INTRODUCTION

Top Abstract Introduction Materials & Methods Results Discussion References

Verticillium dahliae Kleb. causes wilt diseases in agronomic, horticultural, and nursery plant species (18, 19). Microsclerotia, which form in the senescing tissues of the diseased plant, may persist in soil for several years in the absence of a susceptible host. Quantifying the density of microsclerotia in soil is important for the development of disease prediction systems and for assessing the performance of control tactics.

Many methods have been devised for quantifying V. dahliae in soil. These include elements such as concentrating microsclerotia by flotation (2, 9), wet sieving soil samples (11), and using semiselective media, including incorporation of a selective carbon source such as ethanol (15), cellulose (23), sorbose (13), or pectate (10). Cellulose also has been added to nutrient media to improve discrimination between V. dahliae and V. tricorpus Isaac (8, 23). Wet plating (WP) (14, 15) and dry plating (DP) (3, 7) are the two methods of distributing soil onto an agar medium. Both methods involve plating a small amount of soil on a semiselective medium which usually contains pectin as a carbon source. WP assays may include wet sieving as a step to remove particles smaller (e.g., <20 µm) and larger (e.g., >160 µm) than microsclerotia (6). With DP assays, the Andersen air sampler (1, 4) distributes the soil particles as evenly as possible onto agar plates (3). In both assay types, soil is dried prior to analysis to reduce the number of conidia and mycelial fragments of V. dahliae (14), to reduce the inoculum of other fungi, and to facilitate grinding and mixing of the soil sample.

Existing methods of analyzing soil for V. dahliae can be compared either by having a single laboratory following published methods for the same soil sample or by having different laboratories each follow their own protocols with a common set of soil samples. Although several comparative studies using the first approach have been published (3, 6, 16, 21, 25), the question of which method is best remains unresolved. A difficulty encountered with the first approach is how to reproduce exactly the published methods of other workers, since minor details may be important (6). For this reason, we took the second approach in an attempt to identify which method has the highest precision (i.e., the least variation between repeated measurements), recovery percentage (i.e., accuracy, or the nearness to the true value; bias is the reverse of accuracy), and sensitivity (i.e., the detection limit).

Our objectives were as follows: (i) to inventory the methods currently used for estimating V. dahliae in soil and (ii) to compare the results of various methods over a range of soils. Preliminary results of this study have been published previously (24).

	MATERIALS AND METHODS

Top Abstract Introduction Materials & Methods Results Discussion References

Exchange of soil samples. The contacts for the research groups were as follows: J. R. Davis (United States), O. C. Huisman (United States), G. Lazarovits (Canada), T. Locke (United Kingdom), J. M. Melero Vara (Spain), L. Mol (The Netherlands), M. Powelson (United States), D. I. Rouse (United States), A. J. Termorshuizen (The Netherlands), E. J. Paplomatas (Greece), H. W. Platt (Canada), R. C. Rowe (United States), and L. Tsror (Israel). Throughout this paper, the results will be attributed anonymously; groups will be referred to by uppercase letters, and the methods they used and soils they provided will be referred to by lowercase letters. Each group, with the exception of two, collected one soil sample of about 5 kg. Each sample was air dried and ground. Samples were sent by airmail to the senior author (A.J.T.), who divided and distributed the subsamples. A small portion of each sample was retained by each group to determine the effect of transportation. To treat the Dutch samples in a manner similar to the foreign samples, they were airmailed to the United Kingdom and back.

1

Identification of V. dahliae. Four groups reported that they could not distinguish between colonies of V. dahliae and V. tricorpus. The other groups did attempt to discriminate between the two species, but not all were totally confident in their identifications. On average, figures for V. tricorpus were 3.9% of those for V. dahliae plus V. tricorpus. Given the apparently low incidence of V. tricorpus, the uncertainty of several groups in identifying the two species, and the desirability of comparing all the data in one analysis, data for the two fungi were combined.

Analysis of data. Data of two groups, L and M, were omitted from the analysis because the data were anomalous. Group L reported averages of 90 and 89 CFU g of soil¹, respectively, for the samples amended with 5 and 60 mpg. The results obtained by group M were extremely variable and therefore regarded as unreliable. For example, their data showed the largest mean value (490 CFU g of soil¹) of all methods for densities of V. dahliae plus V. tricorpus, but their median was 0.

1

 + _j + _ij + e_ijk

 + _j + _ij + s_ijk + p

	RESULTS

Top Abstract Introduction Materials & Methods Results Discussion References

Differences between transported and nontransported samples were significant (P < 0.05) for three groups by t tests. After correcting for the number of comparisons, using the Bonferroni approach (testing at a significance level of 0.05), two comparisons showed significant effects. Group H reported that numbers of V. dahliae plus V. tricorpus in the transported samples were 3.3 times lower than those in nontransported soil (see Fig. 4). Results of group E (see Fig. 4) and of method b1 (see Fig. 3) of group B showed the opposite effect: the numbers of V. dahliae plus V. tricorpus were 1.8 times (not significant) and 7.8 times (significant) higher in the transported samples than in the nontransported samples. The high degree of recovery in the transported sample measured by method b1 is due to a single extreme value reported for one subsample (Fig. 3).

View larger version (19K): [in this window] [in a new window]   FIG. 1.   Densities of V. dahliae plus V. tricorpus in six subsamples per soil sample (squares) and their averages (asterisks) per method for soil samples that were artificially infested with 5 and 60 mpg. The methods are described in Table 1.

The naturally infested soil samples differed widely in their content of V. dahliae, which ranged on average from 1.1 to 120 CFU g of soil¹ (Table 2). The results obtained by the different groups for the same soils differed markedly (Fig. 1 to 4). The means of the nontransformed data ranged from 0.57 CFU for method a to 67 CFU for method k (Table 3). The duplicate analyses b1 and b2 by group B yielded generally similar data. Group F reported twice as many colonies with its DP assay (method f-2) as with its WP assay (method f-1) (Table 3). Numbers obtained by DP assays were generally higher than those obtained by methods based on WP, although group G (method g) reported obtaining relatively low values by the DP-based method (Table 3).

Median and backtransformed values per method were considerably lower than the arithmetic means (Table 3). This was due to the high frequency of zero values in the subsamples (DP assays, 14%; WP assays, 7.7%) and the petri dish counts (DP assays, 39%; WP assays, 37%).

An ANOVA of log-transformed subsample means with method, sample, and method × soil interaction as sources of variation explained 88% of the total variation (r²), and all means were highly significant (Table 4). The interaction indicates that certain methods work better for certain soils, although this effect is relatively small compared to the main effects (Table 4). Residual plots, used to check the goodness of fit of the model, showed a few studentized residuals with absolute values of larger than 4. This appeared to be due to some outliers of groups B (method b1) and H (Fig. 1 to 4).

View larger version (14K): [in this window] [in a new window]   FIG. 2.   Densities of V. dahliae plus V. tricorpus in six subsamples per soil sample (squares) and their averages (asterisks) per method for soil samples with low infestation levels. Soil and method codes are explained in the text. The methods are described in Table 1. Method labels marked with an asterisk indicate that the results of those of nontransported soil samples.

View larger version (14K): [in this window] [in a new window]   FIG. 3.   Densities of V. dahliae plus V. tricorpus in six subsamples per soil sample (squares) and their averages (asterisks) per method for soil samples with medium infestation levels. Soil and method codes are explained in the text. The methods are described in Table 1. Method labels marked with an asterisk indicate that the results are those of nontransported soil samples.

View larger version (14K): [in this window] [in a new window]   FIG. 4.   Densities of V. dahliae plus V. tricorpus in six subsamples per soil sample (squares) and their averages (asterisks) per method for soil samples with high infestation levels. Soil and method codes are explained in the text. The methods are described in Table 1. Method labels marked with an asterisk indicate that the results are those of nontransported soil samples.

For the samples containing 5 or 60 mpg, maximum mean recoveries were 59% (group H, by the DP assay) and 41% (groups H, I, and K, by the DP assay), respectively (Table 5). The maximum mean recoveries determined by WP assays were 39 and 25% (both group E), respectively. There was an insignificant trend toward a lower percent recovery at the higher population level.

Large fluctuations among groups occurred for the estimates of variance among subsamples (s_s²) and among petri dishes (s_p²) (Table 6). For all methods s_p² was larger than s_s². In all cases except method g, methods based on DP resulted in larger s_p² values than did those based on WP. The large residual for method h seems to be due to variation among subsamples rather than variation among petri dishes.

	DISCUSSION

Top Abstract Introduction Materials & Methods Results Discussion References

Our results show that laboratories differ widely in their ability to quantify V. dahliae in soil. One problem which had not been anticipated was the difficulty that some groups experienced in distinguishing colonies of V. dahliae from those of V. tricorpus. V. tricorpus is a commonly occurring saprophyte or weak pathogen (12) that can be distinguished from V. dahliae by its dark hyphae and chlamydospores (12, 20). Those confident in making this distinction reported low numbers of V. tricorpus compared to V. dahliae on plates; therefore, for the purpose of analyzing the data, we assumed that including this species would have no substantive effect on the conclusions drawn.

Methods were based on either DP or WP of soil, but there were many other variations (Table 1), probably including several which are insignificantfor example, the precise manner in which the Andersen air sampler was handled. Nevertheless, it is clear that with the exception of group G, the DP assays (methods f2 to k) yielded higher numbers of CFU per gram of soil than did the WP assays (methods a to f1). The parallel comparison by group F of WP and DP assays (methods f-1 and f-2) strongly supports this conclusion. The difference in the recovery rates of DP and WP assays may be explained in part by the loss of microsclerotia from wet sieving (16, 21). However, the results of group E, which did not sieve, suggest that other factors also may be involved. Thus, the conditions for germination of microsclerotia may be better if dry, small, discrete aggregates of soil, rather than an aqueous suspension of aggregates, are dispersed over the agar surface. In their comparison of methods, Nicot and Rouse (16) also found that recoveries from naturally infested soils by a DP-type assay were at least twofold higher than those of dilution plating (a WP-type assay). However, the DP-type assay was found to be biased in contrast to the dilution plating method. This remarkable result may be due to an artifact caused by the use of steamed silica sand and artificially produced inoculum to quantify bias.

Certain aspects of the methodology appear to be less important than had been thought, while others may be more important than was realized. Thus, groups I and K, using a pectin carbon source in the plating medium, obtained results similar to those of groups H and J, which used sucrose. The results of group E indicate that wet sieving of soil samples in the WP assays is not as advantageous as previously reported (5, 21). The pH of the plating medium was highly correlated with the percent recovery from the soil to which 60 microsclerotia were added per gram (Pearson correlation coefficient = 0.76; P < 0.01). A medium of pH 5.6 or less was associated with low counts, and media of pH ca. 7.0 were associated with the highest recoveries.

The analysis method × soil sample interaction was significant (Table 4). Thus, for soil sample l, group B obtained counts higher than all other methods, whereas with all of the other soils, the counts were generally low (Fig. 1 to 3). Similarly, for soils i and k, group I obtained lower mean counts than did groups J and K, whereas groups I, J, and K had similar counts for the other soils. The method × soil interaction may be caused by different microbial communities that act according to, for instance, the plating medium used. Recovery percentages could be used to obtain less-biased estimates of the densities in soil samples containing unknown inoculum densities, but this approach is precarious if the analysis × method interaction is significant. For example, the best estimate for the overall mean for group G would be 552 CFU g of soil¹ (the mean [Table 3] divided by R₆₀/100 [Table 5] = 16/0.029), whereas for group K it would be 163 CFU g of soil¹ (67/0.41). Thus, the noncritical use of percent recoveries from samples with known densities is not necessarily the correct approach for improvement of detection assays. In general, however, the interaction effect was small compared to the effect of the analysis method, indicating that the performance of most methods is soil type independent.

The comparison of transported with nontransported samples gave inconsistent results. However, transport of samples was not a major factor influencing the results of this study. The degree to which V. dahliae was affected during transportation would depend on the conditions to which it was subjected, and these conditions would have been unique to each consignment for each destination. In addition, the mode of separating the sample into the transported and nontransported subsamples was not prescribed and may have been biased.

Variation among plate counts was 1.3 to 36 times greater than that among subsamples (Table 6). The generally higher degree of variation among petri dishes for the DP assay than for the WP assay may be attributed to the small subsample (25 to 100 mg) leading to more zero values for the former. Therefore, at levels below 1 CFU g of soil¹, which may still be epidemiologically significant (5, 17), the higher degree of accuracy of the DP assay than the WP assay may be outweighed by its poorer precision and sensitivity limits. Variability was slightly higher among subsamples tested by WP methods than with DP assays, probably because of the sub-subsampling needed to prepare the soil dilutions for the WP assays.

The variance of an estimator for the infestation of a sample can be expressed as the sum of the estimated variances among subsamples, s_s², divided by the number of subsamples, I, and the estimated variance among petri dishes, s_p², divided by I and divided by the number of petri dishes per subsample J. If the number of petri dishes is held constant, the variance decreases only if I is increased. It is possible to calculate the amount of subsamples (I) needed to reach a certain variability at a given probability. Snedecor and Cochran (22) showed that, assuming a normal distribution and known variance ², half the width of a 0.95 confidence interval L = 1.96 ×  × I^0.5. For a given L, this results in I = ² × 1.96² × L² = (/µ)² × 1.96² × (L/µ)² = SMR² × 1.96² × (L/µ)², where SMR is the standard deviation-to-mean ratio, µ is the (unknown) expectation, and L/µ is half the width of the confidence interval expressed as a fraction of the mean. The SMR per sample (n = 6) varied from 0.3 to 0.7 per group. At these extremes, the numbers of samples required to realize an L/µ of 0.3 would be 4 and 21, respectively. It is apparent from Fig. 5 that it is highly advantageous to select for methods that produce low SMR values.

View larger version (22K): [in this window] [in a new window]   FIG. 5.   Sample size as a function of standard deviation-to-mean ratio at four different values of L/µ (i.e., half the width of the confidence interval [L] expressed as a fraction of the mean [µ]).

Our results lead to the conclusion that the data available in literature presenting densities of V. dahliae in soil are difficult to interpret. For this reason, methods should be compared with those of other laboratories and control soil samples containing known densities of microsclerotia should be included. Ideally, research workers should settle on a single well-defined protocol to increase the comparability of results obtained at different laboratories. In conclusion, it is apparent that WP assays are less accurate than DP assays. Additional experimental analysis of the conditions required for microsclerotia to germinate and to develop new microsclerotia in the plating medium is needed to explain the substantial variation noted within the two assay types. Method components that probably deserve experimental attention are plating medium composition and petri dish incubation conditions. The low recovery percentages for both assay types indicate that there is room for improvement of the methodologies.