An Expanded Genome-Wide Association Study of Fructosamine Levels Identifies RCN3 as a Replicating Locus and Implicates FCGRT as the Effector Transcript

Fructosamine is a measure of total glycated proteins in serum. Since the most abundant serum protein is albumin, fructosamine predominately reflects glycation of albumin (1). In contrast to glycated hemoglobin (HbA_1c), which reflects average glycemia during the preceding 3 months, fructosamine measures short-term glycemic control (from 2 to 3 weeks), reflecting the shorter turnover time of serum proteins (1). As it is independent of hemoglobin, fructosamine levels are not affected by red cell turnover or characteristics of hemoglobin, making it a viable alternative to HbA_1c to monitor glycemic control in the presence of anemia or a hemoglobinopathy (1). Another important difference is that whereas fructosamine reflects levels of extracellular glucose, HbA_1c is a measure of intracellular glycation. Determinants of these two measurements may reflect differences in glycation in the two different environments (2). Despite its potential advantages and its association with diabetes incidence, retinopathy, and chronic kidney disease (CKD), independently of baseline fasting glucose (FG) and HbA_1c (1,3), fructosamine has not been widely used as a measure of glucose control (4).

To date, a single study has examined the single nucleotide polymorphism–based heritability of fructosamine, yielding an h² estimate of ∼13% (5). A fructosamine genome-wide association study (GWAS) performed on 8,951 U.S. White individuals (N_discovery = 7,647) and 2,712 Black individuals (N_discovery = 2,104) without a diabetes diagnosis found an association in Whites near RCN3 (rs34459162, P_discovery =5.3 × 10⁻⁹) and an association near CNTN5 (rs2438321, P_discovery = 6.2 × 10⁻⁹) in Blacks but neither variant replicated in additional samples (N_replication = 1,304 and N_replication = 608, respectively). This study also demonstrated that, despite some evidence (P < 2.7 × 10⁻⁴) of association with three established FG and/or HbA_1c loci (TCF7L2, GCK, and SLC2A2), there was no significant (P > 0.05) genetic correlation of fructosamine with FG or HbA_1c.

In this study, we aimed to gain further insight into the genetic architecture of fructosamine by performing a GWAS in 20,731 European-ancestry blood donors from the INTERVAL cohort (6). To increase power for novel locus discovery, we combined our results with association statistics from U.S. White participants from the study by Loomis et al. (7) in a meta-analysis (N_meta = 29,685). Lastly, we explored the heritability of the trait and its genetic relationship with other glycemic and nonglycemic traits to establish the degree of shared genetic influences.

We conducted a GWAS for fructosamine using the INTERVAL cohort (6) and then meta-analyzed our results with those of U.S. White participants from the previously published Atherosclerosis Risk in Communities (ARIC) Study (7). The INTERVAL cohort consists of 47,394 blood donors in the U.K. (6). The ARIC Study consists of 15,792 participants recruited from four U.S. communities (8).

All participants from the INTERVAL cohort were genotyped using the Affymetrix UK Biobank Axiom Array and imputed using a combined UK10K-1000G phase III imputation panel (9) and those from ARIC were genotyped using the Affymetrix 6.0 array and imputed separately by race using the 1000G Project phase I reference panel (7). Genotype quality control for INTERVAL has been previously described in Astle et al. (9). Briefly, samples with poor signal intensity (dish quality control <0.82) or low call rate (<97%) were excluded. Duplicated, contaminated, and non-European samples were also excluded. Variants with low call rate (<95%) and those with cluster statistics indicating poor quality genotyping or hard-to-call multiallelic variants were excluded. Additionally, before imputation, variants were removed using the following filters: 1) Hardy-Weinberg equilibrium P < 5 × 10⁻⁶; 2) call rate <99% over the genotyping batches in which the variant did not fail; and 3) global call rate <75% (over 10 genotyping batches). After imputation, the total number of variants was 87,696,910. In ARIC, samples with high missingness (>5%), sex mismatch, discordance with previous TaqMan assay genotypes, genetic outliers, and relatedness were excluded (9). Low frequency variants (minor allele frequency [MAF] <5%) and those with imputation quality <0.8 were excluded, resulting in 5,446,889 variants (7).

Phenotyping for the INTERVAL cohort was performed by Star-SHL laboratory (https://www.star-shl.nl/), and fructosamine was measured on 28,310 INTERVAL cohort participants using a colorimetric assay (Roche/Hitachi Modular P analyzer system). We performed phenotype quality control in R (10) to prepare the data for association analysis. After adjusting for relevant biometric and technical variables (sex, donation center, height, weight, processing date, number of donations, and attendance date), values were transformed on the natural log scale in order to match the approach taken by Loomis et al. (7). After removal of participants on glucose-lowering medication and phenotype quality control, we kept 20,731 participants with fructosamine and genotype data. Fructosamine in ARIC was measured using a Roche Modular P800 system from serum collected at visit 2.

BOLT-LMM (11) was used to run genome-wide association analysis on 19,100,024 variants with MAF >0.1% and INFO score >0.4. Linkage disequilibrium (LD) score regression results showed no signs of inflation, so no genomic correction was performed (LD intercept = 1.01). Summary statistics for ARIC White participants from Loomis et al. (7) were obtained from the authors. We then performed inverse variance-weighted meta-analysis using a fixed-effects model in METAL (12). In total, 5,200,018 were included in the meta-analysis. Variants were clumped into the same locus if they were within 250 kb of the lead variant and if r² > 0.1. Clumping was performed as implemented in PLINK (13). Variants were declared as genome-wide significant if they met the standard genome-wide significance threshold (P < 5 × 10⁻⁸). To identify potential effector transcripts at the RCN3 locus, expression data from Genotype-Tissue Expression (GTEx) v7 (https://gtexportal.org/) (14) were used to discover colocalized expression quantitative trait locis (eQTLs) in whole blood. For this purpose, we used coloc (15), a software package that calculates the probability of two phenotypes sharing a causal variant in a region by performing approximate Bayes factor colocalization analysis. Protein-coding genes within 1 Mb of the lead variant in the RCN3 locus were tested for colocalization.

LD score regression (16) was used to establish the heritability of the trait. Genetic correlation analyses with glycemic traits, hematological traits, anthropometric traits, and kidney diseases/traits (Supplementary Table 1) were performed using LD Hub (17). Power calculation for genetic correlation analyses was done using the GCTA-GREML Power Calculator (18).

Summary statistics from the genome-wide association analysis in INTERVAL will be available from the GWAS catalog upon publication under accession GCST90017143.

Genome-wide association analysis of fructosamine in 20,731 blood donors from INTERVAL (19,100,024 variants; MAF >0.1%) yielded two genome-wide significant (P < 5 × 10⁻⁸) loci. The ABCB11 locus (rs853777, β = −0.009 [95% CI −0.013 to −0.007]; MAF = 0.35; P = 8.8 × 10⁻⁹) previously associated with HbA_1c and FG (19) and the RCN3 locus (rs111476047, β = 0.013 [95% CI 0.009–0.017]; MAF = 0.21; P = 2.1 × 10⁻¹¹) associated with fructosamine in Loomis et al. (7) (Table 1). Next, to increase power for additional locus discovery, we performed genome-wide meta-analysis of our data set with that of White participants from Loomis et al. (7). Following meta-analysis (Table 1 and Supplementary Figs. 1–4), two loci were genome-wide significant: RCN3 (rs113886122, effect allele = C; β = 0.013 [95% CI 0.010–0.017]; MAF = 0.17; P_meta = 5.71 × 10⁻¹⁸) and GCK (rs3757840, effect allele = T; β = 0.006 [95% CI 0.004–0.008]; MAF = 0.49; P_meta=3.66 × 10⁻⁸), another established glycemic trait locus (19). In contrast, the association at the ABCB11 locus was no longer genome-wide significant (P_meta = 8.50 × 10⁻⁷) due to lack of supporting evidence for association at this locus in ARIC (rs853777, effect allele = T; β = −0.002 [95% CI −0.005 to 0.001]; P = 0.17) (Table 1).

While GCK is known to be the effector transcript at this locus (20), little is known about the RCN3 locus and its relationship with fructosamine. We therefore sought to explore if eQTL information could point toward potential effector transcripts at this locus. Of 42 protein-coding genes within 1 Mb of the lead signal (rs113886122), only the FCGRT eQTL in whole blood displayed convincing evidence of a shared causal variant with same direction of effect (posterior probability 97.7%) (Supplementary Table 2), suggesting it is the likely effector transcript at this locus (Fig. 1).

To estimate the heritability of fructosamine, we next used LD score regression to estimate its heritability explained by common genetic variation (MAF >0.05 in EUR) and to quantify the degree of genetic correlation of fructosamine with other glycemic-related traits. Heritability was estimated to be 7.7% (95% CI 3.6–11.9). Genetic correlation results with anthropometric, glycemic, kidney, and blood cell traits (Supplementary Table 1) showed evidence of moderate negative genetic correlation (rg) (Bonferroni-corrected threshold P < 0.0012) with waist-to-hip ratio (rg = −0.29 [95% CI −0.45 to −0.14]; P = 0.0002), waist circumference (rg = −0.32 [−0.50 to −0.14]; P = 0.0004), body fat percentage (rg = −0.32 [−0.50 to −0.13]; P = 0.0007), and obesity class 1 (rg = −0.29 [−0.45 to −0.12]; P = 0.0006).

In this study, we aimed to further elucidate the genetic architecture of fructosamine by conducting a GWAS in 20,731 European-ancestry blood donors from the INTERVAL cohort. Combining our data in a meta-analysis with that of 7,647 White individuals previously published by Loomis et al. (7), we identified two loci, RCN3 and GCK, associated with fructosamine levels at genome-wide significance level (P < 5 × 10⁻⁸).

GCK (rs3757840) was not previously known to associate with fructosamine levels, but it is a well-established glycemic locus (19); it codes for glucokinase, a key enzyme that plays a role in sensing glucose levels in β-cells (20).

RCN3 (lead variant rs113886122) was shown to associate with fructosamine levels in U.S. White participants by Loomis et al. (7). Variants in this region have previously also been associated with total cholesterol, total protein, albumin, and multiple red cell traits (21,22). In this study, we replicated this association locus in a large sample of European-ancestry blood donors, and, using colocalization analysis with blood eQTL data, we established FCGRT as the likely effector transcript in the region. FCGRT codes for the Fc fragment of the IgG receptor and transporter, which plays a role in maintenance of albumin levels, protecting albumin from degradation (23). In agreement with these results, the rs59774409-C fructosamine-increasing allele was associated with higher FCGRT expression levels in whole blood. In mouse studies, hepatic levels of this protein have been shown to regulate albumin homeostasis and susceptibility to liver injury (24). These results suggest that the locus found in this study could influence fructosamine levels through pathways that regulate albumin levels. As fructosamine normally reflects glycated albumin (1), a shared genetic link is not unexpected.

The ABCB11 locus previously associated with HbA_1c and FG (19) associated with fructosamine at genome-wide significance levels in INTERVAL participants (rs853777, P = 8.80 × 10⁻⁹), but failed to reach this threshold after meta-analysis with White participants from Loomis et al. (7) (P_meta = 8.80 × 10⁻⁷). Given the fact that ABCB11 is an established glycemic locus (19), testing its association with fructosamine in larger numbers and diverse ancestry participants will be important.

In agreement with a previous study (5), fructosamine appears to be a trait with modest heritability (7.7% [95% CI −3.6 to 11.9]), suggesting most of the variation of the trait in this generally healthy population is due to environmental factors. This is in keeping with fructosamine measuring short-term changes in glycemia (25) and its use as a measure of treatment response in patients with diabetes (25). In our data, fructosamine does not show evidence of significant genetic correlation with other glycemic traits, including with HbA_1c (P > 0.05). This is despite both traits normally having a high phenotypic correlation (∼0.61 (26)) and reflecting similar biological processes—namely, the glycation of proteins and having enough power (>80%) to detect a genetic correlation of 0.16. This lack of significant genetic correlation was also observed in Loomis et al. (7).

Interestingly, the only traits for which we observed a Bonferroni significant negative genetic correlation were waist-to-hip ratio, body fat percentage, obesity class 1, and waist circumference. This is consistent with prior studies showing a negative association between BMI and fructosamine (27,28). The effect of adiposity on fructosamine is not fully understood but may impact its use as a clinical measurement of glycemic control.

Lastly, among the genetic correlation results (Supplementary Table 1), nominally significant negative correlations were found with HOMA of β-cell function, platelet count, and estimated glomerular filtration rate, while nominally significant positive genetic correlation was detected with CKD. Given the evidence in the literature linking fructosamine with incident CKD independently of other risk factors in individuals with and without diabetes (2), these correlation results between estimated glomerular filtration rate and CKD provide some interesting hypotheses to explore in future studies.

One limitation of this study is our limited power to detect associations for rarer variants (MAF <1%) due to our sample size (e.g., 28% power to detect an effect size of 0.2-SD units for variants with an MAF of 1%, which is almost double the effect size of the strongest signal in this study).

In conclusion, we have expanded knowledge into the genetic architecture of fructosamine levels by identifying a new genome-wide significant locus (GCK), highlighting FCGRT as the potential effector transcript at RCN3, finding evidence of genetic correlation with obesity-related traits, and replicating the absence of a significant genetic correlation with other glycemic traits in an increased sample size.

Acknowledgments. The authors thank the staff and participants of the ARIC study for their important contributions. The academic coordinating center thanks the blood donor center staff and blood donors for participating in the INTERVAL trial.

Funding. Participants in the INTERVAL randomized controlled trial were recruited with the active collaboration of NHS Blood and Transplant (U.K.; www.nhsbt.nhs.uk), which has supported field work and other elements of the trial. DNA extraction and genotyping was cofunded by the National Institute for Health Research (NIHR), the NIHR BioResource (https://bioresource.nihr.ac.uk), and the NIHR (Cambridge Biomedical Research Centre at the Cambridge University Hospitals NHS Foundation Trust). The sequencing work was supported by Wellcome Trust grant 206194. The academic coordinating center for INTERVAL was supported by core funding from NIHR Blood and Transplant Research Unit in Donor Health and Genomics (NIHR BTRU-2014-10024), UK Medical Research Council (MR/L003120/1), British Heart Foundation (SP/09/002, RG/13/13/30194, and RG/18/13/33946), and the NIHR (Cambridge Biomedical Research Centre at the Cambridge University Hospitals NHS Foundation Trust). A complete list of the investigators and contributors to the INTERVAL trial is provided in Di Angelantonio E, Thompson SG, Kaptoge SK, et al.; INTERVAL Trial Group. Efficiency and safety of varying the frequency of whole blood donation (INTERVAL): a randomised trial of 45,000 donors. Lancet 2017;390:2360–2371. This work was supported by Health Data Research UK, which is funded by the U.K. Medical Research Council, Engineering and Physical Sciences Research Council, Economic and Social Research Council, Department of Health and Social Care (England), Chief Scientist Office of the Scottish Government Health and Social Care Directorates, Health and Social Care Research and Development Division (Welsh Government), Public Health Agency (Northern Ireland), British Heart Foundation, and Wellcome. The ARIC study has been funded in whole or in part with federal funds from the National Heart, Lung, and Blood Institute (NHLBI), National Institutes of Health (NIH), and Department of Health and Human Services (contract numbers HHSN268201700001I, HHSN268201700002I, HHSN268201700003I, HHSN2682017 00005I, and HHSN268201700004I), grants R01HL087641, R01HL059367, and R01HL086694; National Human Genome Research Institute contract U01H G004402; and NIH contract HHSN268200625226C. Infrastructure was partly supported by the NIH and NIH Roadmap for Medical Research grant UL1RR025005. This work was funded in part by an “Expanding excellence in England” award from Research England (to I.B.). I.B. further acknowledges funding from Wellcome (WT206194). F.R.-M. received funding from Consejo Nacional de Ciencia y Tecnología México (489672). J.D. is funded by the NIHR and holds a British Heart Foundation Professorship and an NIHR Senior Investigator Award. E.S. was supported by NIH/NHLBI grant K24 HL152440 and NIH/National Institute of Diabetes and Digestive and Kidney Diseases grant R01DK089174. The GTEx Project was supported by the Common Fund of the Office of the Director of the National Institutes of Health and by the National Cancer Institute, National Human Genome Research Institute, NHLBI, National Institute on Drug Abuse, National Institute of Mental Health, and National Institute of Neurological Disorders and Stroke. The data used for the analyses described in this manuscript were obtained from the GTEx Portal on 9 October 2020.

The views expressed are those of the authors and not necessarily those of the National Health Service, NIHR, or the Department of Health and Social Care.

Duality of Interest. F.R.-M. is a current employee of Genomics Plc. I.B. and/or spouse own stock in GlaxoSmithKline, Incyte Corporation, and Inivata Ltd. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. F.R.M. performed analysis, wrote the paper, and revised, edited, and approved the final version of the paper. J.D. and E.S. contributed data sets and revised, edited, and approved the final version of the paper. D.R., E.D.A., B.Y., N.S., J.D., E.S., and A.S.B. revised, edited, and approved the final version of the paper. I.B. supervised the work, wrote the paper, and revised, edited and approved the final version of the paper. I.B. is the guarantor of this work and, as such, had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.