98 Diabetes Care Volume 44, January 2021

Plasma and Type 2 Ju-Sheng Zheng,1,2 Jian’an Luan,1 Eleni Sofianopoulou,3 Fumiaki Imamura,1 Diabetes: Genome-Wide Isobel D. Stewart,1 Felix R. Day,1 Maik Pietzner,1 Eleanor Wheeler,1 Association Study and Mendelian Luca A. Lotta,1 Thomas E. Gundersen,4 Pilar Amiano,5 Eva Ardanaz,6,7,8 Randomization Analysis in Mar´ıa-Dolores Chirlaque,8,9 Guy Fagherazzi,10,11 Paul W. Franks,12 European Populations Rudolf Kaaks,13 Nasser Laouali,11 Francesca Romana Mancini,11 12 Diabetes Care 2021;44:98–106 | https://doi.org/10.2337/dc20-1328 Peter M. Nilsson, N. Charlotte Onland-Moret,14 Anja Olsen,15,16 Kim Overvad,16,17 Salvatore Panico,18 Domenico Palli,19 Fulvio Ricceri,20,21 Olov Rolandsson,22 Annemieke M.W. Spijkerman,23 Mar´ıa-JoseS´ anchez,´ 24 Matthias B. Schulze,25,26,27 Nuria´ Sala,28 Sabina Sieri,29 Anne Tjønneland,16,30 31 OBJECTIVE Rosario Tumino, 14 Higher plasma vitamin C levels are associated with lower type 2 diabetes risk, but Yvonne T. van der Schouw, Elisabete Weiderpass,32 Elio Riboli,33 whether this association is causal is uncertain. To investigate this, we studied the 3,34,35,36,37,38 association of genetically predicted plasma vitamin C with type 2 diabetes. John Danesh, Adam S. Butterworth,3,34,36,37,38 EPIDEMIOLOGY/HEALTH SERVICES RESEARCH RESEARCH DESIGN AND METHODS Stephen J. Sharp,1 Claudia Langenberg,1 1 1 We conducted genome-wide association studies of plasma vitamin C among 52,018 Nita G. Forouhi, and Nicholas J. Wareham individuals of European ancestry to discover novel genetic variants. We performed Mendelian randomization analyses to estimate the association of genetically predicted differences inplasma vitamin C with type 2diabetes in up to 80,983 caseparticipants and 842,909 noncase participants. We compared this estimate with the observa- tional association between plasma vitamin C and incident type 2 diabetes, including 1Medical Research Council Epidemiology Unit, 8,133 case participants and 11,073 noncase participants. University of Cambridge, Cambridge, U.K. 2Key Laboratory of Growth Regulation and RESULTS Translational Research of Zhejiang Province, 2 We identified 11 genomic regions associated with plasma vitamin C (P < 5 3 10 8), School of Life Sciences, Westlake University, SLC23A1 SLC23A3 Hangzhou, China with the strongest signal at , and 10 novel genetic loci including , 3 CHPT1 BCAS3 SNRPF RER1 MAF GSTA5 RGS14 AKT1 FADS1 British Heart Foundation Cardiovascular Epide- , , , , , , , , and . Plasma vitamin miology Unit, Department of Public Health and C was inversely associated with type 2 diabetes (hazard ratio per SD 0.88; 95% CI Primary Care, University of Cambridge, Cam- 0.82, 0.94), but there was no association between genetically predicted plasma bridge, U.K. 4 vitamin C (excluding FADS1 variant due to its apparent pleiotropic effect) and type 2 Vitas Ltd, Oslo, Norway 5Ministry of Health of the Basque Government, diabetes (1.03; 95% CI 0.96, 1.10). Public Health Division of Gipuzkoa, Biodonostia Health Research Institute, Donostia-San Sebas- CONCLUSIONS tian, Spain These findings indicate discordance between biochemically measured and genet- 6Navarra Public Health Institute, Pamplona, Spain ically predicted plasma vitamin C levels in the association with type 2 diabetes 7 fi IdiSNA, Navarra Institute for Health Research, among European populations. The null Mendelian randomization ndings provide Pamplona, Spain no strong evidence to suggest the use of vitamin C supplementation for type 2 8CIBER in Epidemiology and Public Health (CI- diabetes prevention. BERESP), Madrid, Spain 9Department of Epidemiology, Regional Health Council, Instituto Murciano de Investigation´ Bio- Observational evidence suggests that plasma vitamin C (ascorbic acid) is inversely sanitaria (IMIB)-Arrixaca, Murcia University, Mur- associated with the incidence of type 2 diabetes (1). However, evidence from cia, Spain 10 randomized controlled trials (RCTs) of supplementation with vitamin C suggests no Digital Epidemiology and e-Health Research Hub, Department of Population Health, Luxem- effect of supplementation on glycemia in individuals without diabetes (2) or on type 2 bourg Institute of Health, Strassen, Luxembourg, diabetes incidence among women at high risk of cardiovascular disease (CVD) (3). France care.diabetesjournals.org Zheng and Associates 99

Important limitations of the RCT evidence (rs33972313) at SLC23A1 has been iden- committees, and participants provided have included small sample size of some tified to be associated with plasma vita- written informed consent. individual trials, uncertainty of the optimal min C levels (6), based on a candidate We performed meta-analysis of GWAS dose,baselinestatus,theinabilitytoseparate approach, and no GWAS has been pub- of plasma vitamin C levels in up to 52,018 out the effects of vitamin C when these were lished so far. individuals from the Fenland study (n 5 used in combination with other vitamins/ The objective of this study was to 10,771) (2), European Prospective Inves- antioxidants, and declining compliance with study the relationship between genetic tigation into Cancer and Nutrition (EPIC)- the intervention over time. Conducting fur- variants and vitamin C levels and to de- InterAct study (n 5 16,841) (7), EPIC ther RCTs to assess an impact of vitamin C on velop a genetic instrument to be used in Norfolk study (8) (n 5 16,756, excluding type 2 diabetes is challenging, requiring large an MR analysis of the association of duplicated samples with EPIC-InterAct), sample size, long follow-up duration, and genetically predicted levels of plasma vi- and the EPIC-CVD study (9) (n 5 7,650, high cost. Meanwhile, it remains unclear tamin C with type 2 diabetes risk. excluding duplicated samples with whether there is a causal role of vitamin C EPIC-InterAct or EPIC-Norfolk). in the prevention of type 2 diabetes. RESEARCH DESIGN AND METHODS We then estimated the association of Mendelian randomization (MR) is an Participating Studies and Genotyping lead genetic variants from the vitamin approach that can help to assess the Figure 1 provides an overview of the C GWAS meta-analysis with the risk of potential causal relevance of a risk ex- participating cohorts, and our overall type 2 diabetes in a meta-analysis in- posure for a disease end point, taking approach to data inclusion and analysis. cluding participants (80,983 case par- advantage of the random assortment of The three steps in the process were as ticipants with type 2 diabetes, and allelesatconception(4).Thismeansthat follows: 1) we conducted GWAS to iden- 842,909 noncase participants) from the association between a disease and tify genetic variants associated with UK Biobank (10), DIAbetes Meta-ANalysis a genetically determined phenotype is plasma vitamin C levels; 2)usingthe of Trans-Ethnic association studies (DIA- unlikely to be affected by reverse cau- GWAS-discovered genetic variants, we MANTE) (European) (11), and EPIC-Norfolk sation or by confounding. A limitation of examined the association between ge- study (additional case participants with MR is weak instrument bias if the ge- netically predicted plasma vitamin C and type 2 diabetes not included in the netic variants explain a small fraction of type 2 diabetes risk by MR analysis; and DIAMANTE study: 1,220 case participants the variation in the risk factor of in- 3) for comparison with MR results, we with diabetes and 18,026 noncase partic- terest, but this could be partially ad- performed analyses to evaluate the ob- ipants) (8). dressed through the use of multiple servational association between plasma The details of each participating study independent genetic variants identified vitaminCandincidenttype2diabetes. All and genotyping strategy and quality con- in genome-wide association studies (GWAS) of the studies included in the present trol procedures are described in Sup- (5). Thus far, only one genetic variant analyses were approved by local ethics plementary Table 1 and Supplementary

11Center of Epidemiology and Population Health 24Andalusian School of Public Health, Granada, and Genomics, University of Cambridge, Cam- UMR 1018, INSERM, Paris South - Paris Saclay Spain bridge, U.K. University, Gustave Roussy Institute, Villejuif, 25Department of Molecular Epidemiology, Ger- 35British Heart Foundation Center of Research France man Institute of Human Nutrition Potsdam- Excellence, University of Cambridge, Cambridge, 12Department of Clinical Sciences, Lund Univer- Rehbruecke, Nuthetal, Germany U.K. sity, Malmo,¨ Sweden 26German Center for Diabetes Research (DZD), 36Department of Human Genetics, Wellcome 13Division of Cancer Epidemiology, German Can- Munchen-Neuherberg,¨ Germany Sanger Institute, Hinxton, U.K. cer Research Center (DKFZ), Heidelberg, Germany 27Institute of Nutrition Science, University of 37National Institute for Health Research Cam- 14Julius Center for Health Sciences and Primary Potsdam, Nuthetal, Germany bridge Biomedical Research Center, University of Care, University Medical Center Utrecht, Utrecht 28Unit of Nutrition and Cancer, Cancer Epidemi- Cambridge and Cambridge University Hospitals, University, Utrecht, the Netherlands ology Research Program and Translational Cambridge, U.K. 15Danish Cancer Society Research Center, Copen- Research Laboratory; Catalan Institute of Oncology 38Health Data Research UK Cambridge, Well- hagen, Denmark - ICO, Group of Research on Nutrition and Cancer, come Genome Campus and University of Cam- 16Department of Public Health, Aarhus Univer- Bellvitge Biomedical Research Institute (IDIBELL), bridge, Cambridge, U.K. sity, Aarhus, Denmark L’Hospitalet of Llobregat, Barcelona, Spain 17 29 Corresponding authors: Nita G. Forouhi, nita. Department of Cardiology, Aalborg University Epidemiology and Prevention Unit, Fondazione [email protected], and Nicholas J. fi Hospital, Aarhus, Denmark Istituto di Ricovero e Cura a Carattere Scienti co Wareham, [email protected] 18Dipartimento di Medicina Clinica e Chirurgia, (IRCCS) Istituto Nazionale dei Tumori di Milano Federico II University, Naples, Italy Via Venezian, Milan, Italy Received 2 June 2020 and accepted 15 October 19Cancer Risk Factors and Life-Style Epidemiology 30Institute of Public Health, University of Copen- 2020 Unit, Institute for Cancer Research, Prevention hagen, Copenhagen, Denmark This article contains supplementary material online and Clinical Network - ISPRO, Florence, Italy 31Cancer Registry and Histopathology Depart- at https://doi.org/10.2337/figshare.13105559. 20 Department of Clinical and Biological Sciences, ment, Azienda Sanitaria Provinciale (ASP), Ra- C.L., N.G.F., and N.J.W. contributed equally. University of Turin, Orbassano, Turin, Italy gusa, Italy 21 32 © 2020 by the American Diabetes Association. Unit of Epidemiology, Regional Health Service International Agency for Research on Cancer, Readers may use this article as long as the work is ASL TO3, Grugliasco, Turin, Italy Lyon, France 22 33 properly cited, the use is educational and not for Department of Public Health and Clinical Med- School of Public Health, Imperial College, Lon- fi icine, Family Medicine, Umea˚ University, Umea,˚ don, U.K. pro t, and the work is not altered. More infor- Sweden 34National Institute for Health Research Blood mation is available at https://www.diabetesjournals 23National Institute for Public Health and the and Transplant Research Unit in Donor Health .org/content/license. Environment (RIVM), Bilthoven, the Netherlands 100 Plasma Vitamin C and Type 2 Diabetes Diabetes Care Volume 44, January 2021

residuals (in SD units) for plasma vitamin C were calculated, adjusting for age, sex, and study center (where appropriate). Moreover, in each cohort, GWAS was performed by running linear regression with SNPTEST (v2.5.4) assuming an ad- ditive effect, adjusting for the first 10 ge- netic principal components of ancestry within each cohort. Meta-analysis of the GWAS results was conducted by combining b coefficients and SEs using inverse variance-weighted fixed-effect meta-analysis across the participating studies using METAL (13). Associated loci were identified using the conventional threshold for genome-wide statistical significance (P , 5 3 1028). At each locus, a lead single-nucleotide polymor- phism (SNP) was identified as the SNP Figure 1—Design of the study for MR analysis of plasma vitamin C with type 2 diabetes. with the lowest P value within a 1 mega- base-pair window. Functional annotation Text. Briefly, the Fenland study is an CVD was performed to the Haplotype andpathwayanalysisoftheGWASresults ongoing, population-based cohort study Reference Consortium reference panel were performed using MAGENTA and including 12,435 adults at baseline in using IMPUTE v4 software. DEPICT (14,15) (Supplementary Text). Cambridgeshire, U.K. (2). The GWAS in- The UK Biobank study is a population- Linear regression models were used to cluded 10,771 Fenland participants with based cohort of ;0.5 million U.K. indi- estimate the variance in plasma vitamin C both genotype and vitamin C data. Ge- viduals aged 40–69 years recruited explained by the identified lead SNPs in notype imputation was performed to between 2006 and 2010 across the U.K. each of Fenland, EPIC-InterAct, EPIC- the Haplotype Reference Consortium (10). Genotype data and prevalent type 2 Norfolk, and EPIC-CVD studies. We also reference panel using IMPUTE v4 or the diabetes information were both available calculated the F statistic in the EPIC- Sanger imputation server. among a total of 449,333 individuals in Norfolk study to evaluate the strength EPIC-InterAct is a case-cohort study the UK Biobank data set (24,758 case of the instrument. We estimated the including 12,403 participants with inci- participants and 424,575 noncase par- genetic correlations of vitamin C levels dent type 2 diabetes from among ticipants). DIAMANTE is a consortium that with type 2 diabetes and related glyce- 340,234 participants across eight Euro- published a large-scale meta-analysis of mic markers(fasting glucose,insulin, 2-h pean countries (France, Italy, Spain, U.K., GWAS of type 2 diabetes in individuals of glucose, HOMA of insulin resistance Netherlands, Germany, Sweden, and Den- different ethnicities. Summary results from [HOMA-IR], HOMA of b-cell function mark) and 16,835 subcohort participants the DIAMANTE meta-analyses among Euro- [HOMA-b], and hemoglobin A1c [HbA1c]) randomly selected from the full cohort, of peans were made publicly available (11). using the meta-analyzed GWAS results whom 16,154 remained after exclusion of and linkage disequilibrium score regres- participants with prevalent diabetes (7); Plasma Vitamin C Measurement sion on the LD Hub platform (16), and 16,841 participants with both genotype Plasma vitamin C was measured using the results were corrected for multiple and vitamin C data contributed to the high-performance liquid chromatography testing. present GWAS. EPIC-Norfolk is an ongoing with ultraviolet detection in EPIC-InterAct U.K.-based prospective cohort study with and EPIC-CVD using samples that had Association of Genetically Predicted Levels 25,639menandwomenaged40–79 at been stored at 2196°C (2150°C for Dan- of Vitamin C With Type 2 Diabetes: MR baseline (8), among whom 16,756 par- ish samples) (12). In EPIC-Norfolk and Analysis ticipants had the required data to be Fenland, plasma vitamin C was measured We used two-sample MR analysis to test included in the vitamin C GWAS. EPIC- with a fluorometric assay using samples for an association between a genetically CVD is also a case-cohort study nested storedat 270°C.Foreachofthesestudies, predicted difference in plasma vitamin C within the EPIC study, with a random plasma was stabilized in a standardized levels and type 2 diabetes risk. To this subcohort of 18,249 participants and volume of metaphosphoric acid. The dis- end, we used estimates of each of the 24,557 adults who later developed CVD tributionofplasmavitaminClevelsineach “SNP to vitamin C” and “SNP to diabetes” during the follow-up. EPIC-CVD shared participating cohort is shown in Supple- associations to examine the association the random subcohort with the EPIC- mentary Fig. 1. between a genetically predicted 1-SD InterAct for eight participating countries difference in plasma vitamin C and type 2 (9), and a total of 7,650 EPIC-CVD par- Statistical Analysis diabetes risk. Estimates from multiple ticipants contributed to the present vi- Meta-analysis of GWAS SNPs were pooled using an inverse tamin C GWAS. Genotype imputation of For the GWAS of plasma vitamin C among variance–weighted method (5). MR as- the EPIC-InterAct, EPIC-Norfolk, and EPIC- each participating cohort, standardized sumptions are violated if there is horizontal care.diabetesjournals.org Zheng and Associates 101

pleiotropy.Therefore,inthe EPIC-InterAct We calculated the estimates of a ge- RESULTS study, we constructed a vitamin C-raising netically predicted 1-SD difference in Population Characteristics genetic score based on the identified plasma vitamin C with several glycemic Plasma vitamin C was associated with a genetic variants and examined its asso- traits using data from Meta-Analysis of variety of anthropometric, lifestyle be- ciation with anthropometric, lifestyle, Glucose and Insulin-related traits Con- havioral factors (education, employment dietary factors, and lipid biomarkers sortium (MAGIC): fasting glucose (n 5 status, physical activity, and smoking to examine the potential pleiotropic 133,010), 2-h glucose (n 5 42,854), fast- status), dietary factors, and lipid bio- effect of the genetic score. The poten- ing insulin (n 5 108,557), HOMA-IR (n 5 markers (Supplementary Table 2). In con- b 5 tial pleiotropic effect of the identified 46,186),HOMA- (n 46,186),andHbA1c trast, the vitamin C-raising genetic score SNPs was further examined by investigat- (n 5 123,665). calculated from the vitamin C-raising al- ing their associations with 174 blood me- Observational Association of Plasma leles (generated from the GWAS results) tabolites using GWAS summary statistics Vitamin C With Type 2 Diabetes waspositivelyassociatedonlywithplasma from genome-wide meta-analysis of these For the observational association of vitamin C levels, and it was not associated metabolites across Fenland, EPIC-Norfolk, plasma vitamin C and incident type 2 withotherfactorsincludinganthropomet- and Efficiency and Safety of Varying the diabetes in the EPIC-InterAct study, we ric, lifestyle, diet factors, or lipid biomarkers Frequency of Whole Blood Donation (IN- used Prentice-weighted Cox regression, (Supplementary Table 3). TERVAL) studies, as previously reported accounting for the design of a case- (17). cohort study, to estimate the country- Meta-analysis of GWAS of Plasma In addition, directional pleiotropy for specific hazard ratios and 95% CIs for Vitamin C Genome-wide meta-analysis identified 11 theused SNPs wastested using MR-Egger associations per 1-SD difference (SD cal- independent genomicloci associatedwith intercept (18). We used MR-Egger re- culated in the subcohort) of plasma plasma vitamin C levels (Table 1), of which gression to detect and adjust for poten- vitamin C with incident type 2 diabetes. 10 were novel, while SLC23A1 (6) was a tial unbalanced pleiotropy in the MR The estimates were adjusted for potential known plasma vitamin C locus (Table 1 analysis (18), and the Cochran Q statistic confounding factors of age as the under- and Supplementary Figs. 2 and 3). The to examine the heterogeneity of the as- lying time scale, sex, center, physical ac- tivity, smoking status, employment status, regional plots of the identified loci are sociation between different genetic var- presented in Supplementary Figs. 4–12. iants, because different genetic variants marital status, education level, alcohol in- take, total energy intake, individual plasma The heterogeneity for the association of may represent different biological path- carotenoids, BMI, and waist circumfer- the identified lead SNPs with plasma ways. We also used a weighted median ence. The results across countries were vitamin C across individual cohorts was MR method (19) as a sensitivity analysis pooled by random-effects meta-analysis. low (Supplementary Fig. 13). The stron- and highlighted the weighted median re- Statistical analyses were performed in gest signal was rs33972313 within the sults if significant heterogeneity of the 290 Stata 14 (StataCorp, College Station, TX), SLC23A1 gene (P 5 4.61 3 10 )on associations among different genetic var- R 3.2.2 (R Foundation for Statistical Com- 5, a missense mutation that iants was observed. We conducted a re- puting), and METAL 2011-03-25. wasreportedtobeassociatedwithcircu- verse MR analysis to examine the potential lating vitamin C levels in a previous can- association of genetically predicted risk Data and Resource Availability didate gene study (6). SLC23A1 encodes of type 2 diabetes on plasma vitamin C GWAS summary statistics for the plasma 23 member 1, also concentrations, using a score composed vitamin C levels can be accessed at known as sodium-dependent vitamin C of 231 type 2 diabetes SNPs identified in a figshare.com (https://doi.org/10.6084/ transporter1(SVCT1),which is responsible recent GWAS (11). m9.figshare.13227443.v1). for the uptake of vitamin C into target

Table 1—Lead SNPs identified in the genome-wide meta-analysis of plasma vitamin C across four cohort studies (Fenland, EPIC-Norfolk, EPIC-InterAct, and EPIC-CVD)* Chromosome Position (GRCh37) Lead SNPs Effect allele Other allele EAF b SE P value Nearest gene 1 2330190 rs6693447 T G 0.551 0.039 0.006 6.25 3 10210 RER1 2 220031255 rs13028225 T C 0.857 0.102 0.009 2.38 3 10230 SLC23A3 5 138715502 rs33972313 C T 0.968 0.360 0.018 4.61 3 10290 SLC23A1 5 176799992 rs10051765 C T 0.342 0.039 0.007 3.64 3 1029 RGS14 6 52725787 rs7740812 G A 0.594 0.038 0.006 1.88 3 1029 GSTA5 11 61570783 rs174547 C T 0.328 0.036 0.007 3.84 3 1028 FADS1 12 96249111 rs117885456 A G 0.087 0.078 0.012 1.70 3 10211 SNRPF 12 102093459 rs2559850 A G 0.598 0.058 0.006 6.30 3 10220 CHPT1 14 105253581 rs10136000 A G 0.283 0.040 0.007 1.33 3 1028 AKT1 16 79740541 rs56738967 C G 0.321 0.041 0.007 7.62 3 10210 MAF 17 59456589 rs9895661 T C 0.817 0.063 0.008 1.05 3 10214 BCAS3 EAF, effect allele frequency. *The b coefficients are in SD unit per allele; effect allele is the vitamin C-raising allele. The pooled sample size for the genome-wide meta-analyses is 52,018. 102 Plasma Vitamin C and Type 2 Diabetes Diabetes Care Volume 44, January 2021

Figure 2—Volcano plot for the association of the plasma vitamin C genetic score with blood metabolites. Associations of the 11 variant scores and the score excluding the FADS1 variant were estimated using fixed-effects meta-analyses of individual variant-metabolite associations aligned to the plasma vitamin C-raising alleles. The horizontal dashed line indicates a Bonferroni-corrected significance threshold (corrected for 175 tests), and diameters of points are proportional to abs(effect size)*50, where “abs” is the function to calculate the absolute value of the effect size.

tissues (20). The second strongest signal 3 10214, encoding microtubule associ- pleiotropic effect of the variant at FADS1, rs13028225 (P 5 2.38 3 10230)waswithin ated cell migration factor). which was associated with a large number the SLC23A3 gene on chromosome 2, which Results of pathway analysis and tissue of glycerophospholipids or sphingolipids encodes sodium-dependent vitamin C and gene set enrichment analysis did not (Fig. 2). Thus, we excluded this SNP from transporter 3 (SVCT3), belonging to the reveal a specificpathwayorenriched our MR analysis, using the other 10 SNPs same family as SLC23A3. tissues/gene sets (Supplementary Tables in a genetic instrument to examine the We identified further genomic loci with 4 and 5). The variance of plasma vitamin C causal association of plasma vitamin C with less obvious linkage to the metabolism explained by the 11 lead SNPs was 1.87% type 2 diabetes (Supplementary Table 7). of vitamin C, such as RER1 (rs6693447, P 5 on average, with 2.51% in Fenland, 1.74% Genetic predisposition to a higher 6.25 3 10210, encoding retention in in EPIC-Norfolk, 1.61% in the InterAct level of plasma vitamin C was not asso- endoplasmic reticulum sorting receptor subcohort, 1.63% in the InterAct nonsub- ciated with odds of type 2 diabetes, with 1), RGS14 (rs10051765, P 5 3.64 3 1029, cohort, 1.07% in the EPIC-CVD subcohort, an estimated odds ratio of 1.03 (95% CI encoding regulator of G signaling and 0.7% in the EPIC-CVD nonsubcohort. 0.96, 1.10) per 1-SD difference in plasma 14), GSTA5 (rs7740812, P 5 1.88 3 1029, The Fstatistic of the vitamin C genetic score vitamin C level (Fig. 3). There was no encoding glutathione S-transferase a 5), for MR analysis was 30.5. The estimates of evidence of directional horizontal pleiot- FADS1 (rs174547, P 5 3.84 3 1028, genome-wide genetic correlations of vita- ropy (P 5 0.41 for the test of MR-Egger encoding fatty acid desaturase 1), SNRPF min C levels with type 2 diabetes and intercept) and heterogeneity for the ge- (rs117885456, P 5 1.7 3 10211, encod- glycemic traits were significant only for netic instrument used (Supplementary ing small nuclear ribonucleoprotein poly- fasting insulin (rgenetics 520.22, P50.005) Table 8), which confirmed the validity of peptide F), CHPT1 (rs2559850, P 5 6.3 3 after controlling for multiple testing (Sup- our genetic instrument. Sensitivity anal- 10220, encoding choline phosphotrans- plementary Table 6). ysis using different methods (MR-Egger, ferase 1), AKT1 (rs10136000, P 5 1.33 3 weighted median) of MR did not find a 28 10 , encoding serine-threonine protein Association of Plasma Vitamin C- substantial difference compared with the kinase), MAF (rs56738967, P 5 7.62 3 Raising Alleles With Type 2 Diabetes inverse variance–weighted method. The 10210, encoding MAF bZIP transcription From among 11 SNPs that were signifi- reverse MR analysis suggested that ge- factor), and BCAS3 (rs9895661, P 5 1.05 cant at the genome-wide level, there was a netically higher risk of type 2 diabetes care.diabetesjournals.org Zheng and Associates 103

significant lead SNPs at corresponding loci in a genetic instrument suggested that a genetically higher level of plasma vitamin C was not significantly associated with type 2 diabetes risk or with glyce- mic traits.

Findings in Context of Other Evidence: MR Findings There is no prior published MR analysis of plasma vitamin C and type 2 diabetes or glycemic traits to compare with our current findings. However, our finding of no evidence for an association of genetic markers of plasma vitamin C with type 2 diabetes or with any of the glycemic traits is consistent with prior null findings from RCTs of vitamin C supplementation (2,3). An obvious explanation for the discrep- ancy between the observational and the MR or RCT evidence is likely to be confounding by several factors, including socioeconomic, dietary, and lifestyle be- havioral factors, as previously discussed (21), and as we found in the current analyses.

Findings in Context of Other Evidence: GWAS Findings Our novel GWAS findings stimulate greater understanding of the biology of plasma vitamin C. Prior evidence suggests that vitamin C homeostasis is mainly regu- lated by two categories of . Some — Figure 3 Genetically predicted and observational associations of plasma vitamin C with type 2 genes are involved in direct transport and diabetes. A: MR and observational estimate of plasma vitamin C with type 2 diabetes. MR estimate of the associationbetweena geneticallypredicteddifference of 1 SD in plasma vitaminC and type 2 regulation of vitamin C concentrations, diabetes (odds ratio [OR]) was determined using different methods: inverse-variance weighted, including vitamin C absorption from food, MR Egger, and weighted median. Ten lead genetic variants (excluding FADS1 variant due to renal reabsorption, and cellular uptake, pleiotropic effects) identified in the present genome-wide meta-analysis were used for the MR particularly the sodium-dependent vita- analysis. Observational estimate of the association of plasma vitamin C with type 2 diabetes (hazard ratio [HR]) was estimated in EPIC-InterAct using Prentice-weighted Cox regression, min C transporters of the SLC23 family adjusting for potential confounders. B: Scatter plot of genetic association with type 2 diabetes and (e.g., SLC23A1 and SLC23A2), whereas genetic association with the plasma vitamin C levels. others are related to antioxidant and oxidative stress (22). SLC23A1 encodes a well-known protein SVCT1 affecting the was not significantly associated with plasma HOMA-IR, HOMA-b, and HbA (Supple- 1c circulating vitamin C levels (6,22), and vitamin C levels (Supplementary Fig. 14). mentary Fig. 15). This is consistent with the lead SNP rs33972313 identified in In the observational analysis, a 1-SD dif- the lack of association seen for type 2 the current study also showed consis- ference in plasma vitamin C level was diabetes. The associations of individual tent, reproducible evidence in a pre- associated with a lower hazard of type 2 vitamin C-raising alleles with these traits vious study (6). diabetes (hazard ratio 0.88; 95% CI 0.82, are presented in Supplementary Fig. 16. Unlike SLC23A1, we did not identify 0.94) (Fig. 3). SNPs of genome-wide significance at CONCLUSIONS SLC23A2, although both genes encode Association of Vitamin C-Raising This study represents the first GWAS to SVCTs, which are specificforvitaminC Genetic Risk Score With Glycemic be conducted for plasma vitamin C, iden- transportation. ItmaybethatSVCT2 Traits tifying 11 genomic regions associated encoded by SLC23A2 mainly regulates In the MR analysis, a genetically pre- with plasma vitamin C, including 10 that tissue levels of vitamin C and that its dicted higher plasma vitamin C level was were novel loci and confirming the pre- impact on circulating vitamin C is min- not significantly associated with any of viously identified candidate locus at SLC23A1 imal, as indicated in a recent review (22). the examined glycemic markers, includ- (rs33972313). The results from using Notably, we identified GWAS hits at ing fasting insulin, glucose, 2-h glucose, these newly identified genome-wide SLC23A3 associated with plasma vitamin 104 Plasma Vitamin C and Type 2 Diabetes Diabetes Care Volume 44, January 2021

C levels. SLC23A3 encodes SVCT3, which with a variety of diabetes-related traits, as a nutrient per se, but this does not rule is an orphan transporter, and its functional which also confirmed the observed null out the utility of plasma vitamin C as a valid role is largely unknown (23). There is results for type 2 diabetes. biomarker of fruit and vegetable intake, evidence showing that SVCT3 is mainly Limitations of our work include that reducing measurement error and recall expressed in the kidney (23) and that despite the large GWAS meta-analysis bias of subjective reporting. It is likely the SLC23A3 mRNA expression level is we conducted, the variation of vitamin C that a synergetic effect of various constit- highly correlated with the estimated glo- explained by the SNPs used in the MR uents in fruits and vegetables (including merular filtration rate, a marker of renal was still relatively small. The low vari- vitamins, minerals, numerous phytochem- function (24). Therefore, genetic variation ability of plasma vitamin C explained by icals, and dietary fiber) rather than vitamin at SLC23A3 may affect circulating vitamin theseSNPsmaylimitthestatisticalpower C alone may be responsible for observed C levels by regulating renal function and precision for the MR analysis. Our inverse association of fruit and vegetable andrelated vitamin Creabsorptionby the analyses included White European pop- consumption on type 2 diabetes risk. kidney. In addition, we identified several ulations; hence, our findings may not be In summary, using a genome-wide other genomic regions that might affect generalizable to other populations. It is meta-analysis approach, we identified 10 vitamin C levels potentially through reg- possible that pleiotropic effects of ge- novel genetic loci and confirmed a known ulating renal function and renal reab- netic variants used in MR may potentially locus at SLC23A1 for plasma vitamin C sorption of vitamin C, including RGS14, exist, although we did not find evidence levels. With these newly identified ge- MAF, and BCAS3 (25). of pleiotropic effects (i.e., none of the netic variants as a genetic instrument, Other genetic loci identified in the SNPs used in the MR were associated our MR analysis found that there was no present GWAS may be related to anti- with tested confounders) or heteroge- evidence to support a causal association oxidant and oxidative stress. The GSTA5 neity across the genetic variants. There- between plasma vitamin C and type 2 gene encodes glutathione S-transferase fore, confounding may not be a substantial diabetes or with related metabolic traits (GST)-a5, a detoxifying enzyme belong- concern, and our instrument is specific among European populations. These ing to the GST families, which contributes for vitamin C, which fulfils an important findings suggest that previous observa- to the glutathione–vitamin C antioxidant criterion to identify causality. In addi- tions on higher vitamin C levels and lower cycle (22). Genetic variants at FADS1 tion, the MR estimate may be biased by type 2 diabetes risk may be attributed were associated with lipid peroxides, as gene-environment interaction or cana- to a dietary pattern high in fruit and well as other oxidative stress markers lization, which should be investigated in vegetable intake rather than to vitamin C (26), and might influence vitamin C levels future research. level as an isolated nutrient. The impli- through the oxidative stress-related path- cation of the current study is that vitamin way. The RER1 may be related to endo- Implications of the Current Findings C supplementation aiming to increase plasmic reticulum stress, which is closely Clarification of the potential role of vi- circulating vitamin C levels is unlikely to related to oxidative stress (27). Genetic tamin C in type 2 diabetes development be helpful for the prevention of type 2 variation at CHPT1 may lead to disruption is important as the prevalence of type diabetes among individuals of European of phosphatidylcholine synthesis, which is 2 diabetes continues to expand globally ancestries. closely related to the antioxidant system and identification of modifiable risk fac- (28). Finally, we identifiedgeneticvariants tors is critical to curb the type 2 diabetes at SNRPF gene associated with plasma pandemic. Despite interest in vitamin fi Acknowledgments. The authors thank all EPIC vitamin C, but a potentially functional link C supplement use, the current ndings participants and staff fortheir contribution to the between this gene and vitamin C metab- indicate that there is unlikely to be a study. The authors thank Nicola Kerrison (MRC olism is unclear. causal role of vitamin C as an isolated Epidemiology Unit, Cambridge) for managing the micronutrient on type 2 diabetes and data for the InterAct Project. The authors thank the technical and functional operational teams of Strengths and Limitations its related traits in general populations. the Medical Research Council Epidemiology Unit This study has several strengths. We This should be distinguished from the and laboratory team at VITAS AS, Norway, for the conducted the first GWAS for plasma use of plasma vitamin C as a biomarker measurements of plasma vitamin C. The authors vitamin C using a large sample from mul- to objectively estimate or be a proxy thank staff fromthe EPIC-CVD and EPIC-InterAct tiple cohorts and identified 10 novel indicator for the consumption of fruit Coordinating Centers for performing sample genetic loci for plasma vitamin C in ad- and vegetables (29). The latter provides preparation and data-handling work, particularly fi Sarah Spackman (EPIC-CVD Data Manager), and dition to con rming the single variant support for the previous interpretation Cambridge Genomic Services for genotyping. described previously from candidate that the inverse observational prospec- Funding. The InterAct project was funded by the gene analysis (6). The current work ex- tive associations between plasma vita- European Union Sixth Framework Programme pands our understanding about vitamin C min C and cardiometabolic diseases, (grant number LSHM_CT_2006_037197). In ad- biology, and it enabled the derivation of a including type 2 diabetes (1,30–32), re- dition, InterAct investigators acknowledge fund- ing from the following agencies: Medical genetic instrument to use in MR analysis. flect the potential benefits of promoting Research Council Epidemiology Unit MC_UU_ Our study is the first large MR analysis for higher dietary consumption of fruit and 12015/1 and MC_UU_12015/5 (J.-S.Z., F.I., N.G.F., circulating vitamin C and type 2 diabetes, vegetables. It is important to clarify the N.J.W.), and National Institute for Health Re- with a large number of case participants two distinct phenomena: the current search Biomedical Research Center Cambridge: – fi Nutrition, Diet, and Lifestyle Research Theme in the genotype type 2 diabetes associ- ndings suggest no evidence for an as- (IS-BRC-1215-20014; N.G.F., N.J.W.). The coordina- ation. We examined the association of sociation with type 2 diabetes of genet- tion of EPIC is financially supported by the European genetically predicted vitamin C levels ically predicted plasma vitamin C level Commission (DG-SANCO) and the International care.diabetesjournals.org Zheng and Associates 105

Agency for Research on Cancer. The Spanish A.O., K.O., S.P., D.P., F.R., O.R., A.M.W.S., M.-J.S., column and UV-detection. J Chromatogr B Analyt National cohort is supported by Health Research M.B.S., N.S., S.S., A.T., R.T., Y.T.v.d.S., E.W., E.R., Technol Biomed Life Sci 2005;824:132–138 Fund (FIS) - Instituto de Salud Carlos III (ISCIII), and J.D. contributed to interpretation of data and 13. Willer CJ, Li Y, Abecasis GR. METAL: fast and Regional Governments of Andaluc´ıa, Asturias, revised the article critically for important in- efficient meta-analysis of genomewide associa- Basque Country, Murcia, and Navarra, and the tellectual content. All authors approved the final tion scans. Bioinformatics 2010;26:2190–2191 Catalan Institute of Oncology – ICO (Spain). version of the manuscript. S.J.S., C.L., N.G.F., and 14. Pers TH, Karjalainen JM, Chan Y, et al.; Biomarker measurements for vitamin C were N.J.W. coordinated the InterAct project with Genetic Investigation of ANthropometric Traits funded by the Medical Research Council Cam- N.J.W. as chief investigator. J.-S.Z. and N.G.F. are (GIANT) Consortium. Biological interpretation of bridge Initiative (RG71466, SJAH/004) and the guarantors of the work and, as such, had full genome-wide association studies using pre- EPIC-CVD project, which is supported by the access to all the data in the study and take dicted gene functions. Nat Commun 2015;6:5890 European Union Framework 7 (HEALTH-F2-2012- responsibility for the integrity of the data and the 15. Segre` AV, Groop L, Mootha VK, Daly MJ, 279233), European Research Council (268834), accuracy of the data analysis. Altshuler D; DIAGRAM Consortium; MAGIC in- UK Medical Research Council (G0800270 and vestigators. Common inherited variation in mi- MR/L003120/1), British Heart Foundation (SP/ References tochondrialgenesisnotenrichedforassociations 09/002, RG/08/014, and RG13/13/30194) and 1. Harding AH, Wareham NJ, Bingham SA, et al. with type 2 diabetes or related glycemic traits. the UK National Institute for Health Research Plasma vitamin C level, fruit and vegetable con- PLoS Genet 2010;6:e1001058 (Cambridge Biomedical Research Center at the sumption, and the risk of new-onset type 2 16. Zheng J,Erzurumluoglu AM, Elsworth BL, et al.; Cambridge University Hospitals National Health diabetes mellitus: the European prospective in- Early Genetics and Lifecourse Epidemiology (EA- Service Foundation Trust). This work was sup- vestigation of cancer–Norfolk prospective study. GLE) Eczema Consortium. LD Hub: a centralized ported by Health Data Research UK, which is Arch Intern Med 2008;168:1493–1499 database and web interface to perform LD score funded by the UK Medical Research Council, 2. Ashor AW, Werner AD, Lara J, Willis ND, Mathers regression that maximizes the potential of Engineering and Physical Sciences Research JC, Siervo M. Effects of vitamin C supplementation summary level GWAS data for SNP heritability Council, Economic and Social Research Council, onglycaemiccontrol:asystematicreviewand meta- and genetic correlation analysis. Bioinformatics Department of Health and Social Care (England), analysis of randomised controlled trials. Eur J Clin 2017;33:272–279 Chief Scientist Office of the Scottish Government Nutr 2017;71:1371–1380 17. Lotta AL, Pietzner M, Stewart DI, et al. Cross- Health and Social Care Directorates, Health and 3. Song Y, Cook NR, Albert CM, Van Denburgh M, platform genetic discovery of small molecule Social Care Research and Development Division Manson JE. Effects of vitamins C and E and beta- products of metabolism and application to clin- (Welsh Government), Public Health Agency carotene on the risk of type 2 diabetes in women ical outcomes. 4 February 2020 [preprint]. Bio- (Northern Ireland), British Heart Foundation, at high risk of cardiovascular disease: a random- Rxiv 2020.02.03.932541 and Wellcome. This research has been con- ized controlled trial. Am J Clin Nutr 2009;90:429– 18. Bowden J, Davey Smith G, Burgess S. Men- ducted using the UK Biobank Resource (appli- delian randomization with invalid instruments: cation number 44448). J.-S.Z. was additionally 437 4. Didelez V, Sheehan N. Mendelian randomi- effect estimation and bias detection through supported by the European Union’s Horizon Egger regression. Int J Epidemiol 2015;44:512– 2020Research andInnovation Programme under zation as an instrumental variable approach to 525 the Marie Skłodowska-Curie Actions grant agree- causal inference. Stat Methods Med Res 2007; – 19. Bowden J, Davey Smith G, Haycock PC, ment No 701708, the National Natural Science 16:309 330 Burgess S. Consistent estimation in Mendelian Foundation of China (81903316), and Zhejiang 5. Burgess S, Butterworth A, Thompson SG. randomization with some invalid instruments Ten-thousand Talents Program (2019R52039). Mendelian randomization analysis with multiple using a weighted median estimator. Genet Epi- The views expressed are those of the authors genetic variants using summarized data. Genet – demiol 2016;40:304–314 and not necessarily those of the National Health Epidemiol 2013;37:658 665 20. Shaghaghi MA, Kloss O, Eck P. Genetic Service, the National Institute for Health Re- 6. Timpson NJ, Forouhi NG, Brion MJ, et al. variation in human vitamin C transporter genes search, or the Department of Health and Social Genetic variation at the SLC23A1 locus is asso- in common complex diseases. Adv Nutr 2016;7: Care. ciated with circulating concentrations of – Where authors are identified as personnel of L-ascorbic acid (vitamin C): evidence from 5 in- 287 298 the International Agency for Research on Cancer/ dependent studies with .15,000 participants. 21. Lawlor DA, Davey Smith G, Kundu D, WorldHealthOrganization,the authorsaloneare Am J Clin Nutr 2010;92:375–382 Bruckdorfer KR, Ebrahim S. Those confounded responsible for the views expressed in this article 7. Langenberg C, Sharp S, Forouhi NG, et al.; vitamins: what can we learn from the differences and they do not necessarily represent the de- InterAct Consortium. Design and cohort descrip- between observational versus randomised trial – cisions, policy, or views of the International tionofthe InterActProject:anexaminationof the evidence? Lancet 2004;363:1724 1727 Agency for Research on Cancer/World Health interaction of genetic and lifestyle factors on the 22. Michels AJ, Hagen TM, Frei B. Human genetic fl Organization. incidence of type 2 diabetes in the EPIC Study. variation in uences vitamin C homeostasis by Duality of Interest. J.D. is a member of the Diabetologia 2011;54:2272–2282 altering vitamin C transport and antioxidant International Cardiovascular and Metabolic Ad- 8. Day N, Oakes S, Luben R, et al. EPIC-Norfolk: enzyme function. Annu Rev Nutr 2013;33:45–70 visory Board for Novartis (since 2010), the Steer- study design and characteristics of the cohort. 23. Burzle¨ M, Suzuki Y, Ackermann D, et al. The ing Committee of UK Biobank (since 2011), the European Prospective Investigation of Cancer. Br sodium-dependent ascorbic acid transporter Medical Research Council International Advisory J Cancer 1999;80(Suppl. 1):95–103 family SLC23. Mol Aspects Med 2013;34:436– Group (ING) member, London (since 2013), the 9. Danesh J, Saracci R, Berglund G, et al.; EPIC- 454 Medical Research Council High Throughput Sci- Heart. EPIC-Heart: the cardiovascular component 24. Bunnag S, Einecke G, Reeve J, et al. Mo- ence “Omics” Panel Member, London (since of a prospective study of nutritional, lifestyle and lecular correlates of renal function in kidney 2013), the Scientific Advisory Committee for biological factors in 520,000 middle-aged partic- transplant biopsies. J Am Soc Nephrol 2009;20: Sanofi (since 2013), the International Cardio- ipants from 10 European countries. Eur J Epi- 1149–1160 vascular and Metabolism Research and Devel- demiol 2007;22:129–141 25. Okada Y, Sim X, Go MJ, et al.; KidneyGen opment Portfolio Committee for Novartis, and 10. Collins R. What makes UK Biobank special? Consortium; CKDGen Consortium; GUGC consor- the AstraZeneca Genomics Advisory Board (2018). Lancet 2012;379:1173–1174 tium. Meta-analysis identifies multiple loci asso- No other potential conflicts of interest relevant 11. Mahajan A, Taliun D, Thurner M, et al. Fine- ciated with kidney function-related traits in east to this article were reported. mapping type 2 diabetes loci to single-variant Asian populations. Nat Genet 2012;44:904–909 Author Contributions. J.-S.Z., J.L., and E.S. resolution using high-density imputation and 26. Hong SH, Kwak JH, Paik JK, Chae JS, Lee performed the statistical analyses. J.-S.Z. and islet-specific epigenome maps. Nat Genet 2018; JH. Association of polymorphisms in FADS gene N.G.F.draftedthemanuscript.TheWorkingGroup 50:1505–1513 with age-related changes in serum phospholipid (J.-S.Z.,J.L.,E.S.,F.I.,I.D.S.,F.R.D.,M.P.,E.W.,L.A.L., 12. Karlsen A, Blomhoff R, Gundersen TE. High- polyunsaturated fatty acids and oxidative T.E.G.,A.S.B.,C.L.,andN.G.F.)andP.A.,E.A.,M.-D.C., throughput analysis of vitamin C in human stress markers in middle-aged nonobese men. G.F., P.W.F., R.K., N.L., F.R.M., P.M.N., N.C.O.-M., plasma with the use of HPLC with monolithic Clin Interv Aging 2013;8:585–596 106 Plasma Vitamin C and Type 2 Diabetes Diabetes Care Volume 44, January 2021

27. Hara T, Hashimoto Y, Akuzawa T, Hirai R, and carotenoids as indicators of fruit and veg- 31. Khaw KT, Bingham S, Welch A, et al. Relation Kobayashi H, Sato K. Rer1 and calnexin regulate etable intake: a systematic review and meta- between plasma ascorbic acid and mortality in endoplasmic reticulum retention of a peripheral analysis of randomised controlled trials. Br J Nutr men and women in EPIC-Norfolk prospective myelin protein 22 mutant that causes type 1A 2015;114:1331–1340 study: a prospective population study. European Charcot-Marie-Tooth disease. Sci Rep 2014;4:6992 Prospective Investigation into Cancer and Nutri- 30. Cooper AJ, Sharp SJ, Luben RN, Khaw KT, 28. Lee HS, Kim BK, Nam Y, et al. Protective role tion. Lancet 2001;357:657–663 Wareham NJ, Forouhi NG. The association of phosphatidylcholine against cisplatin-induced 32. Zheng JS, Sharp SJ, Imamura F, et al. Asso- renal toxicity and oxidative stress in rats. Food between a biomarker score for fruit and vege- ciation of plasma biomarkers of fruit and veg- Chem Toxicol 2013;58:388–393 table intake and incident type 2 diabetes: the etable intake with incident type 2 diabetes: EPIC- 29. Pennant M, Steur M, Moore C, Butterworth EPIC-Norfolk study. Eur J Clin Nutr 2015;69:449– InterAct case-cohort study in eight European A, Johnson L. Comparative validity of vitamin C 454 countries. BMJ 2020;370:m2194