Here we performed a GWAS of plasma levels of glucagon during an OGTT in 1899 individuals and identified a previously described type 2 diabetes variant in EYA2, we replicated a previously reported suggestive association with fasting glucagon level and identified a novel genome-wide significant variant, rs4691991 in the MARCH1 locus.
Our results indicate that genetic influence of fasting glucagon levels and suppression of glucagon secretion during an OGTT have a limited impact on type 2 diabetes development. We only identified one of the current known type 2 diabetes risk alleles to associate with glucagon levels during OGTT. The variant in EYA2 was associated with increased glucagon levels at 30 min during the OGTT. EYA2 is involved in DNA repair and has, besides type 2 diabetes, also been reported to associate with triglyceride levels [16] and waist-hip ratio [17].
We plotted all the selected markers in a QQ-plot, to investigate if we had more significant associations of the type 2 diabetes risk increasing alleles to plasma glucagon levels during the OGTT than expected by chance (Additional file 2: Figure S1) and could conclude that it is most probably not the same genetic markers that is associated with type 2 diabetes that also affect glucagon levels.
We were able to replicate one of the suggestive associations with fasting glucagon level reported by Almgren et al. [4]. Importantly, in both this and the present study, the same glucagon assay has been used. One of the variants, rs142179968 in MACF1, for which Almgren et al. [4] found a suggestive association with fasting glucagon, has a MAF of < 0.0005 according to the UCSC Genome Browser (GRCh37/hg19) [18, 19], although it was slightly higher in the meta-analysis of the MDC and PPP-Botnia cohorts described in the paper (MAF = 0.002) [4]. Since we restricted our analyses to include only markers with a MAF > 0.01, this variant was not analyzed in our primary data set. Another variant associated with fasting glucagon in the aforementioned paper, rs140436257, located in TDC, has a reported allele frequency < 0.003, but is ten times more common in the meta-analysis by Almgren et al. (MAF = 0.026). However, this variant was not included in the analysis reported here and neither was the variant rs111485612 with a MAF below 1%.
We identified a genome-wide significant association with 0–30 min glucagon suppression. The closest gene to the signal is MARCH1, which mediates ubiquitination of proteins and promotes their subsequent endocytosis and sorting to lysosomes. Variants in MARCH1 have previously been shown to be nominally associated with BMI and increased risk of type 2 diabetes in the UK Biobank [20]. However, whether this gene or another gene in the locus is responsible for the effect on plasma levels of glucagon remains to be discerned.
We did not find any strong evidence of an impact of previously published type 2 diabetes genetic variants on glucagon levels during OGTT, except for EYA2. Neither did we find a large number of novel genetic variants affecting glucagon levels during OGTT in the GWAS performed here. We have not been able to identify any previous studies investigating the heritability of glucagon levels in the fasting state and during glucose stimulation in other cohorts. The sample size in this study is modest and might be a contributing factor to the lack of strong genetic signals. Larger studies are required to get a more comprehensive picture of the genetic influence on glucagon levels during OGTT. We are also strongly encouraging heritability studies of glucagon levels in family or twin studies in the future, since there is an obvious lack of these kinds of studies published.
A larger sample size would allow for studies of polygenic risk scores (PRS), which might be clinically useful in the future. The PRS could be used to identify individuals genetically prone to deviating glucagon levels and thereby monitor them before any adverse effect manifests.
The accuracy of measurements of plasma glucagon has been debated for many years. Apart from the usual reliability concerns, specificity and sensitivity have been the main issue. Glucagon assays may react with the glucagon-containing products from the gut, glicentin and oxyntomodulin, and with proglucagon 1–61 from the pancreas as well as the gut. Regarding glucagon secretion, inhibition of secretion, e.g. by glucose, is as important as stimulation with e.g. arginine, which puts great demands on the sensitivity of assay systems. In addition, interfering matrix effects have been a problem. The assay used in this study is a radioimmunoassay directed against the C-terminus based on a C-terminal-wrapping antibody, 4305, which grants the assay complete specificity towards molecular forms with a free glucagon C-terminal which characterizes pancreatic glucagon as opposed to the gut derived peptides, as well as a sufficient sensitivity (< 1 pmol/l) to pick up all relevant changes. The assay has been validated in several recent studies against mass spectrometry [21] as well as against sandwich ELISA [22, 23]. However, in patients with major surgical operations on the human GI tract, such as gastrectomy and bypass, the sandwich ELISA has in some cases shown aberrant behavior. The radioimmunoassay (4305) also measures proglucagon 1–61 because of the identical C-terminus, but this component plays a very minor role under normal circumstances and has glucagon-like bioactivity, if any [24]. We are therefore convinced that we have used the ideal assay for measurement of glucagon in this study [11].
A strength of our study is the measurement of the physiological important 30 min glucagon levels during an OGTT. In fact, the strongest genetic influence in our data set, was seen for plasma levels of glucagon at 30 min of the OGTT. However, it is a limitation to the study that we have not been able to validate our suggestive associations at this time point in additional cohorts, since plasma glucagon at 30 min is a rare measurement.