The Modification of Coffee Beans Through a Combination of Microbial and Enzymatic Processes

TL;DR

This study explores how combining microbes and enzymes can improve coffee quality by modifying its chemical and sensory properties.

Contribution

The novelty lies in combining microbial fermentation and enzymatic treatment in controlled conditions to modify coffee beans.

Findings

Acid-producing bacteria increased antioxidant activity and retained bioactive compounds while reducing sucrose.

Lactiplantibacillus plantarum significantly lowered acrylamide levels in roasted coffee beans.

The combined microbial and enzymatic process altered volatile organic compound content and composition.

Abstract

Fermentation with various microorganisms modifies the quality of coffee. In animal-digested coffee, enzymatic activity also affects coffee characteristics. However, limited information is available on in vitro coffee modification employing both mechanisms simultaneously in controlled conditions. In this study, robusta green beans were modified with selected bacterial species (Bacillus subtilis, Gluconobacter sp., Lactiplantibacillus plantarum) and pepsin, which was introduced at the soaking or fermentation stage. The characteristics of green and roasted coffee were analyzed, including the amount of basic aroma precursors, antioxidant activity, acrylamide concentration and volatile organic compound (VOC) content. The number of bacterial cells increased by 1.95–2.64 logCFU/mL during the modification process; pepsin addition did not affect their growth significantly. The use of acid-producing bacteria (APB) resulted in higher consumption of sucrose but also in greater retention of bioactive compounds and higher antioxidant activity. The acrylamide content in fermented and roasted beans was below 10 μg/100 g dry matter; the lowest values were reported after modification with L. plantarum. The combination of tested bacterial and enzymatic processes changed the content and composition of VOCs. Further research should focus on sensory attributes as the result of the combined modification process.

Figures

Full text

1. Introduction

Coffee is one of the most important agricultural crops in the world. Roasted and ground beans are used to produce various drinks, highly demanded by consumers thanks to their high cultural value, bioactive properties and sensory quality. The overall coffee quality results not only from the botanical origins of the plant, cultivation conditions or roasting process but also from the post-harvest processing of coffee fruits, during which fermentation occurs. The activity of various microorganisms helps with the removal of fruit mucilage. More and more studies show that it also influences the amounts of coffee aroma precursors and bioactive compounds in the bean. Furthermore, this process can be used to modify not only fresh coffee material but also green coffee beans [1,2].

Fermentation with selected microbial inocula is considered a way to produce coffee with high quality. Apart from polysaccharide degradation, the selection of yeasts and/or bacterial strains (e.g., lactic acid bacteria (LAB) or Bacillus) aims to better the antioxidant content and modify the aroma profile of the final product [3]. Research shows that the fermentation process, specifically the high diversity of microorganisms and their metabolites, can correlate with the aroma profile of roasted beans [4,5]. Apart from changes in the sugar profile and the migration of microbial metabolites into the coffee material, additional modification influences the stability of the bioactive compounds. Fermentation might accelerate a partial degradation or transformation of those compounds; in many cases, it can increase the availability of their active forms or their metabolites [6].

Another possible way to modify coffee material could involve the use of external enzymes. The use of proteases, hydrolases or decarboxylases allows for controlled degradation of macromolecules and releases bioactive compounds. Enzymatic activity naturally accompanies microbial metabolism in the production of kopi luwak, black ivory coffee and jacu coffee—after consumption, the coffee fruit is modified by agents present in the gastrointestinal tract of animals. In the case of in vitro modifications, enzyme addition at the proper moment of the process might enhance or modify the effect of pure microbial fermentation. In the context of coffee modification, such an approach is not common [7].

The timing as well as the order of enzyme addition relative to inoculum might impact the metabolism of microorganisms. Furthermore, it might change the fermentation environment (pH, substrate availability), thus favoring different groups of microorganisms. The addition of proteolytic enzymes during fermentation might also degrade proteins, increasing the amount of peptides and amino acids in coffee material, and influence the reactions during heat treatment, such as Maillard reactions and caramelization. Since research shows a strong connection between the coffee aroma profile and content of precursor compounds in green beans, both enzymatic and microbial modifications as well as their combination might have an effect on the aroma of the final product [8].

Of the two most popular coffee varieties, Robusta beans are characterized by lower acidity and higher bitterness, thus being less popular compared to Arabica beans. In this aspect, additional processing—by means of inoculation, enzyme addition or both—seems a promising way to improve their quality, both in chemical and sensory aspects [9].

In light of all the information presented above, it is only reasonable to consider combined microbial and enzymatic modification of coffee beans as a method for quality improvement. Knowledge about a proper microbial inoculum as well as the right moment of enzyme addition might help to further optimize the modification process and allow for proper quality control and/or creation. The objective of this study was to analyze the influence of bacterial fermentation combined with enzyme addition at various modification stages on the properties of coffee material before and after roasting. Analysis of two enzymatic modification variants allowed for evaluation of the enzymes’ effect on the fermentation process and coffee composition, antioxidant potential and aroma profile.

2. Materials and Methods

All reagents were of analytical grade. Unless mentioned otherwise, reagents were obtained from SigmaAldrich (Buchs St. Gallen, Switzerland), Chempur (Piekary Ślaskie, Poland), Alchem Group (Toruń, Poland), Avantor Performance Materials Poland S.A. (Gliwice, Poland) and VWR International Sp. z o.o. (Gdańsk, Poland). Microbiological madia were purchased from BTL sp. z o.o. (Łódź, Poland) and Millipore, Merck (Darmstadt, Germany).

For electrophoresis, equipment (the Mini-PROTEAN® Tetra handcast system) and reagents were sourced from Bio-Rad Laboratories Inc. (Hercules, CA, USA). Electrophoresis was performed with the use of a 2× Laemmli buffer as an extraction agent. Stock solutions and gels were prepared according to the manufacturer’s instructions. The stacking and separation gels had 5% and 12% of acrylamide, respectively; the final gel had a thickness of 1.50 mm. As a standard, Blue Prestained Protein Standard, Broad Range 11–250 kDa (New England BioLabs Inc., Ipswich, UK), was used.

2.1. Modification Procedure

Wet-processed green beans of Coffea canephona var. robusta from Rwanda were used in this experiment. The basic modification procedure was described in our previous article [10]. Briefly, coffee beans were mixed with distilled water (ratio 75:100, w:v), soaked for 3 h and sterilized (121 °C, 20 min, 0.1 bar). Bacterial inoculum (20 mL, 0.5 on McFarland’s scale) was added to the prepared coffee sample, and fermentation was carried out (30 °C, 24 h). Further, coffee was washed and dried in an air oven (40 °C) for 12–13 h. Part of the obtained beans was roasted using a sample roaster, iKAWA Pro V2 (IKAWA Ltd., London, UK; max temperature, 213 °C; total time, 5.75 min). Both green and roasted beans were ground using a laboratory grinder (MF 10 basic, IKA, Warsaw, Poland; 3500 rpm, 2 mm sieve). Samples were refrigerated in plastic containers.

During the fermentation step, Bacillus subtilis ATCC 6633 (BS), Gluconobacter sp. KKP 3751 (Gsp) and Lactiplantibacillus plantarum ATCC 4008 (LP), sourced from the Museum of Pure Cultures (Department of Food Biotechnology and Microbiology, Warsaw University of Life Sciences WULS-SGGW), were utilized separately. Bacterial species were cultivated in liquid media (tryptic soy broth, glucose yeast extract and de Man, Rogosa and Sharpe media for BS, Gsp and LP, respectively) for 48 h prior to the experiment. Bacterial cells were separated from the medium with centrifugation (6000 rpm, 4 min) and dispersed in sterile distilled water to obtain the final inoculum (20 mL, 0.5 on McFarland’s scale) for immediate use.

In order to evaluate the impact of combined bacterial and enzymatic modification on the characteristics of green and roasted coffee beans, the following changes were applied to the basic procedure. Based on previous results [7], a pepsin solution (34.86 mg/5 mL of solution in 0.01 M HCl) was selected for this experiment. Two kinds of samples were prepared in order to check the effect of enzyme addition at different steps of the modification process: samples marked as V1 had pepsin added during the soaking of coffee beans, whereas V2 samples had pepsin added together with the bacterial inoculum.

pH values and cell counting were used to monitor this process; pH measurement and plate counting methods were utilized, respectively. Measurements were performed at the beginning and the end of the process. During plate counting, the media used for the cultivation of microorganisms prior to fermentation with the addition of agar (2%) were utilized.

For each kind of sample (V1 and V2), beans without bacterial fermentation were also prepared as controls; furthermore, a sample without enzyme or inoculum addition was prepared and marked as V0. Two separate biological repetitions of the samples were prepared.

2.2. Coffee Extraction

Methanol extracts were prepared from both green and roasted beans using a previously described procedure [10]. In short, a portion of ground coffee was mixed with 80% methanol at a 1:6 ratio, sonicated for 20 min and further extracted (24 h, 5 °C). Extracts were filtered into closed plastic containers.

Proteins from green coffee beans were extracted using a 2× Laemmli buffer. Ground green coffee beans (100 mg), 475 μL of 2× Laemmli buffer and 25 μL of β-mercaptoethanol were combined, vortexed, heated up (95 °C, 10 min) and immediately cooled down on ice. After centrifugation (5 min, 9703 rpm), extracts were transferred to new Eppendorf tubes.

Acrylamide extraction was performed in accordance with Andrzejewski et al. [11] with some modifications. Ground roasted beans (1 g) were mixed with distilled water at a 1:9 ratio; 90 μL of acrylamide-d_3_ solution in water (1 μg/mL) was added as an internal standard. Samples were shaken for 30 min and centrifuged (10 min, 5000 rpm). Water extracts were further cleaned using solid-phase extraction. The cartridge (Discovery DSC-18 SPE Tube 1 g/6 mL; Supelco, Bellefonte, PA, USA) was conditioned with 1 mL of methanol and 1 mL of water. After 1 mL of coffee extract was loaded and passed through, the column was rinsed with 0.5 mL of water. The analyte was eluted with 1 mL each of methanol and water. The collected filtrate was evaporated under nitrogen at room temperature, and the residue was dissolved in 2 mL of a methanol:water mixture (1:1).

All extracts were stored at −20 °C before further analysis. Prior to chromatographic analyses, methanolic extracts were filtered using syringe filters (PTFE, 25 mm/0.22 μm).

For the analysis of the browning index (BI), water extracts from roasted beans were prepared—1.5 g of roasted coffee was mixed with 9 mL of boiling distilled water. Extracts were filtered after 10 min, cooled down to room temperature and used immediately.

2.3. Coffee Analysis

Coffee extracts were analyzed using both spectrophotometric and chromatographic methods. Additionally, soluble proteins were separated based on molecular weight using sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE).

Regarding the first group, the Folin–Ciocalteu method [12] was utilized to evaluate the total phenolic content (TPC), and the Lowry method [13] was used to analyze the content of soluble proteins; gallic acid and bovine serum albumin were used as standards, respectively. Detailed descriptions of these methods are given in a previous article [7]. Analysis of the BI was based on the method by Chung et al. [14] and consisted of absorbance measurements (420 nm) of diluted water extracts from roasted coffee beans.

The antioxidant activity of roasted coffee was evaluated with two spectrophotometric methods. The CUPRAC method (cupric ion reducing antioxidant activity) was performed in accordance with Özyürek et al. [15]. Briefly, 3 mL of copper(II) complex solution was mixed with diluted coffee extracts (1.1 mL). After 30 min of incubation, absorbance (450 nm) was measured. The ABTS method, which utilizes the 2,2’-azinobis-(3-ethylbenzothiazoline-6-sulfonic acid) radical cation, was also performed in accordance with Re et al. [16]. In short, 4 mL of prepared radical solution and 40 μL of coffee extracts were mixed. Absorbance at 734 nm was measured after 6 min of incubation. In both methods, the Trolox solution (1 mg/mL) was used as a standard.

In the case of SDS-PAGE, 3 μL of protein standard or 5 μL of protein extracts was loaded into the wells on a polyacrylamide gel; all sample variants were present on the same gel for comparison. Electrophoresis was run in Tris–Glycine–SDS buffer at 160 V for around 50–54 min, until the last marker of the protein standard was close to the edge of the gel. After separation, the gel was stained with a staining solution containing Coomassie Brilliant Blue R-250 (0.025%) for an hour and later washed three times with washing solution containing acetic acid (7%), each wash lasting 30 min. The gel was left overnight in distilled water to finish the destaining process and scanned using the ChemiDoc MP Imaging System with Image Lab software (version 5.2.1. build 11, 2014).

High-performance liquid chromatography (HPLC) was used to evaluate the contents of basic carbohydrates (fructose, glucose, sucrose) as well as caffeine and chlorogenic acid. The Nexera HPLC system (Shimadzu, Kyoto, Japan) was composed of the following components: SCL-40 system controller, DGU-405 degassing unit, LC-40D XR solvent delivery module, SIL-40C XR auto sampler, CTO-40C column oven and ELSD-LT II low-temperature evaporative light scattering detector (ELSD; used for the carbohydrate analysis) or SPD-M40 photo diode array detector (PDA; used for caffeine and chlorogenic acid analysis). In the case of carbohydrates, analytes were separated using the Supelcoil LC-NH2 column (25 cm × 4.6 mm, 5 μm; Supelco) at a temperature of 30 °C. The mobile phase was a mixture of water and acetonitrile in a 22:78 ratio; it was used in isocratic flow (1 mL/min). Carbohydrates were identified based on comparison with standard solutions analyzed under the same conditions. The abovementioned method is described in detail in our previous article [10].

Analysis of caffeine and chlorogenic acid was performed using a method described by Vazquez et al. [17] with some modifications. A Kinetex C18 column (150 × 4.6 mm, 5 μm, 100 A; Phenomenex, Torrance, CA, USA) was used at a temperature of 50 °C. The mobile phase was a mixture of methanol (A) and acetic acid in water (B; 0.1%) in changing proportions; the flow was 0.2 mL/min. The gradient applied was as follows: 0–5 min, 95% B; 5–16 min, 10% B; 16–35 min, 10% B; and 35–40 min, 95% B. The total time of the analysis was 45 min. The injection volume was 3 μL. The standard solution was analyzed under the same conditions, which allowed for the identification of caffeine and chlorogenic acid in coffee extracts. The retention time and characteristic wavelength for chlorogenic acid were 18.74 min and 325 nm; for caffeine, they were 19.74 min and 272 nm. Furthermore, linear calibration curves were created for quantification. All chromatographs were processed using LabSolutions software (version 5.117).

Liquid chromatography coupled to mass spectrometry (LC-MS) was used to analyze the acrylamide content in roasted beans. The Nexera HPLC system with an LCMS-8050 mass spectrometer (Shimadzu) was utilized in this analysis. Samples (2 μL) were separated on the Synergi Hydro-RP (150 × 2 mm, 4 μm, 80 A; Phenomenex) column at 30 °C. The elution of compounds was conducted with 0.1% formic acid in water at 0.2 mL/min; the total time of analysis was 25 min. In MS, electrospray ionization (ESI) in the positive ion mode and multiple reaction monitoring were utilized. The temperature of ESI was 300 °C, and the voltage was 4.0 kV. The desolvation line temperature was 250 °C, and the heat block temperature was 400 °C. The collision gas was argon; its pressure was 270 kPa. Nitrogen was used as the nebulizing and drying gas with respective flows of 3 L/min and 10 L/min. Compressed air was a heating gas with a flow of 10 L/min. Acrylamide and acrylamide-d_3_ were used as the standard and internal standard, respectively. For both acrylamide and acrylamide-d_3_, three ion transitions were chosen for analysis and optimization of collision energy. Acrylamide (retention time, 4.43 min) had the following transitions: 71.9 → 55.05 (quantitative), 71.9 → 27.1 (reference), and 71.9 → 44.0. Acrylamide-d_3_ (retention time, 4.40 min) was analyzed with the following transitions: 75.1 → 58.05 (quantitative), 75.1 → 30.0 (reference), and 75.1 → 44.05. Standard solutions containing both acrylamide and acrylamide-d_3_ were analyzed under the same conditions in order to create a linear calibration curve of the peak area ratio and concentration ratio between both compounds. LabSolutions software (version 5.99 SP2) was used to analyze the LC-MS data.

Volatile organic compounds (VOCs) from roasted coffee samples were analyzed using headspace solid-phase microextraction and gas chromatography coupled to mass spectrometry (HS-SPME-GCMS). Roasted and ground coffee (500 mg) was weighed in a glass vial (15 mL), and 1 μL of naphthalene-d_8_ solution (0.2 μg/mL) was added as an internal standard. The sample was incubated at 50 °C for 10 min. An SPME fiber (Divinylbenzene/Carboxen™/Polydimethylsiloxane; 1 cm × 50/30 μm; Supelco) was inserted, and the extraction of VOCs was conducted at 50 °C for 10 min. Next, VOCs were desorbed (230 °C, 1 min, splitless mode) in a GCMS system (GCMS-QP2010; Shimadzu) and separated on a ZB-WAXplus polar column (30 m + 5 m guard × 0.25 mm, 0.25 μm; Phenomenex). The column temperature was programmed as follows: 0–3 min, 40 °C; 3–50.5 min, 230 °C; and 50.5–62.5 min, 230 °C. Helium was used as a carrier gas with a flow of 0.8 mL/min. The ion source and interface temperatures were both set at 230 °C. During analysis (1–53 min), mass spectrometry was used to scan ions with a mass-to-charge ratio (m/z) in the range of 45–500. The obtained data was processed using LabSolutions software (version 4.52); VOCs were identified based on a similarity search using the NIST and Wiley libraries and quantified based on the pick area of an internal standard. VOCs were classified based on their affiliation to groups created during roasting, namely furans, pyrazines and pyridines; the categorization of other compounds was based on the chemical structure of their functional groups.

2.4. Statistical Analysis

Results were calculated using MS Excel and expressed as means ± standard deviations. All statistical analyses were performed in R Statistical Software v4.3.2 [18] with the confidence level set at 0.05.

One-way analysis of variance (ANOVA), ANOVA with Welsh correction (ANOVA–Welsh) and the Kruskal–Wallis test (KW) were used to detect significant differences within the analyzed enzyme modification variants (V1 and V2) as well as between all non-fermented control samples. The choice between these methods was based on the fulfillment of assumptions regarding the normal distribution of the model’s residuals and equal variance between the tested groups, which were tested with Shapiro–Wilk and Levene tests, respectively. If the analysis of variance was significant, further tests were used to separate results into groups—Tukey’s HSD, Games–Howell and Wilcox tests for ANOVA, ANOVA–Welsh and KW, respectively. To analyze differences between samples fermented with the same bacterium but from different enzyme modification variants, the t-Student or Wilcox test was used, depending on the fulfillment of the normal distribution of values requirement. Furthermore, correlation between most of the obtained numerical data was calculated using the Hmisc package v5.3-3 [19], and the results were demonstrated as a correlation matrix created with the use of the corrplot package v0.95 [20]. Treeplots and barplots were created using functions from the ggplot2 package v4.0.2 [21].

3. Results and Discussion

3.1. Enzyme Addition and Fermentation

Robusta coffee was modified with the use of proteolytic enzyme (pepsin) and three selected bacterial species. Following the basic modification procedure described in Section 2.1, two possibilities for enzyme addition were identified—during the coffee soaking step (V1 samples) and together with the bacterial inoculum (V2 samples). As for the first option, enzymatic modification lasted 3 h and was ended by the sterilization process. On the other hand, the addition of the enzyme at the beginning of the fermentation step allowed for an extended duration of its activity and the observation of possible interactions between the effects of microorganisms and the enzyme on coffee characteristics. During material preparation, the extraction of coffee components (mainly carbohydrates and soluble proteins) was previously observed [7]. These compounds were likely available to microorganisms and supported their growth.

The fermentation process was monitored by cell count and pH measurements of the fermentative liquid; all measurements were performed at the 0 h and 24 h marks. The results of these analyses are presented in Table 1. In order to observe possible changes connected to enzyme addition at different steps of modification, the increase in the cell number was calculated and statistically analyzed. The results show that the growth of all utilized bacteria was similar in all processes. For V1 samples, LP exhibited the greatest growth (2.37 logCFU/mL); in the case of V2 samples, Gsp was characterized by the highest increase in cell number (2.64 logCFU/mL). Regardless of the enzyme modification variant, the rest of the results varied by around 0.2 logCFU/mL; the differences were not statistically significant. Similar to our previous results, we observed growth for all tested bacterial strains, which confirms that the fermentation mixture composition was rich in carbon and nitrogen sources [10]. While analyzing various bacterial strains isolated from wet coffee fermentation for their growth ability in a medium with coffee pulp, Ribeiro et al. [22] observed a great increase in the population of Leuconostoc mesenteroides and L. plantarum strains (around 2 logCFU/mL) after 48 h of incubation. They also observed a slight decrease in the case of B. subtilis-inoculated samples. In comparison, we observed a similar growth rate in all prepared fermentations, which might be attributed to strain differences and the application of whole beans, from which nutritional compounds could diffuse to the water medium and provide a better environment for bacterial growth.

Comparison of the results from fermentations with the same bacterium and different enzyme modification variants showed that BS and Gsp presented slightly better growth when pepsin was added together with the inoculum, while the results for LP demonstrated the opposite trend; no statistical significance was detected.

The acid–base conditions of the fermentative mixture were assessed by pH measurement. At the start of fermentation, the pH values were in the range of 5.43–5.56. Minor variations were observed, which proved to be statistically significant. Slight fluctuations in physical processes taking place during coffee preparation steps (soaking, sterilization) might explain the observed differences in these values.

After 24 h of fermentation, clear differences in pH values were observed; changes were mostly related to the bacterial species utilized during the process, and the effect of the enzyme modification variant was mostly insignificant. For V1 and V2 groups, the control (5.59 for both variants) and BS-fermented samples (5.56 and 5.59 for V1 and V2, respectively) obtained the highest pH values. Both Gsp and LP lowered the pH values of their fermentative mixtures; the latter one to the lowest values—4.05 and 3.99 for V1 and V2, respectively. The control samples did not differ in this respect from each other; the final pH values were slightly higher (0.01–0.07) when compared to those observed at the beginning of the process, probably as an effect of soaking. These observations are in accordance with our previous research [10] and data available in the literature. Ribeiro et al. [22] observed a decrease in the pH values of coffee pulp culture media inoculated with LAB strains (values lower than pH 4) as well as almost no change in the case of inoculation with B. subtilis (initial pH around 5.5). Similarly, Duque-Buitrago et al. [23] reported a strong ability of the utilized LAB strains (Streptococcus thermophilus and Lactobacillus delbrueckii subsp. bulgaricus) to decrease the pH and increase titratable acidity during fermentation. They also declared that these changes were correlated with an increase in the sensory score of the final product. The use of LAB resulted in the production of mainly lactic acid, which greatly influenced acidity values. According to the literature, B. subtilis strains also produce organic acids (i.e., malic and acetic acids); however, this does not result in a strong lowering effect on pH values [22,24].

3.2. Basic Coffee Compounds—Carbohydrates and Proteins in Green Coffee

Carbohydrates are of great importance during coffee fermentation—on the one hand, they are a source of carbon for microorganisms used for modification; on the other, they act as precursors for numerous aromatic compounds created during the roasting process. The amounts of basic carbohydrates (fructose glucose, sucrose) were analyzed using HPLC, and the results calculated per 100 g of dry matter (dm) are presented in Table 2. The main carbohydrate in green Robusta beans was sucrose (1.80–2.23 g/100 g dm), followed by fructose (177.38–237.73 mg/100 g dm); glucose was quantified only in the V0 control sample in the amount of 198.25 mg/100 g dm. The addition of proteolytic enzyme might explain this phenomenon—pepsin facilitated the diffusion of glucose from the coffee matrix by altering the bean structure, thus allowing for its presence in quantifiable amounts only in the V0 control sample.

Fructose is a monosaccharide easily metabolized by most of the bacteria. In both the V1 and V2 sample groups, beans fermented with the use of Gsp exhibited the highest amount of this carbohydrate; contrarily, the use of LP resulted in a reduction in fructose in beans to trace amounts. The results presented here are similar to those reported in our previous study [10] and point to differences in bacterial metabolisms. The control samples did not differ between each other, suggesting that the addition of pepsin at different steps of the modification procedure alone had no significant effect on the content of this monosaccharide; the use of bacteria was the main factor determining the fructose content in beans. Fermented samples, however, showed a consistent trend where the addition of the enzyme together with the bacterial inoculum (V2) allowed for a higher content of fructose in the final beans; this observation was significant only in the case of Gsp-fermented samples.

According to the literature, sucrose is the most abundant disaccharide present in green coffee beans [25,26]—our results for modified Robusta beans correspond with this statement as well. The highest amount was observed for modification with BS (2.23 and 2.10 g/100 g dm for V1 and V2 variants, respectively). The use of acid-producing bacteria (APB) resulted in the greatest reduction in sucrose content in the V1 sample variant. In the controls, each addition of pepsin during the modification procedure resulted in a decrease in the sucrose present in the final beans; a significant decrease was observed only in the case of the V2 variant, most probably due to greater enzyme activity. Similarly, the timing of pepsin addition had a significant effect on all fermented samples—in the case of BS, a trend similar to the one observed in the controls is present, while the results obtained after fermentation with APB presented the opposite tendency.

Wang et al. [27] observed that, with a longer germination time, the sucrose content decreased in roasted Arabica beans; interestingly, soaking alone resulted in an increase in its content compared with the control sample. The observation regarding the germination time could be related to the results observed in this study (specifically for controls). It could suggest that the observed decrease in sugars during coffee modification under water conditions could be attributed to either extraction from the coffee material or the beans’ metabolism associated with the germination process occurring before the sterilization step. In the case of samples modified with enzyme and bacteria stains, , further changes in the carbohydrate contents signify the metabolic activity of microorganisms. Park et al. [28] observed a continuous decrease in the total sugar content and individual free carbohydrate contents in both coffee pulp and beans during dry fermentation with L. mesenteroides, Saccharomyces cerevisiae and Aspergillus oryzae. Duque-Buitrago et al. [23] also reported a decrease in the soluble solid content of the fermentative liquid after modification with S. cerevisiae and LAB, signifying their activity. Furthermore, the condition of the microbial inoculum (optimized in coffee by-products or directly from sourcing) affected the bacteria’s ability to metabolize sugars and might increase the diffusion of soluble compounds from the beans to the liquid medium. Aditiawati et al. [29] analyzed the effect of a Bacillus spp. consortium on the composition of Indonesian green Arabica beans from various origins. After modifications, higher contents of sugars and amino acids were observed, which significantly distinguished the beans from their controls. Furthermore, with a longer modification time, slight decreases in the sugar contents were observed and attributed to the production of organic acids. This suggests that, at the beginning, the use of microorganisms could increase the amount of carbohydrates due to their enzymatic activity and loosening of the beans’ structure; however, APB utilized these carbohydrates further to produce organic acids in great amounts, resulting in lower pH values after fermentation. Organic acids produced during microbial modification can migrate back into beans and influence the quality of the final product [22,29].

Another important component of green coffee is nitrogen-containing compounds, such as soluble proteins and caffeine. The former might not only support microbial activity but also, together with carbohydrates, participate in aroma compound creation through Maillard reactions and Strecker degradation [25]. Although, in this experiment, pepsin was added at a pH different from its optimal range (pH 2–4), the literature shows that it exhibits proteolytic activity also in weakly acidic conditions (pH < 6) [7,30,31], which further supports the use of this enzyme in the current study. Starting with the control samples, a clear difference between samples can be observed. The addition of the enzyme during the soaking step (V1) resulted in a significant increase in the soluble protein amount compared to the control for V2; the control without any additional modifications (V0) did not differ significantly from either of them, but its value was closer to V2 (Table 2). This shows that the duration of enzymatic modification in the case of the V1 sample variant (3 h) was sufficient to observe significant changes in the amount of soluble proteins in green Robusta beans, which was also concluded in our previous research [7]. It is possible that the addition of pepsin after sterilization allowed for greater enzyme activity on the loosened structure of the coffee bean, in effect accelerating the extraction of this group of compounds; a longer duration of enzymatic treatment might have also had an effect on the extraction rate. The inclusion of pepsin at the soaking stage might have affected proteins in the coffee material, and its breakdown was also facilitated by the subsequent sterilization; the amount of soluble proteins might have been greater than in other coffee variants, and even constant extraction during further modification was not able to lower its amount in the final beans.

The addition of microbial fermentation to enzymatic modification allowed the production of beans with a high content of soluble proteins and prevented the significant decline observed in the control sample of the V2 variant. In the V1 sample group, all results were in a narrow range of values (15.80–18.67 g/100 g dm); the highest amount of soluble proteins was found in the sample fermented with Gsp and the lowest in the sample fermented with LP. Looking at the V2 sample variant, it can be observed that all fermented samples exhibited significantly higher values than their controls. Fermentation with both APB resulted in the highest values in this group—20.12 g/100 g dm and 21.05 g/100 g dm for Gsp and LP, respectively. This could be attributed to the enzymatic activity of bacteria, which could further affect the proteins present in green coffee through degradation or solubility changes. Comparing fermentation samples from the V1 and V2 variants, a significant difference was observed only in the case of LP.

A closer inspection of soluble proteins was conducted with the use of SDS-PAGE. The results of protein separation based on molecular weight are presented in Figure 1. All samples presented two main bands, representing calculated weights of 34.65 ± 0.19 kDa and 22.66 ± 0.22 kDa. These bands indicate the presence of 11 S storage proteins, which during electrophoresis with a reducing agent (β-mercaptoethanol) are degraded into two separate groups: 33 kDa and 24 kDa [32]. Clear differences in the intensity of the bands can also be observed, which correlates with the amount of each protein. The control sample from the V1 variant has bands with the greatest intensity, compared with the other control samples; bands for all fermented samples were also more intense than their controls. The difference between the two variants of the same fermentation process was not visible, apart from BS, where the V1 sample variant seemed more intense compared to the V2 sample. Interestingly, a broad protein smear between 17 kDa and 11 kDa was detected across all samples. This suggests the presence of low-molecular-weight products of protein degradation due to the activity of various enzymes—those natural to coffee and those added during the modification process. It was previously shown that enrichment in protein degradation products (mainly various amino acids) might influence the quality of the final product [29], which has also been reported for animal-processed coffee [33,34]. The presence of a protein smear in the control sample modified without enzyme addition (V0) points to the possible activity of natural enzymes present inside the bean and connected to the germination process, which could occur during the modification process [26]. The greater intensity of the protein smear in the V1 control sample and all fermented samples suggests that additional processing (both enzymatic and microbial) intensified degradation processes initiated by natural coffee enzymes. On the other hand, the loss of intensity in the V2 control sample might be explained by extraction processes occurring during modification.

3.3. Bioactive Coffee Compounds—Caffeine and Polyphenols in Green and Roasted Coffee

Caffeine is another nitrogen-containing compound of great significance for coffee—it is responsible for its stimulating effect on the human body, and its bitter taste plays a role in the creation of the complex coffee flavor [28,35,36].

In green Robusta beans, the caffeine content was in the range of 1.05–1.18 g/100 g dm (Table 2). Samples fermented with LP had the highest content of this alkaloid regardless of the enzyme addition variant; in the V2 modification variant, the results obtained for APB-fermented samples did not differ significantly. In the controls, the addition of pepsin during the fermentation stage (V2) resulted in a significant decrease in the caffeine content; this trend is comparable with the changes observed for the sucrose content in the control samples. It can be speculated that the reduction in caffeine might be related not only to diffusion from the coffee matrix but also to changes associated with the germination process [24]. Similar to the soluble protein changes, the addition of microbial fermentation prevented the loss of caffeine observed in the V2 sample variant. Furthermore, enzyme modification variation had a very slight effect on the results; it was significant only for samples fermented with BS.

Caffeine was also quantified in roasted coffee beans, and the results of this analysis are reported in Table 3. These results are lower than the contents reported for green beans (0.57–0.66 g/100 g dm). Similar to the changes in caffeine content described above, the use of LP resulted in the highest obtained values regardless of the enzyme modification variant. It is worth noting that in the V1 group, both APB had high contents of caffeine, and they did not differ from each other. As for the control samples, they exhibited low caffeine contents—in all cases, similar to the values for BS-fermented samples. Comparing fermented samples from different enzyme addition variants, minor fluctuations can be seen; only Gsp-fermented samples differed significantly in this regard.

Additional processing of coffee can affect the caffeine content in various ways. While analyzing sequential fermentation with L. plantarum and S. cerevisiae, Borém et al. [37] observed little change in the caffeine content between samples. On the contrary, Purwoko et al. [38] reported a decrease in the content of this alkaloid during fermentation with various LAB and fungi; only in the case of 4-day modification with Rhizopus oryzae was the decrease statistically significant. Hájíček et al. [34] reported that animal-processed coffee (kopi luwak, black ivory) exhibited lower amounts of caffeine in comparison to typical roasted Arabica beans, probably due to the passage of beans through the digestive tract of animals and the accompanying absorption and/or degradation of caffeine. While analyzing regular and luwak roasted beans using a metabolomic approach, Farag et al. [33] did not find markers for caffeine and trigonelline in samples processed by civet cats. These data show that the additional animal processing of coffee affects the caffeine content in the final roasted products. The results presented in this study further emphasize this statement and underline the effect of the utilized microorganisms and pepsin addition on this coffee attribute.

The effect of coffee drinking on the human body is not limited to stimulation. Apart from its high caffeine content, coffee is rich in various phenolic compounds, which exhibit antimicrobial, antioxidant and anti-inflammatory abilities [36]. It is reported that coffee consumption can lower the risk of lifestyle-related diseases, such as cardiovascular, neurologic and metabolic diseases [35]. In this study, the amount of chlorogenic acid—one of the most important phenolic compounds in coffee [28]—and TPC was analyzed.

In the green coffee beans, the species of bacterium used in the fermentation had a greater effect on both chlorogenic acid and TPC. The amount of chlorogenic acid was in the range of 1.95–2.17 g/100 g dm, and TPC was in the range of 2.40–2.86 g GA/100 g dm (Table 2). In both enzyme modification variants, the use of APB allowed for the highest observed values in both analyses, which is similar to the trends observed for caffeine content.

These results are in accordance with Therdtatha et al. [36], who observed an increase in the amount of coffee antioxidants in Robusta samples after optimized fermentation with yeasts and L. plantarum. On the contrary, Arabica beans fermented with L. plantarum were characterized not only by higher antioxidant activity but also by a decreased amount of most of the metabolites, including chlorogenic acid. These results suggest that the observed differences are derived not only from inoculum differences but also from coffee matrix variations.

The results for the control samples were either similar to those for BS-fermented samples (in V1) or significantly lower (in V2). Furthermore, the controls exhibited a similar trend to the one observed for soluble proteins—addition of pepsin during the soaking stage (V1) allowed for higher contents of chlorogenic acid and TPC compared to other samples in this group. This might suggest the release of phenolics bound in complex protein structures due to enzymatic activity; however, it was not as strong as observed in fermented samples [36]. The addition of enzymes at a different stage of the modification process again seemed to cause slight fluctuations in values, which were significant only in the case of BS-fermented beans.

After roasting, the contents of chlorogenic acid and TPC decreased (by approximately 75% and 50%, respectively) due to their thermolability (Table 3). In the case of the V1 sample group, modification with Gsp resulted in the highest chlorogenic acid content (0.62 g/100 g dm); the sample fermented with LP also exhibited a high value (0.57 g/100 g dm). As for TPC, the use of either one of the APB also resulted in higher values (1.25 and 1.33 g GA/100 g dm for Gsp and LP, respectively) compared to the BS-fermented sample and control. Samples from the V2 group had comparable contents of chlorogenic acid; however, the use of LP strongly influenced TPC. Aditiawati et al. [29] observed an increase in TPC in both green and roasted beans modified with Bacillus spp. inoculum; however, the effect varied among samples depending on the modification conditions; slight variation due to various origins was also observed. In other studies, roasted luwak beans were characterized by the highest amount of chlorogenic acid, while elephant-digested coffee had the lowest [34]. Apart from possible absorption during the digestion of the coffee fruit, differences in gut microflora and their enzymatic activity might have influenced these results, as shown in this study. In both analyses, the effect of pepsin addition at various stages of modification was significant only in the case of fermentation with Gsp.

3.4. Antioxidant Activity of Roasted Coffee

Coffee exhibits antioxidant properties mainly due to a high concentration of phenolic compounds and melanoidins [39]. In this study, antioxidant activity was analyzed using two in vitro methods, namely CUPRAC and ABTS. The results presented in Table 3 show that values obtained with the CUPRAC method (3.37–4.02 g Trolox/100 g dm) were generally higher than those obtained with the ABTS method (1.18–1.41 g Trolox/100 g dm). This could be explained by differences in their reaction conditions and mechanisms—CUPRAC acts on the basis of single-electron transfer, while ABTS is based on a mixed-mechanism model—as well as the preferred mechanism of action of antioxidant compounds present in roasted coffee. Moreover, the use of Gsp (in V1 group) and LP (in V1 and V2 groups) resulted in the highest activity, regardless of the analytical method. Hájíček et al. [34] showed, however, that civet- and elephant-processed coffee had lower antioxidant activity when tested with various analytical methods; as the exception, luwak coffee was characterized by the greatest result in the ORAC (oxygen radical absorbance capacity) method. Such results might be explained by the extensive effect of the digestion process and loss of antioxidant compounds. On the other hand, Therdtatha et al. [36] showed that a decrease in phenolics is not always connected to lower antioxidant activity. The production of different compounds with antioxidant activity—such as peptides, melanoidins or aromatic compounds — might provide an explanation for this phenomenon [40,41].

Considering the control samples, the addition of pepsin during the fermentation stage (V2) resulted in higher antioxidant activity as compared with pepsin addition at the soaking stage. On the contrary, when considering samples modified with both enzymatic and microbial activity, the V2 samples showed slightly lower values compared to their V1 equivalents; this was only significant in the case of Gsp in both methods.

3.5. Browning Index, Acrylamide and Volatile Compounds in Roasted Coffee

Melanoidins are among the many compounds created during the roasting of coffee beans; they are products of Maillard reactions and are characterized by high molecular weight. Melanoidins are mostly responsible for coffee color, but, as already shown, they can also exhibit antioxidant activity [42]. Brown pigments in coffee were estimated by the measurement of the absorbance of water extracts at 420 nm, obtaining a BI. As shown in Table 3, LP-fermented samples presented the highest values (0.34 and 0.37 for V1 and V2, respectively). Modification with BS and Gsp significantly decreased BI values, resulting in scores lower than those obtained for the controls. Interestingly, samples which showed lower amounts of carbohydrates prior to roasting (obtained with APB) were characterized here by higher BI values. Wang et al. [43] reported similar observations—samples after glucose supplementation and fermentation with Lactococcus lactis subsp. cremoris were darker in color after roasting under various conditions, as compared with other samples, including those fermented without glucose supplementation and controls. They concluded that higher acidification of the coffee after supplemented fermentation accelerated the processes during roasting—not only Maillard reactions but also caramelization—resulting in greater production of brown-colored compounds, such as melanoidins.

No significant differences were found between the control samples, which shows that enzymatic modification alone did not affect the BI of roasted coffee. Interestingly, in all fermented and roasted beans, the same trend regarding pepsin addition is clearly visible—when the enzyme was added together with the bacterial inoculum, higher BI values were obtained.

Another compound created during coffee roasting is acrylamide. It is mainly formed as the result of a reaction between a reducing sugar (fructose, glucose) and asparagine, as a part of Maillard reactions in various foods. The International Agency for Cancer Research (IARC) classified acrylamide as probably carcinogenic to humans (group 2A), and coffee is considered one of the main dietary sources of its exposure due to high consumption [11,41]. The amounts of acrylamide in coffee samples are presented in Table 3. All control samples had more than 10 μg/100 g dm of acrylamide; the highest value was observed for the control sample from the V2 variant. The combination of enzymatic and microbial modification allowed for a reduction in the acrylamide content to below 10 μg/100 g dm, with the exception of the BS-fermented sample from the V2 variant. In both enzymatic modification variants, the lowest results were obtained after fermentation with LP. The amounts of acrylamide in coffee reported in this study are in accordance with the contents reported for conventional roasted ground coffee [11,44]. The observed reduction in its content after additional modification might be connected with the decrease in reducing sugars and asparagine contents due to microbial metabolism. Wang et al. [27] reported a great reduction in acrylamide after soaking and germination of coffee beans; they suspected the degradation of asparagine during these processes as the main reason behind the observed change, which might have occurred in the samples analyzed here as well. Starowicz and Zieliński [45] stated that acrylamide content is correlated with a darker color of the product. On the other hand, Mojska and Gielecińska [44] noted a negative correlation between the color of roasted beans and acrylamide content, which is similar to the results presented in this work. Such observations can be explained by acrylamide decomposition with increasing roasting time, as this compound is created primarily at the beginning of this process [44].

One of the sensory qualities of coffee is its complex aroma, composed of various VOCs created during the heat treatment of green beans. In this study, VOCs present in roasted coffee were analyzed with the use of HS-SPME-GCMS. More than 90 different compounds were identified, and their full list is presented in Table S1. VOCs were separated into groups based on their affiliation with classes of compounds created specifically during roasting and based on their chemical structure. Ten groups were created, with the following numbers representing the amount of compounds in each of them: furan derivatives (22), pyrazine and its derivatives (16), other heterocyclic compounds (14), organic acids (11), ketones (10), phenol and its derivatives (6), esters (5), alcohols (4), and pyridine and its derivatives (3).

The amount of each compound was calculated with the use of an internal standard (naphthalene-d_8_). The total amount of VOCs (in μg/100 g of roasted coffee beans) is presented in Table 3. In the case of the controls, the one from the V1 variant was characterized by the greatest amount of VOCs. The addition of pepsin at the beginning of the fermentation step (V2 variant) resulted in lower values compared to enzyme addition at the soaking stage (V1 variant). Regardless of the timing of pepsin addition, fermentation with LP increased the total amount of VOCs, which is in accordance with reports indicating that the use of L. plantarum enhances coffee aroma [36].

The participation of each group of compounds (in percentage) was calculated and presented in the form of treemaps (Figure 2 and Figure 3). For the three groups with the highest participation, further statistical analysis was performed.

Furan is classified as potentially cancerogenic to humans (2B group) by the IACR; it also exhibits genotoxicity and hepatotoxicity [46]. The presence of furan in heat-treated food and its exposure to humans has raised great concern. Furthermore, some furan derivatives—e.g., methyl derivatives, 2-furanmethanol (also classified in 2B group)—might also present a threat to human health due to their metabolism [46,47]. Nevertheless, from a sensory point of view, furans are valuable aroma and flavor compounds responsible for notes such as sweet, caramelic, bready and woody [35,43].

Furan derivatives were the most abundant group in all analyzed samples (approximately 39.0–52.0%, Figure 2 and Figure 3). The effect of enzyme addition at different modification stages can be seen in the control samples—the addition of pepsin during the soaking stage resulted in an increased amount of furan derivatives, while the addition of the enzyme during the fermentation stage had the opposite effect. Compared to the V1 control, the size of the observed changes was 1.4–2.6 percentage points. In the case of the fermented samples, no clear trend regarding enzyme modification variants was observed; the use of LP generally resulted in a great increase in the furan derivatives’ participation in the overall aroma compounds, exceeding 50% in both the V1 and V2 samples.

The most abundant representatives of furan derivatives were 2-furancarboxaldehyde (furfural), 5-methyl-2-furancarboxaldehyde (5-methylfurfural) and 2-furanmethanol (furfuryl alcohol). For these compounds and the sum of the other furans, further statistical analysis was performed in order to detect dependencies related to the modification process. These results are presented in Figure 4. Fermentation with LP combined with pepsin addition at the beginning of the modification resulted in the highest content of the abovementioned furan derivatives and sum of other furans—605.44 ng/100 g, 534.72 ng/100 g, 467.58 ng/100 g and 287.63 ng/100 g for furfural, 5-methylfurfural, furfuryl alcohol and other furans, respectively. In the case of furfural and its methyl derivative, the control and BS-fermented samples differed significantly from LP-fermented V1 beans. Samples from the V2 modification variant showed lower amounts of analyzed compounds; however, no significance was observed due to the high standard deviation of the results. Among the controls, the V1 sample exhibited the highest amount of furan derivatives, while the V2 sample presented the lowest. The results for furfuryl alcohol and other furans presented the same trends regarding the modification process; for the former, however, this trend was slightly less pronounced.

These results show that additional processing had an effect on the amounts of important volatile compounds in coffee. The choice of microorganisms seems to be of great importance. According to the literature, their ability to produce organic acids and to change the pH value during modification is one of the key factors. Wang et al. [43] also observed increased amounts of furan derivatives in roasted beans after glucose-supplemented fermentation with L. lactis subs. cremoris. They explained this by the acidification of the beans, sufficient for the modification of the Maillard reaction mechanism to produce higher amounts of furfural and its derivatives. Furthermore, Duque-Buitrago et al. [23] performed fermentations with LAB and S. cerevisieae using various media; after principal component analysis, they noted that the lower pH and higher titratable acidity were correlated with higher sensory scores of beans fermented with LAB and combined inoculum of LAB and yeasts. Furthermore, the importance of the Maillard reaction environment was shown by Adams and Kimpe [48]—with a pH value change from 3 to 6, they observed lower generation of furfural during heat treatment of aqueous model mixtures of ascorbic acid and various amino acids. Above pH 6, no furan derivatives formation was reported. The abovementioned changes observed by other researchers explain the results presented in this study.

Pyrazine and its derivatives were the second most abundant group of VOCs identified in coffee—across all samples, they accounted for 14.0–27.0% of the total aroma compound content (Figure 2 and Figure 3). BS-fermented samples were characterized by the highest participation of pyrazine and its derivatives (more than 26.0% in both V1 and V2 sample variants), while the use of LP created samples with the lowest percentages (14.68% and 15.27% for V1 and V2 samples, respectively).

Methylpyrazine was the most abundant derivative in this group (Figure 4). Three ethylmethylpyrazine isomers (namely 2-ethyl-6-methylpyrazine, 2-ethyl-5-methylpyrazine, 2-ethyl-3-methylpyrazine) and two dimethylpyrazine isomers (2,5-dimethylpyrazine, 2,3-dimethylpyrazine) were summarized, and their sums were used in further analysis; in each list, the first compound is the one with the highest concentration. Both the sum of ethylmethylpyrazines and the sum of dimethylpyrazines were at a similar level to other pyrazines; they were also approximately two times lower than the amount of methylpyrazine. In the case of the sum of ethylmethylpyrazines and methylpyrazine, some significant differences can be observed between samples fermented with BS and LP, the latter one exhibiting lower values. According to Adams and Kimpe [46], the addition of a base (K_2_CO_3_) increased the amount of pyrazines created during the dry roasting of model solutions containing ascorbic acid and various amino acids. This might explain the results observed in this study—more pyrazine derivatives were created during the roasting of beans fermented with BS, since this modification did not significantly change the final pH level of the fermentative mixture. The lower amount of this group of VOCs in the controls might be related to the lower amount of soluble proteins compared to BS-fermented beans as well as the different composition of available peptides and amino acids as a result of enzyme activity. Higher amounts of protein degradation products might also explain the lower concentration of furan derivatives in beans modified with BS—according to the literature, carbonyl intermediates produced during dry heating are diverted towards amino acid interactions, reducing their participation in furan formation [48].

Similar to furan derivatives, the addition of pepsin together with bacterial inoculum (V2 variant) seemed to lower the amount of pyrazine derivatives in coffee samples. Statistically, there was no difference regarding the content of dimethylpyrazines and other pyrazines across all samples, but the same trends as mentioned above can be easily seen. In regard to the controls, the addition of pepsin at various modification steps resulted in both a decrease (sums of ethylmethylpyrazines and dimethylpyrazines) and increase (methylpyrazine) in the observed values.

The third group of the most abundant VOCs in the analyzed samples was pyridine and its derivatives. This group accounted for 5.90–10.32% of the total amount of aroma compounds in all samples (Figure 2 and Figure 3). Similar to the pyrazine results, the amount of pyridine and its derivatives was the highest in samples fermented with BS and the lowest where fermentation was conducted with LP. The addition of pepsin at the fermentation stage seemed to slightly lower the relative amounts of pyridine, as can be observed in the BS- and Gsp-fermented samples. The opposite trend is evident in the controls and samples fermented with LP. Nevertheless, the amplitude of the observed changes was lower than one percentage point.

Three compounds included in this group of VOCs are presented in Figure 4. 5-methyl-2-pyridineamine was the compound present in the highest amount in all samples (118–244 ng/100 g); pyridine and its acetyl derivative were present in much lower amounts—36–50 ng/100 g and 8–26 ng/100 g, respectively. The effect of enzyme addition again shows a consistent trend—V2 samples showed lower values than V1 samples. Furthermore, the addition of the enzyme at the soaking stage could cause a slight increase in the observed values, which is the most evident in the amount of pyridine in the control samples. As for the fermentation effect, tendencies similar to those observed in the pyrazines are present—in the V1 sample variant, fermentation with BS resulted in the highest amounts of 5-methyl-2-pyridineamine and N-acetyl-4(H)-pyridine, while the use of LP produced the lowest results. The amount of pyridine did not differ significantly between samples; the same trend regarding enzyme addition is clearly visible.

Recent studies on fermented coffee show that additional processing with the participation of animals, particularly their gut microflora and gastric enzymes, affects the aroma of coffee. Farag et al. [33] reported higher amounts of furans present in luwak coffee as compared to conventional Arabica beans. Hájíček et al. [34] also reported characteristic differences between luwak and elephant-digested coffee. Luwak coffee showed lower amounts of furfural and furfuryl acetate but a higher content of 2-ethyl-5-methylpyrazine; black ivory coffee was richer in pyrazine and 3-ethyl-2,5-dimethylpyrazine. Furthermore, the use of selected bacteria also influences the VOCs present in roasted beans. In their experiment, Ribeiro et al. [22] observed a higher participation of furans in coffee fermented with L. plantarum and L. mesenteroides. The use of the former, together with Microbacterium testaceum, increased the amount of pyrazines and pyridines detected in coffee. Previous studies also show that changes in the VOC content influence the sensory quality of the final product. Roasted beans after glucose-supplemented fermentation with L. lactis subs. cremoris were characterized by a stronger caramel aroma as compared with non-supplemented and control samples [44]. Park et al. [28] observed that additional inoculation with yeast, LAB and fungi separately increased the total score of the obtained coffee in sensory evaluation as well as changed the main sensory notes in each group. They noted that changes in the aromatic compound composition might be related to these results.

This study further emphasizes the ability of LAB to modify coffee VOC contents—an increase in the furan derivative content was observed, similar to data reported in the literature. Variations between the results presented here and those present in the literature might come from different bacterial strains, different modification procedures, or the initial characteristics of the coffee material—cherries/beans, plant and geographic origins, and processing and roasting conditions [24]. The use of BS also changed the composition of VOCs in the roasted coffee; in particular, it increased the amount of both pyrazines and pyridines, which are generally positively correlated with cup sensory quality [22]. Differences in the VOC composition and content might have an effect on the overall quality of coffee; however, further research is required to support this statement.

3.6. Correlations

As the final stage of data analysis, the correlation between most of the obtained data regarding green and roasted coffee was calculated (Figure 5). The final pH of the fermentative mixture was included as a parameter which characterized the modification process; cell count was omitted due to a lack of data for non-fermented control samples. The contents of VOCs discussed in Section 3.5. were also included. Having in mind the wide dispersion of some parameters and potential failure to meet statistical assumptions, Spearman’s rank correlation was chosen for these calculations.

Fermentation with various bacterial species and, connected to this, the pH value of the mixture after fermentation seem to be of great importance—all bioactive compounds (caffeine, TPC, chlorogenic acid) in both green and roasted coffee beans as well as the antioxidant activity of roasted coffee were negatively correlated with the pH value of the mixture at the end of modification. In these cases, the correlation coefficients (r) ranged from −0.68 to −0.78. On the other hand, the acrylamide content was positively correlated with the final pH value (r = 0.83). The sucrose content showed a similar negative correlation (r = −0.67 to −0.88) with most of the abovementioned parameters as well. Correlation with the sucrose content might suggest that a different activity of microorganisms caused the release of phenolic compounds from the matrix, thus allowing for their higher content in green beans and influencing antioxidant activity in roasted beans [25]. Acrylamide was also negatively correlated with the soluble protein content, caffeine content, TPC and chlorogenic acid content in green coffee (r = −0.73 to −0.83). As the soluble proteins are degraded, a higher content of peptides and amino acids (especially asparagine) is expected, thus promoting acrylamide formation.

Interestingly, bioactive compounds in green coffee samples exhibited a strong positive correlation with each other and the caffeine content in roasted samples (r = 0.72 to 0.98). In the case of roasted beans, the caffeine content was correlated with the chlorogenic acid content (r = 0.92), but all bioactive compounds showed a significantly strong correlation with the antioxidant activity results (r = 0.73 to 0.95). Additionally, a very strong positive correlation (r = 0.98) was detected between the results from the ABTS and CUPRAC methods. Apart from the sucrose content in green coffee, scarce significant correlation between the contents of green and roasted beans shows how strongly the roasting process affects coffee composition [33].

Strong and positive correlations were observed in the case of VOCs. The total amount of VOCs was strongly correlated with all furan derivatives, the sum of dimethylpyrazines and pyridine (r = 0.77 to 0.92), suggesting that these compounds were mostly responsible for the changes in this value. Furan derivatives showed strong interdependence (r = 0.72 to 0.95), while a cross-correlation was observed between pyrazines and pyridines (r = 0.68 to 0.95). This observation underlines the importance of the different trends discussed for each of these compound groups in Section 3.5.

There were more cross-correlations worth looking into. The furfural content was the only VOC representative significantly correlated with the final pH level of the fermentative mixture (r = −0.70), caffeine and TPC in green coffee (r = 0.72 and 0.67, respectively), as well as the chlorogenic acid content (r = 0.72) and antioxidant activity analyzed with the ABTS method (r = 0.67) in roasted coffee. The amounts of bioactive compounds in green coffee were also negatively correlated with selected compounds from pyrazines and pyridines (r = −0.67 to −0.77). The BI was significantly correlated only with the methylpyrazine and 5-methyl-2-pyridinamine contents (r = −0.70). Furthermore, most of the furan derivatives showed a significant positive correlation with the pyridine content; only in the case of furfuryl alcohol was the correlation coefficient higher than 0.90.

4. Conclusions

In this paper, the modification of green Robusta coffee beans was performed with the use of chosen microorganisms and a proteolytic enzyme (pepsin), which was added at two different stages of the modification process. Fermentation with different bacterial species had a much greater effect on coffee than pepsin addition. A strong correlation between the final pH of the fermentative mixture and bioactive compound retention in coffee material was observed; numerical data also indicates greater amounts of these compounds in beans fermented with APB, mainly with LP. However, correlation does not fully account for other possible changes in the coffee material, e.g., creation of antioxidant peptides and melanoidins or release of conjugated chlorogenic acids; thus, further study is needed to fully understand the observed relationships. The effect of enzyme addition alone was connected with the timing of its addition—generally higher amounts of coffee compounds were observed in samples modified with pepsin at the soaking stage compared with addition after sterilization. Combination of enzymatic and microbial modification minimized this trend, as bacterial modification exerted a much greater effect. Furthermore, the use of LAB increased the amount of furan derivatives in roasted coffee. This was probably the result of its acid production capabilities. On the other hand, modification with BS allowed for a higher content of pyrazines and pyridines.

This study highlights the potential to modify coffee characteristics by the application of widely accessible materials, including microorganisms and enzymes, in a process that can be performed with green beans in any location. The greatest potential for quality improvement was shown by fermentation with LP. Enzyme addition was of lesser importance; however, it might change some effects of microbial modification in coffee beans. As the investigated process altered the amount and composition of VOCs in roasted coffee, further research should prioritize the sensory evaluation of coffee and the roasting process as a vital step in the development of overall coffee quality.