TOX and ADIPOQ Gene Polymorphisms Are Associated with Antipsychotic-Induced Weight Gain in Han Chinese

To find the genetic markers related to the antipsychotic-induced weight gain (AIWG), we analyzed associations among candidate gene single-nucleotide polymorphisms (SNPs) and quantitative traits of weight changes and lipid profiles in a Chinese Han population. A total of 339 schizophrenic patients, including 86 first-episode patients (FEPs), meeting the entry criteria were collected. All patients received atypical antipsychotic drug monotherapy and hospitalization and were followed for 12 weeks. Forty-three SNPs in 23 candidate genes were calculated for quantitative genetic association with AIWG, performed by PLINK. The TOX gene SNP rs11777927 (P = 0.009) and the ADIPOQ gene SNP rs182052 (P = 0.019) were associated with AIWG (in body mass index, BMI). In addition, the BDNF SNP rs6265 (P = 0.002), BDAF SNP rs11030104 SNP (P = 0.001), and ADIPOQ SNPs rs822396 (P = 0.003) were significantly associated with the change of waist-to-hip ratio (WHR) induced by atypical antipsychotics. These results were still significant after age and gender adjustments. These findings provide preliminary evidence supporting the role of TOX, ADIPOQ and BDNF in weight and WHR gain induced by atypical antipsychotics.

carried the CDKN2A/B rs2811708 G allele, the BMI of the patients with the CDKN2A/B rs2811708 TT genotype increased significantly after the treatment of AAPD (P < 0.05) (Supplemental Figure 3).

Discussion
With the development of molecular genetics during the past decade, intensive research has examined the influence of genetic variations on AIWG. However, as yet no genetic tests for AIWG are endorsed for clinical application. Although there are significant findings for AIWG associations with many other genes, the most consistently replicated findings are with HTR2C, MC4R, and leptin genes 24,25 . A recent meta-analysis reported that 11 SNPs from 8 genes were associated with weight or BMI change, and 4 SNPs from 2 genes were significantly related to categorical weight or BMI increase 14 . Many previous studies have not been confirmed because of varying criteria, differences in frequencies across different populations, or poor statistical power. In addition, compared with the candidate-gene-based approach, GWAS of AIWG are relatively limited [19][20][21] . Genes PTPRD, MC4R, and PMCH were discovered having the strongest associations with AIWG.  AIWG is thought to be multifactorial and polygenic, which may be different from simple obesity per se. In recent years research has pointed to a broader array of genes and pathways hypothesized to underpin AIWG. Attention has turned to genes and pathways harboring variants that might make the risk of energy unbalance increased by the influence of antipsychotics, leading to AIWG 26 . Thus, seeking AIWG genes in Han Chinese would promote the development of predictive genetic tests and help us better understand and guide treatment.
In this study, we analyzed 43 SNPs in 23 candidate genes for quantitative association with AIWG for the first time. Here, we decided to use the candidate gene approach rather than performing a GWAS to select a small set of variants for a more focused study, providing sufficient power for these selected variants given the small sample size. As candidate genes, we chose (a) genes mainly from a previous GWAS for Caucasian obesity-AIWG subjects and (b) genes associated with simple obesity, type 2 diabetes, microvascular complications of diabetes, and AIWG.
In the present study, SNPs in TOX and ADIPOQ yielded the most significant associations for AIWG, and CDKN2A/B showed light to strong associations. We did not replicate associations with several well-known AIWG genes, including FTO, MTHFR, and COMT, in our Han Chinese population. Potential reasons for this include genetic heterogeneity, less covered genes (i.e., too few SNPs were genotyped for certain candidate genes), entry criteria, and the relatively small sample size. Although we were unable to test genome-wide association for all genes, the spectrum of AIWG-associated genes differed between Han Chinese in our study and European populations in the previous study [15][16][17]27 .
The TOX quantitative association with AIWG was first discovered in this study. TOX (thymocyte selection-associated HMG box) is a member of a novel gene family and encodes a novel nuclear DNA-binding protein belonging to a large superfamily of HMG (high-mobility-group) proteins 28,29 . TOX may play a role in regulating expression of genes involved in cell cycle progression, such as the cell division cycle gene and oncogenes 30 . In recent years, emerging studies have found that TOX is aberrantly expressed or mutated in various diseases, such as leukemia 29 or cardiovascular diseases 31 . A previous study by our group first found the TOX gene association with type 2 diabetes 32 . In the present study, we tested 3 SNPs of the TOX gene, rs1526167, rs2726557, and rs11777927 (Supplemental Figure 4). Unfortunately, the SNP rs1526167, which was located in a separate haplotype block, not in linkage disequilibrium with SNPs in the TOX gene coding region and introns, failed the HWE test.  Table 3. Quantitative gene-gene interaction analysis of AIWG (ΔBMI). a All the dependent variables were adjusted for age and gender. b No age and gender adjusted phenotypes were used in subgroups due to limited sample size. Our results showed that the TOX SNP rs11777927 was significantly associated with AIWG (P = 0.009) and that the A allele contributed to the increased risk for AIWG. And gene × gene interaction analyses showed significant epistasis between TOX and PKHD1 and between TOX and RPTOR for AIWG (Δ BMI) in all patients. We found some associations of TOX SNPs with obesity and metabolic-syndrome-related phenotypes. It is stated by Cox et al. in a published US patent application (US 2006/0177847 A1, August 10, 2006), that they found the TOX polymorphism and other 27 DNA sequence variations to be related to OLZ-treatment-emergent weight gain and "metabolic syndrome" in a 1.7 million SNP genome association study.
However, no rs11777927 association has been reported for AIWG. To the best of our knowledge, only one association study of rs11777927 and disease-intracranial aneurysm-has been reported 33 . The biological connections between TOX and AIWG are poorly understood, perhaps owing to the inflammatory responses mediated by immune cells, developmentally regulated by TOX gene, contributing to AIWG.
The ADIPOQ locus has been shown to be the only major gene for plasma adiponectin, which is exclusively expressed in adipose tissue 34 . The ADIPOQ SNP rs1501299 has been reported to be associated with the risk of obesity 35,36 and cardiovascular diseases [37][38][39] . Four candidate SNPs in ADIPOQ in this study were selected from four different haplotype blocks: rs182052, rs822396, rs7649121, and rs1501299 (Supplemental Figure 5). The results showed that the ADIPOQ SNP rs182052 (P = 0.019) was significantly associated with AIWG, and that the ADIPOQ SNPs rs822396 (P = 0.003) and rs1501299 (P = 0.040) were associated with changes in WHR.
Other studies have not delivered consistent findings for ADIPOQ and AIWG. An association of rs1501299 with significant weight gain (> 7% of the baseline weight) was reported in Chinese patients 40 . Another study reported genotypic or allelic association of 6 ADIPOQ variants with AIWG in European patients 41 . However, the latest two studies did not support a major role of ADIPOQ rs1501299 in the regulation of AIWG 42,43 , and the association of rs1501299 with AIWG was not present in Japanese patients 44 . The SNP rs1501299 was a hot spot that had attracted many researchers' interest. Interestingly, our results showed rs1501299 was associated with changes in WHR but not in AIWG, and that rs182052 and rs822396 were significantly associated with AIWG and changes in WHR induced by AAPD, which haven't previously been reported. ADIPOQ is most likely closely related to AIWG, which needs further research.
BDNF can encode the BDNF precursor protein located in the chromosome 11p13 region 45 . And BDNF crossing the blood-brain barrier is the most abundant neurotrophin that modulates synaptic transmission and neuroplasticity in the central nervous system 46,47 . Previous researches suggested that the BDNF rs6265 was associated with many eating disorders 48,49 . In addition, several SNPs in the BDNF and BDAF were strongly associated with obesity in GWAS studies 50,51 .
In the field of the genetics of AIWG, the recent two studies showed BDNF rs6265 was associated with the increased BMI in the psychiatric patients receiving AAPD 18,52 . Moreover, the BDNF haplotypes including SNP rs6265 were associated with AIWG 53,54 . However, a study reported that BDNF rs6265 was not associated with AIWG performed by Tsai et al. 55 . In the present study, we failed to find the association between BDNF rs6265 and AIWG, which was consisted with the results of Tsai's. Interestingly, we found the BDNF SNP rs6265 (P = 0.002) and BDAF SNP rs11030104 SNP (P = 0.001) were significantly associated with the change of WHR induced by AAPD. The WHR is a alternative measure which have been found to be superior to BMI to reflect abdominal obesity in the World Health Organization (WHO) guidelines 56 .
All the above findings, including ours, indicate that the SNPs of BDNF have a significant impact on the obesity schizophrenic patients induced by AAPD. The associated genetic markers from BDNF may have different effects, and accumulated mutations may provide a whole contribution to the obesity induced by AAPD. The SNP BDNF rs6265 may play an important role in this process. Hence, the following study about the function of the genetic variants will be necessary to elucidate the mechanism how the genetic variants in BDNF medicate signaling pathway and lead to obesity.
CDKN2A/B is located in the chromosome 9p21 region, has been highlighted as the strongest genetic susceptibility locus for cardiovascular disease 57,58 and type 2 diabetes 58,59 . CDKN2A/B encodes the CDK inhibitor proteins involved in cell cycle regulation, aging, senescence, and apoptosis. In the present study, we selected 3 SNPs of CDKN2A/B: rs3731245, rs2811708 and rs10811661 (Supplemental Figure 6); rs3731245 and rs2811708 are in a haplotype block. SNPs rs3731245 (P = 0.04) and rs2811708 SNP (P = 0.039) were associated with the AIWG but with low statistical power. We found no association between rs10811661 and AIWG, but this SNP was associated with the FPG changes that induced by OLZ (P = 0.009). These results suggest that glucose and lipid metabolism abnormalities and weight gain are influenced by AAPD through different pathways. Previous studies have found that AAPD may not affect glucose and lipid metabolism directly through weight gain 10,60 . The mechanism by which the CDKN2A/B gene affects susceptibility for AIWG remains to be investigated.
In the future, we will further substantiate our gene results and explore more genetic factors underlying AIWG. However, the heterogeneity of medications and psychiatric disorder factors still exist. We analyzed the genetic association in FEPs, OLZ-treated patients, and RIS-treated patients, but the subgroups were too small for strong statistical power. For this reason, a better designed trial with different AAPDs in larger samples and different populations should also be implemented to validate the genes associated with AIWG.
We also performed binary association studies for AIWG, the discrete analyses showed similar associations with quantitative BMI changes. Replication is essential for association studies, although GWASs for AIWG were relatively limited. We compared our associations with published GWAS for body weight related traits (GWASdb, v2. http://jjwanglab.org/gwasdb), ADIPOQ (rs182052), BDNF (rs6265), and BDAF (rs11030104) gene SNPs all yielded genome-wide associations.
In our study, we used quantitative associations between SNPs of candidate genes and AIWG and other antipsychotic-related phenotypes to find AIWG candidate genes in Han Chinese schizophrenia patients. We measured not only weight but also the WHR and lipid and glycemic profiles, which are a common index of metabolic syndrome. In addition, for the purpose of excluding environmental factors that may affected AIWG 62 , we selected a sample of inpatients with lifestyle, exercise, and diet under unified management by the hospital in order to reduce the heterogeneity. Our comprehensive study of TOX and ADIPOQ is an important contribution to understanding the biology of AIWG.

Patients and Materials
Study participants and design. A total of 339 patients hospitalized with schizophrenia or schizoaffective disorders were recruited. All the subjects were unrelated Han Chinese collected from the Tianjin Metal Health Centre, which is Asia's largest independent psychiatric hospital. The inclusion criteria for this study were as follows: (a) Clinical diagnoses were independently confirmed by two psychiatrists according to the Diagnostic and Statistical Manual of Mental Disorders, fourth edition, text revision. (b) All patients were first-episode drug-naive patients or without any antipsychotic drugs at least 4 weeks before enrollment. (c) The patients were physically healthy with normal hematological and biochemical parameters. (d) All patients were 18-60 years old. Subjects with neurological disorders, eating disorders, and thyroid diseases were excluded from the study. Clinical characteristics of patients were shown in Table 1. All subjects provided written informed consent prior for this study, and the protocol was approved by the Committee on Studies Involving Human Beings at Tianjin Medical University. All experiments were performed in accordance with relevant guidelines and regulations.
All patients received single AAPD intervention on the basis of clinical treatment need. Only trihexyphenidyl for extrapyramidal symptoms and lorazepam for insomnia or agitation were allowed as needed during the study period as concomitant medications.
Patients were followed up during the 12-week treatment course. All patients were measured for body weight and for waist, abdominal, and hip circumferences at baseline and at weeks 2, 4, 6, 8 and 12 after treatment initiation. Body mass index (BMI) and waist-to-hip ratio (WHR) were calculated. Before and after treatment for 4, 8, and 12 weeks, triglyceride, high-and low-density lipoprotein, total cholesterol, total protein, albumin, fasting plasma glucose (FPG), blood creatinine, urea nitrogen (urea), urea/creatinine ratio, and serum C reactive protein were measured for all patients.  Candidate gene and SNP selection. Candidate genes mainly came from a comparative GWAS for Caucasian obesity-AIWG subjects performed by our project team members and colleagues. Genes associated with simple obesity, type 2 diabetes, or microvascular complications of diabetes in our previous studies were also included 32,63 , in a GWAS of simple obesity, which was the largest sample size by far (340,000 persons) 51 , and in other previous candidate gene associations for AIWG that gave inconsistent or controversial results [15][16][17][18]27 . Forty-three (43) SNPs in 23 candidate genes were selected in our study (Table 4). For some genes, tagging SNPs were selected using the HapMap database (phase2 + phase3, release #28, CEU population, Build36; www.hapmap.org) and Tagger in Haploview 64 . Minor allele frequencies of Han Chinese were taken from dbSNP (http://www.ncbi.nlm.nih.gov/snp/). For previously reported associations, we selected SNPs with the most significant association rather than genotyping the whole gene. For less studied genes, multiple SNPs were chosen based on the LD pattern of the gene (r 2 > 0.8).
Genotyping and quality control. Genomic DNA samples were extracted from 5 ml of peripheral whole blood samples using the high-salt method. Samples were stored and processed by the Center for Molecular and Population Genetics at Tianjin Medical University. Genotyping was performed by primer extension of multiplex products with detection by matrix-assisted laser desorption time-of-flight mass spectrometry. For quality control, 10% of the sample was randomly re-genotyped, with a 100% concordant rate. All genotyping was done blind to knowledge of subjects' clinical data.
Statistical analysis. The Hardy-Weinberg equilibrium (HWE) test was carried out before the association analysis ( Table 4). All phenotypes were documented in a Filemaker Pro database. Statistical analyses for phenotypes were performed by SPSS software (version 20.0). We tested for recessive effects for three significant SNPs by comparing the three genotypes or different subgroups using ANOVA, with change in BMI across the 12-week trial as the dependent measure. The level of statistical significance for the above tests was set a priori at P < 0.05. The linear regression model in the PLINK 65 was used to test the association between the chosen SNPs and several phenotypes, where the changes of weight, BMI, WHR, lipid and glycemic profiles were used as phenotypes for the quantitative trait locus analyses. Linear regressions were performed for each quantitative trait against age within sexes, standardized residuals were saved to make mean = 0 and standard deviation = 1 for each phenotype. Outliers (more than 4 standard deviations) were deleted from this study. As 43 SNPs in 23 candidate genes were analyzed, we used Bonferroni corrections for multiple testing, and the nominal P-value should be at least 0.002 to be significant after multiple test corrections.
Pairwise gene-gene interaction analyses (epistasis) were carried out by PLINK 65 among candidate gene SNPs in all patients, the FEP group, the OLZ treatment group, and the RIS treatment group, Bonferroni corrections were also employed for multiple testing.
In addition, we also performed a discrete association study for AIWG by PLINK. A weight gain "case" was defined as a patient who gained 7% or more of his or her baseline body weight in a 12-week trial.

Conclusions
In Conclusion, our findings suggest the role of TOX, ADIPOQ and BDNF in weight and WHR gain induced by atypical antipsychotics in schizophrenia subjects. This study provides a better understanding of genetic factors predisposing individuals to AAPD-induced obesity, dyslipidemia, and abnormal glucose metabolism that may help guide clinical medication intervention and also reveal the pathogenesis of AIWG.