NRF2 Genetic Polymorphism Modifies the Association of Plasma Selenium Levels With Incident Coronary Heart Disease Among Individuals With Type 2 Diabetes

Selenium is an essential trace mineral absorbed from the diet and incorporated into the polypeptides, forming a variety of selenoproteins essential for human health (1,2). As a major component of selenoproteins, selenium exerts metabolic and antioxidant functions in enzymatic form (3). Selenium deficiency is increasingly recognized as a global concern (3). Consistent with a World Health Organization (WHO) report that China is 1 of 40 principal selenium-deficit countries, a nationwide review based on 187,732 soil samples from 30 provinces showed that adverse health effects of selenium deficiency afflicted 51% of the area of China (4,5). Selenium deficiency refers to low physiological selenium levels, resulting in limited ability to withstand oxidative stress, and is associated with various human diseases, including Keshan disease, cancer, and coronary heart disease (CHD) (6,7). Some studies also found that increased plasma selenium levels were associated with lower cardiovascular disease (CVD) risk, particularly CHD (6,8). In our latest research, we also observed the inverse association among people with type 2 diabetes (T2D) (9). However, whether the protective effect of selenium against CHD differs across populations with different genetic predispositions to regulating antioxidant response remains to be elucidated.

T2D is an expanding global health problem characterized by elevated fasting plasma glucose levels, essentially caused by insulin resistance, impaired islet β-cell function, or a combination of both (10). The hyperglycemia state of patients with diabetes may lead to a higher state of oxidative stress (11), leaving the body vulnerable to endothelial damage, impaired angiogenesis, and atheroma formation, eventually translating into CHD (12). In addition, growing evidence demonstrated that nuclear factor erythroid 2–related factor 2 (NRF2) plays a crucial role in oxidative stress–related illnesses. NRF2 has been repeatedly demonstrated to play a central role in antioxidant signaling (13,14). As a critical upstream promoter polymorphism in NRF2, genetic variation of rs6721961 has been widely investigated and reported to be associated with higher oxidative stress levels, diabetes, higher blood pressure, cardiovascular mortality, and coronary artery disease severity (15–17). Other studies also indicated that the mRNA and protein levels of NRF2 in the GT/TT genotype group (i.e., risk-allele carriers) were lower than those in the GG genotype group (i.e., those not carrying the risk alleles) (15,18,19), confirming rs6721961 to be biologically functional. All this evidence supports that rs6721961 in the NRF2 promoter may be a critical variant that profoundly contributes to subsequent disease development by affecting NRF2 expression.

As an indicator of selenium nutritional status, plasma selenium levels were directly associated with higher expression of NRF2-targeted genes in 96 healthy participants in one study (20). Experimental studies also illustrated not only that selenium regulated the NRF2 pathway but selenium status was affected in an NRF2-dependent manner (21,22). Using hepatic selenoprotein mRNA (e.g., SELENOW) as a biomarker of selenium status, when fed a selenium-deficiency diet, SELENOW expression in Nrf2^−/− mice was comparable to that of wild-type mice but was significantly lower under a normal selenium-level diet (22). Additionally, reduced NRF2 expression level due to NRF2 promoter polymorphisms (e.g., T allele in rs6721961) reduces the expression of several selenium-dependent enzymes, including thioredoxin reductase and glutathione peroxidase, disturbing the antioxidant system (23). The impairment of the antioxidant system would prolong oxidative stress and enhance the expression of various inflammatory cytokines, causing endothelial dysfunction and resulting in CHD (24). We hypothesized that high plasma selenium levels would be associated with a decreased risk of CHD among T allele carriers of rs6721961, a functional single nucleotide polymorphism (SNP) of NRF2.

Therefore, our primary aim was to investigate whether there is an association between the NRF2 promoter genetic variant (rs6721961) and incident CHD risk and whether the NRF2 promoter polymorphism modifies the association of plasma selenium levels with incident CHD risk in study participants with diabetes.

The detailed description of the participants in the study has been reported elsewhere (9). In brief, the participants were selected from the Dongfeng-Tongji cohort, an ongoing prospective project in Shiyan, a city in the province of Hubei in China (25). In 2008, a total of 27,009 retirees were recruited, with a response rate of 87%. A total of 25,978 participants at baseline were successfully followed in 2013, with a follow-up rate of 96.2% (6). As Supplementary Fig. 1 illustrates, from baseline in 2008 (n = 5,173) and the first follow-up period in 2013 (n = 1,509), a total of 6,682 participants diagnosed with T2D were included and followed up to December 2018. After giving written informed consent, all participants provided information on sociodemographic characteristics through face-to-face interviews and were administered clinical examinations by professional medical staff. Through inspecting questionnaires or medical records, we excluded participants who self-reported or were diagnosed with CHD and those with self-reported stroke, cancer, or abnormal electrocardiogram results at baseline (n = 2,510). After excluding those with an estimated glomerular filtration rate (eGFR) <30 mL/min/1.73 m² (n = 35) and those without genotype data (n = 1,500) and not enough plasma specimens for baseline metal determination (n = 386), a total of 2,251 participants were finally enrolled for further analysis. The study was approved by the Ethics and Human Subject Committees of the School of Public Health, Tongji Medical College, Huazhong University of Science and Technology (HUST), and Dongfeng General Hospital of the Dongfeng Motor Corporation (DMC).

Blood samples were collected after overnight fasting. Plasma was separated from EDTA-anticoagulated blood after centrifuging and was frozen at −80°C until selenium measurement. Plasma selenium concentrations were measured using inductively coupled plasma mass spectrometry (Agilent 7700× ICP-MS; Agilent Technologies) when study participants with diabetes were included (26). Briefly, 100 μL of plasma, after melting, was diluted to 2 mL using 1% nitric acid, mixed to guarantee homogeneity and ready for determination. Standard reference materials SRM 1,640a (Trace Elements in Natural Water; National Institute of Standards and Technology, Gaithersburg, MD) and certified reference agents (ClinChek human plasma controls for trace elements no. 8,883 and 8,884; Recipe, Munich, Germany) were also detected in every 20 samples for intraday and interday comparison to ensure the reliability and accuracy of detection. The limit of detection (LOD) of selenium was 0.00969 μg/L, and we substituted plasma selenium values below LOD with the value LOD/2.

The NRF2 promoter polymorphism rs6721961 was genotyped using Affymetrix Genome-Wide Human SNP Array 6.0 chips (Santa Clara, CA) and Illumina Infinium OmniZhongHua-8 Kit chips (San Diego, CA). Specific genotyping and quality-control procedures have been reported in previous studies (27,28). The SNP satisfies the Hardy–Weinberg equilibrium (P = 0.32), and the minor allele frequency was 0.281 for rs6721961.

The definition of incident CHD cases was the first occurrence of fatal or nonfatal cardiac events, including nonfatal myocardial infarction, recanalization of acute coronary ischemia (e.g., angioplasty, coronary bypass surgery), stable or unstable angina pectoris, fatal CHD, following the recommendations and guidelines from the American Heart Association (29). Furthermore, CHD cases were diagnosed by the clinical signs, cardiac enzymes, electrocardiogram, or coronary angiography (stenosis of one or more main coronary arteries ≥50%), according to the criteria recommended by WHO (29). All participants could be tracked through the health care service system of the Dongfeng Corporation, making it convenient to trace the specific causes of morbidity and mortality. Suspected cases were further verified by reviewing the medical insurance documents, medical records, and death registration up to December 2018 and were adjudicated by an expert panel of physicians.

Information on demographic and lifestyle covariates was self-reported at baseline, including age; sex; education level (primary school or less, junior high school, senior high school, college or more); cigarette-smoking and alcohol-drinking status (never, former, current); physical activity (yes, no); sleep quality (good, fair, poor, terrible [use of sleeping pills]); night sleep duration (<7 h, 7 to < 8 h, 8 to < 9 h, ≥9 h); family history of CVD (yes, no); and dietary factors, including coarse grains, meat, fish, beans, milk, eggs, vegetables, and fruit intake (≥5 days/week). Trained medical staff measured body weight and standing height. BMI was then calculated by the ratio of body weight in kilograms to the square of standing height in meters. Hypertension (yes, no) or hyperlipidemia (yes, no) was defined by comprehensively considering clinical examination; laboratory test results; self-reported, physician-diagnosed disease history; and oral drug use, which have been described previously (25). We used the glomerular filtration rate–estimating equation modified on the basis of Chinese individuals with chronic kidney disease to compute the eGFR (30). We applied the glucose oxidase method to determine fasting glucose levels by an automatic biochemistry analyzer (ARCHITECT ci8200; Abbott). The duration of diabetes was calculated as the baseline age at entering into cohort minus the age of diabetes onset, which was obtained through self-report in the baseline questionnaires or telephone verification for previously diagnosed T2D, and the date of clinical examination for newly identified patients.

T2D denotes a combined consideration of self-reported diagnosed diabetes, fasting glucose level ≥7.0 mmol/L (126 mg/dL) at recruitment to the study, or currently taking hypoglycemic medication or using insulin, according to the guidelines recommended by American Diabetes Association (31).

Baseline characteristics were compared using ANOVA or Kruskal-Wallis tests for continuous variables and χ² tests for categorical variables and are presented as the mean with SD or median with interquartile range for continuous variables or as numbers with percentages for categorical variables. Person-years for each participant were calculated from the enrollment date to the time of diagnosis of CHD, censoring, or the end of the follow-up (i.e., 31 December 2018), whichever came first.

The Cox proportional-hazards model was primarily performed to estimate hazard ratios (HRs) of plasma selenium levels associated with incident CHD risk among participants with T2D, using chronological age as the time scale. Moreover, plasma selenium levels were introduced in models as quartiles (categorical) or SD-divided transformed (continuous) variables. Quartiles of selenium level were defined according to the distribution of the baseline concentration, and the lowest quartile was assigned as the reference group. Model 1 was adjusted for sex, age, education level, and family history of CVD. Model 2 was adjusted for covariates in model 1 and further adjusted for hypertension, hyperlipidemia, eGFR, baseline antidiabetic medication use, baseline glucose level, BMI, and baseline diabetes duration. Model 3 was adjusted for covariates in model 2 and further adjusted for cigarette-smoking status, alcohol-drinking status, physical activity, sleep quality, sleep duration, and intake of coarse grains, meat, fish, egg, beans, milk, fruits, and vegetables.

The Fisher exact test was performed to examine whether the SNP meets the Hardy–Weinberg equilibrium. The allele frequency of SNP was calculated on the basis of the whole cohort, using R package genetics. A Cox proportional-hazards model was also used to assess the rs6721961 genotype and subsequent CHD risk among participants with T2D under three genetic (dominant, recessive, or additive) patterns. In addition, we performed linear models to investigate the associations between rs6721961 genotype under three patterns and plasma selenium levels with adjustment for sex and age. Moreover, the marginal means in each genotype (GG, GT, and TT) were estimated using the lsmeans function in the R package emmeans.

We examined gene–environment interactions by introducing the multiplicative interaction term between genotypes and plasma selenium levels in the Cox proportional-hazards models under three genetic patterns. In addition, we used the restricted cubic splines with one knot each placed at 10th, 50th, and 90th percentiles of the selenium level to assess the nonlinear relationship of plasma selenium levels with CHD risk among participants with T2D stratified by the dominant pattern of rs6721961. Plasma selenium concentration at the 10th percentile was assigned as the reference. Furthermore, we examined the association between rs6721961 genotype and subsequent CHD risk among participants with T2D stratified by dichotomized plasma selenium levels using the Cox proportional-hazards model. Adjustment factors were the same as in model 3.

The trend test was performed by splitting selenium into intervals, with each interval's median as a continuous variable, and modeling this variable into the Cox proportional-hazards model. A two-sided P < 0.05 was considered statistically significant, and all analyses were conducted using R software (version 3.6.3; R Foundation, Vienna, Austria).

The data sets generated and/or analyzed during the current study are available from the corresponding author upon reasonable request.

Among the 2,251 participants with T2D (women, 52.2%; mean age [SD], 65.13 [7.32] years; median diabetes duration, 0.49 years), the median plasma selenium concentration was 70.7 μg/L. Compared with the lowest quartile (<58.7 μg/L), the participants in the highest quartile (≥85.3 μg/L) were more likely to be female; more educated; never smokers; more prone to have the higher consumption frequency of meat, fish, and eggs; lower BMI; and more likely to have hyperlipidemia, family history of CVD, and higher fasting glucose level (P < 0.05 for all; Table 1).

During a mean follow-up of 6.90 ± 2.96 years, we observed 798 incident CHD cases. As Table 2 shows, plasma selenium was inversely associated with CHD risk among participants with T2D after adjusting for age, sex, education level, and family history of CVD (the HR for each 22.8 μg/L in plasma selenium was 0.92; 95% CI 0.85–0.99). After further controlling for disease status and lifestyle factors, the inverse association remained robust (HR 0.91; 95% CI 0.84–0.98). Compared with participants with selenium levels <58.7 μg/L, the multivariable-adjusted HRs (95% CIs) for those with ≥58.7 to 70.7, ≥70.7 to 85.3, and ≥85.3 μg/L of selenium concentrations were 0.81 (0.67–0.99), 0.82 (0.67–1.00), and 0.72 (0.58–0.88), respectively, for the incident CHD risk among T2D.

Table 3 shows that carrying the rs6721961 T allele in the NRF2 promoter may be associated with a higher risk of incident CHD among people with T2D. Compared with the GG genotype in rs6721961, the GT genotype was significantly associated with a 17% higher risk of CHD among participants with T2D (HR 1.17; 95% CI 1.01–1.35) after adjustment for potential covariates. In contrast, the TT genotype showed a positive but nonsignificant association with CHD risk among participants with T2D. Similar positive associations were observed between rs6721961 genotypes and CHD risk among those with T2D in dominant and additive patterns; the corresponding multivariable-adjusted HRs (95% CIs) were 1.17 (1.02–1.35) and 1.13 (1.01–1.26), respectively. Additionally, the T allele of rs6721961 was marginally associated with lower plasma selenium levels under the additive pattern (P = 0.052) (Supplementary Fig. 2).

Figure 1 demonstrates the potential interaction between plasma selenium levels and the NRF2 promoter polymorphism of rs6721961 on incident CHD risk among participants with T2D. In the dominant pattern, among participants with T2D, those with the highest quartile of plasma selenium levels were associated with a 0.54-fold (95% CI 0.40–0.73) risk of incident CHD compared with the lowest quartile, and the adjusted HR for incident CHD risk among those with T2D with each 22.8-μg/L increase in plasma selenium was 0.80 (95% CI 0.71–0.89) among GT/TT carriers. Nevertheless, the inverse association was not found in participants with GG genotype (P for interaction = 0.004). Moreover, the restricted cubic spline curves show that increased plasma selenium levels were linearly associated with a lower risk of incident CHD risk among participants with T2D with GT/TT genotype (P for overall association = 0.0007; P for nonlinear association >0.05) but not in GG carriers (P for overall association >0.05) (Fig. 2). Interestingly, when further stratifying the risk-allele carriers (i.e., GT/TT), we found that the inverse associations of plasma selenium levels with incident CHD risk among participants with T2D were further enhanced as risk-allele numbers increased. Each 22.8-μg/L increase in plasma selenium levels was associated with a 0.63-fold (95% CI 0.41–0.98) risk of incident CHD risk among those with T2D with the TT genotype. Nevertheless, in the same scenario, the multivariable-adjusted HR was 0.82 (95% CI 0.72–0.92) among those with the GT genotype. Modification effects on the association of plasma selenium levels with incident CHD risk among participants with T2D were also persistently observed in promoter polymorphism rs6721961 of the NRF2 gene under three genetic patterns (P for interaction < 0.05 for all; Fig. 1).

From another perspective, we found that only in cases of low plasma selenium status was carrying the rs6721961 T allele in the NRF2 promoter significantly positively associated with CHD risk among participants with T2D (HR 1.32; 95% CI 1.08–1.60) after adjustment for potential covariates (Fig. 3). Notably, under the additive pattern, we observed a significant dose-dependent association between the risk allele (T) and incident CHD risk among those with T2D with low selenium status (for each T allele increase, HR 1.26; 95% CI 1.08–1.46) (Fig. 3). Compared with the GG genotype in rs6721961, the HRs (95% CIs) of the GT and TT genotypes were 1.28 (1.04–1.57) and 1.54 (1.08–2.20), respectively. By contrast, the positive association between rs6721961 and CHD risk among those with T2D was not found among participants with high selenium status (P for interaction = 0.02) (Fig. 3).

To our knowledge, this is the first study to evaluate the interaction between plasma selenium levels and NRF2 promoter polymorphism on the incident CHD risk among participants with diabetes. Elevated plasma selenium levels were associated with decreased incident CHD risk among participants with T2D with the T allele of SNP rs6721961 but not those with the GG genotype. We also found that the T allele in the NRF2 promoter polymorphism was positively associated with incident CHD risk among participants with T2D, especially those with low selenium status.

Plasma selenium level was often used to represent physiological nutrition selenium status. The median level of plasma selenium (70.7 μg/L) in the present study was relatively lower than the reported values of other studies from southeast (213.0 μg/L) and central (112.7 μg/L) China (32,33). The selenium level in our study was also lower than the mean level (129.7 μg/L) in participants aged >60 years in the National Health and Nutrition Examination Survey (34). To date, mounting evidence suggests that elevated circulating selenium at a relatively low level might be a protective factor for CVDs (6). Additionally, the protective effects of selenium in people were primarily observed in those regions with low selenium levels rather than in regions with sufficient background selenium content, like the United States, consistent with the findings in Asia and Europe, as we reported previously (9). Thus, the findings in the present study would be of pivotal importance for the population with relatively lower physiological selenium status.

We also found a positive association of NRF2 polymorphism with incident CHD risk among participants with T2D. Previous studies have demonstrated that the rs6721961 risk allele (TT or GT genotype) was associated with a higher risk of breast cancer, cerebrovascular disease, CHD, atherosclerotic severity, and diabetic nephropathy (15,18,34,35). Polymorphism rs6721961 is a variant located in the antioxidant response element–like promotor binding sites with an autoregulatory role on NRF2 expression (18,36). NRF2 is a transcription factor essential for redox homeostasis, protecting the critical tissues or cells against oxidation (36). In animal models, Nrf2 activation attenuates hyperglycemia-evoked reactive oxygen species in diabetes by inhibiting the proinflammatory effects (37), reversing endothelial cell dysfunction, and delaying vascular disease development in diabetes (38). Previous findings illustrated that the coronary artery disease–related genotype (TT) had lower expression of NRF2 mRNA, and those findings were validated by the same low NRF2 expression of whole blood in Genotype-Tissue Expression, as illustrated in Supplementary Fig. 3 (39). Additionally, NRF2 is an indispensable component for protecting phagocytic cells from oxidized low-density lipoprotein in diabetic rats (40). All this evidence suggests an essential role of NRF2 in diabetic cardiovascular complications.

Furthermore, our results demonstrate that the T allele of rs6721961 in the NRF2 promoter modified the association between plasma selenium and CHD risk among participants with T2D. In particular, an inverse association between plasma selenium and incident CHD was only observed among those with the risk-allele (i.e., TT or GT) genotype. Previous observational studies directly linked the NRF2 pathway to plasma selenium. Reszka et al. (20) observed that plasma selenium levels were inversely associated with transcriptional levels of NRF2-targeted genes (e.g., GSTP1, PRDXR1, SOD1) among 96 healthy, nonsmoking men in Poland. Experimental studies also supported the interaction between NRF2 genetic variation and plasma selenium, because NRF and selenoproteins both exert essential functions in maintaining redox homeostasis and could exert functions independent of each other (41). Animal studies revealed that sufficient selenium modulates the antioxidant system to maintain redox homeostasis independent of the Nrf2-regulated antioxidant response pathway (42). When mice were fed suboptimal, adequate, and excess selenium, respectively, glutathione peroxidase activity and hepatic selenoprotein mRNA levels exhibited a dose-dependent increase in Nrf2^−/− mice, but this was not observed in wild-type mice (22). Findings from another rodent study also indicated that in Nrf2^−/− mice, nutritional selenium deficiency led to the failure of weight gain during growth at 16 and 22 weeks, but an adequate selenium diet rescued the Nrf2^−/− mice from the growth failure, suggesting adequate selenium compensates for the deficiency of low Nrf2 expression (43). Under low Nrf2 expression status, adequate selenium would induce the synthesis of selenoproteins (e.g., TrxR1), directly exerting antioxidant functions independent of Nrf2 (23,42). These findings suggested that impaired NRF2 expression requires adequate selenium supplementation to maintain normal antioxidant functions by synthesizing selenoproteins to protect the body against damage from oxidative stress.

In addition, we found that only in low selenium status was the T allele in the NRF2 promoter polymorphism associated with increased CHD risk among study participant with T2D. Generally, the transcription factor encoded by NRF2 is required for the body to synthesize antioxidant enzymes to defend against oxidative stress (13,14). However, the NRF2 promoter polymorphism reduced the expression levels of NRF2 and the downstream antioxidant enzymes, thereby compromising antioxidant defense systems (44). Previous experimental studies demonstrated that adequate selenium was required for synthesizing many antioxidant selenoproteins, such as TrxR1, which inhibits Nrf2 activation (23). However, selenium deficiency would downregulate selenoproteins such as TrxR1 and upregulate Nrf2, which exerts antioxidant functions through other Nrf2-dependent pathways (21,23). This mechanism was illustrated in a recent rodent study. In selenium-optimal mice, because of higher levels of basal selenium-dependent antioxidant enzymes (i.e., TrxR1), activities of selenium-dependent antioxidant systems were enhanced to serve as the first-line defense system (45). However, in selenium-deficit mice, the Nrf2 antioxidant system was activated due to the lower basal levels of TrxR1 or other selenium-dependent enzymes (45). If the NRF2 system could not be fully activated because of the NRF2 promoter variant under selenium-deficiency status, prolonged oxidative stress would disturb redox homeostasis, then persistently affect the downstream inflammation response, thus resulting in subsequent CHD development. The underlying mechanism of how plasma selenium modifies the effect of the NRF2 variant on CHD requires further confirmation.

Although the prospective cohort study design with a relatively long follow-up period enabled us to control biases and inform the causal association, several limitations of this study should be noted. First, each selenium speciation could be distinguished and identified by the total plasma selenium level in the study, because the function of selenium might be affected by its speciation. However, plasma selenium was generally considered to represent the nutritional status of selenium and reflect the overall selenium level (46). Second, plasma selenium levels were measured only at baseline, possibly leading to misclassification of exposure, because plasma selenium levels would fluctuate during follow-up. However, a previous study of 138 individuals found a significant correlation between selenium levels in plasma collected at 5-year intervals, suggesting a small probability of exposure misclassification (6). Third, the CHD incidence in this population (47) is relatively high compared with those reported in other Chinese populations (48). This might be because, in the present study, we defined incident CHD according to the WHO guidelines and included both CHD cases and deaths. Also, we verified CHD events through a combination of medical insurance documents, medical records, the death registry, and manual review of all medical records of each individual, making it possible to trace all CHD cases. Finally, although we collected information on lifestyle factors at baseline, these factors may have changed during the follow-up period, leading to potential residual confounding.

In summary, plasma selenium levels are inversely associated, at a low-to-moderate level, with incident CHD risk in individuals with diabetes, and the association may be modified by NRF2 promoter polymorphism. Our findings suggest that genetic susceptibility to defense against oxidative stress should be considered when using selenium as an antioxidant supplement. Mechanistic studies are warranted to verify and confirm these findings.

Acknowledgments. The authors thank all staff at the Dongfeng Central Hospital for their contributions to the fieldwork. We especially appreciate all the participants involved in this study.

Funding. This study was supported by grants from the National Natural Science Foundation (grants 82073656 to M.H. and NSFC-81930092 to T.W.), the Program for HUST Academic Frontier Youth Team (grant 2017QYTD18 to M.H.), and the National Key Research and Development Program of China (grants 2016YFC0900800 to T.W. and 2017YFC0907500 to M.H.).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. C.J. contributed to conception and study design, performed analyses, interpreted results, and critically drafted and revised the manuscript. R.W., T.L., Y.X., and Y.Z. participated in the study design, data collection, laboratory analyses, and manuscript revision. R.P., X.Z., H.G., H.Y., and T.W. participated in the maintenance of the cohort and study design. M.H. 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.