Density functional theory insights into the most reactive sugar and amino acid pair for Amadori rearrangement in the Maillard reaction

Abstract

The Maillard reaction is fundamental to food flavor and color, with the Amadori rearrangement as a critical branching point. Using density functional theory, this study investigated the Amadori rearrangement mechanisms between glucose and 20 amino acids, and between leucine and 10 sugars. Leucine exhibited the lowest energy barrier (93.55 kJ/mol) in the rate-determining condensation step with glucose, significantly lower than other amino acids (133.73–221.22 kJ/mol). For subsequent dicarbonyl formation, leucine maintained the lowest energy barriers across all pH conditions. Among sugars, glucose showed the lowest condensation barrier (93.55 kJ/mol) versus others (128.77–162.98 kJ/mol). Reactivity order with leucine was: glucose > lactose > xylose > erythrose > arabinose > galactose > maltose > fructose > sucrose > mannose. Based on a comparison of DFT-calculated energy barriers, this study identifies leucine and glucose as the most reactive amino acid and sugar in the Maillard reaction, providing theoretical foundations for targeted product regulation and rational selection of sugar-amino acid pairs to control flavor, color, and safety in thermally processed foods.

Graphical abstract

Highlights

• •Leucine is the most reactive amino acid in Amadori rearrangement.
• •Glucose shows the lowest condensation barrier among ten sugars.
• •Sugar reactivity order with leucine: glucose > lactose > xylose > erythrose > arabinose > galactose > maltose > fructose > sucrose > mannose.
• •Leucine-glucose is the optimal pair for pH-dependent Maillard reaction regulation.

1. Introduction

The Maillard reaction, which underlies the formation of flavors, colors, and bioactive compounds during food processing, storage, and even physiological processes, has long been a central topic in food chemistry due to its complex reaction network (Hodge, 1953). Among its various pathways, the Amadori rearrangement serves as the decisive initial step, converting unstable Schiff bases into relatively stable Amadori compounds. This process involves multiple proton transfer steps, which are essential for the tautomerization of the Schiff base to the Amadori rearrangement product (Gao et al., 2023). The efficiency and directionality of these proton transfers determine the overall reaction barrier and product distribution (Ren et al., 2015). Furthermore, this step lays the molecular foundation for subsequent key pathways such as Strecker degradation, cyclization, and polymerization, which are crucial for the development of flavor and color (Martins et al., 2000; Ashoor and Zent, 1984). Therefore, clarifying the intrinsic driving forces and reactivity patterns of the Amadori rearrangement is key to a deeper understanding and targeted control of the Maillard reaction process and its products.

Traditionally, understanding the reactivity of sugars and amino acids in the Maillard reaction has relied primarily on experimental observations, inferring trends from the qualitative and quantitative analysis of products in model systems (Yaylayan and Keyhani, 2000; Ledl and Schleicher, 1990). These studies have accumulated valuable empirical rules; for instance, it is widely recognized that pentoses (e.g., ribose) are generally more reactive than hexoses (e.g., glucose), and basic amino acids (e.g., lysine) typically exhibit higher reactivity (van Boekel, 2006). However, such experimental approaches struggle to clearly reveal the essential differences in reactivity and their underlying physicochemical origins at the atomic and electronic levels (Jaeger et al., 2010). Is reactivity primarily governed by the electrophilicity of the carbonyl carbon in the sugar, steric hindrance, the nucleophilicity of the amino group, or the ease of proton transfer? These microscopic mechanistic questions remain challenging to resolve with macroscopic experimental data alone, limiting our ability to accurately predict and rationally design reaction selectivity (El Hosry et al., 2025).

In recent years, computational chemistry, particularly quantum mechanics-based density functional theory (DFT), has emerged as a powerful tool for unraveling the mechanisms of complex chemical reactions (Hohenberg and Kohn, 1964; Kohn and Sham, 1965). DFT enables precise calculations of the geometric structures, electron density distributions, frontier orbital energies, and thermodynamic and kinetic parameters of reactants, transition states, and products along a reaction path. In the field of food chemistry, DFT has been successfully applied to elucidate mechanisms in various stages of the Maillard reaction and related processes, such as the formation of furfurals (Antal Jr et al., 1990), the Strecker degradation of amino acids (Yu et al., 2012; Ishida et al., 2025), and the formation of specific aroma compounds (Yu et al., 2012). Notably, computational studies have begun to probe the initial condensation and rearrangement steps, providing valuable insights into the energetics of specific model systems (Perez-Locas and Yaylayan, 2008). For example, Huan et al. (2024) investigated the Maillard reaction between glucose and methylamine using DFT, confirming the Amadori rearrangement as the rate-controlling step. Solís-Calero et al. (2013) systematically compared Amadori rearrangements on different amino substrates. However, a systematic and comprehensive DFT investigation that directly compares and ranks the reactivity of a broad range of fundamental sugar and amino acid pairs specifically for the rate-limiting Amadori rearrangement step remains lacking. Most prior computational studies have focused on isolated parts of the reaction network or specific compound pairs, leaving a gap in a unified, comparative analysis that can definitively identify the most reactive pair and generalize the key electronic-structure-based rules governing this selectivity (Huan et al., 2024; Xu et al., 2019).

To address this knowledge gap, this study employs high-accuracy DFT methods to construct detailed reaction models of the Amadori rearrangement, explicitly including the proton transfer processes. We systematically calculated the reaction pathways and energy barriers for the Amadori rearrangement step involving a series of representative sugars (e.g., aldopentoses, aldohexoses with varying configurations) and amino acids (e.g., neutral, acidic, and basic). The primary objectives of this study are: (1) to theoretically identify and rank the most favorable sugar–amino acid pairs for the Amadori rearrangement in the initial stage of the Maillard reaction; (2) to unveil the key electronic structure descriptors and physicochemical factors governing their relative reactivity, such as charge distribution, proton affinity, and strain energy; and (3) to establish a robust theoretical framework for predicting reactivity in the Maillard reaction, offering a new, mechanistic perspective for future rational design in controlling food flavor, color, and nutritional quality. This work is expected to significantly deepen the atomistic understanding of the initial kinetics of the Maillard reaction and promote a shift in food chemistry from empirical observation towards quantitative prediction and mechanism-driven innovation.

2. Experimental

2.1. Computational methods

All quantum chemical calculations in this study were performed using the Gaussian 16 software package. The computational work was conducted within the framework of DFT, employing the B3LYP functional (Becke three-parameter exchange functional combined with the Lee-Yang-Parr correlation functional) and the 6-31G(d,p) basis set for all atoms.

2.2. Geometry optimization

The initial geometries of all reactants, intermediates, transition states, and products were constructed using GaussView 6.0. To account for the vast conformational space of flexible sugars and amino acid side chains, we adopted a systematic protocol: for each reaction system, multiple initial conformations were generated by manually rotating the amino acid side chain and the sugar backbone to avoid unrealistic steric clashes, and the lowest-energy conformation was selected for subsequent optimization. All geometry optimizations were performed at the B3LYP/6-31G(d,p) level using the Gaussian 16 software package. The following convergence criteria were applied: maximum force = 0.00045 au, root-mean-square (RMS) force = 0.00030 au, maximum displacement = 0.0018 au, and RMS displacement = 0.0012 au. No symmetry constraints were imposed during optimization. All optimized structures were confirmed to be local minima by subsequent vibrational frequency analysis.

2.3. Transition state search and verification

All transition state (TS) structures were searched and optimized using the Gaussian 16 software package at the B3LYP/6-31G(d,p) level. The initial guess of each TS structure was constructed using the quadratic synchronous transit (QST2) or QST3 method as implemented in Gaussian 16, which interpolates between the optimized reactant and product geometries to generate a trial TS geometry. For cases where QST2 failed to converge, a stepwise scanning approach was employed: a potential energy surface scan along the presumed reaction coordinate was performed first to obtain an approximate TS geometry, which was then used as the starting point for full TS optimization using the Berny algorithm. All TS optimizations were performed using the default Berny optimizer with the following convergence criteria: maximum force = 0.00045 au, root-mean-square (RMS) force = 0.00030 au, maximum displacement = 0.0018 au, and RMS displacement = 0.0012 au.

To unambiguously verify each TS, harmonic vibrational frequency analysis was performed at the same theoretical level. A valid TS structure must satisfy two essential criteria: (1) the frequency analysis yields exactly one imaginary frequency (i.e., one negative eigenvalue in the Hessian matrix), confirming that the stationary point is a first-order saddle point on the potential energy surface; and (2) the vibrational mode associated with this imaginary frequency must correspond to the intended reaction coordinate. Specifically, the nuclear displacement vectors of the imaginary mode were visually inspected using GaussView 6.0 to ensure that they point toward the reactant and product directions along the targeted bond-breaking and bond-forming processes. No TS was accepted without satisfying both criteria.

2.4. Intrinsic reaction coordinate (IRC) calculation

To confirm that each located transition state correctly connects the intended reactant and product minima, and to map the minimum energy path (MEP) of each elementary reaction, intrinsic reaction coordinate (IRC) calculations were performed for all verified transition states. The IRC calculations were carried out at the B3LYP/6-31G(d,p) level using the Gonzalez-Schlegel second-order integration method as implemented in Gaussian 16. For each TS, the IRC was calculated in both the forward and reverse directions along the eigenvector of the unique imaginary frequency. The following computational parameters were used: step size = 0.1 amu1/21/2 bohr, maximum number of points = 50 in each direction, and the force constants were updated using the Berny algorithm at every step. The convergence criterion for each IRC point was set to a root-mean-square force of 0.00045 au. After the IRC path was obtained, the geometries at the terminal points of the forward and reverse trajectories were re-optimized without constraints to confirm that they correspond exactly to the designated reactant and product local minima. Only those transition states whose IRC paths successfully connected to the correct reactant and product minima were retained for subsequent thermodynamic and kinetic analyses. Representative IRC plots for TS1-TS8 of the glucose-leucine system were provided in Supplementary Fig. S1.

2.5. Construction of the reaction potential energy surface

2.5.1. Reactions with explicit transition states

For each elementary reaction that was validated by IRC calculations, harmonic vibrational frequency analysis was first performed at the B3LYP/6-31G (d,p) level on the optimized geometries of the reactant, transition state (TS), and product. Frequency analysis was carried out under the conditions of 298.15 K and 1 atm to obtain zero-point energy (ZPE) corrections, thermal corrections (including translational, rotational, and vibrational contributions), and entropy contributions. All frequency analyses confirmed that each TS had exactly one imaginary frequency, while reactants and products had none.

Subsequently, to obtain more accurate relative Gibbs free energies, high-precision single-point energy calculations were performed on the optimized geometries using the UB3LYP functional combined with the def2-TZVPP basis set, with solvation effects included via the SMD continuum solvation model. The integration grid for single-point calculations was set to Int = UltraFine, and the SCF convergence criterion was set to 10⁻⁸ au. Finally, the Gibbs free energy of each stationary point was calculated using the following equation:

ΔG_(total) = E + ΔZPE + ΔG – TΔS

where: E is the single-point energy obtained at the UB3LYP/def2-TZVPP//SMD level; ΔZPE and ΔG were extracted from the frequency analysis output using the Shermo program (version 2.4.6), which is based on the ideal gas model and the rigid-rotor harmonic-oscillator approximation. For low-frequency vibrational modes (<100 cm⁻¹), a quasi-harmonic approximation was applied to mitigate errors from the rigid-rotor harmonic-oscillator model.

All energies are reported relative to the total Gibbs free energy of the separated reactants, yielding the relative Gibbs free energy profile along each reaction path.

2.5.2. Reactions without explicit transition states

For certain proton transfer steps, potential energy surface scanning was performed at the B3LYP/6-31G(d,p) level, because such processes often lack a well-defined first-order saddle point that can be located by conventional TS searches.

The proton transfer reaction was monitored using the distance between the donor atom (D) and the transferring hydrogen, i.e., R(D–H), and the distance between the hydrogen and the acceptor atom (A), i.e., R(H···A). The reaction coordinate was defined as the D–H···A distance. During the scan, this distance was incrementally increased from its initial value in the reactant state (typically R(D–H) ≈ 1.0 Å, R(H···A) ≈ 1.8–2.2 Å) to its final value in the product state (R(D–H) ≈ 1.8–2.2 Å, R(H···A) ≈ 1.0 Å). A fixed step size of 0.05 Å was used, with 30–50 points scanned in each direction. At each scan step, all other geometric degrees of freedom were fully optimized while the chosen reaction coordinate was constrained. After each scan point, the total energy of the constrained geometry was extracted.

After obtaining the one-dimensional potential energy profile as a function of the reaction coordinate, the data were fitted using a cubic spline interpolation implemented in the UnivariateSpline function of the Python SciPy library, yielding a smooth energy curve. The maximum energy on the fitted curve was taken as the apparent energy barrier for the proton transfer process. If the curve exhibited a clear single peak located between the reactant and product regions, that peak was treated as a “quasi-transition state” energy. For such barrier-only processes, rate constants were not calculated; only the apparent barrier height was reported in the potential energy diagrams.

2.6. Calculation of reaction rate constants

Reaction rate constants for the key elementary steps were calculated using transition state theory (TST) as implemented in the TSTcalculator program (version 1.2). The calculations were performed over the temperature range of 273.15–483.15 K, which covers typical food processing and thermal reaction conditions. For each elementary reaction with a well-defined transition state (TS), the rate constant k(T) was calculated using the Eyring equation:

$$k\left(T\right)=\mathrm{\kappa }\frac{K_{B}T}{h}\exp \left(-\frac{{\Delta G}^{‡}\left(T\right)}{RT}\right)$$

where: k_B​ is the Boltzmann constant; h is the Planck constant; R is the universal gas constant; T is the absolute temperature; ΔG‡(T) is the Gibbs free energy of activation at temperature T, obtained from the potential energy surface construction described in Section 2.5.1. Specifically, ΔG‡(T) = G^TS​(T)−G^reactant​(T), with both values including zero-point energy, thermal corrections, and entropy contributions at the same temperature.

The tunneling correction factor κ was estimated using the Wigner approximation for reactions involving significant hydrogen or proton transfer:

$$\mathrm{\kappa }=1+\frac{1}{24}{\left(\frac{{hv}^{‡}}{K_{B}T}\right)}^2$$

where: ν‡ is the magnitude of the imaginary frequency of the transition state. This correction is particularly important for the proton transfer and enolization steps where light atoms (hydrogen) are transferred.

For bimolecular reactions, the rate constant was calculated using the same equation, with ΔG‡(T) referenced to the separated reactants. The TSTcalculator program automatically handles both unimolecular and bimolecular reactions by reading the Gibbs free energy differences directly from the Shermo output files. All input files were prepared in the format required by TSTcalculator. The program was run with default settings.

For each reaction pathway, the rate constants at different temperatures were computed and tabulated. To compare the relative kinetic advantages of different amino acids or sugars under identical conditions, we reported the rate constants for the rate-determining step. For reactions without a true transition state, rate constants were not calculated because a well-defined TS could not be located; only the apparent energy barriers are reported in the potential energy diagrams.

3. Results and discussion

3.1. Detailed mechanism of the Amadori rearrangement: the case of leucine and glucose

The overall mechanism of the Amadori rearrangement leading to dicarbonyl compounds is well documented (Hodge, 1953; Martins et al., 2000). Nevertheless, atomistic details, including transition state structures and energy barriers, are best revealed by DFT calculations. In this study, we use the computationally identified optimal pair, leucine and glucose, to describe the complete reaction pathway, which also serves as a template for other sugar-amino acid combinations.

As illustrated in Fig. 1A, the reaction between glucose and leucine proceeds through two stages: the formation of the Amadori rearrangement product (ARP) and its subsequent pH-dependent decomposition to dicarbonyl compounds. During the Amadori rearrangement phase, a hydrogen atom from N14 of the amino acid (2#) attacks the O1 atom of glucose (1#), leading to the formation of a C–N bond via transition state TS1 and generating N-glucosamine (3#). The energy barrier for this step is 93.55 kJ/mol. Subsequently, a hydrogen atom from N14 of N-glucosamine (3#) attacks O4, resulting in the cleavage of the C4–O4 bond and the loss of one water molecule, accompanied by the formation of an N14–C12 double bond. This step proceeds through transition state TS2 to yield the Schiff base (4#), with an energy barrier of 147.72 kJ/mol. Next, a hydrogen atom from C10 of the Schiff base (4#) attacks N14, causing cleavage of the N14–C10 double bond and the formation of a new C10–C12 double bond via transition state TS3, producing compound (5#). The energy barrier for this reaction is 159.90 kJ/mol. Finally, a hydrogen atom from O34 of compound (5#) attacks C11, leading to cleavage of the C10–C11 double bond and the formation of a C10–O34 double bond via transition state TS4, yielding 1-amino-1-deoxy-2-ketose (6#). The energy barrier for this step is 67.84 kJ/mol.

Fig. 1.

Mechanism of DFT-based Amadori rearrangement reaction to generate dicarbonyl compounds. A: Mechanism diagram of glucose with leucine; B: Transition state structures of glucose with different kinds of amino acids.

The subsequent stage involves the thermal decomposition of the Amadori rearrangement product to form dicarbonyl compounds. Specifically, 1-amino-1-deoxy-2-ketose (6#) undergoes deamination under different pH conditions, losing one molecule of leucine to form 1-deoxyglucosone (13#) or 3-deoxyglucosone (10#). Under acidic conditions, a hydrogen atom from C11 of 1-amino-1-deoxy-2-ketose (6#) attacks O43, completing the enolization reaction via transition state TS4 to generate compound (7#), with an energy barrier of 169.50 kJ/mol. Subsequently, the hydroxyl group on C8 of compound (7#) migrates to C11, accompanied by the formation of a C8–C10 double bond, generating compound (8#). A hydrogen atom from C4 of compound (8#) then attacks N3, leading to cleavage of the N3–C4 bond and the elimination of one molecule of leucine (2#). This step proceeds through transition state TS5 to form intermediate (9#), with an energy barrier of 27.46 kJ/mol. Finally, a hydrogen atom from O5 of intermediate (9#) attacks C6 via transition state TS6 to produce 3-deoxyglucosone (10#), with an energy barrier of 123.59 kJ/mol. Under alkaline conditions, a hydrogen atom from C8 of 1-amino-1-deoxy-2-ketose (6#) transfers to O14 via transition state TS7 to generate compound (11#), with an energy barrier of 138.59 kJ/mol. Subsequently, a hydrogen atom from C6 of compound (11#) transfers to N3, resulting in cleavage of the N3–C4 bond and the release of one molecule of leucine (2#) to form intermediate (12#). Intermediate (12#) then proceeds through transition state TS8 to generate 1-deoxyglucosone (13#), with an energy barrier of 98.44 kJ/mol.

3.2. Identification of the most reactive amino acid in the Amadori rearrangement

In the Maillard reaction system, the formation of Amadori rearrangement products (ARPs) and their subsequent decomposition to dicarbonyl compounds represent a critical juncture within the entire reaction network (Gao et al., 2025). ARPs are not only the stable products resulting from the initial condensation between sugars and amino acids but also serve as direct precursors for the generation of various flavor compounds, melanoidins, and reactive intermediates (Lei et al., 2024). Therefore, elucidating the differences in energy barriers and kinetic parameters among different amino acids at this stage is of significant theoretical importance for understanding the selectivity mechanisms of the Maillard reaction and holds practical value for the targeted regulation of product profiles in food processing.

By confirming all transition state structures along the reaction pathways between glucose and 20 amino acids (Fig. 1B) and calculating their potential energy surfaces (Fig. 2), this study found that during the Amadori rearrangement phase, the initial carbonyl-amine condensation reaction (generating N-glucosamine via TS1) is universally the rate-determining step. The height of the energy barrier for this step directly dictates the ease with which an amino acid enters the Maillard reaction channel. Its numerical value reflects the inherent difficulty of the reaction and serves as a core thermodynamic indicator for comparing the reaction priority of different amino acids. The calculated results show that the energy barrier for leucine in this step is 93.55 kJ/mol, significantly lower than those of the other 19 amino acids (ranging from 133.73 kJ/mol to 221.22 kJ/mol). This indicates that in the initial stage of the reaction, leucine can rapidly surmount the energy barrier by virtue of its lower activation energy, preferentially forming the Amadori product. Furthermore, this result strongly suggests that leucine possesses unique advantages in terms of electronic effects and spatial configuration, enabling it to more effectively facilitate the initial nucleophilic attack and proton transfer processes of the Amadori rearrangement. It is thus identified as the amino acid most prone to undergoing the Amadori rearrangement with glucose among the 20 tested.

Fig. 2.

Potential energy diagrams for the reaction of glucose with different kinds of amino acid in Amadori rearrangements to form dicarbonyl compounds. A: Under neutral, weakly basic reaction; B: Under acidic conditions.

In the subsequent thermal decomposition of Amadori rearrangement products to form dicarbonyl compounds, the reaction pathway exhibits a significant pH dependency, which is a crucial manifestation of the Maillard reaction's complexity. Under neutral to weakly alkaline conditions, 2,3-enolization is the key reaction in this phase. The highest energy barrier corresponding to leucine is 138.59 kJ/mol, while the energy barriers for other amino acids range from 214.72 kJ/mol to 393.19 kJ/mol (Fig. 2A). Under acidic conditions, 1,2-enolization becomes the key reaction. The highest energy barrier for leucine is 169.50 kJ/mol, whereas the energy barriers for other amino acids range from 180.71 kJ/mol to 875.96 kJ/mol (Fig. 2B). Observing the overall trend, the reaction energy barriers for the Amadori rearrangement products of most amino acids are significantly lower under neutral to weakly alkaline conditions than under acidic conditions. This implies that, from a thermodynamic perspective, a neutral to weakly alkaline environment is more favorable for the progression of the Maillard reaction during thermal food processing. Notably, regardless of the pH condition, the highest energy barriers for the leucine-glucose system are substantially lower than those for other amino acid systems, and its energy barrier ranking consistently places it at the lowest or second-lowest position among the 20 amino acids. This fully corroborates that leucine also possesses a significant thermodynamic advantage during the ARP decomposition phase.

Further analysis of pH influence on reaction priority reveals pronounced differences in pH sensitivity among amino acids. Under neutral to weakly alkaline conditions, amino acids such as serine, cysteine, and alanine exhibit relatively high energy barriers (330.58 kJ/mol, 289.52 kJ/mol, and 297.98 kJ/mol, respectively), indicating that the ARPs of these amino acids decompose with difficulty via 2,3-enolization in a weakly alkaline environment. In contrast, under acidic conditions, the energy barriers for glutamine, glutamic acid, and serine increase dramatically (reaching 875.96 kJ/mol, 647.99 kJ/mol, and 631.72 kJ/mol, respectively), demonstrating that the 1,2-enolization decomposition of these amino acids' ARPs is extremely challenging in an acidic environment. In comparison, leucine not only maintains the lowest energy barriers under both pH conditions but also exhibits a relatively small difference in energy barriers between acidic and alkaline conditions (approximately 30 kJ/mol), showcasing good adaptability to pH. This suggests that the decomposition pathway of leucine's ARP is less perturbed by pH variations, possessing a wider operational pH window, which confers broader application potential across diverse food processing conditions.

These phenomena can be understood from the perspective of the interplay between amino acid side chain structure and the reaction mechanism. In the 1,2-enolization pathway under acidic conditions, the involvement of a protonated environment may lead to unfavorable electrostatic repulsion or solvation effects for amino acids with polar or charged side chains (e.g., glutamic acid, serine), thereby significantly increasing the transition state energy barrier. In the 2,3-enolization pathway under neutral to weakly alkaline conditions, the steric hindrance of the side chain becomes a predominant factor. Amino acids with bulky side chains (e.g., tryptophan, phenylalanine) may face difficulties in proton transfer due to spatial impediments. Leucine's isobutyl side chain lacks strongly polar groups and possesses a moderate spatial volume, enabling it to effectively stabilize the key transition states under different pH conditions, thus exhibiting optimal and stable reactivity.

This trend is further corroborated by the reaction rate constants. As it is shown in Fig. 3, the rate constants show that the reaction rates for the leucine-glucose system are generally higher than those for other amino acid systems across various temperatures. Moreover, as the temperature increases, its rate enhancement is more pronounced. This indicates that leucine not only possesses a lower energy barrier advantage thermodynamically but also demonstrates a faster reaction progress kinetically. The consistency between thermodynamics and kinetics further solidifies the conclusion that leucine is the most reactive amino acid. The superior reactivity of leucine can be rationally explained by its side chain structure and electronic effects. First, the isobutyl side chain of leucine is non-polar and of moderate volume; this avoids the unfavorable electrostatic interactions or solvation effects that strongly polar side chains (e.g., serine, glutamic acid) might induce in the transition state, as well as the steric hindrance during proton transfer caused by bulky side chains (e.g., tryptophan, phenylalanine). Second, we infer that in the transition state of the condensation step, the formation trajectory of the C–N bond between leucine and glucose exhibits a relatively minor deviation from the ideal orbital overlap direction, which may be one of the reasons for its lower activation energy barrier compared to other amino acids. In contrast, due to unfavorable combinations of side chain polarity, volume, or electronic effects, other amino acids exhibit more significant geometric distortions in their transition states, leading to correspondingly higher energy barriers.

Fig. 3.

Rate constants for the reaction of glucose with different kinds of amino acid in Amadori rearrangements to form dicarbonyl compounds. A: Amadori arrangement; B: Under neutral, weakly basic condition; C: Under acidic condition.

In summary, through a comparative analysis of the thermodynamic and kinetic parameters of 20 amino acids during the Amadori rearrangement and subsequent dicarbonyl compound formation stages, it is unequivocally established that leucine exhibits the highest reaction priority in the Maillard reaction with glucose. This advantage is consistently maintained under neutral, weakly alkaline, and acidic conditions. Concurrently, based on the ranking of the highest energy barriers, the order of priority for amino acids to undergo the Maillard reaction is generally: Leucine > Glutamic acid > Tyrosine > Lysine > Valine > Histidine > Arginine > Isoleucine > Glutamine > Phenylalanine > Threonine > Glycine > Tryptophan > Aspartic acid > Asparagine > Methionine > Cysteine > Alanine > Serine>Proline. This conclusion not only deepens the theoretical understanding of amino acid selectivity in the Maillard reaction at the molecular level but also provides a clear target amino acid for directionally regulating the Maillard reaction process and controlling the formation of key flavor precursors, such as dicarbonyl compounds, based on pH conditions in food processing.

3.3. Reactivity of leucine with different sugars: identification of the most reactive sugar

Having identified leucine as the most reactive amino acid, we examined the reaction of leucine with ten structurally diverse sugars: tetrose (erythrose), pentoses (arabinose, xylose), hexoses (fructose, galactose, glucose, mannose), and disaccharides (maltose, lactose, sucrose). The general mechanism described in Section 3.1 applies to all sugar, and the mechanism of representative erythrose (tetrose), xylose (pentose), lactose (polysaccharide) was shown in Fig. 4 A, B, C, respectively. Meanwhile, Table 1 summarizes the key energy barriers, including those for the condensation TS1, the highest barrier under acidic 1,2-enolization, and the highest barrier under neutral/weakly alkaline 2,3-enolization.

Fig. 4.

DFT-based mechanistic diagram of the reaction of leucine with different kinds of sugar in Amadori rearrangement to generate dicarbonyl compounds. A: mechanism of leucine with erythrose; B: mechanism of leucine with xylose; C: mechanism of leucine with lactose; D: Transition state structures.

Table 1.

Key energy barriers (kJ/mol) for the reaction of leucine with different sugars.

Sugar	Condensation (TS1)	Acidic condition	Alkaline condition
Glucose	93.55	169.49	138.59
Lactose	138.89	249.94	184.30
Xylose	137.78	551.33	189.95
Erythrose	146.12	254.90	195.10
Arabinose	146.92	483.93	223.51
Galactose	162.98	357.64	230.94
Maltose	149.66	343.99	250.27
Fructose	162.78	254.95	253.28
Sucrose	159.90	260.27	348.89
Mannose	128.77	310.13	422.10

In the Maillard reaction network, the structural characteristics of sugars, including carbon chain length, carbonyl position, and glycosidic bond type, significantly influence their reactivity with amino acids. To identify the sugar most susceptible to the Amadori rearrangement or even to the whole Maillard reaction, this study, based on the determination of leucine as the optimal amino acid, systematically compared the reaction pathways, energy barrier characteristics, and kinetic behaviors of leucine with 10 structurally distinct sugars.

Through the optimization and confirmation of all transition state structures (Fig. 4D) and potential energy surface calculations (Fig. 5) for each reaction pathway, it is found that during the Amadori rearrangement phase, the energy barrier of the initial carbonyl-amine condensation reaction is also generally the rate-determining step. Its magnitude directly reflects the ease with which different sugars enter the Maillard reaction channel. The calculated results show that the energy barrier for glucose in this step is 93.55 kJ/mol, significantly lower than those of the other nine sugars (ranging from 128.77 kJ/mol to 162.98 kJ/mol), indicating that glucose possesses a significant thermodynamic advantage even at the initial condensation stage. This reactivity is closely correlated with the accessibility of the aldehyde group and the degree of steric hindrance near the reaction center. As an aldohexose, glucose possesses an electrophilic open-chain aldehyde form; although the equilibrium constant for ring-opening is small, it is sufficient to supply the reactive species (van Boekel, 2001; Yaylayan & Keyhani, 2000). More importantly, the specific configuration of glucose is such that in its most stable ⁴C1 chair conformation, all hydroxyl groups are equatorial (Matsuoka et al., 2013; Gaweda & Plazinski, 2019). This arrangement minimizes steric repulsion during nucleophilic attack by the amino group. Furthermore, this equatorial orientation facilitates the subsequent enolization step by allowing favorable hydrogen bonding interactions with the transition state. In contrast, mannose (the C2 epimer) has an axial hydroxyl group, which introduces steric hindrance and elevates its condensation barrier to 128.77 kJ/mol (Matsuoka et al., 2013; Mayes et al., 2014). For pentoses (xylose and arabinose), the five-membered furanose ring is more puckered and exhibits lower stability under acidic conditions, resulting in significantly higher enolization barriers (Table 1). For disaccharides, the presence of a free reducing end is critical: lactose retains a free reducing glucose unit, whereas sucrose lacks any free aldehyde or ketone group and must undergo prior hydrolysis (Kroh, 1994), as reflected in its high condensation barrier (159.90 kJ/mol).

Fig. 5.

Potential energy diagram for the reaction of leucine with different types of sugar in Amadori rearrangements to form dicarbonyl compounds. A: Under neutral, weakly basic conditions; B: Under acidic conditions.

During the subsequent thermal decomposition of Amadori rearrangement products to form dicarbonyl compounds, the energy barrier distribution across different sugar reaction pathways exhibits pronounced pH dependency and structural correlation. As shown in Fig. 5A, under neutral to weakly alkaline conditions, where 2,3-enolization is the key reaction phase, the highest energy barrier for the glucose-leucine system is 138.59 kJ/mol, followed by lactose (184.30 kJ/mol), xylose (189.95 kJ/mol), erythrose (195.10 kJ/mol), arabinose (223.51 kJ/mol), galactose (230.94 kJ/mol), maltose (250.27 kJ/mol), fructose (253.28 kJ/mol), sucrose (348.89 kJ/mol), and mannose (422.10 kJ/mol). Under acidic conditions (Fig. 5B), where 1,2-enolization is the key reaction phase, the highest energy barrier for the glucose-leucine system is 169.49 kJ/mol, followed by lactose (249.94 kJ/mol), erythrose (254.90 kJ/mol), fructose (254.95 kJ/mol), sucrose (260.27 kJ/mol), mannose (310.13 kJ/mol), maltose (343.99 kJ/mol), galactose (357.64 kJ/mol), arabinose (483.93 kJ/mol), and xylose (551.33 kJ/mol).

Further analysis of the mechanistic differences among various sugars reveals that sugar structural characteristics exert a decisive influence on their reactivity. In the reaction pathway of the tetrose sugar (erythrose), its initial carbonyl-amine condensation energy barrier (146.12 kJ/mol) is already significantly higher than that of glucose, and it exhibits relatively high energy barriers in the subsequent enolization stages (254.90 kJ/mol under acidic conditions, 195.10 kJ/mol under alkaline conditions). This is attributed to differences in ring structure stability resulting from its shorter carbon chain and restricted open-chain aldehyde conformation. Among pentoses, although xylose shows a relatively low energy barrier in the initial condensation step (137.78 kJ/mol), it exhibits an extremely high energy barrier (551.33 kJ/mol) during the 1,2-enolization phase under acidic conditions. This phenomenon arises from two factors. First, the furanose ring of xylose is less flexible than pyranose rings, restricting the conformational rearrangement needed for proton transfer (Serianni & Barker, 1984). Second, under acidic pH, the ring oxygen can be protonated, creating a positive charge that electrostatically repels the incipient positive charge on the transferring proton. In contrast, glucose's pyranose ring can adopt a more favorable chair conformation that aligns the reacting centers with minimal distortion. This explanation is supported by the fact that arabinose (another pentose) also shows a very high barrier (483.93 kJ/mol) under acidic conditions, while hexoses (except mannose) have much lower barriers.

Among hexoses, glucose maintains the lowest energy barriers across all pH conditions, while fructose and galactose show significantly elevated energy barriers due to differences in carbonyl position and hydroxyl configuration (galactose is a C4 epimer). Mannose exhibits an extremely high energy barrier under neutral to weakly alkaline conditions (422.10 kJ/mol), which may be associated with steric hindrance effects resulting from the axial orientation of its C2 hydroxyl group. These observations indicate that subtle structural differences among hexoses can substantially impact reactivity. Among disaccharides, lactose exhibits relatively low energy barriers (249.94 kJ/mol under acidic conditions, 184.30 kJ/mol under alkaline conditions), whereas maltose and sucrose display higher energy barriers (maltose: 250.27 kJ/mol alkaline, 343.99 kJ/mol acidic; sucrose: 348.89 kJ/mol alkaline, 260.27 kJ/mol acidic). This is closely related to their glycosidic bond types (lactose has β-1,4 glycosidic bond, maltose has α-1,4 glycosidic bond, sucrose has α-1,2 glycosidic bond) and the accessibility of the reducing end. These energy barrier differences are further corroborated by reaction rate constants (Fig. 6). Rate constants also show that the reaction rates for the glucose-leucine system are generally higher than those for other sugar-leucine systems across all temperatures, with its rate advantage becoming more pronounced as temperature increases. This indicates that glucose not only possesses the lowest energy barriers thermodynamically but also exhibits the fastest reaction progress kinetically.

Fig. 6.

Rate constants for the reaction of leucine with different types of sugar in Amadori rearrangements to form dicarbonyl compounds. A: Amadori arrangement; B: Under neutral, weakly basic condition; C: Under acidic condition.

In summary, through comparative analysis of thermodynamic and kinetic parameters for leucine with 10 structurally distinct sugars during the Amadori rearrangement and subsequent dicarbonyl compound formation stages, it is unequivocally established that glucose is the optimal sugar for undergoing the Maillard reaction with leucine, and this advantage is consistently maintained under neutral, weakly alkaline, and acidic conditions. Based on the highest energy barriers (primarily under neutral to weakly alkaline conditions, with consideration of performance under acidic conditions), the order of priority for sugars to undergo the Maillard reaction with leucine is: Glucose > Lactose > Xylose > Erythrose > Arabinose > Galactose > Maltose > Fructose > Sucrose > Mannose. This conclusion not only reveals the intrinsic relationship between sugar structural characteristics and their Maillard reaction reactivity at the molecular level but also provides an important theoretical basis for selectively regulating the Maillard reaction process and directionally generating target flavor compounds based on sugar selection in food processing. It should be noted that our two-step screening strategy did not cross-validate all possible sugar-amino acid pairs; therefore, the conclusion that leucine and glucose are the most reactive pair is strictly based on the compared sets.

4. Conclusions

This DFT study investigated the Amadori rearrangement mechanisms in the Maillard reaction between glucose and 20 amino acids, and between leucine and 10 structurally diverse sugars. The rate-determining carbonyl-amine condensation step was analyzed across all pathways. Leucine exhibited the lowest energy barrier among amino acids reacting with glucose, significantly lower than others. During subsequent dicarbonyl formation, leucine maintained the lowest maximum energy barriers under both neutral/weakly alkaline and acidic conditions, with consistently higher rate constants across temperatures, establishing leucine as the most reactive amino acid. For sugars reacting with leucine, structural features including carbon chain length, carbonyl position, and glycosidic bond type profoundly affected reactivity. Glucose showed the lowest condensation energy barrier compared to other sugars, and maintained the lowest maximum energy barriers during dicarbonyl formation under all pH conditions. Based on comprehensive energy barrier analysis, the reactivity order with leucine was: glucose > lactose > xylose > erythrose > arabinose > galactose > maltose > fructose > sucrose > mannose. Leucine's isobutyl side chain enables effective transition state stabilization across pH conditions, while glucose's aldohexose structure provides favorable electrophilicity and spatial orientation for nucleophilic attack. These findings provide theoretical insights into Maillard reaction selectivity and establish a molecular basis for targeted regulation of reaction products in food processing applications.

It should be noted that this study is purely computational and has not yet been experimentally validated. The idealized models used in the calculations differ to some extent from actual food processing systems. Furthermore, our two-step screening strategy did not cross-validate all possible sugar-amino acid pairs; therefore, the conclusion that leucine and glucose are the most reactive pair is strictly based on the compared sets. Consequently, the obtained reactivity orders and energy barrier values await future experimental verification and refinement. We recommend that conventional methods such as UV-Vis spectrophotometry for monitoring browning processes and HPLC-MS for quantifying Amadori rearrangement products be employed to validate the theoretical predictions. We hope that our computational results provide a valuable theoretical reference for subsequent experimental studies.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Handling Editor: Professor Alejandro G.Marangoni

Appendix A. Supplementary data

The following is the Supplementary data to this article: