The strains of Saccharomyces cerevisiae investigated in this research (L13, L37, L63 and L67) were previously isolated and identified by Silva et al. (18) and characterized according to pectinase activity by Carvalho et al. (22). They belong to the culture stock of the Applied Microbiology Laboratory (LABMA/UESC, Ilhéus, Bahia, Brazil) and they are preserved in a 15% (m/V) glycerol solution at -80 °C. A lyophilized commercial yeast was also used for performance comparison: S. cerevisiae CA-11 (recommended for the production of cachaça, a Brazilian distilled spirit made from sugarcane), which was donated by the company LNF (Bento Gonçalves, Rio Grande do Sul, Brazil).

The stock cellular solutions were obtained by cultivating each strain in 25 mL of Sabouraud broth at 28 °C for 48 h (SL-200; Solab, Piracicaba, São Paulo, Brazil). After cultivation, 100-µL aliquots were spread over the surface of the same solid medium that was incubated once again at 28 °C for 24 h. Then, the cells grown on the surface of the Petri dishes were scraped and rinsed with 2-mL aliquots of sterile distilled water. Total cell count in the collected volume was determined in a Newbauer chamber. This value was used to obtain the initial cell number N_o (cell/mL), labeled inoculum 1, which was then used in the preliminary tests and the study of fermentation conditions. For the activation and inoculum preparation of the commercial strain, 3 g of the lyophilized cells were used following the same procedure.

Inoculum 2 was adapted to the fermentation must by using approx. 30 mL of inoculum 1, 50 mL of cocoa honey and 2.5 g of sucrose, with sterile distilled water added until reaching a volume of 100 mL. Cultivation occurred at 30 °C for 12 h (SP-222; SP Labor, Presidente Prudente, São Paulo, Brazil), without stirring, and then the cell mass was separated by centrifugation (15 000×g for 5 min; 16R; Thermo Scientific, Jundiaí, São Paulo, Brazil) and the supernatant was discarded. The cells were resuspended in 15 mL of sterile distilled water and the total cell count was determined as described above. The same procedure was followed for strain CA-11.

The strains (L13, L37, L63 and L67) were cultivated in 250-mL Erlenmeyer flasks containing 150 mL of culture medium. The performance of each strain during incubation was observed for 4–5 days, including cell count (N/(cell/mL)), total soluble solid content (g/100 g), pH and flocculation (visual analysis), while ethanol volume fraction (%) was analyzed only at the end of fermentation. Initially, the strains were cultivated in solutions of commercial pectin (20 g/L) and sucrose (100 g/L) in a phosphate buffer (0.5 M at pH=4.0) with N_o around 1·10⁷–2·10⁷ cell/mL. Incubation was carried out at 30 °C and agitation speed of 150 rpm (SP-222; SP Labor). Then, the cultivation of strains L13 and L63 was evaluated in pure cocoa honey (φ_CH=100%) at initial conditions of pH=3.43 and 16.8 °Brix, under the same temperature and N_o, but without sucrose and agitation.

The best conditions for fermentation of cocoa honey and cocoa pulp by the yeast S. cerevisiae L63 were investigated applying the statistical tool of experimental design (23) with a Plackett-Burman matrix (PB8). For a total volume of 200 mL of medium, the independent variables (factors) and their lowest (-1) and highest (+1) levels were: volume fractions of cocoa honey (with or without cocoa pulp) (φ_CH=80−100%), sucrose concentration (γ_suc=0−100 g/L), temperature (t=28−32 °C) and initial cell count (N_o=10⁶−10⁸ cell/mL). The dependent variables (responses) investigated over 72 h of fermentation were: ethanol volume fraction (%), cell count (cell/mL), pH, total soluble solid content (g/100 g) and absorbance (A) read at 600 nm. In order to use the same cocoa honey batch for all of the PB8 experiments, it was not possible to perform the central points in triplicate but only in one replicate.

The responses obtained with the experimental design were used in the effects analysis (23) performed at 85% confidence using the statistical software Protimiza Experimental Design (24). In this analysis, the individual effects of each factor were considered statistically significant at p<0.15, which indicates that by increasing the factor from its lowest to its highest level, it is possible to obtain a statistically significant alteration (positive or negative) in the average response.

The concentrations of sugar, organic acid and alcohol (γ), number of cells (N), total soluble solids, pH and A were determined throughout the fermentation (V=200 mL) at 28 °C and N_o=6.5·10⁷ cell/mL, with samples being collected between 3 and 48 h. For both values of φ_CH=90 and 80% fermentations were carried out with γ_suc=50 and 100 g/L for 24 h and 48 h, respectively. For comparison, the commercial yeast strain was used under the same conditions. Cocoa honey, cocoa pulp and the volume fractions of both (without the addition of sucrose) were analyzed by chromatography for sugar, organic acid and alcohol concentrations. Due to batch limitation of cocoa honey and cocoa pulp, it was only possible to perform one replicate in the analysis.

The total soluble solids were determined with a digital refractometer (model RHB90; Akso Produtos Eletrônicos Ltda, São Leopoldo, Rio Grande do Sul, Brazil) on a °Brix scale and expressed in g/100 g, and pH was determined with a pH meter (model PHS3BW; Bel Engineering®, Rio de Janeiro, Rio de Janeiro, Brazil). The ethanol volume fraction (%) was determined with a refractometer (RETK-75; Tekcoplus Ltd., Hong Kong, PR China). To assess the clarification of the fermented must, the absorbance (A) was read at 600 nm in a spectrophotometer (SP 2000 UV; Bel Engineering®).

Mass concentrations (g/L) of sucrose, glucose, fructose, ethanol, methanol, glycerol and citric, lactic and acetic acids were determined by high-performance liquid chromatography (HPLC, Hitachi Primaide; Tokyo, Japan) with an ion exchange column (Aminex HPX-87H, 300 mm×7.8 mm; Bio-Rad, Hercules, CA, USA) and an infrared detector (K-9800; Knauer, Berlin, Germany). The samples were diluted 20 times in ultrapure water and filtered in a polyvinylidene difluoride (PVDF) membrane filter (0.45 μm) with an injection volume of 20 µL. For the analytical run, an isocratic system was used with the mobile phase of H₂SO₄ (0.01 M) and a flow rate of 0.6 mL/min at room temperature (approx. 26 °C) and a 20-minute running time. The acquisition and integration of the peaks was performed using the Star Chromatography Workstation v. 6.0 software (25) based on standard curves previously prepared with chromatographic standards (Sigma-Aldrich, Merck, St. Louis, MO, USA).

Fermentations were carried out and the obtained φ_ethanol and N were considered the main responses to be analyzed, while pH, total soluble solids (TSS) and A were analyzed as complementary responses at the beginning (0 h) and at the end of fermentation (72 h). From the results presented in Table 1, it is possible to observe that the highest obtained φ_ethanol corresponded to run 2 (72 h) and run 6 (48 and 72 h). However, it is also important to observe the φ_ethanol value of run 9, which was obtained in a shorter time (24 h) and, consequently, represents a higher productivity. These alcohol volume fractions are in accordance with the range stipulated for fruit wines in Brazil: 4–14% (28). For the response N, it is possible to observe that regardless of N_o, its value did not exceed the factor of 10⁸ (Table 1), which was also confirmed by Dias et al. (14). Thus, a range of N_o=10⁷–10⁸ cell/mL is suggested considering possible difficulties in larger scales for the preparation of inoculum solution and cell count. A similar inoculum (10⁸ cell/mL) was used for fermentation of cocoa pulp by Duarte et al. (16). In relation to the pH values, small variations were once again observed during fermentation (Table 1). The clarification of the must can also be confirmed due to the reduction of the total soluble solid content and A throughout fermentation; the highest reductions (>70%) of total soluble solid content were obtained in runs 5 and 9, and for more than half of the performed runs, A was reduced in more than 95% (Table 1). The clarification associated with fermentation can be understood, for example, when related to the consumption of sugar, due to cell growth, and/or enzymatic hydrolysis of pectin.

Considering the analysis of effects (Table 2) for the response N, the only factor with a positive and statistically significant effect (p<0.15) at all times was N_o, and only at 24 h, it was φ_CH. For φ_ethanol, N_o was also statistically significant and positive for the first part of fermentation (24 and 48 h), with its highest effect in the first 24 h. The concentration of sucrose (γ_suc) also had a positive and statistically significant effect in the second half of fermentation (48 and 72 h). In general, this analysis suggests that N can be increased in the first 24-48 h of fermentation with the increase of γ_suc, N_o and φ_CH, as demonstrated in the obtained responses in Table 1. Additionally, in order to increase φ_ethanol, fermentation can be extended up to 48 h with a higher γ_suc value. Obtaining a fermented product with higher ethanol volume fraction is interesting as it can be used, for example, to obtain a distilled spirit, denominated in Brazil as aguardente de fruta (28).

Considering the best φ_ethanol values (Table 1), run 9 also gave (within 24 h) pH, TSS and A values of: 3.46, 5.4 g/100 g and 0.078, respectively, and run 6 (within 48 h): 3.62, 7.0 g/100 g and 0.088, respectively. Between these two fermentations, run 9 was carried out with a higher φ_CH and lower γ_suc than run 6. The results also indicate that increasing the fermentation time up to 72 h was not advantageous for φ_ethanol since the observed increments did not justify extending the fermentation time to more than 48 h. Runs 6 and 9 were both performed with initial TSS content around 19 g/100 g, which is close to the value used by Leite et al. (17) with cocoa honey. For fermentation of other types of fruit pulp with S. cerevisiae, values from 16 to 20 g/100 g have been reported (1, 29).

In general, the obtained results are in accordance with the results of cocoa honey and cocoa pulp fermentations carried out by other researchers. The fermentation of cocoa honey with a commercial strain (S. cerevisiae AWRI726) was also evaluated by Magalhães-Guedes et al. (21), considering three factors: temperature, total soluble solid content and fermentation time. In this example, the authors obtained the highest alcohol volume fraction (15%) at 25 °C, 13 g/100 g (with sucrose addition) and a longer time (96 h). Furthermore, an increase in temperature to 28 °C and time to 112 h was favorable for fermentation. In a complementary study, this same research group obtained an alcohol content of 16% when fermenting cocoa honey with the same strain but at 20 °C for 144 h (17).

In another example, fermentation of cocoa pulp by S. cerevisiae UFLA CA 1162 resulted in an alcohol content of about 8% and fermentation was carried out for 30 days at 22 °C followed by a period of wine maturation (16). Also, cocoa pulp fermentation by S. cerevisiae CA1183 resulted in approx. 11% ethanol and pH=3.7, after 64 h at 25 °C with an additional period of maturation (14). An alcohol volume fraction of approx. 10% and pH=3.6 were obtained when using 80% cocoa pulp (previously clarified with pectinases) complemented with Hibiscus sabdariffa L. extract fermented at 30 °C for 168 h by S. cerevisiae var. bayanus (15).

As temperature (t) was not statistically significant (p>0.15) in the analyzed range for the two responses (Table 2), the lowest value (28 °C) was chosen. However, based on the preliminary tests and other published results with S. cerevisiae (6, 7, 15, 29), the range of 28–30 °C can be suggested as well. As the increase in φ_CH up to 100% slowed down N, but had no effect on φ_ethanol, both φ_CH (90 and 80%) were chosen due to the possibility of achieving a different sensorial profile of fruit wines. Additionally, 24 h and γ_suc=50 g/L were selected as suitable for obtaining a fruit wine, while 48 h and γ_suc=100 g/L were selected for obtaining higher alcohol volume fractions (e.g. for distillation).

Fig. 1 shows the concentrations of sugar and citric acid in cocoa honey and cocoa pulp, and according to the results, cocoa honey had a higher concentration of sucrose (27.75 g/L), and cocoa pulp had higher concentrations of citric acid (γ_cit=8.17 g/L), fructose (γ_fru=64.12 g/L) and glucose (γ_glu=64.28 g/L). Thus, the two selected φ_CH for the fermentation of must can be easily understood as a reflection of their individual compositions (Fig. 1).

Fermentations conducted for 24 h (Fig. 2) and 48 h (Fig. 3) with both φ_CH were investigated to determine the concentrations of sugars, alcohols and acids (Figs. 2a, 2c, 3a and 3c) and N, TSS content, pH and A (Figs. 2b, 2d, 3b and 3d). For these fermentations, it was not possible to obtain N_o=10⁸ cell/mL as desired; however, the actual value employed (6.5·10⁷ cell/mL) was within the suggested range for inoculum cell count. In general, the obtained fermentation profiles indicated an increase in φ_ethanol and N associated with the consumption of sugars and simultaneous clarification, which is in agreement with other studies (16, 17). Table 3 also shows the compilation of the final values for fruit wines fermented with strain L63 in comparison with the commercial strain for both φ_CH (90 and 80%) and times (24 and 48 h).

In regard to the fermentations performed for 24 h with γ_suc=50 g/L, it was observed (Figs. 2a and 2c) that for both φ_CH there was a marked decrease in γ_suc in the first 12 h of fermentation and lower reductions of glucose (γ_glu) and fructose (γ_fru) after 6 h of fermentation. It is also possible to infer the preferential consumption of glucose by S. cerevisiae L63, since γ_fru remained higher than γ_glu until the end of fermentation. In the cocoa honey fermentation evaluated by Leite et al. (17), for example, total sugar concentration was reduced from 186.78 to 5.02 g/L after 240 h of fermentation with a residual fructose concentration of 4.21 g/L. In this work, the must with φ_CH=90% resulted in a total reduction of 54% of total soluble solid content, a better result than with φ_CH=80%, where total reduction was 38%. In addition, the highest reduction (91%) of A occurred in the first 3 h of fermentation with φ_CH=80%, but the lowest A value was achieved with φ_CH=90% (greatest clarification) at the end of 24 h (Figs. 2b and 2d). Considering the variability, the increase in N was once again greater at higher volume fraction of cocoa honey than of cocoa pulp. Also, the pH and citric acid (γ_cit) values varied little over time (Figs. 2b and 2d) and methanol, acetic and lactic acid were not identified in any of the fermentations.

After 24 h of fermentation, the wine obtained from φ_CH=90%(Table 3) had a higher alcohol volume fraction, while the wine obtained from φ_CH=80% resulted in a higher concentration of residual sugar (79.06 g/L), which is in accordance with the classification of sweet wine based on the Brazilian legislation (28). Regarding the fruit wine fermented with the commercial strain for 24 h, compositions were similar, especially in relation to φ_ethanol. However, higher concentrations of residual sugar (>88 g/L) were obtained (Table 3).

Similar trend was observed for the fermentations conducted for 48 h with γ_suc=100 g/L, where sucrose was almost entirely consumed around 24 h and the concentrations of γ_glu and γ_fru showed a decrease from 12–15 h (Figs. 3a and 3c). The wine obtained from φ_CH=90% also had a reduced γ_glu at the end of fermentation (Table 3). After 48 h of fermentation, the volume fractions of ethanol in both fruit wines increased with time (Figs. 3a and 3c) and the final values of φ_ethanol (Table 3) were similar. Once again, the wine obtained from φ_CH=80% had a higher concentration of residual sugar (79.83 g/L). The TSS content (Figs. 3b and 3d) increased in the first 12 h of fermentation (as a consequence of the change of sugar concentrations) followed by a constant decrease, representing 65–67% reduction until the end of fermentation. Clarification was also achieved since the A values showed reductions between 94–96% at the end of fermentation (Figs. 3b and 3d). In comparison with the fruit wine fermented for 48 h using the commercial strain CA-11 (Table 3), it is also possible to observe similar values of alcohol but much lower values of residual sugars (<11 g/L).

The citric acid (γ_cit) concentrations increased throughout both fermentations for 24 (Figs. 2a and 2c) and 48 h (Fig. 3a and 3c). During the former, the increments were between 15–25%, and during the latter, larger increments (87–94%) were observed. In general, lactic acid, acetic acid and methanol were not detected, except for the small concentrations of acetic acid in the wine obtained from φ_CH=80% with the strain L63, and lactic acid in the wine obtained from φ_CH=90% with the strain CA-11 and after 48 h of fermentation (Table 3). The organic acids naturally found in fruit wines are fundamental to the unique sensory quality of each beverage (16). These acids can come from the used fruit pulp or they can be produced during fermentation. Citric acid is an example that fits these two situations and its presence is generally valued for sensory aspects. However, the presence of lactic and/or acetic acid, for example, can indicate contamination of the fermentation process and is a good indicator of quality (30) or it can be detected as a result of fermentation. For example, the cocoa pulp wine obtained by Dias et al. (14) had 5.5 g/L citric acid, 1.1 g/L acetic acid, but it did not contain lactic acid. Longer fermentations of fruit pulp generally require the addition of sulfur dioxide (up to 100–200 mg/L of free SO₂) in the form of dipotassium disulfite to inhibit bacterial growth (14, 16, 17). This was not applied in this study, but it is also recommended for scaled up fermentations.

In fruit wine production, the presence of glycerol also contributes to sensory characteristics such as sweetness and fullness, as mentioned by Dias et al. (14), who obtained glycerol concentrations of 5.53–9.0 g/L in a cocoa pulp wine. In this study, the glycerol concentrations increased with time (Figs. 2a, 2c, 3a and 3c) and the final concentrations obtained for the wines fermented for 48 h (Table 3) are within the range declared by Gamella et al. (31) for dry wines (4–10 g/L); the wines fermented for 24 h had lower concentrations of glycerol (Table 3). Interestingly, different Saccharomyces strains have been investigated in order to obtain wines with low-ethanol content, higher smoothness and a greater capacity to produce glycerol (32).