Food Microbiology 55 (2016) 32e46

Contents lists available at ScienceDirect

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

Exploring the phenotypic space of non-Saccharomyces wine yeast

biodiversity
Debra Rossouw, Florian F. Bauer*
Institute for Wine Biotechnology, University of Stellenbosch, Private Bag X1, Matieland, 7602, South Africa

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

Article history: Tremendous microbial diversity exists in vineyards, and the potential to harness this diversity for novel
Received 7 April 2015 mixed or pure starter cultures for wine fermentation has received significant attention in recent years.
Received in revised form However, most studies are limited to a small subset of strains and species. Here we present data from a
10 October 2015
systematic screen of 91 yeast isolates from South African grape must and vineyard samples for oeno-
Accepted 26 November 2015
logically relevant traits. One focus area was finding non-Saccharomyces isolates showing both reduced
Available online 30 November 2015
ethanol yields, as well as improved aromatic characteristics. Of the 91 isolates evaluated initially, 21
showed lower ethanol yields when compared to commercial wine yeast strain controls. Collectively, the
Keywords:
Non-Saccharomyces yeast
metabolic data (primary fermentation and secondary aroma compounds) highlight the enormity of the
Low ethanol ‘phenotypic space’ of yeast communities in South African vineyards. The data also emphasise intraspecies
Aroma variability, challenging our concept of species typicity. Of particular oenological interest was the ability of
Sensory analysis several isolates to produce high levels of terpenoid compounds. A few strains were ultimately found
which showed a substantial reduction (>1.5%) in the final ethanol content of sequential fermentations, as
well as unique aroma compound production profiles. Four of these strains were selected for compre-
hensive wine trials in both red and white grape musts, complete with microbial, chemical and sensory
analyses of the red wines. This presents, for the first time, a full bench-to-bottle characterisation of non-
Saccharomyces strains showing the most potential for commercial application.
The findings of this study enlarge the potential range of oenological applications for non-Saccharo-
myces yeast, while also suggesting the potential usefulness of several yeast species that have previously
not been considered for winemaking applications.
© 2015 Elsevier Ltd. All rights reserved.

1. Introduction Pichia, Torulaspora, Metschnikowia etc.) which colonise grape and

The microbiomes of terrestrial environments have only recently
become fully accessible due to the development of culture-
independent identification methods, and constitute vast, under-
explored and underexploited reserves of biodiversity. Of these
environments, the vineyard and wine microbiome have received
significant attention in the past few years due to many perceived
opportunities in terms of biotechnological applications in wine-
making processes. Any single vineyard is indeed populated by
hundreds of different species and strains of yeast and bacteria
(Setati et al., 2012; Bokulich et al., 2013). The scientific and indus-
trial potential of this vast reserve of uncharacterised species is
remarkable and largely untapped. The many indigenous yeast
species (such as those from the genera of Candida, Hanseniaspora,
* Corresponding author.
E-mail address: fb2@sun.ac.za (F.F. Bauer).
winery surfaces can contribute to the transformation of grape
metabolites to compounds with sensory or technological impacts
(for a review see Jolly et al. 2013). In inoculated fermentations the
involvement of non-Saccharomyces yeast strains has been limited
by the use of starter cultures of wine yeast strains (almost exclu-
sively of the Saccharomyces cerevisiae species) which rapidly
outcompete the indigenous yeast population numerically at the
high inoculum levels used commercially (Fleet et al. 1984; Henick-
Kling et al. 1998; Heard and Fleet, 1985).
Traditionally inoculation practices have largely sought to limit
the involvement of non-Saccharomyces yeasts during wine
fermentation as early research showed that certain species are
prone to high production levels of undesirable compounds such as
acetic acid, acetaldehyde, acetoin and ethyl acetate (Van Zyl et al.
1963; Comitini et al. 2011). Unfortunately this exclusion of non-
Saccharomyces yeasts from the fermentation process may result in a
loss of complexity and wines lacking distinctive characteristics

http://dx.doi.org/10.1016/j.fm.2015.11.017
0740-0020/© 2015 Elsevier Ltd. All rights reserved.
 D. Rossouw, F.F. Bauer / Food Microbiology 55 (2016) 32e46
(Pretorius, 2000; Bisson and Kunkee, 1991; Varela et al. 2009). Non-
Saccharomyces yeasts possess enzymatic activities which can ca-
talyse the release of volatile aroma compounds from non-volatile
bound precursors (Fern andez-Gonz ales et al. 2003; Hernandez-
Orte et al., 2008). These yeasts can also impact aroma production
directly by their own metabolic activity (production of alcohols and
esters) or by the release of extracellular enzymes which transform
S. cerevisiae ederived metabolites (Bisson and Kunkee, 1991;
Boulton et al. 1996; Clemente-Jimenez et al. 2004). Strains of
non-Saccharomyces yeasts have also shown potential for producing
aroma compounds not associated with fermentation by many
strains of Saccharomyces cereivisiae, such as various monoterpenes
and other terpenoid compounds (Garcia et al. 2002).
Interest has also been raised in the use of non-Saccharomyces
wine yeast strains to reduce the ethanol levels of wines. World-
wide, growing demands for lower alcohol (yet high quality) wines
are driven by consumer preference and legislation imposing
financial penalties on beverages with alcohol contents above set
thresholds (Howley and Young, 1992; Pickering, 2000; Heux et al.
2006). High alcohol levels may also negatively affect wine quality
due to the influence of ethanol on the volatility of other important
aroma compounds as well as the perception of ‘hotness’ on the
palate (Guth and Sies, 2002). While post fermentation processes do
exist for ethanol removal by physical methods (such as spinning
cone columns or reverse osmosis) these methods are expensive and
generally reduce the quality of the treated wine (Pickering, 2000). A
large research focus over the past two decades has been the
development of yeast strains for reduced ethanol fermentations by
genetic engineering (for a review see Varela et al. 2012). Due to the
problems associated with the increased production of unwanted
by-products by many of these engineered strains, along with cur-
rent restrictions imposed on the use of genetically modified yeast in
winemaking, the pursuit of low ethanol yielding non-Saccharo-
myces yeasts has gained momentum in recent years (Contreras et al.
2014). Several studies have reported lower ethanol yields when
using non-Saccharomyces yeasts, however the decreased ethanol
production was often the result of high residual sugar levels in
these wines (Ciani and Ferraro, 1996; Ciani et al. 2006; Magyar and
Toth, 2011).
In this study we undertook a detailed metabolic screen of 86
non-Saccharomyces and five S. cerevisiae isolates from South African
vineyards (isolated from various regions and vintages) to determine
their sugar utilisation abilities as well as primary and secondary
metabolite production. Isolates were selected across a wide range
of genus and species groups for this purpose. Many of these isolates
represent species and genera that have previously not been
assessed for their possible contribution to winemaking, but which
appear highly represented in our sampling from local South African
vineyards. For many of the species targeted in our study we
included several isolates of each in order to explore intraspecies
variability. The isolates were characterized in detail with regards to
primary fermentation kinetics (ethanol production in particular) as
well as the higher alcohols, esters and monoterpenes produced. A
subset of these isolates (23) showing the most promising results in
terms of application to wine production were evaluated in
sequential inoculation (in 200 ml fermentation volumes) with the
wine yeast VIN13. Based on these outcomes four strains were
further selected for winemaking trials in both red- and white-grape
musts. Detailed sensory analyses were performed for the red wines
produced e a key component which is lacking in previous studies of
non-Saccharomyces yeasts for low ethanol wines. Finding vineyard
isolates with characteristics cross-cutting the targets of sugar uti-
lisation potential, low ethanol yields and improved aroma was a
key aim of this study. The extent of our analyses, including sensory
evaluations, and the number of isolates included, make this the
33
largest study of its kind.
2. Materials and methods
2.1. Strains, media and culture conditions
The yeast isolates used in the initial screen were selected from
the strain collection of the Institute for Wine Biotechnology (IWBT).
The culture collection contains grape and vineyard isolates of both
Saccharomyces and non-Saccharomyces yeasts isolated over
numerous years from various locations across the South African
winelands, some of which have been described previously (Setati
et al. 2012). The isolates were characterised by RFLP analysis
(Esteve-Zarzoso et al. 1999) and ribosomal RNA gene amplification
and sequencing as described by Lee and Taylor (1990).
Pure freeze cultures were streaked out on YPD agar. For liquid
cultures cells were cultivated in YPD synthetic media containing 1%
yeast extract (Biolab, South Africa), 2% peptone (Fluka, Germany),
2% glucose (Sigma, Germany). Solid medium was supplemented
with 2% agar (Biolab, South Africa).
The isolates screened in this study are listed in Table 1. The wine
yeast strains EC1118 and VIN13 were used as controls throughout
for all fermentations.
2.2. Fermentation conditions
Yeasts were maintained on YPD plates and overnight cultures
were grown in YPD broth. Cell proliferation was determined
spectrophotometrically (PowerwaveX, Bio-Tek Instruments) by
measuring the optical density (at 600 nm) of 200 ml samples of
serial dilutions of resuspended cells. Cells from overnight cultures
were harvested by centrifugation, washed and subsequently inoc-
ulated into the synthetic must, the composition of which is based
on the chemically defined medium developed by the Australian
Wine Research Institute (Jiranek et al. 1995) with amino acid ad-
ditions based on the MS300 synthetic must of Bely et al. (1990).
Sugar concentrations in the must were 100 g$L
1 glucose and
100 g$L
1 fructose. The pH was buffered at 3.3 with NaOH.
Initially, pure culture micro-scale fermentations (three inde-
pendent biological repeats) were conducted in 20 ml SPME vials for
a preliminary screen of the yeast isolates shown in Table 1. The
fermentation vials contained 16 ml of synthetic must and were
inoculated at an initial cfu/ml of approximately 2  106 (an OD600
of 0.2  107). Vials were sealed with water-filled airlocks and fer-
mentations conducted at 25  C. The progression of the fermenta-
tions was monitored by weight loss. Fermentations were assessed
as ‘stuck’ when no further weight loss was observed for three
consecutive days. At the end of fermentation the must was sampled
for metabolite analysis (ethanol, glycerol, acetic acid, reducing
sugars and volatile alcohols, esters and monoterpenes).
For the sequential fermentations, 250 ml Erlenmeyer vessels
were used (containing 200 ml of the fermentation medium) and
sealed with a rubber bung and an S-bend airlock. All fermentations
were inoculated with the non-Saccharomyces yeasts (and control
VIN13) at an initial cell density of OD600 ¼ 0.2 (2  106 cfu/ml) and
fermented at 22  C. Fermentations were assessed by weight loss,
and samples were taken at specific intervals for HPLC analysis and
spectrophotometric analysis (OD600) was performed to give an
indication of the growth of the inoculated strains. For this set of
fermentations, the wine yeast VIN13 was inoculated after 7 days (at
an initial OD600 of 0.1) to complete the fermentations. All fer-
mentations were complete by 21 days and the final fermented must
was subjected to chemical analysis. All micro-scale and sequential
fermentations were conducted in triplicate (independent biological
repeats).
34 D. Rossouw, F.F. Bauer / Food Microbiology 55 (2016) 32e46

Table 1 concentrations (g$l
1) or yield (g$g
1

sugar consumed). All data is re-
Species names and isolate ID's of the 91 strains used in this study. Two wine yeast ported as the average of three replicates ± standard deviation.
strains, EC1118 and VIN13 were included as controls (bold, italics). All other yeasts,
including the other five Saccharomyces cerevisiae strains, are isolates from the
Western Cape winelands, South Africa. 2.3.2. GCMS
Volatile aroma compounds (alcohols, esters, fatty acids and
ID ID
essential oils) were measured by headspace GCMS analysis at the
Candida Metchnikowia end of fermentation in both the micro-scale and sequential fer-
Candida albicans Y882 Metchnikowia fructicola Y1005
mentations as follows: Samples (5 ml) were transferred to 20 ml
Candida albicans Y891 Metchnikowia fructicola Y1009B
Candida albicans Y892 Metchnikowia pulcherrima Y1065 SPME vials and 50 ml of the internal standard (Anisole_d8) was
Candida albicans Y894 Metschnikowia pulcherrima Y1093 added. The volatiles were then extracted from the headspace of the
Candida albicans Y867 Metschnikowia pulcherrima Y1094 vial. The vials were subsequently allowed to equilibrate for 5 min at
Candida albicans Y1023 Metschnikowia pulcherrima Y1118
60  C in the CTC (CTC Analytics combiPAL) autosampler incubator
Candida albicans Y1097B Metschnikowia pulcherrima Y1119
Candida albicans Y1040 Metschnikowia pulcherrima Y1122
and after this equilibration time, a 50/30 mm divinylbenzene/car-
Candida apicola Y1091 Metschnikowia pulcherrima Y1103 boxen/polydimethylsiloxane coated fiber (grey) was exposed to the
Candida azyma Y1014 Metschnikowia pulcherrima Y1101B headspace for 30 min at 60  C. After sampling, the volatile com-
Candida parapsilosis Y838 Metschnikowia pulcherrima Y1109B pounds adsorbed on the fiber coating were desorbed for 20 min in
the GC injection port operated in pulsed splitless mode maintained
Cryptococcus Pichia
Cryptococcus terrestris Y1002 Pichia Kudriavzevii Y1130
at 250  C.
Cryptococcus carnescens Y815 Pichia Kudriavzevii Y1134 Separation was performed on a gas chromatograph, Agilent
Cryptococcus magnus Y817 Pichia carribica Y1036 technologies network GC system, model 6890N, coupled with an
Cryptococcus oeirensis Y819 Pichia carribica Y1037 Agilent technologies inert XL EI/CI Mass Selective Detector (MSD)
Cryptococcus oeirensis Y872
model 5975B (Agilent Technologies Inc., Palo Alto, CA). The GCMS
Cryptococcus flavescens Y841 Rhodotorula
Cryptococcus flavescens Y844 Rhodotorula glutinis Y824 system was coupled with a CTC Analytics PAL autosampler. Sepa-
Cryptococcus flavescens Y1006A Rhodotorula sloofiae Y878 ration of the essential oils was performed on a FFAP column (60 m,
Cryptococcus saitoi Y873 Rhodoturula mucilaginosa Y1027 0.25 mm ID, 0.5 mm film thickness) model J&W 122e3263 (Agilent
Cryptococcus terrestris Y1004 Rhodoturula mucilaginosa Y1074 Technologies Inc., Palo Alto, CA).
Cryptococcus terrestris Y1007 Rhodotorula mucilaginosa Y1076
Helium was used as carrier gas at a flow rate of 1 ml/min. The
Cryptococcus terrestris Y1006B Rhodotorula mucilaginosa Y1077
injector temperature was maintained at 250  C and operated in
Hanseniaspora Issatchenkia pulsed splitless mode at a pulse pressure of 300 kpa. The purge flow
Hanseniaspora vineae Y1021 Issatchenkia hanoiensis Y1105 was set at 50 ml/min with both the pulse and purge times of 1 min
Hanseniaspora vineae Y1034 Issatchenkia terricola Y1137 respectively. The oven temperature was as programmed as follows:
Hanseniaspora vineae Y1090 Issatchenkia terricola Y809
40  C for 2 min; and then ramped up to 240  C at a rate of 5  C/min
Hanseniaspora opuntiae Y1041
Hanseniaspora opuntiae Y1056 Aureobasidium and held for 2 min. The total run time was 44.00 min. The MSD was
Hanseniaspora opuntiae Y1055 Aureobasidium pullulans Y1068 operated in full scan mode and the source and quad temperatures
Hanseniaspora opuntiae Y1083 Aureobasidium pullulans Y1010 were maintained at 230  C and 150  C, respectively. The transfer
Hanseniaspora opuntiae Y1085 Aureobasidium pullulans Y1011 line temperature was maintained at 250  C. The mass spectrometer
Hanseniaspora opuntiae Y1043
Hanseniaspora opuntiae Y1126 Saccharomyces
was operated under electron impact mode at ionization energy of
Hanseniaspora opuntiae Y1101A Saccharomyces cerevisiae Y1009C 70 eV, scanning from 35 to 350 m/z.
Hanseniaspora uvarum Y1080 Saccharomyces cerevisiae Y1088 Data integration and compound quantitation was performed
Hanseniaspora uvarum Y1096 Saccharomyces cerevisiae Y1022 using Chemstation 2.0 with the NIST05 (National Institute of
Hanseniaspora uvarum Y1098 Saccharomyces cerevisiae Y1050
Standards and Technology) and/or the Wiley (275) library data-
Hanseniaspora uvarum Y1092 Saccharomyces cerevisiae Y1052
Hanseniaspora uvarum Y1097A Saccharomyces cerevisiae EC1118 bases used for compound identification (95% cut-off).
Hanseniaspora uvarum Y1079 Saccharomyces cerevisiae VIN13
Hanseniaspora uvarum Y1131 2.3.3. GC-FID
Hanseniaspora uvarum Y1104 Other In the grape juice fermentations higher alcohols and esters were
Hanseniaspora uvarum Y1116 Exhophiala xenobiotica Y849
Hanseniaspora uvarum Y1117 Tremella globispora Y1009A
quantified by GC-FID analysis as described by Rossouw et al. (2008).
Hanseniaspora uvarum Y1133 Tremella globispora Y1081
Hanseniaspora uvarum Y1135 Zygoascus meyerae Y1084 2.4. Wine fermentation
Hanseniaspora uvarum Y1138 Sporobolomyces nylandii Y1127
Hanseniaspora uvarum Y1139 Lachancea thermotolerans Y1109A
Four of the non-Saccharomyces isolates, namely H. opuntiae
Hanseniaspora uvarum Y1121 Starmerella bacillaris Y859
Hanseniaspora uvarum Y1044 Meyerozyma caribbica Y852 (Y1055), P. kudriavzevii (Y1030), H. uvarum (Y1035) and Candida
Hanseniaspora uvarum Y1100 Meyerozyma guilliermondii Y880 flavescens (Y844) were selected for trials in two 2013/2014 season
grape musts (Pinotage and Sauvignon Blanc) harvested from the
Welgevallen experimental vineyards at Stellenbosch, South Africa.
The Sauvignon Blanc must presented a YAN of 185 mg/L (adjusted
2.3. Metabolite analysis to 250 mg/L with diammonium phosphate), total initial reducing
sugar concentration of 236 g/L and a pH of 3.5. For the Pinotage
2.3.1. HPLC must the initial sugars were 230 g/L and the YAN and pH were
Glucose, fructose, glycerol and ethanol were monitored by HPLC 266 mg/L and 3.6 respectively. Standard winemaking practices
analysis. Components were separated on a Biorad Aminex HPX-87H were employed for both the red and white wine fermentations. Five
(300  7.8 mm) column, at 55  C with 0.5 mM H2SO4 as mobile litre fermentations were carried out in quadruplicate for these four
phase at a flow rate of 0.5 ml$min
1 (Eye ghe-Bickong et al. 2012). yeasts as well as the control (S. cerevisiae strain VIN13). Fermen-
Agilent RID and UV detectors were used in tandem for peak tations were inoculated at an initial cfu/ml of approximately
detection and quantification. Analysis was carried out using the 3  106 from wet cultures grown in standard YPD. GoFerm protect
HPChemstation software package. Data are either represented as (Lallemand) was added to the yeast cultures pre-inoculation at
 D. Rossouw, F.F. Bauer / Food Microbiology 55 (2016) 32e46 35

Table 2
Residual sugar levels (g$L
1), ethanol yields and glycerol yields (g$g
1 sugars utilised) of selected strains grown in synthetic must as pure cultures at the end of fermentation
(no weight loss for three consecutive days). Strains are grouped by species, and ranked in ascending order in terms of residual sugar levels within individual groups containing
more than one species. Isolate species names and database ID's are indicated in bold font for those isolates that were subsequently evaluated in the sequential fermentations.
All values (for glycerol and ethanol yields) that are statistically significantly different (p < 0.05) from both VIN13 and EC1118 are indicated in bold font (for an increase relative to
VIN13 and EC1118) and in italics (for a decrease relative to VIN13 and EC1118). STDev refers to the standard deviations. Values are the average of three biological repeats.

Species ID Residual sugars STDev Ethanol yield STDev Glycerol yield STDev

S. cereivisiae VIN13 3.8 0.4 0.53 0.003 0.023 0.0018

S. cereivisiae EC1118 4.1 2.1 0.54 0.045 0.028 0.0023
S. cereivisiae Y1009C 4.2 2.6 0.52 0.023 0.026 0.0037
S. cereivisiae Y1052 7.3 1.0 0.46 0.001 0.013 0.0004
S. cereivisiae Y1022 11.8 2.6 0.44 0.000 0.012 0.0007
S. cereivisiae Y1050 19.1 0.2 0.46 0.008 0.013 0.0004
S. cereivisiae Y1088 19.2 5.5 0.45 0.013 0.013 0.0080
H. uvarum Y1080 48.8 5.3 0.53 0.002 0.027 0.0024
H. uvarum Y1117 111.1 4.8 0.40 0.008 0.030 0.0018
H. uvarum Y1044 111.3 13.1 0.52 0.018 0.063 0.0053
H. uvarum Y1116 113.8 1.4 0.37 0.011 0.030 0.0006
H. uvarum Y1100 126.0 4.6 0.52 0.020 0.068 0.0047
H. uvarum Y1135 130.3 5.8 0.31 0.029 0.023 0.0035
H. uvarum Y1121 130.4 3.7 0.28 0.025 0.022 0.0002
H. uvarum Y1131 131.2 1.5 0.31 0.006 0.020 0.0006
H. uvarum Y1104 142.8 0.7 0.35 0.001 0.031 0.0010
H. uvarum Y1096 149.0 2.8 0.48 0.034 0.038 0.0017
H. uvarum Y1138 149.6 8.6 0.50 0.049 0.035 0.0040
H. uvarum Y1097A 151.9 0.7 0.29 0.035 0.041 0.0012
H. uvarum Y1092 153.5 2.1 0.50 0.026 0.041 0.0041
H. uvarum Y1098 158.4 0.7 0.53 0.077 0.046 0.0012
H. vineae Y1034 68.0 8.4 0.42 0.013 0.018 0.0026
H. vineae Y1021 74.2 1.9 0.42 0.011 0.022 0.0016
H. vineae Y1090 105.8 3.1 0.54 0.006 0.035 0.0027
H. opuntiae Y1085 103.2 3.8 0.53 0.010 0.067 0.0020
H. opuntiae Y1055 128.0 5.9 0.44 0.041 0.058 0.0025
H. opuntiae Y1101A 129.9 11.7 0.46 0.025 0.062 0.0094
H. opuntiae Y1043 136.4 12.8 0.54 0.019 0.061 0.0010
H. opuntiae Y1056 143.6 2.3 0.45 0.007 0.111 0.0016
M. pulcherrima Y1101B 124.2 5.8 0.57 0.064 0.034 0.0085
M. pulcherrima Y1094 159.9 5.6 0.68 0.050 0.047 0.0053
M. pulcherrima Y1103 161.1 9.0 0.63 0.056 0.048 0.0011
M. pulcherrima Y1093 171.0 0.6 0.52 0.058 0.101 0.0077
M. pulcherrima Y1118 134.1 7.5 0.48 0.033 0.034 0.0014
C. flavescens Y844 41.4 2.9 0.43 0.027 0.061 0.0027
C. flavescens Y1006A 140.0 11.2 0.30 0.039 0.021 0.0038
M. fructicola Y1005 79.7 8.1 0.32 0.004 0.017 0.0012
M. fructicola Y1009B 146.4 13.6 0.53 0.043 0.033 0.0020
T. globispora Y1009A 137.8 1.7 0.48 0.039 0.066 0.0080
T. globispora Y1081 146.8 15.7 0.52 0.003 0.017 0.0030
L. thermotolerans Y1109A 92.8 14.8 0.51 0.037 0.017 0.0009
C. albicans Y1040 98.0 6.0 0.54 0.014 0.037 0.0058
C. oeirensis Y872 104.8 7.6 0.45 0.011 0.055 0.0024
P. carribica Y1036 116.1 11.3 0.36 0.016 0.018 0.0011
S. bacillaris Y859 128.9 3.7 0.48 0.028 0.049 0.0008
P. kudriavzevii Y1130 146.9 4.5 0.32 0.006 0.027 0.0022

30 g/hL. On day seven the non-Saccharomyces fermentations were
inoculated with VIN13 to complete the fermentations (2  106 cfu/
ml). All fermentations were carried out at 20  C and samples taken
on days 0, 4, 7 and 11 for microbial and chemical analysis in the case
of the Pinotage fermentations and at days 0, 4, 7, 11, 14 and 21 in the
case of the Sauvignon Blanc fermentations. Cell counts of non-
Saccharomyces yeast and S. cerevisiae were determined by plating
serial dilutions of fermentation samples on yeast peptone dextrose
(YPD) agar and lysine media and counting colonies formed after
72 h. Weight loss was likewise monitored throughout all fermen-
tations. At the end of fermentation 50 ppm SO2 was added before
bottling. The final wines were chemically analysed by HPLC and GC-
FID for volatile alcohols and esters.
2.5. Sensory analysis
The wines produced in the Pinotage trials were subjected to
sensory analysis. The sensory analysis was conducted in a well-
ventilated sensory lab with temperature control (20 ± 2  C) and
off-white tasting booths. No communication between panellists
was allowed. Samples were presented with three-digit codes and
the order of presentation was randomised according to a Williams
Latin square design and different for every judge. All panellists
were experienced sensory judges, who take part in sensory analysis
on a regular basis and received training on similar wines during
previous studies. Panellists did not receive any training on the
specific sample sets. Sorting was carried out by 28 panellists as
described by Chollet et al. (2010). Judges were asked to group
similar wines together based on aroma. Assessors were free to
create as many groups as they wanted and to put as many wines in
each group as they wished. The panellists were then required to
ascribe sensory descriptors to the different groupings which they
had identified (free choice profiling). Paper ballots were used and
sensory data was captured using Microsoft excel. Data analysis of
sensory data was conducted using XLStat 2013 (Addinsoft, www.
xlstat.com). An individual similarity matrix was constructed for
36 D. Rossouw, F.F. Bauer / Food Microbiology 55 (2016) 32e46

Fig. 1. Ethanol, glycerol and acetic acid production, and sugar utilisation during and after fermentation with 19 selected yeast isolates and two control commercial yeast strains,
VIN13 and EC1118. Ethanol yields and glycerol yields are also shown. Data for the day 7 time point are indicated by the dark grey bars for all figures, and day 21 at the end of
fermentation by the light grey bars. Values are the average of three repeats ± standard deviation.

each judge, indicating which samples were grouped together. The
individual similarity matrices were summed. The resulting simi-
larity matrix was subjected to multidimensional scaling (MDS).
MDS analysis was performed in order to determine similarities
between samples. The frequencies of sensory attributes ascribed to
the different samples were used to construct a contingency table.
The number of attributes was reduced by a two-step process. Lin-
guistic synonyms were combined as part of data capturing. Se-
mantic grouping, of terms having similar meanings, was done by
the sensory analyst and researcher together as follows: only attri-
butes cited by more than 10% of the judges were kept as separate
terms, attributes cited less than that were either combined with
other attributes or deleted. Correspondence analysis (CA) was
conducted on the resulting contingency table.
2.6. Statistical analysis
Heatmaps were generated based on monoterpene production
levels in the small scale fermentations using MeV version 4.7.1
(Ben-Dor et al. 1999). T-tests and anova analyses were conducted
using Statistica (version 10.2).
3. Results
3.1. Sugar utilisation and ethanol production of vineyard isolates
The 91 isolates shown in Table 1 were evaluated in terms of their
ability to grow and metabolise sugars under fermentative condi-
tions in the synthetic wine-like must. Many of the yeast species
included in this study do not fall within the ‘conventional’ groups of
non-Saccharomyces yeasts commonly isolated from alcoholic
fermentation or considered for oenological application (such as
Candida albicans, or species of Cryptococcus). As such, it was ex-
pected that many yeasts would have a limited to negligible sugar
utilisation capacity under fermentative conditions. For the pur-
poses of our study, the threshold sugar utilisation capacity was the
consumption of at least 20 g/L
1 reducing sugars within 14 days
after inoculation. The residual sugars and ethanol yields of the 46
 D. Rossouw, F.F. Bauer / Food Microbiology 55 (2016) 32e46
Fig. 2. Final ethanol as a percentage in fermentations conducted by the selected strains and sequentially inoculated with VIN13 (data only shown for strains where all fermentations
proceeded to dryness e less than 5 g$L
1 residual sugars). Values are the average of three repeats ± standard deviation.
isolates which met this criterion are shown in Table 2.
Striking variation exists in the rate and extent of sugar uti-
lisation by different isolates within single species. For instance,
total sugar utilisation ranged from just 40 g$L
1 (Y1098) to
150 g$L
1 (Y1080) for two different isolates of H. uvarum. Among
the 17 different H. uvarum isolates the production of ethanol and
glycerol also varied significantly. The same was true for almost
every species included in this study (with the exception of
S. cerevisiae) as vast intraspecies differences were observed with
regards to the primary fermentation kinetics of different isolates.
Most of the non-Saccharomyces isolates yielded more glycerol
per gram of sugar utilised compared to the wine yeast strains
EC1118 and VIN13. The highest yielding glycerol producers were
Y1093 (M. pulcherrima) and Y1056 (H. opuntiae), followed by
Y1009A (T. globispora), Y1043 and Y1085 (H. opuntiae), Y1100
(H. uvarum) and Y844 (C. flavescens) which all yielded more than
0.06 g$g
1 of glycerol compared to the 0.023 g$g
1 glycerol yield of
VIN13.
A large number of isolates yielded significantly less ethanol per
gram of sugar utilised compared to the wine yeast isolates, most
notably Y1005, Y1116, Y1117, Y1036, Y1135, Y1121, Y1131, Y1006A,
Y1130 and Y1097A. These isolates all yielded between 0.3 and 0.4 g
ethanol per gram of sugar utilised. Though many yielded less
ethanol, the sugar utilisation of these isolates was often quite poor
thus their application to decrease ethanol in natural grape juice
fermentations needed further investigation.
3.2. Sequential fermentations of low ethanol yielding yeasts
Nineteen isolates which showed the ability to either utilise
sufficient sugar and yield slightly less ethanol, or which showed
substantial reductions in ethanol yields as well as the ability to
ferment at least 30 g$L
1 sugars on their own, were selected for
follow-up fermentations in sequential inoculation (after seven
days) with a widely used commercial wine yeast strain (VIN13).
Fig. 1 shows the concentrations of primary fermentation metabo-
lites by day 7 (before inoculation of VIN13) and at the end of
fermentation (day 21) for fermentations conducted by these iso-
lates and the control wine yeast strains VIN13 and EC1118.
With the exception of M. fructicola (Y1005), C. flavescens
37
(Y1006A), T. globispora (Y1081) and M. carribica (Y1036), pure cul-
tures of non-Saccharomyces yeasts selected for this part of the study
managed to consume equal to or greater than 60 g$L
1 sugars
within the first 7 days of fermentation (Fig. 1). Ethanol yields for the
non-Saccharomyces yeasts were all substantially lower at this point
(with the exception of H. uvarum strain Y1121). The most significant
increases in glycerol yield, and decreases in ethanol yield were
observed at day 7 before inoculation of the control S. cerevisiae
yeast. Though the impact of the non-Saccharomyces yeasts were
diminished by the end of alcoholic fermentation significant differ-
ences were still evident for many of the isolates. Most notably, two
H. uvarum (Y1131 and Y1135), two H. opuntiae (Y1055 and Y1056),
two H. vinae (Y1021 and Y1034) one P. kudriavzevii (Y1130) and one
C. flavescens (Y844) isolate/s showed significantly reduced ethanol
yields by the end of alcoholic fermentation (p < 0.05). These re-
ductions in ethanol yield resulted in a substantial reduction (>1.5%)
in the final ethanol content of the fermented must (Fig. 2).
In the case of C. flavescens (Y844) a large increase in glycerol
production, and a concomitant increase in acetic acid production
was evident (Fig. 1). The average 12.9 g$L
1 glycerol produced by
this yeast in the sequentially inoculated fermentations was almost
3-fold higher than the average 4.5 g$L
1 glycerol production by the
control VIN13 and EC1118 strains. M. fructicola (Y1005) was another
high-yielding glycerol producer by day 7 of fermentation, but the
relatively slow sugar utilisation of this yeast meant that its influ-
ence on the final glycerol concentration was negligible after
sequential inoculation of S. cerevisiae. The higher glycerol yields
seen by day 7 in the case of the isolates of H. opuntiae and H. uvarum
did indeed translate to increases in the final glycerol levels at the
end of these fermentations.
3.3. Differences in aroma compound production in sequential
fermentations
Though most sugars were consumed, and the fermentation
completed by the S. cerevisiae strain VIN13 inoculated after 7 days
of fermentation, significant differences in the final volatile
composition of the different fermentations were still evident for
the sequential fermentations (Fig. 3). The relative abundance of
selected fatty acids and esters at the end of alcoholic fermentation
38 D. Rossouw, F.F. Bauer / Food Microbiology 55 (2016) 32e46

Fig. 3. Levels of selected fatty acids and esters in the fermented must at the end of fermentation. Only strains showing the greatest differences in aroma compound concentrations
at the end of fermentation compared to the controls are shown. Y-axis values are relative units of compound abundance normalised to the internal standard anisole. Values are the
average of three repeats ± standard deviation.

highlights the notable differences in the aroma forming abilities of
the different non-Saccharomyces species in combination with
VIN13. Interesting results were evident for phenylethyl acetate,
which showed greatly elevated concentrations in fermentations
inoculated with the two H. vineae isolates (Y1021 and Y1034) and
the C. flavescens strain (Y844) compared to the two controls (Fig. 3).
In contrast, phenylethyl acetate was not present above the detec-
tion threshold in P. carribica and I. orientalis -inoculated fermen-
tations. Likewise, the two H. vineae isolates (Y1021 and Y1034) also
produced significantly increased concentrations of ethyl acetate
compared to the control strains and most other non-Saccharomyces
isolates. P. kudriavzevii (Y1130) and H. opuntiae (Y1101A) were
likewise high producers of ethyl acetate, which can be considered a
spoilage compound at high concentrations.
Decanoic acid levels were greatly elevated for the two H. vineae
isolates (Y1021 and Y1034) and decreased for two H. opuntiae iso-
lates (Y1055 and Y1056) as well as for C. flavescens (Y844). Differ-
ences in the final ester levels were also evident for many of the
isolates tested, with several showing reduced ethyl hexanoate and
ethyl hexadecanoate levels compared to the control wine yeast
strains. In addition, the two H. opuntiae isolates Y1055 and Y1056,
as well as C. flavescens Y844 showed reduced ethyl octanoate, ethyl
decanoate and ethyl dodecanoate levels compared to the control
strains.
3.4. Monoterpene production by non-Saccharomyces yeasts
The aroma impact of non-Saccharomyces yeasts in fermentation,
as well as their ability to produce volatile compounds not
commonly associated with fermentations conducted by S. cerevisiae
wine yeast strains (such as high concentrations of terpenoid com-
pounds) has been documented (Cordero-Otero et al. 2003; Sadoudi
et al. 2012). Pure culture micro-scale fermentations conducted by
the isolates listed in Table 1 were also analysed by headspace GCMS
analysis to identify those yeasts showing production of mono-
terpenes and associated compounds (Fig. 4). Several isolates were
responsible for the formation of such compounds in pure culture
fermentations, indicative of de novo synthesis as no precursors for
 D. Rossouw, F.F. Bauer / Food Microbiology 55 (2016) 32e46
Fig. 4. Heatmap showing monoterpene production by strains investigated in this
study. Grey indicates non-production of a particular compound, while increasing
colour intensity in the yellow-orange-red range correlates to the relative concentra-
tions of the monoterpenes (normalised to the internal standard anisole) produced by a
particular strain in pure culture.
these compounds exist in the synthetic must. Isolates of Crypto-
coccus and Candida (particularly C. albicans) produced high
amounts of farnesol and farnesene. Nerolidol, linalool, geraniol and
citronellol were also produced by several isolates in our study.
Importantly, several of the low-ethanol candidate yeasts identified
were monoterpene producers as well, such as the H. opuntiae iso-
lates Y1055 and Y1056, H. vinae Y1034 and C. flavescens Y844.
3.5. Wine trials: Pinotage fermentation kinetics
Four of the non-Saccharomyces isolates, namely H. opuntiae
(Y1055), P. kudriavzevii (Y1030), H. uvarum (Y1035) and C. flavescens
(Y844) were selected for trials in grape must of one red (Pinotage)
and one white (Sauvignon Blanc) cultivar. The weight loss and cell
enumeration data for the Pinotage fermentations are shown in
Fig. 5. Fermentations were complete by Day 7 in the case of VIN13
and by Day 11 for the different non-Saccharomyces treatments.
The initial pinotage must contained 2.2  105 cfu/ml, of which
less than 1% were S. cerevisiae. After inoculation, non-Saccharo-
myces yeasts increased in cell number steadily in the first four days
of fermentation, before declining towards the day 7 sampling point.
By day 4 of fermentation the ratio of non-Saccharomyces yeast to
S. cerevisiae was roughly 1:1. Non-Saccharomyces yeasts in the
39
C. flavescens einoculated fermentations showed the lowest level of
growth, followed by P. kudriavzevii, with faster growth and
improved persistence of non-Saccharomyces yeasts towards the
later stages of fermentation for the H. uvarum and H. opuntia
einoculated fermentations. The inoculated control fermentation
was completely dominated by rapid growth of S. cerevisiae, with
these fermentations almost reaching dryness by day 4 of fermen-
tation. All fermentations were complete by day 11 of fermentation,
with less than 2 g$L
1 residual sugars in all repeats of all treat-
ments. Glycerol and ethanol levels, as well as ethanol yields, are
shown in Fig. 6.
Significantly reduced final ethanol concentrations were
observed for fermentations inoculated with H. uvarum (Y1135) and
H. opuntiae (Y1055). No significant difference was observed for the
C. flavescens treatment, most likely due to the slow growth and
fermentation rate of this strain in the Pinotage must. Final glycerol
concentrations were however still highest (3 g$L
1 higher) for
these fermentations compared to the VIN13 control, highlighting
the fact that even strains of non-Saccharomyces which may not
grow well or persist throughout fermentation can still have a sig-
nificant impact on the final concentrations of major fermentation
metabolites. Despite the higher glycerol concentrations in these
and the other non-Saccharomyces einoculated fermentations no
significant differences in the final acetic acid concentrations of the
wines were observed.
3.6. Wine trials: Sauvignon Blanc fermentation kinetics
The initial Sauvignon Blanc must contained 1.5  105 cfu/ml, of
which less than 1% were S. cerevisiae. Fermentations inoculated
with the control VIN13 proceeded rapidly and residual sugars were
below 3 g$L
1 by day 11 of fermentation (Fig. 7, Frame A). Fer-
mentations were sluggish towards the end of fermentation in the
non-Saccharomyces inoculated fermentations and only reached
complete dryness (less than 2 g$L
1 residual sugars) by day 21 of
fermentation. Growth curves of non-Saccharomyces yeasts show a
general increase in cell numbers up to day 7 of fermentation (the
point at which VIN13 was inoculated) after which non-Saccharo-
myces cell counts slowly decline (rapidly in the case of C. flavescens-
inoculated fermentations) while S. cerevisiae cell numbers steadily
increase up to day 11 in these fermentations (Fig. 7, frames B and C).
Non-Saccharomyces yeasts in the H. opuntiae and H. uvarum
einoculated fermentations once again demonstrate the ability to
grow and persist relatively well in the Sauvignon Blanc must
compared to the other two non-Saccharomyces treatments.
In terms of the impact of the non-Saccharomyces treatments on
the final ethanol concentrations in the wines produced, clear im-
pacts were seen for ethanol yields during the early non-Saccharo-
myces dominated stages of the fermentation (Fig. 8). Even though
S. cerevisiae dominated the later fermentation stages, the overall
ethanol yields (and total ethanol concentration) at the end of
fermentation were lower for the non-Saccharomyces treatments.
Ethanol reductions were most pronounced for fermentations
inoculated with H. opuntiae Y1130 (1.3% reduction) and H. uvarum
Y1055 (1.1% reduction). Glycerol production was significantly
higher compared to the control for all non-Saccharomyces treat-
ments, particularly in the case of the C. flavescens inoculated fer-
mentations (60% increase compared to VIN13 controls). In the
Sauvignon Blanc fermentations the final acetic acid concentrations
were statistically significantly different for all the non-Saccharo-
myces treatments, including P. kudriavzevii (0.28 g$L
1),
C. flavescens (0.41 g$L
1), H. opuntiae and H. uvarum (both
0.43 g$L
1) compared to the control VIN13 (0.09 g$L
1). However
these concentrations all fall within the acceptable range for wine.
40 D. Rossouw, F.F. Bauer / Food Microbiology 55 (2016) 32e46
Fig. 5. Weight loss during fermentation in Pinotage grape must (Frame A), as well as cell proliferation of non-Saccharomyces (Frame B) and S. cerevisiae (Frame C) at different time
points during fermentation for the four non-Saccharomyces treatments (H. opuntiae Y1055, P. kudriavzevii Y1030, H. uvarum Y1035, C. flavescens Y844) and control S. cerevisiae VIN13.
All data are the average of four repeats ± standard deviation.
3.7. Chemical and sensory analysis of final wines
To provide proof-of-concept, sensory analysis was performed
based on the aroma bouquet of the Pinotage wines to determine
whether the different non-Saccharomyces treatments could be
distinguished by panellists. Wines were grouped according to their
perceived similarity by panellists and sensory descriptors provided
for each group. The results for the sorting by aroma for the Pinotage
is shown in Fig. 9.
Clear groupings were evident for the control treatments based
on aroma. Negative descriptors most often associated with the
control VIN13 -inoculated wines were ‘alcohol’, ‘overripe fruit’ and
‘vegetable’. Positive descriptors were ‘spicy’ and ‘black fruits’.
Wines inoculated with the two species from the Hanseniaspora
genus, namely H. uvarum and H. opuntiae, grouped together and
were associated with positive attributes such as ‘hazelnut’, ‘coffee’,
‘caramel’ and ‘cherry’. The only negative attribute that correlated
with this cluster was ‘acetone’, likely due to the higher concen-
trations of ethyl acetate in these fermentations (Supplementary
table S1). ‘Oaky’, ‘floral’ and ‘earthy’ were often used to describe
wines fermented with P. kudriavzevii and C. flavescens. Overall the
sorting task showed that panellists were able to detect clear dif-
ferences between fermentations inoculated with strains of Hanse-
niaspora, P. kudriavzevii and C. flavescens, and the control. The
sensory analysis further shows a generally positive response of
panellists to the non-Saccharomyces fermented Pinotage. Another
noteworthy finding was that distinct aroma impacts of the different
non-Saccharomyces -inoculated musts could be discerned even for
those treatments (i.e. P. kudriavzevii) where the non-Saccharomyces
yeasts were outcompeted by S. cerevisiae relatively early during
fermentation.
Chemical analysis was performed by GC-FID in order to quantify
volatile alcohols and esters in the wines at the end of alcoholic
fermentation. In Tables 3 and 4 only those compounds showing
statistically significant differences for each of the non-Saccharo-
myces inoculated wines compared to the control are shown. Ab-
solute concentrations of the 26 compounds analysed are presented
in supplementary tables S1 and S2.
Clearly significant changes to the aroma profiles of the non-
Saccharomyces einoculated wines were evident, more so for the
Sauvignon Blanc wines than the Pinotage (Tables 3 and 4). Clear
treatment differences are also seen, as certain volatiles are
increased in concentration compared to the control for some
strains, but decreased for others i.e. isoamyl acetate and 2-phe-
nylethyl acetate in the Sauvignon Blanc wines. The non-Saccharo-
myces yeasts thus have a noteworthy impact on the final aroma
profile, regardless of their inability to dominate numerically
throughout the entire fermentation. The metabolic impact of these
yeasts during the early stages of fermentation is sufficient to impart
significant changes to the final balance of volatile alcohols and
esters produced.
The impact of the non-Saccharomyces yeasts is also dependent
on the composition of the starting must: The aroma impacts of any
of the four inoculation treatments differed between the Pinotage
and Sauvignon Blanc wines. However, certain compounds did show
some reproducibility in trends in both the white and red wines for a
given non-Saccharomyces inoculation, including ethyl acetate, iso-
amyl acetate, propanol and isovaleric acid. The impact of non-
Saccharomyces yeasts on aroma will be influenced by various must
parameters including nutrient and nitrogen availability. This
 D. Rossouw, F.F. Bauer / Food Microbiology 55 (2016) 32e46 41

Fig. 6. Fermentation derived primary metabolite concentrations at different time points during fermentation of Pinotage grape must. All data are the average of four
repeats ± standard deviation.

complexity means that the aromatic impact of an individual species
of inoculated yeast (especially in the complex microbial back-
ground of a natural grape must) may be difficult to determine and
predict.
4. Discussion
4.1. Many non-Saccharomyces yeasts appear suitable tools for
lowering ethanol levels in wine
Of the original 91 isolates used in this study, 45 (excluding the
EC1118 and VIN13 controls) utilised more than 30 g$L
1 sugars. It is
important to underscore that our fermentations represent a
defined set of conditions. Adjustments to the fermentation tem-
perature, level of oxygenation or the composition of the synthetic
must may render different outcomes for some of these isolates. Our
data suggest that a large number of species are suitable for ethanol
management approaches, as ethanol yields were significantly
decreased for many of the isolates investigated in this study
(Table 2; Fig. 1).
Non-Saccharomyces yeasts showing the most low-ethanol po-
tential for winemaking were two isolates of H. uvarum (Y1131 and
Y1135), two H. opuntiae (Y1055 and Y1056), two H. vinae (Y1021
and Y1034) one P. kudriavzevii (Y1130) and one isolate of
C. flavescens (Y844), which all showed a substantial reduction
(>1.5%) in the final ethanol content of the sequential fermentations
(Fig. 1). The ability of H. uvarum strains to decrease ethanol yields in
alcoholic fermentation has been shown previously (Ciani and
Picciotti, 1995).
Interestingly, most of the non-Saccharomyces yeasts produced
more glycerol per gram of sugars utilised in pure culture fermen-
tations, which has been previously reported for some of these
species such as L. thermotolerans and Candida zemplinina (Ciani and
Ferraro, 1998; Soden et al. 2000; Comitini et al. 2011). This could be
considered a positive impact, as glycerol contributes to smoothness
(mouth-feel), sweetness and complexity in wines (Ciani and
Maccarelli, 1998). The increased glycerol concentrations by the
end of fermentation were in most cases (with the exception of
C. flavescens) not accompanied by significant increases in acetic acid
concentrations compared to control fermentations.
4.2. Non-Saccharomyces yeasts contribute to aromatic ‘complexity’
Several of the original 91 isolates screened in our study
42 D. Rossouw, F.F. Bauer / Food Microbiology 55 (2016) 32e46
Fig. 7. Weight loss during fermentation in Sauvignon Blanc grape must (Frame A), as well as cell proliferation of non-Saccharomyces (Frame B) and S. cerevisiae (Frame B) at different
time points during fermentation for the four non-Saccharomyces treatments (H. opuntiae Y1055, P. kudriavzevii Y1030, H. uvarum Y1035, C. flavescens Y844) and control S. cerevisiae
VIN13. All data are the average of four repeats ± standard deviation.
produce monoterpenes by de novo biosynthesis of these com-
pounds (based on end-point analysis of micro-scale and sequential
fermentations). Whether these compounds are produced at the
levels required to impart a significant olfactory impact cannot be
established without further chemical and sensory analysis, how-
ever the existence of pathways for de novo synthesis of these
compounds is in itself a significant finding. Previous studies have
reported monoterpene production by non-Saccharomyces (and
some S. cerevisiae) yeasts (Carrau et al. 2005; Garcia et al. 2002;
Sadoudi et al. 2012).
Despite the inability of many non-Saccharomyces yeasts to
persist to the end of fermentation or ferment large amounts of
sugars, these species can still have a significant impact on the
production of aroma compounds during fermentation (Hernandez-
Orte et al. 2008; Garcia et al. 2002; Romano et al. 1992, 2003). Of
the isolates showing potential for use in co-fermentation to reduce
ethanol yields, several also show potentially positive aroma im-
pacts. For example, H. opuntiae Y1056 and Y1055 show significantly
reduced octanoic and decanoic acid levels (Fig. 3) which may mean
a reduction in the fatty, soapy, rancid notes associated with these
medium chain fatty acids. Furthermore, both of these yeasts pro-
duce the monoterpene citronellol, which imparts a pleasant citrus
aroma.
C. flavescens Y844 shows great potential for reducing ethanol
levels in wine, however the 0.6 g$L
1 increase in acetic acid pro-
duction by this strain compared to the control could be problem-
atic. The increase in acetic acid associated with this yeast may be
alleviated to some extent in a natural grape must, which was the
case for the grape must trials conducted in our study. On the pos-
itive side, fermentations conducted with this isolate resulted in a
large increase in phenylethyl acetate levels. In small scale pure
culture fermentations this yeast also showed the ability to syn-
thesise the largest number of different monoterpenes of any of the
isolates in this study (Fig. 4), including nerolidol (woody, fresh
bark), limonene (citrus), linanool (floral, spicy) and farnesene and
farnesol (herbal, fresh, floral).
4.3. Application to fermentation of grape must
Trials in grape must from both red and white cultivars under-
score the potential for non-Saccharomyces yeasts identified in our
study to be applied in sequential inoculation strategies in com-
mercial fermentations with the aim of reducing ethanol levels. In
both the white and red wine fermentations, a direct correlation was
observed between the ability of a given strain to grow and persist
during the course of fermentation, and the final ethanol reduction
observed. In these wine trials, the influence of the indigenous
microflora cannot be excluded, however the clear treatment
especific differences between the four non-Saccharomyces inocu-
lated wines and control suggest that these differences are likely due
to the inoculated yeast.
In the Sauvignon Blanc fermentations all non-Saccharomyces
treatments appeared to have a greater impact on ethanol levels
compared to the Pinotage fermentations. The slower dominance of
the S. cerevisiae over the non-Saccharomyces yeasts in the Sau-
vignon Blanc must likely accounts for the greater metabolic impact
of the non-Saccharomyces yeasts in terms of the final ethanol levels.
In light of the twice daily punch-downs performed during the red
wine fermentations, it was however anticipated that the growth
and metabolic impact of the non-Saccharomyces species would
have been augmented in the Pinotage due to the increased oxygen
availability, which was not the case here (Hansen et al., 2001;
Morales et al., 2015). This highlights the fact that other factors
related to must composition play a significant role in the growth
and persistence of non-Saccharomyces yeasts, independently of
aeration. Research into the nutritional (particularly nitrogen) re-
quirements of non-Saccharomyces yeasts is needed in order to
understand the influence of must composition on fermentation
outcomes in non-Saccharomyces inoculated fermentations.
Despite the complexity of interactions between our inoculated
non-Saccharomyces yeasts, the nutritional matrix of the grape must,
as well as the presence of other competing microorganisms at the
start of fermentation, significant changes in the production of
 D. Rossouw, F.F. Bauer / Food Microbiology 55 (2016) 32e46 43

Fig. 8. Fermentation derived primary metabolite concentrations at different time points during fermentation of Sauvignon Blanc grape must. All data are the average of four
repeats ± standard deviation.

Fig. 9. Results of sensory analysis based on the aroma of Pinotage wines produced by the different non-Saccharomyces treatments and control. In Frame A the results of MDS
analysis shows the outcomes of the sorting task. Samples which cluster closer together were most often grouped together by panellists, whereas those that are further apart were
rarely grouped together (Kruskal's stress of 0.173). Frame B shows the correspondence analysis based on aroma descriptors assigned to these fermentations (free choice profiling).
44 D. Rossouw, F.F. Bauer / Food Microbiology 55 (2016) 32e46

Table 3
Differences in aroma compound composition at the end of fermentation for Pinotage wines inoculated with four strains of non-Saccharomyces yeast. Statistically significant
differences (p < 0.05) were determined for each compound quantified in these fermentations and compared to the control VIN13 fermentations (based on four inde-
pendent biological repeats for each treatment). Compounds which were significantly increased in a particular non-Saccharomyces treatment compared to the control are
indicated in bold font, whilst those that were decreased relative to the control are formatted in italics.

C. flavescens H. opuntiae P. kudriavzevii H. uvarum

Ethyl acetate Ethyl acetate Ethyl acetate Ethyl acetate

Methanol Methanol
Ethyl butyrate
Propanol Propanol
Isobutanol Isobutanol Isobutanol
Isoamyl Acetate Isoamyl Acetate Isoamyl Acetate Isoamyl Acetate
Acetoin
3-ethoxy-1-propanol 3-ethoxy-1-propanol
Propionic Acid Propionic Acid Propionic Acid Propionic Acid
Isobutyric acid
Ethyl Caprate Ethyl Caprate
Iso-Valeric Acid Iso-Valeric Acid Iso-Valeric Acid
Diethyl succinate Diethyl succinate Diethyl succinate
Ethyl phenylacetate Ethyl phenylacetate Ethyl phenylacetate Ethyl phenylacetate
Hexanoic acid

ethanol and glycerol were observed in most of the treatments. In
the Pinotage fermentations, the greatest reduction in final ethanol
concentrations (%) was observed for H. uvarum 1035, followed by
H. opuntiae 1055 (0.8 and 0.6 percent v/v respectively; Fig. 6). In the
Sauvignon Blanc, these strains again led to the lowest final ethanol
concentrations, with a reduction of up to 1.3 percent (v/v)
compared to the control (Fig. 8).
Besides the impact on ethanol yields, inoculation of the non-
Saccharomyces yeasts resulted (in many cases) in arguably positive
aroma impacts and sensory outcomes (Fig. 9). Mostly positive
fruity, flora, and nutty aromas were associated with the non-
Saccharomyces treatments, compared to control fermentations
which were often described as ‘alcoholic’. Concentrations of most
higher alcohols and esters quantified in the finished treatment
wines were also significantly different from one another and/or the
control (Tables 3 and 4).
4.4. Phenotypic variation of non-Saccharomyces wine yeast species
The most striking finding of our work is the very large pheno-
typic space of many of the species analysed here with regard to
sugar utilisation and aromatic features. Many of these species had
previously not been further explored for oenological application
because individual strains which had been assessed had not per-
formed according to minimum requirements. In our case, basic
yeast physiological and oenological features such as the sugar uti-
lisation of isolates from single species showed a very wide range, as
in the case of H. uvarum, H. opuntiae and M. pulcherrima to name a
few. This observation highlights the limitation of approaches based
on single or very limited numbers of strains. Many such previously
analysed species may deserve further consideration for oenological
applications.
The intraspecies variation of H. uvarum isolates have been
demonstrated previously with regards to variations in higher
alcohol, ethyl acetate and acetaldehyde production specifically
(Romano et al., 2003). Large intrastrain differences in particularly
acetate ester production were also noted for Hanseniaspora, Pichia
and Candida species (Viana et al., 2008). In terms of primary
fermentation compounds, intrastrain differences in sugar uti-
lisation and ethanol yields have previously been shown to some
extent for a smaller set of non-Saccharomyces yeasts (Contreras
et al., 2014). The large differences in sugar utilisation for different

Table 4
Differences in aroma compound composition at the end of fermentation for Sauvignon Blanc wines inoculated with four strains of non-Saccharomyces yeast. Statistically
significant differences (p < 0.05) were determined for each compound quantified in these fermentations and compared to the control VIN13 fermentations (based on four
independent biological repeats for each treatment). Compounds which were significantly increased in a particular non-Saccharomyces treatment compared to the control are
indicated in bold font, whilst those that were decreased relative to the control are formatted in italics.

C. flavescens H. opuntiae P. kudriavzevii H. uvarum

Ethyl Acetate Ethyl Acetate Ethyl Acetate Ethyl Acetate

Propanol Propanol Propanol Propanol
Isobutanol Isobutanol Isobutanol Isobutanol
Isoamyl Acetate Isoamyl Acetate Isoamyl Acetate
Butanol Butanol Butanol Butanol
Isoamyl alcohol Isoamyl alcohol Isoamyl alcohol
Hexyl Acetate
Acetoin Acetoin Acetoin
Ethyl Lactate Ethyl Lactate Ethyl Lactate Ethyl Lactate
Hexanol Hexanol Hexanol
3-ethoxy-1-propanol 3-ethoxy-1-propanol 3-ethoxy-1-propanol 3-ethoxy-1-propanol
Ethyl Caprylate Ethyl Caprylate
Propionic Acid Propionic Acid
Isobutyric acid Isobutyric acid
Iso-Valeric Acid Iso-Valeric Acid Iso-Valeric Acid Iso-Valeric Acid
2-Phenylethyl Acetate 2-Phenylethyl Acetate 2-Phenylethyl Acetate
Hexanoic Acid Hexanoic Acid Hexanoic Acid Hexanoic Acid
2-Phenyl Ethanol 2-Phenyl Ethanol
Octanoic Acid Octanoic Acid Octanoic Acid
Decanoic Acid Decanoic Acid Decanoic Acid
 D. Rossouw, F.F. Bauer / Food Microbiology 55 (2016) 32e46
isolates of specifically H. uvarum reported here confirm the pre-
liminary findings of an earlier study which showed significant
differences in sugar utilisation and ethanol yields by three different
H. uvarum strains evaluated (Contreras et al., 2014).
Our findings confirm the significant phenotypic variation in the
yeast species investigated here. In comparison, strains of
S. cerevisiae have similar fermentation phenotypes, as the produc-
tion of fermentation end-products such as glycerol and ethanol fall
within a very narrow range for this species (Piskur et al., 2006).
Non-Saccharomyces yeasts have likely not been subjected to the
same degree of “microbial domestication” by human-made envi-
ronments and activity (Camarasa et al., 2011). Indeed, the narrow
phenotypic range with regards to ethanol production by strains of
the industrial wine yeast S. cerevisiae reflects the significant se-
lection pressures that are associated with relatively recent (in
evolutionary terms) human-made environments. These selection
pressures have been proposed as causative factors leading to a
rapid evolutionary burst, and many recent genotypic and pheno-
typic adaptations in S. cerevisiae. Since species and strains of non-
Saccharomyces yeast have not been subjected to similar selection
pressures as S. cerevisiae, this would suggest significant potential
for the application of directed laboratory evolution to further
improve individual non-Saccharomyces strains for application in
wine.
Acknowledgements
Funding for this work was provided by Winetech, THRIP, the
National Research Foundation, (NRF; South Africa) through Grant
SARChI UID 83471 and the RCA grant. We appreciate the work of
Jeanne Brand, technical officer in sensory science (Department of
Viticulture and Oenology, Stellenbosch University) for the sensory
analyses, and the Central Analytical Facility (Stellenbosch Univer-
sity) for the chemical analyses.
Appendix A. Supplementary data
Supplementary data related to this article can be found at http://
dx.doi.org/10.1016/j.fm.2015.11.017.
References
Bely, L., Sablayrolles, J., Barre, P., 1990. Description of alcoholic fermentation ki-
netics: its variability and significance. Am. J. Enol. Vitic. 40, 319e324.
Ben-Dor, A., Shamir, R., Yakhini, Z., 1999. Clustering gene expression patterns.
J. Comp. Biol. 6, 281e297.
Bisson, L.F., Kunkee, R.E., 1991. Microbial interactions during wine production. In:
Zeikus, J.G., Johnson, E.A. (Eds.), Mixed Cultures in Biotechnology. McGraw-Hill,
Inc., New York, pp. 39e68.
Bokulich, N.A., Thorngate, J.H., Richardson, P.M., Mills, D.A., 2013. Microbial bioge-
ography of wine grapes is conditioned by cultivar, vintage and climate. PNAS
111, E139eE148.
Boulton, R.B., Singleton, V.L., Bisson, L.F., Kunkee, R.E., 1996. Principles and Practices
of Winemaking. Chapman & Hall, New York.
Camarasa, C., Sanchez, I., Brial, P., Bigey, F., Dequin, S., 2011. Phenotypic landscape of
Saccharomyces cerevisiae during wine fermentation: evidence for origin-
dependent metabolic traits. PLoS One 6, e25147.
Carrau, F.M., Medina, K., Biodo, E., Farina, L., Gaggero, C., Dellacassa, E., Versini, G.,
Henschke, P.A., 2005. De novo synthesis of monoterpenes by Saccharomyces
cerevisiae wine yeasts. FEMS Microbiol. Lett. 243, 107e115.
vre, M., Abdi, H., Valentin, D., 2010. Sort and beer: everything you
Chollet, S., Lelie
wanted to know about the sorting task but were afraid to ask. Food Qual. Prefer.
22, 507e520.
Ciani, M., Ferraro, L., 1996. Enhanced glycerol content in wines made with immo-
bilized Candida stellata cells. Appl. Environ. Microbiol. 62, 128e132.
Ciani, M., Ferraro, L., 1998. Combined use of immobilized Candida stellata cells and
Saccharomyces cerevisiae to improve the quality of wines. J. Appl. Microbiol. 85,
247e254.
Ciani, M., Maccarelli, F., 1998. Oenological properties of non-Saccharomyces yeasts
associated with wine-making. World J. Microbiol. Biotechnol. 14, 199e203.
Ciani, M., Picciotti, G., 1995. The growth kinetics and fermentation behaviour of
some non-Saccharomyces yeasts associated with winemaking. Biotechnol. Lett.
45
17, 1247e1250.
Ciani, M., Beco, L., Comitini, F., 2006. Fermentation behaviour and metabolic in-
teractions of multistarter wine yeast fermentations. Int. J. Food Microbiol. 108,
239e245.
Clemente-Jimenez, J.F., Mingorance-Cazorla, L., Martínez-Rodríguez, S., Las Heras-
Vazquez, F.J., Rodríguez-Vico, F., 2004. Molecular characterization and oeno-
logical properties of wine yeasts isolated during spontaneous fermentation of
six varieties of grape must. Food Microbiol. 21, 149e155.
Comitini, F., Gobbi, M., Domizio, P., Romani, C., Lencioni, L., Mannazzu, I., Ciani, M.,
2011. Selected non-Saccharomyces wine yeasts in controlled multistarter fer-
mentations with Saccharomyces cerevisiae. Food Microbiol. 5, 873e882.
Contreras, A., Hidalgo, C., Henschke, P.A., Chambers, P.J., Curtin, C., Varela, C., 2014.
Evaluation of non-Saccharomyces yeast for the reduction of alcohol content in
wine. Appl. Environ. Microbiol. 80, 1670e1678.
Cordero-Otero, R., Ubeda, J.F., Briones-Perez, A.I., Potgieter, N., Villena, M.A.,
Pretorius, I.S., Van Rensburg, P., 2003. Characterization of the b-glucosidase
activity produced by enological strains of non-Saccharomyces yeast. J. Food Sci.
68, 2564e2569.
Esteve-Zarzoso, B., Belloch, C., Uruburul, F., Querol, A., 1999. Identification of yeasts
by RFLP analysis of the 5.8S rRNA gene and the two ribosomal internal tran-
scribed spacers. Int. J. Syst. Bacteriol. 49, 329e337.
Eyeghe -Bickong, H.A., Alexandersson, E.O., Gouws, L.M., Young, P.R., Vivier, M.A.,
2012. Optimisation of an HPLC method for the simultaneous quantification of
the major sugars and organic acids in grapevine berries. J. Chromatogr. B 885,
43e49.
Ferna ndez-Gonza les, M., Di Stefano, R., Briones, A., 2003. Hydrolysis of terpene
glycosides from muscat must by different yeast species. Food Microbiol. 20,
36e41.
Fleet, G.H., Lafon-Lafourcade, S., Ribe reau-Gayon, P., 1984. Evolution of yeasts and
lactic acid bacteria during fermentation and storage of Bordeaux Wines. Appl.
Environ. Microbiol. 48, 1034e1038.
Garcia, A., Carcel, C., Dulau, L., Samson, A., Aguera, E., Agosin, E., Gunata, Z., 2002.
Influence of a mixed culture with Debaryomyces vanriji and Saccharomyces
cerevisiae on the volatiles of a Muscat wine. J. Food Sci. 67, 1138e1143.
Guth, H., Sies, A., 2002. Flavour of wines: towards an understanding by reconsti-
tution experiments and an analysis of ethanol's effect on odour activity of key
compounds. In: Blair, R.J., et al. (Eds.), Proceedings of the 11th Australian Wine
Industry Technical Conference. Australian, Wine Industry Technical Conference
Inc., Adelaide, SA, pp. 128e139.
Hansen, H.H., Nissen, P., Sommer, P., Nielsen, J.C., Arneborg, N., 2001. The effect of
oxygen on the survival of non-Saccharomyces yeasts during mixed culture fer-
mentations of grape juice with Saccharomyces cerevisiae. J. Appl. Microbiol. 91,
541e547.
Heard, G.M., Fleet, G.H., 1985. Growth of natural yeast flora during the fermentation
of inoculated wines. Appl. Environ. Microbiol. 50, 727e728.
Henick-Kling, T., Edinger, W., Daniel, P., Monk, P., 1998. Selective effects of sulfur
dioxide and yeast starter culture addition on indigenous yeast populations and
sensory characteristics of wine. J. Appl. Microbiol. 84, 865e876.
Hernandez-Orte, P., Cersosimo, M., Loscos, N., Cacho, J., Garcia-Moruno, E.,
Ferreira, V., 2008. The development of varietal aroma from non-floral grapes by
yeasts of different genera. Food Chem. 107, 1064e1077.
Heux, S., Sablayrolles, J., Cachon, R., Dequin, S., 2006. Engineering a Saccharomyces
cerevisiae wine yeast that exhibits reduced ethanol production during
fermentation under controlled microoxygenation conditions. Appl. Environ.
Microbiol. 72, 5822e5828.
Howley, M., Young, N., 1992. Low-alcohol wines: the consumer's choice? IJWM 4,
45e56.
Jiranek, V., Langridge, P., Henschke, P.A., 1995. Amino-acid and ammonium utili-
zation by Saccharomyces-cerevisiae wine yeasts from a chemically-defined
medium. Am. J. Enol. Vitic. 46, 75e83.
Jolly, N.P., Varela, C., Pretorius, I.S., 2013. Not your ordinary yeast: non-Saccharo-
myces yeasts in wine production uncovered. FEMS Yeast Res. http://dx.doi.org/
10.1111/1567-1364.12111.
Lee, T., Taylor, S., 1990. PCR protocols: a guide to methods and applications. In:
Innis, M.A., Gelfand, D.H., Sninsky, J.J., White, T.J. (Eds.), Amplification and Direct
Sequencing of Fungal Ribosomal RNA Genes for Phylogenetics. Academic Press,
San Diego, pp. 315e322.
Magyar, I., Toth, T., 2011. Comparative evaluation of some oenological properties in
wine strains of Candida stellata, Candida zemplinina, Saccharomyces uvarum and
Saccharomyces cerevisiae. Food Microbiol. 28, 94e100.
Morales, P., Rojas, V., Quiro s, M., Gonzales, R., 2015. The impact of oxygen on the
final alcohol content of wine fermented by a mixed starter culture. Appl.
Microbiol. Biotechnol. 99, 3993e4003.
Pickering, G.J., 2000. Low-and reduced-alcohol wine: a review. J. Wine Res. 11,
129e144.
Piskur, J., Rozpedowska, E., Polakova, S., Merico, A., Compagno, C., 2006. How did
Saccharomyces evolve to become a good brewer? Trends Genet. 22, 183e186.
Pretorius, I.S., 2000. Tailoring wine yeast for the new millennium: novel approaches
to the ancient art of winemaking. Yeast 16, 675e729.
Romano, P., Suzzi, G., Comi, G., Zironi, R., 1992. Higher alcohol and acetic acid
production by apiculate wine yeasts. J. Appl. Bacteriol. 73, 126e130.
Romano, P., Fiore, C., Paraggio, M., Caruso, M., Capece, A., 2003. Function of yeast
species and strains in wine flavour. Int. J. Food Microbiol. 86, 169e180.
Rossouw, D., Naes, T., Bauer, F.F., 2008. Linking gene regulation and the exo-
metabolome: a comparative transcriptomics approach to identify genes that
46 D. Rossouw, F.F. Bauer / Food Microbiology 55 (2016) 32e46
impact on the production of volatile aroma compounds in yeast. BMC Genom. 9,
530e548.
Sadoudi, M., Tourdot-Marechal, R., Rousseaux, S., Steyer, D., Gallardo-Chacon, J.J.,
Ballester, J., Vichi, S., Guerin-Schneider, R., Caixach, J., Alexandre, H., 2012. Yeast-
yeast interactions revealed by aromatic profile analysis of Sauvignon Blanc wine
fermented by single or coculture of non-Saccharomyces and Saccharomyces
yeasts. Food Microbiol. 32, 243e253.
Setati, M.E., Jacobson, D., Andong, U.-C., Bauer, F.F., 2012. The vineyard yeast
microbiome, a mixed model microbial map. PLoS One 7, e52609.
Soden, A., Francis, I.L., Oakey, H., Henschke, P.A., 2000. Effects of co-fermentation
with Candida stellata and Saccharomyces cerevisiae on the aroma and compo-
sition of Chardonnay wine. Aust. J. Grape Wine Res. 6, 21e30.
Van Zyl, J.A., De Vries, M.J., Zeeman, A.S., 1963. The microbiology of South African
winemaking III. The effect of different yeasts on the composition of fermented
musts. S Afr. J. Agric. Sci. 6, 165e179.
Varela, C., Siebert, T., Cozzolino, D., Rose, L., McLean, H., Henschke, P.A., 2009.
Discovering a chemical basis for differentiating wines made by fermentation
with ‘wild’ indigenous and inoculated yeasts: role of yeast volatile compounds.
Aust. J. Grape Wine Res. 15, 238e148.
Varela, C., Kutyna, D., Solomon, M.R., Black, C., Borneman, A., Henschke, P.,
Pretorius, I.S., Chambers, P.J., 2012. Evaluation of gene modification strategies
for the development of low-alcohol-wine yeasts. Appl. Environ. Microbiol. 78,
6068e6077.
s, S., Valle
Viana, F., Gil, J.V., Genove s, S., Manzanares, P., 2008. Rational selection of
non-Saccharomyces wine yeasts for mixed starters based on ester formation and
enological traits. Food Microbiol. 25, 778e785.