Abnormal glucagon secretion contributes to a longitudinal decline in glucose tolerance

Abstract

Context:

Defects in insulin secretion and action contribute to the progression of prediabetes to diabetes. However, the contribution of α-cell dysfunction to this process has been unclear.

Objective:

Understand the relative contributions of α-cell and β-cell dysfunction to declining glucose tolerance.

Design:

Longitudinal, community-based observational study

Setting:

Clinical Research Unit at an Academic Medical Center

Patients / Participants:

We studied 96 subjects without diabetes (55 ± 1 years; BMI 27.7 ± 0.4 Kg/M²) on 2 occasions, 3 years apart using an oral 75g glucose challenge. Indices for insulin secretion and action were estimated using the oral minimal model. Glucagon secretion rate (GSR) was estimated by deconvolution from peripheral glucagon concentrations.

Intervention:

This was an observational study.

Main Outcome Measure:

Glucose tolerance status (categorical variable) and then symmetrical percent change in peak and 120-minute glucose (post-OGTT) concentrations (continuous variables).

Results:

32 subjects progressed from normal to Impaired Glucose Tolerance (IGT) or from IGT to type 2 diabetes. Disposition Index (DI) declined in the progressors (568±98 vs. 403±65 10^-4 dl/kg/min per μU/ml, baseline vs. 3-years p=0.04). α-cell suppression by glucose (δGSR/δglucose) did not change in the non-progressors (1.5±0.1 vs. 1.3±0.1 nmol/min/L, p=0.37) but decreased (1.0±0.2 vs. 0.8±0.2 nmol/min/L, p<0.01) in those who progressed. Analysis of the entire cohort showed that DI and δGSR/δglucose were independently and inversely correlated with an increase in glycemic excursion.

Conclusions:

These data show that α-cell dysfunction accompanies a decline in β-cell function as IGT or overt type 2 diabetes develops.

Introduction

While loss of β-cell function is central to the metabolic abnormalities seen in type 1 diabetes (1), type 2 diabetes is characterized by both α-cell and β-cell dysfunction (2). It has been widely believed that defects in insulin secretion were necessary for α-cell dysfunction to occur. Moreover, accompanying insulin deficiency was thought to be a necessary condition for dysregulation of glucagon to have any material effect on glycemic control (3). This led to the belief that glucagon dysregulation is a late contributor to the transition from prediabetes to diabetes. That view has been challenged somewhat by findings that have accrued over the past decade (as discussed below).

Large observational studies have suggested that abnormalities of glucagon secretion in the post-prandial period occur independently of changes in insulin secretion although they may be associated with defects in insulin action (4). Analysis of fasting, frequently measured, insulin and glucagon concentrations also suggest independent regulation of their secretion (5). Moreover, people with impaired fasting glucose have been shown to have relatively intact β-cell function in the presence of abnormal suppression of α-cell secretion by glucose (6).

Taken together these data suggest that functional abnormalities of the principal components of the islet can contribute independently to the pathogenesis of diabetes. However, whether this truly occurs in the early development of abnormal glucose homeostasis remains uncertain. One supporting piece of data is the observation that the diabetes-associated allele at rs7903146 in the TCF7L2 locus is associated with impaired α-cell function (7). If this common variant with the largest effect size (amongst common genetic variants) on disease predisposition alters glucagon secretion it would imply that glucagon contributes to the early pathogenesis of prediabetes (8).

Multiple studies have shown that people with one or two copies of the diabetes-associated (T) allele have decreased insulin secretion compared to those with the diabetes-protective allele (9). However, we showed that in subjects without type 2 diabetes, the TT genotype at rs7903146 is also associated with higher glucagon concentrations (7,10) in addition to, the previously described, impaired β-cell function. These cross-sectional studies however provided no information as to the order in which these defects occurred.

We therefore undertook a longitudinal study that used variation at rs7930146 to investigate the temporal relationship of changes in β-cell and α-cell function that contribute to type 2 diabetes (11). We showed that in subjects homozygous for the diabetes-associated allele (TT genotype) at rs7903146 the decline in glucose tolerance over the period of study was not explained by impaired β-cell responsivity to glucose. Instead, nadir glucagon concentrations were inappropriate for the hyperglycemia present. We concluded that, at least initially, genetic variation in TCF7L2 contributes to the development of diabetes via α-cell dysfunction.

These observations may have important corollaries. The conventional oral minimal model measures insulin secretion and action in response to an oral challenge (12). Expressing insulin secretion as a function of the prevailing insulin action in the Disposition Index (DI) enables quantitation of β-cell function (13) and has been used to predict future progression to diabetes. However, the model ignores glucagon secretion and its potential effects on post-prandial glucose metabolism (3,14). We therefore sought to determine whether quantification of glucagon suppression by glucose might help better characterize individuals at risk for developing fasting and postprandial hyperglycemia.

To do so, we reanalyzed the data from our recently published longitudinal study to examine whether a surrogate measure of α-cell dysfunction improves the ability of DI to explain changes in glucose regulation. We first compared islet function in those whose categorical glucose tolerance status worsened (from normal to Impaired Glucose Tolerance (IGT) and from IGT to type 2 diabetes) over the period of observation. A decline in DI was accompanied by impairment in post-challenge glucose suppression of glucagon secretion – the surrogate measure of α-cell dysfunction that we adopted. Subsequently, we demonstrate that changes in glucagon suppression and β-cell function can explain changes in peak post-challenge glucose and glucose tolerance than DI alone.

Methods

Population Studied and Experimental Design

This is a reanalysis of a cohort originally recruited and studied on the basis of their genotype at rs7903146 in the TCF7L2 locus. The results and underlying methodology have been recently published (11). Briefly, after approval from the Mayo Clinic Institutional Review Board, we utilized the Mayo Clinic Biobank, to recruit a new cohort of individuals without diabetes who had the CC or TT genotype at rs7903146. Eligible subjects who expressed interest in participating were invited to the Clinical Research and Trials Unit (CRTU) for a screening visit where written, informed consent was obtained. After a history and physical examination, body composition was measured using dual-energy X-ray absorptiometry (Lunar, Madison, WI). A repeat exam and measurement of body composition was undertaken prior to the second study three years later.

During this interval all participants were contacted by a member of the study team, at 6-month intervals (by phone or e-mail) to review medical history and medications and ensure continued health and the absence of confounding medical illnesses or therapy. The two studies were identical; after an overnight fast (cessation of oral intake other than water at 2000), participants were admitted to the CRTU at 0530 on the day of the study. A dorsal hand vein was cannulated at 0600 and placed in a heated Plexiglas box maintained at 55°C to allow sampling of arterialized venous blood. At 0900 (0 minutes) subjects ingested Jell-O containing 75g of glucose. Blood was collected to allow measurement of glucose and hormone concentrations. At the end of the study (1530, 360 minutes), cannulae were removed; participants consumed a late lunch and left the CRTU.

Analytic Techniques

All blood was immediately placed on ice after collection, centrifuged at 4°C, separated, and stored at −80°C until assay. Plasma glucose concentrations were measured using a Yellow Springs glucose analyzer (Yellow Springs Instruments, Yellow Springs, OH). Plasma insulin concentrations were measured using a chemiluminescence assay (Access Assay, Beckman, Chaska, MN). Plasma C-peptide was measured using a 2-site immunenzymatic sandwich assay (Roche Diagnostics, Indianapolis, IN). Glucagon was measured using a two-site ELISA (Mercodia, Winston Salem, NC) in accordance with the manufacturer’s instructions.

Calculations and Statistical Analysis

Calculations

Net insulin sensitivity (S_i) and β-Cell responsivity to glucose (Ф) were estimated using the oral glucose and the oral C-peptide minimal model respectively (15), incorporating age-associated changes in C-peptide kinetics (16). These models derive their respective indices from the integrated relationship of insulin and glucose concentrations (S_i) and from the relationship of C-peptide and glucose concentrations (Ф). The model also provides an index of basal β-Cell responsivity to glucose (_basal) expressing basal (fasting) C-peptide concentrations as a function of basal glucose concentrations. The Disposition index (DI) for each subject was calculated by multiplying Ф by S_i. Glucagon Secretion Rate (GSR) was calculated from glucagon concentrations using non-parametric deconvolution and the previously described population model of glucagon kinetics. This uses anthropometric characteristics to estimate the volume of distribution and clearance rate for glucagon in a given individual, as previously described (17).

To assess the relationship of glucagon secretion to glucose concentrations we assessed various indices (See (18), Supplementary Figure 1) that expressed GSR as a function of glucose. For the purposes of this manuscript, we utilized the following index: -

$$\delta \mathrm{GSR}/\delta \mathrm{Glucose}(\mathrm{nmol}/\min /\mathrm{l})=\frac{\mathrm{Fasting}\mathrm{GSR}-\mathrm{Nadir}\mathrm{GSR}\left(\mathrm{pmol}\mathrm{per}\min \right)}{\mathrm{Peak}\mathrm{Glucose}-\mathrm{Fasting}\mathrm{Glucose}\left(\mathrm{mmol}\mathrm{per}\mathrm{liter}\right)}$$

Statistical Analysis

All continuous data are summarized as means ± SEM. Area Under the Curve (AUC) and Area Above Basal (AAB) were calculated using the trapezoidal rule. Within-group differences (baseline vs. 3-year visit) were assessed using a two-tailed Student’s paired t-test (parametric) or Wilcoxon matched-pairs signed rank test (non-parametric). We also divided the cohort on the basis of whether changes in glucose concentrations at 120 minutes would have reclassified subjects on the basis of their glucose tolerance status i.e.: progression from normal to IGT or from IGT to type 2 diabetes. To assess between-group differences (progressors vs. non-progressors), we used a two-tailed Student’s unpaired t-test (parametric) or a Wilcoxon test (non-parametric). In addition, to examine changes over time independently of their distribution and underlying baseline value, we calculated the symmetric percent change (19) as 100*Log_e (3-year visit / Baseline visit). BlueSky Statistics software v. 10.3.4 (BlueSky Statistics LLC, Chicago, IL, USA) and Prism 8.0.1 (GraphPad Software, San Diego, CA) were utilized for the statistical analysis. A p-value <0.05 was considered statistically significant.

Results

Subject Characteristics at baseline and then at 3 years of follow-up (Table 1, 2)

Table 1:

Group characteristics at baseline and after 3 years of follow-up.

Characteristics	Baseline	3 Years	P-Value*
N	96
Age (Years)	55 ± 1
Sex (Male / Female)	31 / 65
HbA1c (%)	5.3 ± 0.1
BMI (Kg/M2)	27.7 ± 0.4	28.7 ± 0.5	< 0.01
LBM (Kg)	47.6 ± 1.1	48.6 ± 1.0	< 0.01
TBM (Kg)	80.9 ± 1.5	83.6 ± 1.6	0.37
Fasting Glucose (mmol/L)	5.2 ± 0.1	5.2 ± 0.1	0.61
Peak glucose (mmol/L)	10.7 ± 0.2	10.9 ± 0.2	0.16
120-minute glucose (mmol/L)	8.2 ± 0.2	8.8 ± 0.02	< 0.01
AAB glucose (mmol/L per 6hr)	424 ± 23	480 ± 27	0.01
Si (10−4 dL/kg/min per μU/mL)	10 ± 1	8 ± 1	< 0.01#
ϕbasal (10−9min−1)	7.8 ± 0.4	8.7 ± 0.4	< 0.01
ϕdynamic (10−9)	741 ± 35	742 ± 40	0.50
ϕstatic (10−9min−1)	44 ± 2	42 ± 2	0.54
Φ (10−9min−1)	54 ± 2	51 ± 2	0.52
DI (10−14 dL/kg/min2 per pmol/L)	873 ± 67	702 ± 99	< 0.01#
Fasting glucagon (pmol/L)	6.9 ± 0.4	6.9 ± 0.4	0.73
Nadir glucagon (pmol/L)	2.5 ± 0.2	2.6 ± 0.2	0.18
Fasting GSR (pmol/min)	11 ± 1	11 ± 1	0.81
Nadir GSR (pmol/min)	3 ± 1	4 ± 1	0.58
δ GSR / δ Glucose (nmol/min/L)	7.1 ± 0.5	6.9 ± 0.5	0.99

^*Represents results of a paired two-tailed Student’s t-test, except for ^#representing a Wilcoxon matched-pairs signed rank test. BMI = Body Mass Index; LBM = Lean Body Mass; TBM = Total Body Mass; AAB = Area Above Basal; Si = Insulin Action; ϕ_basal = β-cell responsivity to fasting glucose; ϕ_dynamic = β-cell responsivity to changing glucose concentrations; ϕ_static = β-cell responsivity to static glucose concentrations; Φ = Net β-cell responsivity to glucose; DI = Disposition Index; GSR = Glucagon Secretion Rate.

Table 2:

Group characteristics at baseline and after 3 years of follow-up.

Characteristics	Non-Progressors		Progressors		P-Value*
	Baseline	3 Years	Baseline	3 Years
N	64		32
Age (Years)	54 ± 2		57 ± 2		0.31
Sex (Male / Female)	20 / 44		11 / 21
HbA1c (%)	5.3 ± 0.1		5.4 ± 0.1		0.09
BMI (Kg/M2)	28 ± 1	29 ± 1	28 ± 1	29 ± 1	0.88
	< 0.01#		0.05#
LBM (Kg)	47 ± 1	48 ± 1	50 ± 2	50 ± 2	0.85
	< 0.01#		0.05#
TBM (Kg)	81 ± 2	84 ± 2	81 ± 2	83 ± 2	0.51
	0.74#		< 0.01#
Fasting Glucose (mmol/L)	5.2 ± 0.1	5.2 ± 0.1	5.3 ± 0.1	5.4 ± 0.1	0.03
	0.39#		0.07#
Peak glucose (mmol/L)	10.5 ± 0.2	10.5 ± 0.2	11.4 ± 0.3	12.1 ± 0.3	< 0.01
	0.72#		0.01#
120-minute glucose (mmol/L)	7.9 ± 0.2	8.2 ± 0.2	8.9 ± 0.4	10.3 ± 0.3	< 0.01
	0.53#		< 0.01#
AAB glucose (mmol/L per 6hr)	406 ± 26	453 ± 31	472 ± 47	553 ± 46	0.02
	0.66#		< 0.01#
Si (10−4 dL/kg/min per μU/mL)	11 ± 1	9 ± 1	7 ± 1	5 ± 1	0.24
	0.75#		0.03#
ϕbasal (10−9min−1)	7.8 ± 0.4	8.5 ± 0.5	7.9 ± 0.7	9.3 ± 0.8	0.51
	0.02#		0.01#
ϕdynamic (10−9)	754 ± 44	745 ± 48	704 ± 60	735 ± 73	0.92
	0.78#		0.41#
ϕstatic (10−9min−1)	46 ± 2	43 ± 2	39 ± 2	49 ± 4	0.62
	0.44#		0.80#
Φ (10−9min−1)	56 ± 3	52 ± 3	49 ± 3	49 ± 4	0.26
	0.55#		0.80#
DI (10−14 dL/kg/min2 per pmol/L)	986 ± 87	812 ± 14	568 ± 98	403 ± 65	0.32
	0.94#		0.04#
Fasting GSR (pmol/min)	11.1 ± 0.8	10.5 ± 1.0	8.8 ± 1.2	11.1 ± 1.2	0.05
	0.85#		0.32#
Nadir GSR (pmol/min)	3.6 ± 0.3	3.8 ± 0.4	3.1 ± 0.5	3.6 ± 0.5	0.52
	0.85#		0.99#
δ GSR / δ Glucose (nmol/min/L)	1.5 ± 0.1	1.3 ± 0.1	1.1 ± 0.2	0.8 ± 0.2	< 0.01
	0.22#		< 0.01#

^*Represents results of an unpaired two-tailed Student’s t-test applied to the symmetrical % change between each group with the exception of baseline age and HbA1c.

^#Represents value for a paired t-test or Wilcoxon matched-pairs signed rank test. BMI = Body Mass Index; LBM = Lean Body Mass; TBM = Total Body Mass; Si = Insulin Action; ϕ_basal = β-cell responsivity to fasting glucose; ϕ_dynamic = β-cell responsivity to changing glucose concentrations; ϕ_static = β-cell responsivity to static glucose concentrations; Φ = Net β-cell responsivity to glucose; AAB = Area Above Basal; DI = Disposition Index; GSR = Glucagon Secretion Rate.

A total of 96 subjects completed the study. Over the period of observation there was no net change in mean fasting or peak post-challenge glucose. Lean body mass and BMI increased over the duration of the study. There were no overall changes in post-challenge indices of insulin secretion and action or in glucagon suppression. However, _basal – an index of fasting insulin secretion relative to the prevailing glucose concentration – rose over the period of study.

We subsequently divided the cohort on the basis of whether changes in glucose concentrations at 120 minutes would have reclassified subjects on the basis of their glucose tolerance status (Table 2). Overall, 19 subjects progressed from normal to IGT and 13 subjects progressed from IGT to type 2 diabetes (a total of 32 progressors). Anthropometric variables did not differ between progressors and those with stable glucose tolerance at baseline.

Baseline and 3-year Glucose, Insulin, C-peptide and Glucagon concentrations in Progressors vs. Non-Progressors (Figure 1, Table 2)

Figure 1:

Fasting and post-challenge glucose (Panel A), insulin (Panel C), C-peptide (Panel E) and glucagon (Panel G) concentrations in people whose glucose tolerance status was unchanged at baseline (open circles) and after 3 years (closed circles) of follow-up. The right columns show fasting and post-challenge glucose (Panel B), insulin (Panel D), C-peptide (Panel F) and glucagon (Panel H) concentrations in people whose glucose tolerance status deteriorated at baseline (open diamonds) and after 3 years (closed diamonds) of follow-up. Values plotted are Means ± SEMs.

Fasting glucose concentrations did not differ at baseline between the two groups. However, peak glucose concentrations were higher at baseline in the progressors. Other measures of glucose did not differ between the two groups at baseline (Table 2). Fasting, peak and integrated glucose concentrations did not differ over the 3 years of observation in the non-progressor group (Panel A, Table 2). In contrast, peak and integrated (AAB – Area Above Basal) increased in the progressors (Panel B, Table 2). Fasting glucose concentrations did not change in either group.

Fasting and peak insulin concentrations did not differ significantly at baseline between the two groups. In the non-progressor group, insulin concentrations did not change significantly over the duration of the study (Panels C). In contrast, fasting insulin concentrations increased slightly but significantly in the progressor group (30 ± 4 vs 36 ± 4 pmol/l, p < 0.01, Panel D). Although peak insulin concentrations did not change, AAB increased slightly, but not significantly (51 ± 6 vs 56 ± 6 nmol/l per 6hr, p = 0.07, Panel D) in this group.

Fasting C-peptide concentrations did not differ at baseline between the two groups. However, peak concentrations were higher at baseline in the progressors. Peak and integrated C-peptide concentrations (AAB) increased slightly but significantly in the non-progressor group (Panel E). In the progressor group, fasting C-peptide as well as peak and integrated concentrations increased (Panel F).

Fasting and nadir glucagon concentrations did not differ significantly at baseline between the two groups. Fasting and nadir glucagon concentrations did not change in the non-progressor group (Panel G). There was a slight rise in fasting glucagon (7.0 ± 0.7 vs. 8.3 ± 0.8 pmol/l, p = 0.05, Panel H), although nadir glucagon did not change in the progressor group.

Baseline and 3-year indices of islet function in Progressors vs. Non-Progressors (Figure 2, Table 2)

Figure 2:

Individual values for β-cell function at baseline and at 3 years (DI - Panels A, B) in non-progressors and progressors respectively. Fasting and post-challenge glucagon secretion rates are shown at baseline and at 3 years in the non-progressors (open and closed circles respectively – Panel C) and in the progressors open and closed diamonds respectively – Panel D). Individual values for δ GSR / δ Glucose at baseline and at 3 years (Panels E, F) in non-progressors and progressors respectively. Bars represent Means ± SEMs. *P < 0.05.

Baseline Φ and S_i as well as DI did not differ at baseline between the two groups (Table 2). S_i decreased in the progressor group, and this was accompanied by a decrease in DI (Panel B, Table 2).

Baseline fasting and nadir GSR did not differ between groups (Table 2). Both fasting and nadir GSR did not change significantly in either group (Panels C & D, Table 2).

We utilized δ GSR / δ Glucose as a measure of α-cell responsiveness to glucose and compared between-group baseline values. These did not differ (Panels E & F, Table 2). However, over the 3-year period of observation, α-cell responsiveness to glucose decreased in the progressor group (Panel F, Table 2) but not in the non-progressor group (Panel E, Table 2).

Changes in Islet Function and their correlation with changes in peak post-challenge glucose concentrations (Figure 3)

Figure 3:

Correlation of the symmetrical percent change in peak glucose with symmetrical percent change in DI (Panel A) and with δ GSR / δ Glucose (Panel B).

We used the symmetrical percent change in peak post-challenge glucose to quantify changes in glucose tolerance over time. This was inversely correlated with symmetrical percent change in DI (Panel A, r = −0.62). Similarly, the changes in post-challenge peak glucose were also inversely correlated with symmetrical percent change in our index of α-cell function (Panel B, r = −0.37). A multivariate linear regression also showed an inverse correlation with both variables (r = −0.69, p < 1 × 10⁻¹⁰).

Changes in Islet Function and their correlation with changes in 120-minute post-challenge glucose concentrations (Figure 4)

Figure 4:

Correlation of the symmetrical percent change in 120-minute glucose with symmetrical percent change in DI (Panel A) and with δ GSR / δ Glucose (Panel B).

We also used the symmetrical percent change in 120-minute post-challenge glucose to quantify changes in glucose tolerance over time. This was also inversely correlated with symmetrical percent change in DI (Panel A, r = −0.49). Similarly, the changes in 120-minute post-challenge glucose were also inversely correlated with symmetrical percent change in our index of α-cell function (Panel B, r = −0.30). A multivariate linear regression also showed an inverse correlation with both variables (r = −0.54, p = 2.6 × 10⁻⁷).

Baseline measures of Islet Function and their correlation with changes in peak and 120-minute post-challenge glucose concentrations (Figure 5)

Figure 5:

Correlation of baseline DI with peak (Panel A) and 120-minute (Panel B) glucose at 3 years and the correlation of baseline δ GSR / δ Glucose with peak (Panel C) and 120-minute (Panel D) glucose at 3 years.

To examine the potential utility of baseline measures of islet function to predict future glucose tolerance we correlated baseline DI and δ GSR / δ Glucose with peak (Panels A & C) and 120-minute (Panels B & D) post-challenge glucose concentrations at 3 years.

While DI at baseline was weakly correlated with peak (Panel A, r = −0.39) and 120-minute (Panel B, r = −0.41) glucose at 3 years, δ GSR / δ Glucose was not correlated with glucose concentrations 3 years later (Panels C & D).

Discussion

Reanalysis of the data generated from our recently published study examining the longitudinal effects of diabetes-associated genetic variation at rs7903146 supports the thesis that a decline in α-cell function contributes to the pathogenesis of prediabetes and type 2 diabetes (11). In this cohort, there was no significant longitudinal rise in fasting glucose, however, in the group as a whole there was an increase in _basal implying an increase in fasting insulin secretion despite no changes in glucose (15). This might be a manifestation of decreased insulin action observed over time in the whole cohort. However, this was not accompanied by significant changes in fasting glucagon secretion.

For our initial exploratory analysis, we divided the cohort on the basis of a categorical change in glucose tolerance over time. Intriguingly, at baseline the group that eventually experienced a decrease in glucose tolerance was distinguished from the stable group by a higher post-challenge peak glucose (as has been shown before (20) – Figure 1). The indices of α-cell and β-cell function that we utilized did not clearly differ between the groups at baseline. However, the group which experienced a decline in (categorical) glucose tolerance over the duration of the study exhibited a decline in β-cell function as well as an independent decline in α-cell responsiveness to glucose (Figure 2).

This approach has a significant limitation in that categorical classification of glucose tolerance by an oral glucose challenge is subject to poor reproducibility and a significant incidence of reclassification (21). On the other hand, it provides additional evidence that defective α-cell suppression contributes to worsening glucose tolerance. To overcome the limitations of such categorization we subsequently utilized the whole cohort to examine whether changes (from baseline values) in peak and 120-minute post-challenge glucose correlated with changes in DI and in δ GSR / δ Glucose. This was indeed the case; the combined indices correlating strongly with change in glucose tolerance over the duration of the study (Figures 3 and 4).

Previously, in a slightly younger cohort with otherwise similar anthropometric characteristics, we had suggested that fasting glucagon concentrations predicted a decline in β-cell function and glucose tolerance over an average 7-year period of observation (22). This was not the case in our current study where absolute concentrations of fasting glucagon, GSR or the index δ GSR / δ Glucose did not correlate with changes in post-challenge glucose concentrations (Figure 5). There are several potential explanations for this discrepancy. One is that fasting glucose concentrations were higher in the first cohort perhaps implying baseline defects in fasting α-cell function. This was not the case in the current study where duration of follow up was also shorter and therefore less likely to show longitudinal changes in β-cell function.

Another important difference is that the glucagon assay used in the first study detects proglucagon-derived peptide fragments other than glucagon (23,24). This does not apply to the current assay. It is therefore possible that increased circulating concentrations of one or more of these fragments, rather than glucagon, are the actual biomarker of early α-cell dysfunction. Finally, it is noteworthy that in the current cohort approximately one third expressed progressive glucose intolerance over a relatively short time. At least in the progressor group β-cell dysfunction was already quite impaired relative to non-progressors, perhaps decreasing our ability to utilize fasting glucagon as a marker of early β-cell decline.

We have previously shown that, in response to intravenous glucose infusion, the inverse-exponential relationship of GSR to glucose concentration can be used to derive a measure of α-cell responsivity to glucose (6). We also showed that fasting GSR and glucose measured on a separate day exhibited a similar relationship (6). At present, the limitation of our previous measure of α-cell function (G₅₀ = the rise in glucose concentration necessary to suppress GSR by 50%) is that it is not applicable to OGTT data, because the rate of rise in glucose concentrations differs significantly from the controlled rate observed in a graded glucose infusion (6). We tested various measures as surrogates of α-cell responsivity to glucose including fasting GSR / fasting glucose, nadir GSR / peak glucose or δ GSR / δ Glucose (See (18), Supplementary Figure 1). Taking all of this together, we selected the δ GSR / δ Glucose index for our analyses. This index correlates fasting and postprandial changes in glucagon secretion with the reciprocal changes in glucose concentrations and is therefore a reasonable surrogate of α-cell sensitivity to glucose. However, this index (δ GSR / δ Glucose) assumes a linear relationship of GSR to glucose and does not consider the temporal relationship of these changes. Future indices of α-cell response to an oral glucose challenge will need to account for the current limitations.

In our cohort, age at the start of the study and weight changes during the study did not correlate with a decline in glucose tolerance. More importantly, once we account for changes in islet function, rs7903146 genotype has no effect on progression of glucose intolerance. This indirectly strengthens prior work to show that the predisposition to type 2 diabetes conferred by this variant in the TCF7L2 locus is mediated via changes in islet function alone (10,25). It also implies that the results of our study are generalizable to people without diabetes-associated variation at rs7903146.

One final limitation of our study is that of the 128 subjects who completed the first study, 32 did not return for follow-up or were unable to participate in the second study. This analysis is confined to the remaining 96 subjects. However, as previously discussed, the demographic characteristics of those lost to follow-up did not differ from those completing the study (11).

In our study, changes in α-cell function occurred independently of changes in β-cell function as quantified by DI (See (18), Supplementary Figure 2). This is in keeping with prior work suggesting that changes in glucagon concentrations are unrelated to changes in insulin concentrations (6,7,26). In both cross-sectional and longitudinal studies, insulin secretion does not seem to regulate or restrain α-cell function (4,27). Nevertheless, there has been some suggestion that insulin signaling has a role in controlling glucagon secretion. Impaired insulin signaling has been associated with α-cell dysfunction in rodents (28,29) and in humans (27). Although there was no association of insulin action with δ GSR / δ Glucose, insulin action was weakly correlated with fasting glucagon concentrations (See (18), Supplementary Figure 3) in keeping with our previous observations (27).

In summary, we conclude that a decreased α-cell responsiveness to glucose, quantified by the δ GSR / δ Glucose measure that we adopted correlates with a decline in glucose tolerance over a short period of observation. Although this occurs independently of β-cell dysfunction, as quantified by DI, the combined change in both parameters shown stronger correlation with decreasing glucose tolerance than either parameter alone. It remains to be ascertained whether better measures of α-cell dysfunction, alone or in combination with β-cell function can be used to predict progression to prediabetes and type 2 diabetes.

Data and Resource Availability

The datasets generated during and/or analyzed during the current study are available from the corresponding author upon reasonable request. No applicable resources were generated or analyzed during the current study.