Food Microbiology 28 (2011) 810e817

Contents lists available at ScienceDirect

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

Exploring the lag phase and growth initiation of a yeast culture

by means of an individual-based model
Marta Ginovart a, *, Clara Prats b, Xavier Portell c, Moises Silbert d
a
Department of Applied Mathematics III, Universitat Politècnica de Catalunya, Campus Baix Llobregat, Esteve Terradas 8, 08860 Castelldefels, Barcelona, Spain
b
Department of Physics and Nuclear Engineering, Universitat Politècnica de Catalunya, Campus Baix Llobregat, Esteve Terradas 8, 08860 Castelldefels, Barcelona, Spain
c
Department of Agri-Food Engineering and Biotechnology, Universitat Politècnica de Catalunya, Campus Baix Llobregat, Esteve Terradas 8, 08860 Castelldefels, Barcelona, Spain
d
Institute of Food Research, Norwich Research Park Colney, Norwich NR4 7UA, UK

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

Article history: The performance of fermentation processes is greatly influenced by the size and quality of inocula. The
Received 31 January 2010 characterization of the replicative age is decided by the number of birth scars each yeast exhibits on its
Received in revised form cellular membrane. Yeast ageing and inoculum size are factors that affect industrial fermentation, partic-
1 May 2010
ularly those processes in which the yeast cells are reused such as the production of beer. This process reuses
Accepted 4 May 2010
Available online 11 May 2010
yeast cropped at the end of one fermentation in the following one, in a process called “serial repitching”.
The aim of this study was to explore the effects of inoculum size and ageing on the first stages of the
dynamics of yeast population growth. However, only Individual-based Models (IbMs) allow the study of
Keywords:
Individual-based model
small, well-characterized, microbial inocula. We used INDISIM-YEAST, based on the generic IbM simulator
Lag phase INDISIM, to carry out these studies. Several simulations were performed to analyze the effect of the inoc-
Growth initiation ulum size and genealogical age of the cells that made it up on the lag phase, first division time and specific
Yeast inoculum growth rate. The shortest lag phase and time to the first division were obtained with largest inocula and with
Inoculum size the youngest inoculated parent cells.
Yeast cell age Ó 2010 Elsevier Ltd. All rights reserved.

1. Introduction divided, and so this number constitutes a biomarker of replicative

Biochemical factors have long been recognized as the most
effective elements for determining the yield and productivity of
industrial fermentation. Since these factors are carried forward from
the inocula to the production stages, an accurate characterization of
these inocula, in terms of size and quality, is of importance in
order to improve fermentation consistency and thus increase the
efficiency and/or yield of the biotechnological process (Birol and
Özergin-Ülgen, 1995, Gibson et al., 2007).
The yeast Saccharomyces cerevisiae has a limited replicative
lifespan (Walker, 1998). We are concerned here with yeast budding
reproduction, which leads to scar formation and to cells dividing
unequally. Ageing in budding yeast is not determined by chrono-
logical lifespan (measured as the ability of non-dividing cells to
maintain viability over time) but by the number of times an indi-
vidual cell is capable of dividing (Maskell et al., 2003). Each cell
within a population is only capable of a finite number of divisions
prior to senescence and death. The number of bud scars on the wall
of the mother cell is directly related to the number of times a cell has
* Corresponding author.
E-mail address: marta.ginovart@upc.edu (M. Ginovart).
cell age. In addition, it has been demonstrated that there is an
increase in cell size with age (Walker, 1998). However, this increase
is not considered constant between successive ages, as it may
decrease in older individuals in accordance with the diminution of
the metabolic rate for aged cells. In fact, although aged cells are
known to be larger than younger cells, mixed age yeast populations
can exhibit a greater average size than older populations, depending
on the environmental conditions and other unidentified factors
(Porro et al., 2009). Variety is an intrinsic property of microbial life.
As a consequence of senescence, yeast cells are subject to
morphological, metabolic and genetic modifications (Powell et al.,
2000, 2003). In industrial applications yeast is usually propagated
in a number of steps before being inoculated into the final fermen-
tation medium. Despite the fact that the physiological condition of
the yeast cells may greatly affect the fermentation duration and
outcome the inoculum is not often well-defined (Walker, 1998).
By way of example, the production of beer reuses yeast cropped
at the end of fermentation in subsequent fermentation, so yeast is
maintained and reused a number of times in a process called ‘serial
repitching’. Toward the end of beer fermentation, the yeast sedi-
ments and collects within the fermenter cone (Powell et al., 2003).
The rate at which each cell sediments is believed to vary according

0740-0020/$ e see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.fm.2010.05.004
 M. Ginovart et al. / Food Microbiology 28 (2011) 810e817
to its replicative age. Older cells tend toward faster sedimentation,
and accumulate at the bottom of the cone (although their precise
location may depend on the strain) leading to stratification by
genealogical age. Sedimentation results in the formation of zones
enriched with cells of a particular age (Kurec et al., 2009). At the
end of fermentation a portion of the yeast is removed (‘cropped’)
from the fermentation vessel for serial repitching. Typically this is
the centre-top portion of the yeast crop, theoretically comprising
middle-aged and virgin cells (Powell et al., 2003). However, yeast is
increasingly removed early to decrease process time by means of
a ‘warm’ or ‘early’ cropping regime, which facilitates removal of the
lower portion of the crop that comprises a greater proportion of
aged cells (Powell et al., 2004). Harvesting yeast may therefore
select for a population with an imbalance of young and aged indi-
viduals, depending on the cropping mechanism employed (Powell
et al., 2003). Although a yeast crop may be considered by the
brewer to be a homogenous culture in terms of vitality and viability,
it has been demonstrated that it usually comprises a variety of
individual cells which may differ in respect to their fermentative
capabilities (Walker, 1998, Gibson et al., 2007).
These processes highlight and justify the interest in studying the
main features of a yeast inoculum, obtained from one fermentation
that is going to be used in the following one.
When yeast cells are inoculated into a fresh growth medium,
they enter a brief lag phase in which they are biochemically active
but still do not divide (Gibson et al., 2007). After this lag phase, cells
enter their cell cycle and start dividing. The microbial lag phase has
usually been investigated with continuous population models
using rather high inoculum levels or homogeneous inocula (Baranyi
and Roberts, 1994; Baranyi, 2002; Swinnen et al., 2004), but lately
more and more studies have concentrated on single cell lag times
(McKellar, 2001; Devlieghere et al., 2005; Kutalik et al., 2005;
Metris et al., 2005; Koutsoumanis, 2008; Malakar and Barker,
2008; Baranyi et al., 2009). Nevertheless, when microbial growth
starts from a few accurately characterized cells, the study of this
evolution demands an individual-based approach.
Individual-based Models (IbMs) explicitly simulate individuals,
and the population behavior emerges from their cumulative
behavior and interaction (Grimm and Railsback, 2005). In ecolog-
ical modeling, IbM constitutes a well-established alternative to the
more traditional population-level approach, in which population
parameters are modified directly using model equations. Most
applications of IbMs have been geared to higher trophic levels.
However, advances in microbiology and biochemistry have stimu-
lated an increase in the application of IbMs to microbes as well
(see Hellweger and Bucci, 2009 for a recent review). A model in this
context that is individual-based means that the basic entities are
the individual cells. Thus, every cell is considered as an entity and
has its own properties that change during its life according to the
set of rules declared. Unlike traditional top-down approaches, IbMs
are bottom-up approaches. Sometimes, under certain assumptions,
the two approaches converge in yielding similar results as the size
of the population increases (Gómez-Mourelo and Ginovart, 2009).
Out of the several microbial IbMs that are available nowadays
(Kreft et al., 1998; Dens et al., 2005; Hellweger and Bucci, 2009)
we used INDISIM, the simulator developed by our group (Ginovart
et al., 2002; Ferrer et al., 2008), which has already been used to
study different features of bacterial lag phase yielding an ample
pool of interesting results (Prats et al., 2006, 2008). INDISIM-YEAST
is the adaptation of INDISIM to the study of specific features of the
yeast cell cycle in yeast populations growing in a liquid medium
(Ginovart et al., 2007; Ginovart and Cañadas, 2008; Prats et al., in
press). This simulator has also recently been used in a prelimi-
nary study of some aspects of the influence of cell ageing on
fermentation processes (Ginovart et al., 2009).
811
The aim of this study was to explore the effects of specific
characteristics of the initial inocula on the dynamics of a virtual
yeast culture in liquid medium during the first stages of growth by
means of the individual-based simulator INDISIM-YEAST.
2. Materials and methods
We used INDISIM-YEAST as the individual-based simulator for
this study, which is based on the generic simulator INDISIM (Ginovart
et al., 2002, 2007; Ginovart and Cañadas, 2008; Prats et al., in press).
A short review and selected applications of INDISIM can be found in
Ferrer et al. (2008).
A general presentation and schematic outline of the simulator
INDISIM-YEAST is as follows. It is discrete in space and time and can be
used for modeling yeast populations under different environmental
conditions. It is rule-based, using stochastic variables, and subject to
appropriate boundary conditions. The physical domain where
the virtual fermentation takes place is divided into spatial cubes,
a microcosm of the system: it includes the liquid medium with yeast
cells, glucose particles, as the main nutrient, and excreted ethanol
particles, as the only end product. The temporal evolution of the
population is divided into equal intervals associated with computer
or time steps. The simulator is made up of several elements: 1) the
initialization of the system, in which the input data are entered and
the initial configuration of the population and environment is
determined; 2) the main loop (time step), in which all the rules for
each yeast cell and the medium are implemented and repeated until
the end of the simulations; and 3) the output of results, including the
information obtained from each yeast cell at the end of each time step
stored by the simulation. This makes it possible to obtain the results of
the simulations both at the level of individual cells and at that of the
yeast population throughout the evolutions.
The virtual system models the behavior of each yeast cell in
the following categories: a) random motion; b) glucose uptake; c)
cellular maintenance; d) new biomass production; e) ethanol
excretion; f) budding reproduction (detailed below); and g) cell
viability. The budding reproduction model assumes two differenti-
ated phases, phase 1 or the unbudded phase, when the cell gets
ready to create a new cell (the bud), and phase 2 or the budding
phase, in which the daughter cell-genealogical age 0 (virgin cell)-
grows until it separates from the parent cell, leaving behind another
scar (Fig. 1). Each yeast cell is characterized by its: i) biomass, which
is related by the model to spherical geometry in order to evaluate its
cellular surface; ii) genealogical age measured as the number of scars
on the cellular membrane; iii) state of the reproductive cycle,
Fig. 1. Scheme of the budding reproduction model implemented at individual level
yeast cell. U: phase 1 or the unbudded phase, B: phase 2 or the budding phase. Adapted
from Hatzis and Porro (2006).
812 M. Ginovart et al. / Food Microbiology 28 (2011) 810e817
involving the time spans and masses of both parent and daughter
cells in each of the two phases; and iv) survival times of the
yeast cells not satisfying the metabolic requirements for their
maintenance. During the implementation of the different parts of
the simulator we take into account, amongst other things, the
following: the scars left on the parent cells, as they affect the cellular
membrane and therefore the nutrient uptake and the possibility of
new reproduction cycles; the cell’s chemical stresses caused by the
presence of ethanol in the medium, as it slows down the production
of new biomass (it increases the energy requirements for cellular
maintenance); the increase in volume, or biomass, of parent cells, as
they increase their genealogical age; and the varying duration of
the reproduction cycles (unbudded and budding phases), since
these phases require a minimum time interval of stasis as well as
minimum growth of biomasses in order to change from one phase
into the other. For a more details about INDISIM-YEAST see Ginovart
et al. (2007) and Ginovart and Cañadas (2008), and for further
information about this simulator see also Gómez-Mourelo and
Ginovart (2009), and Prats et al. (in press).
The simulation output may be separated into data related to the
global properties of the system and data concerning the properties
of individual yeast cells. The former includes information on
temporal evolution of the number of nutrient particles (glucose),
the number of metabolites (ethanol particles), the average nutrient
consumption (defined as the number of nutrient particles metab-
olized at each time step divided by the number of viable cells), the
number of viable yeast cells, the number of non-viable yeast cells,
viable yeast biomass, non-viable yeast biomass, heat dissipation
of the system, and maintenance energy expended by the yeast
population (defined as the number of metabolized nutrient parti-
cles not used in the production of new biomass). The latter provides
information on the distributions of genealogical age and of the mass
of the populations, which allows us to observe the structure of
the population throughout the fermentation process. The preceding
separation of output simulation results mirrors the classification
of experimental techniques used to study those properties. The
web page htpps://aneto.upc.es/simulacio/hoja-portada presents
a basic version of INDISIM-YEAST, which allows for simulations of
fermentation processes, and graphically present a few of the vari-
ables controlled by the simulator (Ginovart and Cañadas, 2008).
Other applications and publications about INDISIM-YEAST can also
be found in this web page.
Among the various simulation results at the population level, we
will focus on the temporal evolution of the yeast population that
grows from a completely characterized inoculum. Special attention
was paid during the first stages of its development until the
Fig. 2. Calculation of the following parameters: lag time, first division time and maximum growth rate (exponential phase) performed over the evolution of the growth of some
simulated yeast populations. Left: geometrical method to obtain the lag phase (l) and the maximum specific growth rate (m), where a ¼ LnN0 þ 0.85(LnNMAX  LnN0) and
b ¼ LnN0 þ 0.60(LnNMAX  LnN0); and the first division time (tFD) is indicated. Right: two different simulations showing how negative and positive lag times can be achieved.
population reached the exponential phase. There was no nutrient
limitation at the initial conditions of the simulated culture.
Two parameters were used to characterize the outcome of a given
culture growth during the early stages (Pin and Baranyi, 2008). The
first is the classic lag parameter, defined at the population level of
description and calculated through its geometrical definition; it is
the graphical intersection between the initial level (LnN0) and the
prolongation of the exponential straight line in a semilogarithmic
representation (LnN ¼ mt þ b). Parameters m (maximum growth rate)
and b are determined by means of a logarithmic regression of the
upper interval. Finally, the lag parameter evaluation results in l ¼
(LnN0  b)/m as shown in Fig. 2. In order to better characterize the
initial steps of the yeast population at an individual level of
description, a second parameter was also considered; the time when
the first microbial division (tFD) took place (or time till the first
budding reproduction appeared), also shown in Fig. 2.
The genealogical age of yeast cells and the small size of daughter
cells compared to older cells are two individual characteristics that
can be significant in determining the evolution of a yeast culture
during its first stages. Thus, several simulations to evaluate the
above parameters (l and tFD) were carried out taking inocula with
different characteristics in terms of: i) the genealogical ages of
their cells (namely samples that include daughter cells or virgin
cells with zero scars and/or parent cells with one, two, three or
more scars on their membranes), and ii) their size (the number of
yeast cells of the inocula, ranging from 1 cell to 1000 cells). Lastly,
other relevant simulation results, such as the growth rate achieved
in the exponential phase, are presented.
From a heterogeneous yeast population that emerges from
a simulated fermentation of INDISIM-YEAST, as a virtual “source
culture” (Ginovart et al., 2007, Prats et al., in press), successive sets of
cells were obtained to be used as inocula in subsequent simulations.
The initial groups of cells were randomly taken from the virtual
source culture according to the size and age group that had to be
addressed, and configured the starting point of the successive simu-
lation. The other significant input data required for the simulator in
this case study were the number of glucose particles distributed
uniformly over the spatial cubs of the domain and, fundamentally, the
values that were conducting the yeast cells’ actions or behavior rules
(Ginovart et al., 2007; Ginovart and Cañadas, 2008; Prats et al., in
press). We used dimensionless units. We assumed that units of
length were given in terms of the length of a spatial cell, units of time
in terms of program steps, and units of mass in terms of the “start”
mass of reproduction. Consequently, the simulation results obtained
can be understood as qualitative and available to be compared among
themselves.
 M. Ginovart et al. / Food Microbiology 28 (2011) 810e817 813

Fig. 3. Results of the first simulation series, with 50 independent runs for each case, to evaluate the effect of the genealogical age of a single inoculated cell on the parameters l and
tFD of the growth curve. Lag parameter (left) and first division time (right) versus the number of scars of the single-cell inoculum. The dashed line indicates the mean value for each
genealogical age.

3. Results
Low inocula growth curves have high synchronisms during the
first stages of the temporal evolution. These synchronisms result in
a geometrical effect that disturbs the geometrical evaluation of the
lag phase as it is seen in the simulation results shown in Fig. 2.
As a consequence, in some cases the population lag time is shorter
than the corresponding time for the first bud reproduction (Fig. 2).
This is not consistent at all with the concept of lag as the phase at
which no reproductions take place (i.e. growth rate is equal to zero),
since a part of this non-division period -between lag time and the
first division- is assembled to the subsequent exponential phase.
This effect was also analyzed by Baranyi et al. (2009) in the case of
bacterial cultures.
We simulated the growth of different virtual inocula. From
a heterogeneous population of yeast cells, successive samples were
obtained to be used as inocula in subsequent simulations. The first
series with 50 independent runs for each case was designed to
evaluate the effect of the genealogical age of a single inoculated cell
on the parameters l and tFD of the growth curve (Fig. 3). It can be
seen that both parameters have their minimum value for the
youngest parent yeast. As the genealogical age of the inoculum
increased, the values of these parameters became greater. Daughter
cells also appeared to have longer lag phase than young parent
cells. These outcomes of the simulator at the beginning of
fermentation, during the first stages of population growth, can
basically be explained by means of geometric considerations taking
into account the budding reproduction model (Fig. 1). Cell division
rate is limited by the rate of growth of an individual cell and the rate
of protein synthesis. Cells can pass through the “Start” point only
when they have reached a critical cell size and, daughter cells and
parent cells may be distinguished by both cell size and division rate
(Smart 1999). It has been stated by Powell et al. (2000, 2003) that
with newly budded virgin cells the time taken to reach the critical
size required for the first division would result in a slight delay at
the onset of growth. This reflects the extended period that virgin
cells spent in the first phase prior to proliferation. Virgin cells do
not immediately meet the requirements to pass the “Start” point
that leads to budding reproduction during the cell cycle and, in
order to achieve the specific size, they had to assimilate nutrients
and convert them into biomass, which is time consuming (Powell
et al., 2003). In contrast, mother cells (genealogical age greater
than zero) would satisfy requirements for the “Start” point more
rapidly, and therefore were able to divide quickly, reducing
fermentation lag time and first division time. Seeding fermentation
with aged yeast would result in an extended lag phase due to the
slow progression in the cell cycle. The rate of increase in cell size
may not continue throughout the lifespan, but may even decrease
in older individuals in accordance with the diminution in metabolic
rate reported for aged cells by Motizuki and Tsurugi (1992).
The second series, also with 50 independent runs in each case,
was designed to simulate the yeast population growth from inocula
of different sizes (from 1 to 1000 cells) randomly taken from the
aforementioned population and without age restrictions on the
initial cells. Fig. 4 shows the resulting effect of the inoculum size
on l and tFD. It can be observed that large inocula reached the
exponential phase sooner than small inocula, and that dispersion of
these parameters also decreased as the inoculum size increased.

Fig. 4. Results of the second simulation series, with 50 independent runs in each case, to simulate the yeast population growth from inocula of different sizes randomly taken from
a virtual source culture without age restrictions on the initial cells. Lag parameter (left) and first division time (right) versus inoculum size. The dashed line indicates the mean value
for each inoculum size.
814 M. Ginovart et al. / Food Microbiology 28 (2011) 810e817

Fig. 5. Box plots for the third simulations series where inocula combining the two factors, size and genealogical age were used. Different inoculum sizes (from 1 to 1000 cells) with
only virgin cells (zero scars), middle-aged cells (1 to 5 scars) or old cells (six scars or more) were chosen to perform sets of 50 independent runs for each combination. Lag parameter
versus inoculum size depending on the genealogical age of the set of cells. A middle line inside the box represents the median. Outliers-data that were more than 1.5 times the
interquartile range above or below the box are shown as asterisks. The solid line connects the mean values (C) for each inoculum size.

Large inocula reached the exponential phase sooner than the small
inocula, and this was mainly due to the major probability of having
cells with short tFD in the inoculum greater if the inoculum is large.
For instance, the fermentation and metabolism characteristics
under five initial cell densities of S. cerevisiae were investigated by
Ding et al. (2009), and they found significant effect of inoculum size
on lag phase duration. The prolonged lag phase of yeast cells at low
inoculum size might be envisaged as the statistical effects arising
from the variability in individual lag times, i.e., first division times
(Ding et al., 2009), which is supported by these simulations. Both
parameters, l and tFD, behave in the same way except for one-cell
inocula, in which population l decreases. This is a consequence of
the above-mentioned geometrical effect of the initial synchronism,
which causes an underestimation of the population lag (Fig. 2).
A third series of simulations was performed selecting inocula
that combined the preceding two factors, size and genealogical
ages. Specifically, different inoculum sizes (from 1 to 1000 cells)
with only virgin cells (zero scars), middle-aged cells (1 to 5 scars) or
old cells (six scars or more) were chosen to perform sets of 50
independent runs for each combination. This information made
possible the construction of box plots (also called box-and-whisker
plots) which can be particularly useful for showing the distribu-
tional characteristics of data and for comparing the set of
data corresponding to these three groups of genealogical ages
(Figs. 5e7). Outliers, unusually large or small observations corre-
sponding to certain simulations, can be identified in some of these
graphs. They are the values beyond the whiskers, which in other
cases extend to the highest and lowest data values. The tops of
the boxes are the third quartiles and the bottom ones are the first
quartiles, with the middle lines showing typical values, i.e., the
median values for these sets of data. The simulation results of Figs.
5 and 6 show the expected influence of these initial features of the
inocula on l and tFD in accordance with previous series results and
experimental data (Powell et al., 2000, 2003). It has been observed
that yeast cultures enriched with old or young cells performed in
a different manner during fermentation, and the analysis of the
initial stages of fermentation indicated that virgin cells were slower
to begin utilizing sugars when compared to aged cultures (Powell

Fig. 6. Box plots for the third simulation series, where inocula combining the two factors, size and genealogical ages, were used. Different inoculum sizes (from 1 to 1000 cells) with
only virgin cells (zero scars), middle-aged cells (1 to 5 scars) or old cells (six scars or more) were chosen to perform sets of 50 independent runs for each combination. First division
time versus inoculum size depending on the genealogical age of the group of cells. A middle line inside the box represents the median. Outliers-data that were more than 1.5 times
the interquartile range above or below the box are shown as asterisks. The solid line connects the mean values (C) for each inoculum size.
 M. Ginovart et al. / Food Microbiology 28 (2011) 810e817 815

Fig. 7. Box plots for the third simulation series, where inocula combining the two factors, size and genealogical ages, were used. Different inoculum sizes (from 1 to 1000 cells) with
only virgin cells (zero scars), middle-aged cells (1 to 5 scars) or old cells (six scars or more) were chosen to perform sets of 50 independent runs for each combination. Specific
growth rate versus inoculum size depending on the genealogical age of the group of cells (only the simulations with viable inocula were considered). A middle line inside the box
represents the median. Outliers-data that were more than 1.5 times the interquartile range above or below the box are shown as asterisks. The solid line connects the mean values
(C) for each inoculum size.

et al., 2003). It has also been suggested by Smart (1999) that young
to middle-aged parent cells represented the most active portion of
the pitch population in terms of rapid growth and yeast biomass
production. Fig. 7 shows the effect of inocula characteristics on the
exponential phase maximum specific growth rate, m. In general, all
cultures reached similar potential m values, except for old large
inocula where a slight difference was observed. Non-growing
inocula were ruled out, and this was the case with many small old
inocula which were not counted in these results. Once a cell started
dividing, the culture started growing and reached almost the
maximum population growth rate. Most of the large inocula suc-
ceeded in initiating the culture growth because it was more prob-
able that at least one yeast cell could divide, but it seems that the
presence of a still considerable percentage of old cells along with
the new ones (virgin cells) could even slightly slow down
the population rate (Porro et al., 2009). It has been shown that the
degree of asymmetry of cell division in yeast cells is modulated by
nutritional supply, and that the size distribution of a growing
population is a stable distinctive feature of a given exponential
balanced growth condition. For all the simulations in this study, the
initial amount of glucose was the same and it is feasible to assume
that the available nutrient that was present in the exponential
phase was not identical when the smallest size inocula or greater
size inocula started. Porro et al. (2009) have reported experimental
data on the percentage of cells of different genealogical age as
a function of the specific growth rate of exponentially growing
yeast batch cultures in different media. It was clear from their paper
that the heterogeneity of parent cells was quite significant, showing
that the specific growth rate lowest values corresponded to the
parent cells highest percentages and the virgin cells lowest
percentages. This is in qualitative agreement with the simulation
results achieved with large inocula and middle-aged cells, which
showed only a slight decrease in the specific growth rate, and in
a much more obvious and noticeable way with the large and old
inocula, which showed the major decreases in the specific growth
rate (Fig. 7). The major dispersion of these simulation results was
greater for the parameters l and tFD for old cells and virgin cells
than for middle-aged cells. In addition, in this event, the central
data had major dispersion when very small inocula were used to
start on the simulations.

Fig. 8. Results of the third series of simulations showing the percentage of viable inocula versus the inoculum size, depending on the genealogical age of the group of cells. 50
independent runs for each combination were used.
816 M. Ginovart et al. / Food Microbiology 28 (2011) 810e817
Fig. 8 shows the percentage of viable inocula of this third series
of simulations carried out versus the inoculum size, depending on
the genealogical age of the group of cells. Also 50 independent runs
for each combination were carried out, and only very small inocula
were sometimes not capable of growing. This only happened with
virgin or old cells. In the case of middle-aged cells all the inocula
were successful. A higher number of cells increases the probability
that one or more cells in the population are able to grow. Thus, large
inocula allow growth initiation under conditions where growth of
small inocula is usually difficult. Experimental and quantitative
microbial studies that show the effect of inoculum size on microbial
growth initiation under combination of specific conditions that
control or stop microbial growth has been increasingly recognized
as necessary in food and predictive modeling (Koutsoumanis and
Sofos, 2005, Koutsoumanis, 2008). Experimental research with
small inocula is still scarce, but with the improvement of image
analysis techniques is being more usual.
4. Discussion
Inoculum size is one of the most important factors influencing
industrial fermentation, including lag phase duration, specific growth
rate, biomass yield and final product quality. In addition, investiga-
tions on the effect of cell population in microbial risk assessments,
food pathogen control and inactivation, are examples where the tFD
parameter must be taken into account as those phenomena generally
involve small inocula, and as has already been stated, in these
circumstances, its value differs from the l. With regard to yeast, most
researchers have been restricted to the technology of high cell density
fermentation (Ding et al., 2009). Nevertheless, the metabolic regu-
lation against the variation of inoculum size is still elusive and less
characterized, especially the changes of yeast intercellular metabolite
contents in fermentations with different initial cell densities. It is
necessary to gain further insight into the mechanisms of how yeast
metabolism and growth is affected by inoculum size, although some
recent metabolomic studies have already been performed in this
direction (Ding et al., 2009).
During aging, yeast displays a whole array of changes including
increase in size, cell surface wrinkling, increase of generation time,
increase in bud scar number and decrease in metabolic activity.
Experimental data regarding these phenomena with small inocula
need to deal with traditional techniques of staining and microscopic
counting of bud scars, which are tedious, labor demanding and do not
always provide real-time information on cell populations that are
growing (Powell et al., 2000). Conversely, the flexibility of a flow
cytometry method of yeast bud scar determination was verified
in beer fermentation technology, which offers new possibilities to
follow the development of inocula (Kurec et al., 2009).
We carried out individual-based simulations of a unicellular
fungi (the yeast cell used as a reference was S. cerevisiae) in which we
attempted to look into some aspects of the influence of cell ageing of
the initial inocula, and of their sizes, on the first stages of population
growth. We used the individual-based simulator INDISIM-YEAST to
examine the effects of age selection on yeast fermentation perfor-
mance. Within the stated limitations, this study mimics the indus-
trial production of beer, which reuses yeast cropped at the end of
fermentation in subsequent fermentations, so the immediate and
long term fermentation performance is conditioned by the charac-
teristics of these reused inocula (Motizuki and Tsurugi, 1992, Smart,
1999, Powell et al., 2000, 2003, Ding et al., 2009, Kurec et al., 2009,
Porro et al., 2009).
We have shown that INDISIM-YEAST is capable of distinguishing
the differences in the evolution of a population that emerges from
small inocula, whether started with a single microorganism, or
with a population made up of different genealogical ages or different
sizes. It is particularly useful in the study of small inocula and the
initial steps of the yeast population evolution because of the great
influence of the discrete and asymmetrical nature of yeast cellular
division. It is here that this model has an edge over top-down
continuous models, which are useful when the initial population
contains a large number of cells and the interest is the inspection of
average behaviors exhibited by them. The tendencies found in the
simulations resemble those seen during ‘serial repitching’ in beer
fermentation, although additional work must be done to improve
the simulation model and to integrate new biological features into
yeast cellular activity. For instance, the incorporation of other abiotic
compounds (nitrogen, ammonia, glycerol or oxygen), and the control
of individual internal reservoirs or intracellular storages would
improve the possibilities of the simulator INDISIM-YEAST.
In addition, two remarkable comments regarding the discrete
nature of our simulations are in order. First we would like to note
the uncanny similarities between the shape of the evolution of the
microbial growth of our simulations (expressed as the temporal
evolution of the number of viable yeast cells), and some experi-
mental results conducted by Koutsoumanis (2009), who, with the
time-lapse microscopy method for monitoring colonial growth of
single cells, was able to achieve truthful and accurate growth
curves. The second relates to the somewhat atypical samples or
extreme values for the parameters identified by means of box-and-
whisker plots. This phenomenon, i.e., the presence of outliers, is
also observed in some experiments which are, perhaps, not suffi-
ciently well thought-out when population mean values and/or
individual typical behaviors are the goals of the studies.
We may summarize the main findings of our simulation study as
follows:
i) The lag parameter has minimum values for the youngest
parent yeast cells. As the genealogical age of the inoculum
increases, lag time becomes longer. Daughter cells also have
longer lags than young parent cells.
ii) Large inocula reach the exponential phase sooner than do
small inocula.
iii) First division time has minimum values for the youngest
parent yeast cells. Small inocula with daughter or old cells
have the largest values for tFD.
To our knowledge no experimental results are available in the
literature to match these simulation results exactly. Nevertheless,
some detailed experimental references and empirical knowledge of
yeast cells linked with these topics allow the corroboration and
justification of these simulation results.
This type of study highlights one of the benefits of IbMs. IbM in
microbiology accounts for cellular morphology and metabolism, and
considers both cell variability among the population and uncertainty
in the processes affecting the cellular course. Pioneer studies such as
those of Takamatsu et al. (1985), Porro et al. (1988) and Strässle et al.
(1988), had already pointed out that the dissection of the complex
structure of a growing yeast population (an asymmetric population
with parent cells and daughter cells), with size and genealogical age
distributions evolving during fermentation, were key points that had
to be considered in a comprehensive yeast modeling study. The
choice of a microbial basic modeling approach, either population-
level or individual-based, is an important decision that needs to be
made at the beginning of a modeling project. There are arguments
for and against either approach (Bernaerts et al., 2004, Ferrer et al.,
2009, Hellweger and Bucci, 2009), and the right choice depends on
project-specific aspects, the characteristics of the microbial system
to be simulated and the questions asked to the model. In general,
arguments for the population-level approach to study yeast activity
are its simplicity, computational efficiency, the fact that it has
 M. Ginovart et al. / Food Microbiology 28 (2011) 810e817
been used and tested widely, and the availability of established and
peer-reviewed modeling frameworks. Arguments for the individual-
based approach include the ability to simulate variability, complete
life cycles and behavior adapted to internal and external conditions
(Gómez-Mourelo and Ginovart, 2009, Prats et al., in press). For our
yeast IbM, in this study, the main motivations were the ability to
resolve intra-population variability, the ability to link mechanisms at
the individual level to population-level behavior, and the inappli-
cability of the continuum hypothesis (i.e., the recognition that
individuals are different and that this fact can play an important role
in a biological system).
Acknowledgements
One of us (MG) is grateful to Dr G.C. Barker for hospitality in Nor-
wich where the write up of this paper was completed. We thank Dr R.
Carbó for several comments and suggestions. We gratefully acknowl-
edge the financial support received from the Ministry of Education and
Science (Plan Nacional I þ D þ i) through grant CGL2007-65142/BOS,
and the Universitat Politècnica de Catalunya (UPC).
References
Baranyi, J., 2002. Stochastic modelling of bacterial lag phase. Int. J. Food Microbiol.
73, 203e206.
Baranyi, J., Roberts, T.A., 1994. A dynamic approach to predicting bacterial growth in
food. Int. J. Food Microbiol. 23, 277e294.
Baranyi, J., George, S.M., Kutalik, Z., 2009. Parameter estimation for the distribution
of single cell lag times. J. Theor. Biol. 259, 24e30.
Bernaerts, K., Dens, E., Vereecken, K., Geeraerd, A.H., Standaert, A.R., Devlieghere, F.,
Debevere, J., Van Impe, J.F., 2004. Concepts and tools for predictive modeling of
microbial dynamics. J. Food Prot. 67, 2041e2052.
Birol, G., Özergin-Ülgen, K., 1995. A study on the effect of inoculums on yeast cell
growth and ethanol production. Tr. J. Chemistry 19, 250e257.
Dens, E.J., Bernaerts, K., Standaert, A.R., Kreft, J.U., Van Impe, J.F., 2005. Cell divi-
sion theory and individual-based modeling of microbial lag, Part II. Modeling
lag phenomena induced by temperature shifts. Int. J. Food Microbiol. 101,
319e332.
Devlieghere, F.K., Smet, K., Standaert, A.R., Geeraerd, A.H., Van Impe, J.F.,
Debevere, J., 2005. Modelling the individual cell lag phase: effect of tempera-
ture and pH on the individual cell lag distribution of Listeria monocytogenes. Int.
J. Food Microbiol. 100, 41e53.
Ding, M.Z., Tian, H.C., Cheng, J.S., Yuan, Y.J., 2009. Inoculum size-dependent inter-
active regulation of metabolism and stress response of Saccharomyces cerevisiae
revealed by comparative metabolomics. J. Biotechnol. 144, 279e286.
Ferrer, J., Prats, C., López, D., 2008. Individual-based modeling: an essential tool for
microbiology. J. Biol. Phys. 34, 19e37.
Ferrer, J., Prats, C., López, D., Vives-Rego, J., 2009. Mathematical modelling
methodologies in predictive food microbiology: a SWOT analysis. Int. J. Food
Microbiol. 134, 2e8.
Gibson, B.R., Lawrence, S.J., Leclaire, J.P.R., Powell, C.D., Smart, K.A., 2007. Yeast
responses to stresses associated with industrial brewery handling. FEMS
Microbiol. 31, 535e569.
Ginovart, M., López, D., Valls, J., 2002. INDISIM, and individual based discrete
simulation model to study bacterial cultures. J. Theor. Biol. 214, 305e319.
Ginovart, M., Xifré, J., López, D., Silbert, M., 2007. INDISIM-YEAST, an individual-
based model to study yeast population in batch cultures. In: Méndez-Vilas, A.
(Ed.), Communicating Current Research and Educational Topics and Trends
in Applied Microbiology. Microbiology Book Series n 1, vol. 1, pp. 401e409.
Formatex, Badajoz.
Ginovart, M., Cañadas, J.C., 2008. INDISIM-YEAST: an individual-based simulator on
a website for experimenting and investigating diverse dynamics of yeast
populations in liquid media. J. Ind. Microbiol. Biotechnol. 35, 1359e1366.
Ginovart, M., Portell, X., Silbert, M., 2009. Repitching of yeast in beer fermentations:
individual-based model simulations. In: Troch, I., Breitenecker, F. (Eds.),
Proceedings MATHMOD VIENNA 09. ARGESIM, Vienna, ISBN 978-3-901608-35-
3, pp. 2655e2658. Full Papers CD Volume.
817
Gómez-Mourelo, P., Ginovart, M., 2009. The differential equation counterpart of an
individual-based model for yeast population growth. Comput. Math. Appl. 58,
1360e1369.
Grimm, V., Railsback, S.F., 2005. Individual-based Modeling and Ecology. Princeton
University Press, New Jersey.
Hatzis, C., Porro, D., 2006. Morphologically-structured models of growing budding
yeast populations. J. Biotechnol. 124, 420e438.
Hellweger, F.L., Bucci, V., 2009. A bunch of tiny individuals e individual-based
modeling for microbes. Ecol. Model. 220, 8e22.
Koutsoumanis, K., 2008. A study on the variability in the growth limits of individual
cells and its effect on the behavior of microbial populations. Int. J. Food
Microbiol. 128, 116e121.
Koutsoumanis, K., 2009. Behavioural noise and colonial growth dynamics of single
microbial cells. In: Plenary Lecture of the 6th International Conference on
Predictive Modeling in Foods. http://www.icpmf.org/pp/Plenary%20lecture_
Koutsoumanis.pdf.
Koutsoumanis, K., Sofos, J., 2005. Effect of inoculum size on the combined
temperature, pH and aw limits for growth of Listeria monocytogenes. Int. J. Food
Microbiol. 104, 83e91.
Kreft, J.U., Booth, G., Wimpenny, J.W.T., 1998. BacSim, a simulator for individual-
based modelling of bacterial colony growth. Microbiology 144, 3275e3287.
Kurec, M., Baszczyn  ski, M., Lehnert, R., Mota, A., Teixeira, J.A., Brányik, T., 2009. Flow
cytometry for age assessment of a yeast population and its application in beer
fermentations. J. Inst. Brew. 115, 253e258.
Kutalik, Z., Razaz, M., Baranyi, J., 2005. Connection between stochastic and deter-
ministic modelling of microbial growth. J. Theor. Biol. 232, 285e299.
Malakar, P.K., Barker, G.C., 2008. Estimating single-cell lag times via a Bayesian
scheme. Appl. Environ. Microbiol. 74, 7098e7099.
Maskell, D.L., Kennedy, A.I., Hodgson, J.A., Smart, K.A., 2003. Chronological and
replicative lifespan of polyploid Saccharomyces cerevisiae (syn. S. pastorianus).
FEMS Yeast Res. 3, 201e209.
Metris, A., LeMarc, Y., Elfwing, A., Ballagi, A., Baranyi, J., 2005. Modelling the vari-
ability of lag times and the first generation times of single cells of E. coli. Int.
J. Food Microbiol. 100, 13e19.
McKellar, R.C., 2001. Development of a dynamic continuous-discrete-continuous
model describing the lag phase of individual bacterial cells. J. Appl. Microbiol.
90, 407e413.
Motizuki, M., Tsurugi, K., 1992. The effect of aging on protein synthesis in the yeast
Saccharomyces cerevisiae. Mech. Age Dev. 64, 235e245.
Pin, C., Baranyi, J., 2008. Single-cell and population lag time as a function of cell age.
Appl. Environ. Microbiol. 74, 2534e2536.
Porro, D., Martegani, E., Ranzi, B.M., Alberghina, L., 1988. Oscillations in continuous
cultures of budding yeast: a segregated parameter analysis. Biotechnol. Bioeng.
32, 411e417.
Porro, D., Vai, M., Vanoni, M., Alberghina, L., Hatzis, C., 2009. Analysis and modeling
of growing budding yeast populations at the single cell level. Cytometry Part A
75A, 114e120.
Powell, C.D., Van Zandyke, S.M., Quain, D.E., Smart, K.A., 2000. Replicative ageing
and senescence in Saccharomyces cerevisiae and the impact on brewing
fermentations. Microbiology 146, 1023e1034.
Powell, C.D., Quain, D.E., Smart, K.A., 2003. The impact of brewing yeast cell age on
fermentation performance, attenuation and flocculation. FEMS Yeast Res. 3,149e157.
Powell, C.D., Quain, D.E., Smart, K.A., 2004. The impact of sedimentation of cone on
yeast heterogeneity. J. Am. Soc. Brew. Chem. 62, 8e17.
Prats, C., López, D., Giró, A., Ferrer, J., Valls, J., 2006. Individual-based modelling of
bacterial cultures to study the microscopic causes of the lag phase. J. Theor. Biol.
241, 939e953.
Prats, C., Giró, A., Ferrer, J., López, D., Vives-Rego, J., 2008. Analysis and IbM simu-
lation of the stages in bacterial lag phase: basis for an updated definition.
J. Theor. Biol. 22, 56e68.
Prats, C., Ferrer, J., Gras, A., Ginovart, M. Individual-based modelling and simulation
of microbial processes: yeast fermentation and multi-species composting.
Math. Compu. Model Dyn. Syst., in press, doi:10.1080/13873954.2010.481809.
Smart, K.A., 1999. Ageing in brewing yeast. Brew. Guardian 128, 19e24.
Strässle, C., Sonnleitner, B., Fiechter, A., 1988. A predictive model for the sponta-
neous synchronization of Saccharomyces cerevisiae grown in continuous culture.
I. Concept. 1988. J. Biotechnol. 7, 299e318.
Swinnen, I.A.M., Bernaerts, K., Dens, E.J.J., Geeraerd, A.H., Van Impe, J.F., 2004.
Predictive modelling of the microbial lag phase: a review. Int. J. Food Microbiol.
94, 137e159.
Takamatsu, T., Shioya, S., Chikatani, H., 1985. Comparison of simple population
models in a baker’s yeast fed-batch culture. Chem. Eng. Sci. 40, 499e507.
Walker, G.M., 1998. Yeast Physiology and Biotechnology. JohnWiley & Sons,
Chichester.