General notes on data analysis

All analyses were conducted in R (version 4.4.1) using Rstudio (version 1.4.1717; [24]; The R [25]). In R, the tidyverse packages, especially ggplot2 and dplyr were used for data processing in general and generation of figures [26]. Other, specific packages used are mentioned at the relevant sections. All R scripts and underlying data are available through git (https://git.wur.nl/published_papers/schaveling_2026_pallida_asqpcr). Sequencing data of the 9 selected and unselected G. pallida AMPOP populations was deposited at the European Nucleotide Archive (E-MTAB-16285).

Selection of G. pallida populations on potato variety Seresta and re-sequencing

We generated new sequencing data for nine G. pallida populations that were obtained from Dutch potato fields from 2011 until 2015, including the previously described AMPOP10 and eight other populations: AMPOP03, -06, -08, -09, -10, -13, -15, -16, and -19. After initial propagation on Desiree, these populations were selected on the GpaV_vrn-containing potato variety Seresta, as previously described [4]. In short, populations were propagated either for four generations on the resistant potato variety Seresta (S4), or, as a control, for one generation on the susceptible variety Desiree (D; S1 Table). Populations resulting from these selection steps were taken as input for DNA isolation and sequencing.

First, DNA extraction and sequencing was performed as previously described [4]. In short, for each S4 and D population DNA of approximately 40 cysts was extracted with phenol-chloroform extraction. DNA was sequenced by BGI Genomics at approximately 200x coverage using DNBSeq with 150 bp paired-end reads [4]. For sample AMPOP06D, 27.6 million duplicated reads were removed from FASTQ files with SeqKit (v2.10.0; [27]), after which all data was uploaded to ENA (E-MTAB-16285). This data was supplemented with two pairs of read data for AMPOP02D versus AMPOP02S4 and one pair of AMPOP10D versus AMPOP10S4. These datasets were included in the analysis and previously published by Schaveling et al. [4]. The four independently generated sequencing libraries of AMPOP02 and AMPOP10 were used for method validation.

Bulk segregant analysis

For analysis of genetic variation the reads were mapped to the Rookmaker genome (PRJEB91928; [4]). Variant calling was performed using Bcftools (mpileup & call, v1.14; [28]) as previously described ([4]; https://git.wur.nl/stefan.vanderuitenbeek/dnaseq_variant_calling_snakemake_pipeline). Before bulk segregant analysis, the resulting variant calls were filtered by removing sites with QUAL < 50, variants with a mean read depth outside two standard deviations of the sample average or a minimum read depth of 30, insertions, deletions, and sites with low allelic variation (mean alternative allele frequency ≤ 0.05 or ≥ 0.95). Bulk segregant analysis was performed by conducting a chi-squared test on the read depths for the reference and alternative alleles in the selected and non-selected populations. This is similar to the G-statistic [29], however as we sequence deep and aim for a high minimum coverage, we calculate the statistic depending on the actual read depths. The resulting p-values were FDR-corrected using the p.adjust function, as we are including variants with linkage disequilibrium.

To identify loci under selection, we counted the number of significant variants in consecutive windows of 10 kb across each scaffold. Identification of bins with a larger number of associated variants was done based on the significance of the numbers based on an exponential distribution. To fit the distribution well, the number of counts per bin was square-root transformed. Bins exceeding the significance threshold (p-value < 0.05) were defined as putative virulence loci.

Sequencing data of Dutch, British and French populations

Previously, we identified four non-synonymous SNPs inside the coding sequence of a gene involved in virulence on GpaV_vrn (S2 Table; [4]). To assess how well the AAF of these four SNPs correlate with selection on GpaV_vrn, we used four different DNA sequencing data sets. First, we used the sequencing data of a previous selection experiment with two G. pallida populations (AMPOP02 and AMPOP10) selected on the GpaV_vrn-containing Seresta (E-MTAB-15408; [4]). Second, we used the sequencing data of the nine GpaV_vrn-selected G. pallida populations described above (E-MTAB-16285). Third, we obtained DNA sequencing data of two French GpaV_vrn-selected populations (Saint-Malo and Noirmoutier) from the ENA (PRJEB90550; [30]). Both populations were selected on the GpaV_vrn-containing potato variety Iledher, the GpaV_spl-containing genotype 96D31.51, and the susceptible variety Desiree, in duplo for ten consecutive generations. Fourth, we obtained DNA sequencing data of a British G. pallida population Newton from the ENA (PRJEB41175; [31]). The Newton population was selected for 12 generations on the GpaV_vrn-containing genotype Sv_11305 (Morag) and the H3-containing CPC2802 (Sa_11415) genotype. Mapping and variant calling on the G. pallida Rookmaker genome was performed as described above.

Correlating SNP data with phenotypic data

Two standard PCN resistance tests were conducted in accordance with the [32]. In short, two litre pots with a potato plant were inoculated with approximately 10,000 living larvae. For each population the exact number was determined. At the end of the experiment (after 3 months the plants were watered for the last time and thereafter dried for ~ 1 month), cyst numbers were counted per pot. Per pot, the average number of larvae per cyst was determined and multiplied by the total number of cysts in the pot to get the final number of larvae per pot. The reproduction rate was calculated by dividing the final number of larvae (P_f) by the number of inoculated larvae (P_i):

A total of 8 G. pallida populations were tested across two tests (S1 Table). Both tests included AMPOP02, -03, -06, -09, -10 and -16. In addition, test 1 also included AMPOP13 and -15. Test 1 included fourteen potato varieties, that were grouped into clusters based on their resistance level (as determined previously by [4]). All varieties carry GpaV_vrn. Six varieties were included from the Seresta cluster (Cl_SER; Axion, Ardeche, Arsenal, VD 07–0289, Innovator, and Seresta) and eight of the Festien cluster (Cl_FES; Avarna, Avito, Altus, Supporter, Basin Russet, Libero, HZD 06–1249, and Festien). Test 2 included eight varieties from Cl_SER (Actaro, Aveka, Novano, Avatar, Simphony, Stratos, Vermont, and Seresta) and six from Cl_FES (BMC, Merenco, Saprodi, Sarion, Sereno, and Festien). Each tests contained at least three biological replicates per population (S3 Table). The average reproduction on Seresta of both tests was used as an indication of the virulence level on GpaV_vrn.

An AS-qPCR for the quantification of the alternative allele frequency of T173N

To quantify the AAF of SNP T173N (AAF_T173N) in G. pallida field samples, we decided to develop an allele-specific (AS) qPCR assay. Therefore, we first aimed at specifically amplifying the reference and the alternative alleles. We developed two sets of primers: one standard set and one with a locked nucleic acid (LNA) at the 3’ end. Both allele-specific primers had the SNP at the 3’ end (S4 Table). We calculated the qPCR-based AAFs based on the amplification of the alternative allele relative to the total amplification [33] according to the formula:

with being the Cq-value of the primer pair amplifying the reference allele and being the Cq-value of the primer pair amplifying the alternative allele, reducing the need for housekeeping genes.

To obtain DNA for the AS-qPCR, 60–70 cysts were crushed with bead-beating in 150 μL water for 80 seconds at 30 Hz. Lysis was performed with an in-house lysis buffer [34]. DNA was extracted using the Maxwell RSC PureFood GMO and Authentication Kit in combination with the Maxwell RSC instrument (Promega, USA) following the manufacturer’s instructions. This yielded between 4.1 and 11.3 ng/ μL of DNA per sample. AS-qPCRs were performed using iQ Supermix (Bio-Rad, USA; S5 Table) on a CFX96 C100 Touch thermal cycler (Bio-Rad, USA; S6 Table).

In a first test, these primers were tested on seven G. pallida populations, including three in house reference populations (D383, Rookmaker and AMPOP10) and four populations that were recently isolated from a field with suspected breakdown of GpaV_vrn resistance. As the standard allele-specific primers gave a better indication of the AAF than the LNA primers, we continued our analysis with the standard primers. In a second test, the standard primers were tested on an additional six recently isolated G. pallida field populations.

Blinded test of the AS-qPCR across four laboratories

To validate the AS-qPCR, we designed a blinded test. This test was conducted across four labs: Dutch General Inspection Service for Agricultural Seeds and Seed Potatoes (NAK), Hilbrands Laboratorium (HLB), Nederlandse Voedsel- en Warenautoriteit (NVWA), Wageningen University and Research (WUR). Each lab tested the same blinded set of 12 G. pallida populations (S1 Table). This included 10 GpaV_vrn-unselected and selected populations and two reference populations (D383 and Rookmaker). Each of the labs used its own DNA extraction protocol and qPCR machines. DNA extraction was performed on 20–40 cysts.

The WUR performed the AS-qPCR as described above, but with DNA isolation on 20 cysts, yielding between 101.7 and 281.8 ng/μL of DNA per sample. The NAK crushed the cysts with bead-beating in 100 μL water for 5 minutes at 30 Hz. Lysis was done by adding 300 μL Tissue & Cell Lysis Solution (LGC BioSearch Technologies, USA) and 1 μL proteinase K for 15 minutes at 65 °C shaking at 750 rpm. After 10 minutes on ice, 150 μL pre-cooled MPC Protein Precipitation Solution (LGC BioSearch Technologies, USA) was added and centrifuged at 16,000 rcf for 10 minutes. DNA and RNA was isolated with magnetic beads in a KingFisher Flex system (ThermoFisher Scientific, USA), yielding between 5.7 and 38.7 ng/μL of DNA per sample. AS-qPCRs were performed with PerfeCTa SYBR Green SuperMix (QuantaBio, USA) on an Applied Biosystems 7500 Real-Time PCR system (ThermoFisher Scientific, USA). The HLB crushed the cysts, after which lysis and DNA purification was carried out using the Nexttec Tissue & Cell DNA Isolation Kit (Nexttec, Germany) according to the manufacturer’s instructions. This yielded concentrations ranging from 8.5 to 97.5 ng/μL. AS-qPCRs were performed with SYBR Green PCR Mix (Clear Detections, Netherlands) on a CFX96 thermal cycler (Bio-Rad, USA). The NVWA crushed the cysts with bead-beating in a mix of 50 μL water, 130 μL ATL buffer (Qiagen, Germany) and 20 μL proteinase K for 2 minutes at 30 Hz. Lysis was done at 56 °C for 60 minutes shaking at 800 rpm. DNA extraction was performed with the DNeasy Blood and Tissue Kit (Qiagen, Germany) according to the manufacturers protocol, yielding DNA concentration ranging between 2.4 and 13.8 ng/μL. AS-qPCRs were performed with GoTaq qPCR Master Mix (Promega, USA) on a CFX Opus Real-Time PCR machine (Bio-Rad, USA).

The methods on the assessment of the reliability of the AS-qPCR assay can be found in the S1 Text.

Using Dutch and German G. pallida populations to assess the predictive value of the AS-qPCR assay

To assess the predictive power of the AS-qPCR, six Dutch G. pallida field populations were tested on five potato varieties, including the susceptible variety Desiree and four GpaV_vrn-containing varieties (Seresta, Festien, Axion, and Supporter) in small container tests as previously described [4]. Cysts were counted at 8–12 weeks after inoculation. Relative susceptibilities were calculated for each resistant variety by dividing the number of cysts on a resistant variety by the number of cysts on a Desiree times 100%.

To assess the value of the AS-qPCR on German G. pallida populations, virulence levels of ten German G. pallida populations obtained from the Emsland region (Germany) were determined through bioassays routinely used [35]. In short, eye plugs of the susceptible potato variety Desiree or the GpaV_vrn-containing varieties Seresta and Axion were individually planted and inoculated with cysts. White females were counted at six to eight weeks after inoculation. The relative susceptibility (RS) was calculated by dividing the number of white females on resistant varieties by the number of white females on the susceptible control times 100%. Using this type of bioassay allows for fast evaluation of the virulence of a G. pallida population [36], however, assays that are more artificial can give aberrant estimates of virulence compared standardized pot tests [37]. However, previously we saw good correlations between closed container tests and pot tests for G. pallida virulence on GpaV_vrn [4]. All ten German G. pallida populations were tested by the AS-qPCR assay. From each population, DNA was extracted from 60 cysts using the MasterPure Complete DNA & RNA Purification Kit (Lucigen, USA) following the manufacturers protocol. The DNA was suspended in 10mM Tris-HCl buffer and DNA concentration was determined by a Qubit 4 fluorometer (Life Technologies, Singapore). AS-qPCRs were performed with SSoAdvanced SYBR Green Supermix using a CFX 96 Real-Time PCR machine (Bio-Rad, USA), with 14–23 ng template DNA per 20μl reaction in triplicate.

Bulk segregant analysis on eight G. pallida field populations independently verifies a virulence locus

Previously, we identified a locus that strongly associates with virulence in two G. pallida field populations [4]. This locus lies on scaffold 28 of the G. pallida Rookmaker genome, with the strongest association between 6,371,140 – 6,682,119 bases. For the identification we conducted whole genome sequencing on each generation of a five-generation selection experiment on two G. pallida populations (AMPOP02 and AMPOP10) with one to three replicates per population per generation. The locus was identified based on association of allele-frequency shifts over these five generations. Here we aimed to independently verify that this locus is involved in virulence on GpaV_vrn. To make that feasible, we confirmed in four independent pairs of selected AMPOP02 and AMPOP10 whether bulk-segregant analysis (BSA) on the sequencing pools with only an unselected and a four-generation selected population as input was sufficient to identify the locus in all four tests (S1 Fig). From this experiment, we conclude that BSA based on generations of selection is a feasible approach for locus verification.

To independently verify the selected locus, we propagated eight G. pallida field populations on the GpaV_vrn-containing variety Seresta for four generations and sequenced the genome pool of starting and final populations (Fig 1A). This included G. pallida populations AMPOP03, -06, -08, -09, -13, -15, -16, and -19. Next, we performed bulk segregant analyses (BSAs) on each of the eight G. pallida populations separately. We identified the previous associated locus in six of the eight populations (S2 Fig). To be able to draw an overall conclusion from these separate analyses, we pooled the individual outcomes (number of significantly associated variants per 10kb bin). Over all eight populations we identified a single locus on scaffold 28 from 6.21 – 7.80 Mb (p < 0.05; Fig 1B). The 95% strongest associated loci (>477 significant SNPs counted over the 8 selected populations per 10kb bin) were located between 6.46 – 6.69 Mb. Since the previously identified virulence locus on scaffold 28 ranged from 6.37-6.68 Mb [4], we conclude this analysis independently verifies the virulence locus.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Bulked segregant analysis confirms the previously identified virulence locus.

A A graphical representation of the eight G. pallida AMPOP populations and the selection experiment. Created in BioRender (BioRender.com/v1ycgcu). B The number of significant SNPs (FDR < 0.05) are plotted per 10kb bin over eight bulk segregant analyses. Bins that are significantly enriched (p < 0.05) are coloured in green. The virulence locus is indicated in red.

https://doi.org/10.1371/journal.ppat.1014201.g001

Three SNPs are consistently selected by GpaV_vrn across Dutch, British and French G. pallida populations

Following verification of the virulence locus, we aimed to investigate whether G. pallida virulence on GpaV_vrn has a common genetic basis. Previously, we identified a single gene on the virulence locus, Gp-pat-1, as a gene that contributes to the breakdown of GpaV_vrn [4]. Four non-synonymous SNPs (E98V, T173N, L562V, L562S) inside the coding sequence of Gp-pat-1 correlated with virulence on GpaV_vrn in two G. pallida populations (S2 Table; [4]). To identify allelic variants involved in or tightly linked to virulence on GpaV_vrn, we assessed the alternative allele frequencies (AAFs) of these four SNPs in unselected and GpaV_vrn-selected G. pallida populations. First, we assessed the AAFs in two previously selected G. pallida populations (AMPOP02 and 10; [4]). Over five generations of selection, the four SNPs showed a consistent average increase in AAF of 0.041 per generation (Fig 2A). Second, we compared the AAFs of these four SNPs in nine G. pallida populations (AMPOP03, -06, -08, -09, -10, -13, -15, -16, and -19). These populations were selected on the GpaV_vrn-containing potato variety Seresta for four generations and compared with the unselected populations. For T173N, L562V, and L562S we observed a consistent and significant increase in the AAFs (p-value ≤ 0.00022; Fig 2B). This indicates that we have now observed a similar increase in the AAF for three SNPs in a total of 10 genetically unique Dutch G. pallida populations.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. The alternative allele frequencies of four SNPs that are located within the coding sequence of Gp-pat-1.

A The alternative allele frequencies (AAFs) of two G. pallida populations selected for five generations on the GpaV_vrn-resistant Seresta potato variety. AAFs show a significant increase over time. Do note that the y-axis is not starting at zero. B Nine G. pallida field populations were selected for four consecutive generations. SNPs T173N, L562V, and L562S show a significant correlation with selection. Statistics are based on one-sided paired t-tests. C Two French G. pallida populations (Noirmoutier and Saint-Malo) and an English G. pallida population (Newton) show a similar increase in the AAF over time when selected on a GpaV_vrn-resistant potato variety for 10 and 12 generations, respectively. Black lines indicate linear regression models. Statistics in A and C are based on linear regression models.

https://doi.org/10.1371/journal.ppat.1014201.g002

Under similar experimental conditions, French and British G. pallida populations have shown the potential to break GpaV_vrn resistance as well [38,39]. Two French populations (Saint-Malo and Noirmoutier) were selected on the GpaV_vrn-containing potato variety Iledher for ten generations [30]. A British population (Newton) was selected on GpaV_vrn-containing genotype Sv_11305 (Morag) for twelve generations [31]. To determine the AAFs of the four SNPs in these unselected and GpaV_vrn-selected G. pallida populations, we mapped their sequencing data (PRJEB90550 and PRJEB41175) to the G. pallida Rookmaker genome. Like the Dutch populations, the French and the British populations showed a significant increase in the AAF for the same three SNPs (Fig 2C). In conclusion, three SNPs (T173N, L562V, and L562S) are consistently selected by GpaV_vrn across 13 Dutch, British and French G. pallida populations.

The alternative allele frequency of a single SNP correlates with the reproduction rate on GpaV_vrn

The consistent correlation of three SNPs with selection on GpaV_vrn indicated that these SNPs are either tightly linked to or causal for virulence. To test whether any of our four SNPs gives a robust prediction of virulence, we assessed the reproduction rate (Pf/Pi) of eight G. pallida field populations (AMPOP02, -03, -08, -09, -10, -13, -15, and -16) on 28 potato varieties across two standard PCN resistance tests [32]. These potato varieties were previously grouped into three clusters based on their level of G. pallida resistance. These clusters were named after the susceptible potato variety Desiree (Cl_DES), and the GpaV_vrn-containing potato varieties Seresta (Cl_SER) and Festien (Cl_FES; [4]). The PCN tests included two Cl_DES varieties, thirteen Cl_SER varieties and thirteen Cl_FES varieties.

Since virulence on GpaV_vrn is a recessive trait [40], only individuals homozygous for the alternative allele (aa) are assumed to be virulent, whereas genotypes Aa and AA will be avirulent. If we assume that the populations are in Hardy-Weinberg equilibrium, the proportion of aa genotypes is expected to equal the square of the AAF. Therefore, to assess a potential quadratic relationship, we plotted the AAF² against the average reproduction rates on Cl_DES, Cl_SER, and Cl_FES. No significant correlations were found between the average reproduction rate on Cl_DES varieties and the AAF² of the four SNPs (Fig 3). This indicates that the AAFs are not predictive for the reproduction rate on susceptible varieties, which is in line with our expectations. For Cl_SER, the AAF of T173N significantly correlated with the average reproduction rate on thirteen varieties (p = 0.004). Moreover, when analysed individually, of the six Cl_SER potato varieties that were tested by eight G. pallida populations, five showed a significant correlation between the AAF of T173N and the reproduction rate (p < 0.05; S3A Fig). For the other three SNPs, no significant correlations with reproduction on Cl_SER were found. For Cl_FES, the average reproduction rate also showed a significant correlation with the AAF of T173N (p = 0.041; Fig 3). However, when analysed individually, only in two of the eight Cl_FES varieties tested with eight G. pallida populations showed a significant correlation between the AAF of T173N and the reproduction rate (p < 0.05; S3B Fig). Together, this data does show that a population’s AAF for T173N gives an accurate prediction of its average virulence level on GpaV_vrn-containing Cl_SER and Cl_FES varieties, making a promising quantitative genetic marker for virulence.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. A population’s alternative allele frequency of T173N correlates well with its average reproduction rate on GpaV_vrn.

The allele frequencies of eight Dutch G. pallida field populations were plotted against their average reproduction rates (Pf/Pi) on two susceptible varieties from the Desiree cluster (Cl_DES), thirteen GpaV_vrn-containing potato varieties from the Seresta cluster (Cl_SER) and thirteen GpaV_vrn-containing potato varieties from the Festien cluster (CL_FES). Significant correlations between the AAF of T173N and the average reproduction on Cl_SER and Cl_FES indicate that a population’s allele frequency of T173N is a good prediction for its virulence level on GpaV_vrn. Statistics are based on linear regression models that are indicated by black lines.

https://doi.org/10.1371/journal.ppat.1014201.g003

Allele-specific qPCR confirms GpaV_vrn-mediated selection of T173N

We established that the alternative allele of T173N is consistently selected by GpaV_vrn and the allele frequency in a population correlates with a population’s virulence level on GpaV_vrn. Subsequenly, we aimed at developing a PCR-based assay to quantify the AAF of T173N in G. pallida field populations. We therefore designed allele-specific primers to specifically amplify the avirulent reference allele and the virulent alternative allele in a qPCR (S4a Fig and S4 Table). We set out to verify the reliability and applicability of this assay (S1 Text).

First, we tested two variants of the same allele-specific primer sequences: one with a 3’-terminal locked nucleic acid (LNA), and one standard primer. The AAFs as determined by both primer pairs (AAF_qPCR) significantly correlated with sequencing-based AAF estimates (AAF_seq; p ≤ 0.0014; S4b Fig). However, the LNA-modified primer overestimated the AAFs, particularly at lower AAFs. In contrast, the standard primers provided an accurate reflection of the sequencing-based AAFs across the tested range.

Second, we validated the standard AS-qPCR primers on thirteen G. pallida populations, including three previously described populations (D383, Rookmaker and AMPOP10) and ten field populations that we recently isolated from Dutch potato fields. Linear regression models confirmed a strong correlation between the AAF_qPCR and the AAF_seq (p = 1.5e-10; Fig 4A) and the regression coefficients (β₁ = 1.00 and β₀ = 0.031) indicated that the AS-qPCR assay reliably estimates AAFs across the full AAF range.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. An allele-specific qPCR identifies GpaV_vrn-mediated increases in the alternative allele frequency.

A The alternative allele frequency (AAF) of SNP T173N in 13 G. pallida populations as measured with allele-specific (AS) primers in a qPCR against the AAF according to DNA sequencing. B The validation of the AS-qPCR assay at four different labs. AAFs as determined by AS-qPCR by one lab are plotted against the AAFs as determined by qPCR by the other lab or against the AAF as determined by sequencing. Statistics in A and B are determined by linear regression models. Black dashed lines indicates y = x for visual support. C The AAFs of unselected and selected G. pallida populations as determined by qPCR in four different labs. Unselected populations were propagated on the susceptible variety Desiree and selected populations were propagated for four rounds on the GpaV_vrn-containing variety Seresta. Statistics are based on one-sided t-tests.

https://doi.org/10.1371/journal.ppat.1014201.g004

Third, we assessed the robustness of this test. Across four labs we tested the same set of 22 G. pallida populations, including the unselected and Seresta-selected AMPOP02 and AMPOP10 populations [4], our eight unselected and Seresta-selected AMPOP populations, and reference populations D383 and Rookmaker. The resulting AAF_qPCR values of each lab showed highly significant correlations with the sequencing data and with the AAF_qPCR values of the other three labs (Fig 4B). This indicated that the assay is robust and can accurately quantify the AAF. When we compared the AAF_qPCR of the unselected populations with the GpaV_vrn-selected populations, nine out of ten GpaV_vrn-selected populations showed a significant increase in their AAF of T173N as compared to the corresponding unselected population (p < 0.05; Fig 4C). Fourth, we tested the AS-qPCR on six G. pallida populations previously propagated on potato roots in small container tests. A significant correlation between a population’s average relative susceptibility on GpaV_vrn-containing potato varieties and its AAF² (R = 0.83, p = 0.021) confirmed that the AS-qPCR is predictive for virulence on GpaV_vrn (S5A Fig). Fifth, we tested the AS-qPCR on ten German G. pallida populations. These populations were propagated on the potato varieties Desiree, Seresta, and Axion. A significant correlation between a populations AAF² and the relative susceptibility on Seresta (R = 0.61, p = 0.031) indicates that the AAF of SNP T173N makes a quantitative genetic marker for virulence on GpaV_vrn in German G. pallida populations (S5B Fig). Together, this confirms that GpaV_vrn-mediated selection acts on T173N in each of the tested populations and the AAF of T173N represents a population’s stage of selection.

The allele frequency of a single SNP is a predictor for a population’s level of virulence on GpaV_vrn

Through bulked segregant analyses (BSA) on eight GpaV_vrn-selected and unselected G. pallida populations we identified a major virulence locus on scaffold 28 of the Rookmaker genome (Fig 1B). This confirms the findings of two previous studies in which G. pallida virulence on GpaV_vrn was associated with the same locus on scaffold 28 and the syntenic region on scaffold 2 of the D383 genome [4,30]. Collectively, three independent studies, each using a distinct statistical approach (BSA, pairwise F_ST, association analyses) on separate pools of GpaV_vrn-selected G. pallida populations, converged on the same virulence locus. In addition to the locus on scaffold 28, we found other loci to be under selection. As these scaffolds are less consistently selected, they may have population-specific contributions to virulence. Given that this is the third independent study to identify the same locus on scaffold 28, strongly supports its central role in virulence on GpaV_vrn.

We then assessed the alternative allele frequencies (AAFs) of four SNPs from Gp-pat-1, a gene residing in the associated locus and previously associated with virulence on GpaV_vrn in two G. pallida populations [4]. We found that three of the four SNPs were consistently selected by GpaV_vrn across eight other Dutch, one British, and two French G. pallida populations (Fig 2). This indicates that these SNPs in Gp-pat-1 are either causal for or tightly linked to virulence on GpaV_vrn. If these SNPs are linked to causal variants, the strength of the correlation between the AAF and virulence will depend on the degree of recombination between the SNP and the causal allelic variant. Across two PCN resistance tests with eight G. pallida field populations, we showed that the AAF of SNP T173N significantly correlates with the average reproduction rate on thirteen Seresta-like and thirteen Festien-like potato varieties (Fig 3). However, when assessing the correlations for single potato varieties, part of them is insignificant (S2 Fig). This likely reflects additional resistances in Cl_FES varieties and technical variation inherent to the PCN resistance assay [32]. This is supported by the observation that the only Cl_SER variety tested twice (Seresta), showed the strongest correlation (p = 0.00057). Given the technical variation associated with PCN resistance tests, the AAF of T173N may provide a better indication of a population’s average virulence level than individual PCN resistance tests. Therefore, the AAF of T173N represented a reliable indicator of G. pallida virulence levels in field populations.

To predict virulence in G. pallida field populations, we developed an allele-specific qPCR assay that accurately quantifies the AAF of the T173N SNP and distinguishes populations based on their stage of selection (Fig 4). The developed qPCR assay was based on standard primers, where a test with LNA primers, unexpectedly, gave less consistent results (S4 Fig). It is possible that the LNA primers can outperform the standard primers as much lower DNA concentrations than tested [41]. Namely, this tool was developed and validated with samples of 20 cysts per DNA isolation. This was done with the potential for integration in current practices in mind. The tool has therefore not been tested for sensitivity beyond these constraints (e.g., single cysts or single juveniles). Within the scopes of the tested use, the AAF of G. pallida field populations was predictive for their propagation on GpaV_vrn-containing potato varieties and the AS-qPCR was able to accurately determine the AAFs. Therefore, the AS-qPCR assay can serve as a useful tool to predict virulence on GpaV_vrn in field populations.

Translating allele frequencies to field-tailored advice for potato farmers

With an accurate molecular tool for determining the virulence level of a G. pallida population, the major challenge lies in the translation of AS-qPCR data to actionable field predictions. For simplicity, we assumed that our populations are in Hardy-Weinberg equilibrium. We observed significant linear correlations between the AAF² in field populations and their reproduction on GpaV_vrn (Figs 3, S6, and S5A) supporting a quadratic relation between a population’s AAF of T173N and its virulence levels on GpaV_vrn. Since the unselected populations were propagated on Desiree before genotyping the assumption that our populations are in Hardy-Weinberg equilibrium is valid for the unselected populations. However, our field populations have been under selection of GpaV_vrn, which has masculinizing activity [11], distorting the equilibrium. When assuming perfect masculinizing resistance, from the second generation onwards, the AAF is expected to show a linear relationship with the number of females and the reproduction rate from the second generation onwards [40]. However, in the field, perfect masculinizing resistance is not realistic. The ratios in which the allele combinations are present in a population depend on many factors, including the escape rate, the male fraction among virulent individuals and the total fraction of juveniles that is able to reach maturity [12,40,42]. The relationship between the AAF and the reproduction rate is therefore expected to be in between a linear and a quadratic relationship. Moreover, the effect of the AAF on the reproduction could depend on the infection density and the allele frequency. For example, at low AAFs, having the alternative allele may confer a strong selective advantage, while at higher AAFs intraspecies competition might play a more important role, reducing the benefit of having the virulence allele. Since the overall impact of an infestation depends on both the virulence level of the cysts and their density in the field, translating AS-qPCR-based AAF estimates into actionable field predictions requires integrating AS-qPCR data with cyst densities and phenotypic field data. Future work should therefore establish how infection densities and virulence levels translate into crop loss under field conditions.

Accurate AS-qPCR-based predictions of G. pallida virulence levels in the field can be of great help for potato farmers. For example, besides the mandatory PCN sampling [21], the majority of Dutch starch potato growers have their fields voluntarily sampled and tested for PCN [22]. Upon the detection of G. pallida, farmers can request variety selection assays to determine which variety to grow based on the G. pallida in their fields. However, these assays are costly and time-consuming. Given that soil samples with suspected PCN infestation are already (q)PCR-tested for species and viability determination [43], the infrastructure is there to incorporate a qPCR for virulence determination. In combination with existing qPCR tests, our AS-qPCR test may support service providers in offering farmers informed advice on the sustainable deployment of resistances in the field, tailored to field-specific conditions in a cost-effective manner.

The same virulence allele is selected by GpaV_vrn across Western Europe

Our analyses showed that the same allele is consistently selected across 13 Dutch, British and French G. pallida populations. Moreover, the allele frequency of this same allele correlated with reproduction on GpaV_vrn in Dutch and German G. pallida populations. This strongly supports our hypothesis that G. pallida virulence on GpaV_vrn has a common genetic basis across Western Europe. As the genetic population structure across Western Europe is homogenous, this can also be the case for the spread of the virulence allele. The spread of virulence alleles suggests that the emergence of GpaV_vrn resistance-breaking populations can also be expected in West-European countries where no resistance-breaking populations have yet been found. Moreover, since Europe has functioned as a secondary distribution centre for the invasion of PCN into all other continents [16], the spread of virulence alleles is probably not geographically restricted. This implies that the repeated deployment of GpaV_vrn consistently selects the same virulence locus in G. pallida, leading to the rise of similar resistance-breaking G. pallida populations irrespective of geography.