Methylglyoxal‐glycated apolipoproteins A‐I and A‐ <scp>II</scp> are associated with new‐onset diabetes in individuals with prediabetes

Type 2 diabetes (T2D) is a major health burden worldwide and a significant risk factor for cardiovascular disease.1 Before the onset of T2D, most individuals experience a prediabetic stage characterized by impaired fasting glucose and/or glucose tolerance, accompanied by early insulin resistance and subtle lipid disturbances.2 Prediabetes represents a dynamic and potentially reversible state, underscoring the need for early biomarkers to guide preventive strategies and improve metabolic outcomes. Recent evidence highlights a key role of high-density lipoproteins (HDLs) in glucose homeostasis and β-cell function. HDLs not only mediate reverse cholesterol transport but also exert antioxidant, anti-inflammatory, and cytoprotective effects on pancreatic β-cells, thereby promoting insulin secretion and survival.3 Consistently, pharmacological studies using cholesteryl ester transfer protein inhibitors, which increase HDL cholesterol (HDL-C) concentrations, have shown a reduced risk of new-onset diabetes (NOD), reinforcing the protective link between HDL and glucose metabolism.4 Because HDL functionality depends on its protein cargo, particularly apolipoproteins, examining their structural and metabolic alterations offers critical insight into the mechanisms linking lipid metabolism and diabetes risk. Apolipoproteins are central regulators of lipoprotein metabolism and have been independently associated with T2D risk in individuals with prediabetes.2 Chronic hyperglycaemia also promotes the formation of methylglyoxal (MGO), a highly reactive by-product of glucose metabolism that drives the formation of advanced glycation end products through covalent modifications of proteins. Elevated plasma MGO concentrations have been associated with micro and macrovascular complications in T2D5; however, its clinical use as a biomarker remains limited due to its chemical instability.6 To overcome these limitations, we developed a bottom-up mass spectrometry-based proteomic workflow capable of detecting and quantifying stable peptide biomarkers that reflect specific MGO-induced modifications on plasma proteins. Using this approach, we identified MGH-ALB, a MGO-derived hydroimidazolone (MGH) modification of albumin (ALB), whose plasma concentration was independently associated with NOD in individuals with prediabetes.7 Here, we extended this strategy to identify MGO-derived modifications on apoA-I (APOA1) and apoA-II (APOA2), the two major structural and functional components of HDL. More precisely, we investigate the association of MGH-APOA1 and carboxyethyl-lysine (CEL)-APOA2 peptide biomarkers with the risk of NOD in patients with prediabetes. The IT-DIAB study (NCT01218061) is a five-year, single-centre prospective cohort designed to identify biomarkers of T2D risk in individuals living with prediabetes (fasting plasma glucose [FPG]: 110–125 mg/dL) who were monitored annually. NOD was defined as FPG ≥126 mg/dL and/or a plasma glucose concentration ≥200 mg/dL after a 2-h 75-g oral glucose tolerance test. HbA1c was not used as a diagnostic criterion and no participants were taking glucose-lowering medications at baseline. Study design, recruitment procedures, and biochemical measurements have been described elsewhere.8 The protocol was approved by the institutional ethics committee, and all participants provided written informed consent. For the present analysis, 300 individuals with at least one follow-up visit were included (Figure 1A). MGH-APOA1 (VEPLR+54AELQEGAR) and CEL-APOA2 (VK+72SPELQAEAK) peptide biomarkers were identified and quantified using bottom-up mass-spectrometry-based assays. Initial peptide mapping and characterization were conducted by trypsin digestion followed by untargeted proteomics, as previously reported.7 Candidate peptides were selected based on analytical performance criteria and incorporated into a targeted assay.7, 9 Peptides were then quantified in EDTA plasma (40 μL) after trypsin digestion using a standardized workflow validated for specificity, linearity, repeatability, and short- and long-term stability. Detailed procedures were similar to those established for the MGH-ALB biomarker.7 Categorical variables were expressed as numbers (%), and continuous variables as mean ± standard deviation (SD) or median [interquartile range, IQR], as appropriate. Spearman correlations were used to assess relationships between variables. Diabetes-free survival was visualized using Kaplan–Meier curves across peptide terciles. Associations between peptide biomarkers and incident NOD were evaluated using Cox proportional hazards models, first unadjusted and then adjusted for established diabetes risk factors (age, sex, body mass index [BMI], FPG, HbA1c, HDL-C, low-density lipoprotein cholesterol [LDL-C], and triglycerides). Given the number of events, models were deliberately parsimonious to limit overfitting. Hazard ratios (HR) were standardized per 1 SD increase. Proportional hazards assumptions were verified using Schoenfeld residuals. Analyses were performed on complete cases using R (v4.0.0), with a two-sided significance level of 0.05. Because MGH-APOA1 and CEL-APOA2 were prespecified and biologically related biomarkers, no formal adjustment for multiple comparisons was applied. The baseline characteristics of the study population are shown in Table 1. Over the five-year follow-up (IQR: 25–62 months; 85% retained until visit 5), 71 participants reverted to normoglycaemia, 116 remained prediabetic, and 113 progressed to NOD. At baseline, individuals who later developed NOD exhibited higher BMI and impaired glucose-homeostasis markers, together with lower HDL-C levels, compared with those who did not progress. Baseline plasma concentrations of MGH-APOA1 and CEL-APOA2 did not differ between participants who reverted to normoglycaemia and those who remained prediabetic (Figure 1B). In contrast, both peptide biomarkers were significantly higher in individuals who developed NOD compared with those who reverted or remained prediabetic, whereas total apoA-I and apoA-II concentrations showed no such difference (Figure 1B, Table 1). Spearman correlations revealed that MGH-APOA1 plasma concentrations were positively associated with HDL-C, MGH-ALB, total apoA-I, and apoA-II concentrations, whereas CEL-APOA2 levels were inversely correlated with estimated glomerular filtration rate (eGFR) and positively correlated with plasma triglycerides, total cholesterol, MGH-ALB, and LDL-C (Figure 1C). Neither peptide correlated with FPG, fasting insulin, HbA1c, HOMA-IR, HOMA-β, nor adiponectin, suggesting that their associations with diabetes risk were independent of glycaemic control, including insulin resistance and β-cell function. Of note, plasma levels of the two peptides were significantly inter-correlated (R = 0.15, p = 0.008). Kaplan–Meier analysis further demonstrated that higher plasma concentrations (third tertile) of both peptides at baseline were associated with increased incidence of NOD during the follow-up (log-rank p < 0.004; Figure 1D). In Cox proportional hazards models, MGH-APOA1 and CEL-APOA2 were significantly associated with NOD both before (HR per one SD [95% confidence interval]: 1.41 [1.22–1.63] and 1.44 [1.22–1.70], respectively; p < 0.0001 for both) and after adjustment for age, sex, BMI, FPG, triglycerides, LDL-C, and HDL-C (HR = 1.53 [1.31–1.79] and 1.44 [1.23–1.68], respectively; p < 0.0001 for both). These associations remained significant after further adjustment for HbA1c (Figure 1E). Additional adjustment for eGFR, statin use, or blood pressure did not alter the results. Although MGH-APOA1 was positively correlated with HDL-C, adjustment for HDL-C did not materially alter the associations with incident diabetes, arguing against major collinearity. We identified two specific peptide biomarkers reflecting MGO-induced glycation of apoA-I and apoA-II. Elevated plasma concentrations of MGH-APOA1 and CEL-APOA2 were independently associated with the risk of NOD in individuals with prediabetes, suggesting that MGO-driven modifications of HDL proteins may contribute to diabetes development beyond classical measures of glycaemic control or β-cell function. Besides, the dissociation between total apolipoprotein levels and their MGO-derived modifications suggests that early diabetes risk may relate more to qualitative HDL alterations than to particle abundance. The lack of correlation of MGH-APOA1 and CEL-APOA2 with HbA1c, together with their modest correlation with MGH-ALB, likely reflects the rapid kinetics of MGO-mediated glycation, combined with the much shorter half-lives of apoA-I and apoA-II (~3 days) compared with albumin (~3 weeks) or red blood cells (~3 months), as well as the distinct biochemical specificity of MGO-driven modifications. Interestingly, CEL-APOA2 showed a positive correlation with LDL-C. This surprising association likely reflects a global disturbance of lipoprotein metabolism in dyslipidaemic states, in which HDL-LDL lipid exchange, apoA-II redistribution, and carbonyl stress may interact to promote a more pro-atherogenic lipid profile. Such links reinforce the possibility that these HDL-derived glycated peptides capture broader cardiometabolic alterations relevant not only to glucose dysregulation but also to cardiovascular risk. Taken together, our findings suggest that the associations of these peptides with diabetes risk may reflect early alterations in HDL-related pathways, including impaired cholesterol efflux capacity, reduced antioxidative or anti-inflammatory properties, or interactions with vascular dysfunction, which warrant further investigation. A limitation of this study is that prediabetes and NOD were primarily defined using FPG, which may limit generalizability to broader prediabetes phenotypes. Future studies should determine whether these MGO-glycated peptides are associated with impaired HDL functional properties and whether they may serve as early biomarkers of cardiometabolic risk in individuals with prediabetes. We are grateful to the participants of the IT-DIAB study. We greatly thank Violette Dierry and Jeremy Thureau for their help with data collection. We also thank the biological resource centre for biobanking (CHU Nantes, Hôtel Dieu, Centre de ressources biologiques [CRB], Nantes, F-44093, France [BRIF: BB-0033-00040]). This work was carried out with financial support from the National Agency for Research (ANR-21-CE17-0063), the “Fondation de France” (# 00147726) and the “Fondation Francophone pour la Recherche sur le Diabète”, sponsored by “Fédération Française des Diabétiques (FFD)”, Abbott, AstraZeneca, Eli Lilly, Merck Sharp & Dohme (MSD) and Novo Nordisk. The IT-DIAB study was funded by an ISI grant from OSEO-API, France, and by an internal grant from CHU Nantes (RC17_0470). The authors declare no competing interests. The peer review history for this article is available at https://www.webofscience.com/api/gateway/wos/peer-review/10.1111/dom.70493. The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.