1904

Journal of Food Protection, Vol. 69, No. 8, 2006, Pages 1904–1912

Copyright 䊚, International Association for Food Protection

Microbial Modeling of Alicyclobacillus acidoterrestris CRA 7152

Growth in Orange Juice with Nisin Added
WILMER EDGARD LUERA PEÑA AND PILAR RODRIGUEZ DE MASSAGUER*

Universidade Estadual de Campinas, Faculdade de Engenharia de Alimentos, Departamento de Ciências de Alimentos, CP 6121, CEP 13083-862,
Campinas, São Paulo, Brazil

MS 05-496: Received 30 September 2005/Accepted 10 February 2006

ABSTRACT
The adaptation time of Alicyclobacillus acidoterrestris CRA 7152 in orange juice was determined as a response to pH
(3 to 5.8), temperature (20 to 54⬚C), soluble solids concentration (⬚Brix; 11 to 19 ⬚Brix), and nisin concentration (0 to 70 IU/
ml) effects. A four-factor central composite rotational design was used. Viable microorganisms were enumerated by plating
on K medium (pH 3.7). Two primary models were used to represent growth and adaptation time. A second-order polynomial
model was applied to analyze the effects of factors. Results showed that the Baranyi and Roberts model was better than the
modified Gompertz model, considering the determination coefficient (R2) for experimental data description. Inhibition of
bacteria can be obtained through several studied combinations for at least 47 days of storage. The shortest period of adaptation
was observed between 37 to 45⬚C, with pHs between 4 and 5, yet the longest periods of adaptation could be obtained around
20⬚C with pHs close to 3.0. Statistical analysis of the quadratic model showed that the adaptation time increased as temperature
or pH decreased, and as nisin concentration or soluble solids increased. The model showed that adaptation time has a minimum
value for juice without nisin added, with 13.5% soluble solids, pH 5.0, and incubated at 43.8⬚C. The statistical parameters
that validated this model were an R2 of 0.816, a bias factor of 0.96, and an accuracy factor of 1.14. Manipulation of more
than one factor, as well as the use of an antimicrobial agent, can be an alternative to preventing the development of A.
acidoterrestris in orange juice, thus contributing to increased orange juice shelf life.

The first report on contamination caused by an acido-
philic spore-forming microorganism occurred in apple juice
aseptically packed in Germany in 1982 (5). That microor-
ganism was a strict aerobic, heat-resistant bacteria, able to
survive the thermal treatments normally applied during fruit
juice processing. Since then, the concept of contamination
in acid fruit juices has been changing, and the destruction
of these bacteria has become the priority during preserva-
tion of the product.
Alicyclobacillus acidoterrestris is a thermoacidophilic,
nonpathogenic, spore-forming bacteria that has been isolat-
ed and identified from forest soil as well as from several
fruit juices pasteurized and contaminated, such as orange
(23), apple, and passion fruit (20, 24, 31, 34) juices. Ali-
cyclobacillus causes an off-flavor in fruit juices; the flavor
contaminant was identified as medicinal (35). The chemical
compounds responsible for this flavor are guaiacol (31) and
phenolic compounds (16). Guaiacol and halophenols were
identified as the offensive-smelling agent in many incidents
of Alicyclobacillus spp.–related spoilage. Although the ex-
act formation pathway of these off-flavors are not known,
studies report that the presence of Alicyclobacillus spp. in
the medium may be a major contributor to their formation
(7). This microorganism grows at temperatures between 25
to 60⬚C (26) at pH levels of 3.0 to 5.5 (25). Current studies
on thermal inactivation (12, 20, 21, 29) show that a con-
* Author for correspondence. Fax: 55-19-32894966; E-mail:
esteril@unicamp.br.
ventional thermal pasteurization treatment applied to fruit
juices might not be sufficient to inactivate Alicyclobacillus
spores. Komitopoulou et al. (17) and Yamazaki et al. (37)
studied the use of nisin by means of the MIC method in
controlling the spore germination of A. acidoterrestris
spores and found that the inhibitory effect on spores de-
pended on temperature and pH.
Microbiologists consider the duration of lag phase (␭)
and the specific growth rate coefficient (␮max) as the main
parameters that characterize the bacterial growth curve (3).
Several models have been proposed to estimate these pa-
rameters. Many authors emphasize that difficulties remain
in the estimation of ␭ (1, 19) because of the lack of knowl-
edge of the physiological and biological mechanisms that
underlie the lag phase. Only a few authors were able to put
this information into a model for calculating this parameter
(3).
The goals for our study were to model the growth and
survival of Alicyclobacillus acidoterrestris spores in orange
juice by means of the Baranyi and Roberts and modified
Gompertz models, and to model Alicyclobacillus acidoter-
restris adaptation time in orange juice as a function of pH,
soluble solids concentration, incubation temperature, and
nisin concentration by the response surface polynomial
model.
MATERIALS AND METHODS
Bacterial strain and culture medium. Alicyclobacillus aci-
doterrestris CRA 7152 strain and nisin were provided by Danisco
J. Food Prot., Vol. 69, No. 8 PREDICTIVE MICROBIOLOGY OF ALICYCLOBACILLUS 1905

Cultor. Sporulation agar was made up of Alicyclobacillus acido- TABLE 1. Experimental design for A. acidoterrestris CRA 7152
caldarius medium (21): 0.05% MnCl24H2O; 1.5% agar; 1.0 g of in concentrated orange juice using response surface methoda
yeast extract (Oxoid, Basingstoke, UK); 0.2 g of (NH4)2SO4; 0.5
Variable:
g of MgSO47H2O; 0.25 g of CaCl22H2O; 0.60 g of KH2PO4; 1.0
g of glucose (Oxoid); and 1,000 ml of water. pH was adjusted to Expt T pH Ni ⬚Brix
4.0.
The quantification medium K was as follows: peptone, 5 g N1 28.5 3.7 17.5 13
(Oxoid); glucose, 1 g (Oxoid); yeast extract, 2.5 g (Oxoid); Tween N2 45.5 3.7 17.5 13
80, 1 g (Synth, São Paulo, Brazil); agar, 15 g (Difco, Detroit, N3 28.5 5.1 17.5 13
Mich.); and distilled water, 1,000 ml. Medium was sterilized at N4 45.5 5.1 17.5 13
121⬚C for 15 min and pH adjusted to 3.7 with malic acid (Vetec, N5 28.5 3.7 52.5 13
Rio de Janeiro, Brazil) at 25%. The malic acid was sterilized by N6 45.5 3.7 52.5 13
filtration (34). N7 28.5 5.1 52.5 13
N8 45.5 5.1 52.5 13
Spore suspension preparation. Alicyclobacillus acidoter- N9 28.5 3.7 17.5 17
restris cells were initially grown in four slant tubes containing N10 45.5 3.7 17.5 17
potato dextrose agar, pH 5.6 (Oxoid) and incubated at 44⬚C for 3 N11 28.5 5.1 17.5 17
days. The result of growth was then collected from the tubes by N12 45.5 5.1 17.5 17
scraping the bottle with sterile glass rods with 5 ml sterile distilled N13 28.5 3.7 52.5 17
water per tube. The resulting cell suspension was transferred to a N14 45.5 3.7 52.5 17
sterile screw-cap tube (25 by 200 mm) and activated at 80⬚C for N15 28.5 5.1 52.5 17
10 min, followed by rapid cooling in an ice bath until room tem- N16 45.5 5.1 52.5 17
perature was reached. A portion (0.1 ml) of activated suspension N17 20 4.4 35 15
was inoculated on each of 100 glass bottles (290 ml) containing N18 54 4.4 35 15
60.0 ml of solidified and slanted A. acidocaldarius medium (38). N19 37 3 35 15
These 100 inoculated bottles were incubated for 9 days at 45.0⬚C. N20 37 5.8 35 15
Spores were collected after 90% microscopically confirmed spor- N21 37 4.4 0 15
ulation (21). Collected spores were washed and resuspended in N22 37 4.4 70 15
sterile distilled water after three centrifugations (12,310 ⫻ g for N23 37 4.4 35 11
15 min at 4⬚C), followed by alternate washing. Lysozyme at 0.3 N24 37 4.4 35 19
mg/ml suspension was added after the first washing, after pH ad- N25 37 4.4 35 15
justment to 11 for disruption of vegetative cells (33). Spore sus- N26 37 4.4 35 15
pension was stored at 4⬚C in sterile distilled water until use. Vi- N27 37 4.4 35 15
able spores were enumerated by pour plating on K medium after
thermal activation for 10 min at 80⬚C. The inverted plates were a Cube points (N1 to N16), axial points (N17 to N24), central
incubated at 43⬚C for 5 days. Concentration of the spore suspen- point in triplicate (N25 to N27). T, temperature (⬚C); Ni, nisin
sion was 8 ⫻ 108 spores per ml. concentration (IU/ml); ⬚Brix, soluble solids concentration.

Experimental design. The effect of four factors (pH, tem-

perature, nisin concentration, and soluble solids concentration) on
the adaptation time of Alicyclobacillus acidoterrestris CRA 7152
in concentrated orange juice was determined by response surface
methods. This method is commonly used in predictive microbi-
ology for secondary modeling. Factor ranges were as follows: pH,
3 to 5.8; temperature, 20 to 54⬚C; soluble solids concentration
(⬚Brix), 11 to 19 ⬚Brix; and nisin concentration, 0 to 70 IU/ml. A
four-factor, fractional central composite design, rotational experi-
mental plan (27 experiments), with cube points (experiments N1
to N16), axial points (experiments N17 to N24), and central point
in triplicate (experiments N25 to N27) was used (Table 1). The
repetitions of the central point have two purposes: to supply a
measure of the pure error and to stabilize the variance of the
predicted response (4, 22).
Orange juice pH was adjusted with 5 N NaOH and 25%
malic acid (wt/vol) and measured with a potentiometer (DMPH-
2 Digimed). Malic acid was used for adjusting pH because this is
the main acid component of orange juice, it being at an even
higher concentration than citric acid (13). The various concentra-
tions of soluble solids were adjusted by adding different amounts
of sterile, distilled water to the concentrated orange juice and mea-
sured with an ATAGO HSRO500 refractometer. All samples were
treated at 105⬚C for 10 min to eliminate possible competitors (18).
Nisin was used after a storage solution was prepared containing
104 IU/ml in 0.02 N HCl, sterilized at 121⬚C for 15 min (28). The
initial inoculated microbial load of A. acidoterrestris was 2.0 ⫻
102 spores per ml of orange juice, activated at 80⬚C for 10 min.
For each experimental unit listed in Table 1, 100 ml of orange
juice samples was prepared in sterile glass bottles. The samples
were incubated until growth reached a stationary phase, as verified
by stabilization of the colony count. Each experiment was per-
formed in duplicate and was monitored daily by pour plating in
K medium (pH 3.7), followed by incubation at 43⬚C for 5 days.
Analysis of growth data. Experimental growth data were
adjusted to the Baranyi and Roberts (2) equations as well as mod-
ified Gompertz equations (14), through the use of Baranyi’s
DMFIT program. In this manner, the following bacterial growth
kinetic parameters were estimated: adaptation time, ␭ (days); max-
imum growth rate, ␮ {[log(N/ml)]/day}; and maximum popula-
tion, ␬. ␭ takes into consideration timing for the spore to germi-
nate and the start of cellular reproduction, demonstrated by an
increase in CFU.
The Baranyi and Roberts function. The Baranyi and Rob-
erts (2) function includes substrate and intracellular conditions of
inoculated microorganism, which are important criteria in evalu-
ating the influences of several factors that affect growth (pH, tem-
perature, nisin concentration, and soluble solids), causing changes
in medium and in microorganism metabolism. This function anal-
yses microorganism behavior in a more comprehensive way than
does the Gompertz equation:
1906 PEÑA AND DE MASSAGUER J. Food Prot., Vol. 69, No. 8

y(t) ⫽ y0 ⫹ ␮max An (t) ⫺

1
m [
ln 1 ⫹
e m ␮ max A n (t) ⫺ 1
e m(y max ⫺y 0 ) ] (1)

where y(t) is the natural logarithm of the cell concentration at time

t, and ymax is the natural logarithm of the maximum cell concen-
tration.
The function A(t) plays the role of a gradual delay in time:
ln(e ⫺␮ max t ⫹ e ⫺y0 ⫺ e ⫺v t ⫺y 0 )
A(t) ⫽ t ⫹ (2)
␮max
where
y0 ⫽ ⫺ln ␣0 (3)
The parameter ␣0 is called physiological state of cell when t is
t0; m is the parameter related to curvature after exponential phase.
Modified Gompertz function. The modified Gompertz func-
tion (14) model uses a sigmoid curve to represent growth as a
slightly curvilinear exponential phase and an initial ascending seg-
ment for the lag phase that does not represent the initial path of
microbial growth:
y(t) ⫽ a ⫹ c·exp{⫺exp[⫺b(t ⫺ M)]} (4)
where y(t) is the log of population at time t (CFU per milliliter),
a is the log(N0), c is the log(N ⫺ N0) in the final lag phase, b is
the slope corresponding to the growth rate, M is the time in which
rate of absolute growth is maximum, and t is time.
Response surface polynomial model The quadratic model
was used to describe the influence of pH, temperature, nisin con-
centration, and soluble solids during adaptation time. The follow-
ing model was implemented:
␭ ⫽ b0⫹ b1 T ⫹ b2 T 2 ⫹ b3 pH ⫹ b4 pH 2 ⫹ b5 Ni
⫹ b6 Ni 2 ⫹ b7 Brix ⫹ b8 Brix 2 ⫹ b9 T·pH ⫹ b10 T·Ni
⫹ b11 T·Brix ⫹ b12 pH·Brix ⫹ b13 pH·Ni
⫹ b14 Brix·Ni ⫹ ␧ (5)
where ␭ is the adaptation time (days), T is the incubation tem-
perature (⬚C), Ni is the nisin concentration (IU per milliliter),
⬚Brix is the concentration of soluble solids, b0 . . . b14 are the
coefficients in this model, and ␧ is the error. Analysis of experi-
mental data was performed by the software Statistica 5.0. The
given model was statistically validated through analysis of vari-
ance. Bias and accuracy factors were calculated as described by
Ross (27).
RESULTS AND DISCUSSION
Modeling growth through primary models. On the
basis of DMFIT program, and by using the log of the pop-
ulation per milliliter versus time as entry data in experi-
ments where growth occurred, growth kinetic parameters
were calculated. Input parameters of the Baranyi and Rob-
erts model, such as m (curvature parameter characterizing
the transition of the growth curve to the stationary phase)
and n (curvature parameter after the lag phase), were se-
lected by iteration (between 1 and 10) until best data ad-
justment. Figure 1 represents growth curves for experi-
ments related to axial and central points of the experimental
plan (Table 1). Axial points consider the extreme value of
each factor, whereas other factors remain on a central level.
Experiment N17 was not represented because microbial
growth was not observed.
FIGURE 1. Curves adjusted by the Baranyi and Roberts model
and growth data of A. acidoterrestris CRA 7152 for experiments
corresponding to axial and central points. 䉱, pH 4.4, 37⬚C, nisin
concentration, 0 IU/ml, 15 ⬚Brix; ⽧, pH 5.8, 37⬚C, nisin concen-
tration, 35 IU/ml, 15 ⬚Brix; ⽧, pH 4.4, 37⬚C, nisin concentration,
35 IU/ml, 15 ⬚Brix; 䢇, pH 4.4, 37⬚C, nisin concentration, 35 IU/
ml, 11 ⬚Brix; 䡬, pH 4.4, 37⬚C, nisin concentration, 70 IU/ml, 15
⬚Brix; ⫹, pH 4.4, 37⬚C, nisin concentration, 35 IU/ml, 19 ⬚Brix;
∗, pH 3, 37⬚C, nisin concentration, 35 IU/ml, 15 ⬚Brix; ⫻, pH
4.4, 54⬚C, nisin concentration, 35 IU/ml, 15 ⬚Brix.
Experimental data from experiments N20, N23, and
N25 were well adjusted by the Baranyi model, which dif-
fered from the results of experiments N18, N19, N22, and
N24, the counts of which do not represent characteristic
growth, but rather growth inhibition. In these four experi-
ments, no cell replication was observed, but a slight de-
crease of the inoculated spore load was. Each factor in its
extreme value was determined to inhibit growth. We thus
obtained an adaptation time of at least 47 days for the test
conditions of data sets N19, N22, and N24. On the other
hand, for data set N18, this adaptation time was 27 days.
An important comparison can be established between this
experiment and data set N23 because the increase in tem-
perature from 37 to 54⬚C and soluble solids from 11 to 15
⬚Brix caused a decrease in the microbial population level
after 27 days (experiment N18), contrary to results of ex-
periment N23, which saw an increase in population. A com-
bination of adverse temperature (54⬚C) and 35 IU of nisin
per ml, as in experiment N18, caused the largest population
decrease.
The variation of soluble solids concentration from 11
to 15 ⬚Brix can only be inhibitory to microbial growth for
a certain time, but the increase of temperature from 37⬚C
to 54⬚C, together with nisin addition, may be provoking
death of vegetative cells to some degree, resulting in re-
duction of the microbial population. In the case of experi-
ment N21, in the absence of nisin, a short adaptation time
was observed. Its data was only adjusted by the Baranyi
model; the Gompertz model did not fit the data. Walls and
Chuyate (34) noted faster growth at pHs of 4.5 and 5.0 than
at 3.5 at 35⬚C. On the other hand, the results of Pinhatti et
al. (25) showed an optimum pH around 3.5 to 5.0 for the
growth of these bacteria. This optimum range depends on
J. Food Prot., Vol. 69, No. 8 PREDICTIVE MICROBIOLOGY OF ALICYCLOBACILLUS 1907

TABLE 2. Parameters of A. acidoterrestris CRA 7152 growth in concentrated orange juice calculated by the Baranyi and Roberts and
Gompertz modelsa
Adaptation time Growth rate Maximum population
Expt Model (days) ([log CFU/ml]/day) (log CFU/ml) R2 b

2 Observed 24 ⫾ 1.0
Gompertz 23.3 ⫾ 1.1 0.0735 ⫾ 0.018 2.73 ⫾ 0.012 0.92
Baranyi 23.7 ⫾ 0.7 0.1038 ⫾ 0.09 2.73 ⫾ 0.012 0.92
3 Observed 25 ⫾ 1.5
Gompertz 25.8 ⫾ 1.9 0.0225 ⫾ 0.009 2.70 ⫾ 0.088 0.76
Baranyi 29 ⫾ 1.8 0.0352 ⫾ 0.004 2.86 ⫾ 0.077 0.92
4 Observed 9 ⫾ 0.5
Gompertz 12.4 ⫾ 0.7 0.452 ⫾ 0.102 5.99 ⫾ 0.70 0.98
Baranyi 11.3 ⫾ 0.7 0.345 ⫾ 0.09 5.94 ⫾ 0.61 0.99
6 Observed 31 ⫾ 1.5
Gompertz 21.1 ⫾ 1.3 0.038 ⫾ 0.004 2.65 ⫾ 0.04 0.63
Baranyi 17.4 ⫾ 1.4 0.037 ⫾ 0.003 2.78 ⫾ 0.03 0.77
8 Observed 17 ⫾ 2.1
Gompertz 14 ⫾ 1.7 0.229 ⫾ 0.48 6.13 ⫾ 0.07 0.98
Baranyi 12.8 ⫾ 1.6 0.191 ⫾ 0.035 5.97 ⫾ 0.14 0.98
10 Observed 28 ⫾ 2.0
Gompertz 26.4 ⫾ 1.5 0.03 ⫾ 0.008 3.71 ⫾ 0.13 0.79
Baranyi 32.7 ⫾ 2.0 0.09 ⫾ 0.029 2.97 ⫾ 0.10 0.94
12 Observed 26 ⫾ 2.0
Gompertz 21.1 ⫾ 3.6 0.04 ⫾ 0.027 2.95 ⫾ 0.069 0.90
Baranyi 25.2 ⫾ 1.6 0.07 ⫾ 0.007 2.94 ⫾ 0.062 0.96
16 Observed 36 ⫾ 1.0
Gompertz 26.1 ⫾ 0.4 0.035 ⫾ 0.005 2.95 ⫾ 0.026 0.94
Baranyi 29.0 ⫾ 0.1 0.053 ⫾ 0.002 2.84 ⫾ 0.002 0.96
20 Observed 18 ⫾ 1.0
Gompertz 16.4 ⫾ 1.0 0.151 ⫾ 0.014 5.01 ⫾ 0.06 0.95
Baranyi 15.1 ⫾ 0.1 0.126 ⫾ 0.003 4.84 ⫾ 0.01 0.96
21 Observed 1 ⫾ 0.0
Gompertzc — — — —
Baranyi ⬍1 1.00 ⫾ 0.0128 6.99 ⫾ 0.12 0.97
23 Observed 27 ⫾ 1.0
Gompertz 25.4 ⫾ 0.8 0.158 ⫾ 0.058 3.40 ⫾ 0.30 0.98
Baranyi 27.4 ⫾ 1.86 0.180 ⫾ 0.133 3.40 ⫾ 0.01 0.98
25 Observed 16 ⫾ 1.5
Gompertz 14.5 ⫾ 1.7 0.07 ⫾ 0.069 3.56 ⫾ 0.08 0.95
Baranyi 16.2 ⫾ 1.2 0.16 ⫾ 0.075 3.58 ⫾ 0.04 0.96
a Data include only experiments where growth was evident. Values are expressed as mean ⫾ SD (n ⫽ 2).
b Determination coefficient.
c Model without fit.

both strain and growth medium. This study was conducted
with only one strain, A. acidoterrestris CRA 7152, recog-
nized as guaiacol producer (9), because not all strains are
deteriorative. Therefore, these results are applied only to
the conditions studied for deteriorative strains. Adaptation
time results for data sets N23 and N20 were 27.4 and 16
days, respectively, and for experiment N25, it was 16.2
days. These values of adaptation time were calculated by
the Baranyi and Roberts model, which offers a better ad-
justment of experimental data in relation to the modified
Gompertz model (Table 2). Some differences were ob-
served in the final population detected for several experi-
ments, which results from the effect of the analyzed factors.
Table 2 shows the standard deviation for estimated param-
eters. Figure 2 shows the effect of pH and nisin concentra-
tion in growth curves for experiments performed at 45.5⬚C
and 13 ⬚Brix.
Data presented in Figure 2 indicate growth in four data
sets (N2, N4, N6, N8), at 45⬚C, which is considered to be
the optimum growth temperature (6). The influence of pH
and nisin concentration can be noted. Furthermore, the ex-
periments show a longer adaptation time (31 days for ex-
periment N6) for the combination of lower pH (3.7) and
higher nisin concentration (52.5 IU/ml). On the other side,
shorter adaptation time (11 days; data set N4) can be related
to a pH of 5.1 (higher value) and 17.5 IU/ml (low nisin
concentration), which is insufficient to inhibit growth.
Delves-Broughton (10) showed strong decrease of nisin ac-
tivity during the storage at high pH and temperature values.
The final maximum population was also influenced by this
1908 PEÑA AND DE MASSAGUER
FIGURE 2. Growth curves of A. acidoterrestris CRA 7152 in con-
centrated orange juice: 13 ⬚Brix and 45.5⬚C. ⫻, pH 5.1, nisin
concentration, 17.5 IU/ml; ⽧, pH 5.1, nisin concentration, 52.5
IU/ml; 䢇, pH 3.7, nisin concentration, 17.5 IU/ml; ⫹, pH 3.7,
nisin concentration, 52.5 IU/ml.
factor, which was higher at pH of 5.1 than at pH 3.7. By
using these growth models, it is possible to estimate min-
imum time so that the microbial population reaches 104
CFU/ml, the level in which the presence of guaiacol can
be detected (11, 24). Thus, for data sets (45.5⬚C, pH 5.1,
17.5 IU/ml, and 13% soluble solids), (45.5⬚C, pH 5.1, 52.5
IU/ml, and 13% soluble solids), (37⬚C, pH 5.8, 35 IU/ml,
and 15% soluble solids), and (37⬚C, pH 4.4, 0 IU/ml, and
15% soluble solids), 16.5 days, 23 days, 30 days, and 1
day would be sufficient to reach this level of bacterial pop-
ulation in juices with an initial contamination of 102 CFU/
ml, which is commonly found in orange juice concentrate.
In the case of samples with 17% soluble solids incu-
bated at 45.5⬚C, adaptation time was longer than those sam-
ples at 13% soluble solids—an indication of the effect of
reduced water activity (Fig. 3). For curves in Figure 3, the
Baranyi and Roberts model was also better (R2 between
0.94 and 0.96) at describing the experimental data than the
modified Gompertz model was (R2 between 0.79 and 0.94),
as confirmed by experimental and estimated data. The mod-
ified Gompertz model does not have a lag phase parallel to
the abscise, and the equation does not present a linear in-
crement during the exponential phase (36). The Baranyi and
Roberts model is a mechanistic model and assumes that the
kinetics of a homogeneous bacterial population can be char-
acterized by extracellular, physicochemical, and environ-
mental conditions. All the extracellular physicochemical
quantities are considered to be independent of growth, and
they are represented in an environmental vector variable.
The intracellular environment represents the physiological
state of the bacterial population for critical substances dur-
ing growth (such as RNA or ATP). These characteristics
may explain the advantages of this model over other sig-
moid functions.
Baty and Delignette-Muller (3) showed that the Bar-
anyi model (2) adjusted better than the modified Gompertz
model and the lag exponential models. Nevertheless, Gib-
son et al. (14) showed that both models (Gompertz and
J. Food Prot., Vol. 69, No. 8
FIGURE 3. Growth curves of A. acidoterrestris CRA 7152 in con-
centrated orange juice: 17 ⬚Brix and 45.5⬚C. ⽧, pH 3.7, nisin
concentration, 17.1 IU/ml; 䉱, pH 5.1, nisin concentration, 17.1
IU/ml; 䢇, pH 5.1, nisin concentration, 52.5 IU/ml; ⽧, pH 3.7,
nisin concentration, 52.5 IU/ml
Baranyi) represent good adjustment to describe bacterial
growth curves.
Data presented in Figure 3 show that only experiment
N14 did not produce an increase of the initially inoculated
microbial load. Therefore, by using the same optimal con-
ditions of growth temperature, it was possible to block the
postgerminative development and further duplication of the
microorganisms, with a tendency for numbers to decrease
after 47 days of incubation. Komitopoulou et al. (17)
showed that when 50 IU of nisin per ml is used in single-
strength orange juice, little growth occurred, with approx-
imately double the initially inoculated load (4 ⫻ 102 CFU/
ml) at 5 days of incubation at 44⬚C. However, this bacteria
was inhibited by 52.5 IU/ml of nisin in juice, pH 3.7, and
17 ⬚Brix for up to 47 days of incubation at 45.5⬚C. This is
a relevant result for orange juice transported in abusive
temperature conditions of up to 45.5⬚C.
Comparison of data presented in Figures 2 and 3 in-
dicates that at 17 ⬚Brix there was not a large increase in
final population, in contrast to the two experiments at 13
⬚Brix (experiments N4 and N8), showing influence of the
studied factors. These experiments imply that longer ad-
aptation time periods can be achieved than those verified
at 13 ⬚Brix. Sinigaglia et al. (30) showed stronger viability
at high values of water activity. Considering a high value
of soluble solid concentration (17 ⬚Brix), a long adaptation
time was observed, even at temperature and pH adequate
for growth. Thus, experiment N12 showed a result of 25.2
days of adaptation, yet for experiments N10 and N16, ad-
aptation time was up to 1 month.
In samples incubated at 28.5⬚C, there was only growth
in experiments N3 (pH 5.1, nisin 17.5 IU/ml, and 13 ⬚Brix)
and N11 (pH 5.1, nisin 17.5 IU/ml, and 17 ⬚Brix) (Fig. 4).
Observed adaptation times were long, from up to 36 days
for experiment N11 and up to 29 days for experiment N3.
In other experiments where the same temperature was used,
there was no increase in microbial numbers, and in one
J. Food Prot., Vol. 69, No. 8 PREDICTIVE MICROBIOLOGY OF ALICYCLOBACILLUS 1909

TABLE 3. Experimental adaptation time of A. acidoterrestris

CRA 7152 used for adjustment to the quadratic model
Expt Adaptation time (days)

1 47
2 23.7
3 29
4 9
5 25
6 31
7 47
8 12.8
9 47
10 32.7
11 36
12 25.2
FIGURE 4. Curves and growth data of A. acidoterrestris CRA 13 47
7152 at different values of pH, ⬚Brix, and nisin concentration 14 47
incubated at 28.5⬚C. ⽧, pH 5.1, nisin concentration, 17.5, 13 15 47
⬚Brix; 䡵, pH 5.1, nisin concentration, 17.5, 17 ⬚Brix; ⫻, pH, 3.7, 16 29
nisin concentration, 17.5, 13 ⬚Brix; 䡵, pH 3.7, nisin concentra- 17 47
tion, 52.5, 13 ⬚Brix; ∗, pH 5.1, nisin concentration, 52.5, 13 ⬚Brix; 18 27
䡬, pH 3.7, nisin concentration, 17.5, 17 ⬚Brix; #, pH 3.7, nisin 19 47
concentration, 52.5, 17 ⬚Brix; 䉭, pH 5.1, nisin concentration, 20 15.1
52.5, 17 ⬚Brix. 21 1
22 47
23 27.4
case, we observed a decrease (experiment N5). Thus, we 24 47
took into consideration 47 days of adaptation time for ex- 25 16.2
periments N1, N7, N9, N13, and N15, as well as 25 days 26 16.2
for experiment N5, which from that time on presented a 27 14.7
clear reduction of population. Several authors (11, 24, 35)
have indicated minimum growing temperatures of between
20 to 25⬚C for this microorganism. Nevertheless, different well as some of the interactions: temperature·pH, temper-
pH combinations, together with relatively low nisin levels ature·nisin, temperature·⬚Brix, pH-nisin, and nisin·⬚Brix.
and concentration of soluble solids, resulted in inhibition at Equation 6 presents the quadratic model achieved:
28.5⬚C for up to 47 days of incubation. ␭ ⫽ 15.70 ⫺ 6.34542T ⫹ 5.53635T 2 ⫺ 5.30458pH
In all data curves, a certain lag period could be ob-
served before growth or decrease of microbial population, ⫹ 4.04885pH 2 ⫹ 5.26292Ni ⫹ 2.28635Ni 2
indicating that there was a decrease in nisin activity during ⫹ 5.13792Brix ⫹ 5.58635Brix 2 ⫺ 3.06813T·pH
the storage period. This observation was more apparent at
high pH and temperature, demonstrating the bacteriostatic ⫹ 1.24213T·Ni ⫹ 1.60562T·Brix ⫹ 2.19438pH·Ni
activity of this preservative, which was also confirmed by ⫹ 1.51813Ni·Brix (6)
Komitopoulou et al. (17) and Yamazaki et al. (37).
Modeling of adaptation time through quadratic TABLE 4. Relevant regression coefficients of the polynomial
model. To model the response surface, we used data related model for A. acidoterrestris adaptation time
to adaptation time (Table 3) that were observed and esti- Variable Coefficient Probability
mated from the adjustment of growth data to primary mod-
els. In experiments where the microbial population did not Avg 15.7 0.001013
Temp ⫺6.34542 0.000775
increase during 47-day incubation time, this was considered
Temp2 5.53635 0.001145
to be the adaptation time. This assumption suggests that in pH ⫺5.30458 0.001109
those experiments, the adaptation time was at least 47 days, pH2 4.04885 0.002138
and that the adjusted model would be valid to predict up Nisin 5.26292 0.001126
to this period, although the real adaptation time is longer. Nisin2 2.28635 0.006658
Data related to adaptation time (days) as a function of Brix 5.13792 0.001182
pH, soluble solids, nisin concentration, and temperature Brix2 5.58635 0.001125
main effects and their interactions were adjusted by nonlin- Temp·pH ⫺3.06813 0.004943
ear regression to the model of equation 5. Table 4 indicates Temp·nisin 1.24213 0.029019
adjustment coefficients with a probability level of 0.05. It Temp·Brix 1.60562 0.017701
is evident that all variables studied were relevant (P ⬍ pH·nisin 2.19438 0.009595
Nisin·Brix 1.51813 0.019739
0.05), be it at the linear level be at the quadratic model, as
1910 PEÑA AND DE MASSAGUER J. Food Prot., Vol. 69, No. 8

FIGURE 6. Observed versus predicted values of adaptation time

FIGURE 5. Pareto graph to evaluate the effect of parameters on of A. acidoterrestris.
adaptation time of A. acidoterrestris CRA 7152 in concentrated
orange juice.
by Splittstoesser et al. (32), who demonstrated inhibition of
Alicyclobacillus growth when soluble solid concentration
The Pareto graphic shown in Figure 5 presents stan- was over 18.5 ⬚Brix. They was also observed optimum
dardized effects of the different parameters from the model growth temperatures of approximately 37 to 45⬚C (shortest
(equation 6). The standardized effect was defined as the adaptation time), indicating the mesophilic trends of this
estimated coefficient divided by the standard error (bi/sei). strain. Cerny et al. (6) found similar behavior for the DSM
Any other effect that exceeded the vertical line (P ⫽ 0.05) 2498 strain in orange juice with good growth at 30, 35, and
was considered relevant. Thus, the following factors 45⬚C. In Figure 7b, the influence of nisin concentration is
showed significant effects (in decreasing order of impor- clear in relation to pH. Delves-Broughton (10) indicate that
tance): T, pH, ⬚Brix2, nisin, T2, ⬚Brix, pH2, T·pH, nisin2, at pH 3 to 4, nisin had higher antimicrobial activity, and
pH·nisin, T·⬚Brix, nisin·⬚Brix, T·nisin, and pH·⬚Brix. longer adaptation times can thus be achieved in this range.
The analysis of variance for the response variable in-
dicated that the model was significant (F ⫽ 4.25, R2 ⫽ Quadratic model verification. The model was verified
0.816). Bias and accuracy factors were 1.07 and 1.28, re- by conducting eight experiments under random combina-
spectively, thus showing a certain degree of deviation be- tion of temperature, pH, nisin concentration, and soluble
tween the prediction shown by the model and the experi- solids (Table 5). All the experiments verified that the model
mental values. This deviation was up to an average of 28% was valid. Bias factor (0.96) indicated safe predictions, and
(accuracy factor). Graham et al. (15) also used the quadratic accuracy factor (1.14) showed deviation of only 14%. Thus,
model to investigate the effect of temperature, pH, and so- the results of experiments 1, 2, 3, 4, 6, and 8 indicated that
dium chloride on lag time of nonproteolytic Clostridium the model prediction ‘‘fail-safe’’ and experiments 5 and 7
botulinum, obtaining an R2 value of 0.856. Delignette-Mull- suggested a slight overprediction by this model. These ex-
er et al. (8) reviewed 755 cases of prediction found in lit- periments (Table 5) show that the use of nisin in condition
erature, of which 468 used the polynomial model and 287 of temperature abuse can help to increase shelf life of or-
square root models. They concluded that mean errors were
high for polynomial models describing generation and ad-
aptation times, with mean errors at about 25% and a slight TABLE 5. Verification trials of response surface model for ad-
aptation time of A. acidoterrestris CRA 7152 in orange juice
tendency to the unsafe side. Figure 6 shows a graphic com-
parison between predicted values versus observed values. Predicted Observed
The predictive model did not well adjust the results of ex- Nisin Soluble adaptation adaptation
Trial Temp concn solids time time
periments where an adaptation time of 47 days was used. no. (⬚C) pH (IU/ml) (⬚Brix) (days)a (days)b
This adaptation time represented only the final incubation
time, without increase or decrease of the microbial popu- 1 35 4.0 10 13 23.7 26.1
lation. This fact may be responsible for deviation in the 2 35 4.5 10 13 17.0 19.0
relations between the dependent and independent variables; 3 35 4.0 50 13 25.8 26.0
or it may have caused a lack of adjustment of the model in 4 35 4.5 50 13 22.6 25.3
these conditions. 5 40 4.0 10 14 17.1 13.6
Figure 7 illustrates the adaptation time response sur- 6 40 4.5 10 14 9.1 11.8
7 40 4.0 50 14 22.5 19.6
faces. As was expected, the main impact was temperature.
8 40 4.5 50 14 18.1 19.1
It is clear from Figure 7a that longer adaptation time was
found in relation to high concentration of soluble solids as a Values predicted by response surface model.
well as low temperatures. A similar observation was made b Experimental values.
J. Food Prot., Vol. 69, No. 8
FIGURE 7. Response surface for adaptation time. (a) ⬚Brix versus temperature, pH 4.4, and nisin concentration of 35 IU/ml. (b) Nisin
concentration versus pH, 15 ⬚Brix, and 37⬚C.
ange juice and prevent growth and spoilage by Alicyclo-
bacillus.
In summary, the use of nisin is an option for control-
ling the growth of Alicyclobacillus acidoterrestris in orange
juice. The use of Baranyi and Roberts’ primary model
showed better adjustment to growth data than the modified
Gompertz model. Those models were used to estimate the
minimum time for the microbial population to reach the 104
CFU/ml level at which the presence of guaiacol can be
detected. Thus, for experiments 45.5⬚C, pH 5.1, 17.5 IU/
ml, and 13 ⬚Brix; 45.5⬚C, pH 5.1, 52.5 IU/ml, and 13 ⬚Brix;
37⬚C, pH 5.8, 35 IU/ml, and 15 ⬚Brix; and 37⬚C, pH 4.4,
0 IU/ml, and 15 ⬚Brix; 16.5 days, 23 days, 30 days, and 1
day, respectively, would be enough to reach this level of
bacterial population, causing product deterioration.
Adaptation time was affected mainly by temperature
(20 to 54⬚C), followed by pH (3 to 5.8), nisin (0 to 70 IU/
ml), and concentration of soluble solids (11 to 19 ⬚Brix).
The following interactions were relevant: pH·nisin, temper-
ature·⬚Brix, nisin·⬚Brix, temperature·nisin, and pH·⬚Brix. In
addition, the linear and quadratic levels of the four variables
were also relevant. Adaptation times of at least 47 days
were found experimentally for the various combinations of
factors: 28.5⬚C, pH 3.7, 17.5 IU/ml, and 17 ⬚Brix; 45.5⬚C,
pH 3.7, 52.5 IU/ml, and 17 ⬚Brix; and 28.5⬚C, pH 5.1, 52.5
IU/ml, and 17 ⬚Brix. On the other hand, by using the model
we implemented, circumstances not used in the experimen-
tal plan can be predicted, as long as they are within the
experimental range. Thus, for juice with 12 ⬚Brix, pH 4.0,
it would be necessary for 54 IU of nisin per ml to be added
to keep it microbiologically stable for up to 47, 38, and 32
days when incubated at temperatures of 25, 30, and 35⬚C,
respectively. Manipulation of more than one factor, as well
as the use of an antimicrobial agent, could be alternatives
to avoid the development of Alicyclobacillus acidoterrestris
in orange juice, thus increasing the product’s shelf life.
ACKNOWLEDGMENTS
We thank Fundação de Amparo à Pesquisa do Estado de São Paulo–
Fapesp for its financial support and Danisco Cultor do Brasil.
PREDICTIVE MICROBIOLOGY OF ALICYCLOBACILLUS 1911
REFERENCES
1. Baranyi, J. 2002. Stochastic modeling of bacterial lag phase. Int. J.
Food Microbiol. 73:203–206.
2. Baranyi, J., and A. T. Roberts. 1995. Mathematics of predictive food
microbiology. Int. J. Food Microbiol. 26:199–218.
3. Baty, F., and M. L. Delignette-Muller. 2004. Estimating the bacterial
lag time: which model, which precision? Int. J. Food Microbiol. 91:
261–277.
4. Bruns, R. E., B. Barros Neto, and J. S. Scarminio. 1995. Planning
and optimization of experiments, 2nd ed. Editora da Unicamp, Cam-
pinas, São Paulo, Brazil.
5. Cerny, G., W. Hennlich, and K. Poralla. 1984. Spoilage of fruit juice
by bacilli: isolation and characterization of the spoiling microorgan-
ism. Z. Lebensm.-Unters.-Forsch. 179:224–227.
6. Cerny, Y. G., H. A. Duong, W. Hennlich, and S. Miller. 2000. Ali-
cyclobacillus acidoterrestris: influence of oxygen content on growth
in fruit juices. Food Aust. 52:289–291.
7. Chang, S. S., and D. H. Kang. 2004. Alicyclobacillus ssp. in the fruit
industry: history, characteristics, and current isolation/detection pro-
cedures. Crit. Rev. Microbiol. 30:55–74.
8. Delignette-Muller, M. L., L. Rosso, and J. P. Flandrois. 1995. Ac-
curacy of microbial growth predictions with square root and poly-
nomial models. Int. J. Food Microbiol. 27:139–146.
9. Delves-Broughton, J. 2001. Personal communication.
10. Delves-Broughton, J. 1990. Nisin and its uses as a food preservative.
Food Technol. 44:100–117.
11. Eguchi, S. Y., G. P. Manfio, E. A. Pinhatti, and S. F. Variane. 2001.
Acidothermophilic sporeforming bacteria (ATSB) in orange juices:
detection methods, ecology, and involvement in the deterioration of
fruit juices. Report of the research project, part III. Fruit Proc. 3:
95–101.
12. Eiroa, V. N. M., A. C. V. Junqueira, and L. F. Schmidt. 1999. Ali-
cyclobacillus in orange juice: occurrence and heat resistance of
spore. J. Food Prot. 62:883–886.
13. Farnworth, E. R., M. Lagacc, R. Couture, V. Yaylayan, and B. Stew-
art. 2001. Thermal processing, storage conditions, and the compo-
sition and physical properties of orange juice. Food Res. Int. 34:25–
30.
14. Gibson, A. M., N. Bratchell, and T. Roberts. 1987. The effect of
sodium chloride and temperature on the rate and extent of growth
of Clostridium botulinum type A in pasteurized pork slurry. J. Appl.
Bacteriol. 62:479–90.
15. Graham, A., D. R. Mason, and M. W. Peck. 1996. Predictive model
of the effect of temperature, pH and sodium chloride on growth from
spores of non-proteolytic Clostridium botulinum. Int. J. Food Mi-
crobiol. 31:69–85.
1912 PEÑA AND DE MASSAGUER
16. Jensen, N., and F. B. Whitfield. 2003. Role of Alicyclobacillos aci-
doterrestris in the development of a disinfectant taint shelf-stable
fruit juice. Lett. Appl. Microbiol. 36:9–14
17. Komitopoulou, E., I. S. Boziaris, E. A. Davies, J. Delves-Broughton,
and M. R. Adams. 1999. Alicyclobacillus acidoterrestris in fruit juic-
es and its control by nisin. J. Food Sci. Technol. 34:81–85.
18. Massaguer, P. R., P. C. Pacheco, M. M. Atarassi, W. L. Peña, A. C.
Gonçalves, N. A. Paula, L. H. Geraldini, L. L. Liossi, M. R. Gag-
liazzi, and V. A. Guerra. 2002. Sensibility and specifity of methods
for Alicyclobacillus detection and quantification: a collaborative
study. Fruit Proc. 11:478–482.
19. McKellar, R. C. 1997. A heterogeneous population model for the
analysis of bacterial growth kinetics. Int. J. Food Microbiol. 36:179–
186.
20. McKnight, I. C. S. 2003. Isolation and identification of Alicycloba-
cillus acidoterrestris from pasteurized passion fruit and pineapple
juices, and determination of heat resistance of spores. Ph.D. thesis.
State Universidade of Campinas, Brazil.
21. Murakami, M., H. Tedzuka, and K. Yamazaki. 1998. Thermal resis-
tance of Alicyclobacillus acidoterrestris spores in different buffers
and pH. Food Microbiol. 15:577–582.
22. Neto, B. B., I. S. Scarminio, and R. E. Bruns. 2002. How to perform
experiments, 2nd ed. Editora da Unicamp, Campinas, São Paulo,
Brazil.
23. Parish, M. E., and R. M. Goodrich. 2005. Recovery of presumptive
Alicyclobacillus strains from orange fruit surfaces. J. Food Prot. 68:
2196–2200
24. Pettipher, G. L., M. E. Osmundson, and J. M. Murphy. 1997. Meth-
ods for the detection and enumeration of Alicyclobacillus acidoter-
restris and investigation of growth and production of taint in fruit
juice and fruit juice–containing drinks. Lett. Appl. Microbiol. 24:
185–189.
25. Pinhatti, M. E. M. C., S. Variani, S. Y. Eguchi, and G. P. Manfio.
1997. Detection of Acidothermophilic bacilli in industrialized fruit
juices. Fruit Proc. 9:350–353.
26. Previdi, M. P., P. Colla, and E. Vicini. 1995. Characterization of
J. Food Prot., Vol. 69, No. 8
Alicyclobacillus, a spore-forming thermophilic acidophilic bacteri-
um. Indust. Conserve 70:128–132.
27. Ross, T. 1996. Indices for performance evaluation of predictive mod-
els in food microbiology. J. Appl. Bacteriol. 81:501–508
28. Scott, V. N., and S. L. Taylor. 1981. Temperature, pH and spore load
effects on the ability of nisin to prevent the outgrowth of C. botu-
linum spores. J. Food Sci. 46:121–125.
29. Silva, F. M., P. Gibbs, C. V. Margarida, and C. L. M. Silva. 1999.
Thermal inactivation of Alicyclobacillus acidoterrestris spores under
different temperature, soluble solids and pH conditions for the design
of fruit processes. Int. J. Food Microbiol. 51:95–103.
30. Sinigaglia, M., M. R. Corbo, C. Altieri, D. Campaniello, D.
D’Amato, and A. Bevilacqua. 2003. Combined effects of tempera-
ture, water activity, and pH on Alicyclobacillus acidoterrestris
spores. J. Food Prot. 66:2216–2221.
31. Splittstoesser, C., Y. Lee, and J. J. Churey. 1998. Control of Alicy-
clobacillus in the juice industry. Dairy Food Environ. Sanit. 18:585–
587.
32. Splittstoesser, D. F., J. J. Churey, and Y. Lee. 1994. Growth char-
acteristics of aciduric sporeforming Bacilli isolated from fruit juices.
J. Food Prot. 57:1080–1083.
33. Stumbo, C. R. 1973. Thermobacteriology in food processing. Aca-
demic Press, London.
34. Walls, I., and R. Chuyate. 1998. Alicyclobacillus—historical per-
spective and preliminary characterization study. Dairy Food Envi-
ron. Sanit. 18:499–503.
35. Walls, I., and R. Chuyate. 2000. Spoilage of fruit juice by Alicyclo-
bacillus acidoterrestris. Food Aust. 52:286–288.
36. Whiting, R. C., and R. L. Buchanan. 1997. Predictive modeling, p.
728–739. In M. P. Doyle, L. R. Beuchat, and T. J. Montwille (ed.),
Food microbiology: fundamentals and frontiers. American Society
for Microbiology, Washington, D.C.
37. Yamazaki, K., M. Murakami, Y. Kawi, N. Inoue, and T. Matsuda.
2000. Use of nisin for inhibition of Alicyclobacillus acidoterrestris
in acidic drinks. Food Microbiol. 17:315–320.
38. Yamazaki, K., H. Teduka, and H. Shinano. 1996. Isolation and iden-
tification of Alicyclobacillus acidoterrestris from acidic beverages.
Biosci. Biotechnol. Biochem. 60:543–545.