Genetically predicted circulating vitamin C in relation to cardiovascular disease

Aim We conducted a two-sample Mendelian randomization (MR) study to assess the associations of genetically predicted circulating vitamin C levels with cardiovascular diseases (CVDs). Methods and results Ten lead single-nucleotide polymorphisms associated with plasma vitamin C levels at the genome-wide significance level were used as instrumental variables. Summary-level data for 15 CVDs were obtained from corresponding genetic consortia, the UK Biobank study, and the FinnGen consortium. The inverse-variance-weighted method was the primary analysis method, supplemented by the weighted median and MR-Egger methods. Estimates for each CVD from different sources were combined. Genetically predicted vitamin C levels were not associated with any CVD after accounting for multiple testing. However, there were suggestive associations of higher genetically predicted vitamin C levels (per 1 standard deviation increase) with lower risk of cardioembolic stroke [odds ratio, 0.79; 95% confidence interval (CI), 0.64, 0.99; P = 0.038] and higher risk of atrial fibrillation (odds ratio, 1.09; 95% CI, 1.00, 1.18; P = 0.049) in the inverse-variance-weighted method and with lower risk of peripheral artery disease (odds ratio, 0.76, 95% CI, 0.62, 0.93; P = 0.009) in the weighted median method. Conclusion We found limited evidence with MR techniques for an overall protective role of vitamin C in the primary prevention of CVD. The associations of vitamin C levels with cardioembolic stroke, atrial fibrillation, and peripheral artery disease need further study.


Introduction
Vitamin C as an antioxidant has been proposed to alleviate oxidative stress and affect vascular remodelling, endothelial function, and lipid peroxidation, thereby potentially having a protective role in cardiovascular disease (CVD). [1][2][3] Observational data have indicated that a high circulating level or intake of vitamin C is associated with a reduced risk of CVD and corresponding mortality. [4][5][6][7][8] Several biological processes and signalling pathways have been proposed to be involved in the potential therapeutic effects of vitamin C on CVD. 9 However, the cardio-protective effect of vitamin C has not been vali- with vitamin C alone or together with other antioxidative vitamins. 3,10,11 Thus, any causal relationship between vitamin C and CVD remains unestablished given potential confounding in previous observational findings and certain limitations of RCTs (e.g. a small sample size, imbalanced baseline characteristics, combined supplementation of vitamin C with other nutrients, and low compliance to intervention).
Utilizing genetic variants as instrumental variables for an exposure (e.g. plasma vitamin C levels) allows the Mendelian randomization (MR) design to more plausibly investigate causal inferences by minimizing residual confounding and other biases. Here, we conducted a two-sample MR study to assess the associations of genetically predicted circulating vitamin C levels with risk of a wide range of CVDs.

Outcome data sources
We included 15 cardiovascular endpoints with numbers of cases ranging from 3373 (large artery stroke) to 139 364 (coronary artery disease). Summary-level data for these outcomes were obtained from large genetic consortia, 12-16 the UK Biobank study, 17 and the FinnGen consortium. 18 Detailed descriptions on data sources are presented in Table 1.

Instrument selection
Eleven lead single-nucleotide polymorphisms (SNPs) associated with plasma vitamin C levels at the genome-wide significance level (P < 5 Â 10 À8 ) were identified from a meta-analysis of genome-wide association studies (GWASs) including up to 52 018 individuals of European descent. 19 These SNPs explained approximately 1.87% of variance in circulating vitamin C levels. 19 Rs33972313 in the SLC23A1 gene region encodes the sodium-dependent vitamin C transporter 1 and explained most of the variance in circulating vitamin C levels. One additional effect allele of this variant is associated with an 11% higher plasma vitamin C level. 20 Rs174547 in FADS1 gene region was excluded from analyses due to pleiotropic effects on plasma phospholipid fatty acids, 21 leaving 10 SNPs leveraged as instrumental variables ( Table 2). Rs13028225 was not available in FinnGen data of SNP-CVD associations and was replaced by rs17655123 in high linkage disequilibrium (r 2 = 0.88) with rs13028225. Likewise, rs4074995 (r 2 = 0.83) and rs3809260 (r 2 = 0.98) were used as the proxy SNPs for rs10051765 and rs2559850, respectively, in the analysis of data from the International Stroke Genetics Consortium. Association estimates in the vitamin C GWAS were adjusted for age, sex, the first 10 genetic principal components, and study centre (where applicable).

Statistical analysis
The multiplicative random-effects inverse-variance-weighted method 22 was used as the main statistical method to assess the association between genetically predicted circulating vitamin C levels and CVDs. MR estimates for each CVD outcome from different data sources were combined using the fixed-effects meta-analysis method. We used two supplementary analyses, the weighed median approach 23 and MR-Egger regression, 24 to examine the robustness of the results and possible pleiotropy. The weighted median method can provide consistent causal estimates provided that > _50% of the weight comes from valid SNPs. 23 MR-Egger regression can detect horizontal pleiotropy by P-value for its intercept and generate estimate after correction for pleiotropy. 24 The I 2 statistic was calculated to assess the degree of heterogeneity 25 among estimates of SNPs in each analysis. All reported odds ratios (ORs) and corresponding 95% confidence intervals (CIs) of CVDs were scaled to 1 standard deviation (SD) increase in genetically predicted circulating levels of vitamin C. The Bonferroni method was used to adjust for multiple testing (15 CVDs). Associations with P-values of <0.003 were regarded as significant associations and associations with P-values between 0.05 and 0.003 were deemed as suggestive associations. All analyses were two-sided and performed using the mrrobust package 26 in Stata/SE 15.0.

Results
The associations of genetically predicted circulating vitamin C levels (per 1 SD increase) with the 15 CVDs in the main analysis are presented in Figure 1. We observed suggestive inverse associations of genetically predicted vitamin C levels with risk of any stroke in UK Biobank (OR, 0.84; 95% CI, 0.73, 0.97; P = 0.018) and cardioembolic stroke in MEGASTROKE (OR, 0.79; 95% CI, 0.64, 0.99; P = 0.038). However, the association for any stroke was not replicated in the MEGASTROKE and FinnGen consortia and did not persist in the meta-analysis. There was no association between genetically predicted vitamin C and atrial fibrillation in the GWAS meta-analysis by Nielsen et al. or in the FinnGen consortium, but the meta-analysis results revealed a suggestive positive association (OR 1.09, 95% CI, 1.00, 1.18; P = 0.049). Genetically predicted vitamin C levels were not associated with the other studied CVDs in the main analysis.
Results of the supplementary analyses based on the weighted median and MR-Egger methods showed no association of genetically predicted vitamin C levels with any CVD (Table 3), with the exception for a suggestive inverse association with peripheral artery disease in the FinnGen consortium (OR, 0.71, 95% CI, 0.52, 0.97; P = 0.030) and the meta-analysis (OR, 0.76, 95% CI, 0.62, 0.93; P = 0.009). We observed the modest heterogeneity in several analyses; however, the P-values for the intercept in corresponding MR-Egger regression were >0.05 ( Table 3). In a supplementary analysis using rs33972313 in the SLC23A1 gene region as instrumental variable, the associations of genetically predicted vitamin C levels with peripheral artery disease (OR, 0.67, 95% CI, 0.46, 1.00; P = 0.048) and other CVDs (data not shown) were consistent with the findings from the analyses based on all instrumental variables.

Discussion
This MR study found no clear pattern of associations between genetically predicted vitamin C levels and risk of CVDs ( Figure 2). However, genetically predicted vitamin C levels showed suggestive inverse associations with cardioembolic stroke and peripheral artery disease but a suggestive positive association with atrial fibrillation ( Figure 2).
The overall lack of support for a protective association between vitamin C and CVDs in the present MR study was consistent with most RCTs 3,10,11 and some 27,28 but not all 4-7 observational studies. A recent review article concluded that findings differed between RCTs   for vitamin C supplementation with null findings and observational studies on dietary vitamin C intake suggesting a protective association. 8 Several signalling pathways were highlighted using the network pharmacology approach, 9 whereas no evidence was found to support that vitamin C supplementation reduced the risk of CVD in healthy participants in a systematic review of RCTs. 10 This discrepancy may be caused by residual confounding by other cardio-protective nutrients, such as magnesium, 29,30 from foods rich in vitamin C (e.g. fruit and vegetables), or healthy lifestyle behaviours among individuals with a vitamin C-rich diet. 31 Even though the present MR study did not support cardiovascular benefits of increasing circulating vitamin C levels, our findings did not hint any information on possible health benefits from a diet abundant in vitamin C, suggested by previous evidence. 4,32,33 Instead, the present study did not justify vitamin C supplementation as a primary prevention strategy for CVD.
Higher plasma vitamin C levels were associated with a reduced risk of total stroke in cohort studies. 34 Nevertheless, a daily supplementation of 500 mg of vitamin C together with 400 IU of vitamin E did not decrease the risk of total stroke in an RCT involving 14 641 US male physicians followed-up of for a mean of 8 years. 35 Data on cardioembolic stroke are sparse. Likewise, few studies have investigated whether vitamin C status was associated with incident peripheral artery disease, although a clinical study revealed that acute vitamin C administration might restore peripheral endothelial function in patients with coronary artery disease. 36 Plasma vitamin C was inversely associated with risk of atrial fibrillation in middle-aged women with low baseline intake but not in men 37 and might prevent post-operative atrial fibrillation albeit with heterogeneous findings. 38 Our study, on the contrary, found a possible positive association of genetically predicted circulating vitamin C levels with atrial fibrillation, a finding that needs to be verified in other studies.
A previous MR study found no association between plasma vitamin C proxied by rs33972313 in the SLC23A1 gene region and ischaemic heart disease, 20 which is consistent with our findings. In addition, genetically proxied plasma vitamin C was not associated with certain cardiovascular risk factors, such as type 2 diabetes 19 and urate. 39 The major strength of the present study is the MR design, which diminished residual confounding and other biases, thereby strengthening the causal inference. In addition, we examined the association of genetically predicted vitamin C levels with CVDs using several data sources and the consistency of results the consistency of results supports the robustness of our findings. Along with the use of multiple independent SNPs as instrumental variables for plasma vitamin C, combining results for one outcome from different data sources increased the statistical power to detect weak associations even though we might have overlooked associations for certain outcomes with small number of cases. We confined the analyses to individuals of European ancestry, with the exception for the analysis for coronary artery disease based on consortium data where >80% of participants are individual of European descent. Thus, our findings were less likely to be influenced by population structure bias. Nonetheless, the population confinement limited the generalizability of our findings.
A potential limitation in MR studies is horizontal pleiotropy. We excluded an SNP (rs174547) with pleiotropic effects in the analysis to minimize bias from horizontal pleiotropy. Results were broadly consistent across sensitivity analyses and no evidence of horizontal pleiotropy was detected by MR-Egger regression. We also examined the association of plasma vitamin C with CVD using rs33972313 20 in the SLC23A1 gene, which encodes the sodium-dependent vitamin C transporter 1, as instrumental variable and found consistent results. No evidence of bias from horizontal pleiotropy was detected.    Another limitation is that we could not assess potential interaction effects of vitamin C with other antioxidants (e.g. vitamin E and bcarotene) on CVD. Our findings were based on generally healthy population and therefore cannot be generalized to populations with special features, such as individuals with vitamin C deficiency, patients with diabetes or kidney disease, and heavy smokers. There was small sample overlap in certain analyses. This overlap might have caused minor bias in the estimates towards the observational associations. Potential non-linear associations could not be examined in this MR study based on summary-level data.
In conclusion, this MR study suggests that elevating circulating vitamin C levels may not benefit the primary prevention of most CVDs. Whether increased vitamin C levels may decrease the risk of cardioembolic stroke and peripheral artery disease needs confirmation.

Acknowledgments
Genetic association estimates for CVDs were obtained from a GWAS meta-analysis of atrial fibrillation (Nielsen et al.), CARDIoGRAMplusC4D (Coronary ARtery DIsease Genome-wide Replication and Meta-analysis plus The Coronary Artery Disease Genetics), HERMES (Heart Failure Molecular Epidemiology for Therapeutic Targets) consortium, MEGASTROKE, ISGC (International Stroke Genetic Consortium), the UK Biobank study, and the FinnGen consortium. The authors thank all investigators for sharing these data. The MEGASTROKE project received funding from sources specified at http://www.megastroke.org/acknowledg ments.html. The author list of MEGASTROKE is listed in https:// www.megastroke.org/authors.html. Analyses of UK Biobank data were performed under application 29202.

Ethical approval
All studies included in cited genome-wide association studies had been approved by a relevant review board. The present MR analyses were approved by the Swedish Ethical Review Authority (2019-02793).

Data availability
All data analysed in this study are available OSF data respiratory (https://osf.io/6qd8f/).

Role of the funder
Funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; or the decision to submit the manuscript for publication.