Insights into the Genetic Susceptibility to Type 2 Diabetes from Genome-Wide Association Studies of Obesity-Related Traits

Obesity and type 2 diabetes (T2D) are common and complex metabolic diseases, which are caused by an interchange between environmental and genetic factors. Recently, a number of large-scale genome-wide association studies (GWAS) have improved our knowledge of the genetic architecture and biological mechanisms of these diseases. Currently, more than ~250 genetic loci have been found for monogenic, syndromic, or common forms of T2D and/or obesity-related traits. In this review, we discuss the implications of these GWAS for obesity and T2D, and investigate the overlap of loci for obesity-related traits and T2D, highlighting potential mechanisms that affect T2D susceptibility.


Introduction
Type 2 diabetes (T2D) is a common metabolic disease of increased plasma glucose levels to which individuals are predisposed to by a combination of genes and environmental factors. The hyperglycemia typically results from decreased insulin sensitivity (insulin resistance) in insulindependent tissues (such as skeletal muscle, liver and adipose tissues), which leads to hyperinsulinemia. Subsequently, when the pancreatic beta cells are not capable of producing the amount of insulin required to maintain normal glycemic status, which may be caused by beta-cell dysfunction and/or reduced beta-cell mass, chronic hyperglycemia and T2D occur (reviewed in [1]).
The prevalence of obesity and T2D is currently escalating worldwide as a consequence of a sedentary lifestyle and increased consumption of high-energy content food [2]. Between 1980 and 2014, the worldwide prevalence of obesity more than doubled. In 2014, 11 % of men (>205 million) and 15 % of women (>297 million) in the world were obese, compared with 5 % for men and 8 % for women in 1980 This article is part of the Topical Collection on Genetics [16]. The overall prevalence of obesity is at least four times higher in high-income countries compared to that in lowincome countries. A similar accompanying increase in the prevalence of T2D is seen as obesity is a risk factor for T2D [17,18]. In 2014, it was estimated that there are 387 million people living with diabetes (ages 20-79) with a worldwide prevalence of 8.3 %, and~90 % of these are individuals with T2D. By 2035, this number is expected to increase by 205 million. It is estimated that 77 % of people with diabetes live in low-and middle-income countries [19].
Genetic, but also environmental, factors interact to cause both obesity and T2D as shown by familial aggregation [20][21][22][23], family and twin studies on obesity (heritability (h 2 )~40-70 %) [22,24,25] and T2D (h 2~2 6-69 %) [21,26]. Beyond a sedentary lifestyle, socioeconomic status, poor nutrition, infections and differences in the gut flora have also been added to the list of potential environmental triggers of obesity and T2D [27]. Genetic and environmental evidence is also provided by numerous animal studies. Rodent models for T2D, such as the Lep ob and Zucker mice strains rely on the mutations in genes encoding leptin or its receptor to develop T2D via obesity (reviewed in [28]). Evidence of both environmental and genetic effects in an animal model has been shown to exist in the Agouti A vy mouse, where the obesity phenotype is inherited through an epigenetic effect that is dependent on the maternal diet [29].
Early evidence for the genetic effect in obesity and diabetes was found through linkage studies of monogenic forms of these diseases segregating as Mendelian disorders, in which mutations occurring in a gene lead to extreme and early-onset forms of these conditions. For obesity, these include genes functioning in the leptin-melanocortin pathway, such as the leptin (LEP) and melanocortin 4 receptor (MC4R) genes (reviewed in [30,31]). Similarly, monogenic forms of diabetes are caused by mutations in genes such as GCK, HNF4A and HNF1A with allelic series causing maturity onset of the young (MODY) (reviewed in [32]). Linkage studies have been subsequently accompanied by larger and statistically more powerful genome-wide association studies (GWAS) that are designed to dissect the genetic architecture of common complex traits in a hypothesis-free way [33]. GWAS are useful for identifying common genetic variants (i.e. single-nucleotide polymorphisms (SNPs) with a minor allele frequency (MAF) >5 %) that affect a trait outcome or increase the risk of a disease of interest by comparing frequencies of alleles in thousands of individuals, or between cases and healthy controls, respectively. Many variants associated with complex traits and diseases have been discovered so far through the GWAS approach. SNPs that reach genome-wide significance (p<5×10 −8 after correction for multiple-testing, 0.05/1,000, 000 independent tests among common variants in the human genome) are specifically targeted for replication and further functional experiments [33]. These associations are important for unraveling biological mechanisms and pathways that might lend themselves to informing about new therapeutic targets.
In this article, we review the current GWAS of obesityrelated traits and consider the overlap with T2D-associated loci in order to gain insights into the genetic susceptibility and potential mechanisms that lead to increased risk of T2D.
Since the heritability of BMI is 7 % higher at younger ages and increases with the mean age in childhood studies (+1.2 % per year) [73], GWAS of children, adolescents and young adults have been carried out identifying three loci associated with childhood BMI [66,[74][75][76]. To date,~2.7 % of the phenotypic variation in BMI was explained by the 97 associations in populations of European and non-European ancestry. Furthermore, common genetic variation (MAF>5 %) accounted for~21 % of BMI variation.

Overlap Between GWAS of Obesity-Related Traits and T2D
Body Mass Index The first overall obesity GWAS [54] identified a robust association between BMI and SNPs in the first intron of the fat mass and obesity-associated (FTO) gene that has been widely replicated since [70,72,[77][78][79][80]. FTO variants had previously been associated with T2D (p<5×10 −8 ), but this association disappeared after adjusting for BMI, which showed that FTO association with T2D is largely due its effect on BMI [54]. In line with this, the effect of FTO variants on 23 cardiometabolic traits, in addition to T2D, is mainly mediated through BMI [81]. The FTO locus is not only strongly associated with T2D risk [50••] and higher BMI [52] but also increased fasting insulin and homeostatic model estimated insulin resistance (HOMA-IR; p=9.5×10 −5 ), which is in agreement with insulin resistance playing a part in the FTO association with T2D via increased BMI [52]. The FTO protein has been characterized as a 2-oxoglutarate-and Fe(II)-dependent demethylase, possibly involved in mRNA modification, and it is highly expressed in the brain [82,83]. However, a recent study suggested that the obesity-associated FTO variants affect expression of IRX3, but not FTO, in the human brain, which may mean that FTO is not the causal gene in this region. Functional experiments also supported this finding; body weight of Irx3-deficient mice was reduced by 25-30 %, and hypothalamic expression of a dominant-negative form of Irx3 resulted in the same metabolic phenotype as the Irx3-deficient mice [84]. The precise biological role of the BMI-associated variants at the FTO locus is still unclear and remains to be disentangled.
Previous studies have shown that mutations causing MC4R inactivation lead to severe and monogenic forms of obesity [31]. Low frequency variants in MC4R were identified in morbid obese individuals (BMI>40 kg/m 2 ) and were associated with obesity [88,89]. MC4R is a neural G-protein-coupled melanocortin receptor that is highly expressed in the brain [90]. It plays an important role in the regulation of energy balance, specifically in the regulation of energy intake via the control of satiety and energy expenditure (reviewed in [91]). The MC4R locus was associated with both T2D (rs12970134, odds ratio (OR)=1.08, 95 % confidence interval (CI)=1.03-1.12, European p=0.0002, trans-ethnic p=2.6×  (Fig. 2). In addition, another SNP (rs571312) at the same locus, in strong linkage disequilibrium with rs12970134 (r 2 =0.87, D′=0.96, HapMap2, Utah Residents with European ancestry population (CEU)), was associated with increased fasting insulin (p=5.2×10 −5 ), HOMA-IR (p=7.6×10 −5 ) and T2D (p=0.0004), which is in agreement with insulin resistance playing a part in the MC4R association with T2D through BMI [52]. However, in an exome sequencing study of 6760 Pima Indians, mutations decreasing MC4R activity were detected and these individuals with MC4R defects had increased T2D risk, partially independent of BMI in childhood (BMI-adjusted hazard rate ratio=3.3, 95 % CI= 1.2-9.2, p=0.03). This effect might be due to an increased rate of weight gain compared to adulthood and MC4R affecting downstream insulin signaling [92,93]. Nevertheless, the effect of MC4R variants on T2D risk was completely attributable to BMI in adulthood [94]. Thus, taking into account the changing physiology and hormonal levels during different stages of life would be valuable when considering the biology behind traits and diseases such as BMI and T2D.
The genetic association between T2D and variants in transcription factor 7-like 2 (TCF7L2) was first discovered in a diabetes, adiposity, waist, fat, body mass index, non-alcoholic liver, adiponectin, weight and adipose. P value threshold for association was p<5×10 −8 . Associations are labeled with corresponding trait colors candidate gene study [95]. This association was later detected in GWAS, and it is the strongest known association with the disease to date among common variants (rs7903146, OR=1.  (Fig. 2). It is interesting that the T allele of rs7903146 increases T2D risk while decreasing BMI, opposing the idea that increased BMI leads to insulin resistance and T2D. In comparison to FTO and MC4R variants, TCF7L2 variants have a much larger effect on T2D risk and a smaller effect on BMI, which might indicate that the TCF7L2 variants act via T2D to affect BMI (Fig. 2). TCF7L2 is a transcription factor functioning in WNT signaling, which is crucial for cell proliferation, motility, normal embryogenesis, and regulation of myogenesis and adipogenesis (reviewed in [96]). Although the causal variant is still unclear, the T2D risk allele appears to act via lowering the levels of insulin secretion and influencing beta-cell function (reviewed in [51,96,97]).

GWAS of Fat Percent
In a meta-analysis of 15 GWAS with 36,626 individuals of European and Indian Asian descent informative for fat% (as measured by BI and/or DXA), three loci near FTO, SPRY2 and IRS1 were identified [3]. All of these loci were previously associated with T2D [42,98]. The fat%-decreasing allele of rs2943650 near IRS1 was associated with increased risk of T2D as that allele decreased subcutaneous fat but not visceral fat, which is more health damaging (reviewed in [99]). The T2D risk allele of another IRS1 variant, rs2943640 (r 2 =0.97, D′=1.00, HapMap2, CEU), was also nominally associated with decreased BMI (beta=−0.014, p=1.1×10 −5 ; Fig. 2) [58••]. Furthermore, another variant (rs2943634), strongly correlated with the T2D and fat%-associated rs2943650 (r 2 = 0.80, D′=0.96, HapMap2, CEU), was associated with fasting insulin levels (beta=0.025, p=2.5×10 −14 ) [34,100,101]. Insulin receptor substrate 1 encoded by IRS1 is an important member of the insulin signaling cascade functioning as a docking protein and activating downstream signaling when phosphorylated by the insulin receptor [102]. Given the essential function of IRS1 in insulin signaling and the association of IRS1 variants with T2D as well as fat% and BMI, this gene is likely to be involved in fat distribution, adipocyte biology and/ or insulin resistance [98].

GWAS of Extreme/Early-Onset Obesity
In the polygenic form of extreme/early-onset obesity, mutations in more than one gene may play a role in susceptibility. Individuals with extreme/early-onset obesity are likely to be enriched for genetic variants predisposing the general population to obesity. Out of five GWAS, only three studies identified novel loci that were not discovered by the previous GWAS of BMI [64,65,68]. Except for FTO and MC4R variants, which affect T2D susceptibility through their effect on BMI, none of these loci overlap with the known T2D loci (Fig. 1)

Waist Circumference and Waist-to-Hip Ratio
WHRadjBMI is a measure of fat distribution that indicates the amount of metabolically adverse visceral fat [61], while taking into account the metabolically protective role of gluteal fat [103,104]. Both WC and WHR are associated with T2D risk independently of BMI [17,18], and are correlated with the gold standard MRI measurement of central adiposity (i.e. visceral fat, r 2 =0.6 and r 2 =0.5, respectively). However, when targeting genetic associations independent of BMI, WHRadjBMI is a better measure of central fat distribution given the strong correlation between WC and BMI (WC-BMI r 2 =0.9, WHR-BMI r 2 =0.6) [59].
A number of variants strongly associated (p<5×10 −8 ) with T2D risk exhibit opposite directions of effect on BMI and WHRadjBMI (Figs. 2 and 3). For instance, while the T2D risk allele in GCKR is associated with increased BMI (rs780094, C allele, beta=0.01, p=0.0002) [58••] (Fig. 2), the same variant has an opposite effect on WHRadjBMI (rs780094, beta= −0.01, p=0.004) [62••] (Fig. 3). Interestingly, sexual dimorphism was also observed in WHRadjBMI, with a statistically significant (p <0.05) GCKR association only in women (rs780094, beta=−0.015, p=0.001). Glucokinase regulatory protein (GCKR) regulates glucokinase (GCK), which is a crucial enzyme for glucose metabolism in the liver and glucosestimulated insulin secretion from pancreatic beta cells. It was previously observed that over-expression of GCKR in the liver significantly improved insulin sensitivity and glucose tolerance in mice resulting in decreased leptin and increased triglyceride levels [105]. This finding may provide a possible explanation for the observed genetic association; the effect of GCKR variants may act through leptin to increase BMI, while independently affecting central fat distribution.
GRB14 is an interesting example of a gene with T2D risk alleles causing increased WHRadjBMI and decreased BMI (Figs. 2 and 3). Associations of T2D risk alleles with increased fasting insulin and HOMA-IR implicate GRB14 variants playing a role in insulin resistance [109]. In rodents and humans, expression of GRB14 in adipose tissue was negatively correlated with insulin sensitivity. In addition, prolonged fasting and metformin treatment in mice significantly decreased Grb14 expression in periepididymal adipose tissue [110]. Furthermore, improved glucose homeostasis and enhanced insulin signaling were observed in Grb14-deficient mice [111]. These findings provide evidence for the importance of GRB14 regulation in insulin resistance and show that complete understanding of its regulation is essential for identification of new therapeutic pathways in obesity and T2D [112].

GWAS of Other Obesity-Related Traits: Abdominal Subcutaneous and Visceral Adipose Tissue, Non-alcoholic Fatty Liver Disease and Pericardial Fat
Similar to WHRadjBMI associations, sexual dimorphism was observed in genetic associations with SAT and VAT, highlighting the importance of physiological and hormonal difference in susceptibility to obesity and T2D in men and women [4]. In a GWAS of NAFLD, associations with five loci were identified [15]. One of these loci, the previously reported WHRadjBMI locus at LYPLAL1, was also associated with VAT/SAT ratio in women (rs4846567, p=0.0004) [4] and T2D (rs2820446, p=2.3×10 −6 ; rs4846567-rs2820446 r 2 = 1.00, D′=1.00, HapMap2, CEU) [50••]. Women are known to have more subcutaneous fat but less visceral fat compared to men [5]. Given the protective role of subcutaneous fat in T2D susceptibility, it is plausible to observe more men with T2D. Globally, the prevalence of T2D is higher in men, but the reasons for this observation may not be limited to the amount of subcutaneous fat in men [116]. In a GWAS of pericardial fat, only one locus (TRIB2) reached genome-wide significance, but this locus is also devoid from associations with T2D and other obesity traits (Fig. 1). This lack of overlap between loci associated with T2D, the more commonly used obesity trait measures (BMI, WHRadjBMI, etc.) and the other obesity-related traits such as pericardial fat may indicate that there is a different genetic architecture for pericardial fat and potentially for other ectopic fat depots. Anthropometric and more specifically measured traits might be more distinct than p<0.05 (green) and p≥0.05 (blue). Red, orange and yellow associations are labeled with corresponding gene names the close relationships between these phenotypes indicate, or these observations most likely reflect that there is a power difference in detection of loci between these studies [14].

Conclusions
T2D loci appear to affect susceptibility to T2D via two main mechanisms: (1) through insulin resistance, i.e. insulin sensitivity (measured by fasting insulin and HOMA-IR) and/or (2) through a beta-cell dysfunction (measured by fasting glucose and homeostatic model estimated beta-cell function). In addition, these loci, in general, also exhibit overlapping associations with obesity-related traits and blood lipid levels (HDL, LDL, triglycerides), which might explain the phenotypic overlap with obesity and cardiovascular diseases. However, these associations are often heterogeneous and variants may have opposite directions of effect for different obesity-related traits, reflecting the intricate biology behind them. For instance, most T2D risk alleles seem to be associated with a decrease in BMI, except for the variants in FTO, MC4R and GCKR, two of which are known to affect T2D susceptibility through BMI [50••]. In contrast, most T2D risk variants are associated with increased WHRadjBMI (Figs. 2 and 3). This observation might indicate distinct mechanisms by which (1) WHRadjBMI-and BMI-increasing alleles act on T2D risk, and (2) T2D risk alleles act on BMI. Increased BMI and central adiposity (defined by increased WHRadjBMI) are known to be health damaging and raising T2D risk via insulin resistance. However, there seems to be a second mechanism where risk alleles (e.g. TCF7L2 variants) predominantly act via T2D and decrease BMI, not vice versa. More targeted genetic and functional studies are necessary to explore these mechanisms and biological pathways implicated (reviewed in [130,131]).
The heritability of obesity and T2D is not entirely explained by all the loci discovered so far [ More studies with larger sample sizes, in different ethnicities, employing various approaches such as rare variant analysis, exome sequencing, studies of epigenetics and geneenvironment interactions are necessary to help explain the missing heritability. Identifying actual functional variants may also increase the phenotypic variance explained by the known loci. Identification of novel loci and functional variants is also required to gain a better understanding of the genetic architecture of body shape, fat depots and T2D. Discovery of additional overlapping genetic associations could provide important insights into the role played by obesity in susceptibility to T2D.
Beyond filling out the gaps in the heritability estimates, deciphering biological mechanisms and pathways that mediate effects leading to susceptibility to obesity and T2D is essential for development of new therapeutic strategies, including lifestyle changes. It is noticeable that genes within loci that are BMI-and WHRadjBMI-associated display different gene expression patterns; they have higher expression levels in the hypothalamus and adipose/peripheral tissues, respectively [52,61]. These initial observations were further supported by the evidence from Data-driven Expression Prioritized Integration for Complex Traits (DEPICT) analyses in the recent GIANT BMI and WHRadjBMI meta-analyses [58••, 62••]. For WHRadjBMI, significant pathways and gene sets included adiponectin signaling, insulin sensitivity and regulation of glucose levels, skeletal growth, transcriptional regulation and those functioning in the development of metabolically active tissues such as adipose, liver and muscle [62••]. In contrast, highlighted pathways and gene sets for BMI included those functioning in the central nervous system involved in synaptic function, long-term potentiation and neurotransmitter signaling [58••].
Monogenic obesity genes in the leptin-melanocortin pathway provide the link between adipose tissue and the hypothalamus, which are critical sites for balancing the energy need of the body. Genes functioning in the leptin-melanocortin pathway such as those encoding leptin (LEP), leptin receptor ( LEPR ), melanocortin 4 receptor (MC4R), proopiomelanocortin (POMC) and brain-derived neurotrophic factor (BDNF) have been implicated in the monogenic form of obesity (reviewed in [30]). Leptin is a hormone produced by adipocytes that play an important role in food intake and weight regulation. Increased leptin signaling in the hypothalamus leads to decreased food intake via MC4R and POMCderived peptide alpha-melanocyte stimulating hormone (alpha-MSH). Many of the monogenic obesity genes lie within loci that are also associated with T2D [50••]. The overlap between monogenic obesity genes and obesity genes identified via GWAS (e.g. MC4R and BDNF) might imply a role of hypothalamic dysfunction affecting the regulation of energy balance in polygenic obesity, which can drive T2D.
Gender-specific effects are observed for anthropometric traits, particularly for waist-related phenotypes, and understanding their biological influences is crucial [61,108]. For instance, variants in and around PPARG have been associated with T2D, monogenic obesity and WHRadjBMI. Of these, the WHRadjBMI association exhibited sexual dimorphism with a significantly stronger effect in women (beta women =0.035, beta men =0.005) [62••]. In parallel with that, gender differences were detected in response to PPARG-agonist therapy in patients with T2D which might indicate different mechanisms for insulin resistance in men and women [132]. Even though biological functions of associated loci are not clear for many genes, gender-specific effects are detected during/after puberty and are potentially attributable to sex hormones [133]. In addition, distribution of body fat also affects metabolic pathways, and body fat has an endocrinological role producing hormones such as estrogen, progesterone, leptin and adiponectin, which affect the regulation of energy balance in the hypothalamus and insulin sensitivity [134]. The understanding of sexual dimorphisms is likely to improve exploration of metabolic disease processes and design of better therapeutic approaches.
In summary, the recent GWAS of obesity-related traits and T2D show considerable overlap in associated loci. These identified associations point to potential mechanisms through which obesity traits affect T2D susceptibility.