(C) PLOS One
This story was originally published by PLOS One and is unaltered.
 . . . . . . . . . .



Exploring the molecular basis of the genetic correlation between body mass index and brain morphological traits [1]

['Daniela Fusco', 'Department Of Neurosciences', 'Rita Levi Montalcini', 'University Of Turin', 'Turin', 'Camilla Marinelli', 'Mathilde André', 'Lucia Troiani', 'Martina Noè', 'Fabrizio Pizzagalli']

Date: 2025-05 

Several studies have demonstrated significant phenotypic and genetic correlations between body mass index (BMI) and brain morphological traits derived from structural magnetic resonance imaging (sMRI). We use the sMRI, BMI, and genetic data collected by the UK Biobank to systematically compute the genetic correlations between area, volume, and thickness measurements of hundreds of brain structures on one hand, and BMI on the other. In agreement with previous literature, we find many such measurements to have negative genetic correlation with BMI. We then dissect the molecular mechanisms underlying such correlations using brain eQTL data and summary-based Mendelian randomization, thus producing an atlas of genes whose genetically regulated expression in brain tissues is pleiotropic with brain morphology and BMI. Fine-mapping followed by colocalization analysis allows, in several cases, the identification of credible sets of variants likely to be causal for both the macroscopic phenotypes and for gene expression. In particular, epigenetic fine mapping identifies variant rs7187776 in the 5 ′ UTR of the TUFM gene as likely to be causal of increased BMI and decreased caudate volume, possibly through the creation, by the alternate allele, of an ETS binding site leading to increased chromatin accessibility, specifically in microglial cells.

Obesity is linked to many chronic diseases and its prevalence worldwide is increasing. Susceptibility to obesity is known to be due, to some degree, to genetic factors, and such genetic contribution is mediated in part by the brain through the control of food intake. Here we investigate the genetic variants that affect both obesity and brain morphology, identifying 21 genes whose expression affects both body mass index and morphological measures obtained from brain magnetic resonance imaging data. These results provide new insight into the complex relationships between the brain and obesity, also in view of the recent development of revolutionary anti-obesity drugs, whose mechanisms of action are believed to include effects on the reward-related areas of the brain.

Funding: This work was supported in part by a Fondazione CRT grant (grant 2021.1787 to P.P.). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Data Availability: GWAS summary statistics of brain imaging-derived phenotypes (IDPs) were obtained from the Oxford Brain Imaging Genetics Server – BIG 40 – and specifically from https://open.win.ox.ac.uk/ukbiobank/big40/release2/stats33k/0001.txt.gz . GWAS summary statistics for BMI were obtained from the Neale Lab website ( http://www.nealelab.is/uk-biobank ), and specifically from https://broad-ukb-sumstats-us-east-1.s3.amazonaws.com/round2/additive-tsvs/21001_raw.gwas.imputed_v3.both_sexes.tsv.bgz Individual-level data were obtained from the UK BioBank ( https://www.ukbiobank.ac.uk/ ) under application 86275. Individual-level data for GTEx subjects were obtained from dbGAP ( https://www.ncbi.nlm.nih.gov/gap/ ) under project #18670. Genetic correlation data, gene/trait associations, and chromatin accessibility predictions produced for this work are available in the Supporting information.

Copyright: © 2025 Fusco et al. This is an open access article distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Here, we systematically investigated the genetic correlation between BMI and brain morphology using GWAS summary data from the UK Biobank. To investigate the molecular mechanisms behind these correlations, we used summary-based Mendelian randomization (SMR) to identify the genes whose genetically regulated expression (GReX) in brain tissues is pleiotropic with BMI and brain morphology, suggesting a mediating effect. Finally, we used fine-mapping and colocalization to identify cases in which a single variant is causal of gene expression, BMI, and brain morphology, and epigenetic fine-mapping based on single-cell chromatin accessibility data to formulate mechanistic hypotheses on the relevant regulatory mechanisms and cell types.

Finally, a study integrating single-cell-RNA-sequencing and genome-wide association studies (GWAS) has shown that susceptibility to obesity is enriched for some hypothalamic cell types such as VMH Sf1-expressing neurons, most of which are involved in the integration of sensory stimuli, learning and memory. Despite neuroanatomical differences, these brain cell types share transcriptional patterns related to obesity [ 18 ].

To date, most studies have focused on phenotypic correlation. Regarding genetic correlation, a study explored the effects of a collection of obesity-related SNPs on 164 regional brain volume traits from the UK Biobank, finding that 17 such SNPs were associated with 51 regional brain volumes (both positively and negatively) [ 16 ]. A bivariate linkage and quantitative analysis on Mexican American individuals looked for genetic factors associated with increased BMI and reduced cortical surface area and subcortical volume, localizing two genome-wide significant QTLs at 17p13.1 and 3q22.1. The former pleiotropically affects ventral diencephalon volume and BMI, suggesting the involvement of such region in obesity through leptin-induced signaling in the hypothalamus. The latter involves the surface area of the supramarginal gyrus and BMI and might be relevant to the food-related reward mechanism [ 17 ].

The relationship between BMI and white matter integrity is more difficult to characterize since alterations show a more complex pattern [ 15 ]. Studies involving diffusion tensor imaging (DTI) investigated the influence of obesity on white matter integrity by comparing BMI with DTI parameters in adults, revealing a negative correlation between several microstructure architectures of the white matter, such as in the corpus callosum, and BMI [ 15 ].

On the other hand, a positive correlation between gray matter volume and BMI was found in nucleus accumbens and hypothalamus [ 14 ], while a twin study identified positive correlation with ventromedial prefrontal cortex and the right cerebellum [ 9 ].

Beyond the aforementioned results, several neuroimaging studies investigated the correlation between obesity and structural alterations in brain regions, especially differences in gray matter volumes. While the results of these studies have been somewhat conflicting, with both positive and negative correlations reported between gray matter volumes and obesity, most neuroimaging studies reported reduced gray matter volume in many brain regions, including several involved in executive control, to be associated with higher BMI [ 5 ]. In particular, reduced volume of cerebellum [ 6 , 7 ], basal ganglia [ 8 ], putamen [ 7 ], prefrontal cortex, temporal lobes, and subcortical structures [ 9 ], as well as reduced cortical thickness, were found to be associated with higher BMI [ 10 – 13 ].

The central nervous system has a role in susceptibility to obesity through the control of food intake [ 2 ]. This notion is strengthened by the fact that the heritability of body mass index (BMI), a parameter typically used by the World Health Organization to distinguish normal-weight (18.5 ≤ BMI ≤ 25) from obese people (BMI ≥ 30), was found to be enriched in genes expressed in the brain and central nervous system [ 3 , 4 ]. In particular, a recent enrichment analysis of the genes located near 97 BMI-associated single-nucleotide polymorphisms (SNPs), aimed at identifying tissues and cell types with high expression of such genes, revealed that 27 out of 31 significantly enriched tissues were part of the central nervous system. However, these results are not enough to identify the specific brain regions involved [ 4 ].

In 2022 one out of eight people were diagnosed with obesity worldwide, amounting to over 890 million adults and 160 million children and adolescents living with this chronic complex disease. In the majority of cases, obesity is a combination of environmental, psycho-social, and genetic factors. If the current increasing trend continues, 60% of the entire human population is estimated to be overweight or obese by 2030 [ 1 ].

Results

Genetic correlation In order to assess the shared genetic basis between brain morphology and BMI, we relied on variant-trait associations for 435 cortical and subcortical measurements estimated on more than 30 thousand donors from the UK Biobank [19]. These measurements included volume, thickness, areas from white and pial surfaces, gray-white matter contrast for 54 cortical regions [20]; volumes and mean intensities for subcortical segments [21] (see Methods for details on the selection of traits). Adopting cross-trait LD Score regression [22], the state-of-the-art summary-based approach, we found 108 nominally significant (P < 0.05) genetic correlations ( ) across all categories. Most of the significant correlations turned out to be negative (Fig 1A), including in particular those between BMI and cortical areas (white and pial surfaces), cortical and subcortical volumes (all genetic correlation values are reported in S1 Data). The most significant correlations (P < 0 . 001 in at least one hemisphere) were all negative (Fig 1B), suggesting that, overall, genetic variants associated with increased BMI are also related to reduced size of brain structures. This is evident from the global subcortical volume and the globus pallidus in particular, and the total areas of pial and white surfaces, with especially negative values in cortical regions adjacent to the central sulcus, and in the temporal and cingulate lobes (Fig 1B). As expected, results for the two hemispheres are largely consistent with each other (Fig 1C). PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 1. Genetic correlations between sMRI traits and BMI. (A) Genetic correlations (y axis) of all 435 sMRI traits with BMI separated by category. n, number of traits for each category. (B) Traits with the most significant (P<0.001) genetic correlation (x axis) with BMI, grouped by category. A few subcortical volume traits were removed to avoid redundancy. Measurements refer to the left (blue) or right (red) emisphere, or are global (green). Significant cortical regions are indicated on the brain surfaces on the right, which are colored according to their genetic correlation between pial surface area and BMI (left-right averages). (C) Comparison between genetic correlations in left (x axis) and right (y axis) hemispheres. Dots are colored based on their significance level. Diamonds represent correlations with P<0.001 in either or both hemispheres. https://doi.org/10.1371/journal.pgen.1011658.g001 Subcortical intensity traits, as provided by the UK Biobank (UKB), were analyzed, revealing a significant negative genetic correlation for brainstem intensity (P<0.001). However, interpreting this association is not straightforward. Although intensity values partially reflect what voxel-based morphometry (VBM) measures, VBM leverages the optimized intensity contrast among gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF) tissues from T1-weighted images to classify voxels and estimate tissue volumes. Voxel intensity itself, however, can be influenced by MRI scanner artifacts and other sources of noise, introducing biases which are difficult to control. To address this, studies employing this method typically perform statistical analyses after normalizing tissue intensities, thereby estimating the contrasts between different tissues [23].

Summary-based Mendelian randomization We hypothesized the existence of molecular mechanisms at the level of gene expression underlying these genetic correlations. Mendelian randomization [24] can help elucidate the causal effect of an exposure on an outcome. Thanks to large databases of expression quantitative trait loci (eQTLs), gene expression can be used as an exposure, allowing the identification of putative genes whose regulation is pleiotropically related to a complex trait, suggesting a mediation effect of their expression. We used summary-based Mendelian randomization (SMR) [25] with gene expression in each of 13 brain tissues included in the GTEx database [26] as the exposure and sMRI traits and BMI as outcomes, reasoning that genes associated with both sMRI and BMI in such analysis could provide a basis for the investigation of the molecular mechanisms driving the genetic correlation between the two types of traits. SMR is unable, by itself, to distinguish between a causal relationship and a pleiotropic one, that is when gene expression and complex trait are influenced by the same genetic variants but not causally related. In addition, it can identify relationships which are actually mediated by different causal variants for exposure and outcome, when they are in linkage disequilibrium (LD) with each other. For this reason, SMR is often coupled with a test for heterogeneity in dependent instruments (HEIDI) [25], which can rule out a LD-mediated relationship. A SMR + HEIDI analysis using the 435 sMRI traits and 13 GTEx brain tissues [26] found several genes potentially mediating the shared genetic components of BMI and sMRI traits through their expression (Fig 2A, respectively blue and green dots). A total of 21 genes (Fig 2A, red dots; S2 Data) were associated with both BMI and one or more sMRI traits. We confirmed the gene/trait associations found by SMR through an individual-level study of the association between GReX and traits, as in transcriptome-wide association studies (TWAS) [27]: the genetic component of gene expression for the 21 genes was predicted in the relevant brain tissues for the UKB subjects for which the relevant phenotype was available (BMI or sMRI), and its correlation with the trait was computed. For BMI the TWAS association was nominally significant in 23 out of 34 cases, and the sign of the correlation was concordant with the one found by SMR for all of them. For sMRI traits, nominal significance and sign concordance was achieved in all of the 687 tested associations. We asked whether the overlap between BMI- and sMRI-associated genes was greater than expected by chance. Since the significant genes for all traits are clearly clustered into genomic loci (see e.g. the loci in chromosomes 8 and 16 in Fig 2A, the latter including the well-known obesity associated FTO gene [28]), we devised a circular permutation procedure preserving the gene order (see Methods) which revealed that the overlap between BMI- and sMRI-associated genes is indeed greater than expected by chance (empirical ). Furthermore, the cis- regions around the pleiotropic genes were enriched in genetic covariance between BMI and sMRI traits, although only nominally and for just 10 sMRI traits out of 97 tested (Monte-Carlo P-value from partial genetic covariance estimates). Nevertheless, as shown in Fig 2B, when considering the genetic regions surrounding all genes mediating BMI, i.e. without restricting to the 21 found to also mediate sMRI traits, we could not observe any significant enrichment: their estimated log-fold changes were markedly smaller and never distinguishable from 0. This suggests that the portion of the genome surrounding the 21 identified pleiotropic genes, albeit small, indeed carries a larger-than-expected proportion of the shared genetic determinants of brain morphology and BMI. PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 2. SMR reveals common molecular patterns. (A) Chicago plot showing the posterior probability of association (PPA, y axis) of gene expression on sMRI traits (top) and BMI (bottom) according to SMR. The x axis represents the genomic position. For each gene-sMRI trait, the reported PPA is the highest among all tissue/trait combinations. Red dots represent genes whose expression is associated with BMI and at least one sMRI trait in at least one tissue. (B) Volcano plot showing the local genetic covariance enrichment of each sMRI trait and its respective p-value. The enrichment is referred to the genomic regions surrounding the genes found by SMR to be in pleiotropy with BMI (green dots) or with both BMI and sMRI traits (red dots). Diamonds represent sMRI traits with significant (P<0.001) genetic correlation with BMI. (C) Heatmap representation of the number of genes whose genetically determined tissue expression pleiotropically affects BMI and each sMRI trait. sMRI traits refer to the left (blue) or right (red) emisphere, or are global (green) https://doi.org/10.1371/journal.pgen.1011658.g002 As shown in S1 Fig, there is a dense network of genetic correlations across the measurements from different brain regions for which we found pleiotropic genes with BMI. This is in line with a common genetic basis across many sMRI traits: it is therefore non-trivial to quantitatively pinpoint which morphological traits are the most affected by these genes. However, all categories show at least one pleiotropic gene (Fig 2C), and even though cortical gray-white contrast measurements seem very prominent, this can be readily explained by an especially strong genetic correlation among them (S1 Fig). Similarly, all GTEx tissues are involved, although with a suggestive prominence of subcortical tissues. SMR also provides an estimate of the direction and size of the effect of gene expression on each trait. We observed that in most cases the effects of gene expression on BMI and sMRI traits have opposite signs (S2 Fig), in agreement with the prevalence of negative genetic correlations. Taken together, these results suggest that the genetic correlations between BMI and morphological brain measures are indeed driven in part by regulatory variants acting on brain gene expression, and provide us with a list of candidate genes for the more detailed variant-level investigations described in the following.

[END]
---
[1] Url: https://journals.plos.org/plosgenetics/article?id=10.1371/journal.pgen.1011658

Published and (C) by PLOS One
Content appears here under this condition or license: Creative Commons - Attribution BY 4.0.

via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/