OUP user menu

Occurrence and diversity of yeasts involved in fermentation of West African cocoa beans

Lene Jespersen, Dennis S. Nielsen, Susanne Hønholt, Mogens Jakobsen
DOI: http://dx.doi.org/10.1016/j.femsyr.2004.11.002 441-453 First published online: 1 February 2005

Abstract

Samples of cocoa beans were taken on two separate occasions during heap and tray fermentations in Ghana, West Africa. In total 496 yeast isolates were identified by conventional microbiological analyses and by amplification of their ITS1-5.8S rDNA-ITS2 regions. For important species the identifications were confirmed by sequencing of the D1/D2 domain of the 5′ end of the large subunit (26S) rDNA. Assimilations of organic acids and other carbon compounds were conducted. For dominant yeasts intraspecies variations were examined by determination of chromosome length polymorphism (CLP) using pulsed-field gel electrophoresis.

For the heap fermentations maximum yeast cell counts of 9.1 × 107 were reached, whereas maximum yeast counts of 6.0 × 106 were reached for the tray fermentations. Candida krusei was found to be the dominant species during heap fermentation, followed by P. membranifaciens, P. kluyveri, Hanseniaspora guilliermondii and Trichosporon asahii, whereas Saccharomyces cerevisiae and P. membranifaciens were found to be the dominant species during tray fermentation followed by low numbers of C. krusei, P. kluyveri, H. guilliermondii and some yeast species of minor importance. For isolates within all dominant species CLP was evident, indicating that several different strains are involved in the fermentations. Isolates of C. krusei, P. membranifaciens, H. guilliermondii, T. asahii and Rhodotorula glutinis could be found on the surface of the cocoa pods and in some cases on the production equipment, whereas the origin of e.g. S. cerevisiae was not indicated by the results obtained.

In conclusion, the results obtained show that fermentation of cocoa beans is a very inhomogeneous process with great variations in both yeast counts and species composition. The variations seem to depend especially on the processing procedure, but also the season and the post-harvest storage are likely to influence the yeast counts and the species composition.

Keywords
  • Cocoa bean fermentation
  • Microbial succession
  • D1/D2 sequencing
  • Strain typing

1 Introduction

Cocoa beans are seeds from the fruit pods of the tree Theobroma cacao Linné[1], which is cultivated in plantations in tropical regions throughout the world, West Africa being the major producing region accounting in year 2001–2002 for more than 66% of the world production [2]. In the pods the cocoa beans are embedded in a mass of mucilaginous pulp and after removal of both beans and pulp from the pods the first step in cocoa processing is a spontaneous fermentation [3,4]. The methods of fermentation vary considerably from country to country and even adjacent farms may differ in their processing practices [5]. The cocoa beans are either fermented in heaps, boxes, baskets or in trays. In West Africa, fermentation in heaps varying in sizes from 20 to 1000 kg and covered with banana or plantain leaves is by far the most dominant method [1]. Another method used, which is often claimed to give a better cocoa quality, is fermentation in stacked trays (∼20–100 kg of beans per tray), giving series of thin layers of cocoa beans with air circulating between each layer [6]. Reviews on cocoa fermentation have been given previously [1,79].

The mucilaginous bean pulp is rich in fermentable sugars such as glucose, 2.4–4.1% (w/w), fructose, 4.2–6.2% (w/w) and sucrose, 2.1–3.2% (w/w) and has a high concentration of especially citric acid, 2.1–2.4% (w/w), but also smaller amounts of other organic acids are present such as lactic acid, 0.03% (w/w) and acetic acid, 0.04% (w/w) [4,10,11]. The pH of the pulp has been reported to be just below 4.0 [12]. During fermentation, microbial activity leads to the formation of a range of metabolic end-products such as alcohols, acetic acid and other organic acids, which diffuse into the beans and cause their death. This induces biochemical transformations within the beans that lead to formation of precursors of the characteristic aroma, flavour and colour, which are further developed during drying and finally obtained during roasting and further processing [13,14]. Earlier studies on the microbiology of cocoa bean fermentation have shown that both yeasts, filamentous fungi, lactic acid bacteria, acetic acid bacteria and Bacillus species might contribute to the fermentation [4,15].

Unfortunately, not many of the studies dealing with isolation and identification of yeast species involved during cocoa fermentation have been performed in West Africa, being the main producer of cocoa beans, and not many of the studies performed deal with the microbial succession of yeast species and the biodiversity within species. Further, as many of the publications date many years back, the taxonomy is not up-to-date and the use of molecular identification methods is not included. The general assumption is that many different yeast species are involved in the fermentation of cocoa beans. Thus, in Ivory Coast, Sanchez et al. [12] have found an abundant and varied yeast population, the dominant yeast species being Kloeckera apiculata, K. corticis and Saccharomyces chevalieri, now recognised as S. cerevisiae[16]. In a recent study by Ardhana and Fleet [4] in East Java, Indonesia, amongst others K. apis, S. cerevisiae and Candida tropicalis were found to be the most significant yeast species. In Bahia, Brazil, de Camargo et al. [3] found the most frequent yeast species to be C. krusei, Geotrichum candidum and C. mycoderma (now recognised as C. vivi[17]). Later, Schwan et al. [18] in Bahia, Brazil, identified amongst several other yeasts the dominant species to be S. cerevisiae, K. apiculata, Kluyveromyces marxianus and C. rugosa.

Several molecular methods are now well established for identification and typing of yeast species. Among such molecular techniques are the amplification of the 5.8S rDNA and the two ribosomal intergenic spacer regions (ITS1 and ITS2) used for grouping of yeast isolates e.g. in combination with restriction or sequence analysis [1923]. Within recent years sequencing of the D1/D2 domain at the 5′ end of the large subunit (26S) rDNA has been well acknowledged for its ability to identify yeast to the species level [2426]. Chromosome length polymorphism (CLP) determined by pulsed-field gel electrophoresis (PFGE) has been used successfully for strain typing of different yeast species such as C. krusei, Debaryomyces hansenii, Hanseniaspora spp., S. cerevisiae and Saccharomyces pastorianus[2731].

The aim of the present study has been to identify dominant yeast species and to follow the microbial successions during heap and tray fermentations of cocoa beans in Ghana, West Africa. Samples of pod surfaces and process equipment were included to examine the origin of predominant yeasts. The yeast species were identified by both conventional microbiological analyses and by sequencing of the D1/D2 domain of the 5′ end of the large subunit (26S) rDNA. For dominant yeasts, intraspecies variations were examined by determination of their chromosome length polymorphism (CLP).

2 Materials and methods

2.1 Cocoa bean fermentation

Fermentations from two cocoa producers were followed; one fermenting the cocoa beans in heaps and one using trays for fermentation. Both producers were located in Ghana, West Africa, ∼150 km north of Accra. Samples were collected early December, corresponding to early season, and then repeated mid January. According to normal practice the pods for the heap fermentations were harvested over 18 days before fermentation, whereas the pods for the tray fermentations were harvested over four days. Before fermentation the harvested cocoa pods were stored on the ground. Only matured pods were used for the fermentation. For the heap fermentations 200–1000 kg of cocoa beans were piled, stacked in a heap on plantain leaves, covered with plantain leaves and left to ferment for 72 h without turning of the heap, due to local practice. For the tray fermentation ∼50–100 kg beans were placed in a tray (90 cm × 120 cm, 10 cm deep), eight trays were stacked on top of each other, the top tray covered with plantain leaves and the beans left to ferment for 72 h without mixing.

2.2 Sampling

From the heap fermentation samples were collected from both the inner (∼15 cm) and the outer part of the heap. For the tray fermentation samples were only taken from one representative tray. Additionally samples were taken from production equipments, the cocoa pod surface and the fresh cocoa beans. For all bean samples, 200 g of cocoa beans were taken aseptically in sterile plastic bags, the samples were stored at ambient temperature and analysed within 2 h after sampling. For analysis of fresh cocoa beans the pods were cut aseptically immediately before analysis. For analysis of the micropopulation on production equipment and cocoa pod surfaces swabs corresponding to ∼50 cm2 were taken.

2.3 Determination of fermentation temperature and changes in pH during cocoa bean fermentation

The temperature was determined by inserting a thermometer (calibrated using Labortherm N thermometer, KEBOLab, VWR International, Albertslund, Denmark) in the middle of the fermenting cocoa beans. The pH of the fermenting beans was determined after blending 10 g of beans with 10 g of distilled water in a sterile Waring blender (Waring Products, New Hartford, USA) at maximum speed for 2 min and measuring the pH (Radiometer pH meter PHM82, Radiometer A/S, Brønshøj, Denmark). The pH-meter was calibrated by use of standard buffer pH 4.005 and 7.000 (Radiometer A/S).

2.4 Isolation, purification and maintenance of yeasts

As the pulp was difficult to remove from the beans it was decided to analyse the pulp together with the beans in order to ensure a complete analysis on the yeast population. The pulp and the cocoa beans were crushed aseptically for 2 min in a Waring blender at medium speed. Then 10 g of crushed cocoa beans were added to 90 ml of saline peptone diluent (SPO) (0.1% Bactopeptone (Difco, Detroit, MI, USA), 0.85% NaCl (Merck, Darmstadt, Germany), 0.03% Na2H2PO4, 2H2O (Merck) adjusted to pH 5.6). After mixing in a stomacher (Lab Blender 400, Seward, London) at normal speed for 30 s, 10-fold dilutions were prepared in duplo and 0.1 ml spread onto MYGP agar (3 g yeast extract (Difco), 3 g malt extract (Difco), 5 g Bactopeptone (Difco), 10 g glucose (Merck) and 20 g agar (Difco)) per 1000 ml supplied with 100 mg chloramphenicol (Sigma C-0378, St. Louis, MO, USA) and 50 mg chlortetracycline (Sigma C-4881). Incubation was carried out at 30 °C for 72 h, followed by recording of colony-forming units (cfu). From a suitable dilution of each sample, 25 representative isolates were picked and recultivated in MYGP broth for 48 h and further purified by streaking onto MYGP agar. The swabs were transferred to 10 ml SPO and mixed on a whirlmixer for 1 min, 0.1 ml was then spread on MYGP agar, per 1000 ml supplied with 100 mg chloramphenicol and 50 mg chlortetracycline. Incubation and recultivation were carried out as described above. A total of 496 isolates were obtained. The purified isolates were maintained at −80 °C in MYGP broth containing 20% (v/v) glycerol.

2.5 Phenotypic characterisation

Micro- and macromorphological description of the cultures were performed as described by Yarrow [32]. Growth in liquid media was determined by visually examining a 24 h old culture grown in MYGP broth at 25 °C. Growth on solid media was determined by examining the culture grown on MYGP agar after five days at 25 °C. The morphology of cells and mode of vegetative reproduction were determined by microscopy of a 24-h old culture grown in MYGP broth at 25 °C. Assimilation of carbon compounds was carried out by use of the API ID 32 C kit (Bio Merieux SA, Marcy-L‘Etoile, France), according to the manufacturers instructions inoculated with a 72-h culture pregrown on YPD agar at 25 °C. Assimilation of citrate and d-xylose was determined as described by Yarrow [32].

2.6 Amplification of the ITS1-5.8S rDNA-ITS2 region

The amplification was basically performed as described by Petersen et al. [23]. The following primers: ITS1 (5′-TCC GTA GGT GAA CCT GCG G-3′) and ITS4 (5′-TCC TCC GCT TAT TGA TAT GC-3′) were used for the amplification of the ITS1-5.8S rDNA-ITS2 region. The reactions were performed in an automatic thermal cycler (GeneAmp® PCR System 9700, Perkin–Elmer, Norwalk, CT, USA) under the following conditions: initial denaturation at 94 °C for 3 min; 30 cycles of denaturation at 94 °C for 2 min, annealing at 60 °C for 1 min, extension at 72 °C for 2.5 min; final extension at 72 °C for 7 min; holding at 4 °C. The amplified products were analysed on an Automatic Laser Fluorescence (ALF)-express sequencer (Amersham Pharmacia Biotech, Piscataway, NJ, USA) using a 6% denaturing polyacrylamide gel (6.6 ml Hydrolink Long Ranger (FMC BioProducts, Philadelphia, PA, USA), 21.8 g urea (ICN Biomedicals, Aurora, OH, USA), 7.2. ml 10 × TBE (450 mM Tris-base (Sigma), 440 mM boric acid (Sigma), 10 mM EDTA), and made up to 60 ml with milliQ water) in 0.6 × TBE (27 mM Tris-base (Sigma), 27 mM boric acid (Sigma), 0.6 mM EDTA), buffer under the following conditions: 300 min at 60 mA, 700 V and 55 °C. As marker ALF express™ Sizer™ 50–500 (Amersham Pharmacia Biotech) and 600–1600 (Amersham Pharmacia Biotech) was used. The sizes of the amplified ITS1-5.8S rDNA-ITS2 regions were determined by use of Fragment Manager (Amersham Pharmacia Biotech).

2.7 Sequencing of the D1/D2 domain of the large-subunit (26S) ribosomal DNA

Sequencing of the D1/D2 domain of the large-subunit (26S) ribosomal DNA was performed for major groups of isolates. The analysis was basically performed as described by van der Aa Kühle et al. [30] and van der Aa Kühle and Jespersen [26]. In brief the following primers: NL-1 (5′-GCA TAT CAA TAA GCG GAG GAA AAG-3′) and NL-4 (5′-GGT CCG TGT TTC AAG ACG G-3′) were used for the amplification of the D1/D2 domain. The reactions were performed in an automatic thermal cycler (GeneAmp® PCR System 9700, Perkin–Elmer) under the following conditions: initial denaturation at 94 °C for 3 min; 36 cycles of 94 °C for 2 min, 52 °C for 1 min, 72 °C for 2 min; final extension at 72 °C for 7 min, holding at 4 °C. The amplified products were purified by use of the GFX™ PCR DNA and Gel Band Purification Kit (Amersham Biosciences AB, Uppsala, Sweden). The purified PCR products were sequenced directly using the CEQ™ 2000 Dye Terminator Cycle Sequencing Kit (P/N 608000, Beckman Coulter Inc., Fullerton, CA, USA) following the manufactures instructions, in an automated sequencer (CEQ™ 2000 DNA Analysis System, Bechman Coulter Inc.). Cycle sequencing was performed in an automated thermal cycler (GeneAmp®PCR System 9700, Perkin Elmer, Norwalk, CT, USA) by use of the external primers NL-1 and NL-4 and the internal primers NL-2A (5′-CTT GTT CGC TAT CGG TCT C-3′) and NL-3A (5′-GAG ACC GAT AGC GAA CAA G-3′) as reported by Kurtzman and Robnett [24]. The sequences were assembled by use of ContigExpress (Vector NTI 7, InforMax, Inc., Frederick, MD, USA) and compared to the sequences reported in the GenBank using the BLAST algorithm. Finally the sequences were reported to the GenBank. Accession numbers are given in Table 3.

View this table:
Table 3

Grouping and identification of yeasts isolated during fermentation of cocoa beans by amplification of their ITS1-5.8S rDNA-ITS2 region

GroupsNumber of isolates (n= 496)ITS sizeaIdentificationb
I1340 (−)Cryptococcus humicolus
II1342 (−)Candida intermedia
III26392 (±2.5)Pichia kluyveri/Pichia fermentansc
IV106411 (±3.0)Pichia membranifaciens/Pichia galeiformisd
V196453 (±1.9)Candida krusei
VI2471 (±1.4)Cryptococcus laurentii
VII38472 (±10.4)Trichosporon asahii
VIII3506 (±2.7)Candida quercitrusa
IX11510 (±0.0)Pichia guilliermondii
X3513 (±6.4)Candida parapsilosis
XI3518 (±5.5)Rhodotorula glutinis
XII1530 (−)Candida silvicola
XIII5540 (±1.7)Candida stellimalicola
XIV71659 (±8.3)Hanseniaspora guilliermondii/H. opuntiaee
XV29722 (±4.0)Saccharomyces cerevisiae
  • a The sizes of the ITS1-5.8S rDNA-ITS2 regions were determined by use of a DNA sequencer (ALFexpress) resulting in lower values than those obtained by gel electrophoresis.

  • b The identifications are based on micro- and macromorphological examinations as described by Yarrow [32] and the assimilation profiles obtained by the API ID 32 C kit.

  • c The species was later identified as P. kluyveri.

  • d The species was later identified as P. membranifaciens.

  • e The species was later identified as H. guilliermondii.

2.8 Pulsed-field gel electrophoresis

Analysis of yeast chromosome polymorphism was basically performed as described by Jespersen et al. [28] and van der Aa Kühle et al. [30]. In brief, the yeast cultures were pregrown in YPG broth containing per litre of distilled water 10 g yeast extract (Difco), 20 g Bactopeptone (Difco) and 40 g glucose (Merck), pH 5.6, at 25 °C for 48 h and then successively recultivated twice for 24 h. Yeast chromosomal DNA was then prepared in agarose plugs. For each isolate Embedded Image block was transferred to a 1.2% (w/v) NA-agarose gel (Amersham Biosciences, Uppsala, Sweden). PFGE was performed on a PFGE DR-III unit (Bio-Rad, CA, USA) at 10 °C. During electrophoresis TBE buffer (45 mM Tris-base (Sigma), 44 mM boric acid (Sigma), 1 mM EDTA (Sigma)) was used as running buffer. The buffer was changed every 24 h. Running conditions for the resolution of chromosomal DNA were 100–120 mA combined with: for C. krusei 130 V at a 300 s switch interval for 24 h followed by 90 V at a 1000 s switch interval for 48 h; for S. cerevisiae 165 V at a 90 s switch interval for 14 h, followed by a 105 s switch interval for 12 h and a 120 s switch interval for 14 h; for P. membranifaciens, P. kluyveri and H. guilliermondii 150 V at a 200 s switch interval for 24 h followed by 100 V at a 700 s switch interval for 48 h. Yeast DNA-PFGE markers S. cerevisiae 345S (BioLabs, Beverly, MA, USA) and Hansenula wingei (Bio-Rad Laboratories, Hercules, CA, USA) were used for determination of chromosomal sizes. Finally the gel was stained with 1 mg ethidium bromide (Sigma) per litre TBE buffer for 1 h and rinsed twice with milliQ water for 15 min. The gels were visualised with UV transillumination and photographed. Estimation of chromosomal DNA sizes was done by use of the Kodak 1D Image Analysis Software, version 3.5 (Eastman Kodak company Rochester, NY, USA). The cluster analyses were carried out by use of the computer program BioNumerics version 2.50 (Applied Maths, Kortrijk, Belgium). Similarities between chromosomal profiles were based on the fraction of shared bands determined by Dice coefficient, as better sensitivity and reproducibility were obtained compared to the Pearson coefficient. Clustering was calculated by the unweighted pair group method using arithmetic average (UPGMA).

3 Results

3.1 Change in yeast cell counts, pH and temperature during cocoa bean fermentation

The cocoa beans used for the heap fermentation early December were generally found to be of good quality with an initial yeast cell count of ∼102 cfu g−1 of pulp and beans. An initial pH value of ∼4.2 was observed. As seen in Table 1, yeast cell counts of 106–107 cfu g−1 were observed already after 24 h for the heap fermentation. The increase in yeast cell count was found to vary dependent on the sampling site. For the inner part of the heap (∼15 cm from the outer layer) a rapid increase in yeast cell numbers was found corresponding to a yeast cell number of 3.3 × 107 after 24 h of fermentation. Hereafter the yeast cell count was found to decrease to 1.3 × 105 after 72 h of fermentation. For the outer part of the heap the increase in yeast cell count was slower, but after 72 h of fermentation a maximum yeast cell count of 9.1 × 107 cfu g−1 was reached. Compared to the pH 4.2 of the fresh cocoa beans, an initial decrease in pH was seen after 24 h, especially for the inner part of the heap, i.e. to pH 3.43. For the outer part only a slight decrease was observed, i.e. to pH 4.08. After 48 h of fermentation an increase in pH was observed reaching a maximum of pH 4.58 for the inner part of the heap and a pH value of 4.52 for the outer part. No significant changes in pH were observed up to 72 h of fermentation. The results from the heap fermentation were in principle confirmed during the second investigation mid January. However, as it was late season the heap fermentation was in this case conducted with cocoa beans originating from pods that had been stored for a longer period, which resulted in initial yeast cell counts of ∼105 cfu g−1 and a pH value of ∼4.4 (results not shown). The temperature of the heap fermentation (mid January) was found to increase from initially 30 °C to a maximum of 50 °C obtained after 24 h of fermentation at the inner part of the heap. The temperature increase at the outer part of the heap was somewhat slower, reaching a temperature of 45 °C after 24 h of fermentation. After 72 h of fermentation the inner part of the heap had a temperature of 49 °C while the outer part of the heap had a temperature of 41 °C (results not shown).

View this table:
Table 1

Yeast cell counts (cfu g−1) and pH during heap fermentation of cocoa beans

Fermentation time (h)Yeasts (cfu g−1)a,bpH
Inner part
0c102(−)4.2
243.3 × 107(0.11)3.43
481.4 × 105(0.86)4.58
721.3 × 105(0.69)4.48
Outer part
0c102(−)4.2
247.8 × 106(1.13)4.08
485.4 × 106(2.12)4.52
729.1 × 107(0.64)4.56
  • a Maltagar with antibiotics incubated at 30 °C for 72 h.

  • b Average values of two determinations are given. Standard deviations are shown in brackets.

  • c Samples were taken immediately after opening of the cocoa pods.

The cocoa beans used for the tray fermentation mid January were found to have a yeast cell count of 2.3 × 104 cfu g−1 and a pH value of 4.45. As seen in Table 2 the yeast cell count increased during the first 48 h to 6.0 × 106 cfu g−1, whereafter a decline was observed. A decrease in pH was found within the first 24 h of fermentation to pH 4.08, whereafter an increase to pH 4.64 was observed after 72 h. The temperature at the beginning of the fermentation corresponded to the ambient temperature (29 °C) and did gradually increase during the fermentation to a temperature of 50 °C after 72 h of fermentation. Except for a higher initial yeast cell count of the beans and a corresponding higher pH value the results reflected those obtained for the tray fermentation early December at the same location (results not shown).

View this table:
Table 2

Yeast cell counts (cfu g−1) and pH during tray fermentation of cocoa beans

Fermentation time (h)Yeasts (cfu g−1)a,bpH
0c2.3 × 104(0.05)4.45
243.4 × 104(0.15)4.08
486.0 × 106(0.51)4.51
721.5 × 106(0.45)4.64
  • a Maltagar with antibiotics incubated at 30 °C for 72 h.

  • b Average values of two determinations are given. Standard deviations are shown in brackets.

  • c Samples were taken after stacking of the trays.

3.2 Grouping and identification of isolates

Based on micro- and macroscopic examinations, 496 isolates primarily isolated during the fermentations but also from cocoa pods and process equipment were grouped into 70 groups. For a representative number within each group, assimilation of carbon compounds was determined by use of the API ID 32 C kit in order to assist in the identification and grouping of the isolates (results not shown). Additionally amplification of the ITS1-5.8S rDNA-ITS2 region was performed. Based on the combined results, the 70 groups were reduced to 15 groups that each represented an identified species (Table 3). The size of the ITS1-5.8S rDNA-ITS2 region was determined both by use of an automated DNA sequencer (ALFexpress) and by gel electrophoresis. The sizes of the ITS regions were in general lower determined by DNA sequencing than by gel electrophoresis. By sequencing of the ITS1-5.8S rDNA-ITS2 regions from representative species the sizes obtained by the DNA sequencer appeared to be more accurate than the sizes obtained by gel electrophoresis (results not shown), reason why the results obtained by the DNA sequencer are given in Table 3.

For the most important yeast species the identifications were further confirmed by sequencing of the D1/D2 domain of the large-subunit (26S) rDNA. Within each of these species a representative number of isolates were analysed as shown in Table 4. The sequences were submitted to Genbank and the accession numbers are as indicated. For all sequenced isolates homologies from 99.3–100.0% were obtained with sequences in GenBank. The lowest match was found for isolates of P. kluyveri with homologies of 99.3%. For several species isolates with 100.0% homologies were found (Table 4). Variations between isolates within a given species were never more than 0.2%.

View this table:
Table 4

Identification of dominant yeast species during cocoa fermentation by sequencing of their D1/D2 domain of the large subunit (26S) rDNA

GroupsaLength of D1/D2 sequenceHomology to original GenBank sequence (%)IdentitiesGenBank accession no.Closest GenBank fit (species name)
III58999.3558/562AY529513P. kluyveri
54699.3536/540AY529514
IV587100.0550/550AY529506P. galeiformisb
60599.8559/560AY529507
60299.8559/560AY529508
59499.8559/560AY529509
V620100.0605/605AY529498C. krusei
624100.0600/600AY529499
63599.8599/600AY529500
62199.8606/607AY529501
61599.8607/608AY529502
61099.8604/605AY529503
60899.8604/605AY529504
60199.8589/590AY529505
VII620100.0543/543AY529519T. asahii
VIII60699.8558/559AY529522C. quercitrusa
XIII62899.8576/577AY529520C. stellimalicola
62399.8576/577AY529521
XIV639100.0572/572AY529510H. opuntiaec
62199.8572/573AY529511
58099.8536/537AY529512
XV634100.0572/572AY529515S. cerevisiae
607100.0557/557AY529516
61499.8572/573AY529517
58699.8564/565AY529518
  • a From Table 3.

  • b The species is due to phenotypic criteria identified as P. membranifaciens (and reported as such in GenBank) even though sequence homology to P. galeiformis.

  • c The species is identified as H. guilliermondii (and reported as such in GenBank), as H. opuntiae is not recognised as a separate species [35].

For some of the isolates (i.e. group Nos. III, IV and XIV in Table 3) agreement could not be obtained between the phenotypic identity and the one obtained by sequencing. Group No. III was by phenotypic criteria initially identified as P. fermentans and by comparison with the sequences obtainable from GenBank as P. kluyveri (homology of 99.3%). The two species are reported to have considerable phenotypic similarities but can be differentiated by the strong assimilation of citrate and d-xylose shown by P. fermentans whereas P. kluyveri shows a weak or variable assimilation of citrate and d-xylose [33,34]. In total nine isolates were investigated for their ability to assimilate citrate and xylose. For all isolates no sign of citrate assimilation was observed within four weeks. For xylose no assimilation was observed within one week but for four isolates growth was observed after two to four weeks. Based on these assimilation results the isolates were grouped as P. kluyveri and the sequences reported as such in GenBank (AY529513AY529514). Group No. IV was by phenotypic criteria initially identified as either Pichia membranifaciens or P. galeiformis, two species that can be difficult to differentiate clearly based on phenotypic criteria. The isolates were by comparison to the sequences obtainable from GenBank identified as P. galeiformis (homologies of 99.8–100.0%). While P. galeiformis is not able to assimilate any of either l-sorbose, d-xylose or d-glucosamine, some strains of P. membranifaciens may be able to assimilate l-sorbose, d-xylose or d-glucosamine. Based on the assimilation results obtained by use of the API ID 32 C kit 84% of the isolates were able to assimilate l-sorbose, none of the isolates were able to assimilate d-xylose and all of the isolates were able to assimilate d-glucosamine. Based on the assimilation of d-glucosamine and partly on the assimilation of l-sorbose the group was identified as P. membranifaciens, eventhough the results obtained by comparison with the sequences obtainable from GenBank indicate P. galeiformis. Also some heavily sporulated isolates of P. membranifaciens can be differentiated from P. galeiformis by producing a noticeable reddish-brown colour [34]. For some of the isolates in group No. IV (Table 3) heavy sporulation was combined with a reddish-light brown colouration which further emphasised that the isolates should be grouped as P. membranifaciens. The sequences are therefore reported as P. membranifaciens in GenBank (AY529506AY529509). The isolates in group No. XIV were by phenotypic criteria identified as H. guilliermondii even though comparison with the sequences obtainable from GenBank indicated H. opuntiae (homologies of 99.8–100.0%). However, as H. opuntiae is not recognized as a separate species [35] the sequences are reported as H. guilliermondii in GenBank (AY529510AY529512).

3.3 Dominant yeast species and microbial successions during cocoa bean fermentation

Based on the above identifications the yeast species involved in the fermentations could then be identified as shown for the heap fermentation early December in Table 5. C. krusei (imperfect form of Issatchenkia orientalis) was found to be the dominant species followed by P. membranifaciens (perfect form of C. valida), P. kluyveri, H. guilliermondii (perfect form of K. apis) and T. asahii. C. krusei was in the beginning of the fermentation primarily found in the inner part of the heap but after 72 h of fermentation C. krusei dominated in samples from both the inner and the outer part of the heap. P. kluyveri, H. guilliermondii and T. ashii were primarily found in the samples taken from the outer part of the heap whereas P. membranifaciens was equally present in samples from both the inner and the outer part of the heap. A microbial succession took place during the heap fermentation resulting in a relative increase of P. membranifaciens, contrary to P. kluyveri and H. guilliermondii that in general were present in the highest relative numbers during the early phases of the fermentation. C. krusei was during the entire fermentation present in high numbers whereas T. asahii only was found in one sample. The composition of the microflora was in principle confirmed during the fermentation mid January even though C. krusei was even more dominant during this fermentation and S. cerevisiae was found in low numbers (results not shown).

View this table:
Table 5

Microbial succession of yeast species during heap fermentation of cocoa beans

Fermentation time (h)aYeasts species (% of total population)
C. kruseiP. membranifaciensP. kluyveriH. guilliermondiiT. asahii
Inner part
248301700
48924400
726428080
Outer part
2412019690
48284121244
7250298130
  • a The microbial composition for the 0 h samples were not determined.

As shown in Table 6, S. cerevisiae and P. membranifaciens were the dominant species during tray fermentation, followed by C. krusei, P. kluyveri, H. guilliermondii and some yeast species of minor importance, i.e. C. stellimalicola, C. quercitrusa and Rhodotorula glutinis. As for the heap fermentation (Table 5) a decline in the numbers of P. kluyveri and H. guilliermondii was observed during the tray fermentation. Contrary to the heap fermentation S. cerevisiae, that was only found in low numbers during the second heap fermentation, was for the tray fermentation found to increase significantly in numbers during the first 48 h of fermentation whereafter a decrease was observed. As for the heap fermentation, P. membranifaciens was found to increase during the entire fermentation and was the dominant species after 72 h of fermentation. C. krusei was generally found in low numbers. C. stellimalocola, C. quercitrusa and R. glutinis were only found in one sample each and are not regarded to be of importance for the fermentation. The composition of the microflora was in principle as seen for the tray fermentation early December, even though at that time significantly higher numbers of C. krusei and H. guilliermondii and correspondingly lower numbers of S. cerevisiae were seen (results not shown).

View this table:
Table 6

Microbial succession of yeast species during tray fermentation of cocoa beans

Fermentation time (h)Yeasts species (% of total population)
C. kruseiS. cerevisiaeP. membranifaciensP. kluyveriH. guilliermondiiC. stellimalicolaC. quercitrusaR. glutinis
0a1111031311600
240572740048
480901000000
72848800000
  • a Samples were taken after stacking of the trays.

3.4 Origin of yeast isolates

The yeast population on the surface of the cocoa pods was in order of appearance: H. guilliermondii (53%), P. guilliermondii (22%), C. intermedia (7%), C. parapsilosis (6%), Cryptococcus laurentii (4%), C. silvicola (2%), P. membranifaciens (2%), R. glutinis (2%) and Cryptococcus humicola (2%), whereas inside the cocoa pods only C. krusei (80%) and P. guilliermondii were found (20%). On the surface of cocoa pods with black pod disease high numbers of C. krusei were additionally seen (results not shown). P. membranifaciens, T. asahii and C. krusei were found on the trays used for fermentation (results not shown).

3.5 Assimilation profiles for dominant yeast species

For the dominant yeast species found during cocoa bean fermentation, their assimilation profiles as determined by the API ID 32 C kit are shown in Table 7. Even though it is clear that each species has a characteristic assimilation profile, the assimilation of some of the isolates varied slightly as compared to the assimilation profiles described in current taxonomic keys [36,37], however, in all cases variations were only seen for some isolates within a given species. As citrate assimilation previously has been reported as an important technological parameter in cocoa fermentation, the ability of the dominant yeast species to assimilate citrate was investigated. Of the dominant species only some isolates of C. krusei (4 out of 13 investigated) were able to assimilate citrate within four weeks. None of the isolates within the following species were able to assimilate citrate within four weeks: S. cerevisiae (0/16), P. membranifaciens (0/21), P. kluyveri (0/9) and H. guilliermondii (0/9).

View this table:
Table 7

Assimilation profiles of dominant species during cocoa fermentation

Carbon compoundsaAssimilation (ratio of positive isolates)b
C. kruseiS. cerevisiaeP. membranifaciensP. kluyveriH. guilliermondii
Galactose8/8
Actidione1/2611/12
Saccharose8/8
N-acetyl-glucosamine26/261/825/264/8
dl-lactate26/265/88/8
l-arabinose
Cellobiose1/2612/12
Raffinose7/8
Maltose
Trehalose9/12
2-Keto-gluconate8/12
α-Methyl-d-glucoside
Mannitol
Lactose
Inositol
Sorbitol
d-xylose4/8
Ribose1/26
Glycerol26/2617/267/8
Rhamnose
Palatinose7/8
Erythritol
Melibiose
Glucoronate
Melezitose1/12
Gluconate1/812/12
Levulinate17/264/8
Glucose26/268/826/268/812/12
Sorbose17/2616/265/8
Glucosamine22/2626/268/8
Esculin21/267/818/268/812/12
  • a Assimilation of carbon compounds was determined by use of the API ID 32 C kit.

  • b Number of positive isolates/number of isolates investigated.

3.6 Intraspecies variations as determined by chromosome length polymorphism

In order to investigate the biodiversity within the dominant yeast species chromosome length polymorphism (CLP) was determined for isolates of C. krusei, S. cerevisiae, P. membranifaciens, P. kluyveri and H. guilliermondii. Cluster analyses of the chromosome profiles are seen in Fig. 1. For C. krusei six chromosomal bands were observed, ranging in size from 1320–3206 kb with an estimated average genomic size of 14,142 ± 69 kb. CLP was evident (Fig. 1(a)) and most of the isolates had individual profiles. The isolates could be grouped into clusters. None of the isolates obtained after 24 and 72 h of heap fermentation were grouped in the same cluster, which could either indicate that a microbial succession took place at the strain level or that the strains were not equally distributed in the heap and therefore are specific for a given sample. For S. cerevisiae 16–17 chromosomal bands were separated on the gel with sizes from 188 to 1945 kb and an estimated average genomic size of 13,045 ± 427 kb. As for C. krusei, CLP was evident for the S. cerevisiae isolates (Fig. 1(b)) and also in this case most of the isolates had an individual profile. For isolates of P. membranifaciens 4–5 chromosomal bands varying in sizes from 1318 to 3322 kb could be seen with an estimated genomic size of 10,725 ± 1675 kb. Most of the P. membranifaciens isolates had individual profiles (Fig. 1(c)). Isolates identified at the pod surface or on the trays were similar or identical to those found during the fermentations, indicating that the isolates involved in the fermentation originate from either the pod surfaces or populations established on the trays. For P. kluyveri 6–7 chromosomal bands were observed with sizes from 612 to 3367 kb and an estimated average chromosomal size of 14,190 ± 1207 kb. Also for P. kluyveri CLP was evident (Fig. 1(d)). All investigated isolates had separate profiles (i.e.<92% similarity). The isolates obtained during the heap fermentation were comparable to the isolates obtained during the tray fermentation even though two separate production sites were included. For H. guilliermondii 8–9 chromosomal bands were observed with sizes from 563 to 3075 kb. The average chromosomal size was estimated to 12,142 ± 1063 kb. CLP was evident (Fig. 1(e)) and most isolates had individual profiles. The chromosome profiles of the isolates obtained from the pod surfaces were similar to those obtained during both heap and tray fermentation, indicating that the origin of the H. guilliermondii isolates seen during the fermentations is the pod surface.

Figure 1

Dendrograms showing the clustering of isolates of the dominant yeast species involved in the fermentation of West African cocoa beans. The clustering is based on chromosome length polymorphism (CLP) as determined by pulsed-field gel electrophoresis. The chromosome profiles were evaluated by use of the computer program BioNumerics version 2.50 (Applied Maths, Kortrijk, Belgium). Similarities between chromosomal profiles were based on the fraction of shared bands determined by Dice coefficient and clustering was calculated by the unweighted pair group method using arithmetic average (UPGMA). (a) Candida krusei (b) Saccharomyces cerevisiae (c) Pichia membranifaciens (d) Pichia kluyveri (e) Hanseniaspora guilliermondii. Running conditions varied between the species as described in Section 2.8. Isolate numbers are indicated in brackets.

4 Discussion

In the present study C. krusei, followed by P. membranifaciens, P. kluyveri, H. guilliermondii and T. asahii, was found to be the dominant species during heap fermentation, reaching maximum yeast cell counts of 9.1 × 107 cfu g−1 of pulp and beans, whereas S. cerevisiae and P. membranifaciens were found to be the dominant species during tray fermentation, followed by low numbers of C. krusei, P. kluyveri, H. guilliermondii and some yeast species of minor importance. Maximum yeast cell counts of 6.0 × 106 cfu g−1 of pulp and beans were reached during the tray fermentation. The results are to some extent in agreement with Sanchez et al. [12] who, for fermentation of cocoa beans in plastic boxes in Ivory Coast, obtained maximum yeast cell counts of 3 × 107 cfu g−1 of pulp and reported the occurrence of amongst others the following species: C. krusei, C. valida (imperfect form of P. membranifaciens), S. chevalieri (now recognised as S. cerevisiae) and P. membranifaciens. As for the tray fermentation in this study, Ardhana and Fleet [4] found S. cerevisiae to be dominant together with K. apis (imperfect form of H. guilliermondii) during wooden box fermentation of cocoa beans in East Java, reaching yeast cell counts of 107–108 cfu g−1 in the pulp, whereas none of the other species found in the present study were reported. In agreement with Ardhana and Fleet [4], K. apis (H. guilliermondii) was in the present study found to be the dominant Kloeckera sp. whereas in other studies K. apiculata (imperfect form of Hanseniaspora uvarum) has been reported [12,18]. As in the present study, a decline in the number of K. apis after 24–36 h fermentation was observed by Ardhana and Fleet [4], a trend that also has been reported for K. apiculata[18]. Schwan et al. [18] found S. cerevisiae to be one of the dominant yeast species during the first 72 h of box fermentation in Brazil. Surprisingly, P. membranifaciens, one of the most predominant yeast species found in this study, especially at the later stages of fermentation, was not found by Ardhana and Fleet [4] in East Java. De Camargo et al. [3] also found P. membranifaciens in Brazil, together with C. krusei, S. cerevisiae and other yeast species. Despite the different species reported to occur during cocoa bean fermentation, all studies conducted so far confirm a very rich and varied yeast population.

Of the dominant yeast species, C. krusei, P. membranifaciens, H. guilliermondii and T. asahii could be associated to either the pod surface, the cocoa beans or the production equipment, whereas this was not the case for S. cerevisiae and P. kluyveri, either due to the fact that they were not there or were present below the detection limit. Similarly, during spontaneous wine fermentation S. cerevisiae is rarely found on the grapes but is present on the winery equipment [38] and it is therefore likely, even though not found, that strains of S. cerevisiae will be able to establish themselves on the equipment used for the tray fermentation. Apart from this study, no other studies appear to have dealt with the habitat of the yeast species involved in cocoa fermentation. However, the occurrence of C. krusei inside healthy cocoa pods is a confirmation of the results obtained by Maravalhas [39], who found C. krusei in high numbers inside 1–2% of fresh cocoa beans in Bahia, Brazil.

Despite differences in yeast species composition, microbial successions at both species and strain levels were found to take place during heap as well as tray fermentations. Microbial succession at species and strain level are often reported for spontaneously fermented products [4042]. The microbiological successions seen during cocoa fermentation are likely to be due to changes in the microenvironment, i.e. changes in nutrient availability, pH, temperature, presence and concentration of organic acids, as well as oxygen concentration. The relative increase of P. membranifaciens during cocoa fermentation might be linked to its tolerance towards low pH and high concentration of organic acids, including acetic acid [43,44]. Further P. membranifaciens has been reported to inhibit the growth of Hanseniaspora spp. on fruit, probably due to substrate competition or production of inhibitory substances, as Hanseniaspora spp. are quite sensitive to ethanol and other secondary metabolites [45]. Also strains of P. membranifaciens and P. kluyveri have been reported to produce mycocins that inhibit other yeasts as e.g. S. cerevisiae and Candida spp. [46,47]. Strains of C. krusei isolated from indigenous fermented maize dough in Ghana have previously been shown to be more tolerant against the presence of organic acids than S. cerevisiae[48]. The reasons why C. krusei in this study plays a major role during heap fermentation compared to tray fermentation, and visa versa for S. cerevisiae, are not evident from the present results.

Even though citrate often has been mentioned as an important carbon source during cocoa fermentation [11], except for a minor part of the C. krusei isolates none of the isolates within the dominant yeast species were able to assimilate citrate, which indicates that citrate assimilation is not an important selective parameter among yeast strains during cocoa fermentation in West Africa.

Sequencing of the D1/D2 domain of the large-subunit (26S) ribosomal DNA was in the present study found to be applicable for identification of most species. However, micro- and macro-morphological examinations were needed before a correct species name could be assigned and care should be taken to avoid invalid species names. For all sequenced isolates homologies to species names given by GenBank were >99.3%, which according to Kurtzman and Robnett [24] is typical of conspecific isolates.

Determination of CLP by PFGE was used for strain typing of the dominant species during cocoa bean fermentation. The chromosome profiles of the C. krusei isolates were comparable to the profiles described by Hayford and Jakobsen [27] of C. krusei strains isolated from fermented maize dough in Ghana, and the chromosome profiles for the S. cerevisiae isolates were comparable to those described by van der Aa Kühle et al. [30] for strains of S. cerevisiae isolated from fermented sorghum beer in Ghana and from spontaneously fermented maize dough in Ghana [40]. The chromosome profiles of P. membranifaciens differed somewhat from those reported by Mikata and Ueda-Nishimura [49], as a higher number of chromosomal bands of larger sizes was found in the present study. On the contrary, the sizes of the chromosomal bands found for the P. membranifaciens isolates in the present study were comparable to those found for the type strain of P. membranifaciens as reported by Naumov et al. [50]. The chromosome profiles of the H. guilliermondii isolates were comparable to those reported by Esteve-Zarzoso et al. [29], in the way that 8–9 chromosomal bands could be observed with a similar size distribution. To our knowledge chromosome profiles for isolates of P. kluyveri are for the first time reported. For some of the species the number of chromosomal bands was found to vary within the species, i.e. for some isolates of S. cerevisiae more than 16 chromosomal bands could be observed. This is most likely due to multiple alleles of the same chromosomes separated into two bands of different sizes. Variations in the number of chromosomal bands determined by PFGE have been reported within several different species such as Candida spp. and Hanseniaspora spp. [29,51]. Due to the great variation in the number of chromosomal bands observed and the unknown ploidy of the isolates, the average genomic sizes given are only rough estimates.

The results obtained in the present study show that fermentation of cocoa beans in West Africa is a very heterogeneous process depending on seasonal variations, batch sizes, production methods, etc. The variations are reflected both in maximum yeast cell counts, yeast successions and the identity of the predominant yeast species. The variations are expected to influence the quality of the cocoa beans, including both the organoleptic quality and the growth of filamentous fungi. With the purpose of selecting starter cultures for controlled cocoa fermentation as recently suggested [52], the present study has indicated potential candidates for further studies to clarify their effects on fermentation and cocoa quality. Determination of CLP can be used as a valid method for differentiation and characterisation of strains within all the dominant yeast species.

Acknowledgements

The work performed was financed through the EU INCO project: “Developing biochemical and molecular markers for determining quality assurance in the primary processing of cocoa in West Africa – COCOQual” and by LMC (Centre for Advanced Food Studies). The authors wish to thank Janne Benjaminsen for her excellent and skilful technical assistance.

References

  1. [1].
  2. [2].
  3. [3].
  4. [4].
  5. [5].
  6. [6].
  7. [7].
  8. [8].
  9. [9].
  10. [10].
  11. [11].
  12. [12].
  13. [13].
  14. [14].
  15. [15].
  16. [16].
  17. [17].
  18. [18].
  19. [19].
  20. [20].
  21. [21].
  22. [22].
  23. [23].
  24. [24].
  25. [25].
  26. [26].
  27. [27].
  28. [28].
  29. [29].
  30. [30].
  31. [31].
  32. [32].
  33. [33].
  34. [34].
  35. [35].
  36. [36].
  37. [37].
  38. [38].
  39. [39].
  40. [40].
  41. [41].
  42. [42].
  43. [43].
  44. [44].
  45. [45].
  46. [46].
  47. [47].
  48. [48].
  49. [49].
  50. [50].
  51. [51].
  52. [52].
View Abstract