Introduction

Flavor represents a key sensory characteristic for food’s acceptability1, and food flavorings are often used as an ingredient to improve or enhance the food’s flavor profile. Flavor compounds bind to proteins, potentially influencing sensory perception2. Added flavor compounds interact with proteins in several ways. These include weak and reversible bonds like hydrophobic interactions, hydrogen bonds, van der Waals forces, and ionic/electrostatic forces, as well as stronger, non-reversible covalent linkages3.

Many studies have examined the effects of flavor addition and subsequent binding to animal- or plant-protein-based systems, mainly by using purified and extracted protein fractions4,5,6. Protein isolates (PIs) and concentrates are used as meat and dairy replacers in the food industry for practical reasons. PIs can vary in composition and may not always consist of a single, uniform type of protein. Instead, they are characterized by different protein fractions7. Among the most commonly used ingredients are soy protein isolate (SPI) and pea proteins. Lupin protein isolate (LPI) is gradually gaining popularity because of its beneficial emulsification, foam stabilization, and gel formation properties8.

Protein-flavor interactions have been addressed using mathematical models to unravel the binding parameters, nature, and strength of these interactions. For instance, Harrison and Hill’s9 model relied on the mass transfer theory to predict flavor release from aqueous emulsions. Additionally, the Scatchard plot, or its adaptation as the Klotz plots10,11,12, served as a standard method for interpreting binding data. These studies predominantly involve defatted and single protein fractions within a model system. However, understanding whether this interaction persists under dynamic conditions, such as oral processing during food consumption, is highly significant for optimal food design13. In-mouth interactions can occur between proteins and flavors but also between salivary proteins and flavor compounds. As a result of these interactions, aroma release is slowed down, thus, affecting the food’s aroma perception14,15,16,17,18. Saliva comprises 98% water and ~2% salts, organic compounds, and several proteins. Mucin (M) is the most abundant proteinaceous material (~ 0.3%) in saliva19, together with α-amylase, immunoglobulin, statherin, histatin, proline-rich proteins, and lactoferrin14. It is characterized by an amphiphilic and heavily glycosylated nature. Within the food-flavor interactions domain, there is a lack of data regarding the in vitro interplay between flavored protein-based aqueous model systems (FPBAS), utilizing PIs, and mucin.

Therefore, this study aims to investigate (1) the binding parameters (binding sites, n, and binding constant, K) for each PI when combined with each flavor compound and (2) the contribution of mucin to the protein-flavor binding process. To achieve these goals, two plant-based PIs (SPI and LPI) and one animal-based PI (whey protein isolate (WPI)) were selected. PIs were combined with a homologous series of carbonyl compounds and/or pig gastric mucin. The interactions in all sample combinations were analyzed using static headspace gas chromatography-mass spectrometry (HS-GC-MS).

Results

Determination of the binding parameters of protein-flavor aqueous model systems

Figure 1A–F displays the Klotz plots of the binding of carbonyl compounds to WPI, LPI, and SPI in mucin-free solutions. The binding parameters (n and K) were determined by generating double-reciprocal plots from the Klotz Eq. (1).

Fig. 1: Klotz plots for binding of carbonyl compounds to commercial food protein isolates (PIs).
figure 1

Whey (WPI) (A, B) soy (SPI) (C, D) and lupin (LPI) (E, F) protein isolates. Plots were calculated following Eq. (1). To optimize data visualization, scales are adjusted differently for the X-axis (1/[L], being L the free ligand concentration in the aqueous phase) and the Y-axis (1/v, being v the number of moles of ligand bound per mole of total protein). These adjustments are made based on their respective maximum responses.

Full size image

Thus, Fig. 1A–F illustrates the linearity of the plots between PIs and carbonyl compounds in mucin-free samples.

The binding parameters (n and K) and linear equations for binding carbonyl compounds to PIs were derived from Eq. (1), and their values are shown in Table 1. For most of the protein-flavor combinations (Table 1), the coefficients of determination (R2) were >0.94, indicating that the equations explained >94% of the total variation for the plots. The R2 values of 2-hexanone were excluded due to a low score (R2 < 0.1), indicating a lack of fit between the model and the data. Its high volatility (Table S1) might limit its binding to proteins, leading to inaccurate parameter quantification.

Table 1 Linear equations and binding parameters (n = binding sites and K = binding constant) obtained from Klotz plots for the binding of carbonyl compounds to commercial food protein isolates (PIs) (whey, WPI, soy, SPI, and lupin, LPI)
Full size table

The number of binding sites (n) differed among the different PIs and was influenced by the type of flavor. Experimental values for n varied from n = 0.021 to n = 7.194. While some exhibit patterns, such as increased n with longer chain lengths for ketones and SPI, the opposite holds for aldehydes and SPI (Table 1).

The binding constant (K) values exhibited variation across different PIs and were influenced by the type of flavors. Values varied from K = 6.8·102 M-1 to K = 1.2·106 M-1 (Table 1).

Role of mucin in the protein-flavor binding mechanism

An artificial saliva solution was utilized to better understand the intricate dynamics between proteins and flavors when combined with mucin, mucin being the main ingredient of (artificial) saliva. HS-GC-MS analyses were performed, and the binding percentage was calculated Eqs. (2) and (3), as illustrated in Figs. 2 and 3. Figure 2 illustrates the binding effect of mucin (M) (0.01(w/v)%) on the protein-flavor binding mechanism in protein-aldehyde-based aqueous model systems (PAB) and protein-ketone-based aqueous model systems (PKB), increasing in chain length from C6 to C10. Controls without mucin (e.g., solely protein and flavor) were included for comparison.

Fig. 2: Effect of mucin (M) (0.01(w/v)%) on the protein-flavor binding mechanism.
figure 2

The influence of mucin was studied in A Protein-aldehyde-based aqueous model systems (PAB) and B Protein-ketone-based aqueous model systems (PKB) increasing in chain length from C6 to C10 The abbreviations “A_” and “K_” indicated the chemical class (aldehydes or ketones), followed by the chain length. Binding (%) was calculated following Eqs. (2) and (3). Results are expressed as the mean ± standard deviation. Letters denote significant differences (p  <  0.05). Treatments with the same letter are not significantly different.

Full size image

Comparing mucin-free solutions, PAB exhibited a higher level of binding (Fig. 2A) than PKB (Fig. 2B). The binding (%) ranged from 24.3% ± 2 to 96.0% ± 3 for PAB, whereas for PKB, the binding (%) varied from 4.7% ± 1 to 74.5% ± 2 (Table S2). No significant differences were found across the studied PIs.

As seen in Fig. 2, a greater effect occurred with the most hydrophilic compounds and short-chain flavor compounds (i.e., 2-hexanone, 2-heptanone, hexanal, and heptanal) compared to the most hydrophobic ones and long-chain flavor compounds (i.e., 2-nonanone, 2-decanone, nonanal, and decanal). However, regardless of the flavor compound or protein used (WPI, SPI, or LPI), the addition of mucin (0.01(w/v)%) to either PAB or PKB strongly increased flavor binding (Fig. 2).

The addition of mucin had a more pronounced effect on PKB (Fig. 2B) than on PAB (Fig. 2A). Compared to mucin-free solutions, binding (%) was up to fifteen times higher in protein-ketone-mucin-based systems (PKMB) and up to three times higher in protein-aldehyde-mucin-based systems (PAMB). Binding (%) ranged from 49.5% ± 2 to 86.2% ± 3 for PKMB and 74.8% ± 1 to 93.0% ± 1 for PAMB (Table S2).

To clarify further the interplay between mucin, flavor, and PI, the fluorescence spectra of tryptophan were analyzed. Since no significant differences were previously reported among the studied WPI, SPI, or LPI (Fig. 2), the comparison was limited to the animal protein (WPI) and one plant protein (SPI). Figure S1A–E illustrates the effect of mucin concentrations (0.01(w/v)% and 0.1(w/v)%) on protein-flavor binding upon the addition of each of the flavor compounds (5 mg/L) individually. The results indicated that the addition of mucin results in a decrease in fluorescence intensity, (Fig. S1A–E).

To gain deeper insights into the role of mucin in flavored protein-based aqueous model systems (FPBAS: PAMB and PKMB) and considering the significant variability of mucin levels among individuals20,21, mucin concentrations were increased to 0.1(w/v)%. Figure 3 shows the effect of mucin (0.1(w/v)%) on PKB, increasing in chain length from C6 to C10. Since no significant differences were previously reported among the studied WPI, SPI, or LPI (Fig. 2), the comparison was limited to the animal protein (WPI) and one plant protein (SPI) (Fig. 3).

Fig. 3: Effect of mucin (M) 0.1(w/v)%, on the protein-flavor binding mechanism.
figure 3

The abbreviation “K_” indicated the chemical class (ketones), increasing in chain length from C6 to C10. Binding (%) was calculated following Eq. (3). Results are expressed as the mean ± standard deviation. Letters denote significant differences (p < 0.05). Treatments with the same letter are not significantly different.

Full size image

As seen in Fig. 3, flavor binding is still noticeably higher upon mucin addition (0.1(w/v)%) to PKB as already seen in Fig. 2. Compared to mucin-free solutions, binding (%) was up to ten times higher in PKMB. The binding (%) ranged from 4.7% ± 1 to 74.5% ± 2 for PKB, whereas for PKMB, the binding (%) varied from -22.83% ± 21 to 81.72% ± 5 (Table S2). No significant differences were found across the studied PIs.

Discussion

The influence of mucin in the protein-flavor binding mechanism was evaluated in a model system. However, due to mucin’s high polydispersity and glycosylation nature, it seems unfeasible to estimate binding parameters for this protein. Mucin peptide chains may vary in size, molecular weight (5.0·105-2.0·106 g/mol), and the number of basic units imparting a non-uniform character22,23. Therefore, the number of binding sites (n) and binding constants (K) of commercial food protein isolates and flavors were calculated using Klotz plots. The good linearity of the double reciprocal plot (Fig. 1) indicates that the flavors bind to the proteins independently in non-cooperative interactions. In non-cooperative interactions, the binding of one ligand does not influence the binding strength/affinity of a second ligand. As a result, each binding event occurs independently, resulting in a linear relationship11,24,25,26.

Overall, the n values obtained in Table 1 are in line with previous research findings in which the binding between proteins and various flavor compounds was investigated: soy-ketones (n = 4)11, soy-vanillin (n = 0.48)27, soy-maltol (n = 3.27)12, pea-hexanal (n = 4.84)4, whey-vanillin (n = 0.67)27, soy-citronellol (n = 4.2)28, pea-methyl anthranilate (n = 0.68)24, and coconut-vanillin (n = 1.47)29. The reported n values do not necessarily signify the total potential binding sites. Protein processing, such as denaturation30,31, may alter, change the distribution, or lose some binding sites.

Klotz plots assume that protein binding sites are uniform in number and function32. However, proteins harbor multiple ligand-binding sites that exhibit non-identical subunits differing in structure and spatial arrangement, thus lacking equivalence33. Therefore, proteins may exhibit both high-affinity and lower-affinity binding sites32 due to the different amino acid composition of the binding site34 or structural rearrangement. This could explain the lack of a clear trend regarding the obtained n values (Table 1).

Lower K values indicate a weaker affinity between the proteins and flavors, while higher K values suggest stronger binding. Typically, longer flavor chains are associated with higher K, indicating hydrophobic interactions26. In Table 1, the data shows that aldehydes exhibited higher K values compared to ketones. An inner location of the keto group (i.e., ketones vs. aldehydes) may have led to steric hindrance hindering flavor binding11,35,36,37. Although the experimental conditions may differ across different binding studies12,26,27,28 there is still a clear consistency concerning the significant role of flavor’s structural properties (functional group and location) in determining n and K values.

Despite ketones potentially hindering binding and their known lower affinity for proteins compared to aldehydes38,39, the higher hydrophilic nature of ketones compared to aldehydes (Table S1) might be considered a key factor for its greatest effect on PKMB systems.

It is well known that aldehydes not only bind via reversible and weak hydrophobic interactions to flavor compounds but can also participate in irreversible covalent binding through Schiff base formation, where they react with amino groups of the protein to form an imine linkage3, resulting in stronger bonds. Likewise, as the flavor´s chain length increases, the hydrophobicity also increases, reducing the impact of mucin addition.

Adding mucin to either PAB or PKB resulted in a combined effect that greatly surpassed the individual binding contributions (Fig. 2). While research on flavor binding in protein solutions with mucin is limited, conformational changes have been observed. According to dynamic light scattering results from Çelebioğlu et al.40 the interaction between β-Lactoglobulin (BLG) and mucin resulted in a more compact mucin conformation as BLG integrated with it. Ahmad et al.41 investigated how incorporating citrus peel pectin into a mucin model system resulted in a significant reduction in fluorescence intensity, suggesting that citrus peel pectin induces changes in the microenvironment around mucin fluorophores, likely due to alterations in mucin’s structural conformation.

From the results obtained in Fig. S1, conformational changes in the tryptophan residues occur due to either the binding of tryptophan residues or ligand binding42.

The results of this study show that mucin, interacting with proteins enhances flavor binding. Although the exact mechanism is unclear, structural conformational changes may reveal hidden hydrophobic pockets.

Notably, as mucin levels increase (0.1(w/v)%), the unfolding either persists or intensifies, leading to pronounced aggregation. This aggregation limits the accessibility of flavor molecules to the binding sites.

The negative binding values obtained at higher mucin concentrations (0.1(w/v)%) (Fig. 3) are intriguing and deserve some explanation. Typically, the most hydrophilic flavors exhibit lower affinities for protein-flavor binding. Viry et al.43 and Snel et al.44 elucidated the reduced binding capacities of hydrophilic flavor compounds through a mechanism named the size-exclusion effect. This phenomenon, driven by steric hindrance, excludes small and hydrophilic flavor compounds from the solution, thereby being expelled into the headspace43,44. Therefore, a higher concentration of mucin may have led to the push-out of the volatile compounds into the headspace, which may explain the negative binding value (Fig. 3).

Although mucins have a negative charge, primarily from sialic acid residues and sulfated sugars, mucins typically exhibit low isoelectric points (pI 2–3.0)45, mucin also contains positively charged patches in the non-glycosylated globular regions composed of histidine, arginine, and lysine residues, which may attract the negatively charged WPI45,46,47,48. Figure 4 illustrates the proposed mechanism: when mucin encounters the food proteins, it may adhere to the protein surface by hydrogen bonding, hydrophobic and/or electrostatic interactions45,46,47,48.

Sarkar et al.19 proposed that lactoferrin-stabilized emulsion droplets interacted with mucin through electrostatic forces. They observed excessive mucin in the continuous phase, which led to depletion-type flocculation and more complex aggregations, including the self-association of mucin molecules. This phenomenon would explain that higher aggregation of mucin at higher concentration may cover the food protein binding sites hiding them from flavor interaction.

Fig. 4: Mechanism of mucin interaction in flavored protein-based aqueous systems (FPBAS).
figure 4

A Mucin–protein interaction mechanism in FPBAS at low mucin concentration (0.01(w/v)%). B Corresponding mechanism observed at medium to high mucin levels (0.1(w/v)%). Adapted from Brown et al. [52].

Full size image

In brief, with 0.01(w/v)% mucin concentration, the potential exists for inducing structural alterations, (i.e., partial unfolding) exposing the binding sites, and enhancing flavor binding (Fig. 4A). At 0.1(w/v)% mucin, unfolding continues, causing aggregation that limits flavor molecules’ access to binding sites (Fig. 4B).

In summary, binding parameters (binding sites, n, and binding constant, K) were determined from Klotz plots in FPBAS (PAB and PKB). The results suggested a non-cooperative, independent, and linear relationship between proteins and flavor compounds. The structural characteristics of the flavors were shown to be significant in determining n and K values, with chain length and position of the functional group being key features of the binding mechanism, whereas protein sources showed only a minor impact on flavor binding.

Adding 0.01(w/v)% of mucin to FPBAS significantly increased flavor binding, irrespective of the flavor compound or protein source, potentially suggesting protein unfolding and the exposure of previously hidden hydrophobic pockets. Whereas increasing the mucin concentration to 0.1(w/v)% did not yield a subsequent increase in flavor binding, suggesting the possibility of a coating mechanism occurring on the available food protein binding sites.

These findings aid in comprehending the essential aspects of flavor binding in the presence and absence of mucin, as well as in the presence of different mucin concentrations, which are crucial variables when designing and developing novel food products.

The study has some limitations in mimicking what happens during the physiological mastication process. These results were obtained in an aqueous model system at equilibrium, which may not fully replicate all the intricate processes occurring dynamically in the mouth. Therefore, complementary sensory evaluation techniques might need to be explored to provide a more accurate reflection of the actual impact on consumer perception compared to analytical approaches alone.

Methods

Food flavor stock solutions preparation

Ten analytical grade (≥98%) flavor compounds were selected based on their spatial configuration, chain length, and wide use in the food industry: 2-hexanone, 2-heptanone, 2-octanone, 2-nonanone, 2-decanone, hexanal, heptanal, octanal, nonanal, and decanal, which were purchased from Sigma-Aldrich (Zwijndrecht, the Netherlands) (Table S1). Each of the chosen flavor compounds was individually prepared in a sodium phosphate buffer solution (Na2HPO4 and NaH2PO4·2H2O, from Sigma-Aldrich at pH 7.0, 50 mM) following an adapted protocol based on Barallat-Pérez et al.35. Five initial concentrations were prepared for each flavor: 1, 2.5, 5, 10, and 20 mg/L. To ensure complete dissolution, the flavor stock solutions underwent ultrasonic treatment in a water bath (Elma Schmidbauer GmbH, Singen, Germany) for 1 h at 30 °C.

Food protein solutions preparation

An adapted approach from Wang and Arntfield31 was employed to prepare protein solutions. The investigation used SPI SUPRO® XT219D IP from Solae (St. Louis, Missouri, USA), LPI 10600 from ProLupin (Grimmen, Germany), and WPI BiPro® from Davisco International (Le Sueur, Minnesota, USA). Table S3 displays the manufacturer’s specifications. Proteins were selected based on their chemical structure, physicochemical properties, and prevalence in plant-based food substitutes. The protein batches were packed and stored in a cool (10–15 °C) and dry environment to minimize result discrepancies. SPI, LPI, and WPI were individually utilized at an initial concentration of 2(w/v)% in sodium phosphate buffer (pH 7.0, 50 mM). The samples underwent vortexing for 10–20 s (3200 rpm, using Genie II, GenieTM, Sigma-Aldrich, Florida, USA). Then, they were placed in an ultrasonic water bath at 30 °C for 20 min to ensure thorough solution blending. Multiple vortexing cycles followed, lasting 10–20 s each, to achieve uniform mixture distribution.

Artificial saliva preparation

Artificial saliva was prepared at room temperature (20-22 °C) and pH 7.0, 50 mM, following an adapted version of van Ruth et al.49, which included: NaHCO3, K2HPO4·3H2O, NaCl, KCl, CaCl2·2H2O, NaN3, and pig gastric mucin, provided by Sigma-Aldrich (Zwijndrecht, the Netherlands). To prevent clumping, mucin was added very gradually to the solution. Considering the significant variability of mucin levels among individuals (1190 ± 0.1-3010 ± 1.0 mg/L)20,21 due to differences in e.g., genetics, environmental factors, age, hydration status, oral health, use of medications, and inter-and intra-individual variability, artificial saliva was prepared at two different concentrations of mucin: 0.1(w/v)%, and 0.01(w/v)%.

Binding parameters determination

Binding parameters were computed from the Scatchard plots and/or Klotz plots11,50,51 Eq. (1) to increase understanding of the mechanisms involved.

$$\frac{1}{v}=\frac{1}{n}+\frac{1}{{Kn}[L]}$$
(1)

where [L] is the free flavor concentration in the aqueous phase (mol/L); v is the number of moles of flavor bound per mole of total protein (mol/mol); K is the binding constant (M-1), and n is the number of independent binding sites. The estimated molecular weight for all PIs was set at 100000 g/mol as per findings from prior scientific literature11,50,51. Binding parameters (n and K) were calculated from the y-intercepts (1/n) and slopes of the plots (1/Kn), respectively.

Gas chromatography-mass spectrometry samples preparation

For the HS-GC-MS analysis, samples were prepared using a modified method from Wang and Arntfield, 201531 and Barallat-Pérez et al.35. Control samples comprised food protein and saliva solutions without added flavor. Vials were sealed and kept in a water bath shaker (SW22, Julabo GmbH, Seelbach, Germany) at 30 °C and 125 rpm for 3 h before HS analysis. Triplicate samples were prepared for analysis.

Binding measurement and calculation

Protein-flavor-mucin binding was assessed by HS through GC-MS (Agilent- 7890 A GC coupled to an Agilent 5975 C with triple-axis detector MS, Agilent, Amstelveen, the Netherlands) following a modified method from Wang and Arntfield, 201531 and Barallat-Pérez et al.35. As described by Barallat-Pérez et al.35, the GC operated in 1:10 split mode with a split flow of 8 mL/min. Samples were incubated and agitated for 14 min at 40 °C. A DB-WAX column (20 m × 180 μm × 0.3 μm) maintained a constant flow of 0.8 mL/min, with the temperature programmed to rise at 40 °C/min to a maximum of 240 °C. The mass spectrometer operated at 70 eV and scanned a mass range of 35–200 g/mol. Quantitative flavor analysis was performed using MassHunter Quantitative Analysis software (MSD ChemStation F.01.03.2357), while chemical and physical data for the selected flavor compounds were sourced from the NIST Mass Spectrometry Library (InChI Library v.105). To avoid competitive binding at protein sites, flavor compounds were examined individually. Flavor binding to proteins, expressed as a percentage in the absence and presence of protein, was calculated Eq. (2).

$$Binding\,({\rm{ \% }})=\left(1-\frac{\begin{array}{c}HS\,1\,-HS\,2\end{array}}{HS\,3}\right)\,x\,100$$
(2)

HS1 represents the headspace abundance, corresponding to the peak response of the flavored aqueous protein solution. HS2 indicates the headspace abundance of the protein solution without added flavor. HS3 reflects the headspace abundance when protein is absent, capturing only the response of the flavor compound. For each flavor compound, key ions and retention times were identified and quantified accordingly.

Flavor binding to mucin was calculated and expressed in % in the absence and presence of mucin Eq. (3).

$$Binding\,({\rm{ \% }})=1-\frac{\begin{array}{c}HS\,1\,-HS\,2-HS\,4-HS\,5\end{array}}{HS\,3}\,x\,100$$
(3)

where HS4 (saliva solution + buffer) represents the HS in the absence of flavor and HS5 (protein solution + saliva solution) is the abundance in the HS of the protein-based saliva solution.

Fluorescence quenching and determination of interaction parameters

Fluorescence quenching was performed using a Varian Cary Eclipse (Agilent, Santa Clara, CA) to study the interaction between PI, flavor compounds, and mucin. Fluorescence intensity was recorded at 0.1(w/v)%, and 0.01(w/v)%. mucin concentrations at 25 °C, using an excitation wavelength of 280 nm for tryptophan.

Statistical analysis

Binding data were determined in triplicate, with results presented as mean ± standard deviation. Statistical analysis was performed using RStudio 4.2.1. (Boston, Massachusetts, USA) for a two-way ANOVA on each sample combination. Post-hoc Tukey tests determined significant differences (p < 0.05) between samples.