P. Dantigny Laboratoire de Microbiologie, UMR INRA 1082, ENS.BANA, 1 Esplanade Erasme, 21000 Dijon, France (email: firstname.lastname@example.org).
Aims: The influence of temperature, water activity and pH on the time necessary for germination of 90% of Penicillium chrysogenum conidia inoculated (T90) was determined.
Methods and Results: A new experimental device was developed for easy monitoring of the germination process. Experiments were carried out according to a Doehlert matrix at 11–31°C, 0·86–0·98 water activity (aw) and pH 3·5–6·5. In these conditions, a second order polynomial relationship between T90 and the environmental factors was established for the different humectants used throughout this study (e.g. glycerol and sorbitol) with regression coefficients close to 0·97.
Conclusions: For both humectants, the major effect of temperature and water activity on T90 was highlighted, whereas the effect of pH on T90 in these experimental conditions was not significant. The combined effect of temperature and water activity on T90 was also demonstrated.
Significance and Impact of the Study: Both the experimental set-up and the Doehlert matrix were well suited to determine the influence of environmental factors on mould germination.
Under favourable environmental conditions, moulds can grow on a large variety of substrates. Due to the appearance of visible hyphae and the production of unpleasant odours, mould growth causes economic losses (Bullerman 1984). These drawbacks can be overcome by the addition of preservatives such as sorbic acid to food products, although this approach has come under criticism (Sofos and Busta 1981) as customers demand ‘more natural’ and ‘fresher foods’. Literature dealing with the prevention of mould growth has emphasized the control of factors such as aw, temperature (Ayerst 1969; Marín et al. 1996; Cuppers et al. 1997; Ramos et al. 1998) and pH (Magan and Lacey 1984; Skirdal and Eklund 1993; McQuilken et al. 1997). Due to difficulties in acquiring sufficient, reproducible data, modelling of mould growth has not developed as rapidly as bacterial modelling (Gibson and Hocking 1997). The majority of the models describe the relationship between environmental factors and fungal growth (Gibson et al. 1994; Cuppers et al. 1997). Few models (El Halouat and Debevere 1997) concern the effect of environmental factors on spore germination. However, it is of paramount importance to assess the influence of environmental factors on spore germination as this is a primary step to mould spoilage.
In this paper, the effects of temperature (T), water activity (aw) and pH on the time to obtain 90% germination of Penicillium chrysogenum conidia (T90) were determined. Because of solute effects on fungal germination and growth (Pitt and Hocking 1977), different humectants (e.g. glycerol and sorbitol) were also examined. In order to reduce the number of experiments, a Doehlert design was used. Doehlert matrices (Doehlert 1970) present the advantage of being easily expanded in both the variables space and the experimental space (Quignon et al. 1997). The combined effects were represented by means of a polynomial model (Taragano and Pilosof 1999) for each humectant. A response surface methodology was then applied to represent these effects.
MATERIALS AND METHODS
Penicillium chrysogenum was isolated from a spoiled pastry product and identified according to the descriptions of Samson et al. (1995).
The basal medium used for spore production was Malt Extract Agar (bioMérieux, Marcy l’Etoile, France). The germination media consisted of Yeast Nitrogen Base (YNB; 6·7 g l–1; Difco) supplemented by glucose (20 g l–1) and agar (15 g l–1), adjusted to different pH levels (3·5, 5·0 and 6·5) and buffered (phosphate citrate). The aw in these media was adjusted by substituting part of the water with an identical weight of glycerol or sorbitol (w/w). The aw measurements of the adjusted media were determined using an Aqualab CX2T (Decagon Devices, Pullman, WA, USA).
The device used in each experiment was made from a Petri dish. The sterile Petri dish was opened within a laminar flow cabinet. Three small glass cylinders (diameter 16 mm) were placed on the internal side of the lid and filled with the appropriate sterile germination medium to about 1 mm thickness. After solidification, each surface was ready for inoculation. In order to equilibrate the relative humidity inside each device after inoculation, an appropriate water/glycerol solution (15 ml) was poured into the Petri dish. The aw of this solution was identical to that of the culture medium on the lid. The devices sealed with Parafilm® constituted the closed incubation chambers (Fig. 1).
Spores were collected by flooding the surface of the plates with sterile saline solution (8·5 g NaCl l–1 water) containing Tween 80 (0·1% v/v). Germination of Penicillium conidia was observed after inoculation of 104 spores (10 μl of a suspension of 106 spores ml–1) at the surface of YNB agar media distributed into the small glass cylinders. After inoculation, the glass cylinders could be removed from the lid. Without opening the devices, spores were examined twice daily for 48 h and then daily by microscopic observation (×100) or (×400) through the Petri dish lid. Spores were considered germinated when the length of the germ-tube was equal to one half of the spore diameter (Paul et al. 1992).
An experimental domain was defined over 11–31°C, aw 0·86–0·98 and pH 3·5–6·5. Two humectants were also tested, glycerol and sorbitol. Each experiment was performed in duplicate.
In contrast to a previous study (Sautour et al. 2001), the experimental domain was sufficiently wide to expect an optimal response. Therefore, a second-order polynomial relationship, which includes quadratic terms, was required to surround this optimum. A response surface methodology was established to obtain a predictive model for the whole domain. Two types of experimental designs, amongst the most widely used in response surface analysis, could be chosen for determining the model coefficients, central composite designs (CCD) and Doehlert matrices. The location of the experiments in the experimental domain depends on the design chosen. Accordingly, the values of the coefficients depend on the experimental design. However, whether coefficients are significant or not should not depend on the choice of experimental design.
One advantage of the Doehlert design over the CCD is the possibility of extended the domain by adding another factor or displacing the design towards a new experimental domain.
The Doehlert design allows the description of a region around an optimal response and contains k2 + k + 1 points for k variables. For three variables, a set of 13 experiments is required and, in that case, one of the properties of the Doehlert design is the uniform distribution of the experiments in a three-dimensional space (Fig. 2). Thus, 12 experiments are equidistant from a central experiment having the coded values (0, 0, 0) and are distributed on a sphere with a radius of 1 (Fig. 2).
In this study, T90 was estimated, taking into account the influence of three environmental factors (i.e. variables), temperature (X1), aw (X2) and pH (X3). Each experiment could be located by its three coded values. The coded (Xi) and experimental (natural) values of these three factors are listed in Table 1. The natural values and the corresponding coded values were used for setting up the experiments and the model, respectively.
Table 1. Experimental matrix obtained by applying the Doehlert design methodology for three factors and experimental values for T90 (time (d) to obtain 90% of the conidia of Penicillium chrysogenum germinated) for glycerol and sorbitol as humectants
Analysis and interpretation of the results
Multiple regression analysis based on the least square method was performed using Nemrod software (LPRAI, Marseille, France). The analysis concerned the linear and quadratic effects of the three factors and their interactions. Thus, the equation giving T90 was a second-order polynomial model with 10 coefficients (b0, b1, b12…b23):
where X1, X2 and X3 are coded factors studied.
The significance of the coefficients was evaluated by multiple regression analysis based upon the F-test with unequal variance (P < 0·05, P < 0·01 and P < 0·001).
Effects of temperature, water activity and pH on germination time
The average T90 values obtained in the different conditions are reported in Table 1 for both humectants. Prior to assessing the influence of the experimental conditions on T90, the accuracy of the results should be examined. Amongst a total of 26 different experiments (varying either the environmental conditions or the humectant), the results in duplicate were identical in 15 cases. In the other cases, the difference between the assays (e.g. 0·5 d for T90 less than equal to 2 d and otherwise 1 d) was due to the frequency in readings. For example, in experiment no. 8 using glycerol as humectant, assuming less than 90% germination for both the duplicates at 5 d and 100% and 80% germination at 6 d, the T90 results would be 6 and 7 d, respectively, even if all the spores had been well germinated shortly after 6 d, leading to an average T90 value of 6·5 d. These results exhibit reasonable reproducibility of the experiments realized in duplicate.
For each experiment the differences between the results obtained for glycerol and sorbitol (Table 1 could be explained by the sampling methodology. Therefore, the T90 values obtained for glycerol were not significantly different from those obtained for sorbitol. Furthermore, the differences observed in T90 could not be explained only by the variability in the experimental results. Accordingly, the influence of the environmental factors on T90 can be examined.
The results of the multiple regression analysis which provided the estimates of the model coefficients are listed in Table 2. The regression coefficients, r2, were equal to 0·973 and 0·968 for glycerol and sorbitol, respectively; therefore, about 97% of the fraction of the variation about the mean could be explained by the models (Box and Draper 1987). It can obviously be verified that the response means coefficients, b0, are the same as the experimental values of T90 obtained when all the factors are weighted with the coded level 0, experiment 1. The greater the absolute value of the linear coefficients, b1–b3, the more important was the variable influencing the response. Therefore, in all cases, the influence of aw was greater than the influence of temperature. For glycerol and sorbitol, pH had no significant influence on T90. Figure 3 shows that the germination time of P. chrysogenum decreased with increasing aw from 0·86 and with increasing temperature from 16°C. For both humectants, aw and temperature had a linear negative effect and a quadratic positive effect on T90 (see Table 2). This result suggests that, depending on the parameter values, optimum values for aw in the range 0·86–0·98 and temperature in the range 16–31°C capable of minimizing T90 could be obtained. For glycerol and sorbitol, a significant interaction between aw and temperature was demonstrated. The positive values for b12 suggested a synergistic influence of both variables on the spore germination process.
Table 2. Model coefficients obtained for glycerol and sorbitol as humectants
In order to determine the optimum conditions for germination, surface response contour plots were drawn. One factor was fixed arbitrarily at the centre of the matrix, while the two other factors varied. By fixing the pH at 5, the effect of temperature and aw on T90 is shown in Fig. 3 for the different humectants. The optimum conditions are schematized as an ellipse. The centre of the ellipse (e.g. glycerol, 23·4°C, 0·963 and sorbitol, 23·6°C, 0·966) represents the optimum conditions in terms of temperature and aw at pH 5. In these optimum conditions, the model predicts germination times of about 0·5 d for both glycerol and sorbitol. In order to prevent food spoilage due to P. chrysogenum it is, therefore, recommended to avoid these optimum conditions. Figure 3 shows that the conditions which allowed the slowest germination times were 16°C and aw 0·860 and 16°C and aw 0·865 for glycerol and sorbitol, respectively. In these conditions, the model prediction for glycerol (T90 about 9·5 d) was a little greater than for sorbitol (T90 about 8·5 d).
An experimental set-up to determine germination time has been described. The control of humidity (by means of a solution of the same aw as the medium and hermetic closure of the experimental set-up) was required in order to maintain a constant aw. It appeared that the main interest in the proposed method was the observation of the spores through the Petri dish lid without opening the dishes. A similar system has been described by Magan and Lacey (1984). In contrast, Marín et al. (1996) opened the dishes to aseptically remove three discs at each sampling, thus leading to many manipulations.
The spores were considered germinated when the length of the germ-tube was equal to one half of the spore diameter and a test was considered positive when 90% of the inoculated conidia were germinated. Another protocol (i.e. spore considered germinated when the germ-tube length was equal to or greater than the diameter of a spore and a test considered positive when 10% of the spores had germinated) has been suggested (Magan and Lacey 1984; Marín et al. 1996; McQuilken et al. 1997). Obviously, the protocol does affect the germination time. Magan and Lacey (1984) recommended the figure of 10% in preference to larger percentages to obtain a better estimate of the minimum aw allowing germination. However, this was not the purpose of the present study. The influence of environmental factors on T10 (time to obtain 10% of the germinated conidia) has been assessed. Although the constant values were different, the main effects of temperature and aw and the combined effects of temperature and aw on the conidial germination of P. chrysogenum were also demonstrated (results not shown).
The main effects of aw and temperature have also been reported in the literature for fungal growth (Holmquist et al. 1983) and conidial germination of Fusarium spp. (Marín et al. 1996). The influence of temperature on growth was second after aw in order of significance (Holmquist et al. 1983). The major influence of aw on the germination process is in accordance with the data of Ayerst (1969) and Seiler (1976) on moulds growing on minimally processed foods.
In contrast, in this study no significant influence of pH, in the range 3·5–6·5, was highlighted. These results are in accordance with those of McQuilken et al. (1997) who reported that maximum germination occurred between pH 4·5 and 6·2. In addition, the optimum temperature and aw have been determined through the model equations at pH 5, which is optimum for most moulds (Lacey 1989). The combined effect of temperature and aw has been reported for fungal growth (Ayerst 1969; Horner and Anagnostopoulos 1973; Pitt 1993) and germination (Ayerst 1969; Marín et al. 1996). The present results were in accordance with this literature.
It has been reported that the effect of the changes in aw, temperature and pH on fungal growth was broadly the same for media adjusted by either glycerol or sucrose (Horner and Anagnostopoulos 1973). A similar observation was made by Pitt and Hocking (1977) for germination time. Eventually, these authors concluded (Hocking and Pitt 1979) that, with the notable exception of NaCl, the differences in effects caused by the various solutes were slight. Accordingly, they recommended the use of glycerol for controlling aw.
Experiment designs according to Doelhert matrices were carried out to determine the optimal conditions for conidial germination. According to Henika (1982), this methodological approach is well suited to work on predictive microbiology in the agro-food area. Previously, other authors have used polynomial models to describe Aspergillus development (Fang et al. 1994). The multifactorial analysis allowed the development of a model with a good quality of fit and taking account of possible interaction effects between the environmental factors. However, the prediction given by the model was valid only for the strain/medium and within the particular experimental domain (Delignette-Muller 1997). Any extrapolation to other species or growth substrates, and outside the limits of the considered factors, would be hazardous; consequently, it is important to specify the areas in which the model cannot be used (Baranyi et al. 1996).
As the experiments were mostly performed on optimal culture medium, the fungal development in the laboratory appeared faster than in food; thus, some caution is needed in use of the model. Predictive mycology studies the behaviour of moulds under different physico-chemical conditions. It can help in the identification of critical points in the production and distribution process and the optimization of production and distribution chains (Zwietering et al. 1990). The modelling approach introduced in this paper is in accordance with these objectives and could contribute to improving the microbiological safety and shelf-life of food products.
This research was partially supported by ACTIA (Paris). The authors would like to thank V. Huchet and D. Thuault from ADRIA, V. Stahl from AERIAL M. Sergent and M.C. Guilhem from LPRAI (Marseille).