International Journal of Food Microbiology 161 (2013) 121–133
Contents lists available at SciVerse ScienceDirect
International Journal of Food Microbiology
j ourna l homepage: www.e lsev ie r .com/ locate / i j foodmicro
Spontaneous cocoa bean fermentation carried out in a novel-design stainlesssteel tank: Influence on the dynamics of microbial populations andphysical–chemical properties
Gilberto Vinícius de Melo Pereira a, Karina Teixeira Magalhães a, Euziclei Gonzaga de Almeida b,Irene da Silva Coelho c, Rosane Freitas Schwan a,⁎a Biology Department, Federal University of Lavras (UFLA), CEP 37200-000, Lavras, MG, Brazilb Biology Department, Federal University of Mato Grosso (UFMT), CEP 78060-900, Cuiabá, MT, Brazilc Microbiology and Immunology Department – Federal Rural University of Rio de Janeiro (UFRRJ), CEP 23890-000, Rio de Janeiro, RJ, Brazil
Spontaneous cocoa bean fermentations carried out in a novel-design 40-kg-capacity stainless steel tank (SST)wasstudied in parallel to traditional Brazilian methods of fermentation in wooden boxes (40-kg-capacity woodenboxes (WB1) and 600-kg-capacity wooden boxes (WB2)) using a multiphasic approach that entailed culture-dependent and -independent microbiological analyses of fermenting cocoa bean pulp samples and targetmetabolite analyses of both cocoa pulp and cotyledons. Both microbiological approaches revealed that thedominant species of major physiological roles were the same for fermentations in SST, relative to boxes. Thesespecies consisted of Saccharomyces cerevisiae and Hanseniaspora sp. in the yeast group; Lactobacillus fermentumand L. plantarum in the lactic acid bacteria (LAB) group; Acetobacter tropicalis belonging to the acetic acid bacteria(AAB) group; and Bacillus subtilis in the Bacillaceae family. A greater diversity of bacteria and non-Saccharomycesyeasts was observed in box fermentations. Additionally, a potentially novel AAB belonging to the genus Asaiawasisolatedduring fermentation inWB1. Cluster analysis of the rRNA genes-PCR-DGGEprofiles revealed amore com-plex picture of the box samples, indicating that bacterial and yeast ecology were fermentation-specific processes(wooden boxes vs. SST). The profile of carbohydrate consumption and fermentation products in the pulp andbeans showed similar trends during both fermentation processes. However, the yeast-AAB-mediated conversionof carbohydrates into ethanol, and subsequent conversion of ethanol into acetic acid, was achieved with greaterefficiency in SST, while temperatures were generally higher during fermentation in wooden boxes. With furtherrefinements, the SSTmodel may be useful in designing novel bioreactors for the optimisation of cocoa fermenta-tion with starter cultures.
Fermentation of cocoa beans is the first step in the chocolate-makingchain and consists of a five- to seven-day on-farm process during whichmicroorganisms growwithin the pulp material that surrounds the seeds(beans) of the cocoa fruit (Schwan et al., 1995, 1997; Schwan, 1998;Schwan andWheals, 2004; Nielsen et al., 2007). The cocoa fermentationconsists of well-defined microbial successions that are initially domi-nated by yeasts and subsequently surpassed by lactic acid bacteria(LAB); these species then decline after 48 h of fermentation and arereplaced by acetic acid bacteria (AAB). Various species of Bacillus,other bacteria and filamentous fungi may also grow throughoutfermentation (Ardhana and Fleet, 2003; Schwan and Wheals, 2004;Garcia-Armisen et al., 2010; Lima et al., 2011). This microbial activitygenerates metabolites and conditions that kill the beans, thereby
+55 35 38291100.
rights reserved.
triggering an array of biochemical reactions and chemical changeswithinthe bean, itself, that are essential for the development of the complex(and much-loved) flavour of “chocolate” (Pereira et al., 2012b).
Cocoa fermentation is an empirical procedure and may not bearbeans of consistent quality, which requires that the chocolate industrycontinuously modifies their formulations (Lagunes-Gálvez et al., 2007).Common problems involve levels of acidity, and incomplete fermenta-tion, which results in a lack of cocoa flavour, or off-flavours resultingfrom over-fermentation and spoilage of beans; all of these problems re-duce crop value for the farmer (Schwan, 1998). While experimental ap-plications of defined starter cultures have produced satisfactory results(Samah et al., 1992; Schwan, 1998; Leal et al., 2008), this technique hasnot been widely implemented in the field (Lefeber et al., 2011). Onelimitation to the introduction of starter cultures is that the greatmajorityof cocoa fermentations are conducted over banana leaves or in woodenboxes that facilitates contamination by natural microbiota.
The demand for hygienic production practices has increased theappeal of using stainless steel tanks in industrial bioprocesses
122 G.V. de Melo Pereira et al. / International Journal of Food Microbiology 161 (2013) 121–133
(e.g., the production of yogurt, beer, wine and cider). As cocoa beanfermentation involves a complexmicrobial succession, the use of a stain-less steel tank is a challenging prospect, and specific strategies to this sce-nario need to be developed to avoid an unsuccessful campaign ofprocess-validation. Both the biology and chemistry of cocoa beanfermentation are complex and all aspects of the process need to besubjected to comprehensive, simultaneous, dynamic analyses to quanti-tatively predict the outcomes of defined changes. Understanding themi-crobial ecology of the fermentation process is the first stage to develop anew procedure and optimise both processing efficiency and quality of itsend products (Schwan, 1998; Ardhana and Fleet, 2003; Garcia-Armisenet al., 2010).
In the present study, spontaneous cocoa fermentations wereperformed in a stainless steel tank. In comparison to traditional Brazilianmethods of fermentation inwooden boxes, the stainless steel tankwas anovel design thatwas capable of turning a 40 kg load of beans. Validationof the microbial fermentations was based on both culture-dependentand culture-independent approaches, which were used to assess thedynamics of yeast and bacterial populations during the fermentationprocesses. In parallel, a link was established between the substratesand metabolites found in pulp and beans, and the microbes responsiblefor these reactions were identified.
2. Materials and methods
2.1. Fermentation experiments and sampling
The experiment was conducted at a cocoa farm located in the city ofItajuípe, Bahia State, Brazil. Fully mature cocoa pods of mixed-hybridswere harvested and broken open with unwashed machetes; seeds plustheir surrounding pulp were scooped out manually and immediatelytransferred to the site of fermentation. Fermentation was performed in1 m3 (WB1) and 0.06 m3 (WB2) wooden boxes capable of holding600 kg and 40 kg of cocoa beans, respectively, or in a 40-kg-capacitystainless steel conical tank (SST) (0.06 m3; non-commercial bioreac-tors). The stainless steel tank fermentation was performed in triplicateand was designed as described by Schwan and Wheals (2004). Thewooden boxes and the stainless steel tank were cleaned with detergentand rinsed with tap water, before they were separately filled with eachof the fermentation treatments. The wooden boxes and the stainlesssteel tank contained holes at the bottom to allow drainage of liquid gen-erated during fermentation. The wooden boxes were covered with freshbanana leaves and the stainless steel tank was covered with a polypro-pylene lid to ensure adequate insulation. All fermentations were donein triplicate and turned every 24 h and were performed simultaneouslyto exclude as many environmental factors as possible (e.g., harvest sea-son and external contamination). Natural fermentation proceeded atambient temperature for 6 days.
Every 12 h during the fermentation process, 200 g of each samplewas collected randomly, placed in sterile bags and transferred to thelaboratory. Samples collected for chemical and culture-independentanalyses were sealed in plastic bags and stored in a freezer at −20 °C.Microbiological analyses were performed on the day of samplecollection.
2.2. Cultivation-based quantification of microorganisms
Freshly acquired samples were immediately plated on four differentculture media following appropriate dilution for enumeration of thenumber of colony-forming units (CFU). For culture-based quantification,25 g of cocoa beans and adherent pulp were added to 225 mLsaline-peptone water [(v/v) (0.1% bacteriological peptone (Himedia),0.8% NaCl (Merck, Whitehouse Station, USA)], homogenised in aStomacher at normal speed for 5 min (10-1 dilution) and diluted serially.Lactic acid bacteria (LAB) were enumerated by pour plate inoculation inMRS agar (Merck) containing 0.1% (v/v) cycloheximide (Merck) to
inhibit yeast growth and 0.1% (v/v) cystein-HCl to maintain anaerobicconditions during incubation. Acetic acid bacteria (AAB)were enumerat-ed by surface inoculation on GYC agar [50 g/L glucose (Merck), 10 g/Lyeast extract (Merck), 30 g/L calcium carbonate (Merck) and 20 g/Lagar (Merck), pH 5.6] supplemented with 0.1% cycloheximide to inhibityeast growth and 50 mg/L penicillin (Sigma, St. Louis, USA) to inhibitLAB growth. Yeast were enumerated by surface inoculation on YEPGagar [1% yeast extract (Merck), 2% peptone (Himedia), 2% glucose(Merck) at pH 5.6] containing 100 mg/L chloramphenicol (Sigma) and50 mg/L chlortetracycline (Sigma) to inhibit bacterial growth. Nutrientagar (Merck) containing 0.1% cycloheximide (Merck) was used as ageneral medium for viable mesophilic bacteria populations and Bacillusspp. After spreading, the plates were incubated at 30 °C for 3–4 daysfor MRS, YEPG and Nutrient agar cultures. GYC agar cultures were incu-bated at 25 °C for 5–8 days. Following incubation, the number ofcolony-forming units (CFU) was recorded, and each colony type wasmorphologically characterised and counted. The square root of the num-ber of colonies counted for each type were re-streaked and purified. Pu-rified isolates of GYC, YEPG and Nutrient agar were stored at −80 °C inYEPG broth containing 20% glycerol (w/w). Isolates from MRS agarwere stored at−80 °C in MRS broth containing 20% glycerol (w/w).
2.3. Identification of microorganisms
Phenotypic characterisation of bacterial colonies originating fromMRS, GYC and Nutrient agar plates was performed using conventionalmicrobiological methods, including gram staining in conjunction withmicroscopic examination, determination of catalase and oxidase activi-ties, motility tests and assessment of the following: spore formation;acid and gas production fromglucose; acid and gas production from lac-tate and acetate (only for presumptive AAB isolates from GYC agarplates showing clear zones around the colonies). Yeast colonies werephysiologically characterised by determination of their morphology,spore formation and fermentation of different carbon sources, as de-scribed by Kurtzman and Fell (2011).
Representative microbial strains were identified by sequenceanalysis of the full-length 16S rRNA gene or the ITS region for bacteriaand yeast, respectively. Bacterial and yeast cultures were grownunder appropriate conditions, collected from agar plates with a sterilepipette tip and resuspended in 40 μL of PCR buffer. The suspensionwas heated for 10 min at 95 °C and 1 μL was used as a DNA templatein PCR experiments. For bacterial isolates fromMRS and Nutrient agarplates, 16S-rRNA-PCR was performed using primers 27-F and 1512-R(Magalhães et al., 2010). For bacterial isolates from GYC agar platesshowing clear zones around the colonies, 16Sd and 16Sr primerswere used to amplify the 16S rRNA gene region conserved amongAAB according to the method described by Ruiz et al. (2000). Foryeast isolates, ITS-PCR was performed with primers ITS1 and ITS4(Nielsen et al., 2007). The rRNA gene regions were amplified in aThermo PCYL220 thermal cycler (Thermo Fisher ScientiWc Inc.,Waltham, USA) and the PCR products were sequenced in an ABI3730XL automatic DNA sequencer. The sequences were aligned using theBioEdit 7.7 sequence alignment editor and were compared to theGenBank database using the BLAST algorithm (National Centre for Bio-technology Information, Maryland, USA).
Isolates confirmation as S. cerevisiae was accomplished through aspecies-specific PCR assay with HO gene-derived primers (Pereira etal., 2010). To differentiate AAB species indicated by their rRNA gene se-quences to be closely related, specific-biochemical tests wereperformed as described by Pereira et al. (2012b).
2.4. Nucleotide sequence accession number
The nucleotide sequences of representative isolates were depositedin the GenBank database under access numbers: JQ726599 to JQ726611
123G.V. de Melo Pereira et al. / International Journal of Food Microbiology 161 (2013) 121–133
for yeast isolates and JQ726612 to JQ726644 for bacterial isolates(Tables 1 and 2).
2.5. Microbial community analysis through nested PCR-denaturinggradient gel electrophoresis (DGGE)
2.5.1. Total-community DNA isolationCocoa beans and pulp were physically separated by adding 100 mL
of sterile distilled water to 100 g of beans and adherent pulp in a plasticbag. The beans were homogenised in a Stomacher at normal speed for5 min, and the pulp fraction was recovered via decantation. The pulpfraction (40 mL) was lyophilised and freeze-dried cocoa pulp wasground thoroughly with a sterile pestle. Freeze-dried pulp (30 mg)was mixture homogenised twice in 1.5 mL of phosphate-buffer. Thecombined fluids were mixed for an additional 10 min and were thencentrifuged to remove large particles at 100×g for 10 min at 4 °C. Thesupernatant was then centrifuged at 8000×g for 20 min at 4 °C to pelletthe yeast and bacterial cells, whichwere subsequently frozen at−20 °Cfor at least 1 h; this procedure was performed twice. Bacterial and yeastcells were lysed using the method described by Pereira et al. (2012a)and Pereira et al. (2012b), respectively. After lysis, supernatant DNAwas purified in accordance with the instructions in step 4 (bacteria pro-cedure) and step 2 (yeast procedure), which are described in “Protocol:DNA Purification from Tissues” (Qiaamp DNA Mini Kit, Qiagen). The re-sultant samples were stored at−20 °C for further use.
2.5.2. Nested PCR-DGGE strategyTo increase sensitivity, and to facilitate DGGE by analysing frag-
ments of the same length, a two-stepnested PCR techniquewasutilised.To analyse bacterial diversity, primers 27F and 1512R were used inorder to amplify the nearly complete 16S rRNA-encoding gene in thefirst amplification step, which was performed under conventional PCRconditions (Magalhães et al., 2010). The product of this first PCR reac-tion was used as a template for a nested PCR reaction that amplifiedthe V3 region of the 16S rRNA gene with GC-338f and 518r primers
Table 1Comparison of the growth of yeast species during cocoa bean fermentations in stainless ste
Species identificationa Estimated average levels (log CFU/g)b of the isolates at t
Abbreviations: S.: Saccharomyces, P.: Pichia, H.: Hanseniaspora, C.: Candida, D.: Debaromycesa The BLAST search was based on sequences of type and cultured strains at GenBank (Nat
given species if the similarity between the query rDNA sequence and the sequences in theb The estimated average levels (log CFU/g) were obtained after morphological characteriza
means that no colonies were found.
(Ovreas et al., 1997); this generated a DNA fragment suitable forDGGE analysis. To analyse yeast diversity, the ITS regions were ampli-fied by PCR with primers ITS1-F and ITS4; the amplification productwas then amplified with the nested DGGE primers GC-ITS1-F and ITS2(White et al., 1990). PCR reactions were performed in a Mastercycler(Eppendorf, Hamburg, Germany). The PCR products were analysed byDGGE using a BioRad DCode universal mutation detection system(BioRad, Richmond, CA, USA). The PCR products of the second stepwere separated via electrophoresis in 8% (w/v) polyacrylamide gels inrunning buffer containing 1×TAE (20 mM Tris, 10 mM acetate and0.5 mM EDTA at pH 8.0). Optimal separation of bacterial communitieswas achieved with a 30–55% urea-formamide denaturing gradient,while a 12–60% gradient yielded optimal separation for yeast communi-ties (100% corresponded to 7 M urea and 40% (v/v) formamide).
To estimate the differences inmicrobial community structure and dy-namics for different fermentation processes, band positions on DGGEpatterns were analysed by hierarchical cluster analysis performed withBionumerics version 6.50 (Applied Maths, Sint-Martens-Latem). Den-drograms were calculated on the basis of Dice's Coefficient of Similarityusing the unweighed pair group method with the arithmetic averagesclustering algorithm (UPGMA).
2.6. Sequencing of DGGE bands
DGGE bands of interest were excised from the gel with a sterile scal-pel, disrupted in 60 μL of sterile Milli-Qwater, and left overnight at 4 °Cto allow the DNA to diffuse out of the gel. A total of 10 μL of eluted DNAfrom each DGGE band were subjected to re-amplification reactionsusing appropriate primers and the conditions described above. The se-quencing products were purified with a QIAquick PCR purification kit(Qiagen) and sequenced in an automated DNA sequencer (AppliedBiosystems, Foster City, CA, USA). GenBank BLAST searches wereperformed to determine the closest known relatives of the partial ribo-somal DNA sequences obtained.
.ional Center for Biotechnology Information). The isolates were assumed to belong to adatabases was higher than 97%.tion and molecular identification of each colony type at the sampling time. b1 log CFU/g
Table 2Comparison of the growth of bacterial species during cocoa bean fermentations in stainless steel and wooden box fermenters.
Species identification a Estimated average levels (log CFU/g)b of the isolates at the following fermentation times (h):
Abbreviations: L.: Lactobacillus, W.:Weissella, A.: Acetobacter, G.: Gluconobacter, B.: Bacillus, Mu. Micrococcus: M.: Microbacterium, P. Pantoea, T.: Tatumella, Ac.: Acinetobacter,Ch.: Chryseobacterium, St.: Staphylococcus, K.: Klebsiella, F.: Frateuria.a and b: see Tables 1 and 2.
124 G.V. de Melo Pereira et al. / International Journal of Food Microbiology 161 (2013) 121–133
2.7. Physical–chemical analysis
The ambient temperature and temperatures of the fermenting cocoapulp-bean mass were determined every 2 h with a Delta OHM portabledatalogger, model HD 2105.2, placed in the middle of the cocoa beanmass.
Fig. 1.Microbial succession and temperature dynamics during cocoa bean fermentations peracid bacteria (GYC, ). Temperature inside the stainless steel tank ( ), wooden box 1(▬
For physical–chemical analysis, the beans and pulp were physicallyseparated according to the protocol of Ardhana and Fleet (2003) andNielsen et al. (2007). The aqueous extracts from fermentation sampleswere obtained as described previously (Schwan, 1998). The presenceof alcohols (ethanol and methanol), organic acids (lactic, acetic andcitric acid) and carbohydrates (glucose, sucrose and fructose) were
formed. Microbial cell counts of lactic acid bacteria (MRS, ▀), yeast (YEPG,□) and acetic) and wooden box 2( ). Ambient temperature ( ).
126 G.V. de Melo Pereira et al. / International Journal of Food Microbiology 161 (2013) 121–133
determined for pulp and bean extracts by high-performance liquidchromatography (HPLC) apparatus (HP series 1200, Hewlett-PackardCompany, USA) equipped with an Aminex HPX-87H column (Bio-RadLaboratories, Hercules, USA) that was connected to an RI detector(HPG1362A, Hewlett-Packard Company). All analyses were performedin triplicate; the average values and standard deviations are representedas milligrams per gram of fresh pulp or fermented dry cocoa beans. Thecolumn was eluted with a degassed mobile phase containing 4 mMH2SO4, at 30 °C at a flow rate of 0.6 mL/min.
3. Results
3.1. Cultivation-based quantification of microorganisms
The results of the cultivation-based quantification analyses demon-strated a nearly 2-log increase in the number of yeasts, LAB andAAB dur-ing a period of 12–24 h in all fermentation processes, with the exceptionof AAB growth in samples derived from SST fermentation (Fig. 1). Therewas a marked drop in the post-peak LAB phase followed by a slight in-crease after 84 h of fermentation. In both fermentations performed inboxes, yeasts were present throughout the fermentation process andpeaked at 12 h (7.57 log CFU/g and 8.12 log CFU/g forWB1 andWB2, re-spectively). In contrast, fermentations performed in SST, showedmaximum populations at 24 h (7.69 log CFU/g) and yeasts were notdetected after 108 h. AAB were present throughout WB1 and WB2 fer-mentations, reaching values between 4.79 log CFU/g and 6.36 CFU/g,and between 5.27 log CFU/g and 6.49 log CFU/g, respectively. No AABwere detected in samples obtained at the onset of SST fermentationbut were observed at 24 h (5.47 log CFU/g) and were not present after60 h of fermentation. Although Nutrient agarmediumwas used tomon-itor the growth of aerobic mesophilic bacteria and Bacillus spp., severalgram-negative and aerobic rods that produced catalase without formingspores (later identified as AAB) grew on this medium (data not shown).
3.2. Distribution of microorganisms through cultivation
A total of 1099 isolates were picked up from different culturemedia and characterised in terms of cell morphology and biochemicalfeatures. Of this total, 323 were identified as yeasts, 315 isolates wereidentified as LAB, and 238 were identified as AAB. The remaining 223isolates were recovered from Nutrient agar plates and exhibitedbroad phenotypic diversity. After initial characterisation, 147 yeasts,129 LAB, 104 AAB and 144 NA-isolates were selected according totheir distributions among different times and fermentation processes,as well as their morphological and physiological characteristics, andsubmitted for sequence-based identification (Tables 1 and 2).
S. cerevisiaewas themost prevalent yeast at the start of fermentationand reached maximum populations of 7.69 log CFU/g at 24 h, 7.57 logCFU/g at 12 h and 7.91 log CFU/g at 12 h in SST, WB1 andWB2, respec-tively (Tables 1 and 2). There was a marked drop after 48 h of box fer-mentations in which S. cerevisiae was replaced by non-Saccharomycesyeasts with low-fermentative activity. Conversely, during fermentationconducted in SST, S. cerevisiae dominated the whole fermentation pro-cess and was the sole yeast observed beyond 72 h. The LAB populationwas dominated by L. plantarum and L. fermentum (Tables 1 and 2). Atthe onset of fermentation, L. plantarum dominated the LAB populationin the box fermentations; however, after 36–48 h, L. fermentum becamethe dominant LAB. In regard to SST fermentation, L. fermentum wasinitially present and dominated the entire fermentation process (maxi-mum population size of 7.93 log CFU/g at 48 h)— a position challengedonly by L. plantarum at 72 and 84 h of fermentation (Tables 1 and 2).Occasionally, Weissella sp. isolates were found during initial stages ofSST fermentation. There was a clear prevalence of A. tropicalis amongAAB isolates in all fermentation processes (i.e., more than 50% of AABisolates); this species survived longer during all fermentations withmaximum populations sizes of 5.73 log CFU/g at 48 h, 6.26 log CFU/g
at 12 h, and 6.14 log CFU/g at 24 h in SST, WB1 and WB2, respectively(Table 2). A range of other bacterial species grewonNutrient agar platesthroughout each fermentation process, but these populations rarelyexceeded 5 log CFU/g. In this culturemedium,B. subtilisnearly dominat-ed the box fermentationswhile B.megaterium and B. subtiliswere equal-ly prevalent during SST fermentation. Some bacterial species were onlyisolated at the start of the fermentation processes at 4.00 log CFU/g,such as Bacillus sp. in SST, Tatumella saanichensis, Acinetobacterbaumannii, Brevundimonas sp., Staphylococcus xylosus and Klebsiellavariicola in WB1, and Staphylococcus xylosus and Staphylococcussaprophyticus in WB2 (data not shown).
3.3. Culture-independent microbiological analysis using a nested PCR-DGGE strategy
3.3.1. Comparative sequence analysisIn addition to enumeration of microbial cultures in culture media,
microbial population dynamics were monitored by DGGE. Bacterialand yeast DGGE profiles for each of the three fermentations areshown in Figs. 2 and 3, respectively. In general, bands correspondingto the yeast species S. cerevisiae andHanseniaspora spp. (over 99% iden-tity with H. opuntiae, H. uvarum and H. guilliermondii) and bandscorresponding to the bacterial species L. fermentum and Acetobactersp. (A. tropicalis and A. senegalensis were the closest relatives found bysequence comparison) were found at all fermentation times inboth boxes and SST. This finding supports the results of thecultivation-basedmethod, which establishes these species as dominantthroughout fermentation processes. In regard to bacterial ecology,PCR-DGGE revealed a large diversity of bacterial species within fermen-tations conducted in boxes (Fig. 2). In addition to the dominant speciesL. fermentum and Acetobacter sp., members of the Enterobacteriaceaefamily (T. ptyseos and P. terrea), spore-forming bacteria (B. megateriumand Bacillus sp.), species of Staphylococcus (S. saprophyticus andStaphylococcus sp.), L. plantarum and Acinetobacter sp. were also identi-fied.Moreover, three faint bands corresponding to uncultivable bacteriawere detected. Conversely, only intense bands corresponding to thedominant species, L. fermentum and Acetobacter sp., as well as a faintband corresponding to T. ptyseos, were successfully recovered fromthe DGGE profile of SST fermentation.
As presented in Fig. 3, DGGE profiles with a Eukarya-specific primerwere considerably simpler than their bacterial counterparts.Hanseniaspora sp. was the most abundant species during fermentationprocesses, as revealed through sequencing of the robust DGGE band inall fermentation samples. S. cerevisiae was also detected in all samplesbut the density of its corresponding band was usually low. WB1exhibited a significant increase in the number of bands from 96 h oncorresponding to C. humilis, P. kluyveri, Debaryomyces sp. and Yarrowialipolytica. Moreover, two faint bands corresponding to the fungusWallemia sp., as well as one uncultured Tomentella, were also present.Additionally, P. kluyveri was identified in SST fermentation, and oneband corresponding to the genomic DNA of uncultured ascomycotawas recovered at 72 h during WB1 fermentation.
3.3.2. Cluster analysis of the DGGE profilesBacterial ecology was found to be fermentation-specific (wooden
boxes vs. SST), as samples collected from like fermentations at differenttime points were usually grouped in the same cluster. Thus, fermenta-tions performed in boxes bore greater similarity to each other than toSST fermentation. In fact, samples from SST fermentation formed a sin-gle cluster that was subdivided into two cluster groupings: samplesfrom the start of fermentation (12, 24 and 36 h; similarity of 65%) andsamples obtained during the middle and at the end of fermentation(similarity of 75%). Only samples analysed at 0 h were distinct and didnot belong to any cluster. No clear distinction regarding bacterial ecolo-gy was observed for either box fermentation process, and only samples
Fig. 2. Dendrogram derived from a cluster analysis of the 16S rRNA gene-PCR-DGGE patterns of the bacterial communities associated with cocoa bean fermentation samples fromstainless steel tank, wooden box 1, and wooden box 2 based on the Dice coefficient of similarity (weighted) and generated with the UPGMA clustering algorithm. Prominent bandswere excised from the gels, re-amplified, and sequenced. The closest relatives of the sequenced fragments identified through in GenBank searches for sequences sharing greaterthan 97% similarity: AAB1–AB9 Acetobacter sp; F1–F5 Lactobacillus fermentum; P1–P4 Lactobacillus plantarum; TP1–TP3 Tatumella ptyseos; PT1–PT3 Pantoea terrea; BM1 Bacillusmegaterium; BS1 Bacillus sp. SS1 and SS2 Staphylococcus saprophyticus; SP1 Staphylococcus sp; AC1 Acinetobacter sp.; UL1 Uncultured Lactobacillus sp.; UK1 Uncultured Klebsiellasp; UA1 and UA2 uncultured bacterium isolate from DGGE gel band. N/A=not identified (i.e., bands excised from the gel that were not successfully re-amplified).
127G.V. de Melo Pereira et al. / International Journal of Food Microbiology 161 (2013) 121–133
Fig. 3. Dendrogram derived from a cluster analysis of the ITS-PCR-DGGE patterns of yeast communities associated with cocoa bean fermentation samples from stainless steel tank,wooden box 1, and wooden box 2 based on the Dice coefficient of similarity (weighted) and obtained with the UPGMA clustering algorithm. The closest relatives of the sequencedfragments were determined via GenBank searches for sequences with over 97% similarity: H1-H7 Hanseniaspora sp; S1-S3 Saccharomyces cerevisiae; C1–C4 Candida humilis; P1–P2Pichia kluyveri; D1 Debaryomyces sp; Y1 Yarrowia lipolytica; UT1 uncultured Tomentella; W1 Wallemia sp. N/A=not identified (i.e., bands excised from the gel but not successivelyre-amplified).
128 G.V. de Melo Pereira et al. / International Journal of Food Microbiology 161 (2013) 121–133
collected at the start of WB2 fermentation (0 and 12 h) were identifiedas belonging to an out-group (Fig. 2).
In Fig. 3, the cluster analysis of the yeast DGGE profile is presented.SST fermentation sampleswere distinguishable frombox fermentationsbased on the 40% coefficient of discrimination. Only samples analysed atthe beginning of SST fermentation and those analysed at 144 h duringWB2 fermentation, were distinct from the rest and did not belong toany cluster. WB1 fermentation produced a temporal yeast distribution
with samples divided into two major groups. In contrast, WB2 fermen-tation produced amore stable yeast population throughout the fermen-tation process.
3.4. Temperature
Variations in ambient temperature correlated with day-time andnight-time cycles, ranging from ambient temperatures of 23.93 to
129G.V. de Melo Pereira et al. / International Journal of Food Microbiology 161 (2013) 121–133
32.24 °C and 22.82 to 28.27 °C, respectively (Fig. 1). The temperatureof the bean-pulp mass gradually increased in the SST while it moveddown and up every mixing in the boxes. Box fermentations exceeded50 °C after 35 h of fermentation, which was significantly higher thanthat achieved in SST fermentation. WB1 fermentation reached maxi-mum temperature at 70 h (51.16 °C), followed by WB2 (50.38 °C at137 h) and SST (43.21 °C at 144 h).
3.5. Sugars and fermentation products
To assess the overall metabolic activity occurring in each fermenta-tion process, the formation of free sugars (i.e., sucrose, glucose and fruc-tose) and fermentation products (i.e., ethanol and organic acids) wasassayed in Figs. 4 and 5, respectively. Sucrose was hydrolysed into glu-cose and fructose in the pulp, by yeast invertase activity, and in thebeans, by diffusion of acetic acid, lactic acid, and ethanol into the beansin conjunction with the heat produced therein (Fig. 4). Consumption ofglucose and fructose in the pulp was similar for both box fermentations.The initial glucose concentrations were rapidly consumed up until 36–48 h; afterwards, no significant changes were observed. In contrast,
Fig. 4. Glucose, fructose and sucrose in thepulp and beans fromcocoa bean fermentation sample(● and ). Bars represent standard deviation.
glucose and fructose concentrations in the pulp increased during thefirst 12 h of SST fermentation. After this initial increase, there was amarked decrease that was followed by another increase in glucose andfructose concentrations throughout the remainder of the fermentationperiod. The glucose and fructose profiles inside the beans were similarin all three fermentation processes (Fig. 4).
Ethanol was produced and then consumed (Fig. 5), which was sim-ilar to the trend observed in dynamic yeast populations (Fig. 1). Themaximum ethanol concentration in pulp was achieved during SSTfermentation (71.29 mg/g at 36 h); this was followed by WB1(64.25 mg/g at 12 h) and WB2 fermentations (48.01 mg/g at 60 h)upon conversion into acetic acid by AAB. The concentration of ethanolin the pulp was higher in the SST fermentation from 12 to 96 h thanin either box fermentations, although it reached similar concentrationsat the conclusion of fermentation (an average of 15 mg/g). Ethanol pro-duced in the pulp diffused into the beans where it reached a maximumconcentration of 3.27 mg/g at 72 h of SST fermentation and 3.59 mg/gat 84 h of box fermentations. Acetic acid production in pulp was similarfor both SST and WB1 fermentations; however, after 24 h, the aceticacid concentrations were higher for the SST process (maximum
s derived fromstainless steel tank ( and■),wooden box 1 (▲ and ), andwooden box 2
Fig. 5. Ethanol and organic acids in the pulp and beans from cocoa bean fermentation samples derived from stainless steel tank ( and ■), wooden box 1 (▲ and ) and woodenbox 2 (● and ). Bars represent standard deviation.
130 G.V. de Melo Pereira et al. / International Journal of Food Microbiology 161 (2013) 121–133
concentrations were 57.45 mg/g and 36.20 mg/g at 144 h for SST andWB2, respectively) (Fig. 5). In WB2 fermentation, the concentration ofacetic acid increased during the first 72 h (27.20 mg/g), subsequentlydecreased and then increased again during the last day of fermentation.
Prior to a marked decrease, lactic acid increased in the first 36–48 h,reaching maximum concentrations of 23.86 mg/g, 17.01 mg/g and22.13 mg/g for SST, WB1 and WB2, respectively. Although concentra-tion of lactic acid changed in similar fashion in all three fermentations,
131G.V. de Melo Pereira et al. / International Journal of Food Microbiology 161 (2013) 121–133
its presence in the pulp was significantly higher for SST fermentationfrom 60 to 120 h. Concentrations of 1.16 mg/g, 0.89 mg/g and1.00 mg/g of lactic acid were found in the beans at the end of SST,WB1 and WB2 fermentation processes, respectively. A rapid decreasein citric acid concentrationwas observed after 12 h for all three fermen-tations performed (Fig. 5). The amount of citric acid in the beans wasstable throughout fermentation and was not detected at the conclusionof the process for either box fermentations; in contrast, 2.18 mg/g citricacid was observed in SST fermentation.
4. Discussion
4.1. The microbial ecology of traditional Brazilian spontaneous cocoabean box fermentations
To our knowledge, this study represents the first assessment of yeastisolates from Brazilian cocoa box fermentations using molecularmethods. The culture-dependent and culture-independent approachesindicated the dominance of the species S. cerevisiae and Hanseniasporaspp., followed by low numbers of species belonging to the generaPichia, Candida, Wickerhamomyces, Debaryomyces, Issatchenkia andSchizosaccharomyces; these species have often been detected in sponta-neously fermented cocoa beans from other geographical areas (Schwanet al., 1995; Ardhana and Fleet, 2003; Jespersen et al., 2005; Nielsen etal., 2005; Camu et al., 2007; Papalexandratou et al., 2011). It has beenwidely reported that S. cerevisiae and Hanseniaspora spp. are the mostabundant species in cocoa fermentation (Schwan et al., 1995; Nielsenet al., 2007; Papalexandratou et al., 2011; Pereira et al., 2012b). Inter-estingly, S. cerevisiae yielded only aweak banding in denaturing gels rel-ative to Hanseniaspora spp. (Fig. 4), though it represented the mostcommon yeast species isolated by the plating method for all fermenta-tion processes (Tables 1 and 2). A possible explanation for this may bethat the fragment of the S. cerevisiae ITS region was amplified at lowerefficiency using the protocol described here, relative to other yeast spe-cies present during cocoa fermentation (Nielsen et al., 2005). An accu-rate assessment of the microbial ecology using PCR-DGGE requiresappropriately designed primers, and the use of poorly targeted primerswill skew estimates of the microbial ecology (Pereira et al., 2012b).
Consistent with the results of other cocoa fermentation biodiversitystudies (Camu et al., 2007; Nielsen et al., 2007; Kostinek et al., 2008;Pereira et al., 2012b), the current investigation found that the complexassociations of homo and heterofermentative species, L. plantarum andL. fermentum, respectively, were the most prevalent LAB species inBrazilian cocoa box fermentations. The homolactic metabolism ofL. plantarum has the ability to achieve a high cell density within areasonable fermentation time, and in contrast to L. fermentum, pro-duce high amounts of lactic acid. Conversely, the heterolactic metab-olism of L. fermentum can rapidly convert citric acid and produce nearlyequal masses of lactic acid and acetic acid (Axelsson, 2004). In the for-mer case, L. plantarum contributed to increased acidity, while in the lat-ter, L. fermentum reduced it. However, according to Carr and Davies(1980), carbohydrate catalysis increases total acidity of the cocoamore than breakdown of organic acids.
The culture-dependent microbiological analyses indicated that A.tropicalis may play a significant role during fermentations of Braziliancocoa beans because of its isolation andmetabolic activity during the fer-mentation process. The dominance of this species during cocoa bean fer-mentationmay be explained by its resistance to acidity and heat (Ndoyeet al., 2006). The presence of another representative of Acetobacter andGluconobacter species, previously reported in cocoa bean fermentation(Schwan and Wheals, 2004; Camu et al., 2007), was confirmed in thepresent study. In addition, phylogenetic analysis of the 16S rRNA geneof some isolates, whichwere grown on GYC agar plates during the initialstage of WB1 fermentation, supported that they belonged to the genusAsaia; however, based on their sequence divergence from known spe-cies, these isolates may represent a potentially novel species. The genus
Asaia was introduced with a single species, Asaia bogorensis, as the fifthgenus of the Acetobacteraceae family (Yamada et al., 2000). The naturalhabitats of Asaia spp. are reportedly flowers of the orchid tree (Bauhiniapurpurea), plumbago (Plumbago auriculate), and fermented glutinousrice, all of which occur in hot tropical climates; particularly in Indonesiaand Thailand (Yamada et al., 2000; Katsura et al., 2001). Interestingly,this AAB group has not been reported by any studies of cocoa bean fer-mentation. In contrast to strains of the genera Acetobacter, Asaia strainsare characterised by little or no capacity for oxidation of ethanol intoacetic acid, as well as an inability to grow in 0.35% (v/v) acetic acid(Yamada et al., 2000). Their potential significance for cocoa bean fermen-tation is not known and requires further investigation.
This study represents the first assessment of Bacillus species fromBrazilian cocoa box fermentations usingmolecularmethods. Our resultsdemonstrated that at least three species of Bacillus are involved in cocoafermentation, namely B. subtilis, B. flexus and B. megaterium. The associ-ation of Bacillus species with cocoa fermentation has been recognisedfor some time. Using classical microbiological analysis, Ostovar andKeeney (1973) and Schwan et al. (1986) observed a broad diversity ofBacillus species during fermentation in Trinidad and Brazil, respectively.However, recentmolecular studies have revealed amuch lower diversi-ty, indicating that traditional methods may have misidentified Bacillusspecies and overestimated their abundance in cocoa bean fermentation(Ouattara et al., 2011; this study). The role of Bacillus in cocoa fermen-tation is not well understood, so this bacterium has never been utilisedas a starter culture in an attempt to control the fermentation process.The studies by Zak and Keeney (1976) suggested the involvement ofB. subtilis in the production of tetramethylpyrazines,while other studiesassociated the presence of Bacillus spp. with the occurrence ofoff-flavours that are regularly encountered close to the end of thefermentation, such as C3-C5 free fatty acids and 2,3-butanediol (Lopezand Quesnel, 1973; Schwan et al., 1986). Thus, given their abundancein natural cocoa fermentation (Schwan, 1998; Ardhana and Fleet,2003; Ouattara et al., 2011), it is tempting to speculate that certainBacillus strains might provide beneficial activity by complementingyeast-mediated pulp depectinisation during advanced stages of cocoafermentation (Ouattara et al., 2011).
Some bacterial species, aside from LAB, AAB and Bacillus ssp., werealso detected in Brazilian cocoa box fermentations. The occurrence ofthese previously unreported bacterial species in box cocoa fermenta-tions demonstrates the importance of monitoring the hygiene of fer-mentation procedures to ensure that microbial contamination doesnot spoil the beans (Ardhana and Fleet, 2003). These microorganismscould be associated with pod surfaces, banana/plantain leaves andwith the material used for fermentation, including the porous wood-en structure and farmers' hands. However, these bacteria did notadapt to the matrix during the fermentation process (as indicatedby their absence after 12 h; Tables 1 and 2) and are, therefore, notregarded as important for cocoa fermentation.
Computer-aided analysis of the DGGE profiles showed that bacterialand yeast ecology were fermentation-specific processes (i.e., woodenbox vs. stainless steel tank), as implied by the complex picture displayedon DGGE gels of each box sample. In addition, bacterial DGGE profilesresolved a group of sequences whose phylogenetic allocation indicatedthe presence of uncultivable bacteria belonging to the generaLactobacillus and Klebsiella. Additionally, DGGE profiles obtained withEukarya-specific primers allowed detection of the yeast Yarrowialipolytica, the fungus Wallemia sp., and one uncultured Tomentella dur-ing fermentation in boxes; these strains were not recovered with culti-vation techniques.
4.2. Fermentation performance of stainless steel tank system
Fermentation of cocoa beans continues to be conducted in a tradi-tional way with great diversity in both production methods and theorganoleptic characteristics of their end products. One promising
132 G.V. de Melo Pereira et al. / International Journal of Food Microbiology 161 (2013) 121–133
goal of the chocolate industry is to improve control of the cocoa fer-mentation process to the point that large-scale fermentation canyield high-quality beans. The development of new methods forfermenting cocoa under controlled conditions requires considerablenew research. A promising alternative strategy to conduct continuouscocoa bean fermentation is to use a sterile stainless steel system, inwhich, inoculum, aeration and turn rate can be controlled (Schwanand Wheals, 2004). However, in addition to technology suitable tomeet the conditional requirements of cocoa bean fermentation, athorough understanding of factors that regulate microbial metabo-lism and stability of the microbiota involved in the process are indis-pensable; this is especially true because the quality of chocolate isdictated by these parameters. During the present study, we examinedthe fermentation performance of a novel-design stainless steel tankas an alternative to wooden boxes in cocoa fermentations.
Microbiological analyses revealed that the prevailing speciescreditedwithmajor physiological roles were the same in SST comparedto spontaneous cocoa box fermentations. The most commonyeasts (S. cerevisiae and Hanseniaspora sp.), LAB (L. fermentumand L. plantarum) and AAB (A. tropicalis) cocoa species appearto dominate the cocoa fermentation performed in SST. Thesespecies have been shown to be prevalent during Ghanaian, Brazilianand Ecuadorian spontaneous cocoa bean fermentations in laboratoryand pilot-scale cocoa fermentations (Lefeber et al., 2011; Pereira et al.,2012b). The key community dynamics and metabolic activity of thesewell-adapted species remained unchanged from box fermentations,which entailed proper chemical changes in both pulp and cocoa nibs.These similarities suggested that stainless the steel tankprovided a suit-ablemodel system for larger-scale fermentations, at leastwith regard tomicrobial ecology and activity.
It was shown through metabolite target analyses that similar sub-strate consumption (sucrose, glucose, fructose and citric acid) andmetabolite production kinetics (ethanol, lactic acid and acetic acid)occurred in the SST compared to wooden boxes fermentation pro-cesses. However, higher concentrations of ethanol were recoveredduring SST fermentation. The SST system markedly altered the pro-portion of yeasts relative to box fermentations by decreasing theirpopulation size and diversity. Consequently, the presence of a widevariety of low-fermentative-power non-Saccharomyces yeasts signifi-cantly decreased the ethanol yield in the box conditions, relative toSST; this may have been the result of competition with S. cerevisiaestrains for available nutrients. The amount of acetic acid produceddid not correspond to the number of AAB present. Despite a smallerAAB population, acetic acid was more effectively produced in SST rel-ative to each of the box conditions. This indicates that the acetic acidproduction was more heavily influenced by the yeast-AAB-mediatedconversion of carbohydrates into ethanol, and subsequent conversionof ethanol into acetic acid, than by the AAB population size. The LABgrowth occurred normally in SST fermentation, therefore the dynam-ics of citric acid consumption and lactic acid production exhibited aprofile similar to that of box fermentations. However, high concentra-tions of lactic acid from 48 to 120 h of SST fermentation might alsohave influenced in the reduction of microbial diversity during this fer-mentation process.
The purpose of the SST system was to enhance hygiene and mini-mize initial contamination of the beans so that any starter culturemight have a positive effect, or even, a spontaneous fermentation canoccur more homogeneously in both terms of kinetic and quality underfield conditions. In fact, the aseptic-design of stainless steel tank offersnew perspectives to conduct cocoa fermentations since it enables theutilization of either chemical sterilizing agents or steam sterilizationover long fermentation batches. The suppression of Enterobacteriaceaeand other cocoa-unrelated bacterial species was demonstrated by SSTsystem, as determined by plating on Nutrient agar plates and DGGEanalysis. On the other hand, it seems that the high thermal conductivityand geometry of the SST system resulted in a marked thermal loss that
overcame the heat produced by the oxidation-process. To avoidheat transfer, a stainless steel tank surrounded by rubber or a thermo-statically controlled insulating water jacket can be used (Schwan andWheals, 2004).
5. Conclusion
This research has contributed to a better understanding of the mi-crobial community structure and metabolite production associatedwith cocoa bean fermentation in boxes and stainless steel tank. Theuse of cultivation-independent techniques and cultivation-based ap-proaches to the study of bacterial and yeast communities has confirmedthat microbial ecosystems in box fermentations of Brazilian cocoa sup-port a wide variety of organisms. In addition, we described the first ex-perimental validation of the stainless steel tank method for cocoa beanfermentation. The use of stainless steel tanksmay be of great interest forthose who seek improved control over the cocoa fermentation processand/or to optimise cocoa fermentation through the use of starter cul-tures. Futurework involving sensory quality of cocoa nibs, cocoa liquorsand the resulting chocolates are required prior to implementation ofstainless steel tank in field conditions.
Acknowledgments
The authors wish to thank the technical assistance of Alício Costa,Iara Brito and Ivani Maria Gervásio. GVMP, EGA and KTM thank theBrazilian agencies Conselho Nacional de Desenvolvimento Científicoe Tecnológico (CNPQ), Fundação de Amparo a Pesquisa do Estado deMinas Gerais (FAPEMIG) and Coordenação de Aperfeiçoamento dePessoal de Nível Superior (CAPES) for scholarships. Financial supportfrom the Mars Center for Cocoa Science, Itajuípe, Brazil (Masterfoods)is also gratefully acknowledged.
References
Ardhana, M., Fleet, G., 2003. The microbial ecology of cocoa bean fermentations in In-donesia. International Jounal of Food Microbiology 86, 87–99.
Axelsson, L., 2004. Lactic acid bacteria: classification and physiology, In: Salminen, S.,von Wright, A., Ouwehand, A. (Eds.), Lactic Acid Bacteria: Microbiological andFunctional Aspects, 3rd edn. Marcel Dekker, New York, pp. 1–66.
Camu, N., De Winter, T., Verbrugghe, K., Cleenwerck, I., Vandamme, P., Takrama, J.S.,Vancanneyt, M., De Vuys, t L., 2007. Dynamics and biodiversity of populations oflactic acid bacteria and acetic acid bacteria involved in spontaneous heap fermen-tation of cocoa beans in Ghana. Applied and Environmental Microbiology 73,1809–1824.
Carr, J.G., Davies, P.A., 1980. Cocoa fermentation in Ghana and Malaysia: Part 2. Furthermicrobiological methods and results. A report available at Long Ashton ResearchStation, Long Ashton, Bristol, England.
Garcia-Armisen, T., Papalexandratou, Z., Hendryckx, H., Camu, N., Vrancken, G., DeVuyst, L., Cornelis, P., 2010. Diversity of the total bacterial community associatedwith Ghanaian and Brazilian cocoa bean fermentation samples as revealed by a16 S rRNA gene clone library. Applied Microbiology and Biotechnology 87,2281–2292.
Jespersen, L., Nielsen, D.S., Hønholt, S., Jakobsen, M., 2005. Occurrence and diversity of yeastsinvolved in fermentation ofWest African cocoa beans. FEMS Yeast Research 5, 441–453.
Katsura, K., Kawasaki, H., Potacharoen, W., Saono, S., Seki, T., Yamada, Y., Uchimura, T.,Komagata, K., 2001. Asaia siamensis sp. nov., an acetic acid bacterium in the a-Proteobacteria. International Journal of Systematic and Evolutionary Microbiology51, 559–563.
Kostinek, M., Ban-Koffi, L., Ottah-Atikpo, M., Teniola, D., Schillinger, U., Holzapfel, W.H.,Franz, C.M.A.P., 2008. Diversity of predominant lactic acid bacteria associated withcocoa fermentation in Nigeria. Current Microbiology 56, 306–314.
Kurtzman, C., Fell, J., 2011. The Yeasts, a Taxonomic Study, 5th ed. Elsevier, Amsterdam.Lagunes-Gálvez, S., Loiseau, G., Paredes, J.L., Barel, M., Guiraud, J.P., 2007. Study on the
microflora and biochemistry of cocoa fermentation in the Dominican Republic. In-ternational Jounal of Food Microbiology 114, 124–130.
Leal Jr., G.A., Gomes, L.H., Efraim, P., Tavares, F.C., Figueira, A., 2008. Fermentation ofcocoa (Theobroma cacao L.) seeds with a hybrid Kluyveromyces marxianus strainimproved product quality attributes. FEMS Yeast Research 8, 788–798.
Lefeber, T., Papalexandratou, Z., Gobert, W., Camu, N., De Vuyst, L., 2011. On-farm imple-mentation of a starter culture for improved cocoa bean fermentation and its influ-ence on the flavour of chocolates produced thereof. Food Microbiology 30, 379–392.
Lima, L.J.R., Almeida, M.H., Rob Nout, M.J., Zwietering, M.H., 2011. Theobroma cacao L.,“The Food of the Gods”: Quality Determinants of Commercial Cocoa Beans, withParticular Reference to the Impact of Fermentation. Critical Reviews in FoodScience and Nutrition 51, 731–761.
133G.V. de Melo Pereira et al. / International Journal of Food Microbiology 161 (2013) 121–133
Lopez, A., Quesnel, V.C., 1973. Volatile fatty acid production in cacao fermentation andthe effect on chocolate flavor. Journal of the Science of Food and Agriculture 24,319–326.
Magalhães, K.T., Pereira, G.V.M., Dias, D.R., Schwan, R.F., 2010. Microbial communitiesand chemical changes during fermentation of sugary Brazilian kefir. World Journalof Microbiology and Biotechnology 26, 1241–1250.
Ndoye, B., Lebecque, S., Dubois-Dauphin, R., Tounkara, L., Guiro, A.T., Kere, C., Diawara,B., Thonart, P., 2006. Thermoresistant properties of acetic acids bacteria isolatedfrom tropical products of Sub-Saharan Africa and destined to industrial vinegar.Enzyme and Microbiology Technology 39, 916–923.
Nielsen, D.S., Honholt, S., Tano-Debrah, K., Jespersen, L., 2005. Yeast populations asso-ciated with Ghanaian cocoa fermentations analysed using denaturing gradientgel electrophoresis (DGGE). Yeast 22, 271–284.
Nielsen, D.S., Teniola, O.D., Ban-Koffi, L., Owusu, M., Andersson, T.S., Holzapfel, W.H.,2007. The microbiology of Ghanaian cocoa fermentations analysed using culture-dependent and culture independent methods. International Jounal of Food Micro-biology 114, 168–186.
Ostovar, K., Keeney, P.G., 1973. Isolation and characterization of microorganisms in-volved in the fermentation of Trinidad's cacao beans. Journal of Food Science 38,611–617.
Ouattara, H.G., Reverchon, S., Niamke, S.L., Nasser, W., 2011. Molecular identificationand pectate lyase production by Bacillus strains involved in cocoa fermentation.Food Microbiology 28, 1–8.
Ovreas, L., Forney, L., Daae, F.L., Torsvik, V., 1997. Distribution of bacterioplankton inmeromictic lake Saelenvannet, as determined by denaturing gradient gel electro-phoresis of PCR-amplified gene fragments coding for 16S rRNA. Applied and Envi-ronmental Microbiology 63, 3367–3373.
Papalexandratou, Z., Vranckena, G., De Bruyneb, K., Vandammeb, P., De Vuys, L., 2011.Spontaneous organic cocoa bean box fermentations in Brazil are characterized by arestricted species diversity of lactic acid bacteria and acetic acid bacteria. Food Mi-crobiology 28, 1326–1338.
Pereira, G.V.M., Ramos, C.L., Galvão, C., Dias, E.S., Schwan, R.F., 2010. Use of specific PCRprimers to identify three important industrial species of Saccharomyces genus: Sac-charomyces cerevisiae, Saccharomyces bayanus and Saccharomyces pastorianus. Let-ters in Applied Microbiology 5, 131–137.
Pereira, G.V.M., Magalhães, K.T., Lorenzetii, E.R., Souza, T.P., Schwan, R.S., 2012a. Amultiphasic approach for the identification of endophytic bacterial in strawberry fruitand their potential for plant growth promotion. Microbiology Ecology 63, 405–417.
Pereira, G.V.M., Miguel, M.G.C.P., Ramos, C.L., Schwan, R., 2012b. Microbiological andpysicochemical characterization of small-scale cocoa fermentations and screeningof yeast and bacteria strains to develop a defined starter culture. Applied andEnvironmental Microbiology 78, 5395–5405.
Ruiz, A., Poblet, M., Mas, A., Guillamón, J.M., 2000. Identification of acetic acid bacteriaby RFLP of PCR-amplified 16 rDNA and 16S–23S rDNA intergenic spacer. Interna-tional Journal of Systematic and Evolutionary Microbiology 50, 1981–1987.
Samah, O.A., Ptih, M.F., Selamat, J., 1992. Biochemical changes during fermentation ofcocoa beans inoculated with Saccharomyces cerevisiae (wild strain). Journal ofFood Science and Technology 29, 341–343.
Schwan, R.F., 1998. Cocoa fermentations conducted with a defined microbial cocktailinoculum. Applied and Environmental Microbiology 64, 1477–1483.
Schwan, R.F., Wheals, A.E., 2004. The microbiology of cocoa fermentation and its role inchocolate quality. Critical Reviews in Food Science and Nutrition 44, 205–221.
Schwan, R.F., Vanetti, M.C.D., Silva, D.O., Lopez, A., Moraes, C.A., 1986. Characterizationand distribution of aerobic, spore-forming bacteria from cacao fermentations inBahia. Journal of Food Science 51, 1583–1584.
Schwan, R.F., Rose, A.H., Board, R.G., 1995. Microbial fermentation of cocoa beans, withemphasis on enzymatic degradation of the pulp. Journal of Applied Bacteriology -Symposium Supplement 79, 96S–107S.
Schwan, R.F., Cooper, R.M., Wheals, A.E., 1997. Endopolygalacturonase secretion byKluyveromyces marxianus and other cocoa pulp-degrading yeasts. Enzyme and Mi-crobiology Technology 21, 234–244.
White, T.J., Bruns, T., Lee, S., Taylor, J., 1990. Amplification and direct sequencing of fun-gal ribosomal RNA genes for phylogenetics. In: Innis, M.A., Gelfand, D.H., Sninsky,J.J., White, T.J. (Eds.), PCR Protocols: A Guide to Methods and Applications. Aca-demic Press, San Diego, CA, USA, pp. 315–322.
Yamada, Y., Katsura, K., Kawasaki, H., Widyastuti, H., Saono, S., Seki, T., Uchimura, T.,Komagata, K., 2000. Asaia bogorensis gen. nov. sp. nov., an unusual acetic acid bac-terium in the a-proteobacteria. International Journal of Systematic and Evolution-ary Microbiology 50, 823–829.
Zak, D.L., Keeney, P.G., 1976. Changes in cocoa proteins during ripening of fruit, fer-mentation, and further processing of cocoa beans. Journal of Agricultural andFood Chemistry 24, 483–486.
