Association of a glucagon-like peptide-1 receptor gene variant with glucose response to a mixed meal

Abstract

Dipeptidyl peptidase-4 (DPP-4) inhibitors increase endogenous glucagon-like peptide-1 (GLP-1). We hypothesized that genetic variation in the gene encoding the GLP-1 receptor (GLP1R) could affect the metabolic response to DPP-4 inhibition. To evaluate the relationship between the GLP1R rs6923761 variant (G-to-A nucleic acid substitution) and metabolic responses, we performed mixed meal studies in individuals with type 2 diabetes mellitus and hypertension after 7-day treatment with placebo and the DPP-4 inhibitor sitagliptin. This analysis is a substudy of NCT02130687. The genotype frequency was 13:12:7 GG:GA:AA among individuals of European ancestry. Postprandial glucose excursion was significantly decreased in individuals carrying the rs6923761 variant (GA or AA) as compared with GG individuals during both placebo (P = 0.001) and sitagliptin treatment (P = 0.045), while intact GLP-1 levels were similar among the genotype groups. In contrast, sitagliptin lowered postprandial glucose to a greater degree in GG as compared with GA/AA individuals (P = 0.035). The relationship between GLP1R rs6923761 genotype and therapies that modulate GLP-1 signalling merits study in large populations.

1 ∣. INTRODUCTION

The incretin glucagon-like peptide-1 (GLP-1) increases glucose-mediated insulin secretion but is rapidly inactivated by dipeptidyl peptidase-4 (DPP-4). Whether genetic variability in the GLP-1 receptor (GLP1R) affects responses to DPP-4 inhibitors has not been extensively studied. The GLP1R variant rs6923761 is a single nucleotide polymorphism with a G-to-A nucleic acid substitution.¹ In some studies, the variant was associated with lower baseline weight and triglycerides,^2-5 yet two reports demonstrated decreased responsiveness to DPP-4 inhibition in individuals with the variant as measured by glycated haemoglobin (HbA1c).^6,7 In vitro studies provide further conflicting information. One group reported decreased cell surface expression when stably transfected into cell lines.¹ In the Genotype-Tissue Expression database, the variant is associated with decreased expression in the pancreas.⁸ Other in vitro studies demonstrate normal basal activity, agonist potency, and cAMP signalling of the variant receptor, but decreased intracellular calcium mobilization.^1,9 At this time, the relationship between this variant and metabolic responses in vivo during increased endogenous GLP-1 is unclear.

We examined the relationship between this variant and the metabolic response to a mixed meal in individuals with type 2 diabetes mellitus (T2DM) and hypertension (HTN) treated with placebo and the DPP-4 inhibitor sitagliptin.

2 ∣. METHODS

2.1 ∣. Participants and protocol

This was a substudy of NCT02130687.¹⁰ Adults aged 18 to 80 years, with HTN and T2DM, were enrolled. Patients were excluded if they had taken an antidiabetic medication other than metformin for more than 12 months, or had type 1 diabetes or poorly controlled diabetes with HbA1c ≥71.6 mmol/mol (8.7%). Participants completed an oral glucose tolerance test (OGTT) and were randomized to ramipril, valsartan or amlodipine for 15 weeks. After 4 weeks of anti-hypertensive treatment, participants were randomized to the first of three 1-week crossover therapies: placebo+placebo; sitagliptin+placebo; or sitagliptin+aprepitant (NK1 receptor antagonist). Each crossover treatment was separated by 4 weeks to avoid carryover effects. During the last day of each 1-week treatment, participants presented for a mixed meal (Figure S1 in Appendix S1).

For the present analysis, data from the anti-hypertensive groups were combined, because the metabolic responses to sitagliptin and the mixed meal were similar among anti-hypertensive groups.¹⁰ In addition, because the effect of the GLP1R rs6923761 variant was identical during sitagliptin and sitagliptin+aprepitant treatment, we present only data from the sitagliptin and placebo arms in the present paper. Data from the sitagliptin+aprepitant treatment are shown in Figure S2 in Appendix S1.

2.2 ∣. Statistical analyses

A two-sample t-test or Pearson's chi-squared test were used to compare variables among individuals of different genotypes, and a paired t-test was used to compare measures within individuals between placebo and sitagliptin treatment. For repeated measures of glucose, insulin, GLP-1, PYY1-36 and PYY3-36, we fitted a generalized least squares model with compound symmetric covariance structure to account for within-participant correlation. The model was fitted for the values from 45 to 240 minutes following the mixed meal when measures had reached steady state. Independent variables were genotype (GG vs. GA/AA), drug (placebo vs. sitagliptin), time after mixed meal, time 0 measurement and interaction between genotype and drug. Interaction between genotype and drug was removed from the model if the P value of the term was >0.1. Wald statistics were used to construct confidence intervals (CIs) for the estimates or estimated difference of interest. All reported statistics for repeated measures derive from this model unless otherwise specified, and the generalized least squares model estimate is denoted as “difference”. Subgroup analyses were performed for hydrochlorothiazide (HCTZ) non-users, and sensitivity analysis performed excluding two Hispanic participants. A two-sided P value ≤0.05 was considered significant. Analyses were performed using SPSS and R3.6.1.

3 ∣. RESULTS

Fifty-three participants were randomized as part of the parent study.¹⁰ Given known racial differences in insulin and glucose metabolism,¹¹ we set out to analyse the effect of genotype on metabolic responses within racial groups. The minor allele frequency is 28% to 32% in white people and 6% in black people.¹² The number of black participants, 11 GG and two GA, was insufficient for analysis in this group; therefore, the remainder of the paper focuses on the white participants. Two participants of Asian ancestry were excluded, two dropped out, and four did not have genotyping. Table S1 in Appendix S1 shows the characteristics of the 32 participants analysed. Genotypes were in Hardy–Weinberg equilibrium. There was no difference in age, gender, weight or body mass index among the genotypes. Baseline blood pressure, fasting blood glucose, homeostatic model assessment (HOMA) of insulin resistance, updated HOMA (HOMA2), HbA1c, and lipids were also statistically similar among the groups. Table 1 shows the fasting characteristics prior to the mixed meal in each treatment arm. Sitagliptin decreased DPP-4 activity and increased intact GLP-1 in all genotypes. Sitagliptin also decreased PYY3-36 in all genotypes. We did not detect a statistically significant effect of sitagliptin on fasting glucose, insulin or lipids.

TABLE 1.

Fasting metabolic and haemodynamic parameters prior to mixed meal on each study day

	GG (N = 13)		GA (N = 12)		AA (N = 7)
	Placebo	Sitagliptin	Placebo	Sitagliptin	Placebo	Sitagliptin
DPP-4 activity, nmol/mL/min	25.1 ± 7.2	9.7 ± 5.6*	24.1 ± 6.7	10.4 ± 3.3*	20.4 ± 4.9**	10.9 ± 8.9*
GLP-1, pg/mL	2.8 ± 3.4	16.3 ± 23.0*	5.5 ± 7.2	9.4 ± 6.2*	3.7 ± 6.3	10.1 ± 16.7*
Glucose, mmol/L	7.23 ± 2.24	6.61 ± 1.34	6.88 ± 1.35	6.54 ± 0.97	6.47 ± 1.27	6.18 ± 0.96
Insulin, μU/mL	23.0 ± 8.0	22.4 ± 10.9	27.6 ± 15.1	32.5 ± 19.2	31.2 ± 14.2	28.6 ± 13.7
PYY 1-36, pg/mL	75.4 ± 52.6	95.2 ± 33.0	94.0 ± 57.9	102.2 ± 46.9	111.1 ± 58.3	132.7 ± 66.2
PYY 3-36, pg/mL	79.2 ± 41.8	42.5 ± 17.7*	65.6 ± 34.8	42.9 ± 28.2*	97.9 ± 55.8	43.0 ± 30.6*
Triglycerides, mg/dL	129.5 ± 54.4	140.6 ± 63.0	110.6 ± 31.2	122.9 ± 35.4	124.6 ± 56.6	132.0 ± 46.4
FFA, μg/mL	262.2 ± 64.9	244.0 ± 65.3	236.7 ± 108.9	252.9 ± 96.8	242.9 ± 89.6	240.6 ± 111.5
VLDL, g/dL	24.6 ± 10.0	26.8 ± 12.2	22.5 ± 6.5	25.1 ± 7.2	25.0 ± 11.1	26.6 ± 9.2
MAP calculated, mmHg	100.3 ± 11.4	95.5 ± 8.9	92.5 ± 7.4	91.9 ± 7.3	97.9 ± 8.9	98.9 ± 5.4***
SBP, mmHg	134.4 ± 19.7	129.8 ± 15.1	125.4 ± 12.4	124.8 ± 9.9	134.4 ± 18.2	135.9 ± 9.9***
DBP, mmHg	80.8 ± 8.8	75.7 ± 7.4*	73.1 ± 7.6**	72.5 ± 9.1	78.9 ± 9.0	80.7 ± 5.8***
Heart rate, bpm	74.7 ± 9.7	73.2 ± 9.8	69.5 ± 11.5	69.3 ± 9.6	69.1 ± 8.8	71.0 ± 11.8

Abbreviations: DBP, diastolic blood pressure; DPP-4, dipeptidyl peptidase 4; FFA, free fatty acids; GLP-1, glucagon like peptide-1; MAP, mean arterial pressure; PYY, peptide YY; SBP, systolic blood pressure.

Data are shown as mean ± SD.

^*P <0.05 vs. placebo;

^**P <0.05 vs. GG;

^***P <0.05 vs. GA.

We first examined the effect of genotype on metabolic response to the mixed meal. Postprandial glucose was lower in GA/AA versus GG individuals during placebo treatment (difference −12.69 [95% CI −20.32, −5.06]; P = 0.001) and sitagliptin treatment (difference −7.76 [95% CI −15.35, −0.17]; P = 0.045 [Figure 1A]). We then examined the interaction of drug with genotype. Sitagliptin decreased glucose by an average of 0.63 mmol/L in GG (95% CI 0.43, 0.84; P <0.001), and by 0.36 mmol/L in GA/AA individuals (95% CI 0.29, 0.52; P <0.001), with an estimated difference of 0.27 mmol/L (95% CI 0.02, 0.53; P = 0.035 [Figure 1B]). Therefore, while GA/AA individuals had lower postprandial glucose as compared with GG individuals, DPP-4 inhibition had a greater glucose-lowering effect in GG individuals compared with GA/AA.

FIGURE 1.

Relationship between GLP1R rs6923761 genotype and postprandial glucose, intact glucagon-like peptide-1 (GLP-1) and insulin after 1-week treatment with either placebo or sitagliptin. Plots show mean ± SD for A, glucose, B, decrease in glucose with sitagliptin treatment, C, insulin and D, GLP-1. Time zero marks ingestion of mixed meal. In B, time 0 subtracted average glucose from 45 to 240 minutes was calculated for GG versus GA/AA individuals during placebo and sitagliptin treatment, and subtracted to yield the average decrease in glucose with sitagliptin

We next quantified insulin, GLP-1, glucagon, and calculated measures of glucose homeostasis. As a surrogate of insulin secretion, we calculated the area under the curve (AUC) of insulin-to-glucose (InsAUC30/GluAUC30), which has previously been validated for OGTT.¹³ We found that GA/AA individuals had an increased InsAUC30/GluAUC30 as compared with GG during the screening OGTT, suggesting enhanced insulin secretion (GG: 0.305 ± 0.208; GA/AA: 0.551 ± 0.340, difference 0.246 [95% CI 0.044, 0.448]; P = 0.019, t-test). During the mixed meal, uncorrected insulin levels were not statistically different (Figure 1C). When insulin levels were normalized for glucose by calculating an insulin-to-glucose ratio during the mixed meal, this ratio was significantly higher in GA/AA individuals during placebo, indicating that for a given glucose concentration, GA/AA individuals have higher insulin levels (difference 0.12 [95% CI 0.02, 0.22]; P = 0.019 [Figure S3 in Appendix S1]). The insulin-to-glucose ratio was not statistically different among genotypes during sitagliptin (difference 0.06 [95% CI −0.04, 0.17]; P = 0.226).

We next calculated the Matsuda insulin sensitivity index, which combines hepatic and peripheral tissue sensitivity and functions well in a mixed meal.¹⁴ There was no significant difference in the Matsuda index among the genotypes, suggesting that the variant does not affect tissue sensitivity by this measure (placebo: GG 2.6 ± 0.7, GA/AA 2.8 ± 1.9, difference −0.16 [95% CI −1.13, 0.82]; P = 0.746, t-test; sitagliptin: GG 3.4 ± 1.4, GA/AA 2.9 ± 1.7, difference 0.53 [95% CI −0.63, 1.69]; P = 0.355, t-test). Furthermore, the oral disposition index was not different between genotypes (placebo: GG 0.14 ± 0.14, GA/AA 0.30 ± 0.36, difference −0.17 [95% CI −0.36, 0.02]; P = 0.077, t-test; sitagliptin: GG 0.15 ± 0.13, GA/AA 0.21 ± 0.43, difference −0.06 [95% CI −0.28, 0.16]; P = 0.573, t-test).

We further examined GLP-1 and glucagon levels. GLP-1 was higher during sitagliptin versus placebo by 24.8 pg/mL (95% CI 19.5, 30.1; P < 0.001) but did not differ by genotype (difference 6.1 [95% CI −9.6, 21.8]; P = 0.446 [Figure 1D]). Glucagon levels 30 minutes post-mixed meal were not different among the genotypes (placebo: GG 88.9 ± 33.7 pg/mL, GA/AA 101.0 ± 32.7 pg/mL, difference – 12.06 [95% CI −36.74, 12.62]; P = 0.324, t-test; sitagliptin: GG 84.8 ± 27.2 pg/mL, GA/AA 96.2 ± 33.0 pg/mL, difference – 11.35, [95% CI −33.21, 10.51]; P = 0.297, t-test). There was no difference in triglyceride or free fatty acid levels among genotypes (data not shown).

Dipeptidyl peptidase-4 also degrades the gut hormone peptide YY (PYY) to PYY3-36, and these peptides affect insulin responses in animal models.¹⁵ As reported,¹⁰ PYY1-36 and PYY3-36 increased following meal ingestion. Sitagliptin significantly decreased PYY3-36 concentrations by 46.6 pg/mL (95% CI 29.3, 64.0; P <0.001) for GG and 64.8 pg/mL (95% CI 50.0, 79.7; P <0.001) for GA/AA individuals. PYY levels did not differ by genotype (Figure S4 in Appendix S1).

To ensure that effects of genotype were not confounded by HCTZ use, we performed separate analyses in HCTZ non-users only (GG:GA:AA, n = 8:9:7) and obtained similar results. We also performed sensitivity analysis to exclude two Hispanic participants and obtained similar results. Finally, given our small sample size, we sought to confirm our findings in a separate cohort. In a study of obese individuals with prediabetes, we found lower glucose excursion after an OGTT in GA/AA (n = 23) as compared with GG individuals (n = 11) at 60 minutes (difference 22.4 [95% CI 0.72, 44.1]; P = 0.043, t test) and 90 minutes (difference 29.2 [95% CI 9.2, 49.2]; P = 0.006, t test [Figure S5]; full details of study in Appendix S1).

4 ∣. DISCUSSION

Glucagon-like peptide-1 receptor agonists and DPP-4 inhibitors are widely used to treat T2DM, and polymorphisms in GLP1R may affect responses to these drugs. In vitro studies must be interpreted with caution given the highly artificial system of overexpression in cell lines that is required to study this receptor. Prior studies in humans have suggested a metabolic advantage to the presence of the GLP1R rs6923761 variant. We report for the first time that individuals of European descent with HTN and T2DM carrying this variant exhibit decreased glucose excursion following a mixed meal and have increased surrogate markers of insulin secretion as measured by the InsAUC30/GluAUC30 during OGTT. There was no relationship between genotype and postprandial GLP-1 levels or glucagon, suggesting the effect may be driven by insulin. We hypothesize that individuals with this variant have enhanced postprandial GLP-1 signalling due to increased receptor sensitivity, resulting in higher insulin secretion and lower glucose excursion. However, our study design did not test this hypothesis directly and further studies are needed. In addition, two recent papers by Gasbjerg et al ^16,17 using combinations of glucose-dependent insulinotropic polypeptide (GIP) and GLP-1 receptor antagonists in healthy individuals during OGTT and mixed meal have demonstrated a greater contribution of endogenous GIP, as compared with GLP-1, to postprandial glucose excursion and insulin secretion, whereas GLP-1 antagonism had a greater effect on gastric emptying during OGTT. These studies raise questions regarding the potency of GLP-1 as an incretin hormone which will need to be further investigated in populations with diabetes and in those with the rs6923761 variant.

Notably, we detected a drug-by-genotype effect whereby treatment with sitagliptin led to more significant glucose-lowering in GG as compared with GA/AA individuals, suggesting that the metabolic advantage conferred by the variant is less important in the setting of higher endogenous GLP-1 levels. This has been reported in two prior studies in which DPP-4 inhibition lowered HbA1c less in GA/AA individuals than in GG individuals.^6,7 Thus, while the present study illustrates a metabolic advantage to this variant, it also shows that GG individuals benefit the most from the glucose-lowering effects of DPP-4 inhibition.

A limitation of this study is the small sample size. In addition, because the endpoints in this substudy are secondary or exploratory, we did not apply multiple testing correction. We have included estimates and CIs to facilitate interpretation. In addition, as diabetes is a complex polygenic disease, the effects we observed may be moderated by interactions with other genetic or environmental factors, which the study was not powered to detect. Furthermore, the small number of patients of African ancestry in the original cohort did not permit analysis in this group. In addition, while GA/AA individuals had lower postprandial glucose as compared with GG individuals during both placebo and sitagliptin (Figure 1A), we detected a statistically significant change in insulin-to-glucose ratio only during placebo treatment. This may in part be explained by the drug-by-genotype effect discussed above, whereby the metabolic benefit of the variant is decreased during sitagliptin treatment when endogenous GLP-1 is increased. Alternatively, different mechanisms, such as delayed gastric emptying,^16,18 could contribute to lower postprandial glucose in individuals with the variant during sitagliptin. The present study may not have been powered to detect a smaller effect of the variant during DPP-4 inhibition using calculated measures of β-cell function and insulin sensitivity. Finally, our population had concurrent HTN. We did not detect an effect of HTN treatment in the original study on the metabolic outcomes, and as HTN and T2DM commonly occur in the same patient, this does not affect the generalizability of the findings.

The relationship between GLP1R rs6923761 genotype and response to pharmacological GLP-1 signalling therapy merits study in large populations. Understanding genetic polymorphisms that impact an individual's responses to drugs targeting the GLP-1 system will be crucial as this class of agents becomes more broadly utilized.

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.