Familial partial lipodystrophy (FPLD) is a heterogenous group of syndromes associated with a high prevalence of cardiometabolic diseases. Prior work has proposed DEXA-derived fat mass ratio (FMR), defined as trunk fat percentage divided by leg fat percentage, as a biomarker of FPLD, but this metric has not previously been characterized in large cohort studies. We set out to 1) understand the cardiometabolic burden of individuals with high FMR in up to 40,796 participants in the UK Biobank and 9,408 participants in the Fenland study, 2) characterize the common variant genetic underpinnings of FMR, and 3) build and test a polygenic predictor for FMR. Participants with high FMR were at higher risk for type 2 diabetes (odds ratio [OR] 2.30, P = 3.5 × 10−41) and metabolic dysfunction–associated liver disease or steatohepatitis (OR 2.55, P = 4.9 × 10−7) in UK Biobank and had higher fasting insulin (difference 19.8 pmol/L, P = 5.7 × 10−36) and fasting triglycerides (difference 36.1 mg/dL, P = 2.5 × 10−28) in the Fenland study. Across FMR and its component traits, 61 conditionally independent variant-trait pairs were discovered, including 13 newly identified pairs. A polygenic score for FMR was associated with an increased risk of cardiometabolic diseases. This work establishes the cardiometabolic significance of high FMR, a biomarker for FPLD, in two large cohort studies and may prove useful in increasing diagnosis rates of patients with metabolically unhealthy fat distribution to enable treatment or a preventive therapy.

Article Highlights
  • Across two population-based cohorts, fat mass ratio, a DEXA-based biomarker of fat distribution previously proposed as a means of discriminating partial lipodystrophy, is associated with an increased risk of cardiometabolic diseases, fasting insulin, and fasting triglycerides.

  • Fat mass ratio and its component traits are highly heritable, with a genome-wide polygenic score explaining 3.4–3.7% of trait variability and genome-wide association study in the UK Biobank revealing 61 conditionally independent variant-trait pairs.

  • A genome-wide polygenic score for fat mass ratio is associated with an increased risk of metabolic syndrome and cardiometabolic diseases in the UK Biobank and Fenland cohorts.

Familial partial lipodystrophy (FPLD) is a heterogeneous group of diseases characterized by reduced regional adipose tissue storage capacity, often in the subcutaneous depots of the arms, legs, and hips (1,2). Patients with FPLD are predisposed to type 2 diabetes, hypertriglyceridemia, hepatic steatosis, and other comorbidities, reflecting a state of insulin resistance (3,4). Despite this severe metabolic burden, FPLD remains an underdiagnosed group of diseases (5,6).

In some cases, FPLD is caused by rare DNA variants in specific genes, suggesting that a clinical genetic testing approach could increase diagnosis for these monogenic subtypes. For example, LMNA and PPARG, genes with critical roles in adipocyte function, are causal for FPLD type 2 (FPLD2) and FPLD type 3, respectively (2,5,7,,9). In contrast, FPLD type 1 (FPLD1) is thought to have a significant polygenic contribution rather than a single monogenic mutation (2,10,11). This has contributed to the hypothesis that monogenic and polygenic lipodystrophy lie along a spectrum, marking the severity of maladaptive fat distribution that contributes to insulin resistance and its comorbidities (6). A scalable biomarker that stratifies individuals along this “lipodystrophy axis” may help to improve risk prediction and increase recognition in clinical practice.

Fat mass ratio (FMR), defined as trunk fat percentage (trunk fat %) divided by leg fat percentage (leg fat %), was initially proposed as a single DEXA measurement that could discriminate patients with HIV with and without lipodystrophy (12,,14). Subsequent studies suggested that FMR may similarly have value in discriminating patients with FPLD2 or FPLD1 compared with matched control subjects (15–19). While the clinical diagnosis of FPLD requires integration of history, physical examination, body composition, and metabolic status, FMR is easily measured in biobank-scale cohorts using a standardized DEXA protocol and could serve as one strategy to stratify patients along the lipodystrophy axis.

In this study, we used the UK Biobank imaging study and Fenland prospective cohort study, a population-based prospective study with detailed metabolic phenotyping, to 1) determine the association of high FMR with cardiometabolic diseases in 50,204 individuals, 2) characterize the common DNA variant genetic architecture of FMR, and 3) construct and evaluate the associations of a genome-wide polygenic score for FMR (GPSFMR) with cardiometabolic diseases in up to 446,757 participants in the UK Biobank without DEXA imaging and 6,768 participants in the Fenland study.

Study Population

The current study included participants of the UK Biobank and the Fenland study. The UK Biobank is an observational study that enrolled >500,000 individuals between the ages of 40 and 69 years between 2006 and 2010, of whom 40,796 underwent DEXA imaging between 2014 and 2020 as part of an imaging substudy and had FMR available (20,48). This analysis of data from the UK Biobank was approved by the Mass General Brigham institutional review board and was performed under UK Biobank application #7089.

The Fenland study is a population-based prospective study with baseline measurements conducted between 2005 and 2015. A total of 12,435 people born between 1950 and 1975 were recruited from general practice lists in Cambridgeshire in the U.K. Ethical approval for the study was given by the Cambridge local research ethics committee (reference no. 04/Q0108/19), and all participants gave their written consent prior to entering the study (49).

FMR and Cardiometabolic Trait Definitions

FMR was computed by dividing trunk fat % (UK Biobank field 23286) by leg fat % (UK Biobank field 23276), as previously described (12–14). Details regarding the DEXA procedure and selection of FMR cutoff used in this study are available in the Supplementary Methods. We studied the association between FMR and a GPSFMR with liver fat, anthropometric quantities, cardiometabolic biomarkers, and cardiometabolic diseases in the UK Biobank and Fenland studies. Details regarding the quantification and definition of these measurements is available in the Supplementary Methods.

Genome-Wide Association Studies

Genome-wide association studies (GWAS) of FMR, leg fat %, and trunk fat % were performed using REGENIE version 2.2.4 (25). After genotype filtering, 433,616 single nucleotide polymorphisms (SNPs) were used as input into step 1 of REGENIE, and the 11,174,863 imputed variants following quality control described in the Supplementary Methods were used as input into step 2. Prior to conducting GWAS, each trait was normalized in the following fashion: the trait was residualized against age at the time of DEXA scan, age2, sex, age × sex, age2 × sex, the first 10 principal components of genetic ancestry, and genotyping array and then rank-based inverse normal transformed. The same covariates were included in the REGENIE version 2.2.4 call (25). A threshold of P < 5 × 10−8 was used to denote genome-wide significance. Further details regarding genotyping, quality control, SNP heritability estimation, identification of conditionally independent signals, and polygenic score construction are available in the Supplementary Methods.

Statistical Analysis

The association of FMR with prevalent cardiometabolic diseases in the UK Biobank was studied with logistic regression models adjusted for age, sex, BMI, smoking status (current smoker vs. not), diet (ideal diet vs. not), and physical activity (ideal, intermediate, or poor). Diet and physical activity categories were defined in accordance with a previous study in the UK Biobank (50). Incident disease analyses were done using Cox proportional hazards models adjusted for the same covariates. A similar framework was used when studying the association of GPSFMR with cardiometabolic outcomes with additional adjustment for the first 10 principal components of ancestry. In the Fenland study, a similar framework was used with all FMR models adjusted for age, sex, BMI, and smoking status, and all GPSFMR models were adjusted for the same covariates as well as the first 10 principal components of ancestry.

All analyses in the UK Biobank were performed using R version 3.6.0 software (R Project for Statistical Computing). All analyses in the Fenland study were performed using Stata 17 (StataCorp, College Station, TX).

Data and Resource Availability

This research has been conducted using the UK Biobank resource under application #7089. The raw UK Biobank data, including the anthropometric data reported here, are made available to researchers from universities and other research institutions with research inquiries following institutional review board and UK Biobank approval (https://www.ukbiobank.ac.uk/enable-your-research/apply-for-access). Summary statistics generated in this study will be available to download from the Common Metabolic Diseases Knowledge Portal (https://hugeamp.org) at the time of publication.

FMR, a DEXA-Based Metric for Partial Lipodystrophy

Leg fat %, trunk fat %, and FMR (trunk fat % / leg fat %) were measured in 40,796 participants in the UK Biobank, of whom 52% were female (20) (Table 1 and Supplementary Fig. 1). Mean age was 64 years, and 97% were White on the basis of self-reported race and ethnicity (Table 1). Female participants had lower mean FMR (0.99 vs. 1.42) and substantially higher mean leg fat % (40% vs. 25%) compared with male participants (Fig. 1A). Correlations of FMR with MRI-derived measures of fat distribution such as visceral adipose tissue (VAT) and gluteofemoral adipose tissue (GFAT) were anatomically consistent, with FMR being most strongly correlated with the ratio of VAT to GFAT (Pearson r: males 0.79, females 0.84), while leg fat % was most correlated with GFAT (Pearson r: males 0.83, females 0.80) (Fig. 1B).

Table 1

Characteristics of UK Biobank participants with FMR available

Male (n = 19,771)Female (n = 21,025)
Age (years), mean (SD) 65.1 (7.7) 63.7 (7.5) 
Self-reported ethnicity, n (%)   
 White 19,122 (96.7) 20,371 (96.9) 
 Black 118 (0.6) 138 (0.7) 
 South Asian 229 (1.2) 135 (0.6) 
 Other 302 (1.5) 381 (1.8) 
Current smoker, n (%) 831 (4.2) 626 (3.0) 
BMI (kg/m2), mean (SD) 27.1 (4.0) 26.2 (4.8) 
Waist circumference (cm), mean (SD) 83 (12) 94 (11) 
Hip circumference (cm), mean (SD) 101 (7) 101 (10) 
WHR 0.93 (0.06) 0.82 (0.07) 
Image-derived measurements   
 FMR 1.42 (0.27) 0.99 (0.20) 
 Leg fat % 25 (5) 40 (6) 
 Trunk fat % 35 (9) 40 (10) 
Prevalent diseases   
 Type 2 diabetes 1,035 (5.2) 522 (2.5) 
 Coronary artery disease 1,546 (7.8) 434 (2.1) 
 Heart failure 198 (1.0) 68 (0.3) 
 Stroke 898 (4.5) 368 (1.8) 
 Atrial fibrillation 305 (4.5) 148 (0.7) 
 MASLD/MASH 76 (0.4) 65 (0.3) 
 Hypertension 7,343 (37.1) 5,245 (24.9) 
Male (n = 19,771)Female (n = 21,025)
Age (years), mean (SD) 65.1 (7.7) 63.7 (7.5) 
Self-reported ethnicity, n (%)   
 White 19,122 (96.7) 20,371 (96.9) 
 Black 118 (0.6) 138 (0.7) 
 South Asian 229 (1.2) 135 (0.6) 
 Other 302 (1.5) 381 (1.8) 
Current smoker, n (%) 831 (4.2) 626 (3.0) 
BMI (kg/m2), mean (SD) 27.1 (4.0) 26.2 (4.8) 
Waist circumference (cm), mean (SD) 83 (12) 94 (11) 
Hip circumference (cm), mean (SD) 101 (7) 101 (10) 
WHR 0.93 (0.06) 0.82 (0.07) 
Image-derived measurements   
 FMR 1.42 (0.27) 0.99 (0.20) 
 Leg fat % 25 (5) 40 (6) 
 Trunk fat % 35 (9) 40 (10) 
Prevalent diseases   
 Type 2 diabetes 1,035 (5.2) 522 (2.5) 
 Coronary artery disease 1,546 (7.8) 434 (2.1) 
 Heart failure 198 (1.0) 68 (0.3) 
 Stroke 898 (4.5) 368 (1.8) 
 Atrial fibrillation 305 (4.5) 148 (0.7) 
 MASLD/MASH 76 (0.4) 65 (0.3) 
 Hypertension 7,343 (37.1) 5,245 (24.9) 

Maximum missingness rate for variables reported is 2.6% (BMI). All variables were measured at the time of imaging.

Figure 1

Distribution and correlation structure of FMR in the UK Biobank. A: Density plots showing the distribution of FMR and its two components leg fat % and trunk fat % among male and female participants of the UK Biobank. Dotted lines denote medians, while the solid lines in the FMR density plots denote the cutoffs for high FMR of 1.2 in female participants and 1.7 in male participants. B: Sex-stratified Pearson correlations among FMR, its components, image-derived measures of body fat distribution, and anthropometric measurements in the same participants (n = 33,133). ASAT, abdominal subcutaneous adipose tissue.

Figure 1

Distribution and correlation structure of FMR in the UK Biobank. A: Density plots showing the distribution of FMR and its two components leg fat % and trunk fat % among male and female participants of the UK Biobank. Dotted lines denote medians, while the solid lines in the FMR density plots denote the cutoffs for high FMR of 1.2 in female participants and 1.7 in male participants. B: Sex-stratified Pearson correlations among FMR, its components, image-derived measures of body fat distribution, and anthropometric measurements in the same participants (n = 33,133). ASAT, abdominal subcutaneous adipose tissue.

Close modal

Few studies have proposed an FMR cutoff in FPLD. Based on cutoffs proposed by the largest case-control studies to date of female patients with FPLD2 with DEXA data, we used a cutoff for high FMR in female participants as >1.2 (17,21) (Fig. 1 and Supplementary Fig. 2). This corresponded to the 87th percentile of the female FMR distribution in the UK Biobank (Supplementary Methods and Supplementary Data 1). Similar cutoffs were not available in the literature for male patients, so we used the matched 87th percentile of the male FMR distribution in the UK Biobank, corresponding to 1.7 (Supplementary Data 1). The FMR cutoffs of 1.2 in females and 1.7 in males corresponded to approximately the 95th sex-stratified percentiles in the Fenland study (Supplementary Data 1).

High FMR Is Associated With Cardiometabolic Disease in UK Biobank and Fenland Study Participants

Participants with high FMR had both lower leg fat % and increased trunk fat %, suggesting that high FMR was not exclusively driven by significantly increased trunk fat % (Fig. 2 and Supplementary Data 2). These participants also had higher liver fat % and an increased prevalence of hepatic steatosis (liver fat as assessed by MRI of >5.5%); for example, 40.6% of female participants with high FMR had imaging consistent with hepatic steatosis compared with 8.9% of female participants without (22) (Fig. 2A). Similar patterns were observed among normal or underweight participants with BMI <25 kg/m2 (15.1% vs. 1.5% in females) (Supplementary Fig. 3).

Figure 2

Association of high FMR with cardiometabolic diseases in the UK Biobank. Carriers were defined as participants with high FMR (>1.7 in male participants, >1.2 in female participants). A: Distribution of leg fat %, trunk fat %, MRI-quantified liver fat %, and MRI-defined hepatic steatosis (MRI-quantified liver fat >5.5%) according to high FMR. Boxed regions of box plots denote the interquartile range, while the upper and lower whiskers are 1.5 times the interquartile range away from Q3 and Q1, respectively. Outliers beyond these whiskers were removed for interpretability. B: ORs for prevalent disease correspond to logistic regression models adjusted for age, sex, BMI, healthy diet (ideal or poor), physical activity (ideal, intermediate, or poor), and smoking status (current smoker or nonsmoker).

Figure 2

Association of high FMR with cardiometabolic diseases in the UK Biobank. Carriers were defined as participants with high FMR (>1.7 in male participants, >1.2 in female participants). A: Distribution of leg fat %, trunk fat %, MRI-quantified liver fat %, and MRI-defined hepatic steatosis (MRI-quantified liver fat >5.5%) according to high FMR. Boxed regions of box plots denote the interquartile range, while the upper and lower whiskers are 1.5 times the interquartile range away from Q3 and Q1, respectively. Outliers beyond these whiskers were removed for interpretability. B: ORs for prevalent disease correspond to logistic regression models adjusted for age, sex, BMI, healthy diet (ideal or poor), physical activity (ideal, intermediate, or poor), and smoking status (current smoker or nonsmoker).

Close modal

Next, we studied the association of high FMR with prevalent and incident cardiometabolic diseases in UK Biobank (Supplementary Data 3). In adjusted logistic regression models, participants with high FMR had a significantly elevated risk of type 2 diabetes (OR 2.30, P = 3.5 × 10−41), coronary artery disease (OR 1.33, P = 1.7 × 10−5), metabolic dysfunction–associated liver disease or steatohepatitis (MASLD/MASH) (OR 2.55, P = 4.9 × 10−7), and hypertension (OR 1.41, P = 5.4 × 10−25) (Fig. 2B and Supplementary Data 4). Similar results were observed when FMR was analyzed as a continuous variable (Supplementary Data 5). Choosing a more stringent cutoff for high FMR of 1.5 in females with a percentile-matched cutoff in males revealed even greater risk (e.g., type 2 diabetes OR 4.54, P = 9.2 × 10−23) (Supplementary Data 6). Results were broadly consistent in incident disease analyses and across sex and BMI subgroups (Supplementary Data 4). Notably, high FMR remained an independent predictor of these cardiometabolic diseases after adjustment for waist-to-hip ratio (WHR), a potential anthropometric proxy (Supplementary Data 7).

We sought to replicate the cardiometabolic burden of high FMR in the Fenland cohort. Up to 9,408 participants with FMR (mean age 48 years, 54% were female) were available for analysis, with 447 (4.8%) classified as having high FMR (Supplementary Data 8). In adjusted linear regression models, participants with high FMR had significantly higher fasting insulin (mean 75.9 vs. 46.1 pmol/L, adjusted difference 19.8 pmol/L, P = 5.7 × 10−36) and fasting triglycerides (mean 149 vs. 103 mg/dL, adjusted difference 36 mg/dL, P = 2.5 × 10−28) (Table 2 and Supplementary Data 9). Participants with high FMR also had a significantly higher prevalence of hepatic steatosis (OR 3.09, P = 1.9 × 10−24) (Supplementary Fig. 4). Associations between FMR and cardiometabolic biomarkers in the UK Biobank were limited by DEXA measurements being made a median 9 years after enrollment, though exploratory analyses were consistent with results from Fenland (adjusted nonfasting triglyceride difference between groups 42 mg/dL, P = 1.4 × 10−225). Taken together, these results suggest that individuals with high FMR have a significant cardiometabolic burden evidenced by 1) an increased risk of cardiometabolic diseases, including type 2 diabetes, coronary artery disease, and MASLD/MASH, and 2) a metabolically unhealthy profile of relevant biomarkers.

Table 2

Association of FMR with metabolic biomarkers in the Fenland cohort

BiomarkerRemainder, mean (SD)High FMR, mean (SD)Adjusted β (95% CI)P
Fasting insulin (pmol/L) 46.12 (33.42) 75.94 (76.01) 19.8 (16.7, 22.9) 5.69 × 10−36 
Fasting triglycerides (mg/dL) 102.7 (71.7) 148.8 (89.5) 36.1 (29.6, 42.6) 2.46 × 10−28 
Fasting glucose (mg/dL) 86.7 (11.4) 91.7 (15.3) 3.4 (2.4, 4.5) 1.34 × 10−10 
HbA1c (%) 5.52 (0.46) 5.69 (0.49) 0.10 (0.06, 0.14) 4.01 × 10−6 
Liver score 4.28 (1.17) 5.27 (1.37) 0.56 (0.46, 0.66) 3.37 × 10−27 
Steatosis   3.09 (2.49, 3.84)* 1.86 × 10−24 
BiomarkerRemainder, mean (SD)High FMR, mean (SD)Adjusted β (95% CI)P
Fasting insulin (pmol/L) 46.12 (33.42) 75.94 (76.01) 19.8 (16.7, 22.9) 5.69 × 10−36 
Fasting triglycerides (mg/dL) 102.7 (71.7) 148.8 (89.5) 36.1 (29.6, 42.6) 2.46 × 10−28 
Fasting glucose (mg/dL) 86.7 (11.4) 91.7 (15.3) 3.4 (2.4, 4.5) 1.34 × 10−10 
HbA1c (%) 5.52 (0.46) 5.69 (0.49) 0.10 (0.06, 0.14) 4.01 × 10−6 
Liver score 4.28 (1.17) 5.27 (1.37) 0.56 (0.46, 0.66) 3.37 × 10−27 
Steatosis   3.09 (2.49, 3.84)* 1.86 × 10−24 

High FMR was defined as >1.7 in male participants and >1.2 in female participants. β and 95% CIs are reported from linear regressions testing the association between high FMR status and each biomarker in models adjusted for age, sex, BMI, and smoking status. An additional analysis with the liver score was undertaken after stratifying into 1) normal liver, 2) mild steatosis, and 3) moderate/severe steatosis by ordered logistic regression adjusted for the same covariates as used to estimate the OR.

*

OR (95% CI).

GWAS of FMR

We next sought to evaluate whether FMR and its component traits have an important inherited component by quantifying heritability. Using the previously described BOLT-REML method (23), the SNP heritability (h2g) of FMR was estimated to be 0.36 (SE 0.01), consistent with a previously reported estimate for the MRI-derived VAT/GFAT ratio in an overlapping patient population (24) (Supplementary Data 10). Leg fat % had a higher SNP heritability (h2g = 0.39; SE 0.01), consistent with a previous report of GFAT having higher heritability than other fat depots (24). The genomic inflation factor for FMR was 1.11, with a linkage disequilibrium (LD) score regression intercept of 1.04, consistent with polygenicity rather than significant unaccounted-for population structure.

Next, we conducted GWAS of FMR, leg fat %, and trunk fat % using the REGENIE software package (25). Across all three traits, we identified 61 conditionally independent SNP-trait pairs at a genome-wide significance threshold of 5 × 10−8 (Fig. 3 and Supplementary Data 11), including 13 that were newly identified (Table 3).

Figure 3

Manhattan plots of FMR and its component traits. Manhattan plots (x-axis, location of a given SNP across 22 autosomes; y-axis, –log10P value corresponding to a given SNP-trait association) for GWAS of FMR and its component traits leg fat % and trunk fat %. Lead SNPs were defined as the most significant association in a genetic locus with P < 5 × 10−8 and are labeled by their nearest gene. A full list of conditionally independent GWAS-significant associations are available in Supplementary Data 11.

Figure 3

Manhattan plots of FMR and its component traits. Manhattan plots (x-axis, location of a given SNP across 22 autosomes; y-axis, –log10P value corresponding to a given SNP-trait association) for GWAS of FMR and its component traits leg fat % and trunk fat %. Lead SNPs were defined as the most significant association in a genetic locus with P < 5 × 10−8 and are labeled by their nearest gene. A full list of conditionally independent GWAS-significant associations are available in Supplementary Data 11.

Close modal
Table 3

Newly identified genetic associations with FMR, leg fat %, and trunk fat %

TraitChrGRCh37 positionIDEffect alleleOther alleleEffect allele frequencyβ (SE)PNearest geneConsequence
FMR 1056550 7:1056550_GCACAGCAGCA_G GCACAGCAGCA 0.894 −0.071 (0.013) 3.00 × 10−8 C7orf50 intron_variant 
FMR 20 39235307 rs117822025 0.92 −0.075 (0.013) 3.01 × 10−8 MAFB intergenic_variant 
FMR 22 38607534 rs11450220 TG 0.448 −0.048 (0.007) 5.73 × 10−12 MAFF intron_variant 
Leg fat % 173388948 rs5873402 CA 0.702 0.047 (0.008) 5.75 × 10−10 CPEB4 3_prime_UTR_variant 
Leg fat % 10 126186794 rs74160915 0.972 0.118 (0.021) 2.06 × 10−8 LHPP intron_variant 
Leg fat % 11 69184067 rs12293855 0.873 0.067 (0.01) 1.30 × 10−10 RP11-211G23.2 upstream_gene_variant 
Leg fat % 14 52508842 rs61747585 0.959 0.098 (0.017) 2.43 × 10−8 NID2 missense_variant 
Leg fat % 17 43169109 rs2239918 0.545 0.038 (0.007) 4.55 × 10−8 NMT1 intron_variant 
Leg fat % 18 2846812 rs11664106 0.63 −0.042 (0.007) 1.64 × 10−8 EMILIN2 upstream_gene_variant 
Leg fat % 21 46786029 21:46786029_GT_G GT 0.436 −0.042 (0.007) 4.35 × 10−9 COL18A1 intergenic_variant 
Leg fat % 22 41854446 rs9607805 0.28 0.046 (0.008) 3.08 × 10−9 PHF5A downstream_gene_variant 
Trunk fat % 25984914 rs146298822 0.994 0.264 (0.046) 1.12 × 10−8 ASXL2 intron_variant 
Trunk fat % 122463915 rs145151270 0.995 0.298 (0.051) 4.23 × 10−9 HSPBAP1 intron_variant 
TraitChrGRCh37 positionIDEffect alleleOther alleleEffect allele frequencyβ (SE)PNearest geneConsequence
FMR 1056550 7:1056550_GCACAGCAGCA_G GCACAGCAGCA 0.894 −0.071 (0.013) 3.00 × 10−8 C7orf50 intron_variant 
FMR 20 39235307 rs117822025 0.92 −0.075 (0.013) 3.01 × 10−8 MAFB intergenic_variant 
FMR 22 38607534 rs11450220 TG 0.448 −0.048 (0.007) 5.73 × 10−12 MAFF intron_variant 
Leg fat % 173388948 rs5873402 CA 0.702 0.047 (0.008) 5.75 × 10−10 CPEB4 3_prime_UTR_variant 
Leg fat % 10 126186794 rs74160915 0.972 0.118 (0.021) 2.06 × 10−8 LHPP intron_variant 
Leg fat % 11 69184067 rs12293855 0.873 0.067 (0.01) 1.30 × 10−10 RP11-211G23.2 upstream_gene_variant 
Leg fat % 14 52508842 rs61747585 0.959 0.098 (0.017) 2.43 × 10−8 NID2 missense_variant 
Leg fat % 17 43169109 rs2239918 0.545 0.038 (0.007) 4.55 × 10−8 NMT1 intron_variant 
Leg fat % 18 2846812 rs11664106 0.63 −0.042 (0.007) 1.64 × 10−8 EMILIN2 upstream_gene_variant 
Leg fat % 21 46786029 21:46786029_GT_G GT 0.436 −0.042 (0.007) 4.35 × 10−9 COL18A1 intergenic_variant 
Leg fat % 22 41854446 rs9607805 0.28 0.046 (0.008) 3.08 × 10−9 PHF5A downstream_gene_variant 
Trunk fat % 25984914 rs146298822 0.994 0.264 (0.046) 1.12 × 10−8 ASXL2 intron_variant 
Trunk fat % 122463915 rs145151270 0.995 0.298 (0.051) 4.23 × 10−9 HSPBAP1 intron_variant 

Newly identified loci were identified by comparing lead SNPs identified in the current study with previous GWAS of related anthropometric traits using the Common Metabolic Diseases Knowledge Portal (Research Design and Methods). Chr, chromosome.

Using summary statistics from these three traits along with previously published summary statistics, we estimated genome-wide genetic correlation among FMR, its component traits, and a selection of other anthropometric measures (24,26,27). Genetic correlations were broadly consistent with observational Pearson correlations (Supplementary Fig. 5). Notably, FMR had high genetic correlation with VAT/GFAT ratio (rg = 0.84), and leg fat % had high genetic correlation with GFAT (rg = 0.80), consistent with each pair of observational correlations and suggesting that these DEXA-derived quantities are reasonable proxies of similar quantities derived from MRI.

GWAS of FMR revealed 24 conditionally independent associations spanning 23 loci (Supplementary Data 11). The top GWAS association with FMR was at an RSPO3 locus including two conditionally independent SNPs: rs577721086 (P = 1.8 × 10−37) and rs1936789 (P = 4.4 × 10−22). These SNPs have been previously associated with VAT/GFAT ratio and are in strong LD with rs72959041 (in-sample R2 = 1.00) and rs1936807 (in-sample R2 = 0.86), respectively, two top GWAS associations with WHR adjusted for BMI (WHRadjBMI) that have been validated in prior work (24,28,29). Other top associations included rs7133378, an intronic DNAH10 variant (P = 9.1 × 10−24), and rs998584 downstream of VEGFA (7.5 × 10−24). Both variants have previously been reported to be associated with WHRadjBMI and have strong associations with serum triglycerides and HDL cholesterol (28,30). DNAH10 has been implicated in the pathogenesis of abnormal lipid accumulation in mouse adipocyte models and in total body triglyceride levels in Drosophila models (10,31). VEGFA overexpression has previously been shown in mouse models to facilitate the expansion of beige fat in subcutaneous white adipose tissue with an associated metabolic benefit, particularly during fat mass expansion (32,33). Discussion of GWAS results for leg fat %, external validation of all GWAS associations, and tissue enrichment analyses can be found in the Supplemental Material.

A High Polygenic Score Predisposes to High FMR

Next, we sought to identify a genetic predictor for lipodystrophy using common DNA variants. We used the LDPred2 software package to construct a GPSFMR in the UK Biobank using 1,124,851 HapMap3 SNPs (34,35) (Supplementary Data 14). Among the 20% of genotyped UK Biobank participants who were held out from polygenic score training and validation, GPSFMR explained 3.7% of FMR trait variance (Supplementary Fig. 7 and Supplementary Data 15). Participants with GPSFMR in the top decile were nearly twice as likely to have a high FMR (OR 1.88, P = 3.5 × 10−10) (Fig. 4 and Supplementary Data 16). A greater than threefold gradient was noted in the prevalence of high FMR across deciles of GPSFMR: 5.9% of participants in the bottom decile had high FMR, while the corresponding percentage in the top decile was 20.3% (Fig. 4).

Figure 4

Prevalence of high FMR according to GPSFMR. High FMR was defined as >1.7 in male participants and >1.2 in female participants. High GPSFMR was defined as the top decile in the DEXA imaging holdout cohort (n = 7,798).

Figure 4

Prevalence of high FMR according to GPSFMR. High FMR was defined as >1.7 in male participants and >1.2 in female participants. High GPSFMR was defined as the top decile in the DEXA imaging holdout cohort (n = 7,798).

Close modal

Similar associations were observed in the Fenland cohort, where GPSFMR explained 3.4% of the trait variance (Supplementary Data 15). In the Fenland cohort, participants with GPSFMR in the top decile were more than twice as likely to have a high FMR (OR 2.59, P = 6.4 × 10−10) (Supplementary Data 16). Taken together, these data suggest that individuals with markedly elevated FMR have a significant polygenic contribution to their phenotype.

High GPSFMR Is Associated With Cardiometabolic Diseases

We studied the association of GPSFMR derived from imaged UK Biobank participants in 1) participants of the UK Biobank without DEXA imaging and 2) participants in the Fenland longitudinal cohort study (Supplementary Fig. 1). Nonimaged UK Biobank participants with GPSFMR in the top decile had higher WHR (0.17 sex-standardized SD, P < 1.0 × 10−300), higher serum triglycerides (0.15 SD, P = 9.1 × 10−206), lower HDL cholesterol (−0.12 SD, P = 3.2 × 10−141), and higher serum ALT (0.08 SD, P = 1.2 × 10−56) in adjusted linear regression models (Fig. 5). These participants also had a higher prevalence of metabolic syndrome (OR 1.37, P = 2.4 × 10−117), type 2 diabetes (OR 1.34, P = 6.1 × 10−22), coronary artery disease (OR 1.18, P = 1.9 × 10−11), MASLD/MASH (OR 1.59, P = 1.1 × 10−4), and hypertension (OR 1.14, P = 2.4 × 10−27) in adjusted logistic regression models. Given prior work noting an association among FPLD, polycystic ovary syndrome, and hyperandrogenism, we additionally tested the association between GPSFMR and sex hormones measured at the time of UK Biobank enrollment (36,37). Female participants with GPSFMR in the top decile had an elevated androgen index, a proxy for free testosterone estimated as the ratio of total testosterone to sex hormone–binding globulin, which may be consistent with a more hyperandrogenic state (38) (Supplementary Data 17).

Figure 5

High GPSFMR is associated with cardiometabolic diseases in a nonimaged subcohort of the UK Biobank. Participants who did not have DEXA imaging available and who were unrelated to participants in the GPSFMR training set were included (up to 442,919 participants). Carriers were defined as those with a high GPSFMR (in the top decile in the nonimaged subcohort of the UK Biobank). Effect sizes for anthropometrics and biomarkers are reported from linear regressions adjusted for age, sex, BMI, diet (ideal or poor), physical activity (ideal, intermediate, or poor), smoking status (current smoker or nonsmoker), and the first 10 principal components of genetic ancestry. Anthropometric effect sizes are reported in units of sex-specific SD, while biomarker effect sizes are reported in units of SD. ORs for prevalent disease corresponding to logistic regression models adjusted for the same covariates. HDL-c, HDL cholesterol; LDL-c, LDL cholesterol.

Figure 5

High GPSFMR is associated with cardiometabolic diseases in a nonimaged subcohort of the UK Biobank. Participants who did not have DEXA imaging available and who were unrelated to participants in the GPSFMR training set were included (up to 442,919 participants). Carriers were defined as those with a high GPSFMR (in the top decile in the nonimaged subcohort of the UK Biobank). Effect sizes for anthropometrics and biomarkers are reported from linear regressions adjusted for age, sex, BMI, diet (ideal or poor), physical activity (ideal, intermediate, or poor), smoking status (current smoker or nonsmoker), and the first 10 principal components of genetic ancestry. Anthropometric effect sizes are reported in units of sex-specific SD, while biomarker effect sizes are reported in units of SD. ORs for prevalent disease corresponding to logistic regression models adjusted for the same covariates. HDL-c, HDL cholesterol; LDL-c, LDL cholesterol.

Close modal

In the Fenland study, participants with GPSFMR in the top decile compared with all others had higher fasting insulin (adjusted difference 5.2 pmol/L, P = 9.6 × 10−5), fasting triglycerides (adjusted difference 10.8 mg/dL, P = 7.0 × 10−5), and ultrasound-determined hepatic steatosis (OR 1.44, P = 4.8 × 10−4), mirroring the observational associations between FMR and metabolic abnormalities (Supplementary Data 18).

High FMR Identifies Hypertriglyceridemia and Type 2 Diabetes in the Absence of Severe Obesity

We aimed to determine the association between high FMR or high GPSFMR with a partial lipodystrophy–like syndrome. We initially examined the ICD-10 code for lipodystrophy (E88.1) as a potential phenotype of interest. After excluding individuals with HIV, 16 participants (0.003%) were identified in the UK Biobank with this ICD-10 code (5). On average, these participants had similar WHR and serum triglycerides as noncarriers (Supplementary Data 19). This may be consistent with prior work that noted the heterogeneous nature of the lipodystrophy ICD-10 code, making it difficult to distinguish FPLD-like syndromes from metabolically unrelated disease entities, such as insulin injection–related lipoatrophy, based on electronic health record diagnosis codes alone (39). Accordingly, 6 of 16 (38%) of these carriers also carried an ICD-10 code for type 1 diabetes.

Given the limitation of the ICD-10 code, we sought to define an alternative proxy phenotype in the UK Biobank. Motivated by cutoffs used for inclusion criteria in a prior clinical trial, we defined a syndrome of hypertriglyceridemia and type 2 diabetes in the absence of severe obesity as 1) serum triglycerides >500 mg/dL, 2) type 2 diabetes or hemoglobin A1c between 7–12%, and 3) BMI <35 kg/m2 (40). In the UK Biobank, 564 participants (0.13%) met these criteria and had a significant burden of cardiometabolic disease (Supplementary Data 20). In the subset of participants in UK Biobank with DEXA imaging, participants with high FMR were significantly more likely to have this syndrome (OR 3.16, P = 1.1 × 10−3) (Supplementary Fig. 8). Among those without DEXA imaging available, participants with GPSFMR in the top decile were also more likely to have this phenotype (OR 1.82, P = 4.9 × 10−7). Results were broadly consistent in a sensitivity analysis where a cutoff of BMI <30 kg/m2 was used instead of 35 kg/m2 (Supplementary Data 21).

In this study, we phenotypically and genetically characterized FMR, a DEXA-derived measurement previously proposed as a tool for identifying FPLD, in two population-based cohorts. Individuals with high FMR, defined as >1.2 in females and >1.7 in males in accordance with prior work (17,21), were found to have an elevated risk of type 2 diabetes, coronary artery disease, hypertension, and MASLD/MASH in the UK Biobank and a metabolically deranged profile of biomarkers in the Fenland cohort. Notably, none of the participants with high FMR had an ICD-10 code for lipodystrophy, highlighting the potential diagnostic gains that may be made by a clinical tool for identifying lipodystrophy-like body habitus and its associated metabolic comorbidities. These results have at least three implications.

First, DEXA-based FMR may provide a useful bridge between two paradigms of partial lipodystrophy. In clinical practice, partial lipodystrophy is conceptualized as a rare set of diseases that require expert clinician diagnosis (4). Identifying rare pathogenic variants in lipodystrophy-related genes such as LMNA and PPARG may further support the diagnosis for certain subtypes (2). An alternative paradigm focuses on a shared pathophysiology of impaired adipocyte differentiation underlying partial lipodystrophy and type 2 diabetes, which might suggest a far greater contribution to disease in the general population (41). Several lines of genetic evidence support this notion, including 1) an unsupervised clustering approach for type 2 diabetes genetic loci identifying a cluster with a lipodystrophy-like signature, 2) an insulin resistance polygenic score converging on reduced peripheral adipose tissue mass and showing an association with FPLD1, and 3) rare variants in several FPLD genes, including LMNA, PPARG, PLIN1, and LIPE, showing associations with WHRadjBMI in a population-based cohort (10,42,,44). These findings have led to the proposal of a continuous lipodystrophy axis whereby the degree of adipose tissue dysfunction confers a proportional risk of insulin resistance (6). This study proposes FMR, previously shown to discriminate individuals with FPLD2 and FPLD1 from control subjects, as an imaging-based measure of the lipodystrophy axis, determining a continuous gradient of cardiometabolic risk and identifying a subset of the population with high FMR who are at particularly high risk (15–19). Future studies including patients with expert-adjudicated FPLD alongside paired genetic and imaging data will enable further investigation of the lipodystrophy axis.

Second, this study estimates the prevalence of a lipodystrophy-like phenotype in two population-based cohorts. Recent work aiming to estimate the clinical prevalence of FPLD in a population-based cohort was centered around lipodystrophy ICD-10 codes (5). While some of these participants are likely to have FPLD, this approach may have been prone to misclassification because lipodystrophy ICD-10 codes are commonly used for the metabolically unrelated entity of localized lipoatrophy; indeed, our study provides evidence of this phenomenon in the UK Biobank (39). Alternatively, a genetics-first approach to disease ascertainment using rare variants in genes such as LMNA and PPARG may be a useful strategy, with the genetic prevalence of FPLD noted to be higher than its clinical prevalence (5). However, this approach does not account for the larger number of individuals in the general population who may have a less severe, but still metabolically significant, lipodystrophy-like change in body fat distribution (6). On the basis of previously proposed cutoffs for high FMR, we identified ∼1 in 8 participants of the UK Biobank and ∼1 in 20 participants of the Fenland study with this profile, ∼1,000 times more common than the clinical or genetic prevalence for FPLD (5,39). While this high FMR status should not be equated to FPLD, many of these participants are likely to lie on the higher end of a lipodystrophy axis with a proportional conferred cardiometabolic risk. Notably, the higher cardiometabolic risk among high FMR carriers was maintained even in participants with a normal BMI and across both males and females. The latter point is notable in the context of a female ascertainment bias noted in the FPLD literature (5).

Third, this study contributes a polygenic score to work toward an algorithmic solution for recruiting patients toward the higher end of a lipodystrophy axis, including patients with FPLD. One key challenge in implementing precision medicine approaches to identify patients with lipodystrophy-like body composition is that the necessary medical imaging neither is broadly available nor thought to be clinically justified for determining body composition. This in part explains why clinical practice relies on anthropometric proxies, like BMI and waist circumference, and frames the background for recent innovation aiming to extract body composition from cheaper imaging modalities (19,45,,47). These approaches are likely to be a crucial part of the algorithmic solution to identify patients with a lipodystrophy-like phenotype but are not ideal for identifying a group of individuals in the general population to recruit for further diagnostic workup. To this end, a genetics-first approach could be useful to recruit patients who are candidates for medical imaging to quantify body composition; for example, a panel including rare FPLD-associated variants and polygenic scores pertaining to the lipodystrophy-like phenotype, including GPSFMR reported here, may be useful (2,5,10). This genetics-first approach may be particularly useful for building a preventive medicine approach to FPLD and its comorbidities given that disease-associated adipose tissue changes may begin as early as puberty (1,4).

In conclusion, we studied cardiometabolic disease associations and the genetic basis of DEXA-derived FMR, a candidate biomarker for FPLD, in two population-based cohorts. Our work highlights the cardiometabolic significance of having a high FMR, lends further support to the existence of an underdiagnosed lipodystrophy-like phenotype, and adds to a growing toolbox for identifying patients with metabolically unhealthy fat distribution.

See accompanying article, p. 1039.

This article contains supplementary material online at https://doi.org/10.2337/figshare.25189037.

N.J.W. and A.V.K. jointly supervised this work.

Funding. This work was supported by a Sarnoff Cardiovascular Research Foundation Fellowship (to S.A.), National Human Genome Research Institute grants 1K08HG010155 and 1U01HG011719 (to A.V.K.), a Hassenfeld Scholar Award from Massachusetts General Hospital (to A.V.K.), and a Merkin Institute Fellowship from the Broad Institute of MIT and Harvard (to A.V.K.).

Duality of Interest. S.A. has served as scientific consultant for Third Rock Ventures and Marea Therapeutics. A.V.K. is an employee of Verve Therapeutics; has served as a scientific advisor to Amgen, Novartis, Silence Therapeutics, Veritas International, Color Health, Third Rock Ventures, Marea Therapeutics, and Foresite Laboratories; holds equity in Verve Therapeutics, Marea Therapeutics, Color Health, and Foresite Laboratories; and received a sponsored research agreement from IBM Research. E.J.W. and B.B.C. are employed by and hold equity in Marea Therapeutics. B.B.C. holds equity in Maze Therapeutics. N.J.W. has acted as a scientific advisor for Third Rock Ventures and Marea Therapeutics and holds equity in Marea Therapeutics. S.A. and A.V.K. are listed as coinventors on a patent application for the use of imaging data in assessing body fat distribution and associated cardiometabolic risk. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. S.A. drafted the manuscript. S.A. and J.L. acquired, analyzed, and interpreted the data. S.A., J.L., B.B.C., E.J.W., N.J.W., and A.V.K. critically revised the manuscript for important intellectual content. S.A., B.B.C., E.J.W., N.J.W., and A.V.K. conceived and designed the study. A.V.K. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

1.
Garg
A
.
Lipodystrophies: genetic and acquired body fat disorders
.
J Clin Endocrinol Metab
2011
;
96
:
3313
3325
2.
Lim
K
,
Haider
A
,
Adams
C
,
Sleigh
A
,
Savage
DB
.
Lipodistrophy: a paradigm for understanding the consequences of “overloading” adipose tissue
.
Physiol Rev
2021
;
101
:
907
993
3.
Hegele
RA
.
Insulin resistance in human partial lipodystrophy
.
Curr Atheroscler Rep
2000
;
2
:
397
404
4.
Brown
RJ
,
Araujo-Vilar
D
,
Cheung
PT
, et al
.
The diagnosis and management of lipodystrophy syndromes: a multi-society practice guideline
.
J Clin Endocrinol Metab
2016
;
101
:
4500
4511
5.
Gonzaga-Jauregui
C
,
Ge
W
,
Staples
J
, et al.;
Geisinger-Regeneron DiscovEHR Collaboration
.
Clinical and molecular prevalence of lipodystrophy in an unascertained large clinical care cohort
.
Diabetes
2020
;
69
:
249
258
6.
Mann
JP
,
Savage
DB
.
What lipodystrophies teach us about the metabolic syndrome
.
J Clin Invest
2019
;
129
:
4009
4021
7.
Agarwal
AK
,
Garg
A
.
A novel heterozygous mutation in peroxisome proliferator-activated receptor-γ gene in a patient with familial partial lipodystrophy
.
J Clin Endocrinol Metab
2002
;
87
:
408
411
8.
Agostini
M
,
Schoenmakers
E
,
Mitchell
C
, et al
.
Non-DNA binding, dominant-negative, human PPARgamma mutations cause lipodystrophic insulin resistance
.
Cell Metab
2006
;
4
:
303
311
9.
Shackleton
S
,
Lloyd
DJ
,
Jackson
SN
, et al
.
LMNA, encoding lamin A/C, is mutated in partial lipodystrophy
.
Nat Genet
2000
;
24
:
153
156
10.
Lotta
LA
,
Gulati
P
,
Day
FR
, et al.;
EPIC-InterAct Consortium
;
Cambridge FPLD1 Consortium
.
Integrative genomic analysis implicates limited peripheral adipose storage capacity in the pathogenesis of human insulin resistance
.
Nat Genet
2017
;
49
:
17
26
11.
Rocha
N
,
Payne
F
,
Huang-Doran
I
, et al
.
The metabolic syndrome-associated small G protein ARL15 plays a role in adipocyte differentiation and adiponectin secretion
.
Sci Rep
2017
;
7
:
17593
12.
Bonnet
E
,
Delpierre
C
,
Sommet
A
, et al
.
Total body composition by DXA of 241 HIV-negative men and 162 HIV-infected men: proposal of reference values for defining lipodystrophy
.
J Clin Densitom
2005
;
8
:
287
292
13.
Freitas
P
,
Santos
AC
,
Carvalho
D
, et al
.
Fat mass ratio: an objective tool to define lipodystrophy in HIV-infected patients under antiretroviral therapy
.
J Clin Densitom
2010
;
13
:
197
203
14.
Freitas
P
,
Carvalho
D
,
Santos
AC
, et al
.
Lipodystrophy defined by fat mass ratio in HIV-infected patients is associated with a high prevalence of glucose disturbances and insulin resistance
.
BMC Infect Dis
2012
;
12
:
180
15.
Ajluni
N
,
Meral
R
,
Neidert
AH
, et al
.
Spectrum of disease associated with partial lipodystrophy: lessons from a trial cohort
.
Clin Endocrinol (Oxf)
2017
;
86
:
698
707
16.
Valerio
CM
,
Godoy-Matos
A
,
Moreira
RO
, et al
.
Dual-energy X-ray absorptiometry study of body composition in patients with lipodystrophy
.
Diabetes Care
2007
;
30
:
1857
1859
17.
Valerio
CM
,
Zajdenverg
L
,
de Oliveira
JEP
,
Mory
PB
,
Moyses
RS
,
Godoy-Matos
AF
.
Body composition study by dual-energy X-ray absorptiometry in familial partial lipodystrophy: finding new tools for an objective evaluation
.
Diabetol Metab Syndr
2012
;
4
:
40
18.
Guillín-Amarelle
C
,
Sánchez-Iglesias
S
,
Castro-Pais
A
, et al
.
Type 1 familial partial lipodystrophy: understanding the Köbberling syndrome
.
Endocrine
2016
;
54
:
411
421
19.
Meral
R
,
Ryan
BJ
,
Malandrino
N
, et al
.
“Fat shadows” from DXA for the qualitative assessment of lipodystrophy: when a picture is worth a thousand numbers
.
Diabetes Care
2018
;
41
:
2255
2258
20.
Littlejohns
TJ
,
Holliday
J
,
Gibson
LM
, et al
.
The UK Biobank imaging enhancement of 100,000 participants: rationale, data collection, management and future directions
.
Nat Commun
2020
;
11
:
2624
21.
Vasandani
C
,
Li
X
,
Sekizkardes
H
,
Adams-Huet
B
,
Brown
RJ
,
Garg
A
.
Diagnostic value of anthropometric measurements for familial partial lipodystrophy, Dunnigan variety
.
J Clin Endocrinol Metab
2020
;
105
:
2132
2141
22.
Haas
ME
,
Pirruccello
JP
,
Friedman
SN
, et al
.
Machine learning enables new insights into genetic contributions to liver fat accumulation
.
Cell Genom
2021
;
1
:
100066
23.
Loh
P-R
,
Bhatia
G
,
Gusev
A
, et al.;
Schizophrenia Working Group of Psychiatric Genomics Consortium
.
Contrasting genetic architectures of schizophrenia and other complex diseases using fast variance-components analysis
.
Nat Genet
2015
;
47
:
1385
1392
24.
Agrawal
S
,
Wang
M
,
Klarqvist
MDR
, et al
.
Inherited basis of visceral, abdominal subcutaneous and gluteofemoral fat depots
.
Nat Commun
2022
;
13
:
3771
25.
Mbatchou
J
,
Barnard
L
,
Backman
J
, et al
.
Computationally efficient whole-genome regression for quantitative and binary traits
.
Nat Genet
2021
;
53
:
1097
1103
26.
Shungin
D
,
Winkler
TW
,
Croteau-Chonka
DC
, et al.;
ADIPOGen Consortium
;
CARDIOGRAMplusC4D Consortium
;
CKDGen Consortium
;
GEFOS Consortium
;
GENIE Consortium
;
GLGC
;
ICBP
;
International Endogene Consortium
;
LifeLines Cohort Study
;
MAGIC Investigators
;
MuTHER Consortium
;
PAGE Consortium
;
ReproGen Consortium
.
New genetic loci link adipose and insulin biology to body fat distribution
.
Nature
2015
;
518
:
187
196
27.
Locke
AE
,
Kahali
B
,
Berndt
SI
, et al.;
LifeLines Cohort Study
;
ADIPOGen Consortium
;
AGEN-BMI Working Group
;
CARDIOGRAMplusC4D Consortium
;
CKDGen Consortium
;
GLGC
;
ICBP
;
MAGIC Investigators
;
MuTHER Consortium
;
MIGen Consortium
;
PAGE Consortium
;
ReproGen Consortium
;
GENIE Consortium
;
International Endogene Consortium
.
Genetic studies of body mass index yield new insights for obesity biology
.
Nature
2015
;
518
:
197
206
28.
Pulit
SL
,
Stoneman
C
,
Morris
AP
, et al.;
GIANT Consortium
.
Meta-analysis of genome-wide association studies for body fat distribution in 694 649 individuals of European ancestry
.
Hum Mol Genet
2019
;
28
:
166
174
29.
Loh
NY
,
Minchin
JEN
,
Pinnick
KE
, et al
.
RSPO3 impacts body fat distribution and regulates adipose cell biology in vitro
.
Nat Commun
2020
;
11
:
2797
30.
Graham
SE
,
Clarke
SL
,
Wu
K-HH
, et al.;
VA Million Veteran Program
;
Global Lipids Genetics Consortium
.
The power of genetic diversity in genome-wide association studies of lipids
.
Nature
2021
;
600
:
675
679
31.
Justice
AE
,
Karaderi
T
,
Highland
HM
, et al.;
CHD Exome+ Consortium
;
Cohorts for Heart and Aging Research in Genomic Epidemiology (CHARGE) Consortium
;
EPIC-CVD Consortium
;
ExomeBP Consortium
;
Global Lipids Genetic Consortium
;
GoT2D Genes Consortium
;
InterAct
;
ReproGen Consortium
;
T2D-Genes Consortium
;
MAGIC Investigators
.
Protein-coding variants implicate novel genes related to lipid homeostasis contributing to body-fat distribution
.
Nat Genet
2019
;
51
:
452
469
32.
Park
J
,
Kim
M
,
Sun
K
,
An
YA
,
Gu
X
,
Scherer
PE
.
VEGF-A-expressing adipose tissue shows rapid beiging and enhanced survival after transplantation and confers IL-4–independent metabolic improvements
.
Diabetes
2017
;
66
:
1479
1490
33.
Sun
K
,
Wernstedt Asterholm
I
,
Kusminski
CM
, et al
.
Dichotomous effects of VEGF-A on adipose tissue dysfunction
.
Proc Natl Acad Sci U S A
2012
;
109
:
5874
5879
34.
Privé
F
,
Arbel
J
,
Vilhjálmsson
BJ
.
LDpred2: better, faster, stronger
.
Bioinformatics
2021
;
36
:
5424
5431
35.
International HapMap 3 Consortium; Altshuler DM, Gibbs RA, Peltonen L, et al. Integrating common and rare genetic variation in diverse human populations. Nature
2010
;
467
:
52
58
36.
Vantyghem
MC
,
Vincent-Desplanques
D
,
Defrance-Faivre
F
, et al
.
Fertility and obstetrical complications in women with LMNA-related familial partial lipodystrophy
.
J Clin Endocrinol Metab
2008
;
93
:
2223
2229
37.
Joy
TR
,
Hegele
RA
.
Prevalence of reproductive abnormalities among women with familial partial lipodystrophy
.
Endocr Pract
2008
;
14
:
1126
1132
38.
Bui
HN
,
Sluss
PM
,
Hayes
FJ
, et al
.
Testosterone, free testosterone, and free androgen index in women: Reference intervals, biological variation, and diagnostic value in polycystic ovary syndrome
.
Clin Chim Acta
2015
;
450
:
227
232
39.
Chiquette
E
,
Oral
EA
,
Garg
A
,
Araújo-Vilar
D
,
Dhankhar
P
.
Estimating the prevalence of generalized and partial lipodystrophy: findings and challenges
.
Diabetes Metab Syndr Obes
2017
;
10
:
375
383
40.
Foss-Freitas
MC
,
Akinci
B
,
Neidert
A
, et al
.
Selective targeting of angiopoietin-like 3 (ANGPTL3) with vupanorsen for the treatment of patients with familial partial lipodystrophy (FPLD): results of a proof-of-concept study
.
Lipids Health Dis
2021
;
20
:
174
41.
Danforth
E
Jr
.
Failure of adipocyte differentiation causes type II diabetes mellitus
?
Nat Genet
2000
;
26
:
13
42.
Udler
MS
,
Kim
J
,
von Grotthuss
M
, et al
.
Type 2 diabetes genetic loci informed by multi-trait associations point to disease mechanisms and subtypes: a soft clustering analysis
.
PLoS Med
2018
;
15
:
e1002654
43.
Deaton
AM
,
Dubey
A
,
Ward
LD
, et al.;
AMP-T2D-GENES Consortium
.
Rare loss of function variants in the hepatokine gene INHBE protect from abdominal obesity
.
Nat Commun
2022
;
13
:
4319
44.
Akbari
P
,
Sosina
OA
,
Bovijn
J
, et al.;
Regeneron Genetics Center
;
DiscovEHR Collaboration
.
Multiancestry exome sequencing reveals INHBE mutations associated with favorable fat distribution and protection from diabetes
.
Nat Commun
2022
;
13
:
4844
45.
Ng
BK
,
Sommer
MJ
,
Wong
MC
, et al
.
Detailed 3-dimensional body shape features predict body composition, blood metabolites, and functional strength: the Shape Up! studies
.
Am J Clin Nutr
2019
;
110
:
1316
1326
46.
Majmudar
MD
,
Chandra
S
,
Yakkala
K
, et al
.
Smartphone camera based assessment of adiposity: a validation study
.
NPJ Digit Med
2022
;
5
:
79
47.
Klarqvist
MDR
,
Agrawal
S
,
Diamant
N
, et al
.
Silhouette images enable estimation of body fat distribution and associated cardiometabolic risk
.
NPJ Digit Med
2022
;
5
:
105
48.
Sudlow
C
,
Gallacher
J
,
Allen
N
, et al
.
UK Biobank: an open access resource for identifying the causes of a wide range of complex diseases of middle and old age
.
PLoS Med
2015
;
12
:
e1001779
49.
O’Connor
L
,
Brage
S
,
Griffin
SJ
,
Wareham
NJ
,
Forouhi
NG
.
The cross-sectional association between snacking behaviour and measures of adiposity: the Fenland Study, UK
.
Br J Nutr
2015
;
114
:
1286
1293
50.
Said
MA
,
Verweij
N
,
van der Harst
P
.
Associations of combined genetic and lifestyle risks with incident cardiovascular disease and diabetes in the UK Biobank study
.
JAMA Cardiol
2018
;
3
:
693
702
Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. More information is available at https://www.diabetesjournals.org/journals/pages/license.