Association between Genotype and the Glycemic Response to an Oral Glucose Tolerance Test: A Systematic Review

The inter-individual variability of metabolic response to foods may be partly due to genetic variation. This systematic review aims to assess the associations between genetic variants and glucose response to an oral glucose tolerance test (OGTT). Three databases (PubMed, Web of Science, Embase) were searched for keywords in the field of genetics, OGTT, and metabolic response (PROSPERO: CRD42021231203). Inclusion criteria were available data on single nucleotide polymorphisms (SNPs) and glucose area under the curve (gAUC) in a healthy study cohort. In total, 33,219 records were identified, of which 139 reports met the inclusion criteria. This narrative synthesis focused on 49 reports describing gene loci for which several reports were available. An association between SNPs and the gAUC was described for 13 gene loci with 53 different SNPs. Three gene loci were mostly investigated: transcription factor 7 like 2 (TCF7L2), peroxisome proliferator-activated receptor gamma (PPARγ), and potassium inwardly rectifying channel subfamily J member 11 (KCNJ11). In most reports, the associations were not significant or single findings were not replicated. No robust evidence for an association between SNPs and gAUC after an OGTT in healthy persons was found across the identified studies. Future studies should investigate the effect of polygenic risk scores on postprandial glucose levels.

Genome-wide association studies (GWAS) have identified associations between single nucleotide polymorphisms (SNPs) and fasting glucose levels. For instance, the Meta-Analysis of the Glucose and Insulin-related traits Consortium (MAGIC) reported several independent genetic loci associated with glucose metabolism [5]. Furthermore, a metaanalysis of nine GWAS, with 15,234 participants without type 2 diabetes mellitus (T2DM), revealed five genetic loci that are associated with the 2-hour glucose level after an oral glucose tolerance test (OGTT) [6], indicating that SNPs also affect postprandial glucose metabolism. However, Berry et al. (2020) have recently shown that genotypes play a minor role as predictors of the postprandial response to a standardized meal challenge [1].
The postprandial 2-hour glucose level is frequently used as a clinical parameter for the classification of disturbances of glucose metabolism and is of diagnostic value for T2DM. In this study, we focus on the glucose area under the curve (gAUC) as the primary outcome as an approximation of glucose metabolism and evaluate the genetic contribution to its
In this narrative synthesis, gene loci were included, for which at least three reports were available (49 reports) ( Figure 1). This restriction of gene loci was crucial to increase the informative value and to reduce the presentation of single, not-replicated findings. Information on gene loci, for which one (68 gene loci) or two reports (15 gene loci) were available, are presented in Supplementary Tables S1 and S2.

Study Quality Assessment
The results of the quality assessment are shown in Figure 2. No report was rated as low quality. The quality of 23 reports was judged to be intermediate, since information on the power calculation, correction for multiple testing, adjustment, and/or ethnicity was missing. The remaining 26 studies were rated as high quality (Figure 2).
The association between the SNP rs12255372 and the gAUC was investigated in five different cohorts (Table 1) [18,20,47,58]. Homozygous carriers of the minor allele (T) showed a significantly higher gAUC compared to heterozygous carriers and the wild-type (p = 0.04) in 1697 participants from the Ely study [20]. Similar results were found in 1538 Finnish men, where homozygous and heterozygous carriers of the minor allele (T) showed a higher gAUC than the wild-type (p = 0.039) [47]. These results could not be replicated in the cohort of the Amish Family Diabetes Study [18], the non-diabetic offsprings of persons with T2DM [47], or participants without a family history of T2DM (p > 0.05) [58].
The SNP rs7903146, which is in high LD (r 2 > 0.8) with the SNP rs12255372, was examined in different cohorts (Table 1) [12,13,18,20,21,[41][42][43][44][45][46]48,58]. While in eight cohorts, no significant difference between the genotypes and gAUC was found [12,13,18,[42][43][44]46,58], there was a statistically significant difference between the genotypes in two cohorts. In the Ely study, homozygous carriers of the minor allele (T) showed a significantly higher gAUC compared to heterozygous carriers and the wild-type (p = 0.013) [20]. A significantdifference was found between homozygous and heterozygous carriers of the minor allele (T) compared to the wild-type in 1065 participants of the TÜF cohort (p = 0.001) [21]. In two cohorts, the results for an association between this SNP and gAUC were inconsistent, depending on the selection of participants or the calculation method of the gAUC [41,45,48]. In the first cohort of 120 persons without diabetes, homozygous carriers of the minor allele (T) had a significantly higher gAUC than the wild-type [41,45]. A similar result was found for women (p < 0.05), while no association was found for men (p > 0.05) [41]. In the second cohort, carriers of the minor allele (T) had a significantly higher gAUC compared to the wild-type (Table 1) [48].
In a cohort of Han-Chinese participants, the findings were inconsistent, depending on the included participants or the genetic model [35]. While a significant difference between the genotypes in 667 normoglycemic participants was found in the additive (p = 0.006) and dominant (p = 0.007) models, no difference was observed for the gAUC between the genotypes in the recessive model. However, the significance disappeared after the correction for multiple testing. Independent of the genetic model, no significant association between SNP rs5215 and the gAUC was found in 458 participants with impaired glucose tolerance and impaired fasting glucose [35]. No association was found between genotypes and the gAUC in 669 participants from the Quebec Family Study (Table 3) [13].
An association between SNP rs5219, which is in high LD (r 2 > 0.8) with SNP rs5215, and the glycemic response to glucose was investigated in five cohorts [33][34][35][36]. No significant difference in the gAUC was observed in four cohorts [34][35][36]. In 298 persons without diabetes, carriers of the minor allele (T) had an increased gAUC compared to the wildtype when using the dominant genetic model (p = 0.04) or by comparing homozygous carriers of the minor allele with the wild-type (p = 0.02) [33]. No significant difference was seen when using the additive model (p = 0.05) [33]. In a subgroup analysis of 75 persons who underwent an OGTT and, in addition, a hyperglycemic clamp, the dominant model resulted in a significantly increased gAUC in carriers of the minor allele (T) compared to the wild-type (p = 0.02) ( Table 3) [33].

Further Genes
Findings for further genes are presented in Table 4. The association between four SNPs within the CDKAL1 gene locus and the gAUC was assessed in four cohorts (Table 4) [12,13,21,28]. For the most examined SNP rs7754840, a significant difference in the gAUC was found between homozygous and heterozygous carriers of the minor allele (C) and the wild-type in 846 participants from the EUGENE2 study (p = 0.016) [28]. Similar findings were found for 1065 participants from the TÜF cohort (p = 0.02) [21], while no significant difference between the genotypes and the gAUC was found for 3367 participants without diabetes from the METSIM cohort [28]. In the Quebec Family Study with 669 participants, the rs7756992, which is in a high LD (r 2 > 0.8) with the SNP rs7754840, was not associated with the gAUC [13].
An association between the HNF4α gene locus and the glucose response was studied in three cohorts (Table 4) [17,19,31]. Out of six SNPs, four SNPs were investigated in one cohort and showed no significant association [17,19,31]. SNP rs1884614 was examined in 689 participants from the Amish Family Diabetes Study [19] and 4430 participants from the Inter99 Study [31]. In both cohorts, homozygous and heterozygous carriers of the minor allele (T) showed significantly different gAUC than the wild-type. While a significant difference was seen in the Amish population with the additive genetic model (p = 0.022) [19], no difference was seen in the Danish cohort in the additive (p = 0.05) as well as in the recessive genetic model (p = 0.21) [31]. Associations between SNP rs1885088 and the gAUC were investigated in the Inter99 Study [31] as well as in the Quebec Family Study [17]. In both cohorts, no significant difference was observed between the genotypes. In a sub-analysis within the Quebec Family Study, homozygous carriers of the minor allele (A) with a high physical activity level showed significantdifferences in the gAUC than heterozygous carriers (p = 0.01) or the wild-type (p = 0.01) [17]. No association was detected in participants with a low physical activity level (Table 4) [17].

Study Quality Assessment
The results of the quality assessment are shown in Figure 2. No report was rated as low quality. The quality of 23 reports was judged to be intermediate, since information on the power calculation, correction for multiple testing, adjustment, and/or ethnicity was missing. The remaining 26 studies were rated as high quality (Figure 2). Figure 2. Quality assessment of genetic association studies [11]. The quality ratio was rather high (green), intermediate (yellow), or low (red). Abate et al. 2003 [49]; Baratta et al. 2003 [37]; Baratta
Most reports investigated an association between TCF7L2 SNPs (rs12255372 and rs7903146, LD r2 > 0.8) and the gAUC [12,13,18,20,21,[41][42][43][44][45][46][47][48]58,60]. For both SNPs, reports based on the biggest sample sizes (SNP rs12255372: Ely study: 1697 participants [20], 1538 Finnish men [47], SNP rs7903146: Ely study: 1697 participants [20], TÜF cohort: 1065 participants [21]) found a significantly higher gAUC in carriers of the minor allele (T) compared to heterozygous carriers and/or the wild-type. However, for the TÜF cohort, no information about any statistical adjustment was given [21]. In contrast, no statistical significance was found in most of the smaller cohorts, including sample sizes between 18 and 721 participants [12,13,18,21,[41][42][43][44][45][46][47][48]58,60]. These results indicate that the SNPs rs12255372 and rs7903146 may modify the gAUC after an OGTT. However, false-positive results cannot be excluded since the statistical power to detect significant associations between the SNPs and gAUC is unknown. There is some evidence from GWAS, that were excluded from this narrative synthesis, that the TCF7L2 gene locus influences glucose metabolism not only in the fasting state [6,61] but also in the post-challenge phase [6]. A meta-analysis of several GWAS, including 15,234 participants without diabetes, showed that the SNP rs7903146 was associated with fasting glucose and 2-h glucose level after an OGTT [6]. However, no association could be found between the SNP rs7903146 and the AUC ratio of insulin to glucose [6]. Similar findings were found for an association between the PPARγ SNP rs1801282 and the gAUC [14,15,22,24,25,[37][38][39][40]59]. For example, in the Sapphire cohort with 1713 participants, significant differences were found when comparing homozygous and heterozygous carriers of the minor allele (G) and the wild-type [24]. Nevertheless, in most cohorts, no significant association between rs1801282 and gAUC was found, possibly due to small sample sizes or different ethnicities. A meta-analysis with around 32,000 participants without diabetes revealed no evidence for an association between SNP rs1801282 and the 2-h glucose level; however, data on gAUC were not reported [62]. In addition, this meta-analysis revealed an association between the SNP and fasting glucose in participants with obesity [62]. To the best of our knowledge, there is no evidence so far for an association focusing on postprandial glucose trajectories.
All analyses investigating the association between KCNJ11 SNPs and gAUC were based on cohorts with less than 1000 participants [12,13,[33][34][35][36]. For the most frequently assessed SNP rs5219, one report with 298 participants stated that carriers of the minor allele (T) had an increased gAUC compared to the wild-type [33]. However, the significance disappeared in the additive genetic model. Considering other weaknesses such as low sample sizes, different ethnicities, and missing correction for multiple testing, there is little evidence for a clinically relevant association between SNPs rs5215 or rs5219, and differences in gAUC after an OGTT. In addition, no data from GWAS for an association between the KCNJ11 gene locus and gAUC are available.
The eligible articles included data from the glucose response after a standardized 75 g OGTT in participants without diabetes. Potential confounding factors, e.g., age and BMI, were not considered mandatory for inclusion in this systematic review. Nevertheless, reports investigating the association between SNPs in the TCF7L2, PPARγ, as well as KNCJ11 gene loci and the gAUC were based on participants with a BMI less than 30 kg/m 2 . Furthermore, most of the identified articles considered potential confounders in the adjustment procedure. However, the following differences between reports were obvious: frequency of plasma glucose measurement during the OGTT (every 10 min up to every hour), duration of the OGTT (120 min up to 300 min), sample size, ethnicity, and statistical methods (genetic model, adjustment, power calculation, and correction for multiple testing). Thus, the comparability between eligible reports might be limited not only by the high variability of SNPs investigated but also by these confounders.
Several explanations for the given negative findings exist: firstly, the missing power to detect small effect differences among the genotypes. To detect small genetic effects on the metabolic response, cohorts with large sample sizes are needed. This was the reason for the establishment of large international consortia, namely, to be able to combine genetic data for the identification of SNPs with rather small effect sizes [63,64]. Out of the 39 different cohorts identified in our analysis, only 4 cohorts were found with a sample size above 1000 participants, which is not comparable to genetic association studies with more than, e.g., 35,000 persons [63]. Nevertheless, GWAS investigating the association between SNPs and the gAUC after an OGTT could not be identified, whereas data on GWAS regarding the association with 2-h postprandial glucose levels are frequently found [6,65].
Secondly, other factors with a greater effect on gAUC might have masked any genetic effect. The Personalized Responses to Dietary Composition (PREDICT) study revealed that factors such as meal composition have a greater effect on the gAUC after a meal challenge than the genotype (15.4% vs. 9.5%) [1]. Addiotionally, the assessment of the association between SNPs and gAUC after an OGTT was not the primary aim of most studies, and usually, a post-hoc analysis was performed. Moreover, due to the missing clinical endpoint of the gAUC, the clinical relevance of the investigated association is difficult to determine.
Furthermore, the most frequently studied gene loci, TCF7L2 [66][67][68], PPARγ [69,70], and KCNJ11 [71,72] are candidate genes for T2DM predisposition. This hypothesis-driven approach, with identified candidate genes, turned out to be of limited value in predicting people with early disturbances in glucose metabolism. It is rather likely that other gene loci or combinations thereof may also play a role for the metabolic response after an OGTT. The gastric inhibitory polypeptide receptor (GIPR) gene locus is one of the known genes to affect the metabolic response after an OGTT [6]. The GIPR SNP rs10423928 was associated with the 2-h glucose level and the AUC ratio of insulin and glucose after an OGTT in participants without diabetes [6]. However, the association between the GIPR gene locus and the gAUC could not be identified in any eligible article of this systematic review.
Finally, no main single effect of an SNP on gAUC after an OGTT was found. Therefore, it may be worthwhile to study the effect of a combination of SNPs. In several studies, the association between a polygenetic risk score and gAUC after an OGTT was analyzed [73][74][75][76][77][78]. Depending on the chosen gene loci for the calculation of a risk score, both significant [73,76,77] and non-significant [24,74,75,78] differences were found for gAUC per risk allele. Therefore, research on polygenic risk scores might be more meaningful to evaluate a genetic effect on the metabolic response after an OGTT. So far, most candidate genes for T2DM or gene loci known to interfere with glucose metabolism were used for the calculation of the genetic risk score [73][74][75][76][77][78]. Machine learning approaches and artificial intelligence measures open further possibilities for a more comprehensive understanding of the genetic contribution to metabolic responses after an OGTT. Genome-wide polygenic risk scores may be even more promising in this context [79].

Strengths and Limitations
This systematic review focused on OGTT as the standard method to characterize glucose metabolism. For all included reports, the methodological quality of genetic associations was assessed and presented. This systematic review is limited by focusing on SNPs and by excluding other genetic variants such as copy number variations and haplotypes. Findings are based on hypothesis-driven approaches, including candidate genes. As the gAUC is not a clinical parameter with a defined diagnostic or clinical value, no assessment of the clinical effect can be made. Furthermore, in most of the included cohort studies, the performance of the OGTT was for the classification of participants according to their glucose metabolism, e.g., normoglycemic or diabetic, rather than on the primary or secondary outcomes. This systematic review is focused on persons without diabetes to address the research gap on the association between SNPs and metabolic response on an OGTT in healthy persons to follow the current discussion on the inter-individual variation of metabolic response in a standardized meal challenge as a predictor for personalized nutritional recommendations. Therefore, the considered sample sizes are rather small and a conclusion on gender-specific results was not possible. A narrative synthesis, as indicated in PROSPERO, was conducted since data pooling and performing a meta-analysis were not considered to be appropriate.

Conclusions
In this systematic review, which is based on candidate gene analyses, heterogeneous findings for the association between SNPs and the gAUC after an OGTT in participants without diabetes were detected. The most investigated genetic loci (TCF7L2, PPARγ, and KCNJ11) are known to increase the risk of developing T2DM and have shown single findings for a significant association with gAUC. Therefore, more robust data, including data from hypothesis-free approaches, are needed to exploit the genetic contribution to personalized nutrition.
Supplementary Materials: The following supporting information can be downloaded at: https:// www.mdpi.com/article/10.3390/nu15071695/s1, Table S1: Identified genes for an association between SNPs and gAUC after an OGTT in adults; Table S2: Associations between SNPs and gAUC after an OGTT in adults. References

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