Modulation of Genetic Associations with Serum Urate Levels by Body-Mass-Index in Humans

For decades, geneticists mapped the variants that raise uric acid in the blood, and they consistently found them in study after study. The same loci, the same effect sizes, the same story replicated across populations. Then Huffman and colleagues asked a different question: what happens to those genetic effects when you sort people by body weight? The answer was not what fixed genetic effects are supposed to produce. The strongest known gout gene cut its effect nearly in half. Entirely new loci appeared or flipped direction depending on which body mass index category you looked in. The genome, it turns out, is not reading from the same script in a lean body and an obese one. Serum urate, or the concentration of uric acid circulating in the blood, matters because elevated levels cause gout, associate with kidney stones, and cluster with the components of metabolic syndrome. By the time this study was conducted, genome-wide association studies had already identified twenty-eight loci affecting serum urate, together explaining roughly seven percent of variation between individuals. The two largest loci alone account for about half that genetic variance. One of them, the SLC2A9 gene, encodes a urate transporter with an effect nearly twice as large in women as in men. The other, ABCG2, runs the other way—bigger in men. Sex-specific effects were already known. The question Huffman and colleagues opened was whether body weight produces similar modulation and whether that modulation could be detected genome-wide. Body mass index, or BMI, correlates with serum urate at levels ranging from 0.27 to 0.44 across population studies. This is a substantial relationship. Obesity is described as the strongest modifiable risk factor for hyperuricemia and gout. The mechanistic logic for why BMI might change genetic effects is real: many of the strongest urate-associated variants sit in genes encoding ion transport proteins, and the metabolic shifts that accompany rising BMI—falling serum phosphate, changes in hepatic ATP—can directly or indirectly alter transporter activity. Small earlier studies reported conflicting results about whether this modulation existed, but sample sizes were too limited to settle the question. To resolve it, Huffman and colleagues assembled genome-wide data from twenty-two European-ancestry cohorts, with up to forty-two thousand five hundred sixty-nine participants in total. They pursued two strategies in parallel. The first was BMI-stratified genome-wide association analysis, treating lean participants, defined as a BMI below twenty-five, overweight participants, with a BMI ranging from twenty-five to thirty, and obese participants, defined as a BMI above thirty, as separate discovery samples and running independent meta-analyses within each group. This approach is designed to catch signals that appear or disappear across weight categories. The second strategy was regression-based interaction testing: each model included BMI, the genetic variant, and their product term—the SNP-by-BMI interaction coefficient—which was then meta-analyzed across studies. This catches subtler, linear changes in effect size across the full BMI spectrum, not just categorical jumps. The conventional genome-wide significance threshold of five times ten to the minus eight was applied to both. Interaction tests are harder to power than main-effect tests. As a general rule, detecting a gene-by-environment interaction requires roughly four times the sample size needed for a comparable main effect. The stratified analyses did not uncover new loci with major main effects. No previously unknown variant reached genome-wide significance within a single BMI stratum. What they did reveal was modulation—effect sizes shifting in magnitude, and in some cases direction, as BMI category changed. The interaction-term analyses were more productive. The top result was a signal at the RBFOX3 gene, where the index variant reached genome-wide significance in the combined discovery plus replication meta-analysis, with an interaction coefficient of negative 0.014, a standard error of 0.003, and a p-value of two point six times ten to the minus eight. This represents a small absolute effect, but it is reproducible and clears the hardest statistical bar in the field. RBFOX3 is not a urate gene by any prior accounting. It encodes a neuronal nuclear marker expressed in the arcuate nucleus of the hypothalamus. Its paralog, RBFOX1, has been proposed as an obesity gene, and mouse knockouts show changes in circulating alkaline phosphatase. It is an unexpected address for a BMI-dependent urate signal, which is precisely what makes it interesting—the interaction analysis is pointing somewhere that a straight main-effect search would never have looked. The second top-ranked interaction locus was ERO1LB-EDARADD. In the discovery meta-analysis, the interaction coefficient for the index variant was 0.010 with a p-value of seven point eight times ten to the minus eight; in the combined analysis, it settled at 0.008 with a p-value of two point ninety-four times ten to the minus seven. ERO1LB encodes an endoplasmic reticulum oxidoreductase that responds to unfolded protein stress, a process that ramps up during overnutrition. This provides a plausible biological hook, even if causality remains to be established. Two more loci emerged from the stratified pairwise comparison approach, both in men. A variant in the RBMS1-TANK region reached genome-wide significance for the difference in its urate effect between lean and overweight men, with a p-value of four point seven times ten to the minus eight. The RBMS1-TANK region has prior associations with obesity-related traits, so finding it here with a BMI-dependent urate signal is not random noise. Near TSPYL5, another variant showed a lean versus overweight difference p-value of nine point one times ten to the minus eight in men. TSPYL5 has been linked to the regulation of estradiol produced by adipocytes, giving it a plausible adipose-specific route to urate metabolism. The unifying pattern across all these novel loci is worth pausing on. They show small effects that change direction between BMI strata—not just different magnitudes, but opposite signs. They are essentially invisible in a standard genome-wide association study that pools everyone together because lean individuals and obese individuals pull the signal in opposite directions, and the average comes out near zero. Stratifying by BMI is what lets the signal surface. Now set that pattern against what happened to the most established urate locus. ABCG2, the strongest known gout risk gene, encodes a high-capacity urate transporter in the kidney, liver, and intestines. This gene does not disappear in obese men, but it shrinks dramatically. In lean men, the ninety-five percent confidence interval for the effect of the key ABCG2 variant on standardized urate runs from 0.257 to 0.389. In obese men, that same variant's confidence interval runs from 0.069 to 0.213. The difference between lean and obese men is statistically robust, with a p-value of two times ten to the minus four. The effect is more than halved. A single fixed effect size for ABCG2, applied uniformly across all patients regardless of weight, would substantially overstate the gene's clinical relevance in people with obesity. The pathway analysis adds one more piece to this picture. N-glycan biosynthesis, a metabolic pathway involving the attachment of sugar chains to proteins, ranked at the top of enrichment analyses in the lean stratum specifically. In lean women, the pathway p-value was six times ten to the minus four. It ranked much lower in overweight and obese strata. This suggests that the biological machinery governing urate levels in lean individuals is partly distinct from what operates at higher BMI—a different set of molecular levers, not just the same levers turned up or down. Huffman and colleagues are measured about what their findings establish. The study is limited to European-ancestry populations. While the sample is the largest ever assembled for this question, it is still modest by the standards of interaction detection. Signals at RBFOX3 and ERO1LB-EDARADD are suggestive at best, and some candidates did not replicate. They present these results as hypotheses warranting further investigation, not settled conclusions. The larger point holds even with those caveats in place. The genetic architecture of serum urate shifts with body weight. Some variants matter more in lean individuals. Some matter less or differently in people with obesity. The genome is not immune to its environment. As the authors frame it, studying gene-by-environment interactions is a way to watch biology change in real time as the world around us changes. In an era when obesity rates are rising globally, effect sizes estimated in leaner historical samples may be quietly miscalibrated. The genes have not changed; the environment they are operating in has. This lecture was created by ennepō. Go to https://ennepo.ai to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.