Food Microbiology 46 (2015) 66e73

Contents lists available at ScienceDirect

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

Selection of yeasts with multifunctional features for application

as starters in natural black table olive processing
 pez b, *
 mez b, E.Z. Panagou a, F.N. Arroyo-Lo
S. Bonatsou a, A. Bentez b, F. Rodrguez-Go
a
Laboratory of Microbiology and Biotechnology of Foods, Department of Food Science and Human Nutrition, Agricultural University of Athens, Iera Odos 75,
Athens GR-11855, Greece
b
Food Biotechnology Department, Instituto de la Grasa (CSIC), Avda Padre Garca Tejero n
4, 41012 Seville, Spain

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

Article history: Yeasts are unicellular eukaryotic microorganisms with a great importance in the elaboration on many
Received 10 February 2014 foods and beverages. In the last years, researches have focused their attention to determine the
Received in revised form favourable effects that these microorganisms could provide to table olive processing. In this context, the
20 June 2014
present study assesses, at laboratory scale, the potential technological (resistance to salt, lipase, esterase
Accepted 16 July 2014
Available online 24 July 2014
and b-glucosidase activities) and probiotic (phytase activity, survival to gastric and pancreatic digestions)
features of 12 yeast strains originally isolated from Greek natural black table olive fermentations. The
multivariate classication analysis carried out with all information obtained (a total of 336 quantitative
Keywords:
Enzymatic spectrophotometric assays
input data), revealed that the most promising strains (clearly discriminated from the rest of isolates)
Growth modelling were Pichia guilliermondii Y16 (which showed overall the highest resistance to salt and simulated di-
Multifunctional starters gestions) and Wickerhamomyces anomalus Y18 (with the overall highest technological enzymatic activ-
Principal component analysis ities), while the rest of strains were grouped together in two clearly differentiated clusters. Thus, this
Table olives work opens the possibility for the evaluation of these two selected yeasts as multifunctional starters,
Yeasts alone or in combination with lactic acid bacteria, in real table olive fermentations.
2014 Elsevier Ltd. All rights reserved.

1. Introduction
Table olives are one of the most important and well known
fermented vegetables of the food industry, with an estimated
worldwide production that currently exceeds 2.5 million tons/year
(IOC, 2013). The elaboration of this food is closely related to the
culture and diet of many Mediterranean countries, with Spain,
Turkey, Egypt, Greece and Italy as the main producers. There is also
important production in the USA, Peru, Argentina and Australia.
Thus, table olive processing is widespread around the world and
represents an important economic source for producing countries.
The most usual table olive elaborations are: a) alkali-treated green
olives (the so-called Spanish style), b) ripe olives by alkaline
oxidation (Californian style), and c) naturally black olives (also
known as Greek style) (Garrido-Ferna ndez et al., 1997).
Microorganisms, especially lactic acid bacteria and yeasts, play
an important role during fermentation of the olive fruit
* Corresponding author. Tel.: 34 954692516x115; fax: 34 954691262.
pez).
E-mail address: fnarroyo@cica.es (F.N. Arroyo-Lo
determining the safety, quality and avour of the nal product
(Garrido-Fern andez et al., 1997). For many years, the search for
starters with application in olive fermentation has been practically
focused on the activity of lactic acid bacteria. However, in the last
years, diverse publications have emphasised the importance of
selection of yeasts as starter cultures during table olive processing
(Silva et al., 2011; Arroyo-Lo  pez et al., 2012; Rodrguez-Go mez
et al., 2012; Bevilacqua et al., 2012, 2013; Tofalo et al., 2013).
Although these microorganisms can cause spoilage of the product
due to the production of CO2, bad odours and avours, clouding of
brines or softening of fruits, which is especially harmful in olive
packaging or storage, they also show some desirable activities with
important potential technological and probiotic applications. For
this reason, yeasts may be exploited in table olive processing as
multifunctional starters.
In this context, several authors have studied, among others, the
lipolytic (lipase and esterase) and b-glucosidase activities (tech-
nological characteristics) of different yeast species related to table
olives or oleic ecosystems (Psani and Kotzekidou, 2006; Herna ndez
et al., 2007; Bevilacqua et al., 2009, 2012; Aponte et al., 2010;
Romo-S anchez et al., 2010; Bautista-Gallego et al., 2011;

http://dx.doi.org/10.1016/j.fm.2014.07.011
0740-0020/ 2014 Elsevier Ltd. All rights reserved.
 S. Bonatsou et al. / Food Microbiology 46 (2015) 66e73
Rodrguez-Go  mez et al., 2012). It has also been widely proven that
many yeast species have excellent aromatic proles, can improve
lactic acid bacteria growth and inhibit undesirable microorganisms
(Psani and Kotzekidou, 2006; Querol and Fleet, 2006; Arroyo-Lo pez
et al., 2012). Recently, the probiotic potential of table olive related
yeasts has also begun to be evaluated (Psani and Kotzekidou, 2006;
Silva et al., 2011). However, most of the available information until
now comes only from qualitative tests carried out on agar-assays,
which do not facilitate the comparison and selection of the most
appropriate yeast starters. On the contrary, spectrophotometric
assays have been developed to quantitatively determine the yeast
activity levels for several groups of enzymes (lipases, esterases, b-
glucosidades and phytases, among others) (Rosi et al., 1994; Haros
et al., 2005; Ciafardini et al., 2006; Rodrguez-Go  mez et al., 2012) or
to determine the resistance of microorganism to salt by mathe-
matical modelling (Romero-Gil et al., 2013). These methods are
easily applicable and inexpensive; they do not need complex in-
struments and offer objective information on yeast reponse for
comparison.
Another additional problem is the application of an appropriate
methodology for the management of such a large amount of data,
necessary when researchers have to analyse several physiological
traits (variables) from a considerable number of strains (cases).
Multivariate analysis techniques offer an interesting approach to
solve this drawback. Principal component analysis (PCA) is a
multivariate statistical technique used for data compression, to
capture the main features in the multivariate data sets and to
extract information from them. PCA implies a mathematical pro-
cedure that transforms the overall set of original variables into a
smaller number of mathematical constructs, also called factors.
Thereby, a new set of axes, called factor axes, is obtained in a lower
dimensional space onto which the original space of variables and
cases can be projected and classied into categories. Such factors
can be viewed as linear combinations of the original variables that
are uncorrelated (Jackson, 1991). The new variables thus generated
account for the inherent variation of the data to the maximum
possible extent.
In this work, diverse methodologies (growth modelling, spec-
trophotometric enzymatic assays and in vitro probiotic tests) have
been applied to assess the potential technological (resistance to
salt, lipase, esterase and b-glucosidase activities) and probiotic
(phytase activity and resistance to gastric, pancreatic and overall
digestions) applications of a considerable number of yeast species
never before studied in Greek natural black table olive processing.
One way ANOVA and multivariate analysis based on PCA were then
applied to this quantitative set of data to select the yeast strains
with the most promising global characteristics for their use as
multifunctional starters.
2. Material and methods
2.1. Yeast strains
A total of 12 yeast strains (Pichia guilliermondii Y16, Pichia
kluyveri Y17, Wickerhamomyces anomalus Y18, Candida silvae Y19,
Metschnikowia pulcherrima Y20, Saccharomyces cerevisiae Y22,
Pichia manshurica Y37, Rhodotorula mucilaginosa Y38, Rhodotorula
diobovatum Y39, Aereobasidium pullulans Y40, Debaryomyces han-
senii Y57 and Pichia membranifaciens Y67), representative of the
yeast microbiota usually found in Greek natural black table olives,
were used in the present study. All of them were previously isolated
by Nisiotou et al. (2010) from distinct phases (early, middle and
nal stages) of the spontaneous fermentation of naturally black cv.
Conservolea olives, and were identied at the species level by PCR-
restriction fragment length polymorphism (RFLP) of the 5.8S
67
internal transcribed spacer (ITS) region combined with sequence
analysis of the D1/D2 domain of 26S gene. The previous mentioned
work was focused only on the identication of the succession of
yeasts species during the fermentation of natural black olives, and
for this reason, only one strain per species was randomly selected
and deposited in the microbiological collection of the Laboratory of
Microbiology and Biotechnology of Foods of the Agricultural Uni-
versity of Athens (Greece). Their sequences were also deposited in
the NCBI GenBank data library under accession numbers FJ715430,
FJ649188 to FJ649197 and GQ140297 to GQ140301.
2.2. Modelling the resistance of yeasts to salt
Yeast inocula were prepared by inoculating one single colony of
each isolate into 5 mL of YM broth medium (Difco, Becton and
Dickinson Company, Sparks, USA). After 48 h of incubation at 30
C,
1 mL from each tube was centrifuged at 9000  g for 10 min, the
pellets were washed with sterile saline solution (9 g/L), centrifuged
and re-suspended again in 0.5 mL of a sterile saline solution to
obtain a nal concentration of about 7.2 log10 CFU/mL, which was
conrmed by surface spread on YM agar. These yeast suspensions
were used to inoculate the different modelling experiments.
Growth was monitored in a Bioscreen C automated spectro-
photometer (Labsystem, Helsinki, Finland) with a wideband lter
(420e580 nm). Measurements were taken every 2 h after a pre-
shaking of 5 s. The wells of the microplate were lled with
20 mL of inoculum and 330 mL of medium (according to treatment
as described below), always reaching an initial OD of approxi-
mately 0.2 (inoculum level above 6 log10 CFU/mL). The inocula
were always above the detection limit of the apparatus, which was
determined by comparison with a previously established calibra-
tion curve. Uninoculated wells for each experimental series were
also included in the microplate to determine, and consequently
subtract, the noise signal. All experiments were carried out in
triplicate.
Sterilised YM broth was supplemented with NaCl to obtain the
following nal concentrations of salt in the media: 0, 5, 10, 20, 30,
50, 70, 90, 110, 140, 180, 200 and 250 g/L. Thus, a total of 468 growth
curves (13 levels NaCl  12 yeasts  triplicate) were obtained. The
use of a well known, standardised synthetic laboratory medium to
carry out the experiments was preferred because, in the olive
matrix, the presence of diverse components released by fruits such
as polyphenols, organic acids, etc., may mask the real inhibitory
effect of NaCl.
The basis of the technique used for estimating the NIC (non-
inhibitory concentration) and MIC (minimum-inhibitory concen-
tration) values of the assayed yeast strains for NaCl was the com-
parison of the area under the OD/time curve of a positive control
(absence of salt, optimal conditions) with the areas of the tests
(presence of salt, increasing inhibitory conditions). As the amount
of inhibitor in the well increases, the effect on the growth of the
organism also increases. This effect on growth is manifested by a
reduction in the area under the OD/time curve relative to the
positive control at any specied time. The areas under the OD/time
curves were calculated by integration using OriginPro 7.5 software
(OriginLab Corporation, Northampton, USA). The relative amount of
growth for each NaCl concentration, denoted as the fractional area
(fa), was obtained using the ratios of the test area (areatest) to that of
the positive control of the yeast (areacont), according to the
following formula:
fa areatest =areacont
The plot of the fa versus the natural logarithm (ln) of the NaCl
concentration produced a sigmoid-shape curve that could be well-
68 S. Bonatsou et al. / Food Microbiology 46 (2015) 66e73
tted with a reparameterized modied Gompertz function for
decay, which had the following expression:
 agitated daily for 1 min with a vortex. Then, cells were separated by
y exp  x=lnMIC=exp  lnlnNIC=lnMIC= centrifugation at 9000  g for 15 min. The supernatant was steri-
 unit with 0.2 mm of porosity (Millipore Co., Bedford, Massachusetts,
2:718282:71828=lnlnNIC=lnMIC USA) and used as the extracellular fraction. At the same time, the
where y is the dependent variable (fa), x is the independent variable
(ln NaCl g/L), MIC is the minimum NaCl concentration (g/L) above
which growth is not observed, and NIC is the NaCl concentration (g/
L) above which an inhibitory effect begin to be observed. These
parameters were obtained by a non-linear regression procedure,
minimising the sum of squares of the difference between the
experimental data and the tted model, i.e., loss function
(observed  predicted)2. This task was accomplished using the
non-linear module of the Statistica 7.1 software package (StatSoft
Inc, Tulsa, OK, USA) and its Quasi-Newton option. Fit adequacy was
checked by the proportion of variance explained by the model (R2)
with respect to the experimental data.
2.3. In vitro gastric and pancreatic digestions
Simulated gastric digestion (GD) was performed using the
protocol initially described by Corcoran et al. (2007) with slight
modications. Briey, synthetic gastric juice was prepared in a
buffer solution at pH 2.0 containing NaCl (2.05 g/L), KH2PO4 (0.60 g/
L), CaCl2 (0.11 g/L) and KCl (0.37 g/L). It was adjusted to pH 2.0 with
1 M HCl and autoclaved at 121
C for 15 min. Prior to use, pepsin
(0.0133 g/L) and lysozyme (0.01 g/L) were added. All the compo-
nents were obtained from SigmaeAldrich (St. Louis, USA). The yeast
cultures to be tested were grown to early stationary phase,
centrifuged (10,000  g, 10 min) and the pellet was washed with
the buffer (pH 2.0) mentioned above. Then, the cells were re-
suspended in the synthetic gastric juice to a nal concentration
around 7 log10 CFU/mL and incubated for 2.5 h at 37
C in an orbital
shaker (~200 rpm) to simulate peristaltic movements. Finally, serial
dilutions of the cultures were plated onto YM agar medium and
counted after incubation at 30
C for 2 days.
Simulated pancreatic digestion (PD) was formulated using bile
(3.0 g/L, Oxoid LTD, Basingstoke, England) and pancreatin (0.1 g/L,
SigmaeAldrich) in a buffer at pH 8.0 (adjusted with 1 M HCl) con-
sisting of 50.81 g/L of sodium phosphate dibasic heptahydrate and
8.5 g/L of NaCl. Harvested cells from the previous GD step were
washed in saline and re-suspended in the same volume of the
simulated pancreatic juice. A colony count on YM agar plates at
initial time, to know the initial counts in the freshly prepared
simulated pancreatic juice, was performed by serial dilutions of the
cultures. After shaking at 200 rpm in an orbital shaker during 3.5 h at
37
C, the pellet was washed and then re-suspended in a volume of
isotonic solution in order to avoid carryover of the buffers to the
agar. Serial dilutions were done, and subsequently plated onto YM
agar. A reference probiotic strain (Lactobacillus rhamnosus GG), was
used for comparison of the survival rates after simulated GD and PD,
but using MRS (de Man, Rogosa and Sharpe) (Oxoid) as culture
medium. The use of a bacterium was preferred instead a yeast
because they are the most common probiotic microorganisms found
in the literature. Overall survival was obtained by comparison of the
initial yeast counts (CFU/mL) at the beginning of the simulated GD
and those cells recovered at the end of the simulated PD.
2.4. Assessment of the enzymatic activities
Before enzymatic tests, the 12 yeast cultures were previously
refreshed on YM agar. After 48 h of incubation at 30
C, colonies
were picked and directly used to inoculate 5 mL of new fresh YM
broth. These tubes were then incubated at 30
C for 2 days and
lised through microltration using a sterile syringe-driven lter
cell pellets were washed twice in a sterile 50 mM phosphate buffer
(pH 7.0) to remove liquid culture media for lipase and esterase, in a
100 mM citrate-phosphate buffer (pH 5.0) for b-glucosidase, or in a
buffer sodium acetate-acetic acid 0.1 M (pH 5.5) for phytase de-
terminations. Finally, cells were re-suspended in a suitable volume
of the corresponding sterile buffer solution to reach a population
level between 7 and 8 log10 CFU/mL, according to yeast species.
Both supernatant and whole cells were used for the enzymatic tests
described below.
b-glucosidase, esterase and lipase activities were quantitatively
determined in duplicate following procedures already mentioned
by several authors (Rosi et al., 1994; Ciafardini et al., 2006; Costas
et al., 2007; Rodrguez-Go  mez et al., 2010, 2012) with slight mod-
ications. Briey, these activities were evaluated by measuring the
amount of p-nitrophenol liberated from the following chromogenic
substrates: p-nitrophenyl stearate (SigmaeAldrich) and 4-
nitrophenyl palmitate (Fluka Chemical, Buchs, Switzerland) for
lipase determination, p-nitrophenyl butyrate (SigmaeAldrich) for
esterase, and p-nitrophenyl-b-D-glucoside (SigmaeAldrich) for b-
glucosidase assays.
For lipase and esterase determinations, the corresponding
chromogenic substrates were dissolved in undecane to reach a nal
concentration of 1 mM. The reactions were carried out in a biphasic
system formed by 1 mL of the substrate undecane mix (organic
phase) and 1 mL of the extracellular or cellular fractions (aqueous
phase). For b-glucosidase evaluation, 0.2 mL of the cellular or su-
pernatant fraction was mixed with another 0.2 mL of a solution
formed by the corresponding chromogenic substrate dissolved in
100 mM of a citrate-phosphate buffer to reach a nal concentration
of 5 mM. After 24 h of incubation at 30
C for lipase and esterase, or
only 1 h at 40
C for b-glucosidase, 0.2 M carbonate buffer (pH 10.2)
was added to stop the reaction. The concentration of liberated p-
nitrophenol was estimated from the absorbance obtained at 410
(lipase and esterase) or 400 nm (b-glucosidase) in a spectropho-
tometer model Cary1E UVevis (Varian INC., Palo Alto, CA, USA)
using a suitable blank for each case. Results were expressed as the
amount of enzyme liberating 1 nmol of p-nitrophenol per hour and
milliliter (nmol h1 mL1) under the assay conditions for both
extracellular and cellular fractions.
Phytase activity was determined by measuring the amount of
liberated inorganic phosphate from sodium phytate using the
protocol described by Haros et al. (2005), but with slight modi-
cations. The reaction mixture consisted of 0.5 mL of 0.1 M sodium
acetace-acetic acid (pH 5.5), containing 6 mM sodium phytate and
0.5 mL of sample solution. After incubation at 50
C for 60 min, the
reaction was stopped by adding 0.25 mL of 20% (w/v) trichloro-
acetic acid (TCA) solution. The reaction was cooled in ice for 15 min
and then centrifuged at 9000  g for 10 min. Then, 0.9 mL of the
aqueous phase was mixed with 0.96 mL of phosphorus reagent
(Mo-Va:5 N H2SO4:acetone in 1:1:2 proportion). After 10 min of
incubation at room temperature, absorbance was measured at
405 nm. One unit of phytase activity was dened as the amount of
enzyme liberating 1 nmol of inorganic phosphorous per hour and
milliliter (nmol h1 mL1) under the assay conditions for both
extracellular and cellular fractions.
Finally, commercial lipase (from porcine pancreas), esterase
(from porcine liver), b-glucosidase (from almonds) and phytase
(from wheat) enzymes were purchased from SigmaeAldrich
 S. Bonatsou et al. / Food Microbiology 46 (2015) 66e73 69

(references L3126, E3019, G0395 and P1259, respectively), and used
in the present study at a nal concentration of 5 mg/L as internal
control for comparison with the yeast activities.
2.5. Statistical data analysis
An analysis of variance was performed by means of the one-way
ANOVA module of Statistica 7.1 software (Statsoft Inc., Tulsa, USA)
to check for signicant differences among yeast strains for the
different technological and probiotic tests. For this purpose, a post-
 obtained from three independent experiments. The NIC value,
hoc comparison statistical test was applied by means of the Scheffe
test, which is considered to be one of the most conservative post-
hoc tests (Winer, 1962).
A total of 336 quantitative input data were then subjected to
multivariate analysis. PCA was applied to discriminate yeast strains
with highly desirable properties, using a varimax rotation. For the
selection of the number of Principal Components or Factors, the
Kaiser criterion (Jolliffe, 1986) was followed and only factors with
eigen-values higher than 1.00 were retained. Cases introduced in
the analysis were the 12 assayed yeast strains, while explanatory
variables were: NIC and MIC values, survival to gastric, pancreatic
and overall digestions (GS, PS and OS, respectively), and cellular
lipase (L-C), extracellular lipase (L-E), cellular esterase (E-C),
extracellular esterase (E-E), cellular b-glucosidase (B-C), extracel-
lular b-glucosidase (B-E), cellular phytase (P-C) and extracellular
phytase (P-E) activities. Statistica software version 7.1 was used for
data processing and graphic representation.
3. Results
3.1. Susceptibility and resistance of the assayed yeasts to salt
In this survey, a total of 468 OD curves were obtained in syn-
thetic laboratory media to estimate the effect of NaCl concentration
on the growth of 12 table olive related yeasts isolated from the
fermentation of Greek natural black table olives. Data were tted
using the reparameterized Gompertz equation for decay, with an R2
in many cases above 0.95 (data not shown). As an example, Fig. 1
shows the t for Y40 yeast (A. pullulans), which followed a typical
sigmoid decay function. Consequently, yeast response to NaCl was
non-linearly dose related. The whole sigmoid-shaped curve for this
microorganism could be divided into three sections: i) points cor-
responding to concentrations from zero up to the NIC (concentra-
tions at which no effect of the inhibitor was observed and fa was
around 1), ii) concentrations between NIC and MIC values (within
which growth inhibition progressively occurred and the fa
decreased), and iii) a third section above MIC (where no growth
relative to the control was recorded and fa was around 0).
Table 1 shows the NIC and MIC values obtained for all the yeasts
assayed by this methodology. For each species, average values were
related to salt susceptibility, was widespread among yeast strains
and ranged from 20.2 (P. membranifaciens Y67) to 109.9 g/L
(P. guilliermondii Y16), while the MIC value, related to salt resis-
tance, ranged from 123.7 (R. mucilaginosa Y39) to 274.4 g/L
(S. cerevisiae Y22). Thereby, according to values shown in Table 1,
yeasts Y16 and Y22 were the most resistant microorganisms to salt
in laboratory medium, with no statistically signicant differences
between them according to a Scheffe  post-hoc comparison test,
while yeast Y40 was the less resistant.
3.2. Resistance of the assayed yeasts to in vitro digestions
Table 2 shows the survival (%) of all assayed yeasts to the
simulated in vitro gastric and pancreatic digestions, as well as for
the probiotic bacterium (L. rhamnosus GG) used as control. All yeast
isolates showed a survival percentage to the gastric digestion
higher than the reference probiotic microorganism, which had a
survival of only 0.04%. Two strains, P. guilliermondii Y16 and
D. hansenii Y57, showed values especially high (492.0% and 90.4%,
respectively), indicative that both microorganisms were not
affected by the stressed conditions found in the stomach. On the
contrary, C. silvae Y19, R. mucilaginosa Y38 and A. pullulans Y40
were the yeasts less resistant to the gastric test, with values close to
30%.
After GD, all yeasts were sequentially subjected to the simulated
PD. In general, a lower survival frequency in the PD than in the GD
was noticed (Table 2). Especially, two strains were considerably
affected by this test (A. pullulans Y40 and D. hansenii Y57), which
exhibited a high mortality (survival close to 0%). On the contrary,
P. guilliermondii Y16, C. silvae Y19 and S. cerevisiae Y22 had values
higher than 60%, although only Y22 was statistically different to the
probiotic control (14.4% for LGG).
The overall yeast survival to the simulated digestive process,
comparing the initial CFU/mL with the nal population obtained

Table 1
NIC (non-inhibitory concentration) and MIC (minimum inhibitory concentration)
values obtained for salt (NaCl) with the doseeresponse model.

Yeast reference NIC (g/L) MIC (g/L)

Y16 109.93 (4.93) d 261.60 (1.71) d

Y17 84.69 (5.00) b,d 209.66 (11.04) c
Y18 47.27 (0.94) a,c 163.18 (10.88) a,b,c
Y19 45.82 (2.86) a,c 134.66 (10.26) a,b
Y20 61.08 (5.22) a,b 176.65 (10.26) b,c
Y22 86.30 (19.10) b,d 274.44 (8.10) d
Y37 71.16 (19.18) a,b 152.13 (12.06) a,b
Y38 58.26 (6.01) a,b 266.02 (29.01) d
Y39 49.97 (11.26) a,b,c 123.70 (11.43) a
Y40 44.47 (2.97) a,c 127.09 (3.84) a,b
Y57 50.22 (0.94) a,b,c 153.59 (11.52) a,b
Y67 20.15 (0.98) c 154.41 (8.53) a,b
Fig. 1. Fit of the fractional areas of Aureobasidium pullulans Y40 as a function of ln NaCl
(g/L) using the reparameterized Gompertz equation for decay (see Materials and Note: Standard deviation in parentheses. Values followed by different online letters,
methods) for the estimation of NIC (non-inhibitory concentration) and MIC (mini- within the same column, are signicantly different according to a Scheffe post-hoc
mum inhibitory concentration). comparison test.
70 S. Bonatsou et al. / Food Microbiology 46 (2015) 66e73

Table 2 Regarding phytase activity, the strain with the highest values for
Survival (%) of the assayed yeast strains to the in vitro simulated gastric, pancreatic this test was M. pulcherrima Y20 in both cellular and extracellular
and overall digestions. Data used for analysis were directly CFU/mL.
fractions, while many yeasts (Y18, Y37, Y38, Y39) did not show any
Yeast reference Gastric digestion Pancreatic digestion Overall levels of activity. The great potential of Y20 to produce phytase
Y16 492.05 (319.63) a 63.80 (21.09) d,e 17.04 (4.35) a,b,c enzymes was evidenced when it was compared the commercial
Y17 87.77 (17.01) a 40.84 (5.38) a,b,c,d,e 26.71 (9.26) c enzyme versus the cellular fraction (796 vs 1914 nmol ml1 h1,
Y18 40.48 (8.27) a 13.39 (4.38) a,b,c 3.71 (2.29) a,b respectively). Other isolate with high values in both cellular and
Y19 33.25 (0.26) a 60.22 (2.76) c,d,e 18.68 (0.08) a,b,c
extracellular fraction was C. silvae Y19.
Y20 75.01 (30.91) a 12.10 (11.33) c,d,e 4.43 (3.82) a,b,c
Y22 51.02 (7.60) a 64.28 (2.85) e 26.97 (2.34) c
Y37 76.96 (3.77) a 57.28 (16.97) c,d,e 23.74 (8.89) b,c 3.4. Multivariate analysis
Y38 31.33 (10.83) a 53.38 (7.07) b,c,d,e 26.34 (1.59) c
Y39 47.54 (0.10) a 3.83 (3.66) a,b 0.09 (0.06) a
Y40 35.55 (5.28) a 0.00 (0.00) a 0.00 (a)
In addition to specic tests, a study which simultaneously
Y57 90.37 (109.87) a 0.08 (0.06) a 0.03 (a) considers all of them is of interest for selection of multifunctional
Y67 60.44 (14.62) a 11.29 (0.21) a,b,c 5.04 (0.42) a,b,c starters. For this reason, quantitative data were subjected to PCA in
LGGa 0.04 (0.02) a 14.41 (6.04) a,b,c,d 0.01 (0.00) a order to condense the information into a reduced number of Fac-
Note: Standard deviation in parentheses. Values followed by different online letters, tors. PCA led to the identication of four Factors with eigen-values
within the same column, are signicantly different according to a Scheffe post-hoc higher than 1 (data not shown), indicating that the total number of
comparison test. analysed variables (13) could be grouped into only four Factors
a
Lactobacillus rhamnosus GG was included as probiotic bacterium control.
which explained 84.5% of the total variance (37.0% for Factor 1,
22.9% for Factor 2, 16.3% for Factor 3 and 8.8% for Factor 4).
after both gastric and pancreatic digestions, is also shown in The relationship between the two major factors and the original
Table 2. Three yeast strains, P. kluyveri Y17, S. cerevisiae Y22 and variables can be deduced from Fig. 2 (upper panel). In this way,
R. mucilaginosa Y38, were especially resistant with values around Factor 1 was linearly related (positive correlation in Fig. 2) to many
26% and statistically higher than the probiotic bacterium used as of the enzymatic activities (except phytase). Thus, it was re-named
control (0.01%), while the mortality for yeasts Y40 and Y57 was as technological enzymatic activities. On the contrary, Factor 2
considerable (close to 0% survival). was mainly related (positive correlation) to resistance to digestion
(GS, PS and OS) and salt (MIC and NIC) tests, so it was re-named as
3.3. Enzymatic activities of the assayed yeasts resistance to stress. Fig. 2 (upper panel) can also be used to
establish relationships among variables. According to this graph,
Table 3 shows the average results obtained for the spectro- there was no relationship between stress variables (GS, PS, OS,
photometric enzymatic assays. Data were analysed by means of NIC and MIC) and technological enzymatic activities (L-C, L-E, B-E,
one-way ANOVA analysis to identify the yeast strains with the E-E, E-C), because they formed an angle close to 90
.
highest levels for each enzymatic activity and compared to com- The projection of the cases (yeast strains) onto the planes
mercial enzymes. In this way, strain W. anomalus Y18 was the yeast formed by the two major Factors (59.9% of the total variance) led to
with the highest statistically signicant levels (p  0.05) in four of the segregation of several isolates clearly differentiated from the
six enzymatic technological activities (L-C, L-E, E-E and B-C), with rest (Fig. 2, lower panel). The majority of yeasts were grouped close
also high values for B-E and E-C. However, this microorganism did to the intersection between both factors (one group formed by Y17,
not show any activity for the phytase tests. It can be observed that Y22, Y38 and Y37, slightly in the positive axis of Factor 2, and other
the production of lipase enzymes for the majority of yeasts was group formed by Y19, Y20, Y39, Y40, Y57 and Y67, slightly in the
practically null, except for the above mentioned microorganism negative axis of Factor 2), while two cases were clearly located
that was even above the commercial enzyme. Table 3 also shows outside bouth groups. They were P. guilliermondii Y16 and
that some strains (Y16, Y38 and Y39) did not produce b-glucosidase W. anomalus 18. The former, was characterised by its high resis-
enzymes under the assayed conditions. However, it is clear that the tance to the stress conditions (salt and digestions), while the later
production of esterases was the most widerspread technological was characterised by its considerable high technological enzymatic
property among the yeasts. activities.

Table 3
Cellular and extra-cellular enzymatic activities (nmol ml1 h1) of the assayed yeast strains compared to commercial enzymes.

Yeast reference L-C L-E E-C E-E B-C B-E P-C P-E

Y16 0.00 (0.00) a 0.00 (0.00) a 6.06 (0.93) a 1.79 (0.28) a 0.00 (0.00) a 0.00 (0.00) a 241.86 (4.34) b 0.00 (0.00) a
Y17 0.00 (0.00) a 0.00 (0.00) a 6.01 (0.23) a 1.53 (0.15) a 14.48 (4.55) a 28.75 (1.16) a,c,d 729.34 (69.00) c 0.00 (0.00) a
Y18 9.89 (4.34) b 12.89 (7.57) b 9.76 (0.32) a,b 9.61 (1.50) c 803.51 (270.54) b 67.13 (20.20) b,c,d 0.00 (0.00) a 0.00 (0.00) a
Y19 0.00 (0.00) a 0.00 (0.00) a 4.63 (0.73) a 2.33 (0.28) a,b 0.00 (0.00) a 22.85 (0.00) a,c 113.92 (60.01) a,b 136.41 (67.65) a
Y20 0.00 (0.00) a 0.00 (0.00) a 5.98 (0.82) a 1.63 (0.00) a 15.02 (0.03) a 96.11 (0.61) b 795.98 (30.44) c 146.52 (72.29) a
Y22 0.00 (0.00) a 0.00 (0.00) a 6.05 (0.34) a 1.06 (0.16) a 15.66 (1.48) a 112.82 (0.45) b 164.09 (19.81) b 74.75 (11.03) a
Y37 0.00 (0.00) a 0.00 (0.00) a 14.96 (1.13) b 1.19 (0.11) a 19.04 (1.12) a 0.00 (0.00) a 0.00 (0.00) a 0.00 (0.00) a
Y38 0.00 (0.00) a 0.00 (0.00) a 9.92 (1.94) a,b 1.25 (0.25) a 0.00 (0.00) a 0.00 (0.00) a 0.00 (0.00) a 0.00 (0.00) a
Y39 0.00 (0.00) a 0.00 (0.00) a 9.56 (0.51) a,b 2.08 (0.19) a,b 0.00 (0.00) a 0.00 (0.00) a 0.00 (0.00) a 0.00 (0.00) a
Y40 0.00 (0.00) a 0.00 (0.00) a 8.51 (2.00) a 4.39 (0.14) b 26.69 (6.41) a 113.89 (21.69) b 0.00 (0.00) a 87.75 (71.13) a
Y57 0.00 (0.00) a 0.00 (0.00) a 8.61 (0.68) a 0.97 (0.09) a 6.01 (2.15) a 12.54 (2.87) a 0.00 (0.00) a 89.66 (98.97) a
Y67 0.00 (0.00) a 0.00 (0.00) a 8.39 (1.79) a 0.99 (0.06) a 15.36 (0.63) a 70.21 (0.25) b,d 0.00 (0.00) a 25.69 (36.34) a
CEa 3.02 (0.23) 952.20 (71.70) 6020.55 (312.70) 1914.43 (12.65)
 post-hoc
Note: Standard deviation in parentheses. Values followed by different online letters, within the same column, are signicantly different according to a Scheffe
comparison test.
a
CE, values obtained for the commercial enzymes used as control, using a concentration of 5 mg/mL. See Material and method section for the reference of enzymes.
 S. Bonatsou et al. / Food Microbiology 46 (2015) 66e73 71

Fig. 2. Projection of the two major factors of the PCA as a function of variables (upper panel) and yeast strains (lower panel). GS: gastric digestion, PS: pancreatic digestion, OS:
overall digestion, MIC: minimum inhibitory concentration, NIC: non-inhibitory concentration, L-C: cellular lipase, L-E: extracellular lipase, E-C: cellular esterase, E-E: extracellular
esterase, B-C: cellular b-glucosidase, B-E: extracellular b-glucosidase, P-C: cellular phytase, P-E: extracellular phytase.

4. Discussion microorganisms play especially an important role during fermen-

The presence of yeasts during processing of table olives around
the world is very common, particularly species of the genera
Candida, Pichia, Rhodotorula, Saccharomyces and Debaryomyces
(Garrido-Ferna ndez et al., 1997; Arroyo-Lopez et al., 2012). They can
reach population levels above 6 log10 CFU/mL in this fermented
vegetable, affecting the organoleptic and nutritional characteristics
of the nal product (Garrido-Ferna ndez et al., 1997; Rodrguez-
Go mez et al., 2010; Arroyo-Lo  pez et al., 2012). These
tation of natural black olives, where high salt levels (8e10% or even
higher) are used, resulting in a process that is basically dominated
by yeasts and occasionally by lactic acid bacteria (Heperkan, 2013).
This type of table olive elaboration is very common in Greece,
especially at farmers' level. Modelling data obtained in this work
corroborate the high resistance of these microorganisms to NaCl,
with strains belonging to species P. guilliermondii, P. kluyveri,
S. cerevisiae and R. mucilaginosa as the most resistant yeasts with
MIC values above 200 g/L.
72 S. Bonatsou et al. / Food Microbiology 46 (2015) 66e73
Today, the attention on microbiology of table olive fermenta-
tions has been given almost exclusively focused on lactic acid
bacteria, which stand among the most important groups of mi-
croorganisms found in olive microbiota (Hurtado et al., 2012).
However, recent research has revealed the signicant contribution
that yeasts may have to this fermented vegetable, and how they
could improve the process and the added value of the nal product
(Arroyo Lo  pez et al., 2008, 2012; Rodrguez-Go  mez et al., 2010).
Selection of yeasts as starter cultures is a complex process including
a selection step, validation on laboratory scale and nally demon-
stration at factory level (Bevilacqua et al., 2012). Potential starters
must be assessed on the basis of technological as well as functional
traits. In this direction, the screening of an appropriate number of
strains is critical in successful starter development. It must be noted
that in this work only one strain per yeast species was subjected to
technological and functional characterisation. However, the infor-
mation provided is important for the Greek table olive industry as it
refers to the most frequently encountered yeast species in natural
black olive fermentation and could become a basis for future
research on this issue.
Esterase and lipase enzymes (globally recognised as lipolytic
activity) are desirable in yeasts because they can improve the
avour of olives through the formation of volatile compounds that
can be generated by the catabolism of free fatty acids (Herna ndez
et al., 2007; Rodrguez-Go  mez et al., 2010, 2012; Bevilacqua et al.,
2012). Microorganisms with b-glucosidase activities are also
convenient because they can be used to hydrolyse the glucoside
oleuropein, removing the natural bitterness present in directly
brined olives, without a lye treatment (Psani and Kotzekidou, 2006;
Arroyo-Lo  pez et al., 2012). Unfortunately, very few quantitative
studies are available in the literature about the inuence and ac-
tivity of these microorganisms on real table olive fermentations
(Rodrguez-Go mez et al., 2010, 2012; Bautista-Gallego et al., 2011).
This work offers, for the rst time, quantitative information on the
potential technological and probiotic applications of diverse yeast
species isolated from natural black Greek table olive fermentation,
which were then analysed using PCA for the selection and
discrimination of species with the most promising characteristics.
Especially two strains, W. anomalus Y18 and P. guilliermondii Y16,
showed desirable features and were clearly differentiated from the
rest of isolates. In the last years, many researchers have also used
this methodology for the selection of the most promising yeast
starters related with table olive processing (Bevilacqua et al., 2009,
2013; Rodrguez-Go  mez et al., 2012).
The presence of W. anomalus (previously called Pichia anomala)
in directly brined olives, and in table olives in general, is very
common (Hern andez et al., 2007; Bautista-Gallego et al., 2011;
Arroyo-Lo  pez et al., 2012; Doulgeraki et al., 2012). This species is
well adapted to grow under the environmental conditions that
drive table olive fermentations such as low pH and high NaCl
concentrations. Romero-Gil et al. (2013) reported similar MIC
values for NaCl (166.9 g/L) for diverse W. anomalus strains isolated
from Spanish table olive processing compared to the isolate assayed
in this work (163.2 g/L). In a previous study, Hern andez et al. (2007)
found diverse strains of this species isolated from table olive fer-
mentations with interesting technological applications. Recently,
Bautista-Gallego et al. (2011) and Tofalo et al. (2013) have reported
the high b-glucosidase activity of this microorganism. All these
features make this species one of the best candidates to obtain
isolates with technological applications in table olives, as was also
reported recently by Rodrguez-Go mez et al. (2012) and Bevilacqua
et al. (2013). Data obtained in this work prove again the high po-
tential technological application that this species, and especially
the Y18 isolate, may have during table olive processing because of
its high lipase, esterase and b-glucosidase activities.
P. guilliermondii is another yeast species usually also found in
table olive processing (Nisiotou et al., 2010; Tofalo et al., 2012) and
according to results obtained in this work with the Y16 isolate,
showed interesting potential probiotic properties. According to
FAO/WHO (2001), probiotics are live microorganisms which, when
administered in adequate amounts, confer a health benet to the
host. They should be resistant to gastric juices and be able to grow
in the presence of bile. There is a series of in vitro tests such as acid
and bile tolerance, although requiring further renement, that are
usually applied as a rst approach for the selection of potential
probiotic microorganisms (FAO/WHO, 2001). In the specic case of
table olives, only Psani and Kotzekidou (2006) and Silva et al. (2011)
have started to assess the probiotic potential of yeasts isolated from
this fermented vegetable using these tests. The rst authors found
diverse Torulaspora delbrueckii and D. hansenii strains which toler-
ated high bile salt concentrations and low pH values, while the later
authors found diverse Pichia fermentans and Candida oleophila
strains with similar properties. It is remarkable to notify how many
of the yeast strains assayed in this work showed survival percent-
ages to simulated digestions higher than the probiotic bacterium
used as control, which was especially evident in the case of
P. guilliermondii Y16.
This is the rst time that phytase activity has been studied for
table olive related yeasts. Phytic acid or phytate is the primary
storage form of phosphorous in mature seeds of plants and it is
abundant in legumes, oilseeds and many cereal grains. Phytate has
antinutritional effects because of the strong chelating capacity of this
compound with divalent minerals (calcium, zinc, etc.). Humans lack
the required enzymes in the gastrointestinal tract for the degrada-
tion of phytate complexes, but yeasts are able to desphosphorylate
these antinutritional compounds by the production of phytase en-
zymes, which release free inorganic phosphate, inositol phosphate
esters and minerals. These enzymes are widespread in yeast species
such as Issatchenkia orientalis, S. cerevisiae, T. delbrueckii and K. lactis
(Moslehi-Jenabian et al., 2010; Olstorpe et al., 2009), many of them
usually isolated from diverse table olive processing (Arroyo-Lo pez
et al., 2012). However, in this work, M. pulcherrima Y20 was the
strain with the highest phytase activity levels in both cellular and
extracellular fraction in laboratory medium, followed by P. kluyveri
Y17 and P. guilliermondii Y16 in the cellular fraction.
5. Conclusions
The results obtained in this survey have shown that especially
two isolates (W. anomalus Y18 and P. guilliermondii Y16) show
interesting properties for their use as starters in table olive pro-
cessing, the former with technological applications and the later as
potential probiotic agent. This information provides new perspec-
tives for the industrial sector of natural black olives and enhances
the research on the impact of these microorganisms in the char-
acteristics of this fermented vegetable. The selection of the most
appropriate strains, in combination with/without lactic acid bac-
teria, could transform table olives from a traditional food to a novel
functional commodity with multiple benets for the consumers.
Thus, the real challenge for the Greek table olive industry is to
address this demand of consumers and develop new products with
enhanced sensory and functional properties that may have an
impact on human health in the long run. However, further studies
should be carried out to determine and validate the inuence of
both microorganisms on industrial scale fermentations.
Acknowledgements
The research leading to these results has received funding from
the Spanish Government (Ramo  n y Cajal program), and the Junta de
 S. Bonatsou et al. / Food Microbiology 46 (2015) 66e73
Andaluca (through nancial support to group AGR-125). S.
pez wish to express thanks to the Eu-
Bonatsou and F.N. Arroyo-Lo
ropean Erasmus program and Ramo n y Cajal postdoctoral research
contract (Spanish Government), respectively.
References
Aponte, M., Ventorino, V., Blaiotta, G., Volpe, G., Farina, V., Avellone, G.,
Lanza, C.M., Moschetti, G., 2010. Study of green Sicilian table olive fermen-
tations through microbiological, chemical and sensory analyses. Food
Microbiol. 27, 162e170.
Arroyo Lo pez, F.N., Bautista Gallego, J., Querol, A., Garrido-Fern
andez, A., 2008. Role
of yeasts in table olive production. Int. J. Food Microbiol. 128, 189e196.
Arroyo-Lo  pez, F.N., Romero-Gil, V., Bautista-Gallego, J., Rodrguez-Go mez, F.,
Jimenez-Daz, R., Garca-Garca, P., Querol, A., Garrido-Ferna ndez, A., 2012.
Yeasts in table olive processing: desirable or spoilage microorganisms? Int. J.
Food Microbiol. 160, 42e49.
Bautista-Gallego, J., Rodrguez-Go mez, F., Barrio, E., Querol, A., Garrido-
Fernandez, A., Arroyo-Lo pez, F.N., 2011. Exploring the yeast biodiversity of
green table olive industrial fermentations for technological applications. Int. J.
Food Microbiol. 147, 89e96.
Bevilacqua, A., Perricone, M., Cannarsi, M., Corbo, M.R., Sinigaglia, M., 2009. Tech-
nological and spoiling characteristics of the yeast microora isolated from Bella
di Cerignola table olives. Int. J. Food Sci. Technol. 44, 2198e2207.
Bevilacqua, A., Corbo, M.R., Sinigaglia, M., 2012. Selection of yeasts as starters cul-
tures for table olives: a step-by-step procedure. Front. Microbiol. 3. Art. 194.
Bevilacqua, A., Beneduce, L., Sinigaglia, M., Corbo, M.R., 2013. Selection of yeasts as
starters cultures for table olives. J. Food Sci. 78, M742eM751.
Ciafardini, G., Zullo, B.A., Iride, A., 2006. Lipase production by yeasts from extra
virgin olive oil. Food Microbiol. 23, 60e67.
Corcoran, B.M., Stanton, C., Fitzgerald, G.F., Ross, R.P., 2007. Growth of probiotic
lactobacilli in the presence of oleic acid enhances subsequent survival in gastric
juice. Microbiology 153, 291e299.
Costas, M., Deive, F.J., Longo, M.A., 2007. Lipolytic activity in submerged cultures of
Issatchenkia orientalis. Process Biochem. 39, 2109e2114.
Doulgeraki, A.I., Hondrodimou, O., Iliopoulos, V., Panagou, E.Z., 2012. Lactic acid
bacteria and yeast heterogeneity during aerobic and modied atmosphere
packaging storage of natural black Conservolea olives in polyethylene pouches.
Food Control 26, 49e57.
FAO/WHO, 2001. Health and Nutritional Properties of Probiotics in Food Including
Power Milk with Live Lactic Acid Bacteria. Amerian Co  doba Park Hotel, Cordoba,
Argentina. Accessed at: http://www.who.int/foodsafety/publications/fs_
management/en/probiotics.pdf?ua1.
Garrido-Ferna ndez, A., Ferna ndez Daz, M.J., Adams, R.M., 1997. Table Olives: Pro-
duction and Processing. Chapman & Hall, London, UK.
Haros, M., Bielecka, M., Sanz, Y., 2005. Phytase activity as a novel metabolic feature
in Bidobacterium. FEMS Microbiol. Lett. 247, 231e239.
Heperkan, D., 2013. Microbiota of table olive fermentations and criteria of selection
for their use as starters. Front. Microbiol. 4, 143.
73
Herna ndez, A., Martin, A., Aranda, E., Pe rez-Nevado, F., Co rdoba, M.G., 2007. Iden-
tication and characterization of yeast isolated from the elaboration of
seasoned green table olives. Food Microbiol. 24, 346e351.
Hurtado, A., Reguant, C., Bordons, A., Roze s, N., 2012. Lactic acid bacteria from
fermented table olives. Food Microbiol. 31, 1e8.
IOC (International Olive Oil Council), 2013. Statistical of Table Olive Production.
http://www.internationaloliveoil.org/estaticos/view/132-world-table-olive-
gures (Last accessed 24.01.13.).
Jackson, J.E., 1991. A User's Guide to Principal Components. Wiley, New York, USA.
Jolliffe, I.T., 1986. Principal Component Analysis. Springer-Verlag, New York, USA.
Moslehi-Jenabian, S., Lindegaard Pedersen, L., Jespersen, L., 2010. Benecial effects
of probiotic and food borne yeasts on human health. Nutrients 2, 449e473.
Nisiotou, A.A., Chorianopoulos, N., Nychas, G.-J.E., Panagou, E.Z., 2010. Yeast het-
erogeneity during spontaneous fermentation of black Conservolea olives in
different brine solutions. J. Appl. Microbiol. 108, 396e405.
Olstorpe, M., Schnren, J., Passoth, V., 2009. Screening of yeast strains for phytase
activity. FEMS Yeast Res. 9, 478e488.
Psani, M., Kotzekidou, P., 2006. Technological characteristics of yeast strains and
their potential as starter adjuncts in Greek-style black olive fermentation.
World J. Microbiol. Biotechnol. 22, 1329e1336.
Querol, A., Fleet, H., 2006. Yeasts in Food and Beverages. Springer-Verlag, Berlin,
Germany.
Rodrguez-Go  mez, F., Arroyo-Lo
 pez, F.N., Lo
pez-Lopez, A., Bautista-Gallego, J., Gar-
rido-Ferna ndez, A., 2010. Lipolytic activity of the yeast species associated with
the fermentation/storage phase of ripe olive processing. Food Microbiol. 27,
604e612.
Rodrguez-Go  mez, F., Romero-Gil, V., Bautista-Gallego, J., Garrido-Fern andez, A.,
Arroyo-Lo pez, F.N., 2012. Multivariate analysis to discriminate yeast strains
with technological applications in table olive processing. World J. Microbiol.
Biotechnol. 28, 1761e1770.
Romero-Gil, V., Bautista-Gallego, J., Rodrguez-Go mez, F., Garca-Garca, P., Jime
nez-
Daz, R., Garrido-Fern andez, A., Arroyo-Lo pez, F.N., 2013. Evaluating the indi-
vidual effects of temperature and salt on table olive related microorganisms.
Food Microbiol. 33, 178e184.
Romo-Sa nchez, S., Alves-Baf, M., Are valo-Villena, M., beda-Iranzo, J., Briones-
Perez, A., 2010. Yeast biodiversity from oleic ecosystems: study of their
biotechnological properties. Food Microbiol. 27, 487e492.
Rosi, I., Vinella, M., Domizio, P., 1994. Characterization of b-glucosidase activity in
yeasts of oenological origin. J. Appl. Bacteriol. 77, 519e527.
Silva, T., Reto, M., Sol, M., Peito, A., Peres, C.M., Peres, C., Xavier-Malcata, F., 2011.
Characterization of yeasts from Portuguese brined olives, with a focus on their
potentially probiotic behavior. LWT Food Sci. Technol. 44, 1349e1354.
Tofalo, R., Schirone, M., Perpetuini, G., Suzzi, G., Corsetti, A., 2012. Development and
application of a real-time PCR-based assay to enumerate total yeasts and Pichia
anomala, Pichia guillermondii and Pichia kluyveri in fermented table olives. Food
Control 23, 356e362.
Tofalo, R., Perpetuini, G., Schirone, M., Suzzi, G., Corsetti, A., 2013. Yeast biota
associated to naturally fermented table olives from different Italian cultivars.
Int. J. Food Microbiol. 161, 203e208.
Winer, B.J., 1962. Statistical Principles in Experimental Design. McGraw-Hill, New
York.