Abstract
Background and Objectives
Type 2 diabetes affects an increasing number of people worldwide. Although genome-wide association studies (GWAS) of type 2 diabetes have identified hundreds of loci, their interactions with other risk factors aren’t well understood. We investigated genetic interactions with three sex hormones (total testosterone, bioavailable testosterone, and sex hormone binding globulin (SHBG)) to identify additional type 2 diabetes-related loci that were undetected in traditional GWAS.
Methods
The study population consisted of white European UK Biobank participants. Individuals with type 1 diabetes were excluded. We examined sex-stratified interactions of polygenic risk score (PRS) for type 2 diabetes with sex hormone levels. We analyzed sex-stratified, genome-wide SNP × sex hormone interactions, adjusting for age and the top ten principal ancestry components.
Results
We found significant (P < 0.05) interactions for each of the sex hormones with PRS in both men and women, with the most significant being between SHBG and PRS in women (OR 0.88; 95% CI: 0.85-0.90; P = 1.09E-18). We identified 3 SNP × sex hormone interactions in men and 14 in women that achieved genome-wide significance (GWS; P < 5 × 10-8). Applying a 2-degree of freedom test, we identified GWS loci (10 in men and 23 in women) that were not GWS when testing marginal genetic effects alone.
Conclusion
Including interaction terms in GWAS may identify additional risk loci and improve the understanding of genetic architecture for type 2 diabetes. Different genetic interactions with sex hormones in men and women emphasize the importance of sex-stratified analysis in sex differential diseases.
Author summary
Type 2 diabetes is an increasingly common disease that is caused by both genetic and environmental factors. Though genome-wide association studies (GWAS) have been performed to identify several genes associated with type 2 diabetes, little is known about how an individual’s internal or external environment interacts with their genetics to affect disease outcomes. Here, we performed sex-stratified genome-wide interaction studies with 3 different sex-hormone interaction terms: testosterone, bio-available testosterone, and sex-hormone binding globulin. We identified 22 genomic risk loci for type 2 diabetes that were not reported in the most recent type 2 diabetes GWAS and were not significant using traditional GWAS. By including interaction terms in GWAS, we may be able to identify additional disease-related genes and gain a better understanding of the mechanisms behind complex diseases.
Introduction
Type 2 diabetes is a major cause of blindness, kidney failure, heart attack, stroke, and lower limb amputation. It affects over 11% of the US population, and it is estimated that roughly 23% of individuals with diabetes (3.4% of all U.S. adults) are unaware that they have it [1,2]. According to the Centers for Disease Control National Diabetes Statistics Report, type 2 diabetes was the 8th leading cause of death in the U.S. in 2021 with 31.1 deaths per 100,000 people [1]. Its global prevalence increased from 108 million in 1980–663 million in 2019 [2,3]. Risk factors associated with type 2 diabetes include tobacco smoking, high body mass index (BMI), low physical activity, high blood pressure, and greater age [1,4]. Epidemiological studies have demonstrated sex differences in type 2 diabetes in youth (<18 years old), with two-thirds of children and adolescents diagnosed with type 2 diabetes being girls [5]. However, among adults, prevalence of type 2 diabetes is higher in men [5]. In 2021, the International Diabetes Federation estimated that global diabetes prevalence in men was 10.8%, versus 10.2% in women [6]. In the U.S., the estimated diabetes prevalence from 2017-2020 was 14.2% in men and 12.4% in women [1].
Although both sexes produce the same sex hormones, their levels and bioavailability are vastly different. In men, the testes produce about 7,000 μg of testosterone per day and convert 0.25% into estradiol (E2), whereas in women, the ovaries produce about 300 μg of testosterone per day and convert 50% of it into E2 [7]. Testosterone also induces estrogen sulfotransferase (EST), which inactivates E2 [7]. The amounts of biologically available testosterone and E2 (i.e., testosterone and E2 available to target tissues) are modified by levels of albumin and sex-hormone binding globulin (SHBG). Albumin binds the sex hormones similarly; however, the affinity of SHBG for testosterone that is twice that for E2, and changes in SHBG levels greatly affect the amount of bioavailable testosterone (BAT).
Previous studies have observed that lower levels of testosterone confer greater risk of coronary events and type 2 diabetes in men but lower risk in women, while higher SHBG concentrations protect against coronary disease and type 2 diabetes in both sexes (although to a greater extent in women) [5,8]. A longitudinal study [9] including testosterone measurements for more than 70,000 men and 80,000 women found that men with serum testosterone <7 nmol/L had an adjusted type 2 diabetes incidence rate ratio (aIRR) of 2.71 when compared to men with testosterone ≥ 20 nmol/L; women with serum testosterone ≥ 3.5 nmol/L had an aIRR of 1.98 when compared to women with testosterone < 1.0 nmol/L. The same study reported that the risk of type 2 diabetes increased as the concentration of serum SHBG decreased to <40 in men and <50 nmol/L in women [9]. In a recent Mendelian randomization study, one standard deviation increase in genetically predicted testosterone levels was associated with higher risk of developing type 2 diabetes in women (odds ratio (OR) = 1.37) but a lower type 2 diabetes risk in men (OR = 0.86) [8]. Based on the evidence that testosterone levels affect the risk of type 2 diabetes differently in men and women, the Endocrine Society has recommended measuring free and total testosterone levels in all patients with type 2 diabetes [10].
Genetic factors also contribute to type 2 diabetes risk. The estimated heritability of type 2 diabetes ranges from 30 to 70% [11]. Genome-wide association studies (GWAS) have identified over 700 genetic loci associated with type 2 diabetes [12]. Genetic risk scores and polygenic risk scores (PRS) have been constructed to combine individual genetic effects; however, these genetic scores do not significantly enhance risk prediction over standard clinical risk factors [12,13]. Currently known risk loci explain only 10–20% of type 2 diabetes heritability, and the remaining 80–90% of heritability is likely due to variants that have yet to be detected due to their small effects and rare variants with low minor allele frequencies (MAF) [13]. Gene-gene and gene-environment/risk factor interactions can also contribute to this “missing heritability” [14].
Gene-environment interaction (G × E) analyses of type 2 diabetes-related outcomes have examined interactions with smoking, physical activity, and diet [15–19]. In 2020, Wu et al. [20] identified 5 loci associated with genotype-smoking interactions in type 2 diabetes risk. Kilpeläinen et al. reported 8 loci associated with gene-physical activity interactions and type 2 diabetes-related traits, including the conversion of impaired glucose intolerance to type 2 diabetes and blood lipid profile [17,18]. Gene-diet interaction studies have reported that 13–27% of phenotypic variance in fasting insulin and insulin resistance can be explained by gene-diet interactions [21,22]. Despite the well-established associations between sex hormones and type 2 diabetes [13–19], no published studies have investigated the interactions of genetic factors with sex hormone levels in relation to type 2 diabetes risk.
Here, we analyze data from the UK Biobank (UKB) cohort to assess the effects of gene-sex hormone interactions on type 2 diabetes and to identify genetic variants that would not be found in type 2 diabetes GWAS examining only main genetic effects. A graphical representation of the study design and sampling can be found in Fig 1. We focused on measured total testosterone (TT) and SHBG and calculated BAT, which were available for most UKB participants.
Fig 1. Graphical abstract.
Results
The study cohort
There were 484,127 UKB participants with basic phenotypic and genetic information available. Of these, 362,687 had TT measurements, 363,945 had SHBG measurements, and 294,727 had TT, SHBG, and albumin measurements allowing for the calculation of BAT. Characteristics of the TT sub-cohort, which is representative of all 3 sub-cohorts with sex hormone measures, are summarized in Table 1. The numbers of participants included in the genome-wide Gene Environment in Millions (GEM) analyses varied slightly, depending on the available data: 182,645 men and 180,773 women for TT, 169,872 men and 194,833 women for SHBG, and 150,798 men and 144,517 women for BAT.
Table 1. Characteristics of participants included in analysis of total testosterone x polygenic risk score interaction*.
| WOMEN | MEN | |||||
|---|---|---|---|---|---|---|
| Controls | Cases | All | Controls | Cases | All | |
| N | 169388 | 11068 | 180456 | 162505 | 19726 | 182231 |
| Age (years) | 55.98 (7.96) |
59.02 (7.4) |
56.17 (7.96) |
56.64 (8.12) |
60.14 (6.82) |
57.02 (8.06) |
| BMI (kg/m2) |
26.78 (4.89) |
32.23 (6.54) |
27.11 (5.17) |
27.44 (3.91) |
31.21 (5.14) |
27.84 (4.23) |
| Cholesterol (nmol/L) |
5.94 (1.1) |
5.34 (1.27) |
5.9 (1.12) |
5.59 (1.08) |
4.77 (1.18) |
5.5 (1.12) |
| PRST2D | -0.19 (0.94) |
0.46 (0.94) |
-0.15 (0.96) |
-0.21 (0.93) |
0.43 (0.92) |
-0.14 (0.95) |
| TT (nmol/L) |
1.12 (0.63) |
1.15 (0.68) |
1.12 (0.63) |
12.17 (3.68) |
10.4 (3.57) |
11.98 (3.71) |
| SHBG (nmol/L) |
63.16 (30.09) |
44.24 (26.86) |
61.99 (30.25) |
40.36 (16.63) |
35.62 (16.74) |
39.85 (16.71) |
| BAT (nmol/L) |
0.35 (0.25) |
0.46 (0.31) |
0.36 (0.26) |
5.27 (1.55) |
4.76 (1.56) |
5.21 (1.56) |
| CAD | 9206 (5.43) |
2404 (21.72) |
11610 (6.43) |
21663 (13.33) |
6973 (35.35) |
28636 (15.71) |
| Current smoker | 15288 (9.03) |
1336 (12.07) |
16624 (9.21) |
19361 (11.91) |
2692 (13.65) |
22053 (12.10) |
| Former smoker | 53332 (31.49) |
3918 (35.40) |
57250 (31.73) |
61619 (37.92) |
9880 (50.09) |
71499 (39.24) |
| Non-smoker | 100205 (59.16) |
5763 (52.07) |
105968 (58.72) |
81025 (49.86) |
7024 (35.61) |
88049 (48.32) |
| Post-Menopause† | 102215 (60.34) |
7354 (66.44) |
109569 (60.72) |
NA | NA | NA |
*Analysis based on logistic regression model incorporating total testosterone x polygenic risk score as an interaction term, along with age and the first 10 principal components as covariates. The participants included in this model are representative of the participants included in all analyses reported in this study.
† Menopause status was available for only 153,947 women in this cohort.
Data are presented as count (percentage) or mean (standard deviation) as appropriate. Cases refer to participants with type 2 diabetes; controls refer to participants with no diabetes. N- number of participants, BMI – body mass index, PRST2D – polygenic risk score for type 2 diabetes, TT – total testosterone, SHBG – sex hormone binding globulin, BAT – bioavailable testosterone, CAD – coronary artery disease.
PRST2D × sex hormone interactions in type 2 diabetes
Interactions between PRST2D and the sex hormones (TT, SHBG, and BAT) were examined separately in men and women (Table 2). The PRST2D was significantly associated with higher type 2 diabetes risk in all 6 models (ORs 1.85-2.34 per SD of PRST2D). Higher TT and BAT levels were significantly associated with increased odds of type 2 diabetes in women (TT OR 1.08, C.I. 1.06-1.10; BAT OR 1.30, C.I. 1.27-1.32) and decreased odds in men (TT OR 0.61, C.I. 0.60-0.62; BAT OR 0.76, C.I. 0.75-0.78). Higher SHBG level was protective in both women (OR 0.45 C.I. 0.44-0.46) and men (OR 0.67, C.I. 0.66-0.69). The PRST2D × sex hormone interaction term was nominally significant (p < 0.05) in all 6 models, with the most significant interaction being between PRST2D and SHBG in women (OR 0.88, 95% C.I. 0.85-0.90, p < 2 × 10-16). A closer examination of the PRST2D × SHBG interaction in women showed that having both high SHBG and high PRST2D risk confers higher odds of developing type 2 diabetes than would be expected from additive or multiplicative effects and that an interaction may indeed be present (S1 Text, S2 Table).
Table 2. Logistic regression model results for sex hormone x PRST2D interaction analysis, stratified by sex.
| Sex Hormone | Predictor | Women | Men | ||||||
|---|---|---|---|---|---|---|---|---|---|
| OR (95% C.I.) | P-value | N Cases |
N Controls | OR (95% C.I.) | P-value | N Cases | N Controls | ||
| TT | PRST2D | 2.12 (2.05-2.19) | <1E-320 | 11,068 | 169,388 | 1.90 (1.80-2.01) | 1.84E-109 | 19,726 | 162,505 |
| TT | 1.08 (1.06-1.10) | 3.52E-19 | 0.61 (0.60-0.62) | <1E-320 | |||||
| PRST2D x TT | 0.98 (0.97-1.00) | 0.01 | 1.02 (1.01-1.04) | 0.01 | |||||
| SHBG | PRST2D | 2.34 (2.23-2.46) | 6.01E-265 | 12,502 | 181,955 | 2.15 (2.06-2.25) | 1.21E-233 | 19,018 | 150,470 |
| SHBG | 0.45 (0.44-0.46) | <1E-320 | 0.67 (0.66-0.69) | <1E-320 | |||||
| PRST2D x SHBG | 0.88 (0.85-0.90) | 1.09E-18 | 0.97 (0.95-0.99) | 2.17E-03 | |||||
| BAT | PRST2D | 1.91 (1.84-1.98) | 1.04E-262 | 8,954 | 135,312 | 1.85 (1.73-1.97) | 4.81E-79 | 16,681 | 133,600 |
| BAT | 1.30 (1.27-1.32) | 2.19E-174 | 0.76 (0.75-0.78) | 1.09E-133 | |||||
| PRST2D x BAT | 1.04 (1.02-1.06) | 3.17E-06 | 1.04 (1.02-1.06) | 4.11E-04 | |||||
TT- total testosterone, SHBG - sex hormone binding globulin, BAT- bioavailable testosterone, PRST2D – polygenic risk score for type 2 diabetes, T2D- type 2 diabetes, OR- odds ratio of T2D, C.I.- confidence interval. The interaction variable in the table is represented by PRST2D x hormone. The odds ratios shown are per increase in standard deviation. Age and principal components 1–10 were covariates in all models.
Genome-wide SNP × sex hormone interactions in type 2 diabetes
GEM analysis of genome-wide significant (GWS; p < 5 × 10-8, 1-df) interactions identified fourteen relevant loci in women (4 interacting with SHBG, 10 with BAT), and two in men (1 interacting with TT and 1 with SHBG) (Table 3). Among women, 3 loci interacting with SHBG and 5 with BAT were also GWS in the analysis of joint (2 df) effects; the same was true for the locus interacting with TT among men. None of the loci identified via their interaction effects had marginal effects that were GWS.
Table 3. Genomic risk loci associated with type 2 diabetes through their interactions with sex hormones.
| SNPID | CHR | Position (hg19) | Nearest Gene | EA | NEA | EAF | P value | Beta (SE) | |
|---|---|---|---|---|---|---|---|---|---|
| Men | |||||||||
| TT | rs118089799 | 9 | 80839160 | GNAQ | A | G | 9.1E-03 | 2.19E-09 | -0.11 (0.02) |
| SHBG | rs738408 | 22 | 44324730 | PNPLA3 | T | C | 0.23 | 4.09E-13 | 0.01 (0.001) |
| Women | |||||||||
| SHBG | rs72847791 | 2 | 129479940 | AC012451.1 | T | C | 1.2E-02 | 2.40E-09 | -0.02 (0.003) |
| SHBG | 10:134562791:T:TCACG* | 10 | 134562791 | INPP5A | T | TCACG* | 1.0E-02 | 3.46E-09 | -0.02 (0.003) |
| SHBG | rs9804606 | 11 | 129889590 | RP11-567M21.3 | T | C | 0.03 | 4.49E-08 | -0.01 (0.003) |
| SHBG | rs140069179 | 17 | 11651941 | DNAH9 | C | A | 0.05 | 3.83E-08 | -0.01 (0.002) |
| BAT | rs34789050 | 3 | 51790029 | IQCF6 | A | G | 0.05 | 1.02E-10 | 0.90 (0.14) |
| BAT | rs146037870 | 4 | 101861217 | EMCN | T | C | 0.03 | 4.15E-08 | 0.90 (0.16) |
| BAT | rs4476958 | 7 | 52954110 | RP11-398K14.1 | A | G | 0.95 | 9.57E-09 | -0.83 (0.15) |
| BAT | 8:17032214:CTT:C | 8 | 17032214 | ZDHHC2 | C | CTT | 0.93 | 3.46E-11 | -0.85 (0.13) |
| BAT | rs117878783 | 8 | 72988412 | TRPA1 | A | G | 0.03 | 1.12E-08 | 1.05 (0.19) |
| BAT | rs80123595 | 10 | 119556155 | RP11-355F22.1 | T | C | 0.03 | 1.50E-08 | 0.73 (0.13) |
| BAT | rs12268706 | 10 | 132418020 | Y_RNA | C | T | 0.06 | 9.93E-09 | 1.03 (0.18) |
| BAT | rs117530301 | 11 | 91819588 | RPL7AP57 | G | A | 0.03 | 1.34E-08 | 1.13 (0.20) |
| BAT | rs12825501 | 12 | 30134559 | LOC105369715 | C | T | 0.07 | 1.06E-08 | 0.67 (0.12) |
| BAT | rs17256233 | 14 | 61922461 | PRKCH | T | C | 0.10 | 2.32E-08 | 0.64 (0.11) |
CHR- chromosome, SE – Standard error, EA – effect allele, NEA – non-effect allele, EAF – effect allele frequency, SHBG – sex hormone binding globulin, BAT – bioavailable testosterone.
* TCACGGTTGCAGCCACGCTTTGTTGGGGAGCAGCCAGCCTCAGGCCCTTCGC
More genomic loci with significant marginal or joint effects with TT, SHBG, or BAT were found among men than among women (S3 Table). Most loci with joint effects were also found to have marginal effects; those identified only by joint effects are described further in Table 4. There was no overlap between GWS loci identified for men and women within any hormone interaction analysis. Manhattan and QQ plots for each analysis are included in S1 Fig.
Table 4. Genomic risk loci associated with type 2 diabetes through their joint effects with sex hormones that do not overlap genomic risk loci identified through marginal effects.
| SNPID | CHR | Position (hg19) | Nearest Gene | EA | NEA | EAF | P value Joint |
BETA (SE) Interaction |
P value Interaction |
BETA (SE) Genetic Main Effects |
P value Genetic Main Effects |
|
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Men | ||||||||||||
| TT | rs141469161 | 7 | 2410296 | EIF3B | C | T | 0.99 | 3.29E-08 | 0.04 (0.02) | 1.43E-02 | 0.29 (0.05) | 6.00E-09 |
| TT | rs118089799 | 9 | 80839160 | GNAQ | G | A | 0.99 | 1.60E-08 | 0.11 (0.02) | 2.19E-09 | 0.04 (0.06) | 0.476 |
| TT | rs145010109 | 15 | 60620292 | ANXA2 | G | A | 0.97 | 4.27E-08 | 0.05 (0.01) | 3.01E-06 | 0.16 (0.03) | 1.76E-06 |
| TT | rs738409 | 22 | 44324727 | PNPLA3 | G | C | 0.22 | 6.42E-09 | 0.02 (0.01) | 4.48E-07 | 0.08 (0.02) | 6.87E-08 |
| SHBG | rs62580110 | 9 | 133865370 | LAMC3 | T | A | 0.99 | 3.62E-10 | 0.02 (0.004) | 7.39E-08 | 0.21 (0.05) | 7.28E-05 |
| SHBG | rs78328312 | 14 | 61842791 | PRKCH | G | T | 0.99 | 1.19E-09 | 0.01 (0.003) | 4.46E-05 | 0.23 (0.04) | 1.29E-07 |
| SHBG | rs72720262 | 14 | 69682013 | EXD2 | G | A | 0.33 | 6.42E-09 | -0.003 (0.001) | 2.05E-03 | 0.05 (0.01) | 3.35E-05 |
| SHBG | rs3747207 | 22 | 44324855 | PNPLA3 | A | G | 0.21 | 3.86E-16 | 0.01 (0.001) | 4.47E-13 | 0.09 (0.01) | 1.63E-10 |
| BAT | rs142442003 | 3 | 452463 | CHL1 | A | G | 0.99 | 8.62E-09 | 0.15 (0.04) | 6.10E-05 | 0.26 (0.05) | 3.55E-07 |
| BAT | rs271130 | 6 | 133246241 | HMGB1P13 | A | G | 0.48 | 1.12E-08 | -0.03 (0.01) | 3.99E-03 | 0.05 (0.01) | 6.04E-05 |
| Women | ||||||||||||
| TT | rs1182395 | 7 | 157032924 | UBE3C | T | G | 0.65 | 4.98E-08 | -0.04 (0.018) | 3.82E-02 | 0.08 (0.01) | 6.29E-08 |
| SHBG | rs72847791 | 2 | 129479940 | ISCA1P6 | C | T | 0.99 | 2.81E-09 | 0.02 (0.003) | 2.40E-09 | 0.35 (0.07) | 1.27E-06 |
| SHBG | rs13266634 | 8 | 118184783 | SLC30A8 | C | T | 0.69 | 1.37E-08 | -0.003 (0.001) | 4.04E-03 | 0.03 (0.02) | 0.258 |
| SHBG | rs112045042 | 9 | 71551592 | PIP5K1B | C | T | 0.99 | 1.05E-08 | 0.01 (0.003) | 3.93E-07 | 0.34 (0.06) | 3.58E-08 |
| SHBG | 10:134562791_ TCACGGTTGCAGCCACGCTTTGTT GGGGAGCAGCCAGCCTCAGGCCCTTCGCA_T |
10 | 134562791 | INPP5A | TCACGGTTGCAGCCACGCTTTGTT GGGGAGCAGCCAGCCTCAGGCCCTTCGCA |
T | 0.99 | 7.77E-09 | 0.02 (0.003) | 3.46E-09 | 0.34 (0.08) | 6.11E-06 |
| SHBG | rs9804606 | 11 | 129889590 | PRDM10 | C | T | 0.98 | 2.38E-09 | 0.01 (0.003) | 4.49E-08 | 0.34 (0.06) | 2.99E-09 |
| SHBG | rs184922766 | 12 | 59633138 | CCND2 | G | C | 0.99 | 4.48E-09 | 0.02 (0.003) | 3.82E-07 | 0.42 (0.07) | 5.47E-09 |
| SHBG | rs71386896 | 16 | 85191033 | CTC-786C10.1 | G | A | 0.98 | 2.69E-08 | 0.01 (0.003) | 3.90E-07 | 0.31 (0.06) | 2.97E-08 |
| SHBG | rs117547596 | 22 | 34824038 | SYN3 | C | T | 0.99 | 4.35E-08 | -0.01 (0.004) | 0.031038 | 0.13 (0.09) | 0.143 |
| BAT | rs114701560 | 2 | 30587510 | Lnc-CAPN13–2 | A | G | 0.97 | 2.74E-08 | -0.73 (0.16) | 2.91E-06 | 0.15 (0.04) | 3.57E-04 |
| BAT | rs150147479 | 2 | 45057311 | SIX3 | A | G | 0.98 | 1.76E-09 | -0.68 (0.15) | 3.27E-07 | 0.14 (0.04) | 3.88E-04 |
| BAT | rs13020842 | 2 | 122477115 | NIFK | G | A | 0.96 | 4.38E-08 | -0.96 (0.19) | 7.94E-06 | 0.17 (0.05) | 1.23E-04 |
| BAT | rs34789050 | 3 | 51790029 | IQCF6 | G | A | 0.95 | 2.50E-10 | -0.9 (0.14) | 1.02E-10 | 0.05 (0.04) | 0.138 |
| BAT | rs149522 | 7 | 130449842 | LOC105375508 | T | C | 0.40 | 3.28E-08 | 0.33 (0.08) | 7.40E-05 | 0.06 (0.02) | 1.84E-04 |
| BAT | 8:17032214_CTT_C | 8 | 17032214 | ZDHHC2 | C | CTT | 0.93 | 2.32E-10 | -0.85 (0.13) | 3.46E-11 | 0.03 (0.03) | 0.360 |
| BAT | rs147952627 | 10 | 119543721 | LINC02674 | C | G | 0.97 | 3.56E-08 | -0.73 (0.13) | 2.00E-08 | 0.07 (0.03) | 0.077 |
| BAT | rs12268706 | 10 | 132418020 | LOC100419870 | C | T | 0.06 | 9.03E-09 | 1.03 (0.18) | 9.93E-09 | 0.09 (0.05) | 0.027 |
| BAT | rs61729748 | 11 | 5020933 | OR51L1 | A | T | 0.98 | 2.28E-08 | -0.87 (0.20) | 1.30E-05 | 0.18 (0.05) | 1.34E-04 |
| BAT | rs12825501 | 12 | 30134559 | LOC100422352 | T | C | 0.93 | 3.88E-08 | -0.67 (0.18) | 1.06E-08 | 0.04 (0.03) | 0.145 |
| BAT | rs118086999 | 15 | 38403041 | LOC100507568 | C | A | 0.97 | 4.99E-08 | -0.92 (0.18) | 1.39E-07 | 0.10 (0.04) | 0.021 |
| BAT | rs551795351 | 17 | 76977993 | CANT1 | A | G | 0.99 | 2.64E-08 | -1.57 (0.33) | 1.86E-06 | 0.20 (0.07) | 5.28E-03 |
| BAT | rs139416564 | 19 | 15172209 | CASP14 | T | C | 0.99 | 4.21E-09 | -1.83 (0.34) | 9.80E-08 | 0.17 (0.07) | 0.011 |
| BAT | rs2545962 | 19 | 20591603 | LOC100288623 | A | C | 0.99 | 4.22E-08 | -1.66 (0.32) | 3.34E-07 | 0.20 (0.07) | 7.82E-03 |
Chr – chromosome, EA – effect allele, NEA – non-effect allele, EAF – effect allele frequency, TT – total testosterone, SHBG – sex hormone binding globulin, BAT – bioavailable testosterone.
Sensitivity analyses
We performed two sensitivity studies in women to consider menopausal status. Firstly, menopausal status was added as a covariate in the GEM model, along with age at enrollment, PC1–10, and the G × SHBG or G × BAT interaction term (S1 Text). Participants without menopause information were excluded, reducing the sample size from 194,833–163,975 for SHBG and 144,517–122,581 for BAT. The interaction effects (beta coefficients) for the models with and without menopause as a covariate were highly correlated (r = 0.86 for SHBG and r = 0.93 for BAT) and the correlations between genome-wide significant SNPs was even higher (r = 0.98 for SHBG and r = 1.0 for BAT). These results indicated that the interactions we observed were not sensitive to a woman’s menopausal status (S2 Fig).
The second menopause sensitivity analysis was performed in only post-menopausal women. GEM models were built with age at enrollment, PC1–10, and the G × SHBG or G × BAT interaction terms as covariates (S1 Text). The sample sizes for the SHBG and BAT cohorts were 120,700, and 88,518, respectively. The interaction effects (beta coefficients) for the models in post-menopausal women vs all women were highly correlated (r = 0.76 for SHBG and r = 0.84 for BAT) and the correlations between genome-wide significant SNPs was even higher (r = 0.98 for SHBG and r = 1.0 for BAT). These results further indicated that the interactions we observed were not sensitive to a woman’s menopausal status (S3 Fig).
We performed a third sensitivity study to examine interaction effects in persons with incident type 2 diabetes. Incident cases were identified by comparing the date of initial assessment with the first date of type 2 diabetes diagnosis (S1 Text). Of the 50,647 type 2 diabetes cases in the UKB, 10,178 were incident. Beta coefficients for marginal and interaction effects in each G × hormone analysis for incident cases were compared to results from the corresponding analysis including both incident and prevalent type 2 diabetes. With the exception of SHBG interactions in men (r = 0.74), the correlations of genome-wide significant SNPs in all analyses were ≥ 0.94. Plots are shown in S4 Fig. The results of this analysis indicated that the interactions we observed were not sensitive to whether a person had incident or prevalent type 2 diabetes.
Replication in African and South Asian cohorts
We attempted a small replication study in South Asian and African ancestry cohorts (S1 Text). In total, there were 7,502 participants in the South Asian cohort, 2,145 of whom had type 2 diabetes and 7,588 participants in the African cohort, 1,572 of whom had type 2 diabetes (S5 Table). The beta coefficients and p values for genomic risk loci identified via 1df and 2df interaction tests in the European cohort were then compared to their counterparts in the South Asian and African sub-cohorts. Generally, the coefficients showed very little correlation between ancestry groups (S6 Table), likely due to inaccurate estimates in the much smaller sample sizes with South Asian or African ancestry.
Discussion
Observational studies of testosterone in relation to type 2 diabetes have found opposite effects in men and women: lower testosterone increases the risk of type 2 diabetes in men but decreases the risk in women. Biological processes underlying this phenomenon are not well understood. We undertook large-scale, sex-specific analyses of genetic interactions with measured testosterone and SHBG levels and calculated BAT in UK Biobank participants. Although interactions of all three biomarkers with a type 2 diabetes polygenic risk score were nominally statistically significant in both men and women, we found the best evidence for interaction with SBHG in women. We observed negative interaction between PRST2D and total testosterone (TT), which indicates an antagonistic interaction given the positive associations of both PRST2D and TT. Such statistical interaction suggests a potential shared etiology between PRST2D and TT which can mask out their effects on T2D. In a genome-wide analysis of SNP- sex hormone interactions, we identified 40 genomic risk loci for type 2 diabetes, of which only 18 overlap those reported in the most recent type 2 diabetes GWAS [23] (S4 Table). None of these SNPs had marginal effects that reached genome-wide significance.
Two protein coding genes, PNPLA3 and GNAQ, mapped to the newly identified loci in men. Several studies have identified PNPLA3 as being associated with testosterone and SHBG levels [8,24], and it has also been linked to type 2 diabetes [25]. The SNP rs738408 on PNPLA3 is a synonymous variant that has been associated with Non-Alcoholic Fatty Liver Disease [26]. GNAQ has been associated primarily with body mass index (BMI) [27], which is also associated with both sex hormones and type 2 diabetes. In women, seven protein coding genes mapped to GWS interaction loci: INPP5A, DNAH9, ZDHHC2, EMCN, PRKCH, TRPA1, and IQCF6. GWAS have identified four of these (INPP5A, ZDHHC2, IQCF6, and PRKCH) to be associated with BMI [27,28] or obesity [29]. INPP5A has also been associated with of cardiovascular disease [28]. EMCN has been associated with left ventricular ejection fraction [30] and glomerular filtration rate [31]. DNAH9 and TRP1 have not been previously associated with any diabetes or sex-hormone related traits. Four of the 16 loci identified through interaction effects in either men or women -rs9804606, rs34789050, rs117530301, and rs12825501- are expression quantitative trait loci (eQTLs) which affect gene expression levels (S7 Table). The eQTL, rs34789050, is associated with genes TEX264, GRM2, and MANF and alter gene expression in several sex-associated tissues including vagina, testis, ovary, and mammary tissue.
These findings provide deeper insight into the modulating effects of sex and sex hormones on type 2 diabetes and emphasize the power of including both sex stratification and interaction terms in GWAS. Despite the known sex disparity for testosterone’s association with type 2 diabetes, it has not been examined in the context of gene-environment interactions. Previous GWAS have identified several loci showing sex-specific associations with type 2 diabetes-related traits. One 2012 study conducted by Morris et al. [32] found that sex stratification identified two additional loci associated with type 2 diabetes--CCND2 in men and GIPR in women--that did not reach statistical significance when the sexes were combined. A 2014 study by Hara et al. [33] found evidence for 23 genes and regions displaying sexual dimorphism in associations with type 2 diabetes. In 2021, a study by Lagou et al. [34] reported 2 loci (IRS1 and ZNF12) that display sexual dimorphism in effects on fasting insulin, a marker for diabetes.
Our analysis of genetic interactions with sex hormones in type 2 diabetes was limited in several important respects. First, hormone measurements for each participant were made at a single visit, without accounting for longitudinal variations. We were able to study only testosterone, sex hormone-binding globulin, and bioavailable testosterone because estradiol measurements were missing for most UK Biobank participants. G × Sex hormone analysis of sex chromosomal variants was not included, limited by data access and analysis issue. Finally, we were unable to access suitable data for replication analysis. We attempted to replicate our findings in Black and South Asian cohorts but were unsuccessful because of their small sample sizes. Future studies of sex hormones in type 2 diabetes would benefit from additional measurements in more diverse cohorts and could consider age stratification because hormone levels, particularly in women, vary greatly throughout the lifetime.
Methods
Ethics statement
The UKB received ethical approval from the National Information Governance Board for Health and Social Care and the National Health Service North West Centre for Research Ethics Committee (REC reference number 21/NW/0157) [23].
Study cohort and phenotypes
The UKB includes about 500,000 individuals in the United Kingdom (UK) for whom detailed phenotypic and genotypic data were collected [35]. All participants provided written informed consent for their data to be used for health-related research at the time of recruitment.
Type 2 diabetes cases were identified using UKB hospital inpatient admission data and the self-reported data field “diabetes diagnosed by a doctor”. Hospital inpatient admission data were queried for type 2 diabetes diagnoses using International Classification of Diseases, 10th edition (ICD 10) codes E11-E14 and ICD-9 code 250. Participants with ICD-10 code E10 and ICD-9 codes 25001 and 25011 (type 1 diabetes) were excluded, as were any participants with self-reported type 1 diabetes. Participants with either prevalent or incident type 2 diabetes were considered cases for this analysis.
Plasma levels of total testosterone (TT), SHBG, and albumin measurements were attempted in all UKB participants at enrollment though some measurements fell outside the range of detection for their assay or did not pass quality control [24]. These individuals (69,024 missing TT; 69,544 missing SHBG; 65,150 missing albumin measurements) were excluded from the association analysis. Bio-available testosterone was calculated according to the Vermeulen equation [35–37] (S1 Text). We were unable to examine estradiol in this study because its measurement was missing for 92% of men and 79% of women.
Genetic data
Genotyping was done using the Applied Biosystems UKB Axiom Array followed by data processing and quality control procedures considering batch, plate, and array effects, missing rate, Hardy-Weinberg Equilibrium, and sex mismatch, which are described in detail in the UKB genotyping documentation [38]. The dosages of genomic variants were then imputed by the UKB using the Haplotype Reference Consortium and UK10K + 1000 Genomes reference panels [38]. Only unrelated white Europeans with kinship < 0.0884 were included in our analyses. Ethnicity was self-reported during the patients’ initial Assessment Centre visit, and genetic data were used to estimate kinship coefficients, as described in the UKB genotyping documentation [38]. The first ten principal components (PC1–10) of ancestry were computed by the UKB, first on a subset of high-quality unrelated samples before the loadings were projected onto all samples [38], to account for population structure. Standard PRS for type 2 diabetes (PRST2D) for the UKB participants (field ID 26285) were generated by the UKB from three GWAS datasets external to the UKB using genetic variants with INFO scores of imputation quality > 0.7 [39]. Briefly, PRS algorithms were built from trait-specific meta-analyses using a Bayesian approach, and per-individual PRS values were calculated as the genome-wide sum of the per-variant posterior effect size multiplied by allele dosage [39].
G × Sex hormone interaction analysis
We used logistic regression to model the associations of PRST2D and each sex hormone level with type 2 diabetes status, stratified by sex. Age at enrollment and PC1–10 were included as covariates. We used this model to assess the marginal effects of PRST2D and sex hormones, as well as PRST2D × sex-hormone interaction terms. The GLM function of R version 4.0.3 was used to calculate raw and standardized estimates, which were then converted to odds ratios. In the standardized models, both the PRST2D and sex hormone variables were treated as continuous, and the odds ratios represent the change in type 2 diabetes risk per standard deviation increase in each respective variable.
The Gene-Environment interaction analysis for Millions of samples (GEM) approach enables GWAS that incorporates multiple exposures, controls for genotype-covariate interactions, and supports robust inference [40]. GEM considers a generalized linear model with interaction terms. In the present study, we implemented the following model: T2D ~ β0 + βCC + βGG + βEE + βGE(G × E) where T2D is the type 2 diabetes status, β0 is the intercept for the model; βC and C represent the coefficients and values, respectively, for the covariates (age and PC1–10); βG and G represent the coefficients and values for the genotype; βE and E represent the coefficients and values for the exposure (sex hormone); and βGE represents the coefficient for the interaction term G × E. Based on this equation, GEM calculates coefficient estimates and standard errors for interaction effects (G × E; 1 degree of freedom [1-df]) and genetic main effects (G), and also conducts joint tests (2 degrees of freedom [2-df]) of the genetic main and interaction effects (G + G × E) and reports the p values. Within the same analysis, GEM also considers a standard GWAS model with no interaction terms (T2D ~ β0 + βCC + βGG + βEE) and reports coefficient estimates, standard errors, and p values for the marginal genetic effects (G) of this model. It should be noted that within this manuscript, “marginal effects” will refer to genetic effects from a model with no interaction terms and “genetic main effects” will refer to genetic effects from a model with interaction terms. GEM version 1.4.2 was used to conduct sex-stratified gene-hormone interaction analysis with a type 2 diabetes outcome. Only common variants with MAF > 0.01 were included in the present study.
Bioinformatic analysis
Functional Mapping and Annotation of Genome-Wide Association Studies (FUMA) is an online platform that can be used to annotate, prioritize, visualize, and interpret GWAS results [41]. The SNP2GENE function of FUMA version 1.5.1 was used under default settings to annotate the genomic loci identified from the robust marginal, joint, and interaction effects summary statistics output from GEM. For each analysis, FUMA identified genomic risk loci, lead SNPs, independently significant SNPs, and candidate GWAS-tagged SNPs. The definitions of these terms can be found in the FUMA documentation and are included in S1 Table [41]. Using the FUMA output, non-overlapping loci between the marginal (i.e., standard GWAS) and joint 2-df tests were identified based on chromosomal locations using the GenomicRanges package from Bioconductor [42].
Supporting information
Examination of the SHBG × PRST2D interaction in women. Menopause sensitivity analysis. Incident type 2 diabetes sensitivity analysis. Replication study in South Asian and African ancestry cohorts.
(DOCX)
(DOCX)
(DOCX)
(DOCX)
(DOCX)
S1.2 A-C Fig. Manhattan and QQ Plots for the G x Total Testosterone analysis in women. S1.3 A-C Fig. Manhattan and QQ Plots for the G x SHBG analysis in men. S1.4 A-C Fig. Manhattan and QQ Plots for the G x SHBG analysis in women. S1.5 A-C Fig. Manhattan and QQ Plots for the G x Bioavailable Testosterone analysis in men. S1.6 A-C Fig. Manhattan and QQ Plots for the G x Bioavailable Testosterone analysis in women.
(DOCX)
(DOCX)
(DOCX)
(DOCX)
(DOCX)
(DOCX)
(DOCX)
Acknowledgments
This research has been conducted using the UK Biobank Resource under Application Number “34031”. Yan V. Sun is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Data Availability
The data that support the findings of this study will be publicly available from the UNC Dataverse with the identifier https://doi.org/10.15139/S3/UG2NLU.
Funding Statement
This work is supported in part by funding from the National Institute of Health (HL154996 to YVS, DK139632 to YVS, and HL156991 to YVS). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. YVS received salary support from the National Institute of Health.
References
- 1.National Diabetes Statistics Report. Diabetes. 2022 [cited 2023 May 2]. https://www.cdc.gov/diabetes/data/statistics-report/index.html
- 2.Diabetes. n.d. [cited 2023 May 2]. https://www.who.int/news-room/fact-sheets/detail/diabetes
- 3.Ali MK, Pearson-Stuttard J, Selvin E, Gregg EW. Interpreting global trends in type 2 diabetes complications and mortality. Diabetologia. 2022;65(1):3–13. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.National Institute of Diabetes and Digestive and Kidney Diseases. Risk Factors for Type 2 Diabetes. n.d. [cited 2023 May 2]. https://www.niddk.nih.gov/health-information/diabetes/overview/risk-factors-type-2-diabetes
- 5.Huebschmann AG, Huxley RR, Kohrt WM, Zeitler P, Regensteiner JG, Reusch JEB. Sex differences in the burden of type 2 diabetes and cardiovascular risk across the life course. Diabetologia. 2019;62(10):1761–72. doi: 10.1007/s00125-019-4939-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.International Diabetes Federation. IDF Diabetes Atlas. 10th edn. Brussels, Belgium. 2021. [cited 2024 Jan 24]. https://diabetesatlas.org/ [Google Scholar]
- 7.Gambineri A, Pelusi C. Sex hormones, obesity and type 2 diabetes: is there a link?. Endocr Connect. 2018;8(1):R1-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Ruth KS, Day FR, Tyrrell J, Thompson DJ, Wood AR, Mahajan A, et al. Using human genetics to understand the disease impacts of testosterone in men and women. Nat Med. 2020;26(2):252–8. doi: 10.1038/s41591-020-0751-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.O’Reilly MW, Glisic M, Kumarendran B, Subramanian A, Manolopoulos KN, Tahrani AA, et al. Serum testosterone, sex hormone-binding globulin and sex-specific risk of incident type 2 diabetes in a retrospective primary care cohort. Clin Endocrinol (Oxf). 2019;90(1):145–54. doi: 10.1111/cen.13862 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Testosterone Therapy for Hypogonadism Guideline Resources. n.d. [cited 2023 May 2]. https://www.endocrine.org/clinical-practice-guidelines/testosterone-therapy
- 11.Almgren P, Lehtovirta M, Isomaa B, Sarelin L, Taskinen MR, Lyssenko V, et al. Heritability and familiality of type 2 diabetes and related quantitative traits in the Botnia Study. Diabetologia. 2011;54(11):2811–9. doi: 10.1007/s00125-011-2267-5 [DOI] [PubMed] [Google Scholar]
- 12.DeForest N, Majithia AR. Genetics of Type 2 Diabetes: Implications from Large-Scale Studies. Curr Diab Rep. 2022;22(5):227–35. doi: 10.1007/s11892-022-01462-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Xue A, Wu Y, Zhu Z, Zhang F, Kemper KE, Zheng Z, et al. Genome-wide association analyses identify 143 risk variants and putative regulatory mechanisms for type 2 diabetes. Nat Commun. 2018;9(1):2941. doi: 10.1038/s41467-018-04951-w [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Lamri A, De Paoli M, De Souza R, Werstuck G, Anand S, Pigeyre M. Insight into genetic, biological, and environmental determinants of sexual-dimorphism in type 2 diabetes and glucose-related traits. Front Cardiovasc Med. 2022;9:964743. doi: 10.3389/fcvm.2022.964743 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Francis M, Li C, Sun Y, Zhou J, Li X, Brenna JT, et al. Genome-wide association study of fish oil supplementation on lipid traits in 81,246 individuals reveals new gene-diet interaction loci. PLoS Genet. 2021;17(3):e1009431. doi: 10.1371/journal.pgen.1009431 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Hartiala JA, Hilser JR, Biswas S, Lusis AJ, Allayee H. Gene-Environment Interactions for Cardiovascular Disease. Curr Atheroscler Rep. 2021;23(12):75. doi: 10.1007/s11883-021-00974-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Kilpeläinen TO, Lakka TA, Laaksonen DE, Laukkanen O, Lindström J, Eriksson JG, et al. Physical activity modifies the effect of SNPs in the SLC2A2 (GLUT2) and ABCC8 (SUR1) genes on the risk of developing type 2 diabetes. Physiol Genomics. 2007;31(2):264–72. doi: 10.1152/physiolgenomics.00036.2007 [DOI] [PubMed] [Google Scholar]
- 18.Kilpeläinen TO, Bentley AR, Noordam R, Sung YJ, Schwander K, Winkler TW. Multi-ancestry study of blood lipid levels identifies four loci interacting with physical activity. Nat Commun. 2019;10(1):376. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Lee YC, Lai CQ, Ordovas JM, Parnell LD. A database of gene-environment interactions pertaining to blood lipid traits, cardiovascular disease and type 2 diabetes. J Data Min Genomics Proteomics. 2011;2(1):106. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Wu P, Rybin D, Bielak LF, Feitosa MF, Franceschini N, Li Y, et al. Smoking-by-genotype interaction in type 2 diabetes risk and fasting glucose. PLoS One. 2020;15(5):e0230815. doi: 10.1371/journal.pone.0230815 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Zheng J-S, Lai C-Q, Parnell LD, Lee Y-C, Shen J, Smith CE, et al. Genome-wide interaction of genotype by erythrocyte n-3 fatty acids contributes to phenotypic variance of diabetes-related traits. BMC Genomics. 2014;15(1):781. doi: 10.1186/1471-2164-15-781 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Zheng J-S, Arnett DK, Lee Y-C, Shen J, Parnell LD, Smith CE, et al. Genome-wide contribution of genotype by environment interaction to variation of diabetes-related traits. PLoS One. 2013;8(10):e77442. doi: 10.1371/journal.pone.0077442 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Suzuki K, Hatzikotoulas K, Southam L, Taylor HJ, Yin X, Lorenz KM, et al. Genetic drivers of heterogeneity in type 2 diabetes pathophysiology. Nature. 2024;627(8003):347–57. doi: 10.1038/s41586-024-07019-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Fantus RJ, Na R, Wei J, Shi Z, Resurreccion WK, Halpern JA. Genetic Susceptibility for Low Testosterone in Men and Its Implications in Biology and Screening: Data from the UK Biobank. Eur Urol Open Sci. 2021;29:36–46. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Mahajan A, Wessel J, Willems SM, Zhao W, Robertson NR, Chu AY, et al. Refining the accuracy of validated target identification through coding variant fine-mapping in type 2 diabetes. Nat Genet. 2018;50(4):559–71. doi: 10.1038/s41588-018-0084-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Najafi M, Rafiei A, Ghaemi A, Hosseini V. Association between rs738408, rs738409 and rs139051 polymorphisms in PNPLA3 gene and non-alcoholic fatty liver disease. Gene Rep. 2022;26:101472. [Google Scholar]
- 27.Huang J, Huffman JE, Huang Y, Do Valle Í, Assimes TL, Raghavan S, et al. Genomics and phenomics of body mass index reveals a complex disease network. Nat Commun. 2022;13(1):7973. doi: 10.1038/s41467-022-35553-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Kichaev G, Bhatia G, Loh PR, Gazal S, Burch K, Freund MK. Leveraging Polygenic Functional Enrichment to Improve GWAS Power. Am J Hum Genet. 2019;104(1):65–75. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Wheeler E, Huang N, Bochukova EG, Keogh JM, Lindsay S, Garg S, et al. Genome-wide SNP and CNV analysis identifies common and low-frequency variants associated with severe early-onset obesity. Nat Genet. 2013;45(5):513–7. doi: 10.1038/ng.2607 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Deng X, Sabino EC, Cunha-Neto E, Ribeiro AL, Ianni B, Mady C, et al. Genome wide association study (GWAS) of Chagas cardiomyopathy in Trypanosoma cruzi seropositive subjects. PLoS One. 2013;8(11):e79629. doi: 10.1371/journal.pone.0079629 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Liu H, Doke T, Guo D, Sheng X, Ma Z, Park J, et al. Epigenomic and transcriptomic analyses define core cell types, genes and targetable mechanisms for kidney disease. Nat Genet. 2022;54(7):950–62. doi: 10.1038/s41588-022-01097-w [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Morris AP, Voight BF, Teslovich TM, Ferreira T, Segrè AV, Steinthorsdottir V, et al. Large-scale association analysis provides insights into the genetic architecture and pathophysiology of type 2 diabetes. Nat Genet. 2012;44(9):981–90. doi: 10.1038/ng.2383 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Hara K, Fujita H, Johnson TA, Yamauchi T, Yasuda K, Horikoshi M, et al. Genome-wide association study identifies three novel loci for type 2 diabetes. Hum Mol Genet. 2014;23(1):239–46. doi: 10.1093/hmg/ddt399 [DOI] [PubMed] [Google Scholar]
- 34.Lagou V, Mägi R, Hottenga JJ, Grallert H, Perry JRB, Bouatia-Naji N, et al. Sex-dimorphic genetic effects and novel loci for fasting glucose and insulin variability. Nature Communications. 2021;12:24. doi: 10.1038/s41467-020-20112-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Conroy MC, Lacey B, Bešević J, Omiyale W, Feng Q, Effingham M, et al. UK Biobank: a globally important resource for cancer research. Br J Cancer. 2023;128(4):519–27. doi: 10.1038/s41416-022-02053-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Watts EL, Perez-Cornago A, Knuppel A, Tsilidis KK, Key TJ, Travis RC. Prospective analyses of testosterone and sex hormone-binding globulin with the risk of 19 types of cancer in men and postmenopausal women in UK Biobank. Int J Cancer. 2021;149(3):573–84. [DOI] [PubMed] [Google Scholar]
- 37.Vermeulen A, Verdonck L, Kaufman JM. A critical evaluation of simple methods for the estimation of free testosterone in serum. J Clin Endocrinol Metab. 1999;84(10):3666–72. doi: 10.1210/jcem.84.10.6079 [DOI] [PubMed] [Google Scholar]
- 38.Bycroft C, Freeman C, Petkova D, Band G, Elliott LT, Sharp K, et al. The UK Biobank resource with deep phenotyping and genomic data. Nature. 2018;562(7726):203–9. doi: 10.1038/s41586-018-0579-z [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.UK Biobank release and systematic evaluation of optimised polygenic risk scores for 53 diseases and quantitative traits. medRxiv. n.d. [cited 024 June 24]. https://www.medrxiv.org/content/10.1101/2022.06.16.22276246v1
- 40.Westerman KE, Pham DT, Hong L, Chen Y, Sevilla-González M, Sung YJ. GEM: scalable and flexible gene–environment interaction analysis in millions of samples. Bioinformatics. 2021;37(20):3514–20. doi: 10.1093/bioinformatics/btab204 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Watanabe K, Taskesen E, van Bochoven A, Posthuma D. Functional mapping and annotation of genetic associations with FUMA. Nat Commun. 2017;8(1):1826. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Lawrence M, Huber W, Pagès H, Aboyoun P, Carlson M, Gentleman R, et al. Software for computing and annotating genomic ranges. PLoS Comput Biol. 2013;9(8):e1003118. doi: 10.1371/journal.pcbi.1003118 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Examination of the SHBG × PRST2D interaction in women. Menopause sensitivity analysis. Incident type 2 diabetes sensitivity analysis. Replication study in South Asian and African ancestry cohorts.
(DOCX)
(DOCX)
(DOCX)
(DOCX)
(DOCX)
S1.2 A-C Fig. Manhattan and QQ Plots for the G x Total Testosterone analysis in women. S1.3 A-C Fig. Manhattan and QQ Plots for the G x SHBG analysis in men. S1.4 A-C Fig. Manhattan and QQ Plots for the G x SHBG analysis in women. S1.5 A-C Fig. Manhattan and QQ Plots for the G x Bioavailable Testosterone analysis in men. S1.6 A-C Fig. Manhattan and QQ Plots for the G x Bioavailable Testosterone analysis in women.
(DOCX)
(DOCX)
(DOCX)
(DOCX)
(DOCX)
(DOCX)
(DOCX)
Data Availability Statement
The data that support the findings of this study will be publicly available from the UNC Dataverse with the identifier https://doi.org/10.15139/S3/UG2NLU.