Food Research International 136 (2020) 109588

Contents lists available at ScienceDirect

Food Research International

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

Coffee flavour modification through controlled fermentations of green T

coffee beans by Saccharomyces cerevisiae and Pichia kluyveri: Part I. Effects
from individual yeasts
⁎
Chenhui Wanga, Jingcan Sunb, Benjamin Lassabliereb, Bin Yub, Shao Quan Liua,c,
a
Department of Food Science and Technology, National University of Singapore, Science Drive 2, 117546, Singapore
b
Mane SEA PTE LTD, Biopolis Drive 3, 138623, Singapore
c
National University of Singapore (Suzhou) Research Institute, No. 377 Linquan Street, Suzhou Industrial Park, Suzhou, Jiangsu 215123, China

A R T I C LE I N FO A B S T R A C T

Keywords: Direct fermentations of sterilised green coffee beans by monocultures of Saccharomyces cerevisiae and Pichia
Coffee flavour biotransformation kluyveri were investigated for coffee flavour biotransformation. During fermentation, fruity esters were gener-
Wine yeasts ated in the green coffee beans by yeasts. 2-Phenylethyl acetate was elevated by 1.1 mg/kg and 0.03 mg/kg in P.
Isoamyl acetate kluyveri- and S. cerevisiae-fermented green beans, respectively, as compared to the untreated sample. Ethyl oc-
2-Phenylethyl acetate
tanoate (0.51 mg/kg) and isoamyl acetate (1.69 mg/kg) only existed in S. cerevisiae- and P. kluyveri-fermented
Fruity esters
Coffee roasting
green beans, respectively. After roasting, higher levels of 2-phenylethyl acetate were detected in fermented
coffees, and ethyl octanoate was found only in the S. cerevisiae-fermented sample, despite the loss of isoamyl
acetate in P. kluyveri-fermented coffees during roasting. The fruity esters generated by the yeasts during green
coffee bean fermentations were directly transferred to the volatile profiles formed after roasting and enhanced
the fruity attribute in the roasted coffees, with a more noticeable effect observed from S. cerevisiae fermentation.
Higher productions of N-heterocyclic volatiles occurred during roasting of S. cerevisiae-fermented coffees and
contributed to elevated nutty and roasted aromas. S. cerevisiae and P. kluyveri are considered suitable starter
cultures for controlled coffee flavour biotransformation through controlled fermentations of green coffee beans.

1. Introduction roasting process and contribute to the fruity and floral attributes of the
roasted coffee (de Carvalho Neto et al., 2017; Gonzalez-Rios et al.,
Yeast is an important group of the indigenous coffee microorgan- 2007a).
isms both in terms of microbial population and functionalities in on- Coffee aroma modulation associated with yeast metabolism is evi-
farm processing that assists the release of dry green coffee beans from dent from several studies on adding selected yeasts into indigenous
the rest of the coffee cherry matrix (Haile & Kang, 2019). Diverse yeast microbiota of coffee cherries in on-farm processing. For example, the
species from various genera, such as Pichia, Candida, Saccharomyces and inoculation of Saccharomyces cerevisiae and non-Saccharomyces yeasts
Torulaspora, have been detected throughout the coffee cherry proces- from the genera Pichia, Candida, and Torulaspora during spontaneous
sing with an abundance (de Bruyn et al., 2017; de Carvalho Neto et al., coffee cherry fermentation reportedly enhanced coffee sensory qualities
2017; Evangelista, Miguel, Silva, Pinheiro, & Schwan, 2015; Zhang (Bressani, Martinez, Evangelista, Dias, & Schwan, 2018; Evangelista
et al., 2019). Metabolic activities of yeasts contribute to the modulation et al., 2014; Martinez, Bressani, Miguel, Dias, & Schwan, 2017; Pereira
of soluble carbohydrates, organic acids, alcohols, volatile compounds et al., 2015; Ribeiro et al., 2017; Saerens & Swiegers, 2016; Takahashi,
and other substances in the pulp and mucilage of coffee cherries, al- Minami, Kanabuchi, Togami, & Mitsuhashi, 2007). Particularly, P.
tering the flavour-related constituents in green coffee beans through kluyveri was reported to elevate fruity characteristics and S. cerevisiae
substance exchange and subsequently modifying the final roasted coffee contributed to winey and fruity aroma in final roasted coffee (Pereira
flavour (de Carvalho Neto et al., 2017; Gonzalez-Rios et al., 2007b; et al., 2015; Saerens & Swiegers, 2016; Takahashi et al., 2007). S. cer-
Vilela, Pereira, Silva, Batista, & Schwan, 2010). The yeast volatile evisiae is the primary ethanol producer and also plays a part in aroma
metabolites, particularly aromatic esters, can partially remain after the formation in wine fermentation (Stewart, 2017b). P. kluyveri is superior

⁎
Corresponding author at: Department of Food Science and Technology, National University of Singapore, Science Drive 2, 117546, Singapore.
E-mail address: fstlsq@nus.edu.sg (S.Q. Liu).

https://doi.org/10.1016/j.foodres.2020.109588
Received 3 February 2020; Received in revised form 7 July 2020; Accepted 21 July 2020
Available online 24 July 2020
0963-9969/ © 2020 Elsevier Ltd. All rights reserved.
C. Wang, et al. Food Research International 136 (2020) 109588

in the production of aromatic esters but has limited ability to generate 100 °C for 30 min to achieve sterilisation and then cooled to room
ethanol (Benito et al., 2019). The non-Saccharomyces yeast has been temperature. The sterility efficiency was checked with total plate count
demonstrated to enhance the fruity-floral attributes of various types of agar (Oxoid, Basingstoke, UK) plates. The S. cerevisiae and P. kluyveri
alcoholic beverages (Chua, Lu, & Liu, 2018; Lu, Huang, Lee, & Liu, inocula were each inoculated into the sterilised fermentation matrix in
2016) and reported to improve the overall aroma of chocolate through blue-cap bottles for FSc and FPk, respectively, at 1% v/v to reach an
cocoa fermentation (Crafack et al., 2014). The positive findings on the initial cell population of around 5.6 log CFU mL−1. For the Control, the
yeast-mediated coffee cherry fermentation and other fermentation same volume of the 0.9% NaCl solution was added. All the three
processes shine light on the potential for using S. cerevisiae and P. treatments were incubated at 20 °C for 6 days. The fermentation time
kluyveri in controlled fermentation of green coffee beans. was determined with a preliminary experiment for an optimal aroma
Without participation of coffee cherry matrix or indigenous micro- formation. Fermentation juice (liquid phase of the fermentation media,
organisms, controlled fermentation of sterilised green coffee beans is 10 mL) and green coffee beans (4 g) were sampled from the fermen-
distinguished as a direct and targeted coffee flavour biotransformation tation media every 24 h. After the 6 day-incubation, the fermentation
technique (Lee, Cheong, Curran, Yu, & Liu, 2015). The use of com- juice was drained, and the treated green coffee beans were dried at
mercial starter cultures is favourable in this process due to their 45 °C in an oven until the moisture content dropped to 10%. The treated
availability and flavour potential (Takahashi et al., 2007). In the pre- (G-FSc, G-FPk, and G-Control) and untreated (G-Blank) dry green
vious studies on controlled fermentation of green coffee beans, fila- coffee beans, were stored under − 30 °C before analyses.
mentous fungi and lactic acid bacteria have been used as starter cul-
tures and mainly contributed to indirect coffee flavour biomodification 2.3. Cell count and pH measurement
via metabolising flavour precursors during green coffee bean fermen-
tations and changing the directions of pyrolytic volatile production Fermentation liquid (5 mL) and whole green coffee beans (4 g) were
during roasting (Lee et al., 2017; Lee, Cheong, Curran, Yu, & Liu, 2016; vortexed well in a peptone solution (0.1% w/v, 20 mL), serially diluted
Wang et al., 2019; Wang, Sun, Lassabliere, Yu, & Liu, 2020). In contrast, with the peptone solution, and spread-plated on potato dextrose agar
studies on yeast-mediated coffee cherry fermentation demonstrated (Oxoid, Basingstoke, UK) plates. The plates were then incubated at
that S. cerevisiae and P. kluyveri have the ability to produce aromatic 20 °C for 48 h before counting. The pH of the fermentation juice (5 mL)
volatile compounds that had contributed to aroma enhancement in the was measured with a Metrohm FEP20 pH meter (Herisau, Switzerland).
roasted coffee (Pereira et al., 2015; Saerens & Swiegers, 2016;
Takahashi et al., 2007). The direct application of these two yeasts in 2.4. Roasting of green coffee beans
controlled fermentation of green coffee beans could also result in en-
hancement of coffee flavour complexity, which was the objective of the Coffee beans of G-Blank, G-Control, G-FCs, and G-FPk (100 g) were
current study. roasted into R-Blank, R-Control, R-FSc and R-FPk, respectively, with a
In this study, we investigated the yeast modification of coffee fla- FZ-94 Lab Roaster (Coffee-Tech, Moshav Mazliach, Israel) at 150 °C for
vour through direct fermentation of sterilised green coffee beans with 10 min to obtain a light-medium roasting degree by coffee roasting
monocultures of S. cerevisiae and P. kluyveri. The effects of bio- professionals at Bettr Barista PTE LTD. The roasting condition was se-
transformation of green coffee bean constituents and the subsequent lected to develop sensory characteristics that would best differentiate
modifications on final roasted coffee flavour were also assessed. the green coffee beans treatments and to reach a light-medium roast
suitable for coffee sensory evaluation. The roasted coffee beans were
2. Materials and methods stored at − 30 °C prior to further analyses. Roasting degrees of ground
coffee (aperture diameter: 426 μm) were assessed with L* (lightness)
2.1. Yeasts and preparation of inocula values of the roasted ground coffee under D65 illuminant, measured
using a Konica Minolta CM-3500d spectrophotometer (Osaka, Japan).
The two yeasts used in this study were Saccharomyces cerevisiae
MERIT.ferm and Pichia kluyveri FROOTZEN, which were originated 2.5. Compositional analyses of green and roasted coffee beans
from grape juice and purchased from Christian Hansen (Hoersholm,
Denmark) in freeze-dried form. The freeze-dried yeast powders were 2.5.1. Volatile analysis
separately inoculated into sterile yeast nutrient broth (20 g of glucose, The green coffee beans and roasted coffees were ground and sieved
2.5 g of yeast extract, 2.5 g of bacteriological peptone and 2.5 g of malt (aperture diameter: 426 μm) for a headspace-solid phase microextrac-
extract in 1 L of water, adjusted to pH 5.0 with 1 M HCl) and incubated tion-gas chromatography-mass spectrometry (HP-SPME-GC–MS) ana-
statically at 20 °C for 48 h. The pure cultures were mixed with equal lysis on a GC system (mode 7890 N, Agilent Technologies, Palo Alto,
volumes of glycerol in water solution (15% v/v) and stored at −80 °C. USA) as described elsewhere with modifications (Wang et al., 2019).
Prior to fermentation, frozen stock cultures were inoculated into the Ground green and roasted coffee beans (1 g) were spiked with 5 μL of
yeast nutrient broth at 10% v/v and subcultured twice at 20 °C for 48 h. an internal standard solution of methyl salicylate in acetone (155 mg L-
1
The resulted cultures were centrifuged, and the cell pellets were washed ), added into a headspace vial (20 mL), and subjected to the SPME
twice and then reconstituted with an 0.9% NaCl solution to form in- using a Carboxen/ Polydimethylsiloxane fibre (Carboxen/PDMS, 85-μm
ocula with cell populations of around 7.6 log CFU mL−1 for each yeast film thickness) (Supelco, Bellefonte, USA) at 60 °C for 30 min. The in-
culture. jector was operated at 250 °C. Volatile compounds were separated on a
DB-FFAP column (60 m × 0.25 mm, 0.25-μm film thickness) using
2.2. Fermentation and drying of green coffee beans helium as carrier gas (1.2 mL min−1). The oven temperature of column
rose from 50 °C (5 min) to 230 °C (30 min) at 5 °C min−1. The MS
Green coffee beans (Coffea arabica, semi-dry processed) from Mount (mode 5975, Agilent Technologies) was operated in the full scan mode
Sinabung, Indonesia with an original moisture content of around 10% (m/z 40 to 500). Identification of volatile compounds was done by
were purchased from Yong-In Traders PTE LTD (Singapore). The comparing their linear retention indices (LRI) calculated in relation to a
treatments of green coffee beans, including sterilised but unfermented C8–C20 n-alkane mixture (Sigma-Aldrich, Barcelona, Spain) to the lit-
control (Control), fermentation with S. cerevisiae (FSc), and fermenta- erature values and matching their mass spectra with the NIST library
tion with P. kluyveri (FPk) were performed in triplicate. For each 2017 using Agilent MassHunter Quantitative Analysis software for
treatment, a mixture of whole green coffee beans (1 kg) and deionised GC–MS (version B.07.00, Agilent Technologies). The concentrations of
water (2.3 kg) was loaded into blue-cap bottles and autoclaved at identified volatile compounds (mg/kg) were calculated as:

2
C. Wang, et al. Food Research International 136 (2020) 109588

Concentration of volatile = (Concentration of standard × MS Peak coffee tasters with certified Q-grader coffee certificate (one male and
area of volatile) / MS Peak area of standard. one female) from Bettr Barista PTE LTD following methods described
previously with modifications (Wang et al., 2020). The ground roasted
2.5.2. Non-volatile analysis coffee samples (8.25 g, 70% − 75% particles with size < 841 μm) were
Non-volatile extractions and analysis were conducted by following mixed with hot water (150 mL, 93 °C) and tasted following the methods
methods described previously with modifications (Wang et al., 2020). from SCAA (Specialty Coffee Association of America) (SCAA, 2015).
Ground of green coffee beans (G-Blank, G-Control, G-FPk, and G-FSc) The flavour qualities of the coffee brews were evaluated on five aroma
and roasted coffee beans (R-Blank, R-Control, R-FPk, and R-FSc) were attributes, including nutty, roasted, caramel, fruity and smoky, and two
subjected to non-volatile extractions. Soluble carbohydrates and phe- taste attributes, including sweet and acidic, on a scale of 0–10 with 0.5
nolic compounds were extracted together by vortex of ground coffee increments (“0″ indicating “not detected”, “5” indicating “moderate”,
(2 g) in a methanol solution (80% v/v, 3 × 20 mL) for 30 min at “10” indicating “extreme”). The flavour attributes were identified due
200 rpm, and solvent was removed using a rotary evaporator and re- to their distinctiveness in the roasted coffee samples.
constituted with aqueous methanol (80% v/v, 2 mL) (Maness, 2010).
Organic and free amino acids were co-extracted by vortex of the ground 2.7. Statistical analysis
coffee (1 g) in deionised water (10 mL) for 30 min at 200 rpm and
centrifuging to separate the liquid phase (Lee et al., 2016). The extracts Analysis of variance (ANOVA, p < 0.05) and Tukey’s test were
from green and roasted coffee beans, together with the fermentation performed with OriginPro 2018 (OriginLab Corporation, Northampton,
juices collected after the 6-day treatments (J-Control, J-FSc, and J-FPk), USA).
were filtered with 0.2-μm polytetrafluoroethylene (PTFE) membranes.
The soluble carbohydrates, organic acids and phenolic compounds 3. Results
from the extracts of ground green and roasted coffees and fermentation
juices were separated using high performance liquid chromatography 3.1. Yeast growth and changes in pH during green coffee bean fermentation
(HPLC) analyses using a Shimadzu Prominence system (Kyoto, Japan)
(Wang et al., 2020). Non-volatile qualification and quantification were With similar initial cell populations (5.6–5.7 log CFU mL−1),
conducted with authentic standards and their external calibration growth kinetics of S. cerevisiae and P. kluyveri during the 6-day green
curves. Resolution of soluble carbohydrates was performed on a Zorbax coffee bean fermentations are shown in Fig. 1(a). In FSc fermentation,
carbohydrate column (150 mm × 4.6 mm) coupled with a Zorbax NH2 cell counts of S. cerevisiae increased by 2.3 log CFU mL−1 within the
guard column (12.5 mm × 4.6 mm) (Agilent Technologies) at 30 °C first 2 days and by 0.2 CFU mL−1 on day 3. Following a plateau at
using acetonitrile (80% v/v) as the mobile phase (flow rate of around 8.1 log CFU mL−1 from day 4 to day 5, the cell counts of S.
1 mL min−1), and peak detection was performed with an evaporative cerevisiae dropped slightly to 7.8 log CFU mL−1 on the last day. Steady
light scattering detector (ELSD). Identified soluble carbohydrates in- growth of P. kluyveri of 2 log CFU mL−1 was observed from day 0 to day
cluded sucrose, fructose, and glucose, and external calibration curves 3 in FPk fermentation, and cell populations of this non-Saccharomyces
for quantification were constructed in the range of 0.04–10.00 g/L. yeast stabilised from day 3 to day 6. Slight acidification occurred in
Organic acids were separated on a Supelcogel C-610H column both fermentations (Fig. 1(b)), and the pH values of the liquid phase in
(30 mm × 7.8 mm) (Supelco) at 40 °C, using sulfuric acid solution fermentation media dropped from 5.8 to 5.5 in FSc and from 5.8 to 5.7
(0.1%, v/v) flowing at 0.4 mL min−1 as the mobile phase, and peaks in FPk (P < 0.05).
were detected using a photodiode array (PDA) detector at 210 nm.
Identified organic acids included citric, malic, quinic, succinic, lactic, 3.2. Changes of non-volatile compounds in fermented green coffee beans
formic, and alpha-ketoglutaric (alpha-KT) acids, and external calibra-
tion curves for quantification were constructed in the range of Contents of the non-volatile compounds in the untreated green
0.05–2.00 g/L.
Phenolic compounds were resolved together using a Zorbax Eclipse
C18 column (150 mm × 4.6 mm) at 40 °C with gradient elution of a
binary mobile phase containing an acetic acid solution (0.1% v/v) and
pure methanol as described in the previous study (Wang et al., 2020).
Phenolic compounds were detected with a PDA detector at 320. Iden-
tified phenolic compounds included chlorogenic, caffeic, ferulic, and
quinic acids, and external calibration curves for quantification of
chlorogenic acid and the other identified phenolic acids were con-
structed in the range of 0.008–0.8 g/L and 0.003–0.300 g/L, respec-
tively.
Free amino acid analysis was performed on an ARACUS Amino Acid
Analyzer (MembraPure, Berlin, Germany) with a pre-designed hydro-
lysate analysis method (Chua, Lua, & Liu, 2017). Resolution was
achieved on a lithium-cation exchange column, followed by the post-
column derivatization with ninhydrin. Amino acid detection was made
at 440 nm for proline and 570 nm for others. Identification and quan-
tification by calibration factors were conducted with an authentic
standards mixture containing 100 nmol of each amino acid. Alcohol Fig. 1. Yeast growth (a) and changes of pH (b) during the fermentations of
contents of the fermentation juices were determined with an Anton green coffee beans with S. cerevisiae and P. kluyveri. changes in cell
Paar density meter DMA 4500 M coupled with Alcohlyzer ME (GmbH, count during the fermentation with S. cerevisiae (FSc); changes in cell
Austria) count during the fermentation with P. kluyveri (FPk); changes in pH
during the fermentation with S. cerevisiae (FSc); changes in pH during
2.6. Evaluation of flavour profiles of roasted coffees the fermentation with P. kluyveri (FPk). Cell counts and pH are the mean values
of triplicate fermentations (n = 3), with error bars representing the standard
Flavour profiles of roasted coffees were analysed by two expert deviation of the mean.

3
C. Wang, et al. Food Research International 136 (2020) 109588

Fig. 2. Effects of S. cerevisiae and P. kluyveri fermentations for 6 days on sugars (a), organic acids (b), phenolic compounds (c), and amino acids (d) in green coffee
beans. untreated green coffee beans (G-Blank); sterilised but unfermented control green coffee beans (G-Control); P. kluyveri-fermented green coffee
beans (G-FPk); S. cerevisiae-fermented green coffee beans (G-FSc). Alpha-KT, alpha-ketoglutaric acid. Concentrations are the mean values of triplicate fer-
mentations (n = 3), with error bars representing the standard deviation of the mean. Columns with different letters (a–d) indicate statistical differences among
different samples (P < 0.05).

coffee beans (G-Blank) were mostly higher than those in the control both yeast fermentations relative to G-Control (Fig. 2(b)). Alpha-ke-
green coffee beans (G-Control) which had been subjected to sterilisation toglutaric acid (alpha-KT) was only generated after sterilisation and
through submerged heat-treatment. Fermentation of the sterilised green further increased by fermentation. The chlorogenic acid was the pre-
coffee bean-water media with S. cerevisiae and P. kluyveri decreased the dominant phenolic compound in green coffee beans and existed in two
contents of soluble carbohydrates and amino acids, while increasing orders of magnitude higher than the free phenolic acids, including
organic acids and free phenolic acids, in both green coffee beans (Fig. 2) caffeic, ferulic, and coumaric acids (Fig. 2(c)). The chlorogenic acid was
and fermentation liquid (Fig. S1). The following data presentation and unaffected by fermentation (P < 0.05). Compared to G-Control, caffeic
discussion will focus on the changes in green coffee bean constituents, and ferulic acids were elevated by 14–42% in G-FSc and G-FPk, while
which were translated into modifications of final coffee flavour upon larger increases occurred after S. cerevisiae fermentation (P < 0.05).
roasting. The ethanol content in the liquid phase of the FSc fermentation
Soluble carbohydrates detected in green coffee beans included su- media was higher than that of the control media by 1.76% v/v, while
crose, glucose and fructose (Fig. 2(a)). Sucrose was predominant, ac- increases in the ethanol content were not observed in the FPk media
counting for about 95% of the total soluble carbohydrate content in all (P < 0.05) (Table S1).
green coffee bean samples. S. cerevisiae fermentation reduced the su-
crose content from 51.3 g/kg dry mass (G-Control) to 12.4 g/kg dry
3.3. Volatile compounds derived from yeast fermentations of green coffee
mass (G-FSc), while P. kluyveri fermentation did not exert any sig-
beans
nificant impact on the disaccharide (P < 0.05). As minor soluble
carbohydrates, glucose and fructose levels in G-Blank and G-Control
Volatile compounds that originated from yeast coculture fermenta-
were lower than 2 g/kg dry mass. The contents of the two hexoses in G-
tion of green coffee beans included 1 ketone, 3 esters, 2 phenolic
FSc and G-FPk were lower than those in G-Control by 30–41%, with
compounds, and 3 acids (Table 1). For the 5 fermentation-associated
higher residual amounts of hexoses observed in G-FSc (P < 0.05).
volatile compounds detected in G-Blank, 3-methylbutanoic acid, 2-
Similar to soluble carbohydrates, amino acids in green coffee beans
phenylethyl acetate, and 2-phenylethanol existed in lower contents in
were reduced to a larger extent by S. cerevisiae compared to P. kluyveri
G-Control, and the contents of acetic acid and 4-vinylguaiacol in G-
(P < 0.05) (Fig. 2(d)). All the amino acids in G-FSc were lower than
Control were not significantly different from those in G-Blank
those in G-Control by 46% (His) – 89% (Ser), and Met was not detected
(P < 0.05). 2,3-Butanedione (i.e. diacetyl), which was absent in G-
in G-FSc. In contrast, eight amino acids (Glu, Asp, Pro, Ser, Tyr, Gly,
Blank and G-FPk, was detected in G-Control and further elevated by
Thr, and Cys) were unaffected by P. kluyveri fermentation (P < 0.05).
28% (0.11 mg/kg) in G-FSc. Isoamyl acetate and ethyl octanoate were
Largest amino acid reduction by P. kluyveri fermentation occurred in
only detected in G-FPk and G-FSc, respectively. Contents of the other
Met, the content of which in G-FPk was lower than that in G-Control by
ester, 2-phenylethyl acetate, in G-FSc and G-FPk were 1.2- and 2.4-folds
23%.
of that in G-Blank, respectively. 4-Vinylguaiacol concentrations in G-
Organic acids including citric, malic, quinic, succinic, lactic and
FPk and G-FSc were around 1.7-folds of those in G-Control and G-Blank.
formic acids in green coffee bean were elevated by 7–20% in one or
The highest level of acetic acid occurred in G-FSc, followed by G-Blank

4
 C. Wang, et al.

Table 1
Volatile compounds derived from S. cerevisiae and P. kluyveri fermentations for 6 days in green and roasted coffee beans.
Volatile compounds LRI* Concentration (mg/kg)** Identification‡
(FFAP) Green coffee beans Roasted coffee beans

G-Blank† G-Control G-FPk G-FSc R-Blank R-Control R-FPk R-FSc

Ketones
a,c-e,g,h
2,3-Butanedione 979 ND 0.39 ± 0.03bB ND 0.50 ± 0.00aB 1.55 ± 0.02b 1.96 ± 0.15aA 1.96 ± 0.15a 1.73 ± 0.14aA MS, LRI
Esters
Isoamyl acetate 1113 ND ND 1.69 ± 0.00 ND ND ND ND ND MS, LRIk,l
Ethyl octanoate 1423 ND ND ND 0.51 ± 0.00B ND ND ND 0.63 ± 0.05aA MS, LRIl
dB
2-Phenylethyl acetate 1811 0.15 ± 0.00cA 0.10 ± 0.00dA 0.36 ± 0.00aA 0.18 ± 0.00bB 0.11 ± 0.01 0.23 ± 0.01cA 0.66 ± 0.00aA 0.30 ± 0.02bB MS, LRIk
Acids
Acetic acid 1434 2.29 ± 0.02bB 2.30 ± 0.03bB 1.44 ± 0.11cB 3.31 ± 0.07aB 14.86 ± 1.35aA 11.03 ± 0.30cA 11.98 ± 0.32bA 7.45 ± 0.72dA MS, LRI a-d,f,g,h
3-Methylbutanoic acid 1654 9.26 ± 0.09aA 6.55 ± 0.53cA 6.67 ± 0.01cA 8.48 ± 0.44bA 1.57 ± 0.07aB 1.17 ± 0.04bB 1.00 ± 0.04 dB 1.11 ± 0.01cB MS, LRI b,e,f,g,i,j

5
Octanoic acid 2040 ND ND ND 0.44 ± 0.00A ND ND ND 0.19 ± 0.05B MS, LRIk,l
Phenolics
2-Phenylethanol 1910 0.73 ± 0.04aA 0.21 ± 0.01cB 0.53 ± 0.04bA 0.17 ± 0.00dA 0.08 ± 0.00bB 0.27 ± 0.00aA 0.07 ± 0.00bB 0.08 ± 0.01bB MS, LRI a,d,g-j
4-Vinylguaiacol 2196 0.19 ± 0.00bB 0.19 ± 0.00bB 0.29 ± 0.01aB 0.32 ± 0.03aB 1.31 ± 0.02bA 1.02 ± 0.09cA 0.94 ± 0.09cA 1.95 ± 0.11aA MS, LRIb-d,g-j

ND, not detected.

* Linear retention indices calculated based on a C8–C20 n-alkane mixture on a FFAP column.
** Values of concentrations are standard deviation ± standard deviation of triplicate fermentations (n = 3); values in the same row with different letters (a-d) indicate statistical differences within the four types of
green or roasted coffee beans (P < 0.05); values in the same row with different letters (A-B) indicate statistical differences between the green or roasted coffee beans from the same treatment (P < 0.05).
†
G, green coffee beans; R, roasted coffee beans; Blank: untreated; Control, sterilised but unfermented; FPk, P. kluyveri-fermented; FSc, S. cerevisiae-fermented.
‡
Identification: MS = mass spectra, LRI = linear retention indices from literature values; LRIa refer to values obtained from a reference (Nebesny, Budryn, Kula, & Majda, 2007); LRIb refer to values obtained from a
reference (Moon & Shibamoto, 2009); LRIc refer to values obtained from a reference (Gonzalez-Rios et al., 2007); LRId refer to values obtained from a reference (Ochiai, Tsunokawa, Sasamoto, & Hoffmann, 2014); LRIe
refer to values obtained from a reference (Steen, Waehrens, Petersen, Münchow, & Bredie, 2017); LRIf refer to values obtained from a reference (Lee et al., 2017); LRIg refer to values obtained from a reference (Bressanello
et al., 2018); LRIh refer to values obtained from a reference (Lee et al., 2016); LRIi refer to values obtained from a reference (Scheidig et al., 2007); LRIj refer to values obtained from a reference (Cantergiani et al., 2001);
LRIk refer to values obtained from a reference (Chua et al., 2018); LRIl refer to values obtained from a reference (Chua et al., 2017).
Food Research International 136 (2020) 109588
C. Wang, et al. Food Research International 136 (2020) 109588

Fig. 3. Effects of S. cerevisiae and P. kluyveri fermentations for 6 days on sugars (a), organic acids (b) and phenolic compounds (c) in the roasted coffees beans.
roasted but untreated coffee beans (R-Blank); roasted sterilised but unfermented control coffee beans (R-Control); roasted P. kluyveri-fermented coffee
beans (R-FPk); roasted S. cerevisiae-fermented coffee beans (R-FSc). Alpha-KT, alpha-ketoglutaric acid. Concentrations are the mean values of triplicate fer-
mentations (n = 3), with error bars representing the standard deviation of the mean. Columns with different letters (a–d) indicate statistical differences among
different samples (P < 0.05).

and G-Control and G-FPk (P < 0.05). Octanoic acid was only detected S. cerevisiae and P. kluyveri fermentation-associated volatile com-
in G-FSc. pounds in green coffee beans were also detected in some roasted coffee
beans (Table 1). The contents of 2,3-butanedione in the roasted coffees
were higher than those in their respective green coffee beans
3.4. Non-volatile profiles and colours in roasted coffees
(P < 0.05). For the three esters detected in the green coffee beans:
isoamyl acetate was not detected after roasting; ethyl octanoate was
Soluble carbohydrates in the green coffee beans were reduced after
only detected in R-FSc, and its content in R-FSc was 20% higher than
roasting. The highest soluble carbohydrates content was obtained in R-
that in G-FPk; 2-phenylethyl acetate was detected in both roasted cof-
Blank, followed by R-FPk, R-FSc, and R-Control (P < 0.05) (Fig. 3(a)).
fees, and its contents in R-FPk and R-FSc were increased by 80% and
The roasting process generated succinic and formic acids, but reduced
60% by roasting, respectively. Roasting led to increase in acetic acid
citric, quinic, malic, and chlorogenic acids (Fig. 3(b)). Alpha-ketoglu-
and 4-vinylguaiacol, together with the decrease in 3-methylbutanoic
taric, free phenolic, and amino acids were not detected in the roasted
acid and octanoic acid (P < 0.05). The content of 2-phenylethanol was
coffees (Fig. 3(c)).
reduced in most of the coffees but increased in R-FSc (P < 0.05).
Among the four samples of roasted coffee under the same roasting
conditions, R-Blank and R-FSc exhibited the darkest and lightest colors,
respectively (Table S2). 3.6. Flavour profiles of the roasted coffees

Different treatments of the green coffee beans resulted in roasted

3.5. Volatile profiles of roasted coffees
coffees with distinctive flavour profiles (Fig. 6). Without treatments, the
aroma of R-Blank was characterized by moderate caramel and smoky
Volatile profiles of four roasted coffees including R-Blank, R-
attributes, weaker nutty and roasted attributes, and a particularly mild
Control, R-FPk, and R-FSc are presented in Figs. 4 and 5. Volatile
fruity attribute. The taste attributes, including sweetness and acidity, of
analysis of roasted coffees identified 2 esters, 4 acids, 8 ketones, 2 al-
R-Blank surpassed that of the other roasted coffees. Compared to R-
dehydes, 8 furfurals, 14 furans and furanones, 6 pyrroles, 20 pyrazines,
Blank, the flavour of R-Control featured stronger nutty and roasted
2 pyridines, 5 phenolic compounds, 3 sulfur-containing compounds,
attributes but weaker acidic and sweet tastes. R-FSc was distinguished
and 3 miscellaneous volatile compounds (Fig. 4). Most of the detected
by strongest nutty and roasted aroma attributes. The aroma profiles of
volatile compounds are O- and N-heterocycles.
R-FPk featured moderately elevated nutty and roasted, but the incre-
Furfural and its derivatives were the major O-heterocyclic volatile
ments were close to those of R-Control. Compared to R-Blank and R-
compounds in the roasted coffees. Dominated by furfural, 5-methyl-
Control, a noticeable elevation in fruitiness was noted in the fermented
furfural, and furfural alcohol, total furfurals in the four roasted coffees
roasted coffees, especially in R-FSc.
showed no significant variation (P < 0.05) (Figs. 4 and 5). Total
content of furans and furanones in R-Blank were higher than those in
other roasted coffees by about 60% (Fig. 5), with most of the individual 4. Discussion
furans and furanones also being at higher levels in R-Blank (P < 0.05)
(Table S3). This study investigated the submerged fermentation of sterilised
N-heterocyclic volatile compounds in the roasted coffees were green coffee beans with two commercial wine yeasts, S. cerevisiae and P.
dominated by pyrazines both in terms of content and number of species kluyveri with the aim of modulating coffee flavour. The sterilisation of
(Figs. 4 and 5). The pyrazines detected were mainly methyl- and ethyl- green coffee beans led to some losses of non-volatile compounds, in-
substituted, and the predominant species was 2-methylpyrazine cluding soluble carbohydrates, amino acids, organic acids, and phenolic
(Fig. 4). Unlike O-heterocyclic volatile compounds, the highest and compounds, into liquid phase of the fermentation matrix Figs. 2 and
lowest contents of N-heterocyclic volatile compounds were found in R- S1). Constituents from sterilised green coffee beans were able to sup-
FSc and R-Blank, respectively (P < 0.05). The total content of pyr- port the growth and metabolism of monocultures of S. cerevisiae and P.
azines in R-FSc was 1.8-folds of those in R-Blank and 1.2-folds of those kluyveri. Metabolic activities of the two yeasts on non-volatile and vo-
in R-FPk, while the contents of pyrazines in R-FSc and R-Control had no latile compounds enabled the biomodification of flavour profiles of
significant difference (P < 0.05) (Fig. 5). The total content of pyrroles green coffee beans even after roasting. The invertase activity in S.
in R-FSc was 1.5- to 1.8- folds of those in the other roasted coffees, with cerevisiae enables the Saccharomyces yeast to hydrolyze sucrose into
a similar trend observed for the total content of pyridines (Fig. 5). glucose and fructose (Lu, Chan, Li, & Liu, 2018). The three detected

6
C. Wang, et al. Food Research International 136 (2020) 109588

cerevisiae and P. kluyveri) and from sucrose hydrolysis (S. cerevisiae) to

pyruvate through glycolysis, and the resulted pyruvate was used for
ethanolic fermentation or converted into acetyl-CoA and channelled
into tricarboxylic acid (TCA) cycle for aerobic respiration (Stewart,
2017a). Pyruvate and acetyl-CoA can also be supplied by catabolic
other nutrients such as amino acids (Stewart, 2017a). Although S. cer-
evisiae strain has outstanding capacity to convert sugars into ethanol,
ethanol generation in the green coffee bean fermentation was rather
limited (< 2 v/v %) by the low sugar content. With a low capacity for
ethanolic fermentation, P. kluyveri did not contribute to ethanol pro-
duction in the media for FPk (Table S1).
The production of acids by S. cerevisiae and P. kluyveri during the
fermentations of sterilised green coffee beans was in accordance with
previous studies on the fermentations of alcoholic beverages by the two
yeasts (Table 1 and Fig. 2) (Chen & Liu, 2014; Chua et al., 2018; Chua
et al., 2017; Lu et al., 2016, 2018). Acetic acid is formed as pyruvate
was converted into acetyl-CoA (Stewart, 2017a). Citric, alpha-ketoglu-
taric, malic, and succinic acids are generated in TCA cycle and partially
excreted from the cells of the yeasts. Alpha-ketoglutaric acid plays an
important role in volatile production by converting branched and aro-
matic amino acids into corresponding fusel aldehydes, higher alcohols,
and fusel acids (Bell & Henschke, 2005). Such microbial activities were
reflected by the elevation of organic acids in yeast-fermented green
coffee beans, as compared to the sterilised but unfermented control
beans (Fig. 2). Nonetheless, the contents of citric, malic, quinic, and
succinic acids in the fermented green coffee beans were still lower than
those in the blank sample due to the loss of these organic acids into the
liquid phase of fermentation media during sterilisation (Figs. 2 and S1).
The organic acid production contributed to the acidification of the
fermentation media. Higher organic acid production in FSc corre-
sponded to its larger pH reduction, as compared to FPk.
Metabolic activities of S. cerevisiae and P. kluyveri led to production
of fruity esters. Yeast transamination of leucine (Leu), isoleucine (ILe)
and phenylalanine (Phe) led to the production of isoamyl alcohol (3-
methylbutanol) and 2-phenylethanol, respectively (Table 1). The two
higher alcohols enzymatically esterified with acetyl-CoA to form iso-
amyl acetate with remarkable fruity aroma and 2-phenylethyl acetate
with fruity and floral attributes, which were subsequently excreted
from the yeast cells (Stewart, 2017b). The outstanding abilities of P.
kluyveri to generate fruity-smelling acetate esters, particularly isoamyl
acetate, has been well documented (Benito et al., 2019). The formation
of 2,3-butanedione (diacetyl) was promoted by the higher utilisation of
valine and other amino acids by S. cerevisiae (Stewart, 2017b). The
detection of 2,3-butanedione in the control green coffee beans was in
accordance with the finding in a previous study (Wang et al., 2020).
However, 2,3-butanedione in the sterilised green coffee beans might
have been reduced to acetoin and possibly, 2,3-butanediol by P. kluy-
veri. The smoky-smelling 4-vinylguaiacol was the product from cou-
maric acid decarboxylation by the yeasts in green coffee beans
(Table 1).
S. cerevisiae and P. kluyveri significantly modified flavour-related
constituents in green coffee beans, which were translated into direct
and indirect modifications on the final coffee flavour upon roasting
(Table 1, Figs. 3, 6, and 7). Controlled green coffee bean fermentation
Fig. 4. Heat map of volatile profiles of roasted coffees. R-Blank, roasted but
untreated coffee beans; R-Control, roasted sterilised but unfermented control by S. cerevisiae and P. kluyveri yielded coffees with fruity esters that
coffee beans; R-FPk, roasted P. kluyveri-fermented coffee beans; R-FSc, roasted were uncommon in the volatile profiles of coffee from traditional on-
S. cerevisiae-fermented coffee beans. farm processing (Cantergiani et al., 2001; Lee, Kim, & Lee, 2017;
Scheidig, Czerny, & Schieberle, 2007). Metabolic activities of S. cere-
visiae and P. kluyveri during fermentation of green coffee beans not only
soluble carbohydrates of coffee, glucose, fructose and sucrose, were all
generated ethyl octanoate and 2-phenylethyl acetate, but also the pre-
carbon sources for S. cerevisiae (Fig. 2). However, P. kluyveri was unable
cursors of the two esters. Further esterification of ethanol and octanoic
to utilize sucrose due to a lack of invertase (Lu et al., 2016). The
acid (or its Co-A form) from metabolism by S. cerevisiae may explain the
available carbon sources in green coffee beans were therefore much
increased ethyl octanoate in R-FSc after roasting. Similarly, the pro-
more limited for the non-Saccharomyces yeast, restricting its growth and
duction of 2-phenylethyl acetate in both fermented coffee samples after
metabolism to a certain extent.
roasting could be attributed to the condensation of 2-phenylethanol and
The yeasts metabolized the monosaccharides in free form (S.
acetyl Co-A. Nonetheless, isoamyl acetate from P. kluyveri fermentation

7
C. Wang, et al. Food Research International 136 (2020) 109588

Fig. 5. Effects of S. cerevisiae and P. kluyveri

fermentations for 6 days on contents of volatile
categories in the roasted coffees beans.
roasted but untreated coffee beans (R-Blank);
roasted sterilised but unfermented control
coffee beans (R-Control); roasted P. kluy-
veri-fermented coffee beans (R-FPk);
roasted S. cerevisiae-fermented coffee beans (R-
FSc). Concentrations are the mean values of tri-
plicate fermentations (n = 3), with error bars
representing the standard deviation of the mean.
Columns with different letters (a–d) indicate
statistical differences among different samples
(P < 0.05).

was not conserved during roasting.

Besides the esters in yeast-fermented green coffee beans that re-
mained after the roasting process and contribute to the elevated fruity
aroma in the roasted coffees (Fig. 6), the majority of volatile com-
pounds in roasted coffees were formed during roasting through the
pyrolysis of non-volatile green bean constituents (Shibamoto, 2015). As
a principal pyrolytic reaction during coffee roasting, Maillard reaction
between reducing sugars and amino acids leads to the production of the
brownish pigments and various volatile compounds which cause the
characteristic aroma of roasted coffees (Mottram, 2007). Apart from the
free glucose and fructose in green coffee beans, the two hexoses from
sucrose degradation under high temperatures are also sources of re-
ducing sugars during coffee roasting (Neves, Pool, Kok, Kuipers, &
Santos, 2005). The released glucose and fructose from sucrose hydro-
lysis were not completely degraded and partially contributed to the
residual sugars in the roasted coffees (Wang et al., 2020). The higher
content of soluble carbohydrates in blank green coffee beans contribute
to the higher residual sugar level and sweetness in its roasted coffee
Figs. 3 and 6). Sugar pyrolysis produces acetic and formic acids
Fig. 6. Effects of S. cerevisiae and P. kluyveri fermentations on flavor profiles of
coffee brews prepared from roasted coffees. roasted but untreated coffee
(Majcher, 2011; Parker, 2015). Nevertheless, the overall reduction in
beans (R-Blank); roasted sterilised but unfermented control coffee beans total soluble carbohydrates and organic acids from the sterilisation and
(R-Control); roasted P. kluyveri-fermented coffee beans (R-FPk); subsequent fermentation of green coffee beans was reflected in the
roasted S. cerevisiae-fermented coffee beans (R-FSc). milder acidity and sweetness of the resulted roasted coffees (Fig. 6).

Fig. 7. Proposed mechanisms of coffee flavour modification by S. cerevisiae and P. kluyveri fermentations of green coffee beans (based on this study and Parker, 2015,
Cerny, 2010, Majcher, 2011).

8
C. Wang, et al. Food Research International 136 (2020) 109588

The two major volatile production pathways of Maillard reaction CRediT authorship contribution statement
include 1- and 3-deoxyosone routes. The 1-deoxyosone route results in
the production of furanones, maltol, organic acids, including acetic and Chenhui Wang: Conceptualization, Methodology, Investigation,
formic acids, and dicarbonyls, such as 2,3-butanedione (Parker, 2015). Formal analysis, Writing - original draft, Visualization. Jingcan Sun:
The subsequent Strecker degradation involving dicarbonyls and amino Methodology, Investigation, Writing - review & editing. Benjamin
acids leads to the production of N-heterocyclic volatile compounds, Lassabliere: Methodology. Bin Yu: Methodology, Writing - review &
including pyrazines, pyrroles, and pyridines, with roasted and nutty editing. Shao Quan Liu: Conceptualization, Writing - review & editing,
attributes (Parker, 2015). The heat-treatment for sterilisation and me- Supervision.
tabolism by S. cerevisiae raised the 2,3-butanedione content in S. cere-
visiae-fermented green coffee beans, which could heighten the pro- Declaration of Competing Interest
duction of the N-heterocyclic volatile compounds during roasting and
enhance the roasted and nutty attributes in the resulted roasted coffee The authors declare that they have no known competing financial
(Low, Parker, & Mottram, 2007) (Fig. 6). The effect seemed to over- interests or personal relationships that could have appeared to influ-
weigh the loss of amino acids in S. cerevisiae-fermented green coffee ence the work reported in this paper.
beans (Fig. 2(d)). The 2,3-butanedione produced during the pre-treat-
ment can also be associated with the moderate elevation in nutty- Acknowledgments
smelling pyrazines in the roasted control coffee (Figs. 5 and 6). The
enhancement in the production of N-heterocycles from the elevated The authors would like to thank the Q graders, Mr Shaun Ong and
content of 2,3-butanedione was also reported in a previous study in- Ms Gloria Soh from Bettr Barista PTE LTD (Singapore) for their con-
volved non-supplemented green coffee bean fermentation with Lc. lactis tributions to the sensory evaluation.
subsp. cremoris (Wang et al., 2020).
The 3-deoxyosone route of Maillard reaction leads to the production Appendix A. Supplementary material
of furfural, which is then converted into other furan derivatives (Fig. 4)
(Parker, 2015). Besides Maillard reaction, caramelization of sugars also Supplementary data to this article can be found online at https://
generates various O-heterocyclic volatile compounds, as well as ke- doi.org/10.1016/j.foodres.2020.109588.
tones, aldehydes, organic acids and brown pigments during coffee
roasting (Majcher, 2011). With the higher levels of organic acids and References
higher sugar content (Table 1 and Fig. 4), the blank green coffee beans
had a higher acidity that promoted both caramelization and Maillard Bell, S. J., & Henschke, P. A. (2005). Implications of nitrogen nutrition for grapes, fer-
reaction through the 3-deoxyosone route (Fig. 3) (Cerny, 2010; mentation and wine. Australian Journal of Grape and Wine Research, 11(3), 242–295.
Majcher, 2011). As a result, the blank roasted coffee was distinguished https://doi.org/10.1111/j.1755-0238.2005.tb00028.x.
Benito, S., Ruiz, J., Belda, I., Kiene, F., Beisert, B., Navascués, E., ... Rauhut, D. (2019).
by elevated contents of furans and furanones (Fig. 4) (Flament, 2002). Application of non-Saccharomyces yeasts in wine production. In A. Sibirny (Ed.).
Such enhancement in O-heterocycles production in green coffee beans Non-conventional yeasts: From basic research to application (pp. 75–89). Cham:
with higher acidity was also observed in a previous study on glucose- Springer.
Bressanello, D., Liberto, E., Cordero, C., Sgorbini, B., Rubiolo, P., Pellegrino, G., ... Bicchi,
supplemented lactococcus fermentation (Wang et al., 2020). C. (2018). Chemometric modeling of coffee sensory notes through their chemical
Maillard reaction and caramelization also leads to the production of signatures: Potential and limits in defining an analytical tool for quality control.
the brownish pigments which are responsible for the characteristic Journal of Agricultural and Food Chemistry, 66(27), 7096–7109. https://doi.org/10.
1021/acs.jafc.8b01340.
colour of roasted coffees (Majcher, 2011; Parker, 2015). Such browning Bressani, A. P. P., Martinez, S. J., Evangelista, S. R., Dias, D. R., & Schwan, R. F. (2018).
process from sugar degradation was restricted in the S. cerevisiae-fer- Characteristics of fermented coffee inoculated with yeast starter cultures using dif-
mented coffee due to the considerable consumption of soluble carbo- ferent inoculation methods. LWT, 92, 212–219. https://doi.org/10.1016/j.lwt.2018.
02.029.
hydrates by the yeast during the green coffee bean fermentation.
Cantergiani, E., Brevard, H., Krebs, Y., Feria-Morales, A., Amado, R., & Yeretzian, C.
To summarise, the modulation of flavour-related constituents in (2001). Characterisation of the aroma of green Mexican coffee and identification of
green coffee beans from fermentation with S. cerevisiae and P. kluyveri mouldy/earthy defect. European Food Research and Technology, 212(6), 648–657.
led to considerable modifications of coffee flavour upon roasting. The Cerny, C. (2010). Thermal generation of aroma-active volatiles in food. In A. Herrmann
(Ed.). The chemistry and biology of volatiles (pp. 231–252). Chichester: John Wiley &
aromatic esters produced by both yeasts partially survived the roasting Sons.
process and directly enhanced the fruitiness of the roasted coffees. The Chen, D., & Liu, S. Q. (2014). Chemical and volatile composition of lychee wines fer-
S. cerevisiae fermentation promoted the pyrolytic formation of N-het- mented with four commercial S accharomyces cerevisiae yeast strains. International
Journal of Food Science & Technology, 49(2), 521–530. https://doi.org/10.1111/ijfs.
erocyclic volatile compounds through production of dicarbonyl com- 12332.
pound (2,3-butanedione), which indirectly enhanced the roasted and Chua, J. Y., Lu, Y., & Liu, S. Q. (2018). Evaluation of five commercial non-Saccharomyces
nutty attributes in the resulted roasted coffee (Fig. 7). yeasts in fermentation of soy (tofu) whey into an alcoholic beverage. Food
Microbiology, 76, 533–542. https://doi.org/10.1016/j.fm.2018.07.016.
Chua, J. Y., Lua, Y., & Liu, S. Q. (2017). Biotransformation of soy whey into soy alcoholic
beverage by four commercial strains of Saccharomyces cerevisiae. International Journal
5. Conclusions of Food Microbiology, 262, 14–22. https://doi.org/10.1016/j.ijfoodmicro.2017.09.
007.
Crafack, M., Keul, H., Eskildsen, C. E., Petersen, M. A., Saerens, S., Blennow, A., ...
The current study investigated the coffee flavour modulation Nielsen, D. S. (2014). Impact of starter cultures and fermentation techniques on the
through fermentation of green coffee beans with monocultures of S. volatile aroma and sensory profile of chocolate. Food Research International, 63,
306–316. https://doi.org/10.1016/j.foodres.2014.04.032.
cerevisiae and P. kluyveri. Both yeasts directly enriched the flavour
de Bruyn, F., Zhang, S. J., Pothakos, V., Torres, J., Lambot, C., Moroni, A. V., ... de Vuyst,
profile of the roasted coffees by imparting favorable aroma attributes L. (2017). Exploring the impacts of postharvest processing on the microbiota and
which were absent without the biotransformation. The metabolites metabolite profiles during green coffee bean production. Applied and Environmental
from S. cerevisiae also indirectly enhanced the volatile production Microbiology, 83(1), e02398–02316. https://doi.org/10.1128/AEM.02398-16.
de Carvalho Neto, D. P., de Melo Pereira, G., Tanobe, V., Thomaz Soccol, V., G da Silva,
through modulating the pathway of pyrolytic reactions during roasting. B., Rodrigues, C., & Soccol, C. (2017). Yeast diversity and physicochemical char-
S. cerevisiae and P. kluyveri are therefore considered suitable starter acteristics associated with coffee bean fermentation from the Brazilian Cerrado
cultures for controlled coffee biotransformation for flavour improve- Mineiro region. Fermentation, 3(1), 11. https://doi.org/10.3390/
fermentation3010011.
ment. Evangelista, S. R., Miguel, M. G. C. P., Silva, C. F., Pinheiro, A. C. M., & Schwan, R. F.
(2015). Microbiological diversity associated with the spontaneous wet method of
coffee fermentation. International Journal of Food Microbiology, 210, 102–112. https://

9
C. Wang, et al. Food Research International 136 (2020) 109588

doi.org/10.1016/j.ijfoodmicro.2015.06.008. headspace composition of Robusta coffee bean aroma. European Food Research and
Evangelista, S. R., Silva, C. F., Miguel, M. G. P.d. C., Cordeiro, C.d. S., Pinheiro, A. C. M., Technology, 225(1), 9–19. https://doi.org/10.1007/s00217-006-0375-0.
Duarte, W. F., & Schwan, R. F. (2014). Improvement of coffee beverage quality by Neves, A. R., Pool, W. A., Kok, J., Kuipers, O. P., & Santos, H. (2005). Overview on sugar
using selected yeasts strains during the fermentation in dry process. Food Research metabolism and its control in Lactococcus lactis—the input from in vivo NMR. FEMS
International, 61, 183–195. https://doi.org/10.1016/j.foodres.2013.11.033. Microbiology Reviews, 29(3), 531–554. https://doi.org/10.1016/j.fmrre.2005.04.005.
Flament, I. (2002). Coffee flavor chemistry. England: Jonhy Wiley & Sons. Ochiai, N., Tsunokawa, J., Sasamoto, K., & Hoffmann, A. (2014). Multi-volatile method
Gonzalez-Rios, O., Suarez-Quiroz, M. L., Boulanger, R., Barel, M., Guyot, B., Guiraud, J.- for aroma analysis using sequential dynamic headspace sampling with an application
P., & Schorr-Galindo, S. (2007a). Impact of “ecological” post-harvest processing on to brewed coffee. Journal of Chromatography A, 1371, 65–73. https://doi.org/10.
coffee aroma: II. Roasted coffee. Journal of Food Composition and Analysis, 20(3–4), 1016/j.chroma.2014.10.074.
297–307. https://doi.org/10.1016/j.jfca.2006.12.004. Parker, J. K. (2015). Thermal generation or aroma. In J. K. Parker, J. S. Elmore, & L.
Gonzalez-Rios, O., Suarez-Quiroz, M. L., Boulanger, R., Barel, M., Guyot, B., Guiraud, J.- Methven (Eds.). Flavour development, analysis and perception in food and beverages (pp.
P., & Schorr-Galindo, S. (2007b). Impact of “ecological” post-harvest processing on 151–185). Oxford: Elsevier.
the volatile fraction of coffee beans: I. Green coffee. Journal of Food Composition and Pereira, G. V. M., Neto, E., Soccol, V. T., Medeiros, A. B. P., Woiciechowski, A. L., &
Analysis, 20(3–4), 289–296. https://doi.org/10.1016/j.jfca.2006.07.009. Soccol, C. R. (2015). Conducting starter culture-controlled fermentations of coffee
Haile, M., & Kang, W. H. (2019). The role of microbes in coffee fermentation and their beans during on-farm wet processing: Growth, metabolic analyses and sensorial ef-
impact on coffee quality. Journal of Food Quality, 2019. https://doi.org/10.1155/ fects. Food Research International, 75, 348–356. https://doi.org/10.1016/j.foodres.
2019/4836709. 2015.06.027.
Lee, L. W., Cheong, M. W., Curran, P., Yu, B., & Liu, S. Q. (2015). Coffee fermentation and Ribeiro, L. S., Ribeiro, D. E., Evangelista, S. R., Miguel, M. G.d. C. P., Pinheiro, A. C. M.,
flavour – An intricate and delicate relationship. Food Chemistry, 185, 182–191. Borém, F. M., & Schwan, R. F. (2017). Controlled fermentation of semi-dry coffee
https://doi.org/10.1016/j.foodchem.2015.03.124. (Coffea arabica) using starter cultures: A sensory perspective. LWT, 82, 32–38.
Lee, L. W., Cheong, M. W., Curran, P., Yu, B., & Liu, S. Q. (2016). Modulation of coffee https://doi.org/10.1016/j.lwt.2017.04.008.
aroma via the fermentation of green coffee beans with Rhizopus oligosporus: II. Effects Saerens, S., & Swiegers, J. H. (2016). Enhancement of coffee quality and flavor by using
of different roast levels. Food Chemistry, 211, 925–936. https://doi.org/10.1016/j. pichia kluyveri yeast starter culture for coffee fermentation. US Patent:
foodchem.2016.05.073. US20160058028A1.
Lee, L. W., Tay, G. Y., Cheong, M. W., Curran, P., Yu, B., & Liu, S. Q. (2017). Modulation SCAA (Specialty Coffee Association of America) (2015). SCAA protocols: Cupping spe-
of the volatile and non-volatile profiles of coffee fermented with Yarrowia lipolytica: cialty coffee.
II. Roasted coffee. LWT, 80, 32–42. https://doi.org/10.1016/j.lwt.2017.01.070. Scheidig, C., Czerny, M., & Schieberle, P. (2007). Changes in key odorants of raw coffee
Lee, S. J., Kim, M. K., & Lee, K. G. (2017). Effect of reversed coffee grinding and roasting beans during storage under defined conditions. Journal of Agricultural and Food
process on physicochemical properties including volatile compound profiles. Chemistry, 55(14), 5768–5775. https://doi.org/10.1021/jf070488o.
Innovative Food Science & Emerging Technologies, 44, 97–102. https://doi.org/10. Shibamoto, T. (2015). Volatile chemicals from thermal degradation of less volatile coffee
1016/j.ifset.2017.07.009. components. In V. R. Preedy (Ed.). Coffee in health and disease prevention (pp. 129–
Low, M. Y., Parker, J. K., & Mottram, D. S. (2007). Mechanisms of alkylpyrazine for- 135). Oxford: Elsevier.
mation in a potato model system containing added glycine. Journal of Agricultural and Steen, I., Waehrens, S. S., Petersen, M. A., Münchow, M., & Bredie, W. L. (2017). Influence
Food Chemistry, 55(10), 4087–4094. https://doi.org/10.1021/jf070044s. of serving temperature on flavour perception and release of Bourbon Caturra coffee.
Lu, Y., Chan, L. J., Li, X., & Liu, S. Q. (2018). Effects of sugar concentration on mango Food Chemistry, 219, 61–68. https://doi.org/10.1016/j.foodchem.2016.09.113.
wine composition fermented by Saccharomyces cerevisiae MERIT.ferm. International Stewart, G. G. (2017a). Energy metabolism by the yeast cell. Brewing and distilling yeasts
Journal of Food Science & Technology, 53(1), 199–208. https://doi.org/10.1111/ijfs. (pp. 77–107). Cham: Springer.
13574. Stewart, G. G. (2017b). Flavour production by yeast. Brewing and distilling yeasts (pp. 325–
Lu, Y., Huang, D., Lee, P. R., & Liu, S. Q. (2016). Assessment of volatile and non-volatile 355). Cham: Springer.
compounds in durian wines fermented with four commercial non-Saccharomyces Takahashi, K., Minami, Y., Kanabuchi, Y., Togami, K., & Mitsuhashi, M. (2007). US
yeasts. Journal of the Science of Food and Agriculture, 96(5), 1511–1521. https://doi. Patent: US20070190207A1.
org/10.1002/jsfa.7253. Vilela, D. M., Pereira, G. V.d. M., Silva, C. F., Batista, L. R., & Schwan, R. F. (2010).
Majcher, M. (2011). Saccharides-derived flavor compounds. In H. Jelen (Ed.). Food fla- Molecular ecology and polyphasic characterization of the microbiota associated with
vors: Chemical, sensory, and technological properties (pp. 95–120). Boca Raton: CRC semi-dry processed coffee (Coffea arabica L.). Food Microbiology, 27(8), 1128–1135.
Press. https://doi.org/10.1016/j.fm.2010.07.024.
Maness, N. (2010). Extraction and analysis of soluble carbohydrates. In R. Sunkar (Ed.). Wang, C., Sun, J., Lassabliere, B., Yu, B., & Liu, S. Q. (2020). Coffee flavour modification
Plant stress tolerance (pp. 341–370). Clifton: Humana Press. through controlled fermentation of green coffee beans by Lactococcus lactis subsp.
Martinez, S. J., Bressani, A. P. P., Miguel, M. G. C. P., Dias, D. R., & Schwan, R. F. (2017). cremoris. LWT, 120, 108930. https://doi.org/10.1016/j.lwt.2019.108930.
Different inoculation methods for semi-dry processed coffee using yeasts as starter Wang, C., Sun, J., Lassabliere, B., Yu, B., Zhao, F., Zhao, F., ... Liu, S. Q. (2019). Potential
cultures. Food Research International, 102, 333–340. https://doi.org/10.1016/j. of lactic acid bacteria to modulate coffee volatiles and effect of glucose supple-
foodres.2017.09.096. mentation: Fermentation of green coffee beans and impact of coffee roasting. Journal
Moon, J. K., & Shibamoto, T. (2009). Role of roasting conditions in the profile of volatile of the Science of Food and Agriculture, 99(1), 409–420. https://doi.org/10.1002/jsfa.
flavor chemicals formed from coffee beans. Journal of Agricultural and Food Chemistry, 9202.
57(13), 5823–5831. https://doi.org/10.1021/jf901136e. Zhang, S. J., de Bruyn, F., Pothakos, V., Torres, J., Falconi, C., Moccand, C., ... de Vuyst, L.
Mottram, D. S. (2007). The maillard reaction: Source of flavour in thermally processed (2019). Following coffee production from cherries to cup: Microbiological and me-
foods. In R. G. Berger (Ed.). Flavours and fragrances: Chemistry, bioprocessing and tabolomic analysis of wet processing of Coffea arabica. Applied and Environmental
sustainability (pp. 269–283). Berlin: Springer. Microbiology, 85(6), e02635–02618. https://doi.org/10.1128/aem.02635-18.
Nebesny, E., Budryn, G., Kula, J., & Majda, T. (2007). The effect of roasting method on

10