Mycol. Res. 103 (7) : 887–895 (1999) 887 Printed in the United Kingdom Detection and estimation of conidial abundance of Penicillium verrucosum in soil by dilution plating on a selective and diagnostic agar medium (DYSG) S U S A N N E E L M H O L T1, R O D R I G O L A B O U R I A U2, H E L L E H E S T B J E RG1 A N D J Ø R G E N M. N I E L S E N1 " Department of Crop Physiology and Soil Science # Biometry Research Unit, Danish Institute of Agricultural Sciences, Research Centre Foulum, P.O. Box 50, DK-8830 Tjele, Denmark Penicillium verrucosum is one of the main producers of ochratoxin A (OA) in agricultural commodities. To forecast the risk of OA contamination, there is a need to improve our knowledge on the ecology of P. verrucosum in the field. Dilution plating on ‘ dichloran yeast extract sucrose 18 % glycerol agar ’ (DYSG) offers a simple and very sensitive method of detecting P. verrucosum propagules in soil. The properties of DYSG are illustrated in a suspension mixture experiment and confirmed in a soil mixture experiment. In the latter, P. verrucosum could be detected in conidial concentrations below 200 colony forming units (cfu) g−" soil even when it constituted no more than 0±3 % of the cfu. Furthermore, the DYSG method can be used to estimate the abundance of P. verrucosum propagules in soil with good precision. In some of the analysed cases, however, it was necessary to use appropriate mathematical models to treat results with high numbers of cfu on the Petri dishes. Penicillium verrucosum Dierckx belongs to the subgenus Penicillium and is characterized in part by its production of ochratoxin A (OA) and verrucolone (Frisvad & Filtenborg, 1989). Based on a major research effort, taxonomists today agree that P. verrucosum is the only known species of the genus to produce OA (Pitt, 1987 ; Frisvad & Samson, 1991). P. verrucosum can be divided into two physiologically distinct chemotypes (Ciegler et al., 1973 ; Frisvad & Filtenborg, 1989). Chemotype I is found on processed meat products and produces OA. Chemotype II produces OA and citrinin and is found on grain and on cereal products. The mycotoxin OA has nephrotoxic, carcinogenic, teratogenic, and immunosuppressive properties and constitutes a significant health risk to humans and domestic animals (Krogh, 1987 ; Boorman, 1989 ; Smith et al., 1994). OA contamination of plant products has been reported regularly from a wide range of countries, especially within the temperate regions of the world (e.g. Ho$ kby et al., 1979 ; Sinha, Abramson & Mills, 1986 ; Jørgensen, Rasmussen & Thorup, 1996). Although some species of Aspergillus produce OA, there are numerous indications that OA contamination in temperate regions can be ascribed solely to P. verrucosum. Many mycotoxin-producing fungi, including P. verrucosum, are regarded as storage fungi and their occurrence and behaviour have been studied almost exclusively under storage conditions. In a review on OA contamination of agricultural commodities, Lilleho$ j & Elling (1983) drew attention to the lack of knowledge of the fungal cycle from soil to grain and back to soil. To the best of our knowledge, no such studies have been performed for P. verrucosum. Better knowledge of the natural occurrence and behaviour of P. verrucosum in the field ecosystem is needed to forecast and obviate the problems that arise from this fungus. Although P. verrucosum grows readily on many agar media, it has not been found in Danish arable soils with soil washing and dilution plating on general media (Elmholt & Kjøller, 1989 ; Elmholt, Frisvad & Thrane, 1993). One reason could be that it occurs in very low propagule concentrations, implying that a more sensitive method of detection is required. Selective and diagnostic media have been developed for the examination of P. verrucosum in stored cereal products. They are all based on ‘ yeast extract sucrose agar ’ (YES), which was developed for the production of aflatoxins (Davies, Diener & Eldridge, 1966) and a range of other mycotoxins, among them OA (Scott, Lawrence & van Waalbeek, 1970). The YES medium furthermore enables a differentiation of chemotypes I and II (Frisvad, 1981 ; Frisvad & Filtenborg, 1989). Chemotype I has a cream-coloured Petri dish reverse, while chemotype II has a violet brown- to brownish red-coloured reverse. The selectivity of YES has been improved by the addition of rose bengal and either dichloran in ‘ DRYES agar ’ (Frisad, 1983) or pentachloronitrobenzene in ‘ PRYES agar ’ (Frisvad, 1986). Preliminary studies indicated that neither the selective nor the diagnostic properties of YES and DRYES were good enough to detect and enumerate P. verrucosum (chemotype II) in a population of soil fungi (Elmholt & Hestbjerg, 1996). They did, however, point to another medium which has been developed for the detection of P. verrucosum (chemotype II) on grain. It is called ‘ dichloran yeast extract sucrose 18 % glycerol Detection and enumeration of P. verrucosum in soil 888 agar ’ (DYSG) and was introduced by Frisvad et al. (1992). DYSG is a modified DRYES agar in which rose bengal is replaced by 220 g l−" of glycerol. This paper reports the properties of DYSG as a selective medium for detecting and estimating the abundance of P. verrucosum (chemotype II) in soil, using the classical dilution plating technique. The following questions were addressed : (i) Does the method produce consistent estimates of the abundance of P. verrucosum in soil suspensions at a range of different conidial concentrations ? (ii) Does a very strong competition from the indigenous soil fungi on the Petri dishes affect the estimated abundance of P. verrucosum ? (iii) Does competition from P. verrucosum on the Petri dishes affect the estimated abundance of indigenous soil fungi ? (iv) Can P. verrucosum be detected when constituting only a very small proportion of the viable propagules in a soil suspension ? (v) What is the recovery and lower limit of detection of P. verrucosum conidia in soil, using the DYSG method ? Results were treated statistically according to methods based on a Poisson distribution of the number of cfu per Petri dish, as originally proposed by Fisher, Thornton & Mackenzie (1922). The statistical techniques used here were, however, specially designed to take into account that the amount of soil added per Petri dish was not constant, as in the classical experiment treated by Fisher and co-workers. We thus used a polynomial Poisson regression model, which was flexible to treat situations of crowding on the Petri dish. MATERIALS AND METHODS A portion of the soil was infested with a strain of P. verrucosum (IBT 5010), isolated from Danish-grown barley (Lund et al., 1992). It was verified by tlc that the strain produces OA and citrinin, confirming that it belongs to chemotype II. The fungus was grown on ‘ malt extract agar ’ (MEA) (Pitt, 1979) and incubated for two weeks at 20 °C. Using a Drigalsky spatula, conidia were washed off the dishes with a hydrous dilution medium of 8±5 g NaCl, 1 g peptone, and one drop of Tween 804 l−". The conidial suspension was filtered twice (140 µm) to exclude hyphae. To break up conidial chains, 2 mm glass beads and five drops of Tween 804 l−" were added. The suspension was rotated for 1 h on a circular rotator (250 rpm) and afterwards overhead-rotated for 3 h (40 rpm). After centrifugation (10 min at 4° and 4000 rpm), most of the supernatant was discarded. The pellet was very carefully resuspended into the remaining liquid. A microscopic analysis confirmed that the conidial chains were broken and that no germination had taken place. To facilitate a homogeneous incorporation of the conidia into soil, the suspension was first meticulously mixed with autoclaved river sand. Following this, the infested sand was very thoroughly mixed into one portion of the soil in an amount of 1 % (w}w). The same amount of non-infested sand was mixed into another portion of the soil, which served as a reference. In the following, the terms ‘ infested soil ’ (IS) and ‘ non-infested soil ’ (NIS) will be used for the two soils. Dilution medium and agar media Fungi were isolated from the soil by dilution plating on 9 cm diam. Petri dishes, using a hydrous dilution medium with 8±5 g NaCl and 1 g peptone l−". ‘ V8-juice agar ’ (V8) (Diener, 1955) and DYSG were used as nutrient media. Both media were amended with 50 ppm chloramphenicol and 25 ppm chlortetracycline to inhibit bacterial growth. Soil sampling and infestation with P. verrucosum The soil was sampled in a recently harvested wheat field at The Danish Institute of Agricultural Sciences, Research Centre Foulum. It is a sandy loam with 31 % coarse sand, 11 % clay, 1±78 % C, and a C}N ratio of 12. The soil was slowly air-dried to a moisture content of approx. 14 % and sieved (2 mm). Incubation and enumeration The Petri dishes were incubated at 25° in the dark. The total number of cfu was enumerated after 4 d and all colonies were marked with a thin felt-tip pen on the Petri dish reverse. At day 6 of the incubation, P. verrucosum (chemotype II) colonies Table 1. P. verrucosum and indigenous soil fungi on Petri dishes, prepared from the infested (IS) and non-infested (NIS) soil suspension. Results are shown as number of colonies (cfu) on five replicate Petri dishes, using two dilutions and different volumes for plating Soil concentration in suspension (mg .. ml−") Volume plated (µl) Soil on dishes (µg ..) P. verrucosum and soil fungia (cfu dish−") IS 0±574 5±74 5±74 5±74 100 40 80 120 57 230 459 689 9}3 34}13 54}19 89}24 8}2 30}17 60}26 79}24 100 40 80 120 136 544 1090 1630 0}6 0}38 0}61 0}85 0}9 0}24 0}47 0}73 9}3 35}8 77}16 81}32 12}2 30}13 52}21 78}20 10}3 21}10 57}25 77}16 0}6 0}28 0}64 0}65 0}5 0}29 0}67 0}87 NIS 1±36 13±6 13±6 13±6 a Five replicates. Left side of slash : P. verrucosum. Right side of slash : Indigenous soil fungi. 0}8 0}37 0}49 0}106 Susanne Elmholt and others 889 Table 2. Preparation of M1 to M10. Two dilutions of the infested soil (IS) and one dilution of the non-infested soil (NIS) were used. The table shows how volumes (µl) of these suspensions were mixed on the Petri dishes to obtain concurrent high numbers of indigenous fungi and varying numbers of P. verrucosum Table 3. Procedure to obtain soil mixtures SM1 to SM7. Two individual thinning series (A and B) were prepared from field-moist, infested (IS) and non-infested soil (NIS). The moisture content of the IS was 10±8 % and of the NIS 9±5 % Volume of soil suspension, mixed and plated on Petri dish (µl) IS M10 M9 M8 M7 M6 M5 M4 M3 M2 M1 SM1 SM2 SM3 SM4 SM5 SM6 SM7 NIS 5±74 mg soil ml−" 0±574 mg soil ml−" 136 mg soil ml−" 200 160 120 100 80 40 20 0 0 0 0 0 0 0 0 0 0 100 40 20 100 100 100 100 100 100 100 100 100 100 a Thinning series IS (g) NIS (g) IS (%) Moisture content (%) A B A B A B A 30a 6a 6 g SM1 6 g SM2 6 g SM3 6 g SM4 6 g SM5 30 54 54 54 54 24 24 50 10 5 1 0±5 0±2 0±1 9±9 9±9 10±0 10±2 9±9 9±6 9±7 Non-thinned infested soil. liquid. The suspension mixtures were termed Mixture 1 to Mixture 10 (M1 to M10). P. verrucosum was enumerated as described above. Soil fungi were too numerous to count. The soil mixture experiment had developed their characteristic terracotta-coloured pigmentation on the DYSG reverse (7D7-7D8, according to Kornerup & Wanscher, 1969). An enumeration of P. verrucosum in Petri dishes with a relatively high number of colonies was possible, because the felt-tip marks made at day 4 of the incubation facilitated a discrimination of colonies which had become confluent at day 6. The number of indigenous soil fungi was calculated by subtracting the number of P. verrucosum cfu from the total number of cfu on a plate. The suspension mixture experiment Preliminary dilution platings of the IS and the NIS were performed on DYSG and V8 to establish appropriate dilution levels and assure that the NIS held a typical arable soil microbiota. Based on these results (not shown), dilutions with 5±74 and 0±574 mg soil (..) ml−" were prepared from the IS. From the NIS, suspensions with 136, 13±6 and 1±36 mg soil (..) ml−" were prepared. Calibration. A calibration for the suspension mixture experiment was performed in order to estimate the abundance of P. verrucosum in the IS and the abundance of soil fungi in both soils. For each soil, different volumes of two suspensions were plated onto Petri dishes, giving a range of four different amounts of soil per plate and five replicate dishes for each amount (Table 1). P. verrucosum and soil fungi were enumerated as described above. Main experiment. A range of 10 suspensions was prepared by mixing the above suspensions of the IS with that of the NIS according to the scheme in Table 2. Table 4 shows the resulting amount of infested soil added to the Petri dishes. Five replicate DYSG Petri dishes were prepared from each suspension mixture. Mixing was performed with a Drigalsky spatula, directly on the Petri dish during spreading of the In this experiment, the IS was gradually thinned with the NIS to produce seven soil mixtures with decreasing numbers of P. verrucosum conidia g−" of soil. The IS and the NIS originate from the same soils as in the suspension mixture experiment. They were, however, treated differently prior to dilution plating and, therefore, the abundance of fungi in the IS and the NIS is different in the two experiments. Two thinning series, A and B, were prepared according to Table 3. The soil samples were mixed very carefully and kept at 2°C until analysis. They were termed Soil Mixture 1 to Soil Mixture 7 (SM1 to SM7). The soil moisture content was determined in triplicate for the IS, the NIS and the SM1 to SM7 by drying samples of approx. 0±5 g at 105° for 24 h (Table 3). Three replicate subsamples were drawn from the IS and the NIS as well as from each of the seven soil mixtures. For each subsample, a dilution was prepared containing 70 mg (..) soil ml−" (Dilution 1). This dilution was used for plating SM4 to SM7. For the IS, NIS and SM1 to SM5, a second dilution (Dilution 2) was prepared from Dilution 1. Dilution 2 contained 17±5 mg (..) soil ml−" (Table 6). From each subsample, ten replicate DYSG dishes were plated with 0±1 ml of the chosen dilution. This gives a total of 30 Petri dishes from each soil except for SM4 and SM5. From these two mixtures, both dilutions were used, giving a total of 60 Petri dishes. The Petri dishes were incubated and P. verrucosum enumerated as described above. Soil fungi were enumerated in the NIS. Statistics The statistical analysis was performed by assuming that the counts of cfu on a Petri dish, Y, follow a Poisson distribution with an expectation depending on the amount of soil added to the Petri dish, x. The adherence to the Poisson distribution was verified in each case. Two models, corresponding to two different curves, were considered in order to describe the relation between the amounts of soil used and the expected Detection and enumeration of P. verrucosum in soil 890 Table 4. Estimated abundance (cfu) and density (cfu cm−#) of P. verrucosum (P. verr.) and indigenous soil fungi on the Petri dishes of M1 to M10. The estimates were based on the calibration result (see text) and the plated amount (µg ..) of infested (IS) and non-infested soil (NIS) on the Petri dishes M10 M9 M8 M7 M6 M5 M4 M3 M2 M1 IS on dishes (µg ..) Estimated P. verr. cfu on dishesa Estimated soil fungi cfu on dishesb Estimated density P. verr.c (cfu cm−#) Estimated density soil fungi (cfu cm−#) P. verr.}total (%)d 1148 918 689 574 459 230 115 57 23 12 144 115 86±6 72±1 57±7 28±9 14±4 7±2 2±9 1±4 763 754 744 739 734 725 720 718 716 716 2±27 1±81 1±36 1±13 0±91 0±45 0±23 0±11 0±05 0±02 12±0 11±9 11±7 11±6 11±5 11±4 11±3 11±3 11±3 11±3 15±9 13±3 10±4 8±9 7±3 3±8 2±0 1±0 0±4 0±2 based on the calibration result for P. verrucosum and the amount of IS per dish. based on the calibration results for soil fungi and the amount of IS and NIS per dish. Due to the fixed amount of NIS suspension on all Petri dishes (13±6 mg ..), the estimated contribution of soil fungi from this suspension is the same on all dishes (715 cfu), while the amount of soil fungi from the IS varies with the varying amount of IS suspension. c Petri dish area : 63±6 cm#. d P. verrucosum¬100}(P. verrucosum­indigenous soil fungi). a b counts. In the first model, the ‘ Proportional model ’, the expectations of the counts, E(Y), were assumed to be proportional to the amount of soil added i.e. E(Y) ¯ λx. (1) Here, the parameter λ is interpreted as an estimate of the abundance of cfu in the soil This model is thought to represent the ideal situation where neither intra- or interspecies competition nor systematic errors occur. The analysis of the experimental data showed that in some cases, the curve representing the expected counts as a function of the amount of soil did not correspond to a straight line crossing the origin, as in (1). Instead, the curve was bent to the right compared with the straight line in (1), indicating a lack of fit of the proportional model. In order to correct the model for this deviation, an alternative model was considered. According to this ‘ Corrected model ’, the expectation of the counts is assumed to be a homogeneous fourth degree polynomial function of the amount of soil added to the Petri dish, i.e. E(Y) ¯ λx­λ x#­λ x$­λ x%. (2) # $ % Note that the curve described in (2) also crosses the origin. Moreover, the straight line (1) is a particular case of (2), obtained when λ , λ and λ are zero. Here, the parameter λ # $ % is the derivative of the curve at the origin (i.e. the inclination of the tangent of the curve at the point x ¯ 0), and it is interpreted as an estimate of the abundance of cfu in the soil. The other parameters are understood as nuisance parameters used to correct the curve for deviations from a straight line. Both the proportional and the corrected model are generalized linear models (McCullagh & Nelder, 1989). The analysis was performed using the procedure ‘ genmod ’ in SAS. The adequacy of the models was verified by the standard techniques for validation of generalized linear models (McCullagh & Nelder, 1989). In particular, the deviance standardized residuals were used to detect outliers and other possible abnormalities. A likelihood ratio test has been used to verify whether the parameters λ , λ and λ are all equal to zero, i.e. to test for # $ % a possible linearity of the curve. This test will be referred to in the following as ‘ the test of linearity ’. The models (1) and (2) were embedded sometimes in a structure of covariance analysis in order to test the effect of factors of interest (e.g. thinning series, replications, dilutions). The adequacy of the curve, fitted for the two models, was tested by a likelihood ratio test comparing the likelihood of the current model with the likelihood of a model with a free form of the curve. More precisely, the current model was compared with a model for which a common mean was assumed for the Poisson distribution corresponding to the observations with the same amount ‘ x ’ of soil added. Nothing was assumed concerning the relation between the means corresponding to observations with different values of ‘ x ’. This test was performed (but not shown) for each of the final models used. In no case was a lack of fit detected. All confidence intervals and tests used in relation to models (1) and (2) were based on the likelihood ratio (LR). RESULTS In the following, all results are reported as cfu per Petri dish or as cfu mg−" or g−" soil (..). The suspension mixture experiment Calibration. The Results in Table 1 were used to estimate the abundance of propagules of P. verrucosum and indigenous soil fungi in the IS and NIS. For P. verrucosum in the IS, the proportional model fit the data reasonably well, and the test of linearity proved that the proportional model needed no correction (P-value " 0±10). The abundance of P. verrucosum in the IS was estimated to 126 cfu mg−" soil with a 95 % LR based confidence interval of (118 ; 134). No P. verrucosum was found on Petri dishes with suspensions of NIS. The test of linearity confirmed that the proportional model was also Susanne Elmholt and others 891 Table 5. P. verrucosum on Petri dishes from M1 to M10. Results are shown as cfu dish−" (five replicates) with means and .. The last column shows the difference between the observed mean (O) and the estimated mean (E) in percentage of the latter. The estimated mean is given in Table 4. P. verrucosum (cfu dish−") M10 M9 M8 M7 M6 M5 M4 M3 M2 M1 108 96 83 57 48 24 10 7 2 2 124 101 79 65 45 31 9 7 7 1 111 96 68 59 39 23 18 7 2 2 123 99 65 58 45 26 14 8 2 2 99 90 77 45 34 28 8 5 1 1 Mean (..) (cfu dish−") (O-E)}E (%) 113 (11) 96 (4) 74 (8) 57 (7) 42 (6) 26 (3) 12 (4) 7 (1) 3 (2) 2 (1) ®22 ®16 ®14 ®21 ®27 ®9 ®18 ®6 ®3 11 appropriate (P-value ¯ 0±07) to analyse the number of indigenous soil fungi in the IS and the NIS in a joint model. Their abundance in the IS was estimated to be 42 cfu mg−" soil with a 95 % LR confidence interval of (37 ; 47). Their abundance in the NIS was estimated to be 53 cfu mg−" soil with a 95 % LR confidence interval of (49 ; 56). The difference between the number of cfu in the two soils was statistically significant (P-value ! 0±001). Main experiment. The amount of infested soil per Petri dish varied from 12 µg in M1 to 1148 µg in M10 (Table 4). As a result of this and based on the calibration, the estimated number of P. verrucosum on a Petri dish varied from 1±4 in M1 to 144 in M10 or from 0±02 to 2±27 cfu cm−#. The contribution of indigenous soil fungi from the IS varied according to the amount of soil on the dishes. But due to the large ( " 700) and fixed contribution of soil fungi from the NIS suspension, the estimated number of indigenous soil fungi was very high on all Petri dishes, ranging from 716 in M1 to 763 in M10 or 11–12 cfu cm−# in all mixtures. Expressed as percentages, P. verrucosum was estimated to constitute from 0±2 % to 15±9 % of the total number of cfu on the plates. Table 5 reports the observed number of P. verrucosum on the Peri dishes of each suspension mixture. Means with .. have been calculated as well as the difference between the observed and the estimated values. The data in Table 5 indicated that the proportional model would not fit the data and the test of linearity confirmed that the corrected model should be used (P-value ¯ 0±02). The adequacy of the corrected model was confirmed by a likelihood ratio test against a free regression curve model (P-value ¯ 0±672). A likelihood ratio test showed that the corrected model should not be reduced to a third order polynomial model. Using the corrected model, the abundance of P. verrucosum in the IS was estimated to 129 cfu mg−" soil with a 95 % LR confidence interval of (101 ; Table 6. Abundance of P. verrucosum (cfu g−" soil (..)) in three replicate subsamples, each consisting of 10 replicate Petri dishes, of the infested soil (IS) and the SM1 to SM7. The mean value of all replicates (n ¯ 30) is shown with the P-value of the proportional model, used to test differences among replicates. Numbers in italic give the .. (n®1). The last columns show the estimated means and the difference between the observed mean (O) and the estimated mean (E) in percentage of the estimated mean. Penicillium verrucosum (cfu g−" .. soil) Dil.b IS 2 SM1 2 SM2 2 SM3 2 SM4 1 SM4 2 SM5 1 SM5 2 SM6 1 SM7 1 Sample a Sample b Sample c 44 300 5600 28 600 2120 6630 853 2600 1370 641 366 298 314 399 231 245 429 156 156 0 47 700 4680 26 400 2640 5950 2130 1550 887 1040 304 1090 742 353 179 508 416 160 127 15 48 000 4170 ®c 6380 2270 3950 1580 481 198 481 553 226 227 120 381 120 138 NDc Mean P-value Estimated P. verrucosuma (E) (cfu g−" .. soil) 46 700 4980 27 500 2600 6210 2060 2700 1610 721 373 622 643 326 219 291 428 145 137 8 0±232 55 657 ®16 0±203 27 829 ®1 0±829 5566 12 0±000 2783 ®3 0±001 557 30 0±014 557 12 0±178 278 17 0±101 278 5 0±795 111 30 56 NDd NDd (O®E)}E (%) The estimated abundance of P. verrucosum is based on the statistical analysis of results from all soil mixtures, except SM7 and eight outliers (three from SM3, four from SM4-Dilution 1 and one from SM4-Dilution 2. b Dilution 1 : 70 mg soil (..) ml−". Dilution 2 : 17±5 mg soil (..) ml−". c Due to an experimental error during substrate preparation, Petri dishes in this dilution series lacked glycerol. Therefore no Petri dishes in SM7-Sample c, and only two in SM2-Sample a, could be enumerated. d ND : not determined due to insufficient data (see text). a Detection and enumeration of P. verrucosum in soil 892 160). For comparison, its abundance according to the rejected proportional model was estimated to 102 cfu mg−" (98 ; 107). The soil mixture experiment The results of the soil mixture experiment are shown in Tables 6 and 7. The proportional model was applied to the results from the NIS (not shown) and produced an estimate of 35 soil fungi cfu mg−" soil with a 95 % LR confidence interval of (33 ; 38). This result was used to calculate the estimated number and density of soil fungi on the Petri dishes as well as the percentage of P. verrucosum as a proportion of the total number of fungi on the dishes (Table 7). Table 6 shows the observed abundance of P. verrucosum in the IS and SM1 to SM7. For each soil it gives the result of each of the three replicate subsamples (a–c) and the mean of the three replicates. P. verrucosum was found in all soil mixtures. In SM7, however, only one colony was detected on the total of 30 Petri dishes, and SM7 was not included in the statistical analyses. At first, the effect of using different dilutions of the same soil mixture was tested. The proportional model was used on the data from SM4 and SM5 to compare the results from the two dilutions (n ¯ 120). A likelihood ratio test showed no statistically significant difference (P-value ¯ 0±41). Secondly, the proportional model was used to test the difference between the two soil thinning series A and B Table 3), using SM2 and SM4–Dil 1 from thinning series B and SM3 and SM5–Dil1 from thinning series A (n ¯ 120). A likelihood ratio test showed no statistically significant dif- ference between A and B (P-value ¯ 0±13). The homogeneity of P. verrucosum abundance in the soil mixtures was tested by applying the proportional model to the results from the replicate subsamples, the P-values of which are shown in Table 6. In SM3 and SM4, a statistically significant difference was found between these replicates. By removing eight outliers (Table 6, note a) from a total of 235 observations, the P-values were increased to 0±035 (SM3), 0±027 (SM 4–Dil1) and 0±065 (SM4–Dil2). These eight points presented high values for the standardized deviance-residuals and they were removed in all following analyses. The above analyses indicate a homogenous distribution of P. verrucosum propagules in all soil mixtures, including SM6, from which an average of 145 cfu g−" soil (..) were isolated with no significant difference between the replicate subsamples. Therefore, it was considered appropriate to pool the results from SM1 to SM6 to produce an estimate on the abundance of P. verrucosum in the IS. A test confirmed that there was no need to use the corrected model. Using the proportional model, the abundance of P. verrucosum in the IS, as based on the results from SM1 to SM6, was estimated to 56 cfu mg−" IS with a 95 % LR based confidence interval of (53 ; 58). Based on this result, the estimated abundance of P. verrucosum was calculated in all SMs (Table 6), ranging from 27 829 cfu g−" soil in SM1 to 56 cfu g−" in SM7. The proportional model applied to the 30 observations from the IS showed a lower abundance of P. verrucosum, i.e. 47 cfu mg−" soil with a 95 % LR based confidence interval (45 ; 49). Table 6 also shows the difference between the observed mean (O) and the estimated mean (E) in percentage of the latter. Table 7. Estimated abundance (cfu) and density (cfu cm−#) of P. verrucosum (P. verr.) and indigenous soil fungi on the Petri dishes of SM1 to SM7. The total amount of soil (µg ..) on Petri dishes is shown as the mean of the three replicate subsamples. Calculations to show the amount of soil, originating from the IS were based on the thinning procedure (Table 3). Numbers in italic give the .. (n ¯ 3). Dil.a NIS 2 IS 2 SM1 2 SM2 2 SM3 2 SM4 1 2 SM5 1 2 a b SM6 1 SM7 1 Soil on dishes (µg) IS on dishes (µg) 1519 95 1649 19 1713 88 1686 46 1648 29 6534 250 1633 62 6755 302 1689 75 6853 192 6609 121 0 0 1649 19 856 44 168 4 83 2 65 2 17 1 34 2 8 !1 14 1 7 !1 Estimate P. verr. cfu on dishes Estimate soil fungi cfu on dishes Density P. verr. (cfu cm−#) Density soil fungi (cfu cm−#) P. verr.}total (%)b 0 53 0 0±83 0 91±8 57 1±44 0±90 61±6 47±7 59 0±749 0±93 44±5 9±37 58 0±147 0±92 13±8 4±60 57 0±072 0±90 7±5 3±64 226 0±057 3±56 1±6 0±93 57 0±015 0±89 1±6 1±89 234 0±030 3±68 0±8 0±46 59 0±007 0±92 0±8 0±76 238 0±012 3±73 0±3 0±39 229 0±006 3±60 0±2 Dil. : Dilution 1 : 70 mg .. soil ml−". Dilution 2 : 17±5 mg .. soil ml−". P. verrucosum¬100}(P. verrucosum­soil fungi). Susanne Elmholt and others Table 7 shows the estimated number of cfu and density of P. verrucosum and the soil fungi on the Petri dishes for the NIS, IS, and SM1 to SM7. The amount of infested soil on each Petri dish varied from 7 µg in SM7 to 1649 µg in the IS. As a result of this, the estimated number of P. verrucosum on a Petri dish varied from 0±39 in SM7 (0±006 cfu cm−#) to 91±8 in the IS (1±44 cfu cm−#). The estimated abundance of soil fungi on each Petri dish was approx. 60 cfu when Dil2 was used (about 0±90 cfu cm−#) and approx. 230 cfu when Dil1 was used (about 3±5 cfu cm−#). P. verrucosum was estimated to constitute from 0±2 to 61±6 % of the total number of cfu on the Petri dishes. DISCUSSION Most fungal colonies on soil dilution plates originate from conidia or other fungal propagules and not from hyphal fragments (Warcup, 1957). In view of this, dilution plating can be used to assess the fungal spore content in soil samples (Parkinson, 1994). In the present study, dilution plating was regarded as an appropriate method for recovering P. verrucosum, which had been added to the soil in the form of conidia. It should be noted that the indigenous soil fungi only represent those, present in the soil as propagules, with the ability to grow on DYSG. The selectivity of DYSG is reflected in the relatively low numbers of indigenous fungi (35–53 cfu mg−" soil). In Danish soils grown to wheat, 100–500 cfu mg−" will typically be found on general media such as V8 (Elmholt, 1991 ; Elmholt, 1996). In a preliminary study, DYSG was shown to reduce the number of cfu to about 30 % of the numbers on V8 (Elmholt & Hestbjerg, 1996). This is in accordance with the abundance of soil fungi obtained in the present study. The statistical models The statistical methods used in the present study rely on the assumption that the number of cfu per Petri dish is Poissondistributed. This is in accordance with the classic literature, presuming that the dilution plating is well-performed (Fisher et al., 1922). Fisher & co-workers argued that the homogeneity of the suspensions used in the sequence of dilutions is the crucial point for obtaining Poisson-distributed responses without overdispersion (see also McCullagh & Nelder, 1989). Furthermore, non-homogeneity most likely implies that the data are not Poisson-distributed. Therefore, the good adjustment to the Poisson-distribution and the absence of overdispersion in all the experimental data from the suspension mixture experiment indicate that the dilution process was well-performed. This was confirmed in the soil mixture experiment, in which no significant difference was found between the two dilution levels, analysed for SM4 and SM5. In principle, the average number of cfu per plate can be expected to be proportional to the amount of soil added to the dish. This situation was met in some of the examples presented in this paper. In those cases, the proportional model was sufficient. In other cases, however, relatively large amounts of soil added to a Petri dish led to lower cfu recordings than expected from the proportional model, i.e. the 893 graph of the expected number of cfu, µ(x), against the amount of soil, x, bends to the right with respect to a straight line. This will be referred to as ‘ non-proportionality ’. This phenomenon has been described previously and treated in different ways. Roberts & Coote (1965) used essentially a proportional model together with a cut-off value for the amount of soil added per dish. The cut-off was chosen as the largest value for which the proportional model still fits the data reasonably well, and values above this were ignored. When applying this method to our data, the estimates on fungal abundance come close to the estimates obtained by the corrected model, introduced in the present paper, but with larger asymptotic variances and wider confidence intervals (results not shown). Ridout & Harris (1997) treated the non-proportionality by introducing an alternative model, which assumes an exponential form of the response curve, µ(x), properly located and scaled to cross the origin. Unfortunately, this model does not fit our data. The ‘ Corrected model ’, used in this paper, introduces a new alternative to treat non-proportionality. In fact, one can assume the function µ(x) to be smooth and take a polynomial approximation, using a homogenous polynomium of degree 4 (determined empirically) to approximate µ(x). That is, the corrected model can be seen as an approximation to a curve contained in a large class of curves (the continuous curves crossing the origin). The negative sign of the polynomial coefficients in the data analysis indicates that there is a reduction in the number of cfu on very crowded dishes, i.e. non-proportionality. This may be due to substrate antagonism as defined by Lockwood (1986), antibiosis, colony confluence or a combined effect of these factors. Note especially that the results of the suspension mixtures are based on several hundred cfu per dish (Table 4). Further studies are needed to elucidate the exact reasons for the non-proportionality phenomenon. Non-proportionality may be observed at much lower numbers of colonies, i.e. 25–50 cfu on 9 cm plates (Jensen, 1962), because the growth pattern of filamentous fungi causes strong interference between colonies on most general substrates. The reason why the high cfu numbers did not lead to inconsistent results in the present study is probably a combination of (i) the DYSG medium, which restricts colony growth rate and diameter, (ii) the short incubation time, (iii) the special enumeration technique, and (iv) the use of an appropriate statistical model. Preliminary experiments had demonstrated that after four days of incubation, the majority of cfu had reached detectable size (results not shown). To obtain the most precise estimate on propagule abundance in a given soil, the incubation time giving the highest number of counts can be stated prior to dilution plating as demonstrated for bacteria (Hattori, 1988) The suspension mixture experiment The suspension mixture experiment was designed to address the four questions listed in the introduction. The answer to Question (i) is positive : The DYSG method produces consistent estimates on the abundance of P. verrucosum propagules in soil suspensions provided that the appropriate statistical method is used. In the calibration, there was no indication of non-proportionality (P-value " 0±10), and the Detection and enumeration of P. verrucosum in soil proportional model could estimate the abundance of P. verrucosum at a range of 8–89 cfu on the Petri dishes. This indicates that both intra- and inter-species competition was low. By mixing varying amounts of the IS suspension with a large, fixed amount of the NIS suspension in the main experiment, the consistency of the DYSG method was tested under conditions in which the indigenous soil fungi exerted a very high competition pressure (Question (ii)). The results showed that by using the relatively simple ‘ Corrected model ’, the information on P. verrucosum provided by the DYSG method could be recovered. The DYSG method produced a result of 129 P. verrucosum cfu mg−" soil under high competitive pressures (0±2–15±9 % of more than 700 cfu), using this model. This is relevant, because P. verrucosum is regarded as a rare inhabitant in soil compared to other species of Penicillium. The result of the main experiment is quite comparable to the result obtained in the calibration part of the experiment (126 P. verrucosum cfu mg−" soil), using the proportional model and operating at conditions of much less competition from soil fungi. Question (iii) on whether competition from P. verrucosum on the Petri dishes affects the estimated abundance of indigenous soil fungi was tested in the calibration part of the suspension mixture experiment. The answer is positive. A significantly lower number of indigenous soil fungi was found in the IS as compared with the NIS (P-value ! 0±001), indicating that P. verrucosum suppresses some of the indigenous soil fungi. This effect may be due to antibiosis or substrate antagonism as mentioned above, and it can be easily overcome by a further dilution of the suspension before plating. Question (iv) deals with the sensitivity of the method : Can P. verrucosum be detected when constituting less than one percent of the cfu in a soil suspension ? The answer is positive. In the suspension mixtures, the estimated number of P. verrucosum on a Petri dish varied from 0±02–2±27 cfu cm−#. Due to the large and fixed contribution of indigenous soil fungi from the NIS, the total estimated abundance of soil fungi was 11–12 cfu cm−# in all mixtures. Yet there were no difficulties in the enumeration of P. verrucosum due to its characteristic coloration of the agar reverse. Although the observed numbers were consistently lower than the estimated numbers, this difference never exceeded 27 % (M6). In fact, it decreased with a decreasing density of P. verrucosum on the Petri dishes. In M1–M3, containing 100–500 times as many propagules of soil fungi as of P. verrucosum, the difference between the observed and estimated abundance was lower than 10 %. In an M1 Petri dish, the estimated abundance of P. verrucosum was 1±4 cfu as compared to an estimated abundance of more than 700 cfu of soil fungi. Nevertheless, P. verrucosum was detected on all five replicate Petri dishes of M1 (Table 5). So even when constituting no more than 0±2 % of the estimated cfu (Table 4), P. verrucosum was detected with a very high consistency and in numbers that were differing by no more than 11 % from the estimate. In conclusion, the suspension mixture experiment showed that it is possible to detect P. verrucosum at very low concentrations and at the same time obtain a very precise estimate of its 894 abundance, even when it constitutes only a very small proportion of the cfu on a Petri dish. The soil mixture experiment The soil mixture experiment was performed to confirm the above results under conditions where the low numbers of P. verrucosum on the Petri dishes were caused by low propagule concentrations of the fungus in the soil. To obtain this, IS was gradually thinned with NIS in order to study recovery from soil and approach the lower limit of P. verrucosum detection in soil. The results, reported in Table 6, show that the IS had been mixed very homogenously into the NIS during the thinning procedure (no statistically significant difference in P. verrucosum abundance between replicate subsamples and no difference between the two thinning series). The statistical analysis confirmed that the results of SM1 to SM6 could be pooled to produce an estimate of the abundance of P. verrucosum in the IS, implying that the recovery of P. verrucosum from the soil mixtures with a low concentration of conidia has been as good as from the soil mixtures with a high concentration. The results show that by means of the DYSG method, P. verrucosum can be detected and quantified in soil in concentrations of less than 200 cfu g−" soil. Below 100 cfu g−" soil (SM7), far fewer P. verrucosum than predicted were observed. The DYSG method may, however, be further improved to include the Most Probable Number (MPN) technique in the determination of conidial abundance at extremely low concentrations. When comparing the observed abundance of P. verrucosum in the IS with its abundance, estimated on the basis of SM1 to SM6, only 84 % of the estimated propagules were recorded (Table 6). As discussed for the suspension mixture experiment, this may be due to intra-species inhibition or confluent colonies of P. verrucosum, as it constitutes 62 % of the cfu on the Petri dishes of the IS (Table 7). For comparison, it constitutes 0±3–1±6 % on the Petri dishes of SM4 to SM6, which offer the major contribution to the result on the estimated abundance of P. verrucosum in the IS. In this study, no other fungi developed a coloration on DYSG that was confused with P. verrucosum. In cases of doubt, another advantage of the method is the possibility of recultivating the fungi and using the strains for further studies, e.g. for confirmation of their ability to produce OA. Recultivation was done successfully even at high competitive pressures. In conclusion, the soil mixture experiment confirmed the findings of the suspension mixture experiment, i.e. that the sensitivity of the DYSG method is very high. Less than 200 cfu g−" soil can be detected and quantified with good precision using this simple and classical method. It can be performed in most microbiological laboratories and requires no expensive investments. The presented method enables detailed studies on the ecology of P. verrucosum and a possibility to conduct large surveys of the distribution of the fungus in soil, also when it occurs in very low conidial concentrations compared to other soil fungi. Dr J. C. Frisvad is gratefully acknowledged for supplying the strain of Penicillium verrucosum used in this work as well as for Susanne Elmholt and others fruitful discussions on the qualities of DYSG. The work was funded by the Research Programme on Food and Technology (FØTEK 1), initiated by the Danish Ministry of Food, Agriculture and Fisheries. REFERENCES Boorman, G. A. (1989). Toxicology and carcinogenesis studies of ochratoxin A (Cas no. 303-47-9) In F344}N rats. (gavage studies). NTP Technical Report 358, U.S. Department of Health and Human Services. Research Triangle Park, NC. Ciegler, A., Fennell, D. I., Sansing, G. A., Detroy, R. W. & Bennett, G. A. (1973). Mycotoxin-producing strains of Penicillium viridicatum : classification into subgroups. Applied Microbiology 26, 271–278. Davis, N. D., Diener, U. L. & Eldridge, D. W. (1966). Production of aflatoxins B and G by Aspergillus flavus in a semisynthetic medium. Applied " " Microbiology 14, 378–380. Diener, U. L. (1955). Sporulation in pure culture by Stemphylium solani. Phytopathology 43, 141–145. Elmholt, S. (1991). Side-effects of propiconazole on non-target soil fungi in a field trial compared to natural stress effects. Microbial Ecology 22, 99–108. Elmholt, S. (1996). Microbial activity, fungal abundance, and distribution of Penicillium and Fusarium as bioindicators of a temporal development of organically cultivated soils. Biological Agriculture and Horticulture 13, 123–140. Elmholt, E., Frisvad, J. C. & Thrane, U. (1993). The influence of fungicides on soil mycoflora with special attention to tests of fungicides effects on soilborne pathogens. In Pesticide Interactions in Crop Production (ed. J. Altman), pp. 227–243. CRC Press : Boca Raton. Elmholt, S. & Hestbjerg, H. (1996). Recovery and detection of deuteromycete conidia from soil. In Monitoring Antagonistic Fungi Deliberately Released into the Environment (ed. D. F. Jensen, H. B. Jansson & A. Tronsmo), pp. 49–55. Kluwer Academic Publishers : Dordrecht, The Netherlands. Elmholt, S. & Kjøller, A. (1989). Comparison of the occurrence of the saprophytic soil fungi in two differently cultivated field soils. Biological Agriculture and Horticulture 6, 229–239. Fisher, R. A., Thornton, H. G. & Mackenzie, W. A. (1922). The accuracy of the plating method of estimating the density of bacterial populations. Annals of Applied Biology 9, 325–359. Frisvad, J. C. (1981). Physiological criteria and mycotoxin production as aids in identification of common asymmetric penicillia. Applied and Environmental Microbiology 41, 568–579. Frisvad, J. C. (1983). A selective and indicative medium for groups of Penicillium viridicatum producing different mycotoxins in cereals. Journal of Applied Bacteriology 54, 409–416. Frisvad, J. C. (1986). Selective medium for Penicillium viridicatum in cereals. In Methods for the Mycological Examination of Food (ed. A. D. King, Jr, J. I. Pitt, L. R. Beuchat & J. E. L. Corry), pp. 132–135. Plenum Press : New York. Frisvad, J. C. & Filtenborg, O. (1989). Terverticillate penicillia : chemotaxonomy and mycotoxin production. Mycologia 81, 837–861. Frisvad, J. C., Filtenborg, O., Lund, F. & Thrane, U. (1992). New selective media for the detection of toxigenic fungi in cereal products, meat and cheese. In Modern Methods in Food Mycology (ed. R. A. Samson, A. D. (Accepted 11 September 1998) 895 Hocking, J. I. Pitt & A. D. King), pp. 275–284. Elsevier Science Publishers : Amsterdam. Frisvad, J. C. & Samson, R. A. (1991). Mycotoxins produced by species of Penicillium and Aspergillus occurring in cereals. In Cereal Grain (ed. J. Chelkowski), pp. 441–276. Elsevier Science Publishers : Amsterdam. Hattori, T. (1988). The Viable Count. Quantitative and Environmental Aspects Science Tech Publishers : Madison, Wisconsin. Ho$ kby, E., Hult, K., Gatenbeck, S. & Rutqvist, L. (1979). Ochratoxin A and citrininin in 1976 crop of barley stored on farms in Sweden. Acta Agriculturæ Scandinavica 29, 174–178. Jensen, V. (1962). Studies on the microflora of Danish beech forest soils. I. The dilution plate count technique for the enumeration of bacteria and fungi in soil. Zentralblatt fuX r Bakteriologie, Parasitenkunde, Infektionskrankheiten und Hygiene, II Abteilung 116, 13–32. Jørgensen, K., Rasmussen, G. & Thorup, I. (1996). Ochratoxin A in Danish cereals 1986–1992 and daily intake by the Danish population. Food Additives and Contaminants 13, 95–104. Kornerup, A. & Wanscher, J. H. (1978). Methuen Handbook of Colour. 3rd ed. Eyre Methuen : London. Krogh, P. (1987). Ochratoxins in food. In Mycotoxins in Food (ed. P. Krogh), pp. 97–121. Academic Press : London. Lillehoj, E. B. & Elling, F. (1983). Environmental conditions that facilitate ochratoxin contamination of agricultural commodities. Acta Agricultura Scandinavica 33, 113–128. Lockwood, L. H. (1986). Soilborne plant pathogens : Concepts and connections. Phytopathology 76, 20–27. Lund, F., Thrane, U., Frisvad, J. C. & Filtenborg, O. (1992). IBT List of Fungal Cultures and their Secondary Metabolites. 1st ed., Department of Biotechnology : Technical University of Denmark, Lyngby. McCullagh, P. & Nelder, J. A. (1989). Generalized Linear Models. 2nd Edition. Chapham & Hall : London. Parkinson, D. (1994). Filamentous Fungi. In Methods of Soil Analysis, Part 2Microbiological and Biochemical Properties (ed. R. W. Weaver et al.), pp. 329–351. SSSA Book Series 5. SSSA : Madison, Wisconsin. Pitt, J. I. (1979). The Genus Penicillium and its Teleomorphic States Eupenicillium and Talaromyces. Academic Press : New York. Pitt, J. I. (1987). Penicillium viridicatum, Penicillium verrucosum, and production of ochratoxin A. Applied and Environmental Microbiology 53, 266–269. Ridout, M. S. & Harris, D. C. (1997). Estimating the concentration of propagules of a plant pathogen from soil dilution plate counts when the response is non-linear. Applied Statistics 46, 111–121. Roberts, E. A. & Coote, G. G. (1965). The estimation of concentration of viruses and bacteria from dilution counts. Biometrics 21, 600–615. Scott, P. M., Lawrence, J. W. & van Walbeek, W. (1970). Detection of mycotoxins by thin-layer chromatography : application to screening of fungal extracts. Applied Microbiology 20, 839–842. Sinha, R. N., Abramson, D. & Mills, J. T. (1986). Interrelations Among Ecological Variables in Stored Cereals and Associations with Mycotoxin Production in the Climatic Zones of Western Canada. Journal of Food Protection 49, 608–614. Smith, J. E., Lewis, C. W., Anderson, J. G. & Solomons, G. L. (1994). Mycotoxins in Human Nutrition and Health. European Commission. Science, Research and Development. EUR 16048 EN. Warcup, J. H. (1957). Studies on the occurrence and activity of fungi in a wheat-field soil. Transactions of the British Mycological Society 40, 237–262.