Food Microbiology 28 (2011) 964e973

Contents lists available at ScienceDirect

Food Microbiology
journal homepage: www.elsevier.com/locate/fm

Comparison of the bacterial species diversity of spontaneous cocoa bean

fermentations carried out at selected farms in Ivory Coast and Brazil
Zoi Papalexandratou a, Nicholas Camu b, Gwen Falony a, Luc De Vuyst a, *
a
Research Group of Industrial Microbiology and Food Biotechnology (IMDO), Faculty of Sciences and Bio-engineering Sciences, Vrije Universiteit Brussel (VUB),
Pleinlaan 2, B-1050 Brussels, Belgium
b
Innovation Fermentation, Barry Callebaut Belgium, Aalstersestraat 122, B-9280 Lebbeke, Belgium

a r t i c l e i n f o a b s t r a c t

Article history: To compare the spontaneous cocoa bean fermentation process carried out in different cocoa-producing
Received 16 August 2010 regions, heap and box (one Ivorian farm) and box (two Brazilian farms) fermentations were carried out. All
Received in revised form fermentations were studied through a multiphasic approach. In general, the temperature inside the fer-
19 November 2011
menting mass increased throughout all fermentations and reached end-values of 42e48
C. The main end-
Accepted 25 November 2011
Available online 1 February 2011
products of pulp carbohydrate catabolism were ethanol, lactic acid, acetic acid, and/or mannitol. In the case
of the fermentations on the selected Ivorian farm, the species diversity of lactic acid bacteria (LAB) and
acetic acid bacteria (AAB) was restricted. Lactobacillus fermentum and Leuconostoc pseudomesenteroides
Keywords:
Cocoa
were the predominant LAB species, due to their ethanol and acid tolerance and citrate consumption. The
Fermentation levels of mannitol, ascribed to growth of L. fermentum, were fermentation-dependent. Also, enterobacterial
Diversity species, such as Erwinia soli and Pantoea sp., were among the predominating microbiota during the early
PCR-DGGE stages of both heap and box fermentations in Ivory Coast, which could be responsible for gluconic acid
Metabolite target analysis production. Consumption of gluconic acid at the initial phases of the Ivorian fermentations could be due to
yeast growth. A wider microbial species diversity throughout the fermentation process was seen in the case
of the box fermentations on the selected Brazilian farms, which differed, amongst other factors, regarding
pod/bean selection on these farms as compared to fermentations on the selected Ivorian farm. This
microbiota included Lactobacillus plantarum, Lactobacillus durianis, L. fermentum, Lactobacillus mali,
Lactobacillus nagelii, L. pseudomesenteroides, and Pediococcus acidilactici, as well as Bacillus subtilis that was
present at late fermentation, when the temperature inside the fermenting mass reached values higher than
50
C. Moreover, AAB seemed to dominate the Brazilian box fermentations studied, explaining higher acetic
acid concentrations in the pulp and the beans. To conclude, it turned out that the species diversity and
community dynamics, influenced by local operational practices, in particular pod/bean selection, impact
the quality of fermented cocoa beans.
Ó 2011 Elsevier Ltd. All rights reserved.

1. Introduction farm in different regions of the equatorial zone are necessary, as to

Cocoa beans are the principal raw material for chocolate
production. However, their quality depends on several factors, not
only with respect to crop cultivation but also concerning post-
harvest processing (Wood and Lass, 2001). Recently, the demand
and interest of chocolate manufacturers for high quality cocoa
beans has expanded, as fine cocoa varieties are increasingly popular
and health benefits of cocoa and chocolate are getting known
(Visioli et al., 2009). Therefore, comparative studies of the cocoa
bean fermentation process regarding operational practices on the
* Corresponding author. Tel.: þ32 2 6293245; fax: þ32 2 6292720.
E-mail address: ldvuyst@vub.ac.be (L. De Vuyst).
assess their impact on cocoa bean quality.
Ghana remains the world producer of the best quality bulk
cocoa, thanks to excellent post-harvest handling of cocoa beans by
Ghanaian farmers (Baker et al., 1994). Ivory Coast is the world’s
leading exporter of bulk cocoa (40% of the world cocoa production;
www.icco.org). Brazil used to be one of the biggest cocoa producers
in the world, but nowadays, it represents only 5% of the annual
world cocoa production, because of its reputation to produce cocoa
of inferior quality and as many of its plantations were destroyed by
witches’ broom disease (Passos et al., 1984; Wood and Lass, 2001;
Leiter and Harding, 2004).
Cocoa bean fermentation is the main stage in cocoa post-harvest
processing. It is generally carried out in a traditional manner
through a spontaneous four- to six-day fermentation process

0740-0020/$ e see front matter Ó 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.fm.2011.01.010
 Z. Papalexandratou et al. / Food Microbiology 28 (2011) 964e973
(Wood and Lass, 2001; Schwan and Wheals, 2004; De Vuyst et al.,
2010), either in heaps, boxes, baskets, trays, or on platforms.
Successful cocoa bean fermentation requires a succession of activ-
ities of different microorganisms, which are prevalent in the
surrounding environment (cocoa pod surfaces, leaves, pollen,
insects, baskets, knives, etc.) and accidentally contaminate the
sterile fresh cocoa pulp-bean mass (pulp and cocoa beans) after
manual opening of the cocoa pods at the fields (Schwan and
Wheals, 2004; De Vuyst et al., 2010). During the initial phases of
the cocoa bean fermentation process, there is a colonization by
yeasts (depectinization and ethanol formation), lactic acid bacteria
(LAB, citric acid fermentation and lactic acid production), and acetic
acid bacteria (AAB, oxidation of ethanol into acetic acid and over-
oxidation of acetic acid and lactic acid into carbon dioxide and
water) (Schwan et al., 1995; Ardhana and Fleet, 2003; Camu et al.,
2007; Nielsen et al., 2007). During fermentation, ethanol and ace-
tic acid diffuse into the beans, and this, in combination with the
heat produced along fermentation, causes the death of the seed
embryo as well as the end of the fermentation. As a result of these
microbial activities and diffusions, complex physical processes and
biochemical reactions are initiated in the beans to form the
required flavor and color precursors (Schwan and Wheals, 2004;
Thompson et al., 2007; De Vuyst et al., 2010).
Most recent studies deal with the identification and metabolic
activities of the yeast and bacterial communities associated with
cocoa bean fermentations in Ghana and Indonesia (Ardhana and
Fleet, 2003; Jespersen et al., 2005; Nielsen et al., 2005, 2007;
Camu et al., 2007, 2008; Daniel et al., 2009). Some of these and
in particular earlier studies of the cocoa bean fermentation process
elsewhere are based on culture-dependent methods for microbi-
ological analysis, including traditional cultivation methods in
combination with phenotypic and/or genotypic identification of
the isolates. However, classical plating methodologies only allow
the isolation of a limited number of microbial taxa (Giraffa, 2004;
Cocolin and Ercolini, 2008). The application of molecular biological
techniques to detect and identify microorganisms through the use
of certain molecular markers, for instance the 16S rRNA gene in
the case of bacteria, is more frequently used now, such as dena-
turing gradient gel electrophoresis (DGGE) and temperature
gradient gel electrophoresis (TGGE), as these techniques do not
always require prior isolation of the microorganisms (Ercolini,
2004; Giraffa, 2004; Cocolin and Ercolini, 2008; Justé et al.,
2008). Moreover, these methods allow a combined qualitative
and semi-quantitative approach of the characterization of micro-
bial communities involved in complex microbial ecosystems.
Alternatively, molecular techniques are not free of biases and/or
limitations (Sekiguchi et al., 2001; Holben et al., 2004; Justé et al.,
2008). In general, culture-independent methods are of assistance
to traditional methods for a successful overall study of the
community dynamics and species diversity associated with
complex matrices, such as the fermenting cocoa pulp-bean mass
(Nielsen et al., 2005, 2007; Camu et al., 2007, 2008). Also, they
allow to follow the dynamics of microbial communities as a func-
tion of time, cultivable or not and applying cultivation methods
or not.
The aim of the present study was to monitor and analyze the
species diversity of the bacteria involved in traditional spontaneous
cocoa bean fermentations in two important cocoa-producing
regions, namely Lagunes (Ivory Coast) and Bahia (Brazil), focusing
on the impact of different methods of fermentation (in particular
heap and box) and concomitant operational practices on the farm
on the quality of fermented (dry) beans. The study was based on
a culture-independent approach solely to monitor LAB and AAB
species in parallel with metabolite target analyses of whole
fermentation samples. In addition, a comparison of these
965
fermentations with spontaneous cocoa bean heap fermentations
performed in the New Tafo region of Ghana (Camu et al., 2007,
2008) was made.
2. Materials and methods
2.1. Cocoa bean fermentations
To compare the cocoa bean fermentation process carried out in
different cocoa-producing regions, four spontaneous fermenta-
tions were performed in parallel during the main crop of 2006
(OctobereNovember 2006), namely two in Ivory Coast (Barry
Callebaut-Ivory Coast, Abidjan, Lagunes) and two in Brazil (Barry
Callebaut-Brasil, Ilhéus, Bahia). Cocoa pods, mostly from Criollo
and Forastero hybrid cocoa trees, were harvested by traditional
methods and opened within one to three days. In Ivory Coast, one
heap (H) and one box (B) fermentation were performed at the
same farm, with cocoa pods harvested from the plantation of that
farm and carefully selected while opening (only healthy pods were
used). For the heap fermentation, 150 kg of fresh cocoa pulp-
bean mass was piled into a heap onto plantain leaves laid on the
ground and the heap was covered with additional plantain leaves.
For the box fermentation, 1200 kg of fresh cocoa pulp-bean mass
was placed in a wooden box (1.0 m  1.0 m  1.0 m) and covered
with plantain leaves and jute sacks. In Brazil, two box fermenta-
tions (B1, B2) of 1500 kg of fresh cocoa pulp-bean mass derived
from cocoa pods harvested from two different plantations were
performed at the corresponding farms. Healthy and infected pods
were not separated. Wooden boxes of 1.0 m x 1.0 m  1.0 m and
1.2 m  1.2 m  1.0 m, respectively, were used. During all fermen-
tations, 500-g samples were taken at certain time points, namely
at the start of the fermentation (0 h) and after 6, 24, 30, 48, 54, 72,
78, 96, 102, 120, 126, 144, and 150 h of fermentation in Ivory Coast
and after 6, 12, 18, 24, 30, 36, 42, 48, 54, 60, 66, 72, 84, 96, 120, and
144 h of fermentation in Brazil. Temperature and pH (only for the
Brazilian box fermentations) of the fermenting cocoa pulp-bean
mass were measured at the sampling times. The environmental
temperature was measured online with temperature recorders
(DeltaTRAK, Pleasanton, CA, USA). In each region, each set of
fermentations was performed at the same time to exclude the
influence of weather conditions. Natural sun-drying of the fer-
mented cocoa beans was carried out on cane platforms for
between five and ten days, depending on the weather conditions,
and samples were taken at the end of the drying process. All
samples were cooled, frozen, and sent on dry ice to Belgium for
a multiphasic analysis approach.
2.2. Extraction of DNA from cocoa bean fermentation samples
In Belgium, direct extraction of DNA from the cocoa bean
fermentation samples was performed. Collection and lysis of the
cells and phenolechloroform extraction of their DNA were
according to the protocol of Camu et al. (2007). To get rid of cocoa
pulp compounds potentially inhibiting polymerase chain reactions
(PCR), such as polysaccharides, proteins, enzymes, and poly-
phenols, a NucleoSpin column (Macherey Nagel GmbH, Düren,
Germany) was used for further purification of the DNA-containing
aqueous phase, following the manufacturer’s instructions. The
final DNA samples (50 ng ml1) were stored at 20
C until further
use.
2.3. 16S rRNA-PCR-DGGE
A pair of LAC primers that targets the V3eV4 region of the 16S
rRNA gene was used to amplify DNA from LAB species (LAC1eLAC2;
966 Z. Papalexandratou et al. / Food Microbiology 28 (2011) 964e973
Walter et al., 2001) and a pair of universal primers (357fe518r;
Ercolini et al., 2001) was used to amplify DNA of the V3 region of
the 16S rRNA gene of bacteria in general. A GC-rich sequence
(50 -CGCCCGCCGCGCCCCGCGCCCGGCCCGCCGCCCCCGCCCC-30 ) was
attached to the LAC2 and 357f primers to prevent early separation
of double stranded DNA during DGGE. PCR conditions and DGGE
analyses were as reported previously (Camu et al., 2007;
Vasilopoulos et al., 2008). Gel processing, DGGE band pattern
cluster analysis (with the band-based Dice coefficient), control of
the purity of DNA from excised bands, DNA band sequencing, and
BLAST (Basic Local Alignment Search Tool) analysis of the nucleo-
tide sequences in the GenBank database (http://www.ncbi.nlm.nih.
gov/BLAST) were performed as described previously (Camu et al.,
2007).
2.4. Metabolite target analysis
2.4.1. Sample preparation
Frozen cocoa bean fermentation samples as well as samples of
fermented dry cocoa beans were used to prepare aqueous extracts
for metabolite target analysis, following the protocol described
previously (Camu et al., 2007). All sample preparations and anal-
yses mentioned below were performed in triplicate and the mean
values  standard deviations are presented.
2.4.2. High performance anion exchange chromatography (HPAEC)
To measure the concentrations of lactic acid, acetic acid, citric
acid, and gluconic acid in the aqueous extracts, HPAEC with
conductivity under ion suppression (CIS) was applied, using an ICS
3000 chromatograph (Dionex, Sunnyvale, CA, USA), equipped with
an AS-19 column (Dionex), as described previously (Camu et al.,
2007). The concentrations of sucrose, glucose, fructose, and
mannitol in the aqueous extracts were determined by HPAEC with
pulsed amperometric detection (PAD; Dionex), using an ICS 3000
chromatograph equipped with a CarboPacÔ PA10 column (Dionex),
as described previously (Camu et al., 2007). To prepare the samples,
the aqueous extracts (300 ml) were treated with 900 ml of acetoni-
trile for protein removal, microcentrifuged (16,060g, 15 min, 4
C),
appropriately diluted, and filtered (0.2 mm, Minisart RC4; Sartorius,
Goettingen, Germany) into 0.3-ml microvials with slitted screw
caps (VWR International, Darmstadt, Germany) prior to injection
(10 ml). Samples were run together with the appropriate external
standards.
2.4.3. Gas chromatography (GC)
The concentrations of ethanol present in the aqueous extracts
were measured using a gas chromatograph (Focus GC; Interscience,
Breda, The Netherlands), equipped with an AS 3000 autosampler,
a Stabilwax-DA column (Restek, Bellefonte, PA, USA), and a flame
ionization detector (FID). Hydrogen gas was used as carrier gas with
a constant flow rate of 1 ml min1 and nitrogen gas was used as
make-up gas. The injection volume was 1 ml and the injection was
performed in split mode with a split ratio of 20:1. The injector and
detector temperatures were set at 150
C and 250
C, respectively.
The column temperature was programmed as follows: 0 min,
40
C; 10 min, 140
C; 12 min, 230
C; and 22 min, 230
C. Data
processing was performed using the ChromCard for Windows
(version 1.18) software (Thermo Electro Corp., Milan, Italy). Before
injection, samples of aqueous extracts were treated with four
volumes of acetonitrile to remove proteins. After centrifugation
(16,060g, 15 min, 4
C), the supernatant was filtered (0.2 mm,
Minisart RC4) into 1.5-ml glass vials (Agilent Technologies, Santa
Clara, CA, USA) with screw caps (VWR International) prior to
injection. External standards of ethanol were run before and after
the samples.
3. Results
3.1. Physical changes
3.1.1. Ivory Coast
The temperature of the fermenting cocoa pulp-bean mass
gradually increased throughout the H and B fermentations, from an
initial 23e24
C to 42.5
C and 48
C, respectively, after 117 and
96 h of fermentation, respectively. The ambient day and night
temperatures were 25e32.5
C and 22e25
C, respectively.
3.1.2. Brazil
The ambient day and night temperatures during the two box
fermentations were 28e32
C and 21e24
C, respectively. The
temperature of the fermenting cocoa pulp-bean mass increased
from an initial 26e30
C to 50
C after 72 h of fermentation, fol-
lowed by a slight decrease to 48
C in the case of the B2 fermen-
tation, while a drop to 42
C occurred during the B1 fermentation.
In both cases, the pH of the fresh pulp was about 2.7  0.1. The pH
pattern of the fermenting cocoa pulp-bean mass was similar for
both fermentations, increasing gradually, and reaching a value of
4.6 and 3.8 at the end of the B1 and B2 fermentations, respectively
(data not shown).
3.2. Culture-independent microbiological analysis through 16S
rRNA-PCR-DGGE
3.2.1. Ivory Coast
The DGGE profiles (LAC primers) of the H (Fig. 1) and B (data not
shown) fermentations showed a restricted LAB species diversity
involved in cocoa bean fermentation. The B fermentation was
dominated by Lactobacillus fermentum after 30 h. The intensity of
the DGGE band of L. fermentum increased between 48 and 78 h and
was visible until the end of the B fermentation. The H fermentation
showed a slightly different profile, with Leuconostoc pseudome-
senteroides dominating the first 54 h, thereafter being replaced by
L. fermentum. L. fermentum was detectable until the end of the H
fermentation, whereas the band of L. pseudomesenteroides was only
visible during the first 78 h (Fig. 1).
Using universal primers, bands of the LAB communities were
visible at the upper part of the gels and corresponded with the LAB
species found in the DGGE band patterns for the LAC primers at the
same time points (Fig. 2). A strong band in the middle of the gel in
the initial phases of all fermentations showed 100% identity through
BLAST analysis with chloroplast plant DNA (Fig. 2). This band was
found to be responsible for a decrease of the intensity of the L.
pseudomesenteroides band in the DGGE gels (universal primers), due
to contamination of the PCR reaction mixtures with plant DNA.
Erwinia soli and Pantoea sp. were detected during the first 54 h of the
H and B fermentations, the bands being more intensive in the case of
the H fermentation (Fig. 2). Sequencing of band (xii) in Fig. 2
revealed the presence of AAB (Acetobacter) after 78 h and upon in
the case of the B fermentation. BLAST analysis showed 100% identity
with more than one species of Acetobacter (e.g., Acetobacter pas-
teurianus, Acetobacter lovaniensis, Acetobacter syzygii), indicating
that these Acetobacter species shared the sequence of the V3 region
of the 16S rRNA gene that was amplified with the pair of universal
primers. This was confirmed by co-migration of the three Aceto-
bacter species used in the reference ladder (band f), namely A.
pasteurianus LMG 1262T, Acetobacter ghanensis LMG 23848T, and
Acetobacter fabarum LMG 24244T, whereas the DGGE band of Ace-
tobacter senegalensis LMG 23690T (band g) was visible at the bottom
of the gel (lane L in Fig. 2).
Cluster analysis of the PCR-DGGE profiles (LAC primers)
revealed an influence of the method of fermentation, carried out in
 Z. Papalexandratou et al. / Food Microbiology 28 (2011) 964e973 967

Fig. 1. PCR-DGGE profiles with LAC1eLAC2 primers (35e60% denaturant gradients) of amplified 16S rRNA gene fragments of the LAB present in the cocoa bean fermentation
samples taken from the Ivorian heap and the Brazilian box fermentations. The reference ladder (L) consisted of (a) L. plantarum LMG 6907T; (b) L. acidophilus LMG 9433T; (c) L.
fermentum LMG 6902T; (d) L. mesenteroides subsp. mesenteroides LMG 6893T; (e) P. acidilactici LMG 11384T; and (f) L. casei LMG 6904T. The lane numbers represent the time (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) L. plantarum (100%; EU637402.1); (ii) and (iii) L. mali (99%; AB326352.1); (iv) L. durianis (100%; AJ315640.1); (v) L. mali (100%;
AB326352.1); (vi) L. fermentum (100%; EU688978.1); (vii) L. nagelii (99%; AB370876.1); (viii) L. pseudomesenteroides (100%; DQ523483.1); (ix) P. acidilactici (100%; GU904688.1); and
(x) L. nagelii (100%; AB370876.1).

the same region, on the dynamics of the LAB communities (data not
shown). No similarity was shown in the LAB communities between
the B and H fermentations during the first three days of fermen-
tation, whereas 100% similarity was observed during the last three
days, indicating final dominance of the cocoa bean fermentations
by the same LAB species. Comparison of the DGGE profiles (LAC
primers) of the H fermentation with representative Ghanaian heap
fermentations carried out before (Camu et al., 2007, 2008) showed
a high similarity (80%e100%) between samples of the heap
fermentations performed in different regions of West Africa
(Fig. 3A), indicating the dominance of West African cocoa bean
fermentations by the same LAB species. Cluster analysis of PCR-
DGGE profiles obtained with universal primers revealed three
clusters, representing the beginning (0e24 h), the middle

Fig. 2. PCR-DGGE profiles with 357f-518r universal primers (35e70% denaturant gradients) of the amplified 16S rRNA gene fragments of the bacteria present in the cocoa bean
fermentation samples taken from the Ivorian heap and the Brazilian box fermentations. The reference ladder (L) consisted of (a) L. plantarum LMG 6907T; (b) W. ghanensis 215 (Camu
et al., 2007); (c) L. fermentum LMG 6902T; (d) E. casseliflavus 555 (Camu et al., 2007); (e) L. pseudomesenteroides 274 (Camu et al., 2007); (f) A. pasteurianus LMG 1262T/A. ghanensis
LMG 23848T/A. fabarum LMG 24244T; and (g) A. senegalensis LMG 23690T. The lane numbers represent the time (h) of sampling during the 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) L. plantarum
(100%; EU637402.1); (ii) L. durianis (100%; AJ315640.1); (iii) L. fermentum (100%; EU688978.1); (iv) and (vi) L. mali (98%; AB326352.1); (v) L. pseudomesenteroides (100%;
DQ523483.1); (vii), (viii), (xii), and (xiii) AAB (100%; GQ246703.1 - HM190252.1 - AB485746.1); (ix) L. nagelii (98%; AB370876.1); (x) L. nagelii (100%; AB370876.1); (xi) chloroplast
DNA (100%; DQ231562.1); (xiv) B. subtilis (100%; GQ392055.1); (xv) P. acidilactici (100%; GU904688.1); (xvi) Erwinia soli (100%; EF540893.1); (xvii) Pantoea sp. (100%; FJ560468.1); (?)
no satisfied sequencing.
968 Z. Papalexandratou et al. / Food Microbiology 28 (2011) 964e973

Fig. 3. Dendrogram derived from the cluster analysis of PCR-DGGE profiles of the bacterial communities associated with the Ivorian heap H and box B fermentations, the Brazilian
box B2 fermentation, and Ghanaian heap (H3, H4, H5, H7, H10, H11, H12) fermentations (Camu et al., 2007, 2008) based on the Dice coefficient of similarity (weighted) and obtained
through an UPGMA clustering algorithm. (A) Comparison of the same fermentation method in West Africa (Ghana and Ivory Coast). (B) Shifts of the bacterial communities during
cocoa bean fermentation. (C) Influence of operational practices. Clusters are based on the PCR-DGGE profiles with LAC1eLAC2 primers (A and C) or universal primers (B).

(30e72 h), and the end (78e126 h) of the B fermentation, respec-
tively (Fig. 3B). Comparable results were found for the H fermen-
tation (data not shown).
3.2.2. Brazil
PCR-DGGE (LAC primers) of the B1 and B2 fermentations
revealed a wide LAB species diversity (Fig. 1). Despite the large
number of bands in each lane, and especially at the initial phases of
the fermentations, all bands were separated and no co-migration
occurred. However, due to the presence of several LAB bands
without notable differences in intensities, a dominant species could
not be indicated in general. Only the B1 fermentation showed
a band of high intensity corresponding with Lactobacillus nagelii
between 18 and 66 h of fermentation, while L. fermentum domi-
nated the LAB communities at 30 h into fermentation. In general,
the same LAB species were determined during the two box
fermentations, namely Lactobacillus. plantarum, Lactobacillus mali,
Lactobacillus durianis, L. nagelii, L. fermentum, Pediococcus acid-
ilactici, and L. pseudomesenteroides (Fig. 1)
Using universal primers, LAB bands were visible in the upper part
of the DGGE patterns (Fig. 2). Chloroplast plant DNA was detected in
the initial phases of the Brazilian box fermentations as well (Fig. 2).
AAB were found in both box fermentations, but could not be
discriminated at species or genus level according to NCBI data.
A strong AAB band was visible throughout the B1 and B2 fermen-
tations (Fig. 2), with higher intensity between 48 and 72 h (during
the B2 fermentation), underlining the dominance of AAB during
these fermentations. Also, a band corresponding with Bacillus sub-
tilis was detected after 144 h in both fermentations (Fig. 2).
Cluster analysis of the PCR-DGGE patterns (LAC primers) of the
B1 and B2 fermentations revealed no influence of the plantation on
the microbiota involved. However, comparison of the B2 fermen-
tation with the Ivorian B fermentation through cluster analysis of
the PCR-DGGE profiles showed no similarity between these two
fermentations, indicating an apparent influence of operational
practices on the same method of fermentation (Fig. 3C).
3.3. Metabolite target analysis
Fig. 4e6 show representative results of the kinetics of residual
and produced compounds (carbohydrates, mannitol, ethanol, and
organic acids) during the Ivorian H and Brazilian B2 fermentations.
 Z. Papalexandratou et al. / Food Microbiology 28 (2011) 964e973 969

Fig. 3. (continued).

3.3.1. Ivory Coast
3.3.1.1. Pulp. Fresh pulp contained glucose, fructose, and citric acid,
but no sucrose. A fast consumption of the pulp glucose and fructose
occurred, being almost exhausted after 50 h, during both H and B
fermentations (Fig. 4). A rapid decrease of the citric acid concen-
tration was noticed after 24 h and 48 h for the B and H fermenta-
tions, respectively (Fig. 5), paralleling with the dominance of the
fermentations by L. fermentum (Fig. 1). In both cases, mannitol was
slightly produced after 30 h of fermentation, reaching a maximum
concentration of 7.0  0.09 mg g1 after 54 h into the B and
1.7  0.03 mg g1 after 96 h into the H fermentations (Fig. 4). The
initial level of gluconic acid in the pulp was approximately
1.5 mg g1. It disappeared after 54 h during both fermentations,
followed by an increase after 80 h during the B and after 102 h
during the H fermentation (Fig. 5). The initial level of pulp acetic
acid was approximately 4.0 mg g1. In the case of the B fermenta-
tion, it increased slightly after 30 h together with the presence of
L. fermentum (data not shown), followed by a pronounced increase
after 72 h, paralleling with AAB detection (data not shown),
reaching a maximum level of 24.1  0.50 mg g1 after 126 h of
970 Z. Papalexandratou et al. / Food Microbiology 28 (2011) 964e973

70 70
Ivory Coast, heap (H) Brazil, Box 2 (B2)
60 60

50 50

40 40
mg/g

mg/g
30 30

20 20

10 10

0 0
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160
Time (h) Time (h)

Fig. 4. Course of glucose (C), fructose (:), and mannitol (-) in the pulp during the Ivorian heap and the Brazilian box fermentations.

fermentation. The level of acetic acid was stable during the first
48 h of the H fermentation (Fig. 6), which corresponded with
a dominance of the LAB communities by L. pseudomesenteroides
(Fig. 1). This was followed by a slight increase until 96 h of
fermentation, when L. pseudomesenteroides was replaced by L. fer-
mentum. Finally, it increased to a maximum of 15.1  0.14 mg g1
after 126 h of fermentation, paralleling AAB growth (Figs. 6 and 2,
respectively). The production of lactic acid in the pulp started at 6 h
and 24 h into the H and B fermentations, respectively, and corre-
lated with the presence of LAB (Figs. 5 and 1, respectively). The
lactic acid concentration decreased after 78 h of the B fermentation,
but it followed a linear increase during the H fermentation. During
the B fermentation, ethanol was produced in the pulp during the
first 72 h, to reach a maximum concentration of 12.9  0.26 mg g1,
followed by a decrease through conversion into acetic acid by the
AAB (data not shown). The maximum amount of ethanol
(20.0  0.49 mg g1) was produced faster (30 h) in the pulp during
the H fermentation compared to the B fermentation (Fig. 6).
3.3.1.2. Beans. The concentration of sucrose in the fresh beans was
9.6  0.10 mg g1. Sucrose was hydrolyzed into glucose and fructose
in the beans, the levels of which increased after 54 h and reached
a maximum of 3.0  0.50 mg g1 and 2.5  0.30 mg g1, respec-
tively, in the H and B fermentations. Mannitol was not found in the
beans. The initial level of acetic acid was approximately 4.0 mg g1,
as in the pulp, and it reached a maximum concentration faster in
the case of the B fermentation compared to the H fermentation.
Lactic acid remained at low levels in the beans throughout the H
and B fermentations. The diffusion of ethanol into the beans started
after 6 h, with a rapid increase after 24 h, followed by a linear
decrease to zero at the end of the fermentations. The maximum
concentration of ethanol in the beans was found after 54 h in the
B (12.9  0.70 mg g1) and H (15.2  0.26 mg g1) fermentations.
3.3.1.3. Dry beans. Metabolite target analysis performed on the
sun-dried, fermented beans revealed a full evaporation of ethanol.
The levels of fructose and glucose were lower in the fermented dry
beans of the H fermentation compared to the B fermentation and
almost no mannitol was found in either case. No gluconic acid and
almost no lactic acid were present either. Citric acid and acetic acid
concentrations were 4.6  0.72 mg g1 and 10.2  0.31 mg g1,
respectively, in the H fermentation. Acetic acid in the dried beans of
the B fermentation achieved a rather high concentration of
12.9  0.27 mg g1 and the citric acid concentration was
7.8  0.14 mg g1.
3.3.2. Brazil
3.3.2.1. Pulp. The fresh pulp consisted of glucose, fructose, and
citric acid, and no sucrose. After 24 h of the B1 fermentation,
glucose and fructose reached almost zero levels, whereas their
concentration in the pulp of the B2 fermentation remained low
after 66 h (Fig. 4). The citric acid concentration decreased to
approximately 1.0 mg g1 after 60 h of fermentation in both cases,
paralleling with the presence of LAB (Figs. 5 and 1, respectively).
However, the presence of AAB from the beginning of these
fermentations (Fig. 2) could also be responsible for carbohydrate
consumption. The concentration profile of gluconic acid in the pulp
showed fluctuations throughout both fermentations and reached

20 20
Ivory Coast, heap (H) Brazil, Box 2 (B2)
18 18
16 16
14 14
12 12
mg/g

mg/g

10 10
8 8
6 6
4 4
2 2
0 0
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160
Time (h) Time (h)

Fig. 5. Course of lactic acid (D), citric acid (B), and gluconic acid (,) in the pulp during the Ivorian heap and the Brazilian box fermentations.
 Z. Papalexandratou et al. / Food Microbiology 28 (2011) 964e973 971

30 30
Ivory Coast, heap (H) Brazil, Box 2 (B2)
27 27
24 24
21 21

18 18

mg/g
mg/g

15 15

12 12

9 9

6 6

3 3

0 0
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160
Time (h) Time (h)

Fig. 6. Course of ethanol (A) and acetic acid (>) in the pulp during the Ivorian heap and the Brazilian box fermentations.

high concentrations (approximately 19 mg g1) during the first
96 h of B2 fermentation. Mannitol was slightly produced in the pulp
after 96 h and 42 h of the B1 and B2 fermentations, respectively
(Fig. 4). A small increase of the acetic acid concentration
was noticed during the first 24 h and 12 h of the B1 and B2
fermentations, respectively, followed by a marked increase to
28.4  0.71 mg g1 (B1, after 72 h) and 27.9  0.39 mg g1 (B2, after
48 h; Fig. 6), paralleling with the appearance of AAB (Fig. 2). Then,
a drop of the acetic acid concentration was found and after 96 h it
stabilized to 11.0  1.50 mg g1 (Fig. 6). The profile of lactic acid
production by LAB in the pulp (Fig. 5) was the same for both
fermentations, but the concentrations were lower for B1. The
concentration of lactic acid increased the first 60e66 h, followed by
a remarkable drop; there was again an increase in the concentra-
tion during the last two days of the B2 fermentation (Fig. 5). In the
B1 fermentation, ethanol was produced during the first 24 h,
reaching a maximum concentration of 22.8  1.07 mg g1, when its
conversion into acetic acid by AAB started, to decrease almost to
1.0 mg g1 after 84 h of fermentation. In the B2 fermentation,
19.9  1.61 mg g1 of ethanol was produced, which was completely
disappeared after 48 h (Fig. 6).
3.3.2.2. Beans. The initial level of sucrose in the beans was
9.5  0.05 and 12.6  0.04 mg g1 for the B1 and B2 fermentations,
respectively; it was hydrolyzed into glucose and fructose
thereafter, reaching maximum concentrations of 4.0  0.22 and
4.0  0.05 mg g1, respectively, after 96 h of fermentation. Gluconic
acid and mannitol were absent in the beans of both fermentations.
The concentration of acetic acid in the fresh beans was
4.0  0.50 mg g1 and it was stable during the first 24 and 30 h for
the B1 and B2 fermentations, respectively. As in the pulp, a notable
increase followed this stable phase, reaching high acetic acid
concentrations in the beans after 54 h of fermentation
(22.4  1.15 mg g1 for the B1 and 19.0  0.01 mg g1 for the B2
fermentation), indicating diffusion of the pulp acetic acid into the
beans. A slight but almost linear increase of the concentration of
lactic acid was found. After 24 h, the ethanol concentration was at
a maximum in the beans of the B1 (7.7  0.39 mg g1) and B2
(5.2  0.33 mg g1) fermentations.
3.3.2.3. Dry beans. Similar levels of residual carbohydrates and
organic acids were observed in the sun-dried, fermented beans of
the B1 and B2 fermentations. Gluconic acid was at a rather high
concentration of 3.1  0.12 mg g1 to 4.8  0.07 mg g1 in the beans
of both fermentations and must have originated from the dried
pulp on the shell of the dry beans. Also, residual fructose and
mannitol were at high levels in the beans of both fermentations. No
ethanol was present in the fermented dry beans.
4. Discussion
Cocoa bean fermentation has a serious impact on the formation
of the characteristic flavor and color precursors of cocoa (Schwan
and Wheals, 2004; De Vuyst et al., 2010). In addition, fermenta-
tion practices such as method and duration of the fermentation
may influence the nature and course of local cocoa bean fermen-
tations (Wood and Lass, 2001). Therefore, extensive studies are
going on to determine the influence of external factors on the
fermentation process and to improve traditional practices, so that
the best final products can be achieved (Schwan, 1998; Nielsen
et al., 2007; Camu et al., 2007, 2008). The present paper
compared heap and box fermentations carried out within the same
or a different cocoa-producing region through a combination of
a culture-independent microbiological approach (16S rRNA-PCR-
DGGE) and metabolite target analysis of cocoa pulp-bean mass
samples taken during fermentation. Although PCR-DGGE of cocoa
bean fermentation processes can only give a general estimation of
the size of the microbial communities as a function of time, its
unique combination with metabolite target analysis of the
ecosystem as a whole enables the study of the dynamics of the
fermentation process, also in the absence of culturing results, as
the quantitative evolution of the compounds measured can be
linked with the dynamics of the main microbial groups/species
responsible for their consumption or production.
A successful cocoa bean fermentation process is characterized
by a correct succession of the activities of three different groups of
microorganisms, namely yeasts, LAB, and AAB, being responsible
for the production of mainly ethanol, lactic acid, and acetic acid,
respectively (Schwan and Wheals, 2004). This mainly depends on
the composition and quality of the fresh cocoa pulp and is made
possible through changes in chemical composition, pH, tempera-
ture, and oxygen tension of the fermenting cocoa pulp-bean mass
(Camu et al., 2007, 2008). Similarities were seen in the composition
of Ivorian and Ghanaian fresh pulp (present study; Camu et al.,
2007). Although no influence could be seen of the plantation on
the main cocoa bean fermentation process (Camu et al., 2007),
which was confirmed during the present study, an impact of
operational practices (separation of healthy and infected pods/
beans, and method and duration of the fermentation, which
differed in the cases investigated), beyond bean mixing (Camu
et al., 2008), was found. Hence, the present study underlined
the importance of good fermentation practices such as careful
972 Z. Papalexandratou et al. / Food Microbiology 28 (2011) 964e973
pod/bean selection, amongst other factors, to get successfully fer-
mented cocoa beans, independent of method and region. In Brazil,
the fresh pulps with low initial pH reflected the use of immature
cocoa pods (high citric acid concentrations), which resulted in
atypical microbial successions and a high temperature inside the
fermenting cocoa pulp-bean mass (>50
C) too fast into fermen-
tation, indicating early and intensive ethanol oxidation by AAB. The
environmental temperature did not influence the course of the
fermentations directly. In contrast, environmental contamination
did influence the process. Whereas a restricted LAB species diver-
sity (L. pseudomesenteroides and L. fermentum) was seen in the case
of a fermentation carried out with care, e.g. regarding pod/bean
selection, as exemplified by the Ivorian heap and box fermentations
during the present study and the Ghanaian cocoa bean heap
fermentations reported elsewhere (Camu et al., 2007, 2008; Nielsen
et al., 2007), a wide LAB species diversity was seen in the case of the
box fermentations at the selected Brazilian farms that were carried
out with less care (amongst others, no separation of healthy and
infected pods/beans). In the latter case, 16S rRNA-PCR-DGGE (both
LAC and universal primers) revealed the presence of L. plantarum,
L. durianis, L. fermentum, L. mali, L. nagelii, L. pseudomesenteroides,
and P. acidilactici. A wide LAB species diversity has been found
before in Brazilian box fermentations (Passos et al., 1984; Schwan
et al., 1995). It could be responsible for the unpleasant flavor of
Brazilian cocoa, as both homo- and heterofermentative LAB
occurred during fermentation, besides enhanced concentrations of
acetic acid produced by AAB. Poor pod/bean selection as a cause of
a wide LAB species diversity is of course an operational practice on
the farm that is not necessarily region-dependent, but will depend
on the farms visited in the region, and hence there will be places
with good pod/bean selection in Brazil and lack of it in Ivory Coast
too.
Yeasts and LAB consumed carbohydrates, thereby producing
ethanol, lactic acid, and CO2. Additionally, LAB were responsible for
the assimilation of citric acid supported by the succession of DNA
bands corresponding with relevant LAB species in the DGGE gels
and taking into account that prevailing cocoa-specific yeasts do not
ferment citrate (Camu et al., 2007; Daniel et al., 2009). Alterna-
tively, oxidative metabolism of lactic acid and citric acid by yeasts
may occur. In general, fermentations carried out in West Africa are
dominated by few LAB species, being represented by L. pseudome-
senteroides in the beginning and strictly heterofermentative L. fer-
mentum throughout fermentation, as shown for the Ivorian heap
fermentation of the present study and for Ghanaian heap fermen-
tations as well (Nielsen et al., 2007; Camu et al., 2007, 2008). The
latter species was mainly responsible for the assimilation of citric
acid, as during the presence of L. pseudomesenteroides in the Ivorian
heap fermentation the concentration of citric acid remained stable.
Also, L. fermentum was responsible for mannitol production, using
fructose as alternative external electron acceptor (Camu et al.,
2008). In addition, acetic acid production was delayed when
L. pseudomesenteroides and L. fermentum were prevalent, indicating
the importance of the succession of bacterial activities by dedicated
species to obtain fermented cocoa beans of high quality. As L. fer-
mentum did not dominate the two Brazilian box fermentations,
citric acid was consumed slowly in these cases. Whereas most
species detected in the Brazilian box fermentations have been
found before, L. durianis and L. nagelii were not. The fact that
L. durianis was even visible during six days of the B1 box fermen-
tation showed its acid and ethanol tolerance.
Universal primers were used in the present study, as Camu et al.
(2007) have reported that WBAC primers do not allow detection of
AAB species in cocoa bean fermentation samples. However, recent
studies of Garcia-Armisen et al. (2010) have shown that the use of
WBAC primers can unravel the AAB species diversity in cocoa bean
fermentation samples through 16S rRNA-PCR-DGGE, when DNA of
low concentration is subjected to PCR amplification. As such, AAB
were present from the beginning of the Brazilian box fermentations
and their enhanced growth was responsible for an enhanced
ethanol-to-acetic acid oxidation, compared with heap fermenta-
tions. Although 16S rRNA-PCR-DGGE with universal primers did
not allow the detection of Gluconacetobacter species, Garcia-
Armisen et al. (2010) have revealed the presence of these species
throughout Brazilian box fermentations by constructing a 16S rDNA
clone library. Also, Schwan and Wheals (2004) have shown the
involvement of Glucon(acet)obacter species in Brazilian cocoa bean
box fermentations before. Hence, AAB could be involved in carbo-
hydrate consumption in the beginning of the fermentations too. In
all cases, a considerable increase of acetic acid concentrations took
place when AAB bands became visible in the DGGE gels. Although
co-migration of DNA bands representing several AAB species
occurred, it is likely that A. pasteurianus dominated the heap (Ivory
Coast) and box (Ivory Coast and Brazil) fermentations, as it was the
predominating AAB species during Ghanaian cocoa bean heap
fermentations (Camu et al., 2007, 2008). Actually, it is the main AAB
species during several cocoa bean fermentations worldwide (De
Vuyst et al., 2010).
High concentrations of acetic acid in pulp and beans, obtained
early into fermentation, are typical for cocoa bean box fermenta-
tions from Brazil (this study; Schwan et al., 1995) and the Domin-
ican Republic (Lagunes-Gálvez et al., 2007) as compared with cocoa
bean heap fermentations from Ghana (Camu et al., 2007) and Ivory
Coast (this study), suggesting a high acidity of fermented dry beans
derived from boxes. However, in the case of a wide LAB species
diversity, representing a fermentation carried out with less care as
mentioned above, acetic acid production could be ascribed to
a combined action of LAB and AAB, in particular in the beginning of
the fermentation. Less effective fermented dry beans were hence
characterized by relatively high lactic acid and citric acid
concentrations.
The increased AAB growth led to a high fermentation temper-
ature at the end of the Brazilian box fermentations, hence
explaining the appearance of B. subtilis at the end of a cocoa bean
fermentation, as reported before (Ardhana and Fleet, 2003). The
role of Bacillus species, which come into fermentation when the
temperature is high, is not clear. Former studies have reported on
their possible impact on off-flavor development (Schwan and
Wheals, 2004). However, Garcia-Armisen et al. (2010) did not
retrieve any Bacillus clones in cocoa bean fermentation samples of
Ghana and Brazil. In contrast, Ouattara et al. (2008) have reported
on low Bacillus counts during cocoa bean fermentation in Ivory
Coast. The microenvironment of the fermenting cocoa pulp-bean
mass may not be appropriate enough for optimal growth and
activity of bacilli. In general, it seems that cocoa-associated LAB and
AAB species are highly adapted to the microenvironment of the
cocoa bean fermentation and its nutrient availability, although
under laboratory conditions their growth requires specific condi-
tions (low incubation temperature, neutral pH).
Interestingly, enterobacterial species, encompassing Pantoea
sp. and E. soli, were found during the early stages of both the heap
and box fermentations in Ivory Coast, possibly derived from the
environmental soil or plant material (Kageyama et al., 1992; Pujol
and Kado, 2002). The occurrence of bands of high intensity in the
DGGE gels corresponding with Enterobacteriaceae during the first
day of the Ivorian fermentations could indicate their involvement
in the initiation of the fermentations, for instance contributing to
depectinization of the pulp and/or breakdown of pulp citrate. In
addition, the increased production of gluconic acid in the beginning
and at the end of the Ivorian cocoa bean fermentations could be
ascribed to enterobacterial and fungal species, respectively, both
 Z. Papalexandratou et al. / Food Microbiology 28 (2011) 964e973
being reported for conversion of carbohydrates (mainly glucose)
into gluconic acid (Adachi et al., 2003; Magnuson and Lasure,
2004). The dominance of the Brazilian box fermentations by
several LAB and AAB species seemed to have avoided the detection
of enterobacterial species through 16S rRNA-PCR-DGGE. Indeed,
these species could be present in low numbers and hence fall under
the detection limit of 16S rRNA-PCR-DGGE. However, Tatumella
sp. and Erwinia tasmaniensis were among the clones obtained by
Garcia-Armisen et al. (2010) in the case of Brazilian box fermen-
tation samples, indicating their possible role in gluconic acid
production. Consumption of gluconic acid, as in the beginning of
the Ivorian cocoa bean fermentations, could be the result of yeast
activities, as Saccharomyces cerevisiae e a yeast species involved in
cocoa bean fermentation (Daniel et al., 2009) e has been reported
to assimilate gluconic acid (Peinado et al., 2003).
In conclusion, this study revealed that a wide LAB species
diversity, enhanced growth of AAB, and the presence of Bacillus
species in the later phase of the cocoa bean fermentation process
may be responsible for the inferior character of Brazilian cocoa
beans fermented in boxes compared with West African cocoa beans
fermented in heaps, as a result of different operational practices on
the farm and not as a result of the box-type process, at least for the
selected farms involved in this study. Consequently, fermentation
practices such as careful pod/bean selection seem to be required to
obtain fermented dry beans of high quality, whatever the tech-
niques used or the region produced in, provided raw material of
high quality is used and treated with care. Also, the present study
unraveled the bacterial species diversity of Ivorian cocoa bean
fermentations for the first time. Finally, it showed the usefulness of
the unique combination of a fast, culture-independent microbio-
logical approach and multiple metabolite target analysis to monitor
cocoa bean fermentations successfully, independent of the
fermentation method or region. Such an approach, can be of
industrial value, for fermented food processes in general. A good and
clear understanding of the total microbial species diversity of cocoa
bean fermentations worldwide through a combination of classical
and molecular techniques forms the basis to control and even steer
this fermentation process with appropriate starter cultures.
Acknowledgments
The research was funded by the Research Council of the Vrije
Universiteit Brussel, the Fund for Scientific Research-Flanders, the
Flemish Institute for the Encouragement of Scientific and Techno-
logical Research in the Industry (Contracts 0400433 and 050818),
and Barry Callebaut N.V. The authors thank the Centre National
de Recherche Agronomique (Abidjan, Ivory Coast) for their
cooperation.
References
Adachi, O., Moonmangmee, D., Toyama, H., 2003. New developments in oxidative
fermentation. Appl. Microbiol. Biotechnol. 60, 643e653.
Ardhana, M., Fleet, G., 2003. The microbial ecology of cocoa bean fermentations in
Indonesia. Int. J. Food Microbiol. 86, 87e99.
Baker, D.M., Tomlins, K.I., Gray, C., 1994. Survey of Ghanaian cocoa farmer
fermentation practices and their influence on cocoa flavour. Food Chem. 51,
425e431.
Camu, N., De Winter, T., Verbrugghe, K., Cleenwerck, I., Vandamme, P., Takrama, J.S.,
Vancanneyt, M., De Vuyst, L., 2007. 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, 1809e1824.
Camu, N., González, Á, De Winter, T., Van Schoor, A., De Bruyne, K., Vandamme, P.,
Takrama, J.S., Addo, S.K., De Vuyst, L., 2008. Influence of turning and environ-
mental 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, 86e98.
Cocolin, L., Ercolini, D., 2008. Molecular Techniques in the Microbial Ecology of
Fermented Foods. Springer, New York.
973
Daniel, H.M., Vrancken, G., Takrama, J.F., Camu, N., De Vos, P., De Vuyst, L., 2009.
Yeast diversity of Ghanaian cocoa bean heap fermentations. FEMS Yeast Res.
9, 774e783.
De Vuyst, L., Lefeber, T., Papalexandratou, Z., Camu, N., 2010. The functional role of
lactic acid bacteria in cocoa bean fermentation. In: Mozzi, F., Raya, R.R.,
Vignolo, G.M. (Eds.), Biotechnology of Lactic Acid Bacteria: Novel Applications.
Wiley-Blackwell, Ames, pp. 301e326.
Ercolini, D., Moschetti, G., Blaiotta, G., Coppola, S., 2001. Behavior of variable V3
region from 16S rDNA of lactic acid bacteria in denaturing gradient gel elec-
trophoresis. Curr. Microbiol. 42, 199e202.
Ercolini, D., 2004. PCR-DGGE fingerprinting: novel strategies for detection of
microbes in food. J. Microbiol. Methods 56, 297e314.
Garcia-Armisen, T., Papalexandratou, Z., Hendryckx, H., Camu, N., Vrancken, G., De
Vuyst, L., Cornelis, P., 2010. Diversity of the total bacterial community asso-
ciated with Ghanaian and Brazilian cocoa bean fermentation samples as
revealed by a 16 S rRNA gene clone library. Appl. Microbiol. Biotechnol.
87, 2281e2292.
Giraffa, G., 2004. Studying the dynamics of microbial populations during food
fermentation. FEMS Microbiol. Rev. 28, 251e260.
Holben, W.E., Feris, K.P., Kettunen, A., Apajalahti, J.H., 2004. GC fractionation
enhances microbial community diversity assessment and detection of minority
populations of bacteria by denaturing gradient gel electrophoresis. Appl.
Environ. Microbiol. 70, 2263e2270.
Jespersen, L., Nielsen, D.S., Hønholt, S., Jakobsen, M., 2005. Occurrence and diversity
of yeasts involved in fermentation of West African cocoa beans. FEMS Yeast Res.
5, 441e453.
Justé, A., Thommad, B.P.H.J., Lievens, B., 2008. Recent advances in molecular tech-
niques to study microbial communities in food-associated matrices and
processes. Food Microbiol. 25, 745e761.
Kageyama, B., Nakae, M., Yagi, S., Sonoyama, T., 1992. Pantoea punctata sp. nov.,
Pantoea citrea sp. nov., and Pantoea terrea sp. nov. isolated from fruit and soil
samples. Int. J. Syst. Bacteriol. 42, 203e210.
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. Int. J. Food Microbiol. 114, 124e130.
Leiter, J., Harding, S., 2004. Trinidad, Brazil, and Ghana: three melting moments in
the history of cocoa. J. Rural Stud. 20, 113e130.
Magnuson, J., Lasure, L., 2004. Organic acid production by filamentous fungi. In:
Tkacz, J.S., Lange, L. (Eds.), Advances in Fungal Biotechnology for Industry,
Agriculture and Medicine. Kluwer/Plenum, New York, pp. 307e340.
Nielsen, D.S., Hønholt, S., Tano-Debrah, K., Jespersen, L., 2005. Yeast populations
associated with Ghanaian cocoa fermentations analysed using denaturing
gradient gel electrophoresis (DGGE). Yeast 22, 271e284.
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. Int. J. Food Microbiol.
114, 168e186.
Ouattara, H.G., Koffi, B.L., Karou, G.T., Sangaré, A., Niamke, S.L., Diopoh, J.K., 2008.
Implication of Bacillus sp. in the production of pectinolytic enzymes during
cocoa fermentation. World J. Microbiol. Biotechnol. 24, 1753e1760.
Passos, F.M.L., Lopez, A.S., Silva, D.O., 1984. Aeration and its influence on the
microbial sequence in cacao fermentation in Bahia, with emphasis on lactic acid
bacteria. J. Food Sci. 49, 1470e1474.
Peinado, R.A., Moreno, J.J., Ortega, J.M., Mauricio, J.C., 2003. Effect of gluconic acid
consumption during simulation of biological aging of sherry wines by a flor
yeast strain on the final volatile compounds. J. Agric. Food Chem.
51, 6198e6203.
Pujol, C.J., Kado, C.I., 2002. Genetic and biochemical characterization of the pathway
in Pantoea citrea leading to pink disease of pineapple. J. Bacteriol. 182,
2230e2237.
Schwan, R.F., Rose, A.H., Board, R.G., 1995. Microbial fermentation of cocoa beans,
with emphasis on enzymatic degradation of the pulp. J. Appl. Bacteriol. Symp.
Suppl. 79, 96Se107S.
Schwan, R.F., 1998. Cocoa fermentations conducted with a defined microbial
cocktail inoculum. Appl. Environ. Microbiol. 64, 1477e1483.
Schwan, R.F., Wheals, A.E., 2004. The microbiology of cocoa fermentation and its
role in chocolate quality. Crit. Rev. Food Sci. Nutr. 44, 205e221.
Sekiguchi, H., Tomioka, N., Nakahara, T., Uchiyama, H., 2001. A single band does not
always represent single bacterial strains in denaturing gradient gel electro-
phoresis analysis. Biotechnol. Lett. 23, 1205e1208.
Thompson, S.S., Miller, K.B., Lopez, A.S., 2007. Cocoa and coffee. In: Doyle, M.P.,
Beuchat, L.R. (Eds.), Food Microbiology: Fundamentals and Frontiers, third ed.
ASM Press, Washington, pp. 837e849.
Vasilopoulos, C., Ravyts, F., De Maere, H., De Mey, E., Paelinck, H., De Vuyst, L.,
Leroy, F., 2008. Evaluation of the spoilage lactic acid bacteria in modified
atmosphere-packaged artisan-type cooked ham using culture-dependent and
culture-independent approaches. J. Appl. Microbiol. 104, 1341e1353.
Visioli, F., Bernaert, H., Corti, R., Ferri, C., Heptinstall, S., Molinari, E., Poli, A.,
Serafini, M., Smit, H.J., Vinson, J.A., Violi, F., Paoletti, R., 2009. Chocolate, lifestyle,
and health. Crit. Rev. Food Sci. Nutr. 49, 299e312.
Walter, J., Hertel, C., Tannock, G.W., Lis, C.M., Munro, K., Hammes, W.P., 2001.
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, 2578e2585.
Wood, G.A.R., Lass, R.A., 2001. Cocoa, fourth ed. Blackwell Science, Oxford.