Gender-specific genetic associations of polymorphisms in ACE, AKR1C2, FTO and MMP2 with weight gain over a 10-year period

Abstract

Weight gain, when it leads to overweight or obesity, is nowadays one of the major health problems. ACE, FTO, AKR1C2, TIMP4 and MMP2 genes have been implicated in previous studies on weight regulation. This study investigated the contribution of polymorphisms in these five candidate genes to the risk of weight gain over a 10-year time period. Two groups were selected from participants of the Doetinchem cohort study who were followed over a 10-year period: A stable weight group (±2 kg/10 year; n = 259) and a weight gainers group who increased their body weight by roughly 10 % (>8 kg/10 year; n = 237). Starting BMI was between 20 and 35 kg/m² and baseline age between 20 and 45 years. Selected SNPs and insert/deletion in candidate genes were measured in each group. In men, the allelic distribution of FTO rs9939609 (χ ² p = 0.005), ACE rs4340 (χ ² p = 0.006) and AKR1C2 rs12249281 (χ ² p = 0.019) differed between the weight stable and weight gainers group. Interaction between FTO rs9939609 and ACE rs4340 was observed. In women, the allelic distribution of MMP2 rs1132896 differed between the weight stable and weight gainers group (χ ² p = 0.00001). The A-allele of FTO was associated with a 1.99× higher risk of gaining weight in men (OR 1.99, p = 0.020), while in women, the C-allele of MMP2 was associated with a 2.50× higher risk of weight gain (OR 2.50, p = 0.001) over the 10-year period. We found that FTO in men and MMP2 in women are associated with weight gain over a 10-year follow-up period.

Introduction

Weight gain, when it leads to overweight or obesity, is nowadays one of the major health problems (Wyatt et al. 2006). Unfortunately, the etiology of weight gain is complex and has remained largely unclear. Studies have indicated that lifestyle factors like physical activity, eating behavior and the way to spend leisure time are of influence (Brug et al. 2012). Moreover, body weight seems also to be determined by genetic background, which may be important as an offset of the energy balance as MacLean reported (Maclean et al. 2011).

Several genetic variants have been associated with body weight or weight gain. Polymorphisms in the angiotensin converting enzyme (ACE) gene have been found to be associated with body weight, body mass index (BMI), overweight and obesity (Wang et al. 2012). Recent data suggest that the renin-angiotensin system might be involved in the pathophysiology of obesity and associated hypertension (Darimont et al. 1994). ACE gene polymorphisms influence the levels of plasma ACE, and thereby the level of angiotensin II which is known to play a role in adipocyte growth and the differentiation of preadipocytes to mature adipocytes (Passaro et al. 2011). The best studied ACE gene polymorphism in this respect is the insertion/deletion (I/D) located in intron 16. Additionally, rs9939609 in the gene coding for fat mass and obesity-associated protein (FTO) is a well-known predictor for increased body weight and BMI (Frayling et al. 2007; Scuteri et al. 2007). In adults, individuals with the AA genotype of rs9939609 weigh on average 3 kg more than those with the TT genotype, and each A-allele increases the BMI with approximately 0.4 kg/m² (Frayling et al. 2007). How FTO influences the development of obesity remains unclear, but recent studies suggest that FTO may influence adiposity by affecting appetite (den Hoed et al. 2009). Another suggested mechanism is that obesity-associated SNPs in the intron region of the FTO gene act on body weight via regulation of the expression of IRX3, a homeobox gene playing a role in early neuron development (Ragvin et al. 2010; Smemo et al. 2014). Another candidate gene is the gene for aldo–keto reductase 1C2 (AKR1C2). We have observed correlations between AKR1C2 protein level in adipose tissue and parameters of adiposity (Bouwman et al. 2009). It has also been reported that the level of AKR1C2 mRNA expression in adipose tissue of women positively correlates with adiposity (Blouin et al. 2005). Similarly, Wake et al. (2007) showed that central fat distribution, measured by waist/hip ratio, was associated with mRNA expression for AKR1C2 in white adipose tissue.

An association has also been found between obesity and genetic variation in the gene for matrix metalloproteinase-2 (MMP2) (Han et al. 2008), of which the activity is controlled by tissue inhibitor of metalloproteinase 4 (TIMP4). MMP2 and TIMP4 are proteins involved in the remodeling of the extracellular matrix (ECM) of adipocytes, which is essential for the development of adipocytes (Bouloumie et al. 2001). An imbalance between MMP2 and TIMP4 could have an effect on ECM remodeling and as such on fat mass development (Chavey et al. 2003).

The aim of the present study was to investigate if preselected genetic polymorphisms in these candidate genes influence weight gain over a 10-year period of time.

Materials and methods

Study population

The present study selected subjects from the ongoing Doetinchem cohort (Lu et al. 2008; van den Berg et al. 2009). The design of the Doetinchem study and detailed methods have been reported earlier (Verschuren et al. 2008). Participants were eligible for the present study if he/she has participated in three consecutive rounds of examinations covering a 10-year period, were stable in smoking habits and not pregnant, starting BMI between 20 and 35 kg/m², and baseline age between 20 and 45 years (Fig. 1). Based on a relative risk of two, a power calculation indicated that 208 subjects in each group are required for a power of 0.8. We selected a stable weight group (<±2 kg/10 year) and a weight gainers group (>8 kg/10 year) from these eligible subjects. The two groups were group-matched for baseline age, BMI and gender, resulting in 282 participants per group. In the stable weight group, 259 DNA samples of the participants were available and in the weight gainers group, 237 DNA samples of the participants (Fig. 1). The 8 kg weight gain is roughly 10 % of the initial body weight. The study was approved by the external Medical Ethics Committee (MEC) of the Netherlands Organization of Applied Scientific Research according to the guidelines of the Declaration of Helsinki. A written informed consent, approved by the MEC, was obtained from all participants. At the start of the study, a blood sample was drawn, fractionated to obtain buffy coats for isolation of DNA and subsequently stored at −30 °C.

Fig. 1

Flowchart of the different selection steps and the number of participants

DNA isolation, SNP selection and genotyping

Genomic DNA was isolated from stored buffy coats using a salting out method as described before (van den Berg et al. 2009). To ensure a considerable fraction of carriers of the minor allele, only SNPs with a minor allele frequency in Europeans of at least 25 %, as indicated in the SNP public database (dbSNP; http://www.ncbi.nlm.nih.gov/SNP), were considered as candidates. Preferably, SNPs were considered that had earlier been shown to be associated with relevant phenotypes, such as BMI or body composition. Polymorphisms in angiotensin converting enzyme (ACE), fat mass and obesity-associated protein (FTO), aldo–keto reductase 1C2 (AKR1C2), tissue inhibitor of metalloproteinase 4 (TIMP4) and matrix metalloproteinase-2 (MMP2) were chosen as candidate genes for weight gain, also based on previous research (Bouwman et al. 2009; Wang et al. 2012). Genotyping was performed using commercially available TaqMan SNP genotyping assays from Applied Biosystems (Foster City, CA, USA). The procedure was performed according to the manufacturer’s protocol and measured on an Applied Biosystems 7900 HT fast real-time PCR system. Allelic calls were determined semiautomatically using the SDS 2.3 software. The ACE deletion/insertion (rs4340) was determined with the following primers: Forward primer 5′-CCTTTCTCCCATTTCTCTAGACCTG-3′ DEL-Reverse Primer 5′-GCTCAGAGAATTTCAGAGCTGGA-3′ INS-Reverse Primer 5′-GATCCCGCCACTGCACTC-3′ DEL-Probe 5′-VIC-TGCCTATACAGTCACTTTTATGTGGTTTCGC-TAMRA-3′ and INS-Probe 5′-FAM-CTCGCTCTGTCGCCCAGGCT-TAMRA-3′. The PCR was performed using the following PCR program: denaturation 10 min at 95 °C followed by 40 cycles of denaturation at 95 °C for 15 s and annealing/extension at 62 °C for 1 min (Koh et al. 2003). The allelic calls were determined semiautomatically using the SDS 2.3 software.

Statistical analysis

Hardy–Weinberg equilibrium (HWE) for the different polymorphisms was calculated using the Chi square test. Linkage disequilibrium (LD) was calculated using SNAP which is a web-based tool for identification and annotation of SNPs in LD using HapMap 3 (release 2) (Johnson et al. 2008). SNAP is publicly available at http://www.broad.mit.edu/mpg/snap. We used Pearson’s Chi square test to examine differences in the frequencies of the genotypes between the weight stable and weight gainers group. The inheritance model with the most significant p value was chosen using the SNPStats program (Sole et al. 2006). SNPStats is publicly available at http://bioinfo.iconcologia.net/snpstats. All statistical analyses were performed for men and women separately and adjusted for matching criteria, i.e., baseline age and baseline BMI. Logistic regression was performed using the logit function in generalized linear model to examine the effect of the polymorphisms on weight gain, using baseline age and BMI as covariates. Pearson’s Chi square test and logistic regression analyses were done using the SPSS v20.0 statistical software. To identify epistatic interactions, the multifactor dimensionality reduction (MDR) method was used (Heidema et al. 2007). The MDR software is publicly available at http://www.epistasis.org. A p value <0.05 was considered statistically significant. To evaluate the gene-dietary interaction, the cohort was stratified to low, medium and high intake level for each dietary factor in each sex group. Logistic regression was used to examine the effect of the interaction between allelic type and dietary intake level for weight gain on each dietary factor and each SNP. To correct for multiple testing, Bonferroni correction was applied, a significance threshold of p value = 0.007 was used in the Chi square test for examining the relation between seven SNPs and weight gain.

Results

Subject characteristics

On average, weight gainers increased their body weight with 11.8 kg in men and 13.2 kg in women. This was roughly 10 % or more of their baseline weight and resulted in a 3.6 and 4.9 kg/m² increase in BMI, respectively (Table 1). Information on dietary intake is given in Table 2. In women, carbohydrate and protein intake differed between the weight stable and weight gainers group. No differences were observed in men.

Table 1 Characteristics of the study population

Table 2 Average dietary intake in men and women over the 10-year follow-up period

Genetic analysis

The genotypic and allelic distributions of the determined polymorphisms are shown in Table 3. All SNPs were in Hardy–Weinberg equilibrium. None of the SNPs were in linkage disequilibrium (LD), indicating that all SNPs provided independent information and thus can be used separately for further statistics.

Table 3 Genotype and allele frequencies per polymorphism in the whole study population

ACE deletion/insertion rs4340

The distribution of the genotypes between the weight stable and weight gainers group was significantly different (p = 0.006) in men, but not in women (p = 0.84, Table 4). Although borderline statistically significant when using a recessive model, that fitted best, the I/I genotype increased the risk of weight gain over the 10 years in men [OR 1.91, 95 % CI (0.93–3.93), Table 5].

Table 4 Relationship between frequency of the different polymorphisms and weight gain over a 10-year period

Table 5 Results of logistic regression analysis with weight gain as dependent variable (stable weight is reference) and the genotypes as independent variable adjusted for baseline age and BMI

FTO rs9939609

The A-allele of this SNP was associated with weight gain in men, but not in women. The A-allele frequency was significantly higher in the weight gainers than in the weight stable group (0.45 vs 0.35, p = 0.001, Table 4). Also, the genotype distribution differed between the weight gainers and weight stable group (χ ² p = 0.005). Logistic regression analysis shows that the combined A/A and A/T genotypes of FTO in men were associated with weight gain (Table 5). Being a male A-allele carrier increased the risk of gaining weight over the 10-year period by 1.99-fold (95 % CI 1.16–3.39, Table 5).

AKR1C2 rs2854482, rs4143631 and rs12249281. rs2854482 is until now the only known functional SNP in the NCBI database of AKR1C2. This amino acid variation has an effect on the enzyme activity and on cofactor binding (Arthur and Reichardt 2010). The A-allele exchanges the amino acid F into Y, which decreases the enzyme activity. rs4143631 is in the promoter inside the CNV region, while rs12249281 is in the promoter but outside the CNV region. Figure 2 shows a schematic overview of the relative position of the AKR1C2 gene, the CNV and the investigated SNPs. The T/T homozygote of rs2854482 was not present in our study population (Table 4). For this SNP as well as for rs4143631, no significant differences in the genotype and allele frequencies between the weight stable and weight gainer groups were observed. In contrast, in men, the A/A genotype of rs12249281 was more frequent and the A/G genotype less frequent among weight gainers than among the weight stable group (χ ² p = 0.019 Table 4). However, linear regression based on a recessive model did not show significant results (Table 5). In women, no association with any of the AKR1C2 genotypes was observed.

Fig. 2

Schematic overview of the position of the AKR1C2 gene, the CNV region (copy number variation) and the investigated SNPs

TIMP4 rs3755724

In both men and women, no significant differences were found in the genotype or allele frequencies between the weight stable and weight gainers group (Table 4).

MMP2 rs1132896

Significantly, different allele frequencies were observed between the female weight stable and weight gainers group with the C-allele being more frequent among the weight gainers (43 vs 34 %, p = 0.005, Table 4). Also, the genotype distribution between the weight stable and gainers differed (χ ² p = 0.00001). The frequency of the C/G genotype was higher in the weight gainers (65 %) than in the weight stable group (43 %). The G/G genotype was lower (25 vs 45 %, Table 4). Logistic regression analysis showed being a carrier of the C-allele of MMP2 in women conferred a 2.50-fold (95 % CI 1.44–4.35) higher risk for weight gain over a 10-year period than being a carrier of the G/G genotype (Table 5). No significant differences between weight gainers and the weight stable group were observed for men.

SNP–SNP interactions

MDR analysis revealed a significant SNP–SNP interaction for weight gain between FTO and ACE in men (Fig. 3). Logistic regression shows that men with the T/T genotype for FTO in combination with the D-allele for ACE were protected against weight gain (OR 0.47, 95 % CI 0.27–0.85, p = 0.011). In contrast, the TT genotype for FTO in combination with the II genotypes for ACE seemed to increase the risk of weight gain, though not statistically significant. No SNP–SNP interactions were found in women.

Fig. 3

Numbers of weight stable and weight gainers group in men for the different multi-locus genotypes of the best model obtained by multifactor dimensionality reduction (MDR) (p = 0.05). Dark gray cells indicate increased risk of weight gain. Light gray cells indicate decreased risk of weight gain, in each segment, the left bar indicates the number of weight stable subjects and the bar on the right the number of weight gainers. The odds ratio is shown in the left upper corner. *p < 0.05

Gene-dietary interaction

We tested the possible contribution to the weight gain from the interactions of gene-dietary intake. However, we could not detect any significant contribution. The energy intake in women was 1.06× higher (p = 0.032) in the MMP2 C/C–C/G versus G/G carriers. This suggests that the MMP2 SNP works synergistically with energy intake on weight regulation.

Discussion

Here, we present a study of several gene polymorphisms in relation to weight gain over a longtime period. It should be noted that our study population is rather defined as being composed of a group that remain at normal weight over 10 years and a group that develops overweight during that period. We defined a weight gain of more than 8 kg over 10 years, as an obvious cutoff for weight gain. Because gender differences are regularly seen in matters of weight and weight regulation (Bradshaw et al. 2013; Heidema et al. 2010; Stevens et al. 2010), we decided to study men and women separately. Indeed we found sex-specific genetic associations. AKR1C2, ACE and FTO were associated with long-term weight gain in men, while MMP2 was associated with long-term weight gain in women over a 10-year period. It should be noted that about 30 % of the women passed the age of 50 during the 10-year period and therefore menopause might have affected weight gain. However, as we matched for age, the proportion was the same for the weight stable (39/127) and the weight gainers group (40/117) suggesting a minimal effect on our results.

AKR1C2 plays a role in the crosstalk between glucocorticoids and androgens as part of the regulation of adipogenesis and lipid accumulation (Veilleux et al. 2012). The androgen dihydrotestosterone (DHT) inhibits adipocyte differentiation in humans (Blouin et al. 2009b). Studies show that an increased expression or activation of AKR1C2 induces adipocyte differentiation by inactivating DHT (Blouin et al. 2009a; Veilleux et al. 2012). Previously, we found that AKR1C2 protein expression after weight loss/maintenance was increased and that the increase correlated with changes in parameters of adiposity including weight, BMI, waist and plasma LDL (Bouwman et al. 2009). Takahashi reported that the F46Y polymorphism (rs2854482) has a lowering effect on the enzyme activity (Takahashi et al. 2009). However, we did not find a genetic effect on weight gain, which might be explained by the low frequency of the T-allele in our population. In fact, only rs12249281 showed an association with long-term weight gain in men. This variation may have an influence via promoter activity of the AKR1C2 gene, but this requires confirmation in another cohort in addition to functional studies.

The Alu insertion/deletion (I/D) polymorphism in intron 16 of the ACE gene has been shown to be associated with body weight, BMI, overweight and obesity in various ethnic groups (Mao and Huang 2013). However, weight gain over a longer period has rarely been studied. Strazzullo et al. (2003) found that the DD genotype was associated with more spontaneous body weight gain and adiposity over 20 years in middle-aged men. Although in our study also an association between the ACE I/D polymorphism and weight gain was observed in men, it was the II-genotype that seems associated with long-term weight gain in men. Associations between ACE I/D and obesity parameters show the same controversy regarding the risk allele. For instance, in the study of Bienertova-Vasku et al. (2009), the II-genotype was found associated with the prevalence of obesity, whereas Settin et al. (2009) reported a significantly higher BMI in D-allele carriers. It has also been shown that subjects with the D/D genotype lose less body fat with similar amounts of weight loss than I-allele carriers (Hamada et al. 2011). It must be noted that rs4340 and rs1799752 are located in the same 288-bp insertion/deletion. In fact, rs1799752 is a 2-bp indel, compassing the 3′nucleotides of the 288-bp indel of rs4340. It is also known that the ACE I/D polymorphism influences plasma ACE concentration and activity. D-allele carriers have a higher mean plasma ACE concentration than I-allele carriers (Rigat et al. 1990). Interestingly, a change in ACE plasma concentration during weight loss might predict the risk for weight regain (Wang et al. 2012). It is tempting to speculate that D-allele carriers can more easily reduce their plasma ACE levels during weight loss, which is favorable for weight maintenance and would be in line with our results. However, no evidence has shown whether such concentration change and eventually the risk for weight regain is associated with ACE genetic variation.

The FTO gene has been repeatedly reported for being associated with parameters of adiposity (Day and Loos 2011; Vimaleswaran et al. 2012). Here, we found that the A-allele conferred increased risk of weight gain over 10 years, but only in men. At the moment, we do not have a clear explanation for this sex difference. Interestingly, it has been shown that physical inactivity was associated with a BMI increase (1.95 ± 0.3 kg/m²) in homozygous A-allele carriers compared with homozygous T-allele carriers (Andreasen et al. 2008). The FTO genotype has also been shown to interact with weight loss (Delahanty et al. 2012), and the A-allele seems to complicate weight maintenance in severe obese patients (Woehning et al. 2013). Studies also show that the A-allele is associated with increased energy intake (Speakman et al. 2008), but this could not be confirmed in the present study. Alternatively, the FTO gene may be involved in the control of appetite as FTO expression is decreased in the hypothalamus of fasted mice when compared to fed mice (Stratigopoulos et al. 2008). Also, in humans, associations with appetite have been found (Karra et al. 2013). It suggests that genetic influence of FTO on long-term weight gain in men is not only a matter of physical activity or energy intake per se, but could follow another mechanism involving the brain.

In men, genetic interaction was observed between FTO and ACE (Fig. 3). This may point to fat storage/turnover capacity as the major process for long-term weight gain in men. In genetically obese diabetic mice, it was shown that ACE-inhibition blocks PAI-1 activity indicating that ACE is necessary for the formation of active PAI-1 (Zaman et al. 2001). ACE converts angiotensin I into angiotensin II, and in humans, angiotensin II increases the expression of PAI-1, whereas inhibition of ACE decreases PAI-1 in plasma and tissues (Brown et al. 1998, 1999). In the mouse, inhibition of PAI-1 has been associated with reduced weight of fat depots and adipocyte size (Crandall et al. 2006). Influence of ACE on tissue plasminogen activator (tPA) has also been reported, but remains controversial (Brown et al. 1999). Overall, the adipose-specific renin-angiotensin system appears involved in positive regulation of adiposity (Kalupahana and Moustaid-Moussa 2012). Mechanistically, ACE may influence adipocyte ECM dynamics which is an important cellular parameter of weight loss and weight regain. In this respect, we have proposed that the ECM has a lower metabolic plasticity than the cell, leading during weight loss to the accumulation of stress between the shrinking cell and the rigid ECM, which forms a driving force for weight regain because re-storage of fat may alleviate this stress (Mariman 2011, 2012). The genetic interaction between FTO and ACE opens the possibility that also FTO may exert its effect somehow through the ECM dynamics.

Matrix metalloproteinases (MMPs) are essential for proper ECM remodeling (Bouloumie et al. 2001; Chavey et al. 2003). MMP2 is an endopeptidase that degrades the basement membrane surrounding adipocytes and thus may facilitate hypertrophic development of adipocytes and formation of adipocyte clusters (Brown et al. 1997). An induction of the MMP2 transcription in adipose tissue of obese mice and during human adipocyte differentiation has been reported (Chavey et al. 2003; Dubois et al. 2008). In this study, we found that the C-allele of MMP2 in women is associated with weight gain over a 10-year period, which is in line with a study of Han et al. (2008) who showed that the C-allele is associated with obesity. The polymorphism is located in a codon for glycine but is not accompanied by a change of this amino acid. Therefore, the biology behind the observed association remains unknown. Interestingly, a role for estrogens in the expression and activity of MMP2 has been proposed (Marin-Castano et al. 2003; Wingrove et al. 1998), which may explain the fact that the association was found only in women. Maybe, a different plasticity of the adipose tissue through MMP2 under increased energy intake can explain the difference in risk of long-term weight gain in women.

The interaction between genes and environmental factors, including nutrition and life style, can be a contribution to the long-term weight gain. Therefore, in the present study, we have also investigated the interaction of SNPs and macronutrient intake levels, but we have not detected any significant contribution for these SNPs. Our interaction results are in agreement with several reports for FTO and energy intake, fat, carbohydrate and protein (Lappalainen et al. 2012; Liu et al. 2010), but in contrast to other reports (Ahmad et al. 2011; Corella et al. 2011). An overview of interactions between FTO and environmental factors is given in supplementary Table 1. No known interactions of environmental factors with the genes for AKR1C2, MMP2 and TIMP4 have been reported. However, it cannot be excluded that other nutrients may play a role in the interaction with these SNPs in the long-term weight gain.

In a former study with a different cohort, we have investigated other genetic influences on weight gain over a period of almost 7 years (Heidema et al. 2010). CNTF was found associated with weight gain in men, whereas in women, IL6 appeared influencing long-term weight gain. Obviously, together with the present findings, this corroborates the concept that long-term weight regulation is influenced by genetic factors in a sex-specific manner. As a mechanism, the activity of the sympathetic nervous system was proposed, which has a strong effect on the balance between fat storage in and release from adipocytes. Our present results suggest ECM dynamics as a possible mechanistic factor working through this same balance of storage and release of fat in adipocytes.

All taken together, genetic association was found between SNPs in ACE, FTO and AKR1C2 and weight gain over 10 years in men, and between an SNP in MMP2 and weight gain over the same period in women. Our results indicate that the investigated SNPs in FTO and MMP2 contribute to the risk of long-term weight gain in men and women, respectively. Epistasis was observed for the SNPs in FTO and ACE in men. Further research can now be conducted to validate our findings concerning gender-specific genetic influences on long-term weight gain during adulthood.