What this is
- This research investigates the association between genetic variants in the fat mass and obesity-associated (FTO) gene and type 2 diabetes mellitus (T2DM).
- It includes a spatial analysis and meta-analysis of data from over 60,000 T2DM patients and 90,000 controls across various regions.
- The study specifically examines the rs9939609 and rs8050136, finding that their associations with T2DM risk vary by region.
Essence
- rs9939609 and rs8050136 in the FTO gene are associated with increased T2DM risk, particularly in East and South Asia. The associations are region-dependent and highlight the importance of geographic factors in genetic studies.
Key takeaways
- The rs9939609 is associated with a 15% increased risk of T2DM ( = 1.15, 95% CI 1.11â1.19). This association remains significant after adjusting for body mass index (BMI), indicating a consistent risk across different populations.
- The rs8050136 also shows a significant association with T2DM risk ( = 1.14, 95% CI 1.10â1.18). Similar to rs9939609, this association persists after BMI adjustment, emphasizing its relevance in T2DM susceptibility.
- No association was found for rs1421085 and rs17817499 with T2DM, suggesting these SNPs do not contribute to diabetes risk. The limited studies on these variants indicate a need for further research.
Caveats
- The study primarily focuses on Asian populations, with fewer data from European and North American groups, which may limit the generalizability of the findings.
- Few studies examined rs1421085 and rs17817499, potentially biasing the results towards negative findings for these SNPs.
- The analysis did not account for other potential confounders like age, sex, and gene-environment interactions, which could influence T2DM risk.
Definitions
- polymorphism: A genetic variation where a particular gene has multiple forms, which can affect traits or disease susceptibility.
- odds ratio (OR): A measure of association between exposure and an outcome, indicating the odds of the outcome occurring in the exposed group compared to the non-exposed.
AI simplified
1. Introduction
Diabetes is a growing global health problem; more than 300 million people live with diabetes worldwide [1], and the prevalence of diabetes is estimated to rise [2]. Type 2 diabetes mellitus (T2DM) is the most common type of diabetes, as it accounts for more than 90% of diabetes cases [3]. Although the pathogenesis mechanisms of T2DM have not been clearly defined, a combination of genetic and environmental factors is believed to lead to the disease [4].
The fat mass and obesity-associated (FTO) gene is located on chromosome 16 (16 q12. 2), containing nine exons and several single-nucleotide polymorphisms (SNPs) [5]. In 2007, a genome-wide association study (GWAS) searching for type 2 diabetes-susceptibility genes confirmed a common variant (rs9939609) in the FTO gene that predisposes European populations to diabetes [6]. Since then, a large number of studies have focused on the association between FTO polymorphisms, expression and T2DM in different populations [7,8,9,10]. Meanwhile, some meta-analyses have been performed to elucidate the relationship between FTO polymorphisms and T2DM risk. For instance, a meta-analysis utilizing data from studies prior to 2010 identified an association between rs9939609 and T2DM in East and South Asians [11]. Additionally, a Norwegian population-based Nord-TrÞndelag Health Study (HUNT study) [12], including three cohorts (HUNT, Malmö Diet and Cancer (MDC) and Malmö Preventive Project (MPP)), reported strong association between rs9939609 and T2DM risk in Scandinavians after adjustment for age, sex and body mass index (BMI). Another meta-analysis of association between obesity/BMI-associated loci and T2DM risk [13], using data from studies conducted between 2007 and 2012, revealed that FTO rs9939609 significantly associated with T2DM which also remained significant following adjustment for BMI; Analysis by Vasan et al. [14] has provided evidence that rs9939609 is associated with obesity and T2DM in Asian Indians, with modest attenuation observed when adjusting for BMI. These and the majority of other previous meta-analyses have focused on single population or one FTO loci without consideration of population-specific environmental influences among different regional subgroups. As such, the results of these meta-analyses cannot be generalized to the world.
More recently, geographic information systems (GIS) and spatial analysis are increasingly applied in the investigation of disease spatial pattern, including diabetes [15].
To more comprehensively clarify the association between FTO polymorphisms and T2DM risk, we performed this spatial analysis and meta-analysis to include most, if not all, eligible studies published before January 2017.
2. Materials and Methods
2.1. Search Strategy
Eligible articles were selected by searching up to January 2017 in PubMed and EMBASE using the following keywords: âFTO or fat mass and obesity-associated geneâ and âvariant or variation or polymorphismâ and âtype 2 diabetes or type 2 diabetes mellitus or T2D or T2DMâ. Articles obtained from the initial search were then screened based on the inclusion criteria described below. Only publications with English language were included. If more than one population was included in a given article, results were considered as separate studies.
2.2. Study Selection Criteria and Data Extraction
The selected studies met all of the following inclusion criteria. The studies had to: (1) evaluate the association between FTO polymorphisms and T2DM risk; (2) have a caseâcontrol or cohort design; and (3) provide odds ratios (OR) with a 95% confidence interval (CI) or sufficient data for calculation. From each study, the following information was collected: (1) name of the first author; (2) year of publication; (3) country of origin; (4) ethnicity of the samples; (5) sample size of cases and controls; (6) HardyâWeinberg equilibrium (HWE) in control groups; and (7) data of SNPs. Data were independently extracted from eligible articles by two authors (YY and HYL) according to the criteria described. Discrepancies were resolved by discussion with a third reviewer (SML), and a consensus approach was used.
2.3. Spatial Analysis
The ArcGIS v10.3 software is a GIS tool that has become increasingly prevalent in public health research to understand the spatial pattern of diseases and genetic biodiversity [15]. This software was utilized to depict the geographic distribution of the association studies. R was used to calculate Moranâs I, a statistic for evaluating the spatial autocorrelation [16,17]. By constructing the spatial weight matrix, Moranâs I coefficient can be calculated as follows:I=Nâiâjwijâiâjwij(XiâXÂŻ)(XjâXÂŻ)âi(XiâXÂŻ)2
N is the number of spatial units indexed by i and j; X is the variable of interest; XÂŻ is the mean of X; and wij is an element of a matrix of spatial weights. In this study, we constructed the spatial weight matrix by making a distance threshold h. If the distance between point i and point j is smaller than h, wij will be 1. Otherwise, wij will be 0. It is worth noting that all diagonal elements of matrix w are all 0. Monte Carlo simulations were used to test for the significance of Moranâs I.
2.4. Statistical Analysis
The strength of association between FTO SNPs and T2DM risk was expressed as a pooled OR and 95% CI. A z-test was performed to evaluate the significance of the pooled OR (p < 0.05 was considered statistically significant). The Ï2-test-based Q test and I2 were performed to assess the heterogeneity of the studies. A value of I2(%) > 50% or p †0.10 indicated significant heterogeneity. A random-effects model (DerSimonianâLaird method) [18] was used to determine the pooled OR in the presence of heterogeneity; otherwise a fixed-effects model (MantelâHaenszel method) [19] was used. Subgroup analyses were performed by region. Sensitivity analyses were performed to assess the stability of the combined results by excluding the studies with unknown HWE in controls. Publication bias was evaluated by Beggâs test [20] and Eggerâs test [21] (p < 0.05 was considered statistically significant). Data analyses were conducted using STATA 12.0 (Stata-Corp LP, College Station, TX, USA).
3. Results
3.1. Study Characteristics and Quality
A total of 202 potentially relevant papers were identified from PubMed and EMBASE. After reading the title and abstract, 148 articles were excluded because they addressed topics that did not match the inclusion criteria. The full texts of the remaining 54 articles were carefully screened. We excluded five meta-analyses or reviews, three articles that explored the association between FTO polymorphisms and gestational diabetes, two articles that did not include the full text, and three papers with insufficient data. In total, 41 articles met the inclusion criteria. A flow chart describing the article selection for our meta-analysis is shown in Figure 1. Of the articles included, 29 studies investigated rs9939609, 26 studies explored rs8050136, four studies investigated rs1421085 and three studies explored rs17817499. Other SNPs that were assessed in only one study were not analyzed. The detailed characteristics of the included studies are shown in Table 1.
3.2. Region-Related Associations Exist between rs8050136, rs9939609 and T2DM
For rs8050136, a total of 33,889 T2DM cases and 45,490 controls were included in the final data analysis. The overall results showed a significant association between rs8050136 and T2DM risk (OR = 1.14, 95% CI 1.10â1.18, p (z-test) < 0.001, I2 = 37.4%) (Table 2, Figure 2a), with the association remaining statistically significant after adjustment for BMI (OR = 1.08, 95% CI 1.03â1.12, p (z-test) < 0.001, I2 =27.1%) (Table 2, Figure 2b). To more clearly understand the association between rs8050136 and T2DM in different regions, we performed the subgroup analyses by region. Consequently, without BMI adjustment, a significant association between rs8050136 and T2DM was uncovered in East Asia (OR = 1.15, 95% CI 1.10â1.20), West Asia (OR = 1.17, 95% CI 1.05â1.29) and Europe (OR = 1.19, 95% CI 1.14â1.25) (Table 2, Figure 3a), with no such association in North America (OR = 1.06, 95% CI 0.93â1.19) or South Asia (OR = 1.19, 95% CI 0.91â1.48). After adjustment for BMI, significant association was only observed in East Asia (OR = 1.13, 95% CI 1.05â1.20) (Table 2, Figure 3b). More importantly, as seen in Figure 4, the majority of studies on rs8050136 were distributed in East Asia. Several other studies were scattered throughout Europe, Northern America, South Asia and West Asia. More data for these regions may be required to detect an association.
For rs9939609, a total of 32,771 T2DM cases and 50,161 controls were included in the meta-analysis. The overall results indicated that rs9939609 was significantly associated with an increased risk of T2DM (OR = 1.15, 95% CI 1.11â1.19, p (z-test) < 0.001, I2 = 53.2%) (Table 2, Figure S1a). After adjustment for BMI, the association remained statistically significant (OR = 1.11, 95% CI 1.05â1.17, p (z-test) < 0.001, I2 = 56.1%) (Table 2, Figure S1b). Due to the heterogeneity that existed between studies, we performed stratified analyses grouped by region. In the subgroup analyses, similar results were found in East Asia (without BMI adjustment: OR = 1.11, 95% CI 1.05â1.17; with BMI adjustment: OR = 1.11, 95% CI 1.02â1.20) and South Asia (without BMI adjustment: OR = 1.19, 95% CI 1.10â1.29; with BMI adjustment: OR = 1.19, 95% CI 1.06â1.31), whereas no such association existed between rs9939609 and T2DM in North America (without BMI adjustment: OR = 1.11, 95% CI 0.89â1.32; with BMI adjustment: OR = 1.02, 95% CI 0.81â1.22) (Table 2, Figure S2). Additionally, in Europe, a significant association between rs9939609 and T2DM was observed without BMI adjustment (OR = 1.18, 95% CI 1.14â1.22), whereas no association was uncovered with BMI adjustment (OR = 1.11, 95% CI 0.93â1.29). Similar to the distributions of rs8050136 studies, the geographic distribution of researches on rs9939609 were concentrated in East Asia and South Asia, where the association was found to be significant.
As illustrated in Figure 5, when the spatial scale was smaller than 1,000,000 meters, there was significant positive spatial autocorrelation in terms of both rs9969309 and rs8050136. It turned out that in relative small spatial scale (h < 1,000,000 meters), the studies with significant correlations tended to be clustered, which indicated that the correlation between rs9969309 and rs8050136, and T2DM risk was strongly associated with the geographic factors. With the h increasing, Moranâs I showed no positive spatial autocorrelation of these two SNPs and T2DM risk, which meant we cannot reject the null hypothesis of completed spatial randomness. Our results follow Toblerâs first law of geography: âEverything is related to everything else, but near things are more related than distant thingsâ (pp.236, [56]). It seemed that in Asia, there was a strong positive-positive (significant-significant) spatial autocorrelation while in Europe there may be some negative-negative (non-significant-non-significant) spatial autocorrelation. In North America, the spatial autocorrelation was not significant, maintaining a relatively random spatial pattern.
For rs1421085 and rs17817499, a total of 4,285 T2DM cases with 16,279 controls and 2,634 T2DM cases with 15,482 controls, respectively, were identified for data analysis. The results indicated that neither rs1421085 nor rs17817499 were associated with T2DM, independent of BMI adjustment (Table 2, Figures S3 and S4). Compared with rs9939609 and rs8050136, studies that focused on rs1421085 and rs17817499 were relatively fewer and were distributed in North America and North Africa.
3.3. Sensitivity Analyses
To assess the stability of the combined results obtained by excluding studies of unknown HWE in controls [7,25], a sensitivity analysis was conducted (Figure S5). The analysis confirmed that the rs9939609 polymorphism conferred a predisposition to T2DM.
3.4. Assessment of Publication Bias
To evaluate the publication bias, we performed Beggâs test and Eggerâs test. The results showed that there was no publication bias for the associations between the four FTO polymorphisms and T2DM risk (p > 0.05 for Beggâs test and Eggerâs test) (Table S1).
4. Discussion
Our meta-analysis and spatial analysis are based on a large sample size, including over 60,000 and 90,000 subjects for rs9939609 and rs8050136, respectively, spanning regions across Asia, Europe and Northern America. In line with previous meta-analyses of Asian populations [14,36,45], we further demonstrated a strong association between rs9939609 and rs8050136, and T2DM regardless of adjustment for BMI (Table 2, Figure 2 and Figure 3, Figures S1 and S2). Notably, the associations are region-related.
Indeed, some statistics such as Moranâs I [16,17], and local indicators of spatial autocorrelation (LISA) [57] can be used to quantitatively study spatial autocorrelation. However, due to obstacles including the modifiable areal unit problem (MAUP) (i.e., some papers only provide a country location while some papers have the city location) and the low data volume, it is difficult to perform spatial statistics for rs1421085 and rs17817499 to further explore the spatial pattern. Nevertheless, our data still indicate the geographic factor may play an important role in the correlations between T2DM risk and rs8050136 (Figure 4 and Figure 5), rs9939609 (Figure 5).
Initially, the articles we reviewed contained more than 10 types of FTO SNPs in T2DM patients and controls, but we eventually chose the four most common SNPs, namely rs9939609, rs8050136, rs1421085 and rs17817499. All four SNPs are located in intron 1 of the FTO gene, a region of strong linkage disequilibrium [40]. Some studies have found no direct connection between the variants and FTO expression or function [9], while other studies have suggested that variants of FTO play an important role in regulating body weight and fat mass by influencing food intake [6]. A recent report revealed that SNPs in FTO could influence obesity by altering the expression of the adjacent genes IRX3 and RPGRIP1L [58]. Although mechanisms regarding how these noncoding variants affect T2DM are not yet clear, Smemo et al. have demonstrated that variants within FTO can form long-range functional connections with IRX3, representing a determinant of body mass and composition [59]. Additionally, recent studies have suggested hepatic FTO contributes to glucose homeostasis [60,61,62], indicating that FTO may play a role in the regulation of carbohydrate metabolism.
Of note, the overall heterogeneity of rs9939609 increased slightly after BMI adjustment (I2 = 53.2%, p < 0.001 without BMI adjustment vs. I2 = 56.1%, p = 0.003 with BMI adjustment) (Table 2), suggesting that BMI may not primarily account for heterogeneity. To this end, we performed additional subgroup analyses by region and found that heterogeneity still existed in the group of North America and South Asia independent of BMI adjustment. We then excluded each study in South Asia and North America and performed subgroup analyses, respectively. When omitting studies by Fawwad et al. or Chauhan et al. in South Asia, as well as Bressler et al. (African-Americans) in North America [24,32,34], the heterogeneity disappeared in the South Asian (I2 = 34.6%, p = 0.141 and I2 = 37.2%, p = 0.121) and North American (I2 = 0.0%, p = 0.667) subgroups, respectively, without BMI adjustment (Table S2). Of note, the heterogeneity showed no change by removing other studies in South Asian or North American subgroup. Alternatively, only removing the study by Ali et al. [27], heterogeneity in the South Asian subgroup also attenuated sharply (I2 = 20.3%, p = 0.288) after adjustment for BMI (Table S2). These results demonstrated that these studies mentioned above were the main source of heterogeneity in South Asia and North America. Unlike rs9939609, owing to the low data volume of the studies, the heterogeneity in rs1421085 and rs17817499 showed no change by subgroup analyses.
BMI is widely considered as a confounder of T2DM risk. In this study, the overall associations between the four SNPs and T2DM risk were not affected by BMI adjustment. (Table 2), indicating that the overall associations were BMI-independent. Nevertheless, in Europe for rs9939609 and West Asia for rs8050136, the BMI adjustment altered the associations (Table 2). In agreement with previous reports [11,12], our data showed that rs9939609 was also associated with T2DM risk somewhat independently of BMI in East and South Asia as well as in Europe. Interestingly, different regions showed different associations between rs9939609 and rs8050136, and T2DM risk, demonstrating that the associations were region-dependent. Generally, a race/ethnicity population might live in the same region in most of the non-immigrant countries. Thus, our results might reflect the influence of different races/ethnicities to some extent.
The rs9939609 was the first SNP discovered within the FTO gene that showed a strong association with BMI and as such is the most widely investigated SNP of FTO [63]. Additionally, the A allele of rs9939609 is known to indicate a predisposition to obesity, T2DM, polycystic ovary syndrome (PCOS) and some cancers [41,64,65]. Our results of rs9939609 are not only consistent with earlier reports [11,12,13,14], but also include more recent studies with greater geographical coverage [7,8,9,22,23,34] (Table 2, Figure S2), providing stronger evidence for these associations. Similarly, rs8050136 was also found to function as a susceptible SNP to rs9939609-related diseases. Unlike rs9939609 and rs8050136, studies on rs1421085 and rs17817499 are scarce, and have limited regional coverage; lack of association maybe due to smaller sample size and less studies involved.
The study we present here still possesses several limitations. First, a large proportion of the studies focused on Asian populations, with European and Northern American populations only accounting for a small part. Second, there were relatively few studies on rs1421085 and rs17817499, which may lead to bias in negative results (Table 2, Figures S3 and S4). Lastly, except for BMI, we used genotype data without considering other possible confounders (such as age and sex) or gene-gene and geneâenvironment interactions. Although BMI is widely used to measure obesity, it has been suggested that different criteria (not necessarily > 30) may be used in different ethnic populations. Adiposity (or specific distribution of fat) rather than body weight (or BMI) may play a critical role in the regulation of insulin sensitivity and the development of diabetes. This may lead to an inconsistency in the effect of BMI on the association between FTO variants and T2DM risk. Therefore, further studies that adjust for more concomitant factors and cover more regions should be conducted.
5. Conclusions
The spatial analysis and meta-analysis showed that the associations between genetic polymorphisms in FTO and T2DM are region-related and that shedding light on spatial variations can provide new insights into well-established relationships. The rs9939609 and rs8050136 SNPs contributed to an increased risk of T2DM, which could provide new solutions for T2DM prevention and therapy. This study presented an initial step in spatial analysis for genetic and regional factors in the development of diabetes, although more work remains to be done before we can understand the impact of genetics, environment, geography, BMI and fat distribution on diabetes as well as how these associations may vary across space.