Introduction

In European countries, potato seed production is monitored for viruses and other pathogens by law due to the potentially high yield loss (Hepting, 2002). The six main viruses can cause a yield loss of: 20–90% Potato Leaf Roll Virus (PLRV), 50–90% Potato Virus Y (PVY), 20–50% Potato Virus M (PVM), 5–75% Potato Virus S (PVS), up to 40% Potato Virus A (PVA) and up to 10% Potato Virus X (PVX) (Bauch et al., 2012; Kolychikhina et al., 2021).

Introduced about forty years ago (Gugerli & Gehriger, 1980), the standard procedure for virus detection in the seed potato certification process was, and in some regions still is, the Double Antibody Sandwich-Enzyme Linked Immunosorbent Assay (DAS-ELISA). As detection of PVY is enhanced when sprouting is induced, tubers were treated with rindite gas (2-chloroethanol, 1,2 dichloroethane and Tetrachlorcarbon, volume ratio 7:3:1). Sprouts then were tested using DAS-ELISA. Since rindite was known to be cancerogenic it has been regulated by EU occupational safety law. And because DAS-ELISA on tubers is unreliable (Spiegel & Martin, 1993) the procedure changed to growing on DAS-ELISA in the respective countries. Growing on DAS-ELISA (ELISA) detects potato virus in extracts obtained from leaves of four-weeks-old, cultivated tuber-eye-cuttings. This method is time-consuming and requires big greenhouse capacities. For that reason, various potato virus diagnosis protocols were developed using PCR and thus are suitable for dormant tubers (RT-PCR: Singh et al., 2000; RT-qPCR: Hühnlein et al., 2016; Boonham et al., 2009; Agindotan et al., 2007; Mortimer-Jones et al., 2009; Hühnlein et al., 2013; Macroarray: Agindotan & Perry, 2007; RT-IC-PCR: Ahouee et al., 2010; NASBA: Leone et al., 1997; LAMP: Ju, 2011). In Switzerland, conventional RT-qPCR has been established for potato virus diagnosis (Schumpp et al., 2016; Schumpp et al., 2021; Agroscope video: https://www.youtube.com/watch?v=o2jXuCzVdYk).

However, most conventional RT-qPCR methods use expensive extraction methods and consequently need to reduce the number of reactions. Therefore, only a small amount of seed pools that contain a high number of tubers are tested in the respective certification process. A small number of pools, even if they consist of many tuber heel ends, leads to a lower statistical resolution of the percentages of virus per seed lot. Thus, for distinction of different seed categories of some cases in the certification process, especially for the low percentages for pre-breeding and basic seed lots (1 or 2% virus infestation) and the distinction between 8 or 10% (ZA and ZB certificates in Germany) virus per field sample, a higher number of tested pools is necessary but very costly. Therefore, we set out to find more cost-efficient methods to use instead of the conventional RT-qPCR methods, that use expensive chemicals for RNA extraction, for the analysis of higher virus rates.

Two triplex RT-qPCRs, that test for PLRV, PVY and an internal control in one set and PVA, PVM and an internal control in another, are commercially available. They also contain a rapid extraction kit for virus RNA for each pooled sample (Bioreba AG, Basel, Switzerland). This commercial kit includes special extraction buffers and boiling of samples. It is used in some countries, e.g., Luxemburg (personal communication Serena Rauch, ASTA Luxembourg) for seed certification. But the latter RT-qPCR method has the disadvantage that it cannot test for all six viruses. The complete method is called RE-kit henceforth.

In Stammler (2020) and Prinz et al. (2022) we developed a method, called multiplex real-time DiRT-PCR (= direct reverse transcription PCR). This method can detect all six important viruses plus controls in two quadruplex RT-qPCRs without prior RNA extraction. This is possible by using the KAPA3G™ Plant PCR kit (former KAPA Biosystems, USA; now Roche, Switzerland, Schori et al., 2013; Stammler, 2020) and adding a robust reverse transcriptase, i.e. SIII (Merck - Sigma Aldrich, Germany) as described in Stammler et al., 2014, 2018, Stammler, 2020 or highScriber (HighQu, Germany) as described in Prinz et al., 2022. So far, we validated this new method comparing single tuber heels to ELISA on leaves coming from the same tuber. The agreement between methods ranged from the best rating “almost perfect” to the third best rating “fair” according to Gwet’s AC1 agreement index, depending on the type of virus, being the best for PLRV and the worst for PVM (Prinz et al., 2022). However, single tuber testing (heel end) is not practical for routine testing and comparison to ELISA on leaves not ideal as it is not the same part of the plant and virus might be enriched unproportionally in the leaf (Stammler et al., 2018; Stare et al., 2015) or unequally distributed within the tuber (Dupuis, 2017). Hence, in the following study, we decided to use the protocol of the multiplex real-time DiRT-PCR as in Prinz et al. (2022), improve the method-protocol further for pool testing, (in short DiRT) and compare it to the results of ELISA on leaves of the same tubers and the commercial RE-kit on the same heel ends.

Material and Methods

Plants and tubers

For the positive control, we used infected in vitro plants provided by Dr. Petyr Dedič (Potato Research Institute Havlíčkův Brod, Department of Virology, Czech Republic) to extract virus RNA (PureLink®, Invitrogen, USA).

For the validation itself, we processed 157 official sample lots taken for the Bavarian state seed certification process, each consisted of 120 field grown potato tubers. These 157 lots represent approximately 40 different varieties (see Supp1). Tubers were stored in cool temperatures until used. Sampling and processing took place from September 2019 until January 2020.

Sample preparation and assay conduction of PCR based methods

120 tubers of each official seed certification sample-lot were washed together in tap water using a potato washing machine (Flottenwerk H.-J. Dames K.G., Germany). In total we collected 12 times ten heel ends of tubers, with one cone measuring ~2 mm in diameter and ~ 2–3 mm in depth (less than 0.2 g for 10 heels, Fig. 1). Each 10 heel ends were collected in extraction bags with a membrane (Bioreba AG, Switzerland). In a primary homogenization step without buffer, we pressed the sample in one bag once for 3 s with a machine that used air pressure (developed together with the Institute for Agricultural Engineering of the Bavarian State Research Center for Agriculture, Germany). Alternatively, bags with heel ends were put in the freezer (−20 °C) overnight or stored there for up to one month. In a second step we homogenized the sample using Homex6 (Bioreba AG, Switzerland) together with 1 mL extraction buffer 1 (EB1 from RE- kit, Bioreba AG, Switzerland) in the same bag.

Fig. 1
figure 1

Schematic description of the sample preparation for the multiplex real-time DiRT-PCR (lab pictures by ©Tobias Hase, StMELF; others by M. Prinz, A. Kellermann and J. Schwarzfischer)

Full size image

A 100 μL of the heel-homogenate was transferred into a 96-well PCR plate (Brand, Germany) and treated according to the rapid extraction and RT-qPCR protocol for PLRV/PVY and PVM/PVA by Bioreba AG (RE-kit). 2 μL of the extract was used for RT-qPCR.

The remaining of the homogenate (approximately 300 μL) was transferred into a 96-deep-well plate (VWR, Germany) and centrifuged at 4000 x g for 30 min at 4 °C. The supernatant was further diluted 2 μL in 198 μL extraction buffer General (10 X diluted to 1 X, Bioreba AG, Switzerland) and 2 μL of this dilution were used for the DiRT.

The mastermix for the multiplex DiRT-PCRs were done using the same primers, probes, fluorophore-combinations as described in Prinz et al. (2022) using an optimized concentration protocol version with HighScriber reverse transcriptase and HighScriber Buffer (HighQu, Germany, for details, see Supp3). The DiRT-temperature-protocol and setting was the same as in Prinz et al. (2022).

Sample preparation and assay conduction for ELISA

After cutting the heel, rose-ends of the same tubers were cut with a tuber cone cutting machine (Institute for Agricultural Engineering of the Bavarian State Research Center for Agriculture, Germany) with clean, disinfected knifes. Cones were kept in the same order as for PCR pools to be used for virus detection via ELISA. The tuber dormancy was broken by dipping the rose-end cones into 0.0001% (w/v) gibberellic acid (GA3, Carl Roth, Germany) solution for 20 min. After air drying, the cones were cultivated in peat-sand substrate (Stender GmbH, Germany) in the greenhouse under artificial light (Philips IP55, Germany) with a photoperiod of 16 h light at 22 °C. After 4–6 weeks cultivation, the third fully developed leaf was sampled for virus detection using ELISA. Three drops of leaf sap from a sap press (Institute for Agricultural Engineering of the Bavarian State Research Center for Agriculture, Germany) were diluted in 1 mL Extraction buffer General supplemented with 0.1% (w/v) egg albumin (Bioreba AG, Switzerland).

For each virus tested (PLRV, PVY, PVM, PVS), 150 μL of the centrifuged leaf sap was added to antibody coated microtiter plates (GreinerBio, Germany). Every step was done according to the manufacturers protocol (antibodies and coating protocol: Bioreba AG, Switzerland).

Extraction buffer without any sample material was added to each ELISA as no template control (NTC). As positive control, we used sap of greenhouse-plants previously tested virus positive.

DiRT analysis

For analyzing the DiRT, BioRad CFX™ Manager 3.1 was used with the same settings as in Prinz et al. (2022): Baseline subtracted curve fit, fluorescence drift correction and Ct determination mode “single threshold”. Due to the high fluctuation in background fluorescence, the threshold line was set manually above background fluorescence. Samples with a negative internal control in the DiRT or invalid curves, were not included into the respective analysis. As a cut-off-Ct value, we used the values empirically determined in our analysis in Prinz et al. (2022), which were Ct 36 for PLRV, PVA and PVX, Ct 40 for PVY and Ct 32 for PVM and PVS. The cut-off for the relative fluorescence unit (RFU) value was 100 RFU. As negative controls, we used the buffer solutions EB1 and EB2 (both from the Bioreba AG rapid extraction kit), Extraction buffer General (10 X diluted to 1 X, Bioreba AG, Switzerland), virus free potato sap and mastermix without additions.

RE- kit analysis

To obtain results for the RE- kit (Bioreba AG), a cut-off Ct of 37 was determined by a dilution series with positive controls for PLRV, PVY, PVM and PVA (extracted virus RNA). Curves above an RFU (relative fluorescence unit) of 300 were positive. Samples with negative internal controls were excluded from further analysis.

ELISA analysis

Leaf material was tested in one reaction per plant and was evaluated optically. This was according to the standard procedure of the growing-on ELISA for seed potato certification of the Working Group of the Certification Agencies for Agricultural and Horticultural Seed and Plant Material in Germany. Plants that did not grow out were not included into the analysis.

Method comparison

To compare results for two out of three methods at a time, respectively, we evaluated the number of positive and negative pools that were the same or different in both methods. ELISA “pools” of the ten matching single samples in which one or more plants did not grow, were not included into pool comparison analysis with ELISA but could be included in the DiRT versus RE-kit. The same was true for PCR results with invalid controls.

To determine the percentage of virus for the whole seed lot we used the statistical tool SeedCalc8, programmed by the international seed testing association (ISTA, https://www.seedtest.org/en/services-header/tools/statistics-committee/statistical-tools-seed-testing.html), using the number of virus positive pools, number of tubers per pool and total number of tubers to calculate the percentage of virus in the seed lot in case of the PCR-based methods. To derive the percentage of virus per seed lot for ELISA, we took the number of virus plants divided by the number of grown plants times hundred.

We also used each seed lot as one sample with the percentage of virus infestation for comparison of certified versus not certified samples. Therefore, we used the certification categories present in the German certification system, which uses the categories prebasic with max. 0,5% basic class S (BS) with up to 1%, basic class SE and basic class E (BSE) with up to 2%, certified class A (ZA) with up to 8% and certified class B (ZB) with up to 10% virus. When the seed lot reaches percentages below or equal to the respective category, it is accepted as certified and if it exceeds the percentage it is called not certified in the results part.

According to the numbers of the comparison analysis, we calculated Gwet’s AC1 coefficient as a measure of inter-method agreement for experiments with unequal numbers of positive and negative samples (Gwet, 2008; Wongpakaran et al., 2013). In addition, we used the rating categories developed by Landis and Koch (1977).

Results

About 1875 pools of 10 tuber heels, derived from 157 official samples of different varieties for seed certification, were tested with three methods, the newly developed DiRT, the commercial RE-kit (Bioreba) and ELISA. Two reactions were needed for each pool to obtain results for PLRV/PVY and PVM/PVA with the RE-kit, three reactions for each pool PLRV/PVY/PVM and PVS/PVA/PVX for DiRT and four reactions per single plant for PLRV, PVY, PVM and PVS for ELISA. Positive and negative results, including Ct-values for each pool and varieties and certification grade for each sample, are shown in Supp1 for each method and virus type. We then compared positive/negative results of single pools or overall percentages of virus of the whole sample from two of those methods at a time through Gwet’s AC1. Results for PVA and PVX were always negative in DiRT, except positive controls. Results for PVA in the RE-kit were negative for all pools. Comparison of DiRT and ELISA results in the previous paper were both negative for PVA and PVX (Prinz et al., 2022), performing over 3000 tests. Thus, no ELISA and no further analysis were done for these viruses.

Comparison of the DiRT with a commercial RE-kit

The percentage of pools that were rated virus positive or negative with both methods was over 94% and the agreement index Gwet’s AC1 was over 0.9 for PLRV, PVY and PVM. That means, the agreement of DiRT with a commercial RE-kit was “almost perfect”, the highest rating category (for more detail see Fig. 2).

Fig. 2
figure 2

Comparison of the same single pools of 10 tuber heels rated virus positive or negative using either the commercial rapid extraction RT-qPCR-kit (RE-kit, Bioreba AG) or the multiplex-real-time DiRT-PCR (DiRT). The percentage of pools that were rated positive with both methods are depicted in green, those rated negative in both methods in red, those solely rated positive testing with RE-kit in yellow and those solely rated positive testing with DiRT in blue. The percentages are shown for each virus (PLRV, PVY, PVM on the x-axis) separately. The light green boxes on the left of each column illustrates how many pools were rated the same in both methods confirming the agreement index Gwet’s AC1 in the table below for each virus separately. An agreement of 1–0.8 is rated as “almost perfect”. Additionally, the table below shows the exact percentages of the respective category and the total number of pools that were analyzed (n for PVY is lower due to two invalid curves)

Full size image

Furthermore, we took a closer look at the Ct-values in our analysis (Fig. 3). For PLRV, the cut-off Ct’s were 37 for RE-kit and 36 for DiRT. Looking at boxplots of the Ct’s of those pools where both methods detected PLRV (agreement red boxplots, Fig. 3), medians were at Ct 21 for RE-kit and four cycles later at Ct 25 for DiRT. Ct values in boxes did not overlap. However, Cts of pools that were positive with solely one method are higher in general with median Cts at 34 for RE-kit and 33 for DiRT and boxes overlap (Fig. 3, no agreement red boxplots). Further, we saw similar results for PVY, except that cut-offs were at Ct 37 for RE-kit and Ct 40 for DiRT. Medians were lower for the RE-kit at Ct 23 compared to Ct 28 for DiRT with non-overlapping boxes (Fig. 3, agreement blue boxplots). Also, median Cts of positive results with solely one method were higher, with 34 for RE-kit and 35 for DiRT, than the Cts of method agreement. And boxes did not overlap (Fig. 3, no agreement blue boxplots). For PVM positive agreement values, with cut-offs at Ct 37 for RE-kit and Ct 32 for DiRT, the median was higher for DiRT at Ct 26 compared to Ct 23 for RE-kit. But boxes slightly overlapped this time (Fig. 3, agreement green boxplots). The boxplot figure for PVM also differs from that of PLRV and PVY in the no agreement part. The median Ct for RE-kit with 35 is higher than the median of the agreement values of RE-kit and higher than the no agreement values of DiRT with 25. Additionally, Ct-values of those tested positive with solely one method did not overlap (Fig. 3, no agreement green boxplots). Moreover, boxplots DiRT did overlap in the agreement versus the no agreement part for PVM. Looking up the ELISA PVM results for the no agreement DiRT positive Cts, we found positive results for all available data points. Whereas for the no agreement RE-kit positive Cts, we found negative ELISA PVM results for all available data points.

Fig. 3
figure 3

Boxplots of virus positive Ct-values (y-axis) of RE-kit and DiRT pools grouped by agreement (left of each graph) and no agreement (right of each graph) between RE-kit and DiRT method, and virus type (PLRV: upper red graph; PVY: middle blue graph; PVM: lower green graph). The number of Ct-values is depicted below each pair of boxplots

Full size image

Comparison of DiRT with ELISA

We performed the same comparison analysis for DiRT and ELISA for PLRV, PVY, PVM and PVS (Fig. 4). We rated the “pool” for ELISA positive, when one or more plants of ten were virus positive. 674 pools, in which one or more eye cuttings did not grow were excluded, thus we analyzed in total 1201 pools. In general, more than 92% of pools had the same virus status in both methods and the overall agreement was in the rating category “almost perfect” with a Gwet’s AC1 between 0.98 and 0.87. The agreement was greatest for PLRV and PVM and lowest for PVY and PVS (for more details see Fig. 4).

Fig. 4
figure 4

Comparison of the same single pools of 10 tuber heels rated virus positive or negative using either growing on DAS-ELISA (ELISA) or the multiplex-real-time DiRT-PCR (DiRT). The percentage of pools that were rated positive with both methods are depicted in green, those rated negative in both methods in red, those solely rated positive testing with ELISA in yellow and those solely rated positive testing with RE-kit in blue. The percentages are shown for each virus (PLRV, PVY, PVM and PVS on the x-axis) separately. The light green boxes on the left of each column illustrates how many pools were rated the same in both methods confirming the agreement index Gwet’s AC1 in the table below for each virus separately. An agreement of 1–0.8 is rated as “almost perfect”. Additionally, the table below shows the exact percentages of the respective category and the total number of pools that were analyzed

Full size image

The more plants in a pool were tested positive with ELISA the better the agreement between the two methods and vice versa (see Supp1).

Comparison of the RE- kit with ELISA

This was the same analysis as above but for RE-kit and ELISA. Due to the lesser virus types tested in the RE-kit, we compared PLRV, PVY and PVM results for the 1201 pools. For PLRV und PVM, 96% of pools had the same results in both methods an agreement of 0.96 which is “almost perfect”, the highest rating category. For PVY, 89% of the results were the same in both methods agreement of 0.8, which is the second rating category “good” (for more details see Fig. 5). The more plants in a pool were positively tested with ELISA the better the agreement between the two methods (see Supp1).

Fig. 5
figure 5

Comparison of the same single pools of 10 tuber heels rated virus positive or negative using either growing on DAS-ELISA (ELISA) or the commercial rapid extraction RT-qPCR-kit (RE-kit, Bioreba AG). The percentage of pools that were rated positive with both methods are depicted in green, those rated negative in both methods in red, those solely rated positive testing with ELISA in yellow and those solely rated positive testing with RE-kit in blue. The percentages are shown for each virus (PLRV, PVY, PVM on the x-axis) separately. The light green boxes on the left of each column illustrates how many pools were rated the same in both methods confirming the agreement index Gwet’s AC1 in the table below for each virus separately. An agreement of 1–0.8 is rated as “almost perfect”. An agreement of 0.8–0.6 is rated as “good”. Additionally, the table below shows the exact percentages of the respective category and the total number of pools that were analyzed

Full size image

Comparison analysis grouped by starch content of varieties

Additionally, we asked if the results of DiRT-PCR are negatively influenced by the starch content of potatoes. That is, if the percentage of starch potato pools (starch potato varieties see Supp1) in virus positive and virus negative tested pools with DiRT differs from the overall percentage of tested starch potato pools. Again, we compared DiRT to each of the standard methods RE-kit and ELISA for each virus, respectively (Fig. 6). In addition, the standard methods were compared to each other. The percentage of starch potato pools over all tested pools (= 100%) varied between 27 and 29% depending on pools that were excluded or included in the analysis for each method (see material and methods). For most categories, the percentage of starch potato pools grouped around these values (Fig. 6) with a few exceptions: For PLRV, the percentage starch potato pools in the category solely DiRT positive increases to 80% compared to RE-kit and 44% compared to ELISA. However, the number of tested pools in that category is 7 and 16, respectively (see table, Fig. 6). For PVY, starch potato pool percentages drop down to 4 to 10% in the category virus positive with both methods. For PVM, starch potato pool percentages increase to 40 to 86% in the categories only one method positive and RE-kit and DiRT positive. PVS was solely detected with DiRT and ELISA. Here, starch potato pool percentages for ELISA and DiRT negative had the lowest and ELISA and DiRT positive the highest starch potato pool content. Overall, starch content has no effect on the method used for virus testing in potatoes. However, all three methods showed that a lower percentage of starch potato pools were PVY positive compared to all other varieties.

Fig. 6
figure 6

Percentage of starch potato pools (y-axis) grouped by agreement category: RE-kit compared to DiRT (upper left graph); RE-kit compared to ELISA (upper right graph); DiRT compared to ELISA (lower left graph). Bars are color coded for PLRV in red, PVY in blue, PVM in green and PVS in yellow, respectively. For additional information, the lower right shows a table with associated numbers of total pools in each category

Full size image

Comparison DiRT, RE-kit and ELISA on the level of state recognition

The last step was to compare (pairwise) how many samples would be differently certified or not certified by the state (see material and methods), depending on the test results of the different methods (Fig. 7, Supp2). We compared those samples in which all 120 grown out eye-cuttings were available for this analysis. Therefore, the number of compared samples for ELISA sank to 40 samples. The agreement was 92.5% for DiRT versus ELISA, 89.8% for RE-kit versus DiRT and 85% for RE-kit versus ELISA. The agreement indices in DiRT versus ELISA and DiRT versus RE-kit were „almost perfect”, 0.87 and 0.8. For RE-kit versus ELISA, the agreement index was 0.72, rating as “good “(Fig. 7).

Fig. 7
figure 7

Comparison of the samples total percentage of virus infestation that results in a certified or not certified sample, in the respective seed category, depending on the method tested by, i.e., multiplex real-time DiRT (DiRT), growing on DAS-ELISA (ELISA) or rapid extraction RT-qPCR-kit (RE-kit, labled on the x-axis). Included are only ELISA results where 120 plants grew. The percentage of samples were certified with both methods are depicted in green, those not certified in both methods in red. For DiRT versus ELISA those solely certified testing with ELISA in blue and those solely certified testing with DiRT in yellow. For RE-kit versus ELISA, those solely certified by RE-kit in yellow and those solely certified with ELISA in blue. For DiRT versus RE-kit, those solely certified with DiRT in yellow and those solely certified with RE-kit in blue. The light green boxes on the left of each column illustrate how many pools were rated the same in both methods confirming the agreement index Gwet’s AC1 in the table below for each virus separately. An agreement of 1–0.8 is rated as “almost perferct”, 0.8–0.6 as “good”. Additionally, the table below shows the exact percentages of the respective category and the total number of pools that were analyzed

Full size image

Discussion

The new DiRT method involves direct testing on tuber sap for three viruses in one reaction. DiRT consequently reduces either glasshouse capacities needed for ELISA or extraction time and associated expensive chemicals needed for conventional RT-qPCR methods. Here we validated the multiplex real-time DiRT-PCR, introduced in Prinz et al. (2022), for pool testing in comparison to ELISA on leaves and a commercial RT-qPCR RE-kit.

In comparison to the agreement of single tuber testing with DiRT and ELISA (Prinz et al., 2022; Stammler et al., 2018), the agreement in the pool testing stayed in the same rating category or has even improved. For example, for PLRV it stayed in the “almost perfect” agreement, for PVY it changed from “good” to “almost perfect” and for PVM in multiplex from “fair” to “almost perfect”. This improvement might be also due to the optimization in the size of heel cuttings, amount and type of buffer which were 1 ml buffer general on 0.2–0.5 g of a single tuber heel and 1 ml EB1 on 0.2 g of 10 tuber heels. In addition, we applied a better homogenization through developing our own machine or freezing and homogenizing compared to a single homogenizing step on raw tuber heels in single tuber testing.

Looking at all tested pools (1201–1875), the agreement of DiRT with ELISA or RE-kit was in the highest rating category “almost perfect” and the overlap of results over 90% for all tested viruses, PLRV, PVY, PVM and PVS. In addition, RE-kit and ELISA had an “almost perfect” agreement for PLRV and PVM, except for PVY, which was in the second agreement category “good”. As the RE-kit is already used for routine testing in Luxemburg (personal communication with Serena Rauch, EAPR conference 2017) and gave similar agreement results with ELISA, we suggest DiRT can be used for routine testing as well.

The comparison of ELISA on leaves with DiRT and RE-kit on tubers on certification level produces a lower agreement index than the virus specific comparison on the pool level. This could be due to a relatively low number of 40 or 157 seed lots to 1201 or 1875 pools compared statistically and more importantly, the combination of results for PLRV, PVY and PVM into one result. In addition, not all 120 plants grew out for most ELISA seed-lots and could be statistically compared to DiRT and RE-kit. There could be many reasons why not all 120 plants grew out, e.g. too much virus, too many bacteria or fungus reduced the growth or stopped it completely. It was not possible for us to deduce if this in turn might confound the certification results for seed-lots tested with ELISA. Hence, testing on the level of tuber is one advantage for using DiRT and other PCR-based methods for routine testing in the future.

Although we had to use the extraction buffer of the RE-kit due to the experimental design, DiRT could be further validated using other buffers to reduce costs in the future. This could for example be the general buffer (Bioreba AG, Switzerland) that gave reliable results for DiRT in Prinz et al., 2022 or the phosphate buffer from Schumpp et al., 2021. In the future, the number of heel ends per pool in DiRT could also be increased for decreasing costs further. Nevertheless, the current number of pools is ideal for seed lots coming from regions with higher virus content because it leads to a better percentual resolution. This in turn helps the decision to certify or not certify the seed lot in the respective seed category (ranges of percentages can be calculated and downloaded here: https://www.seedtest.org/en/services-header/tools/statistics-committee/statistical-tools-seed-testing.html). For example, seed potato categories that differentiate seeds with 8 versus 10% virus need the respective number of pools for appropriate statistical results in the SeedCalc table, i.e. 12 pools of 10 heels.

We find a higher disagreement between DiRT and RE-kit at higher PCR-cycle numbers for PLRV and PVY that might correspond to a low number of infected plants per pool. This reflects the fact that the probability of detection decreases with the number of DNA/RNA copies in the sample (Uhlig et al., 2015). It should also be noted that SeedCalc statistically accounts for the increasing disagreement between results of seed pools that contain only one or two virus positive plants per pool. Additionally, it is known that the false positive/false negative rates (FPR and FNRs) rise with each cycle, especially close to the cut-off. Nowadays, FPRs and FNRs of qPCR results are extensively discussed for detecting SARS COVID 19 virus in humans (Braunstein et al., 2021). There might appear nonspecific reactions towards the higher cycle numbers which cannot be exactly determined for the commercial RE-kit nor the DiRT nor any other qPCR. Hence, due to the FPR/FNR, it is impossible to say, which method is more sensitive in detecting which virus. Nevertheless, on one hand, having done dilution series and looking at the very good agreement between methods, we think the cut-off we chose was appropriate for PLRV and PVY. On the other hand, we also found disagreement between DiRT and RE-kit at lower qPCR-cycle numbers for DiRT PVM detection. As all datapoints are also virus positive when tested with ELISA, we can conclude that PVM detection with DiRT is more sensitive than PVM detection with RE-kit. Additionally, the cut-off for PVM detection with RE-kit might have been too high because, all data points that did not agree were negative with ELISA. Another reason could be that ELISA was not sensitive enough in these cases.

Generally raised doubts that starch content of the potato variety might influence the PCR reaction in the DiRT method can be refuted. We sorted the results for potato varieties with high starch content versus low starch content. Comparing all three methods, we found no difference in the percentage of starch varieties in virus positive versus virus negative categories for PLRV. In fact Demeke and Jenkins (2010) already mention that neutral polysaccharides such as starch have no inhibitory effect on the PCR reaction.

Nevertheless, we did find a difference in PVY positive versus negative pools for starch potato content comparing all three methods. But that might be because most tested starch potato varieties are resistant to PVY and thus we see that a lower percentage of starch potatoes are PVY positive.

Curiously, the percentage of starch potatoes rises in the PVM positive category, especially when we compare the two PCR-based methods. Possible reasons for this could be: (1) Potato breeders let these varieties grow on the field until autumn where temperatures decrease (personal communication between A. Kellermann and breeders). This gives PVM a higher chance to infect the plant because PVM infection is temperature sensitive, that is, more infections occur at lower temperatures around 20 °C (Tatarowska et al., 2020) (2) Fertilization or metabolism might be different in these varieties changing, for example, cell wall structure which in turn facilitates PVM infection (Kozieł et al., 2021).

Finally, laboratories all over the world have compared several PCR-related methods with ELISA and found similar or slightly worse overlaps of results. Their comparison results range from 85 to 99% overlap according to a survey done between 2017 and 2019 (UNECE secretariat, 2017–2019) compared to 92 to 98% overlap in this study. The most recent publication about pool testing of conventional multiplex RT-qPCR in comparison to ELISA showed similar agreements for both methods, also on seed certification level, and is now introduced into routine testing in Switzerland (Schumpp et al., 2021). Additionally, immunological and qPCR tests are frequently compared for human virus diseases and results are rated suitable for routine testing, when there is an overlap of over 90% in both methods (Sanou et al., 2020; Schnurra et al., 2020). DiRT also reaches over 90% overlap with ELISA, even on the certification level, when comparing only the fully grown out lots. Thus, DiRT can be reliably used for routine testing for fast and cost-efficient results in seed certification.

Conclusion

To conclude, our newly developed multiplex real-time DiRT-PCR on tuber pools shows an “almost perfect” agreement with two standard methods used for virus certification. Thus, it is now ready for fast and low-cost routine testing as compared to either ELISA or conventional and rapid RT-qPCR protocols.