Introduction

Obesity is a complex disorder that is caused by multiple factors [1]. Importantly, obesity is an outcome of interplay between environment and genetic makeup of an individual or a population. The risk genotypes of genetic variants may have more pronounced expression in the presence of obesogenic factors such as energy-dense foods and physical inactivity [2]. Nonetheless, obesity is primarily genetically determined; therefore, contribution of genetic factors cannot be overlooked. Genetic makeup of an individual greatly influences body mass index via affecting metabolism, appetite, and tendency for physical activity [3]. Twin and adoption studies provided significant evidence for a genetic component of obesity [4]. Several genes are involved in the process of energy regulation as well as in adipogenesis. Thus, variations in these genes may predispose an individual to obesity [5,6,7]. Genetic heterogeneity and allelic diversity exist throughout different populations of the world even in the populations living exclusively at a very restricted geographical area [8,9,10]. A study has confirmed an indisputable impact of common genetic variants on obesity predisposition in a population-specific manner [11]. Therefore, the association of the genetic variants with respective phenotypes must be studied at population level. Pakistani population provides an opportunity for the research community to explicate the interaction between obesogenic environmental modifications and obesity-predisposing genes due to its unique characteristics such as complex demographic history, ethnic diversity, consanguinity, large pedigrees, and current nutritional shift [12].

In genome-wide association studies (GWAS), among numerous obesity-linked genes, the strongest signal has been detected with the fat mass and obesity-associated (FTO) gene [13, 14]. The FTO was found to be expressed in many tissues in humans and animals. However, FTO mRNA is abundantly found in the hypothalamus of brain that controls energy balance [15, 16]. Thus, any functional variation in the FTO gene may affect energy balance and may lead to manifestation of overweight/obese phenotype. The rs1421085 of the FTO gene (involving T → C transition) has recently been predicted as the causal variant for obese phenotype [17]. However, the association of the FTO rs1421085 variant with obesity and obesity-related parameters has never been studied in Pakistani population. Moreover, the interactive effects of the FTO rs1421085 and overweight/obesity (FTO rs1421085 × overweight/obesity) on metabolic parameters were never explored before in Pakistani as well as in any other population. Therefore, the current study was carried out to find the association of the FTO rs1421085 with overweight/obese phenotype and anthropometric and metabolic parameters in a sample of Pakistani individuals. In addition, aforementioned effects of interaction have also been sought in this study.

Materials and methods

Study population

It was a case–control study, which was performed at Dr. Panjwani Center for Molecular Medicine and Drug Research (PCMD), University of Karachi, Pakistan, after approval from Advanced Studies and Research Board and ethical committee of the institute. The study encompassed a total of 612 Pakistani subjects including 306 overweight/obese cases and equal number of their age- (± 5 years) and gender-matched non-obese subjects (n = 306). Simple random sampling technique (without replacement) was employed for the recruitment of subjects. The study subjects were between the ages of 12–63 years. Reference ranges for BMI (≥ 25 kg/m2 for overweight and ≥ 30 kg/m2 for obesity) in adult subjects (≥ 20 years) were followed as recommended by World Health Organization (WHO), whereas for children and adolescents (< 20 years), BMI reference ranges (85th–94th percentile for overweight and ≥ 95th percentile for obesity) were followed according to the growth charts given by Center for Disease Control and Prevention (CDC). Subjects with any type of endocrinopathy and medication history of drugs that may increase body weight were not included in the study. Moreover, the study involved those overweight/obese cases, whose diabetic status was not known, so that we can also find the association of the variant with the obesity-mediated impaired metabolic parameters that may eventually lead to diabetes. Informed consents were obtained from all study subjects before participation in the study.

Blood sample collection

Blood sample (5 mL) was collected following an overnight fasting of 8–12 h. The isolated blood was utilized for subsequent serum and DNA isolation.

Anthropometric parameters

Anthropometric measurements including body height, weight, body mass index (BMI), waist circumference (WC), hip circumference (HC), and skinfold thicknesses from various sites (biceps, triceps, abdomen, suprailliac, sub-scapular, and thighs) were taken from each subject by following standard procedures. The waist-to-stature ratio (WSR) was calculated via dividing the values of WC by the values of height. The measurements of waist and hip circumference were used to determine waist-to-hip ratio (WHR = WC ÷ HC). Sum of skinfold thicknesses (SFTs) was used to compute body fat percentage (% BF) via gender-specific formulae [18]. For males: % BF = [(0.29288 × sum of skin folds) − {0.0005 × (sum of skin folds)2} + {(0.15845 × age) − 5.76377.2}], for females: % BF = [(0.29669 × sum of skinfolds) − {0.00043 × (sum of skinfolds)2} + {(0.02963 × age) + 1.4072}].

Metabolic estimations

Systolic and diastolic blood pressure was measured from right arm of the subject using mercury sphygmomanometer (CR-2001, Certeza medical, Germany) with 1 mm Hg accuracy. Concentration of fasting blood glucose was evaluated using blood glucose monitoring system (Abbott, UK). Serum insulin levels were determined by enzyme-linked immunosorbent assay (ELISA) using commercial kit (DIA source, Belgium) as per instructions of manufacturer’s protocol. Homeostatic model assessment–insulin resistance (HOMA–IR) was calculated by computing values of fasting glucose and insulin in the corresponding formula [19]. All lipid parameters including serum triglycerides (TG), total cholesterol (TC), high-density-lipoprotein cholesterol (HDL-C), and low-density-lipoprotein cholesterol (LDL-C) were measured using parameter-specific assay kits (Merck, Darmstadt, Germany) involving enzymatic colorimetric assay on an automatic chemistry analyzer (Roche Hitachi 902, Tokyo, Japan). The values of very-low-density-lipoprotein cholesterol (VLDL-C) were calculated by dividing TG with 5 (TG ÷ 5). In addition, by computing the values from different anthropometric and metabolic estimations, various metabolic indices and ratios such as visceral adiposity index (VAI) (mmol L−1), lipid accumulation product (LAP) (mmol L−1), product of triglyceride and glucose (TyG) index, coronary risk index (CRI), atherogenic index (AI), and triglyceride-to-HDL-C ratio (TG/HDL-C) were calculated. Gender-specific formulae were used to calculate VAI and LAP [20, 21]. The TyG index was calculated using general formula [22]. For calculating VAI and LAP, the values of TGs and HDL-C were converted into mmol L−1. CRI was calculated as total TC ÷ HDL-C, whereas AI was calculated as LDL-C ÷ HDL-C [23]. TG/HDL-C ratio was calculated as TGs ÷ HDL-C.

DNA extraction and genotyping

DNA from blood samples was extracted by spin column method using genomic DNA purification kit (Bio Basic, Canada). Allelic discrimination assay for genotyping the FTO rs1421085 variant was run on Applied Biosystems rtPCR machine (ABI 7500, Thermofisher Scientific, USA). For this purpose, the TaqMan® primers/probes assay (Assay ID C___8917103_10, cat. # 4351379, Applied Biosystems Thermo Fisher Scientific Inc., USA) and master mix (part # 4371355, Applied Biosystems Thermo Fisher Scientific Inc., foster city, USA) were utilized. For each experiment, two non-template/negative controls (NTCs) and three positive controls (PCs) for each genotype were included. The twenty percent (20%) of the samples were genotyped twice for reproducibility.

Statistical analysis

Genotypic frequencies in cases and controls were calculated via Chi square (χ2) test whereas allelic frequencies were calculated by direct count. Genotypic and allelic frequencies were reported as counts and percentages. Hardy–Weinberg Equilibrium (HWE) test was carried out for both cases and controls to determine whether the allelic and genotypic frequencies are in HWE. HWE test was performed using an online program [24]. The normality of all quantitative variables was checked via Shapiro–Wilk test. Rank-based inverse normal transformation was carried out for all non-normal quantitative variables. For subsequent analysis, transformed values of quantitative variables were used. All study variables were presented as mean (standard deviation). The association of the FTO rs1421085 with overweight/obesity was sought by assuming multiple genetic models including co-dominant, dominant, recessive, and over-dominant model. For dominant, recessive and over-dominant models, association was determined by binary logistic regression. However, in case of co-dominant model, multinomial logistic regression was used. Age and gender adjusted odd ratio (OR) and 95% confidence intervals (CI) were calculated to determine the risk of overweight/obesity associated with the rs1421085 variant. To find the appropriate mode of inheritance in case of getting association in more than one genetic model, h-index (degree of dominance index) was calculated [25]. The association of the rs1421085 variant with all study variables was determined by linear regression. All tests for finding association between the FTO rs1421085 and obesity-related secondary outcomes were adjusted for confounders such as age, gender, and BMI except anthropometric traits, which were adjusted for age and gender only. Effect size (β) with 95% confidence interval (CI) was determined for all variables. The effects of interaction between the FTO rs1421085 and BMI on metabolic variables were also determined by linear regression. The transformed values of BMI were centered for convenience in the interpretation of interaction effects [26]. Scatter plots were drawn for validation of significant interaction effects. In scatter plots, X-axis showed independent variable (BMI) and Y-axis showed dependent variable (metabolic parameters), whereas intersecting lines represented respective genotypes. To plot interactions, age and gender adjusted predicted values were determined for each dependent variable. All statistical tests having p value < 0.05 were considered significant. All association and interaction analyses were corrected for multiple comparisons by Benjamin–Hochberg method [27]. All statistical analyses were performed using Statistical Package for Social Sciences version 19.0 (SPSS, Inc. Chicago, IL, USA IBM statistics).

Results

The distribution of allelic and genotypic frequencies of the FTO rs1421085

The distribution of allelic and genotypic frequencies of the FTO rs1421085 variant among cases and controls is shown in Table 1. Minor allele frequency (MAF) of the FTO rs1421085 variant was 34% in cases and 31% in controls. The genotypic frequencies of controls (p = 0.14) were in compliance with Hardy–Weinberg Equilibrium (HWE), whereas those of cases were in disequilibrium (p = 0.04). The genotypic frequencies of the rs1421085 in cases were 41.2% TT (wild type), 50% CT (heterozygous mutant), and 8.8% CC (homozygous mutant). Similarly, the genotypic frequencies of the rs1421085 in controls were 50.0% TT (wild type), 38.9% CT (heterozygous mutant), and 11.1% CC (homozygous mutant). Thus, the genotypic frequencies of the FTO rs1421085 variant differed significantly between cases and controls (Table 1).

Table 1 Genotypic frequencies of the rs1421085 in cases and controls and association between rs1421085 and overweight/obesity
Full size table

Association of the FTO rs1421085 with obesity risk and anthropometric parameters

The association between the FTO rs1421085 and obesity risk was observed in co-dominant, dominant, and over-dominant models, which remained significant even after age and gender adjustment (Table 1). Thus, the degree of dominance (h-index) was calculated to identify mode of inheritance which turned out to be over-dominant (hadjusted = 11.85). Thus, according to over-dominant model, the CT genotype of the FTO rs1421085 variant increased 1.583 times the risk of having overweight or obese phenotype (ORadjusted = 1.583, 95% CI 1.147–2.185, padjusted = 0.005). The FTO rs1421085 variant (CT genotype) was also found to be significantly associated with almost all anthropometric measurements (p < 0.05) except height and waist-to-hip ratio (Table 2). All associations retained significance after age and gender adjustment (p < 0.05) and correction for multiple comparisons.

Table 2 Association of the FTO variant rs1421085 with anthropometric parameters
Full size table

Lack of association of the FTO rs1421085 with obesity-related metabolic parameters

No association of the rs1421085 with obesity-related metabolic parameters (Table 3) (p  > 0.05) was observed.

Table 3 Association of the FTO variant 1421085 with obesity-related metabolic traits
Full size table

Interaction between the FTO rs1421085 variant and obesity influences multiple metabolic traits

A significant effect of interaction (pinteraction < 0.05) between the FTO rs1421085 and obesity on obesity-related metabolic parameters (except HDL-C) was observed in initial analysis (Table 4). All the interactions retained statistical significance after age and gender adjustments and correction for multiple comparisons. Further validation of the interaction effects by comparing regression slopes (Fig. 1) in follow-up analysis suggested that the CT genotype of the FTO rs1421085 interacts with higher BMI (overweight/obesity) to increase SBP, DBP, TyG index, TG, LDL-C, VLDL-C, CRI, AI, and TG/HDL-C ratio. However, this interaction effect (rs1421085 × BMI) could not be validated for the rest of the parameters including LAP, FBG, fasting insulin, HOMA–IR, and TC in follow-up analysis.

Table 4 Interactive effect of the rs1421085 and BMI on metabolic parameters
Full size table
Fig. 1
figure 1

Effect of interaction between the FTO rs1421085 and BMI on obesity-related metabolic phenotypes. Estimated centered values of BMI were given on X-axis. Age and gender adjusted effect sizes for SBP (a), DBP (b), TyG index (c), TG (d), LDL-C (e), VLDL-C (f), CRI (g), AI (h), and TG/HDL-C (i) were given on Y-axis. Black and grey dots represent all individual observed values of the corresponding parameter for CT and CC/TT genotypes of the rs1421085, respectively. Regression lines show the effect of genotypes (CT and CC/TT) of the rs1421085 on dependent variables (metabolic parameters) in the presence of increasing BMI

Full size image

Discussion

Obesity is a multifactorial disorder that necessitates environmental influences for its expression and the higher predisposition of some individuals to weight gain in an obesogenic environment is mainly determined by genetic factors [28]. Obesity-predisposing genetic factors strongly interact with environment or behavior throughout the life of an individual and in a wide range of situations [29]. Thus, the effects of genetic variants that increase susceptibility to obesity can vary from population to population. The variants rs17782313 near MC4R gene and rs9939609/rs1421085 in the FTO gene are considered among key contributors to human polygenic obesity [30]. We previously reported the gender-specific association of the variants MC4R rs17782313, LEP rs7799039, and FTO rs9939609 with overweight/obese phenotype and related anthropometric traits in Pakistani females [31,32,33]. In addition, we also previously reported lack of association of the BDNF rs6265 variant with obesity and related traits [34]. However, we could not study association of the FTO rs1421085 variant with overweight/obesity and obesity-related parameters before. In addition, this association has not been explored by other investigators from Pakistan so far despite the fact that the variant (rs1421085) has recently been predicted as the causal variant for obese phenotype [17]. Though the association of the FTO rs1421085 variant with overweight/obesity has not been previously explored in Pakistani population, however, there are few studies, which investigated the association of FTO variants other than rs1421085 with overweight/obesity in Pakistanis. For instance, Shabana and Hasnain observed the association of three FTO variants rs9939609, rs1121980, and rs9926289 with obesity in Pakistani subjects independent of age and gender [35]. Similarly, another study by Shahid et al. reported the association of rs9939609 with obesity in Pakistanis [36]. On the other hand, Fawwad et al. reported lack of association of the variant rs9939609 with obesity but the variant was found to be associated with the risk of developing type 2 diabetes [37]. By keeping the aforementioned scenario in mind, the present study was undertaken to seek the association of the FTO rs1421085 variant with overweight/obesity and related traits (anthropometric and metabolic) in Pakistani individuals. Among these traits, WSR, % BF, TyG index, CRI, AI, and TG/HDL-C ratio and visceral adiposity markers (VAI and LAP) have never been studied in relation to association with FTO rs1421085 variant in our and in other populations. Moreover, the interaction effects of the FTO rs1421085 and overweight/obesity on obesity-related metabolic traits were also investigated in the current study, which were never sought before in Pakistani as well as in any other population.

The present study reported a significant association of the rs1421085 with overweight/obesity in Pakistani population. The minor allele frequency (MAF) of the FTO variant rs1421085 (C allele) observed in the current study (32%) was comparable with that reported in north Indians (29%) [38]. However, the MAF observed in our study was higher than those observed in Caucasians (11.4%), American Samoans (18.2%), and Samoans (16.8%), but lower as compared to that seen in Chinese (44.8%) [39]. Differences in allele frequencies among various populations located in different geographic regions are attributable to natural selection and neutral genetic drift. Differential selective pressure of environmental factors may also create differences in allele frequencies between populations [40,41,42,43]. In the current study, the frequency of CT genotype differed significantly between overweight/obese and non-obese subjects, while the difference for other genotypes (CC or TT) was not significant. Thus, heterozygotes or CT carriers of the FTO rs1421085 variant showed a greater risk of having obesity as compared to both homozygotes (CC and TT carriers) in our study. Similarly, other studies conducted on Asian [38, 44,45,46], European  [47, 48], and American [49, 50]  populations also reported a significant association of the rs1421085 with obesity. However, it is important to note that none of aforementioned studies assumed multiple genetic models in association tests and did not report appropriate mode of inheritance. We considered four models of inheritance including co-dominant, dominant, recessive, and over-dominant in association test and found association in more than one model (co-dominant, dominant, and over-dominant). Thus, we then calculated degree of dominance or h-index to find the appropriate mode of inheritance, which turned out to be the over-dominant in our case. In disagreement to the aforementioned studies [38, 44, 46,47,48, 50] and our study, the variant rs1421085 was not found to be associated with obesity in Turkish and Oceanic populations [51, 52]. In genetics, over-dominance is a condition, wherein heterozygote shows different phenotype than any of the homozygote. It is generally explained as heterosis or hybrid vigor, which advantageously increases the survival of heterozygous individuals in the natural selection. Heterosis is progressively being identified in humans [53, 54], but heterotic associations are generally not identified by allele-based linkage techniques [55]. The occurrence of excessive heterozygosity is consistent with over-dominance which might be due to various biological (disassortative mating) and socio-demographic factors that may vary among populations [56]. Moreover, fat storage improves survival chances during famine and scarcity of food that occurred throughout the history of man and it also possibly supports thrifty gene hypothesis [57].

The current study revealed a significant association of the rs1421085 with obesity-related anthropometric traits. Among these traits, the association of the FTO rs1421085 variant with weight, height, BMI, WC, HC, and WHR was investigated in very few studies [39, 47, 49, 58]. Among these studies, Zhang et al. in Croatians, while Albuquerque et al. in Portuguese children reported an association of the rs1421085 with weight, BMI, WC, and HC [47, 58]. Moreover, the rs1421085 was also found to be associated with BMI and WC in White- and African-Americans [49]. In contrast, lack of association between the rs1421085 and obesity-related anthropometric phenotypes was observed in Polynesians [39]. Only Zhang et al. [58] explored the association of the rs1421085 variant with skinfold thicknesses, as we did in our study. However, in contrast to our study, no association of the rs1421085 variant with biceps SFT, triceps SFT, sub-scapular SFT, abdominal SFT, supra-iliac SFT, and thigh SFT was seen. The variability of associations across different populations may be due to their different ethnic backgrounds. Furthermore, in line with the previous reports [47, 49, 58], we found lack of the association of the rs1421085 with height and WHR in our study. In contrast, a significant association of the rs1421085 with WHR was observed in White Americans [49]. According to our knowledge, no study examined the association of the rs1421085 with WSR and % BF. Like WC, WSR is also a useful predictor of central obesity and cardiovascular health [59,60,61]. In addition, WSR was also found to be an effective indicator of mortality risk [62]. Thus, our study highlighted that the rs1421085 may increase the risk of cardiovascular disease and mortality via increasing WC and WSR. Moreover, the association of the genetic variant rs1421085 with skinfold thicknesses and % BF indicated a possible role of the rs1421085 in fat deposition.

Obesity has been strongly linked to hypertension, diabetes, cardiovascular diseases, and dyslipidemia. Therefore, in this study, we determined the association of the rs1421085 with various obesity-related metabolic traits. No association of the FTO rs1421085 was seen with any parameter related to glucose (FBG, fasting insulin levels, HOMA–IR, and TyG index) and lipid (TC, TG, HDL-C, LDL-C, VLDL-C, CRI, AI, and TG/HDL-C ratio) metabolism, visceral adiposity markers (VAI and LAP), and blood pressure. In line with our findings, several studies observed lack of the association of the rs1421085 variant with SBP, DBP, FBG, insulin levels, HOMA–IR, TC, TG, HDL-C, and LDL-C in different populations including Han Chinese, Koreans, European, French Croatians, and Egyptian females [44, 46, 63,64,65,66,67]. However, Karns et al. found the association of the rs1421085 variant with HDL-C in Croatians [64]. Moreover, significant association was seen between this FTO variant and impaired fasting glucose among Japanese and Tunisians [68, 69]. Likewise, Hotta et al. also reported the association of the rs1421085 with hypertension (↑SBP and/or ↑DBP) in Japanese [68]. The observed differences among various studies could be due to several reasons. A potential reason for lack of the association of the FTO rs1421085 with various parameters related to glucose metabolism (FBG, fasting insulin levels, HOMA–IR and TyG index) could be that we did not include diabetic overweight/obese subjects in our study. Moreover, some studies were exclusively carried out on children and females. Thus, age and gender differences among studies may be the cause of discrepancy between outcomes of various studies. It is also noteworthy that we did not find any study in literature that explored the association of the FTO rs1421085 with TyG index, CRI, AI, and TG/HDL-C ratio and visceral adiposity markers (VAI and LAP). Thus, the current study indicated that this FTO variant might have a role in the expression of obese phenotype among Pakistanis but could not influence metabolic traits directly. However, it was observed in the current study that the rs1421085 variant interacted with obesity (FTO rs1421085 × overweight/obesity) to influence the metabolic traits including SBP, DBP, TyG index, TG, LDL-C, VLDL-C, CRI, AI and TG/HDL-C ratio. In the presence of higher BMI, carriers of CT genotype of the rs1421085 showed aberrant metabolic traits than TT/CC homozygotes. It suggests that through interaction with overweight/obesity, this FTO variant may increase the risk of dyslipidemia and hypertension irrespective of age and gender. Consequently, this interaction may increase the proneness of an individual to coronary heart disease via anomalous levels of lipid profile components and elevated values of atherogenic indices (CRI and AI). We did not find any parallel and contrary study in the literature in comparison with our study regarding the effect of interaction between the rs1421085 with BMI (FTO rs1421085 × overweight/obesity) on metabolic traits. It is noteworthy that different metabolic traits may be influenced by genetic variations only in the presence of predisposing environment such as increased BMI [70,71,72].

At this point, some limitations of the current study are indicated. First, the recruited sample size for the present study is modest, while larger sample size provides better investigation of the real effect for SNP association studies. Second, we considered only one variant of the FTO gene; therefore, we cannot be certain about the presence and the association of the other variants of the FTO gene and their association with the overweight or obese phenotype, though we discussed and compared the previously reported associations of FTO variants other than rs1421085 from Pakistan and also our own previously reported studies regarding genetic variants other than the FTO gene variants with the results of the current study. The aforementioned limitations of the current study were attributable to cost and time constraint.

Conclusions

In conclusion, heterozygous genotype (CT) of the FTO rs1421085 may be associated with the risk of having overweight/obese phenotype and related aberrant anthropometric parameters in Pakistani population. In addition, the interaction between CT genotype and overweight/obesity may lead to obesity-related aberrant metabolic parameters.