Introduction: Globally, 462 million (1 in 16) people have type 2 diabetes (T2D), and the prevalence is higher in Samoa, affecting 1 in 4 Samoans. Polynesians are underrepresented in biomedical research, yet they may have population-specific variants important for understanding T2D. To better understand the genetics of metabolic traits relevant to T2D in Samoans, we performed meta-analysis GWASs of fasting glucose, fasting insulin, and homeostatic model assessment for insulin resistance (HOMA-IR).
Methods: We recruited 3 cohorts from Samoa and/or American Samoa in 1990, 2002, and 2010, totaling 5,174 participants. We genotyped the participants with the Illumina Global Screening Array or Affymetrix 6.0 array and performed genome-wide imputation using a Samoan-specific haplotype reference panel. Fasting glucose and fasting insulin levels were determined, and HOMA-IR was calculated from fasting glucose and insulin values. Participants who reported using diabetes medications were excluded. We performed genome-wide association testing on the remaining 4,333 participants using linear mixed models adjusted for the fixed effects of age, sex, and ancestral principal components derived from PC-AiR, and for random effects of kinship derived from PC-Relate. The results from each of the 3 cohorts were combined with p-value-based meta-analysis with METAL.
Results: No SNVs were statistically significant (p ˂ 5×10-8) in the 3 GWASs. The most significant variant associated with fasting glucose was rs373863828 (p = 8.2×10-8, MAF = 27%), a missense variant in CREBRF, which is known to be associated with higher BMI and lower odds of T2D in Samoans. Following the CREBRF finding, fasting glucose had a suggestive association with an intronic variant in LINC00877 (rs902318467, p = 1.9×10-7, MAF = 2.4%), a lncRNA gene which contains many enhancer-like signals and is located < 500kb upstream of PROK2 and FOXP1, two genes involved in glucose homeostasis. The first two introns of PARD3B contained 86 variants that were statistically suggestive for an association with fasting insulin. Lastly, the peak variant associated with HOMA-IR was in AKR1C2, a gene involved in aldehyde and ketone metabolic pathways.
Conclusion: We identified several potential genetic determinants of metabolic traits in Samoans, a population at high risk for T2D, through GWASs of fasting glucose, fasting insulin, and HOMA-IR. While the CREBRF finding is concordant with prior research, the others require further investigation. The findings from this study may shed light on the etiology of metabolic disorders and highlight the importance of including underrepresented groups in biomedical research.

Introduction
It is critical to understand the molecular basis for well-recognized disparities in type 2 diabetes (T2D) risk. We aimed to examine plasma proteomic profile variations among self-identified racial and ethnic groups and assess the mediating impact of race and ethnic-specific proteomic signatures on T2D.
Methods
We used proteomic data from 2339 individuals who self-identified as Non-Hispanic-White, Hispanic, or Non-Hispanic Black in the Women’s Health Initiative (N=1,400) and Multi-Ethnic Study of Atherosclerosis (N=939). To identify race- and ethnicity-specific associations we used linear models adjusting by confounders; site of recruitment, age, BMI, and eGFR. We employed structural equation modeling to examine the mediation of race and ethnic-specific proteins in the relationship between self-reported race/ethnicity and prevalent T2D. Models were stratified by sex; results were meta-analyzed and corrected for multiple comparisons (P<10-5).
Results
We tested associations among 324 proteins. A total of 138 proteins (41.04%) were associated with self-reported race/ethnicity (P<10-5). Seven proteins independently mediated the association between self-reported race/ethnicity and T2D (P<10-6). Top significant proteins include Growth/differentiation factor 8, Insulin-like growth factor-binding protein 2, and Apolipoprotein M. We identified that 8-11% of the association between self-reported race and prevalent T2D can be attributed to the direct mediating effect of race and ethnic-specific proteomic signatures.
Conclusion
More than 40% of protein levels tested varied significantly by race and ethnicity. Seven protein levels were found to significantly mediate the association between race and ethnicity and T2D prevalence. These results provide insights into the molecular pathways that contribute to racial and ethnic disparities in T2D prevalence.

Type 2 diabetes (T2D) is a prominent risk factor for prevalent cardiovascular diseases (CVD). Specifically, people with T2D are twice as likely to have heart disease or all-type stroke compared to people without T2D, though the reason behind this excess risk remains unknown. To better understand the molecular mechanisms underlying CVD in T2D, we studied and compared the metabolomics profiles of T2D participants with or without CVD event. Using TOPMed LC/MS metabolite data, we collected a total of 410 blood metabolite level measurements common to 1010 T2D participants across three TOPMed cohorts: the Framingham Heart Study (FHS), Multi-Ethnic Study of Atherosclerosis (MESA), and Women’s Health Initiative (WHI). We performed an association analysis for each metabolite using a mixed effects logistic regression model, with log-transformed and standardized metabolite level predicting a CVD event, adjusting for age at the last exam, sex, study population, and ancestry. We used a kinship matrix to account for relatedness, and a Bonferroni corrected alpha level of 1.2x10^-4 to detect significant associations. Subsequently, we conducted a two-sample mendelian randomization (MR) to examine causal relationships between each significant metabolite and CVD in T2D.

As a result, we identified 20 significantly associated metabolites with CVD in T2D. Most of these metabolites were lipids, belonging to the subclasses phospholipids, triglycerides, diglycerides, and cholesterol esters. Three of those metabolites, all cholesterol esters, were nominally significant from the two-sample MR (P<0.05), but not after adjusting for multiple testing. Overall, our results showed association of cholesterol esters with CVD in T2D participants. Larger sample sizes will help to more strongly establish metabolite associations with CVD in T2D in TOPMed.