DAS ELISA test establishment

DAS (double-antibody sandwich) - ELISA test is a variant of direct ELISA test, developed for the detection of plant viruses with a specific enzyme-antibody conjugate for each particular virus. In this case the test was established with samples of ground seed (2 g of seed) in 96-well polystyrene plates. The reagents used thus the protocol were from Agdia^®, DAS Elisa (2013) for BSMV.

Sensitization and plate washing

The antibody coverage was diluted 1: 200 for this where mixed 10 000 µl of the coverage buffer (1X), pH 9.6, with 50 µl of antibody coverage concentrated; except for blank wells, in each well 100 µl of antibody dilution coverage were added and incubated for 4 h at room temperature or overnight at 4 °C, in humid chamber. After incubation, the content of the plate was emptied and washed with PBST (1X) buffer wash with a multichannel wash bottle. This process was repeated three times and between each wash the plate was strongly shacked down on a folded paper towel to remove excess liquid and bubbles.

Loading samples

To each of the tubes with ground seed samples were added 20 ml of extraction buffer (1X) 1:10 ratio (sample weight in g: buffer volume in ml). Except blank wells, added 100 µl per well of each sample per duplicate, thus positive control, negative control and the buffer according to the preset load diagram. The plate was incubated for 2 h at room temperature or overnight at 4 °C, in humid chamber. Upon completion of the incubation the plate was emptied and washed eight times with wash buffer, in order to remove any residue from the sample. Between each wash the plate was shacked, down on a folded paper towel to remove excess liquid and bubbles.

Addition of enzyme conjugate

Shortly before using the enzyme conjugate (alkaline phosphatase) it was diluted to a ratio 1: 200. For this 10 000 µl of ECI buffer (1X) with 50 µl of enzyme conjugate concentrate were mixed; except blank wells, in each well 100 µl of enzyme dilution were added. After 2 h incubation in a moist chamber, the plate was emptied and washed eight times with wash buffer, in order to remove any residual sample. Between each wash the plate shacked, down on a folded paper towel to remove excess liquid and bubbles.

Reveil plate

For revealed a PNP (p-nitrophenyl phosphate) solution at 1 mg per ml with two PNP tablets (substrate) dissolved in 10 ml PNP buffer (1X) in a shaker Vortex Scientific Industries G-560 was prepared. It avoided touching the tablets and expose the PNP solution to intense light to avoid contamination and false positive development by color development in the wells of negative samples. Immediately after preparation, 100 µl of the PNP solution were added in each plate well and incubated in the dark at room temperature for 60 min. Elapsed time of incubation, 70 µl of sodium hydroxide (NaOH) 3 molar per well were placed to stop the reaction in the plate. Reading the plate was performed in a BioTek ELx808 spectrophotometer at 405 nm.

Preparation of single samples

From two seed batches of bread wheat (Triticum aestivum L.), one naturally infected seed (SI) with barley stripe mosaic virus (BSMV and healthy seed (SS), single samples (MS) 20 g of seed were formed. Each MS sample was prepared according to the level of infection (NI): NI 15%, 3 g of SI with 17 g of SS; NI 10%, 2 g of SI with 18 g of SS; and NI 5%, 1 g of SI with 19 g of SS. Once MS samples are formed these were deposited separately in paper envelopes properly identified. In total 210 MS samples (70 MS samples per NI) were prepared of which 2 g of seed each were taken to be ground and analyzed separately by DAS-ELISA according to the protocol described above.

Preparation of sample size and composite samples

Of each of the 210 MS samples were taken separately, 1.0, 1.9 and 2.56, g of seed to form sample sizes (TM) of 5, 10 and 15% respectively (Figure 1). For this from an initial batch of 20 g, 1 g was taken to form TM of 5%; the remaining 19.0 g, 1.90 g were taken to form TM 10%; and from the remaining 17.1 g, 2.56 g were weighted to form TM 15%.

Figure 1. Diagram for composition of single samples (MS) with their level of infection and structuring of composite samples (MC= NMS + TM + NI). 

The composite samples (MC) were formed with different numbers of samples MS (25, 20, 15 and 10 NMS). In total 36 MC samples, each MC sample with three NI and three TM and a total weight according to TM (Figure 1) were formed. For practical management purposes, the above MC were labeled considering first the code to number of MS samples, then the code for TM and last NI. For example, the sample labeled MC25-15-15 indicates that is about a MC sample with 25 MS samples with TM 15% and NI 15%. While the MC25-10-15 indicates that is a MC sample of 25 MS samples with TM 10% and NI 15%. While MC25-5-15 indicates that is a MC sample with 25 MS samples, with TM 5-% and NI 15%. This methodology was applied to the remaining 33 samples.

Figure 1 and Table 1 illustrates the formation scheme MS and MC samples.

Evaluation of the number of MS, TM and NI on composite samples

To determine whether NMS, the increase or reduction in the size TM and NI level that make MC samples affects the sensitivity of DAS-ELISA test to BSMV, these were analyzed separately each of the 36 MC samples. Of each MC sample, three replications of 2 g of seed each (108 subsamples in total) were taken, which were deposited in falcon tubes type of 50 ml properly identified to be ground separately in a mill Peter Instruments. Ground seed was recovered in the same tube. Between each grinding the mill was cleaned with compressed air and seed fragments stuck in the mill discs were removed with a dissecting needle to prevent contamination between samples. The procedure for analysis of the subsamples was made as described in the establishment of DAS-ELISA test for sensitization and plate washing, adding samples, addition of the enzyme conjugate and plate reveil, Agdia protocol, Das Elisa (2103) for BSMV.

Statistical analysis

Absorbance data obtained from the evaluation of MS and MC samples were analyzed with the statistical program SAS system for Windows 9.4, using the analysis of variance the lineal model corresponding to a randomized complete block design (RCBD) with sub - sampling, with the procedure of general linear model (GLM) and multiple comparisons test using mean honest significant difference (HSD) from Tukey, with a significance level of 5% (α= 0.05).

NMS and NI effect on sensitivity of DAS ELISA test

Table 2 shows the average values and mean comparison of absorbance in different NMS used to form MC samples with different NI from BSMV, using honest significant difference (HSD) of Tukey with a significance level of 5% (α= 0.05). With levels 15, 10 and 5% NI recording samples with values above the threshold value (0.22). Similarly, the mean values per group and general means were also higher than the threshold. However, when analyzing the results by level and between levels differences were found.

The overall average value (0.92) recorded with NI level 15% was significantly different (α= 0.05) by to that registered with NI 10 and 5%, which showed statistically similar values (0.73 and 0.63, respectively). This indicates that by reducing the NI level from 15% to 10 and 5% was detected a percentage of decrease of 20 and 34%, respectively, of the sensitivity of DAS-ELISA test to BSMV. These results confirm the results found by ^{Scott and Zummo (1995)} who report that as the level of natural incidence of a pathogen is reduced in the seed the likelihood of not detecting it increases.

By analyzing the comparison of means of absorbance at different NMS, it was found that in NMS of 25, 20 and 10 forming the MC samples have no statistically significant differences (α= 0.05) except for NMS 15 (0.85) that showed differences with the previous. When comparing the NMS with lower absorbance value to the threshold (0.22), samples with NI 10 and 5% recorded the highest number of negative MS samples (14 and 24 samples, respectively) compared to samples with NI 15% (6 samples). That is, as NI decreased from 15% to 10 and 5% the number of negative samples to BSMV increases (data not shown). In short with a level of infection of 15% a percentage of 91.5% probability of detecting BSMV by DAS-ELISA test was detected; that is, from the 70 MS samples with NI 15%, 64 of them were positive to the virus; instead with NI 10% detected a percentage of 80% and with NI 5% detected a percentage of 66% probability of detecting BSMV.

The level of infection of a seed or incidence of the pathogen in the crop is key for its detection (^{Scott and Zummo, 1995}) also determines the sample size or composite sample, as found by ^{Priou et al. (2001)}, that reducing the composite sample size of potato tubers are based on the level of incidence of Ralstonia solanacearum in crops.

Effect of NMS, TM and NI on composite sample

Table 3 shows the absorbance average results of 36 MC with DAS-ELISA test. Overall average absorbance values of MC samples were above the threshold value of absorbance (0.21) only in the sample MC10-10-5 the absorbance value (0.18) remained below the threshold, as in 4 of 6 individual observations (replications) obtained a lower absorbance than threshold. Although MC10-15-15, MC10-10-10 and MC10-5-5 recorded two replications below the threshold, by averaging the six replications this exceeded the absorbance threshold, marking it as positive to BSMV. This indicates that when MC samples are formed with a NMS lower to 15 or equal to 10 decreases the sensitivity of DAS-ELISA test to BSMV and the probability of recording false negatives increases.

Because when comparing the 36 values showed in Table 3, obtained from the combination of the three factors, as if they were individual treatments does not make much sense, an analysis of variance was performed considering the factorial structure of the treatments, 3 factors: i) number of single samples (NMS) with 4 levels (25, 20, 15 and 10); ii) Sample size (TM) with three levels (15, 10 and 5); and iii) level of infection (NI) with three levels (15, 10 and 5). Table 4 shows the results of multiple comparisons of means using honest significant difference (HSD) Tukey significance level of 5% (α= 0.05), for each of the principal effects. Now the averages presented come from a higher number of individual values (N).

The results from analysis of the comparison of means for the overall principal effect of absorbance between MC samples with different NMS, TM and NI indicate statistical differences (α= 0.05) between samples. Specifically highlights differences between the averages recorded by MC samples with NMS 25 (2.46a) and NMS 20 (2.11a) with MC samples with NMS 15 (1.72b) and NMS 10 (0.79c) and differences between the latter. This indicates that as NMS decreases in the MC sample, statistically decreases the absorbance value, confirming the findings by ^{Fernández (2007)} mentioning that large sample sizes recorded higher confidence level and degree of accuracy in contrast to what was presented by small samples.

When comparing the mean absorbance between MC samples with different TM statistical differences (α = 0.05) between samples were found. It is noteworthy that these differences occurred only between MC samples with TM 15 (1.9a) and 5 (1.8a) with MC with TM 10 (1.5b). That is, between the sizes of 15 and 5% are no statistical differences but between these and the size 10% are significant differences. This may be related to the fact that MC10-10-5 sample had two samples (replications) with value (0.15 and 0.13) below the threshold (0.21).

In this case a linear relationship between TM from MC sample with the absorbance value was not found; that is, there is no tendency between means from composite samples according to sample size. These results agree with those found by ^{Moreno and Castillo (1978)} for classification of coffee trees, who indicate that a sample of 250 g used for grain size can be reduced to 100 or 50 g without causing any alteration. As well as those reported by ^{Thomas et al. (2005)} who mentioned that there is no difference in the infection rate between 40 and 80 composite samples obtained from wheat batches of 200 and 500 t, respectively, infected with Microdochium nivale.

Several researchers point out that the analysis of small sample sizes results in lower confidence level and degree of accuracy in contrast with those registered with larger sample sizes which record better results (^{Lawrence et al., 1995}; ^{Fernández, 2007}; ^{Miyamoto et al., 2008}; ^{Montesinos et al., 2010}). However, ^{Thomas et al. (2005)}) indicate that infected seed batches with Tilletia tritici is the level of infection and not the NMS from MC samples which determines the capability of a test to detect the fungus.

According ^{Nyrop et al. (1999)} when sampling plans are developed it should start by checking what is known about the distribution of sampling and perform a sensitivity analysis to determine whether a refinement of this information is justified, and that the sample size should be based on information from the representative sample, as this is a good indicator of the incidence of pest management unit.

By analyzing the means for the principal effects of absorbance between MC samples with different NI statistical differences (α= 0.05) between the samples were found. In this case means of 2.18a, 1.71b and 1.42c for MC samples with NI 15, 10 and 5%, respectively (Table 4) were recorded. As NI decreased, the absorbance value from MC sample significantly decreases. Figure 2 shows the linear trend of the average absorbance values of MC samples according to NMS, TM and NI with only the MC10-10-5 value below the threshold line.

Figure 2. Trend of average absorbance values of composite samples (MC) of wheat seed. Each MC sample was formed with different number of single samples (NMS, 25, 20, 15 and 10), sample size (TM, 15, 10 and 5%) and level of infection (NI, 15, 10 and 5%) from BSMV. The values from the brown line indicate the average of three MC samples; dotted purple, the linear trend of values; and blue, the average absorbance threshold (0.21). 

^{Priou et al. (2001)} found that as the level of disease incidence by Ralstonia solanacearum decreases in culture, fungus detection decreases in tuber. Therefore these authors suggest to analyze large composite samples when disease levels are low in the crop. Similarly, ^{Scott and Zummo (1995)} observed that when the level of natural infection with Aspergillus flavus is high (3%) the size of sample to be analyzed is reduced but when the level of infection is very low not even with larger sample sizes detects the fungus.

The results described above are evidence of the importance that should be given to sampling as a fundamental activity in a seed testing program (^{Morrison, 1999}). ^{ISTA (2013)} points out that a batch can be composed of harvested seeds from a single field or several sections of land and the objective of sampling from seed will be to obtain a representative sample of adequate size so that the results of the laboratory analysis reflect the quality of the seed batch. Knowing the importance that parameters of sample size and level of infection in samplings of seed batches is essential to determine the probability of selecting positive or negative samples to the pathogen of interest which will result in reliability of results and in making decisions about the final destination of the batch assessed.