0.00007) between three markers localized within or near the INSIG2 gene (rs17587100, rs10490624 and rs17047764) and antipsychotic-related weight gain. Our finding is supported by the recent involvement of the INSIG2 gene in obesity in the general population and implicates SREBP-controlled lipogenesis in drug-induced metabolic adverse effects. Molecular Psychiatry (2009) 14, 308–317; doi:10.1038/sj.mp.4002133; published online 15 January 2008 Keywords: functional convergent genomics; metabolic syndrome; cholesterol; gene; antipsychotic

Introduction propensity to cause a variety of metabolic adverse effects. These include dyslipidemia, weight gain, type 2 Antipsychotic drugs represent a cornerstone in the diabetes and the metabolic syndrome,1–5 with con- treatment of schizophrenia. Unfortunately, many sequent increased risk of cardiovascular disorders.6,7 atypical antipsychotics but also some typical drugs In addition, the weight gain may reduce treatment have recently been much criticized because of their compliance in patients that otherwise respond efficiently to treatment. Major research efforts have Correspondence: Dr S Le Hellard, Dr Einar Martens’ Research therefore been launched to better understand and Group for Biological Psychiatry, Center for Medical Genetics and prevent this problem.6,8 Molecular Medicine, Haukeland University Hospital, N-5021, Various mechanisms underlying these drug-in- Bergen, Norway. duced metabolic adverse effects have been proposed, E-mail: [email protected] Received 19 August 2007; revised 5 October 2007; accepted 8 most of them focused around the hypothalamic October 2007; published online 15 January 2008 control of appetite regulation and energy expenditure INSIG2 and antipsychotic-related weight gain S Le Hellard et al 309 modulated through histamine and serotonin recep- tors. An association between the affinity of antipsy- chotic drugs to histamine H1- and serotonin 5-HT2C receptors and their potency to induce weight gain was described.9,10 Furthermore, Kim et al.11 observed that orexigenic antipsychotics (for example, clozapine and olanzapine) but not weight neutral drugs (for exam- ple, ziprasidone) activate the appetite-stimulating AMP-kinase in the hypothalamus through histamine H1 receptors. The weight gain induced by antipsychotic drugs is highly variable among individuals,12 and a twin study has provided evidence that the initial change in body mass index (BMI) during treatment with antipsycho- Figure 1 Activation of cellular lipogenesis by the sterol tics is more strongly correlated between monozygous regulatory element-binding protein (SREBP) transcription twins as compared to same-sex sib pairs,13 suggesting factor system. Cellular sterol levels control the SREBP that genetic factors play a major role. Several transcription factor activity. The SREBPs (red) are synthe- pharmacogenetic studies have investigated the role sized as inactive precursors in the endoplasmic reticulum of different plausible candidate genes in antipsycho- (ER), they reside in a complex with the SREBP-cleavage activating protein (SCAP) (green) and the insulin-induced tic-mediated weight gain (for review, see Mu¨ ller and 14 15 gene (INSIG) (two different isoforms, INSIG1 and INSIG2) Kennedy and Chagnon 2006 ). Most of the reports (light blue). In the presence of sterols an interaction have failed to identify predisposing factors, but a between SCAP and INSIG retains the SCAP/SREBP complex promoter variant (À759C/T) of the serotonin 5-HT2C in the ER. When sterol levels are low, SCAP undergoes a receptor gene seems to be associated with weight gain conformational change and is released from INSIG. SCAP in antipsychotic-treated patients in several popula- assists the transport of the 120 kDa precursor SREBP to the tions,16–20 although a subgroup of the cohort used in Golgi apparatus as the initial step in SREBP activation, the present study failed to show association with this followed by a two-step proteolytic cleavage involving the promoter variant.21 Golgi-specific S1P and S2P proteases, thus releasing a 60– In an attempt to search for alternative explanations, 70 kDa transcriptionally active N-terminal domain. The mature transcriptionally active basic-helix-loop-helix (red we have used a cellular model of cultured human hexagonal) is subsequently translocated to the nucleus glial and liver cells exposed to antipsychotics. We where it activates expression of genes involved in choles- demonstrated that such drugs activate the sterol terol and fatty acid biosynthesis via binding to their SRE. regulatory element-binding protein (SREBP) trans- cription factors, with subsequent transcriptional upregulation of downstream cholesterol and fatty acid biosynthesis.22,23 There were marked differences function, but SREBP1c controls the expression of between the drugs, with clozapine and chlorproma- fatty acid biosynthesis genes,31 SREBP2 mainly zine being the most potent SREBP activators in a regulates cholesterol biosynthetic genes,32 whereas context of therapeutically relevant concentrations, SREBP1a regulates the expression of both cholesterol whereas ziprasidone had minimal effect.24 We there- and fatty acid biosynthesis genes, with the later more fore hypothesized that antipsychotic-induced stimu- efficiently activated.33 lation of cellular lipogenesis could be involved in the Based on our demonstration of antipsychotic- increased risk of dyslipidemia and weight gain. induced transcriptional activation of the SREBP The SREBP transcription factors normally regulate system and lipid biosynthesis in cultured human the expression of a number of genes involved in cells, we considered SREBF1, SREBF2, SCAP, INSIG1 biosynthesis and uptake of cholesterol, fatty acids, and INSIG2 as promising candidate genes for study- triglycerides and phospholipids.25–28 The SREBPs are ing interindividual variation in weight gain in synthesized in the endoplasmic reticulum (ER) where patients treated with antipsychotic drugs. Interest- they reside and form a complex with the SREBP- ingly, the SREBF1 and INSIG2 genes have recently cleavage activating protein (SCAP) and the insulin- been associated with obesity in several different induced gene (INSIG) proteins29 (which has two populations.34–36 The aim of this study was different isoforms, INSIG1 and INSIG2). At low sterol therefore to examine tagging single nucleotide poly- levels, the SREBPs are released from the membrane, morphisms (tagSNPs) in these five SREBP-linked activated by cleavage30 and translocated to the genes in a sample of clozapine-treated patients with nucleus (Figure 1) where they bind to the SRE in schizophrenia. the promoter of numerous SREBP target genes, hence activating lipogenic There are two Materials and methods different SREBP isoforms, SREBP1 (with two splice variants, SREBP1a and -1c, both encoded by the Clinical sample and DNA extraction SREBF1 gene) and SREBP2 (encoded by SREBF2 We studied a total of 160 German patients (97 male gene). The different SREBP isoforms overlap in and 63 female; 21.9±8.9 years (mean±s.d.); range

Molecular Psychiatry INSIG2 and antipsychotic-related weight gain S Le Hellard et al 310 10–64 years) who met ICD-10 criteria for schizophre- analysis for INSIG1, 53 for INSIG2, 17 for SCAP, 20 for nia spectrum disorders; all of Central European SREBF1 and 58 for SREBF2. origin. Patients were ascertained at the Department Tagging SNPs were selected using the Tagger of Child and Adolescent Psychiatry and the Depart- function implemented in Haploview with the follow- ment of Psychiatry at the University of Marburg, for ing criteria: aggressive tagging; use two- and three- initiation of clozapine treatment. We have pre- marker haplotypes and r2 threshold 0.8. Five markers viously21,37,38 reported on the characteristics of a were suitable for analysis for INSIG1, 23 for INSIG2,3 subsample of patients included in the current study. for SCAP, 6 for SREBF1 and 14 for SREBF2. The Briefly, all patients were first time users of clozapine, design assay failed for one marker that could not be of whom 98 had been treated with typical antipsy- replaced by equivalent markers in linkage disequili- chotics prior to the use of clozapine, 26 patients had brium (LD). The markers were genotyped on been pretreated > 2 weeks with an atypical antipsy- a Sequenom Massarray platform at the Center chotic only, and 36 had been pretreated with both for Integrative Genetics (http://www.umb.no/, the atypical and typical antipsychotics, the mean period national SNP typing platform in Norway, Functional of antipsychotic treatment prior to entry in the study Genomics in Norway (FUGE) funded). was 2.7±5.3 years (range 0.02–31 years). The mean In addition, the marker rs756605, localized near clozapine dose was 285±121 mg per day (range 50– INSIG2 and recently implicated in susceptibility to 800 mg per day). BMI of each patient was measured obesity,35 was genotyped by sequencing PCR products just before initiation of clozapine treatment (BMI-2). amplified by: the primers 50-TGATCGCTGCGCTGC In addition, BMI at the start of the first antipsychotic TATGG-30 and 50-TGAGAGTCAGTGCGATGTCC-30, treatment (BMI-1) was obtained retrospectively using AmpliTaq Gold (Applied Biosystems, Foster through medical records, and a third BMI value City, CA, USA) according to the manufacturer’s (BMI-3) was measured prospectively 12±1.2 weeks instructions with 25 cycles: 94 1C for 10 s, 55 1C for after initiation of clozapine treatment. We are aware 30 s, 72 1C for 30 s (initial denaturation 94 1C of the problems inherent in investigations based on for 10 min). The PCR products were sequenced with the analysis of medical records. Nevertheless, in the BigDye v3.1 (Applied Biosystems) according to the light of the restrictions attached to the use of manufacturer’s instructions and the sequences were clozapine and the personal responsibility of physi- aligned with phredPhrap program and read in cians, patients were carefully observed and clinical Consed39 (http://bozeman.mbt.washington.edu/consed/ relevant data carefully noted. Mean BMI values were consed.html#documentation). 21.3±3.2 (range 15.4–35.1), 23.5±3.8 (range 16.9– 35.6) and 25.4±3.8 (range 16.6–39.7) for the BMI-1, Data analysis BMI-2 and BMI-3 recordings, respectively. Exclusion Genotyping results were analyzed using the Helix criteria were other medical or neurological disorders Tree software (Golden Helix: http://www.goldenhelix. and pregnancy. All patients gave written informed com/index.jsp). Standard w2-test of independence consent; the study was approved by the ethics was used to exclude any markers with deviation from committee of the Philipps University of Marburg, the Hardy–Weinberg equilibrium. The remaining Germany. markers were analyzed as single markers for a linear regression where the values of BMI (BMI-1, -2 and -3) Marker selection and genotyping or the change in BMI (BMI-2 versus BMI-1, BMI-3 The markers were selected using the phase II (Oct 05) versus BMI-2, and BMI-1 versus BMI-3) were the version of HapMap (http://www.hapmap.org/cgi-perl/ outcome predicted by the genotypes. The nominal gbrowse/gbrowse/hapmap_phaseI/). SNP genotype P-values given are uncorrected for multiple testing. data for the CEPH (Utah residents with ancestry from An experiment-wide threshold was calculated using northern and western Europe) were downloaded for the method by Nyholt et al.40 (http://fraser.qimr. the genomic region of each gene plus 10 kb in 50 and 30: edu.au/general/daleN/SNPSpDfraser/). In addition, INSIG1 (chr7q36.3: 154447282–154503683), INSIG2 single marker associations were analyzed similarly (chr2q14.1: 118 931194–118972740), SCAP (chr3p21.31: with the STATA for pc version 9.0 (StataCorp, USA) 47 405 775–47 488 036), SREBF1 (chr17p11.2: 17 905 951– for a linear regression where the change in BMI (see 17 950 890) and SREBF2 (chr22q13.2: 40 462 144–40 above) was the outcome parameter predicted by the 555 411). The most centromeric and telomeric HapMap genotypes. The genotypes were given codes of 0, markers downloaded were INSIG1 rs9692071- 1 and 2 for the co-dominant model; 0, 1 and 1 for the rs12381375; INSIG2: rs11679259-rs11675138; SCAP: dominant model and 0, 0 and 1 for the recessive rs3736177-rs11716763; SREBF1 rs11649804-rs3744115; model. SREBF2: rs4822056-rs2229442. HapMap markers We also calculated haplotype trend regression on were analyzed with Haploview 3.2 (www.broad. two- and three-markers sliding window haplotypes, mit.edu/mpg/haploview) with the following criteria estimated using the expectation-maximization of marker selection: HW P-value cutoff: 0.001; algorithm implemented in Helix Tree (1000 itera- minimum genotypes: 85%; maximum number of tions), using sex, age and BMI-1 as co-variates, for the Mendelian errors: 1; minimum minor allele whole data set. In addition, for the INSIG2 markers frequency: 0.001. Fifteen markers were suitable for haplotype trend regression, we calculated a permuted

Molecular Psychiatry INSIG2 and antipsychotic-related weight gain S Le Hellard et al 311 P-value based on 10 000 shuffles as implemented in highly significant association (P = 0.0003–P = 0.00007) Helix Tree Regression analysis module. This option between three markers localized within or near shows the permuted P-value as the percentile of the the INSIG2 gene (rs17587100, rs10490624, and permutation tests in which the P-value is less or equal rs17047764) and weight gain from the initiation of to the observed P-value (which is included in the set antipsychotic drug treatment (BMI-1) to the time of 10 000 shuffles). In order to avoid misleading of inclusion (BMI-2) (that is, the retrospectively results caused by multiple rare haplotypes, haplo- recorded change in BMI); an additional two markers types with a frequency p5% were clumped together had P-values of 0.005 and 0.002. Three of these into one haplotype. markers were also associated with weight gain under The post hoc sample power was calculated with the clozapine treatment (BMI-3 versus BMI-2, that is, STATA for pc version 9.0 (StataCorp, USA), where prospective record), but the association was not as the sample size was 160, a = 0.05 and effect size was strong as for the BMI-1 versus BMI-2 difference determined post hoc using the observed mean and (P = 0.003–0.04; see Table 1). Due to the remarkably standard deviation for each genotype at each marker low P-values, the results obtained initially with the position. Helix Tree software were double-checked by analy- zing the data with STATA, which confirmed the strong association observed between the three Results markers located in INSIG2 and antipsychotic-induced Marker selection and genotyping weight gain (data not shown; available upon request). Fifty-two tagSNPs in SREBPF1, SREBF2, SCAP, One INSIG2 marker (rs10490624) also demonstrated INSIG1 and INSIG2 were selected using the HapMap association with weight gain over the whole anti- CEPH data. One SNP assay failed on design and could psychotic treatment period (BMI-1 versus BMI-3), not be replaced by any other HapMap SNP. The although the P-values were lower than those seen in remaining tagSNPs were genotyped on a sample of the first set (BMI-1 versus BMI-2). 160 individuals. Five SNPs had genotyping success We observed no significant association with the too low for analysis. Two SNPs (rs12623648 in INSIG2 marker rs7566605 near INSIG2, which has been and rs2413660 in SREBF1) were in Hardy–Weinberg associated with obesity in several independent disequilibrium (P < 0.001), probably due to an error in populations. The single marker analyses of the the selection of the markers as they both appeared to SREBF1, SREBF2, SCAP and INSIG1 genes turned be located in repeat rich regions (the other markers out negative, except for three nominally significant were also checked and none were in repeat regions). SNPs, one in INSIG1 (rs13223383; P = 0.005) and one One marker was monomorphic in our population. In in SREBF2 (rs4822064; P = 0.04) with BMI-2 versus addition, we genotyped by sequencing the SNP BMI-3, and one in SCAP (rs12490383; P = 0.01) with rs7566605, localized 10.034 kb from the start of BMI-1 versus BMI-3 (Table 1). INSIG2, which was reported recently to be associated We performed a considerable number of tests in a with obesity in several populations.35,36,41–48 In total total of five different genes, and although the we analyzed 44 SNPs: 3 SNPs in INSIG1, 21 SNPs in Bonferroni correction is over-conservative (as many INSIG2, 3 SNPs in SCAP, 4 SNPs in SREBF1 and 13 of the tests are correlated), the strongest associations SNPs in SREBF2. observed still remain significant at the Bonferroni experience-wide threshold of P = 0.0011. The experience- Association analysis wide threshold calculated with the Nyholt method, The sample of 160 clozapine-treated patients with P = 0.0012, is actually very similar to the one obtained schizophrenia had been characterized for their BMI at by Bonferroni method. This is expected if the selected three time points: (1) BMI at the very first anti- markers are in low LD and it reflects that 41.3 out of psychotic treatment (BMI-1), (2) BMI at the time of the 43 markers are effectively independent (as entry in the study just prior to the switch to clozapine calculated by the Nyholt method). treatment (BMI-2) and (3) BMI measured after an We calculated a post hoc power at each marker average of 12 weeks (±1.2) on clozapine (BMI-3). We position under a dominant model for BMI changes first performed single marker analysis with the Helix (BMI-1 versus BMI-2 and BMI-2 versus BMI-3), using Tree software. We observed no association between 160 for the sample size, a = 0.05, and the observed the markers tested and weight before initiation of mean and standard deviation for each genotype to antipsychotic treatment (BMI-1). After antipsychotic determine the effect size. Each marker was regarded treatment (BMI-2), we found a modest association as independent of the others. For the markers that with two markers in INSIG2: rs17587100 and were significantly associated with antipsychotic- rs10490624 (P-values 0.01 and 0.006, respectively). related weight gain (see above), the power ranged No marker showed association with BMI-3. from 0.6 for rs17587100 to 0.9 for rs17047764 (BMI-2 The data were then analyzed for association with versus BMI-3). mean change in BMI with a linear regression analysis using sex, age and BMI-1 as co-variates. The results of Haplotype analysis the single marker analysis along with the SNP The four markers associated at the single marker level information are displayed in Table 1. We observed a showed some degree of LD (D0 0.7–1) but the

Molecular Psychiatry INSIG2 and antipsychotic-related weight gain S Le Hellard et al 312 Table 1 Single marker linear regression analysis of association with change in body mass index (BMI), with sex, age and BMI-1 as co-variates

Gene in dbSNP rs SNP Single marker BMI-2 Single marker BMI-3 Single marker BMI-3 proximity numbera localization versus BMI-1 versus BMI-2 versus BMI-1

INSIG1 rs9692071 chr7:154495522 — — — rs9690040 chr7:154477456 — — — rs13223383 chr7:154495522 — 0.005 — INSIG2 rs7566605 chr2:118931160 — — — rs11679259 chr2:118931668 — — — rs11679303 chr2:118931749 — — — rs2422166 chr2:118931821 — — — rs2161830 chr2:118933426 — — — rs17587100 chr2:118933555 0.00007* 0.04 — rs4849676 chr2:118938384 — — — rs12464355 chr2:118944995 — — — rs2042492 chr2:118950069 — — — rs2161829 chr2:118952059 — — — rs889904 chr2:118955616 0.002 0.03 — rs10490624 chr2:118957637 0.00008* 0.007 0.025 rs13409050 chr2:118958915 — — — rs17047764 chr2:118963727 0.0003* 0.003 — rs10207953 chr2:118967046 — — — rs12151787 chr2:118968732 0.005 — — rs11681903 chr2:118968781 — — — rs2113485 chr2:118969291 — — — rs13013460 chr2:118970714 — — — rs11673894 chr2:118971217 — — — rs11675138 chr2:118971529 — — — SCAP rs3736177 chr3:47412003 — — — rs12490383 chr3:47412003 — — 0.01 rs17784714 chr3:47472290 — — — SREBF1 rs8067439 chr17:1798820 — — — rs2297508 chr17:17915883 — — — rs11656665 chr17:17925355 — — — rs1108511 chr17:17927035 — — — SREBF2 rs5758487 chr22:40469434 — — — rs9607849 chr22:40472906 — — — rs7287886 chr22:40481066 — — — rs17002706 chr22:40501616 — — — rs11702960 chr22:40512571 — — — rs4822063 chr22:40519781 — — — rs4822064 chr22:40522545 — 0.04 — rs13055841 chr22:40522692 — — — rs2269658 chr22:40523657 — — — rs9623466 chr22:40524188 — — — rs2267443 chr22:40530493 — — — rs4822066 chr22:40532539 — — — rs2269661 chr22:40539009 — — —

Abbreviations: INSIG, insulin-induced gene; SCAP, SREBP-cleavage activating protein; SNP, single nucleotide polymor- phism. —, not significant (data not shown). *, significant after Bonferroni correction. BMI-1, BMI before the initiation of antipsychotic treatment (retrospectively recorded). BMI-2, BMI measured at time of entry in study, prior to initiation of clozapine treatment. BMI-3, BMI measured after approximately 12 weeks of clozapine treatment. aAccording to HapMap phase II (Oct 05 version).

r2 remained weak (0.1–0.5). We therefore chose to antipsychotic-related weight gain between BMI-1 examine the data also at the haplotype level using a and BMI-2 (Table 2), and between BMI-2 and haplotype trend regression (HTR) analysis with BMI-3 (data not shown). Sex, age and BMI-1 sliding windows of two and three markers and were included as co-variates and we performed

Molecular Psychiatry INSIG2 and antipsychotic-related weight gain S Le Hellard et al 313 Table 2 INSIG2 two- and three- MSW haplotype trend regression for weight gain during first antipsychotic treatment (BMI-1 versus BMI-2) dbSNP rs 2 Two MSW sample Two MSW Three MSW Three MSW number localizationa P-value permutations P-value sample P-value permutations P-value rs7566605 118931160 0.05 0.13 — — rs11679259 118931668 ———— rs11679303 118931749 ———— rs2422166 118931821 — — 0.0008 0.0007 rs2161830 118933426 0.04 0.002 0.003 0.003 rs17587100 118933555 0.001 0.002 0.003 0.003 rs4849676 118938384 ———— rs12464355 118944995 ———— rs2042492 118950069 0.04 0.1 0.0002 0.0001 rs2161829 118952059 0.01 0.02 0.0005 0.0003 rs889904 118955616 0.00004 0.0001 0.00002 0.0001 rs10490624 118957637 0.00002 0.0001 0.0002 0.0001 rs13409050 118958915 0.001 0.0003 0.00003 0.0001 rs17047764 118963727 0.00006 0.002 0.0001 0.0001 rs10207953 118967046 0.009 0.02 0.00009 0.0001 rs12151787 118968732 0.00005 0.0001 0.04 0.06 rs11681903 118968781 ———— rs2113485 118969291 ———— rs13013460 118970714 ———— rs11673894 118971217 ———— rs11675138 118971529 ————

Abbreviation: MSW, markers sliding window. HTR was calculated using sex, age and BMI-1 as co-variates. The permuted P-values displayed were estimated with 10 000 shuffles, including the unpermuted set as one of the tested set. —, not significant (data not shown). aAccording to HapMap phase II (Oct 05 version).

permutation analysis with 10 000 shuffles for the HTR, as compared to the HTR with two-markers, it is INSIG2 markers. No haplotype in the other genes likely that the last marker is beyond the limit of the (SREBF1, SREBF2, SCAP and INSIG1) showed haplotype block associated. Six out of these ten significant association (data not shown). markers formed a haplotype block located between Table 2 summarizes the P-values observed in the intron 3 and 30UTR of INSIG2 (see Figure 2). The sample in comparison to the permuted P-value additional four markers from this haplotype block obtained from the 10 000 shuffles for the weight gain were in weaker LD with the other markers. Interest- between BMI-1 and BMI-2 with the INSIG2 markers. ingly, we also observed association between the We observed significant association of two haplotype prospectively observed clozapine-induced weight blocks in INSIG2 and antipsychotic-related increase gain (BMI-3 versus BMI-2) and seven of the same in BMI. The first block included markers rs2161830, markers as for BMI-1-BMI-2 but weaker (P-values rs17587100 and rs4849676 in the two-markers sliding 0.008–0.04; data not shown). window HTR and extended to rs2422166 and rs12464355 in the three-markers sliding window The effect of INSIG2 genotype on weight gain HTR, although the association with these two later In the total sample, mean BMI values were 21.3±3.2 markers probably reflects the association with the (range 15.4–35.1), 23.5±3.8 (range 16.9–35.6) and other markers (Table 2). We imported the genotype 25.4±3.8 (range 16.6–39.7) for the BMI-1, BMI-2 and data in Haploview to generate solid spine of LD BMI-3 recordings, respectively. The genotype distri- blocks and observed that these five markers are part of bution of the most significant weight gain-associated a common block located in 50 of the transcription INSIG2 marker (rs17587100, A/C polymorphism; start of INSIG2 (Figure 2). The second block Table 1) was 84.8 % AA (N = 129), 14.4% AC included nine markers (rs2402492, rs2161829, (N = 22) and 0.8 % CC (N = 1). Interestingly, the mean rs889904, rs10490624, rs13409050, rs17047764, increase in BMI (difference in BMI-1 versus BMI-2; rs10207953, rs12151787 and rs11681903) under the DBMI-1–2) was 1.8±2.3 for the AA carriers as two-markers sliding window HTR and extended to compared to 4.1±3.9 for the AC heterozygotes rs2113485 with the three-markers sliding window. (P = 6.6 Â 10À8), which represents an actual weight However, considering how the P-value becomes less gain of about 5.2±6.6 kg and 11.8±11.2 kg, respec- significant when this last marker is included in the tively, for a person of 170 cm height (the CC

Molecular Psychiatry INSIG2 and antipsychotic-related weight gain S Le Hellard et al 314

118,552,255 118,972,740

Figure 2 Human insulin-induced gene 2 (INSIG2) gene, linkage disequilibrium (LD) map and localization of associated markers. The upper panel represents the INSIG2 genomic organization on . The black and yellow boxes represent non-coding and coding exons, respectively. In the medium panel the tagging single nucleotide polymorphisms (tagSNPs) typed in this study are localized in respect of their genomic distances. The lower panel displays the LD map generated from the 21 tagSNPs genotyped in this study. LD was calculated by Haploview version 2.5 using solid spine of LD > 0.8, (see www.broad.mit.edu/mpg/haploview/using.php#lddisplay for blocks and color scheme definition). The red lines represent the associated haplotypes on the LD map and genomic structure; markers strongly associated are represented by a full line and markers more weakly associated (P > 0.01 or associated in marker sliding window (MSW) only) are represented with a dotted line.

individual had a DBMI-1–2 of 5.1). Among the 25% due to its involvement in antipsychotic-induced fraction of patients with the highest increase in BMI SREBP-mediated activation of lipid biosynthesis in (N = 40; DBMI-1–2, range: 3.5–10.9), the genotype cultured cells, (2) the association was observed with distribution was 67.5% AA (N = 27), 30% AC several markers which are not in complete LD, (3) the (N = 12) and 2.5% CC (N = 1), whereas the 25% association was found in both the retrospective and subjects with the lowest increase in BMI (N = 40; prospective recording of antipsychotic-related weight DBMI-1–2, range: À2.2–0.28) displayed 87.5% AA gain in the sample and (4) INSIG2 has recently been (N = 35), and 12.5% AC (N = 5) genotypes. independently implicated as a susceptibility gene in obesity in several but not all populations (see below). The present study is founded on our recent Discussion observation that antipsychotic drugs stimulate the Here we demonstrate a highly significant association SREBP transcription factors in cultured human glial between several markers within two haplotype blocks and liver cells, with subsequent transcriptional in INSIG2 and weight gain during antipsychotic upregulation of downstream cholesterol and fatty treatment in a German sample of patients with acid biosynthesis.22,23 There are marked differences schizophrenia. The finding is supported by the fact between the drugs, with clozapine and chlorproma- that (1) this candidate gene was selected a priori by us zine being the most potent SREBP activators in a

Molecular Psychiatry INSIG2 and antipsychotic-related weight gain S Le Hellard et al 315 context of therapeutically relevant concentrations, corroborates a recent report that most of the anti- whereas ziprasidone has minimal effect,24 which psychotic-induced weight gain is displayed in the matches the potency of these drugs to induce weight first year of treatment, with much less increase during gain. INSIG2 was one of five selected candidate genes the next 7 years.52 Still, it should be considered due to its involvement in the control of SREBP that the effect we observed at least in part is due to activation, and our present findings suggest that the the influence of INSIG2 on weight gain itself, INSIG2 gene polymorphisms could play a role in independent of drug treatment. In this context, the the interindividual variation in weight gain observed prospective data pertaining to the association for in patients treated with antipsychotic drugs. DBMI-2–3 most likely reflect the effect of medication The role of INSIG2 in drug-induced weight gain is and in particular clozapine. Other limitations in the supported by several studies. A common variant present study are sample size, variability of (rs7566605) upstream of INSIG2 was first reported the antipsychotics used in the first period and the by Herbert et al.35 to be associated with adult and length of this treatment. In terms of genetic associa- childhood obesity. This association has now been tion studies, the number of subjects is rather small, confirmed in five cohorts in a large scaled project by but still represents a comparably large sample for the Lyon et al.,36 but lack of confirmation was observed in specific phenotype. The association we report should three cohorts from this study and in seven additional therefore be valuable, novel information in under- populations.41–48 Although the association of this SNP standing weight gain related to antipsychotic to obesity is not seen in all population it is likely that treatment. Further studies are now needed for this marker reflects the effect of a nearby genetic replication purposes as well as to investigate the factor in some populations. Our study was designed effect of INSIG2 allelic forms on the lipogenic to examine specifically the five SREBP system genes response of cells to antipsychotic exposure. In and we therefore covered the genomic region more addition, it would be interesting to see other pros- thoroughly than in the studies mentioned above (and pective studies looking at the effect of INSIG2 below). The marker rs7566605 did not show associa- polymorphisms on weight gain in populations that tion with any trait we analyzed. have not been treated with drugs. A recent study by Skelly et al.48 looked at the marker rs7566605 (located 10 kb from INSIG2) and 10 other nearby markers in the CATIE sample.49 They Acknowledgments observed a ‘non-significant statistical trend’ for association between rs7566605 and BMI in the The present study is grounded on an initial micro- antipsychotic-treated patients, whereas no associa- array-based gene expression analysis on antipsychotic tion was seen with the other markers they typed. drug action, using the infrastructure provided by the However, out of the 10 markers, only one (rs7589375) Norwegian Microarray Consortium FUGE technology was closer to INSIG2 than rs7566605 (3 kb from platform (www.microarray.no), funded by the FUGE INSIG2), and both markers are in weak LD (r2 < 0.4) programme of the Research Council of Norway. The with the INSIG2 markers that we found to be study has been supported by grants from the Research associated with drug-related weight gain in the Council of Norway (including FUGE grant no. 151904, present study. It is therefore possible that the trend ‘Senter for grunnleggende sykdomsmekanismer’, they observed reflects an association with other Mental Health program and PSYKISKHELSE grant markers within INSIG2. no. 175345), Helse Vest RHF and Dr Einar Martens In the mouse, a high-density SNP analysis identi- Fund. MMN received support for this work from the fied Insig2 as a strong candidate susceptibility gene Alfried Krupp von Bohlen und Halbach-Stiftung. JH for variation in total plasma cholesterol levels.50 From and AH receive funding via the German National a functional point of view, the two INSIG are Genome Net. We are indebted to the patients for their both located in the membrane of the ER, where they participation in this study. We also thank CIGENE, the negatively regulate SREBP activity and cholesterol FUGE/Research Council of Norway-funded national biosynthesis through their sterol-dependent binding SNP platform, for genotyping our samples. to SCAP.29 In contrast to INSIG1, INSIG2 is usually expressed at low levels and SREBP is not required for 51 its expression. It could be speculated that the References different INSIG2 haplotypes are linked to variation in its level of expression, thereby affecting the SREBP- 1 Allison DB, Mentore JL, Heo M, Chandler LP, Cappelleri JC, mediated lipogenic potential of antipsychotic drugs Infante MC et al. Antipsychotic-induced weight gain: a compre- hensive research synthesis. Am J Psychiatry 1999; 156: 1686–1696. in various patients. 2 Lindenmayer JP, Czobor P, Volavka J, Citrome L, Sheitman B, The association observed by us was most McEvoy JP et al. Changes in glucose and cholesterol levels in pronounced for the retrospectively recorded increase patients with schizophrenia treated with typical or atypical in BMI (that is, from the very first use of antipsychotic antipsychotics. Am J Psychiatry 2003; 160: 290–296. 3 Meyer JM, Koro CE. The effects of antipsychotic therapy on serum drugs to the inclusion in the study) but was lipids: a comprehensive review. Schizophr Res 2004; 70: 1–17. also clearly evident for the prospectively measured 4 Saari KM, Lindeman SM, Viilo KM, Isohanni MK, Ja¨rvelin MR, weight gain following clozapine therapy. This trend Laure´nLHet al. A 4-fold risk of metabolic syndrome in patients

Molecular Psychiatry INSIG2 and antipsychotic-related weight gain S Le Hellard et al 316 with schizophrenia: the Northern Finland 1966 Birth Cohort CNS-related cell lines: marked differences between various study. J Clin Psychiatry 2005; 66: 559–563. antipsychotic drugs. BMC Neurosci 2006; 7: p 69. 5 Lamberti JS, Olson D, Crilly JF, Olivares T, Williams GC, Tu X et al. 25 Dietschy JM, Turley SD. Thematic review series: brain Lipids. Prevalence of the metabolic syndrome among patients receiving Cholesterol metabolism in the central nervous system during clozapine. Am J Psychiatry 2006; 163: 1273–1276. early development and in the mature animal. J Lipid Res 2004; 45: 6 Marder SR, Essock SM, Miller AL, Buchanan RW, Casey DE, 1375–1397. Davis JM et al. Physical health monitoring of patients with 26 Edwards PA, Tabor D, Kast HR, Venkateswaran A. Regulation of schizophrenia. Am J Psychiatry 2004; 161: 1334–1349. gene expression by SREBP and SCAP. Biochim Biophys Acta 2000; 7 Osborn DP, Levy G, Nazareth I, Petersen I, Islam A, King MB. 1529: 103–113. Relative risk of cardiovascular and cancer mortality in people with 27 Horton JD, Goldstein JL, Brown MS. SREBPs: activators of the severe mental illness from the United Kingdom’s General Practice complete program of cholesterol and fatty acid synthesis in Research Database. Arch Gen Psychiatry 2007; 64: 242–249. the liver. J Clin Invest 2002; 109: 1125–1131. 8 Arranz MJ, de Leon J. Pharmacogenetics and pharmacogenomics of 28 Nakamura MT, Cheon Y, Li Y, Nara TY. Mechanisms of schizophrenia: a review of last decade of research. Mol Psychiatry regulation of gene expression by fatty acids. Lipids 2004; 39: 2007; 12: 707–747. 1077–1083. 9 Kroeze WK, Hufeisen SJ, Popadak BA, Renock SM, Steinberg S, 29 Yang T, Espenshade PJ, Wright ME, Yabe D, Gong Y, Aebersold R Ernsberger P et al. H1-histamine receptor affinity predicts short- et al. Crucial step in cholesterol homeostasis: sterols promote term weight gain for typical and atypical antipsychotic drugs. binding of SCAP to INSIG-1, a membrane protein that facilitates Neuropsychopharmacology 2003; 28: 519–526. retention of SREBPs in ER. Cell 2002; 110: 489–500. 10 Matsui-Sakata A, Ohtani H, Sawada Y. Receptor occupancy-based 30 Rawson RB. Control of lipid metabolism by regulated intra- analysis of the contributions of various receptors to antipsycho- membrane proteolysis of sterol regulatory element binding tics-induced weight gain and diabetes mellitus. Drug Metab proteins (SREBPs). Biochem Soc Symp 2003; 70:221. Pharmacokinet 2005; 20: 368–378. 31 Shimano H, Horton JD, Shimomura I, Hammer RE, Brown MS, 11 Kim SF, Huang AS, Snowman AM, Teuscher C, Snyder SH. Goldstein JL. Isoform 1c of sterol regulatory element binding Antipsychotic drug-induced weight gain mediated by histamine protein is less active than isoform 1a in livers of transgenic mice H1 receptor-linked activation of hypothalamic AMP-kinase. Proc and in cultured cells. J Clin Invest 1997; 99: 846–854. Natl Acad Sci 2007; 104: 3456–3459. 32 Horton JD, Shimomura I, Brown MS, Hammer RE, Goldstein JL, 12 Basile VS, Masellis M, McIntyre RS, Meltzer HY, Lieberman JA, Shimano H. Activation of cholesterol synthesis in preference to Kennedy JL. Genetic dissection of atypical antipsychotic-induced fatty acid synthesis in liver and adipose tissue of transgenic mice weight gain: novel preliminary data on the pharmacogenetic overproducing sterol regulatory element-binding protein-2. J Clin puzzle. J Clin Psychiatry 2001; 62: 45–66. Invest 1998; 101: 2331–2339. 13 Theisen FM, Gebhardt S, Haberhausen M, Heinzel-Gutenbrunner M, 33 Shimano H, Horton JD, Hammer RE, Shimomura I, Brown MS, Wehmeier PM, Krieg JC et al. Clozapine-induced weight gain: a Goldstein JL. Overproduction of cholesterol and fatty acids causes study in monozygotic twins and same-sex sib pairs. Psychiatry massive liver enlargement in transgenic mice expressing truncated Genet 2005; 15: 285–289. SREBP-1a. J Clin Invest 1996; 98: 1575–1584. 14 Mu¨ ller DJ, Kennedy JL. Genetics of antipsychotic treatment 34 Eberle´ D, Cle´ment K, Meyre D, Sahbatou M, Vaxillaire M, Le Gall A emergent weight gain in schizophrenia. Pharmacogenomics et al. SREBF-1 gene polymorphisms are associated with obesity 2006; 7: 863–887. and type 2 diabetes in French obese and diabetic cohorts. Diabtes 15 Chagnon YC. Susceptibility genes for the side effect of anti- 2004; 53: 2153–2157. psychotics on body weight and obesity. Curr Drug Targets 2006; 7: 35 Herbert A, Gerry NP, McQueen MB, Heid IM, Pfeufer A, Illig T 1681–1695. et al. A common genetic variant is associated with adult and 16 Reynolds GP, Zhang ZJ, Zhang XB. Association of antipsychotic childhood obesity. Science 2006; 312: 279–283. drug-induced weight gain with a 5-HT2C receptor gene poly- 36 Lyon HN, Emilsson V, Hinney A, Heid IM, Lasky-Su J, Zhu X et al. morphism. Lancet 2002; 359: 2086–2087. The association of a SNP upstream of INSIG2 with body mass 17 Reynolds GP, Zhang Z, Zhang X. Polymorphism of the promoter index is reproduced in several but not all cohorts. PLoS Genet region of the serotonin 5-HT(2C) receptor gene and clozapine- 2007; 3: e61. induced weight gain. Am J Psychiatry 2003; 160: 677–679. 37 Bromel T, Blum WF, Ziegler A, Schulz E, Bender M, Fleischhaker 18 Templeman LA, Reynolds GP, Arranz B, San L. Polymorphisms of C et al. Serum leptin levels increase rapidly after initiation of the 5-HT2C receptor and leptin genes are associated with clozapine therapy. Mol Psychiatry 1998; 3: 76–80. antipsychotic drug-induced weight gain in Caucasian subjects 38 Theisen FM, Linden A, Geller F, Schafer H, Martin M, with a first-episode psychosis. Pharmacogenet 2005; 15: 195–200. Remschmidt H et al. Prevalence of obesity in adolescent and 19 Lane HY, Lin CC, Huang CH, Chang YC, Hsu SK, Chang WH. young adult patients with and without schizophrenia and in Risperidone response and 5-HT6 receptor gene variance: genetic relationship to antipsychotic medication. J Psychiatr Res 2001; 35: association analysis with adjustment for nongenetic confounders. 339–345. Schizophr Res 2004; 67: 63–70. 39 Gordon D. Viewing and Editing Assembled Sequences Using 20 Ryu S, Cho EY, Park T, Oh S, Jang WS, Kim SK et al. 759 C/T Consed. Current Protocols in Bioinformatics, In: Baxevanis AD, polymorphism of 5-HT2C receptor gene and early phase weight Davison DB (eds). Current Protocols in Bioinformatics. John Wiley: gain associated with antipsychotic drug treatment. Prog Neurop- New York, 2004, pp 11.12.11–11.12.43. sychopharmacol Biol Psychiatry 2007; 31: 673–677. 40 Nyholt DR. A simple correction for multiple testing for SNPs in 21 Theisen FM, Hinney A, Bromel T, Heinzel-Gutenbrunner M, linkage disequilibrium with each other. Am J Hum Genet 2004; 74: Martin M, Krieg JC et al. Lack of association between the -759C/T 765–769. polymorphism of the 5-HT2C receptor gene and clozapine- 41 Dina C, Meyre D, Samson C, Tichet J, Marre M, Jouret B et al. induced weight gain among German schizophrenic individuals. Comment on ‘A common genetic variant is associated with adult Psychiatry Genet 2004; 14: 139–142. and childhood obesity’. Science 2007; 315: 187. 22 Ferno J, Raeder MB, Vik-Mo AO, Skrede S, Glambek M, Tronstad KJ 42 Loos RJ, Barroso I, O’rahilly S, Wareham NJ. Comment on ‘A et al. Antipsychotic drugs activate SREBP-regulated expression of common genetic variant is associated with adult and childhood lipid biosynthetic genes in cultured human glioma cells: a novel obesity’. Science 2007; 315: 187. mechanism of action? Pharmacogenomics 2005; 5: 298–304. 43 Rosskopf D, Bornhorst A, Rimmbach C, Schwahn C, Kayser A, 23 Raeder MB, Ferno J, Glambek M, Stansberg C, Steen VM. Kruger A et al. Comment on ‘A common genetic variant is Antidepressant drugs activate SREBP and up-regulate cholesterol associated with adult and childhood obesity’. Science 2007; 315: and fatty acid biosynthesis in human glial cells. Neuroscience Let 187. 2006; 395: 185–190. 44 Hall DH, Rahman T, Avery PJ, Keavney B. INSIG-2 promoter 24 Ferno J, Skrede S, Vik-Mo AO, Ha˚vik B, Steen VM. Drug-induced polymorphism and obesity related phenotypes: association study activation of SREBP-controlled lipogenic gene expression in in 1428 members of 248 families. BMC Med Genet 2007; 30: 83.

Molecular Psychiatry INSIG2 and antipsychotic-related weight gain S Le Hellard et al 317 45 Kumar J, Sunkishala RR, Karthikeyan G, Sengupta S. The common 49 Lieberman JA, Stroup TS, McEvoy JP, Swartz MS, Rosenheck RA, genetic variant upstream of INSIG2 gene is not associated with Perkins DO et al. Effectiveness of antipsychotic drugs in patients obesity in Indian population. Clin Genet 2007; 71: 415–418. with chronic schizophrenia. N Engl J Med 2005; 353: 1209–1223. 46 Smith AJ, Cooper JA, Li LK, Humphries SE. INSIG2 gene 50 Cervino AC, Li G, Edwards S, Zhu J, Laurie C, Tokiwa G et al. polymorphism is not associated with obesity in Caucasian, Afro- Integrating QTL and high-density SNP analyses in mice to identify Caribbean and Indian subjects. Int J Obes 2007; 31: 1753–1755. Insig2 as a susceptibility gene for plasma cholesterol levels. 47 Feng Y, Dong H, Xiang Q, Hong X, Wilker E, Zhang Y et al. Lack of Genomics 2005; 86: 505–517. association between rs7566605 and obesity in a Chinese popula- 51 Yabe D, Komuro R, Liang G, Goldstein JL, Brown MS. Liver-specific tion. Hum Genet 2007; 120: 743–745. mRNA for Insig-2 down-regulated by insulin: implications for fatty 48 Skelly T, Pinheiro AP, Lange LA, Sullivan PF. Is rs7566605, a SNP acid synthesis. Proc Natl Acad Sci 2003; 100: 3155–3160. near INSIG2, associated with body mass in a randomized clinical 52 Bai YM, Lin CC, Chen JY, Lin CY, Su TP, Chou P. Association of trial of antipsychotics in schizophrenia? Mol Psychiatry 2007; 12: initial antipsychotic response to clozapine and long-term weight 321–322. gain. Am J Psychiatry 2006; 163: 1276–1279.

Molecular Psychiatry