Elucidating odorant synergy in red wine: through olfactory receptor-based profiling

Abstract

The flavor complexity of red wine stems from diverse aroma compounds and their interactions. Key odorants 2,3,5-trimethylpyrazine (TMP) and furfuryl alcohol (FA) contribute roasted/earthy and sweet/caramel-like notes, respectively, collectively shaping the wine’s sensory profile. However, the mechanism underlying their olfactory interaction remains unclear. In this study, human sensory evaluation revealed a mutual synergistic enhancement between TMP and FA. Cellular models expressing cognate receptors OR5K1 (TMP-specific) and OR2W1 (FA-responsive) were established, with intracellular cAMP detection confirming receptor-level synergy. Further molecular simulations predicted non-competitive binding mechanisms, as both odorants occupied distinct sites within their respective receptor pockets, consistent with observed non-antagonistic synergy. This work elucidates the molecular perceptual mechanism underlying key wine aroma interactions, providing a novel theoretical framework for understanding red wine flavor complexity.

Introduction

The intricate and captivating aroma profile of wine is a symphony composed of a vast array of volatile compounds. This complex matrix encompasses hundreds of molecules, including monoterpenes responsible for floral and citrus notes, esters contributing to fresh fruity aromas, volatile phenols imparting spicy or smoky nuances, and various heterocyclic compounds that often provide roasted or caramel-like dimensions^1,2. These diverse odorants are traditionally classified into three categories based on their origin: primary aromas derived from the grape itself (e.g., varietal thiols in Sauvignon Blanc)³, secondary aromas formed during alcoholic and malolactic fermentation (e.g., esters like ethyl acetate)⁴, and tertiary aromas that develop during aging and maturation in bottle or oak barrels, often involving chemical transformations like oxidation and esterification⁵. Among this chemical diversity, pyrazine compounds and furfuryl alcohol (FA) represent two significant heterocyclic compounds^6,7. 2,3,5-Trimethylpyrazine (TMP) ranks among the most significant pyrazine compounds, characterized by its distinctive nutty and roasted coffee aroma⁸. Furfuryl alcohol is characterized by its potent sweet, caramel-like aroma, primarily formed through thermal degradation of pentose sugars or Maillard reactions during winemaking^6,7. In summary, pyrazines and furfuryl alcohol represent key aromatic components that contribute significantly to the fermented and roasted flavor profiles in red wine. Therefore, investigating the interactions between these two compound classes is of considerable scientific interest.

Flavor interaction research represents an interdisciplinary field integrating instrumental analysis with sensory science. The application of sophisticated analytical tools, particularly gas chromatography-olfactometry-mass spectrometry (GC-O-MS), has enabled precise identification of key odorants and their interactions in complex alcoholic beverages including wine, whiskey, and baijiu^9–11. The multifaceted nature of wine flavors arises from a dynamic interplay between odorant molecules, with synergistic intensification and competitive masking of specific aroma notes being principal manifestations. To better understand the complex and integrated perception of aroma substances, particularly through methods such as the S-curve model, has become critical for understanding the complex sensory profiles of food products. Aroma compounds interact synergistically or antagonistically to influence the overall perception of flavor, for example, acetate can enhance the floral and fruity character of red wine¹². However, relying solely on sensory evaluation to study key aroma interactions in wines introduces unavoidable subjectivity and variability, limiting mechanistic understanding. Therefore, molecular biological approaches are becoming crucial for deciphering precise odorant interaction mechanisms.

As a complex food matrix, red wine contains a diverse array of aroma components. However, there is a paucity of research on the effects of interactions between key components (e.g., TMP and FA) on human olfactory perception, especially studies that leverage human olfactory receptors as the experimental basis. The molecular mechanisms underlying human olfactory perception and odorant interactions can be elucidated at the receptor level through advanced cellular biological approaches. Currently, olfactory response receptors for a few key aroma components in red wine have been identified, including sotolone in red wine, which can activate the olfactory receptor OR8D1, and 1,1,6-trimethyl-1,2-dihydronaphthalene (TDN)—the compound responsible for the gasoline-like aroma in Riesling wine—that selectively activates the olfactory receptor OR8H1^13,14. Furthermore, researchers identified OR2T11 as a receptor specifically responsive to small-molecule thiols, highlighting the essential function of metal ions in mediating odorant-receptor interactions¹⁵. Following this discovery, further research confirmed that some floral odorants can effectively suppress the smell of sulfides, with β-ionone serving as a potent antagonist through its ability to downregulate OR2T11 response to sulfide compounds¹⁶. These research findings can provide new insights into the olfactory perception of key sulfur-containing aroma compounds in red wine, as well as their perceptual interactions with other types of aroma compounds. In general, systematic characterization of ORs activation patterns has significantly advanced our understanding of food flavor^17–19.

Based on the aforementioned background, this study systematically investigated the interaction between pyrazines and furfuryl alcohol in red wine. The research commenced with sensory experiments to determine the olfactory detection thresholds of TMP and FA, along with their perceptual interactions. Subsequently, an olfactory receptor heterologous expression system was established and the modulatory effects of FA on TMP-induced OR5K1 activation were further examined. Parallel experiments employed the broad-spectrum receptor OR2W1 to assess TMP’s influence on FA response. Finally, molecular modeling provided mechanistic insights into these odorant interactions at both receptor binding sites.

Results

Determining interaction effects of TMP and FA by sensory evaluation

The phenomenon of olfactory interaction between TMP (exhibiting a bakery-like odor) and FA (characterized by a fermented aroma) was demonstrated through designed sensory experiments. In this section, the olfactory detection thresholds were determined for both the individual compounds (TMP and FA) and the binary mixture, enabling their interaction effects to be systematically compared. TMP and FA olfactory detection threshold concentrations were 9.37 mmol/L and 33.12 mmol/L, respectively (Fig. 1a, d; Table S3). Subsequently, the thresholds of TMP and FA were examined again after equimolar mixing (each odorant maintains the same concentration as its single-substance solution). In the binary system, the detection thresholds were 3.05 mmol/L for TMP and 7.07 mmol/L for FA (Fig. 1b, e; Table S3). Based on these results, equimolar mixing leads to a significant decrease in the detection thresholds of both TMP and FA, and their S-curves shifts to the left distinctly (Fig. 1b, e). For TMP, the measured detection threshold is less than the theoretical detection threshold, with D (experimental threshold/theoretical threshold) = 0.33, showing a synergistic effect (Table 1); for FA, the measured detection threshold is less than the theoretical detection threshold, with D (experimental threshold/theoretical threshold) = 0.21, also showing a synergistic effect (Table 1).

Fig. 1. Results of olfactory threshold detection by the 3-AFC method.

(a, d) represent the respective thresholds for TMP and FA, respectively; (c) represents the changes in TMP after mixing with 1000 mmol/L FA; (b, e) represent the changes in the respective thresholds of TMP and FA after equimolar mixing; (f) represents the changes in FA after mixing with 1000 mmol/L TMP. TMP 2,3,5-trimethylpyrazine, FA Furfuryl alcohol.

Table 1.

3-AFC-S curve discrimination of interactions between TMP and FA

Composition combination	Experimental value (mmol/L)	Theoretical value (mmol/L)	D (experimental/theoretical)	Action Effect
TMP + FA (concentration ratio 1:1)	3.06	9.37	0.33	Synergistic
TMP + FA (FA 1000 mmol/L)	1.57	9.37	0.17	Synergistic
FA + TMP (concentration ratio 1:1)	7.07	33.12	0.21	Synergistic
FA + TMP (TMP 1000 mmol/L)	2.57	33.12	0.08	Synergistic

To further investigate the interaction pattern, an alternative experimental approach was employed wherein the concentration of one odorant remained fixed in the binary mixture system while the detection threshold of another one was assessed. First, the concentration of FA was fixed at 1000 mmol/L, and the concentration of TMP was varied to achieve FA: TMP = 10000:1/2000:1/1000:1/200:1/100:1/20:1/10:1/2:1/1:1. The olfactory detection thresholds of TMP was measured as 1.57 mmol/L and the measured detection threshold was less than the theoretical detection threshold, with D (experimental threshold/theoretical threshold) = 0.17, showing a synergistic effect. In turn, to investigate the effect of TMP on FA, the TMP concentrations were fixed at 1000 mmol/L, and the concentration ratio of TMP to FA was varied to achieve TMP: FA = 10000:1/2000:1/1000:1/200:1/100:1/20:1/10:1/2:1/1:1. The thresholds of FA were measured to be 2.57 mmol/L and the measured detection threshold was less than the theoretical detection threshold, D (experimental threshold/theoretical threshold) = 0.08, which also showed a strong synergistic effect (Fig. 1c, f; Table 1).

Effect of odorants on cell viability

To ensure reliable assessment of odorant interactions via olfactory receptors, preliminary cytotoxicity screening was conducted at varying concentrations. Cells were exposed to odorants for 10 min at 24 h post-transfection — a duration consistent with cAMP signal detection protocols. Results indicated no cytotoxicity for TMP and FA at concentrations up to 1000 μmol/L in OR5K1-expressing cells (Fig. 2a–c). An identical viability profile was observed in OR2W1-expressing cells (Fig. 2d–f). No significant cytotoxicity was observed when TMP and FA were administered in combination at different concentration gradients. Based on these findings, subsequent experiments could utilize a concentration range of 1–1000 μmol/L for testing both compounds, as this represents a safe concentration range.

Fig. 2. Determination of cell viability after application of odorants.

(a–c) represent the cell viability of OR5K1-expressing cells treated with varying concentrations of TMP, FA, and TMP + FA, respectively; (d–f) represent the cell viability of OR2W1-expressing cells treated with varying concentrations of TMP, FA, and TMP + FA, respectively. The “Blank” group represents no treatment was performed; the “CON” group represents the treatment with DMSO solvent alone. Results that did not achieve statistical significance were remained unmarked. TMP 2,3,5-trimethylpyrazine, FA Furfuryl alcohol.

Binary interactions between TMP and FA based on OR5K1

The previous sensory experiments have demonstrated that eatable alkanols specifically FA (wine-like) could modulate the sensory perception of TMP (Fig. 1; Table 1). However, the underlying molecular mechanisms remain unexplored at the cellular level. The present study provides mechanistic evidence for odorant-induced modulation of TMP detection at the single-receptor level, elucidating the dynamic regulation of receptor activation. Previous studies have shown that systematic screening of olfactory receptors for TMP identified OR5K1 as a narrowly tuned receptor²⁰. Therefore, OR5K1 serves as an ideal target for studying the interaction between TMP and other odorants. Its potent and selective response to TMP enables systematic investigation of the modulation mechanisms of odorant mixtures at the receptor level. A concentration–response relationship for TMP acting on OR5K1 was established to allow quantitative analysis of the response pattern. The results demonstrated that OR5K1 activation by TMP was dose-dependent, with gradually increasing response amplitudes over the concentration range of 0.1–1000 μmol/L. Nonlinear regression analysis yielded an EC₅₀ value of 65.03 ± 8.65 μmol/L for TMP (Fig. 3a; Table S4). In contrast, FA did not elicit significant receptor activation at any tested concentration, as evidenced by the absence of detectable response signals and the inability to establish a meaningful concentration–response relationship (Fig. 3a).

Fig. 3. Concentration-response curves for individual and binary systems on OR5K1.

(a) represents the concentration-response curves for TMP and FA; (b) represents the concentration-response curves for TMP, FA, and TMP + FA (in equimolar concentrations); (c) represents the time-response curves for TMP, FA, and TMP + FA (in equimolar concentrations) at 1000 μmol/L. All data were mock control-subtracted and normalized to the maximum response, and are shown as mean ± SD (n = 3 or 4). RLU relative luminescence unit, TMP 2,3,5-trimethylpyrazine, FA Furfuryl alcohol.

A concentration series of FA and TMP binary mixtures was prepared at equimolar ratios, ranging from 0.1 μmol/L to 1000 μmol/L for both components (e.g., 0.1 μmol/L FA + 0.1 μmol/L TMP; 0.5 μmol/L FA + 0.5 μmol/L TMP; …; 1000 μmol/L FA + 1000 μmol/L TMP). Compared to the TMP-treated group, under equimolar concentrations of FA and TMP, the cellular response level increased, and this enhancement of receptor signaling was observed at higher concentrations. In the concentration–response curve (Fig. 3b), the binary system exhibited slightly higher response values than TMP treatment alone. In the presence of high concentrations of FA, the activation response of OR5K1 to equimolar TMP was significantly elevated (Fig. 3b, c; Table S4). The enhancing effect of FA on receptor activation by TMP began at an FA concentration of 500 μmol/L and was also present when both compounds reached 1000 μmol/L, resulting in an approximately 20% increase. This confirms that FA enhances the agonistic effect of TMP on OR5K1 at the same concentration. It is worth noting that no potent enhancement on TMP signals was observed at low FA concentrations, indicating that in the binary system, TMP still plays a dominant role in activating its high-affinity receptor OR5K1 at low FA concentrations (Fig. 3b).

Binary interactions between FA and TMP based on the broadly tuned receptor OR2W1

OR2W1 has been reported to be a broadly tuned receptor for key odorants of foods, responding to aldehydes, alcohols, ketones, and so on^21,22. In addition, OR2W1 could respond to FA but not TMP^23,24. Based on this, OR2W1 was therefore selected as an additional key receptor to verify whether TMP influences its agonism by FA. Our experimental results showed that the response of OR2W1 to FA was concentration-dependent from 0.1 to 1000 μmol/L, with an EC₅₀ value of 216.62 ± 16.29 μmol/L (Fig. 4a; Table S4). The ligands that OR2W1 responds to are FA but not TMP, so the core substances investigated on this receptor were FA. In the presence of equimolar TMP, the cellular response to FA was significantly enhanced (Fig. 4b, c; Table S4), characterized by a concentration-dependent relationship in the 0.1–1000 μmol/L range (similar to observations with OR5K1) and culminating in an approximate 40% increase in signal. This suggests that the two odorants will mutually enhance each other from the receptor perspective of TMP and FA.

Fig. 4. Concentration-response curves for individual and binary systems on OR2W1.

(a) represents the concentration-response curves for FA and TMP; (b) represents the concentration-response curves for FA, TMP, and FA + TMP (in equimolar concentrations); (c) represents the time-response curves for FA, TMP, and FA + TMP (in equimolar concentrations) at 1000 μmol/L. All data were mock control-subtracted and normalized to the maximum response, and are shown as mean ± SD (n = 3 or 4). RLU relative luminescence unit, TMP 2,3,5-trimethylpyrazine, FA Furfuryl alcohol.

Molecular modeling elucidates binary odorant interactions

Although assays with olfactory receptor-expressing cells have confirmed the functional interaction between TMP and FA, the underlying structural basis remains elusive. To provide a possible explanation for the synergistic mechanism of such odor molecules, computational molecular modeling has been employed. AlphaFold3 was used to predict the co-folding of odorants and proteins and directly predict binding sites and modes (Fig. 5a, b). For OR5K1, TMP establishes stable hydrogen bonding with S203^5.43, π-π stacking with H155^4.56, and hydrophobic interaction with L255^6.51, potentially constituting its activation mechanism, while FA binds loosely to individual residues and lacks comprehensive interactions, preventing receptor activation (Fig. 5 c, e). As a receptor responsive to FA, molecular docking revealed that the Y259^6.55 residue of OR2W1 forms a stable hydrogen bond with the hydroxyl group of FA. In addition, the furan ring of FA engages in hydrophobic interactions with residue I255^6.51 (Fig. 5d). This interaction may represent a potential activation mechanism. In contrast, TMP only establishes weak hydrophobic interactions with Y259^6.55 in OR2W1, which are insufficient to trigger its activation (Fig. 5f).

Fig. 5. Molecular docking elucidates interactions between odorants and olfactory receptors.

(a) represents the overlapping state in which TMP and FA co-interact with OR5K1; and (b) represents the overlapping state in which FA and TMP co-interact with OR2W1; (c, e) represent OR5K1 interaction modes with FA and TMP, respectively; (d, f) represent OR2W1 interaction modes with FA and TMP, respectively. In the 3D diagram, the green dashed line indicates hydrophobic interactions and the purple dashed line indicates hydrogen bonding. TMP 2,3,5-trimethylpyrazine, FA Furfuryl alcohol.

Based on the AlphaFold 3-predicted structures, 200 ns molecular dynamics (MD) simulations were performed for all odorant-receptor systems. First, the MM-PBSA method was employed to calculate binding free energies (Table 2). The results showed that OR2W1 exhibited lower binding energy with FA than with TMP, suggesting higher affinity for FA. In contrast, TMP showed the lowest binding energy and highest affinity for OR5K1. Decomposition of binding free energies identified the top 10 contributing residues. The key residues of OR2W1 (e.g., M105^3.33, I255^6.51) and OR5K1 (S203^5.43, V206^5.46, L255^6.51, L199^5.39) showed interactions consistent with AlphaFold 3 predictions (Fig. 6a, b). Additionally, the root-mean-square deviation (RMSD) of ligand heavy atoms was calculated following alignment of protein Cα atoms using the AlphaFold 3-predicted structure as a reference. In OR5K1, both ligands remained bound initially, but FA departed from its original binding site around 200 ns, whereas TMP binding remained relatively stable (Fig. 6c). In OR2W1, FA maintained an RMSD of ~0.5 nm, indicating stable binding, while TMP displayed larger RMSD fluctuations and eventually dissociated from the binding pocket (Fig. 6d). Based on the divergent binding modes and lack of competitive inhibition observed in OR5K1 and OR2W1 systems, it is proposed that FA and TMP could co-exist at the binding site to mediate a synergistic co-activation effect. The computational findings are in agreement with both cellular assays and molecular modeling data, collectively supporting the reliability of the docking predictions generated by AlphaFold 3.

Table 2.

Docking binding energy of olfactory receptors to odorants

ΔG Mean ± SEM (kcal/mol)	TMP	FA
OR2W1	−2.65 ± 0.09	−4.42 ± 0.10
OR5K1	−13.34 ± 0.09	−0.92 ± 0.18

Fig. 6. Molecular dynamics elucidates binding of odorants and olfactory receptors.

(a) represents the 10 most important key residues in the interaction of TMP with OR5K1; (b) represents the 10 most important key residues in the interaction of FA with OR2W1; (c) represents the RMSD of TMP (green), FA (blue), and OR5K1 within 200 ns; and (d) represents the RMSD of TMP (green), FA (blue), and OR2W1 within 200 ns. TMP 2,3,5-trimethylpyrazine, FA Furfuryl alcohol.

Discussion

Sensory evaluation serves as a universal and crucial method for studying the interactions of flavor compounds, and scientifically designed experimental protocols facilitate objective verification of odorant interactions^25–27. However, human olfactory perception is a complex process involving receptor binding, neuronal signaling, and central nervous system integration²⁸. Furthermore, environmental and psychological factors can potentially influence the outcomes²⁹. Therefore, physiological data, such as receptor and neural signals, are required to substantiate the findings. A combination of three complementary approaches was employed to investigate the synergistic effects among odorant molecules: psychophysical sensory experiments revealed the existence of the phenomenon, heterologous expression of olfactory receptors verified it at the signal transduction level, and molecular modeling provided structural insights at the molecular level.

At the compounds molecular level, interactions between odorants are governed by structural features and olfactory characteristics. The odor synergy investigated in this study always occurs between molecules that are structurally similar and possess complementary odor characteristics, such as fruity esters (Fig. 1)³⁰. For instance, in red wine, a series of roasted pyrazine compounds interact through complex regulatory mechanisms, resulting in a more authentic and multi-dimensional flavor profile³¹. Key structural features—such as hydrogen-bonding groups (e.g., hydroxyl) and hydrophobic moieties—further modulate the interaction outcomes by determining binding stability and activation efficacy³². Odor type also plays a critical role: perceptually congruent categories (e.g., floral-floral pairs) more readily exhibit synergy, while contrasting qualities (e.g., sweet vs. burnt) often induce masking³³. For instance, fresh aldehydes can enhance the olfactory perception of musky odors³⁴. However, intricate human perception and chemical complexity can influence interactions, as exemplified by the decanal-limonene pair, where synergism and inhibition coexist at different concentration levels³⁵. Therefore, the molecular structure, chemical class, and perceptual properties of odorants collectively determine the overall aroma profile presented by a mixture.

Beyond the inherent properties of odorant compounds, their specific modes of action on ORs profoundly influence intermolecular interactions. Narrowly tuned receptors (e.g., OR5K1) are of particular value owing to their high selectivity, a characteristic that maximizes the reduction of non-ligand interference. OR5K1 has been identified as a receptor with specific responsiveness to pyrazine compounds³⁶, and the potent and specific activation of OR5K1 by 2,3,5-trimethylpyrazine (TMP) constitutes a paradigmatic example. To date, no other in vitro expressed olfactory receptors have been documented to exhibit robust responsiveness to TMP. In cellular assays with binary odorant mixtures, low concentrations of FA showed no significant effect on TMP-induced OR5K1 activation (Fig. 3b, c; Figure S3). A synergistic effect was observed only at high FA concentrations, suggesting that the receptor’s high sensitivity and preferential response to TMP initially dominate its output^20,36. By contrast, while OR2W1 is the most prominent among the receptors responsive to furfuryl alcohol, this receptor has also been demonstrated to exhibit a broader tuning range, being capable of responding to a variety of compounds with distinct chemical structures²³. This further highlights the potential permissiveness of its binding pocket in ligand recognition²⁰. Consequently, odorant interactions, such as synergy, were more readily apparent in OR2W1 assays compared to those involving OR5K1 (Fig. 4b, c; Figure S3). Molecular modeling was employed to anticipate this phenomenon, showing that distinct molecules appear to bind to the same pocket of the olfactory receptor (Fig. 5a, b; Figure S3). This observation aligns with previous studies by Aier et al., suggesting this phenomenon may be attributed to the homology and inclusivity of the olfactory receptor’s binding pocket³⁷. However, entry into the binding pocket does not necessarily lead to receptor activation, as activation requires allosteric changes of specific key residues in addition to odorant binding (Fig. 5; Fig. 6).

The combination of sensory experiments, cellular assays, and molecular modeling can provide a research paradigm for the interaction modes of aroma compounds in food matrices, yet there still remains room for further refinement. There exist certain discrepancies between the experimental concentrations of odorants used in this study and their natural concentrations in real wine matrices, which stem from the inherent technical limitations of current in vitro cellular and sensory detection assays. In real red wine matrices, quantifiable concentrations of pyrazines and furfural typically fall within 100 ng/L (i.e., the nanomolar level), a range that cannot be effectively detected by in vitro cellular assay systems^38,39. Specifically, interaction signals between odor molecules and olfactory receptors (Fig. 3, Fig. 4), as well as synergistic effects among flavor compounds, are difficult to stably capture and quantify at physiologically relevant low concentrations, and a concentration amplification strategy has become a general technical approach for investigating the molecular mechanisms underlying flavor compound interactions in the current research stage^12,24. It should be emphasized that the high concentrations in this study were only used to amplify the intrinsic molecular interaction mechanisms of odorants rather than induce artificial, non-physiological effects. A notable limitation of this study is its failure to cover the full concentration gradient of the target compounds, particularly the low concentrations present in real red wine matrices, which restricts the comprehensiveness of the characterized molecular interaction profiles. In future research, we will establish a high-sensitivity detection platform and develop an artificial intelligence-enabled odor interaction model to investigate the interactions of flavor compounds based on practically relevant concentrations, aiming to further elucidate the olfactory action mechanisms of flavor substances in wine^11,40. Notably, while our receptor-level findings advance the understanding of peripheral olfactory synergy mechanisms, extrapolation to holistic human sensory experiences requires utmost caution. Critical factors including perceived aroma intensity, refined aroma quality, food matrix context, and sensory acceptability are all pivotal for food science applications and remain unaddressed in the present study⁴¹.

The study of aroma compound interactions based on olfactory receptors is effective for explaining results from human sensory evaluations. However, while cellular assays measuring receptor responses via second messenger cAMP levels provide functional insights, the underlying molecular mechanisms require further elucidation, and the limitations of extrapolating cellular receptor responses to human sensory perception warrant in-depth exploration. This is because cellular models only replicate the single-receptor signaling cascade in isolation, failing to recapitulate the intricate spatial and temporal integration of olfactory signals in the native olfactory epithelium and the subsequent neural processing in the olfactory bulb and higher brain regions that underpin human olfactory perception. Techniques such as site-directed mutagenesis of olfactory receptor genes and cryo-EM microscopy can directly identify binding sites of different aroma compounds and their effects on the receptor’s active pocket⁴². Furthermore, tissue-based and whole-organism experiments should be incorporated into research planning. With over 400 olfactory receptors identified, odor coding exhibits remarkable diversity and complexity³⁶. Consequently, ex vivo neuronal experiments and brain imaging can provide deeper insights into how odor mixtures influence perception through neural signal tuning^43,44. Moreover, heterologously expressed human olfactory receptors in cellular systems may exhibit functional differences from their native counterparts in olfactory tissues, likely due to variations in host-specific accessory proteins and signaling cascades⁴⁵, which further complicates the direct translation of in vitro cellular data to in vivo human sensory experiences and adds another layer of uncertainty to such extrapolation. Therefore, research solely relying on olfactory receptors has limitations for studying complex odor interactions, necessitating more multi-dimensional approaches.

In conclusion, synergistic effects between aroma compounds constitute a key phenomenon in natural and processed food systems. This study sought to systematically investigate the reciprocal potentiation between TMP and FA, two typical flavor compounds in red wine, via an integrated approach combining sensory analysis, cellular functional assays, and computational simulations. Sensory assessments utilizing the 3-AFC protocol initially identified evidence of perceptual synergy between these two red wine flavor compounds in human olfactory perception. Cellular functional assays further revealed a mutual potentiation effect in their activation responses, and these cellular-level observations were consistent with the sensory results obtained from sensory evaluation. Molecular docking-based structural analyses suggested non-competitive binding modes of the two red wine flavor compounds at their respective binding sites, which may provide a potential mechanistic rationale for the observed synergistic effects. This integrated investigative approach offers new insights into the characterization of flavor interactions in red wine, and the multi-layered elucidation of the underlying flavor potentiation mechanisms provides actionable insights for the targeted design and optimization of flavor profiles in wine production practices.

Methods

Chemicals

Chemicals required for the cellular experiments: DMEM medium, fetal bovine serum (FBS) (Premium Plus), 1 × PBS buffer, 10,000 U/mL penicillin/streptomycin, trypsin/EDTA solution, 1 M HEPES, 1 × HBSS, and Opti-MEM medium were purchased from Invitrogen (California, America). Lipofectamine™ 3000 purchased from Thermo Fisher Scientific (Shanghai, China) was used for cell transfection. The reagents used for odorant stimulation include 2,3,5-trimethylpyrazine and furfuryl alcohol sourced from Aladdin (Shanghai, China). The odorant solvent is dimethyl sulfoxide (Beyotime, Shanghai, China).

Sensory conditions and sensory panel

All sensory evaluation procedures were conducted in strict compliance with relevant regulatory guidelines (ISO 8586:2012; ISO 8589:2007) and institutional ethical standards. The sensory team consisted of 20 trained members, including 10 males and 10 females, with an age range of 20–25 years old. Participants must have been in good health and free from any conditions that might have impaired olfactory function. During the training process, panelists were required to undergo olfactory threshold and identification tests to ensure olfactory sensitivity and exclude individuals with olfactory abnormalities. Upon completing a structured 60-hour pre-experimental training program (2 h daily) focused on odor recognition, descriptive terminology, intensity scaling, and difference testing, only those panelists who met stringent criteria were selected for formal trials (see the elaboration by Martin et al. for details)^46,47. This regimen was designed to solidify their perceptual acuity and long-term memory of each reference odorant. Additionally, all samples were evaluated at a controlled room temperature of 25 °C. Black 10 mL ISO glass bottles with lids were used, containing about 2 mL of test sample or blank and coded with two-digit random numbers. All experimental samples were prepared in full compliance with national food safety regulations, with tested concentrations maintained within safety thresholds to ensure human health. Research involving human research participants, material, or data have been performed in accordance with the Declaration of Helsinki. The sensory evaluation protocol has been approved by the Ethics Committee of Shanghai Jiao Tong University (Approval Number: B20250330I), and all participants have been informed of the experimental purpose, protocol, and ethical review outcome.

Sensory experiments on odor interaction

Odor interactions among target aromatic compounds were evaluated using a systematic concentration-gradient framework, in accordance with our previously established procedures for binary mixture analysis^35,48. Detection thresholds of individual aroma compounds were determined via the three-alternative forced-choice (3-AFC) method, consistent with the ISO 13301:2018 standard. Each target test group comprised ten concentration levels, all of which were assessed using the previously referenced methodology⁴⁹.

Data analysis was conducted based on the S-curve method to derive olfactory thresholds and characterize odorant interactions. The detection probability was modeled using the following equation⁵⁰:

$$P=\frac{1}{1+e^{-\frac{x-c}{D}}}$$ 1

where P denotes the chance-corrected detection probability, x represents odorant concentration, c indicates the olfactory threshold, and D describes the function slope⁵⁰. The experimentally observed detection probability P(A) was compared against the theoretical detection probability p(A) for odorant A presented alone. A dose–response curve was constructed by maintaining the original A–B ratio while varying total concentration. The interaction threshold was defined as the concentration at which P = 0.5 on the fitted curve. A value of P(A) lower than p(A) at this point indicated that B masked A; conversely, P(A)>p(A) suggested a synergistic enhancement⁵⁰.

Olfactory receptor expression

The encoding sequences of ORs were amplified from human genomic DNA via polymerase chain reaction (PCR) using specific primers and subsequently ligated into the pCI mammalian expression vector using T4-DNA ligase (Figure S1; Table S1). Hana-3A cells, an engineered derivative of the human embryonic kidney cell line (HEK293), were used for olfactory receptor expression¹⁷. Hana-3A cells were cultured in 10 cm dishes, with DMEM (with 4.5 g/L D-glucose) containing 10% FBS and 1% penicillin/streptomycin at 37 °C, 5% CO₂, and 100% humidity. Cells were plated in 35 mm dishes and transfected with Lipofectamine™ 3000 when reached 60–70% confluence. 800 ng of the corresponding ORs and pGloSensor-22F (Promega, Shanghai) plasmid, and 400 ng of mRTP1s and Gα_olf plasmid were co-transfected⁵¹.

Odorants preparation

All odorants were prepared by diluting pure reagents with dimethyl sulfoxide (DMSO). Primary stock solutions (2 M) of TMP and FA were initially prepared and subsequently serially diluted to achieve working concentrations. For odor interaction experiments, the required odorant mixtures were pre-configured by combining stock solutions in predetermined ratios prior to dilution. Compositional analysis of the pre-mixed binary odorants systems was performed using gas chromatography-quadrupole/orbitrap mass spectrometry (GC-Q/Orbitrap MS; Thermo Fisher Scientific, USA), which confirmed the absence of chemical reactions or formation of new substances (Figure S2).

Cell viability assay

The cytotoxicity of the administered odorants was evaluated using the methylthiazolyldiphenyl-tetrazolium bromide (MTT) assay⁵². Cell viability was measured post-transfection and odorant exposure, with results from the odorant-treated group compared to both an untreated blank control and a DMSO solvent control.

Luminescence assay

Transfected cells were harvested after 24 hours and resuspended in a solution containing 87% HBSS, 10% FBS, and 3% GloSensor cAMP Reagent (Promega, Wisconsin, America), followed by plating in white 96-well plates (Thermo Scientific™ Nunc™ F96 MicroWell™)⁵³. Luminescence measurements were performed using a Varioskan LUX microplate reader (Thermo Fisher) controlled by Skanlt Software 6.1 RE (version 6.1.0.51). Following a 2 h incubation at room temperature in darkness, baseline luminescence was recorded for all wells prior to the addition of odorants solutions. Real-time luminescence signals were subsequently monitored at 30-second intervals using kinetic cycle detection to capture dynamic response profiles¹⁹.

Analysis of cAMP luminescence signals

Baseline and post-stimulation data points were averaged separately for each measurement. Luminescence intensity was calculated as a dimensionless multiplicity constant using the formula 2.

$$L\mathit{um}=\frac{x-x_0}{\mathit{Baseline}}$$ 2

Where x represents the raw luminescence value corrected by subtraction of the mock control (odorant solvent), and x₀ denotes the basal level.

The resulting dataset was normalized to the maximum amplitude of the reference odorant-receptor pair²³, and concentration-response curves were fitted using the Logistic function in Origin 2019 software. All data are presented as mean ± SD, with error bars displayed on all figure data points.

Molecular modeling

The interactions between aroma compounds and olfactory receptors were predicted using AlphaFold3 1.0. AlphaFold3 enables de novo prediction of complex structures from sequence without requiring a pre-defined receptor structure or binding pocket. In contrast to conventional docking approaches, it co-models the full 3D conformations of both the protein and ligand simultaneously, capturing their mutual structural adaptation. The systems for simulation were built via CHARMM-GUI website, where the bilayer was composed of ~169 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC)⁵⁴. The bilayer and protein complex were solvated into periodic hexagonal type TIP3P water box containing 0.15 M KCl. Specifically, the binding patterns of two aroma compounds (TMP and FA) with OR5K1 (UniProt ID: Q8NHB7) and OR2W1 (UniProt ID: Q9Y3N9) were investigated. Protein sequences and compound simplified molecular input line entry system (SMILES) notations were input into AlphaFold3 for individual ligand-receptor pair predictions, yielding four structural complexes. Intermolecular interactions within each complex were analyzed using Maestro Viewer 14.3 to generate 2D interaction diagrams, while 3D visualizations were created with PyMOL 2.3.0.

Molecular dynamics simulations were performed using GROMACS 2020, with the CHARMM36m force field for proteins and the CHARMM General Force Field (CGenFF) for ligands⁵⁵. Energy minimization and equilibration utilized the default parameters generated by the CHARMMGUI webserver. In the production, the systems were simulated for 200 ns with a 2-fs time step. During the simulation, the v‑rescale thermostat was used to keep the temperature at 303.15 K and the Parrinello‑Rahman semi-isotropic barostat controlled pressure at 1 atm (NPT ensemble)⁵⁶. Van der Waals interactions were force-switched off from 0.8 to 1.2 nm. The Particle Mesh Ewald method was used with the 1.2 nm cutoff⁵⁷. The LINCS algorithm was applied to constrain all bonds associated with hydrogen atoms⁵⁸. The root mean square deviation (RMSD) of the ligand heavy atoms was calculated using the GROMACS rms tool, with the predicted initial conformation as the reference. The binding free energy was calculated using the molecular mechanics/Poisson Boltzmann surface area (MM/PBSA) method⁵⁹. In this work, the change in conformational entropy is omitted owing to its high computational cost and limited predictive accuracy. Binding free energies for all protein–ligand complexes were calculated using gmx_MMPBSA 1.6.2⁶⁰, based on 200 ns (2000‑frame) molecular dynamics trajectories.

Data analysis

All of the experiments were performed in appropriate biological replicates and the results of statistical analysis were presented as mean ± SEM (standard error of mean). Microsoft Office Excel was used for data processing, and GraphPad Prism 10 and Origin 2019 software were used to graph the data. For comparisons between two independent groups, a two-tailed unpaired Student’s t-test was employed. For comparisons involving three or more groups, one-way analysis of variance (ANOVA) was conducted, followed by Dunnett’s post-hoc test to determine specific group differences. Statistical significance was denoted as follows: *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001. Results that did not achieve statistical significance were labeled as “ns” (not significant) or remained unmarked. All units appearing in the experimental results have been declared (Table S2).

Author contributions

Boyong Hu: Methodology, Data curation, Visualization, Software, Writing - original draft. Haochen Zheng: Data curation and Visualization. Yanfei Shen: Formal analysis, Validation. Hao Wang: Visualization, Software. Zhihua Liu: Visualization, Software. Haoming Lv: Data curation. Lijun Han: Formal analysis. Yinuo Ma: Formal analysis. Heng Wang: Conceptualization, Supervision, Writing - review & editing. Zuobing Xiao: Funding acquisition, Resources, Project administration.

Data Availability

All relevant data supporting this study are included within the manuscript and supplementary data. We will provide any necessary supplements if required.

Competing interests

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

Supplementary information

The online version contains supplementary material available at 10.1038/s41538-026-00793-9.