Molecular Psychiatry (2007) 12, 474–482 & 2007 Nature Publishing Group All rights reserved 1359-4184/07 $30.00 www.nature.com/mp ORIGINAL ARTICLE Physiogenomic comparison of weight profiles of olanzapine- and risperidone-treated patients G Ruan˜o1, JW Goethe2, C Caley2, S Woolley2, TR Holford3, M Kocherla1, A Windemuth1 and J de Leon4 1Genomas, Inc., Hartford, CT, USA; 2Institute of Living, Hartford Hospital, Hartford, CT, USA; 3Department of Epidemiology and Public Health, Yale University School of Medicine, New Haven, CT, USA and 4Department of Psychiatry, College of Medicine, Eastern State Hospital, Mental Health Research Center, University of Kentucky, Lexington, KY, USA

Atypical antipsychotics induce pre-diabetic symptoms in some but not all patients, characterized most notably by elevated weight. The side effect profiles of the various drugs in the class differ, however, raising the possibility of drug-specific mechanisms for similar side effects. We used physiogenomic analysis, an approach previously employed to study the genetics of drug and diet response, to discover and compare genetic associations with weight profiles observed in patients treated with olanzapine and risperidone as an approach to unraveling contrasting mechanistic features of both drugs. A total of 29 single nucleotide polymorphisms (SNPs) were selected from 13 candidate relevant to two potential pharmacological axes of psychotropic-related weight profiles, appetite peptides and peripheral homeostasis. We applied physiogenomic analysis to a cross-section of 67 and 101 patients being treated with olanzapine and risperidone, respectively, and assessed genetic associations with the weight profiles. Weight profiles in patients treated with olanzapine were significantly associated with SNPs in the genes for E, apolipoprotein A4 and scavenger receptor class B, member 1. Weight profiles in patients treated with risperidone were significantly associated with SNPs in the genes for leptin receptor, neuropeptide Y receptor Y5 and paraoxonase 1. These results are consistent with contrasting mechanisms for the weight profile of patients treated with these drugs. Genes associated with olanzapine weight profiles may be related to peripheral lipid homeostatic axes, whereas those associated with risperidone’s may be related to brain appetite peptide regulation. Future physiogenomic studies will include neurotransmitter receptor SNPs and validation in independent samples. Molecular Psychiatry (2007) 12, 474–482. doi:10.1038/sj.mp.4001944; published online 2 January 2007 Keywords: obesity; antipsychotic agents; pharmacogenetics; appetite; lipid metabolism

Introduction case–control studies have shown that olanzapine is an independent risk factor for diabetes mellitus in Atypical antipsychotics (AAPs) are associated with patients with schizophrenia.7,8 Olanzapine is a thie- weight gain, although excess weight does not develop nobenzodiazepine with high affinities for the D ,D, in all patients.1–3 Hypertriglyceridemia ( > 600 mg/dl) 1 2 D , 5HT , 5HT , 5HT ,H, alpha and M receptors. may develop in as little as 3 months after initial expo- 4 2A 2C 3 1 1 1 It is a substrate for the cytochrome P450 1A2 (CYP sure to these drugs.4 The AAPs are much more 1A2) and for the UDP-glucuronosyltransferase 1A4 frequently associated with diabetes than are the (UGT 1A4)9 and has no active metabolites.10,11 ‘typical’ antipsychotics. In a study of more than Risperidone, another drug in this class, has a 38 000 patients, those on AAPs were 9% more likely different side effect profile. Weight gain is less pro- to have diabetes than those treated with ‘typical’ nounced than with olanzapine,12 but it is much more antipsychotics.5 In September 2003, the Food and likely to increase prolactin. Risperidone is a benzi- Drug Administration required re-labeling of all AAPs soxazole with high affinities for the D , 5HT , 5HT , to inform physicians about these risks. 2 2A 2C a and a receptors. It is a substrate for the cytochrome Olanzapine and clozapine carry the highest risk for 1 2 P450 2D6 and 3A (CYP 2D6 and CYP 3A)13 and has these side effects.6 Several population-based nested an active metabolite, 9-hydroxyrisperidone, with a comparable receptor affinity profile.10,11 Correspondence: Dr J de Leon, Department of Psychiatry, Study of the AAP pharmacological feature, which College of Medicine, Eastern State Hospital, Mental Health is responsible for the development of pre-diabetic Research Center, University of Kentucky, 627 West Fourth Street, complications in vulnerable populations, has princi- Lexington, KY 40508, USA. pally centered around weight gain caused by these E-mail: [email protected] Received 13 July 2006; revised 2 October 2006; accepted 2 medications. Histamine 1 (H1) and serotonin 2C November 2006; published online 2 January 2007 (5HT2C) receptor antagonism were among the most Physiogenomics weight olanzapine risperidone G Ruan˜o et al 475 important pharmacologic mechanisms discussed.14 of variation are analyzed to discover statistical 26 Antagonism of H1 and 5HT2C receptors is believed to associations with physiological characteristics. stimulate appetite and prevent satiety, respectively.15 The phenotypes are measured in populations of Other mechanisms that may lead to weight gain are individuals either at baseline or after they have been drug-induced elevations in serum leptin, neuro- similarly exposed to environmental challenges or peptide Y (NPY) or ghrelin.16,17 Patients taking AAPs interventions, including drugs and diet. The physio- are also known to be at risk for elevations in serum genomic approach affords testing broad hypotheses triglycerides and glucose.4,18 The mechanism behind about mechanistic features of drug effects. By relating hypertriglyceridemia is not known. In one case the physiological response to natural gene variation, report, olanzapine-induced hyperlipidemia resolved we gain valuable mechanistic insight. We also gain a when olanzapine levels disappeared, whereas weight practical way of predicting the response when the persisted.19 Genetically tractable metabolic differ- gene variation is known. In this report, we have ences could account for the variability in drug- extended physiogenomics to a comparison of two induced weight gain.1,2,18,20–23 AAP medications, olanzapine and risperidone. The availability of comprehensive genomic data- bases and massively parallel SNP genotyping techno- logy cannot be adequately handled by traditional Materials and methods statistical methods; it requires the development of Subjects and study design new statistical methods. This report is the first use in Patients were recruited as part of ongoing pharmaco- psychiatry of one of these new methods, which we genetic studies at the Institute of Living (Hartford, have called physiogenomics.24–26 Physiogenomics is a CT, USA) and at three Kentucky state hospitals medical application of sensitivity analysis and sys- (Lexington, KY, USA). These hospitals treat patients tems engineering, which we had previously applied with severe and persisting mental illness. From this to the study of drug and diet responses.24,25 Sensiti- study, 67 patients taking olanzapine and 101 patients vity analysis is the study of the relationship bet- taking risperidone who were willing to provide ween input and output as determined by each system a blood sample were selected for genotyping. The component.27 Physiogenomics utilizes genes as the demographic characteristics of the patient population components of the system. Gene variability, measured are presented in Table 1. To be eligible, patients had to by single nucleotide polymorphisms (SNPs), is have a current prescription of olanzapine or risper- correlated with physiological characteristics in a idone. All patients had an examination during which population, the output. With the advent of nanotech- their weight was determined. nology-based gene arrays, parallel processing of gene variability is practically possible at the level of Candidate gene selection physiological systems. Although single-gene muta- We selected eight genes related to peripheral lipid tions are the basis of most inborn errors of meta- homeostasis and five related to central appetite bolism, partial penetrance of genes interlinked in regulation, as summarized in Table 2. From the networks is more relevant to phenotypes common in lipid homeostasis axis, we selected representative clinical neuroscience. In physiogenomics, markers low-density (LDL) and high-density

Table 1 Demographic characteristics of the olanzapine- and risperidone-treated patients in this study

Variable Value Olanzapine Risperidone

Subjects Mean weight (kg) Subjects Mean weight (kg)

All All 67 82.9 101 78.4 Gender Female 19 82.4 38 73.4 Gender Male 48 83.1 63 81.5 Age (years) < 20 1 65.0 3 97.7 Age (years) 20–30 20 87.1 20 77.8 Age (years) 30–40 22 83.2 26 78.0 Age (years) 40–50 20 79.3 41 77.8 Age (years) 50–60 3 95.7 10 76.5 Age (years) 60–70 0 1 87.6 Age (years) 70–80 1 44.4 0 Heritage African-American 14 87.4 10 80.5 Heritage Caucasian 50 82.0 91 78.2 Heritage Hispanic 1 110.9 0 Heritage Other 2 60.5 0 Site Connecticut 14 84.2 0 Site Kentucky 53 82.6 101 78.4

Molecular Psychiatry Physiogenomics weight olanzapine risperidone G Ruan˜o et al 476 Table 2 Genes and SNPs analyzed for associations with weight profiles for olanzapine- and risperidone-treated groups

Area Pathway Gene Symbol SNP Type

Lipid metabolism LDL APOE rs405509 B200 bp upstream rs439401 B1.5 kbp downstream rs446037 B1.5 kbp upstream rs7412 Exon 3, C176R rs429358 Exon 3, R130C APOB rs3791981 Intron 18 (including Ag(x) antigen) rs1801701 Exon 26, Q3638R rs676210 Exon 26, L2739P

HDL Apolipoprotein A-I APOA1 rs5070 Intron rs4225 Downstream Apolipoprotein A-II APOA2 rs5085 Intron 2 Apolipoprotein A-IV APOA4 rs5092 Exon 2, T29T rs675 T367S, exon 3 Paraoxonase 1 PON1 rs705381 B200 bp upstream rs662 Exon 6, R192Q rs854572 B900 bp upstream rs3917550 Intron 7 Scavenger receptor SCARB1 rs10846744 Intron 1 class B, member 1 rs4765623 Intron 1 Scavenger receptor SCARB2 rs3853188 Intron 2 class B, member 2 rs894251 Intron 1

Appetite peptides Hormones Neuropeptide Y NPY rs1468271 Intron 1 Ghrelin precursor GHRL rs26312 B1 kb upstream Galanin GAL rs694066 Intron 1

Receptors Neuropeptide Y NPY5R rs6837793 B9 kb upstream receptor Y5 rs11100494 Intron 3 Leptin receptor LEPR rs7602 Intron 1 (30 UTR on another gene) rs1171276 Intron 1 (untranslated) rs8179183 Exon 12, N656K

Abbreviation: SNP, small nucleotide polymorphisms.

lipoprotein (HDL) pathways. For LDL, apolipoprotein peptide hormone receptors involved in appetite E (APOE) is essential for the normal catabolism of regulation. triglyceride-rich lipoprotein constituents and apoli- poprotein B (APOB) is the main apolipoprotein Laboratory analysis of chylomicrons and LDL. From the HDL pathways, A blood sample was obtained for each patient and we selected the A-I, A-II and A-IV DNA extracted.31 Genotyping was performed using (APOA1, APOA2, APOA4), paraoxonase (PON1) the Illumina BeadArray platform and the GoldenGate and the scavenger receptor class B, members 1 and assay.32,33 Table 3 lists the assay information and 2 (SCARB1, SCARB2). APOA1, APOA2 and APOA4 observed allele frequencies for the SNPs used in are components of HDL in plasma. PON1 is this study. an HDL-associated enzyme.28 SCARB1 and SCARB2 mediate selective uptake of cholesteryl esters from Data analysis HDL particles.29 On the axis of appetite regulation, Statistical analysis utilized the R Statistics Language we focused on hypothalamic hormones and their and Environment.34–37 Covariates were analyzed receptors. The hormone genes we genotyped were using multiple linear regression, and selected using NPY, ghrelin precursor (GHRL) and galanin (GAL). the stepwise procedure. To test for association NPY is a key appetite regulator which interacts with SNPs, the residual of weight from the covariate with GHRL, leptin, GAL and other hormones to model was tested using linear regression on the SNP orchestrate feeding behavior and appetite.30 Leptin genotypes. SNP genotype was coded numerically receptor (LEPR) and NPY receptor Y5 (NPY5R) are according to carrier status for the minor allele: 0 for

Molecular Psychiatry Physiogenomics weight olanzapine risperidone G Ruan˜o et al 477 Table 3 Assay DNA sequences for the SNPs analyzed

SNP Gene Chr Maj Min Freq Assay sequence rs405509 APOE 19 A C 0.45 GAGGACACCTCGCCCAGTAAT[A/C]CAGACACCCTCCTCCATTCT rs439401 APOE 19 T C 0.34 GAGAACTGAGGGGGTGGGAGG[A/G]GAAGAGAGTGCCGGCGGCTC rs446037 APOE 19 A C 0.01 AGACACAGGTGACCCAACTCC[A/C]ATGGCTGGCCTAGGCCCCTC rs7412 APOE 19 T C 0.12 CGGCCTGGTACACTGCCAGGC[A/G]CTTCTGCAGGTCATCGGCAT rs429358 APOE 19 T C 0.15 GGTACTGCACCAGGCGGCCGC[A/G]CACGTCCTCCATGTCCGCGC rs3791981 APOB 2 A G 0.13 TTTTCCAAAGATGATCTCTCC[A/G]GAGCTATTGTTTCTTCATTC rs1801701 APOB 2 A G 0.09 TCAGATGGAAAAATGAAGTCC[A/G]GATTCATTCTGGGTCTTTCC rs676210 APOB 2 A G 0.20 ATGTGGGGAAGCTGGAATTCT[A/G]GTATGTGAAGGTCAGGAACT rs5070 APOA1 11 A G 0.33 GCCACGGGGATTTAGGGAGAA[A/G]GCCCCCCGATGGTTGGCTCC rs4225 APOA1 11 T G 0.44 CTTTTAAGCAACCTACAGGGG[A/C]AGCCCTGGAGATTGCAGGAC rs5085 APOA2 1 G C 0.17 CAGACTCTAGAGACTGAAATT[C/G]AAGGCCCAGTTCTTGCTGTT rs5092 APOA4 11 A G 0.19 CAGTGCTGACCAGGTGGCCAC[A/G]GTGATGTGGGACTACTTCAG rs675 APOA4 11 T A 0.15 GAGAAAGAGAGCCAGGACAAG[A/T]CTCTCTCCCTCCCTGAGCTG rs705381 PON1 7 T C 0.26 GGTGGGGGCTGACCGCAAGCC[A/G]CGCCTTCTGTGCACCTGGTC rs662 PON1 7 A G 0.29 ATTTTCTTGACCCCTACTTAC[A/G]ATCCTGGGAGATGTATTTGG rs854572 PON1 7 C G 0.49 GGTGCCTCTGTACAACCATGT[C/G]TCTCTTCTCTGCTGTCTGCT rs3917550 PON1 7 T C 0.10 AGCAACGTCTTGCTGTTTTTC[A/G]GAGGTAGAGGGCTGCTTTCT rs10846744 SCARB1 12 C G 0.19 TAGCTTATCAGGTTTATTGCT[C/G]TCCATCTGTATCACCTGCCT rs4765623 SCARB1 12 T C 0.34 GATTTTGCCCAGTGGCTCTCC[A/G]AGGTGGCTGTACTGATGGAC rs3853188 SCARB2 4 A C 0.08 TTCACATACTGGGGAGTTCAG[A/C]ATAGTAATGTTTTTGGAAAA rs894251 SCARB2 4 T C 0.14 CTCAGGAGGCCTTACTGTGCC[A/G]TGGTTCTTGCCCTTTGATTT rs1468271 NPY 7 A G 0.03 GACCCTGTAATTTTCAGAAAC[A/G]CACATAGGAGTGGGTGTCTG rs26312 GHRL 3 A G 0.12 GCTGTTGCTGCTCTGGCCTCT[A/G]TGAGCCCCGGGAGTCCGCAG rs694066 GAL 11 A G 0.13 TTCTAAGTCCTCTGCCATGCC[A/G]GGAAAGCCTGGGTGCACCCA rs6837793 NPY5R 4 A G 0.10 ATGAATTGTCACTCAGAAGAA[A/G]CTTAATAGGCATTAATACTA rs11100494 NPY5R 4 A C 0.08 CAGAAAGATGTCATCATCCAG[A/C]ATTGCGTCCACACAGTCAAC rs7602 LEPR 1 A G 0.25 CTTGGAGAGGCAGATAACGCT[A/G]AAGCAGGCCTCTCATGACCC rs1171276 LEPR 1 A G 0.24 AGTTTCATGTACATTAAATAT[A/G]AATTTCTTTTGGCTGGAAAT rs8179183 LEPR 1 C G 0.18 TAATGGAGATACTATGAAAAA[C/G]GAGAAAAATGTCACTTTACT

Abbreviations: APOA, apolipoprotein A; APOB, apolipoprotein B; APOE, apolipoprotein E; Chr, location of the gene; Freq, frequency (0.00–1.00) of the minor allele in the study population; GAL, galanin; GHRL, ghrelin precursor; LEPR, leptin receptor; Maj, sequence of the most common allele, major; Min, sequence of the least common allele, minor; NPY, neuropeptide Y; NPY5R, neuropeptide Y receptor Y5; PON1, paraoxonase 1; SCARB, scavenger receptor class B.

non-carriers, 1 for single carriers and 2 for double the number of subjects, f the carrier proportion and D carriers. The F-statistic P-value for the SNP variable the difference in average weight between carriers and from the analysis of variance (ANOVA) was used non-carriers expressed relative to the s.d.41 to evaluate the significance of association. To test the validity of the ANOVA P-values, we performed an independent calculation of the P-values using permu- LOESS representation tation testing. The ranking of the first three SNPs was We used a locally smoothed function of the SNP identical under permutation and ANOVA analyses frequency as it varies with weight to visually repre- (data not shown). To account for the multiple testing sent the nature of an association. LOESS (LOcally of 29 SNPs, we calculated adjusted P-values using wEighted Scatterplot Smooth) is a method to smooth Benjamini and Hochberg’s false discovery rate (FDR) data using a locally weighted linear regression.42,43 At procedure.38–40 In addition, we evaluated the power each point in the LOESS curve, a quadratic poly- for detecting an association based on the Bonferroni nomial is fitted to the data in the vicinity of that multiple comparison adjustment. We calculated for point. The data are weighted such that they contribute each SNP the effect size relative to the s.d. necessary less if they are further away, according to the tricubic for detection of an association at a power of 80% function ! (20% false-negative-rate) using the formula 3 3 x À xi za= þ zb wi ¼ 1 À ; D ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffic ; dðxÞ Nf ð1 À f Þ where a is the desired false-positive rate (a = 0.05), where x is the abscissa of the point to be estimated, b the false negative rate (b =1ÀPower = 0.2), c the the xi are the data points in the vicinity, and d(x)is number of SNPs, z a standard normal deviate, N the maximum distance of x to the xi.

Molecular Psychiatry Physiogenomics weight olanzapine risperidone G Ruan˜o et al 478 Results NPY5R) and for the olanzapine-specific associations (APOE, APOA4, SCARB1) these values are, respec- Figure 1 depicts the weight profiles in the olanzapine tively, < 1 and < 7%. We attribute these differences to and risperidone study populations. These distri- the larger sample size of the risperidone-treated butions are approximately normal. The mean weight group. All remaining genes showed no significant was 83 kg for olanzapine vs 78 kg for risperidone. This association in either treatment group, and no gene difference was not statistically significant. We tested showed significant associations in both. We also list the potential covariates of age, gender, race and site in Table 4 the effect size D relative to s.d. needed to (Kentucky or Connecticut) for association with weight reach significance under the Bonferroni adjustment. using multiple linear regression. Only gender was Figures 2 and 3 show detailed representations of found to be significantly associated (P = 0.033) in the the physiogenomic analysis for olanzapine and risperidone-treated group where 4.5% of the variance risperidone, respectively. Each plot contains three is explained by gender, with men being 8 kg heavier components: the distribution of the phenotype (thin on average. No other significant covariate was found line), the genotype of each individual patient (circles) in either treatment group. and the LOESS fit of the allele frequency as a function Table 4 lists the results of the association tests, com- of phenotype (thick line). In each of three graphs, the paring olanzapine and risperidone treated groups. We abscissa represents each patient’s weight (kg), which found that SNPs in APOE, APOA4 and SCARB1 were is the phenotype. The ordinate indicates the allele significantly associated with the weight profile in the frequency for the LOESS curve. For clarity’s sake, axis olanzapine-treated group but not in the risperidone labels for the genotypes and distribution curve are not counterpart. Conversely, we found that SNPs in LEPR, shown. The genotypes are drawn on three levels: PON1 and NPY5R were significantly associated with non-carriers of the minor allele at the bottom, single- the weight profile in the risperidone group but not in carriers in the middle and double-carriers at the top. the olanzapine counterpart. The FDR-corrected P- The scale for the distribution curve is arbitrary. values yield an estimate of the false-positive rate. For The first panel in Figure 2 shows the LOESS curve the risperidone-specific associations (LEPR, PON1, for SNP rs7412 of the APOE gene. The frequency of

Figure 1 Distribution of body weight for olanzapine- and risperidone-treated groups. The vertical axis indicates the number (no.) of patients observed within a given 10 kg interval up to 150 kg on the horizontal axis.

Table 4 Significance levels of gene SNPs associated with weight profiles for olanzapine- and risperidone-treated groups

Area Marker P-value Coefficient Freq FDR D

SNP Gene Olanz. Risp. Olanz. Risp. Olanz. Risp.

Lipids rs7412 APOE 0.0056 0.3036 10.93 5.44 0.12 0.065 1.42 1.15 rs5092 APOA4 0.0116 0.0775 À10.64 À6.13 0.19 0.065 1.17 0.96 rs705381 PON1 0.1324 0.0012 5.20 À9.19 0.26 0.008 1.05 0.85 rs4765623 SCARB1 0.015 0.2954 7.12 2.93 0.34 0.065 0.97 0.79 rs3853188 SCARB2 0.6315 0.0224 2.39 9.67 0.08 0.073 1.70 1.38

Appetite peptides rs6837793 NPY5R 0.7259 0.0024 1.90 12.83 0.10 0.010 1.53 1.25 rs8179183 LEPR 0.9701 0.001 À0.15 À11.14 0.18 0.008 1.20 0.98

Abbreviations: APOA, apolipoprotein A; APOE, apolipoprotein E; FDR, false discovery rate (P-value); Freq, frequency (0.00–1.00); LEPR, leptin receptor; NPY5R, neuropeptide Y receptor Y5; Olanz., olanzapine; PON1, paraoxonase 1; Risp., risperidone; SCARB, scavenger receptor class B; D effect size relative to standard deviation. Bold type denotes significance at a = 0.05.

Molecular Psychiatry Physiogenomics weight olanzapine risperidone G Ruan˜o et al 479

Figure 2 Physiogenomic representation of the most significant genetic associations found in the olanzapine-treated group. Each plot contains three components: the distribution of the phenotype (thin line), the genotype of each individual patient (circles) and the LOESS fit of the allele frequency as a function of phenotype (thick line). In each of three graphs, the abscissa represents each patient’s weight (kg) which is the phenotype. The ordinate indicates the allele frequency for the LOESS curve. For clarity’s sake, axis labels for the genotypes and the distribution curve are not shown. The genotypes are drawn on three levels: non-carriers of the minor allele at the bottom, single-carriers in the middle and double-carriers at the top. The scale for the distribution curve is arbitrary.

Figure 3 Physiogenomic representation of the most significant genetic associations found in the risperidone-treated group. Each plot contains three components: the distribution of the phenotype (thin line), the genotype of each individual patient (circles) and the LOESS fit of the allele frequency as a function of phenotype (thick line). In each of three graphs, the abscissa represents each patient’s weight (kg), which is the phenotype. The ordinate indicates the allele frequency for the LOESS curve. For clarity’s sake, axis labels for the genotypes and the distribution curve are not shown. The genotypes are drawn on three levels: non-carriers of the minor allele at the bottom, single-carriers in the middle and double-carriers at the top. The scale for the distribution curve is arbitrary. the minor allele is below 20% in subjects with a ApoA4 that risperidone did not. Thus, physio- weight of 40–60 kg, whereas it approaches 40% in genomic analysis affords the derivation of a ‘genetic subjects above 100 kg in weight. This finding indi- contour’ for a drug, a series of mechanistic genetic cates a strong association between the APOE marker links which can differentiate even drugs in the same and weight. As the frequency of the minor allele is class. higher in the high spectrum of the weight distribu- The strongest association (P < 0.001) was found in tion, APOE SNP rs7412 is considered a risk marker the risperidone-treated group with SNP rs8179183 in for olanzapine. In contrast, the first panel in Figure 3 the leptin receptor, which causes an amino-acid shows the LOESS curve for SNP rs8179183 of change from lysine to asparagine (Lys 656 Asn). the LEPR gene. The frequency of the minor allele is Leptin has an obvious role in regulating body 20% in subjects with weight of 40–60 kg, whereas it weight.44 This SNP in the leptin receptor was approaches 0% above 100 kg. This finding points to previously found to be associated with body weight another strong association, in this case between LEPR in a sample of 1 873 subjects from 405 Caucasian and weight. However, as the frequency of the minor nuclear families.45 The leptin receptor also encom- allele is higher in the low spectrum of the weight passes other SNPs that have been associated with distribution, LEPR SNP rs8179183 is considered a weight and related phenotypes, in particular Gln 223 protective marker for risperidone. Arg.46 There is significant linkage between these SNPs, and any or none of them could be the actual functional variation. Discussion Another equally strong association (P < 0.001) in the This study shows that genetic associations with risperidone group was with SNP rs705381 upstream weight profiles established for genes in pathways of the gene PON1. PON1 is an arylesterase enzyme encompassing appetite peptides and peripheral lipid expressed in the and found in the bloodstream homeostasis differentiate olanzapine and risperidone. associated with APOA1 and HDL-c. The enzyme pro- Risperidone weight profiles had central appetite rela- tects against organophosphate poisoning and pre- tionships through leptin and NPY Y not found with vents the accumulation of oxidized in LDL olanzapine. Conversely, olanzapine weight profiles in vitro. Oxidized lipids are involved in atherosclero- had apolipoprotein associations through ApoE and sis, and genetic variation in PON1 has been associated

Molecular Psychiatry Physiogenomics weight olanzapine risperidone G Ruan˜o et al 480 with .47 SNP rs705381 is map- drugs clinically rather than genetically. We are helped ped upstream of the gene in the promoter area (À161 by the unique properties of SNP markers. First, SNPs C/T). This SNP has previously been associated with are inherited and stable DNA markers, which Alzheimer’s disease.48 eliminates the ambiguity about cause and effect so A third association in the risperidone group was common in epidemiological studies. Second, treat- with SNP rs6837793 in the NPY5 receptor. NPY plays ment decisions are independent of SNP markers an important role in the integration of appetite (unknown by treating physicians), diminishing the and energy expenditure through the NPY Y1 and Y5 possibility of treatment bias affecting the physio- receptor subtypes. Moreover, the NPY Y1 receptor genomic analysis. is highly expressed on human adipocytes, where it This study is focused on a limited set of genes inhibits lipolysis. SNP rs6837793 is located in the related to lipid metabolism and appetite peptides. promoter region between the Y1 and Y5 receptor The more recent genetic studies of antipsychotics genes, which are believed to be co-regulated by a and weight gain have focused on neurotransmitter 55,56 single promoter. An other SNP in the same promoter receptors including serotonin receptor 5HT2C, a region has been linked to serum triglyceride and adrenergic receptors,57 combinations of serotonin, HDL-c levels.49 dopamine and a adrenergic receptors58 but a few of In the olanzapine group, the strongest association them have also explored peptides such as the brain- (P < 0.006) was found with SNP rs7412 in apolipo- derived neurotrophic factor58 and leptin.55 A genome- protein E, which causes an amino-acid change (Arg wide scan study in patients taking antipsychotics 176 Cys) and is known as the e2 allele. The e2 allele is pointed toward another appetite peptide, the peptide known to be associated with decreased LDL-c and pro-melanin-concentrating hormone.59 At this stage of increased triglyceride levels,50 and is protective our physiogenomic study, we have not included any for Alzheimer’s disease.51 The e4 allele has also of the neurotransmitter receptor gene variations been linked to weight loss in Alzheimer’s disease because they will be the subject of subsequent patients.52 research looking at entire families of serotonin, To investigate whether the observed associations dopamine, adrenergic and histamine receptors. Some might be due to linkage disequilibrium with neigh- prior studies56,57 have associated neurotransmitter boring genes, we performed a chromosome locus receptor gene variation with olanzapine weight gain. analysis using the HapMap data.53 Using the r2- crite- These gene variations may influence appetite during rion for linkage, we observed linkage blocks, genomic olanzapine treatment. Unfortunately, our study only regions within which most SNPs are physically tested peptide-related genes, some for the peptides linked. The following genes were linked with other themselves and others for their receptors. The possi- genes in the same linkage block: APOE with TOMM40 bility that neurotransmitter receptor gene variations (mitochondrial outer membrane protein TOM40) may influence weight in our olanzapine patients and APOC1 (apolipoprotein C-I precursor); APOA4 cannot be excluded, but will have to wait for our with ZNF259 (zinc-finger protein 259), APOA5 future studies. (apolipoprotein A-V), APOC3 (apolipoprotein C-III The associations reported here are suggestive, and precursor) and APOA1 (apolipoprotein A-I prepro- three of them (PON1, NPY5R and LEPR) are statisti- tein); and PON1 with PON3. SCARB1 and the cally significant after correction for multiple compar- NPY1R/NPY5R promoter do not show significant isons. As most subjects were Caucasians (only 27% linkage with other genes. Our HapMap analysis out of 168 individuals or 16% of the study were non- utilized the Utah reference population with Western Caucasians), we lack power to detect racial asso- European ancestry, which is the best match for the ciations or racial confounding effects. It is important predominantly Caucasian patient populations in this that these results are validated in different popula- study. tions before they can be accepted as fact. We are As the patients were not randomized to the treat- pursuing such studies and continue recruiting addi- ment, it is possible that the results have been tional subjects. influenced by prior multiyear treatment with anti- Ascertaining genetic risk differences for patients psychotics and other medications, other potential on AAPs would have significant clinical utility. The confounders and, particularly, treatment decisions. specter of obesity may reduce compliance with As it is well known that olanzapine can exacerbate antipsychotic regimens and lead to low self-esteem obesity,54 prescription of this drug may be biased and social withdrawal in already marginalized towards lower weight patients. The lack of a signifi- patients. In a recent study, it was shown that 75% cant difference in weight between olanzapine- and of patients on AAPs discontinue therapy within 18 risperidone-treated patients is most likely owing to months of treatment initiation.60 The observed varia- this confounding effect, which tends to oppose the bility among individuals highlights the need for actual drug effect. Our focus on contrasting factors personalized recommendations of drug choice to between two drug treatment groups may minimize minimize potential side effects, particularly for AAPs baseline effects that are shared between treatment that are widely prescribed for many psychiatric dis- groups. Confounding factors would be a much more orders61 and in children.62 Obesity and pre-diabetic serious problem if we were trying to differentiate the syndromes introduce serious vascular complications,

Molecular Psychiatry Physiogenomics weight olanzapine risperidone G Ruan˜o et al 481 underscoring the urgency of avoiding the progression 3 Ruan˜ oG,Zo¨llner S, Goethe JW. Drug-induced metabolic syndrome to diabetes in the first place. (DIMS) in psychiatry: a diagnostic need uniquely suited to Personalized medicine and DNA-guided drug regi- pharmacogenomics. In: Wong SHY, Linder MW, Valdes Jr R (eds). Pharmacogenomics and Proteomics: Enabling the Practice mens allow physicians to tailor therapies based on of Personalized Medicine. AACC Press: Washington, DC, 2006, individual patient characteristics, rather than the pp 277–282. average properties of drugs in clinical trials.63–66 4 Meyer JM. Novel antipsychotics and severe hyperlipidemia. J Clin A drug with a higher average risk, for example, may Psychopharmacol 2001; 21: 369–374. be safe for those patients who inherited an ensemble 5 Sernyak MJ, Leslie DL, Alarcon RD, Losonczy MF, Rosenheck R. Association of diabetes mellitus with use of atypical neuroleptics of DNA markers with the most protective factors and in the treatment of schizophrenia. Am J Psychiatry 2002; 159: the least risk factors. The ability to match a drug’s 561–566. genetic ‘contour’ with the DNA profile of an indivi- 6 Koro CE, Fedder DO, L’Italien GJ, Weiss SS, Magder LS, dual patient marks the beginning of high-resolution Kreyenbuhl J et al. Assessment of independent effect of olanzapine personalized pharmacotherapy for prevention of and risperidone on risk of diabetes among patients with schizo- phrenia: population-based nested case-control study. BMJ 2002; adverse drug reactions. 325:243. A recent review has stressed that off-patent typical 7 Gianfrancesco FD, Grogg AL, Mahmoud RA, Wang RH, Nasrallah antipsychotics have been largely abandoned in deve- HA. Differential effects of risperidone, olanzapine, clozapine, and loped countries owing to their tardive dyskinesia conventional antipsychotics on type 2 diabetes: findings from a large health plan database. J Clin Psychiatry 2002; 63: 920–930. risk. However, typical antipsychotics are frequently 8 Kornegay CJ, Vasilakis-Scaramozza C, Jick H. Incident diabetes used in developing countries and pharmacogenetics associated with antipsychotic use in the United Kingdom general may help to reduce their risk of developing tardive practice research database. J Clin Psychiatry 2002; 63: 758–762. dyskinesia.67 In 2006, risperidone became generic in 9 Linnet K. Glucuronidation by cDNA-expressed human UDP- some European countries and will soon become glucuronyltransferases and human liver microsomes. Hum Psychopharmacol 2002; 17: 233–238. generic in the US. Olanzapine will become off-patent 10 Markowitz JS, Brown CS, Moore TR. Atypical antipsychotics Part only a few years later. It is likely that the marketing of I: pharmacology, pharmacokinetics and efficacy. Ann Pharmaco- generic risperidone and olanzapine will be associated ther 1999; 33: 73–85. with substantial reductions in price, which may make 11 Ereshefsky L. Pharmacokinetics and drug interactions: update for new antipsychotics. J Clin Psychiatry 1996; 57(Suppl 11): 12–25. pharmacogenetic testing for associated weight gain 12 Ganguli R, Brar JS, Ayrton Z. Weight gain over 4 months in cost-effective. schizophrenia patients: a comparison of olanzapine and risper- idone. Schizophr Res 2001; 49: 261–267. 13 Bork J, Rogers T, Wedlund P, de Leon J. A pilot study of Acknowledgments risperidone metabolism: the role of cytochrome P450 2D6 and 3A. J Clin Psychiatry 1999; 60: 469–476. This research was supported in part by NIH 14 Leadbetter R, Shutty M, Pavalonis D, Vieweg V, Higgins P, Small Business Innovation Research Grant 1 R43 Downs M. Clozapine-induced weight gain: prevalence and clinical relevance. Am J Psychiatry 1992; 149: 68–72. MH073291-01 ‘Gene Markers: Antipsychotic-Induced 15 Lamberti JS, Bellnier T, Schwarzkopf SB. Weight gain among Metabolic Syndrome’ to Genomas, Inc. Principal schizophrenic patients treated with clozapine. Am J Psychiatry investigator is G Ruan˜ o, MD, with a subcontract to J 1992; 149: 689–690. Goethe, MD. The work at the Mental Health Research 16 Fitzgerald PB, Scaffidi A, Morris MJ, de Castella AR, Kulkarni J. Center at Eastern State Hospital was supported by The relationship of changes in leptin, neuropeptide Y and reproductive hormone to antipsychotic-induced weight gain. internal research funding from J de Leon, MD. In the Hum Psychopharmacol Clin Exp 2003; 18: 551–557. past 2 years, J de Leon, MD. has been on the advisory 17 Palik E, Birkas KD, Faludi G, Karadi I, Cseh K. Correlation of board of Bristol-Myers Squibb and Roche Molecular serum ghrelin levels with body mass index and carbohydrate Systems Inc.; he received investigator-initiated grants metabolism in patients treated with atypical antipsychotics. Diabetes Res Clin Pract 2005; 68(Suppl 1): S60–S64. from Roche Molecular Systems Inc. and the E Lilly 18 Atmaca M, Kuloglu M, Tezcan E, Ustundag B. Serum leptin and Research Foundation; he has lectured once supported triglyceride levels in patients on treatment with atypical anti- by E Lilly, once supported by Bristol–Myers Squibb, psychotics. J Clin Psychiatry 2003; 64: 598–604. twice supported by Janssen and five times by Roche 19 Markham-Abedi C, de Leon J. Hypertriglyceridemia associated Molecular Systems Inc. We are thankful to G Pearlson, with direct effects of olanzapine rather than with weight gain: a case report (letter). J Clin Psychiatry 2006; 67: 1473–1474. MD, Institute of Living, Hartford, CT, USA, for his 20 Green AI, Patel JK, Goisman RM, Allison DB, Blackburn G. Weight contribution to our neuropsychiatric research pro- gain from novel antipsychotic drugs: need for action. Gen Hosp grams. We are grateful to M Johnson, RN, MT Susce, Psychiatry 2000; 22: 224–235. RN, MLT, B Szarek, RN and E Murray-Carmichael, 21 Basile VS, Masellis M, McIntyre RS, Ozdemir V, Meltzer HY, BSc, who helped with recruitment of patients and Macciardi FM et al. Genetic dissection of atypical antipsychotic- induced weight gain: novel preliminary data on the pharmaco- establishment of the DNA repository. genetic puzzle. J Clin Psychiatry 2001; 62(Suppl 23): 45–66. 22 Sussman N. The implications of weight changes with antipsycho- tic treatment. J Clin Psychopharmacol 2003; 23(Suppl 1): S21–S26. References 23 Wirshing DA, Wirshing WC, Kysar L, Berisford MA, Goldstein D, Pashdag J et al. Novel antipsychotics: comparison of weight gain 1 Muller DJ, Muglia P, Fortune T, Kennedy JL. Pharmacogenetics liabilities. J Clin Psychiatry 1999; 60: 358–363. of antipsychotic-induced weight gain. Pharmacol Res 2004; 49: 24 Ruan˜ o G, Thompson PD, Windemuth A, Smith A, Kocherla M, 309–329. Holford TR et al. Physiogenomic analysis links serum creatine 2 Correll CU, Malhotra AK. Pharmacogenetics of antipsychotic- kinase activities during statin therapy to vascular smooth muscle induced weight gain. Psychopharmacology 2004; 174: 477–489. homeostasis. Pharmacogenomics 2005; 6: 865–872.

Molecular Psychiatry Physiogenomics weight olanzapine risperidone G Ruan˜o et al 482 25 Ruan˜ o G, Windemuth A, Kocherla M, Holford T, Fernandez ML, 48 Erlich PM, Lunetta KL, Cupples LA, Huyck M, Green RC, Baldwin Forsythe CE et al. Physiogenomic analysis of weight loss induced CT, et al., MIRAGE Study Group. Polymorphisms in the PON gene by dietary carbohydrate restriction. Nutr Met 2006; 3: 20–30. cluster are associated with Alzheimer’s disease. Hum Mol Genet 26 Ruan˜ o G, Windemuth A, Holford TR. Physiogenomics: Integrating 2006; 15: 77–85. systems engineering and nanotechnology for personalized medi- 49 Blumenthal JB, Andersen RE, Mitchell BD, Seibert MJ, Yang H, cine. In: Bronzino JD (ed). The Biomedical Engineering Handbook, Herzog H et al. Novel neuropeptide Y1 and Y5 receptor gene 3rd edn. CRC Press: Boca Raton, FL, 2005, pp 281–289. variants: associations with serum triglyceride and high-density 27 Saltelli A, Chan K, Scott EM. Sensitivity Analysis. John Wiley and lipoprotein levels. Clin Genet 2002; 62: 196–202. Sons: Chichester, 2000. 50 Dallongeville J, Lussier-Cacan S, Davignon J. Modulation of 28 Rozek LS, Hatsukami TS, Richter RJ, Ranchalis J, Nakayama K, plasma triglyceride levels by ApoE phenotype: a meta-analysis. McKinstry LA. The correlation of paraoxonase (PON1) activity J Lipid Res 1992; 33: 447–454. with lipid and lipoprotein levels differs with vascular disease 51 Rubinsztein DC, Easton DF. Apolipoprotein E genetic variation status. J Lipid Res 2005; 46: 1888–1895. and Alzheimer’s disease: a meta-analysis. Dement Geriatr Cogn 29 Eckhardt ER, Cai L, Sun B, Webb NR, van der Westhuyzen DR. Disord 1999; 10: 199–209. High density lipoprotein uptake by scavenger receptor SR-BII. 52 Vanhanen M, Kivipelto M, Koivisto K, Kuusisto J, Mykkanen L, J Biol Chem 2004; 279: 14372–14381. Helkala EL et al. APOE-Epsilon4 is associated with weight loss in 30 Kalra SP, Kalra PS. NPY and cohorts in regulating appetite, obesity women with AD: a population-based study. Neurology 2001; 56: and metabolic syndrome: beneficial effects of gene therapy 655–659. (review). Neuropeptides 2004; 38: 201–211. 53 The International HapMap Consortium. A haplotype map of the 31 de Leon J, Susce MT, Pan RM, Fairchild M, Koch W, Wedlund PJ. . Nature 2005; 437: 1299–1320. The CYP2D6 poor metabolizer phenotype may be associated 54 American Diabetes Association (ADA), American Association of with risperidone adverse drug reactions and discontinuation. Clinical Endocrinologists (AACE), American Psychiatric Asso- J Clin Psychiatry 2005; 66: 15–27. ciation (APA), North American Association for the Study of 32 Oliphant A, Barker DL, Stuelpnagel JR, Chee MS. BeadArray Obesity (NAASO). Consensus development conference on anti- technology: enabling an accurate, cost-effective approach to high- psychotic drugs and obesity and diabetes. Diabetes Care 2004; 27: throughput genotyping. Biotechniques 2002; 32: S56–S61. 596–601. 33 Fan JB, Oliphant A, Shen R, Kermani B, Garcia F, Gunderson KL 55 Templeman LA, Reynolds GP, Arranz B, San L. Polymorphisms of et al. Highly parallel SNP genotyping. Cold Spring Harb Symp the 5-HT2C receptor and leptin genes are associated with Quant Biol 2003; 68: 69–78. antipsychotic drug-induced weight gain in Caucasian subjects 34 R Development Core Team. A Language and Environment for with a first-episode psychosis. Pharmacogenet Genomics 2005; 15: Statistical Computing. R Foundation for Statistical Computing: 195–200. Vienna, Austria, 2004; http://www.R-project.org. 56 Ellingrod VL, Perry PJ, Ringold JC, Lund BC, Bever-Stille K, 35 Dalgaard P. Introductory Statistics with R. Springer: New York, Fleming F et al. Weight gain associated with the À759C/T 2002. polymorphism of the 5HT2C receptor and olanzapine. Am J Med 36 Maindonald J, Braun J. Data Analysis and Graphics Using R. Genet B Neuropsychiatr Genet 2005; 134: 76–78. Cambridge University Press: Cambridge, 2003. 57 Park YM, Chung YC, Lee SH, Lee KJ, Kim H, Byun YC et al. Weight 37 Faraway JJ. Linear Models with R. Chapman & Hall/CRC: Boca gain associated with the alpha2a-adrenergic receptor-1,291 C/G Raton, FL, 2004. polymorphism and olanzapine treatment. Am J Med Genet B 38 Reinere A, Yekutiele D, Benjamini Y. Identifying differentially Neuropsychiatr Genet 2006; 141: 394–397. expressed genes using false discovery rate controlling procedures. 58 Lane HY, Liu YC, Huang CL, Chang YC, Wu PL, Lu CT et al. Bioinformatics 2003; 19: 368–375. 39 Benjamini Y, Hochberg Y. Controlling the false discovery rate: Risperidone-related weight gain: genetic and nongenetic predic- a practical and powerful approach to multiple testing. J Roy Stat tors. J Clin Psychopharmacol 2006; 26: 128–134. Soc Series B 1995; 57: 289–300. 59 Chagnon YC, Merette C, Bouchard RH, Emond C, Roy MA, 40 Benjamini Y, Hochberg Y. On the adaptive control of the false Maziade M. A genome wide linkage study of obesity as secondary discovery rate in multiple testing with independent statistics. effect of antipsychotics in multigenerational families of eastern J Educ Behav Stat, 2000; 25: 60–83. Quebec affected by psychoses. Mol Psychiatry 2004; 9: 1067–1074. 41 Rosner B. Fundamentals of Biostatistics, 4th edn. Duxbury Press: 60 Lieberman JA, Stroup TS, McEvoy JP, Swartz MS, Rosenheck RA, Belmont, 1995. Perkins DO et al. Clinical antipsychotic trials of intervention 42 Cleveland WS. Robust locally weighted regression and smoothing effectiveness (CATIE) investigators. Effectiveness of antipsychotic scatterplots. J Am Stat Assoc 1979; 74: 829–836. drugs in patients with chronic schizophrenia. N Engl J Med 2005; 43 Cleveland WS, Devlin SJ. Locally weighted regression: an 353: 1209–1223. approach to regression analysis by local fitting. J Am Stat Assoc 61 Buckley PF. Broad therapeutic uses of atypical antipsychotic 1988; 83: 596–610. medications. Biol Psychiatry 2001; 50: 912–924. 44 Licinio J, Caglayan S, Ozata M, Yildiz BO, de Miranda PB, 62 Olfson M, Blanco C, Liu L, Moreno C, Laje G. Mutational trends O’Kirwan F et al. Phenotypic effects of leptin replacement on in the outpatient treatment of children and adolescents with morbid obesity, diabetes mellitus, hypogonadism, and behavior antipsychotic drugs. Arch Gen Psychiatry 2006; 63: 679–685. in leptin-deficient adults. Proc Natl Acad Sci USA 2004; 101: 63 Ruan˜ o G. Quo Vadis personalized medicine? Personalized Med 4531–4536. 2004; 1: 1–7. 45 Liu Y-J, Rocha-Sanchez SM, Liu PY, Long J-R, Lu Y, Elze L et al. 64 Malhotra AK, Murphy Jr GM, Kennedy JL. Pharmacogenetics of Tests of linkage and/or association of the LEPR gene polymorph- psychotropic drug response. Am J Psychiatry 2004; 161: 780–796. isms with obesity phenotypes in Caucasian nuclear families. 65 Arranz MJ, Collier D, Kerwin RW. Pharmacogenetics for the Physiol Genomics 2004; 17: 101–106. individualization of psychiatric treatment. Am J Pharmaco- 46 Quinton ND, Lee AJ, Ross RJ, Eastell R, Blakemore AI. A single genomics 2001; 1: 3–10. nucleotide polymorphism (SNP) in the leptin receptor is asso- 66 Ozdemir V, Shear NH, Kalow W. What will be the role of ciated with BMI, fat mass and leptin levels in postmenopausal pharmacogenetics in evaluating drug safety and minimising Caucasian women. Hum Genet 2001; 108: 233–236. adverse effects? Drug Saf 2001; 24: 75–85. 47 Sanghera DK, Aston CE, Saha N, Kamboh MI. DNA Polymor- 67 Ozdemir V, Aklillu E, Mee S, Bertilsson L, Albers LJ, Graham JE phisms in two paraoxonase genes (PON1 and PON2) are associated et al. Pharmacogenetics for off-patent antipsychotics: reframing with the risk of coronary heart disease. Am J Hum Genet 1998; 62: the risk for tardive dyskinesia and access to essential medicines. 36–44. Expert Opin Pharmacother 2006; 7: 119–133.

Molecular Psychiatry