Predicting mechanisms of action at genetic loci associated with discordant effects on type 2 diabetes and abdominal fat accumulation
Curation statements for this article:-
Curated by eLife
Evaluation Summary:
This study reports candidate causal genes in genome-wide association studies that exhibit a discordant pattern of association, namely a higher waist-hip ratio simultaneously with a lowered risk of type 2 diabetes. Identification of such genes could provide insights into why some individuals with obesity are not developing type 2 diabetes, knowledge that ultimately could shed light on the complex interplay between fat distribution and type 2 diabetes. The work is of relevance to the fields of genetics of diabetes and obesity.
(This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #1 and Reviewer #2 agreed to share their name with the authors.)
This article has been Reviewed by the following groups
Listed in
- Evaluated articles (eLife)
Abstract
Metabolic syndrome (MetSyn) is a cluster of dysregulated metabolic conditions that occur together to increase the risk for cardiometabolic disorders such as type 2 diabetes (T2D). One key condition associated with MetSyn, abdominal obesity, is measured by computing the ratio of waist-to-hip circumference adjusted for the body-mass index (WHRadjBMI). WHRadjBMI and T2D are complex traits with genetic and environmental components, which has enabled genome-wide association studies (GWAS) to identify hundreds of loci associated with both. Statistical genetics analyses of these GWAS have predicted that WHRadjBMI is a strong causal risk factor of T2D and that these traits share genetic architecture at many loci. To date, no variants have been described that are simultaneously associated with protection from T2D but with increased abdominal obesity. Here, we used colocalization analysis to identify genetic variants with a shared association for T2D and abdominal obesity. This analysis revealed the presence of five loci associated with discordant effects on T2D and abdominal obesity. The alleles of the lead genetic variants in these loci that were protective against T2D were also associated with increased abdominal obesity. We further used publicly available expression, epigenomic, and genetic regulatory data to predict the effector genes (eGenes) and functional tissues at the 2p21, 5q21.1, and 19q13.11 loci. We also computed the correlation between the subcutaneous adipose tissue (SAT) expression of predicted effector genes (eGenes) with metabolic phenotypes and adipogenesis. We proposed a model to resolve the discordant effects at the 5q21.1 locus. We find that eGenes gypsy retrotransposon integrase 1 ( GIN1 ), diphosphoinositol pentakisphosphate kinase 2 (PPIP5K2), and peptidylglycine alpha-amidating monooxygenase ( PAM ) represent the likely causal eGenes at the 5q21.1 locus. Taken together, these results are the first to describe a potential mechanism through which a genetic variant can confer increased abdominal obesity but protection from T2D risk. Understanding precisely how and which genetic variants confer increased risk for MetSyn will develop the basic science needed to design novel therapeutics for metabolic syndrome.
Article activity feed
-
-
-
-
Evaluation Summary:
This study reports candidate causal genes in genome-wide association studies that exhibit a discordant pattern of association, namely a higher waist-hip ratio simultaneously with a lowered risk of type 2 diabetes. Identification of such genes could provide insights into why some individuals with obesity are not developing type 2 diabetes, knowledge that ultimately could shed light on the complex interplay between fat distribution and type 2 diabetes. The work is of relevance to the fields of genetics of diabetes and obesity.
(This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #1 and Reviewer #2 agreed to share their name with the authors.)
-
Reviewer #1 (Public Review):
Aberra et al.'s work is focused on identifying genes that exhibit opposing effects on type 2 diabetes and abdominal obesity. Identification of such genes would provide stepping stones for a better understanding of why some individuals with obesity are not developing type 2 diabetes, knowledge that ultimately could shed light on the complex interplay between fat distribution and type 2 diabetes.
Aberra et al. use a number of computational tools to identify genetic variants associated with both type 2 diabetes and waist-hip ratio (both adjusted for body mass index). They identify six genetic loci that associate with both phenotypes, but exhibit discordant effects.
To better understand which tissues and genes are potentially mediating the discordant effects, the authors use GTEx data to co-localize eQTLs with …
Reviewer #1 (Public Review):
Aberra et al.'s work is focused on identifying genes that exhibit opposing effects on type 2 diabetes and abdominal obesity. Identification of such genes would provide stepping stones for a better understanding of why some individuals with obesity are not developing type 2 diabetes, knowledge that ultimately could shed light on the complex interplay between fat distribution and type 2 diabetes.
Aberra et al. use a number of computational tools to identify genetic variants associated with both type 2 diabetes and waist-hip ratio (both adjusted for body mass index). They identify six genetic loci that associate with both phenotypes, but exhibit discordant effects.
To better understand which tissues and genes are potentially mediating the discordant effects, the authors use GTEx data to co-localize eQTLs with genetic variants at the six discordant loci. They identified four genes, at two of the discordant loci, that are regulated by an eQTL co-localizing with one of the discordant variants. Using the Finnish METSIM cohort and correlation analysis, the authors show that expression of these genes is associated with both glycemic and obesity risk-phenotypes.
The manuscript is very concise and well-written. All computational analyses seem well thought through and executed. I have two suggestions that potentially could help the authors to improve their work.
The authors write that they "[...] predict the mechanisms of action at discordant loci" (L. 286), which seems too strong a claim given their data. Potentially the following points could help to provide more evidence on the functional context to the four prioritized genes and more guidance on how mechanistic insights could be advanced further:
Aberra et al. indicate that the 2p21 locus harboring the THADA gene and its antisense RNA are differentially open during preadipocyte development. Are these RNAs differentially expressed during specific stages of adipocyte development and are they differentially expressed in certain human adipocyte clusters? Relevant datasets to address these questions could be (https://www.nature.com/articles/s41467-020-16019-9, https://pubmed.ncbi.nlm.nih.gov/32066997/, https://pubmed.ncbi.nlm.nih.gov/35296864/, and https://pubmed.ncbi.nlm.nih.gov/33116305/.
The authors' work may serve as an example on how to shortlist relevant genetic variants for variant-to-function approaches. It could be instructive to the metabolism community if the authors' in the Discussion could dedicate a paragraph to carefully discuss how one best could further explore the function of the discordant variants they identify and the genes they implicate. For instance, how could one (i) experimentally prove that the given variants regulate the predicted effector genes, (ii) further understand the mechanisms with which they impact adipocyte biology, and (iii) further establish evidence that they have a discordant effect on glycemic and lipid traits.
-
Reviewer #2 (Public Review):
In this paper, the authors used co-localization analyses to identify association signals that show directionally contradictory associations (from a phenotypic perspective) with abdominal adiposity (i.e., waist-to-hip ratio adjusted for BMI [WHRajdBMI]) and risk of type-2 diabetes adjusted for BMI (T2DadjBMI). Of the 79 co-localizing association signals identified, six show seemingly discordant associations, where the allele associated with lower WHRajdBMI is associated with higher risk of T2DadjBMI. The authors state that genes in these six loci could explain the phenomenon where some individuals are obese yet metabolically healthy, while some others are lean but insulin resistant. By subsequently examining co-localization of 99% credible variants in these loci (i.e., variants with a high posterior …
Reviewer #2 (Public Review):
In this paper, the authors used co-localization analyses to identify association signals that show directionally contradictory associations (from a phenotypic perspective) with abdominal adiposity (i.e., waist-to-hip ratio adjusted for BMI [WHRajdBMI]) and risk of type-2 diabetes adjusted for BMI (T2DadjBMI). Of the 79 co-localizing association signals identified, six show seemingly discordant associations, where the allele associated with lower WHRajdBMI is associated with higher risk of T2DadjBMI. The authors state that genes in these six loci could explain the phenomenon where some individuals are obese yet metabolically healthy, while some others are lean but insulin resistant. By subsequently examining co-localization of 99% credible variants in these loci (i.e., variants with a high posterior probability of being causal) with expression and splicing quantitative trait loci (eQTLs and sQTLs), the authors aimed to pinpoint genes that may be responsible for the uncoupling of abdominal lipid deposition and glucose metabolism.
The authors used state of the art statistical approaches throughout to identify variants that show unexpected directions of association with abdominal obesity vs. risk of type-2 diabetes, as well as to pinpoint potential causal genes within these loci. However, there are several factors that limit the conclusions one can draw from the current results that can still be addressed.
Firstly, it is difficult to appreciate exactly what associations mean for an outcome that is a ratio of two traits adjusted for a third trait. Also, adjusting for a heritable trait (BMI) may introduce collider bias. Hence, in addition to associations with WHRadjBMI and T2DadjBMI, it would be informative to also see summary statistics for associations of these loci with waist circumference, hip circumference, BMI and T2D. Furthermore, in a best-case scenario, WHRadjBMI reflects abdominal obesity, but it represents the net sum of adiposity across all individual depots. It does not tell us which lipid depot(s) drive(s) the association with WHRadjBMI. It may be possible to get one step closer to answering this important question using MRI derived information from e.g. UK Biobank data. Furthermore, associations don't provide insights into causal pathways or directions of effect. That is, if we assume that a putative causal variant acts through only one causal gene, then higher expression of that gene may increase abdominal adiposity and reduce the risk of type-2 diabetes, or it may reduce abdominal adiposity and increase the risk of type-2 diabetes, for example. Even if there is only one causal gene, such effects may be mediated by different tissues and cell types, i.e., different mechanisms. Since all six loci showing directionally discordant associations with WHRadjBMI and T2DadjBMI are located in non-coding regions, these loci likely have a regulatory role, e.g. as enhancers. Hence, the seemingly directionally discordant associations may well be mediated by more than one causal gene, which would not reflect uncoupling. Furthermore, co-localization of credible variants with sQTLs and eQTLs does not imply that such genes are causal for the traits of interest, as the association between the expression of that gene and the outcome can be confounded by the expression of the actual causal gene, with the expression of both genes influenced by the causal variant in the regulatory region. To get one step closer to such causal inference, it would be informative if the authors could show chromatin-chromatin interaction data in relevant cell types for credible variants in all six discordant loci. In addition, gene prioritization is a fast-evolving field, and a range of bioinformatics approaches - but no gold standard - are available to prioritize candidate genes. Results using proof-of-concept loci have not shown very convincing results for the predictive ability of eQTL associations to flag causal genes, even in the presence of co-localization, so it would be informative to see which genes are prioritized in these loci by other available approaches - including but not limited to chromatin accessibility QTLs in relevant tissues - and if there is consensus across those approaches as to which gene(s) are likely causal. Finally, even if there are loci that uncouple genetic effects on abdominal adiposity and risk of type-2 diabetes, the causal beneficial effect of weight loss is so large in comparison that it would no doubt remain advisable for most if not all individuals to lose weight to reduce the risk of T2D and its complications.
-
Reviewer #3 (Public Review):
The authors have used GWAS summary results for WHR adj. BMI and T2D-risk adj. BMI to identify genome-wide significant loci that show a discordant pattern of association with the traits: higher WHRadjBMI and lower risk of T2Dadj.BMI. They identify 6 discordant loci, for which they perform a series of follow up analyses to connect the genetic variants to their causal genes and their target tissues. They find evidence that THADA-AS and GIN1/PAM may be causal genes in two of these discordant loci.
The strength of the study is the extensive work done by the authors to ensure that the discordant associations between WHRadjBMI and T2DadjBMI are colocalized, to fine-map the genetic loci, and to link the genetic variants to their target genes and tissues. The main weakness is the lack of clear biological and clinical …
Reviewer #3 (Public Review):
The authors have used GWAS summary results for WHR adj. BMI and T2D-risk adj. BMI to identify genome-wide significant loci that show a discordant pattern of association with the traits: higher WHRadjBMI and lower risk of T2Dadj.BMI. They identify 6 discordant loci, for which they perform a series of follow up analyses to connect the genetic variants to their causal genes and their target tissues. They find evidence that THADA-AS and GIN1/PAM may be causal genes in two of these discordant loci.
The strength of the study is the extensive work done by the authors to ensure that the discordant associations between WHRadjBMI and T2DadjBMI are colocalized, to fine-map the genetic loci, and to link the genetic variants to their target genes and tissues. The main weakness is the lack of clear biological and clinical rationale for the analyses that have been performed. Furthermore, there are some remaining concerns about the possibility of allele mismatching, as well as specific gaps in the analysis pipeline and unclear statements in the text, which will require clarification. The paper could be of interest to human geneticists and molecular biologists interested in understanding the function of genetic risk variants of cardiometabolic disease.
-