Use of physiologically based kinetic modeling to predict neurotoxicity and genotoxicity of methylglyoxal in humans

Methylglyoxal (MGO), a highly reactive α-oxoaldehyde, is found in various foods and endogenously produced. Exogenous sources include Maillard reactions, sugar autoxidation, caramelization, and lipid oxidation during food processing. Endogenous formation, primarily from glycolysis, produces approximately 3 mmol daily in healthy adults. The glyoxalase system (Glo1 and Glo2) detoxifies MGO to D-lactate, maintaining low levels (0.05–0.6 µM in plasma, 1–4 µM in cells). However, conditions like hyperglycemia elevate MGO, potentially up to 30% higher in type 2 diabetics. MGO is a precursor for advanced glycation end products (AGEs), linked to chronic diseases. Elevated MGO increases AGE formation, interacting with DNA and proteins, causing structural and functional changes, and potentially contributing to diabetes and neurodegenerative disorders like Alzheimer's and Parkinson's disease. Clinical studies associate higher serum MGO with cognitive decline, with elevated MGO and AGEs found in cerebrospinal fluid of Alzheimer's patients and nigra neurons of Parkinson's patients. In vitro studies demonstrate MGO's cellular injury and toxicity in neuronal cells, impacting cell viability, mitochondrial redox activity, ROS production, and apoptosis. MGO's genotoxic potential is also a concern, with studies showing DNA adduct formation and increased mutation rates. This study aimed to use PBK modeling-facilitated reverse dosimetry to translate in vitro neurotoxicity and genotoxicity data into in vivo dose-response predictions for humans, assessing the impact of both exogenous and endogenous MGO.

The literature extensively documents methylglyoxal's presence in food and its endogenous production. Studies have linked elevated methylglyoxal levels to various adverse health outcomes, including neurodegenerative diseases and cancer. The glyoxalase system's role in methylglyoxal detoxification is well-established. In vitro studies have shown methylglyoxal's toxicity in neuronal cells and its genotoxic potential through DNA adduct formation. However, a significant gap exists in translating these in vitro findings to predict in vivo human risk. The current study addresses this gap by applying physiologically based kinetic (PBK) modeling and reverse dosimetry as a new approach methodology (NAM) to bridge the gap between in vitro and in vivo data, thereby enabling a more accurate risk assessment.

This study employed a PBK modeling-facilitated reverse dosimetry approach. First, a mouse PBK model for MGO was developed, incorporating compartments for the stomach, intestine, liver, fat, blood, rapidly and slowly perfused tissues. The model utilized first-order kinetics for processes like gastrointestinal transport, absorption, and clearance. Physiological and anatomical parameters were obtained from literature, including a Caco-2 *P*_app value (1.09 × 10⁻⁶ cm/s) extrapolated to in vivo values for mice and humans. Intestinal absorption was modeled based on *P*_app, intestinal surface area, and luminal MGO concentration. A 55-min lag time was included to match in vivo observations. Tissue/blood partition coefficients were estimated using the Rodgers and Rowland method via the QIVIVE tool. Clearance was modeled based on tissue volume fractions and literature-reported apparent total body clearance (*CL*_app), using the lowest dose value (7.69 L/h/kg) to avoid potential biases from dose-dependent clearance. Fraction absorbed (*f*_a) correction was incorporated to improve model fit at higher doses. The model was evaluated by comparing predictions to in vivo mouse data. A sensitivity analysis assessed the impact of model parameters on predicted *C*_max. A human PBK model was then created, adapting the mouse model using human physiological parameters and the same *CL*_app value. In vitro concentration-response data for MGO neurotoxicity (from human neuronal-like cells: hNLCs) and genotoxicity (from WM-266-4 human melanoma cells) were used for in vivo extrapolation. The in vitro unbound concentration was equated to in vivo unbound *C*_max, correcting for protein binding differences. Reverse dosimetry determined the oral MGO dose levels needed to achieve the effective *C*_max concentrations. Benchmark dose modeling (BMD) analysis determined points of departure (PODs), using BMDL₁₀ values for risk assessment. Margins of exposure (MOEs) were calculated by dividing the BMDL₁₀ by estimated daily MGO intake and endogenous formation in healthy individuals and diabetic patients. The highest reported values for both dietary intake and endogenous formation in diabetic patients were used to represent a worst-case scenario.

The mouse PBK model accurately predicted MGO blood concentrations in mice across different doses, with predictions falling within a twofold difference of in vivo data. Sensitivity analysis revealed that gastrointestinal transport and absorption parameters (*SA*_in, *V*_in, *k*_sto, *P*_app, *k*_in) and *CL*_app most significantly influenced *C*_max predictions. The human PBK model, using in vitro data from hNLCs and WM-266-4 cells, predicted in vivo dose-response curves for neurotoxicity (mitochondrial function, cytotoxicity, apoptosis) and genotoxicity (DNA adduct formation). BMD analysis yielded BMDL₁₀ values: 1366 mg/kg bw (mitochondrial function, 48 h), 590 mg/kg bw (cytotoxicity, 48 h), 304 mg/kg bw (apoptosis, 48 h), 251 mg/kg bw (R-N²-CEdG formation), and 254 mg/kg bw (S-N²-CEdG formation). MOE calculations revealed that dietary MGO intake posed no neurotoxicity concern (MOEs > 100), but genotoxicity concerns couldn't be ruled out (MOEs < 10,000). Endogenous MGO formation, particularly in diabetics, raised concerns for apoptosis and cytotoxicity (MOEs < 100), and genotoxicity concerns persisted (MOEs < 10,000). Endogenous MGO levels in healthy individuals exceeded the safety threshold for apoptosis. Diabetic patients exceeded thresholds for both apoptosis and cytotoxicity.

This study successfully demonstrated the application of PBK modeling-facilitated reverse dosimetry as a NAM for extrapolating in vitro data to predict in vivo toxicity of MGO in humans. The PBK model accurately predicted MGO kinetics. Dietary MGO intake did not raise neurotoxicity concerns, but genotoxicity risk couldn't be ruled out. Endogenous MGO, especially in diabetics, presented significant concerns for both neurotoxicity and genotoxicity. The model's sensitivity analysis highlighted the importance of accurate gastrointestinal and metabolic parameters. The study's limitations include the reliance on existing in vitro datasets, the assumption that blood *C*_max drives neurotoxicity (without a CNS compartment in the model), and the use of melanoma cells for genotoxicity assessment. Future studies should improve the accuracy by independently determining clearance rate constants, incorporating Monte Carlo simulations, and using more physiologically relevant in vitro models like human peripheral blood lymphocytes for genotoxicity testing. The relatively low contribution of dietary MGO to the total body burden compared to endogenous formation emphasizes the need to focus on endogenous MGO formation.

This study successfully integrated in vitro data and in silico PBK modeling to predict in vivo MGO toxicity in humans. Dietary MGO poses minimal neurotoxicity risk, but genotoxicity concerns remain. Endogenous MGO, especially in diabetics, raises significant concerns for both neurotoxicity and genotoxicity. The model serves as a valuable tool for QIVIVE and risk assessment, though further refinements and data are needed to enhance its accuracy and scope.

The study's limitations include the use of a mouse-based PBK model extrapolated to humans, the absence of a dedicated CNS compartment in the model, and the use of melanoma cells for genotoxicity assessment, which might not fully reflect in vivo conditions. Furthermore, uncertainties remain in the extrapolation of in vitro data to in vivo scenarios, and the sensitivity analysis highlights the influence of various parameters that warrant further investigation. The limited availability of in vivo kinetic data for MGO and the need for more comprehensive in vitro data sets to enhance the reliability of QIVIVE should also be considered.