Interaction of the TRIM46 / MUC1 locus with cigarette smoking may inuence the risk of gout.

Objectives: Some studies suggest that current-smoking may be protective against gout and smoking cessation associated with higher incidence of gout. Our study assessed potential interactions between smoking, genetic variants and gout prevalence. Methods: Four loci (ABCG2, GCKR, TRIM46, HNF4G) with evidence of smoking-inuenced associations with serum urate were tested for non-additive interaction with current-smoker or ex-smoker status that associates with gout in Aotearoa New Zealand (NZ) East and West Polynesian participants with (n=520) and without (n=629) gout. Results: Ex-smoker status was associated with higher prevalence of gout in people with East Polynesian but not West Polynesian ancestry. No association was detected between current smoking and gout. An interaction between TRIM46 (rs11264341) and ex-smoker status that associates with gout was observed in meta-analysis of NZ East and West Polynesians [OR Interaction = 0.58 (0.37-0.92)]. Never-smokers who were homozygous for the rs11264341 C-allele had higher odds of gout [OR= 2.43 (1.27; 4.64)], but not never-smoker heterozygotes [1.20 (0.73; 1.97]. The C allele was not associated with gout in ex-smokers [0.73 (0.36-1.47)]. No interactions involving current-smoker status were detected. Conclusions: We provide evidence for a non-additive interaction between TRIM46 (rs11264341) and ex-smoker status that associated with gout prevalence. MUC1, which encodes a transmembrane mucin in the lungs affected by cigarette smoke, is a possible candidate gene at this locus. No interaction involving current-smoker status was observed raising uncertainties about the relevance of an interaction specic to ex-smokers.

Conclusions: We provide evidence for a non-additive interaction between TRIM46 (rs11264341) and exsmoker status that associated with gout prevalence. MUC1, which encodes a transmembrane mucin in the lungs affected by cigarette smoke, is a possible candidate gene at this locus. No interaction involving current-smoker status was observed raising uncertainties about the relevance of an interaction speci c to ex-smokers.

Background
Gout arises from an interplay between genetic variants and the environment (1). Genome-wide association studies have identi ed dozens of genomic loci associated with serum urate concentrations (2, 3) and gout (1).
Multiple studies have found associations of current-smoking with lower gout incidence/prevalence (5-7)).
However, the evidence is equivocal, with others reporting associations of current-smoking with higher gout prevalence (8,9). There is also evidence of higher gout incidence in ex-compared to never-and current-smokers (6), and of serum urate increasing following smoking cessation (10).

Study participants
Cross-sectional data was from NZ participants aged 18 to 88, recruited 2010-2018. Participants were categorised into East Polynesian (Cook Island and NZ Māori; 381 non-gout and 352 gout) and West Polynesian (Samoa, Tonga, Niue, Tuvalu and Tokelau; 248 non-gout and 168 gout) groups based on selfreported ancestry of their grandparents. Ancestry was con rmed using genotypic principal component (PC) vectors calculated from 2858 ancestry-informative markers that cluster these genetically distinct ancestral groups (13). The study included a group of Māori (44 gout and 33 non-gout) from the rohe (area) of Ngati Porou iwi (tribe) from the Tairawhiti region, recruited in collaboration with Ngati Porou Hauora (health service). Gout was de ned by American Rheumatism Association classi cation criteria (14). All controls self-reported no previous gout diagnosis. Ethical

Data collection
Participants were asked if they were a current-smoker (Yes/No) or an ex-smoker. Participants who reported any or no alcohol consumption in the past week were categorised as drinkers or non-drinkers, respectively. Participants missing data for smoking, BMI and alcohol were excluded.

Statistical analysis
All analyses were performed using R statistical software version 4.0.2. (R Core Team 2018) (15). Allele frequencies and results of the Hardy-Weinberg Equilibrium exact test were calculated using SNPassoc (version 1.9.2) (Supplementary File 1: Table S1, S2) (16). Two binary smoking variables were analysed in separate regressions -current-smoker compared to never-smoker and ex-smoker compared to neversmoker. Multivariable logistic regression was used for association analyses of smoking variables and gout, and to estimate interaction between individual SNP genotypes and smoking in gout prevalence, using a SNP x smoker-variable interaction term. Interaction analysis of SLC2A9 (rs7442295) was not performed because the minor allele frequency was too low (<0.03) (Table S1). TRIM46 (rs11264341) was further analysed by strati cation of participants according to smoking status (never-smoker or exsmoker) and minor allele genotype. Multivariable logistic regression was performed with the genotypesmoker strati ed independent variable and gout as the dependent variable. All regression models were adjusted for age (continuous), sex (dichotomous), BMI (continuous) and alcohol intake (dichotomous), and the rst 10 genotypic PC vectors, to adjust for genetic admixture and population substructure.
Adjusted odds ratios (ORs) for gout and their 95% con dence intervals (CIs) were calculated. Metaanalysis was performed using meta package (version 4.13-0 ) (17). Heterogeneity was calculated using the Q (chi squared, χ 2 ) test. When P Heterogeneity <0.10 the random-effect model is shown instead of the xed-effect. Individual SNP vs gout association analyses are shown in Supplementary File 1: Table S3.

Results
Association analyses of smoking categories with gout.

Smoking and SNP interaction
There was interaction between ex-smoker status and TRIM46 (rs11264341) associated with gout prevalence in the combined East and West Polynesian cohort [OR Meta-Interaction (95% CI): 0.58 (0.37; 0.92), p=0.021,  Figure S2). No interaction was observed for rs11264341 with current-smoker status [OR Meta : 0.77 (0.43; 1.40), p=0.40, Table 3]. No interaction was observed at any of the other loci tested.
We strati ed participants according to TRIM46 (rs11264341) genotype and smoking status, and looked at the relationship of each genotype-smoker combination with gout (Supplementary File 3: Figure S3  gout, compared to the T/T reference group ( Figure S4).

Discussion
We identi ed a GxE interaction of the TRIM46 (rs11264341) locus with ex-smoker status associated with gout prevalence in a NZ population of East and West Polynesian ancestry, independent of age, alcohol consumption, BMI and sex.
In main effect analysis ex-smoker status was independently associated with higher prevalence of gout in East Polynesians (6,7). No association was observed between current-smoking and gout prevalence in East or West Polynesian cohorts. It is unclear why ex-smoker status is strongly associated with gout in East Polynesians and interacts with TRIM46 (rs11264341) to associate with gout in the combined Polynesian cohort, while current-smoker status does not. The smaller number of current-smokers compared to ex-smokers in the study cohort may in part explain why an association was not detected.
Previous associations have been found for current-smoking with lower serum urate and gout prevalence (5)(6)(7). It is possible that greater genetic predisposition to gout in people with Polynesian ancestry negates any protective effect of current-smoking. Higher prevalence of gout in ex-smokers may be due to an increase in serum urate following smoking cessation (10). It is possible that ex-smokers gained weight following smoking cessation and/or are more likely to be older, however our results were adjusted for age and BMI. It is also possible that higher prevalence in ex-smokers re ects gout onset that occurred before smoking cessation.
When considering the potential nature of the TRIM46 (rs11264341) interaction, never-smokers homozygous for the C allele had higher prevalence of gout than never-smokers without the C allele. Exsmokers had higher prevalence of gout compared to T/T never-smokers irrespective of genotype and there was no association of rs11264341 genotype with gout within the ex-smoker subgroup. These results could suggest that rs11264341 genotype in uences gout risk in never-smokers, whereas the risk effect of past smoking overrides the effect of genotype.
Causal candidate genes were identi ed at the TRIM46 (rs11264341) locus (18). Current knowledge would suggest MUC1 as the most likely candidate to interact with smoking. MUC1 encodes a transmembrane mucin, expressed on lung epithelium, whose expression, localisation and function is affected by cigarette smoke (19).
Limitations of this study include the relatively small number of participants. We did not know when smoking was commenced or ceased in relation to gout onset. Furthermore, information on smoking duration, intensity or a biochemical measure of smoking were not available for the cohort.

Conclusions
In this study of NZ Polynesian people we identi ed an interaction of the TRIM46 (rs1126434) locus and past smoking that is associated with gout prevalence, supporting our hypothesis that GxE interactions involving smoking associate with gout.