Increased virulence of Globodera pallida during repeated rearing on different resistant potato cultivars explained by a simple model

Selection for virulence of Globodera pallida on potato cultivars was studied for four generations under controlled conditions. The reproduction rate (Pf/Pi) of a mixed Pa2/3 population increased by a factor of 61 during rearing on the partially resistant potato cv. Darwina compared to rearing on the susceptible cv. Irene. This was a result of selection for virulence on cv. Darwina, and achieving the Hardy–Weinberg equilibrium on cv. Irene. Increased virulence also significantly raised the reproduction rate on several other Solanum genotypes. These changes could be explained reasonably well by the monogenic inheritance of a virulence factor breaking the Grp1 locus. The virulence changes were probably mainly evoked by this gene only, inherited from S. vernei 1-3 or S. vernei 24/20. The Grp1 locus has probably provided the differential S. vernei hybrid (VTn)2 62-33-3 with its resistance to the Pa2 group and not to the Pa3 group. Alternation of cultivars did not halt selection if the cultivars highly differentiated between the Pa2 and Pa3 populations. Only when alternation was with cultivars that harboured a different resistance gene against Pa3 was selection for virulence delayed. Differences in virulence levels (i.e. reproduction rates) within the nematode population determined the rate of selection, not the resistance level itself. Selection of a Pa3 population for three generations on cv. Karakter not only increased the reproduction rate on cv. Karakter itself by a factor 4.2, but also raised the reproduction on other potato genotypes. A simple monogenic model could explain these changes in virulence.


Introduction
Potato cyst nematodes, including Globodera pallida, are a serious threat for potato production, especially in short and/or narrow rotations. In naturally infected potato fields, eggs in the cysts of G. pallida hatch in the spring, and the emerging larvae feed on roots of potatoes. If the potato cultivar is resistant to the nematode, the larvae cannot develop a good feeding structure in the roots, required for maturing females, and so the larvae become males; this is known as epigenetic sex determination (Schouten, 1993). A susceptible potato root allows the formation of a feeding structure, leading to both female and male nematodes. After mating, the females develop into cysts, containing 200-400 eggs. In autumn/winter after the potato harvest, the cysts will remain in the soil, and the eggs become dormant. Every year in spring, without the host, about 30% of the eggs hatch spontaneously. When potatoes are grown again, the surviving eggs will hatch, leading to a new reproduction cycle (Seinhorst et al., 1995;Been & Schomaker, 2000;Eves-van den Akker et al., 2015).
Until now, two different pathotypes have been distinguished in the Netherlands, i.e. Pa2 and Pa3 (Hockland et al., 2012). More recently, Niere et al. (2014) reported the presence of highly virulent G. pallida populations in potato fields in Emsland, Germany, possibly caused by selection for virulence. Turner et al. (1983) were the first to report on selection for virulence in G. pallida, reared in pots over five generations on Solanum vernei hybrids (VT n ) 2 62-33-3 and MPI 65.346/19. Turner continued this experiment for up to 11 generations, and provided further proof for selection for virulence, leading to an increase in the relative susceptibility of (VT n ) 2 62-33-3 from 10% before selection to 89% after selection. These selected populations were genetically distinct from their unselected counterparts and exhibited similar levels of environmental fitness under field-type conditions (Turner, 1990). Turner & Fleming (2002) used these populations for additional experiments on effects of alternation of cultivars. Inoculation of these populations onto hosts from different sources of resistance (S. multidissectum, S. sanctae-rosae and S. andigena) resulted in significantly reduced reproduction. However, in most cases these populations selected on the S. vernei progenitor also resulted in selection for virulence on their subsequent alternate host. Whitehead (1991) also showed selection for virulence in G. pallida populations on the S. vernei-based resistant cultivars Glenna or Morag. Beniers et al. (1995) found that in field plots, 8 years of continuous cultivation with cv. Darwina increased the virulence from 8.5% to 30%. The virulence of the G. pallida populations from the Darwina plots to (VT n ) 2 62-33-3, one of the parents of cv. Darwina, had also increased significantly. Phillips & Blok (2008) studied selection for virulence and effectiveness of alternation of hosts. They detected significant selection, both on S. vernei-derived potato progenitors and on clones derived from S. tuberosum subsp. andigena CPC2802. However, selection on S. vernei did not lead to significantly increased virulence to the S. andigena-derived progenitors. It is probable that the S. vernei-derived potato progenitors contained other resistance genes than S. tuberosum subsp. andigena CPC2802, and therefore other virulence alleles emerged. In this context it is interesting that Rouppe van der Voort et al. (2000) mapped a major-effect QTL conferring Pa3 resistance from S. vernei on chromosome 5, whereas Moloney et al. (2010) mapped resistance of S. tuberosum subsp. andigena on chromosome 4. However, it is still speculative whether the resistance genes that exerted the selection in the experiments of Phillips & Blok (2008) are the same genes that were mapped by Rouppe van der Voort et al. (2000) and Moloney et al. (2010).
A previous paper (Schouten & Beniers, 1997) analysed the increase in the reproduction rate of G. pallida populations that were reared for four generations on partially resistant potato cultivars. Reproduction rate is measured as the number of newly formed eggs per inoculated egg on a test host, also denoted as P f /P i . Reproduction rate was split into aggressiveness (fitness) and virulence, where aggressiveness was defined as reproduction rate (P f /P i ) on the susceptible cultivar Irene, and virulence as reproduction rate on the partially resistant cultivar relative to reproduction rate on the susceptible cultivar, expressed as a percentage. Rearing G. pallida on the cultivar Elkana with a low level of resistance did not increase reproduction rate significantly, but rearing on the more resistant cultivars Darwina and Karakter did. This was caused by an increase in virulence, in turn caused by an enhanced ability of the eggs to develop into cysts, and not by an increase in the number of eggs in the new cysts. Reproduction rate on the susceptible host generally was not significantly affected (Schouten & Beniers, 1997). Fournet et al. (2013) also observed a very significant increase in virulence after selection on resistant hosts, some with progenitor AM 78-3778 as parent, carrying the GpaV vrn QTL from S. vernei on chromosome 5, synonymous with the Grp1 locus, rather than to Gpa5. The latter has been mapped in a diploid derived from AM 78-3704 (Rouppe van der Voort et al., 1998Voort et al., , 2000. Counterintuitively, in their case, selection for virulence also led to an increase in fitness on the susceptible host, as shown by increased cyst sizes and elevated numbers of larvae per cyst (Fournet et al., 2016).
In a previous paper (Schouten & Beniers, 1997), selection experiments were described on a Pa3 population and on a mixture of a Pa2 and Pa3 population, which were very distinct regarding virulences. The observed increased virulence for the mixed population could be predicted rather well by a simple numerical model, using assumptions that Jones used in his models for selection in G. rostochiensis (Jones et al., 1981;Jones, 1985). Validation of models for selection for virulence requires the variation in virulence within a population to be known. However, for natural populations this is generally not known. For that reason, an 80% Pa2:20% Pa3 mixed population was made (Schouten & Beniers, 1997).
The present study looks at the effect of alternation of potato cultivars on the rate of selection to increased reproduction rates of the mixed Pa2/3 population and the Pa3 population. This was studied in two ways. First, after rearing a G. pallida population for four generations on one partially resistant cultivar A, the reproduction rate on another potato cultivar B was determined, and compared to the reproduction rate of the unselected population on cultivar B. This is the relative reproduction rate on cultivar B after selection on cultivar A. Secondly, two or more cultivars were alternated when rearing a population for four generations. After this selection process, the selected population was inoculated on test hosts, and relative reproduction rates were determined. Working definitions are given in Table 1.
By measuring the reproduction rates of the Pa2, Pa3 and mixed population before and after selection, a simple monogenic model for selection for virulence could be validated, as described earlier (Schouten & Beniers, 1997). Here, this study is extended to effects on virulence to other cultivars.

Nematode populations
The G. pallida populations D383 (Pa2) and Coll. 1112 (Pa3) were used, as in a previous study (Schouten & Beniers, 1997). These two populations were mixed at an initial ratio of 80% Pa2:20% Pa3 by crushing approximately 8000 and 2000 cysts, respectively, and mixing the suspensions of eggs. Selection experiments were also performed using the Pa3 population alone. Table 2 shows the sequence of alternation of the cultivars. Cultivars Elles and Darwina were chosen, as at the time when the experiments started, these cultivars were used in regions in the Netherlands for production of starch potatoes, and were known for their resistance to Pa2. Cultivar Karakter was used as it has resistance to both Pa2 and Pa3. Cultivar Irene was used as a susceptible reference cultivar, and the generally less susceptible cv. Elkana was also included. More detailed levels of resistances are shown in Table 3.

Production of successive nematode generations
The nematode populations were reared on the susceptible cv. Irene to provide the unselected populations, and on partially resistant cultivars to provide the selected populations. The populations were produced in 'production series'. Each plastic pot was filled with 5 kg dry weight of Seinhorst soil-mixer, consisting of silver sand, hydrogravel and clay powder (12:3:2), nutrients and 100 g water per kg dry weight (Teklu et al., 2018). Suspensions of eggs were obtained by crushing approximately 1400 cysts. Four eggs per g of dry soil were inoculated into the soil, using 20 cm-long needles. After inoculation, two potato sprouts were planted per pot and placed in an air-conditioned glasshouse with a day length of 16 h and a temperature of 20°C. Seventeen weeks after inoculation, the watering was stopped. The cysts were elutriated from the soil and stored at approximately 20°C to obtain two generations per year.

Measurement of the reproduction factors
In addition to these 'production series', there were also 'measurement series', in which the reproduction of the selected and unselected populations were measured simultaneously in four replicates under the same conditions in a temperaturecontrolled glasshouse, on roots from tubers from the same potato stocks, in a fully randomized design. Unlike the 'production series', only one sprout per pot was planted, and the moisture content was adjusted more accurately to between 10% and 15% by weighing all the pots individually once a week. The pots were moved twice a week to diminish position effects. The numbers of newly formed eggs per inoculated egg (P f /P i ) were estimated from the collected cysts, as was the number of eggs in the newly formed cysts. Further details are described by Schouten & Beniers (1997).

Statistical analysis
As reproduction rate and virulence depend on environmental conditions and quality of the mother tubers, these parameters were analysed, comparing selected and unselected populations within a generation, and not between generations. In order to obtain normal distributions and homogeneous variances, data were log-transformed, and pooled variances were obtained within a generation for comparison of selected and unselected populations by means of Student's t-test. As virulence was not directly estimated from the reproduction data, but from a quotient (i.e. reproduction rate on the test host:reproduction rate on the susceptible host), an extra step for estimating the pooled variance was applied, as described by Schouten & Beniers (1997). The statistical package GENSTAT (Payne, 2004) was used.

Reproduction rate of Pa2/3 mixture reared on cv. Darwina
The mixed 80% Pa2:20% Pa3 population reared for four generations on cv. Darwina resulted in a 61 times higher reproduction rate on Darwina compared to rearing for the same number of generations on the susceptible cultivar Irene (Table 3). The main reason for this very clear increase was selection for virulence by cv. Darwina, as the Pa3-nematodes had a more than 1000 times higher reproduction rate (P f /P i ) on this cultivar compared to the Pa2-nematodes (Table 3). The second reason was gaining the Hardy-Weinberg equilibrium in the unselected population after one generation. This can be illustrated by a simple calculation: if, for simplicity, monogenic, recessive virulence and homozygosity of the starting populations is assumed (AA for Pa2 and aa for Pa3), the initial mix consisted of 80% Pa2-nematodes Table 2 The sequence of potato cultivars over four consecutive generations of the mixed Globodera pallida Pa2/3-population and the Pa3-population in the production series. and 20% Pa3-nematodes (80% AA + 20% aa), but after one generation the alleles are redistributed among the new nematodes according to the Hardy-Weinberg equilibrium (64% AA, 32% Aa, 4% aa), reducing the Pa3nematodes (aa) from 20% to 4% in one generation, i.e. a 5-fold reduction, without selection. Therefore, the 61fold increase of the reproduction rate of the selected population compared to the unselected population can be explained by selection for virulence in the selected population, combined with the Hardy-Weinberg equilibrium in the unselected population.
Reproduction rate of Pa2/3 mixture on other cultivars Table 3 shows that rearing on cv. Darwina not only raised the reproduction rate (P f /P i ) on cv. Darwina itself, but also the reproduction rate on other potato genotypes. The effect was strongest for cv. Darwina, but the relative reproduction rates (i.e. the reproduction rate of the selected population relative to the unselected population) on AM 78-4102 and cv. Elles were also very high (58 and 41, respectively). Table 3 also reveals these increases of relative reproduction rates were mostly predominant for the hosts that strongly differentiated between the Pa2 and Pa3 population. Initially, the Pa2/3 population was a mixture of 80% Pa2 and 20% Pa3. During selection, cv. Darwina pushed the reproduction rate of Pa2/3 towards the level of Pa3. As a result, genotypes that were more susceptible to Pa3 also became more susceptible to the selected Pa2/ 3 mix.

Selection for virulence and aggressiveness
Rearing the Pa2/3 mixture on cvs Darwina and Karakter did not decrease the reproduction on cv. Irene (Table 3), and therefore did not reduce the aggressiveness (fitness) of these nematodes. Apparently, the selection for increased virulence did not occur at the expense of fitness on a susceptible host.

Selection for virulence of the Pa3 population
Selection not only occurred in the mixed Pa2/3 population, but also in the Pa3 population (Table 3), though to a far lesser extent. Continuous rearing of the Pa3 population on cv. Darwina increased the relative reproduction rate on Darwina itself, but also on cvs Elles, Karakter and several other resistant genotypes.
Rearing of Pa3 on cv. Karakter raised the relative reproduction rate on cv. Karakter itself to 4.2, and on cv. Darwina to 1.7. It also raised the relative reproduction rate on other resistant genotypes (Table 3).

Alternation of cultivars
Tables 4 and 5 show the effects of alternation of cultivars on relative reproduction rate. The rearing of the

Modelling selection for virulence
Several models for selection for virulence have been presented (Jones, 1985;Spitters & Ward, 1988;Schouten, 1993Schouten, , 1994Schouten, , 1996Schouten, , 1997. In a previous paper (Schouten & Beniers, 1997) a simple monogenic model was described for simulation of selection for virulence. Virulence is the reproduction rate of the nematodes on a resistant cultivar relative to the reproduction rate on the susceptible reference cultivar. Table 6 shows the calculation procedure according to this model. The model predicted that after four generations of rearing the mixed Pa2/3 population on cv. Darwina, the frequency of the virulent nematodes (the Pa3-nematodes) is 0.9, and the frequency of the avirulent nematodes is 0.1. If a certain cultivar has a resistance gene effective to Pa2, but ineffective to Pa3 populations, then selection by cv. Darwina should also raise the virulence to that cultivar. In mathematical terms: the virulence (Vir) of the Pa2/3 population would have been pushed from (0.8 9 Vir Pa2 + 0.2 9 Vir Pa3 ) before selection to (0.1 9 Vir Pa2 + 0.9 9 Vir Pa3 ) after four generations of the selection on cv. Darwina. Table 3 allows the estimation of Vir Pa2 and Vir Pa3 , as virulence is defined as the reproduction rate on the resistant host divided by the reproduction rate on the susceptible host. From these values, the expected final virulence level could be calculated according to the model, as 0.1 9 Vir Pa2 + 0.9 9 Vir Pa3 , and could be compared with the observed levels, shown in Table 3.
In Figure 1 these expected and observed virulence levels are plotted for a set of Solanum genotypes. Despite the simplicity of the underlying assumptions, the model predicted reasonably well. Heterosis effects that arose from mixing Pa2 and Pa3 were filtered out by using virulence data instead of reproduction rates (Schouten & Beniers, 1997).

Modelling the effect of the Hardy-Weinberg equilibrium on virulence
The effect of obtaining the Hardy-Weinberg equilibrium during rearing on the susceptible cultivar Irene was also modelled (Table 7). Although no selection occurred, according to the monogenic model the virulence would decrease from (0.8 9 Vir Pa2 + 0.2 9 Vir Pa3 ) Data within a row followed by a different letter are significantly different (P < 0.05). The italicized data refers to rRR on a cultivar after continuous cultivation on that cultivar. a Continuous cultivation of one cultivar. Number of generations in parentheses. b Alternation of cultivars. The exact sequence of alternation is shown in Table 2. Karakter (3), Irene (1) Darwina (2), Elles (2) Karakter (2) to (0.96 9 Vir Pa2 + 0.04 9 Vir Pa3 ) after rearing on cv.
Irene. This was validated for the same eight Solanum genotypes as shown in Figure 1. The results in Figure 2 again show a reasonable agreement between the expected outcome of the model and the observed virulence levels, despite the simplicity of the model, and the rigid assumptions underlying it. Figures 1 and 2 indicate that the selection on Pa2/3 by cv. Darwina and obtaining the Hardy-Weinberg equilibrium can to a large extent be explained by monogenic virulence to one resistance gene.

Discussion
After four generations on cv. Darwina, the mixed Pa2/3 population had not only increased its relative reproduction rate on cv. Darwina by a factor of 61, but had also increased considerably the relative reproduction rate on other hosts such as Elles, Karakter and several AM 78selections. For cv. Elles this can be understood from the sources of resistance, as cvs Darwina and Elles have been derived from the same three resistance sources.
Both cvs Elles and Darwina differentiated strongly between the Pa2 and Pa3 populations, by ratios of 719 and 1036, respectively. This explains the strong selection pressure towards Pa3 by cv. Darwina, and the resulting increase of relative reproduction rate on both cvs Elles and Darwina. The higher the differentiation by a host between the Pa2 population and Pa3 population, the stronger the relative reproduction rate of the Pa2/3 population to that host after selection by cv. Darwina. This is studied in more detail in the modelling paragraph.   The source of resistance that might have caused the differences in virulence of the Pa2 and Pa3 populations can be derived from Table 3. Cultivar Elkana has been derived from the resistance sources S. tuberosum subsp. andigena CPC 1673 and S. edinense, but was hardly resistant to the Pa2 and Pa3 populations. As a result, there was no significant selection of the mixed Pa2/3 population on cv. Elkana. Selection ODV 22731 has only S. vernei I-3 in its pedigree, and apparently did not differentiate between Pa2 and Pa3. Both cv. Elkana and ODV 22731 showed no increased relative reproduction rate after selection of Pa2/3 on cv. Darwina. Therefore it is concluded that Elkana and ODV 22731 did not exert a clear selection pressure on Pa2/3.
The cultivars Darwina and Elles have been derived from sources 1, 2 and 3 (Table 3). These cultivars both have the source S. vernei 24/20, which is not present in either cv. Elkana or ODV 22731. It is probable that S. vernei 24/20 is responsible for the strong ability to differentiate between the Pa2 and Pa3 population, and for the very significant selections for virulence by cvs Elles and Darwina. These genotypes contain the resistance gene Grp1 (van Eck et al., 2017). Grp1 contributes to a major resistance to Pa2 and partial resistance to Pa3 (Rouppe van der Voort et al., 2000). Most likely, the selection exerted by cv. Darwina on the Pa2/3 population is caused by Grp1, providing resistance to Pa2. This gene probably descends from S. vernei 24/20.
The differential between the Pa2 group and the Pa3 group, as proposed by Kort et al. (1977), is S. vernei hybrid 62-33-3, also referred to as (VT n ) 2 62-33-3. This differential has been derived from sources 2 and 3 (Table 3). From Table 3 it can be deduced that the ability of the S. vernei hybrid (VT n ) 2 62-33-3 to differentiate between the Pa2 group and Pa3 group is probably also caused by the Pa2 resistance provided by Grp1. Turner et al. (1983) and Turner & Fleming (2002) used the same differential for selection to increased virulence in G. pallida. This selection can probably be mainly ascribed to the same resistance gene (Grp1) that caused the selection in the present experiments with Pa2/3. Selection on cv. Darwina also increased virulence to other genotypes that have been derived from sources 2 or 3 (cv. Karakter,. These genotypes probably also harbour Grp1.  Schouten & Beniers (1997). The The relationship between the expected virulence of the Pa2/3 population after rearing on cv. Irene according to the model described in  (2000) showed that Gpa5 and Gpa6 together provide high levels of resistance to Pa3. This suggests that Gpa5 and Gpa6 are responsible for the stronger selection pressure. Selection on cv. Karakter also increased the reproduction rate on cv. Darwina and all other tested Solanum genotypes.
The rather constant reproduction rate of Pa3 on the susceptible cv. Irene indicates that the selection had no clear impact on aggressiveness. As shown earlier (Schouten & Beniers, 1997), the increased reproduction rate was a result of increased virulence, not of increased aggressiveness.
As a measure for selection pressure in Pa2/3 towards Pa3 at the expense of Pa2, the quotient (reproduction rate of Pa3):(reproduction rate of Pa2) may be used. This quotient appears to be predictive for the selection rate of Pa2/3 and for the efficiency of alternation of the cultivars for slowing down selection in this population.
Alternation of cvs Elles and Darwina was not effective, whilst alternating with the nonselective cv. Elkana was more effective. This indicates that alternation of cultivars did not halt selection if the cultivars differentiated strongly between Pa2 and Pa3, in favour of Pa3. Only alternations with cultivars that did not provide a selective advantage for Pa3 were helpful in delaying selection. Unfortunately, no cultivars were available that allowed stronger reproduction of Pa2 compared to Pa3.
Globodera pallida nematode movements in the soil are very limited. As a result, populations of this nematode tend to inbreed, thus loosing heterozygosity. This inbreeding tendency favours recessive virulence (Montarry et al., 2015). This was not accounted for in the modelling here, nor in the experiments, as the alleles were mixed randomly. The inbreeding tendency will boost the increase in virulence in the field.
Several conclusions can be drawn from this study. Rearing Pa2/3 on cv. Darwina not only raised the reproduction rate on cv. Darwina itself, but also the reproduction rate on other potato varieties. The increased reproduction rates on hosts other than Darwina were mostly predominant for the hosts that highly differentiated between the Pa2 population and the Pa3 population. Both cvs Elles and Darwina discriminated very significantly between these two populations by a ratio of 719 and 1036, respectively. This explains the strong selection pressure towards Pa3 by cv. Darwina, and the resulting increase in reproduction rate on both cvs Elles and Darwina. These very high increases in relative reproduction rate after rearing Pa2/3 on cv. Darwina are caused not only by selection for virulence, but also by a decrease in virulence after rearing on cv. Irene, therewith achieving the Hardy-Weinberg equilibrium. These changes in virulences for Pa2/3, due to selection and the Hardy-Weinberg equilibrium, could be predicted reasonably well by a simple monogenic model. This indicates that for Pa2/3, these changes could be explained to a large extent by one major resistance gene and one complementary avirulence gene only.
Based on pedigree information, selection rate and the monogenic model, the selection exerted by cv. Darwina on the Pa2/3 population is probably caused by the resistance gene Grp1 from (VT n ) 2 62-33-3, the resistant parent of Darwina (Rouppe van der Voort et al., 1998;van Eck et al., 2017). This gene provided resistance to the Pa2 population but to a far lesser extent to the Pa3 population.
The genes Grp1, Gpa5 and Gpa6 are likely to be present in cv. Karakter, and the Solanum genotypes AM 78-3778 and AM 78-4102, in view of the pedigrees.
Alternating cultivars did not halt selection if the cultivars differentiated heavily between Pa2 and Pa3, in favour of Pa3. Only alternating with cultivars that did not provide a selective advantage for Pa3 were helpful in delaying selection. Selection occurred not only in the mixed Pa2/3 population, but also in the Pa3 population, though to a far lesser extent. Selection to a more virulent 'pathotype' took place ('Pa4/Emstland'; Niere et al., 2014). Selection of Pa3 on cv. Karakter also increased reproduction rate to cv. Darwina and all other tested Solanum genotypes.
As appeared from selection of Pa2/3 on cvs Darwina and Karakter, it is not the resistance level of a host to a nematode population that determines the rate of selection, but the differences in virulence levels within the nematode population are the main factor. This is in agreement with the classical population genetics (Fisher, 1930). Apparently, for cv. Karakter the variance for virulence was larger within Pa3 than within Pa2/3. For cvs Elles and Darwina the opposite was true. The simple monogenic model could also explain the virulence change of the Pa3 group to a more virulent 'pathotype'. The differentials S. vernei hybrids AM 78-3787 and AM 78-4102 with resistance to the Pa2 and Pa3 group in particular showed selection to a more virulent 'pathotype' (Hockland et al., 2012).