Posts Tagged ‘FTO’

Obesity Variant Circuitry

Larry H. Bernstein, MD, FCAP, Curator



FTO Obesity Variant Circuitry and Adipocyte Browning in Humans

Melina Claussnitzer,  Simon N. Dankel, Kyoung-Han Kim,  Gerald Quon,  Wouter Meuleman,  Christine Haugen,  Viktoria Glunk,  Isabel S. Sousa, et al.

N Engl J Med 2015; 373:895-907  Sept 3, 2015    DOI: http://dx.10.org1/056/NEJMoa1502214


Genomewide association studies can be used to identify disease-relevant genomic regions, but interpretation of the data is challenging. The FTO region harbors the strongest genetic association with obesity, yet the mechanistic basis of this association remains elusive.

Full Text of Background…



We examined epigenomic data, allelic activity, motif conservation, regulator expression, and gene coexpression patterns, with the aim of dissecting the regulatory circuitry and mechanistic basis of the association between the FTO region and obesity. We validated our predictions with the use of directed perturbations in samples from patients and from mice and with endogenous CRISPR–Cas9 genome editing in samples from patients.

Full Text of Methods…



Our data indicate that the FTO allele associated with obesity represses mitochondrial thermogenesis in adipocyte precursor cells in a tissue-autonomous manner. The rs1421085 T-to-C single-nucleotide variant disrupts a conserved motif for the ARID5B repressor, which leads to derepression of a potent preadipocyte enhancer and a doubling of IRX3 and IRX5 expression during early adipocyte differentiation. This results in a cell-autonomous developmental shift from energy-dissipating beige (brite) adipocytes to energy-storing white adipocytes, with a reduction in mitochondrial thermogenesis by a factor of 5, as well as an increase in lipid storage. Inhibition of Irx3 in adipose tissue in mice reduced body weight and increased energy dissipation without a change in physical activity or appetite. Knockdown of IRX3 or IRX5 in primary adipocytes from participants with the risk allele restored thermogenesis, increasing it by a factor of 7, and overexpression of these genes had the opposite effect in adipocytes from nonrisk-allele carriers. Repair of the ARID5B motif by CRISPR–Cas9 editing of rs1421085 in primary adipocytes from a patient with the risk allele restored IRX3 and IRX5 repression, activated browning expression programs, and restored thermogenesis, increasing it by a factor of 7.


Effect of the FTO Locus on IRX3 and IRX5 in Human Adipocyte Progenitor Cells

To identify the cell types in which the causal variant may act, we examined chromatin state maps15,16 of the FTO obesity region across 127 cell types. An unusually long enhancer (12.8 kb) in mesenchymal adipocyte progenitors indicated a major regulatory locus (Figure 1B; and Fig. S1A, S1B, and S1C in the Supplementary Appendix). Haplotype-specific enhancer assays showed activity in association with the risk haplotype that was 2.4 times as high as that associated with the nonrisk haplotype in human SGBS adipocytes (i.e., adipocytes derived from a patient with the Simpson–Golabi–Behmel syndrome), which indicated genetic control of enhancer activity (Figure 1C). Enhancers in brain cells and other cell types were considerably shorter than those in mesenchymal adipocyte progenitors and lacked allelic activity (Fig. S1C and S1D in the Supplementary Appendix).


Figure 1. Activation of a Superenhancer in Human Adipocyte Progenitors by the FTO Obesity Risk Haplotype.

Panel A shows the genetic association with body-mass index (BMI) for all common FTO locus variants,14 including the reported single-nucleotide variant (SNV) rs1558902 (red diamond) and the predicted causal SNV rs1421085 (red square). Gray shading delineates consecutive 10-kb segments. CEU denotes a population of Utah residents with northern and western European ancestry, and LD linkage disequilibrium. Panel B shows chromatin state annotations for the locus across 127 reference epigenomes (rows) for cell and tissue types profiled by the Roadmap Epigenomics Project.15,16 For information on the colors used to denote chromatin states, see Figure S1A in the Supplementary Appendix. Vertical lines delineate the consecutive 10-kb segments shown in Panel A. ESC denotes embryonic stem cell, HSC hematopoietic stem cell, and iPSC induced pluripotent stem cell. Panel C shows human SGBS adipocyte enhancer activity, for 10-kb tiles, of the risk and nonrisk haplotypes with the use of relative luciferase expression. The boxes indicate means from seven triplicate experiments, and T bars indicate standard deviations.

To predict putative target genes, we examined large domains that had long-range three-dimensional chromatin interactions surrounding FTO and identified eight candidate genes (Figure 2A and 2B)



Figure 2. Activation of IRX3 and IRX5 Expression in Human Adipocyte Progenitors by the FTO Obesity Risk Genotype.

Panel A shows gene annotations and LD with array tag variant rs9930506 in a 2.5-Mb window; LD is expressed as r2 values in the CEU population. Arrows indicate the direction of transcription of annotated genes in the locus. Panel B shows chromosome conformation capture (Hi-C) interactions contact probabilities in human IMR90 myofibroblasts,22 revealing a 2-Mb topologically associating domain, and LD mean r2 statistics for all SNV pairs at 40-kb resolution. Panel C shows box plots for expression levels, after 2 days of differentiation, in human adipose progenitors isolated from 20 risk-allele carriers and 18 nonrisk-allele carriers, evaluated by means of a quantitative polymerase-chain-reaction analysis for all genes in the 2.5-Mb locus. The horizontal line within each box represents the median, the top and bottom of each box indicate the 75th and 25th percentile, and I bars indicate the range.

Among them, the developmental regulators IRX3 and IRX5 had genotype-associated expression, which indicated long-range (1.2-Mb) genetic control in primary preadipocytes (Figure 2C). Genotype-associated expression was not observed in whole-adipose tissue, a finding consistent with previous reports23,24; this indicated that the effect was cell type–specific and restricted to preadipocytes, which represent a minority of cells in adipose tissue (Fig. S2A in the Supplementary Appendix).


Effect of the FTO Locus on Mitochondrial Thermogenesis and Lipid Storage

To identify the biologic processes affected by altered IRX3 and IRX5expression in adipocytes, we used genomewide expression patterns in brown adipocyte–containing perirenal adipose tissue from a separate cohort of 10 nongenotyped, healthy kidney donors to identify genes with expression that was positively or negatively correlated with IRX3 and IRX5 expression. Genes that are associated with mitochondrial functions were found to have a negative correlation with IRX3 and IRX5, and genes with FXR and RXR lipid-metabolism functions were found to have a positive correlation, which suggests thatIRX3 and IRX5 may play roles in energy dissipation and storage


Figure 3A


Regulation of Obesity-Associated Cellular Phenotypes in Human Adipocytes by IRX3and IRX5., and Table S1 in the Supplementary Appendix). IRX3 and IRX5 had consistently higher mean expression in white adipose tissue from nine participants, as well as negative correlation with PGC1A and UCP1expression, as assessed with the use of interindividual expression patterns in perithyroid brown adipose tissue (Fig. S2B and S2C in the Supplementary Appendix); these findings indicated potential roles for IRX3 and IRX5 in the repression of thermogenesis.

To examine the trans-eQTL genetic control of energy balance by the FTOobesity locus, we used primary preadipocytes from risk-allele carriers and nonrisk-allele carriers to evaluate the genes with mitochondrial and FXR and RXR functions that had expression patterns most closely correlated with those of IRX3 and IRX5, as well as several known markers of energy-balance regulation (Fig. S2D and S2E in the Supplementary Appendix). As compared with nonrisk-allele carriers, risk-allele carriers had lower expression of mitochondrial, browning, and respiration genes and higher expression of lipid-storage markers, which indicated a shift from energy dissipation to energy storage.

These differences in expression were also reflected in the cellular signatures of obesity. Risk-allele carriers had increased adipocyte size, reduced mitochondrial DNA content, and a loss of UCP1 response to β-adrenergic stimulus or cold exposure (Figure 3B and 3C, and Fig. S2F in theSupplementary Appendix), as well as resistance to isoproterenol-mediated uncoupling, a decreased basal oxygen consumption rate, and a reduction in mitochondrial thermogenesis by a factor of 5 (Fig. S2G in the Supplementary Appendix); this indicated excessive accumulation of triglycerides, reduced mitochondrial oxidative capacity, reduced white adipocyte browning, and reduced thermogenesis.

Adipocyte-Autonomous Effects of IRX3 and IRX5 on Energy Balance

We next quantified the effect that manipulation of IRX3 and IRX5 expression had on thermogenesis in primary preadipocytes that were isolated from both risk-allele carriers and nonrisk-allele carriers. In preadipocytes from risk-allele carriers, IRX3 and IRX5 knockdown restored oxygen consumption and thermogenesis response to nonrisk levels, increased thermogenesis by a factor of 7 (Figure 3D), and restored UCP1 expression levels (Fig. S3A in the Supplementary Appendix). In preadipocytes from nonrisk-allele carriers, IRX3 and IRX5 overexpression reduced basal respiration and thermogenesis to risk-allele levels (with thermogenesis reduced by a factor of 8) (Figure 3D) and decreased the expression of UCP1, other regulators of mitochondrial function and thermogenesis (PGC1A, PGC1B, and PRDM16), and the β-adrenergic receptor (ADRB3), which also regulates UCP1-independent thermogenesis programs (Fig. S3B and S3C in the Supplementary Appendix). These manipulations had no significant effect on preadipocytes from participants with the reciprocal genotypes, which indicated that IRX3 and IRX5 levels recapitulate the effect that the FTO genetic variant has on thermogenesis.

To examine the organism-level effects of the repression of Irx3 in adipose tissue, we used adipose Irx3 dominant-negative (aP2-Irx3DN) mice. These mice had pronounced antiobesity characteristics, including reduced body size, body weight, fat mass, white and brown fat depots, and adipocyte size (Fig. S4A through S4G in the Supplementary Appendix). These aP2-Irx3DN mice also had resistance to weight gain on a high-fat diet, increased energy expenditure both at night and during the day, and increased oxygen consumption both at room temperature (22°C) and in thermoneutral conditions (30°C), but they did not have significant differences from control mice in food intake or locomotor activity (Fig. S4A and S4H through S4L in the Supplementary Appendix). At the molecular and cellular levels, these mice had increased mitochondrial activity and thermogenesis marker expression, reduced lipid-storage marker expression in both white and brown fat compartments, and markedly smaller adipocytes than did control mice (Fig. S4M, S4N, and S4O in the Supplementary Appendix).

Figure 4. Disruption of a Conserved ARID5B Repressor Motif by Causal SNV rs1421085 in Humans.

Panel A shows disruption of an ARID5B repressor motif in the evolutionarily conserved motif module surrounding rs1421085. The sequences shown at the top of the panel indicate the frequencies of each nucleotide, with the size scaled to indicate the information content (measured as entropy) at each position. Panel B shows adapted phylogenetic module complexity analysis (PMCA)25 scores in the FTO region for all 82 noncoding SNPs in LD (r2≥0.8) with tag SNV rs1558902, which was identified in a genomewide association study26; rs1421085 had the maximal score. Chromatin state annotation is shown for Roadmap Epigenomics reference genome E025, which corresponds to adipose-derived mesenchymal stem cells; for information on the colors used to denote chromatin states, see Figure S1A in the Supplementary Appendix. Panel C shows increased endogenous expression of IRX3 and IRX5 on single-nucleotide T-to-C editing of rs1421085 in the nonrisk haplotype of a nonrisk-allele carrier, using CRISPR–Cas9 (five clonal expansions). CRISPR–Cas9 re-editing from the engineered C risk allele back to a T nonrisk allele with the use of an alternative single guide RNA restores low endogenous IRX3 and IRX5 gene expression. Panel D shows reduced expression of IRX3 and IRX5 on C-to-T editing of the risk allele in adipocyte progenitors from a risk-allele carrier. Knockdown of ARID5B increases IRX3 and IRX5 levels, as compared….

We next evaluated the tissue-autonomous versus brain-mediated roles of Irx3 by comparing the aP2-Irx3DN mice with hypothalamus dominant-negative Ins2-Irx3DN mice.19 The aP2-Irx3DN mice had a reduction in fat-mass ratio that was 3 times as great as that in Ins2-Irx3DN mice (a reduction of 57% vs. 19%), despite the fact that transgene expression in the hypothalamus was 3 times lower than that in Ins2-Irx3DN mice (Fig. S4P and S4Q in the Supplementary Appendix), which indicated that Irx3 has a hypothalamus-independent regulatory role in whole-body energy regulation. The phenotypic effects of Irx3 repression in aP2-Irx3DN mice were also stronger than those in whole-body Irx3 knockout mice, which suggested potential dominant repressor effects in adipocytes or other tissues, and were independent of Fto gene expression, which did not change (Fig. S4P and S4R in the Supplementary Appendix).

Our findings indicate that both Irx3 and Irx5 have cell-autonomous roles: manipulation of Irx3 andIrx5 led to energy-balance differences in three mouse cellular models, including mouse embryonic fibroblast–derived adipocytes, white 3T3-L1 preadipocytes, and β-adrenergic–stimulated beige ME3 preadipocytes (Fig. S5 in the Supplementary Appendix). In each case, our results indicated that Irx3 and Irx5 induced adipocyte lipid accumulation and repressed thermogenesis in a cell-autonomous way.


Determination of the Causal Variant and Disruption of Repression by ARID5B

To predict the causal variant, the disruption of which is necessary and sufficient to cause IRX3 andIRX5 dysregulation in human preadipocytes, we used phylogenetic module complexity analysis (PMCA)25

(Figure 4A 


Disruption of a Conserved ARID5B Repressor Motif by Causal SNV rs1421085 in Humans., and Fig. S6A and S6B in the Supplementary Appendix). The highest PMCA score was found for the rs1421085 T-to-C SNV, which is in perfect linkage disequilibrium with the most significant reported SNV, rs1558902, across multiple populations (1000 Genomes Phase 1 data), a finding that is consistent with a potentially causal role.

To evaluate whether rs1421085 plays a causal role in enhancer activity, we introduced the C allele into the nonrisk haplotype in our luciferase reporter assay. The T-to-C single-nucleotide alteration increased enhancer activity levels for 10-kb and 1-kb segments centered on the variant, in both orientations and both upstream and downstream of the transcription start, which indicated a gain of enhancer activity in association with the rs1421085 risk allele (Fig. S6C and S6D in the Supplementary Appendix).

To evaluate the effect of the variant on regulator binding, we used electrophoretic mobility-shift assays (EMSAs) of adipocyte nuclear extract with probes for the risk allele and the nonrisk allele of rs1421085. We found binding for the nonrisk allele, T, which lacked enhancer activity, but no binding for the risk allele, C; this indicated that the increased enhancer activity associated with the risk allele is probably due to a loss of repressor binding rather than to a gain of activator binding (Fig. S6E in the Supplementary Appendix).

We examined disrupted motifs and regulator expression to identify potential upstream regulators. The T-to-C substitution disrupted conserved motifs for NKX6-3, LHX6, and the ARID family of regulators (Figure 4A). Among them, ARID5B had the highest expression in adipose tissue and adipocytes and was bound specifically to the nonrisk allele in EMSA competition experiments (Fig. S6E and S6F in the Supplementary Appendix). ARID5B is known to play both repressive and activating roles and was previously implicated in adipogenesis and lipid metabolism in mice.27,28. Among nonrisk-allele carriers, expression of ARID5B was negatively correlated with expression ofIRX3 and IRX5, a finding consistent with ARID5B having a repressive role. No correlation was found in risk-allele carriers, which indicates a loss of ARID5B regulation (Fig. S6G in the Supplementary Appendix).

To evaluate the causal role of ARID5B, we next examined the effects of its knockdown and overexpression on IRX3 and IRX5. ARID5B knockdown increased IRX3 and IRX5 expression in primary preadipocytes from nonrisk-allele carriers to risk-allele levels, which indicates a loss of repression, but it had no effect on preadipocytes from risk-allele carriers, which indicates epistasis with the obesity-risk haplotype (Fig. S6H in the Supplementary Appendix). Consistent with this finding, in SGBS enhancer assays, ARID5B knockdown increased the activity of preadipocytes with the nonrisk allele to risk-allele levels, which indicates a loss of repression, but had no effect on risk-allele constructs, indicating epistasis with the rs1421085 risk allele (Fig. S6I in the Supplementary Appendix). ARID5B overexpression further reduced IRX3 and IRX5 levels in nonrisk-allele carriers, which indicated that repression was strengthened, but had no significant effect on risk-allele carriers, a finding consistent with impaired ARID5B repression in association with the risk haplotype (Fig. S6J in the Supplementary Appendix).

We also evaluated the cellular effects of ARID5B-directed perturbations in primary preadipocytes from risk-allele carriers and nonrisk-allele carriers. In preadipocytes from nonrisk-allele carriers,ARID5B knockdown reduced basal oxygen consumption and lipolysis (Fig. S6K and S6L in theSupplementary Appendix) and shifted expression patterns from mitochondrial to lipid markers (Fig. S2E in the Supplementary Appendix), which indicated that ARID5B plays causal roles in energy-balance regulation. In contrast, ARID5B knockdown had no effect on preadipocytes from risk-allele carriers, a finding consistent with a loss of ARID5B control.

These results suggest that the FTO obesity variant acts through disruption of ARID5B binding in the risk haplotype, leading to a loss of repression, a gain of enhancer activity, and increases inIRX3 and IRX5 expression (Fig. S6M in the Supplementary Appendix).


C-to-T Editing of the rs1421085 Risk Variant and the Effect on Thermogenesis

Targeted genome editing technology involving CRISPR–Cas929 makes it possible to test the phenotypic effect of altering the predicted causal nucleotide rs1421085 in its endogenous genomic context, in isolation from the other obesity-associated genetic variants in the same haplotype. We used CRISPR–Cas9 in primary preadipocytes with two separate guide RNAs, one for rs1421085 C-to-T rescue of the ARID5B motif disruption in risk-allele carriers and one for rs1421085 T-to-C disruption of the ARID5B motif in nonrisk-allele carriers.

We first evaluated the effect of rs1421085 editing on IRX3 and IRX5 expression levels. Starting from preadipocytes of a nonrisk-allele carrier, T-to-C editing doubled endogenous IRX3 and IRX5expression, to levels seen in risk-allele carriers; starting from the edited preadipocytes, C-to-T re-editing back to the nonrisk allele restored low expression levels (Figure 4C). Starting from the risk haplotype, C-to-T editing reduced IRX3 and IRX5 to nonrisk-allele levels, but only in the presence of ARID5B (Figure 4D); this established that disruption of ARID5B repression by rs1421085 is the mechanistic basis of the IRX3 and IRX5 dysregulatory event that mediates the effects of the FTOlocus on obesity.

Next, we evaluated the role of rs1421085 editing during differentiation of white and beige adipocytes, by studying differences in expression between edited and unedited preadipocytes during differentiation. Unedited adipocytes from a risk-allele carrier had a peak in IRX3 and IRX5expression during days 0 and 2 of preadipocyte differentiation into adipocytes; expression during early differentiation was reduced to nonrisk-allele levels by rs1421085 editing, which indicated a causal role of rs1421085 in developmental gene expression programs.

(Figure 5A


Rescue of Metabolic Effects on Adipocyte Thermogenesis through Editing of SNV rs1421085 in a Risk-Allele Carrier. The causal role of rs1421085 was further reflected in a significant increase in the expression of thermogenesis regulators (ADRB3, DIO2, PGC1A, and UCP1) and mitochondrial markers (NDUFA10, COX7A, and CPT1) in differentiating preadipocytes (Figure 5B), which indicated that C-to-T editing of the risk allele rescued thermogenesis regulatory programs.

Last, we evaluated the role of rs1421085 editing in cellular signatures of obesity by quantifying phenotypic differences between edited and unedited adipocytes. A causal role in the regulation of energy balance was indicated by the fact that C-to-T rescue of rs1421085 in edited adipocytes resulted in a reduction in gene expression for lipid storage and lipolytic markers (Fig. S2E and S8A in the Supplementary Appendix), an increase by a factor of 4 in basal metabolic rate and β-adrenergic oxygen consumption, and an increase by a factor of 7 in thermogenesis (Figure 5C, and Fig. S7B in the Supplementary Appendix). In particular, rescue of the ARID5B motif in C-to-T edited preadipocytes restored the strong dependence of mitochondrial respiration on ARID5B that is seen in nonrisk-allele carriers (Fig. S7C in the Supplementary Appendix).

These results indicate that the rs1421085 T-to-C single-nucleotide alteration underlies the association between FTO and obesity by disrupting ARID5B-mediated repression of IRX3 andIRX5. This disruption leads to a developmental shift from browning to whitening programs and loss of mitochondrial thermogenesis (Figure 5D).


Our work elucidates a potential mechanistic basis for the genetic association between FTO and obesity and indicates that the causal variant rs1421085 can disrupt ARID5B repressor binding; this disruption results in derepression of IRX3 and IRX5 during early adipocyte differentiation. This process could lead to a cell-autonomous shift from white adipocyte browning and thermogenesis to lipid storage, increased fat stores, and body-weight gain.

To translate the results of genomewide association studies into mechanistic insights, we combined public resources (epigenomic annotations, chromosome conformation, and regulatory motif conservation), targeted experiments for risk and nonrisk haplotypes (enhancer tiling, gene expression, and cellular profiling), and directed perturbations in human primary cells and mouse models (regulator–target knockdown and overexpression and CRISPR–Cas9 genome editing). These methods are specific to the elucidation of noncoding variants, which constitute the majority of signals in genomewide association studies; 80% of the trait-associated loci identified in such studies lack protein-altering variants, and 93% of the top hits are noncoding.30

The FTO association with obesity is unusual in many ways. First, rs1421085 has both a high frequency and a strong effect size,31 which suggests positive selection or bottlenecks (e.g., 44% frequency in European populations vs. 5% in African populations). Second, rs1421085 has switchlike behavior in enhancer activity, target-gene expression, and cellular phenotypes, possibly because of selective pressures on energy-balance control for rapid adaptation. Third, rs1421085 acts specifically in the early differentiation of preadipocytes, which emphasizes the importance of profiling diverse tissues, cell types, and developmental stages. Fourth, enhancer activity is found only for the risk allele, which emphasizes the importance of profiling both alleles. Finally, rs1421085 leads to a gain of function (increased enhancer, IRX3, and IRX5 activity); this is a rare property in protein-coding variants but may be common in noncoding variants.

The apparent genetic link between obesity and cell-autonomous adipocyte browning suggests a central role of beige adipocyte thermogenesis in whole-body energy metabolism in humans, a role that is consistent with that suggested in recent reports on PRDM16 in mice.9 IRX3 and IRX5 have evolutionarily conserved roles, and the ARID5B motif lies in a module that is functionally conserved across multiple mammalian species; this indicates that adaptive thermogenesis circuits are conserved, and IRX3 and IRX5 probably play both UCP1-dependent and UCP1-independent roles. Even though IRX3 and IRX5 dysregulation by rs1421085 was restricted to early differentiation, their effects persisted in mature adipocytes, and the targeting of these genes can have broader effects.

Last, we found that direct manipulation of the ARID5B–rs1421085–IRX3/IRX5 regulatory axis in primary cell cultures of adipocytes from patients reversed the signatures of obesity. This indicates that in addition to changes in physical activity and nutrition, manipulation of mitochondrial thermogenesis26 offers a potential third pathway for shifting between energy storage and expenditure in a brain-independent and tissue-autonomous way in humans.

In summary, our work elucidates a mechanistic basis for the strongest genetic association with obesity. Our results indicate that the SNV rs1421085 underlies the genetic association between theFTO locus and obesity. The SNV disrupts an evolutionarily conserved motif for the ARID5B repressor, which leads to loss of binding, derepression of a potent preadipocyte superenhancer, and activation of downstream targets IRX3 and IRX5 during early differentiation of mesenchymal progenitors into adipocyte subtypes. This results in a cell-autonomous shift from white adipocyte browning to lipid-storage gene expression programs and to repression of basal mitochondrial respiration, a decrease in thermogenesis in response to stimulus, and an increase in adipocyte size. Manipulation of the uncovered pathway, including knockdown or overexpression of the upstream regulator ARID5B, genome editing of the predicted causal variant rs1421085, and knockdown or overexpression of target genes IRX3 and IRX5, had a significant effect on obesity phenotypes.


Read Full Post »