Genes that make you fat, but keep you healthy

Abstract

Obesity prevalence continues to rise worldwide, posing a substantial burden on people’s health and wellbeing. However, up to 45% of obese individuals do not suffer from cardiometabolic complications, also referred to as the metabolically healthy obese (MHO). Concurrently, up to 30% of normal weight individuals demonstrate cardiometabolic risk factors that are typically only seen in obese individuals; the so called metabolically obese normal weight (MONW). Besides lifestyle factors (physical activity, diet, smoking, …) and demographic (age, sex, ancestry) factors, innate biological mechanisms are known to contribute to the etiology of the MHO and MONW phenotypes, as well. Experimental studies in animal models have shown that adipose tissue and adipocyte biology are key players, and mechanisms such as adipose tissue expandability, fat distribution, adipogenesis, vascularization of adipose tissue, inflammation, and mitochondrial function are the main mechanisms that uncouple adiposity from its cardiometabolic comorbidities. We reviewed the current studies that take advantage of genetic association data to expand insights into the biology of MHO/MONW phenotypes. At least four genetic loci were identified through genome-wide association studies for body fat percentage (BF%) of which the BF%-increasing allele was associated with a protective effect on glycemic and lipid outcomes. For some – but not all – this association was mediated through favorable effect on body fat distribution. Other studies that aimed to characterize the genetic susceptibility of insulin resistance, found that a higher genetic susceptibility was associated with lower overall adiposity due to less fat accumulation at hips and legs, suggesting that an impaired capacity to store fat in adipose tissue may be metabolically harmful. While these genetic association studies have started to provide new insights into the biology of MHO/MONW phenotypes, purposefully designed genome-wide association studies to discover new genes in a hypothesis-free manner have not yet been performed. Clearly, a lot more work remains to be done in this field, first through gene discovery, and subsequently through functional follow-up of identified genes.

Introduction

Over the past four decades, the prevalence of obesity among adults worldwide has nearly tripled--from 4.8% in 1975 to almost 13% (or 650 million people) in 2016 [1–3]. Particularly alarming is the steady increase in middle- and low-income, often densely populated countries that have only been exposed to an increasingly obesogenic environment in recent years, the full impact of which remains to unfold [1, 3].

Obesity is a major risk factor for many metabolic diseases, such as type 2 diabetes (T2D), cardiovascular disease (CVD), dyslipidemia, and hypertension [4–10], and poses an enormous burden on people’s personal health and well-being [10, 11]. However, not all individuals who are obese suffer from metabolic comorbidities. An estimated 15–45% of obese individuals are considered to be metabolically healthy, despite excess adiposity (the so-called “metabolically healthy obese”, MHO) [12–32]. Conversely, not all normal weight individuals are protected from metabolic disease; i.e. an estimated 6–30% of normal weight people demonstrate cardiometabolic abnormalities that are typically seen in obese individuals (the so-called “metabolically obese normal weight”, MONW) [17–34]. These subgroups of obese and normal weight individuals were first described in the early 1980’s [35–37], but in the past decade, research into their etiology, correlates, and future disease risk has increased exponentially.

Here, we review the factors that explain why some obese individuals are metabolically healthy, why some normal weight individuals are at risk, and how this affects their risk of future disease. The main focus of this review is on the genetic etiology; we review the literature on genetic variations that are associated with increased adiposity and that, at the same time, have favorable effects on cardiometabolic outcomes. Through studying the genetic contributors to this apparently paradoxical relationship, we aim to gain insight into the biology that couples (or uncouples) adiposity to cardiometabolic comorbidities.

What is MHO and MONW ?

As noted above, the prevalence of MHO and MONW varies substantially across studies. Primarily, this wide range is due to the lack of a uniform definition to capture these complex conditions. Across different studies, different phenotypes and thresholds have been used to define “obesity” and “metabolic health” [34, 38]. In most studies, “normal weight” and “obese” are defined by using body mass index (BMI) cut offs defined by the WHO (BMI < 25 kg/m² and BMI ≥ 30 kg/m², respectively) [2], but other adiposity traits, such as body fat percentage and waist circumference, are being used, as well [34, 38]. The definition of “metabolic health” is more diversified. Most often, metabolic health is defined by the absence or presence of any of the four risk factors that constitute the metabolic syndrome (high blood pressure, high fasting glucose, triglycerides, and low HDL-cholesterol levels). However, the number of risk factors that have to be absent or present to define a person as metabolically “healthy” or “unhealthy”, respectively, varies across studies [34, 38]. Other markers of metabolic health, such as insulin sensitivity measures (HOMA), markers of inflammation, and even cardiorespiratory fitness, are sometimes included, as well [34, 38].

The phenotypes and criteria used to define MHO and MONW typically depend on the data available and the research questions to be answered. While the absence of a consensus hampers comparison of observations between studies and their generalizability, it is debatable whether a single definition can capture these complex and heterogeneous conditions. It is reasonable to assume that the MHO and MONW subgroups are themselves not homogenous. For example, a commonly used definition for MHO requires individuals to be both obese (BMI ≥ 30 kg/m²) and not have more than one metabolic syndrome risk factor present. As such, obese individuals with no risk factors are grouped with obese individuals who are mostly metabolically healthy, but may just have elevated fasting glucose levels, or elevated triglyceride levels, or low HDL-cholesterol levels. While all of these individuals are considered to be MHO, they clearly constitute one clinically heterogeneous group. In particular, when one aims to identify underlying genes, and thus underlying biology of MHO and MONW, such heterogeneity may make this task almost impossible.

Taken together, while the existence of MHO and MONW individuals is well recognized, the diversity of definitions makes the comparison of findings not straightforward. On the other hand, a single definition to capture these heterogeneous conditions could be a limiting factor when aiming to study the underlying biology that links adiposity and cardiometabolic health.

Disease risk associated with MHO and MONW

The disease risks associated with MHO and MONW vary across studies and across disease outcomes studied. Some studies found that the disease risk for MHO individuals is similar to that of normal weight individuals (at least for cardiovascular disease [15–18], stroke [15], peripheral vascular disease [20], dyslipidemia [39], type 2 diabetes [16]) and mortality [27–29]), even after many years of follow-up. However, cumulating evidence from several large-scale prospective studies suggest that MHO individuals have a higher risk than healthy normal weight individuals for many cardiometabolic outcomes (cardiovascular disease [14, 18–25, 40], cerebrovascular disease [20], hypertension [39], insulin resistance [39], type 2 diabetes [14, 24–26, 41]), yet their risk is lower than that of metabolically unhealthy obese individuals. Other studies suggest that the MHO-state may be transient, and that over time, MHO individuals will transition to metabolically unhealthy obese [14].

Similarly, MONW individuals were found to be at increased risk for future cardiometabolic disease, compared to healthy normal weight individuals, but for most outcomes (cardiovascular disease [14–25, 40], cerebrovascular disease [20], peripheral vascular disease [20], type 2 diabetes [14, 16, 24–26] and mortality [28, 29]), their risk is generally not as high as that of metabolically unhealthy obese individuals.

Differences in definitions, duration of follow-up, sample size, and disease outcomes studied likely contribute to the discrepancies in disease risks reported for MHO and MONW individuals [42, 43]. In general, it seems disease risks of MHO and MONW are intermediate to healthy normal weight and unhealthy obesity individuals.

Factors contributing to MHO and MONW

Lifestyle and demographic factors have been shown to explain at least part of the difference between metabolically healthy and metabolically unhealthy obese individuals. MHO individuals are generally younger, female, more often of African ancestry, have a relatively lower BMI and waist circumference, and tend to live a healthier lifestyle (smoke less, more physically active, healthier diet) compared to their unhealthy obese peers [13–32]. Similar factors partially explain why some normal weight individuals are metabolically unhealthy; i.e. MONW individuals are often older, male, less likely of African ancestry, have a higher BMI and larger waist circumference and live a less healthy lifestyle, compared to healthy normal weight individuals [14–32].

Nevertheless, even after accounting for these demographic and environmental correlates, MHO individuals continue to have a more favorable cardiometabolic risk profile than the unhealthy obese individuals, and MONW individuals continue to be at greater risk than the healthy normal weight individuals [14–28]. This suggests that innate, biological mechanisms also contribute to the cardiometabolic differences among obese and normal weight individuals.

Current insights into the underlying mechanisms

The mechanisms that determine why some obese individuals remain free from cardiometabolic complications, while others do not, are not fully understood. Current insights have come mainly from rodent models, which aim to identify pathways through the manipulation of candidate genes and proteins, such as adiponectin [44, 45], mitochondrial membrane protein (mitoNEET) [46], vascular endothelial growth factors [47–49], glucose transporter type 4 [50], and others [51–53]. From these studies, adipose tissue and adipocyte biology have emerged as key players underlying the MHO and MONW phenotypes. Specifically, adipose tissue expandability and lipotoxicity, fat distribution and ectopic fat accumulation, adipogenesis and lipogenesis, oxygen availability and vascularization of adipose tissue, adipose tissue inflammation, and mitochondrial function have been proposed as the main mechanisms involved [44–57]. These observations from animal studies have been corroborated by findings from studies in humans, which are often small-scale, due to the invasiveness of the methods used [58–68].

These in-depth experimental and physiological studies in animals and humans make it possible to focus on specific pathways or manipulate presumed mechanisms involved, but they are always “hypothesis-driven” – i.e, they build upon a biology that is already partially known. As such, these types of studies may prevent the discovery of totally new mechanisms that were not previously believed to play a role.

Using human genetics to identify new pathways and mechanisms

Complimentary to the hypothesis-driven experimental animal and human studies is the genome-wide association approach, which aims to identify new mechanisms through gene discovery. Genome-wide association studies (GWAS) are hypothesis-free and have the potential to point to new genes that have not been previously implicated in a given disease [69]. As genes translate into proteins and proteins constitute biological pathways, the discovery of genes and genetic variations that influence obesity and metabolic traits can lead to new insights in the biology that underlies MHO and MONW.

Over the past decade, GWAS have led to the identification of thousands of genetic loci robustly associated with a wide range of diseases and traits [70], including >500 genetic loci associated with adiposity traits [71], mainly with BMI (as a proxy of overall obesity) [72–74] and waist-to-hip ratio (WHR; proxy of fat distribution) [75], but also with more refined adiposity traits, such as body fat percentage (BF%) [76, 77], leptin [78], specific fat depots, such as visceral and subcutaneous adipose tissue (VAT, SAT) [79–81], and extreme and early-onset obesity [82–84]. In addition, GWAS have also identified hundreds of loci associated with cardiometabolic traits, including glycemic traits [85, 86], circulating lipid levels [87–89], blood pressure [90, 91], coronary artery disease (CAD) [92], and type 2 diabetes [93, 94].

So far, no GWAS have been performed to specifically search for genetic variants associated with MHO or MONW. This may be due to the absence of a uniform definition, but more importantly, it may be because the MHO and MONW represent heterogeneous sub-groups, which impedes gene discovery [69]. Despite the lack of specific GWAS for MHO or MONW, genetic variants exist that are associated with increased adiposity (e.g. higher BMI), and – at the same time – with a favorable cardiometabolic profile (e.g. lower glucose levels, lower triglyceride and higher HDL-cholesterol levels, lower blood pressure, and/or lower risk of T2D and CAD), and thus mirror the MHO phenotype. Because genetic variants in GWAS are typically bi-allelic (two variations), the alternate allele is associated with decreased adiposity and a poorer cardiometabolic profile, resembling the MONW phenotype. Here, we review the genetic variants identified so far and the insights that they have contributed to our understanding of the biology that either couples or uncouples obesity and cardiometabolic disease.

Genome-wide association studies for adiposity outcomes

Our interest in using genetic association data to identify adiposity-increasing variants that have protective effects on cardiometabolic outcomes was sparked by a GWAS on body fat percentage (N=76,150) that identified three loci that reached genome-wide significance (FTO, near IRS1, near SPRY2) [77]. Interestingly, the BF%-increasing allele (rs2943650, P=3.8×10⁻¹¹) of the IRS1 locus (insulin receptor substrate 1) was significantly associated with a lower cardiometabolic risk profile, including lower risk of T2D and CAD (Table 1), which is consistent with a MHO phenotype. In follow-up analyses, we showed that this – at first unanticipated –association signature was explained by an effect on fat deposition; i.e. the BF%-increasing allele favors fat deposition in subcutaneous, but not in the metabolically harmful visceral adipose tissue. These findings mirror Irs1 knockout mice that are lean but insulin resistant [95, 96], and whose cell lines suggest a role in adipocyte differentiation [97, 98]. The variant (rs2943650) that we identified is not located in IRS1, but at ~500k basepairs upstream of the gene. However, as its BF%-increasing allele is associated with higher IRS1 expression in adipose tissue and muscle, we believe that IRS1 is likely the causal gene is this locus [77]. Together, these data suggest that genetic variation near IRS1 may impair people’s ability to store fat subcutaneously, which may result in adverse cardiometabolic outcomes, such as insulin resistance and dyslipidemia. In contrast to IRS1, the BF%-increasing alleles of the variants in FTO and near SPRY2, the two other loci identified in the GWAS for BF%, were associated with a higher cardiometabolic risk [77].

Table 1.

Association results (direction of effect, P-value) of four loci, identified through genome-wide association studies for body fat percentage, that show a MHO/MONW association signature.

Trait			IRS1 rs2943650		COBLL1/GRB14 rs6738627		PLA2G6 rs3761445		TOMM40 rs6857
	Source (PMID)	Nmax	BMI increasing (C-) allele	P value	BMI increasing (A-) allele	P value	BMI increasing (G-) allele	P value	BMI increasing (C-) allele	P value
BMI	25673413	335,397	increase	7.1 × 10−3	increase	6.1 × 10−4	increase	2.4 × 10−3	increase	10−4
Body Fat%	26833246	99,614	increase	3.8 × 10−11	increase	5.7 × 10−9	increase	1.8 × 10−7	increase	6.8 × 10−9
VAT/SAT	27918534	18,205	decrease	3.5 × 10−3	decrease	0.014	increase	0.37	increase	0.25
VAT (visceral adipose tissue)	27918534	18,312	increase	0.21	increase	0.05	increase	1.6 × 10−3	increase	8.6 × 10−5
SAT (subcunatenous adipose tissue)	27918534	18,206	increase	1.2 × 10−5	increase	1.7× 10−3	increase	0.031	increase	0.014
WHRadjBMI	25673412	221,453	decrease	0.38	decrease	6.7 × 10−9	decrease	0.78	increase	7.8 × 10−5
WHR (waist-to-hip ratio)	25673412	223,839	increase	4.3 × 10−3	decrease	3.6 × 10−5	increase	0.54	increase	1.3 × 10−5
Waist Circumference	25673412	241,458	increase	8.4 × 10−3	increase	0.20	increase	0.012	increase	8.3 × 10−5
Hip Circumference	25673412	224,459	increase	2 × 10−4	increase	1.7 × 10−6	increase	1.1 × 10−3	increase	0.19
Height	25282103	253,006	decrease	0.38	increase	0.72	increase	6.7 × 10−5	decrease	0.17
Triglycerides	24097068	175,846	decrease	2 × 10−5	decrease	3.3 × 10−5	decrease	8.1 × 10−12	decrease	4.6 × 10−19
HDL-Cholesterol	24097068	185,166	increase	1 × 10−7	increase	4.9 × 10−9	increase	3.9 × 10−4	increase	2.6 × 10−17
LDL-Cholesterol	24097068	171,168	decrease	0.045	decrease	0.035	decrease	0.04	decrease	5.1 × 10−112
Fasting glucose	22885924	133,010	decrease	0.11	decrease	0.22	increase	0.93	increase	2.1 × 10−3
Fasting Insulin	22885924	108,557	decrease	1.4 × 10−7	decrease	4.1 × 10−6	decrease	0.86	increase	0.28
Fasting Insulin adjusted for BMI	22885924	108,557	decrease	3.2 × 10−19	decrease	1.7 x 10−12	decrease	10−3	increase	0.35
T2D risk (cases/controls)	24509480	26,488/83,964	decrease	2.7 × 10−14	decrease	5.4 × 10−11	decrease	0.43	increase	2.8 × 10−3
CVD risk (cases/controls)	26343387	60,801/123,504	decrease	1.6 × 10−7	decrease	0.039	increase	0.74	decrease	6.8 × 10−8

Associations that reach nominal significance (P<0.05) are in bold

Associations that are considered favorable are in green, and those that are unfavorable for health are in red.

Associations are shown in function of the BF% increasing allele.

In a larger, subsequent GWAS (N=100,716), we identified 12 loci significantly associated with BF% [76]. Five loci had been previously identified for association with BMI; their effect on BMI was more pronounced than on BF%. Consistent with expectations based on phenotypic correlations between adiposity and cardiometabolic traits, the BMI-increasing alleles of these five loci were associated with a generally poorer glycemic and lipid profile. The other seven loci – four of which were novel – had a larger effect on BF% than on BMI. Interestingly, four of these loci (including IRS1) showed an association signature consistent with a MHO/MONW phenotype; i.e. the BF%-increasing alleles had protective effects on glycemic and lipid traits (Table 1). For example, the BF%-increasing allele of the COBLL1/GRB14 variant (rs6738627) is associated with lower triglyceride levels, higher HDL-cholesterol, and lower fasting insulin levels, as well as a lower risk of T2D. Similar to the locus near IRS1, the protective effect of the COBLL1/GRB14 variants may be mediated through an effect on body fat distribution; i.e. the BF%-increasing allele is associated with increased subcutaneous, but not visceral, adipose tissue. In addition, the BF%-increasing allele is associated with lower BMI-adjusted WHR and larger hip circumference, suggestive of a relatively greater gluteal than abdominal fat deposition. The causal gene in the locus remains to be pinpointed, but current evidence points to GRB14, as the BF%-increasing allele is associated with a lower expression of GRB14 in subcutaneous and omental fat, whereas no association is observed with COBLL1 expression [76]. GRB14 encodes a protein that interacts with insulin receptors and insulin-like growth-factor receptors. Grb14 expression is increased in genetically obese (ob/ob) mice [99] and grb14-deficient mice show improved glucose homeostasis and enhanced insulin action through increased irs1 phosphorylation in muscle and liver [100]. Together, the genetic association data and the animal data suggest that GRB14 and IRS1 act in the same pathway and influence insulin sensitivity through an effect of fat distribution and storage.

The BF%-increasing allele of the PLA2G6 locus (rs3761445) is also associated with lower triglyceride and higher HDL cholesterol levels, lower fasting insulin levels, and a lower risk of T2D (Table 1) – a similar association signature to the near-IRS1 and GRB14 loci [76]. However, the role of body fat distribution as a potential mediator is less evident; the BF%-increasing allele is associated with increased SAT and hip circumference, which is considered to be protective, but the same allele is also associated with increased VAT and waist circumference (Table 1). In addition, the BF%-increasing allele is associated with a significantly greater height. This suggests that the mechanisms that underlie the MHO/MONW phenotypes for this locus, may be different than for the near-IRS1 and GRB14 loci. The lead variant (rs3761445) is located in PLA2G6, a potential candidate gene in this locus, as it encodes a phospholipase A2 enzyme that catalyzes the release of fatty acids from phospholipids.

Another BF% associated locus, nearest to TOMM40, is more complex. The BF%-increasing allele (rs6857) is associated with a significantly healthier lipid profile (lower triglyceride and LDL-cholesterol levels, higher HDL-cholesterol levels) and protects against coronary artery disease, whereas no associations with glycemic traits were observed (Table 1). Interestingly, despite its protective effects on lipid levels, the BF%-increasing allele of rs6857 was associated with a significantly increased WHR, waist (but not hip) circumference and visceral adipose tissue deposition, suggesting a proportionally greater fat accumulation in the abdominal area, which is metabolically more harmful. The locus is located near APOE, in which variants are known to be highly significantly associated with lipid levels. However, follow-up analyses showed that the associations observed for rs6857 are mostly independent from those observed for APOE variants. Similar to the PLA6G2 locus, the BF%-increasing allele of the TOMM40 locus are associated with a proportionally greater VAT than SAT, and a higher WHR, which in contrary to the typical hypothesis that fat distribution mediates the MHO/MONW phenotypes (Table 1). The BF%-increasing alleles of variants highly correlated with rs6857 and thus representing the same locus, have been shown to associate with lower risk of Alzheimer’s disease [101], slower cognitive decline [102], and increased longevity [103]. TOMM40 encodes a protein that is embedded in the outer membranes of mitochondria and is critical for protein transport in mitochondria [104]. BF%-increasing allele of rs6857 is associated with increased expression of TOMM40 in omental and subcutaneous fat [76]. TOMM40 has been studied primarily in the context of Alzheimer’s disease; the precise mechanism through which it may affect adiposity and cardiometabolic disease remains to be elucidated.

Another locus with a MHO/MONW association signature was identified through a GWAS for BMI in a relatively small population of Samoans (N_max=3,072) [105], whose obesity prevalence ranks among the highest in the world. Samoans constitute a founder population with distinct demographic and evolutionary characteristics, which provide unique opportunities for gene discovery. The GWAS identified one genome-wide significant association: the minor allele of a coding variant (rs373863828, p.Arg457Gln) in CREBRF was associated with a 1.4 kg/m² higher BMI (equivalent to ~4 kg/allele for a 1.7m tall person) and with a 1.3-fold increased risk of obesity. Interestingly, the obesity-increasing allele was significantly associated with 1.6-fold lower risk of T2D and lower fasting glucose levels [106]. This obesity-increasing and T2D-decreasing association signature has since been replicated in other Pacific populations [107], with effects on body size seen as early as age 4 [108], and does not seem to be mediated through favorable fat distribution (assessed by WHR) [106]. The BMI-increasing allele is common among Polynesian populations (MAF>10%) [106–109], in particular among Samoans (MAF=26%) [106], and is present among some Melanesian and Micronesian populations (MAF>2%) [109], but is practically non-existent in other populations [110]. CREBRF is widely expressed, including in adipose tissue, and the Arg-to-Gln change at rs373863828 is predicted to have functional implications for the gene/protein [106, 110]. Using a cellular model, it was shown that the Gln-(BMI-increasing) allele promotes lipid storage but reduces energy use in adipocytes. While this explains the potential biology underlying the obesity susceptibility, it does not explain the protective effect on T2D [105]. More work will be needed to further elucidate CREBRF’s paradoxical effect on obesity and T2D susceptibility, and to assess its role in non-Pacific Islanders.

Follow up of genetic variants associated with cardiometabolic traits

The GWAS studies described above aimed – in the first place – to identify loci associated with adiposity traits. The fact that some of these loci have a MHO/MONW association signature was a secondary – and rather serendipitous – finding. In other studies, reviewed here, follow-up analyses were performed on GWAS-identified loci for cardiometabolic traits to gain insight into how they affect health. The first three studies focused on genetic loci identified in GWAS for fasting insulin levels, and a fourth study identified novel loci by combining three GWAS (fasting insulin, HDL-cholesterol, and triglyceride levels). All four studies show that these insulin resistance associated loci – individually or aggregated – are associated with higher cardiometabolic risk, despite being associated with decreased adiposity [111–114].

The first three studies focused on genetic variants identified in a GWAS for insulin levels, as a proxy for insulin resistance [85]. In the first study (N_max=18,565), Scott et al. created a genetic risk score (GRS) that combined 10 variants that, besides being associated with fasting insulin levels, were also associated with lower HDL-cholesterol and higher triglyceride levels [113]. In the second study (N_max = 70,000), Yaghootkar et al. [111] considered all 19 insulin-associated variants identified in the same GWAS and examined their associations with eight adiposity and cardiometabolic traits. Hierarchical clustering of these association results split the 19 variants into two clusters, with the main cluster grouping 11 insulin-associated variants (in/near LYPLAL1, GRB14, IRS1, PPARG, FAM12A1, PDGFC, ARL15, ANKRD55, RSPO3, PEPD, and TET2) [111]). Apart from TET2, the variants in this main cluster fully overlapped with the 10 variants examined by Scott et al. [113]. In the third paper, Yaghootkar et al. replicated their observations for the same 11 variants in 164,609 indviduals of the UK Biobank and 5 additional studies [112]. As the three studies ended up using largely the same variants and similar approaches (i.e. GRS) to characterise the genetic susceptibility to insulin resistance, their overall observations are consistent. All three studies created a GRS that summed the risk alleles of insulin-associated variants that were also known to associate with a poorer lipid profile. As expected, a higher GRS was associated with increased insulin resistance, lower HDL cholesterol and higher triglyceride levels, and increased risk of T2D and CAD. Follow-up analyses, using more refined measures of body composition and fat distribution, showed that the association with lower BMI and BF% was linked to a higher WHR and visceral-to-subcutaneous adipose tissue (VAT/SAT) ratio, which was driven by a smaller hip circumference, less subcutaneous (but not visceral) adipose tissue, and less leg and gynoid (but not trunk or android) fat mass [111–113]. Thus, all three studies show – across different populations – that the lower adiposity observed with increased genetic susceptibility to insulin resistance is due to less fat accumulation at hips and legs. The 11 genetic variants, used to assess the genetic susceptibility to insulin resiatnce, are located in or near genes that may play a role in adipocyte biology (e.g. IRS1, PPARG, LYPLAL1, PEPD, GRB14, COBLL1, ANKRD55, FAM13A), but the mechanisms through which each associate with reduced adiposity and increased insulin resistance may be different and requires further functional follow up.

In the fourth study, using a more exploratory approach to move beyond already established insulin-associated variants, Lotta et al. aligned the summary statistics of three GWAS, for fasting insulin (BMI-adjusted) [86, 113], HDL-cholesterol [87], and triglyceride levels [87] to identify additional variants and to gain insight into the genetic and molecular mechanisms underpinning insulin resistance and adiposity [114]. They identified 53 genetic loci associated with higher insulin, higher triglyceride, and lower HDL-cholesterol levels (at P<0.005 for each trait), which included the 10 loci included in the previous studies, and 43 novel loci. A GRS based on these 53 loci was associated with increased risk of T2D (P<10⁻⁶⁰) and – to a lesser extent – with risk of coronary heart disease (P<10⁻¹³) (Figure 1). Consistent with previous studies, a higher GRS for insulin resistance was associated with lower BMI (P<10⁻⁷) and BF% (P<10⁻¹⁵), but with higher WHR (P<20⁻⁸⁷), which was mainly due to a proportionally smaller hip (P<10⁻³³) than larger waist (P=0.003) circumference (Figure 1). Follow-up analyses using more refined measurements showed that the association with lower adiposity was driven by lower gynoid (P<10⁻¹²) and leg (P<10⁻¹⁵) fat mass (but not lean mass), with no effect on trunk fat mass (P=0.42) [114] (Figure 1). As before, these findings suggest that a genetic susceptibility to insulin resistance – as assessed by these 53 loci – is mediated through an inability to store fat in peripheral adipose depots. Tissue and cell type enrichment analyses showed that genes located in the 53 loci are significantly more often expressed in adipose tissue and adipocytes (Figure 2), and 31 of the 53 lead SNPs (P<10⁻⁴) overlap enhancer elements that are active in adipose tissue [114]. Three candidate genes among the 53 loci (PPARG, PIK3R1 and INSR) are known to cause monogenic forms of lipodystrophy, and another candidate, LPL, is a key lipolysis regulator. Eight of the 53 loci were previously identified – at genome-wide significance – in GWAS for BF% and/or hip circumference (COBLL1/GRB14, FAM13A, ITPR2, KLF14, L3MBTL3, LYPLAL1, TNFAIP8, PIK3R1) and the adiposity-lowering alleles at these loci are associated with increased insulin resistance and T2D risk [114].

Figure 1.

Association between the GRS, based on the insulin-increasing alleles of 53 variants, with adiposity and cardiometabolic outcomes. The per-allele effects for continuous traits are expressed in SD. Adapted from Lotta et al. [114].

Figure 2.

Tissue and cell type enrichment, using DEPICT, based on expression patterns in 37,427 human microarray samples. The y axis represents the −log10 (P value) for enrichment of signal in a cell or tissue type; enrichment is significant at P_Bonferroni ≤ 0.00072. From Lotta et al. [114].

These four studies -- focusing on insulin resistance-associated genetic variants -- suggest that the unexpected association between a lower overall adiposity and higher insulin resistance and risk of T2D is mediated through an unfavorable fat distribution. Specifically, these studies show that less subcutaneous fat storage and/or less fat accumulation at legs and hips is metabolically harmful and that an impaired capacity to expand peripheral fat depots might contribute to increased insulin resistance and T2D risk in the general population – features that are reminiscent of those observed in more extreme monogenic forms of lipodystrophy. The finding that relatively larger hip and thigh circumferences are metabolically protective corroborates observations from general and genetic epidemiological studies [115–122].

While these studies support a role for fat distribution and fat storage as one mechanism underlying insulin resistance, it is important to note that a GRS based on other (or more) insulin resistance-associated variants may not show the same association with lower (gluteofemoral) adiposity. After all, the variants included in the GRS were identified in GWAS for BMI-adjusted insulin levels and were also associated with HDL-cholesterol and triglyceride levels. Therefore, the GRSs may represent only one part of the genetic susceptibility to insulin resistance and T2D risk. Other variants of which the association with increased insulin levels is mediated through an effect on increased BMI (e.g. FTO, YSK4, HIP1) or that are not associated with lipid levels (e.g. UHRF1BP1, IGF1, HIP1), were not included in the GRS.

A limitation of using GRSs is that contributing genetic variants and/or genes are only assessed in aggregate, as a proxy for genetic susceptibility to insulin resistance, whereas the biological candidacy of the individual genes is not considered. Besides examining the GRS, Lotta et al. performed experimental follow up on a handful of genes (IRS1, CCDC92, DNAH10, L3MBTL3, and FAM13A) near four of the 53 variants included in their analyses [114]. They aimed to validated these genes’ putative role in adipogenesis, given that enrichment analyses had implicated adipocytes as a likely effector cell type underlying the observed associations (Figure 3). Except for Fam13a, knockdown of these genes in an adipocyte mouse model resulted in impaired lipid accumulation, consistent with the genetic association data showing that a higher genetic susceptibility to insulin resistance correlates with less peripheral fat and lower expression of these genes in subcutaneous adipocytes [114]. While these experiments in mouse models validate the candidacy of the genes, in-depth functional follow-up is needed to further pinpoint the underlying pathways.

Figure 3.

Summary of insights from current genetic association studies. Adiposity-increasing alleles that have protective effects on cardiometabolic outcomes have a favourable effect on fat distribution; fat accumulates peripherally, at thighs, hips and in subcutaneous adipocytes.

Conclusions

Environmental factors (diet, physical activity, smoking, etc.) and demographic factors (age, sex, ancestry) are important contributors to why some obese individuals are protected from cardiometabolic complications and why some normal weight individuals are at risk. In-depth functional studies in animals and humans support a role for innate biological mechanisms to underlie these paradoxical phenotypes, as well. As such, gene discovery analyses can be used to identify new genes, and thus reveal new pathways and mechanisms, that link increased adiposity with reduced risk of cardiometabolic complications. Nevertheless, insights gained from human genetic studies in the context of the MHO and MONW phenotypes have been limited, thus far.

Genome-wide association studies that aimed to identify variants for BF%, discovered that for some of these variants (in/near IRS1, COBLL1/GRB14, PLA2G6, TOMM40), the BF%-increasing allele had protective effects on cardiometabolic outcomes (Table 1) [76, 77]. These observations were secondary to the original goal of the studies and in-depth experimental follow up is still needed to gain insight into the underpinning biology. Even though, the adiposity-increasing allele for all four loci has protective effects on cardiometabolic outcomes, their association signatures differ and suggest that they each may represent different parts of the biology (Table 1).

Other studies aimed to characterize the genetic susceptibility for insulin resistance, using GRS based on genetic variants associated with BMI-adjusted insulin levels, and found that a higher susceptibility was coupled to lower adiposity [111–114]. Follow-up analyses, using more refined body composition measures, showed that this “paradoxical” observation was primarily due to less peripheral fat storage. This observation is consistent with the expandability hypothesis that suggests an impaired capacity to store fat in (subcutaneous) adipose tissue, can lead to metabolically harmful ectopic fat storage (e.g. in liver, muscle, …) and may result in comorbidities. The role of fat distribution and adipose tissue expandability has been highlighted before by experimental studies in animals and humans before. The value of the genetic association studies is that genes contributing to the GRS may reveal new proteins and pathways that had not been considered previously.

While the current studies provide some new insights to the field, observations are secondary to the main objective, and progress has been generally limited. Currently, what is missing are purposefully designed genome-wide discovery studies that aim to identify loci for MHO and/or MONW. This would allow discovery of genetic loci without any a priori hypothesis and may reveal new biology that (un)couples obesity and cardiometabolic complications.