Association of CNR1 and INSIG2 polymorphisms with antipsychotics-induced weight gain: a prospective nested case–control study

Weight gain is a frequent and severe adverse reaction in patients taking antipsychotics. The objective was to further investigate in a natural setting influential risk factors associated with clinically significant weight gain. An observational follow-up study was conducted. Patients when initiating treatment with whatever antipsychotic were included; a structured questionnaire was applied at baseline, 3 and 6 months later; a blood sample was obtained. In a nested case–control approach, patients with an increase ≥ 7% of their initial weight were considered as cases, the remaining, as controls. The results showed that, out of 185 patients, 137 completed the 6-month follow-up (cases, 38; controls, 99). Weight gain gradually and significantly increased in cases (baseline, 65.0 kg; 6 months, 74.0 kg) but not in controls (65.6 kg and 65.8 kg, respectively). Age (adjusted OR = 0.97, 95% CI = 0.96–0.99, p = 0.004), olanzapine (adjusted OR = 2.98, 95% CI = 1.13–7.80, p = 0.027) and quetiapine (adjusted OR = 0.25, 95% = 0.07–0.92, p = 0.037) significantly associated with weight gain. An association was also found for the CNR1 (rs1049353) and INSIG2 (rs7566605) polymorphisms. In conclusion, an increased risk of antipsychotics-induced weight gain was observed for younger age and olanzapine, and a relative lower risk for quetiapine. A potential role of CNR1 rs1049353 and INSIG2 rs7566605 polymorphisms is suggested.

Purified DNAs were quantified by Picogreen. The amplicons were prepared using the Access Array System for Illumina Platform (Fluidigm). Briefly, this method consist of the following phases: (1) quantification by Picogreen and normalization of the DNA samples to 50 ng/ul, (2) combination of diluted DNA samples with primers (5′-AAT GAT ACG GCG ACC ACC GAG ATC TAC ACT GAC GAC ATG GTT CTACA-3′ and 5′-CAA GCA GAA GAC GGC ATA CGA GAT -[barcode]-TAC GGT AGC AGA GAC TTG GTCT-3′) of the Access Array Barcode Library for Illumina Sequencers (Fluidigm), (3) preparation of 20X specific target primers solutions, (4) loading of sample inlets and primer inlets of a primed Access Array Integrated Fluidic Circuit (IFC) and running of the PCR AA48 × 48 Standard v1 protocol and, finally, (5) harvest of PCR products from the IFC. For more details consult the protocol: "Access Array System for Illumina Platform User Guide_G1".
The finally obtained amplicons were validated and quantified by an Agilent 2100 Bioanalyzer using High Sensitivity chips, and an equimolecular pool of these amplicons was purified by agarose gel electrophoresis to eliminate primers/dimers, and titrated by quantitative PCR using the "Kapa-SYBR FAST qPCR kit forLightCy-cler480" and a reference standard for quantification. The pool of amplicons was denatured prior to be seeded on a flowcell at a density of 10 pM, where clusters were formed and sequenced using a "MiSeq Reagent Kit v3", in a 2 × 300 pair-end sequencing run on a MiSeq sequencer. Raw sequences were filtered according to quality under standard Illumina, parameters and positive sequences were mapped against human genome. Following mapping, DNA variant detection was performed using the MiSeqReporter software (Illumina).

Statistics analysis.
For the case-control approach, patients who, at 180 days, showed an increase ≥ 7% of their initial weight were considered as cases 18 ; the remaining were considered as controls. Values of qualitative variables are presented as absolute frequencies and percentages; quantitative variables are presented as means with Standard Deviation (SD) or median with interquartile ranges [Q1-Q3]. For comparison among qualitative variables, including Hardy-Weinberg equilibrium, the Pearson's chi-square test, or the Fisher's exact test if the expected frequency was lower than 5 in any cell, were used. Genotypes distribution in the two groups in comparison was assessed with the Monte Carlo method for multiple testing 19 ; the estimated p values in this form are in Table 3; for comparison among groups for quantitative variables, a Student t test or a Mann-Whitney U were calculated according to the Kolmogorov-Smirnov or Shapiro-Wilk test. To explore possible correlation between some quantitative variables without normal distribution, Spearman´s Rho was calculated.
Based on the literature, covariates initially considered for adjusting in the analysis were: age, sex, BMI at baseline and the different SNPs considered. Information on exposure to the explanatory variables was that consigned at baseline. Other possible confounders were selected among those with a p < 0.1 significance level in a previous univariate analysis; a forward stepwise multivariate logistic regression was performed with those variables. With the significant variables in this first regression, a second regression analysis was performed, minimizing in this way the possible effect of the missing values in variables which finally were non-significant. A Hosmer-Lemeshow goodness-of-fit test was performed. To learn the influence of each of the several factors on weight gain, adjusted odds ratios (adjusted OR) and their 95% confidence intervals (95% CI) were calculated.

Results.
Out of 185 patients of the ongoing cohort, 137 completed at least a follow-up period of 6 months and were included in the current analysis; reasons for dropouts are presented in a flow-chart ( Fig. 1). Of those patients who completed the specified period, 38 showed a clinically significant increase of their initial weight and were accordingly considered as cases; the remaining 99, who did not, were considered as controls.
The whole sample was of European origin. The distribution of the main sociodemographic and clinical variables in the two groups is presented in Table 1. Mean age in the whole sample was 55.2 years, it ranged from 15 to 100 years; age was higher in controls than in cases (59.7 years vs 43.4 years, p = 0.001). Though there were more female patients in the cohort (63.5%), no significant differences between cases and controls were found (p = 0.654). As an average, patients had 1.1 psychiatric diagnoses and were having 1.2 antipsychotics; risperidone, followed by quetiapine, olanzapine and aripiprazole were the antipsychotics most frequently used. More cases than controls received two or more antipsychotics during some time in the follow-up period (34.2% vs. 20.2%, p = 0.086); cases received higher doses than controls (median chlorpromazine equivalence, 237.74 mg/d vs. www.nature.com/scientificreports/ 117.24 mg/d, respectively; p = 0.002). More psychotic diagnoses were observed within the cases without statistically significant difference (39.5% vs. 24.2%, p = 0.08). Previous history of metabolic disease was associated with a minor risk of weight gain (p = 0.018); similarly, higher Charlson index scores were associated with less risk (p = 0.008; Table 1). No differences in overall medications were found between cases and controls (p = 0.072). Cases were taller than controls (mean values (SD), 1.64 (0.10) m vs. 1.59 (0.12) m, respectively; p = 0.01); body weight was similar in the two groups at baseline (65.05 (12.71) kg vs. 65.65 (14.49) kg, respectively; p = 0.824), Table 1. Distribution of the main characteristics in the two groups in comparison. *p < 0.05. a No data, 7 patients (1 case, 6 controls); Basic education: reading and writing, primary and secondary education. b Psychiatric diagnoses were made by medical specialists. Psychotic disorders cover schizophrenia, schizotypal, delusional, and other non-mood psychotic disorders (F20 to F29 codes). Non-psychotic disorders cover mental disorders due to known physiological conditions (F00 to F09 codes), mental and behavioral disorders due to psychoactive substance use (F10 to F19 codes), mood [affective] disorders (F30 to F39 codes) and other mental disorders (F40 to F99 codes), according to the International Classification of Diseases, 10th edition. c Amisulpride, asenapine, sulpiride, tiapride, ziprasidone. d Include both low and high comorbidity. e Medications influencing somehow weight gain were metformin (cases, 0; controls, 5), levothyroxine (cases, 3; controls, 6), topiramate (cases, 3; controls, 5), digoxin (cases, 1; controls, 2) and methylphenidate (cases, 0; controls, 2). Information for these drugs is provided in the Summary of Product Characteristics. f No tobacco abstinence cases were registered.  26.07 (4.27) kg/m 2 ; p = 0.082). No differences in physical activity existed between cases and controls, neither at baseline (p = 0.379) nor at the end of the follow-up period (p = 0.608) when measured by the activity factor (Table 1). In our multivariate model only age, along with exposure to olanzapine or quetiapine, showed a significant association with weight gain (Table 2). Sex was not significant in the model, not even for olanzapine or quetiapine exposure as an interaction term. Dose was not significant either; when taken independently a significant correlation was observed between dose in chlorpromazine equivalents and percentage of weight gain (Rho = 0.31; p = 0.0003). A lower risk of weight gain was observed with increasing patients' age and for those treated with quetiapine, this similarly occurs when choosing an age range of 18-65; risk of weight gain significantly increased for patients who were treated with olanzapine.
When examined by a 2 × 3 table Pearson's chi-square test (two degrees of freedom), a trend of significant differences between subjects who gained 7%, or higher, of their baseline body weight, and those who did not, emerged for CNR1 (rs1049353) and INSIG2 (rs7566605) polymorphisms (Tables 3 and 4

Discussion
Our observational study following a nested case-control approach showed that one in three patients on antipsychotics developed an increase of ≥ 7% of their initial weight after a period of 6 months (cases) when compared with those who did not (controls). This significant and severe weight gain associated with antipsychotics has been largely observed elsewhere: at 6 weeks in 45.8% of schizophrenic patients on olanzapine 20 and in 20.2% of bipolar patients with the same medication 21 ; at 3 months in 35% of schizophrenic first-episode medicationnaive patients 22 . In a large systematic review and meta-analysis, a significantly higher risk of gaining ≥ 7% of the baseline weight (RR = 2.04, 95% CI = 1.54-2.71, p < 0.001) was also found after 3 to 12 weeks of treatment, and the most severe weight-gain was caused by olanzapine 5 . Weight gain also has been found in longer follow-up studies    www.nature.com/scientificreports/ at 6, 12 and 36 months [23][24][25] . Our figures are in this range; follow-up duration, antipsychotics used, psychiatric disease, comorbidities and, particularly, age certainly account for these differences in percentages. Taking as a reference the sample studied by Verma et al. 24 , with a similar 6-month period of observation, it can be observed that differences in percentages of patients who gained 7% or more than the baseline weight (ours, 27.7%; Verma's, 65%) inversely correlated with average age differences in the two samples (ours, 55.1 years; Verma´s, 29.8 years). Although percentage of male are generally higher in schizophrenia samples using antipsychotics, female patients are majority in our sample; this is probably due to the more frequent non-psychotic disorders in women, including neurocognitive and affective disorders. The influence of sex in antipsychotic-induced weight gain is still controversial; moreover, sex did not appear to emerge as a critical predictive factor for beneficial effect 26 . We did not observe differences in weight gain between sexes, what is coincident with previous observations by Zipursky et al. 27 . Though most of the studies found a higher risk in females 1,24,28-30 and lower in males in the first months of treatment 23 , the figures by Zipursky et al. 27 remain chiefly reliable because they are based on a survival analysis at 2 years in a study specifically designed to find out predictors of weight gain in people with first-episode psychosis.
As far as age is concerned, younger age seems to be consistently associated with a higher risk of weight gain associated with antipsychotics 28,30,31 ; being children and adolescents even at a greater risk 32,33 . In fact, younger age, along with non-Caucasian ethnicity, low baseline BMI, along with other variables, are considered one of the 16 items for Weight Gain Risk Factor Assessment Checklist 30 . In our sample, a higher proportion of patients younger than 35 years showed an increase of weight (43.7%) compared to those of 35-55 years (39.1%), or those older than 55 years (10.2%). According to our estimates, for each year of increasing age, the risk of weight gain diminishes 3%; this estimate remained stable, barely without changes from the crude estimate. Thus, increasing age appears as a protective factor for antipsychotics-associated weight gain in the studied sample. This has been consistently observed in other studies 24,29,31,34 ; it must be underlined the susceptibility observed in children and adolescents to antipsychotics-associated obesity 32,35,36 . Age is also associated with certain diseases, comorbidities, and the corresponding treatments: i.e., doses of antipsychotics in elderly populations are usually lower than those used in the youngest ones. Moreover, age is related to physiological changes, which could be related to a less propensity for weight gain. None of these factors seem to have a significant modifying effect in our estimate.
Up to eleven different antipsychotics were used by patients in our sample (Table 1); all, but haloperidol, were atypical. Olanzapine was independently and strongly associated with weight gain in our sample ( Table 2); this is consistent with previous findings 1,21,23,31,[37][38][39] and with two meta-analyses, where olanzapine showed the highest risk of weight gain 5,40 . Although in our study quetiapine increased the initial weight in 0.4 kg as an average, this increase was lower than that observed with the other antipsychotics; this is coincidental with results of a recent meta-analysis 5 . Most of the patients in the cohort received only one psychiatric diagnosis and had one antipsychotic, though receiving several antipsychotics is a common practice in schizophrenia 41,42 ; this lower use found might be related to the lower proportion of psychotic patients in our sample.
In this study exploring 38 polymorphisms located in 14 candidate genes, two were found to be associated with antipsychotics-induced weight gain in CNR1 and INSIG2 genes ( Table 3). The minor T-allele of the frequently studied synonymous variant in the final exon of CNR1 (rs1049353) conferred some protection; it was observed that carriers of the minor allele gained less weight while treated with antipsychotics than the CC homozygotes (OR = 0.41, 95% CI = 0.18-0.94, p = 0.034). To our knowledge, four studies have explored this association so far [43][44][45][46] . Our results are coincidental with those found by Nurmi et al. 43 in a relatively large sample of children with autism treated for the first time with antipsychotics (n = 181); they differ, however, from the results found by the others. In the study by Monteleone et al. 44 , carried out similarly in a Mediterranean population like ours (MAF, 17% and 22%, respectively), comparable associations would be expected; notwithstanding, subtle differences in design might account for the disparities observed: i.e., patients who had been previously exposed to antipsychotics being already obese could participate in that study (n = 83). Regarding the study by Park et al. 45 (n = 78), the differences in allelic distribution of the Asian population they analyzed, with a MAF of 7%, by far lower than ours (22%), make it difficult to detect any association when the exposure in the cases is low, as it was the case (6.7%). The study by Tiwari et al. 46 in samples exposed to clozapine and olanzapine with an elevated initial body weight (n = 183) neither found any association in a shorter follow-up period of 14 weeks, though it was found with other polymorphisms in the same gene; all in all, it is concluded that the CNR1 gene may be associated with antipsychotics-induced weight gain in chronic schizophrenia. As for the INSIG2 polymorphism (rs7566605), we found an association with weight gain in a recessive model. A genome-wide study by Skelly et al. 47 looked at the same marker in a large sample of patients treated with antipsychotics from the CATIE study 1 ; for the same recessive model, it was observed an analogous trend in magnitude to that first reported by Herbert et al. 48 in the general population [in subjects of European ancestry, mean BMI (SD) was 31.9 (0.94) for the CC genotype versus 30.8 (0.43) for combined CG/GG genotypes]. As denoted by these figures, a possible explanation for these inconclusive results might be the elevated baseline BMIs of the patients included in the CATIE study 1 ; otherwise, a study primarily intended to assess a different outcome. Also, the study by Tiwari et al. 49 found certain association for the same allele with this variant in a subset of patients with African ancestry who received either clozapine or olanzapine (n = 54). Conversely, two studies exploring the same variant did not find any statistically significant relationship (n = 160 and n = 128, respectively) 50,51 ; patients in both studies, with shorter durations than ours, had been previously exposed to antipsychotics, being stated in one of the studies that it could bias the identified amount of body weight gain 51 .
The cannabinoid receptor (CNR1 or CB1) is related to food intake and lipogenesis 52 . Cnr1 knockout mice presents a lean phenotype [53][54][55] , and, consistently, administration of the inverse agonist rimonabant leads to decreased body weight 56,57 ; on the contrary, endogenous ligands of CNR1 increases dietary food intake 58  www.nature.com/scientificreports/ involved in the processing of sterol regulatory element binding proteins (SREBPs): CNR1 receptor activity leads to increased lipogenesis through augmented expression of these SREB proteins 60,61 and, on the contrary, INSIG2 blocks these proteins acting as a negative regulator of cholesterol biosynthesis 62 . Hence, it is conceivable that variants in genes encoding these proteins may alter their normal functioning, accounting for metabolic dysregulation and yielding to weight gain; this is what has been observed in our study: protection with a CNR1 variant and risk with other INSIG2 variant. Among the study limitations, sample size and heterogeneity are the most remarkable. Heterogeneity is linked to natural settings in which patients are quite different in diagnoses, ages, and treatments; thus, it makes it difficult to identify minor risk factors since larger samples are needed for stratification. Conversely, this setting and the observational nature of our study provides the actual variety of patients on these medications; in addition, a close follow-up has permitted a more detailed collection of the main features of all individuals in the study which includes adherence-dose collected were those that patients were taking. As for genetic association analysis, the stringent inclusion criteria upon use of antipsychotics-most of our patients were drug naive prior to the beginning of the study limiting the influence of one of the most confounding factors in previous studies-and weight along with the prospective character of this study make it suitable to clearly observe the development of the condition studied and assess the genetic risk factors; though, in terms of genetic association studies, the number of subjects is rather small, it still represents a comparably large sample for the specific phenotype. Hidden population always remains as a possible confounder; however, all individuals in our sample were from European origin, they come from the same geographical region, Castile (Spain), without cultural differences. No migrants or adopted children were found in the recruited patients; moreover, the Spanish population is generally like that of Northern and Western Europe origin, but also largely homogeneous within itself 63,64 .
In summary, our study has identified two major risk factors for weight gain when on antipsychotics, younger age and olanzapine, as well as a third relative protective factor, quetiapine; these influential factors are fully discussed in the light of more information. Additionally, the observations made in the large screening of this study indicate a possible role of certain polymorphisms in CNR1 and INSIG2 genes in antipsychotic-induced weight gain; at this regard, new approaches in the study design considering the likely polygenic character of the genetics influence for weight gain should be adopted. All in all, our results should be interpreted as exploratory and require replication in an independent and larger data set. www.nature.com/scientificreports/