The effectiveness and costs of pathogen resistance strategies in a perennial plant
Summary
- Plants have evolved different strategies to resist pathogens, but little is known about how effective, stable and costly these mechanisms are in perennial plants across multiple growing seasons.
- We conducted a laboratory experiment to assess resistance variation in Plantago lanceolata against the powdery mildew-causing fungus Podosphaera plantaginis and to measure possible trade-offs between the different resistance strategies. To test stability and costs of resistance, we established common garden populations of plants possessing three different resistance strategies: qualitative resistance that is the ability to block pathogen infection, quantitative resistance that is the ability to mitigate pathogen development and susceptibility. We measured their performance with and without disease for 3 years.
- There were no trade-offs between qualitative and quantitative resistance, and the components of quantitative resistance were positively correlated.
- Throughout the 3-year common garden study, pathogen loads were significantly affected by host resistance in the study populations. Qualitative resistance efficiently blocked infections but contrary to laboratory obtained results; quantitative resistance did not hinder epidemic growth.
- We detected costs in plant performance for qualitative and quantitative resistance compared to susceptible plants in absence of the pathogen. The costs of infection varied according to resistance strategy, pathogen load and plant age.
- Synthesis. In perennial plants, the costs and benefits of resistance need to be evaluated over multiple years as they may change with plant age. Our results give new insights into how polymorphism in resistance can be maintained through costs of resistance in plant performance and through shifts in resource allocation between vegetative growth and reproduction under infection.
Introduction
Most known wild plant species are hosts to some pathogens, yet we rarely witness devastating epidemics in native plant–pathogen interactions. One possible reason for this is the extensive variation in resistance that wild plant populations have against their pathogens (Thrall, Burdon & Young 2001; Koskela et al. 2002; Zhang et al. 2008; Antonovics et al. 2011; Laine et al. 2011; Lebeda et al. 2013). Theory proposes co-evolutionary dynamics through negative frequency-dependent selection to maintain this variation (Frank 1992; Kirchner & Roy 2000) as the rare alleles are favoured over the common ones. Despite being a widely accepted concept, to date, there is relatively limited evidence for negative frequency-dependent selection in wild plant–pathogen associations (Thrall et al. 2012). Scarcity of data may partly result from the difficulties of carrying out co-evolutionary studies (Gaba & Ebert 2009), but it is also worth noting that while some of the underlying assumptions of co-evolutionary selection are well supported by data, other fundamental aspects are not. Firstly, it is relatively well-established that there is variation in the traits of hosts and pathogens (Salvaudon, Giraud & Shykoff 2008; Laine et al. 2011; Tack et al. 2012) and that many host–pathogen interactions are governed by specificity (Thompson & Burdon 1992; Laine 2007). However, theory also assumes (1) resistance to be effective and stable through time, (2) infection to reduce host fitness and (3) resistance is costly in the absence of disease (but see Laine & Tellier 2008; Tellier & Brown 2011). To date, empirical support for these assumptions is surprisingly scarce and variable (Bergelson & Purrington 1996; Mauricio & Rausher 1997; Kniskern & Rausher 2007), particularly for perennial plants.
In their defence against pathogens, plants rely on a two-step mechanism. The first step qualitatively blocks the infection, and if that fails, the second step quantitatively restricts pathogen growth and reproduction (Jones & Dangl 2006; Dodds & Rathjen 2010). Qualitative resistance is usually based on R-genes (resistance genes) with major effects that are responsible for pathogen recognition and induction of defence mechanisms (Bergelson et al. 2001), whereas the quantitative resistance mechanisms are proposed to be under polygenic control (Poland et al. 2009). The value of quantitative resistance in disease management is in the ability to delay pathogen generation time which is expected to result in reduced transmission (McDonald & Linde 2002; Poland et al. 2009). Qualitative defence is often based on a hypersensitive response that induces localized programmed cell death that blocks pathogen growth (Jones & Dangl 2006) and can be detected in the laboratory when a host plant is challenged with pathogen strains (Laine 2007). Fine-scale monitoring of pathogen development during laboratory inoculation assays allows detecting the effects of the quantitative trait loci on infection development (Azzimonti et al. 2013; Le Van et al. 2013). To date, the factors affecting the stability of both qualitative and quantitative resistance phenotypes during the growing season are poorly understood. The resistance phenotype may depend on both plant developmental phase and plant senescence (Boege & Marquis 2005; Whalen 2005; Marçais & Desprez-Loustau 2012; Rodewald & Trognitz 2013). Moreover, the full resistance repertoire may not be expressed constantly (Boege & Marquis 2005). Prior contact with a pathogen or herbivore can alter the resistance reaction against later arriving antagonists (van Hulten et al. 2006; Conrath 2011; Laine 2011).
When a plant has become infected, there are several limitations in measuring and generalizing the costs (expressed as reduced growth, reproduction and/or life span) of infection for the host. Firstly, the outcome of any given host genotype–pathogen genotype interaction can be influenced by abiotic conditions such as temperature and nutrient status (Laine 2007; Wolinska & King 2009) as well as simultaneously occurring microbes (Hodgson et al. 2004; Laine 2011). Secondly, any association between pathogen damage and reduction in host performance can be masked if infection alters host development (Shykoff & Kaltz 1998; Pan & Clay 2003; Korves & Bergelson 2004), or if the resistance reaction itself, successful or not, changes resource allocation to growth and development via plant hormones that are also entangled in disease resistance (Navarro et al. 2008; Alcazar et al. 2011; Robert-Seilaniantz, Grant & Jones 2011; Denancé et al. 2013b; Vos, Pieterse & van Wees 2013). Finally, generalization of the reduction in the host performance is hardly feasible as plant genotypes may differ in their tolerance to damage with some genotypes being able to minimize the negative effects on the performance due to infection, while others are more vulnerable (Baucom & de Roode 2011).
In theory, polymorphism in resistance within populations is maintained by costs of resistance in the absence of the pathogen, whereas under pathogen attack, the resistant hosts outperform the susceptible ones (Frank 1992; Antonovics & Thrall 1994; Tellier & Brown 2007). The rationale behind the theory is that plant performance (expressed as reduced growth, reproduction or life span) is assumed to be weakened by infection and resistance, as under infection the host resources are redirected for pathogen growth and reproduction, and possessing resistance requires resources that are not available for growth or reproduction (Crawley 1997). In reality, detection of resistance costs in the presence or absence of disease depends on the underlying resistance mechanisms (Heil & Baldwin 2002). Costs can arise either through resource allocation or pleiotropy of the existence or expression of resistance genes (Tian et al. 2003; Todesco et al. 2010; Denancé et al. 2013a) or autotoxicity of the defence mechanisms (Baldwin & Callahan 1993) that harm the host itself or its mutualists (Heil & Baldwin 2002). To date, the evidence of costs associated with resistance remains mixed (Bergelson & Purrington 1996; Brown & Rant 2013).
Here, we study the efficacy and stability of qualitative and quantitative resistance and their costs in the perennial Plantago lanceolata against its powdery mildew pathogen Podosphaera plantaginis. Previous studies have revealed considerable diversity in resistance against P. plantaginis both within and between local host populations (Laine 2004). The level of host population resistance changes with spatial structure with direct effect on epidemiological dynamics; well-connected host populations experience less pathogen colonization and higher pathogen extinction rates due to their higher resistance level (Jousimo et al. 2014). Hence, understanding what maintains resistance diversity even in the absence of disease has proven to be one of the key questions for understanding the drivers of disease dynamics. For this purpose, we first screened resistance variation across 41 host genotypes originating from seven natural populations and tested for possible trade-offs between the resistance traits. We then established multiyear common garden populations to understand how effective and stable different resistance strategies are, how resistance and infection affect plant performance (e.g. flower and leaf production) and whether resource allocation between growth and reproduction arises (Crawley 1997). Jointly, our results shed much needed light on the eco-evolutionary dimensions that govern plant–pathogen interactions in the wild.
Materials and methods
Plantago–Podosphaera System
Plantago lanceolata L., ribwort plantain, is a monoecious perennial rosette-forming herb (Sagar & Harper 1964). The plant is a wind-pollinated obligate outcrosser (Ross 1973), and the seeds drop close to the mother plant, forming a long-term seed bank (Bos 1992).
Podosphaera plantaginis (Castagne; U. Braun and S. Takamatsu) is an obligate biotroph belonging in the order Erysiphales within the Ascomycota. It is a specialist pathogen infecting only P. lanceolata. Like all powdery mildews, it completes its whole life cycle as localized lesions on host leaves only the haustorial feeding roots penetrating the host tissue to feed nutrients from its host (Bushnell 2002). The pathogen is an obligate biotroph and requires living host tissue through its life cycle (Bushnell 2002). The harm experienced by the host is assumed to result from decreased photosynthesis and loss of nutrients (Jarvis, Gubler & Grove 2002).
The interaction between P. lanceolata and P. plantaginis is based on a two-step mechanism typical for plant–pathogen association (Jones & Dangl 2006; Laine 2007). The first step is the plant's ability to recognize the pathogen and block its growth (Laine 2004). This interaction is strain-specific, suggesting a gene-for-gene type control (Thompson & Burdon 1992; Laine 2004, 2007). In the second step, the host can mitigate pathogen growth and reproduction, and the infection outcome is governed by both host and pathogen genotypes (Laine 2004, 2007). Previous studies have detected considerable variation in resistance and infectivity within host and pathogen populations, respectively (Laine 2004).
The Plantago–Podosphaera interaction has been intensively studied since 2001 in the Åland islands in SW Finland where the pathogen persists as metapopulation with frequent extinctions and (re) colonizations (Laine & Hanski 2006; Ojanen et al. 2013). Annually, P. plantaginis is present in 2–16% out of the 4000 host populations (Laine & Hanski 2006; Jousimo et al. 2014). Infection is visible to the naked eye as lesions consisting of whitish mycelia and chasmothecia, resting structures that are seen as black dots within mycelia. Podosphaera plantaginis spreads via asexual conidial spores that are carried by wind. Typically, the epidemics start in early June with 6–9 asexual cycles during growing season until the epidemics seize by September (Laine 2004). In the autumn, when the host plants die back to the rootstock, the pathogen populations crash, and overwintering takes place as chasmothecia on decaying leaves (Tack & Laine 2014).
Plant and Fungal Material
Plantago lanceolata seeds from seven natural populations across Åland Islands (IDs 4 (eight plants), 325 (one plant), 511 (seven plants), 1062 (four plants), 1413 (eight plants), 2220 (eight plants) and 9031 (five plants)) were collected in September 2010. The source populations represented different regions of the metapopulation and were infected with P. plantaginis in year 2010. Between 2001 and 2009, the pathogen was present for three different years within population 4, for 2 years within populations 511 and 2220, and 1 year within populations 325 and 1413. In populations 1062 and 9031, the pathogen was detected for first time in 2010. Podosphaera plantaginis strains were collected from six allopatric natural pathogen populations in July (IDs (years of pathogen presence given in parentheses, 877 (8), 689 (6), 228 (3), 2821 (4), 9066 (2)) and in September (ID 9609 (0)) of year 2010. We collected one strain from each population except from population 689 where two strains were collected. The source populations were selected to be allopatric to all the plants used in the experiment to get comparable results as pathogen local adaptation may affect the infection outcome (Susi & Laine 2013). The seeds were sown on 1:1 mixture of potting soil and sand and grown in the greenhouse at 20 ± 2 °C with 16:8 L:D photoperiod. The fungal strains were purified through three cycles of single-colony inoculations and maintained on live leaves in Petri dishes in a growth chamber 20 ± 2°C with 16:8 L:D photoperiod. The strains were transferred to fresh susceptible P. lanceolata every 2 weeks. To produce adequate stocks of sporulating fungal material, repeated cycles of inoculations were performed.
Characterization of Resistance and Infectivity
To characterize plant resistance and fungal infectivity and aggressiveness, an inoculation experiment was set up. We used seven fungal strains and 41 plant genotypes resulting in 287 host genotype–pathogen genotype combinations. Detached leaves placed on moist filter paper on Petri dishes were inoculated with conidial spores from 1 cm2 sized lesions by evenly brushing the spores on a leaf. Pathogen development was observed daily under a dissecting microscope starting at fourth day post-inoculation (DPI) until day 12. Qualitative resistance was recorded as 0 = no resistance (infection) and 1 = resistance (no infection). We measured three components of quantitative resistance that correspond to pathogen aggressiveness: time to pathogen germination, time to pathogen sporulation and pathogen lesion development at day 12 DPI. The first day when fungal hyphae were observed was considered as the day of germination, and the first day when spores were observed was considered as the day of sporulation. Pathogen lesion development was scaled from 0 to 4: 0 = no mycelium, 1 = only mycelium; 1.5 = mycelium and spores visible only through microscope; 2.5 = mycelium and sparse sporulation visible with naked eye; 3 = abundant sporulation, colony size < 0.5 cm2; and 4 = abundant sporulation and colony size > 0.5 cm2 (Bevan, Crute & Clarke 1993).
Out of the seven pathogen strains used in the inoculation study, two strains – strain 3 originating from population 877 and strain 10 originating from population 2821 – were chosen for the common garden experiment based on their infectivity profiles (Table S1 in Supporting Information). As infection and resistance are strain-specific in this system (Laine 2004), we chose pathogen strains that were infective on most of the plant material. Based on their responses in the inoculation experiment, plant material was selected for the experiment and assigned to three resistance classes: qualitatively resistant, quantitatively resistant and susceptible (Table S2 in Supporting Information). Qualitative resistance in the plant was defined as the number of pathogen strains the plant was able to resist, and quantitative resistance was measured as pathogen development rates. Qualitatively resistant plants expressed resistance against at least two pathogen strains and were susceptible to at least one of the pathogen strains used in the experiment to sustain infection in the population. This allowed estimating the cost of also those R-genes that are unnecessary. For quantitatively resistant and susceptible common garden populations, we chose plants that were qualitatively susceptible to both of the pathogen strains used in the experiment but differed in their quantitative response to infection, with as few qualitative resistance responses as possible expressed in the laboratory experiment. The quantitatively resistant plants were able to delay the development of the pathogen strains as their specific and average development rate was not higher than 2.5 in the Bevan's scale. This score describes visually conspicuous lesions that have very low spore production. Susceptible plants allowed the two chosen fungal strains to develop lesions of 3 or 4 on the Bevan's scale (i.e. high sporulation) and were generally susceptible to other strains as well (lesion development rates between 2.7 and 3.6 on Bevan's scale).
Cloning the Plant Material
The plant genotypes were cloned up to 30 ramets in the greenhouse (Laine 2004). The mother plants were placed on top of vermiculite-filled pots, and the roots were grown through the vermiculite to water. After 8 weeks, the roots were cut and started to sprout. Four weeks after sprouting, the ramets were planted to 8 cm × 8 cm pots in greenhouse. When the ramets were 8 weeks old, they were given 1 week acclimation outside and subsequently planted in the outdoor common garden populations. From the study area, soil was removed and replaced with a mixture of garden soil and sand to create homogenous growth medium.
The Common Garden Populations
To measure the effectiveness and cost of plant resistance and cost of infection for plant performance under natural conditions, we established multiyear common garden populations at the Lammi Biological Station (61°05′28′'N, 25°03′90′'E) in southern Finland by planting P. lanceolata plants in June 2011 and 2012. We continued the experiment until 2013, and during that time, we measured the plants and pathogen load on each growing season resulting in measurements of first 12 populations for 3 years and the second 12 populations for 2 years. Lammi Biological Station was an ideal place for experiment as there is no danger of contamination from wild since neither the plant nor its pathogen occurs there naturally, yet both are capable of growing and overwintering there. The plants were planted into 24 plant populations each consisting of one resistance type (qualitative, quantitative and susceptible). Each population consisted of 20 plants with 4–7 different genotypes of the same resistance type. The total amount of plants used in the experiment was 480. The resistance profiles and the genotypes of the plants planted each year are given in Table S2.
In years 2011 and 2012, four different pathogen treatments were given: strain 3 singly, strain 10 singly, mixture of strains 3 and 10 and no inoculation as a control. After the first year, we found no significant effect of the pathogen strain used (Susi, H. and Laine, A.-L. in prep). Therefore, in year 2013, only two pathogen treatments were given: all populations except the controls received mixture of 3 and 10, and the controls plots received no pathogen treatment. The pathogen overwintered in the common garden populations in both winters 2012 and 2013, but to ensure disease establishment in all populations, the inoculation treatment was given every year. The populations were separated using plastic fences to prevent cross-contamination between plots. The control plots were placed at 8 metres distance from the pathogen-treated plots and separated with plastic fences to prevent pathogen contamination.
To mimic natural epidemics, the pathogen treatment was given in mid-July. A two-step procedure was used to inoculate the study populations. Firstly, we placed transmission plants containing two infected leaves into the populations. Prior to the experiment, these transmission plants were inoculated with pathogen strains 3 or 10 and kept in separate growth chambers. Each study population received four powdery mildew-infected transmission plants. In the co-infection treatment, two plants of strain 3 and two plants of strain 10 were used. One week later, a second inoculation was performed to ensure initial infection of the plots using fresh conidial spores that were transferred from P. plantaginis colonies maintained on Petri dishes. Each population received an equal amount of inoculum as the same amount of transmission plants, and the conidial spores were brushed to leaves of four randomly chosen plants in the populations. To test efficiency of each resistance type during the growing season and to measure pathogen load on each plant, the epidemics were monitored at four time points 17, 31, 53 and 76 DPI by counting infected leaves of each plant. The pathogen load was measured as the fraction of leaves that contained powdery mildew of all leaves on a plant. The control populations were monitored to ensure that there was no disease spread between plots. When there were diseased leaves found in the control plots, those leaves were removed and pine soap was sprayed to prevent the further spread of the pathogen (once in 2012, three times in 2013). Cutting of the leaves did not have an effect on the size of the plant (PC1 d.f. = 1,39, F = 0.25, P = 0.6191; PC2 d.f. = 1,39, F = 1.19, P = 0.2822; PC3 d.f. = 1,39, F = 0.01, P = 0.9229).
To assess the effect of pathogen infection on plant performance and the potential costs of resistance, plant size was measured in July before the pathogen treatment was applied, and again at the end of the season in late September by counting the number of leaves, and measuring the length and width of the longest leaf on each plant. To measure seed production, in 2011 the number of flowering stalks was counted in early July, and seeds from five stalks of each plant were collected in August and counted. The seeds were stored in paper envelopes at room temperature and subsequently counted and weighed. In 2012 and 2013, seed stalks were counted and collected as they matured and their height was measured. Given the size of the experiment, it was impossible to count all seed which may be considered a key fitness component for P. lanceolata. However, based on seed data from 2011, the measurements we obtained are strongly positively correlated with seed production. There were significant positive correlations between the number of seeds per stalk and height of the seed stalks (F = 19.79, P < 0.0001), between number of stalks per plant and seed weight (F = 9.14, P = 0.0029) and between number of the seed stalks per plant and number of seeds per stalk (F = 9.78, P = 0.0020). Other studies on P. lanceolata have also found a positive correlation between plant size and reproduction (Shefferson & Roach 2010) and number of leaves and length of the longest leaf and reproduction (Mook et al. 1992). Overwintering of the plants was monitored in June. In subsequent years, regeneration from seed was prevented by covering the ground with plastic.
To ensure that there was no recombination between pathogen strains, we genotyped the pathogen populations each year in September and again from the overwintered populations in July 2012 and 2013 by collecting leaf samples (77 in 2011, 277 in 2012, 792 in 2013) and subsequently genotyping them using a set of 27 SNPs as previously described (Tollenaere et al. 2012).
Statistical Analyses
In order to reduce the number of plant trait measures – many of which were correlated – we performed two independent principal component analyses (PCAs). This method allows reducing the number of variables by summarizing them into a few principal components. We performed two PCAs, first one using measurements of year 2011 and second one using the measurements of years 2012–2013. PCAs were run in jmp Version 9 (SAS Institute Inc., Cary, NC) using correlation matrices. In the first PCA, we used measurements of mean seed number, mean seed weight, leaf width and length and leaf number of each plant in year 2011. The PCA performed on traits of year 2011 showed that seed traits were assigned to PC1 correlating positively with leaf width and length, while leaf number was assigned separately to PC2 (Table S3 in Supporting Information). Therefore, in subsequent years, we were able to reduce the traits measured and use the plant leaf length, width and number based growth measures together with flower number and height in years 2012–2013. Only PCs with eigenvalues greater than 0.9 were kept for further analysis. For each plant resistance type and pathogen treatment, the scores for PC components were stored and used to describe the distribution along the axes defined by the estimated PCs. The PC scores were used to analyse the effect of pathogen load, age and resistance strategy on plant performance.
We ran our analyses as generalized linear mixed models in sas 9.1 PROC MIXED (SAS Institute Inc., Cary, NC.) To test whether there were trade-offs between resistance components in the laboratory measured traits, we analysed correlations between the mean values of the resistance traits measured for each of the 41 plant genotypes: qualitative resistance as the fraction of pathogen strains plant genotype was able to resist and quantitative traits measured as pathogen development, that is pathogen germination time, sporulation time and lesion development as measured according to the Bevan scoring (Bevan, Crute & Clarke 1993). When a plant genotype showed qualitative resistance, the subsequent quantitative resistance values (i.e. 0) were not used for the calculation of the mean. Population of origin was defined as a random effect in the model. For the analyses, qualitative resistance was arcsine-transformed to meet model assumptions of normally distributed errors.
To understand the factors affecting pathogen epidemics within season and between seasons, we performed two different analyses. Firstly, we tested the efficiency of resistance strategy against infection during the growing season. The fraction of diseased leaves was defined as the response variable and was analysed with measurement time as a covariate (17, 31, 53 and 76 DPI) separately in each growing season (2011, 2012 and 2013). Plant genotype (nested under resistance strategy), resistance type, population age and pathogen strain were defined as explanatory class variables. In both models, plant individual and plot were defined as random effects, with plant individual hierarchically nested under genotype and resistance type. Secondly, to understand the factors affecting pathogen loads between years, we performed another analysis on all the plots in years 2012–2013 at the peak of epidemics (53 DPI). In these years, the pathogen load at the onset of epidemics was likely to vary among plots following pathogen overwintering from the previous season. The pathogen overwintered in the populations in 2012 and 2013, and therefore, we separated the 2011 dynamics from the subsequent analysis. The pathogen treatments were different in these 2 years: in 2012, we used single inoculation of strains 3 and 10 and a co-infection treatment of the strains 3 and 10, and in 2013, we gave all populations coinfection treatment of strains 3 and 10, and therefore, we did not use pathogen treatment in the analysis. In the 2-year analysis, the age of the plant and year were analysed as explanatory class variables. The data on host plant–pathogen load were arcsine-transformed to meet model assumptions of normally distributed errors.
We then performed four different analyses on the effect of disease and resistance strategy on plant performance. To control the possible variation caused by the location of the population in the study area plot was used as random variable in all these models. In order to evaluate the potential cost of the different resistance strategies in absence of the disease, we investigated the effects of the resistance type of the plant on the plant performance in control populations, using selected PCs coordinates as response variable. The plant genotype was used in the analysis nested under resistance type explaining the amount of variation caused by the genotype within a resistance type. To identify differences within significant main effects, we used post hoc comparisons. To understand the effect of disease and resistance strategies on the performance of plant, we analysed the effect of plant resistance types, age, genotype nested under resistance type, year and pathogen treatment as explanatory class variables. To test how plant performance and resource allocation between vegetative growth (leaves) and reproduction (flowers) respond to varying pathogen load on the susceptible and quantitatively resistant plants, we analysed the effect of plant resistance types, age, genotype nested under resistance type, year and pathogen load as explanatory class variables using a generalized linear mixed model. For this analysis, the pathogen load was divided to five classes depending on the fraction of the leaves diseased (1 = 0% diseased, 2 = 0.01–10% diseased, 3 = 10.01–25% diseased 4 = 25.01–50% diseased 5 = 50.01–100% diseased). To test whether resistance strategy and pathogen treatment had an effect on host overwintering, we analysed the effect of plant resistance types, age, genotype nested under resistance type, year and pathogen treatment as explanatory class variables using a generalized linear mixed model.
Results
Characterization of Plant Resistance in the Laboratory
We found that the majority of plant genotypes (53.6%) possessed no qualitative resistance against the seven pathogen strains used in the study, and only 12.2% of the plants were resistant against 4 or more strains (Fig. 1a). In the components of quantitative resistance, there was more variation (Fig. 1b–d). There were no trade-offs between resistance components of the plants in laboratory study (Table 1), but we found positive correlation between qualitative resistance and one of the quantitative components, sporulation time (Table 1). There were significant positive correlations between all components of quantitative resistance that affect the pathogen traits, lesion development, germination time and sporulation time (Table 1).
| Source | F 1,33 | P |
|---|---|---|
| Qualitative resistance | ||
| Lesion development | 3.98 | 0.0542 |
| Qualitative resistance | ||
| Germination time | 0.62 | 0.4354 |
| Qualitative resistance | ||
| Sporulation time | 5.04 | 0.0318 |
| Lesion development | ||
| Germination time | 23.68 | < 0.0001 |
| Lesion development | ||
| Sporulation time | 26.71 | < 0.0001 |
| Germination time | ||
| Sporulation time | 19.33 | 0.0001 |
The Efficiency of Resistance in Field Conditions
In contrast to the laboratory trials, the quantitatively resistant plants did not result in lower pathogen load at the peak of the epidemics (Table 2; Fig. 2). We found that during the 3-year experiment, the quantitatively resistant plants became as heavily diseased as the susceptible plants at the peak of epidemics at 53 DPI and the pathogen load increased as the plants became older (Table 2; Fig. 2). During all seasons, we found that the qualitatively resistant plants were the least infected (Table 2; Fig. 2).
| Source | 2011 | 2012 | 2013 | 2012–2013 All | P | |||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| d.f. | F | P | d.f. | F | P | d.f. | F | P | d.f. | F | ||
| Resistance type | 2,128 | 59.99 | < 0.0001 | 2,277 | 5.16 | 0.0063 | 2,313 | 13.8 | < 0.0001 | 2,256 | 10.38 | < 0.0001 |
| Genotype | 14,128 | 8.21 | < 0.0001 | 16,277 | 8.72 | < 0.0001 | 16,313 | 13.22 | < 0.0001 | 16,256 | 5.49 | < 0.0001 |
| Days | 3,468 | 200.92 | < 0.0001 | 3,944 | 38.61 | < 0.0001 | 3,920 | 5.99 | < 0.0001 | |||
| Pathogen treatment | 2,468 | 44.51 | < 0.0001 | 2,944 | 0.19 | 0.8248 | ||||||
| Age | 1,944 | 8.47 | 0.0037 | 1,920 | 6.57 | 0.0105 | 2,245 | 6.82 | 0.0013 | |||
| Year | 1,245 | 1.72 | 0.1906 | |||||||||
| Resistance type × days | 6,468 | 30.85 | < 0.0001 | 6,944 | 4.74 | < 0.0001 | 2,920 | 10.51 | < 0.0001 | |||
| Resistance type × pathogen treatment | 4,468 | 6.48 | < 0.0001 | 6,944 | 0.28 | 0.8893 | ||||||
| Genotype × days | 42,468 | 3.13 | < 0.0001 | 48,944 | 2.13 | < 0.0001 | 16,920 | 2.29 | < 0.0001 | |||
| Genotype × pathogen treatment | 28,468 | 1.67 | 0.0181 | 32,944 | 2.93 | < 0.0001 | ||||||
| Days × pathogen treatment | 6,468 | 29.98 | < 0.0001 | 6,944 | 3.62 | 0.0015 | ||||||
| Days × age | 3,944 | 2.28 | 0.0774 | 1,920 | 27.64 | < 0.0001 | ||||||
| Genotype × age | 7,944 | 2.94 | 0.0047 | 21,245 | 1.81 | 0.0185 | ||||||
| Resistance type × age | 2,944 | 1.35 | 0.2586 | 2,920 | 3.15 | 0.0434 | 4,245 | 3.81 | 0.005 | |||
| Resistance type × days × pathogen treatment | 12,468 | 5.57 | < 0.0001 | 12,944 | 3.94 | < 0.0001 | ||||||
| Resistance type × days × age | 6,944 | 7.46 | < 0.0001 | 2,920 | 7.85 | < 0.0001 | ||||||
When measuring within season epidemics, we found that the effect of resistance type was dependent on time (as days; Table 2). For example, in first year of the study in 2011 when there was no overwintered pathogen population present, there were differences between the resistance types at 31 DPI (Table 2; Fig. 2). Qualitatively resistant plants were the least infected (0.0018 ± 0.0003 leaves infected), quantitatively resistant plants supported low pathogen loads (0.0089 ± 0.0009), and as expected, the susceptible ones were most infected and their average pathogen load was 0.0308 ± 0.0069 (Table 2; Fig. 2). The highest peak of the epidemics was measured at 53 DPI, when qualitatively resistant plants were least infected (0.028 ± 0.0025), while the pathogen loads between susceptible (0.275 ± 0.021) and quantitatively resistant (0.268 ± 0.017) had become quite similar (Table 2; Fig. 2). Overall, the efficiency of resistance depended on time, pathogen load increasing rapidly in the quantitatively resistant populations towards the end of the season while in susceptible populations the pathogen load was highest at 53 DPI and in qualitatively resistant plants the pathogen load only moderately increased towards the end shown by the significant interaction between resistance type and days in the analysis (Table 2; Fig. 2). Finally, the interaction between resistance type and time was dependent on the age of the population (Table 2).
Although there was a significant main effect of resistance type (Table 2), there was also significant variation among genotypes within the resistance groups. The genotype-specific expression of the resistance type had a significant effect on pathogen load in all years (Table 2). Moreover, there was a significant interaction between host genotype and time in all seasons (Table 2), and in 2011 and 2012, the interaction between pathogen treatment and genotype was significant (Table 2).
The pathogen treatment effect was large in 2011, but in the second season, the effect of strain diminished (Table 2). In 2012–2013, there was also a significant plant age and genotype interaction (Table 2). Efficiency of resistance also depended on pathogen treatment (Table 2; Fig. 2) as well as on the interaction between pathogen treatment and time (Table 2; Fig. 2).
Principal Component Analysis of Plant Traits
In the PCA of years 2012–2013, the first three PC axes explained well the variation (82.1%; Table S4 in Supporting Information). PC1 was good indicator for the plant size, it consisted of high loadings of leaf width and length positively correlating with flower height and number, and therefore, it was interpreted as ‘plant size’ component (Table S4). PC2 consisted of high loadings of leaf width negatively correlating with flower number and was interpreted as ‘leaf width’ component (Table S4); moreover, PC2 was further used to assess resource allocation between vegetative growth (leaf width) and reproduction (flower number). PC3 was interpreted as ‘leaf number’ variable because leaf number had the highest loading (Table S4). We compared the scores of principal components on the plant resistance strategies in pathogen-treated plants and in control plants (Fig. 3). In both PC1 and PC2 components, the quantitatively resistant plants scored lowest (Fig. 3).While there were generally no overall significant effects of pathogens treatment, the control susceptible plants seemed to perform better than the diseased susceptible plants in PC2 (Table 4; Figs. 3 and S1a,b in Supporting Information). However, there were no statistically significant interaction between resistance type and pathogen treatment, and therefore, the interaction was not included in the model.
Costs of Resistance
We expected the costs for resistance to become visible as reduced performance in the qualitatively and quantitatively resistant plants in the absence of the pathogen. Post hoc comparisons revealed that qualitatively resistant plants displayed evidence for a cost of resistance in performance only along leaf width component PC2 (P = 0.0199), while quantitatively resistant plants displayed evidence for a cost along both plant size component PC1 (P = 0.0039) and the leaf width component PC2 (P = 0.0021; Table 3; Fig. 3) suggesting that quantitative resistance is the costliest strategy. For the leaf number component, PC3, the resistance types did not differ from each other (Table 3; Fig. S1a,b).
| Source | PC1 | PC2 | PC3 | ||||||
|---|---|---|---|---|---|---|---|---|---|
| d.f. | F | P | d.f. | F | P | d.f. | F | P | |
| Resistance type | 2,116 | 4.73 | 0.0106 | 2,116 | 5.27 | 0.0064 | 2,116 | 0.79 | 0.4555 |
| Genotype | 15,116 | 2.42 | 0.0043 | 15,116 | 7.10 | < 0.0001 | 15,116 | 1.62 | 0.0778 |
| Age | 2,116 | 10.10 | < 0.0001 | 2,116 | 0.52 | 0.5969 | 2,116 | 21.85 | < 0.0001 |
| Year | 1,116 | 15.84 | 0.0001 | 1,116 | 14.58 | 0.0002 | 1,116 | 16.63 | < 0.0001 |
Plant Age, Annual Variation and Phenotypic Responses to Pathogen Infection
We compared the effect of plant resistance strategy and pathogen treatment on the three PCs (Fig. 4a–c). In PC1, plant size and age were negatively correlated and the effect of plant resistance strategy varied with age class (Table 4; Fig. 4a). The effect of pathogen treatment on PC1 was also dependent on plant age (Table 4; Fig. 4a). There were noticeable interactions between the age and the effect of disease treatment in PC2 (Table 4; Fig. 4b). The overall performance in PC2 was highest in 1-year-old plants where no difference between diseased and control plants was shown (Fig. 4b). In 2-year-old plants, the control plants had higher performance than diseased in all resistance types (Fig. 4b). In 3-year-old plants, the pattern changed and only susceptible plants had higher performance in control plants, whereas qualitatively and quantitatively resistant plants had lower performance in control plants (Fig. 4b). In PC2, the leaf width correlated negatively with flower production and can be interpreted as an increase in flower production and a shift in resource allocation between vegetative growth and reproduction after a long history with disease (Fig. 4b). PC3 correlated positively with age, but the pathogen treatment and resistance type were not significant (Table 4; Fig. 4c). There was large variation between years in plant performance in all three PCs (Table 4). In the leaf width component (PC2), the effect of pathogen treatment also varied between years (Table 4).
| Source | PC1 | PC2 | PC3 | ||||||
|---|---|---|---|---|---|---|---|---|---|
| d.f. | F | P | d.f. | F | P | d.f. | F | P | |
| Resistance type | 2,573 | 2.97 | 0.0523 | 2,577 | 5.01 | 0.0069 | 2,579 | 0.16 | 0.8548 |
| Genotype | 17,573 | 8.95 | < 0.0001 | 17,577 | 15.28 | < 0.0001 | 17,579 | 3.57 | < 0.0001 |
| Pathogen treatment | 1,573 | 0.83 | 0.3614 | 1,577 | 1.67 | 0.1969 | 1,579 | 2.06 | 0.1514 |
| Age | 2,573 | 22.24 | < 0.0001 | 2,577 | 10.13 | < 0.0001 | 2,579 | 94.13 | < 0.0001 |
| Year | 1,573 | 68.28 | < 0.0001 | 1,577 | 63.53 | < 0.0001 | 1,579 | 86.07 | < 0.0001 |
| Resistance type x age | 4,573 | 2.46 | 0.0446 | ||||||
| Treatment x year | 1,577 | 8.06 | 0.0047 | ||||||
| Treatment x age | 2,573 | 10.51 | < 0.0001 | 2,577 | 11.35 | < 0.0001 | |||
The Effect of Infection on Plant Performance
To test how plant performance changes under increasing pathogen load, we compared the two heavily diseased resistance types, susceptible and quantitatively resistant, across the different pathogen load levels (Table 5; Fig. 5a–c). The plant size component (PC1) responded to increasing pathogen load, and in general, the quantitatively resistant plants were smaller (Table 5; Fig. 5a). There was a negative relationship between pathogen load and plant size component (PC1; Table 5; Fig. 5a). Also, quantitatively resistant plants had lower growth than susceptible plants when they were infected (significant effect of resistance type Table 5; Fig. 5a). When we tested the effect of resistance type on the leaf width component PC2 on increasing pathogen load, we found that the response varied according to resistance strategy (Table 5; Fig. 5b). The quantitatively resistant plants had lower overall values and a negative correlation between PC2 and pathogen load, whereas susceptible plants responded strongly to increasing pathogen load and had positive correlation with pathogen load suggesting different resource allocation between vegetative growth and reproduction under infection. The leaf number component (PC3) increased in response to pathogen load (Table 5; Fig. 5c). Quantitatively resistant plants had higher performance in PC3 than susceptible plants (significant effect of resistance type; Table 5; Fig. 5c).
| Source | PC1 | PC2 | PC3 | ||||||
|---|---|---|---|---|---|---|---|---|---|
| d.f. | F | P | d.f. | F | P | d.f. | F | P | |
| Resistance type | 1,300 | 3.90 | 0.0208 | 1,300 | 2.42 | 0.1209 | 1,304 | 7.41 | 0.0069 |
| Genotype | 10,300 | 14.58 | < 0.0001 | 10,300 | 13.02 | < 0.0001 | 10,304 | 3.75 | < 0.0001 |
| Pathogen load | 4,300 | 4.58 | 0.0327 | 4,300 | 0.70 | 0.5940 | 4,304 | 4.55 | 0.0014 |
| Age | 2,300 | 37.31 | < 0.0001 | 2,300 | 31.79 | < 0.0001 | 2,304 | 65.11 | < 0.0001 |
| Year | 1,300 | 118.68 | < 0.0001 | 1,300 | 47.57 | < 0.0001 | 1,304 | 60.77 | < 0.0001 |
| Pathogen load × year | 4,300 | 3.48 | 0.0085 | ||||||
| Resistance type × pathogen load | 4,300 | 4.28 | 0.0022 | ||||||
Plant Overwintering
There was no significant effect of neither pathogen treatment nor resistance type or plant age on P. lanceolata overwintering (Table S5 in Supporting Information). However, genotypes differed in their overwintering (Table S5), and in year 2013, the overwintering success of the plants was significantly higher than in 2012 (89% vs. 65%, respectively; Table S5).
Discussion
We tested the efficiency and costs of different plant resistance strategies and the costs of P. plantaginis infection in the perennial host, P. lanceolata. We found differences in the stability of plant resistance strategies: laboratory measured quantitative resistance did not reduce pathogen load at the epidemic peaks in a multiyear study, but qualitative resistance remained efficient and stable throughout the study. We detected costs in plant performance for both quantitative and qualitative resistance. The cost of infection on host performance varied according to plant age, resistance type and the level of pathogen load.
Qualitative resistance was rare among the plant genotypes collected from the natural populations. However, in the common garden study, it was an efficient and stable strategy through the years. This was in line with previous findings from this system documenting that the qualitative component of the interaction remains stable across environments, while the quantitative interaction is altered by environmental variation (Laine 2007). Indeed, the disease symptom delaying effect of quantitative resistance measured under laboratory conditions was observed only in the first season and did not hinder epidemic growth. At the peak of epidemics, pathogen load in quantitatively resistant plants was not lower than in the susceptible plants. One reason for this might be that quantitative resistance is the sum of minor gene effects (Poland et al. 2009) that can be traced to components (Azzimonti et al. 2013; den Boer et al. 2013; Le Van et al. 2013) which can have development phase-specific effects on the disease progress (Azzimonti et al. 2013). Moreover, some of these components can even enhance susceptibility (den Boer et al. 2013). There was an increasing trend in the pathogen load of quantitatively resistant plants over the years which is in line with previous study (Rodewald & Trognitz 2013), suggesting that ageing or senescence may decrease host resistance. However, we cannot exclude the possibility that the observed outcome is due to pathogen adaptation or mutation, although no recombination was observed in our study. Theoretically quantitative resistance can select for higher virulence, that is the amount of damage the pathogen causes for the host (Gandon & Michalakis 2000), and breakdown in quantitative resistance by increase in pathogen aggressiveness has been reported empirically (Palloix, Ayme & Moury 2009; van den Berg et al. 2014). In our study, however, the pathogen load in susceptible plants did not increase suggesting that pathogen aggressiveness remained stable. In all our analyses, the phenotypic variation of the quantitatively resistant plants differed significantly from both susceptible and qualitatively resistant plants, suggesting that our initial classifying was relevant.
Pathogens are considered a selective force due to the negative effects they impose on the performance of their hosts (Rausher 1996), and host–pathogen co-evolutionary theory expectations assumes pathogens to cause harm for the host (Frank 1992; Antonovics & Thrall 1994). In this study, the costs of infection varied according to host age, host resistance strategy and the amount of pathogen load. Given that the pathogen load on the hosts tended to increase over years, we found interesting trajectories in host performance. Firstly, compared to control populations, the overall cost of disease was not large. The difference in plant size (PC1) observed in first year between diseased and control populations attenuated as plants aged, whereas in diseased populations, the advantage of qualitative resistance became more pronounced through time (Fig. 4a). The pattern of leaf width component PC2 did not change between resistance types over the years, but the sensitivity of this component to pathogen treatment changed over years by direction and magnitude (Fig. 4b). Lack of strong benefits for resistance for plant performance has been also reported in Ipomoea (Kniskern & Rausher 2007). In diseased populations, the performance of quantitatively resistant and susceptible plants differed. Quantitative resistance seemed to be costly when infected in both plant size component PC1 and leaf width component PC2. Furthermore, there was a negative effect of increasing pathogen load on quantitatively resistant plant performance. However, the phenotypic changes described by PC2 include negative correlation of leaf width and flower production. Under increasing pathogen load, resources are increasingly used for flower production which is in line with prior observations on P. lanceolata (Dudycha & Roach 2003). Although our study does not explore the actual viability of seeds, this proposes an explanation how this strategy could be maintained. The susceptible plant performance seemed to respond strongly to pathogen infection. In this resistance category, we did not observe directional change in the effect of disease, whereas in other resistance categories, the effect of disease on plant performance was age specific (Fig. 4a–c).
The costs of resistance in the absence of disease were not in relation to the achieved protection when infected; quantitative resistance was both costly and inefficient compared to qualitative resistance, whereas qualitative resistance was efficient with small benefits and costs in host performance over the years. However, through time, both the costs and benefits of qualitative resistance became more pronounced. Moreover, the cost of qualitative and quantitative resistance compared to susceptible plants was in currency of leaf size which negatively correlates with inflorescence production. To date, the evidence of resistance costs has been scarce and mixed (Bergelson & Purrington 1996; Brown & Rant 2013), and our results show that while the effects of disease are age-dependent, the qualitative resistance becomes beneficial through time and the selection for resistance operates over multiple growing seasons. Empirical data of the costs of quantitative resistance have been lacking, but it has been postulated that due to its polygenic inheritance, net costs of quantitative resistance would be low (Brown & Rant 2013), whereas others suggest it to have large pleiotropic effects on plant development (Poland et al. 2009). Our results support the latter assumption; there was a cost in the absence of pathogen compared to qualitative resistant and susceptible plants in plant size (PC1) and compared to susceptible plants in leaf width (PC2). Moreover, the annual variation in the plant performance was large potentially masking the effect of infection. Our results also suggest that the cost of infection may vary depending on the year and therefore may be context-dependent. The lack of strong benefits and costs in plant performance for resistance polymorphism can be possibly explained by metapopulation dynamics and seed dormancy (Tellier & Brown 2007). In a dynamic metapopulation, the selection pressure exerted by pathogen epidemics may vary between years (Thompson 2005; Jousimo et al. 2014), with pollen and seed dispersal modifying the genetic landscape (Laine & Tellier 2008). As P. lanceolata seeds are capable for remaining viable in soil for several years (Lotz 1992), the germinating seeds may have been produced in periods that differ in abiotic and pathogen epidemiological conditions. Moreover, in a spatially heterogeneous environment, the spatial structure (Damgaard 1999) and genotype × environment interaction may allow maintenance of variation in populations (Thompson 2005). Alternatively, trade-offs between defence-related traits have been suggested to maintain resistance polymorphism. Trade-offs between constitutive and induced resistance traits (Kempel et al. 2011; Moreira et al. 2014), different resistance pathways (Navarro et al. 2008), and between resistances to different enemies (McGrann et al. 2014) are shown to shape resistance evolution. Our goal was to search for trade-offs between qualitative and quantitative resistance. We did not detect trade-offs between the resistance strategies; instead, there was a positive correlation between qualitative and quantitative resistance mechanisms as well as between the components of the quantitative resistance traits. It is possible that the potential trade-offs between resistance traits are condition-dependent, mediated by, for example local adaptation (Susi & Laine 2013) or temperature (Luong & Polak 2007), although several authors have also suggested that quantitative resistance and qualitative resistance are conditioned by the same genetic mechanisms (Poland et al. 2009).
Conclusions
Based on the results on this study, we now know more about the endurance of laboratory measured resistance reactions and their epidemiological consequences in the field. Careful consideration over years and seasons is needed when applying quantitative resistance in disease management. This is one of the very first reports exploring the eco-evolutionary foundations of perennial plant resistance in the wild. While variation in resistance has been recognized by a large body of studies, long-term studies measuring the contribution of the plant defences leading to observed pathogen loads in the host have been lacking. By showing that allocation costs of resistance and infection can change dynamically over the years, our study contributes to understanding of the fundamentals of host–pathogen co-evolution. Our results give new insights into how polymorphism in resistance can be maintained through costs and shifts in resource allocation between vegetative growth and reproduction under infection.
Acknowledgements
We want to acknowledge the Lammi Biological Station for hosting the common garden populations and Krista Raveala for technical assistance. Anu Hakkarainen, Saara Hartikainen, Pauliina Hyttinen and Aapo Salmela provided valuable help in the common garden study. This work was supported by funding from the Academy of Finland (Grant Nos 250444, 136393 and 133499) and European Research Council (PATHEVOL; 281517) to A.L.L.
Data accessibility
Data available from the Dryad Digital Repository: http://doi.org/10.5061/dryad.3p35d
