Post–Glucose Load Plasma α-Dicarbonyl Concentrations Are Increased in Individuals With Impaired Glucose Metabolism and Type 2 Diabetes: The CODAM Study

Type 2 diabetes and impaired glucose metabolism (IGM) are associated with the development of microvascular complications and cardiovascular disease (CVD) (1). Prospective studies in diabetes have shown a strong correlation between mean glucose levels, as reflected by HbA_1c, and diabetes complications. However, in recent years, postprandial glucose excursions, rather than fasting glucose concentrations or mean glucose levels, have increasingly been recognized as a contributing factor to the increased risk of vascular complications (2,3). Controlled clinical trials, such as the Diabetes Control and Complications Trial (DCCT) in type 1 diabetes (4) and the UK Prospective Diabetes Study (UKPDS) in type 2 diabetes (5), have established that intensive therapies that reduce HbA_1c levels reduce and delay the development and progression of diabetes-related long-term vascular complications. Posttrial analyses of the DCCT revealed that postmeal glucose levels contribute more to HbA_1c than fasting plasma levels (6), suggesting that postprandial hyperglycemic spikes may accelerate the onset of diabetes complications. Moreover, many epidemiological data support this concept, showing that glucose levels after an oral glucose tolerance test (OGTT) are an independent risk factor for CVD while fasting glucose levels are not or are less so (2,7–11). In addition, postprandial glucose in type 2 diabetes predicts myocardial infarctions (12) and is a risk factor for CVD (13). However, it is largely unknown how these postprandial glucose peaks cause the increased risk for diabetes complications. A possible mechanism through which they have a more damaging effect than high fasting or mean glucose levels might be via the formation of α-dicarbonyls.

The α-dicarbonyls, methylglyoxal (MGO), glyoxal (GO), and 3-deoxyglucoseone (3-DG), are mainly formed as glycolytic intermediates by metabolic conversion of glucose. We and others have shown the importance of α-dicarbonyls in the development of nephropathy (14,15), retinopathy (16), and neuropathy (17,18) and in macrovascular complications (19), using rodent models of diabetes.

We hypothesized that the postprandial glucose concentrations, rather than fasting plasma glucose, determine α-dicarbonyl levels. Although Beisswenger et al. (20) already demonstrated, in a small study of 21 individuals with type 1 diabetes, that both MGO and 3-DG plasma levels increase after postprandial glycemic excursions, the question of whether α-dicarbonyl levels are elevated in individuals with impaired glucose metabolism (IGM) and type 2 diabetes remains unanswered.

In the current study, we investigated, in a large cohort study, whether plasma levels of α-dicarbonyls are elevated in individuals with IGM and type 2 diabetes. Both fasting levels and incremental areas under the curve (iAUCs) obtained from an OGTT were analyzed.

The current study comprises participants from the Cohort on Diabetes and Atherosclerosis Maastricht (CODAM) study, which included 574 participants who were selected from the general population as previously described in detail (21). The CODAM study was designed to investigate cardiovascular and metabolic function and was enriched for IGM status, as previously described (22). Current main analyses were performed on 491 participants who underwent a full OGTT and were not on insulin treatment. The study was approved by the local Medical Ethics Committee of Maastricht University Medical Centre, and all participants gave written informed consent.

The glucose tolerance status of the participants was determined by an OGTT. After an overnight fasting period (10–12 h), study participants underwent a standard 75-g OGTT (82 g dextrose monohydrate; AVEBE, Veendam, the Netherlands) and venous blood was obtained prior to and at 30, 60, and 120 min after the glucose load. Fasting and postload plasma glucose concentrations (in millimoles per liter) were measured with a hexokinase glucose-6 phosphate dehydrogenase method (ABX Diagnostics, Montpellier, France). Fasting and 2-h postload glucose concentrations were used to classify the study participants’ glucose metabolism status (GMS) according to the World Health Organization criteria. Briefly, individuals were classified as having normal glucose metabolism (NGM) when they had normal fasting (<6.1 mmol/L) and 2-h postload (<7.8 mmol/L) glucose concentrations. Individuals with impaired fasting glucose (6.1–7.0 mmol/L), impaired 2-h postload glucose levels (7.8–11.1 mmol/L), or both were classified as having IGM. When individuals had high fasting plasma glucose levels (≥7.0 mmol/L) and/or high 2-h postload glucose levels (≥11.1 mmol/L) or when they used glucose-lowering medication or insulin, they were classified as having type 2 diabetes (22). Individuals with known type 2 diabetes or those with fasting glucose levels >8.5 mmol/L were excluded from undergoing an OGTT.

Plasma levels of α-dicarbonyls were measured in EDTA plasma samples from the OGTT at baseline and 30, 60, and 120 min after the glucose load. Blood samples were collected in EDTA tubes, which were stored on ice prior to blood sampling to ensure rapid cooling of the blood. After withdrawal of the blood sample, tubes were stored on ice immediately and were spun within 2 h at 3,000 rpm, 4°C. Plasma samples were stored at −80°C until analysis. Reversed-phase ultra-performance liquid chromatography–tandem mass spectrometry was used to analyze the plasma samples for MGO, GO, and 3-DG as previously described (23). The interassay variations for MGO, GO, and 3-DG were 4.3, 5.1, and 2.2%, respectively. Current analyses were performed with fasting α-dicarbonyl levels and OGTT iAUCs.

The area under the curve for the OGTT levels of the α-dicarbonyls and glucose was calculated according to the trapezoidal method (24), where baseline (fasting) levels were subtracted from each individual data point to specify the post–glucose load increases. These data are referred to as the iAUC.

Waist circumference and prior CVD were assessed as previously described (25). Questionnaires were used to assess smoking behavior (never, ever, or current smoker) and use of medication (lipid-, glucose-, and blood pressure–lowering medication). Plasma creatinine levels were measured with the Jaffe diagnostic test (Roche Diagnostics, Mannheim, Germany), and the estimated glomerular filtration rate (eGFR) was calculated using the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation (26). Systolic blood pressures were measured in the brachial arteries and twice in both the tibialis posterior and dorsalis pedis arteries of the lower extremities with a standard Doppler device (Mini Dopplex D900; Huntleigh Diagnostics Ltd., Harmelen, the Netherlands) (27). Total and HDL cholesterol and triglycerides were measured in EDTA plasma by enzymatic methods (Roche Diagnostics, Mannheim, Germany) (27).

The general characteristics of the study population were compared across the three groups of GMS. Continuous variables were analyzed with one-way ANOVA, and discrete variables were analyzed with χ² tests. Fasting levels and iAUCs from the OGTT were normally distributed for all α-dicarbonyls. For increase in statistical power, missing values on HbA_1c (3.7% missing), eGFR (0.8% missing), and BMI (0.2% missing) were imputed by creating five data sets using multiple imputation. Skewed variables (fasting plasma glucose, fasting plasma insulin, HOMA of insulin resistance, and triglycerides) were log_e transformed prior to analyses. Two-way repeated-measures ANOVA with Bonferroni correction was used to compare the curves of the α-dicarbonyls during the OGTT between NGM, IGM, and type 2 diabetes over time. Fasting levels and iAUCs of the α-dicarbonyls were compared between the groups using one-way ANOVA with Bonferroni correction. Multiple linear regression analysis was used to investigate potential influence of confounding factors on the association between GMS and α-dicarbonyl levels. The main independent variables were GMS, analyzed as dummies for IGM and type 2 diabetes, fasting glucose levels, iAUC glucose, and HbA_1c, and the main dependent variables were fasting levels and iAUCs of the α-dicarbonyls. Data are presented as standardized regression coefficients (β). Analyses were initially adjusted for age and sex; analyses with fasting glucose levels, iAUC glucose, and HbA_1c were also adjusted for GMS (model 1). Analyses were further adjusted for smoking status, eGFR, log_e triglycerides, total-to-HDL cholesterol ratio, waist circumference, prior CVD, systolic blood pressure, and use of medication (lipid-, glucose-, and blood pressure–lowering medication) (model 2). Finally, analyses for GMS were additionally adjusted for glucose (fasting or iAUC glucose, appropriate to the α-dicarbonyl measure) as a potential source of α-dicarbonyl formation (model 3), and to investigate the extent to which iAUCs were dependent on fasting α-dicarbonyl levels, model 2 was additionally adjusted for fasting plasma levels of the appropriate α-dicarbonyl (model 4). A P value of <0.05 was considered statistically significant.

As increased levels of α-dicarbonyls can be a result of both increased formation and decreased detoxification, we investigated whether individuals with IGM or type 2 diabetes had a decreased capacity to detoxify α-dicarbonyls. To test this, we analyzed whether the association between postload glucose excursions and α-dicarbonyl iAUCs differed between individuals with NGM, IGM, or type 2 diabetes. To this end, the interaction terms IGM × iAUC glucose and type 2 diabetes × iAUC glucose were added to the linear regression model (adjusted for covariates in model 2). P_interaction < 0.1 was considered statistically significant. Significant P values for interaction would indicate a different detoxification capacity of α-dicarbonyls in individuals with IGM or type 2 diabetes for the same change in glucose concentration.

All statistical analyses were performed with IBM SPSS Statistics Software, version 20 (IBM Corporation, Armonk, New York).

General characteristics of the study population stratified according to GMS are shown in Table 1. Overall, participants with type 2 diabetes were slightly older; had a higher BMI, higher blood pressure, and lower HDL cholesterol; and had a higher incidence of CVD and microalbuminuria. Curves of the OGTT in Fig. 1 demonstrate increased levels of glucose and MGO, GO, and 3-DG after the glucose load. Plasma glucose levels were ∼5,000-fold higher compared with the plasma α-dicarbonyl levels.

When we analyzed the fasting α-dicarbonyl levels, we found ∼1.1-fold higher plasma concentrations of MGO in individuals with type 2 diabetes, but not with IGM (Table 1) (P < 0.001), compared with NGM. Fasting levels of 3-DG were increased in both individuals with IGM and individuals with type 2 diabetes compared with those with NGM (Table 1) (both P < 0.001). GO levels at baseline did not differ between the groups (Table 1).

When we additionally included individuals with known type 2 diabetes who were not allowed to undergo an OGTT, we observed significantly higher fasting levels of plasma α-dicarbonyls compared with levels from those who did undergo the OGTT (Supplementary Fig. 1).

iAUCs for the α-dicarbonyls MGO, GO, and 3-DG were significantly higher (1.6-, 2.3- and 1.8-fold change, respectively) in participants with IGM (all P < 0.001) and were even higher (2.1-, 3.4-, and 2.6-fold change, respectively) in those with type 2 diabetes (all P < 0.001) (Fig. 1B, D, and F) compared with NGM. For glucose, both fasting levels and iAUC of the OGTT were increased in individuals with IGM and type 2 diabetes (Table 1 and Fig. 1H) (all P < 0.001).

The presence of IGM was not associated with higher fasting plasma levels of MGO or GO after adjustment for age and sex (Table 2) (model 1). Fasting plasma 3-DG levels in individuals with IGM were 0.255 SD higher than in those with NGM (β = 0.255 [95% CI 0.136–0.375], P < 0.001). Additional adjustment for smoking, eGFR, triglycerides, total-to-HDL cholesterol ratio, waist circumference, medication use, prior CVD, and systolic blood pressure did not materially change this association (Table 2) (model 2). Addition of fasting glucose levels to the regression model completely attenuated the association between the presence of IGM and higher fasting levels of 3-DG (Table 2) (model 3).

In contrast, the presence of type 2 diabetes was associated with higher fasting levels of both MGO (β = 0.542 [95% CI 0.322–0.761], P < 0.001) and 3-DG (β = 1.517 [95% CI 1.385–1.649], P < 0.001) after adjustment for age and sex compared with NGM. After additional adjustment for multiple covariates in model 2, associations between the presence of type 2 diabetes and higher fasting levels of MGO (β = 0.492 [95% CI 0.216–0.769], P < 0.001) and 3-DG (β = 1.376 [95% CI 1.211–1.542], P < 0.001) were largely unchanged. Further adjustment for fasting glucose levels resulted in an 82% attenuation of the association between type 2 diabetes and fasting MGO levels (Table 2) (model 3). In line, the association between type 2 diabetes and fasting 3-DG levels was attenuated by 61% but remained statistically significant (Table 2) (model 3). Also, after adjustment for potential confounders, we found no associations between either IGM or type 2 diabetes and fasting GO levels.

In age- and sex-adjusted analyses, the presence of IGM was associated with significantly higher iAUCs for all three α-dicarbonyls (MGO β = 0.203 [95% CI 0.123–0.283], P < 0.001; GO β = 0.268 [95% CI 0.160–0.377], P < 0.001; and 3-DG β = 0.319 [95% CI 0.257–0.382], P < 0.001), compared with NGM. For type 2 diabetes, the associations with higher iAUCs of all three α-dicarbonyls appeared to be even stronger (MGO β = 0.359 [95% CI 0.271–0.448], P < 0.001; GO β = 0.486 [95% CI 0.366–0.606], P < 0.001; and 3-DG β = 0.645 [95% CI 0.576–0.714], P < 0.001). After further adjustment for the covariates in model 2, both IGM and type 2 diabetes remained significantly associated with higher iAUCs of all three α-dicarbonyls (Table 2) (model 2) compared with iAUCs of individuals with NGM. However, when analyses were further adjusted for the iAUC from glucose, the associations of both IGM and type 2 diabetes with MGO and 3-DG iAUCs disappeared completely (Table 2) (model 3). Similarly, although the association between type 2 diabetes and higher iAUC of GO (β = 0.180 [95% CI 0.001–0.358], P = 0.049) remained significant, it was also attenuated by 66%. Adjustment for fasting levels of the appropriate α-dicarbonyls in model 4 did not change the strength of the association compared with model 2 (Table 3) (model 4 vs. 2).

Since the associations between IGM and type 2 diabetes and higher α-dicarbonyl iAUCs almost completely disappeared after adjustment for glucose, we hypothesized that the higher iAUCs of MGO, GO, and 3-DG were a direct result of higher iAUCs of glucose in IGM and type 2 diabetes rather than a decreased capacity to detoxify α-dicarbonyls. Indeed, when adjusted for all covariates in model 2, the glucose iAUC was strongly associated with the iAUC of all three α-dicarbonyls (Table 3) (MGO β = 0.195 [95% CI 0.148–0.243], P < 0.001; GO β = 0.231 [95% CI 0.165–0.297], P < 0.001; and 3-DG β = 0.342 [95% CI 0.318–0.365], P < 0.001). These associations did not differ between NGM, IGM, or type 2 diabetes for all three α-dicarbonyls (P_interaction > 0.1). Overall, associations between the iAUC of glucose and the iAUC of α-dicarbonyls were stronger than associations between HbA_1c or fasting plasma glucose and the iAUC of α-dicarbonyls (Table 3).

This study demonstrates that iAUCs of the α-dicarbonyls MGO, GO, and 3-DG, as calculated from an OGTT, were higher in individuals with IGM and type 2 diabetes, independently of potential confounders. Fasting plasma levels of α-dicarbonyls were predominantly higher in individuals with type 2 diabetes, although fasting 3-DG levels were also slightly increased in individuals with IGM. After adjustment for glucose, these associations disappeared almost completely for MGO and 3-DG and to a large, but slightly lesser, extent for GO, indicating that glucose is the major source of α-dicarbonyls. To our knowledge, this is the first time that α-dicarbonyls have been measured in a post–glucose load setting in individuals with NGM, IGM, and type 2 diabetes.

In the current study, we found that iAUCs of MGO, GO, and 3-DG, as calculated from an OGTT, were higher in individuals with IGM and type 2 diabetes, independently of potential confounders. Our data are in accordance with findings from a previous study by Beisswenger et al. (20), in which they showed increased MGO and 3-DG levels during the postprandial period in patients with type 1 diabetes. As glucose serves as a primary source for the formation of α-dicarbonyls, transient glucose excursions during the postprandial period may give rise to increases of α-dicarbonyl levels, which in turn may induce long-term damage to the vasculature. Indeed, El-Osta et al. (28) demonstrated that even transient exposures to high glucose levels induce persistent changes in cultured endothelial cells, which could be prevented by an overexpression of glyoxalase I, the major enzyme detoxifying MGO and GO. We and others have demonstrated that increased levels of α-dicarbonyl compounds are directly associated with vascular complications (14,29). MGO particularly has attracted a lot of attention as a key player in vascular dysfunction as a result of its capacity to induce oxidative stress (29), cell death (30), and endothelial dysfunction (14). Therefore, the observed increased levels of α-dicarbonyls with postchallenge glucose excursions in IGM and type 2 diabetes may link fluctuations in blood glucose levels in these patients with persistent increases in risk of vascular complications and CVD.

The α-dicarbonyls in the plasma can originate from various sources, including in situ formation in the plasma, release from cells and external sources (31). Owing to their rapid increase during the OGTT, the postload plasma α-dicarbonyl levels most likely originate from intracellular compartments that come directly into contact with plasma glucose. As we found the largest increase in α-dicarbonyls in the individuals with IGM and with type 2 diabetes, it is likely that plasma α-dicarbonyls originate from insulin-independent cells, specifically, red blood cells and endothelial cells. Indeed, we demonstrated in human endothelial cells that hyperglycemia produced higher levels of MGO (32) and that α-dicarbonyl levels are much higher in circulating cells than in plasma (23). We assume that α-dicarbonyl levels after a glucose challenge would increase even further in circulating cells and endothelial cells, as only a small percentage of dicarbonyl compounds leaks into the circulation (23). To what extent this increase of dicarbonyls in plasma is similar in other tissues that are sensitive to diabetes complications, such as the kidney, nerves, retina, and atherosclerotic plaques, remains to be elucidated. Next to release from cells, plasma α-dicarbonyls may also be a result of autoxidation of glucose in the plasma, but it is not known to what extent this process contributes to plasma levels of α-dicarbonyls. It is unlikely, however, that the iAUCs reflect any autoxidation, as this is a slow process and we observed an increase of α-dicarbonyls already 30 min after the glucose load. In addition, dicarbonyl compounds can also originate from exogenous sources (31), for example, from the glucose drink we used in this study. However, we previously observed no increased plasma MGO levels in a healthy volunteer within 2 h after drinking coffee, a drink with very high levels of MGO (55 µmol/L) but without any glucose. As the glucose drink used for the OGTT only contained 1.6 µmol/L MGO, it is highly unlikely that the low levels of α-dicarbonyls in the glucose drink contribute to their plasma levels after the glucose challenge. The question remains whether the increased α-dicarbonyl iAUCs are a reflection of decreased detoxification potency in IGM and type 2 diabetes. The glyoxalase system is the major pathway to detoxify MGO and GO (33). This system consists of the rate-limiting enzyme glyoxalase-1 and glyoxalase-2. These enzymes convert MGO and GO, with the involvement of reduced glutathione, to their end product d-lactate (33). Several experimental studies have linked the presence of diabetes to dysfunction of the glyoxalase system (34). However, additional adjustment for glucose in our analyses attenuated the associations between GMS and the iAUC of α-dicarbonyls by 60–97%, suggesting that the elevated postchallenge dicarbonyl levels are the result of increased formation from its substrate glucose. Although dysfunction of α-dicarbonyl detoxification cannot be eliminated as a contributor to increased plasma α-dicarbonyl levels, the lack of interaction between dicarbonyl iAUCs and GMS indicates that detoxification mechanisms are not differently active in any of the three GMS groups.

Our new observation that postload glucose levels are closely associated with α-dicarbonyl formation is of high clinical relevance because of the current developments in the field of glucose-lowering therapies. Dipeptidyl peptidase-4 inhibitors form a very new treatment strategy that has been shown to regulate postprandial glucose concentrations (35). Whether dipeptidyl peptidase-4 inhibitors can reduce α-dicarbonyls is unknown. In addition, the bionic pancreas is a state-of-the-art intervention that has been demonstrated to regulate glycemic control very strictly in type 1 diabetes (36) and may prove valuable in type 2 diabetes as well. Directly lowering α-dicarbonyl levels may also be a mechanism to reduce postprandial carbonyl stress and the putative association with vascular damage long-term. One intervention that is currently highly under investigation is pyridoxamine. Pyridoxamine is a chemical scavenger of reactive α-dicarbonyls and has been shown to inhibit formation of AGEs. Several experimental and clinical studies have already demonstrated beneficial effects of pyridoxamine with regard to diabetes microvascular complications (37–39).

The major strengths of this study are that we were able to perform our analyses in a large and well-defined cohort study, and in addition, we measured plasma α-dicarbonyls with state-of-the-art techniques based on ultra-performance liquid chromatography–tandem mass spectrometry. There are also a few limitations of this study. First, individuals with known type 2 diabetes did not undergo an OGTT, while they had higher fasting levels of α-dicarbonyl compounds compared with the patients with type 2 diabetes who did undergo a full OGTT. Therefore, their α-dicarbonyl levels during the OGTT are expected to increase even more than in newly diagnosed subjects, indicating that our observations may be an underestimation of the true effect in type 2 diabetes. Furthermore, the glucose load in the OGTT is not completely comparable with postprandial glucose excursions. For investigation of the effect of postprandial glucose excursions on α-dicarbonyl levels, a mixed-meal test needs to be done, although it has been described that the level of glycemia 2 h after an OGTT is closely related to the level of glycemia after a standardized meal (40). Moreover, a major advantage of the OGTT used in this study is that it allowed us to specifically investigate the hypothesis that glucose spikes cause formation of MGO without confounding of postprandial changes in lipid and protein levels.

In conclusion, we found that significant increases of MGO, GO, and 3-DG levels occurred during an OGTT in individuals with IGM and type 2 diabetes in comparison with control subjects. These increases were strongly associated with postload glucose excursions. These findings, together with the known harmful biological effects of these α-dicarbonyls, underline the potential importance of dicarbonyl stress as a functional candidate to explain the increased risk of diabetes complications in individuals with postprandial hyperglycemia. Prospective analyses on micro- and macrovascular complications are necessary to associate our current findings with vascular outcome.

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

Author Contributions. D.E.M. and N.M.H. researched data, contributed to discussion, and wrote and edited the manuscript. J.L.S. performed laboratory analyses and reviewed and edited the manuscript. C.J.v.d.K., M.M.v.G., C.D.S., and C.G.S. contributed to discussion and reviewed and edited the manuscript. C.G.S. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.