Species diversity, community dynamics, and metabolite kinetics of the microbiota associated with

Transcription

1 AEM Accepts, published online ahead of print on 16 September 2011 Appl. Environ. Microbiol. doi: /aem Copyright 2011, American Society for Microbiology and/or the Listed Authors/Institutions. All Rights Reserved. 1 2 Species diversity, community dynamics, and metabolite kinetics of the microbiota associated with traditional Ecuadorian spontaneous cocoa bean fermentations Zoi Papalexandratou 1, Gwen Falony 1, Edwina Romanens 1, Juan Carlos Jimenez 2, Freddy Amores 2, Heide-Marie Daniel 3, and Luc De Vuyst 1* Research Group of Industrial Microbiology and Food Biotechnology (IMDO), Faculty of Sciences and Bio-engineering Sciences, Vrije Universiteit Brussel, Pleinlaan 2, B-1050 Brussels, Belgium 1 ; Estación Experimental Tropical Pichilingue, Instituto Nacional Autónomo de Investigaciones Agropecuarias (INIAP), Cassilla Postal 24, Quevedo, Los Ríos, Ecuador 2 ; Mycothèque de l Université catholique de Louvain (MUCL), Belgian Coordinated Collection of Microorganisms (BCCM), Earth and Life Institute, Applied Microbiology, Mycology, Université catholique de Louvain, Croix du Sud 3, bte 6, B-1348 Louvain-la-Neuve, Belgium 3 ; 1 is Partner of the Flanders Research Consortium on Fermented Foods and Beverages Running Title: Cocoa bean fermentations in Ecuador * Correspondent footnote Mailing address: Research Group of Industrial Microbiology and Food Biotechnology (IMDO), Vrije Universiteit Brussel (VUB), Pleinlaan 2, B-1050 Brussels, Belgium. Phone: Fax:

2 ABSTRACT Traditional fermentations of the local Ecuadorian cocoa type Nacional with its fine flavor are carried out in boxes and on platforms during short time. A multiphasic approach, encompassing culturedependent and -independent microbiological analysis of fermenting cocoa pulp-bean samples, metabolite target analyses of both cocoa pulp and beans, and sensory analysis of chocolates produced from the concomitant fermented dry beans was applied for the investigation of the influence of these fermentation practices on the yeast and bacterial species diversity and community dynamics during cocoa bean fermentation. A wide microbial species diversity was found during the first three days of all fermentations carried out. The prevailing ethanol-producing yeast species were Pichia kudriavzevii and Pichia manshurica, followed by Saccharomyces cerevisiae. Leuconostoc pseudomesenteroides (glucoseand fructose-fermenting), Fructobacillus tropaeoli-like (fructose-fermenting), and Lactobacillus fermentum (citrate-fermenting, mannitol-producing) represented the main lactic acid bacteria species of the fermentations studied, resulting in intensive heterolactate metabolism of the pulp substrates. Tatumella saanichensis and Tatumella punctata were among the Enterobacteriaceae present during the initial phase of the cocoa bean fermentations and could be responsible for the production of gluconic acid in some cases. Also, a potential new yeast species was isolated, namely Candida sorbosivoranslike. Acetic acid bacteria, with as main representative Acetobacter pasteurianus, generally appeared later during fermentation and oxidized ethanol to acetic acid. However, acetic acid bacteria were not always present during the main course of the platform fermentations. All data taken together indicated that short 2

3 46 47 box and platform fermentation methods caused incomplete fermentation, which had a serious impact on the quality of the fermented dry cocoa beans Keywords: cocoa bean fermentation, Ecuador, platform, box, diversity, (GTG) 5 -PCR, PCR-DGGE, chocolate, sensory analysis. INTRODUCTION Today s consumers increasingly require high-quality standard cocoa and chocolate products, singleorigin and other premium chocolates, and chocolates with health-promoting properties (4, 15). Whereas cocoa processing and chocolate manufacturing benefit from the most recent technologies, the quality of the bulk and fine cocoa raw materials for chocolate making can still be improved through better controlled post-harvest processing, such as fermentation and drying (7-9, 22, 32, 38). After opening of the cocoa pods, the fruits of the cocoa tree (Theobroma cacao L.), the collected cocoa pulp-bean mass undergoes a spontaneous four- to six-day fermentation process during which the pulp substrates are broken down through microbial action; in parallel, biochemical changes inside the cocoa beans are initiated, which contribute to a reduction of their bitterness and astringency and an improvement of their color and flavor (15, 45). Fermentation is carried out by yeasts, Enterobacteriaceae, lactic acid bacteria (LAB), and acetic acid bacteria (AAB) (1, 7, 15, 22, 37). The cocoa bean flavor is primarily governed by the genetic constitution of the cocoa tree variety. This genetically inherent flavor potential is released and developed through both agricultural practices and the fermentation process. Whereas it is impossible to improve genetically inferior raw materials by superior processing, a careless or inadequate fermentation and/or drying process could irretrievably damage 3

4 good-quality fermented and/or dry cocoa beans (4, 47). Therefore, it is worth investigating the effect of traditional fermentation processes making use of superior raw material with up-to-date methodologies. Being only the seventh largest producer of cocoa worldwide (with a contribution of 3%), Ecuador is the largest producer of fine cocoa, accounting for more than half of the world production (4, 26). In general, Ecuador produces two types of cocoa, namely the complex Nacional x Trinitario (fine cocoa, Arriba flavor, high-yielding) and CCN-51 (bulk cocoa, hybrid of Trinitario cocoas, high-yielding, resistant to diseases) (11, 17, 26, 44). Nacional x Trinitario cocoa is grown as traditional cocoa crop (low-input systems) by small- and medium-sized farms; it roughly represents 85% of all cocoa cultivated in Ecuador. CCN-51 cocoa is cultivated by large farms too (high-input systems). Pure Nacional cocoa represents only 2-3% of the total annual Ecuadorian cocoa production, cultivated by small- and mediumsized farms solely (11). It has almost disappeared and high-yielding (resistant) hybrids between Nacional and Trinitario cocoas represent the current cocoa varieties with Arriba flavor (4), mostly cultivated on large-scale plantations (16). Unfortunately, Ecuadorian cocoa is usually poorly fermented and of inconsistent quality, which explains its low price on the world market (4). In addition, the economic role of cocoa in Ecuador is undeveloped compared to its potential and there is a need to increase the quality and productivity of the country s cocoa sector (26). Cocoa bean fermentation practices vary considerably according to region and/or country and, in many instances, even adjacent farms in the same region of a country may adopt different curing (fermentation and drying) methods (15). Approximately half of the world s cocoa crop is fermented in wooden boxes of several sizes and the remaining half is fermented in heaps or by using other primitive methods (47). A traditional technique applied in Ecuador for the production of Arriba cocoa is fermentation on drying platforms. According to this procedure, the beans are customarily spread out on the platforms during the day and piled into heaps during the night to get rid of the surrounding pulp, to aid their drying, and to 4

5 easily handle and transport them (33). Hence, fermentations are short (96 h instead of the usual 144 h). However, this fermentation technique is disappearing and only carried out for small quantities of beans on family-hold farms nowadays. The box fermentation is generally practiced on large farms, as it is more suitable for large quantities of cocoa beans (33). Recently, the microbial species diversity of Ghanaian and Brazilian cocoa bean fermentations have been studied in detail through a combination of traditional microbiological techniques (enumeration of cultivable microorganisms and subsequent isolation of colonies for species identification) and molecular PCR-based methods for both community dynamics and identifications, in particular to overcome biases and limitations of any of these methods (7-9, 22, 28, 37). Additional data have become available on cocoa bean fermentation processes performed in Indonesia, the Dominican Republic, and Nigeria, albeit partially carried out multiphasically (3, 30, 31). Despite the reputation of Ecuadorian cocoa regarding its fine flavor, little is known about the microbial communities active in Ecuadorian cocoa bean fermentations. Most of the latest studies on Ecuadorian cocoa have focused on agricultural practices (5, 43), economical issues (11, 24, 26), and flavor characterization (16, 36, 50). The present study aimed at the characterization of the microbiota involved in traditional cocoa bean fermentations carried out in Ecuador to estimate its impact on fine cocoa flavor development by comparing box and platform cocoa bean fermentation processes. Therefore, a multiphasic approach was applied, including culture-dependent and -independent microbiological analysis of fermenting cocoa pulp-bean samples, metabolite target analyses of both cocoa pulp and beans, and sensory analysis of chocolates produced from the concomitant fermented dry cocoa beans. MATERIALS AND METHODS 5

6 Field experiments. Four spontaneous cocoa bean fermentations were performed in a traditional manner in Ecuador during the main crop of 2008 (April 2008). During a first fermentation round, cocoa pods from hybrid trees (Nacional x Trinitario) were harvested at a farm situated in Mocache (Los Rios Province). They were manually opened one day after harvesting. A cocoa bean box fermentation was carried out in a fermentary room of the Estación Experimental Tropical (EET) Pichilingue of INIAP, while a platform fermentation was performed at the farm in open air (15 km south of EET Pichilingue), at the same time using the same batch of cocoa pods. The second fermentation round consisted of both fermentations two weeks later with cocoa beans from cocoa pods harvested from a second farm (Valencia, 10 km north of EET Pichilingue). The platform fermentations (P1 and P2) were performed on a bamboo mattress with 100 kg of fresh cocoa beans each. In the case of P1, the platform mattress was not used for a long time, in contrast with the P2 bamboo mattress that was used frequently. To initiate the fermentations, a flat heap was formed, covered with plantain leaves, and protected with a plastic cover from unfavorable weather conditions. During P1, the whole fermenting cocoa pulp-bean mass was spread out on a concrete floor after 50 h and 70 h of fermentation for 3.0 h and 3.5 h of sun-drying, respectively, before further fermentation in a flat heap. The spreading of the fermenting cocoa pulp-bean mass of P2 on the bamboo mattress took place after 54 h and 72 h for 3.0 h (interrupted because of heavy rain) and 4.0 h (because of less sunshine), respectively. After a fermentation period of four days (96 h), the cocoa beans were transported to the EET Pichilingue of INIAP and spread out for final sun-drying on a wooden tray for 5-7 days. For each box fermentation (B1 and B2), 100 kg of fresh cocoa beans were placed in a wooden box of 60 cm 60 cm 60 cm. After 24 h and 72 h of fermentation, the fermenting cocoa pulp-bean mass was transferred to a second and third identical box, respectively, for additional mixing of the cocoa pulp-bean 6

7 mass. After another fermentation period of four days (96 h), the cocoa beans were spread on a wooden tray for sun-drying. Because of the presence of high amounts of pulp around the cocoa beans at the end of the P2 and box fermentations, the fermenting cocoa pulp-bean mass was reassembled into a heap and the fermentation process was continued for 24 h, before the beans were finally spread onto the wooden tray for 7-10 days of actual sun-drying. During all cocoa bean fermentations, temperature (of the environment and inside the fermenting cocoa pulp-bean mass) and ph (of the fermenting cocoa pulp-bean mass and inside the beans) were measured. Temperature and ph were measured on-line in the middle of the fermenting cocoa pulp-bean mass with a digital ph 340i sensor (WTW GmbH, Weilheim, Germany). Additional temperature sensors were inserted at the surface of the fermenting cocoa pulp-bean mass for the platform fermentations (5 cm depth) and at the bottom and in the middle of the fermenting cocoa pulp-bean mass for the box fermentations (50 cm depth; close to the box wall). Environmental temperatures were measured with temperature recorders (DeltaTRAK, Pleasanton, CA). To measure the ph inside the beans, 10 g of fermented cocoa beans were collected every 24 h. The beans were peeled manually with a blade and grounded in a mill in the presence of 100 ml of ultrapure water for approximately 1 min. The ph of this aqueous solution was measured using a ph-meter (PB-11; Sartorius AG, Göttingen, Germany). Rainfall data were recorded by the meteorological station of EET Pichilingue of INIAP. Samples of 500 g were collected according to a fixed time schedule, namely at the start of the fermentation (time 0, fresh cocoa beans) and after 6, 12, 18, 24, 30, 36, 42, 48, 54, 60, 66, 72, 84, and 96 h of fermentation. Additional samples were taken immediately after transfer and mixing during the box fermentations and after 120 h in the case of prolonged fermentations. Sampling was always done at the same depth of the fermenting cocoa pulp-bean mass (approximately 30 cm from the upper surface) but in 7

8 different points of the fermenting cocoa pulp-bean mass. All samples were stored at -20 o C; an aliquot was used for immediate culture-dependent microbiological analysis. Swab samples corresponding to a surface of 25 cm 2 were taken from the environment (pod surfaces; leaves; workers hands; machetes; platforms and fermentation boxes; sacks; plastic covers; and buckets). At the occasion of the second fermentation round, drainage samples were collected during the first 54 h and 36 h of the platform and box fermentations, respectively. Culture-dependent microbiological analysis. The fermentation and swab samples were immediately transported to the laboratory for plating on malt extract agar (MEA, yeasts; Oxoid, Basingstoke, Hamphsire, UK), plate count agar (PCA, total aerobic bacteria; Oxoid), de Man-Rogosa-Sharpe agar (MRS, LAB; Oxoid), deoxycholate mannitol sorbitol agar (DMS, AAB), and acetic acid medium agar (AAM, AAB), as described before (8), except that MRS, DMS, and AAM agar media were supplemented with 200 mg of pimaricin (Sigma-Aldrich, Steinheim, Germany) per liter of medium. All media were incubated at 37 o C for 2-4 days. For a statistically relevant investigation of the LAB and AAB species diversity, 35% of the colonies were picked up from MRS, DMS, and AAM agar media of an appropriate dilution. They were transferred into tubes containing 10 ml of the respective liquid medium for subsequent growth under the same conditions as for the isolation. In the case of slow growth, the tubes were further incubated at 28 o C. This cultivation method resulted in a total of 1,914 bacterial isolates: 776 isolates from MRS agar, 573 from DMS agar, and 565 from AAM agar. MRS, DMS, and AAM isolates were checked for purity as described before (9). All bacterial isolates were stored at -80 C in cryovials in the appropriate medium supplemented with 25% (vol/vol) of glycerol as a cryoprotectant. In addition, colonies were washed off from MRS, DMS, AAM, and PCA agar media of a low sample dilution with 10 ml of saline (0.85%, wt/vol, NaCl); colony washes were stored at -20 C. Also, for a preliminary investigation of the dominant yeast species involved in the fermentations, 150 8

9 yeast colonies were randomly picked up from MEA plates that contained 10 to 100 colonies and streaked on the same medium for transport to Belgium (incubation at 30 C for 1-2 days). Further, they were tested for purity and morphology by streaking on yeast-glucose (YG) agar (5 g l -1 of yeast extract, 20 g l - 1 of glucose) and incubation at 30 C for 1-2 days. The pure isolates were grown in YG medium under the same conditions and subsequently stored on slants of the same medium (+4 C) as well as in cryovials supplemented with 10% (vol/vol) of glycerol (-80 C). For MRS, DMS, and AAM isolates, total genomic DNA was extracted from the corresponding cultures and subjected to (GTG) 5 -PCR fingerprinting, following the protocol described previously (7, 39). High-molecular-mass DNA was extracted from yeast cultures in YG medium and subjected to M13- PCR fingerprinting (12). Numerical cluster analysis of PCR fingerprints was performed with BioNumerics version 5.0 software (Applied Maths, Sint-Martens-Latem, Belgium) for bacteria and BioloMICS version software (Bioware SA, Hannut, Belgium) for yeasts. Dendrograms were obtained by means of the unweighted pair group method with arithmetic averages (UPGMA) clustering algorithm, with correlation levels expressed as percentage values of the Pearson correlation coefficient or the CloseSym algorithm, respectively. The identity of the bacterial isolates was confirmed by partial sequencing of the 16S rrna gene of representative isolates of each (GTG) 5 -PCR cluster. Representative yeast isolates of all M13-PCR clusters were subjected to sequencing of the D1/D2 region of the rrna gene of the large ribosomal subunit (26S), and if needed also to ITS and/or ACT1 gene sequencing (12). The BLAST (Basic Local Alignment Search Tool) program ( was used to search in the GenBank database for the closest known relatives of the gene sequences obtained. Culture-independent microbiological analysis through denaturing gradient gel electrophoresis (DGGE). (i) Fermentation samples. Direct extraction of DNA from the fermentation samples was performed as described previously (7) with minor modifications (32, 38). The LAC1-LAC2 (48) and 9

10 universal 357f-518r (21) primers, targeting the V3-V4 and V3 regions of the 16S rrna gene, respectively, were used to amplify DNA from LAB and bacterial species in general, respectively, through PCR with 1 μl of DNA (50 ng) (22). Also, a pair of primers (WBAC1-WBAC2; 35) that targets the V7-V8 region of the 16S rrna gene was applied to monitor AAB and LAB simultaneously, following the optimized protocol of Garcia-Armisen et al. (22). A GC-clamp was attached to the LAC2, 357f, and WBAC2 primers to prevent early separation of double stranded DNA during DGGE, for which a denaturing gradient from 35-60%, 35-70%, and 45-70%, respectively, was used. DGGE bands of interest were identified through sequencing and BLAST analysis, as described previously (22). (ii) Plate washes. To analyze DNA from bulk cells washed off from MRS, DMS, AAM, and PCA agar media, DNA was extracted as described above for cultures of single isolates, and PCR amplicons obtained with the LAC1-LAC2 primers (bulk cells from MRS agar), WBAC1-WBAC2 primers (bulk cells from DMS and AAM agar), and a pair of universal prokaryotic primers (bulk cells from PCA) were subjected to DGGE, as described above for the fermentation samples. Metabolite target analysis. For the metabolite target analysis, aqueous extracts were prepared from fermentation samples, from both pulp and beans, as well as from fermented non-deshelled dry cocoa beans, as described previously (7). Drainage samples were used untreated. (i) High performance anion exchange chromatography. To measure the concentrations of acetic acid, citric acid, lactic acid, and gluconic acid in the aqueous extracts, high performance anion exchange chromatography with conductivity under ion suppression (HPAEC-CIS) was applied. The concentrations of glucose, fructose, sucrose, and mannitol of the aqueous extracts were determined by HPAEC with pulsed amperometric detection (PAD). Apparatus, conditions, and gradients applied for the runs were as reported before (7). To quantify metabolites, a standard addition protocol was used (32). 10

11 (ii) Gas chromatography. The concentrations of ethanol present in the aqueous extracts were measured by gas chromatography. Apparatus, process parameters, sample preparations, and quantification with external standards were as described previously (38). Quality assessment of fermented dry cocoa beans, chocolate production, and sensory analysis. A cut test was performed on the fermented dry cocoa beans by EET Pichilingue of INIAP to estimate their quality (9). Also, the appearance of the dry cocoa beans (weight of 100 beans, bean size, physical damage, insect penetration) was checked. In addition, their moisture content was determined (25). Thirty kilograms of fermented dry cocoa beans were used for dark chocolate production by Barry Callebaut Belgium (Wieze, Belgium). The recipe and the procedure followed for chocolate making were as described previously (9). The chocolates were subjected to a sensory analysis carried out by a trained tasting panel of eight members of Barry Callebaut Belgium. The flavor descriptors were cocoa, heavy, sweet, sour, fruity, herbaceous, flowery, intensity, and aftertaste intensity, expressed as numerical values from 0-100, and are represented as averages. The chocolates were compared with a commercial reference sample (811NV) with standard flavor descriptors to indicate flavor and taste differences. General comments and characteristics were given by the panel members. The chocolate samples were brought to a temperature of 45 o C prior to analysis and water was used for rinsing the mouth after each tasting. RESULTS Physical changes. During the first two days of P1, the environmental temperature followed the day (28-30 C) and night (21-23 C) cycles, but because of heavy rainfall (25 mm) after 48 h of fermentation the temperature dropped to ± 23 C to increase to 30 C during the last 12 h. The temperature inside the 11

12 fermenting cocoa pulp-bean mass increased from 26 C to 47 C in 36 h. The two spreadings of the cocoa pulp-bean mass after 50 h and 70 h caused a temperature drop of 9 to 12 C. The maximal fermentation temperature of 50.5 C was reached after 48 h. The ph value inside the fermenting cocoa pulp-bean mass was constant (around ph 4.00) during the initial stage of the fermentation. After 50 h, the time point of the first spreading, the ph increased and reached a maximal value of 5.00 after 76 h. The final ph was 4.57 after 96 h. The ph inside the beans decreased from 6.45 (0 h) to 5.75 (96 h). During P2, the environmental temperature was almost constant around C, because of heavy rainfall during this period. The temperature of the fermenting cocoa pulp-bean mass increased from 24 C at the beginning to 47.5 C at the end of the fermentation, being influenced by the spreadings of the fermenting cocoa pulp-bean mass (drop of 9 to 12 C). The ph inside the fermenting cocoa pulp-bean mass had an initial value of 3.40 (0 h) and it slightly increased to 3.75 (96 h). The ph inside the beans decreased from 6.86 (0 h) to 4.70 (96 h). B1 started with a temperature of 28.7 C inside the fermenting cocoa pulp-bean mass, which remained stable until the time of the first mixing at 24 h, after which it increased continuously up to 48 C after 70 h. The second mixing at 72 h resulted in a small temperature decrease. The maximal fermentation temperature of 49 C was reached after 82 h. The ph inside the fermenting cocoa pulp-bean mass had an initial value of 3.67 and increased to a value of 4.20 after 96 h. The ph inside the beans decreased from 6.50 (0 h) to 4.24 (96 h). The temperature inside the fermenting cocoa pulp-bean mass of B2 remained almost stable for the first 24 h (25-27 C), followed by an increase to 48.8 C after 73 h. The ph of the fresh pulp of B2 was 3.71 (0 h) and it reached a value of 4.01 at the end of the fermentation. The ph inside the beans decreased from 6.84 (0 h) to 4.21 (96 h). 12

13 The temperature at the surface of P1 and P2 was 1 to 4 C and 4 to 8 C lower than in the center of the fermenting cocoa pulp-bean mass during the first 24 h and between 30 and 54 h of fermentation, respectively. In the case of the B1 and B2 fermentations, the environmental temperature outside the fermentary room followed the day (38-41 C) and night (23-25 C) cycles. However, the temperature inside the fermentary room was always 3-5 C lower and higher during day and night, respectively. The temperature at the bottom and the middle of the box walls differed 0 to 6 C and 0 to 3 C, respectively, compared to the temperature in the middle of the fermenting cocoa pulp-bean mass. Community dynamics: culture-dependent microbiological analysis. (i) Cell counts. Low initial yeast counts on MEA agar [ log (CFU g -1 )] were determined during all fermentations; they reached a maximum of 7.5 log (CFU g -1 ) after 12 h in the case of P1 and P2 and after 36 to 42 h in the case of B1 and B2 (Fig. 1). In general, LAB counts (MRS agar) increased intensively during the first 30 h of all fermentations from to log (CFU g -1 ). The yeast and LAB counts remained at high levels throughout the platform fermentations but they started to decrease after 40 h and 54 h of the box fermentations, respectively (Fig. 1B,D). The initial AAB counts on DMS and AAM agar were approximately log (CFU g -1 ). During all fermentations, counts on AAM agar were higher than those on DMS agar, with the AAM counts following the same trend as the MRS counts. Maximum AAB counts obtained on DMS and AAM agar were 7.5 and 8.5 log (CFU g -1 ), respectively (Fig. 1). Spreading or mixing of the fermenting cocoa pulp-bean mass had no direct impact on the cell counts. However, reassembling the fermenting cocoa pulp-bean mass at the end of the box and P2 fermentations for a prolonged fermentation caused a direct increase of the counts on all media (Fig. 1B,C,D). Concerning the swab samples, low cell counts [< log (CFU per 25 cm 2 )] were found on the surfaces of the plant material (leaves, pods) and the washed equipment (machetes, buckets). In the case of unwashed equipment, such us fermentation boxes, platforms, and sacks, the cell counts were high [ log 13

14 (CFU per 25 cm 2 )]. High PCA counts [4.5 log (CFU per 25 cm 2 )] and low MEA, DMS, and AAM counts were found on the workers hands [< 2.0 log (CFU per 25 cm 2 )]. (ii) Identification of bacterial and yeast isolates: selection of isolates. Excluding 332 isolates ( 17%), which could not be recovered out of the 1,914 bacterial isolates that were picked up, 1,582 pure isolates were tested for catalase activity. The 163 catalase-positive MRS isolates (23.5% of all recovered MRS isolates) and 5 catalase-negative DMS isolates (1.5% of all recovered DMS isolates) were excluded from further analysis, as they did not represent potential LAB and AAB, respectively. Also, 262 (57%) of the recovered AAM isolates were catalase-negative. As randomly selected isolates of this group were identified as LAB species (mainly Lactobacillus fermentum) by (GTG) 5 -PCR fingerprinting, this group was excluded from further analysis too. In addition, a high number of the catalase-positive DMS and AAM isolates ( 50%, 340 isolates) did not represent potential AAB. Gram-staining, growth on yeast-selective YG agar, examination under the light microscope, unusual (GTG) 5 -PCR fingerprints, and lack of DNA amplification with the 16S rrna primers allowed their discrimination as yeasts. Their growth in the presence of 200 mg l -1 of pimaricin, included in the MEA or YG agars, revealed their resistance towards this antifungal agent. Further sequencing of representative isolates of this group of pimaricin-resistant yeasts allowed their identification as Pichia kudriavzevii and Saccharomyces cerevisiae (data not shown). Lactic acid bacteria. The 529 LAB (GTG) 5 -PCR fingerprints were clustered in one dendrogram to illustrate the LAB species diversity during both fermentation rounds (Fig. 2). Also, Table 1 shows the source of the LAB species found. Leuconostoc pseudomesenteroides and Lb. fermentum were the LAB species isolated from all fermentations. However, important differences were seen in the LAB species diversity of the two fermentation rounds, which was most pronounced during platform cocoa bean 14

15 fermentations at the farm site. In general, the fermentations were characterized by the presence of several LAB species during the first 60 h of the process. The largest group of LAB was formed by 261 isolates of Lb. fermentum (49.3% of all LAB isolates), which were derived from several time points during the platform and box fermentations. Leuconostoc pseudomesenteroides (99 isolates; 18.7% of all LAB isolates) was mainly isolated during the first 60 h of the platform and box fermentations of both fermentation rounds, as well as from swab samples. Fructobacillus tropaeoli-like, Lactobacillus fabifermentans, and Lactococcus lactis subsp. lactis were mostly present at the initial stages of P1, while Lactobacillus nagelii, Lactobacillus cacaonum, and Enterococcus casseliflavus were only isolated from fermentation and swab samples of P2, respectively. Fructobacillus tropaeoli-like was also isolated from several samples of B1. A previously uncultured bacterium/lactobacillus sp. together with Lactobacillus amylovorus occurred after 84 to 96 h of the B1 fermentation. Lactobacillus farraginis was only obtained from swab samples of B2. Lactobacillus plantarum was accidentally picked up from both platform fermentations. The remaining isolates of the first fermentation round were identified as Enterococcus saccharolyticus, Fructobacillus ficulneus, Lactococcus garvieae, Leuconostoc mesenteroides, and Weissella cibaria, all from P1; Lactobacillus coryniformis from B1; and Streptococcus salivarius from both fermentations. Leuconostoc pseudomesenteroides and Leuconostoc fallax were isolated after the first spreading of the cocoa pulp-bean mass of P2 and Weissella fabaria was found in the initial samples of the platform and B2 fermentations. Acetic acid bacteria. The 334 AAB (GTG) 5 -PCR fingerprints were clustered in one dendrogram to illustrate the AAB species diversity during both fermentation rounds (Fig. 3). In Table 1, the origin of each AAB species is reported too. The largest group of AAB was formed by 183 isolates of Acetobacter pasteurianus (54.8% of all AAB isolates), mainly derived from the second fermentation round. In 15

16 general, the platform fermentations showed a wide AAB species diversity during the first hours, including Acetobacter fabarum, Acetobacter peroxydans, Acetobacter cibinongensis, and Acetobacter malorum/indonesiensis [(98% similarity in the 16S rrna gene with that of A. malorum ATi6 (JF718421) and A. indonesiensis GY1-1 (HQ711341)] from P1, and Acetobacter syzygii, Acetobacter malorum/cerevisiae [98% similarity in the 16S rrna gene with that of A. malorum GY4-9 (HQ711343) and A. cerevisiae Ac3 (JF718424)], Acetobacter orientalis, and Acetobacter lovaniensis/fabarum from P2. The low 16S rrna gene similarity is worth further investigation. The species Acetobacter pomorum was exclusively isolated from swab samples collected from farm 1. Despite the wide initial AAB species diversity, no AAB were isolated during the main stage of P1; only single isolates of Acetobacter ghanensis and A. peroxydans were obtained, which were probably opportunistic contaminants. However, the prevalent species collected from the AAM and DMS agars of this fermentation were the yeast isolates mentioned above. The image of the AAB species diversity of P2 was different, representing mainly A.pasteurianus as well as Acetobacter senegalensis and A. ghanensis that were collected from several samples throughout this fermentation. The cocoa bean box fermentations carried out in the fermentary room showed the same AAB species diversity. Gluconobacter species and A. pasteurianus were the AAB species present at the beginning and during the main stage of these box fermentations, respectively. Also, the AAB-like species Frateuria aurantia was among the AAB species of the first stage of both box fermentations and was isolated only from DMS agar. Finally, two isolates from the last samples of P2 were identified as Microbacterium lacticum. Less dominant species, such as Gluconobacter spp., A. cibinongensis, and A. orientalis, were isolated only from AAM agar. Yeasts. Out of the 150 MEA isolates, 132 could be recovered after transport. Cluster analysis of the M13-PCR-fingerprints revealed two main clusters, which were identified through D1/D2 LSU sequencing as P. kudriavzevii (64 isolates; 48.5% of all yeasts) and Pichia manshurica (25 isolates; 16

17 % of all yeasts), respectively. They were isolated from all fermentations. The particular origin of each species is also presented in Table 1. Twenty-three S. cerevisiae isolates were found as two subclusters of 18 and 5 isolates, respectively, representing 17.4% of all yeast isolates. These isolates came from samples of B1, B2, and P2. Six isolates originating from swab samples (fermentation box and plastic cover) of the first fermentation round showed morphological and physiological similarities with Candida magnoliae and Candida sorbosivorans. Three isolates showed identical D1/D2 LSU sequences to C. sorbosivorans and were distinguished from C. magnoliae by six nucleotide substitutions. However, a high number of substitutions in the ITS sequence of these six isolates in comparison with the type strain of C. sorbosivorans (26 to 38 substitutions, up to 4 insertions/deletions) calls for a further taxonomic study of these strains. ACT1 gene sequences were highly similar among the six strains, varying only in two substitutions, but were not available for the relevant type strains. Six isolates from the end of the fermentations could be identified as Kluyveromyces marxianus after D1/D2 LSU and ACT1 gene sequencing. Other species found in low percentages were Hanseniaspora opuntiae (swabs/beginning of fermentations), Candida tropicalis (swabs/end of fermentations), Pichia kluyveri (swabs), Rhodotorula minuta (swabs), and Torulaspora delbrueckii (beginning of fermentations). Community dynamics: culture-independent approach. (i) PCR-DGGE with LAC primers. The application of PCR-DGGE with the LAC primers allowed the verification of the identification results of the culture-dependent approach as a function of time (Fig. 4). In general, the initial stage of the platform cocoa bean fermentations was characterized by a wide LAB species diversity. The species detected during the first 48 h, however, differed for P1 and P2, including Lb. plantarum, Lb. fermentum, Lc. pseudomesenteroides, and Fr. tropaeoli-like (P1), and Lb. cacaonum, Lb. fermentum, Lc. pseudomesenteroides, and Lb. nagelii (P2). Spreading of the beans had a positive impact on the stability of the LAB species diversity of the platform fermentations. Moreover, the Lb. fermentum band was only 17

18 visible after 48 h of the platform fermentations, with a Leuconostoc sp. band of very low intensity becoming visible in parallel during the last two days of P1. The LAB pattern of the box fermentations carried out at the EET Pichilingue of INIAP was different, with even less similarity between these two fermentations. For B1, an intense Weissella beninensis band was visible during the first 18 h of fermentation, which was replaced by three bands that were dominant for the main course of the fermentation, representing the species Lb. fermentum, Lc. pseudomesenteroides, and Fr. tropaeoli-like, respectively. The LAB species diversity of B2 was restricted to Lb. fermentum and Lc. pseudomesenteroides. Mixing of the fermenting cocoa pulp-bean mass had no influence on the LAB species diversity. Reassembling the fermenting cocoa pulp-bean mass on the drying platform allowed the growth of uncultured bacteria, as detected by PCR-DGGE, during this short extended fermentation period. (ii) PCR-DGGE with universal prokaryotic primers. PCR-DGGE with universal prokaryotic primers (Fig. 5) did not give extra information concerning the LAB species diversity. However, it allowed the detection of enterobacterial species at the start of all fermentations. Bands of Tatumella species, such as Tatumella saanichensis, Tatumella punctata, Tatumella terrea, and Tatumella morbirosei, were present during the first 36 to 42 h. Preliminary experiments during this study showed that T. punctata LMG T was able to assimilate citrate and produce gluconic acid in Luria Bertani medium (10 g l -1 tryptone, 5 g l -1 yeast extract, 10 g l -1 NaCl) supplemented with 25 mm citric acid or with 50 mm glucose and 25 mm citric acid, under aerobic conditions, at ph 7.5 and 30 o C. Although the detection of AAB species through PCR-DGGE is not always easy, a band with 100% similarity toward several AAB species (based on DNA sequencing results) could be detected. This band was of strong intensity in the case of the second fermentation round and was visible from 42 and 48 h of the B2 and P2 fermentations, respectively. In the case of B1, the AAB band appeared after 84 h of fermentation. In P1, 18

19 an AAB band appeared only after 12 h of the spreading but was absent from the subsequent sample. The universal prokaryotic primers caused amplification of chloroplast DNA present in the initial samples of all fermentations. (iii) PCR-DGGE with WBAC primers. The identification of AAB at species level was possible with the use of WBAC primers. Concerning LAB, a similar species diversity as with the LAC primers could be detected, although less information could be obtained on the prevailing species. In the case of P1, only LAB species were detected. Bands belonging to Lb. fermentum appeared at 30 and 60 h into fermentation, whereas Lc. pseudomesenteroides could only be detected after 12 to 30 h of fermentation. Also, Lb. plantarum was present from 12 to 30 h of fermentation of and at 48 h into the P1 fermentation process. Bands belonging to A. ghanensis or A. syzygii were detected at 6 and 66 h into fermentation of B1. In the sample collected after 120 h, A. pasteurianus displayed a strong band, while F. aurantia was detected at 18 h into fermentation of B1. In the case of P2, bands of Gluconobacter sp., A. lovaniensis/fabarum, A. pasteurianus/pomorum, Lb. cacaonum, Lc. pseudomesenteroides as well as F. aurantia were visible during the first 24 h, whereas Lb. fermentum and A. pasteurianus bands were mostly visible between 60 and 120 h of fermentation. Frateuria aurantia and Lc. pseudomesenteroides bands were visible during the first 24 h of B2, the latter remaining present up to 60 h of fermentation together with Lb. fermentum. Acetobacter pasteurianus/pomorum formed a strong DGGE band at the last sampling point. (iv) PCR-DGGE of bulk cells. The PCR-DGGE results of bulk cells harvested from MRS, DMS, AAM, and PCA agars were in accordance with the (GTG) 5 -PCR classification and identification of the single isolates as well as with the results obtained through PCR-DGGE of fermentation samples. Eventually, Lb. plantarum could be detected by PCR-DGGE (LAC primers) in MRS plate washes of B1, which was not the case by PCR-DGGE of DNA from the fermenting cocoa pulp-bean samples. The 19

20 bands corresponding with Enterobacteriaceae (Klebsiella sp., Providencia stuartii, Enterobacter sp.) and species of the genera Bacillus, Stenotrophomonas, Xanthomonas, Acinetobacter, Rhizobium, and Aeromonas were mostly present in the case of bulk cells from PCA agar of all fermentations, with a wider species diversity for P1 and P2 and a decreasing diversity from 6 h to 54 h of all fermentations (PCR-DGGE with universal prokaryotic primers). Bands of the prevailing LAB could be found only in the gels of the second fermentation round. These results confirmed the non-selective isolation of AAB and their absence from P1. No bands of A. pasteurianus/pomorum were present among the bulk cells of DMS and AAM agars of low dilutions during the main course of this fermentation (PCR-DGGE with WBAC primers). Exceptionally, bands of A. lovaniensis/fabarum and A. ghanensis/syzygii were present at the beginning of P1. In contrast, Acetobacter sp. and A. pasteurianus/pomorum were present among the bulk cells of DMS and AAM agars of the second fermentation round (Fig. 6). In addition, LAB bands could be detected mainly for bulk cells from AAM agar of low dilutions, underlining the inappropriate character of this medium for AAB isolation (Fig. 6). Metabolite target analysis. (i) Fermentation samples. Glucose and fructose, which were in higher concentrations in the fresh pulp of the first fermentation round (Fig. 7A,C), were consumed simultaneously within 50 h and 36 h of the platform and box fermentations, respectively. High amounts of pulp mannitol (29.5 ± 6.4 mg g -1 to 35.1 ± 8.0 mg g -1 ) were produced during the first fermentation round, mainly during P1 (Fig. 7A). No sucrose was found in the pulp. The bean sucrose concentration varied between 9.4 ± 2.0 mg g -1 to 19.0 ± 4.4 mg g -1 in the fresh beans of all fermentations. The mannitol concentrations inside the beans were low ( 1.5 mg g -1 ). The initial citric acid concentration was higher in the fresh pulp of the cocoa harvested from the second farm ( 10 mg g -1 ) (Fig. 7F,H). Bean citric acid concentrations ( mg g -1 ) remained stable during all fermentations. High concentrations of ethanol ( mg g -1 ) were produced in the pulp and reached a maximum after 36 to 48 h (Fig. 7B,D,F,H). 20

21 P1 showed an unexpected profile with little acetic acid production (Fig. 7B). During the first 96 h of fermentation, acetic acid was not oxidized in any case (except P2 after 84 h), indicating incomplete fermentations. Instead, gluconic acid was produced in the fermenting pulp with maximum concentrations of 11.5 ± 1.0 mg g -1 to 21.3 ± 2.1 mg g -1 and 8.9 ± 1.3 mg g -1 to 14.8 ± 3.4 mg g -1 during the platform and box fermentations, respectively. Lactic acid was produced in the pulp up to concentrations of 5.0 ± 0.4 mg g -1 to 6.1 ± 0.3 mg g -1 and 1.7 ± 0.3 mg g -1 to 4.5 ± 0.5 mg g -1 during the first and second fermentation rounds, respectively. Spreading of the fermenting cocoa pulp-bean mass during the platform fermentations impacted the course of the ethanol and gluconic acid concentrations. Mixing of the fermenting cocoa pulp-bean mass of the box fermentations only influenced the course of ethanol. Reassembling the fermenting cocoa pulp-bean mass into a heap on the drying platform caused an increase of the gluconic acid concentration and a decrease of the acetic acid concentration. (ii) Drainage samples. Concerning carbohydrates, glucose and fructose concentrations in the drainage of the first 12 h of P2 were approximately 70.0 ± 2.3 mg ml -1 to 80.0 ± 13.4 mg ml -1 each, whereas those concentrations were 19.9 ± 9.0 mg ml -1 to 51.5 ± 3.8 mg ml -1 in the drainage of B2. Sucrose was absent and mannitol was hardly found in the drainage samples. Later into fermentation, drainage samples contained 0.3 ± 0.5 mg ml -1 to 5.5 ± 3.2 mg ml -1 of carbohydrates. The citric acid concentrations of the drainage samples of both fermentations were 8.0 ± 1.3 mg ml -1 to 19.3 ± 6.1 mg ml -1 until 12 h of drainage and 0.4 ± 0.3 mg ml -1 to 4.5 ± 0.1 mg ml -1 afterwards. Lactic acid was found in low concentrations (0.5 ± 0.1 mg ml -1 to 2.6 ± 0.1 mg ml -1 ) after 12 h of B2 drainage. In the drainage samples of this fermentation, acetic acid occurred in a concentration of 6.5 ± 0.1 mg ml -1 after 36 h, whereas it reached a concentration of 3.0 ± 0.2 mg ml -1 after 54 h of P2. High amounts of gluconic acid (4.0 ± 0.8 mg ml -1 to 40.0 ± 5.6 mg ml -1 ) were present in the drainage too. 21

22 (iii) Fermented dry cocoa beans. Sucrose was still found in the fermented dry beans of all fermentations in concentrations of mg g -1, indicating incomplete fermentation (Fig. 8). Glucose and fructose were present in lower and higher concentrations, respectively, in the fermented dry beans of the second fermentation round compared to the first fermentation round. Citric acid and acetic acid concentrations were approximately the same in all cases ( mg g -1 and 6.0 mg g -1, respectively), except for the acetic acid concentration of fermented dry beans of B1 (13.3 ± 3.1 mg g -1 ). Gluconic acid was found in the fermented dry beans of the platform fermentations. Quality assessment of fermented dry cocoa beans and sensory analysis of chocolates. Sun-drying of the fermented cocoa beans resulted in a drop of the moisture level to the acceptable level of % (Table 2). However, high percentages of violet and semi-fermented beans were encountered, indicating an incomplete fermentation process (Table 2). All chocolates were characterized as heavy and sour with high intensity (Table 3). Comparison of the chocolates of the present study with the commercial reference chocolate 811NV of Barry Callebaut reflected high scores of fruitiness and floweriness for the first. Panelists comments confirmed the image of incomplete fermentations (Table 3). DISCUSSION The impact of the cocoa bean fermentation process on the development of flavor precursors in the cocoa beans as a result of microbial and biochemical activities in pulp and beans, respectively, has already been shown (2, 8-9; 18, 42). However, also the cocoa cultivar may be responsible for the flavor capacity of the beans (1, 6, 15). The present study dealt with traditional Ecuadorian cocoa bean fermentations and compared platform and box fermentation methods, making use of Nacional (fine) cocoa. 22

23 Although it has often been shown that several microbial groups are isolated from or found in cocoa bean fermentation samples (15, 45), the key players for a successful evolution of cocoa bean fermentation processes are yeasts (responsible for pectin degradation and ethanol production from glucose and/or fructose), LAB (glucose and/or fructose and citric acid co-fermentation, lactic acid production, and mannitol production), and AAB (acetic acid production and overoxidation of acetic acid and lactic acid) (8, 22). Enterobacterial activity may play a role (whether or not desirable) in pectin degradation, citric acid assimilation, and gluconic acid production. The enterobacterial species T. punctata and T. saanichensis were among the initial microbiota of all Ecuadorian fermentations carried out during the present study. Recently, the practical involvement of enterobacterial species during the anaerobic phases of Ghanaian heap, Brazilian box, and Ivorian vessel fermentations of the cocoa pulpbean mass has been suggested (22, 32, 38). The prevailing yeast species found through M13-PCR fingerprinting and sequencing in all Ecuadorian fermentations of the present study were P. kudriavzevii and P. manshurica, followed by S. cerevisiae. The species P. kudriavzevii and H. opuntiae were revealed through 26S rrna gene-pcr-dgge too (40). Ghanaian heap fermentations are dominated by S. cerevisiae and P. kudriavzevii, besides the occurrence of H. opuntiae or the closely related Hanseniaspora guilliermondii (12, 28, 37). All these yeast species reflect adaptation to the fermenting cocoa pulp-bean mass. Leuconostoc pseudomesenteroides and Lb. fermentum (P1) as well as Fr. tropaeoli-like, Lb. nagelii, and Lb. cacaonum (P2) were the prevailing LAB species found in the platform fermentations performed at the farm site and Lc. pseudomesenteroides, Fr. tropaeoli-like, and Lb. fermentum characterized the box fermentations carried out in the fermentary room. The composition of the LAB communities of cocoa bean fermentations may influence the end-products of carbohydrate fermentation, as was seen in the present study. For instance, the strong presence of Leuconostoc spp. and fructophilic LAB, such as Fr. tropaeoli-like, resulted in the production of high amounts of mannitol 23

24 from fructose. Also, the metabolism of Fructobacillus species is usually enhanced under aerobic conditions, as those established during spreading of the fermenting cocoa pulp-bean mass during platform fermentations; oxygen can then be used as alternative external electron acceptor, allowing these obligately fructophilic LAB species to ferment glucose too (19). Fructobacillus tropaeoli, recently described as a novel fructophilic LAB species, has first been isolated from the flower Tropaeolum majus (19). During the present study, a new fructose-loving bacterium, showing high identity with Fr. tropaeoli within the 16S rrna gene, was found among the prevailing LAB communities of the first fermentation round in both box and platform fermentations. The presence of citrate-fermenting LAB species, such as Lb. fermentum, at the beginning of cocoa bean fermentation is important. Assimilation of citric acid results in a ph increase, allowing the growth of less acid-tolerant LAB species, the facilitated growth of AAB, and optimal expression of several microbial activities, such as pectinolytic activity by yeasts (27). Among the organic acids formed, acetic acid produced by AAB species is most important, as its diffusion into the beans is necessary for the initiation of flavor development (15). Initial A. senegalensis communities were replaced by A. pasteurianus and represented the main AAB species of the Ecuadorian fermentations. This species is mainly responsible for ethanol oxidation into acetic acid and its further overoxidation into carbon dioxide and water. Several other LAB (Lc. pseudomesenteroides, W. fabaria, Lb. cacaonum) and AAB species (A. ghanensis, A. senegalensis), typically contaminating the cocoa pulp-bean mass at the start of cocoa bean fermentation (10, 13-14), as well as cocoa non-specific species (F. aurantia, M. lacticum) were found during the present study as well. Lactobacillus nagelii was present during P2 and it has been associated with Brazilian cocoa bean box fermentations before (22). Frateuria aurantia was isolated from the environment and during the first day of the box fermentations. Frateuria possesses the phenotypical features of the family Acetobacteraceae (α-proteobacteria), but it belongs to the γ-proteobacteria (46). It 24

25 produces acetic acid from ethanol and glucose and is able to oxidize lactate (46). It has been isolated from several flowers and fruits (29, 49, 34). Hence, it is no surprise to find it in fermenting cocoa pulpbean mass as well. This wide microbial species diversity of the Ecuadorian fermentations of the present study could be associated with unusual fermentation practices, as has also been seen for Brazilian cocoa bean box fermentations that were performed poorly (22, 41). The present study showed that traditional techniques of Ecuadorian cocoa bean fermentations, characterized by a short fermentation period (96 h instead of 144 h during a common cocoa bean fermentation process) and spreading of the fermenting cocoa pulp-bean mass during platform fermentations, caused an incomplete fermentation process and thus had a serious impact on the quality of the fermented dry cocoa beans. The sudden changes of the microenvironment inside the fermenting cocoa pulp-bean mass (evaporation of volatile compounds, escape of heat, loss of carbohydrate-rich pulp) caused by these fermentation practices did not allow a common microbial succession during the platform fermentations, resulting in the strong presence of yeasts, a wide LAB species diversity, and/or absence (P1) or too early appearance (P2) of AAB species. This is in contrast with the restricted microbial species diversity of common Ghanaian heap (7, 12) and Brazilian box (22, 41) fermentations carried out before. The unusual succession of the cocoa-specific microorganisms during the platform fermentations carried out was reflected in the metabolite profiles, with low acetic acid concentrations (P1), high mannitol concentrations (P1), and high gluconic acid concentrations (P1 and P2). The short term resulted in under-fermentation of the cocoa beans, as only low amounts of acetic acid could diffuse into the beans and, hence, the ph inside the beans was kept at high levels avoiding optimal enzymatic reactions (23). Also, the temperature of the fermenting cocoa pulp-bean mass did not increase extensively, which impacts the nature and counts of the fermenting microbiota and bean enzyme activities. Furthermore, the high production of mannitol by strictly heterofermentative LAB resulted in 25

26 the suppression of additional supply of pyruvate and its subsequent conversion into volatile flavor compounds, such as diacetyl, acetoin, and 2,3-butanediol. High gluconic acid concentrations indicated that fewer carbohydrates were available for ethanol production by yeasts and fermentation by LAB and that other microorganisms were highly active, such as Enterobacteriaceae and nonethanol-producing yeasts. Mixing of the cocoa pulp-bean mass seemed to have a minor impact on the course of the box fermentations, but it ensured its homogeneity (8). Execution of the box fermentations inside a fermentary room provided good protection from environmental contamination and influence of weather conditions, in contrast with the platform fermentations (especially P2). Interruption of the box fermentations after 96 h obliged reassembling of the fermenting cocoa pulp-bean mass into heaps on the drying platform, an unusual practice that was necessary because of the short fermentation time. This practice caused additional contamination of the under-fermented cocoa beans and, therefore, influenced their final quality, both microbiologically and sensorially. The consequences of poor operational practices on the farm during fermentation and drying on chocolate flavor have been reported before (4, 22, 38). In this case, the strong fruity and floral notes present in the fermented dry cocoa beans, obviously related to the fine cocoa genotype used as compared to the scores for the bulk reference sample, were masked by astringency, impurity, and bitterness, as a result of under-fermentation and therefore, incomplete flavor development. In conclusion, the study of traditional Ecuadorian cocoa bean platform and box fermentations allowed a further understanding of the course of these fermentation processes and unraveled, through a multiphasic approach, the influence of unusual fermentation practices on their success. In particular, the short fermentation time and interruption of the fermentation process had a negative impact on the optimal succession of microbial activities and hence on the chocolate flavor produced of the concomitant 26

27 fermented dry cocoa beans. Avoiding spreading of the cocoa pulp-bean mass during fermentation and extension of the actual fermentation process on platforms or in boxes from 96 h to h, with regular mixing every two days, followed by common sun-drying for 6-10 days, would significantly improve the quality of fermented dry cocoa beans in Ecuador. Nacional cocoa is one of the finest cocoa varieties with an inherent desirable (floral) flavor potential and, hence, more care will need to be taken during its post-harvest curing to maximally exploit this flavor and to provide fermented dry cocoa beans of excellent quality. ACKNOWLEDGEMENTS This research was funded by the Research Council of the Vrije Universiteit Brussel, the Federal Research Policy (Contract C3/10/003), the Fund for Scientific Research-Flanders, the Flemish Institute for the Encouragement of Scientific and Technological Research in the Industry, and Barry Callebaut N.V. The cooperation of the local farmers is highly appreciated. G.F. was recipient of a Ph.D. grant of the IWT. H.-M.D. thanks M.-C. Moons and P. Evrard for morphological and physiological analyses as well as S. Huret and C. Bivort for sequence analyses. The authors thank the INIAP for infrastructural and logistic support. REFERENCES 1. Afoakwa, E. O Chocolate science and technology. Wiley-Blackwell, Oxford. 2. Adeyeye, E. I., R. O. Akinyeye, I. Ogunlade, O. Olaofe, and J. O. Boluwade Effect of farm and industrial processing on the amino acid profile of cocoa beans. Food Chem. 118:

28 Ardhana, M., and G. Fleet The microbial ecology of cocoa bean fermentations in Indonesia. Int. J. Food Microbiol. 86: Beckett, S. T Industrial chocolate manufacture and use. 4 th ed., John Wiley & Sons, Ltd., Chichester. 5. Bentley, J. W., E. Boa, and J. Stonehouse Neighbor trees: shade, intercropping, and cacao in Ecuador. Hum. Ecol. 32: Cambrai, A., et al Differentiation of chocolates according to the cocoa s geographic origin using chemometrics. J. Agric. Food Chem. 58: Camu, N., et al Dynamics and biodiversity of populations of lactic acid bacteria and acetic acid bacteria involved in spontaneous heap fermentation of cocoa beans in Ghana. Appl. Environ. Microbiol. 73: Camu, N., et al. 2008a. Influence of mixing and environmental contamination on the dynamics of populations of lactic acid and acetic acid bacteria involved in spontaneous cocoa bean heap fermentation in Ghana. Appl. Environ. Microbiol. 74: Camu, N., et al. 2008b. Fermentation of cocoa beans: influence of microbial activities and polyphenol concentrations on the flavour of chocolate. J. Sci. Food Agric. 88: Cleenwerck, I., et al Acetobacter ghanensis sp nov., a novel acetic acid bacterium isolated from traditional heap fermentations of Ghanaian cocoa beans. Int. J. Syst. Evol. Microbiol. 57: Collinson, C., and M. Leon Economic viability of ethical cocoa trading in Ecuador. Report Natural Resources Institute, Chatham. 12. Daniel, H.-M., et al Yeast diversity of Ghanaian cocoa bean heap fermentations. FEMS Yeast Res. 9:

29 De Bruyne, K., N. Camu, L. De Vuyst, and P. Vandamme Lactobacillus fabifermentans sp. nov. and Lactobacillus cacaonum sp. nov., isolated from Ghanaian cocoa fermentations. Int. J. Syst. Evol. Microbiol. 59: De Bruyne, K., N. Camu, L. De Vuyst, and P. Vandamme Weissella fabaria sp. nov., isolated from a Ghanaian cocoa fermentation. Int. J. Syst. Evol. Microbiol. 60: De Vuyst, L., T. Lefeber, Z. Papalexandratou, and N. Camu The functional role of lactic acid bacteria in cocoa bean fermentation, p In F. Mozzi, R. R. Rayak, and G. M. Vignolo (ed.), Biotechnology of lactic acid bacteria: novel applications. Wiley-Blackwell, Ames, IA. 16. Deheuvels, O., B. Decazy, R. Perez, G. Roche, and F. Amores The first Ecuadorian Nacional cocoa collection based on organoleptic characteristics. Tropical Sci. 44: Despréaux, D Phytophthora diseases of Theobroma cacao, p In C. Cilas and D. Despréaux (ed.), Improvement of cocoa tree resistance to Phytophthora diseases. Cirad, Montpellier. 18. Efraim, P., et al Influência da fermentação e secagem de amêndoas de cacau no teor de compostos fenólicos e na aceitação sensorial. Ciênc. Tecnol. Aliment. 30: Endo, A., Y. Futagawa-Endo, and L. M. T. Dicks Isolation and characterization of fructophilic lactic acid bacteria from fructose-rich niches. Syst. Appl. Microbiol. 32: Endo, A., et al Fructobacillus tropaeoli sp. nov., a novel fructophilic lactic acid bacterium isolated from a flower. Int. J. Syst. Evol. Microbiol. 61: Ercolini, D., G. Moschetti, G. Blaiotta, and S. Coppola Behavior of variable V3 region from 16S rdna of lactic acid bacteria in denaturing gradient gel electrophoresis. Curr. Microbiol. 42:

30 Garcia-Armisen, T., et al Diversity of the total bacterial community associated with Ghanaian and Brazilian cocoa bean fermentation samples as revealed by a 16S rrna gene clone library. Appl. Microbiol. Biotechnol. 87: Hansen, C. E., M. del Olmo, and C. Burri Enzyme activities in cocoa beans during fermentation. J. Sci. Food Agric. 77: Henderson, P Cocoa, finance, and the state in Ecuador, Bull. Latin Am. Res. 16: ISO 2291:1980. Cocoa beans-determination of moisture content (routine method). International Organization for Standardization. Geneva, Switzerland. 26. Jano, P., and D. Mainville The cacao marketing chain in Ecuador: Analysis of chain constraints to the development of markets for high-quality cacao. IAMA, Parma. 27. Jayani, R. S., S. Saxena, and R. Gupta Microbial pectinolytic enzymes: a review. Process Biochem. 40: Jespersen, L., D. S. Nielsen, S. Hønholt, and M. Jakobsen Occurrence and diversity of yeasts involved in fermentation of West African cocoa beans. FEMS Yeast Res. 5: Kondo, K., and M. Ameyama Carbohydrate metabolism by Acetobacter species. Part I. Oxidative activity for various carbohydrates. Bull. Agric. Chem. Soc. Jpn. 22: Kostinek, M., et al Diversity of predominant lactic acid bacteria associated with cocoa fermentation in Nigeria. Curr. Microbiol. 56: Lagunes-Gálvez, S., G. Loiseau, J. L. Paredes, M. Barel, and J. P. Guiraud Study on the microflora and biochemistry of cocoa fermentation in the Dominican Republic. Int. J. Food Microbiol. 114:

31 Lefeber, T., W. Gobert, G. Vrancken, N. Camu, and L. De Vuyst Dynamics and species diversity of communities of lactic acid bacteria and acetic acid bacteria during spontaneous cocoa bean fermentations in vessels. Food Microbiol. 28, Lehrian, D. W., and G. R. Patterson Cocoa fermentation, p In G. Reed (ed.), Biotechnology, a comprehensive treatise, vol. 5. Verlag Chemie, Basel. 34. Lisdiyanti, P., Y. Yamada, T. Uchimura, and K. Komagata Identification of Frateuria aurantia strains isolated from Indonesian sources. Microbiol. Cult. Coll. 19: Lopez, I., et al Design and evaluation of PCR primers for analysis of bacterial populations in wine by denaturing gradient gel electrophoresis. Appl. Environ. Microbiol. 69: Luna, F., D. Crouzillat, L. Cirou, and P. Bucheli Chemical composition and flavour of Ecuadorian cocoa liquor. J. Agric. Food Chem. 50: Nielsen, D. S., et al The microbiology of Ghanaian cocoa fermentations analysed using culture-dependent and culture-independent methods. Int. J. Syst. Evol. Microbiol. 114: Papalexandratou, Z., N. Camu, G. Falony, and L. De Vuyst Comparison of the bacterial species diversity of spontaneous cocoa bean fermentations carried out at selected farms in Ivory Coast and Brazil. Food Microbiol. 28: Papalexandratou, Z., I. Cleenwerck, P. De Vos, and L. De Vuyst (GTG) 5 -PCR reference framework for acetic acid bacteria. FEMS Microbiol. Lett. 301: Papalexandratou, Z., and L. De Vuyst Assessment of the yeast species composition of cocoa bean fermentations in different cocoa-producing regions using denaturing gradient gel electrophoresis. FEMS Yeast Res. In press. 41. Papalexandratou, Z., G. Vrancken, K. De Bruyne, P. Vandamme, and L. De Vuyst Spontaneous organic cocoa bean fermentations in Brazil are characterized by a restricted species 31

32 diversity of lactic acid bacteria and acetic acid bacteria. Food Microbiol. doi: /j.fm Payne, M. J., W. J. Hurst, K. B. Miller, C. Rank, and D. A. Stuart Impact of fermentation, drying, roasting, and Dutch processing on epicatechin and catechin content of cacao beans and cocoa ingredients. J. Agric. Food Chem. 58: Phillips-Mora, W., M. C. Aime, and M. J. Wilkinson Biodiversity and biogeography of the cacao (Theobroma cacao L.) pathogen Moniliophthora roreri in tropical America. Plant Pathol. 56: Sanches, C. L. G., L. R. M. Pinto, A. W. V. Pomella, S. D. V. M. Silva, and L. L. Loguercio Assessment of resistance to Ceratostytis cacaofunesta in cacao genotypes. Eur J. Plant. Pathol. 122: Schwan, R. F., and A. E. Wheals The microbiology of cocoa fermentation and its role in chocolate quality. Crit. Rev. Food Sci. Nutr. 44: Swings, J The Genus Frateuria, p In M. Dworkin, S. Falkow, E. Rosenberg, K. M. Schleifer, and E. Stackebrandt (ed.), The prokaryotes, 3rd ed. Springer, New York, NY. 47. Thompson, S. S., K. B. Miller, and A. S. Lopez Cocoa and coffee, p In M. P. Doyle, L. R. Beuchat, and T. J. Montville (ed.), Food microbiology: fundamentals and frontiers, 2nd ed. ASM Press, Washington, DC. 48. Walter, J., et al Detection of Lactobacillus, Pediococcus, Leuconostoc, and Weissella species in human feces by using group-specific PCR primers and denaturing gradient gel electrophoresis. Appl. Environ. Microbiol. 67: Yamada, Y., Y. Okada, and K. Kondo Isolation and characterization of polarly flagellated intermediate strains in acetic acid bacteria. J. Gen. Appl. Microbiol. 22:

33 Ziegleder, G Linalool contents as characteristic of some flavor grade cocoas. Z. Lebensm. Unters. Forsch. 191: FIGURES FIG. 1. Microbial succession of yeasts (MEA counts, ), lactic acid bacteria (MRS counts, ), acetic acid bacteria (DMS counts, Δ, AAM counts, x) and total aerobic bacteria (PCA counts, ) during the P1 (A), B1 (B), P2 (C), and B2 (D) cocoa bean fermentations carried out in Ecuador. Full arrows indicate spreading (platform fermentations) or mixing (box fermentations) of the fermenting cocoa pulp-bean mass; the dashed arrow indicates the end of the fermentation. FIG. 2. Cluster analysis of the generated and digitized (GTG) 5 -PCR fingerprints of the 529 LAB isolates from MRS agar, isolated during the two fermentation rounds in Ecuador. Banding patterns were clustered including reference strains (not shown) by using UPGMA, with correlation levels expressed as percentage values of the Pearson correlation coefficient. Species validation of representative strains of each cluster was carried out by partial 16S rrna gene sequence analysis (indicated by *). Percentages of identity between sequenced isolates of the present study and well characterized strains in the BLAST DataBank are reported in the case of < 100% identity. Two identities are given in the case of nonaccurate identification. FIG. 3. Cluster analysis of the generated and digitized (GTG) 5 -PCR fingerprints of the 334 AAB isolates from DMS and AAM agars, isolated during the two fermentation rounds in Ecuador. Banding patterns were clustered together with reference strains (not shown) by using UPGMA, with correlation levels 33

34 expressed as percentage values of the Pearson correlation coefficient. Species validation of representative strains of each cluster was carried out by partial 16S rrna gene sequence analysis (indicated by *). Percentages of identity between sequenced isolates of the present study and well characterized strains in the BLAST DataBank are reported in the case of < 100% identity. Two identities are given in the case of non-accurate identification. FIG. 4. PCR-DGGE profiles with LAC1-LAC2 primers (35%-60% denaturing gradients from top to bottom of the gels) of the amplified 16S rrna gene fragments of the LAB present in the cocoa bean samples of the P1 (A), B1 (B), P2 (C), and B2 (D) cocoa bean fermentations carried out in Ecuador. The reference ladder (L) consisted of (a) Lactobacillus plantarum LMG 6907 T, (b) Lb. acidophilus LMG 9433 T, (c) Lb. fermentum LMG 6902 T, (d) Leuconostoc pseudomesenteroides 274 (7), (e) Pediococcus acidilactici LMG T, and (f) Lb. casei LMG 6904 T. Full arrows indicate spreading (platform fermentations) or mixing (box fermentations) of the fermenting cocoa pulp-bean mass; the dashed arrow indicates the end of the fermentation. The lane numbers represent the time (in h) of sampling during fermentation. The closest relatives of the fragments sequenced (% of identical nucleotides compared to sequences retrieved from the GenBank database and accession numbers between brackets) were: (i) Lb. plantarum (100%; EU637402); (ii) Weissella beninensis (100%; EU439435); (ii ) W. fabaria (100%; FM179679); (iii) Lb. fermentum (100%; EU688978); (iv) Lb. cacaonum (100%; AM905389); (v) Leuconostoc sp. (100%; HM224479); (v ) Lc. mesenteroides subsp. mesenteroides (100%; HM218817); (vi) Lc. pseudomesenteroides (100%; DQ523483); (vii) Fr. tropaeoli (98%; AB542054); (viii) a previously uncultured bacterium (100%; DQ980802); and (ix) Lb. nagelii (100%; AB370876). 34

35 FIG. 5. PCR-DGGE profiles with 357f-518r primers (35%-70% denaturing gradients from top to bottom of the gels) of the amplified 16S rrna gene fragments of bacteria present in the cocoa bean samples of the P1 (A), B1 (B), P2 (C), and B2 (D) cocoa bean fermentations carried out in Ecuador. The reference ladder (L) consisted of (a) Lactobacillus plantarum LMG 6907 T, (b) Lb. fermentum LMG 6902 T, (c) Leuconostoc pseudomesenteroides 274 (7), (d) Acetobacter pasteurianus LMG 1262 T, (e) Gluconacetobacter europaeus LMG T, and (f) A. senegalensis LMG T. Full arrows indicate spreading (platform fermentations) or mixing (box fermentations) of the fermenting cocoa pulp-bean mass during fermentation; the dashed arrow indicates the end of the fermentation. The lane numbers represent the time (in h) of sampling during fermentation. The closest relatives of the fragments sequenced (% of identical nucleotides compared to sequences retrieved from the GenBank database and accession numbers between brackets) were: (i) Lb. plantarum (100%; EU637402); (ii) Lb. fermentum (100%; EU688978); (iii) Lc. pseudomesenteroides (100%; DQ523483); (iv) Fr. tropaeoli (98%; AB542054); (v) T. saanichensis (100%; EU215774); (vi) Tatumella sp./t. morbirosei (100%; FJ617235); (vii) chloroplast DNA (100%; HQ244500); (viii) T. punctata (100%; FJ756351); (ix) Tatumella sp./t. terrea (100%; FJ756353); and (x) acetic acid bacteria species (100%; GQ246703, HM190252, AB485746). FIG. 6. PCR-DGGE profiles with WBAC1-WBAC2 primers (45%-70% denaturing gradients from top to bottom of the gels) of the amplified 16S rrna gene fragments of the LAB and AAB present in bulk cells washed off from AAM and DMS agars of the P2 and B2 fermentations. The reference ladder (L) consisted of (a) Lactobacillus plantarum LMG 6907 T, (b) Leuconostoc pseudomesenteroides 274 (7), (c) Lb. fermentum LMG 6902 T, (d) A. senegalensis LMG T, (e) Acetobacter pasteurianus LMG 1262 T, and (f) Glunonacetobacter europaeus LMG T. The lane numbers represent the time (in h) 35

36 of sampling during fermentation. The closest relatives of the fragments sequenced (% of identical nucleotides compared to sequences retrieved from the GenBank database and succession numbers between brackets) were: (i) Lc. pseudomesenteroides (100%; DQ523483); (ii) A. malorum/orleanensis (100%; FJ831444); (iii) Lb. fermentum (100%; EU688978); (iv) A. pasteurianus/pomorum (100%; FJ227313); (v) Bacillus sp. (100%; FJ772025); (vi) Acetobacter sp. (100%; HM804024); and (vii) E. durans/lc. lactis subsp. lactis (100%; FJ917740, CP002365). FIG. 7. Course of glucose ( ), fructose ( ), and mannitol ( ) in the pulp during the P1 (A), B1 (C), P2 (E), and B2 (G) cocoa bean fermentations carried out in Ecuador. Course of citric acid ( ), lactic acid ( ), gluconic acid ( ), acetic acid ( ), and ethanol ( ) in the pulp during the P1 (B), B1 (D), P2 (F), and B2 (H) cocoa bean fermentations. Full arrows indicate spreading (platform fermentations) or mixing (box fermentations) of the fermenting cocoa pulp-bean mass; the dashed arrow indicates the end of the fermentation. FIG. 8. Residual carbohydrates, mannitol, and organic acids in the fermented dry cocoa beans of the P1, B1, P2, and B2 cocoa bean fermentations. 36

37 Table 1. Species diversity of yeasts, lactic acid bacteria (LAB), and acetic acid bacteria (AAB), indicated as number of isolates, and their isolation source [swabs or fermentation day (D)] during the Platform 1, Box 1, Platform 2, and Box 2 cocoa bean fermentations carried out in Ecuador, as revealed by M13-PCR (yeasts) and (GTG) 5-PCR (LAB and AAB) fingerprinting classification and identification through gene sequencing of representative isolates of each cluster. Yeasts Species Fermentation/Source First fermentation round Second fermentation round Platform 1 (P1) Box 1 (B1) Platform 2 (P2) Box 2 (B2) Swabs D0 D1 D2 D3 D4 Swabs D0 D1 D2 D3 D4 D5 Swabs D0 D1 D2 D3 D4 Swabs D0 D1 D2 D3 D4 Candida tropicalis 1 1 Candida sorbosivorans-like 1 5 Hanseniaspora opuntiae 2 1 Kluyveromyces marxianus Pichia kluyveri 1 Pichia kudriavzevii Pichia manshurica Rhodotorula minuta 1 Saccharomyces cerevisiae Torulaspora delbrueckii 1 Lactic acid bacteria Enterococcus casseliflavus 12 Enterococcus saccharolyticus 1 Enterococcus saccharolyticus 97% 2 Enterococcus sp. 1 Fructobacillus durionis 1 Fructobacillus ficulneus 98% 1 Fructobacillus tropaeoli 98% Lactobacillus amylovorus 1 1 Lactobacillus cacaonum 6 Lactobacillus coryniformis 1 Lactobacillus fabifermentans Lactobacillus farraginis 4 Lactobacillus fermentum Lactobacillus garvieae 2 Lactobacillus nagelii 11 Lactobacillus plantarum Lactobacillus sp./lactobacillus satsumensis Lactococcus lactis subsp. lactis Leuconostoc fallax 1 Leuconostoc mesenteroides 1 Leuconostoc pseudomesenteroides Streptococcus salivarius 1 1 Weissella cibaria 1 Weissella fabaria Acetic acid bacteria 2 1 Acetobacter cibinongensis Acetobacter lovaniensis/fabarum Acetobacter malorum/cerevisiae 98% 4 2 Acetobacter malorum/indonesiensis 98% 2 1 Acetobacter fabarum Acetobacter ghanensis Acetobacter orientalis 2 1 Acetobacter pasteurianus Acetobacter peroxydans 6 1 Acetobacter pomorum 3 1 Acetobacter senegalensis Acetobacter syzygii Frateuria aurantia Microbacterium lacticum 1 1 Gluconobacter oxydans Gluconobacter sp

38 Table 2. Quality assessment of fermented dry cocoa beans by means of a cut test. Cut test parameters P1 B1 P2 B2 Moisture (%) Violet beans (%) Well-fermented beans (%) Semi-fermented beans (%) Weight of 100 beans (g) Downloaded from 38 on August 22, 2018 by guest

39 Table 3. Average flavor scores (on a scale of 0 to 100) of the dark chocolates made in comparison with the commercial reference 811NV. Averages of eight taste panel members are presented. 811NV P1 B1 P2 B2 Intensity Cocoa Heavy Sweet Sour Fruity Herbaceous Flowery Aftertaste intensity Remarks impure, astringent green, alcoholic impure, lactic acid chemical Downloaded from 39 on August 22, 2018 by guest

40 Log (CFU/g) Log(CFU U/g) A Time (h) Log (CFU/g) Log(CFU U/g) B Time (h) C 9 9 D Time (h) Time (h)

41 Similarity (%) M1386, M1385, M1397, M1285, M1284, M1276, M1297, M1420, M1383, M1290, M1252, M1302, M1326, M1337, M1242, M1400, M1288, M1404, M1272, M1331, M1403, M1218, M1339, M1262, M1359, M1323, M1220, M1330, M1344, M1327, M1353, M1278, M1219, M691, M1280, M779*, M966, M697a, M792, M770, M776, M791a, M784, M811, M861, M783, M782, M701b, M786, M840, M1291, M1332, M1354, M1322, M1352, M1358, M1343, M1340, M1224, M1232, M1258, M1247a, M1350, M1329, M1282, M1355, M1261, M1198, M1360, M1356, M1357, M1351, M814, M1257, M1334, M1271, M1259, M767b, M777, M888a, M752a, M1239*, M1222, M1281, M1277b, M1221, M831, M804, M817, M828, M827, M854, M912, M675, M913b, M798, M796, M810, M761b, M808, M763a, M826a, M678, M823, M922, M835, M1361a, M1249, M839, M1324, M771, M918, M822, M833a, M769c, M841b, M806b, M664, M837, M849, M679, M819, M753, M915a, M646, M829, M676, M686a, M797, M821b, M914, M1270, M1375a, M1266, M1263, M1293, M1338, M1335, M1341, M1253, M820a, M921, M1246a, M913a, M809, M815b, M802b, M852, M777a, M824, M813, M769b, M758, M803, M759a, M687, M848a, M685, M816b, M848b, M769a, M715a, M923, M756, M832, M830b, M825, M909, M911, M846, M754, M760*, M768, M807, M681, M919, M822a, M636, M709a, M763b,M674a, M805, M1236, M1246b, M1213, M1301, M1407, M1314, M1412, M1391, M1415, M1289, M1279, M1305, M1245, M1304, M1379, M1310, M1308, M1275, M1409, M1384, M1283, M1315, M1336, M1217, M1254, M1333, M1416, M1286, M1325, M1410, M1292, M1073, M1320, M1307, M1287*, M1376, M1306, M1421, M1299, M1319, M1312, M1255, M1260, M1317, M1393, M1295b, M1406, M1405, M1056, M1243, M1244, M1237, M1241, M1238, M1234, M1240, M1047, M1233, M1265, M1267, M1235, M1349, M1347, M1300, M996, M1402, M930, M1256, M1247, M1318, M1346, M1408, M797a, M917, M690, M704, M684, M642, M836*, M834 M738* M1066* Lactobacillus fermentum Enterococcus saccharolyticus Enterococcus sp. M1230, M1228, M1229, M1223, M1141, M1214, M1225, M1192, M1216, M1185, M1183, M1182, M1186, M1147, M1177, M1176, M1169, M1202, M1175, M1226, M1208, M1132, M1125, M1204, M1180, M1179, M1061, M1231, M1211, M1207, M1210, M1215, M1171, M1150, M1227, M1209, M1123, M1124, M1139, M1151, M1120, M1059 *, M1295a, M1142, M1149, M1251, M1174, M1126, M1140, M878, M740b, M1162, M1181, M693, M879, M877*, M692, M967, M700, M744, M736, M858, M699, M1212, Leuconostoc pseudomesenteroides M1166, M1268, M1164, M1153, M624, M628*, M626, M650, M939, M936a, M627, M706, M707, M705, M643, M959, M1145, M1144, M1005, M1203, M1173, M1156, M689, M1129, M1201, M1178, M1206, M1137, M1148, M1194, M1191, M1187*, M1152a, M1195, M1165 M638* M905* M633, M634, M665, M653, M639, M863, M876, M651, M641, M645, M1189* M1025, M1037, M1039 *, M1036, M1033, M1035, M1042, M1050b, M1031, M1034, M1026, M1048 * M667*, M670, M1010*, M1099b M871, M869* M870 *, M795, M897, M790, M974, M952, M964, M886, M885, M960, M972, M884, M799, M955, M881, M792b, M792a, M785, M954, M774, M892, M973, M896, M965, M775, M801, M963, M893, M780, M969, M788, M903 *, M968 M1160 M660, M672 *, M983, M745, M728 *, M632, M721, M724, M718, M748, M741, M725, M721b, M719, M750, M723, M737, M1190, M743, M730, M714, M732, M735, M708a, M713, M712a, M731, M733 *, M751, M741a, M722, M746 *, M661b, M673a, M662 *, M926b, M739, M734, M655a, M926a, M710 *, M658, M720, M747, M729, M715, M671, M640, M657, M656, M711, M637 *, M654, M629, M644 *, M749a* M1199a* M859* M1184, M1168*, M1134, M696b, M1146 M936, M941*, M874, M940, M888b M1127*, M1167, M1188 M1019, M1013*, M1012, M1008 M648* M1365, M1368*, M1372, M1362 M1364, M1370, M1363, M1366*, M1369, M1373, M1377 M927*, M931 M933* M925* M629a, M697b M956*, M880b M742* M1163, M1152b, M1200 *, M1196, M1199b *, M1197b Fructobacillus ficulneus 98% Fructoabacillus durionis Lactobacillus fabifermentans Enterococcus casselifvavus Lactobacillus plantarum Lactobacillus garvieae Lactobacillus sp./lb. satsumensis Weissella fabaria Fructobacillus tropaeoli 98% Leuconostoc fallax Lactobacillus fabifermentans Lactococcus lactis subsp. lactis Lactococcus lactis subsp. lactis Weissella fabaria Lactobacillus farraginis Weissella cibaria Lactobacillus nagelii Lactobacillus nagelii Enterococcus saccharolyticus 97% Leuconostoc mesenteroides Lactobacillus coryniformis Streptococcus salivarius Lactobacillus amylovorus Leuconostoc pseudomesenteroides Lactobacillus cacaonum

42 Similarity (%) D747a, D734b, D694b, D682a, D684b, D681,, D741, D752, D1012*, D1007, D755, D815*, D976*, A1128*, A1134, D725b, D1085, D1024, D1003, D1010, D995, D1022, D997, D1001*, D1004, D999, D994, A1024, D1009, D1055.D1014, A1129a, A1194.D1000*, D978, D817, D1017, D1023, D1077, D1037, D992, D1018, D1006a, D1016, D998, D1006b, D986, D1020, D975, A1023, A1245, A1251*, D1087, A1113, D1038, D1062, D1081, A1249, A1206, A1210, A1218, A1244, A1129b, A1243, A1217, D1068, D968, A1120, A1214, D971, D1073, D1039, D969, A1109, A1204, A1083, D972, D1082, D1028, A1216, A1222, A1247, D938, A1215, A1250, A1223, A1086, D1043, D1021a, D1075, D1069, D1061, D1065, A1252, A1213, A1226*, D1078, D1030, D1080, A1119, D1084, A1246, D1060, D1053*, D988, D1051, D1011,A1230, A1236, A1240, D1057, D1002, D1005*, D1058, D985, D811, A1233, A1029*, D1035, D895, D1021b, D1027, D1054, D982, D730b*, D966, D691, D762b, D980, D1050, D1008, A1238, A1235, A1239, A951, D744a, D983, D991, D748*, D760a, D757a, D740, D764, D685, D690a, D732, D656a, D683, D655, D654, D653b, D742a, D693, D735, D759a, D738b, D692b, D711, D743, D629a, D538, D751, D756, D736, D754b, D680, D532, A935, D728a, D749, D759, A950, D657a, D687, D659, D763a, D753, D746a, D739, D758, D727b, D737a, D750, D663*, D762a, D990 D1051* A750* A715 A1066, A1048 D1026*, D1064 A1228* D858, D848* A760*, A762a, A761b D916, D1071, D960*, A1124, D964, D919, D939, A1136, D955, D846, D766, D778, D901,*, D1034, D1040, D667, A1221, D1044, D921, D931, D937, D1041, A1199, D873 A1132, D965, A1220, D1059* A966*, A845*, A1028, A817* A824*, A723* A827, A828* A813* D973* A854, A905*, D953, D1046*, D1049, D648, D649, D526*, D1032, D1045, D1047, D1070*, D1048, D965, A1026, A1118, A1205, A1060 D646, D489a, D455, D456, A748, A753, A744 A1063 A728* A1005 *, A1012, A1050, A1065 *, A1013, A1052, A954, A1049 Acetobacter pasteurianus Acetobacter malorum/cerevisiae 98% Acetobacter syzygii Acetobacter syzygii Acetobacter malorum/cerevisiae 98% Microbacterium lacticum Acetobacter pasteurianus Acetobacter pomorum Acetobacter senegalensis Acetobacter pasteurianus Gluconobacter oxydans Gluconobacter sp. Gluconobacter oxydans Gluconobacter sp. Acetobacter senegalensis Acetobacter t ghanensis Acetobacter peroxydans Acetobacter pasteurianus Acetobacter peroxydans Acetobacter syzygii A762, A763 *, D453, D459, A747, A742, D548, D745 *, D547, D550, D543, D549, D569, D552, D568, D544, D553, D452, D559a, D555 *, D551, D566, D542, D891* A770*, A722, D451* A1248*, A1025 A729*, A718, A799 D488, D514, D515*, D513, D517b A740, D458*, D641, A720 D460* D559*, D493, D506, D494, D558, D501*, D563, D853*,D840 D527* A1010*, A1004, A1219 D898, D900*, A1064 Acetobacter fabarum Acetobacter malorum/indonesiensis 98% Acetobacter senegalensis Acetobacter cibinongensis Frateuria aurantia Acetobacter lovaniensis/fabarum Gluconobacter sp. Acetobacter lovaniensis/fabarum Acetobacter senegalensis Acetobacter orientalis Acetobacter malorum/cerevisiae 98%

43 A C Time (h) L L L L a i i ii b c iii d vi v e vii vii f L L L L ii iv v vi ix i iv iii Time (h) ix iii B D Time (h) L L L L a b c d e f ii vii vi iii Time (h) L L L ii vi iii vi iii iii vii vi iii viii viii viii viii viii viii viii viii

44 A C Time (h) L L L a b c d e f i vii viii vi ii ii ii iii Time (h) L L L L xiii xiii B Time (h) L L L L a ii ii ii b ii ii ii c iii iii d e f vii v ix x x x a b c d e f D vii iv v vi ix ii iii ix Time (h) L L L vii v vi x ii x x

45 AAM medium Platform fermentation 2 Box fermentation 2 Platform fermentation 2 Box fermentation L i ii iv vii i Time (h) L a DMS medium L v b vi c ii iii iv d iv iv e f ii

46 Concentrati on (mg/g) Con ncentration (mg/g) Time (h) Time (h) A C g) oncentration (mg/g Co ion (mg/g) Concentrati Time (h) Time (h) B D

47 g) oncentration (mg/g Co Concentrati on (mg/g) Time (h) Time (h) E G g) oncentration (mg/g Co Concentratio on (mg/g) 30 F Time (h) 30 H Time (h)

48 g) oncentration (mg/g C Glucose Fructose Sucrose Mannitol P1 B1 P2 B2 g) oncentration (mg/g C Lactic acid Citric acid Acetic acid Gluconic acid P1 B1 P2 B2