The concentration, solubility, volatility, and chemical properties of all molecules in food matrices vary widely, resulting in no fixed method for the simultaneous analysis of all substances. Thus, solid phase microextraction (SPME), liquid–liquid extraction (LLE), stir bar sorptive extraction (SBSE), solid phase extraction (SPE), dispersive liquid–liquid microextraction (DLLME), agitator adsorption, supercritical carbon dioxide extraction, and others have been extensively used for the studies in the field of flavor analysis (Table 1). Considering the extremely low content of FFT in matrices, it is difficult to achieve efficient isolation by using conventional extraction/enrichment techniques to analyze FFT due to its instability and possible interactions between FFT and other compounds (54). Therefore, it is of great significance to develop and apply efficient enrichment methods before analysis. SPME is a common, solvent-free technology developed based on SPE and can effectively solve the problems of blocking and channeling existing in SPE and other traditional pre-treatment technologies and requires less samples to handle shorter processing time (47, 59). Generally, SPME can be categorized into headspace solid phase microextraction (HS-SPME) and direct immersion solid phase microextraction (DI-SPME), in which the former is widely used in comparison to DI-SPME (51). HS-SPME has been widely used in various matrices to enrich target odor substances with relatively good results, for instance, HS-SPME has been introduced in wine (12, 48) and coffee (60). Furthermore, it is necessary to select the appropriate extraction fiber due to the discrepancy in extraction efficiency and capacity of the various types of analyte extraction fibers (13). Moreover, it is worth noting that temperature is a non-negligible factor that cannot be ignored due to the tendency of FFT to oxidize or degrade at high temperatures (61, 62).

Moreover, LLE can extract the analyte and squash the matrix through the analyte solubility divergence between an aqueous sample and water-immiscible organic solvent; ether and dichloromethane are commonly used as extraction solvents. In addition, sensitivity is improved by blowing nitrogen and injecting the final sample for analysis during extraction (51, 63). Pre-concentration FFT of brewed coffee prior to LLE analysis was reported by Casas et.al. Notably, this operation is laborious, tedious, and time-consuming. Especially, the operator will be exposed to hazardous chemicals and, despite the advantage of the small sample size required, some compounds with similar properties may be removed at the same time (64). In addition, the addition of large amounts of organic solvents and inorganic salts to improve the recovery of target compounds may interfere with subsequent chromatographic analysis (52). Fortunately, implantation of liquid–liquid microextraction (LLME) can reduce the volume of solvent required. The choice of solvent depends on the compounds to be extracted and must be safe for the operator and the environment.

Similar to LLE, solvent-assisted flavor evaporation (SAFE) involves the use of organic matter and long extraction times. In contrast to the former, it is a gentle but comprehensive method for extracting volatiles from complex food matrices devised by Engel et al. (58). The SAFE system is a compatible combination of a distillation unit and a high vacuum pump to reduce the loss of heat-sensitive substrates from the sample and to maximize the preservation of the original flavor of the analytes and is particularly suitable for the separation and analysis of volatile molecules in complex matrices (42, 65). SAFE is a highly time-intensive and laborious practice in flavor analysis. The flavor profiles of wine were concentrated before further analysis, despite this method requires significant time and may reduce the loss of target analytes with highly volatile (i.e., FFT) characteristics from the sample (44, 66–68). In addition, unlike SPME, which is commonly used for sample preparation and combined with GC-O analysis, SAFE requires the preparation of a large number of samples (42).

Recently introduced for FFT analysis in wine (56), stir bar sorptive extraction (SBSE) is a novel method for SPME sample pre-treatment. In parallel to SPME, it has a large volume of stationary phase and extraction capacity, allowing for the simultaneous extraction and enrichment of a greater number of trace components without the need for additional stirrers. However, the selection of the type and thickness of the coating on the extraction head is equivalent to the selection of the column in chromatographic analysis (59). Furthermore, in SBSE, only two coatings of polydimethylsiloxane (PDMS) and ethylene glycol silicone (EG silicone) are commercially available (59, 69). In addition, there is a balance between coating thickness and equilibrium extraction time as the adsorption capacity of the analyte is determined. Furthermore, the detection limits are positively correlated with coating thickness and vice versa for the equilibrium extraction time (49, 59). The limited available adsorbents may become a bottleneck for the SBSE technique in FFT analysis, while the availability of new materials (e.g., carbon, metal-organic frameworks, polymers) with high concentration capacity for some specific analytes offers great opportunities for the development of efficient SBSE coatings (51, 70).

In addition to the pre-treatment methods mentioned above, other extraction methods have been used to extract FFT in foodstuffs such as solid phase extraction (SPE) (71, 72), derivatization (64, 73), and simultaneous distillation extraction (SDE) (74). Each method has advantages and disadvantages when used. Therefore, it is necessary to consider the physical–chemical properties of the target compound and the variability of the matrix in order to select the appropriate extraction method to achieve a comprehensive and accurate analysis of FFT in different matrices.

Sample preparation is followed by instrumental separation and analysis. With advances in theory and technology, there have been significant improvements in instrumentation for the detection of trace components in the field of flavor analysis. Generally, FFT in various matrices can be separated from the mixture by diverse columns in gas chromatography (GC) or liquid chromatography (LC) and further analyzed by high-sensitivity detectors such as mass spectrometry (MS) and flame ionization detector (FID).

Sulfur compounds are a kind of pivotal trace component in foodstuffs, often produced from sulfur-containing amino acids, and as such, many advanced analytical instruments are used to analyze various sulfur-containing compounds (e.g., FFT). GC-MS technology has become the most widely used technology in flavor component analysis due to its highly commercialized, powerful separation, and qualitative ability (75). Moreover, the qualification and quantification of aroma odorants in GC-MS are based on the divergence of different compounds and the concentration of molecules being related to peak intensity. GC-MS was used to identify FFT from roasted duck and it was regarded as a key aroma molecule (8). However, the detection limit for trace amounts of material is not sufficient, although it is fast, simple, and has fewer solvents. Other instruments are also suitable for the analysis of FFT in complex matrices, such as the gas chromatography pulsed flame photometric detector (GC-PFD), a detector with good selectivity and sensitivity for the analysis of sulfur-containing compounds such as FFT, at low cost. Recently, GC-PFPD was used to qualify and quantify the FFT in baijiu (10). Similar to GC-PFPD, GC-FID and gas chromatography-flame photometric detector (GC-FPD) have been also employed to analyze FFT in diverse matrices. For instance, Zhang et al. (37) applied GC-MS and GC-FPD coupled with GC-O to analyze sesame-flavored volatile compounds in baijiu and detected seven sulfur-containing compounds, including FFT and six other sulfur-containing chemicals (37). In a recent study, volatile thiols, FFT, benzenemethanethiol, and three compounds of ethyl 2-mercaptopropionate were quantified by GC-MS and GC-FPD in four types of heat-treated soy sauce and raw soy sauce (76). Gas chromatography-sulfur chemiluminescence detector (GC-SCD) is the most efficient detector for sulfur-containing compounds with features of high sensitivity, selectivity, and equimolar concentration compared to other detectors such as FPD and PFPD. This stems from the chemiluminescent detection mechanism that sulfur-containing compounds are burned into SO at high temperature and then react with O₃ to generate an excited state SO2*, and when it degrades to the ground state, the excited state SO2^* emits a characteristic spectrum (77–79). Due to the complexity of the actual sample composition, samples need to be pretreated for separation before SCD testing. In most applications, researchers combine SCD with GC to improve the purity and sensitivity of the analytes (80). Additionally, gas chromatography-time-of-flight mass spectrometry (GC-TOF/MS) is a versatile detector that is of increasing interest to analysts and industry investors for its fast response, high sensitivity, high accuracy, and high upper limits of mass for the determination of macromolecules (81).

Recently, GC-SCD and GC-TOF/MS were utilized to analyze FFT in wine (67, 74). To overcome the complexity of matrices, and improve peak capacity and separation of aroma compounds, GC × GC systems offer the advantages of increased peak capacity and high resolving power, which are essential for analyzing complex samples (82). For instance, comprehensive two-dimensional gas chromatography/time-of-flight mass spectrometry (GC × GC-TOFMS) and comprehensive two-dimensional gas chromatography-sulfur chemiluminescence detector (GC × GC-SCD) were utilized to characterize the FFT of baijiu and wine (11, 67, 83), respectively. Despite the extremely high sensitivity and response values of SCD detectors for sulfur-containing compounds, their maintenance costs need to be reduced and detection stability needs to be further improved in future studies. Furthermore, their application in practical research less popular since these apparatus (for example, SCD detector) are quite expensive and sensitive to external environmental changes, as well as the high maintenance costs. Additionally, the content of FFT may also experience loss caused by high-temperature GC inlet (generally 250°C) due to the instability when exposed to high temperatures.

The development of analytical approaches and the adaptation of derivatization to achieve more selective, efficient, complex and simplified thiol separation procedures, and/or thiol stabilization have been major developments in thiol separation and resolution. On the one hand, derivatization tempts to hide the sulfhydryl group (or shield a carbonyl group, take 4-mercapto-4-methylpentan-2-one as an example), and the thiol derivatives formed are chemically settled for isolation, as well as thermally stable for GC analysis (84–86). On the other hand, the introduction of substitutions implies that thiol derivatives indicate greater hydrophobicity, lower polarity, or higher proton affinity and may lead to better liquid chromatography (LC) separations and signal enhancement for mass spectrometry (MS)-based detection (54, 85).

In order to retard thiol degradation during analytical procedures, selective derivatization stabilizes the free thiol group. Similarly, for a more selective, efficient, and simplified study of FFTs in flavor analysis, derivatization and/or selective extraction in combination with chromatographic analysis is widely used, as derivatization is more easily extracted, chromatographed, and detected. For this reason, the analysis of FFT frequently implies the use of thiol-specific derivatization agents, such as p-hydroxymercury benzoate (p-HMB) (18) and 4,4′-dithiodipyridine (DTDP) (71).

2,3,4,5,6-pentafluorobenzyl bromide (PFBBr) is a derivative reagent for thiol compounds and is commonly used in the study of thiol compounds. This can be attributed to that the bromine atom is particularly susceptible to nucleophilic substitution by thiols in the presence of a base, thus obtaining PFBBr thiol derivatives desirable for not only stabilizing thiols but also providing electron capture capability and MS detection (87). The derivatization reaction between volatile thiols, such as FFT in wine, and PFBBr has been evaluated in several formats: automated headspace on-fiber derivatization (87), derivatization in organic reagent system (88), and derivatization in SPE column (89). SPME for fiber derivatization is fast, automated, and solvent-free. A polydimethylsiloxane/divinylbenzene (PDMS/DVB) SPME fiber is exposed successively to the vapors of tributylamine (5 min), PFBBr solution (5 min), subsequent, and pre-incubated wine sample (containing ethylenediaminetetraacetic acid, salt, and internal standard) to extraction for 10 min at 55°C (87). This approach provides convenience and less potential interference by using an autosampler and HS-SPME fiber, whereas the linear range of the studied thiols is not very wide and only two thiols (FFT and 3-mercaptohexyl acetate) can be analyzed by this method (87).

Furthermore, several sample preparations may require pH adjustment prior to injection. The p-hydroxymercuribenzoate (p-HMB) selective extraction process is tedious, laborious intensive, and hazardous as it is quite time-consuming due to the need to adjust pH and the use of organic reagents. More importantly, FFT is prone to oxidation during the laborious phase of sample preparation, which may further affect its quantification (23, 90). The FFT of misos were extracted by p-HMB and percolated on a basic anion-exchange column, then washed with high-purity water, released from the FFT-p-HMB complex in the column by using a cysteamine solution (500 mg/50 mL, pH 7), and concentrated and dried in a final step (18). Although this operation can be completed by GC-MS, there is instead room for future modifications of the method due to a large amount of solvent used, time-consuming, and labor-intensive.

Fortuitously, there is a derivatization reagent that can react with FFT in a wide pH range (55, 91). Thiol derivatization with DTDP has also been devised for wine analysis owing to the excellent derivatization ability of sulfhydryl groups at acidic pH, whereby DTDP selectively and rapidly directly reacts with thiols in the natural wine pH range (71). It has been reported that DTDP was used to react with FFT in baijiu (20) and wine (71), followed by SPE coupled with high-performance liquid chromatography-mass spectrometry (HPLC-MS/MS), resulting in excellent result with a limit of detection of 0.7~1.5 ng/L. In coffee, the derivatization of FFT with ebselen has been reported (41). After extraction and derivatization, the concentration of FFT in coffee powder was quantified by applying high-performance liquid chromatography-high resolution mass spectrometry (HPLC-HRMS). Notably, the presence of some important sulfur-containing molecules in coffee powder, such as 4-mercapto-1-butanol and 4-methoxy-2-methyl-2-buthiol, was also determined by the above method (41). Moreover, HRMS provides information on the molecular structure and composition of the molecule, which is crucial in food analysis and contributes to accuracy and reliability (75). The procedure is based on a selective and efficient reaction between thiols and ebselen (selenium-containing reagents), which enables the derivatization and separation of thiols in a time-saving and reduced sample manipulation manner. In summary, a number of key factors need to be considered when selecting the available derivatization reagents, namely reaction specificity and efficiency, matrix complexity and compatibility, sample manipulation required, introduction of interferences, and whether the interferences occur before or after the extraction of the analyte.

Moreover, to mitigate the loss of FFT caused by isolation procedures and decrease quantification errors, solvent extraction and concentration combined with stable isotope dilution assay (SIDA) was utilized (14, 24, 76), another approach is SPME (92, 93). This technique allows for the precise quantification of FFT in samples by using internal standards labeled with stable isotopes, which are often synthesized or purchased using expensive reagents. Recommendations are provided for analytical chemists interested in developing better methods for quantifying FFT in various matrices such as wine, baijiu, and other related matrices. Table 2 provides an overview of general published strategies with sample preparation, types of analysis, and major advantages and drawbacks.

The formation mechanism of FFT is not fully understood. It is generally accepted that sulfur-containing amino acids are key precursors and sources of sulfur for reactions with sugars and other minor compounds during manufacture (101–103). As early as 20 years ago, a theory was put forward that Maillard-type reactions play an indispensable role in FFT formation (104, 105). In accordance with previous literature, pentoses or hexoses and L-cysteine could produce FFT by producing furfural and H₂S in a model system (105–107), as demonstrated experimentally by bionic beans (102). In addition, the disturbance factors associated with their production, i.e., the yield of FFT, were also investigated in relation to the factors mentioned above. Temperature is one of the most significant factors in thermal treatment and/or fermentation and dominates manufacturing (108, 109). By adding L-cysteine and ribose to yeast extracts and heating them at high temperatures (100°C and 160°C, respectively), a model of the Maillard reaction was developed, showing that the yield of FFT increased with increasing temperature (27). Model experiments performed by L-cysteine and various carbohydrates at 145°C and 180°C, respectively, confirmed similar results (110). Moreover, pH is an essential factor in deciding the production of FFT. Generally, the amount of FFT is negatively correlated with the change in pH and its production generally increases with decreasing pH. This could also explain, to some extent, the fact that most FFT-rich beverages are acidic matrices (110, 111). Subsequently, the Maillard reaction between L-cysteine and xylose at different pH values from 4.0 to 7.0 confirmed this conclusion and clarified that the yield of FFT increased significantly when the pH was reduced from 7 to 4 (112). Similarly, the yield of FFT, the product of the Maillard reaction, varied depending on the type of carbohydrate. The effect of sugar type on FFT yields has been previously investigated and, in general, the Maillard reaction occurs following high-temperature treatment of L-cysteine and sugar, and the order of concentration of FFT obtained is ribose > xylose > fructose > glucose > rhamnose (107, 110). This phenomenon may be explained by the fact that L-cysteine mainly provides a sulfur source, carbohydrates produce furfural by dehydration, and the structure of pentose seems to have a higher yield/yield of furfural through reaction compared with hexose in terms of structure and composition (113–116). However, it is regrettable that these studies failed to fully investigate the pathways/reactions of furfural and L-cysteine for FFT production at the molecular level. The carbon module labeling (CAMOLA) technical approach is a powerful method to enucleate the formation mechanism of flavor molecules by labeling experiments and isotopomeric quantitation (117).

Based on the aforementioned studies, a comprehensive, appropriate, and in-depth study was performed on the model reaction system and real sample system (99). The main pathway for FFT formation involves the Maillard reaction, via dehydration of sugars, as presented in Figure 3 (113, 114). During high-temperature heating, glucose and L-cysteine undergo Amadori rearrangement and 1,2-enolisation to form 3-deoxyglucosone. Further hydroxyl-aldol condensation reactions are carried out, removing one molecule of alcohol and two molecules of water in turn, gradually producing 3-deoxypentosone and the final product furfural. The furfural undergoes a reduction reaction to generate 2-furfuryl alcohol. Finally, H₂S from the pyrolysis of L-cysteine reacts with 2-furfuryl alcohol to remove one molecule of water, which in turn produces FFT. In addition, previous investigations have suggested that furfural can form FFT by reacting with H₂S in significantly higher yields than the reaction between L-cysteine and furfural (110). This may be related to the fact that H₂S is a hydrolysis product of L-cysteine, omitting the hydrolysis reaction step. In addition to thermochemical reactions under thermal treatment conditions, microbial metabolism or enzyme catalysis is essential in the production of FFT. Therefore, the contribution of related microorganisms or enzymes during fermentation will be discussed in the following section.

Microorganisms constitute the complicated micro-ecological environment of the fermentation system, and flavor metabolites generated in the presence of these microflora are closely related to the quality and flavor of fermented beverages (21, 119–121). Yeast, molds, and bacteria are the main microorganisms in the fermentation process. Generally, these members can offer the impetus for the development of fermentation and flavor components. For instance, Pichia, Saccharomyces, Aspergillus, Rhizomucor, and Rhizopus in jiuqu (a sort of starter in Chinese baijiu), molds in soy sauce, and Bacillus subtilis in natto would secrete enzymes to hydrolyze starches and proteins of raw materials (38, 122, 123), respectively, and leads to hydrolysates undergoing Maillard reactions under appropriate conditions. Additionally, microorganisms also consume carbohydrates to produce corresponding metabolites, such as yeast and lactic acid bacteria, which could convert glucose into alcohol and lactic acid (124–126). Consequently, the formation of FFT during fermentation can be a dynamic transversion motivated by the action of microorganisms and enzymes. Conversely, there is still a lack of systematic and comprehensive knowledge regarding the exact role of microbes in the formation of FFT although several studies have been conducted on the formation mechanism of FFT, and some achievements have been obtained.

Based on the preliminary literature research, the biotransformation mechanism of 2-furfurylthiol was summarized and a metabolic production pathway was mapped as follows (Figure 4). In previous reports, furfural and L-cysteine could be used as precursors and produce FFT through yeast metabolism (127). Pichia, Schizosaccharomyces, Saccharomyces, and Zygosaccharomyces are dominant yeast species in baijiu fermentation, which are essential for the formation of flavor compounds (128). An isolated S. cerevisiae strain G20 showed the highest production capacity of FFT with a yield of 3.03 mg/L, and the genes STR3 and CYS3 (encoded cystathionine β-lyase, and cystathionine γ-lyase, respectively) were found to be closely related to FFT synthesis by gene knockout and overexpression verification (94, 129). In addition to flavor compound-producing strain, microbial interactions were found to be an alternative strategy to regulate the formation of flavor compounds during fermentation. Bacillus subtilis is an indispensable strain in the fermentation process of baijiu. Amino acids are mainly converted from proteins via the hydrolysis of protease, and Bacillus subtilis was one of the major microorganisms producing protease (130). There are synergistic effects between Bacillus subtilis and other functional microorganisms. For example, in the fermentation of baijiu inoculated with Bacillus subtilis LBM 10019 and Bacillus vallismortis LBM 10020, the FFT content in baijiu increased from 1.29 μg/L to 2.44 μg/L after inoculation compared to the non-inoculated group (33). This phenomenon can be explained by the inoculation of Bacillus. Bacillus inoculation provides a constant supply of L-cysteine and more sulfur sources to the associated FFT-producing strains, thus increasing the FFT content of the final baijiu (131, 132).

In general, there are two main pathways to generate FFT. (1) In the process of alcohol fermentation, mercaptan compounds are released from their non-aromatic precursors through the function of yeast. Although the exact mechanisms by which yeast convert cysteinylated and glutathionylated precursors into the corresponding volatile thiols are still not fully understood, some consensus exists regarding these mechanisms. Previous studies have reported that the uptake of cysteine-3-mercaptohexan-1-ol (3MH) adduct precursors is structurally similar to L-cysteine and is induced by amino acid transport proteins; therefore, it is suggested that furfuryl-cysteine adducts, precursors of FFT, are also induced by amino acid transporters (133, 134). Thus, L-cysteine reacts with furfural to produce furfural-cysteine conjugates (129), which are subsequently transported into the cell by amino acid permeases encoded by genes MUP1, Bap2, Bap3, Gnp1, and OPT1 (encoding Mup1p, branched-chain amino acid permease, amino acid transporter, glutamine permease, oligopeptide transporter, respectively) (134–138). The furfuryl-cysteine adduct is cleaved in vitro or after transport into the cell by the C-S lyase encoded by the genes CYS3 and STR3 to produce FFT, pyruvate, and ammonia (129). (2) L-cysteine is transferred into cells under the action of amino acid transporters and then produced as hydrogen sulfide under the guidance of α- and β-elimination reactions or cystathionine β-synthase and cystathionine γ-cleavage enzymes encoded by genes CYS4 and CYS3, respectively (139, 140). Meanwhile, it is worth noting that although L-cysteine is an essential sulfur source of FFT, a high concentration of L-cysteine will restrain the growth of yeast (141). In most cases, the dehydration of carbohydrates produces furfuryl alcohol correspondingly produced by reduction reactions, and the hydroxyl group is replaced by the HS⁻ ion in hydrogen sulfide, ultimately producing FFT, which is secreted extracellularly by the action of carrier proteins (113, 142, 143). It also has been reported that hydrogen sulfide can be used as a precursor to produce FFT from furfural in yeast (100, 142, 144). Many enzymes secreted by microbes in the fermentation environment can catalyze the substrate of materials to generate consequential flavor compounds during growth and metabolism. Lipase from Candida albicans served as a biocatalyst to react with S-2-furfuryl thioacetate, and obtained the corresponding product, FFT (145). This finding provides an alternative, environmentally friendly approach to producing natural thiol species and offers theoretical support and technical guidance for generating characteristic flavors in fermented beverages. Future research can focus on the metabolic mechanisms and driving factors of microorganisms and enzymes during catalysis to better regulate the production of FFT during fermentation.

Aroma plays a crucial role in determining the quality and flavor of a product, as well as influencing individual preferences. However, various external and internal factors lead to the degradation or loss of aroma components in beverages during processing, storage, and transportation. These factors can significantly affect the homogeneity of flavor and quality of the product, ultimately affecting the sensory experience of the consumer and the economic interests of the manufacturer. Correspondingly, we propose to investigate the mechanisms and regulation of aroma deterioration associated with FFT in the following paragraphs.