Modelling the effect of water activity and temperature on growth rate and aflatoxin production by two isolates of Aspergillus flavus on paddy

Authors

  • W. Mousa,

    1.  Centre of Excellence for Food Safety Research (CEFSR), Faculty of Food Science and Technology, University Putra Malaysia, Serdang, Selangor Darul Ehsan, Malaysia
    Search for more papers by this author
  • F.M. Ghazali,

    1.  Centre of Excellence for Food Safety Research (CEFSR), Faculty of Food Science and Technology, University Putra Malaysia, Serdang, Selangor Darul Ehsan, Malaysia
    Search for more papers by this author
  • S. Jinap,

    1.  Centre of Excellence for Food Safety Research (CEFSR), Faculty of Food Science and Technology, University Putra Malaysia, Serdang, Selangor Darul Ehsan, Malaysia
    Search for more papers by this author
  • H.M. Ghazali,

    1.  Department of Food Science, Faculty of Food Science and Technology, University Putra Malaysia, Serdang, Selangor Darul Ehsan, Malaysia
    Search for more papers by this author
  • S. Radu

    1.  Centre of Excellence for Food Safety Research (CEFSR), Faculty of Food Science and Technology, University Putra Malaysia, Serdang, Selangor Darul Ehsan, Malaysia
    Search for more papers by this author

Farinazleen Mohd. Ghazali, Centre of Excellence for Food Safety Research (CEFSR), Faculty of Food Science and Technology, University Putra Malaysia, 43400 Serdang, Selangor Darul Ehsan, Malaysia. E-mail: farinazleen@putra.upm.edu.my

Abstract

Aims:  This study was conducted to characterize the growth of and aflatoxin production by Aspergillus flavus on paddy and to develop kinetic models describing the growth rate as a function of water activity (aw) and temperature.

Methods and Results:  The growth of A. flavus on paddy and aflatoxin production were studied following a full factorial design with seven aw levels within the range of 0·82–0·99 and seven temperatures between 10 and 43°C. The growth of the fungi, expressed as colony diameter (mm), was measured daily, and the aflatoxins were analysed using HPLC with a fluorescence detector. The maximum colony growth rates of both isolates were estimated by fitting the primary model of Baranyi to growth data. Three potentially suitable secondary models, Rosso, polynomial and Davey, were assessed for their ability to describe the radial growth rate as a function of temperature and aw. Both strains failed to grow at the marginal temperatures (10 and 43°C), regardless of the aw studied, and at the aw level of 0·82, regardless of temperature. Despite that the predictions of all studied models showed good agreement with the observed growth rates, Davey model proved to be the best predictor of the experimental data. The cardinal parameters as estimated by Rosso model were comparable to those reported in previous studies. Toxins were detected in the range of 0·86–0·99 aw with optimal aw of 0·98 and optimal temperature in the range of 25–30°C.

Conclusions:  The influences of aw and temperature on the growth of A. flavus and aflatoxin production were successfully characterized, and the models developed were found to be capable of providing good, related estimates of the growth rates.

Significance and Impact of the Study:  The results of this study could be effectively implemented in minimizing the risk of aflatoxin contamination of the paddy at postharvest.

Introduction

Aspergilli are one of the most widely distributed fungi in nature. This genus is capable of spoiling agricultural products, producing toxic metabolites and causing disease (Klich 2007a). They produce aflatoxins, which are secondary metabolites, which are extremely carcinogenic, hepatotoxic, immunosuppressive and antinutritional contaminants (Williams et al. 2004). These metabolites are known to be produced by several members of Aspergillus species (Klich 2007b). However, Aspergillus flavus and Aspergillus parasiticus are the most common and considered the major producer being responsible for the contamination of several agricultural products (Klich 2007b). In some regions, the risk of acute hepatitis and liver cancer had been correlated with the consumption of cereal products heavily contaminated with aflatoxins (Li et al. 2001).

Generally speaking, reports on the mycotoxin contamination of rice are lesser than those on many other cereal products (Tanaka et al. 2007). However, in the last few years, the number of reports dealing with mycotoxin contamination of rice has been growing (Sales and Yoshizawa 2005a,b; Fredlund et al. 2009; Mazaheri 2009; Reddy et al. 2009; Reiter et al. 2010). According to Gummert et al. (2009), the existing figure on the mycotoxin problem in rice does not reflect the actual situation in the humid tropics where a great amount of low-quality rice is being produced and consumed by local farmers. Recent reports highlighted that up to this time, the infection of rice grains with the aflatoxigenic Aspergillus is still a chronic problem in some regions and that consumers in these countries can be at risk owing to exposure to high levels of aflatoxins (Hussaini et al. 2007; Reddy et al. 2009). In view of the report of Gummert et al. (2009), which underscored that up to 90% of the rice in the world is produced and consumed in Asia where more than 80% of it is being sun-dried, the issue becomes unsurprising. Risk assessment studies found that even at a low level of contamination, rice is the main source for exposure to aflatoxins in the communities where it is consumed as staple food (Park et al. 2005; Park and Kim 2006).

Like is the case with most cereals, mycotoxigenic fungi invasion and sequent contamination of rice with mycotoxins starts in the field, and the scenario becomes worst with the poor postharvest practices. Migration of mycotoxins from the husk and the bran layers to the interior endosperm during the parboiling process of rice has been reported (Dors et al. 2009). On the other hand, milling proved to be able to reduce mycotoxin contamination of cereals, including rice (Scudamore et al. 2003; Castells et al. 2007, 2008). The polishing stage, which consists of removing the bran layers from brown rice by friction, is the key step in this regard (Sales and Yoshizawa 2005b; Castells et al. 2007). Nevertheless, no reports indicated that the milling process can eliminate the mycotoxin contamination completely. Moreover, the growing trend of consuming unpolished brown rice, which is believed to be a healthier alternative to the polished rice, may increase the risk of exposure to aflatoxins (Sales and Yoshizawa 2005b). The contaminated by-products of rice milling can threat animals directly and humans indirectly. A positive correlation between contaminated bovine milk with aflatoxin M and contaminated rice bran used in feed making had been stated in the literature (Nordkvist et al. 2009).

Fungal growth on food commodities is influenced by several intrinsic, extrinsic, implicit and processing factors (Sinha 1995). However, among all, temperature, water activity (aw) and the gas composition are considered as the limiting factors for fungal growth and for mycotoxin production in the postharvest stage (Magan et al. 2003). Within this context, characterizing the marginal and optimal aw levels and temperatures for fungal growth and mycotoxin production is prerequisite for any control strategy intended to deal with mycotoxigenic and spoilage fungi. The imperative need for characterizing the effects of the factors that govern fungal growth during the preharvest and postharvest stages has triggered interest in the application of mathematical approaches to describe and predict the response of moulds to the different growth-affecting factors (Lahlali et al. 2005). The main objectives of the present work were (i) to characterize the effects of aw and temperature on the radial growth rate of two isolates of A. flavus and on the attendant aflatoxin formation on paddy and (ii) to develop several models describing the effects of temperature and aw on the growth of both isolates and to assess the performance of each of these models.

Materials and methods

Fungal isolates

Two aflatoxigenic isolates of A. flavus, DISF15 and DISF10, were isolated from paddy and used in the current study. Both strains were deposited in the Laboratory of Food Safety and Quality, University Putra Malaysia. The aflatoxigenic potential of both isolates was verified on Czapek yeast extract agar using the method of Bragulat et al. (2001). Briefly, three circular agar discs of 5 mm were removed from each growing colony, placed in amber glass vials together with 1·0 ml of methanol and vortexed. After 30 min, the extracts were filtrated with a Millipore® filter (Ø 0·45 μm; Bedford, MA, USA) and finally, 20 μl of the filtrate was injected into an HPLC instrument. The aflatoxins produced by DISF15 and DISF10 after 1 week of incubation were 327 and 473 ng g−1, respectively.

Experimental design

The experimental design corresponded to a full factorial design with two factors (temperature and aw). The incubation temperatures studied were 10, 15, 20, 25, 30, 37 and 43°C, while the aw levels investigated were 0·82, 0·86, 0·89, 0·92, 0·95, 0·98 and 0·99. Six plates per conditions, per isolate were prepared. By the end of the third week of incubation, two plates were randomly selected for aflatoxin analysis.

Preparation of the paddy grains

The paddy grains were prepared and inoculated following the procedure of Ramirez et al. (2006). The dried paddy grains (MR219) were obtained from Bernas (Sekinchan, Malaysia). The initial aw and moisture content of the paddy were 0·69 and 12·79% (dry basis), respectively. The paddy grains were sterilized with 12 kGy of gamma irradiation in the Malaysian Nuclear Agency (Bangi, Malaysia). The sterilized paddy grains (free of mycotoxin contamination) were then stored aseptically at 4°C until needed. The paddy grains were adjusted to the desired aw levels (0·82, 0·86, 0·89, 0·92, 0·95, 0·98 and 0·99) by pouring a suitable volume of sterile distilled water into sterilized flasks containing 400 g of sterilized paddy grains each. The moisture adsorption isotherm developed for this paddy variety was used to estimate the amounts of water added to the grains. To ensure uniform absorption of water, samples were then held for 48 h at 4°C with regular mixing. The actual aw values of the samples were then verified with an Aqualab Series 3 water activity meter (Decagon Devices, Inc., Pullman, WA, USA). About 20 g of the moistened paddy was aseptically transferred to sterile 90-mm Petri dishes to form a monolayer of the hydrated grains. Using sterile cork borer, each plate was centrally inoculated with a circular agar disc of 5 mm width obtained from the edge of the colonies of both isolates, which were formerly cultured on PDA for 10 days at 30°C. To maintain fixed relative humidity level during the 5-week incubation period, the six plates of each treatment and a beaker filled with a mixture of glycerol and water having an aw identical to that of the respective paddy treatment were placed in a sealed plastic container. The containers were incubated at 10, 15, 20, 25, 30, 37 and 43°C.

Assessment of fungal growth

Depending on the growth responses of fungi under the experimental conditions, the growth was measured either daily (fast fungal growth) or at an appropriate time interval (slow fungal growth). For each grown colony, two diameters were measured using a ruler at perpendicular positions until the growing colony reached the edge of the plate or the experiment was terminated. The mean diameter of the four cultured plates was used to estimate the change in colony diameter with time. For aflatoxin analysis under the various experimental conditions, two plates were destructively sampled at random by the end of the third week of incubation and kept at −18°C until analysis.

Modelling fungal growth

With the aim of developing models capable of predicting the fungal growth rate as a function of aw and temperature, the common modelling approach in the field of predictive microbiology, which consists of two levels of modelling known as primary and secondary modelling, was applied. Throughout primary modelling, the maximum colony growth rate was estimated using the flexible Baranyi’s function, which describes the change in colony diameter with time (Baranyi and Roberts 1994) using the expression:

image(1)
image(2)

where y is the colony diameter (mm); y0 is the initial colony diameter (mm); ymax is the maximal colony diameter (mm); μmax is the maximum specific colony growth rate (mm day−1); λ is lag phase (days, time for diameter >5mm); and t is time in days. Dividing μmax by two provides the maximum radial growth rate (Rmax).

For secondary modelling, three suitable models were employed to describe Rmax as a function of temperature and aw. Firstly, the second-order response surface model with Gibson transformation was used. The values of aw were transformed to bw for better hyperbolic fitting, and the variance in the radial growth rate was stabilized using the natural logarithm transformation (Gibson et al. 1994). The mathematical expression of the fitted model was

image(3)
image

where a0, a1, a2, a3, a4 and a5 are coefficients to be estimated using nonlinear regression.

The linear Arrhenius–Davey model (Davey 1989), second model employed, was based on the following formula:

image(4)

where T is expressed in absolute temperature (°K); aw is water activity; and a0, a1, a2, a3 and a4 are coefficients to be fit using nonlinear regression. The last model studied was the Rosso model (Rosso et al. 1995; Rosso and Robinson 2001) which takes the form:

image(5)
image
image

where Tmin, Tmax, aw,min and aw,max represent the values of temperature and aw below and above which growth is not possible, respectively. The Topt and aw,opt are the optimal temperature and aw for growth where Rmax approached its optimal value (Ropt).

Modelling aflatoxin production

Aflatoxin production by each isolate as a function of aw and temperature was described using a second-order polynomial equation. The fitted model had the following form:

image(6)

where aw and T are water activity and temperature, respectively, and a1, a2, a3, a4 and a5 are coefficients to be estimated. The nonlinear regression function of statistica ver. 7.0 (Statsoft, Tulsa, OK, USA) was used to construct the primary model of Baranyi, all secondary models and the polynomial model.

Model validation

The models were validated on data obtained from repeated experiments on paddy using the same temperatures employed for model development and five aw levels: 0·99, 0·97, 0·93, 0·90 and 0·87. Performances of the developed models were assessed using the bias factor (Bf), accuracy factor (Af) and the root mean squared error (RMSE) as proposed by Ross (1996).

image(7)
image(8)
image(9)

where n is the number of observations and p is the number of parameters to be estimated in the model.

Determination of aflatoxins

The contents of the plates sampled for aflatoxin analysis were milled to obtain fine paddy. Subsequently, a 25-g subsample of finely milled paddy and 5 g NaCl were blended with 125 ml MeOH/H2O (80/20 v/v) for 2 min. After filtration of the mixture through Whatman filter paper (Whatman, Maidstone, UK), a 15-ml volume of the extract was transferred to a flask and to which 30 ml of deionized H2O were added. The diluted aliquot was then filtered using a Whatman glass microfiber filter (Whatman GF/B). A 15-ml volume of the filtrated extract was introduced to the immunoaffinity columns (Aflatest; Vicam, Milford, MA, USA) at a flow rate of 1 ml min−1. After two subsequent steps of washing with 10 ml deionized H2O, the aflatoxins were eluted with 1·5 ml MeOH followed by 1·5 ml of deionized H2O at a flow rate of 1 ml min−1. Thereafter, the extracts were collected in glass vials and 20 μl of each extract was directly injected into an HPLC system.

The HPLC system

The HPLC system (Waters 600; Milford, MA, USA) was consisting of Waters™ 600E pump, an auto-sampler (Waters 717), in-line degasser (Waters AF) and fluorescence detector (Waters 2475). The whole system was running under the control of Empower software (Waters). Separation of the aflatoxins was achieved using the isocratic mode with a mobile phase consisting of H2O/MeOH/ACN (54 : 29 : 17, v/v/v) running on C18 analytical column (Purospher® STAR RP-18, 5 μm to 250 × 4 mm; Merck, Darmstadt, Germany). The intensity of signals was enhanced using a photochemical reactor (model PHRED; Aura Industries, New York, NY, USA), which was placed between the LC column and the detector. The fluorescence detector was operated at the emission and excitation wavelengths of 365 nm and of 435 nm, respectively. The limit of detection of the methods was 0·02 ng g−1 for both the aflatoxins B1 and B2.

Results

The growth study

Both isolates failed to grow at the aw of 0·82, regardless of temperature, and at 10°C and 43°C, regardless of aw. However, the isolates were found to behave differently at 15°C where DISF10 was able to grow at minimum aw of 0·92, while DISF15 was not. The growth curves obtained by plotting the extension in colony diameters against time were distinctive of fungal growth where for both strains, a linear pattern of growth was noticed after a lag phase.

Applying the modelling procedure introduced above, the primary model of Baranyi was fitted to the related experimental data to estimate the maximum colony growth rate (μmax) based on tracking the daily increases in colony diameters (mm day−1). The model fit well to the growth curves of both strains and produced R2 values >0·97 in all cases (data not shown). Dividing μmax by two provides the maximum colony radius growth rate (Rmax). Table 1 lists the estimated Rmax for both isolates under the different combinations of temperature and aw. On the other hand, the coefficients of the secondary models developed along with their corresponding standard errors of estimation, coefficients of determination (R2), and RMSEs are presented in Tables 2. Furthermore, the response surface plots for all models are presented in Figs 1 and 2. The R2 values of all fitted models were above 0·97, thus indicating good model fit. With the exception of the interaction term in the polynomial model, all coefficients in the polynomial and Davey models were indicative of significant effects on the growth rate. The impacts of temperature and aw can be seen in the response surface plots of all models where generally the colony growth rate tended to increase with increase in temperature and aw. However, beyond the optimal temperature (30–32°C), the growth rates of both isolates tended to decline gradually upon any further increase in temperature as can be observed in the 3D surfaces of the three models. Statistical significance of the quadratic terms together with the opposite signs of the temperature coefficients in the Davey and polynomial models for both isolates points to curvature in the predicted growth rate (Table 2). Inspection of the 3D surface plot of Davey model showed that the parabolic curves of colony growth rate had relatively equal slopes, hence indicating that aw and temperature act independently but additively. The cardinal model estimated an optimal aw between 0·988 and 0·990, minimum aw between 0·834 and 0·845 and optimal temperature of 30°C for both isolates. Nevertheless, this model provided unrealistic values for Tmin (13°C) and Tmax (46–47°C) for both strains.

Table 1.   The maximal radial growth rate (Rmax) of the two isolates of Aspergillus flavus, DISF10 and DISF10, on paddy at different aw and temperature combinations studied
Temperature (°C)awDISF15DISF10
  1. NG, no growth.

430·99NGNG
430·98NGNG
430·95NGNG
430·92NGNG
430·89NGNG
430·86NGNG
430·82NGNG
370·994·9745·606
370·986·0415·0535
370·953·71953·275
370·921·75251·79
370·890·76650·787
370·860·21150·2945
370·82NGNG
300·996·14856·2105
300·986·50755·9785
300·954·1423·722
300·921·87551·831
300·890·89350·8415
300·860·23450·4045
300·82NGNG
250·994·7875·9725
250·986·12055·165
250·953·52653·5105
250·921·8121·693
250·890·74850·7545
250·860·15350·3565
250·82NGNG
200·994·65355·25
200·985·544·7815
200·952·83753·3395
200·921·51351·545
200·890·590·6655
200·860·13250·277
200·82NGNG
150·990·9260·95
150·981·02450·9425
150·950·4160·4585
150·92NG0·249
150·89NGNG
150·86NGNG
150·82NGNG
100·99NGNG
100·98NGNG
100·95NGNG
100·92NGNG
100·89NGNG
100·86NGNG
100·82NGNG
Table 2.   Coefficients, root mean squared error (RMSE) and coefficient of determination of the kinetic models used to describe the effects of aw and temperature on the growth rate of the Aspergillus flavus species DISF15 and DISF10
ModelCoefficientsDISF15DISF10
Estimated valueEstimated value
  1. *Not significant (P > 0·05).

Daveya0−8·44 × 10± 9·92 × 101−8·22 × 10± 9·52 × 101
a13·35 × 10± 5·80 × 1011·54 × 10± 5·61 × 101
a2−1·67 × 10± 3·13 × 101−7·11 × 10± 3·02 × 101
a34·11 × 1 0± 5·75 × 1044·49 × 10± 5·39 × 104
a4−6·24 × 10± 8·61 × 106−6·80 × 10± 8·07 × 106
R20·9790·975
RMSE0·2420·241
Polynomiala0−6·49 × 10± 9·28 × 10−1−6·08 × 10± 8·68 × 10−1
a11·74 × 10± 7·27 × 1008·56 × 10± 3·89 × 100
a2−6·10 × 10± 7·98 × 100−3·75 × 10± 7·56 × 100
a34·82 × 10−1 ± 6·52 × 10−25·12 × 10−1 ± 5·81 × 10−2
a4−7·95 × 10−3 ± 1·26 × 10−3−8·45 × 10−3 ± 1·11 × 10−3
a5−1·93 × 10−3 ± 8·9 × 10−2*−3·48 × 10−2 ± 8·10 × 10−2*
R20·9760·972
RMSE0·2630·252
RossoRmax (mm day−1)7·45 × 10± 1·59 × 1006·71 × 10± 4·52 × 10−1
Tmin1·32 × 10± 1·54 × 102*1·32 × 10± 9·74 × 103*
Tmax4·71 × 10± 1·74 × 1004·62 × 10± 1·53 × 100
Topt3·01 × 10± 5·18 × 10−12·97 × 10± 4·67 × 10−1
aw,min8·46 × 10−1 ± 1·01 × 10−28·34 × 10−1 ± 1·18 × 10−2
aw,max9·90 × 10−1 ± 3·55 × 10−39·93 × 10−1 ± 3·29 × 10−2
aw,opt9·87 × 10−1 ± 9·06 × 10−39·89 × 10−1 ± 1·98 × 10−2
R20·9870·986
RMSE0·3520·347
Figure 1.

 Response surface plot of the predicted growth rate of Aspergillus flavus (DISF15) as a function of aw and temperature using the (a) Davey, (b) polynomial and (c) cardinal models. Data points represent the mean radial growth rates.

Figure 2.

 Response surface plot of the predicted growth rate of Aspergillus flavus (DISF10) as function of aw and temperature using the (a) Davey, (b) polynomial and (c) cardinal models. Data points represent the mean radial growth rates.

Model validation

The results obtained for the statistical and mathematical criteria employed in evaluating the performance of the kinetic models in describing the fungal growth rate as a function of aw and temperature are shown in Table 3. For both strains, the three models had Bf ≤ 1·117 and Af values ranging from 1·179 to 1·330, therefore indicating that these models were able to predict the mean radial growth rates of these fungi. A wide range was obtained for the accuracy factor probably because the models were validated with combined data from both isolates.

Table 3.   Mathematical indices used to evaluate the performances of the models which describe the effects of aw and temperature on the growth rates of the Aspergillus flavus species DISF15 and DISF10
ModelMathematical indicesDISF15DISF10
DaveyBias factor1·0521·001
Accuracy factor1·1961·179
PolynomialBias factor1·0701·008
Accuracy factor1·2201·184
RossoBias factor1·1171·004
Accuracy factor1·3301·210

The outcomes of graphical validation of the models are presented in Figs 3–5. Figures 3 and 4 demonstrate that deviations of model predictions from the observed data points are more pronounced at low aw than at high levels. In Fig. 5, the location and distribution of the data points relative to the line of equivalent of all models indicated that overly the Davey model had better interpolation potential followed by Rosso and finally the polynomial models.

Figure 3.

 Comparison of the predicted radial growth rates between the (a) Davey, (b) polynomial and (c) Rosso models based on DSF15 models and the observed radial growth rates at 30°C (□), 37°C (⋄), 25°C (△), 20°C (○) and 15°C (×). Validated data points represent the mean radial growth rates for DSF15 and DISF10 ± standard deviation.

Figure 4.

 Comparison of the predicted radial growth rates between the (a) Davey, (b) polynomial and (c) Rosso models based on DSF10 models and the observed radial growth rates at 30°C (□), 37°C (⋄), 25°C (△), 20°C (○) and 15°C (×). Validated data points represent the mean radial growth rates for DSF15 and DISF10 ± standard deviation.

Figure 5.

 Validation of the predictive models for the effects of aw and temperature on the growth of Aspergillus flavus DSF15, [(a) Davey, (c) polynomial and (e) Rosso] and DISF10 [(b) Davey, (d) polynomial and (f) Rosso].

Aflatoxin production

The mean concentrations of aflatoxins B1 and B2 produced by the two isolates of A. flavus on paddy at the different examined aw and temperature values after 3 weeks of incubation are presented in Table 4. Aflatoxin production was observed in the aw range of 0·86–0·99, and the toxin levels were increasing with aw. As to temperature, aflatoxin formation took place in the temperature range between 15 and 37°C but two temperatures were conducive to particularly high aflatoxin production: 25 and 30°C. The lowest levels of toxins were detected at 15°C. Overly, the highest level of aflatoxin (1344 ng g−1) was produced by DISF10 at 25°C and an aw of 0·98.

Table 4.   Mean aflatoxin concentrations (ng g−1) on paddy inoculated with the two Aspergillus flavus species DISF15 and DISF10 as function aw and temperature
 43°C37°C30°C25°C20°C15°C10°C
  1. ND, not detected.

  2. (± standard error of mean).

A. flavus DISF15
 0·99ND339 (±19)793 (±20)811 (±7)565 (±18)32 (±3)ND
 0·98ND446 (±10)573 (±7)844 (±17)523 (±3)25 (±2)ND
 0·95ND327 (±14)789 (±15)483 (±10)272 (±10)NDND
 0·92ND55 (±4)214 (±10)195 (±4)73 (±7)NDND
 0·89NDND84 (±3)60 (±4)22 (±3)NDND
 0·86NDND33 (±3)30 (±2)NDNDND
 0·82NDNDNDNDNDNDND
A. flavus DSF10
 0·99ND483 (±27)1271 (±46)1224 (±28)834 (±11)47 (±8)ND
 0·98ND576 (±21)1173 (±18)1344 (±23)723 (±13)30 (±6)ND
 0·95ND318 (±11)689 (±20)883 (±10)482 (±14)12 (±3)ND
 0·92ND81 (±6)214 (±13)395 (19)113 (±60)NDND
 0·89NDND84 (±7)116 (±5)32 (±2)NDND
 0·86NDND33 (4)46 (4)NDNDND
 0·82NDNDNDNDNDNDND

The results obtained for the two strains were fitted to polynomial model describing the individual effects of aw and temperature, as well as their interaction effects on aflatoxin production. The coefficients of the models developed for both isolates and their significance are presented in Table 5. All the parameters in the models, except the interaction term, had significant effects on aflatoxin production. The contour mapping generated using the polynomial models for each strain is presented by Fig. 6. Coinciding with the earlier growth results, the contour plots identified the temperature range between 25 and 30°C as optimum for aflatoxin production by both isolates and the maximum level of aflatoxin production was achieved at the aw of 0·98. As described by the model, the slopes of plots were more dependent on aw, which indicates that aflatoxin production is more affected by aw than by temperature whose effect becomes more important at low aw values. All negative values in the graph should be thought of as null.

Table 5.   Coefficients and coefficient of determination of the polynomial model used to describe the effect of water activity and temperature on aflatoxin production by Aspergillus flavus (DISF15) and (DISF10)
CoefficientDISF15DISF10
Estimated valueEstimated value
  1. *Not significant (P > 0·05).

a01·42 × 10± 6·75 × 1032·22 × 10± 1·01 × 104
a1−4·08 × 10± 1·45 × 104−6·23 × 10± 2·15 × 104
a22·49 × 10± 7·85 × 1033·77 × 10± 1·16 × 104
a31·95 × 10± 6·84 × 1012·71 × 10± 9·86 × 101
a4−3·25 × 10± 4·46 × 10−1−4·12 × 10± 6·96 × 10−1
a5−1·17 × 10± 5·86 × 10−1*−8·52 × 10± 8·70 × 101*
R20·9200·924
Figure 6.

 Contour plots showing the effects of aw and temperature on aflatoxin production by Aspergillus flavus (a) DISF15 and (b) DISF10.

Discussion

The tropical nature of the major rice-producing countries in Asia and the poor postharvest handling of crops create favourable conditions for growth of the invading fungi. Our study had confirmed the effects of temperature and aw on the radial growth rate of A. flavus and aflatoxin formation on paddy. This finding is consistent with findings of earlier studies stressing on the impacts of aw and temperature on fungal growth (Sautour et al. 2002; Magan et al. 2003; Magan and Aldred 2007).

Models with and without biological interpretations were used to explicate the influences of temperature and aw on the rates of fungal growth on inoculated paddy. The cardinal model estimated the optimal aw for the growth of both isolates as 0·990, which is comparable to estimates reported in the literature. As an example, Gibson et al. (1994) reported that the optimal growth of A. flavus occurred in the aw range of 0·980–0·990. The two strains failed to grow at the aw of 0·820 at all the temperatures studies. On the other hand, the minimal aw estimated by Rosso model was between 0·834 and 0·845. In accordance with these values, the minimum aw for the growth of A. flavus as predicted by the same model on chilli powder extract agar was in the range of 0·83–0·85 (Marín et al. 2009). Results of the present study differ from those of Pitt and Miscamble (1995) who reported that A. flavus was able to grow at an aw of 0·82 on malt extract agar where aw was adjusted by adding glucose and fructose to the basal medium. This implies that the availability of the nutrients and structure of the medium might influence the minimum aw for the proliferation of fungi (Pardo et al. 2004), while they have no effects on proliferation of fungi under the optimal conditions (Marín et al. 1999). Paddy grains are sealed with husks, which render them not easily accessible for the invading moulds. The role of husk in protecting the grain is particularly important at suboptimal aw and temperature conditions for fungal growth (Liu et al. 2006).

The estimated optimal, minimal and maximum temperatures as predicted by the cardinal model were 30, 13 and 46–47°C, respectively. Pitt and Hocking (2009) reported that growth of A. flavus can occur over the temperature range of 10–43/48°C, with an optimum between 32 and 33°C. Probably, the omission of all cases with no growth from the data to enable fitting of the models with logarithmic transformation leads to the unrealistic estimation of the Tmin and Tmax by the Rosso model (Tassou et al. 2009). Under the optimal aw and temperature conditions, the estimated optimal radial growth rate for DISF15 was 7·45 mm day−1, which is slightly higher than the value obtained for DISF10 (6·71 mm day−1). In agreement with our finding, the estimated maximal colony diameter growth rate on red chilli powder extracts agar at 25°C was in range of 14·54–14·72 mm day−1 (Marín et al. 2009). The major advantage of Rosso model is that the estimated parameters have a physiological interpretation and that the initial values needed for estimation of fungal growth are usually available, which facilitate parameter estimation.

The R2 values of all models were above 0·970 for both isolates, which highlights that these models fit well to the data (Table 2). Another meaningful and simple measure of the performance of the model is the RMSE (Ross and Dalgaard 2004). The RMSE values were in the range of 0·347–0·352, 0·241–0·242 and 0·252–0·263 for the Rosso, Davey and polynomial models, respectively (Table 2). In terms of the RMSE, accuracies of model predictions of the fungal growth rates followed the order Davey > polynomial > Rosso. Values of the bias factor in the three models and for both strains were close to 1·0, consequently indicating that the three models were able to predict the true means of the fungal growth rates. However, when the accuracy indices of all models were inspected, it was found that the growth rates predicted by the Davey, polynomial and Rosso models deviated from the observed data on the average by 18·7%, 20·2% and 27·0%, respectively. The estimated values for the bias and accuracy factors underlined that the Davey equation performed better than the Rosso and polynomial models in describing the experimental data. This can also be inferred from Fig. 5, which plots the predicted values against the observed ones. The bias and accuracy factors obtained in the present study were close to the values obtained by the polynomial (Lahlali et al. 2007) and Davey models (Samapundo et al. 2005) in similar studies. However, in average, the Rosso model provided the highest RMSE, bias and accuracy factor, which may be due to logarithmic transformation of the radial growth rate in the polynomial and Davey, but not the Rosso, models.

The results showed that each of temperature and aw were significantly affecting aflatoxin production by both isolates. It is well established that aw and temperature, as well as their interaction, had significance influences on aflatoxin production by A. flavus (Gqaleni et al. 1997). For both isolates, the optimal aw values for aflatoxin production as predicted by the model were close to 0·98 and the optimal temperature for toxin production was between 25 and 30°C with notable decline of aflatoxin production at 15°C. Giorni et al. (2007) reported that the optimal aw and temperature for aflatoxin production by Aspergillus section Flavi on Czapek agar were 0·99 and 25°C, respectively, with a little amount of toxins produced at the aw of 0·83 and 15°C. The authors suggested that 15°C and an aw of 0·83 may be considered as limiting conditions for toxin production by Italian isolates. In another study, Kheiralla et al. (1992) found that the optimal temperature for aflatoxin production by A. flavus was between 25 and 30°C. The differences between the various published studies in the optimal temperature and minimal aw supporting growth of A. flavus and subsequent production of toxins can be attributed principally to the differences in (i) the types of media used and (ii) the fungal strains studied (Klich 2007b). Careful inspection of the counter plot in Fig. 6b leads to the conclusion that the model predictions at marginal aw (below 0·86 aw in the temperature range of 25–30°C) are not much sound. These predictions correspond to models over-fitting the data. Over-fitting is well-known limitation associated with the polynomial models. Therefore, predictions of the polynomial model at the marginal conditions should be used with caution.

Our results show that the aw and temperature conditions apt to toxin production on paddy are the conditions leading to growth of A. flavus (aw = 0·86–0·99). Therefore, wet-harvested paddy or poorly dried with moisture contents corresponding to aw in range of 0·86–0·99 represent suitable medium to be colonized by A. flavus and allow aflatoxin production. From the food safety point of view, attention should be paid towards the prevention of fungal growth rather than looking at mycotoxin production. This is definitely true when we considered the uncertainty associated with the kinetics of mycotoxin together with inherent variability in the toxin production by particular isolate, in particular media (Marín et al. 2009). Therefore, when delay in drying of paddy is expected, it is highly appreciated to cool down the paddy to temperature level at which growth will not occur (10°C).

In conclusions, the present work demonstrated that all models developed in this study were able to predict the growth rates of the two investigated A. flavus isolates on paddy. However, performance of the Davey model in simulating the fungal growth rate as a function of aw and temperature was the best. Rosso model provided the cardinal growth parameters for aw (aw,min, aw,max and aw,opt) and temperature (Tmin, Tmax, and Topt), which contribute to additional understanding of the ecology of A. flavus. Formation of aflatoxins occurred at a minimal aw of 0·86. Therefore, wet-harvested paddy or paddy with moisture content corresponding to aw of 0·86 or above represents a good medium for fungal growth and aflatoxin production. When rice-drying facilities are overloaded with grains that need to be dried, as a control measure, the paddy has to be cooled down to a temperature of 10°C to make sure that no A. flavus will grow on it. The models and related results obtained from this study can be used as an effective tool for predicting the extent of fungal growth on paddy and minimizing the risk of paddy contamination with aflatoxins.

Acknowledgement

The authors are grateful to the UPM for sponsoring this research through the Research University Grant Scheme (RUGS Project no. 02-01-070024RU).

Ancillary