Effects of Different Probiotics on the Volatile Components of Fermented Coffee Were Analyzed Based on Headspace-Gas Chromatography-Ion Mobility Spectrometry

Headspace-gas chromatography-ion mobility spectrometry (HS-GC-IMS) was used to study the effects of four kinds of probiotics on the volatile components of fermented coffee. The fingerprints showed that 51 compounds were confirmed and quantified, including 13 esters, 11 aldehydes, 9 alcohols, 6 ketones, 3 furans, 5 terpenes (hydrocarbons), 2 organic acids, 1 pyrazine, and 1 sulfur-containing compound. After fermenting, the aroma of the green beans increases while that of the roasted beans decreases. After roasting, the total amount of aroma components in coffee beans increased by 4.48–5.49 times. The aroma differences between fermented and untreated roasted beans were more significant than those between fermented and untreated green beans. HS-GC-IMS can distinguish the difference in coffee aroma, and each probiotic has a unique influence on the coffee aroma. Using probiotics to ferment coffee can significantly improve the aroma of coffee and provide certain application prospects for improving the quality of commercial coffee beans.


Introduction
Volatile flavor components in coffee are one of the key factors affecting beverage quality and a decisive consumer parameter. The quality of coffee can be affected by many factors, including growth conditions [1] and post-harvest treatment [2]. After harvesting, during this on-farm processing, the coffee cherries are subjected to a natural fermentation in which the coffee pulp is hydrolyzed by microbial growth [3].
Fermentation triggers an array of chemical changes within the beans that are precursors to volatile compounds formed during roasting [4]. Microorganisms play a key role in the formation of coffee flavor. Spontaneous fermentations have numerous disadvantages compared to fermentations with starter cultures. The structural and sensory characteristics of the product are improved using starter cultures, and the growing risk of harmful organisms can be prevented [5]. Lactobacillus rhamnosus has the potential to alter various flavor-related constituents in green coffee beans, which may lead to the modification of the final coffee flavor after roasting [6]. The inoculation of the Lactobacillus plantarum LPBR01 strain also significantly increased the formation of volatile aroma compounds during the fermentation process [7]. In addition, it has been found to have positive impacts on the coffee aroma when it optimizes the fermentation conditions and parameters [8]. For example, inoculates pectinolytic yeasts Pichia guilliermondi and Candida parapsilosis onto coffee cherries to facilitate mucilage removal [9].
In the present study, although several novel methods of coffee aroma modulation involving microbial fermentation have been reported, previous studies mainly attempted to (20 mL) and kept for 20 min (40 • C) with two replicates for HS-GC-IMS analysis. The top gas (200 µL) was analyzed by GC. GC was performed with a 15 m gas chromatographic column (FS-SE-54-CB-1, ID: 0.53 mm) to separate volatile components and couple them to IMS. A nitrogen of 99.999% purity was used as the carrier gas at programmed flow rates as follows: 2 mL·min −1 for 2 min, 2-15 mL·min −1 for 8 min, 20-80 mL·min −1 for 10 min, and 100-130 mL·min −1 for 20 min. The analytes were eluted and separated at 45 • C and then ionized in the IMS ionization chamber containing a tritium ionization source, which uses hydrated protons generated by the tritium source as its calibrant/reactant ion. An IMS electric field strength of 500 V/cm and a 150 mL·min −1 drift gas flow were applied in a drift tube of 98 mm length and operated at 45 • C. IMS was performed at ambient pressure.

Coffee Sample Preparation
Five kilograms of coffee cherries were washed three times in sterile, ultra-pure water, then put into a high-temperature cooking bag with a vacuum seal and sterilized under the conditions of 121 • C, 0.12 MPa, and 15 min. After sterilization, the samples were divided into 5 parts (A-E) and put into a sterile operation table, including non-treated coffee beans (A), coffee beans inoculated with Lactobacillus acidophilus (B), coffee beans inoculated with Bifidobacterium longum subsp. Infantis (C), coffee beans inoculated with Streptococcus thermophilus (D), and coffee beans inoculated with Lactobacillus paracasei (E). Each kilogram sample was subjected to inoculation with 5 mL of probiotic solution (≥1 × 10 6 CFU/mL). The inoculated coffee beans were placed in a sterile fermentation bag with a one-way ventilation valve, and anaerobic fermentation was conducted at 18 • C for 48 h. After 24 h of fermentation, coffee cherries were rubbed to remove the skin in a fermenting bag, and the gas was expelled every three hours. Following fermentation, the samples were treated at 100 • C for 10 min, and screened out, and oven-dried in a thermostatic air-drying box at 35 • C until the moisture content was 12%. They were stored at room temperature in a one-way breathable bag before use. Non-treated (A) and treated green beans (B-E) were roasted with a SANTOKER R500E coffee roaster (Beijing Sandouke Technology Co., LTD., Beijing, China) under a constant air temperature of 185 ± 5 • C for 18 min to achieve a dark roast level. The agtron value of coffee beans is set at 35-40. Roasted samples were ground with an electronic coffee grinder (EK-43, MAHLKÖNIG, Hamburg, Germany) and then collected in a gas blocking packaging and vacuum.
The first letter of the sample name indicates the fermentation method, and the second letter indicates whether it was roasted or not (G for green beans and R for roasted coffee beans).

Statistical Analysis
All data were acquired and processed using the Laboratory Analytical Viewer (LAV) software (version 2.0.0, G.A.S., Beijing, China). GC-IMS Library Search software supplied by G.A.S. (Gesellschaft für analytische Sensorsysteme mbH, Beijing, China) was employed to identify unknown compounds. The software used the NIST and IMS databases to qualitatively analyze the components. IMS was an analytical technique for characterizing ionic chemical substances based on the difference in migration velocities of different gaseous ions in an electric field. According to the retention index (RI) and migration time (Dt), two-dimensional qualitative analysis is an excellent proposal for compound identification. The substances in the library can be qualitatively matched easily, which is more accurate than the degree of matching obtained by GC-MS.
The matching matrix was processed using Origin software version 9.1 (OriginLab, Northampton, MA, USA), and the data were presented as mean values ± standard deviation. Statistical analysis was conducted by SPSS software version 24.0 (SPSS Inc., Chicago, IL, USA). All samples were measured in duplicate.
As shown in Table 1, the contents of the top five compounds in green coffee beans that have been treated differently are different. The aroma components ranking in the top five in AG samples were acetone, dimethyl disulfide, hexanal, ethanol, and 5-methylfurfural. In BG samples, there were dimethyl disulfide, acetone, hexanal ethanol, and 3-methylbutanal. In CG samples, there were acetone, dimethyl disulfide, hexanal, 2-methylbutanal, and 3-methylbutanal. In DG samples, there were acetone, dimethyl disulfide, 2-methylbutanal, hexanal, and 3-methylbutanal. In EG samples, there were acetone, dimethyl disulfide, hexanal, ethanol, and 2-methylbutyric acid. The types and content of aromatic compounds in roasted coffee beans treated in different ways are essentially the same.
The aroma components ranked in the top five in the AR sample were furfural, 5-methylfurfural, acetone, furfuryl alcohol, and 2-butanone. In the BR sample were 5-methylfurfural, furfural, acetone, furfuryl alcohol, and 2-butanone. In the CR sample were 5-methylfurfural, furfural, acetone, 2-butanone, and furfuryl alcohol. In the DR sample were 5-methylfurfural, furfural, acetone, furfuryl alcohol, and 2-butanone. In the ER sample were 5-methylfurfural, furfural, acetone, furfuryl alcohol, and 2-butanone. In conclusion, from the total aroma components, the aroma content in green beans treated with probiotics (BG, CG, DG, EG) increased compared with the untreated one (AG), while the aroma content in roasted beans (BR, CR, DR, ER) decreased compared with the untreated one (AR); roasted coffee (R) has 4.48 to 5.49 times more aroma than green coffee (G).

Effects of Fermented Coffee with Single Probiotics on the Characteristic Components of Coffee Flavor
As shown in Figure 1, combined with Table 2, we observed the effects of the characteristic components of coffee flavor that were fermented with single probiotics. Firstly, comparing sample AG and sample BG, it was found that green beans fermented with Lactobacillus acidophilus, whose 15 signal peaks increased and 12 additional signal peaks decreased, which also significantly increased the content of ethyl acrylate (p < 0.05) while the content of 3-pentanone decreased, but the other strains did not have this ability. Secondly, focusing on samples AG and CG, we can draw the conclusion that green beans fermented with Bifidobacterium longum subsp. Infantis, whose 19 signal peaks increased significantly and 9 additional signal peaks reduced, also significantly increased the content of benzaldehyde, cyclohexanone, and propyl acetate-D, while the content of ethyl acrylate (as opposed to Lactobacillus acidophilus) and (E)-3-hexen-1-ol reduced. In addition, comparing samples AG and DG, it could be seen that green beans fermented with Streptococcus thermophilus, whose 18 signal peaks increased and 12 signal peaks decreased, which also significantly decreased the content of 2-methylpyrazine. Finally, comparing sample AG and sample EG, it was found that green beans fermented with Lactobacillus paracasei, whose 15 signal peaks increased and an additional 11 signal peaks decreased, which also significantly increased 1-heptanal, 4-methyl-2-pentanone-D, and 3-methyl-3-buten-1-ol while the content of cyclohexanone, hexanal-D, pentanal-M, and pentanal-D reduced.
"*" represents that the probiotics can single-mindedly and significantly (p < 0.05) increase or decrease the content of the compound when green or roasted beans fermented with different probiotics were compared between these samples; " # " represents that the probiotics can single-mindedly and significantly (p < 0.05) increase or decrease the content of the compound when green or roasted beans fermented with the same probiotics were compared between these samples; "-" nothing; "↑" indicates the number of compounds with increased content. "↓" indicates the number of compounds with reduced content.
Focusing on samples AR and BR, it was found that coffee beans fermented with Lactobacillus acidophilus after roasting had 8 signal peaks that increased and an additional 15 signal peaks that decreased, which also significantly increased the content of 2-methylbutan-1-ol content (p < 0.05). Through comparing samples AR and CR, it was found that coffee beans fermented with Bifidobacterium longum subsp. Infantis after roasting had 13 signal peaks that significantly increased and an additional 20 signal peaks that decreased, which also significantly increased the content of benzaldehyde and reduced the content of 5-methylfurfural-M, furfuryl alcohol, methyl acetate, 2-ethylfuran-M, and other compounds. Comparing sample AR and DR, it could be seen that coffee beans fermented with Streptococcus thermophilus after roasting, whose 8 signal peaks increased and an additional 20 signal peaks decreased, which also significantly increased 3-methylbutanol and ethyl 3-methylbutanoate-M. Through comparing samples AR and ER, it was found that coffee beans fermented with Lactobacillus paracasei after roasting had 12 signal peaks that increased and 15 signal peaks that reduced, with no significant difference between compound increase and decrease.

Common Effects of Fermented Coffee with Probiotics on Characteristic Components of Coffee Flavor
According to the fingerprints of all samples generated by the gallery plot in Figure 1, there were significant differences (p < 0.05) between untreated green beans (AG) and an additional four fermented samples (BG-EG). Furthermore, the signal peak intensity was enhanced by some probiotics while the other probiotics decreased it (39 signal peaks marked with "*"). No matter what kind of probiotics were used to ferment coffee, there were 12 signal peaks that changed, among which 6 signal peaks increased significantly, including dimethyl disulfide, 2-methylbutanal, 2-methyl butyric acid, ethyl 2-methylpropanoate-M, 4-methyl-2-pentanone-M, and propyl acetate-M; moreover, 6 signal peaks decreased (p < 0.05), including 2-ethylfuran-D, acetic acid, methyl acetate, ethyl 3-methylbutanoate-M, 2-ethylfuran-M, 2-propanol, and other compounds. Among the increasing compounds, dimethyl disulfide features a strong onion odor [29]; 2-methylbutanal features a powerful and choking odor [30], with peculiar cocoa and coffee flavors after diluting, and it was reported that the compound imparts chocolate, sweet, and fruity odors to coffee. The aroma was defined as intense but not typically cheese-like [31]; 2-methylbutyric acid features spicy qualities similar to roquefort cheese, which has a pleasant fruity aroma at low concentration; the aroma profile of ethyl 2-methylpropanoate-M features fruit and cream; 4-methyl-2-pentanone has green, herbal, and fruity characteristics [32]; and the aroma profile of propyl acetate features fruit such as pear and raspberry with some pleasant and bittersweet qualities. In addition, as to the decreasing compounds, 2-ethylfuran-M has a rubber-like and smoky burnt odor [33], with a warm and sweet flavor similar to coffee after diluting; acetic acid is an important odor-active compound having strong and pungent vinegar characteristics [34]; methyl acetate has a pleasant and slightly bitter flavor with an apple aroma [35]; and ethyl 3-methylbutanoate-M has strong vinous and liquor characteristics. In short, probiotics play an important part in the formation of flavor in green beans. Meanwhile, the changes in different signal intensity result in the production of a variety of coffee flavors.
Similarly, there were significant differences (p < 0.05) between untreated roasted beans (AR) and the other four fermented samples after roasting (BR-ER). Some probiotics increase the signal intensity of some compounds while making the others decrease (45 signal peaks marked with "*"). No matter what kind of probiotics were used to ferment coffee, there were 17 signal peaks that changed, among which 4 signal peaks increased significantly, including 2-pentanone, 2-butanone, heptanal, and 1-pentanol, and 13 signal peaks decreased (p < 0.05), including gamma-butyrolactone-D, 2-acetylfuran-M, furfural, 2-methylpyrazine, ethyl acrylate, 3-methylbutanal, ethyl acetate, acetone, ethyl propanoate, ethyl butanoate, 4-methyl-2-pentanone-M, 4-methyl-2-pentanone-D, and propyl acetate-M. Among the increasing compounds, 2-pentanone has a slight fruit odor; the aroma profile of 2-butanone features fragrant milk [36]; heptanal is related to the pungent, fishy, and unpleasant nutty odor; and 1-pentanol has a sweet and pleasant wax odor whose aroma profile features burning. Additionally, for those decreasing compounds, the aroma profile of gammabutyrolactone-D featured faint, sweet, and aromatic buttery and milky [37]; 2-acetylfuran-M has the odor of sweet balsam, almond, cocoa, and caramel coffee [38]; The aroma profile of furfural characterizes almond, bitter, and toasted notes [39]; 2-methylpyrazine has nutty, cocoa, roasted, chocolate, popcorn, and coffee-like aroma [40]; ethyl acrylate emits rum smell; 3-methylbutanal features a powerful, acrid, pungent and choking odor, which was reported that the compound imparts fruity, fatty, animal and malty odors to coffee; ethyl acetate has strawberry and pineapple odor with pleasant ethereal fruity and brandy; acetone has a pleasant odor; ethyl propanoate has an odor reminiscent of rum and pineapple; ethyl butanoate has a fruity odor with pineapple and sweet taste. In a word, after fermenting with probiotics and roasting, the changes in decreasing signal peak intensities were more than those in increasing signal peak intensities. The differences in aroma between probiotic-fermented roasted beans and untreated roasted beans were more significant than those between probiotic-fermented green coffee beans and untreated green coffee beans.

Aroma Matching Matrix for All Samples
The greater the data match, the more coffee aroma will be similar to each other in the matching matrix. When the matching degree of two samples ranged between 90 and 100, the aromas in the two samples were extremely similar to each other and difficult to distinguish. As is shown in Figure 2, the aroma components of green beans fermented with Bifidobacterium longum subsp. Infantis (CG) were extremely similar to those of green beans fermented with Streptococcus thermophilus (DG), whose matching degree ranged between 85 and 90. Moreover, making a comparison between these 5 green beans, it was found that the aromas of samples BG, CG, and DG were similar to each other, whose matching degrees ranged between 90 and 95. There was a unique aroma found in sample EG whose matching degree ranged between 80 and 85 compared with other samples. In addition, there was a difference between green beans (AG) and additional fermented beans (CG-EG), and the matching degree ranged between 80 and 85.
Foods 2023, 12, x FOR PEER REVIEW 11 of 14 malty odors to coffee; ethyl acetate has strawberry and pineapple odor with pleasant ethereal fruity and brandy; acetone has a pleasant odor; ethyl propanoate has an odor reminiscent of rum and pineapple; ethyl butanoate has a fruity odor with pineapple and sweet taste. In a word, after fermenting with probiotics and roasting, the changes in decreasing signal peak intensities were more than those in increasing signal peak intensities. The differences in aroma between probiotic-fermented roasted beans and untreated roasted beans were more significant than those between probiotic-fermented green coffee beans and untreated green coffee beans.

Aroma Matching Matrix for All Samples
The greater the data match, the more coffee aroma will be similar to each other in the matching matrix. When the matching degree of two samples ranged between 90 and 100, the aromas in the two samples were extremely similar to each other and difficult to distinguish. As is shown in Figure 2, the aroma components of green beans fermented with Bifidobacterium longum subsp. Infantis (CG) were extremely similar to those of green beans fermented with Streptococcus thermophilus (DG), whose matching degree ranged between 85 and 90. Moreover, making a comparison between these 5 green beans, it was found that the aromas of samples BG, CG, and DG were similar to each other, whose matching degrees ranged between 90 and 95. There was a unique aroma found in sample EG whose matching degree ranged between 80 and 85 compared with other samples. In addition, there was a difference between green beans (AG) and additional fermented beans (CG-EG), and the matching degree ranged between 80 and 85. Among the roasted samples, making a comparison between these 5 roasted beans, the matching degree of sample AR and other fermented beans (BR-ER) was less than 80, indicating that the aroma components of roasted beans changed significantly compared with untreated ones after fermentation. The aroma components of roasted beans treated with different probiotics (BR-ER) were similar to each other but slightly different, with matching degrees ranging between 80 and 90. Among the roasted samples, making a comparison between these 5 roasted beans, the matching degree of sample AR and other fermented beans (BR-ER) was less than 80, indicating that the aroma components of roasted beans changed significantly compared with untreated ones after fermentation. The aroma components of roasted beans treated with different probiotics (BR-ER) were similar to each other but slightly different, with matching degrees ranging between 80 and 90.
Moreover, comparing green and roasted beans, it could be seen that the matching degree of all samples was less than 40%, indicating that the aroma of coffee changed greatly after roasting.

Conclusions
The present study concluded that the effects of four different probiotics on fermented arabica coffee's volatile components. The fingerprints by HS-GC-IMS showed that 61 signal peaks for 51 compounds were confirmed and quantified in all types of coffee, including 13 esters, 11 aldehydes, 9 alcohols, 6 ketones, 3 furans, 5 terpenes (hydrocarbons), 2 organic acids, 1 pyrazine, and 1 sulfur-containing compound. From the total aroma components, the aroma content in green beans treated with probiotics increased compared with untreated ones (AG), while the aroma content in roasted beans decreased compared with untreated ones (AR), which resulted in a 4.48-5.49-fold increase in aroma component contents after roasting. The differences in aroma between probiotic-fermented roasted beans and untreated roasted beans were more significant than those between probioticfermented green coffee beans and untreated green coffee beans. In conclusion, HS-GC-IMS can distinguish the difference in coffee aroma, and different probiotics have a unique influence on coffee aroma. Fermenting coffee with probiotics can significantly improve coffee aroma, which has excellent application prospects for improving the quality of commercial coffee beans.