ResourceProfound Perturbation of the Metabolome in ObesityIs Associated with Health RiskGraphical AbstractHighlightsd Obesity results in a profound perturbation of the plasmametabolomed At any given BMI, abnormal metabolomes associate withdifferent health outcomesd At any given BMI, different genetic obesity risks do notchange the metabolomed Ametabolome signature effectively tracks changes in obesityCirulli et al., 2019, Cell Metabolism 29, 488–500February 5, 2019 ª 2018 The Author(s). Published by Elsevier Inhttps://doi.org/10.1016/j.cmet.2018.09.022AuthorsElizabeth T. Cirulli, Lining Guo,Christine Leon Swisher, ...,Bernard Thorens, J. Craig Venter,Amalio TelentiCorrespondenceliz.cirulli@gmail.com (E.T.C.),atelenti@scripps.edu (A.T.)In BriefObesity is a heterogeneous and complexdisease that is imprecisely measured byBMI. Cirulli et al. used non-targetedmetabolomics and whole-genomesequencing to identify metabolic andgenetic signatures of obesity and find thatthe metabolome captures clinicallyrelevant phenotypes of obesity and is abetter health predictor than genetic risk.c.Cell MetabolismResourceProfound Perturbation of the Metabolomein Obesity Is Associated with Health RiskElizabeth T. Cirulli,1,8,* LiningGuo,2 Christine Leon Swisher,1 Naisha Shah,1 Lei Huang,1 Lori A. Napier,1 Ewen F. Kirkness,1Tim D. Spector,3 C. Thomas Caskey,4 Bernard Thorens,5 J. Craig Venter,6 and Amalio Telenti7,*1Human Longevity, Inc., San Diego, CA 92121, USA2Metabolon, Inc., Durham, NC, USA3Department of Twin Research and Genetic Epidemiology, King’s College London, London, UK4Molecular and Human Genetics, Baylor College of Medicine, Houston, TX, USA5Center for Integrative Genomics, University of Lausanne, Lausanne, Switzerland6J. Craig Venter Institute, La Jolla, CA 92037, USA7The Scripps Research Institute, La Jolla, CA 92037, USA8Lead Contact*Correspondence: liz.cirulli@gmail.com (E.T.C.), atelenti@scripps.edu (A.T.)https://doi.org/10.1016/j.cmet.2018.09.022SUMMARYObesity is a heterogeneous phenotype that is crudelymeasured by body mass index (BMI). There is aneed for a more precise yet portable method ofphenotyping and categorizing risk in large numbersof people with obesity to advance clinical care anddrug development. Here, we used non-targetedmetabolomics and whole-genome sequencing toidentify metabolic and genetic signatures of obesity.We find that obesity results in profound perturbationof the metabolome; nearly a third of the assayedmetabolites associated with changes in BMI. Ametabolome signature identifies the healthy obeseand lean individuals with abnormal metabolomes—these groups differ in health outcomes and underly-ing genetic risk. Specifically, an abnormal metabo-lome associated with a 2- to 5-fold increase incardiovascular events when comparing individualswho were matched for BMI but had opposing metab-olome signatures. Because metabolome profilingidentifies clinically meaningful heterogeneity inobesity, this approach could help select patientsfor clinical trials.INTRODUCTIONObesity is one of the most widespread problems facing our so-ciety’s health today. Excessive weight significantly increasesrisk for conditions like diabetes mellitus and cardiovascular dis-ease (Hales et al., 2017; Whitlock et al., 2009). Worldwide, theprevalence of obesity has nearly tripled since 1975, with 39%of the world’s adults being overweight and 13% being obese(WHO, 2018). The high prevalence can partially be attributed toincreasing consumption of hypercaloric foods and sedentarylifestyles (WHO, 2018). Previous studies have identified meta-bolic signatures associated with obesity, including increasedlevels of branched-chain and aromatic amino acids as well asglycerol and glycerophosphocholines (Butte et al., 2015; Chenet al., 2015; Ho et al., 2016; Menni et al., 2017; Park et al.,2015; Piening et al., 2018). However, prior work has been limitedby focusing on a relatively small numbers of metabolites, individ-uals, or obesity phenotypes.The characterization of the metabolites that are associatedwith obesity can provide insights into the mechanisms thatlead to this disease and associated consequences. Longitudinalassessment of weight gain and weight loss over time may indi-cate whether there are metabolomic changes that causeobesity—meaning that current metabolite levels could predictfuture weight changes—or whether all metabolite changes asso-ciated with obesity are a consequence of weight changes. Draw-ing genetics into this assessment allows the determination ofwhether genetic variation leads to metabolite changes that sub-sequently result in obesity, allowing the further delineation of thecausal pathway to obesity. Finally, research in this area mayidentify biomarkers of obesity and of different types of obesity,for example biomarkers of so-called healthy obesity (Neelandet al., 2018).There are recent calls to improve phenotyping in very largenumbers of obese people with the goals of understanding fac-tors that make people susceptible to (or protected from)obesity, accompanied by a better elucidation of the factorsthat account for variability in success of different obesity treat-ments (Yanovski and Yanovski, 2018). Here, in an effort to un-derstand the relationship between metabolic perturbations andthe obese state, we analyzed 2,396 individuals with longitudinalmeasurements of body mass index (BMI), anthropomorphicdata, whole-body DEXA scans, and metabolome, combinedwith baseline genetic risk. The metabolome assay covered upto 1,007 metabolites at up to three distinct time points foreach individual over the course of study. We identified associ-ations between nearly a third of the metabolome and BMI, andwe show that metabolite levels can predict obesity status with80%–90% specificity and sensitivity. The metabolome profileis a strong indicator of metabolic health compared to the poly-genic risk assessment and anthropomorphic measurementsof BMI.488 Cell Metabolism 29, 488–500, February 5, 2019 ª 2018 The Author(s). Published by Elsevier Inc.This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).RESULTSProfound Perturbation of Metabolome by ObesityMetabolites Associated with BMIWe compared the levels of individual metabolites to the BMIs of832, 882, and 861 unrelated individuals of European ancestry inthe TwinsUK cohort (Moayyeri et al., 2013) at three time pointsspanning a total range of 8–18 years. We identified 284 metabo-lites that were significantly associated (p < 5.5 3 105) with BMIat one ormore time points (Table S1).We focused on 110metab-olites that were significantly associated with BMI at all 3 timepoints and sought to replicate the associations in an indepen-dent sample of 427 unrelated individuals of European ancestryparticipating in the Health Nucleus cohort (Perkins et al., 2017).In total, our analyses identified 307 metabolites that were signif-icantly associated with BMI in at least one cohort and time point(Table S1). We identified 83 metabolites that showed directionsof effect that were consistent between the two cohorts, of which49 were statistically significant replications (Figure 1; Table 1).While this set of 49 metabolites was the most stringently associ-ated with BMI, the majority of the implicated metabolites (292 of307, 95%) had directions of effect that were consistent betweentime points/cohorts, indicating that many of the remaining me-tabolites may reach our stringent cutoffs in a larger study.The 49 metabolites that associated with BMI were primarilylipids (n = 23, accounting for 7.5% of all lipids assayed acrossboth cohorts) and amino acids (n = 14, 9.3% of all amino acids)but also included nucleotides (n = 3, 12.0% of all nucleotides),peptides (n = 3, 12% of all peptides), and other categories(n = 6; Figure 1; Table 1). The most significantly associatedmetabolite was urate (uric acid; p value 1.23 1040 for combinedanalysis of TwinsUK time point 1 and Health Nucleus data).Patterns in Metabolite Change According to BMIThe majority of the 49 BMI-associated metabolites increasedwith increasing BMI (n = 35) (Figure 1; Table S1). This includedglucose and, notably, mannose, which has recently been high-lighted as playing a role in insulin resistance (Lee et al., 2016).Most metabolites change linearly (both proportionally andinversely) with BMI (Figure 1C). Branched-chain and aromaticamino acids as well as metabolites related to nucleotide meta-bolism like urate had the most rapid increases. Those thatdecreased (n = 14) included phospholipids and lysolipids, aswell as the amino acids asparagine and N-acetylglycine andthe xenobiotic cinnamoylglycine, which has been identified asa product of the microbiome (Wikoff et al., 2009). The negativelyassociated lipids tended to reflect HDL (high-density lipoprotein)levels, while the positively correlated lipids were more represen-tative of triglyceride levels (Tables 1 and S1). Of particularinterest was the association between BMI and cortisone, ametabolite of the steroid hormone cortisol. We identified lowerlevels among the obese individuals, which is consistent with pre-vious reports (Bjo¨rntorp and Rosmond, 2000; Praveen et al.,2011; Walker et al., 2000; Wirix et al., 2017). We examined theoverall composition of the distributions of these metabolites viaFigure 1. Pathway Categories of Metabolites Associated with BMI(A and B) Shown are the pathway categories of (A) the 307 metabolites significantly associated with BMI and (B) the 49-metabolite signature.(C) The values of each of the 49 BMI-associatedmetabolites are plotted with a Loess curve against the BMI for time point 1 in TwinsUK. Only unrelated individualsof European ancestry are included, and the small number of individuals with BMI below 20 (n = 31) or above 40 (n = 10) are removed to keep the ends of the graphsfrom being skewed. The apparent inversion of the relationship between one cofactor/vitamin metabolite and BMI at higher BMIs is an artifact that is correctedonce morbidly obese individuals are included.Cell Metabolism 29, 488–500, February 5, 2019 489Table 1. Metabolite Signature Associated with BMISuper Pathway MetaboliteSubpathway (CorrelatedBlood Lipidsc)Direction of Effect(Ranka) BMI r2bNucleotide urate Pur.Met. [ (1) 16.4%N2,N2-dimethylguanosine Pur.Met. [ (6) 8.8%N6-carbamoylthreonyladenosine Pur.Met. [ (28) 7.3%Amino acid glutamate Glu.Met. [ (2) 11.5%N-acetylglycine Gly.Met. Y (9) 9.0%5-methylthioadenosine (MTA) Poly.Met. [ (10) 7.5%valine Leu.Met. [ (11) 8.8%aspartate Ala.Met. [ (16) 7.0%N-acetylvaline Leu.Met. [ (18) 7.3%kynurenate Try.Met. [ (19) 6.0%alanine Ala.Met. [ (23) 5.3%asparagine Ala.Met. Y (26) 3.7%N-acetylalanine Ala.Met. [ (31) 6.6%tyrosine Phe.Met. [ (34) 1.8%leucine Leu.Met. [ (37) 6.8%N-acetyltyrosine Phe.Met. [ (40) 4.2%2-methylbutyrylcarnitine (C5) Leu.Met. [ (41) 8.3%Lipid 1-(1-enyl-palmitoyl)-2-oleoyl-GPC Plas. (HDL, TG) Y (3) 7.1%1-stearoyl-2-dihomo-linolenoyl-GPC Phos.Met. (TG, Chol) [ (4) 9.8%1-eicosenoyl-GPC Lysolipid Y (5) 6.2%1-arachidoyl-GPC Lysolipid Y (7) 8.6%1-(1-enyl-stearoyl)-2-oleoyl-GPC Phos.lip. (HDL) Y (8) 6.5%propionylcarnitine BCAA.Met [ (12) 9.9%1-nonadecanoyl-GPC Lysolipid Y (14) 4.2%1-linoleoyl-GPC Lysolipid Y (15) 4.9%sphingomyelin Sph.Met. (Chol) [ (20) 6.8%1-palmitoyl-2-dihomo-linolenoyl-GPC Phos.Met. (TG, Chol) [ (21) 5.1%1-(1-enyl-palmitoyl)-2-linoleoyl-GPC Phos.Met. (HDL) Y (22) 5.7%1-palmitoyl-3-linoleoyl-glycerol Phos.Met. (TG) [ (24) 7.6%1-oleoyl-2-linoleoyl-GPC Phos.Met. Y (27) 5.6%1-(1-enyl-stearoyl)-2-docosahexaenoyl-GPC Phos.Met. Y (29) 2.5%1-oleoyl-3-linoleoyl-glycerol Di.gly. (TG, HDL) [ (30) 6.3%carnitine Car.Met. [ (33) 7.5%1-palmitoyl-2-linoleoyl-glycerol Phos.Met. (TG, HDL) [ (36) 7.2%1-oleoyl-2-linoleoyl-glycerol Di.gly. (TG, HDL) [ (38) 5.9%1,2-dilinoleoyl-GPC Phos.Met. Y (39) 4.2%1-palmitoleoyl-2-oleoyl-glycerol Phos.lip. (TG) [ (42) 5.6%1-palmitoleoyl-3-oleoyl-glycerol Phos.lip. (TG) [ (45) 6.0%1-palmitoyl-2-adrenoyl-GPC Phos.Met. (TG) [ (47) 2.9%cortisone Plas. (HDL, TG) Y (49) 2.5%Energy succinylcarnitine TCA [ (13) 9.8%Carbohydrate mannose Fru.Met. [ (17) 6.6%glucose Pyr.Met. [ (48) 6.3%Xenobiotics cinnamoylglycine Food Y (43) 3.5%Cofactors/ vitamins gulonic acid Asc.Met. [ (46) 3.2%quinolinate Nic.Met. [ (44) 8.4%(Continued on next page)490 Cell Metabolism 29, 488–500, February 5, 2019principal component analysis and found complex underlyingcorrelations; in particular, the first principal component ex-plained 20% of the total variation in the levels of these 49 me-tabolites.We additionally examined each of the 49metabolites injust those of normal weight, overweight, or obese separately,and we found directions of effect that were largely consistentwith those seen in the group as a whole (Table S1).Modeling the Metabolome of ObesityWe used ridge regression to build a model that would predictBMI from the 49 BMI-associated metabolites (Figure 2). Wechose this method to focus on the most stringently associatedmetabolites and to remove effects of collinearity, and we foundsimilar results when using lasso regression. We combined ourdata for the first visit of the TwinsUK cohort and the HealthNucleus cohort and trained with 10-fold cross-validation on arandom half of the population. In our test set of the other halfof the data, we found that the model could explain 39.1% ofthe variation in BMI (Figure 2A). In predicting whether partici-pants were obese (BMI R 30) or normal weight (BMI 18.5–25), the model had an area under the curve (AUC) of 0.922,specificity of 89.1%, and sensitivity of 80.2% (Figure S1). Themodel based on the metabolite signature was thereafter usedas a tool to define mBMI, the predicted BMI on the basis ofmetabolome.Richer models using the full set of available metabolites(n = 650 measured in both cohorts) improved the accuracy ofthe model (47%–49% of the variance explained) and could beconsidered as the optimal approach by accepting the additionalcost of a full untargeted metabolome as compared to the moretargeted panel of 49 metabolites. This performance shouldbe contrasted with the results of models using routine clinicallaboratory determinations: regression analysis predicting BMIfrom age, sex, HDL, LDL, total cholesterol, and total triglyceridesexplained 31% of the variation in BMI, whereas a model usingage, sex, and the 49-metabolite signature explained 43%of the variation. In fact, even though mBMI was modeled bytraining on BMI, thismetabolite signature had a better correlationthan BMI with most health-related phenotypes measured here(Table S2).Identification and Characterization of Metabolic BMIOutliersHaving established a model to predict BMI using the metabo-lome (mBMI), we split the participants into five groups (Fig-ure 2A). Three groups included individuals whose metabolomeaccurately predicted their BMI, as defined by having a residualbetween 0.5 and 0.5 from a regression analysis of mBMIwith age, sex, and BMI included as predictors. These criteriadelineated 80% of individuals as having an mBMI relativelyconcordant with actual BMI. They were characterized as havinga normal BMI (18.5–25), overweight (25–30), or obese (>30). Twogroups were characterized as outliers: these included individ-uals whose metabolome predicted a substantially lower mBMIthan the actual BMI (mBMI << BMI, residual < 0.5) or a sub-stantially higher mBMI than the actual BMI (mBMI >> BMI, resid-ual > 0.5). While these two outlier groups had the same weightand age distribution (Figure 2B), they had very different valuesfor many of the phenotypes of metabolic health collected fromthese cohorts (Figures 2B and 2C). Individuals with an mBMIprediction that was substantially lower than their actual BMIhad levels of insulin resistance, total triglycerides, HDL, bloodpressure, waist/hip ratio, android/gynoid ratio, percent bodyfat, percent visceral fat, and percent subcutaneous fat thatwere similar to normal-weight individuals with healthy metabo-lomes. Individuals with an mBMI prediction that was substan-tially higher than their actual BMI had levels for these traitsthat were similar to those of obese individuals with obese me-tabolomes. Evaluating these data from a more clinical perspec-tive, with individuals separated into clinical categories such asnormal BMI with obese metabolome and obese BMI withhealthy metabolome, generally confirmed these effects (Fig-ures 3 and S2). Our findings suggest that the metabolome canbe used as a clinically meaningful instrument, where obesity isanalyzed in the context of its metabolome perturbation ratherthan just on BMI alone. Thus, our results are important in theframe of the current debate on the metabolically ‘‘healthy’’obese and also for the identification of individuals with normalBMI (Chen et al., 2015) but poor metabolic health (Caleyachettyet al., 2017; Iglesias Molli et al., 2017).Table 1. ContinuedSuper Pathway MetaboliteSubpathway (CorrelatedBlood Lipidsc)Direction of Effect(Ranka) BMI r2bPeptide N-acetylcarnosine Dipep. [ (25) 6.9%gamma-glutamylphenylalanine Gam. [ (32) 6.0%gamma-glutamyltyrosine Gam. [ (35) 4.6%Subpathway acronyms: Ala.Met., alanine and aspartate metabolism; Asc.Met., ascorbate and aldarate metabolism; BCAA.Met, branched-chainamino acid metabolism; Car.Met., carnitine metabolism; Di.gly., diacylglycerol; Dipep., dipeptide derivative; Food, food component/plant; Fru.Met.,fructose, mannose, and galactose metabolism; Gam., gamma-glutamyl amino acid; Glu.Met., glutamate metabolism; Gly.Met., glycine, serine, andthreonine metabolism; Leu.Met., leucine, isoleucine, and valine metabolism; Lysolipid, lysolipid; Nic.Met., nicotinate and nicotinamide metabolism;Phe.Met., phenylalanine and tyrosine metabolism; Phos.lip., phospholipid; Phos.Met., phospholipid metabolism; Plas., plasmalogen; Poly.Met., poly-amine metabolism; Pur.Met., purine metabolism; Pyr.Met., glycolysis, gluconeogenesis, and pyruvate metabolism; Sph.Met., sphingolipid meta-bolism; Steroid, steroid; TCA, TCA cycle; Try.Met., tryptophan metabolism.aRank indicates order of significance of association with BMI.bMean r2 indicates the percent variation in BMI explained by each metabolite in univariate analysis for a combined analysis of the first time point of theTwinsUK cohort and the Health Nucleus data.cBlood labs for TG (triglycerides), Chol (cholesterol), HDL (high-density lipoprotein), or LDL (low-density lipoprotein) that had an r2 > 0.1 with themetab-olite are indicated in parentheses.Cell Metabolism 29, 488–500, February 5, 2019 491Having characterized these outliers, we revisited their metab-olome differences. As the 49 metabolites were chosen fortheir association with BMI, there is no a priori expectationthat they would all show differences between the group withmBMI >> BMI and the group with mBMI << BMI, as thesetwo groups have the same BMI distribution. However, we foundthat these two outlier groups did indeed have statistically signif-icant differences in their metabolite levels for all of the 49metabolites but two: asparagine and cortisone (Figure S3).This lack of association of cortisone with mBMI/BMI outliersshows that its correlation with BMI does not appear to extendto metabolic health, which presents interesting questions aboutthe underlying biology. Although cortisone’s parent compoundcortisol has been reported as associated with obesity, there areinconsistent relationships between cortisol and metabolic pa-rameters in the literature (Abraham et al., 2013; IncollingoRodriguez et al., 2015). It is also known that cortisol levelschange throughout the day, and this factor could have influ-enced our results as the participants were not all measuredat the same time. We additionally investigated the associationbetween each of the BMI-associated metabolites and insulinresistance, as many previously reported markers of obesityhave also been markers of diabetes (Ho et al., 2016; Parket al., 2015). We had quantitative insulin resistance measure-ments for 515 unrelated, European-ancestry participants. Aftercontrolling for BMI, we found that 12 of the 49 BMI-associatedmetabolites were also significantly associated (correcting for49 tests requires p < 0.001) with insulin resistance, all with pos-itive directions of effect: tyrosine, alanine, kynurenate, gamma-glutamyltyrosine, 1-oleoyl-3-linoleoyl-glycerol (18:1/18:2), sixphospholipids, and, as expected, glucose (Table S1). Mannose,which recently underwent extensive study with regard toinsulin resistance (Lee et al., 2016), was nominally associatedwith insulin resistance after controlling for BMI in our study(p = 0.004).Evolution of Obesity and Metabolome Clinical ProfilesGiven recent work suggesting that obese individuals who aremetabolically healthy may remain at higher risk of negativehealth outcomes than are normal weight individuals who aremetabolically healthy (Caleyachetty et al., 2017), we next askedwhether the outlier groups were more likely to become obeseover time. Focusing on the 1,458 individuals from TwinsUKwho had weight measurements at all three time points, we foundthat those who had an mBMI that was higher than their BMIwere marginally more likely to gain weight and convert to anobese phenotype (BMI > 30) over the 8–18 years of follow up.For example, 32.8% of those of normal weight but with an over-weight or obese metabolome converted to being overweight orobese by time point 3 compared to 24.8% of those who were ofnormal weight and had a healthy metabolome (p = 0.01; Fig-ure 4). However, this association was not strongly supported,and overall, the mBMI states of the individuals remained fairlystable with time and were a function of BMI changes (Figure 4).For example, 68% of the individuals who began the studywith an obese metabolome ended the study with an obese me-tabolome. When an individuals’ weight increased and thendecreased, their mBMI followed suit, and no single metabolitewas significantly predictive of subsequent BMI changes. Insummary, our results are consistent with a favorable long-termhealth benefit for the overweight and obese individuals with ahealthy metabolome.Cardiovascular Disease OutcomesObesity is considered a risk factor for cardiovascular diseaseand ischemic stroke (Rhee, 2018). In the TwinsUK study, 32 of1,573 individuals reported having had a cardiovascular event(myocardial infarct, angina, or angioplasty) or stroke prior tobaseline. We found significant differences between mBMI/BMIgroups in their reported baseline cardiovascular events (Fisher’sexact p = 0.01). In particular, the mBMI << BMI outliers had nohistory of cardiovascular events, whereas the mBMI >> BMI out-liers had a proportion of historical events that was similar tothose of overweight and obese individuals (Figure 4C). Informa-tion about blood pressure medication use was also available for379 individuals at baseline, and those with mBMI >> BMI hadvery high rates of use compared to mBMI << BMI (Fisher’s exactp = 0.00003; Figure 4C). The longitudinal nature of the TwinsUKstudy also allowed the collection of clinical endpoints in theseunselected participants. The age of participants at the first visitranged from 33 to 74 years old (median 51), and 42 to 88 yearsold (median 65) at the last visit. During the follow up (median13 years), the study recorded 53 cardiovascular events (myocar-dial infarct, angina, or angioplasty) or strokes for 1,573 individ-uals. We calculated that our study had 80% power to identifyeffects with a hazard ratio of at least 1.5 for differences in cardio-vascular event outcomes between the different mBMI/BMIgroups. We found that participants with a healthy metabolome(normal BMI, overweight, or obese) had 2.6 events per hundredindividuals. Individuals with an obese metabolic profile, mBMI,had 3.4 (normal/overweight BMI) and 4.4 events (in obese indi-viduals) per hundred individuals. Separated analysis of thevarious endpoints confirmed the trends, more accentuated forcardiovascular than for diagnosis of stroke. We then performeda formal survival analysis for participants to have any cardiovas-cular event after the first time point, and we found thosewith healthier metabolomes to have fewer/later cardiac events(p = 0.003; Figure 4D).Correlations between TwinsBecause twin studies are important to analyze the heritability oftraits, we reassessed the BMI model predictions and obesitystatus of 350 sets of twins where either both twins had normalBMI (n = 244), both twins were obese (n = 67), or one was obeseand the other had normal BMI (n = 39). To keep the categoriesclear, individuals with BMIs between 25 and 30 (overweight)and their twins were excluded. As asserted by the model’shigh specificity and sensitivity, the metabolite-based obesitypredictions reflected the actual obesity status of the individuals.This was even the case when only one twin was obese: theobese twin was generally predicted by their metabolome tobe obese, while the normal weight twin was not (Figure S1).The correlations between the metabolite-based obesity predic-tions were also substantially higher between the monozygotictwins than the dizygotic twins, as expected. Interestingly, weidentified three sets of twins where both twins were predictedfrom the metabolome to be of normal weight, but bothwere obese, and eight sets of twins where the reversewas true. These outliers were thought to represent healthyobese and normal weight, metabolically unhealthy individualsdescribed above.492 Cell Metabolism 29, 488–500, February 5, 2019Figure 2. Variables Associated with BMI and Predicted BMI from the Metabolome(A) Correlation between ridge regression model prediction of BMI and actual BMI for all unrelated individuals of European ancestry in the TwinsUK and HealthNucleus dataset. The identification of outliers is defined below: the pink box shows individuals with a much lower predicted BMI (mBMI) than actual BMI, and theyellow box shows individuals with a much higher mBMI than actual BMI.(B) Factors associatedwith being anmBMI outlier. Participants were split into five groups: those whosemetabolome accurately predicted their BMI (residual afteraccounting for age, sex, and BMI between 0.5 and 0.5) whose BMIs were either normal (18.5–25), overweight (25–30), or obese (>30), and those whosemetabolome predicted a substantially higher mBMI than the actual BMI (residual < 0.5) or a substantially lower mBMI than the actual BMI (residual > 0.5). Ally axis values are scaled to a range from 0 to 1 to allow comparison across groups.(legend continued on next page)Cell Metabolism 29, 488–500, February 5, 2019 493Genetic AnalysesKnown Genetics of ObesityWe first investigated the known genetic factors contributing tohigh BMI. We calculated polygenic risk scores for BMI usingknown associations from the considerable literature of obesityand BMI GWAS (genome-wide association study) (Locke et al.,2015). As previously reported, we found that polygenic risk scoreonly explained 1.2%–2.2% of the variation in BMI at each of thethree TwinsUK time points and in Health Nucleus for unrelatedparticipants of European ancestry (Figure S4). We investigatedwhether unique individuals with the highest polygenic risk scoreswould have a significant perturbation of the metabolome andanthropomorphic, insulin resistance, and DEXA measurements(Figure 5). While the data did not support a strong role for poly-genic risk, there were trends for higher polygenic risk scores tobe associated with a higher android/gynoid ratio (p = 0.04),waist/hip ratio (p = 0.03), and triglycerides (p = 0.01). However,there was no statistical association between the polygenic scoreand mBMI (p = 0.07).Studies of rare variants in obesity have identified melanocortin4 receptor (MC4R) mutations as having effects large and clearenough to be appropriate for study in our dataset (Collet et al.,2017; Turcot et al., 2018). We therefore identified members ofthe study populations who were carrying rare (MAF < 0.01%)coding variants in MC4R. We identified eight such carriers inthe subset of unrelated participants (Table 2). Each variant wasobserved in one unrelated individual, and six of the eight hadalready been annotated as causing obesity in clinical databasesHGMDor ClinVar (Table 2). As a group,MC4R carriers had signif-icantly higher BMI (p = 0.02) than did non-carriers as well as non-significant trends toward a higher diastolic blood pressure,insulin resistance, and percent body fat (Figure 5). However, itis known that not all rare variants are deleterious, and the meta-bolic impact could have been greater for the true subset of func-tional variants. The BMI data in the participants supported apathogenic role for at least five of the variants (Met292fs,Arg236Cys, Ser180Pro, Ala175T, and Thr11Ala), but did notcorroborate a role of Ile170V, which is defined in HGMD, ClinVar,and the literature as pathogenic (Cle´ment et al., 2018; Colletet al., 2017; Landrum et al., 2018; Stenson et al., 2003). Impor-tantly, of the five sets of twins who both carried the sameMC4R variant, three sets included twins who were both over-weight or obese. In the two cases where a carrier’s twin didnot have theMC4R variant, their BMI was lower than their twin’s.(C) The same process is used to show DEXA imaging values associated with metabolic BMI outliers. The mBMI >> BMI and mBMI << BMI groups had acomparable measured BMI and age; however, these two groups were statistically significantly different from each other (p < 0.01) for all modalities except bloodpressure, LDL, total cholesterol, and polygenic risk score. Additionally, the mBMI >> BMI group was statistically significantly different from normal weight,metabolically healthy people (p < 0.01) for all traits (except LDL), while mBMI << BMI individuals were only statistically different in diastolic blood pressure(p = 0.005). In contrast, outliers with mBMI << BMI were always statistically significantly different from obese, metabolically obese individuals (except systolicblood pressure, LDL, and total cholesterol), while mBMI >> BMI individuals were only statistically different in percent body fat (p = 4.43 106). LDL, low-densitylipoprotein; HDL, high-density lipoprotein. The correlations of each modality with mBMI can be found in Table S2.Whiskers of boxplot extend to the most extreme points no greater than 1.5 times the interquartile range (distance between the first and third quartiles).Figure 3. Body Composition Profiles from Dixon Magnetic Resonance Imaging for Four Outlier Individuals(A) Correlation between ridge regression model prediction of BMI and actual BMI for all unrelated individuals of European ancestry in the TwinsUK andHeath Nucleus dataset. Outliers highlighted in (B) and (C) are marked with corresponding colors. All individuals highlighted are from the outlier mBMI >> BMI ormBMI << BMI categories shown in Figure 2.(B) Body composition profiles (red, visceral adipose tissue; yellow, subcutaneous adipose tissue; cyan, muscle).(C) Waist to hip cross-sections (hip, mid femoral head; waist, top of ASIS).494 Cell Metabolism 29, 488–500, February 5, 2019We observed an enrichment of MC4R variant carriers amongobese individuals with low polygenic risk scores (Figure S4).Out of 37 participants whowere obesewith polygenic risk scoresin the lowest quartile, 5.4%wereMC4R variant carriers, while thecarrier frequency was just 0.4% in those of normal weight.We also searched the study population for rare variantsassociated with Mendelian lipodystrophy syndromes. Lipody-strophies are heterogeneous, genetic, or acquired disorderscharacterized by selective loss of body fat and associatedmetabolic complications such as diabetes mellitus, hypertrigly-ceridemia, and hepatic steatosis (Garg, 2011). We identified sixindividuals with rare heterozygous variants in four genes:ZMPSTE24, AGPAT2, LIPE, and BSCL2 (Table 2). The individ-uals with LIPE or BSCL2 variants were of normal weight butwere in the mBMI >> BMI outlier category. The three individualswith ZMPSTE24 p.Leu362fs caused by a 1-bp insertion hadnormal to low BMI andmBMI. We also observed an unannotated1-bp deletion at this site in one obese individual (Table 2; Fig-ure S5). While these observations are interesting, there are nodata in the literature to support a phenotypic role for these vari-ants in carriers (heterozygotes).Genetics of the Metabolically Healthy ObeseIndividuals with an mBMI that was substantially lower than theiractual BMI had a higher polygenic risk score for BMI than didother groups. In contrast, those whose mBMI was substantiallyhigher than their actual BMI had low polygenic risk scoresFigure 4. Progression of Different mBMI/BMI Categories(A) Alluvial plot showing the proportion of participants who remained in the same weight category or transitioned to a different weight category over the course ofthe 8–18 years of the TwinsUK study. Red individuals have an obesemetabolome, orange individuals have an overweight metabolome, and gray individuals havea normal metabolome.(B) Alluvial plot showing the proportion of participants who remained in the same mBMI category or transitioned to a different mBMI category over the course ofthe 8–18 years of the TwinsUK study. Red individuals begin the study with an obese BMI, orange overweight, and gray normal weight.(C) Bar graph showing baseline rates of stroke, infarction, angina, angioplasty, any of the above, and blood pressure medication use at baseline for TwinsUK (withcolors indicating groups as in D; n = 1,573 for information on cardiovascular and stroke events and 379 for information on blood pressure medication). Those withmBMI << BMI had none of these conditions at baseline except stroke, while those with mBMI >> BMI had rates similar to those of overweight and obese in-dividuals.(D) Survival plot showing age until cardiac event (infarction, angina, or angioplasty). The plot is divided into those whose mBMI corresponds with their BMI(normal weight (n = 598), overweight (n = 461), and obese (n = 192) categories) as well as the two outlier groups: those with mBMI << BMI (n = 158) and those withmBMI >> BMI (n = 164) (p = 0.003 for a difference between these categories in cardiovascular outcomes).Cell Metabolism 29, 488–500, February 5, 2019 495(Figure 2B; p = 0.08 for a difference between these two groups).This result would support the notion that the polygenic risk scorefor BMI may capture an anthropomorphic phenotype (larger-framed individuals) rather than a unique association with obesityas a disease trait.Genetics of Metabolome DifferencesLast, we investigated whether obese individuals with differentgenetic backgrounds had different metabolomes from otherobese individuals. We first searched for metabolites, BMI-asso-ciated or otherwise, that could distinguish individuals withdifferent BMI polygenic risk scores or MC4R variant carriers.Linear regression showed no significant associations betweenany single metabolites and polygenic risk orMC4R carrier statusin either the entire population or in only the obese individuals.This result implies that metabolites are unlikely to be intermedi-ate phenotypes that explain the underlying genetics of obesity.To check for more specific signals beyond the compiled poly-genic risk score, we also performed separate analyses ofeach of the 97 variants that are used to calculate the polygenicrisk score. We found no evidence for any of these knownGWAS variants to be more strongly associated with a metabolitethan with BMI itself, though our power for discovery was limitedgiven the very small effect sizes of most individual GWASvariants. In summary, although it is known that there is astrong genetic component to metabolite levels (Long et al.,2017), most of the metabolic perturbations that occur in theobese state are a response to obesity as opposed to shared ge-netic mechanisms.DISCUSSIONThe results of the present study highlight the profound disruptionof the metabolome in obesity and identify a metabolome signa-ture that serves to examine metabolic health without the limita-tions of anthropomorphic measurements or the cost, time, andequipment requirements of imaging technologies. Nearly one-third of the approximately 1,000 metabolites measured in thestudy were associated with BMI, and 49 were selected as aFigure 5. Genetic Risk Compared to BMI-Relevant Variables(A) Correlation between polygenic risk score (PG) category, MC4R carrier status, and BMI and anthropomorphic and clinical measurements for all unrelatedindividuals of European ancestry in the TwinsUK and HN dataset. All y axis values are scaled to a range from 0 to 1 to allow comparison across groups.(B) The same process is used to show bloodwork and DEXA imaging values.While there was a trend for genetic risk to be associated with variousmeasurements,the polygenic risk score achieved nominal p < 0.05 for BMI, waist/hip ratio, android/gynoid ratio, and triglycerides, and MC4R carrier status achieved nominalp < 0.05 for BMI. LDL, low-density lipoprotein; HDL, high-density lipoprotein.Whiskers of boxplot extend to the most extreme points no greater than 1.5 times the interquartile range (distance between the first and third quartiles).496 Cell Metabolism 29, 488–500, February 5, 2019Table 2. Variants Identified in MC4R or Lipodystrophy Genes in Participants of European AncestryGene Variant Protein Change Study MAFGlobalGnomad MAF Known Functional AnnotationCarrier BMI(mBMI)Carrier TwinBMI (mBMI)Non-carrier TwinBMI (mBMI)TwinZygo-sityObesityMC4R chr18:60371541 G/A p.Ser270Phe 0.04% 0.00% none 25.7 (26.1) 24.8 (24.7) N/A MZMC4R chr18:60372307 G/A p.Leu15Phe 0.04% 0.00% none 23 (22.6) 22.6 (22.4) N/A MZMC4R chr18:60371474 CA/C p.Met292fs 0.04% <0.003% (complete LOF)b 32.8 (26) N/A 28.8 (24.3) DZMC4R chr18:60371644 G/A p.Arg236Cys 0.04% 0.00% HGMD highC DM (not LOF)b 34.5 (32.3) 30 (27.5) N/A DZMC4R chr18:60372319 T/C p.Thr11Ala 0.04% <0.003% HGMD lowC DM (not LOF)b 36 (31.7) N/A N/A N/AMC4R chr18:60371812 A/G p.Ser180Pro 0.04% <0.003% ClinVar LP (partial LOF)b 34.2 (30.2) 34.4 (33.5) N/A DZMC4R chr18:60371827 C/T p.Ala175Thr 0.04% 0.02% ClinVar P HGMD highC DM(partial LOF)b29 (25.6) 28.5 (24.4) N/A MZMC4R chr18:60371842 T/C p.Ile170Val 0.04% 0.01% ClinVar P HGMD highC DM(partial LOF)b22.6 (23.7) N/A 21.3 (24.7) DZLipodystrophyZMPSTE24 chr1:40290870 G/GT p.Leu362fs 0.11% 0.03% ClinVar P 18 (18.9) N/A N/A N/AZMPSTE24 chr1:40290870 G/GT p.Leu362fs 0.11% 0.03% ClinVar P 22 (20.3) N/A N/A N/AZMPSTE24 chr1:40290870 G/GT p.Leu362fs 0.11% 0.03% ClinVar P 22.4 (26.1) N/A 22.6 (25.1) DZZMPSTE24 chr1:40290870 GT/G p.Leu362fs 0.04% <0.003% not annotateda 30.7 (27.5) N/A 24.9 (24.1) DZAGPAT2 chr9:136673876 G/C p.Ala238Gly 0.04% 0.00% HGMD highC DM 20 (23.3) N/A N/A N/ALIPE chr19:42401821CCCCCCGCAGCCCCCGTCTA/Cp.Val1068fs 0.04% 0.07% ClinVar P 23 (27.4) N/A N/A N/ABSCL2 chr11:62692371 C/T c.863+5G>A 0.04% <0.003% ClinVar P 24 (29.9) N/A N/A N/AMAF,minor allele frequency; HGMD highCDM, HumanGeneMutation Database high-confidence disease-causingmutation; lowC, low confidence; ClinVar LP, likely pathogenic; P, pathogenic; MZ,monozygotic; DZ, dizygotic. Each variant was only seen once in the unrelated participants of this study. See alignments in Figure S5.aThis deletion at the same site as a lipodystrophy insertion has not previously been annotated.bFunctional annotation according to Collet et al. (2017), where LOF = loss of function.CellMetabolism29,488–500,February5,2019497strong signature for the study of the relationship between BMI,obesity, metabolic disease, and the genetics of BMI.Consistent with previous studies and earlier work in theTwinsUK cohort, branched-chain and aromatic amino acids,and metabolites involved in nucleotide metabolism, such asurate and pseudouridine, are strongly perturbed by obesity(Butte et al., 2015; Ho et al., 2016; Menni et al., 2017; Parket al., 2015). The underlying reason for the perturbation ofbranched-chain amino acid metabolism in obese individualsand thosewith insulin resistance is thought to be related to differ-ences in the amino acid catabolism in adipose tissue (Newgard,2012). The single metabolite with the most significant associa-tion with BMI was urate, as we previously reported (Menniet al., 2017). It is well known that uric acid increases with obesity,due to insulin resistance reducing the kidneys’ ability to eliminateuric acid, but previous work has not emphasized the power ofurate to predict BMI (Butte et al., 2015; Facchini et al., 1991;Ho et al., 2016). We also found a strong signal for lipids to beassociated with BMI, with an enrichment of associations foundfor glycerol lipids. These results are consistent with previousstudies showing that sphingomyelins and diacylglycerols in-crease with BMI while lysophosphocholines decrease withBMI, with other various phosphatidylcholines having effects inboth directions (Butte et al., 2015; Ho et al., 2016; Park et al.,2015). A number of BMI-associated metabolites (12 of the 49-metabolite signature) were associated with insulin resistance af-ter controlling for BMI. As previously observed (Caleyachettyet al., 2017; Oberbach et al., 2011; Reinehr et al., 2015; Wahlet al., 2015), the metabolome abnormalities associated withhigh BMI corrected with loss of weight. However, our studyfound that metabolite levels did not provide predictive powerfor future weight changes. Overall, the metabolome perturba-tions appear as a consequence of changes in weight as opposedto being a contributing factor.While BMI correlates well and to a large extent with individualhealth outcomes, it does not have the sensitivity to identify out-liers, some of which carry unique health consequences. Theselimitations to BMI are recognized in the literature (Neelandet al., 2018). Here, the metabolome signature identified individ-uals whose predicted mBMI was either substantially lower orhigher than their actual BMI, i.e., individuals with lesser or greatermetabolic consequences of obesity, using a technology that isamenable to large-scale implementation. These individualsinclude the metabolically healthy obese, but we also emphasizethe importance of the metabolome anomalies in identifyingunhealthy individuals with a normal BMI. These profiles weregenerally stable over the prolonged follow up. An abnormal me-tabolome was associated in the present study with an approxi-mately 5-fold increase in baseline cardiovascular events andadditional 2-fold increase in cardiovascular events during followup when comparing individuals who were matched for BMI butwho had opposing metabolome signatures. Thus, while our find-ings are in line with the known relationships between metaboli-cally healthy obese status and health-related traits likemetabolicsyndrome and body fat (Brochu et al., 2001; Caleyachetty et al.,2017; Karelis et al., 2005), we extend this relationship to thebroader category of metabolically healthy and unhealthy individ-uals on the basis of the disparity between mBMI and BMI. Forexample, we observed differences in waist/hip ratio, percentvisceral fat, and blood pressure between mBMI/BMI outliersdespite having the same BMI distribution. The fact that themeta-bolically healthy obese have a high BMI polygenic risk scorealso supports the concept that some of the genetic studiesmay capture anthropomorphic associations—body size—ratherthan obesity sensu stricto. Overall, the health consequencesobserved across the various mBMI groups indicate that thereis a durable benefit of maintaining a healthy metabolome signa-ture and point to an ongoing risk for the individuals that have anunhealthy metabolome despite stability of BMI.In line with the observation that a richer set of biomarkers ismore accurate than a narrower panel of single biomarkers, weobserved that 650 metabolites could explain ~50% and a 49-metabolite signature of the best markers could explain 43% ofvariance in BMI. In contrast, a conventional model that includedage, sex, HDL, LDL, total cholesterol, and triglycerides couldexplain 31%, while the best single metabolic marker (urate)could explain 16% of the variance in BMI. The message thatemerges from these types of observations—and that willbecome central to the implementation of machine and deeplearning in medicine—is that there is a substantial loss of signalassociated with the quest for a single biomarker. The medicalsystem would, however, need to validate and establish thehealth economics of rich data biomarkers to foster adoption inclinical diagnostic routine.In contrast with metabolomics analyses, the present studydoes not support a strong association between metabolomechanges and the genetics of BMI defined by a 97-variant poly-genic risk score (Locke et al., 2015). This may be explained bythe fact that known BMI GWAS loci explain only a small fraction(3%) of BMI heritability (Locke et al., 2015). Despite this overalllack of explanatory power, there was a clear signal for individualswith higher polygenic risk scores to have greater rates of obesity.Because the genetic risk score does not include rare variants, wealso identified individuals who carried rare functional variants inthe known obesity geneMC4R, which is the single best exampleof a gene where rare coding variants have a large effect onobesity (Turcot et al., 2018). The carriers of these variants wereoften obese individuals, but their metabolome was not categor-ically different from that of other obese individuals. The lack ofmetabolome differences for carriers of variants in this gene isnot surprising given that MC4R variants cause obesity byincreasing appetite. However, we did find that obese carriersofMC4R variants often had low polygenic risk scores for obesity;out of 31 participants who were obese with polygenic risk scoresin the lowest quartile, 5.4%wereMC4R variant carriers, while thecarrier frequency was just 0.4% in those of normal weight. Thus,our study shows the interest of sequencing obese individualswith low polygenic risk scores because of the apparent enrich-ment for monogenic contributions. As we completed this study,a large consortium provided additional detail on the role of vari-ants in pathways that implicate energy intake and expenditure inobesity (Turcot et al., 2018). Future studies should also explorethe role of rare functional variants in lipodystrophy genes onthe metabolic traits in the general population.In summary, the present study highlights the health risks of theperturbed metabolome. The study also indicates that the ge-netics of BMI are separate from metabolic health and serves toprioritize a subset of individuals for genetic analysis. The498 Cell Metabolism 29, 488–500, February 5, 2019assessment of themetabolome and genome of BMI lays ground-work for future studies of the heterogeneity of obesity and treat-ment of its endophenotypes. Specifically, the methodology usedhere can be applied to build similar models for insulin resistance,fat distribution, or any other number of clinical traits, and themBMImetabolome signature can act as a biomarker of responseto the new therapeutics that target patients with MC4R muta-tions (K€uhnen et al., 2016). In the future, metabolic profiling couldhelp select patients for clinical trials beyond genetic sequencing,thus expanding drug utility (Yanovski and Yanovski, 2018).Limitations of StudyWhile the uniquely rich metabolomic and phenotypic data in thisstudy permitted a thorough investigation of obesity-relatedhealth, there were nonetheless limitations. We chose to focusthis study on BMI, the most widely used measurement ofobesity. However, future studies building models for other traitsof metabolic health, for example diabetes or visceral fat, may beof greater utility. Our study also had limited longitudinal informa-tion onmedication use, which precluded an analysis of how drugprescription influenced mBMI/BMI categories and vice versa.Additionally, the association we observed between mBMI/BMIcategories and longitudinal cardiovascular outcomes requiresconfirmation in a replication cohort.STAR+METHODSDetailed methods are provided in the online version of this paperand include the following:d KEY RESOURCES TABLEd CONTACT FOR REAGENT AND RESOURCE SHARINGd EXPERIMENTAL MODEL AND SUBJECT DETAILSd METHOD DETAILSB PhenotypingB Metabolite ProfilingB Genome Sequencing and Analysisd QUANTIFICATION AND STATISTICAL ANALYSISd DATA AND SOFTWARE AVAILABILITYSUPPLEMENTAL INFORMATIONSupplemental Information includes five figures and two tables and can befound with this article online at https://doi.org/10.1016/j.cmet.2018.09.022.ACKNOWLEDGMENTSWe thank E. Muse for valuable comments. We acknowledge N. Schenker-Ahmed, L. Huang, M. Tyagi, and P. Sheth for aiding with data collection.TwinsUK receives funding from the Wellcome Trust, European Community’sSeventh Framework Programme (FP7/2007-2013, 277849, 201413, and259749), the National Institute for Health Research (NIHR) Clinical ResearchFacility at Guy’s and St Thomas’ NHS Foundation Trust, and NIHR BiomedicalResearch Centre based at Guy’s and St Thomas’ NHS Foundation Trust andKing’s College London. A.T. is supported by the Qualcomm Foundation andthe NIH Center for Translational Science Award (CTSA, grant numberUL1TR002550).AUTHOR CONTRIBUTIONSA.T. and E.T.C. conceived of the experiment(s); E.T.C. conducted the experi-ment(s); L.G. supported metabolome analyses and interpretation; and E.T.C.,J.C.V., and A.T. analyzed the results. C.T.C. and B.T. provided detailed insightinto metabolic diseases; C.L.S., N.S., L.H., E.F.K., and L.A.N. aided with datacollection and interpretation; and T.D.S. is responsible for the TwinsUK study.All authors reviewed the manuscript.DECLARATION OF INTERESTSE.T.C., C.L.S., N.S., L.H., L.A.N., E.F.K., C.T.C., and J.C.V. are currently orrecently employed at Human Longevity, Inc. L.G. was recently an employeeof Metabolome, Inc. A.T., B.T., and T.D.S. declare no conflict of interest.Received: April 7, 2018Revised: June 27, 2018Accepted: September 25, 2018Published: October 11, 2018REFERENCESAbraham, S.B., Rubino, D., Sinaii, N., Ramsey, S., and Nieman, L.K. (2013).Cortisol, obesity, and the metabolic syndrome: a cross-sectional study ofobese subjects and review of the literature. Obesity (Silver Spring) 21,E105–E117.Bjo¨rntorp, P., and Rosmond, R. (2000). Obesity and cortisol. Nutrition 16,924–936.Brochu, M., Tchernof, A., Dionne, I.J., Sites, C.K., Eltabbakh, G.H., Sims, E.A.,and Poehlman, E.T. (2001). What are the physical characteristics associatedwith a normal metabolic profile despite a high level of obesity in postmeno-pausal women? J. Clin. Endocrinol. Metab. 86, 1020–1025.Butte, N.F., Liu, Y., Zakeri, I.F., Mohney, R.P., Mehta, N., Voruganti, V.S.,Go¨ring, H., Cole, S.A., and Comuzzie, A.G. (2015). Global metabolomicprofiling targeting childhood obesity in the Hispanic population. Am. J. Clin.Nutr. 102, 256–267.Caleyachetty, R., Thomas, G.N., Toulis, K.A., Mohammed, N., Gokhale, K.M.,Balachandran, K., and Nirantharakumar, K. (2017). Metabolically healthyobese and incident cardiovascular disease events among 3.5 million menand women. J. Am. Coll. Cardiol. 70, 1429–1437.Chen, H.H., Tseng, Y.J., Wang, S.Y., Tsai, Y.S., Chang, C.S., Kuo, T.C., Yao,W.J., Shieh, C.C., Wu, C.H., and Kuo, P.H. (2015). The metabolome profilingand pathway analysis in metabolic healthy and abnormal obesity. Int. J.Obes. 39, 1241–1248.Cle´ment, K., Biebermann, H., Farooqi, I.S., Van der Ploeg, L., Wolters, B.,Poitou, C., Puder, L., Fiedorek, F., Gottesdiener, K., Kleinau, G., et al.(2018). MC4R agonism promotes durable weight loss in patients with leptin re-ceptor deficiency. Nat. Med. 24, 551–555.Cohen, I.V., Cirulli, E.T., Mitchell, M.W., Jonsson, T.J., Yu, J., Shah, N.,Spector, T.D., Guo, L., Venter, J.C., and Telenti, A. (2018). Acetaminophen(paracetamol) use modifies the sulfation of sex hormones. EBioMedicine 28,316–323.Collet, T.H., Dubern, B., Mokrosinski, J., Connors, H., Keogh, J.M., Mendesde Oliveira, E., Henning, E., Poitou-Bernert, C., Oppert, J.M., Tounian, P.,et al. (2017). Evaluation of a melanocortin-4 receptor (MC4R) agonist(Setmelanotide) in MC4R deficiency. Mol. Metab. 6, 1321–1329.Facchini, F., Chen, Y.D., Hollenbeck, C.B., and Reaven, G.M. (1991).Relationship between resistance to insulin-mediated glucose uptake, urinaryuric acid clearance, and plasma uric acid concentration. JAMA 266,3008–3011.Garg, A. (2011). Clinical review#: lipodystrophies: genetic and acquired bodyfat disorders. J. Clin. Endocrinol. Metab. 96, 3313–3325.Hales, C.M., Carroll, M.D., Fryar, C.D., and Ogden, C.L. (2017). Prevalence ofobesity among adults and youth: United States, 2015-2016. NCHS DataBrief, 1–8.Ho, J.E., Larson, M.G.M., Ghorbani, A., Cheng, S., Chen, M.-H., Keyes, M.,Rhee, E.P.E.E.P., Clish, C.B., Vasan, R.S.R., Gerszten, R.E., and Wang, T.J.(2016). Metabolomic profiles of body mass index in the Framingham HeartStudy reveal distinct cardiometabolic phenotypes. PLoS One 11, e0148361.Cell Metabolism 29, 488–500, February 5, 2019 499Iglesias Molli, A.E., Penas Steinhardt, A., Lo´pez, A.P., Gonza´lez, C.D., Vilarin˜o,J., Frechtel, G.D., and Cerrone, G.E. (2017). Metabolically healthy obese indi-viduals present similar chronic inflammation level but less insulin-resistancethan obese individuals with metabolic syndrome. PLoS One 12, e0190528.IncollingoRodriguez,A.C.,Epel, E.S.,White,M.L.,Standen,E.C.,Seckl, J.R., andTomiyama, A.J. (2015). Hypothalamic-pituitary-adrenal axis dysregulation andcortisol activity in obesity: a systematic review. Psychoneuroendocrinology 62,301–318.Karelis, A.D., Faraj, M., Bastard, J.P., St-Pierre, D.H., Brochu, M.,Prud’homme, D., and Rabasa-Lhoret, R. (2005). The metabolically healthybut obese individual presents a favorable inflammation profile. J. Clin.Endocrinol. Metab. 90, 4145–4150.K€uhnen, P., Cle´ment, K., Wiegand, S., Blankenstein, O., Gottesdiener, K.,Martini, L.L., Mai, K., Blume-Peytavi, U., Gr€uters, A., and Krude, H. (2016).Proopiomelanocortin deficiency treated with a melanocortin-4 receptoragonist. N. Engl. J. Med. 375, 240–246.Landrum, M.J., Lee, J.M., Benson, M., Brown, G.R., Chao, C., Chitipiralla, S.,Gu, B., Hart, J., Hoffman, D., Jang, W., et al. (2018). ClinVar: improving accessto variant interpretations and supporting evidence. Nucleic Acids Res. 46 (D1),D1062–D1067.Lee, S., Zhang, C., Kilicarslan, M., Piening, B.D., Bjornson, E., Hallstro¨m, B.M.,Groen, A.K., Ferrannini, E., Laakso, M., Snyder, M., et al. (2016). Integratednetwork analysis reveals an association between plasma mannose levelsand insulin resistance. Cell Metab. 24, 172–184.Lippert, C., Listgarten, J., Liu, Y., Kadie, C.M., Davidson, R.I., and Heckerman,D. (2011). FaST linear mixed models for genome-wide association studies.Nat. Methods 8, 833–835.Locke, A.E., Kahali, B., Berndt, S.I., Justice, A.E., Pers, T.H., Day, F.R., Powell,C., Vedantam, S., Buchkovich, M.L., Yang, J., et al.; LifeLines Cohort Study;ADIPOGen Consortium; AGEN-BMI Working Group; CARDIOGRAMplusC4DConsortium; CKDGen Consortium; GLGC; ICBP; MAGIC Investigators;MuTHER Consortium; MIGen Consortium; PAGE Consortium; ReproGenConsortium; GENIE Consortium; International Endogene Consortium (2015).Genetic studies of body mass index yield new insights for obesity biology.Nature 518, 197–206.Long, T., Hicks, M., Yu, H.-C., Biggs,W.H., Kirkness, E.F., Menni, C., Zierer, J.,Small, K.S., Mangino, M., Messier, H., et al. (2017). Whole-genomesequencing identifies common-to-rare variants associated with human bloodmetabolites. Nat. Genet. 49, 568–578.Matthews, D.R., Hosker, J.P., Rudenski, A.S., Naylor, B.A., Treacher, D.F., andTurner, R.C. (1985). Homeostasis model assessment: insulin resistance andbeta-cell function from fasting plasma glucose and insulin concentrations inman. Diabetologia 28, 412–419.Menni, C., Migaud, M., Kastenm€uller, G., Pallister, T., Zierer, J., Peters, A.,Mohney, R.P., Spector, T.D., Bagnardi, V., Gieger, C., et al. (2017).Metabolomic profiling of long-term weight change: role of oxidative stressand urate levels in weight gain. Obesity (Silver Spring) 25, 1618–1624.Moayyeri, A., Hammond, C.J., Hart, D.J., and Spector, T.D. (2013). The UKAdult Twin Registry (TwinsUK Resource). Twin Res. Hum. Genet. 16, 144–149.Neeland, I.J., Poirier, P., and Despre´s, J.-P. (2018). Cardiovascular and meta-bolic heterogeneity of obesity: clinical challenges and implications for man-agement. Circulation 137, 1391–1406.Newgard, C.B. (2012). Interplay between lipids and branched-chain aminoacids in development of insulin resistance. Cell Metab. 15, 606–614.Oberbach, A., Bl€uher, M., Wirth, H., Till, H., Kovacs, P., Kullnick, Y.,Schlichting, N., Tomm, J.M., Rolle-Kampczyk, U., Murugaiyan, J., et al.(2011). Combined proteomic andmetabolomic profiling of serum reveals asso-ciation of the complement system with obesity and identifies novel markers ofbody fat mass changes. J. Proteome Res. 10, 4769–4788.Park, S., Sadanala, K.C., and Kim, E.-K. (2015). A metabolomic approach tounderstanding the metabolic link between obesity and diabetes. Mol. Cells38, 587–596.Perkins, B.A., Caskey, C.T., Brar, P., Dec, E., Karow, D., Kahn, A., Hou, C.,Shah, N., Boeldt, D., Coughlin, E., et al. (2017). Precision medicine screeningusing whole genome sequencing and advanced imaging to identify diseaserisk in adults. bioRxiv. https://doi.org/10.1101/133538.Piening, B.D., Zhou, W., Contrepois, K., Ro¨st, H., Gu Urban, G.J., Mishra, T.,Hanson, B.M., Bautista, E.J., Leopold, S., Yeh, C.Y., et al. (2018). Integrativepersonal omics profiles during periods of weight gain and loss. Cell Syst. 6,157–170.e8.Praveen, E.P., Sahoo, J.P., Kulshreshtha, B., Khurana, M.L., Gupta, N.,Dwivedi, S.N., Kumar, G., and Ammini, A.C. (2011). Morning cortisol is lowerin obese individuals with normal glucose tolerance. Diabetes Metab. Syndr.Obes. 4, 347–352.Reinehr, T., Wolters, B., Knop, C., Lass, N., Hellmuth, C., Harder, U., Peissner,W., Wahl, S., Grallert, H., Adamski, J., et al. (2015). Changes in the serummetabolite profile in obese children with weight loss. Eur. J. Nutr. 54, 173–181.Rhee, E.J. (2018). Being metabolically healthy, the most responsible factor forvascular health. Diabetes Metab. J. 42, 19–25.Stenson, P.D., Ball, E.V., Mort, M., Phillips, A.D., Shiel, J.A., Thomas, N.S.T.,Abeysinghe, S., Krawczak, M., and Cooper, D.N. (2003). Human GeneMutation Database (HGMD): 2003 update. Hum. Mutat. 21, 577–581.Telenti, A., Pierce, L.C.T., Biggs,W.H., di Iulio, J., Wong, E.H.M., Fabani, M.M.,Kirkness, E.F., Moustafa, A., Shah, N., Xie, C., et al. (2016). Deep sequencingof 10,000 human genomes. Proc. Natl. Acad. Sci. USA 113, 11901–11906.Turcot, V., Lu, Y., Highland, H.M., Schurmann, C., Justice, A.E., Fine, R.S.,Bradfield, J.P., Esko, T., Giri, A., Graff, M., et al.; CHD Exome+ Consortium;EPIC-CVD Consortium; ExomeBP Consortium; Global Lipids GeneticConsortium; GoT2D Genes Consortium; EPIC InterAct Consortium;INTERVAL Study; ReproGen Consortium; T2D-Genes Consortium; MAGICInvestigators; Understanding Society Scientific Group (2018). Protein-alteringvariants associated with body mass index implicate pathways that control en-ergy intake and expenditure in obesity. Nat. Genet. 50, 26–41.Wahl, S., Vogt, S., St€uckler, F., Krumsiek, J., Bartel, J., Kacprowski, T.,Schramm, K., Carstensen, M., Rathmann, W., Roden, M., et al. (2015). Multi-omic signature of body weight change: results from a population-based cohortstudy. BMC Med. 13, 48.Walker, B.R., Soderberg, S., Lindahl, B., and Olsson, T. (2000). Independenteffects of obesity and cortisol in predicting cardiovascular risk factors inmen and women. J. Intern. Med. 247, 198–204.Whitlock, G., Lewington, S., Sherliker, P., Clarke, R., Emberson, J., Halsey, J.,Qizilbash, N., Collins, R., and Peto, R.; Prospective Studies Collaboration(2009). Body-mass index and cause-specific mortality in 900 000 adults:collaborative analyses of 57 prospective studies. Lancet 373, 1083–1096.WHO (2018). Obesity and overweight. http://www.who.int/news-room/fact-sheets/detail/obesity-and-overweight.Wikoff, W.R., Anfora, A.T., Liu, J., Schultz, P.G., Lesley, S.A., Peters, E.C., andSiuzdak, G. (2009). Metabolomics analysis reveals large effects of gut micro-flora on mammalian blood metabolites. Proc. Natl. Acad. Sci. USA 106,3698–3703.Wirix, A.J.G., Finken, M.J.J., von Rosenstiel-Jadoul, I.A., Heijboer, A.C.,Nauta, J., Groothoff, J.W., Chinapaw, M.J.M., and Kist-van Holthe, J.E.(2017). Is there an association between cortisol and hypertension in over-weight or obese children? J. Clin. Res. Pediatr. Endocrinol. 9, 344–349.Yanovski, S.Z., and Yanovski, J.A. (2018). Toward precision approaches forthe prevention and treatment of obesity. JAMA 319, 223–224.500 Cell Metabolism 29, 488–500, February 5, 2019STAR+METHODSKEY RESOURCES TABLECONTACT FOR REAGENT AND RESOURCE SHARINGFurther information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, ElizabethCirulli (liz.cirulli@gmail.com).EXPERIMENTAL MODEL AND SUBJECT DETAILSThe study included 1,969 largely European ancestry twins enrolled in the TwinsUK registry, a British national register of adult twins(Moayyeri et al., 2013). We previously reported a detailed study of the genetic variants influencing the human metabolome in thiscohort (Long et al., 2017). Serum samples were collected at three visits, 8-18 (median 13) years apart. The cohort is mainly composedof females (96.7%), and the sample set we used included 388 monozygotic twin pairs, 519 dizygotic twin pairs, and 155 unrelatedindividuals. The age of participants at the first time point ranged from 33 to 74 years old (median 51); 36 to 81 years old (median 59) atthe second time point; and 42 to 88 years old (median 65) at the third time point. The BMI values measured at each metabolome timepoint were taken within two years of the blood draw date. At baseline, 36.3% of the female participants and 53.8% of the male par-ticipants were overweight, and 16.9% of the females and 10.8% of the males were obese. The twins study was approved bySt. Thomas’ Hospital Research Ethics Committee, and all participants provided informed written consent. BMI data were availablefor 1743 participants within two years of the time point for metabolome time point 1, 1834 for within two years of time point 2, and1777 for up to 2 years before time point 3 or 4 years after this time point; 1,458 individuals had all three datapoints.For independent validation and studies of phenotypes correlated with metabolic BMI outliers, we enrolled 427 unselected adultsmore than 18 years old who were able to come to the Health Nucleus in La Jolla, CA for a clinical research protocol (Perkins et al.,2017). Participants underwent a verbal review of the institutional review board-approved consent (Western Institutional ReviewBoard). Participants ranged in age from 18-89 years old (median 53), were 32.9% female, and had BMI data measured at onetime point: 16.7% of the female participants and 47.5% of the male participants were overweight, and 7.2% of the females and23.7% of the males were obese.METHOD DETAILSPhenotypingIndividuals in the TwinsUK cohort and Health Nucleus both underwent dual-energy X-ray absorptiometry (DEXA) imaging. The datafrom these scans were used to calculate android/gynoid ratio, percent body fat, visceral fat, and subcutaneous fat. DEXA is very ac-curate in themeasurement abdominal visceral adipose tissue (VAT). High levels of VAT are associated with atherogenic dyslipidemia,hyperinsulinemia, and glucose intolerance (Neeland et al., 2018). TwinsUK cohort participants were additionally measured forcircumference at the waist and hip using a measuring tip to calculate the waist/hip ratio. TwinsUK participants self-reported infor-mation about whether they were taking high blood pressure medication at their first visit and about cardiovascular events and theirtiming via a survey at the final visit. For a selected number of Health Nucleus participants, images of fat and water (imaging of muscle)were available from symmetrical chemical shift Magnetic Resonance Imaging (MRI) via the Dixon method. Quantitative insulinREAGENT or RESOURCE SOURCE IDENTIFIERDeposited Datagnomad, Genome Aggregation Database http://gnomad.broadinstitute.org/ RRID: SCR_014964ClinVar https://www.ncbi.nlm.nih.gov/clinvar/ RRID: SCR_006169HGMD, Human Gene Mutation Database http://www.hgmd.cf.ac.uk/ac/index.php RRID: SCR_001888Study data deposited into Mendeley Data This paper; https://data.mendeley.com/ https://doi.org/10.17632/kpjsnwd2mc.1Study data and scripts deposited into GitHub This paper; https://github.com https://github.com/humanlongevity/mbmiSoftware and AlgorithmsR Project for Statistical Computing http://www.r-project.org/ RRID: SCR_001905FaST LMM https://github.com/MicrosoftGenomics/FaST-LMMRRID: SCR_015506Cell Metabolism 29, 488–500.e1–e2, February 5, 2019 e1resistance (homeostatic model assessment, HOMA) was calculated as fasting insulin x fasting glucose / 405, and being insulin resis-tant was defined by HOMA scoreR 3 (http://gihep.com/calculators/other/homa/) (Matthews et al., 1985).Metabolite ProfilingThe non-targeted metabolomics analysis of 901 metabolites in the TwinsUK cohort and 1,007 metabolites in the Health Nucleuscohort was performed at Metabolon (Durham, North Carolina, USA) on a platform consisting of four independent ultra high perfor-mance liquid chromatography-tandem mass spectrometry (UPLC-MS/MS) methods. The detailed descriptions of the platform canbe found in our previous publications (Cohen et al., 2018; Long et al., 2017). For the TwinsUK cohort, blood serum after fasting wasused for analysis, and the resulting raw values were transformed to z scores using the mean and standard deviation. For the HealthNucleus cohort, blood plasma after fasting was used for analysis, and values from multiple experimental batches were normalizedinto Z-scores based on a reference cohort of either 42 (n = 457) or 300 (n = 176) self-reported healthy individuals run with each batch.The 42 and 300-normalized batches were converted to the same scale using linear transformation based on the values obtained from7 runs that included both the 42 and 300 controls. Samples with metabolite measurements that were below the detection thresholdwere imputed as the minimum value for that metabolite.Genome Sequencing and AnalysisAs previously described (Telenti et al., 2016), DNA samples were sequenced on an Illumina HiSeqX sequencer utilizing a 150 basepaired-end single index read format. Reads were mapped to the human reference sequence build HG38. Variants were called usingISIS Analysis Software (v. 2.5.26.13; Illumina). A linear mixed model was applied to account for family structure in the cohort whiletesting for associations between genetic variants and the different phenotypes: BMI; BMI prediction model values and residuals afteraccounting for BMI, age, sex; and levels of the 49 BMI-associated metabolites. A genetic similarity matrix (GSM) was constructedfrom 301,556 variants that represented a random 20% of all common (MAF > 5%) variants genome-wide after linkage-disequilibrium(LD) pruning (r2 less than 0.6, window size 200 kb) and was used to model the random effect in the linear mixed model via a ‘‘leave-out-one-chromosome’’ method for each tested variant. Each of 97 known BMI-associated variants was tested independently usingcustomized Python scripts wrapping the FaST-LMM package (FaST LMM, RRID: SCR_015506) (Lippert et al., 2011; Locke et al.,2015). Principal component axes were calculated to check ethnicity using plink, and the first principal component for those ofEuropean ancestry was used as a covariate in analyses of unrelated individuals in R described below. Polygenic risk scores werecalculated using genotypes for 97 variants whose associations and betas had been published previously (Locke et al., 2015).Rare variants in the gene MC4R were defined as coding and splice variants with MAF < 0.1%. Rare lipodystrophy variants weredefined as those achieving a pathogenic or likely pathogenic categorization in ClinVar (ClinVar, RRID: SCR_006169) or HGMD(HGMD, RRID: SCR_001888). Allele frequencies were calculated in our set of samples and also retrieved from gnomad (GenomeAggregation Database, RRID: SCR_014964).QUANTIFICATION AND STATISTICAL ANALYSISR (R Project for Statistical Computing, RRID: SCR_001905) was used for the analysis and data manipulation. Bonferroni correctionwas used for all analyses, and most statistical analyses were restricted to unrelated individuals of European ancestry, in accordancewith field standards for ensuring that ancestry differences do not cause bias or skew in the results. For each quantitative analysis ofBMI or other traits, the subset of BMI values or other outcome variables used were rank-ordered and forced to a normal distribution.Analyses comparing metabolites to BMI were performed in R using the lm function, and age, sex, and the first genetic principalcomponent were included as covariates. The obesity prediction model was built using ridge regression (alpha = 0) with glmnet inR. The residuals used to separate participants into the five categories shown in Figure 2 were calculated using age, sex, and initialBMI. Heatmaps were generated in R using the pheatmap package. Survival analysis was performed using coxph in R with age at firstvisit included as a covariate. Power calculation was performed using the power.stratify command in the powerSurvEpi package in R.DATA AND SOFTWARE AVAILABILITYThe R analysis scripts and data files used are available at https://github.com/humanlongevity/mbmi. Data files have been depositedto Mendeley Data and are available at https://doi.org/10.17632/kpjsnwd2mc.1.e2 Cell Metabolism 29, 488–500.e1–e2, February 5, 2019

Obesity profoundly perturbs the human metabolome

Abstract

Extracted data

Obesity profoundly perturbs the human metabolome

Abstract

Extracted data

Related items

Related items