International Journal of Food Microbiology 210 (2015) 102–112

Contents lists available at ScienceDirect

International Journal of Food Microbiology

journal homepage: www.elsevier.com/locate/ijfoodmicro

Microbiological diversity associated with the spontaneous wet method of

coffee fermentation
Suzana Reis Evangelista a, Maria Gabriela da Cruz Pedroso Miguel a, Cristina Ferreira Silva a,
Ana Carla Marques Pinheiro b, Rosane Freitas Schwan a,⁎
a
Biology Department, Federal University of Lavras, CEP 37200-000, Lavras, MG, Brazil
b
Food Sciences Department, Federal University of Lavras, CEP 37200-000, Lavras, MG, Brazil

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

Article history: The evaluation of the microbiota present during coffee wet fermentation was done in two distinct regions of
Received 17 March 2015 Minas Gerais, Brazil: one farm in the South of Minas Gerais (Lavras = L) and another farm in the savannah region
Received in revised form 5 June 2015 (Monte Carmelo = MC). The yeast population ranged from 2.48 to 4.92 log CFU/g and from 2 to 4.81 log CFU/g,
Accepted 7 June 2015
the mesophilic bacteria population ranged from 3.83 to 8.47 log CFU/g and from 5.37 to 7.36 log CFU/g, and the
Available online 17 June 2015
LAB population ranged from 2.57 to 5.66 log CFU/g and from 3.40 to 4.49 log CFU/g in the L and MC farms, respec-
Keywords:
tively. Meyerozyma caribbica and Hanseniaspora uvarum were the dominant yeasts in coffee wet fermentation at L
Arabica coffee farm, and Torulaspora delbrueckii was the dominant yeast at MC farm. The species Staphylococcus warneri and
Wet fermentation Erwinia persicina were the predominant bacteria at L farm, and Enterobacter asburiae and Leuconostoc
Microbiota mesenteroides were the dominant species at MC farm. Lactic acid was the principal acid detected, reaching
PCR–DGGE 2.33 g/kg at L farm and 1.40 g/kg at MC farm by the end of the process. The volatiles composition was similar
Yeasts for roasted coffee from the two different regions and furans, acids, and alcohol were the main groups detected.
Bacteria Temporal Dominance Sensation (TDS) analyses showed that the coffee beverage from L farm was dominated
by citrus and herbaceous sensory characteristics, while the coffee from MC farm was dominated by citrus, herba-
ceous, and nuts sensory characteristics. Evaluating the microbiota in these two regions was important in improv-
ing the knowledge of the microbial species present during coffee wet fermentation in Brazil.
© 2015 Elsevier B.V. All rights reserved.

1. Introduction
The coffee fruit usually consists of an outer layer of skin (peel), called
the exocarp under which there is a layer of pulp, followed by mucilagi-
nous mesocarp (mucilage) firmly adhered to the rigid layer called
parchment (endocarp). The parchment protects two seeds surrounded
by a thin membrane known as silver skin. The coffee fruit are subject
to a fermentation process in order to remove the mucilage and pulp be-
fore drying.
Three methods might be used to process coffee: the dry, semi-dry,
and wet processes (Brando and Brando, 2014; Silva, 2014). In the dry
method, the newly harvested whole fruit is fermented and dried on
platforms after which the coffee beans are removed by hulling and
polishing (removed the husk layer that cover the dry coffee beans). In
the semi-dry method the coffee peel, pulp and part or all mucilage are
removed mechanically and then the coffee is fermented and dried.
The amount of mucilage removed depends on the characteristic of the
machine used. In the wet method the peel and pulp are removed
⁎ Corresponding author.
E-mail address: rschwan@dbi.ufla.br (R.F. Schwan).
mechanically, leaving the mucilage adhered to the beans. These pulped
coffees are then transferred to water tanks where they are allowed to
ferment for 6 to 72 h (depending on the environmental temperature),
during which the remaining mucilage are degraded and solubilized.
The beans are then removed from the tanks and sun dried.
Brazil is the largest producer and exporter of coffee. The State of
Minas Gerais, in the Southeast of Brazil, produces about 53% of all the
coffee in Brazil, of which 85% is processed by dry method. In the last de-
cade, coffee producers in the southern and central areas of Minas Gerais
began to process coffee via wet fermentation (Sindcafé-MG, 2014). This
processing produces a coffee with a different composition and sensory
characteristics than coffee processed by the semi-dry or dry methods,
which are alternatives used by the producer to suit different markets.
Microorganisms are naturally present during coffee processing and
use the various compounds in the pulp and mucilage as nutrients during
the fermentation stages. They secrete organic acids and other metabo-
lites that may affect the final sensory characteristics of the beverage
(Silva, 2014). This microbiota may vary according to several factors: re-
gional characteristics, coffee fruit composition and the method of fer-
mentation. It is important to know the microbiota present during the
processing of coffee, especially when selecting starter cultures that can

http://dx.doi.org/10.1016/j.ijfoodmicro.2015.06.008
0168-1605/© 2015 Elsevier B.V. All rights reserved.
 S.R. Evangelista et al. / International Journal of Food Microbiology 210 (2015) 102–112
be used in producing differentiated final products and inhibiting the
growth of mycotoxigenic fungi (Masoud and Jespersen, 2006; Massawe
and Lifa, 2010; Silva et al., 2012).
Studies have been performed to identify the microbiota present in wet
processing of coffee fruits in India, Hawaii, and Mexico (Agate and Bhat,
1966; Avallone et al., 2001; Frank et al., 1965). Classical morphological
and biochemical characterization has been used to identify the microbio-
ta, and species of Erwinia, Klebsiella, and Leuconostoc and the yeasts
Kloeckera, Candida, Saccharomyces bayanus, Saccharomyces cerevisiae var.
ellipsoideus, and Cryptococcus have been identified as contributing to
these fermentations. Using a molecular approach, PCR–DGGE, Masoud
et al. (2004) identified Pichia kluyveri, P. anomala, Hanseniaspora uvarum,
Kluyveromyces marxianus, Candida pseudointermedia and Torulaspora
delbrueckii, during wet coffee fermentation in Arusha, Tanzania (East
Africa).
Recently, yeasts species found in coffee processed by the wet meth-
od in Brazil were reported by Pereira et al. (2014). Pichia fermentans
(YC5.2) and Saccharomyces sp. (YC9.15) were studied as having a poten-
tial for use as starter cultures for coffee wet fermentation. The bacteria
population was not mentioned in their work (Pereira et al., 2014).
Studies still are needed in Brazil to improve the knowledge of the
microbiota present in coffee processing for each producing region due
to the variation in climate and altitude. Our aim was to study the micro-
bial diversity involved in wet coffee fermentation in two main produc-
ing regions of Brazil having distinct environmental characteristics using
culture-dependent and culture-independent methods. The target me-
tabolites present during the fermentation process were also analyzed
using headspace solid-phase microextraction/gas chromatography
(HS–SPME/GC) and high-performance liquid chromatography (HPLC).
The sensory profile of the final coffee beverage was also evaluated.
2. Materials and methods
2.1. Processing of coffee fermentation
The coffee fruits were fermented by the wet method. The experiments
were performed in two geographically different regions of Minas Gerais,
at a farm located in Monte Carmelo (MC), 870 m above sea level in the
savannah (Cerrado) region, and at a farm located in Lavras (L), 919 m
above sea level in the Atlantic Forest region. The experiments were
done during two consecutive harvesting seasons (2012 and 2013).
The fruit (60 kg) of Coffea arabica L. var. Acaiá was harvested at the
mature stage (cherries) and was mechanically depulped in a horizontal
machine (model BDSV-04; Pinhalense, São Paulo, Brazil), followed by
for 48 h fermentation in a tank with 60 l of water to remove the muci-
lage. At the L farm, the temperature of fermentation was between 14
and 23 °C, and at the MC farm it was between 20 and 28 °C. Fermenta-
tions were done in duplicate at each farm. After fermentation, the cof-
fees were placed on suspended platforms for sun drying until they
reached approximately 11% moisture (224 h at L farm and 336 h at
MC farm), measured using Moisture Meter G600i (GEHAKA, São
Paulo, Brazil). The following samples were collected: cherries before fer-
mentation, along the fermentation (0, 6, 12, 24, 36, and 48 h), and dur-
ing drying (60, 112, 224, and 336 h). Samples (500 g) were placed
aseptically in sterile plastic bags, in triplicate, and immediately trans-
ferred to the laboratory in iceboxes for microbiological analyses. For
physicochemical analyses, coffee samples were frozen at −20 °C until
analyzed. Sensory analysis and volatile compounds were evaluated in
dried coffee beans.
2.2. Characterization and identification of microbiota
2.2.1. Quantification, isolation, and phenotypic characterization
Coffee cherry sample (10 g) was added to 90 mL of sterile peptone
water (in g/L: 1 bacteriological peptone [Himedia, Mumbai, India]), ho-
mogenized for 2 min in a Stomacher (Mayo Homogenius HG 400, São
103
Paulo, Brazil), and used for decimal serial dilution. Bacteria and yeasts
were enumerated by spread plating on MRS agar (Merck) for lactic acid
bacteria, on plate count agar (PCA) (in g/L: 5 tryptone [Himedia], 2.5
yeast extract [Merck], 1 glucose [Merck], 15 agar Merck]), for total bacte-
ria and on YPD agar (in g/L: 10 yeast extract [Merck], 10 peptone
[Himedia], 20 dextrose [Merck], 20 agar [Merck]) for yeasts. MRS plates
were incubated in an anaerobic jars at 30 °C for 3–4 days and PCA and
YPD plates were incubated at 30 °C for 3–7 days and the morphological
properties of the colonies (cell size, cell shape, edge, color, and bright-
ness) were recorded and the square root of the number of colonies
counted for each morphotype was purified by streaking on new agar
plates (same culture media used for plating) (Senguna et al., 2009).
The pure cultures were stored in an ultra freezer at −80 °C in the same
broth culture media used for plating, containing 20% glycerol (w/w).
The phenotypic characterization of the bacterial colonies was per-
formed using Gram staining, catalase and oxidase activities, motility
tests, growth in culture media with 1% and 5% (w/v) NaCl (salt toler-
ance), protease production, spore formation, and the ability to ferment
glucose, sucrose, and xylose (Sigma, St. Louis, USA), as recommended
in Bergey's Manual of Determinative Bacteriology (Holt et al., 1994).
Yeast colonies were characterized for morphology and biochemical as-
sessments as described by Kurtzman et al. (2011).
A total of 435 isolates was obtained; 251 were from samples from L
farm and 184 isolates were obtained from MC farm.
The microbiota identified was the same in both years analysed.
However, variation of less than 5% was found in the counting.
2.2.2. Genotypic identification
The isolates (435) were grouped by phenotypic characteristics, as
mentioned above, and representatives of each phenotypic group were
subjected to molecular analyses (236 isolates). The genotypic character-
ization of the selected isolates was first performed by rep-PCR and sub-
sequently by DNA sequencing.
The bacterial and yeast cultures were grown under appropriate con-
ditions, collected from agar plates with a sterile pipette tip, and resus-
pended in 50 μL of sterile Milli-Q water. The suspension was heated
for 10 min at 95 °C, and three μL was used as a DNA template in the
PCR experiments.
The fingerprints of the genomic DNA were obtained via the PCR am-
plification of repetitive bacterial and yeast DNA elements (rep-PCR)
using the (GTG)5 primer, as described by Nielsen et al. (2007). Amplified
PCR products were separated via 2% (w/v) agarose gel electrophoresis
at 70 V for 3 h, and the images were visualized and photographed
using a transilluminator LPixImage (LTB 20 × 20 HE, LPix®, Brazil).
The bacteria and yeasts representative of each group were subjected
to 16S rRNA gene and internal transcribed spacer (ITS) region gene se-
quencing, respectively. The amplification of the 16S rRNA gene used the
primers 27 F and 1512R (Devereux and Willis, 1995). The ITS region was
amplified using the primers ITS1 and ITS4 (Nielsen et al., 2007). The am-
plified PCR products were sent for sequencing at the Advanced Genetics
Technologies Center — AGTC (Kentucky, USA); the ABI3730 XL auto-
matic DNA sequencer was utilized.
The sequences were aligned using the BioEdit 7.7 sequence align-
ment editor and were compared to the GenBank database using the
Basic Local Alignment Tool (BLAST) program (National Center for Bio-
technology Information, Bethesda, MD) for the identification of isolates.
2.3. PCR–DGGE analysis
Samples of coffee cherries (3 g) were mixed with 5 mL of Milli-Q
water for 10 min then the coffee cherries were removed and the liquid
phase was centrifuged at 100 ×g for 10 min at 4 °C. The pellet was used
for DNA extraction. The total DNA was extracted from samples using the
“DNA Purification from Tissues” protocol (QIAamp DNA Mini Kit
[Qiagen, Hilden, Germany]) in accordance with the manufacturer's in-
structions. The DNA from the bacterial community was amplified with
104 S.R. Evangelista et al. / International Journal of Food Microbiology 210 (2015) 102–112
the primers 338fgc and 518r (Ovreas et al., 1997). A fragment of the D1-
region of the 26S rRNA gene was amplified with the eukaryotic univer-
sal primers NL1GC and LS2 (Cocolin et al., 2000). Reactions were per-
formed according to Ramos et al. (2010). Aliquots (3 μL) of the
amplification products were analyzed via electrophoresis on 1% agarose
gels before they were used for DGGE.
The PCR products were separated in polyacrylamide gels (8% (w/v)
acrylamide:bisacrylamide [37.5:1]) in 1 × TAE buffer with a DCode sys-
tem apparatus (BioRad Universal Dcode Mutation Detection System,
Richmond, CA, USA). Denaturation gradients were used that varied
from 15 to 55% for the bacterial products (100% corresponded to 7 M
of urea and 40% [v/v] formamide) and from 20 to 60% for the yeast prod-
ucts. Electrophoresis was conducted at a constant voltage of 130 V for 6 h
and at a constant temperature of 60 °C. Following electrophoresis, the
gels were stained with SYBR-Green I (Molecular Probes, Eugene, UK)
(1:10.000 v/v) for 30 min. The images were visualized and photographed
using a transilluminator LPixImage (LTB 20 × 20 HE, LPix®, Brazil).
Selected bands were excised and their DNA amplified using the
primers 338fgc and 518r for bacteria and NL1 and LS2 for yeast. The
PCR products were purified and sequenced at the Advanced Genetics
Technologies Center — AGTC (Kentucky, USA); the ABI3730 XL auto-
matic DNA sequencer was utilized. The sequences were compared to
Fig. 1. The populations of yeast (YEPG counts ), mesophilic bacteria (PCA counts
populations of yeast (YEPG counts ), mesophilic bacteria (PCA counts
the GenBank database using the BLAST algorithm (National Center for
Biotechnology Information, Maryland, USA).
2.4. Analyses of organic acids
The organic acids from the beans were analyzed using a high-
performance liquid chromatography system (Shimadzu Corp., Japan)
equipped with detection system consisting of a UV–Vis detector (SPD
10Ai). The extraction and operating conditions were performed accord-
ing to the methodology described by Evangelista et al. (2014).
2.5. Analysis of volatile compounds
Volatile compounds were extracted from the beans and analyzed ac-
cording to Evangelista et al. (2014). The volatile compounds from each
headspace analysis were defined by integrating the peak areas of all
the identified compounds. Relative percent areas were used for semi-
quantitation of the target compounds. The relative percentage of the in-
dividual compounds was calculated from the total peak area of the vol-
atile compounds on the chromatograms (Petisca et al., 2013).
), and lactic acid bacteria (MRS counts ) in coffee wet processing at Lavras farm (A). The
), and lactic acid bacteria (MRS counts ) in coffee wet processing at Monte Carmelo farm (B).
 S.R. Evangelista et al. / International Journal of Food Microbiology 210 (2015) 102–112 105

2.6. Analysis of sensory characteristics
The samples of roasted coffee from each region were prepared ac-
cording to the Specialty Coffee Association of America (SCAA, 2013).
The coffee was roasted in a laboratory roaster (Probatino, Leogap model,
Brazil) with a capacity of 150 g. The temperature used ranges from 180
to 200 °C and the time was around 10 min until the beans reach light-
medium roast. The coffee beans were ground in an electric mill
(Pinhalense ML-1, Brazil). For sample uniformity evaluation, five cups of
each sample were prepared. Clean and odor-free water was used. Water
was brought to approximately 92–94 °C and poured directly on the
ground coffee. The ratio was 8.25 g of coffee per 150 mL of water. Grounds
were steeped for 3–5 min before the sample was evaluated. A panel of
three trained coffee experts with Q-Grader Coffee Certificates evaluated
the samples. The first sensorial evaluation was conducted to identify the
most relevant of the coffee attributes, and the eight most cited terms
were selected and retained for further Temporal Dominance of Sensations
(TDS) evaluations (Pineau et al., 2009). The attributes selected by the
panel were as follows: chocolate, bitter chocolate, caramel, citric, tobacco,
butter, herbaceous, and nuts. Samples coded with three digits were sub-
mitted in a balanced order (Wakeling and MacFie, 1995) and evaluated
in three replicates. The analysis was performed in 5 sequential sips,
where each one last 15 s. For each sip, the panelist imbibed the coffee,
moved it around in their mouths for 3 s, and swallowed it. The evaluation
continued until no sensation was perceived or 15 s had passed.
Data collection began upon tasting and consisted of selecting the
most dominant attribute (among the eight attributes) perceived at
that time using SENSOMAKER Software (Nunes and Pinheiro, 2012)
and plotting as TDS curves showing the percentage of subjects which se-
lected the attribute as dominant at a specific time — that is, the domi-
nance rate (Pineau et al., 2009).
3. Result
3.1. Quantification and characterization of microbial population
The mesophilic aerobic bacteria, LAB, and yeast populations found
during coffee processing at the two farms, Lavras (L) and Monte Carmelo
(MC), are shown in Fig. 1. The yeast population in coffee from L farm
ranged from 2.5 to 4.9 log CFU/g and that for MC farm ranged from 2
to 4.8 log CFU/g.
The population of mesophilic bacteria varied from 3.8 to 8.5 log CFU/g
at L farm and from 5.4 to 7.4 log CFU/g at MC farm (Fig. 1). The LAB pop-
ulation varied from 2.6 to 5.7 log CFU/g at L farm and from 3.4 to
4.5 log CFU/g at MC farm.

Table 1
Yeasts and bacteria associated with fermentation and drying of coffee beans at Lavras farm (L) and Monte Carmelo farm (MC).

Species identification Population (log CFU/g) of the isolates from fruit coffee before and during processing

Fermentation (hours) Drying (hours)

Coffee fruit 0 6 12 24 36 48 60 112 224 336

Yeast L farm
H. uvarum 2.9 3.7 3.4 4.1 3.6 3.8 b2 b2 2.8 3.0 3.6
M. caribbica 4.4 4.2 3.6 3.8 3.3 4.9 2.9 2.5 3.5 3.9 4.0
P. fermentans 3.4 2.3 2.0 2.0 b2 b2 b2 b2 b2 3.0 b2
D. hansenii b2 b2 2.6 b2 b2 b2 b2 b2 b2 b2 b2
T. delbrueckii b2 b2 b2 b2 b2 b2 b2 b2 b2 2.3 b2
C. railenensis 3.1 b2 b2 b2 b2 b2 b2 b2 b2 b2 b2
C. quercitrusa b2 2.3 b2 b2 b2 b2 b2 b2 b2 b2 b2
W. ciferrii b2 b2 b2 b2 b2 b2 b2 b2 b2 b2 3.6
Yeast MC Farm
H. uvarum 3.0 2.5 b2 b2 b2 b2 b2 b2 b2 b2 ⁎–
M. caribbica 3.8 b2 b2 b2 4.0 b2 b2 b2 b2 b2 –
P. fermentans 2.3 b2 2.0 2.5 b2 b2 b2 b2 b2 b2 –
T. delbrueckii 5.3 b2 b2 b2 4.8 4.5 3.8 4.3 2.7 2.9 –
C. glabrata 2.8 2.3 b2 b2 b2 b2 b2 b2 b2 b2 –
W. anomalus b2 3.4 b2 b2 b2 b2 b2 b2 b2 b2 –
Bacteria L farm
Enterococcus sp. b2 b2 b2 b2 b2 b2 b2 b2 b2 4.6 b2
O. pseudogrignonense b2 4.4 b2 b2 b2 b2 b2 b2 b2 b2 3.6
C. taichungense 4.9 5.0 b2 b2 b2 b2 b2 b2 b2 b2 2.8
C. bovis b2 5.0 4.0 b2 b2 b2 b2 b2 b2 b2 b2
E. persicina 4.9 4.5 4.0 5.5 b2 b2 6.1 6.1 b2 b2 b2
S. warneri 5.2 5.6 6.1 6.7 7.1 8.5 7.2 6.6 6.1 b2 b2
Pseudomonas sp. 4.3 b2 b2 b2 b2 b2 b2 b2 b2 b2 b2
P. amylolyticus b2 b2 b2 b2 b2 b2 b2 b2 b2 b2 2.6
Curtobacterium sp. 4.0 b2 b2 6.2 b2 b2 b2 b2 b2 b2 2.9
K. oxytoca 4.3 b2 b2 b2 5.6 b2 b2 b2 b2 b2 b2
B. amyloliquefaciens 4.8 b2 b2 b2 b2 b2 b2 b2 b2 b2 b2
E. hermannii 4.0 b2 b2 b2 b2 b2 b2 b2 b2 b2 b2
Bacteria MC farm
E. asburiae 4.7 5.3 6.0 6.2 7.3 7.4 6.8 5.9 6.2 b2 –
E. ludwigii 5.7 4.3 b2 b2 b2 b2 b2 b2 4.5 b2 –
C. pallidum 4.6 b2 b2 b2 b2 b2 b2 b2 b2 b2 –
P. agglomerans b2 4.0 b2 b2 b2 b2 b2 b2 b2 b2 –
P. dispersa 5.0 4.7 b2 b2 b2 b2 b2 4.6 b2 b2 –
K. oxytoca b2 5.6 b2 b2 b2 b2 b2 b2 b2 b2 –
S. marcescens 4.7 b2 b2 b2 b2 b2 b2 b2 b2 b2 –
Microbacterium sp. b2 4.8 5.0 b2 b2 b2 b2 b2 b2 b2 –
M. laevaniformans b2 4.8 b2 b2 b2 b2 b2 b2 b2 b2 –
W. cibaria b2 b2 6.1 6.5 b2 b2 b2 b2 b2 b2 –
Kocuria sp. 4.8 b2 b2 b2 b2 b2 b2 b2 b2 b2 –
L. mesenteroides b2 b2 6.0 b2 b2 b2 b2 5.9 4.2 5.4 –
⁎ Processing ends with 224 h in Monte Carmelo farm.
106 S.R. Evangelista et al. / International Journal of Food Microbiology 210 (2015) 102–112

A total of 251 microbial isolates were obtained from the samples of L
farm: 59.4% yeasts and 40.6% bacteria (50.9% Gram-positive, and 49%
Gram-negative). A total of 184 isolates were obtained from MC farm:
33.7% yeasts and 66.3% bacteria (33.6% Gram-positive, and 66.3%
Gram-negative).
The isolates were grouped by phenotypic characteristics (cell mor-
phology and biochemical features) (data not shown), and 236 selected
isolates were subjected to rep-PCR for genotypic characterization.
Representative isolates from each rep-PCR-profile were identified by se-
quencing. The identified species and the population during processing
are shown in Table 1.
The yeast species present in the coffee beans of L farm were
Hanseniaspora uvarum (accession no KM402039–KM402045),
Meyerozyma caribbica (accession no KM402046–KM402058),
P. fermentans (accession no KM402059–KM402064), Debaryomyces
hansenii (accession no KM402067), T. delbrueckii (accession

Fig. 2. PCR–DGGE patterns of the yeast and bacteria communities present during coffee wet processing at Lavras farm. Prokaryote (A) and Eukaryote (B). S = coffee cherries, 0 = 0 h, 1 =
4 h, 2 = 15 h, 3 = 22 h, 4 = 27 h, 5 = 40 h, 6 = 50 h, 7 = 140 h, 8 = 236 h, 9 = 360 h, H2O = water fermentation tank.
 S.R. Evangelista et al. / International Journal of Food Microbiology 210 (2015) 102–112 107

no KM402068–KM402082), Candida railenensis (accession no
KM402083), Candida quercitrusa (accession no KM402088), and
Wickerhamomyces ciferrii (accession no KM402084) (Table 1).
M. caribbica was the most prevalent yeast in the coffee fruit, during
the fermentation and drying reached a maximum population of
4.9 log CFU/g at 36 h of fermentation. It was followed by
H. uvarum with a population of 4.1 log CFU/g at 36 h of
fermentation.
The yeast species present in the coffee beans of MC farm were
H. uvarum (accession no KM402039–KM402045), M. caribbica (accession
no KM402046–KM402058), P. fermentans (accession no KM402059–
KM402064), T. delbrueckii (accession no KM402068–KM402082),

Fig. 3. PCR–DGGE patterns of the yeast and bacteria communities present during coffee wet processing at Monte Carmelo farm. Prokaryote (A) and Eukaryote (B). S = coffee cherries, 0 =
0 h, 1 = 9 h, 2 = 20 h, 3 = 28 h, 6 = 54 h, 7 = 174 h, 8 = 202 h, H2O = water fermentation tank.
108 S.R. Evangelista et al. / International Journal of Food Microbiology 210 (2015) 102–112

Table 2
Species identified by PCR–DGGE using universal primers for yeast and bacteria.

Bands Acess number Similarity (%) Prokaryote

1 AF515228 95 Leuconostoc pseudomesenteroides

2,3, 27, 28, AB854267 98 Weissella confusa
4, 5 HM756486 97 Lysinibacillus fusiformis
6 KF673524 97 Lactobacillus fermentum
7, 8, 11, 12, 15, 16 KF031439 100 Uncultured bacterium
9, 10 HQ683968 97 Klebsiella oxytoca
13,14 JX120129 98 Actinobacterium sp.
29, 30, 31, 32, 40, 41, 47, 48 FJ406528 95 Uncultured bacterium
33 JX315564 94 Acinetobacter schindleri
34, 35 HG798481 100 Lactococcus lactis
36, 37 KF453765 97 Enterobacteriaceae bacterium
38, 39 KF891342 97 Enterobacter cloacae
42, 43, 44 KC430956 95 Enterobacter sp.
45, 46 CP002272 97 Enterobacter lignolyticus
49 JX242698 94 Uncultured actinomycete

Eukaryote
17, 18, 19, 20, 50, 51, 52 EU568995 95 Meyerozyma caribbica
21, 22, 58, 59 JQ417238 98 Mitchella repens
23, 24, 60, 61, 62 KC510080 100 Pichia fermentuns
25, 26, 63, 64 AY314792 100 Uncultured fungus
53 EU650386 98 Coffea arabica
54, 55 AY520372 97 Candida sp.
56, 57 HE799671 97 Torulaspora delbrueckii

C. glabrata (accession no KM402089–KM402091), and W. anomalus (ac-
cession no KM402085–KM402087) (Table 1). T. delbrueckii was the
predominant yeast. It was detected in the coffee fruit with a high
population of 5.3 log CFU/g and during the fermentation the maxi-
mum population was of 4.9 log CFU/g. M. caribbica was the second
most significant yeast found in the coffee fruit (3.8 log CFU/g) and
during fermentation with a population of 4 log CFU/g.
The diversity of bacterial species was greater than that of the yeast: 23
bacteria and 10 yeast species were identified (Table 1). Klebsiella oxytoca
(accession no KM402125–KM402126) was the only bacterial species
found at both farms. Enterococcus sp. (accession no KM402100),
Ochrobactrum pseudogrignonense (accession no KM402101–KM402102),
Chryseobacterium taichungense (accession no KM402103–KM402104),
C. bovis (accession no KM402105–KM402106), Erwinia persicina (acces-
sion no KM402108–KM402111), Staphylococcus warneri (accession no
KM402118–KM402119), Pseudomonas sp. (accession no KM402120),
Paenibacillus amylolyticus (accession no KM402121), Curtobacterium
sp. (accession no KM402123–KM402124), Bacillus amyloliquefaciens
(accession no KM402133) and Escherichia hermannii (accession no
KM402097), were found in samples from L farm (Table 1). S. warneri
was the most prevalent bacteria, and was identified in the coffee
fruit and during fermentation, reaching maximum population of
8.5 log CFU/g at 36 h of fermentation. E. persicina was the second most
significant specie reaching the maximum value of 5.5 CFU/g at 12 h of
fermentation.
Enterobacter asburiae (accession no KM402092–KM402096),
E. ludwigii (accession no KM402098–KM402099), C. pallidum (accession
no KM402107), Pantoea agglomerans (accession noKM402112),
P. dispersa (accession no KM402113–KM402117), Serratia marcescens
(accession no KM402127), Microbacterium sp. (accession no

Table 3
Organic acids present in coffee during wet processing at Lavras (A) and Monte Carmelo.

Time (hours) Compounds (g/Kg)

Citric Malic Succinic Lactic Acetic

L farm
Fermentation tank
0 0.08 ± 0.02 0.73 ± 0.35 0.48 ± 0.25 nd Nd
24 0.02 ± 0.01 0.16 ± 0.01 0.05 ± 0.01 0.62 ± 0.02 Nd
48 0.07 ± 0.02 0.09 ± 0.01 nd 0.88 ± 0.07 0.01 ± 0.00
Drying
60 0.10 ± 0.02 0.09 ± 0.03 0.02 ± 0.00 1.07 ± 0.02 Nd
224 0.05 ± 0.00 0.08 ± 0.00 0.02 ± 0.00 1.09 ± 0.09 Nd
336 0.20 ± 0.01 0.20 ± 0.01 0.08 ± 0.00 2.34 ± 0.45 Nd

MC farm
Fermentation tanks
0 0.18 ± 0.03 0.14 ± 0.00 0.40 ± 0.03 nd Nd
24 0.20 ± 0.01 0.06 ± 0.00 0.14 ± 0.02 0.34 ± 0.10 Nd
48 0.08 ± 0.00 0.02 ± 0.00 0.05 ± 0.00 0.57 ± 0.05 0.02 ± 0.00
Drying
60 0.03 ± 0.00 0.01 ± 0.00 0.04 ± 0.00 1.03 ± 0.19 Nd
112 0.02 ± 0.00 nd 0.12 ± 0.01 1.40 ± 0.12 Nd
224 0.04 ± 0.01 nd 0.09 ± 0.00 1.08 ± 0.09 Nd

nd = not detected; ± = Standard deviation.

 S.R. Evangelista et al. / International Journal of Food Microbiology 210 (2015) 102–112 109

KM402128–KM402129), M. laevaniformans (accession no KM402130),
Weissella cibaria (accession no KM402131), Kocuria sp. (accession no
KM402132), and Leuconostoc mesenteroides (accession noKM402134)
were the bacteria isolated from MC farm (Table 1). E. asburiae was the
most prevalent bacteria found in coffee fruit and during fermentation
reaching maximum value of 7.4 log CFU/g at 36 h of fermentation.
L. mesenteroides was the second most significant bacteria identified
reaching the maximum value of 6 CFU/g at 12 h of fermentation. The mi-
crobiota identified was the same in both years analyzed. However, var-
iation of less than 5% was found in the counting.
curves are presented in Fig. 4. In the coffee from L farm, sip 3 had dom-
inant herbaceous and citric sensation characteristics, and sips 2 and 5
had a dominant citric sensation characteristic; in sips 1 and 4, no sensa-
tion characteristic was dominant (Fig. 4A). In the coffee from MC farm,
sips 1 and 3 had a dominant citric sensation characteristic, sips 2 and
4 had a dominant herbaceous sensation characteristic, and sip 5 had a
dominant nuts sensation characteristic (Fig. 4B). The sensations choco-
late, bitter chocolate, caramel, tobacco, and butter did not trigger the
most attention from the panelists. These results are represented in the
graph below the line of significance (Fig. 4B).

3.2. Culture-independent microbiological analysis using PCR–DGGE 4. Discussion

Figs. 2 and 3 show the PCR–DGGE fingerprints of the bacteria and
yeast communities. In general, species of bacteria present in the coffee
before fermentation remained throughout the process and were also
present in the water tank at the end of fermentation.
The sequencing of the bands from L farm indicated the presence of
Leuconostoc pseudomesenteroides (band 1), Weissella confusa (bands 2
and 3), Lysinibacillus fusiformis (bands 4 and 5), uncultured bacteria
(bands 7, 8, 11, 12, 15, and 16), Lactobacillus fermentum (band 6),
The mesophilic aerobic bacteria were dominant throughout the pro-
cess and showed a high level of diversity (Figs. 1–3). A difference in the
diversity of species of the prokaryotic and eukaryotic groups was ob-
served among the regions studied, the main species of yeast and
Table 4
Relative percentage of volatile compounds identified in green and roasted coffee.

K. oxytoca (bands 9 and 10), and Actinobacterium (bands 13 and 14) Compounds Green (%) Roasted (%) Reference target
(Fig. 2A and Table 2). MC farm samples presented W. confusa (bands compoundsa
Coffee Coffee Coffee L Coffee
27 and 28), uncultured bacteria (bands 29, 30, 31, 32, 40, 41, 47, and L farm MC farm farm MC farm
48), Acinetobacter schindleri (band 33), Lactococcus lactis (bands 34
Ketones
and 35), Enterobacteriaceae bacterium (band 36 and 37), Enterobacter 2.3-butanedione nd nd 0.07 nd Buttery
cloacae (bands 38 and 39), Enterobacter sp. (bands 42, 43, and 44), 2-nonanone nd nd 0.80 1.28
E. lignolyticus (bands 45 and 46), and uncultured actinomycete (band Total ketones nd nd 0.87 1.28
49) (Fig. 3A and Table 2). Alcohols
The yeast species found in the coffee cherries before being placed Methanol 16.10 12.24 0.10 0.09
into the tank remained throughout the fermentation process and 1-propanol nd 1.17 nd nd
some new species appear (Figs. 2B and 3B and Table 2). Both MC and 1-pentanol nd nd 1.35 1.24 Green
2-heptanol nd nd 1.83 2.11
L farms showed the presence of M. caribbica (bands 17, 18, 19, 20, 50, 3-methyl-1-pentanol 0.95 nd 0.55 0.67
51, and 52), Candida sp. (bands 54 and 55), T. delbrueckii (bands 56 1-hexanol nd 1.79 nd nd
and 57), P. fermentans (bands 23, 24, 60, 61, and 62), and uncultured a-Terpineol 1.03 nd nd nd
fungi (bands 25, 26, 63, and 64). The plant species C. arabica (band 2-phenylethanol 1.41 0.65 0.08 0.06 Floral
Total alcohols 19.49 15.85 4.91 4.18
53) and Mitchella repens (bands 21, 22, 58 and 59) were also identified
because the primers NL1 and LS2 are universal. Therefore, they can am- Aldehydes
plify the DNA of various eukaryotic organisms, including plants. Acetaldehyde 0.17 0.28 0.10 0.14 Acrid/egg
Hexanal nd nd 0.01 nd Green
Octanal nd nd 0.44 0.29
3.3. Chemical analyses Nonanal nd nd 0.55 1.06
Butyraldehyde 0.91 nd 0.66 0.47
The main acid present in the coffee during fermentation and drying Decyl aldehyde nd nd 0.34 0.32
was lactic acid (Table 3), which showed increasing concentrations and Total aldehydes 1.07 0.28 2.10 2.29

reached a maximum value of 2.33 g/kg for L farm, while for MC farm, Acids
the maximum concentration was 1.07 g/kg. Acetic acid was detected Isobutyric acid nd 3.11 8.15 7.45
in the coffee (0.02 g/kg) only at the end of fermentation in both exper- Hexanoic acid 1.44 2.36 0.11 0.09
Nonanoic acid 0.80 0.95 0.02 0.02
iments. Citric, malic and succinic acids were detected during fermenta- Propanoic acid nd nd 0.54 0.35
tion tank. Propionic and butyric acids, which may impair the sensory Total acids 2.23 6.42 8.81 7.1
characteristics of coffee were not detected.
Esters
A total of 30 volatile compounds were detected via HS–SPME/GC Propyl acetate 0.52 nd 0.61 0.76
(Table 4). Among these compounds, 18 were detected in green coffee Ethyl butyrate 3.27 0.80 nd nd
and 25 were detected in roasted coffee. The composition of green coffee Diethyl malonate nd nd 0.28 0.21
beans from the two farms was different: the green coffee from L farm Phenyl acetate nd nd 0.32 0.29
Furfuryl acetate 4.13 nd 2.38 2.72 Nutty
showed alcohol (methanol), esters (ethyl butyrate and furfuryl acetate),
Diethyl succinate nd 1.14 nd nd
and acids (hexanoic and nonanoic acid) as the main compounds, while Total esters 7.92 1.95 3.59 3.98
the green coffee from MC farm showed alcohol (methanol), acids
Phenol
(isobutyric and hexanoic acid), and furan (furfuryl alcohol and furfural).
Guaiacol nd nd 0.58 0.64 Burnt
The composition of the majority of volatile compounds in roasted coffee
was similar. The main compounds were furans (furfuryl alcohol and fur- Furans
Furfuryl alcohol nd 2.72 21.93 19.03 Burnt
fural), acids (isobutyric acid), and alcohol (1-pentanol and 2-heptanol).
Furfural nd 2.70 13.26 13.90 Almond/bitter
5-methylfurfural 0.94 0.58 0.42 0.77 Caramel
3.4. Sensorial analyses Total furans 0.94 5.99 35.61 33.70
Total GC area 7163 9870 288,508 215,900
The most relevant attributes of the coffee were selected in the previ- nd = not detected.
ous analysis and subsequently evaluated using TDS analysis. The TDS a
Czerny and Grosch (2000), Gonzalez-Rios et al. (2007).
110 S.R. Evangelista et al. / International Journal of Food Microbiology 210 (2015) 102–112

bacteria were not the same between them (Table 1; Figs. 1 and 2). This
difference could have been due to environmental temperature, humid-
ity, composition of the remaining pulp and mucilage surrounding the
coffee beans, and the altitude (Leong et al., 2014; Silva et al., 2008).
The majority of the microbiota present throughout the wet process-
ing was also detected in coffee fruit. The wet processing favored bacteria
development that showed a high increase of the population in both re-
gions. Studies of the microbiota from coffee undergoing different types
of processing also observed that some species already are naturally
present in the fruit (Silva et al., 2000, 2008). Therefore, harvesting
methods and fermentation process will directly interfere with microbi-
ota (Avallone et al., 2001; Silva et al., 2000; Vilela et al., 2010).
H. uvarum and T. delbrueckii were detected in higher numbers in cof-
fee from L farm and MC farm, respectively, and were also reported in
others studies that examined the wet and semi-dry processes
(Masoud et al., 2004; Vilela et al., 2010). The presence of some yeasts,

Fig. 4. TDS curves of coffee from L farm (A) and coffee from MC farm (B). Chocolate ( ), Bitter chocolate ( ), Caramel ( ), Citric ( ), Tobacco ( ), Butter ( ),
Herbaceous ( ), Nuts ( ). Chance level ( ) represents the dominance rate that an attribute can obtain by chance (one/number of attributes) Significance level ( ) expresses
the smallest value of the proportion that is significantly (p = 0.05) higher than the chance level. Each slip duration of 15 s.
 S.R. Evangelista et al. / International Journal of Food Microbiology 210 (2015) 102–112
such as Saccharomyces, Pichia, Candida and Meyerozyma provides bene-
fits for coffee processing due to their important role in the degradation
of mucilage rich in pectin and because they may inhibit the growth of
mycotoxigenic fungi (Masoud and Jespersen, 2006; Massawe and Lifa,
2010; Silva et al., 2012; Evangelista et al., 2014). Therefore, these yeasts
are interesting strains that could be used as starter cultures (Evangelista
et al., 2014). P. fermentans was detected in both coffee regions. This spe-
cie produces volatile compounds that might interfere with the quality of
coffee beverages (Pereira et al., 2014).
Some bacteria identified in this work belong to genera commonly
found during coffee processing, such as Erwinia, Klebsiella, Leuconostoc,
Weissella, Enterococcus, Enterobacter, Serratia, and Bacillus. These genera
have been identified during the natural, semi-dry, and wet fermentation
of coffee (Avallone et al., 2001; Silva et al., 2000, 2008; Vilela et al.,
2010). L. mesenteroides detected in coffee from MC farm is an important
lactic acid bacteria present during the wet processing and species of
Erwinia and Klebsiella are reported as important producers of pectinases
(Avallone et al., 2001).
Some species of yeast and bacteria found using the culture-
dependent method were not detected via PCR–DGGE (Figs. 2 and 3
and Table 2). Masoud et al. (2004) also reported that some microbial
species were only detected using the culture-dependent method in
the samples taken during wet processing. This fact could be explained
by the difficulty of obtaining high-quality DNA suitable for PCR directly
from the samples. The initial template DNA and template competition
may affect the detection of rare microorganisms in the microbial popu-
lation (Muyzer et al., 1993). All yeasts species detected in DGGE gels
were also isolated from plating.
Some bacterial species were detected only through PCR–DGGE.
L. pseudomesenteroides, W. confuse, L. fusiformis, L. fermentum and
Actinobacterium sp. were the species detected in L farm. A. schindleri,
W. confusa, L. lactis, E. bacterium, E. cloacae and E. lignolyticus were the
species detected in MC farm. This may have occurred because their pop-
ulation densities were lower than 2 log CFU/g, making them impossible
to detect via the culture-dependent method (Figs. 2 and 3 and Table 2).
Vilela et al. (2010) also reported that some microbial species were only
detected using the DGGE method in the samples taken during semi-dry
processing. The species L. pseudomesenteroides, W. confusa, L. lactis and
Enterobacter sp. have been found in others studies (Leong et al., 2014;
Vilela et al., 2010). The species L. pseudomesenteroides and W. confusa
contribute in the fermentation process and might inhibit filamentous
fungi growth (Leong et al., 2014). Therefore, molecular methods should
always be used in conjunction with traditional methods to evaluate
biodiversity.
Acetic and lactic acids were produced throughout the fermentation.
Lactic acid was the main acid found (Table 3). The presence of lactic acid
bacteria, such as bacteria of the genera Weissella, Leuconostoc, and Lacto-
bacillus identified throughout the fermentation process, may have con-
tributed to this fact. The other acids detected were malic, citric, and
succinic acids (Table 3). These acids are already naturally present in
the coffee bean, and their concentrations decreased during fermenta-
tion, but succinic acid showed a slight increase at the end of the process.
Succinic acid may be produced by Bacillus spp. (Silva et al., 2012) and by
heterofermentative lactic acid bacteria (LAB) (Swiegers et al., 2005).
The variation of acids during fermentation due to microbial metabo-
lism changed the pH value and influences the microbiota present in the
tank (Table 3). The diffusion of theses acids into the bean could influ-
ence the final beverage flavor and quality (Evangelista et al., 2014;
Silva et al., 2012; Silva, 2014).
The volatile compounds identified in the green coffee beans differed
between the regions studied (Table 4) that may have influenced this dif-
ference is the microbiota involved during the processing of the coffee
beans (Evangelista et al., 2014; Silva et al., 2012). Some compounds
are known to play a role in aroma development during fermentation,
such as the esters that were detected in the green coffee beans in both
regions.
111
However, after roasting, less difference in the volatile composition
was observed (Table 4). The mechanisms of coffee aroma formation
are extremely complex, and there is clearly a wide range of interactions
between all the pathways involved. Maillard reactions occur between
reducing carbohydrates and proteins and are responsible for the forma-
tion of many volatile compounds during roasting (López-Galilea et al.,
2006).
Each coffee sample was evaluated on the following attributes:
aroma, flavor, aftertaste, acidity, body, balance, uniformity, clean cup,
sweetness, and overall and was given a score for each of these attributes.
The coffee of L farm received a score of 81 and MC farm of 80 (data not
shown), being considered as very good special coffee according to SCAA
(2013).
Although both coffees were well accepted by the panelists, some dis-
tinguished flavors were found in the TDS analysis (Fig. 4). The coffee
from L farm presented a nice citrus and herbaceous flavor, and the
coffee from MC farm presented, addition to these flavors, nuts sensation
characteristics. All these sensations are desirable in the final coffee
beverage. The presence of volatile compounds differed between the
regions and affected the beverages' final flavors. The presence of
different microbial species in each region may have contributed to
this difference. The microbial activity during fermentation produces vol-
atile compounds that might influence the final taste of the beverage
(Evangelista et al., 2014).
Some volatile compounds detected in this study are described in
the literature by providing aromatic notes such as caramel, nutty,
burnt, buttery, floral, and almond (Table 4) (Czerny and Grosch, 2000;
Gonzalez-Rios et al., 2007). The presence of furans and ketones were de-
tected in both samples, which contributed to the citric and herbaceous
flavors that were dominant in the sensory analysis (Fig. 4). The presence
of furans can provide herbal or fruity notes, and ketones are described as
providing buttery, caramel-like, musty, mushroom like, or fruity notes
(López-Galilea et al., 2006).
The evaluation of these two coffee-producing regions contributed to
a better understanding of the microbiota present during coffee wet pro-
cessing and some characteristics of this processing. In conclusion, this
study evaluated coffee wet processing, which involved a variety of bac-
teria and yeasts, some of which prevailed during the fermentation pro-
cess. The combination of two techniques — culture-dependent and
independent methods — proved to be efficient for achieving an under-
standing of the microbial population responsible for wet coffee fermen-
tation. Future work should be conducted to evaluate the production of
enzymes by the main species identified and to select appropriate strains
as starter cultures in coffee fermentation.
Acknowledgments
The authors thank the Brazilian agencies Conselho Nacional de
Desenvolvimento Científico e Tecnológico of Brasil (CNPQ), Fundação
de Amparo a Pesquisa do Estado de Minas Gerais (FAPEMIG), and
Coordenação de Aperfeiçoamento de Pessoal de Nível Superior
(CAPES). We also thank the Juliana farm, located in Monte Carmelo
city, and the Resfriado farm, situated in Lavras in the state of Minas
Gerais, Brazil, for collecting samples.
References
Agate, A.D., Bhat, J.V., 1966. Role of pectinolytic yeasts in the degradation of mucilage
layer of Coffea robusta cherries. Appl. Microbiol. 14 (2), 256–260.
Avallone, S., Guyot, B., Brillouet, J.-Marc, Olguin, E., Guiraud, J.-Pierre, 2001. Microbiolog-
ical and biochemical study of coffee fermentation. Curr. Microbiol. 42, 252–256.
Brando, C.H.J., Brando, M.F., 2014. Methods of coffee fermentation and drying. In: Schwan,
R.F., Fleet, G.H. (Eds.), Cocoa and Coffee Fermentation. CRC Taylor & Francis, Boca
Raton, FL, pp. 367–398.
Cocolin, L., Bisson, L.F., Mills, D.A., 2000. Direct profiling of the yeast dynamics in wine fer-
mentations. FEMS Microbiol. Lett. 189, 81–87.
Czerny, M., Grosch, W., 2000. Potent odorants of raw Arabica coffee. Their changes during
roasting. J. Agric. Food Chem. 48 (3), 868–872.
112 S.R. Evangelista et al. / International Journal of Food Microbiology 210 (2015) 102–112
Devereux, R., Willis, S.G., 1995. Amplification of ribosomal RNA sequences. In: Akkermans,
A.D.L., Van Elsas, J.D., Bruijn, F.J. (Eds.), Molecular microbial ecology manual. Kluwer
Academic, Dordrecht, pp. 1–11.
Evangelista, S.R., Silva, C.F., Miguel, M.G.P.C., Cordeiro, C.S., Pinheiro, A.C.M., Duarte, W.F.,
Schwan, R.F., 2014. Improvement of coffee beverage quality by using selected yeasts
strains during the fermentation in dry process. Food Res. Int. 61, 183–195.
Frank, H.A., Lum, N.A., Cruz, A.S.D., 1965. Bacteria responsible for mucilage-layer decom-
position in kona coffee cherries. Appl. Microbiol. 13, 201–207.
Gonzalez-Rios, O., Suarez-Quiroza, M.L., Boulanger, R., Barel, M., Guyot, B., Guiraud, J.P.,
Schorr-Galindo, S., 2007. Impact of “ecological” post-harvest processing on coffee
aroma: II. Roasted coffee. J. Food Compos. Anal. 20, 297–307.
Holt, J.G., Krieg, N.R., Sneath, P.H.A., Staley, J.T., Williams, S.T., 1994. Bergey's manual of de-
terminative bacteriology. ninth ed. W. & Wilkins, Baltimore.
Kurtzman, C.P., Fell, J.W., Boekhout, T., 2011. The Yeasts, a Taxonomic Study. fourth ed.
Elsevier, Amsterdan.
Leong, K., Chen, Y., Pan, S., Chen, J., Wu, H., Chang, Y., Yanagida, F., 2014. Diversity of lactic
acid bacteria associated with fresh coffee cherries in Taiwan. Curr. Microbiol. 68,
440–447.
López-Galilea, I., Fournier, N., Cid, C., Guichard, E., 2006. Changes in headspace volatile
concentrations of coffee brews caused by the roasting process and the brewing pro-
cedure. J. Agric. Food Chem. 54, 8560–8566.
Masoud, W., Jespersen, L., 2006. Pectin degrading enzymes in yeasts involved in fermen-
tation of Coffea arabica in East Africa. Int. J. Food Microbiol. 110, 291–296.
Masoud, W., Cesar, L.B., Jespersen, L., Jakobsen, M., 2004. Yeast involved in fermentation of
Coffea arabica in East Africa determined by genotyping and by direct denaturating
gradient gel electrophoresis. Yeast 21, 549–556.
Massawe, G.A., Lifa, S.J., 2010. Yeasts and lactic acid bacteria coffee fermentation starter
cultures. Int. J. Postharvest Technol. Innov. 2, 41–80.
Muyzer, G., Waal, E.C., Uitterlinden, A.G., 1993. Profile of complex microbial populations
by denaturing gradient gel electrophoresis analysis of polymerase chain reaction:
amplified genes coding for 16S rDNA. Appl. Environ. Microbiol. 59, 695–700.
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, 168–186.
Nunes, C.A., Pinheiro, A.C.M., 2012. Sensomaker Free user-friendly interface for sensory
analysis Available from: www.ufla.br/sensomaker (Accessed at: Out. 2012).
Ovreas, L., Forney, L., Daae, F.L., 1997. Distribution of bacterioplankton in meromictic Lake
Saelenvannet, as determined by denaturing gradient gel electrophoresis of PCR-
amplified gene fragments coding for 16S rRNA. Appl. Environ. Microbiol. 63,
3367–3373.
Pereira, G.V.M., Soccol, V.T., Pandey, A., Medeiros, A.B.P., Lara, J.M.R.A., Gollo, A.L., Soccol,
C.R., 2014. Isolation, selection and evaluation of yeasts for use in fermentation of cof-
fee beans by the wet process. Int. J. Food Microbiol. 188, 60–66.
Petisca, C., Pérez-Palacios, T., Farah, A., Pinho, O., Ferreira, I.M.P.L.V.O., 2013. Furans and
other volatile compounds in ground roasted and espresso coffee using headspace
solid-phase microextraction: effect of roasting speed. Food Bioprod. Process. 91,
233–241.
Pineau, N., Schlich, P., Cordelle, S., Mathonnière, C., Issanchou, S., Imbert, A., Rogeaux, M.,
Etiévant, P., Koster, E., 2009. Temporal dominance of sensations: construction of the
TDS curves and comparison with time–intensity. Food Qual. Prefer. 20, 450–455.
Ramos, C.L., de Almeida, E.G., Pereira, G.V.D.M., Cardoso, P.G., Dias, E.S., Schwan, R.F., 2010.
Determination of dynamic characteristics of microbiota in a fermented beverage pro-
duced by Brazilian Amerindians using culture-dependent and culture-independent
methods. Int. J. Food Microbiol. 140, 225–231.
Senguna, I.Y., Nielsen, D.S., Karapinar, M., Jakobsen, M., 2009. Identification of lactic acid
bacteria isolated from Tarhana, a traditional Turkish fermented food. Int. J. Food
Microbiol. 135, 105–111.
Silva, C.F., 2014. Microbial activity during coffee fermentation. In: Schwan, R.F., Fleet, G.H.
(Eds.), Cocoa and Coffee Fermentation. CRC Taylor & Francis, Boca Raton, FL,
pp. 398–423.
Silva, C.F., Schwan, R.F., Dias, E.S., Wheals, A.E., 2000. Microbial diversity during matura-
tion and natural processing of coffee cherries of Coffea arabica in Brazil. Int. J. Food
Microbiol. 60, 251–260.
Silva, C.F., Batista, L.R., Abreu, L.M., Dias, E.S., Schwan, R.F., 2008. Succession of bacterial
and fungal communities during natural coffee (Coffea arabica) fermentation. Food
Microbiol. 25, 951–957.
Silva, C.F., Vilela, D.M., Souza, C.C., Duarte, W.F., Dias, D.R., Schwan, R.F., 2012. Evaluation
of a potential starter culture for enhance quality of coffee fermentation. World
J. Microbiol. Biotechnol. 29, 235–247.
Sindicato da Indústria de Café do Estado de Minas Gerais (Sindcafé-MG), 2014. Available
in: http://sindicafe-mg.com.br/plus/modulos/conteudo/?tac=tipos-de-cafe (Accessed
at: Jan. 2014).
Specialty Coffee Association of America (SCAA), 2013. Available in: http://www.scaa.org
(Accessed at: April. 2013).
Swiegers, J.H., Bartowsky, E.J., Henschke, P.A., Pretorius, I.S., 2005. Yeast and bacterial
modulation of wine aroma and flavor. Aust. J. Grape Wine Res. 11, 139–173.
Vilela, D.M., Pereira, G.V.D.M., Silva, C.F., Batista, L.R., Schwan, R.F., 2010. Molecular ecolo-
gy and polyphasic characterization of the microbiota associated with semi-dry proc-
essed coffee (Coffea arabica L.). Food Microbiol. 27, 1128–1135.
Wakeling, I.N., Macfie, H.J.H., 1995. Designing consumer trials balanced for first and
higher orders of carry-over effect when only a subset of k samples from t may be test-
ed. Food Qual. Prefer. 6, 299–308.