Diabetes Volume 70, August 2021 1843

Genetic Evidence for Different Adiposity Phenotypes and Their Opposing Influences on Ectopic Fat and Risk of Cardiometabolic Disease

Susan Martin,1 Madeleine Cule,2 Nicolas Basty,3 Jessica Tyrrell,1 Robin N. Beaumont,1 Andrew R. Wood,1 Timothy M. Frayling,1 Elena Sorokin,2 Brandon Whitcher,3 Yi Liu,2 Jimmy D. Bell,3 E. Louise Thomas,3 and Hanieh Yaghootkar1,3

Diabetes 2021;70:1843–1856 | https://doi.org/10.2337/db21-0129 GENETICS/GENOMES/PROTEOMICS/METABOLOMICS

To understand the causal role of adiposity and ectopic a cluster of events often referred to as the metabolic syn- fat in type 2 diabetes and cardiometabolic diseases, we drome(1).However,inthegeneralpopulation,15–40% of aimed to identify two clusters of adiposity genetic var- individuals categorized as obese do not present any obesity- iants: one with “adverse” metabolic effects (UFA) and related metabolic conditions or diseases and are the other with, paradoxically, “favorable” metabolic “metabolically benign” at the specific time point of measure- effects (FA). We performed a multivariate genome-wide ment, supporting the existence of metabolically benign obe- association study using body fat percentage and meta- sity (2,3). fi bolic biomarkers from UK Biobank and identi ed 38 Previously we showed that a genetic predisposition to UFA and 36 FA variants. Adiposity-increasing alleles storing excess fat in subcutaneous (SAT) is fi were associated with an adverse metabolic pro le, associated with a reduced propensity to store fat in the higher risk of disease, higher CRP, and higher fat in sub- liver, consequently reducing risk of disease (4). The iden- cutaneous and visceral adipose tissue, liver, and pan- tification of “favorable adiposity” variants, with their adi- creas for UFA and a favorable metabolic profile, lower posity-increasing alleles paradoxically associated with risk of disease, higher CRP and higher subcutaneous lower risk of type 2 diabetes, heart disease, and hyperten- adipose tissue but lower liver fat for FA. We detected no – sexual dimorphism. The Mendelian randomization stud- sion (4 7), provided genetic evidence for the paradox of fi ies provided evidence for a risk-increasing effect of UFA metabolically benign obesity. These genetic ndings sug- and protective effect of FA for type 2 diabetes, heart dis- gest that there are at least two types of variants associ- ease, hypertension, stroke, nonalcoholic fatty liver dis- ated with higher adiposity: one with favorable metabolic fi ease, and polycystic ovary syndrome. FA is distinct pro le (favorable adiposity [FA]) and the other with an from UFA by its association with lower liver fat and pro- unfavorable metabolic profile (unfavorable adiposity tection from cardiometabolic diseases; it was not asso- [UFA]). ciated with visceral or pancreatic fat. Understanding Although our previous studies suggested an important the difference in FA and UFA may lead to new insights in role for liver fat, we have been unable to determine preventing, predicting, and treating cardiometabolic whether pancreatic fat deposition or liver and pancreas vol- diseases. umes were similarly implicated due to lack of data, and it has not been possible to investigate mechanisms imposed by each variant individually. Clarification of the underlying Obesity is a significant risk factor for various conditions pathophysiologic mechanisms that link adiposity to higher including type 2 diabetes, heart disease, and hypertension— risk of type 2 diabetes and other cardiometabolic disease is

1Genetics of Complex Traits, University of Exeter Medical School, University of This article contains supplementary material online at https://doi.org/10.2337/ Exeter, Royal Devon & Exeter Hospital, Exeter, U.K. figshare.14555463. 2Calico Life Sciences LLC, South San Francisco, CA S.M., M.C., E.L.T., and H.Y. contributed equally. 3Research Centre for Optimal Health, School of Life Sciences, University of © Westminster, London, U.K. 2021 by the American Diabetes Association. Readers may use this article as long as the work is properly cited, the use is educational and not for Corresponding author: Hanieh Yaghootkar, [email protected] profit, and the work is not altered. More information is available at https:// Received 11 February 2021 and accepted 6 May 2021 www.diabetesjournals.org/content/license. 1844 Types of Adiposity in Cardiometabolic Disease Diabetes Volume 70, August 2021

À critical to understanding disease progression and remis- We identified 254 variants at P < 5 Â 10 8 associated sion, especially given the rising prevalence of obesity and with both our univariate GWAS of body fat percentage the rapid rise of type 2 diabetes in an aging population. (n = 620 variants previously published [4]) and our com- The availability of both metabolic markers and MRI scan posite metabolic phenotype as estimated by the above data in the UK Biobank (8) has enabled us to test in more multivariate GWAS model. This represents an increase in detail the characteristics of adiposity variants and the role 221 signals compared with the 33 previously reported of ectopic fat in disease mechanism. using a similar approach (4). This increase was largely In this study, we focused on how higher adiposity is attributable to the availability of the metabolic bio- associated with ectopic fat, metabolic derangements, and markers in 451,099 individuals of European ancestry cardiometabolic risk. Specifically, we aimed to 1)identify from UK Biobank, whereas previous studies were limited distinct clusters of FA and UFA variants, 2) investigate the to smaller separate data (e.g., 100,000 with HDL and tri- relation between FA and UFA variants and ectopic fat glycerides, 21,800 with SHBG, and 55,500 with ALT). deposition in the liver and pancreas, 3)examinehowFA and UFA variants are associated with circulating markers Step 2: Classification of Adiposity Variants of inflammation, 4) determine whether sexual dimorphism We applied a k-means algorithm on the 254 variants and is a factor in the association between the clusters and met- their effects on the values of the six phenotypes from the fi abolic biomarkers, fat distribution, and disease risk; and 5) rst step and used the parameter k = 3 to group them use Mendelian randomization (MR) to determine the into FA and UFA. We considered a third cluster of “ fl ” potential causal role of “favorable” and “unfavorable” adi- con icting to group any variants that do not belong to posity in different components of metabolic syndrome. the FA or UFA clusters and did not pursue these variants in the rest of the analyses to minimize false discovery. Within UFA and FA clusters, we inspected whether the RESEARCH DESIGN AND METHODS loci are driven by colocalization of signals from a combi- Discovery Data Set—UK Biobank nation of traits or represent a strong univariate signal. UK Biobank recruited >500,000 individuals aged 37–73 years (99.5% were between 40 and 69 years of age) Step 3: Validation of FA and UFA Variants between 2006 and 2010 from across the U.K. (8) To validate FA and UFA variants, we assessed their effects (Supplementary Table 1). The UK Biobank has approval on risk of type 2 diabetes using data from GWAS of 31 from the North West Multicenter Research Ethics Com- studies, excluding UK Biobank, which included 55,005 mittee (https://www.ukbiobank.ac.uk/ethics/), and these case and 400,308 control subjects of European ancestry ethics regulations cover the work in this study. Written (12). We expected to observe adiposity-increasing alleles informed consent was obtained from all participants. as associated with lower risk of type 2 diabetes for FA The steps performed to identify variants associated variants and higher risk of type 2 diabetes for UFA with adiposity but with different effects on metabolic variants. traits are outlined in Supplementary Fig. 1 and, briefly, are as follows. Imaging Study A subcohort of 100,000 subjects were selected for the imag- Step 1: Genetic Variants Associated With Both Body Fat ing enhancement of the UK Biobank, currently at 49,938. Percentage and Composite Metabolic Biomarkers Abdominal MRI scans were obtained with a MAGNETOM We performed a multivariate genome-wide association Aera 1.5T scanner (software version syngo MR D13) (Sie- study (GWAS) of relevant metabolic biomarkers that were mens Healthineers, Erlangen, Germany) (13). Image-derived available in individuals of European ancestry from the UK phenotypes were generated from the three-dimensional Biobank, including HDL cholesterol (HDL) (n 5 392,965), Dixon neck-to-knee acquisition, the high-resolution T1- sex hormone–binding globulin (SHBG) (n = 389,354), tri- weighted three-dimensional pancreas acquisition, and liver glycerides (n 5 429,011), AST (n 5 427,778), and ALT and pancreas single-slice multiecho acquisitions. Images for (n 5 429,203), using BOLT-LMM v2.3.4 (9) and metaCCA this study were obtained through UK Biobank application software (10) as described previously (4). Specifically, no. 44584. Following automated preprocessing of the differ- metaCCA uses canonical correlation analysis to identify ent sequences, volumes of organs of interest (including the the maximal correlation coefficient between genome-wide liver, pancreas, and SAT and visceral adipose tissue [VAT]) genetic variants and a linear combination of the above were segmented using convolutional neural networks (14). phenotypes, based on the computed phenotype-pheno- Fat content of the liver and pancreas was obtained from the type Pearson correlation matrix. We chose these specific multiecho acquisitions after preprocessing where the proton metabolic biomarkers to be consistent with our previous density fat fraction was estimated (15). approach (4). These biomarkers are used to discriminate between three monogenic forms of insulin resistance: lip- GWAS of Imaging-Derived Phenotypes odystrophy (disorders of fat storage), monogenic obesity, We used the UK Biobank Imputed Genotypes v3 (16), and insulin signaling defects (6,11). excluding single nucleotide polymorphisms with minor diabetes.diabetesjournals.org Martin and Associates 1845 allele frequency <1% and imputation quality <0.9. We performed MR within each data source and then meta-a- excluded participants not recorded as European, exhibit- nalyzed the results across the two data sets using a ran- ing sex chromosome aneuploidy, with a discrepancy dom-effects model with the R package metafor (21). We between genetic and self-reported sex, heterozygosity out- ran the same models in UK Biobank data for comparison. liers, and genotype call rate outliers. We used BOLT-LMM For more information on definition of diseases and ICD (9) v2.3.2 to conduct the genetic association study. We codes, please see Supplementary Material. included age at imaging visit, age squared, sex, imaging center, and genotyping batch as fixed-effects covariates, Tissue Enrichment Analyses in addition to scan date scaled and scan time scaled and We used DEPICT (Data-Driven Expression-Prioritized genetic relatedness derived from genotyped single nucleo- Integration for Complex Traits) v.1 rel194 b (22) to iden- tide polymorphisms as a random effect to control for pop- tify tissues and cells in which the from associated ulation structure and relatedness. We performed GWAS loci are highly expressed. Using 37,427 human Affymetrix for VAT (n = 32,859), SAT (n = 32,859), VAT-to-SAT ratio HGU133a2.0 platform microarrays, DEPICT assesses (n = 32,859), pancreatic fat (n = 24,673), liver fat (n = whether genes at the relevant loci are highly expressed in 32,655), pancreas volume (n = 31,758) and liver volume any of the 209 tissues, cell types, and physiological sys- (n = 32,859) (Supplementary Table 1). tems annotated by Medical Subject Headings (MeSH).

Replication Data Set Data and Resource Availability To replicate the effect of FA and UFA variants on measures The data sets analyzed during the current study are avail- of adiposity, metabolic biomarkers, and C-reactive able from FinnGen (20) and the relevant published GWAS (CRP), we used summary statistics from published GWAS, (Supplementary Table 1). The data that support the find- which were independent of UK Biobank (Supplementary ings of this study are available from UK Biobank, but Table 1). To replicate the effects on subcutaneous and restrictions apply to the availability of these data, which ectopic fat, we used data from a combined multiethnic were used under license for the current study (UK Bio- sample size–weighted fixed-effects meta-analysis of SAT bank project application nos. 9072, 9055, and 44584) and (n 5 18,247), VAT (n 5 18,332), VAT-to-SAT ratio (n 5 therefore are not publicly available. No applicable resour- 18,191), and pericardial adipose tissue (n 5 12,204) measured ces were generated or analyzed during the current study. by computed tomography (CT) or MRI (17) (Supplementary Table 1). RESULTS Clusters of Adiposity Variants À Genetic Score Analysis Among 254 variants associated (at P < 5 Â 10 8)with We studied the association of individual variants and of both body fat percentage (4) and a composite metabolic genetic scores with cardiometabolic traits and diseases in phenotype, we identified two distinct clusters of adiposity the UK Biobank using our GWAS results. We performed variants: 1) 38 variants grouped as UFA and 2) 36 var- GWAS with BOLT-LMM to account for population struc- iants grouped as FA (Supplementary Tables 2–4and ture and relatedness using covariates such as age, sex, Supplementary Fig. 2). UFA genetic score was associated genotyping platform, and study center in the model. For with higher body fat percentage and higher BMI and an genetic score analysis, we used the inverse variance– adverse metabolic profile including lower HDL and SHBG weighted method (IVW), assigning a weight of 1 to each and higher triglycerides, ALT, and AST. FA genetic score variant. This method approximates the association of an was also associated with higher body fat percentage and unweighted genetic score (18). BMI but, in contrast, a favorable metabolic profile includ- ing higher HDL and SHBG and lower triglycerides, ALT, MR Studies and AST (Table 1 and Fig. 1A). There was no sex differ- We investigated the causal associations between FA and ence in the association of FA and UFA genetic scores with UFA using FA and UFA clusters as instruments and six adiposity measures or biomarkers at the multiple testing– cardiometabolic disease outcomes (type 2 diabetes, heart corrected significance threshold (0.05 of 44 tests = disease, hypertension, stroke, nonalcoholic fatty liver dis- 0.0011) (Supplementary Table 5 and Supplementary Fig. ease [NAFLD], and polycystic ovary syndrome [PCOS]) by 3). The association between UFA adiposity-increasing performing two-sample MR analysis (19). We used IVW alleles and an adverse metabolic profile and FA adiposi- as our main analysis and MR-Egger and weighted median ty-increasing alleles and a favorable metabolic profile was as sensitivity analyses in order to detect unidentified plei- replicated in independent published GWAS of these bio- otropy of our genetic instruments. We used two sources markers (Supplementary Table 6). of data: FinnGen GWAS summary results (20) and pub- The mean (SD) UFA and FA genetic scores in the UK lished GWAS of the same diseases, excluding UK Biobank Biobank were 36.99 (3.83) and 37.32 (3.75) respectively. to separate it from our discovery data set and allow us to The distributions of UFA and FA genetic scores among run two-sample MR (Supplementary Table 1). We the UK Biobank participants with and without type 2 1846 Types of Adiposity in Cardiometabolic Disease Diabetes Volume 70, August 2021

Table 1—Sex-combined effects of FA and UFA genetic scores on BMI, biomarkers, MRI-derived measures of fat distribution, and cardiometabolic diseases UFA FA Outcome Effect (95% CI) P Effect (95% CI) P BMI (SD) 1.27 (1.01, 1.53) 4EÀ22 0.67 (0.50, 0.84) 3EÀ15 HDL (SD) À0.64 (À0.89, À0.40) 2EÀ7 1.26 (0.96, 1.56) 3EÀ16

SHBG (SD) À0.47 (À0.62, À0.32) 1EÀ9 1.10 (0.44, 1.76) 0.001

Triglycerides (SD) 0.49 (0.34, 0.64) 7EÀ11 À1.54 (-1.92, À1.15) 4EÀ15

ALT (SD) 0.45 (0.36, 0.54) 4EÀ21 À0.90 (-1.12, À0.69) 6EÀ17

AST (SD) 0.35 (0.19, 0.50) 1EÀ5 À0.54 (-0.71, À0.36) 2EÀ9

CRP (SD) 0.52 (0.38, 0.65) 3EÀ14 0.30 (0.09, 0.51) 0.005

SAT (SD) 1.09 (0.87, 1.31) 1EÀ22 0.96 (0.74, 1.18) 1EÀ17

VAT (SD) 0.56 (0.41, 0.72) 2EÀ12 À0.02 (-0.35, 0.31) 0.92

VAT-to-SAT ratio (SD) À0.34 (À0.50, À0.18) 3EÀ5 À0.85 (-1.16, À0.53) 1EÀ7

Liver fat (SD) 0.46 (0.30, 0.63) 2EÀ8 À0.72 (-1.01, À0.43) 9EÀ7

Pancreas fat (SD) 0.52 (0.36, 0.69) 7EÀ10 0.16 (-0.20, 0.51) 0.38

Liver volume (SD) 0.64 (0.44, 0.85) 3EÀ10 À0.43 (À0.73, À0.13) 0.006

Pancreas volume (SD) 0.06 (À0.15, 0.28) 0.57 À0.46 (À0.76, À0.15) 0.003

Type 2 diabetes (OR) 1.06 (1.04, 1.07) 6EÀ16 0.93 (0.91, 0.96) 2EÀ9

Heart disease (OR) 1.05 (1.03, 1.07) 2EÀ7 0.94 (0.92, 0.97) 3EÀ5

Hypertension (OR) 1.13 (1.08, 1.18) 4EÀ8 0.90 (0.85, 0.95) 0.0001

Stroke (OR) 1.01 (1.01, 1.02) 0.0005 0.99 (0.98, 1.00) 0.17

Fatty liver disease (OR) 1.01 (1.00, 1.01) 0.004 0.99 (0.98, 0.999) 0.03

PCOS (OR) 1.01 (1.00, 1.01) 7EÀ5 0.995 (0.99, 0.999) 0.02

Effects are shown per 1-SD-higher body fat percentage as estimated by “favorable adiposity” and “unfavorable adiposity” genetic scores using data from UK Biobank. ASAT, abdominal SAT; VATSAT, VAT-to-SAT ratio.

diabetes are shown in Supplementary Fig. 4. The UFA and false discovery rate <5%) in adipocyte-related cells and FA variants explained 0.6% and 0.2% variance in body fat tissues and in physiological systems labeled as “digestive” percentage in the UK Biobank, respectively. (small intestine, esophagus, pancreas, upper gastrointesti- We used data from the latest GWAS of type 2 diabetes nal tract, and ileum) and “cardiovascular” (arteries), while excluding UK Biobank (12) to validate the paradoxical UFA variants were enriched in mesenchymal stem cells association between the adiposity-increasing alleles at FA and in physiological systems labeled as cardiovascular and UFA variants and risk of type 2 diabetes. Among 38 (aortic valve, heart valves) (Fig. 3 and Supplementary UFA variants, 33 adiposity-increasing alleles were corre- Tables 7 and 8). lated with higher risk of type 2 diabetes (Ptwo-tailed binomial 5 4EÀ6), with 24 at P < 0.05. Among 36 FA variants, all Association With MRI-Derived Measures of Abdominal adiposity-increasing alleles were correlated with lower risk Fat Distribution À of type 2 diabetes (Ptwo-tailed binomial =3E 11), including To investigate the relation between UFA and FA variants 23 variants at P < 0.05 (Fig. 2). This paradoxical associa- and abdominal fat distribution, we looked at the effect of tion was consistent with the pattern of association with FA and UFA variants on SAT, VAT, and ectopic fat in the type 2 diabetes with use of data from the UK Biobank liver and pancreas in addition to liver and pancreas vol- (Supplementary Fig. 5). ume in 32,859 individuals of European ancestry from the To explore whether the UFA and FA variants represent UK Biobank. While both UFA and FA genetic scores were biologically meaningful entities, we searched whether associated with higher SAT with similar effect size, UFA genes at the relevant loci were enriched for expression in score was associated with higher liver and pancreatic fat, certain tissues or pathways. FA variants were enriched (at higher VAT, and increased liver volume, but FA score was diabetes.diabetesjournals.org Martin and Associates 1847

Figure 1—The sex-combined and sex-specific effects and 95% CIs per 1-SD-higher body fat percentage as estimated by FA and UFA genetic scores for measures of adiposity and biomarkers (A), MRI-derived measures of fat distribution (B), and cardiometabolic diseases in UK Biobank (C). ASAT, abdominal SAT; CRP, C-reactive protein; GRS, genetic risk score; VATSAT, VAT-to-SAT ratio.

associated with lower liver fat and smaller liver and pan- CT or MRI in up to 18,332 individuals for some of the mea- creas volume and had no effect on pancreatic fat (Table 1 sured phenotypes. UFA genetic score was associated with and Fig. 1B). Both UFA and FA genetic scores were associ- higher SAT (P =3EÀ8), higher VAT (P =6EÀ7), and higher ated with lower VAT-to-SAT ratio. pericardial adipose tissue (P = 0.003) but had no effect on We replicated these associations using independent data VAT-to-SAT ratio (P = 0.70), while FA genetic score was asso- from the published GWAS of abdominal fat as measured by ciated with higher SAT (P = 6E-6) and lower VAT-to-SAT ratio 1848 Types of Adiposity in Cardiometabolic Disease Diabetes Volume 70, August 2021

Figure 1—Continued.

(P =2EÀ6) and had no effect on VAT (P = 0.92) or pericar- rs4684847 (PPARG), rs12130231 (LYPLAL1/SLC30A10), dial adipose tissue (P = 0.50) (Supplementary Table 6). rs11664106 (EMILIN2), rs13389219 (GRB14/COBLL1), There was no sex difference in the association between rs2943653 (NYAP2/IRS1), rs30351 (ANKRD55), rs4450871 FA and UFA clusters and MRI-derived measures of fat dis- (CYTL1), and rs7133378 (DNAH10). Among these variants, tribution at the multiple testing–corrected significance the FA alleles at only two variants were associated with threshold except for VAT, where FA score was associated pancreatic fat at P < 0.05: near GRB14,withlowerpancre- with higher VAT in men versus lower VAT in women atic fat, and near PPARG, with higher pancreatic fat. (Pdifference = 0.0006 [Supplementary Table 5 and Supplementary Fig. 3]). Association With CRP Levels With the data from the UK Biobank MRI subcohort, To understand the role of inflammation in the mecha- among 38 UFA variants, adiposity-increasing alleles at 31 nisms that link higher adiposity to risk of cardiometabolic variants (Ptwo-tailed binomial = 0.0001) were correlated with disease, we looked at the association between FA and higher ectopic liver fat, including 7 variants with P < 0.05 UFA variants and CRP levels as an inflammatory marker. (Supplementary Fig. 6), and 31 adiposity-increasing alleles In the UK Biobank, both UFA and FA genetic scores were were correlated with higher pancreatic fat (Ptwo-tailed binomial associated with higher CRP (Table 1 and Fig. 1A). These = 0.0001), including 7 variants at P < 0.05 associations were replicated using an independent GWAS (Supplementary Fig. 7). Of the 36 FA variants, 29 adiposi- of CRP (23) (Supplementary Table 6). There was no sex ty-increasing alleles were correlated with lower liver fat difference in the association between UFA and FA var- (Ptwo-tailed binomial = 0.0003), including 9 variants at P < iants and CRP levels (Supplementary Table 5 and 0.05 (Supplementary Fig. 6). FA variants had a mixed effect Supplementary Fig. 3). Of 38 UFA adiposity-increasing on pancreatic fat, as only 14 adiposity-increasing alleles alleles, 35 (Ptwo-tailed binomial =7EÀ8) and, of 36 FA adipo- were correlated with lower pancreatic fat (Ptwo-tailed binomial sity-increasing alleles, 27 (Ptwo-tailed binomial = 0.004) were = 0.24), including two alleles associated with higher and correlated with higher CRP, including 32 and 15 variants two with lower pancreatic fat at P < 0.05 (Supplementary with P < 0.05, respectively (Supplementary Fig. 9). To Fig. 7). further understand the role of higher adiposity on the Results on interesting individual FA variants with para- association between UFA and FA genetic scores and doxical effects where adiposity-increasing alleles are asso- higher CRP, we ran our statistical models, but we addi- ciated with lower risk of type 2 diabetes (from UK tionally adjusted for BMI or body fat percentage. This Biobank–independent GWAS) and lower liver fat (from adjustment removed the association with higher CRP for the UK Biobank) at P < 0.05 are illustrated as forest plots both genetic scores, indicating that the effect was medi- in Supplementary Fig. 8. These include eight variants: ated by higher adiposity (Supplementary Table 9). diabetes.diabetesjournals.org Martin and Associates 1849

Figure 2—Adiposity-increasing alleles were correlated with lower risk of type 2 diabetes for all 36 FA variants, and 33 adiposity-increasing alleles of 38 UFA variants were correlated with higher risk of type 2 diabetes. Effects on x-axis are from the GWAS of body fat percentage in UK Biobank and on y-axis from the GWAS of type 2 diabetes published by Mahajan et al. (12) excluding data from UK Biobank. OR, odds ratio.

Association With Cardiometabolic Disease Risk 13.90]; P =8EÀ9) (Table 2, Fig. 4, and Supplementary For UK Biobank data, UFA genetic score was associated Table 10). In contrast, a 1-SD-higher genetically instru- with higher risk of type 2 diabetes (P =6EÀ16), heart dis- mented FA was associated with lower risk of type 2 diabe- ease (P =2EÀ7), hypertension (P =4EÀ8), stroke (P = tes (0.11 [0.08, 0.16]; P =4EÀ33), heart disease (0.34 0.0005), fatty liver disease (P = 0.004), and PCOS (P = [0.25, 0.47]; P =2EÀ11), hypertension (0.34 [0.21, 0.55]; 7EÀ5) (Table 1 and Fig. 1C). In contrast, FA genetic score P =1EÀ5), stroke (0.65 [0.52, 0.83]; P =0.0004),and was associated with lower risk of type 2 diabetes (P =2E- NAFLD (0.14 [0.03, 0.79]; P = 0.03). There was a trend 9), heart disease (P =3EÀ5), hypertension (P = 0.0001), toward an association with lower odds of PCOS (0.51 fatty liver disease (P = 0.03), and PCOS (P = 0.02) (Table [0.21, 1.23]; P = 0.13) but with wider CIs due to smaller 1 and Fig. 1C). These findings were replicated with use of sample size in the UK Biobank–independent GWAS (Table UK Biobank–independent GWAS data (Supplementary 2, Fig. 4, and Supplementary Table 10). Table 6). There was no sex difference in the association of UFA and FA clusters with risk of disease at the multiple DISCUSSION testing–corrected significance threshold (Supplementary In this study, we used a unique genetic approach to Table 5 and Supplementary Fig. 3). understand the role of body adiposity in relation to the To understand the causal nature of these associations, fat content and volumes of the liver and pancreas, as well we took an MR approach and used two UK Biobank–inde- as pathogenicity of cardiometabolic disease. We have pendent data sets (published GWAS and FinnGen). A 1- identified two distinct clusters of variants associated with SD-higher genetically instrumented UFA was associated higher adiposity: one with a favorable metabolic profile with higher risk of type 2 diabetes (IVW odds ratio 5.50 (FA), consisting of 36 variants, and the other with an [95% CI 4.29, 7.05]; P =4EÀ41), heart disease (1.66 unfavorable metabolic profile (UFA), which included 38 [1.08, 2.54]; P = 0.02), hypertension (3.03 [2.18, 4.22]; P variants. Although the adiposity-increasing alleles in both =5EÀ11), stroke (1.43 [1.23, 1.67]; P =3EÀ6), NAFLD clusters are associated with increased SAT, the FA alleles (3.70 [1.22, 11.17]; P = 0.02), and PCOS (7.13 [3.66, are specifically associated with a lower liver fat and appear 1850 Types of Adiposity in Cardiometabolic Disease Diabetes Volume 70, August 2021

Figure 3—Tissue-specific expression for UFA and FA variants with DEPICT. Results with false discovery rate <0.05 are highlighted in black. Results are grouped by type and ordered by -log10(P-value) within cell types (A), tissues (B), and specific systems (C) (details in Supplementary Tables 7 and 8).

to provide protection against risk of cardiometabolic dis- different obesity phenotypes related to the distribution of eases, whereas the UFA alleles are associated with higher body fat (24,25). The two adiposity phenotypes that we deposition of all fat depots including liver, pancreas, and have described in the current study highlight the role of visceral fat and are associated with higher risk of cardio- SAT as a metabolic sink in obesity. In FA, this metabolic metabolic disease. sink can accommodate excess triglycerides to specifically The results of our genetic analysis support the observa- protect the liver from ectopic fat accumulation and pre- tions from phenotyping studies that have proposed vent or delay pathogenic processes; in UFA, the excess diabetes.diabetesjournals.org Martin and Associates 1851

Figure 3—Continued.

triglycerides appear to exceed the capacity of the SAT bigger liver size is most likely biased by the accumulation metabolic sink and are consequently being deposited in of liver fat (29). alternate sites, including the VAT depot, liver, pancreas, The pattern of association between FA and UFA var- and pericardial adipose tissue (26). iants and MRI-derived measures of ectopic fat can help Our data, consistent with previous findings, provide with understanding the role of each ectopic fat depot in evidence that accumulation of fat in the liver, which is an the pathophysiology of cardiometabolic diseases. While organ integral to glucose, insulin, and , the FA cluster is associated with lower ectopic liver fat, it directly contributes to the development of metabolic has no effect on VAT. This is consistent with previous derangements associated with higher adiposity (27,28). studies showing the effect of thiazolidinediones, a class of Using a small subset of FA variants and a limited sample medicines that improve insulin sensitivity, on promotion size with MRI scans (n = 9,510), we previously showed of differentiation of new adipocytes in SAT without that FA alleles were associated with lower ectopic liver fat changing VAT (30). In the light of strong association in women but not men (4). The availability of MRI scans between VAT and development of metabolic dysfunction of liver fat in 32,859 UK Biobank participants allowed us (31), our data suggest that VAT may reflect the ectopic to demonstrate that there is no sex-specific association fat deposition in the liver (r between the two measures = with liver fat. However, both FA and UFA variants had a 0.5 [14]) but itself may not be causally related to the greater overall effect on liver fat in women than men, development of cardiometabolic diseases. The sex-specific which could indicate the confounding effects of other fac- association between the FA cluster and lower VAT in tors in the measured liver fat in men. The association of women and higher VAT in men is also consistent with both FA and UFA clusters with, respectively, smaller and the more general sex-specific pattern of VAT distribution

Table 2—IVW two-sample MR meta-analysis of cardiometabolic diseases from published GWAS and FinnGen for FA and UFA clusters FA UFA Outcome OR Lower 95% CI Upper 95% CI P OR Lower 95% CI Upper 95% CI P Type 2 diabetes 0.11 0.08 0.16 4EÀ33 5.50 4.29 7.05 4EÀ41 Heart disease 0.34 0.25 0.47 2EÀ11 1.66 1.08 2.54 0.02

Hypertension 0.34 0.21 0.55 1EÀ05 3.03 2.18 4.22 5EÀ11

Stroke 0.65 0.52 0.83 0.0004 1.43 1.23 1.67 3EÀ06

NAFLD 0.14 0.03 0.79 0.03 3.70 1.22 11.17 0.02

PCOS 0.51 0.21 1.23 0.13 7.13 3.66 13.90 8EÀ09

OR, odds ratio. 1852 Types of Adiposity in Cardiometabolic Disease Diabetes Volume 70, August 2021

Figure 4—The IVW two-sample MR meta-analysis of published GWAS and FinnGen and one-sample MR of UK Biobank for FA and UFA clusters on risk of disease. The error bars and width of diamonds represent the 95% CIs of the IVW estimates in odds ratio (OR) per SD change in genetically determined FA and UFA.

as previously shown by genetic studies of waist-to-hip and after the onset of type 2 diabetes showed that liver ratio (WHR) (32). Furthermore, while FA alleles are asso- fat was the main mediator associated with glycemic con- ciated with lower VAT in women and higher VAT in men, trol (41). Furthermore, a recent genetic study of pancre- they are associated with protection from cardiometabolic atic fat and liver fat in the UK Biobank showed that diseases with similar effect size between men and women. genetic variants associated with pancreatic fat did not Although the UFA cluster was associated with higher have a significant impact on metabolic disease (14), sug- pancreatic fat, we did not detect any association between gesting that pancreatic fat has no direct role in pathoge- the FA cluster and this fat depot. FA variants individually nicity of type 2 diabetes and other metabolic disease. had a mixed effect on pancreatic fat with the FA allele Longitudinal imaging studies of individuals prior to and near PPARG, which is the most prominent example of FA after clinical disease onset in addition to MR studies of variants mimicking the effect of thiazolidinediones, being pancreatic fat in type 2 diabetes and other metabolic dis- associated with higher pancreatic fat. The role of pancre- eases will help to unravel the cause and consequence of atic fat in the pathogenicity of type 2 diabetes is currently this relationship. not clear-cut. Although many cross-sectional studies have The FA cluster was associated with a smaller pancreas reported higher pancreatic fat in subjects with type 2 dia- volume, whereas the UFA cluster had no association with betes compared with age-matched control subjects pancreas volume. Studies of individuals with type 2 diabe- (33–36), there are conflicting views regarding whether tes using CT or ultrasound have shown 7–33% lower pan- pancreatic fat is itself a driver of type 2 diabetes, with creas volume compared with that of control subjects some studies showing no association between type 2 dia- (39,42–44). Given that only 1–2% of the adult pancreas is betes and pancreatic fat using either CT or histology at composed of endocrine islets, changes in exocrine cell autopsy (37–40). Moreover, a recent study testing the so- number may contribute more to lower pancreas volume called “twin cycle model” (liver and pancreatic fat) before as shown in studies of type 1 diabetes (45). Since insulin diabetes.diabetesjournals.org Martin and Associates 1853 also acts as a growth stimulation hormone and maintains our approach and these two studies is that they have tissue mass (46,47), a decline in pancreas size in diabetes started with variants associated with BMI or WHR as could be due to a loss of the trophic effect of insulin on measures of adiposity. Recent studies have demonstrated exocrine cells (48,49). We previously showed that variants that BMI is a poor proxy for body adiposity (61), particu- associated with FA are associated with lower fasting insu- larly at an individual level, providing limited insight into lin levels (4–6), which could explain why the FA cluster body fat distribution, VAT, or nonadipose deposition of was associated with a smaller pancreas volume. fat (25). For example, only 7 and 12 of 36 FA variants are Subclinical inflammation is another factor that has associated with BMI and WHR, respectively. Similarly, been shown to be associated with components of meta- while UFA variants are enriched for BMI variants, 7 var- bolic syndrome and vascular disease (50–54). Our data iants are not associated with BMI and only 14 of 38 UFA provided no evidence for any direct role of CRP in mecha- variants are associated with WHR (Supplementary Fig. nisms that link FA and UFA clusters to, respectively, 10). Furthermore, using only BMI and type 2 diabetes to lower and higher risk of disease, since both clusters were identify variants with opposite effects can induce index associated with higher CRP levels, consistent with the event bias (62), e.g., TCF7L2 (60). findings of the largest MR study of CRP and risk of meta- There are several limitations to our study. First, we did bolic disease (23). We observed that the association not have independent studies to replicate the association between FA and UFA genetic scores and higher CRP levels with liver and pancreas fat and volume measurements. disappeared after adjustment for adiposity (BMI or body However, we used the largest data set on MRI phenotypes fat percentage), indicating that their effect on higher CRP available from the UK Biobank, with 32,859 samples, and was mediated by higher adiposity. This pattern of associa- replicated the association with some fat depots available tion could suggest that higher CRP levels are secondary to from a published GWAS (17). Second, our study popula- higher adiposity. Data on other markers of inflammation, tion was limited to Europeans only; it is unclear how our including tumor necrosis factor-a and interleukin-6, could findings can be generalized to other populations and clarify further the role of inflammation in cardiometabolic whether they can explain the excess risk of cardiometa- disease mechanisms. bolic disease in non-Europeans (63). Third, we lacked data Our tissue enrichment analysis provided further evi- on ectopic fat accumulation in muscle in our samples; dence that UFA and FA are biologically two different sub- future studies of MRI-derived muscle fat in the UK Bio- types of adiposity. FA loci were enriched for genes bank will enable the role of this ectopic fat in the patho- expressed in adipose tissue and adipocytes, while UFA loci physiology of cardiometabolic disease to be investigated. were enriched for genes expressed in mesenchymal stem Fourth, lower-body subcutaneous fat mass in the gluteo- cells. The enrichment of genes in adipose-related tissues femoral or leg region has previously been associated and cells was previously shown for loci associated with favorably with obesity-related cardiometabolic diseases WHR (55) in contrast to BMI loci enriched in the central (64). It would also have been of interest to study whether nervous system (56). However, this is the first time mes- the FA cluster is protective of cardiometabolic diseases, enchymal stem cells have been highlighted in tissue particularly via increasing gluteofemoral and leg fat mass. enrichment analysis to be associated with adiposity. Mes- Fifth, we used ALT and AST in our discovery pipeline to enchymal stem cells are major sources of adipocyte gener- identify FA and UFA variants, which could have biased our ation, and in addition to adipose tissue, they also exist in findings toward those variants that influence liver fat more skeletal muscle, the liver, and pancreas, which could sug- than pancreatic fat. Finally, in comparison with our previ- gest that they are responsible for ectopic fat formation in ous study (4), we did not have fasting insulin and adipo- these organs (57,58). Further experiments and data are nectin in our composite metabolic phenotype, since these necessary to determine the relationship between mesen- two biomarkers are not available in the UK Biobank. How- chymal stem cells and ectopic lipid accumulation. ever, our 36 FA variants include all 14 variants previously There have been few approaches to identify variants identified as FA (4) and the comparison of the multivariate associated with FA and UFA using different combinations GWAS P values for these variants (Supplementary Fig. 11) of traits. Winkler et al. (59) used 159 variants associated indicates additional power gained in the current study with BMI, WHR, or WHR adjusted for BMI and described largely attributable to the availability of other metabolic 24 FA variants as those associated with both lower WHR biomarkers in 451,099 individuals in a single cohort, the and higher BMI and 82 UFA variants as those associated UK Biobank. with both higher WHR and higher BMI. Pigeyre et al. (60) One of the major strengths of this study was the used polygenic correlation between BMI and type 2 diabe- unique approach to understanding different mechanisms tes to identify genetic regions where BMI-increasing effect underlying the association between adiposity and risk of was linked to a corresponding increase, decrease, or neu- cardiometabolic diseases. This unique approach is coupled tral effect on type 2 diabetes risk. Our FA and UFA var- with gold standard measurements of organ volume and iants that overlap with these studies are listed in content from MRI scans for understanding the role of dif- Supplementary Table 11. The main difference between ferent fat depots in pathogenicity. The availability of the 1854 Types of Adiposity in Cardiometabolic Disease Diabetes Volume 70, August 2021

UK Biobank made it possible to study the sex-specific review, or approval of the manuscript; and decision to submit the manuscript effects of our variants against metabolic biomarkers, MRI for publication. measures of fat distribution and ectopic fat, and risk of Authors Contributions. S.M. performed the statistical analyses. S.M. fi disease. We used the largest published GWAS and Finn- and H.Y. designed the study and wrote the rst draft of the manuscript. N.B., B.W., Y.L., J.D.B., and E.L.T. created the MRI-derived phenotypes and con- Gen study and independently replicated our results against tributed to the writing of the manuscript. M.C. and E.S. performed the GWAS risk of disease and performed MR studies. Unlike previous of MRI-derived phenotypes. J.T., R.N.B., A.R.W., and T.M.F. contributed to the studies that examined the role of ectopic fat and pancreas analysis of biomarkers from the UK Biobank. H.Y. is the guarantor of this size in small groups of participants categorized by diabetes work and, as such, had full access to all the data in the study and takes status, we investigated the role of these phenotypes in a responsibility for the integrity of the data and the accuracy of the data population-based study of 32,859 participants, which mini- analysis. mizes the effect of confounding factors and statistical bias. Finally, our sets of FA and UFA variants provide two References important genetic instruments for any MR study to exam- 1. Lindsay RS, Howard BV. Cardiovascular risk associated with the ine the causal role of adiposity on risk of disease uncoupled metabolic syndrome. Curr Diab Rep 2004;4:63–68 from its metabolic effect. 2. Stefan N, Kantartzis K, Machann J, et al. Identification and characterization of metabolically benign obesity in humans. Arch Intern Med 2008;168:1609–1616 3. Primeau V, Coderre L, Karelis AD, et al. Characterizing the profile of Conclusion obese patients who are metabolically healthy. Int J Obes 2011;35:971–981 This study provides genetic evidence for two types of adi- fi 4. Ji Y, Yiorkas AM, Frau F, et al. Genome-wide and abdominal MRI data posity: one coupled with a favorable metabolic pro le and provide evidence that a genetically determined favorable adiposity phenotype fi the other with an unfavorable pro le. Both FA and UFA is characterized by lower ectopic liver fat and lower risk of type 2 diabetes, variants were associated with higher CRP levels. We dem- heart disease, and hypertension. Diabetes 2019;68:207–219 onstrated that reduced liver fat, but not VAT or pancre- 5. Yaghootkar H, Lotta LA, Tyrrell J, et al. Genetic evidence for a link atic fat, is on the pathway that links FA to lower risk of between favorable adiposity and lower risk of type 2 diabetes, hypertension, diseases related to metabolic syndrome. We determined and heart disease. Diabetes 2016;65:2448–2460 no sexual dimorphism in the way the FA and UFA var- 6. Yaghootkar H, Scott RA, White CC, et al. Genetic evidence for a normal- iants are associated with metabolic profile, abdominal fat weight “metabolically obese” phenotype linking insulin resistance, hypertension, distribution, and risk of disease. Future MR, longitudinal, coronary artery disease, and type 2 diabetes. Diabetes 2014;63:4369–4377 and independent studies are required to elucidate whether 7. Lotta LA, Gulati P, Day FR, et al.; EPIC-InterAct Consortium; Cambridge FPLD1 Consortium. Integrative genomic analysis implicates limited peripheral higher pancreatic fat and smaller pancreas volume are a adipose storage capacity in the pathogenesis of human insulin resistance. Nat consequence of the ongoing pathological processes or Genet 2017;49:17–26 causal of cardiometabolic disease outcomes. Better under- 8. Collins R. What makes UK Biobank special? Lancet 2012;379:1173–1174 standing of the difference between FA and UFA may lead 9. Loh PR, Tucker G, Bulik-Sullivan BK, et al. Efficient Bayesian mixed- to new insights in preventing and predicting, and treating model analysis increases association power in large cohorts. Nat Genet people who suffer from, cardiometabolic diseases. 2015;47:284–290 10. Cichonska A, Rousu J, Marttinen P, et al. metaCCA: summary statistics- based multivariate meta-analysis of genome-wide association studies using Acknowledgments. This research has been conducted using data canonical correlation analysis. Bioinformatics 2016;32:1981–1989 from the UK Biobank resource and carried out under UK Biobank project 11. Semple RK, Savage DB, Cochran EK, Gorden P, O’Rahilly S. Genetic application numbers 9072, 9055, and 44584. UK Biobank protocols were syndromes of severe insulin resistance. Endocr Rev 2011;32:498–514 approved by the National Research Ethics Service Committee. The authors 12. Mahajan A, Taliun D, Thurner M, et al. Fine-mapping type 2 diabetes loci acknowledge the participants and investigators of the FinnGen study. The to single-variant resolution using high-density imputation and islet-specific authors thank Amoolya Singh and Kevin Wright, Calico Life Sciences LLC, for epigenome maps. Nat Genet 2018;50:1505–1513 their feedback on the manuscript. The authors would like to acknowledge the 13. Littlejohns TJ, Holliday J, Gibson LM, et al. The UK Biobank imaging use of the University of Exeter High-Performance Computing (HPC) facility in enhancement of 100,000 participants: rationale, data collection, management carrying out this work. We acknowledge use of high-performance computing and future directions. Nat Commun 2020;11:2624 funded by an MRC Clinical Research Infrastructure award (MRC Grant: MR/ 14. Liu Y, Basty N, Whitcher B, Bell JD, Sorokin EP, van Bruggen N, Thomas M008924/1). EL, Cule M. Genetic architecture of 11 organ traits derived from abdominal Funding. S.M. is funded by the Medical Research Council. H.Y. is funded MRI using deep learning. Elife. 2021 Jun 15:10:e65554 by a Diabetes UK RD Lawrence fellowship (17/0005594). J.T. is supported by 15. Bydder M, Ghodrati V, Gao Y, Robson MD, Yang Y, Hu P. Constraints in an Academy of Medical Sciences (AMS) Springboard Award, which is sup- estimating the proton density fat fraction. Magn Reson Imaging 2020;66:1–8 ported by the AMS, the Wellcome Trust, the Global Challenges Research 16. Bycroft C, Freeman C, Petkova D, et al. The UK Biobank resource with Fund, the U.K. Government Department of Business, Energy & Industrial deep phenotyping and genomic data. Nature 2018;562:203–209 Strategy, the British Heart Foundation, and Diabetes UK (SBF004\1079). 17. Chu AY, Deng X, Fisher VA, et al. Multiethnic genome-wide meta-analysis Duality of Interest. M.C., E.S., and Y.L. are employees of Calico Life of ectopic fat depots identifies loci associated with adipocyte development Sciences LLC. M.C., E.S., and Y.L. are funded by Calico Life Sciences LLC. and differentiation. Nat Genet 2017;49:125–130 No other potential conflicts of interest relevant to this article were reported. 18. Burgess S, Butterworth A, Thompson SG. Mendelian randomization The funders had no role in the design and conduct of the study; collec- analysis with multiple genetic variants using summarized data. Genet tion, management, analysis, and interpretation of the data; preparation, Epidemiol 2013;37:658–665 diabetes.diabetesjournals.org Martin and Associates 1855

19. Pierce BL, Burgess S. Efficient design for Mendelian randomization 41. Koivula RW, Atabaki-Pasdar N, Giordano GN, et al.; IMI DIRECT studies: subsample and 2-sample instrumental variable estimators. Am J Consortium. The role of physical activity in metabolic homeostasis before and Epidemiol 2013;178:1177–1184 after the onset of type 2 diabetes: an IMI DIRECT study. Diabetologia 20. FinnGen documentation of R4 release, 2020. Available from https:// 2020;63:744–756 finngen.gitbook.io/documentation/releases 42. Alzaid A, Aideyan O, Nawaz S. The size of the pancreas in diabetes 21. Viechtbauer W. Conducting meta-analyses in R with the metafor package. mellitus. Diabet Med 1993;10:759–763 J Stat Softw 2010;36:1–48 43. Lim S, Bae JH, Chun EJ, et al. Differences in pancreatic volume, fat 22. Pers TH, Karjalainen JM, Chan Y, et al.; Genetic Investigation of content, and fat density measured by multidetector-row computed tomography ANthropometric Traits (GIANT) Consortium. Biological interpretation of genome-wide according to the duration of diabetes. Acta Diabetol 2014;51:739–748 association studies using predicted gene functions. Nat Commun 2015;6:5890 44. Macauley M, Percival K, Thelwall PE, Hollingsworth KG, Taylor R. Altered 23. Ligthart S, Vaez A, V~osa U, et al.; LifeLines Cohort Study; CHARGE volume, morphology and composition of the pancreas in type 2 diabetes. Inflammation Working Group. Genome analyses of >200,000 individuals PLoS One 2015;10:e0126825 identify 58 loci for chronic inflammation and highlight pathways that link 45. Wright JJ, Saunders DC, Dai C, et al. Decreased pancreatic acinar cell inflammation and complex disorders. Am J Hum Genet 2018;103:691–706 number in type 1 diabetes. Diabetologia 2020;63:1418–1423 24. O’Donovan G, Thomas EL, McCarthy JP, et al. Fat distribution in men 46. Adler G, Kern HF. Regulation of exocrine pancreatic secretory process by of different waist girth, fitness level and exercise habit. Int J Obes insulin in vivo. Horm Metab Res 1975;7:290–296 2009;33:1356–1362 47. M€ossner J, Logsdon CD, Williams JA, Goldfine ID. Insulin, via its own 25. Thomas EL, Frost G, Taylor-Robinson SD, Bell JD. Excess body fat in receptor, regulates growth and amylase synthesis in pancreatic acinar AR42J obese and normal-weight subjects. Nutr Res Rev 2012;25:150–161 cells. Diabetes 1985;34:891–897 26. Unger RH. Minireview: weapons of lean body mass destruction: the role of 48. Henderson JR, Daniel PM, Fraser PA. The pancreas as a single organ: the ectopic lipids in the metabolic syndrome. Endocrinology 2003;144:5159–5165 influence of the endocrine upon the exocrine part of the gland. Gut 27. Krssak M, Falk Petersen K, Dresner A, et al. Intramyocellular lipid 1981;22:158–167 concentrations are correlated with insulin sensitivity in humans: a 1H NMR 49. Kusmartseva I, Beery M, Hiller H, et al. Temporal analysis of amylase spectroscopy study. Diabetologia 1999;42:113–116 expression in control, autoantibody-positive, and type 1 diabetes pancreatic 28. Fabbrini E, Magkos F, Mohammed BS, et al. Intrahepatic fat, not visceral tissues. Diabetes 2020;69:60–66 fat, is linked with metabolic complications of obesity. Proc Natl Acad Sci USA 50. Libby P. Inflammation in atherosclerosis. Arterioscler Thromb Vasc Biol 2009;106:15430–15435 2012;32:2045–2051 29. Kromrey ML, Ittermann T, vWahsen C, et al. Reference values of liver 51. Schmidt MI, Duncan BB, Sharrett AR, et al. Markers of inflammation and volume in Caucasian population and factors influencing liver size. Eur J Radiol prediction of diabetes mellitus in adults (Atherosclerosis Risk in Communities 2018;106:32–37 study): a cohort study. Lancet 1999;353:1649–1652 30. Adams M, Montague CT, Prins JB, et al. Activators of peroxisome 52. Festa A, D’Agostino R Jr, Tracy RP; Insulin Resistance Atherosclerosis proliferator-activated receptor gamma have depot-specific effects on human Study. Elevated levels of acute-phase and plasminogen activator preadipocyte differentiation. J Clin Invest 1997;100:3149–3153 inhibitor-1 predict the development of type 2 diabetes: the insulin resistance 31. Lee JJ, Pedley A, Hoffmann U, Massaro JM, Levy D, Long MT. Visceral and atherosclerosis study. Diabetes 2002;51:1131–1137 intrahepatic fat are associated with cardiometabolic risk factors above other ectopic 53. Hundal RS, Petersen KF, Mayerson AB, et al. Mechanism by which high- fat depots: the Framingham Heart Study. Am J Med 2018;131:684–692.e12 dose aspirin improves glucose metabolism in type 2 diabetes. J Clin Invest 32. Pulit SL, Stoneman C, Morris AP, et al.; GIANT Consortium. Meta-analysis 2002;109:1321–1326 of genome-wide association studies for body fat distribution in 694 649 54. Pickup JC. Inflammation and activated innate immunity in the individuals of European ancestry. Hum Mol Genet 2019;28:166–174 pathogenesis of type 2 diabetes. Diabetes Care 2004;27:813–823 33. Tushuizen ME, Bunck MC, Pouwels PJ, et al. Pancreatic fat content and 55. Shungin D, Winkler TW, Croteau-Chonka DC, et al.; ADIPOGen Consortium; beta-cell function in men with and without type 2 diabetes. Diabetes Care CARDIOGRAMplusC4D Consortium; CKDGen Consortium; GEFOS Consortium; 2007;30:2916–2921 GENIE Consortium; GLGC; ICBP; International Endogene Consortium; LifeLines 34. Lingvay I, Esser V, Legendre JL, et al. Noninvasive quantification of Cohort Study; MAGIC Investigators; MuTHER Consortium; PAGE Consortium; pancreatic fat in humans. J Clin Endocrinol Metab 2009;94:4070–4076 ReproGen Consortium. New genetic loci link adipose and insulin biology to body 35. Steven S, Hollingsworth KG, Small PK, et al. Weight loss decreases fat distribution. Nature 2015;518:187–196 excess pancreatic triacylglycerol specifically in type 2 diabetes. Diabetes Care 56. Locke AE, Kahali B, Berndt SI, et al.; LifeLines Cohort Study; ADIPOGen 2016;39:158–165 Consortium; AGEN-BMI Working Group; CARDIOGRAMplusC4D Consortium; 36. Wang CY, Ou HY, Chen MF, Chang TC, Chang CJ. Enigmatic ectopic fat: CKDGen Consortium; GLGC; ICBP; MAGIC Investigators; MuTHER prevalence of nonalcoholic fatty pancreas disease and its associated factors in Consortium; MIGen Consortium; PAGE Consortium; ReproGen Consortium; a Chinese population. J Am Heart Assoc 2014;3:e000297 GENIE Consortium; International Endogene Consortium. Genetic studies of body 37. Clark A, Wells CA, Buley ID, et al. Islet amyloid, increased A-cells, mass index yield new insights for obesity biology. Nature 2015;518:197–206 reduced B-cells and exocrine fibrosis: quantitative changes in the pancreas in 57. Uezumi A, Fukada S, Yamamoto N, Takeda S, Tsuchida K. Mesenchymal type 2 diabetes. Diabetes Res 1988;9:151–159 progenitors distinct from satellite cells contribute to ectopic fat cell formation 38. Gilbeau JP, Poncelet V, Libon E, Derue G, Heller FR. The density, contour, in skeletal muscle. Nat Cell Biol 2010;12:143–152 and thickness of the pancreas in diabetics: CT findings in 57 patients. AJR 58. Matsushita K, Dzau VJ. Mesenchymal stem cells in obesity: insights for Am J Roentgenol 1992;159:527–531 translational applications. Lab Invest 2017;97:1158–1166 39. Saisho Y, Butler AE, Meier JJ, et al. Pancreas volumes in humans from 59. Winkler TW, G€unther F, H€ollerer S, et al.; A joint view on genetic variants birth to age one hundred taking into account sex, obesity, and presence of for adiposity differentiates subtypes with distinct metabolic implications. Nat type-2 diabetes. Clin Anat 2007;20:933–942 Commun. 2018;9(1):1946 40. Yamazaki H, Tsuboya T, Katanuma A, et al. Lack of independent 60. Pigeyre M, Sjaarda J, Mao S, et al. Identification of novel causal blood association between fatty pancreas and incidence of type 2 diabetes: 5-year biomarkers linking metabolically favorable adiposity with type 2 diabetes risk. Japanese cohort study. Diabetes Care 2016;39:1677–1683 Diabetes Care 2019;42:1800–1808 1856 Types of Adiposity in Cardiometabolic Disease Diabetes Volume 70, August 2021

61. Nuttall FQ. Body mass index: obesity, BMI, and health: a critical review. 63. Yaghootkar H, Whitcher B, Bell JD, Thomas EL. Ethnic differences in Nutr Today 2015;50:117–128 adiposity and diabetes risk - insights from genetic studies. J Intern Med 62. Yaghootkar H, Bancks MP, Jones SE, et al. Quantifying the extent to 2020;288:271–283 which index event biases influence large genetic association studies. Hum Mol 64. Stefan N. Causes, consequences, and treatment of metabolically Genet 2017;26:1018–1030 unhealthy fat distribution. Lancet Diabetes Endocrinol 2020;8:616–627