An Exome-Chip Association Analysis in Chinese Subjects Reveals a Functional Missense Variant of GCKR That Regulates FGF21 Levels

Fibroblast growth factor 21 (FGF21) is a circulating metabolic hormone predominantly secreted from the liver and from other tissues, such as adipose tissue, pancreas, and skeletal muscle (1,2). Numerous animal studies have demonstrated its favorable effects on insulin sensitivity, glucose and lipid metabolism, and body weight in obese mice and diabetic monkeys (1–5). High circulating levels of FGF21, however, have been observed in patients with obesity and a range of obesity-related conditions, including type 2 diabetes (T2D), coronary artery disease, and metabolic syndrome (1). Our group previously showed that elevated levels of FGF21 independently predicted the development of T2D (6), carotid atherosclerosis (7), and kidney disease progression in T2D subjects (8). Observations of the paradoxical increase of FGF21 in the above diseases may represent a compensatory response to FGF21 resistance, which has been demonstrated in a previous study of mice with diet-induced obesity (9), or to combat the metabolic disturbances in these disease conditions.

The circulating levels of FGF21 have been reported to be moderately heritable in a twin study of young adults, with 40% of variation attributed to genetic determinants (10). So far, only a few genetic variants have been reported to show weak associations with circulating FGF21 levels (11,12). No previous genome-wide or exome-wide association analyses on circulating FGF21 levels have been published to date. The heritability of circulating FGF21 levels remains largely unexplained. This study sought to identify the genetic determinants of circulating FGF21 levels. We conducted an exome-chip association analysis on circulating FGF21 levels using a specially designed Illumina HumanExome BeadChip (Asian exome-chip) in a Chinese population (13,14). In consideration of the possible difference in circulating FGF21 levels between individuals from the patient groups and those from the general population, possibly due to the intrinsic metabolic stress under disease conditions and the treatment effects on FGF21 expression, we first conducted the association analysis in a community-based and two clinic-based cohorts separately, followed by a combined analysis involving a total of 5,169 individuals.

An exome-chip association study on circulating FGF21 levels was performed in 5,169 Chinese individuals who were recruited from the Hong Kong Cardiovascular Risk Factor Prevalence Study (CRISPS) (n = 1,172), the Hong Kong West Diabetes Registry (HKWDR) (n = 2,884), and the Hong Kong Chinese Coronary Artery Disease (HK-CAD) study (n = 1,113). Details of the study cohorts and measurement of FGF21 levels are described in the Supplementary Data.

All subjects were genotyped using the custom Asian exome-chip (13,14) with an add-on content of 58,317 single nucleotide polymorphisms (SNPs), including a custom panel of over 30,000 missense or nonsense coding variants on top of the standard content of the Infinium HumanExome BeadChip (HumanExome-12v1_A; Illumina, San Diego, CA). Genotyping of the exome array was performed using the Illumina iScan system platform at the Centre for Genomic Sciences of The University of Hong Kong. The GenTrain version 2.0 in GenomeStudio V2011.1 (Illumina) was used to perform genotyping calling. Details of genotype quality control are described in the Supplementary Data. After all quality control measures, a total of 1,172 subjects and 76,389 variants remained in the association analysis of the CRISPS cohort, whereas the analyses of the clinic-based cohorts included a total of 2,884 subjects and 77,598 variants from the HKWDR and 1,113 subjects and 75,468 variants from the HK-CAD study. The combined analysis included only polymorphic SNPs in all three cohorts and had minor allele frequency (MAF) ≥0.1%, totaling 5,169 subjects and 70,444 variants. The P value informed linkage disequilibrium (LD)-based clumping approach with the “–clump” command implemented in PLINK was performed to address the between-SNP LD. Each index SNP represented the strongest association P value from each clumped region. Each index SNP formed clumps with other variants, which were within ±500 kb from the index SNP and in LD with the index SNP (r² ≥0.2).

For each cohort, we carried out single-variant association tests, under the additive genetic model, for all markers that passed quality controls using PLINK version 1.9 (15). The FGF21 levels were first transformed to rank-based inverse normal residuals before analysis to ensure normality and minimize the possible effect of outliers. Age, sex, and the first two principal components (PCs) were included as covariates in the multiple linear regression model (model 1) (Supplementary Fig. 1). To assess the adiposity-independent association, BMI was included as an additional covariate in model 2. Meta-analysis of the association results of the three cohorts was conducted using GWAMA (Genome-Wide Association Meta Analysis) (http://www.geenivaramu.ee/en/tools/gwama; accessed 20 December 2016) (16). The inverse variance fixed-effect method was used to meta-analyze the summary statistics of the three cohorts. The heterogeneity of effect was assessed using Cochran Q test and I² index. All 70,444 SNPs showed no evidence of heterogeneity in the combined analysis with a Cochran Q test P value ≥ 1 × 10⁻⁶ (a value selected to take into account the multiple testing and the stringent Bonferroni correction) (17). Exome-wide significance was defined as P < 7.10 × 10⁻⁷ (0.05/70,444). As visualized in a quantile-quantile plot (Fig. 1), the test statistics appeared well calibrated. Associations of GCKR rs1260326 with fasting plasma glucose (FPG) and triglyceride (TG) were examined by multiple linear regression analyses in individuals without T2D and in subjects not taking lipid-lowering medications, respectively, with adjustment for age, sex, BMI, PC1, and PC2.

Of the 70,444 polymorphic SNPs (Fig. 2) examined for associations with circulating FGF21 levels in 5,169 Chinese individuals (Table 1), 45.06% altered protein composition and 19.96% were Asian-specific variants with MAF 0.1–5%.

The strongest association was detected at a missense variant rs1260326 (p.Pro446Leu) of GCKR (Table 2 and Supplementary Fig. 2). The T (Leu446) allele of rs1260326 was significantly associated with a higher level of FGF21 levels at genome-wide significance (model 1, P_combined = 1.61 × 10⁻¹², β [SE] = 0.14 [0.02], P_{heterogeneity} = 0.018, I² = 0.75; model 2: P_combined = 3.01 × 10⁻¹⁴, β [SE] = 0.15 [0.02], P_{heterogeneity} = 0.041, I² = 0.69 [Table 2 and Fig. 2]). GCKR rs1260326 was significantly associated with TG (P < 0.001, β [SE] = 0.08 [0.005]). A modest association between rs1260326 and FPG (P = 0.013, β [SE] = −0.01 [0.004]) was also observed. The associations between rs1260326 with FGF21 levels remained significant when FPG (P_combined = 1.60 × 10⁻¹⁴, β [SE = 0.16 [0.02]) or TG (P_combined = 1.26 × 10⁻⁷, β [SE] = 0.11 [0.02]) or both (P_combined = 5.42 × 10⁻⁷, β [SE] = 0.10 [0.02]) were included in the adjustment models. The proportion of variance in FGF21 levels explained by rs1260326 was estimated to be 0.97%.

In the combined analysis, we also observed suggestive associations of an intronic SNP rs17817964 at FTO (P_combined = 1.44 × 10⁻⁶, β [SE] = 0.13 [0.03], P_{heterogeneity} = 0.690, I² = 0) and a missense variant rs2273983 (p.Asn719Ser) at TSR1 (P_combined = 4.61 × 10⁻⁶, β [SE] = 0.48 [0.10], P_{heterogeneity} = 0.590, I² = 0) with FGF21 levels in adjustment model 1. The associations of these two variants were slightly attenuated on further adjustment for BMI (Table 2). Supplementary Table 1 shows the 27 index SNPs that showed associations with FGF21 level at P_combined < 5 × 10⁻⁴ in the combined analysis.

The current study reports the first exome-chip association analysis on circulating FGF21 levels. By genotyping 5,169 Chinese individuals using a custom Asian exome-chip, we identified a genome-wide significant association between GCKR rs1260326, a proline to leucine substitution at amino acid 446 (p.Pro446Leu), and circulating FGF21 levels. Our study has shed light on the possible role of GCKR on the regulation of circulating FGF21 levels. GCKR, encoding the glucokinase regulator (also known as glucokinase regulatory protein [GKRP]), has been implicated in a wide range of important metabolic pathways, such as glucose homeostasis and lipid metabolism (18). Both FGF21 and GCKR are highly expressed in the liver (1,18). GCKR acts as a competitive inhibitor of glucokinase (GCK), the principal modulator of glucose uptake and release in the liver (18). GCKR forms an inhibitory complex with GCK (GCKR–GCK complex) and allosterically controls its activity in the presence of fructose. Dissociation of the GCKR–GCK complex facilitates the translocation of GCK to cytoplasm where hepatic glycolysis is stimulated (18).

The Pro446 residue of GCKR is conserved across various species, including human and rat. GCKR p.Pro446Leu is located in close proximity to Asp413 and Gln443, which forms a critical salt bridge with Arg186 of GCK (19). Thus, the variant could interfere directly with the binding of GCKR to GCK (19), thereby leading to reduced inhibition of GCK. This in turn contributes to an enhanced glycolytic flux and promoted uptake of glucose into the liver. Accompanying this increased rate of glycolysis is the raised levels of other liver metabolites, such as malonyl-CoA, which ultimately lead to enhanced production of TG (20). This explains the reduced FPG and raised TG levels observed in individuals carrying the Leu446 (T) allele as previously reported by the French longitudinal cohort study D.E.S.I.R. (Data from an Epidemiological Study on the Insulin Resistance Syndrome) (21) and in the current study. GCKR Leu446 may influence FGF21 expression via its ability to increase GCK activity. This leads to elevated phosphorylation of glucose to glucose-6-phosphate (G6P) and the subsequent inhibition of carnitine/palmitoyl-transferase, the rate-limiting enzyme of β-oxidation, by malonyl-CoA (22). In turn, such changes may result in enhanced FGF21 expression via increased G6P-mediated activation of ChREBP (23) and via elevated fatty acid synthesis (1), which was known to regulate FGF21 gene expression.

GCKR has been recognized as a highly pleiotropic gene (1,2). Common variants of GCKR, including rs1260326 or its intronic proxy SNP rs780094 (r² = 0.92), have been shown to be associated with multiple metabolic traits, including FPG and TG (2,13,21). The current study further demonstrated that rs1260326 may play a role in the regulation of FGF21 levels that is independent from the effect of BMI, FPG, and TG. Future functional studies elucidating the role of GCKR in the in vivo regulation of FGF21 levels are warranted.

Recent clinical trials have demonstrated that treatments with FGF21 analogs in obese patients with T2D resulted in obvious improved lipid profiles and lowered fasting insulin and body weight, similar to what have been observed in animals (24,25). In particular, FGF21 treatment has demonstrated a rapid and prominent reduction in circulating TG, which occurred as soon as 2 days after treatment (24). However, there was lack of significant effect on glycemic control in humans (24,25), contradictory to the glucose-lowering effect of FGF21 observed in animals (4,5). In line with these reports, the GCKR rs1260326 was found to be strongly associated with TG but only modestly correlated with FPG in the current study (24,25). Taken together, these findings suggest that FGF21 may play a more prominent role in regulating lipid metabolism and adiposity rather than blood glucose. However, our finding that rs1260326 could possibly influence the circulating levels of FGF21 via enhanced GCK activity also suggests that alterations in hepatic carbohydrate metabolism can impact on FGF21 expression and its circulating levels.

The major limitation of the current study was the lack of external validations. Further replication analysis in independent cohorts in other Asians or European populations would serve to validate our findings. The design of the exome-chip has allowed us to examine more coding variants with much lower MAF compared with conventional genome-wide association studies. However, the relatively small sample size has limited our study power to detect the association of rare or low-frequency variants. With a view to achieve a larger sample size, a meta-analysis of genome/exome-wide association studies would be useful to identify additional genetic variants influencing FGF21 levels.

In summary, we conducted an exome-chip association analysis and identified, for the first time, a highly significant association of a functional variant of GCKR, rs1260326 (p.Pro446Leu), with circulating FGF21 levels, independent of obesity and other metabolic traits. Our findings further highlighted the pleiotropic role of GCKR and provided insights into the regulation of FGF21 levels and its relationship with metabolic diseases.

Acknowledgments. The authors thank all the study participants and clinical and research staff of the CRISPS, HKWDR, and HK-CAD study cohorts for their contribution in this research study.

Funding. This work was supported by the Hong Kong Research Grants Council’s Theme-based Research Scheme (T12-705/11) and Hong Kong Research Grants Council’s Collaborative Research Fund (HKU2/CRF/12R).

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

Author Contributions. C.Y.Y.C. wrote the first draft of the manuscript. C.Y.Y.C. and C.S.T. analyzed and interpreted the data. C.Y.Y.C., A.X., C.-H.L., K.-W.A., L.X., C.H.Y.F., K.H.M.K., W.-S.C., Y.-C.W., M.M.A.Y., J.H., B.M.Y.C., K.C.B.T., and T.-H.L. were involved in the sample collection, selection, and phenotype data preparation. C.S.T., S.S.C., and P.-C.S. provided useful comments to data analysis. H.-F.T. and K.S.L.L. were involved in the database management for the study cohorts. H.-F.T., P.-C.S., and K.S.L.L. conceived the study and undertook project leadership. All authors contributed to the drafting and critical revision of the manuscript. All authors approved the final version of the manuscript. H.-F.T., P.-C.S., and K.S.L.L. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented in poster form at the Endocrine Society Annual Meeting, Orlando, FL, 1–4 April 2017.