Genetic basis of pathogen community structure for foundation tree species in a common garden and in the wild

Authors


Correspondence author. E-mail: busby@post.harvard.edu

Summary

  1. Genetic variation within and among foundation plant species is known to affect arthropod, plant and soil microbial communities. We hypothesized that the same would be expected for pathogen communities, which have typically been studied only as individual pathogen species.

  2. In a common garden in Utah, USA, we first tested how genetic differences within and among Populus angustifolia, P. fremontii and their interspecific hybrid P. × hinckleyana affect a fungal leaf pathogen community. Next, we tested how Populus genetic differences at the level of species and hybrids affect fungal leaf pathogen communities in the wild, specifically in a natural Populus hybridization zone (13 river km) and throughout the larger Weber River watershed (150 river km).

  3. In the common garden, genetic variation both within and among Populus species and hybrids significantly affected the structure (i.e. species abundances and composition) of pathogen communities. In the wild, genetic variation among Populus species and hybrids also significantly affected pathogen communities, though not as strongly as was found in the common garden environment. Stand-level density of the susceptible Populus species most strongly affected the structure of pathogen communities in the hybrid zone.

  4. Synthesis. Plant species and genotypic variation can affect the local and geographic distribution of pathogen communities in a similar fashion as other diverse organisms (e.g. arthropods, plants, soil microbes), both within a relatively controlled common garden environment and in the wild.

Introduction

A central aim of community genetics is to determine how genetic variation in foundation species – those that create locally stable conditions for many associated species (Dayton 1972) – affects the structure (i.e. species abundances and composition) of associated communities. Major advances in this field have been made by common garden studies showing that plant genetic variation can influence diverse foliar arthropod communities (Fritz 1988; Dungey et al. 2000; Van Zandt & Agrawal 2004; Wimp et al. 2005; Barbour et al. 2009; Keith, Bailey & Whitham 2010). However, we know much less about how plant genetic variation affects pathogen communities, in common garden environments or in the wild. Answers to these questions are important as they address the potential to merge the findings of different communities from pathogens to arthropods within the context of the genetic structure of their host plants (Whitham et al. 2012).

Most plants support multiple pathogen species (Farr & Rossman 2012) of large, small or potentially synergistic effects on their plant hosts (Strauss, Sahli & Conner 2005; Morris et al. 2007). The plant immune system has evolved different mechanisms to resist pathogens, but genetic resistance has primarily been studied in the context of individual host species (or genotypes) and individual pathogen species (or genotypes) (Jones & Dangl 2006). Investigating the interactions occurring between different host genotypes and their pathogen communities can tell us a lot about the ecology and evolution of these complex interactions (Rigaud, Perrot-Minnot & Brown 2010). Such knowledge will become increasingly important to unravel as pathogen impacts on the host plant can define much larger communities (P. Busby, unpubl. data). Furthermore, anthropogenic effects, such as climate change, are predicted to affect pathogen severity and the probability of their occurrence (Harvell et al. 2002).

Previous research evaluating plant genetic resistance to individual pathogen species provides some insight into how genetic variation in foundation plant species may influence entire pathogen communities. Non-host resistance completely prevents a particular pathogen from infecting a particular host species and thereby drives differences in pathogen communities among species (Wapshere 1974; Bernays & Graham 1988; Heath 2000; Gilbert & Webb 2007). Two different types of host resistance, major resistance genes and quantitative genetic resistance, are expressed within genotypes of a susceptible species. Major genes prevent infection by specific pathogens (Flor 1955) and potentially alter the composition (i.e. the presence or absence of particular pathogens) of the pathogen community of an individual plant. In contrast, quantitative genetic resistance involves many genes acting together to reduce the pathogen latent period, infection efficiency or spore production (Geiger & Heun 1989). Because quantitative genetic resistance is neither complete nor pathogen-specific, it may be more likely to affect the abundance of pathogens within the community.

The genus Populus is a model woody plant in genomics (Tuskan et al. 2006), a pre-eminent crop for biomass energy production (Ellis 2012), and includes foundation species that affect many other organisms in riparian ecosystems (Whitham et al. 2006). The genetic basis of resistance to many pathogens of Populus has been demonstrated; major resistance genes are important for artificial hybrids (Newcombe 1996, 2005; Duplessis et al. 2009). A critical extension of this research is to elucidate the extent to which Populus genetic resistance affects pathogen communities. Here, we investigate how genetic variation within and among Populus species and their naturally occurring F1 hybrids influences the structure of fungal leaf pathogen communities in a common garden where environmental variation is minimized, and in natural forests in the Weber River drainage system (Utah, USA).

Our first objective was to test the hypothesis that genetic differences within and among Populus species and their F1 hybrids are important for structuring pathogen communities in a common garden. Plant hybridization could enable pathogens to overcome non-host resistance by creating genetic intermediates, or stepping-stones, for host shifting from one plant species to another (Floate & Whitham 1993). Alternatively, hybrids may lose the pathogens of their parental species due to hybrid vigour (Fritz, Moulia & Newcombe 1999). Because hybrid plants represent a major evolutionary pathway in plant speciation, it is important to evaluate how communities respond to genetic variation generated by hybridization. In addition, testing whether genotypic variation contributes to pathogen community structure under controlled conditions (e.g. in a common garden) is a critical first step toward understanding how genetic variation within foundation plant species affects the ecology and evolution of interacting pathogen communities. Our second objective was to test the hypothesis that species and hybrid-level genetic effects on pathogen communities are operative in the wild. We examined pathogen distributions within a narrow plant hybridization zone along the Weber River (13 river km) and throughout the larger Weber River watershed (150 river km). We also evaluated whether particular stand-level biotic and abiotic factors (e.g. host density, elevation, stand openness) contribute to pathogen community structure in the wild.

Materials and methods

Study system

Our study was conducted near Ogden, Utah, where Populus angustifolia, P. fremontii and their naturally occurring interspecific hybrid, P. × hinckleyana (Eckenwalder 1984), are found in mixed stands along a 13 km stretch of the Weber River (hereafter hybrid zone). These stands are separated by natural or human-made boundaries. The overstorey in these stands is dominated by Populus species and hybrids representing c. 70% of the individuals and 90% of the biomass (Adams et al. 2011). Populus angustifolia occurs in pure stands at elevations above the hybrid zone (> 1400 m), and P. fremontii occurs in pure stands below the hybrid zone (< 1300 m).

In 1991, a common garden was established within the hybrid zone. Replicated genotypes of P. angustifolia, P. fremontii and P. × hinckleyana were planted in the garden using cuttings taken from randomly selected trees growing in the hybrid zone and in nearby pure stands (within 100 km of the common garden). At the time of this study (September 2010), the common garden trees were sexually mature in a mostly closed gallery forest. Measures of growth in the garden show a strong genetic component to productivity and other physiological traits (Lojewski et al. 2009).

Pathogen community

We define the fungal leaf pathogen community by those species causing visible symptoms of disease. Busby, Aime & Newcombe (2012) extensively sampled this community in the Weber River Populus hybrid system over two field seasons and utilized both morphological and DNA sequence data to identify the following taxa: Drepanopeziza populi, Phyllactinia populi and Mycosphaerella spp. (orders Helotiales, Erysiphales and Capnodiales, respectively). Each pathogen is able to infect both Populus species and their hybrids, though P. angustifolia is the most susceptible host for all three taxa (Busby, Aime & Newcombe 2012).

In our pathogen community surveys for the present study, the three pathogen taxa were identifiable in the field without magnification. Drepanopeziza populi was easily identified by its characteristic dendritic lesions and white acervuli (Fig. 1a). In contrast, Mycosphaerella lesions were often angular, and the acervuli were black at the surface (Fig. 1a). P. populi mycelia and fruiting bodies were visible on the underside of leaves (Fig. 1b). At the onset of the current study, three species of Mycosphaerella occurring in the Populus study systems had not yet been distinguished: M. angustifoliorum, M. sp. 1 and M. sp. 2 (Busby, Aime & Newcombe 2012). These species remain indistinguishable in the field; thus, we were unable to distinguish between species of Mycosphaerella in pathogen surveys for the present study.

Figure 1.

Fungal leaf pathogens of the hybrid Populus study system. (a) Mycosphaerella lesion with black pycnidia (right arrow) and Drepanopeziza populi dendritic lesions with white acervuli (left arrow), and (b) magnified Phyllactinia populi mycelia and cleistothecia (scale bar represents 500 um).

Mycosphaerella and D. populi are necrotrophic (possibly hemibiotrophic) pathogens: they kill host tissue and feed on the remains. These pathogens are known to cause reduced growth, premature leaf abscission, and shoot and branch death (Ostry 1987). Damage over multiple years can result in tree death (Ostry & McNabb 1986). Phyllactinia populi is a biotrophic pathogen: it feeds on live host tissue. Phyllactinia populi may be less likely to affect Populus fitness because it occurs late in the growing season (Sinclair & Lyon 2005).

There have been no formal studies of Populus genetic resistance to any of the specific pathogens evaluated in our study. However, major resistance genes are known to be particularly effective against biotrophic pathogens like the powdery mildew fungus, P. populi, whereas quantitative resistance is often more effective against necrotrophic pathogens like Mycosphaerella and D. populi (Oliver & Ipcho 2004). A previous study found evidence for quantitative resistance to Mycosphaerella in Populus trichocarpa × P. deltoides (Newcombe & Bradshaw 1996), and we found evidence of quantitative, genetic resistance to D. populi in a gene expression study (P. Busby, unpubl. data).

Pathogen communities in a common garden

In September 2010, we surveyed fungal pathogens on leaves collected from 10 P. angustifolia genotypes, 6 P. × hinckleyana genotypes and 4 P. fremontii genotypes growing in the common garden. For each genotype, we sampled 3–10 replicate clones. The uneven sample size of both clones and genotypes was a constraint of the original garden design.

Our pathogen community surveys consisted of scoring the severity of symptoms associated with each pathogen present on leaves of each tree sampled. Analysis of ecological community structure typically utilizes presence/absence or abundance data on individual species within communities (McCune & Grace 2002). In our pathogen community analyses, we use symptom severities associated with individual pathogens as proxies for their relative abundance in the community. Genetic resistance that does not prevent infection but does limit colonization by a pathogen (i.e. quantitative resistance) should be inversely correlated with pathogen symptom severity (Geiger & Heun 1989). An inoculation experiment with at least one pathogen in this community (i.e. D. populi) confirmed that observed severity on common garden trees is correlated with resistance (P. Busby, unpubl. data).

For each tree, we quantified symptom severity for each necrotrophic pathogen present by visually estimating leaf area damaged by that pathogen for 24 leaves per tree standardized by leaf age (leaf plastochron indexes 3, 4, 5 and 6), collected from six haphazardly selected terminal shoots in the lower canopy. For all leaves, the severity of each pathogen was scored on a scale from 0 to 5 reflecting the percentage of leaf area damaged: 0 = no damage, 1 = 1–6%, 2 = 7–12%, 3 = 13–25%, 4 = 26–50% or 5 = > 50%. Damage scores were then used to calculate a single weighted damage score (Dirzo & Domínguez 1995). The damage score was calculated as: ∑ ni (Ci)/N, where ni is the number of leaves in the ith category of damage, Ci is the midpoint of each category (C0 = 0, C1 = 3.5, C2 = 9, C3 = 18.5, C4 = 37.5, C5 = 75%), and N is the total number of leaves sampled. Because the biotrophic pathogen P. populi did not cause leaf necrosis, which is easily quantified as the percentage of leaf area affected, this pathogen was scored as present or absent at the shoot level.

In our surveys, we only recorded per cent area damaged by Mycosphaerella or D. populi if diagnostic fruiting bodies were present; otherwise, damage was recorded as caused by an unknown species. However, in most cases, such lesions resembled Mycosphaerella or D. populi with immature fruiting bodies. Therefore, our estimates of Mycosphaerella and D. populi severity could be conservative and damage caused by truly unknown species rare. Unknown pathogen damage could have been caused by Fusicladium romellianum, which was found infrequently but did not produce diagnostic characteristics during our summer survey, or other unidentified pathogens. Taken together, unknown damage accounted for an average of 25% of total recorded leaf area damaged.

Pathogen communities in the wild

In September 2010, using the method previously described (Dirzo & Domínguez 1995), pathogen communities were sampled on Populus species and their F1 hybrids in 19 natural stands along the Weber River spanning lower elevation pure P. fremontii stands (= 4), through the hybrid zone (= 7), and into upper elevation pure P. angustifolia stands (= 8). This represents a sampling effort of c. 150 km along the Weber River and 1000 m in elevation (with distance from the P. fremontii zone and elevation being positively correlated). In each natural stand, we haphazardly selected 10 trees of each Populus taxon present (i.e. 30 trees per hybrid zone stand, 10 trees per pure zone stand). We distinguished Populus species and their F1 hybrids in the field by leaf morphology. However, leaf morphology does not definitively distinguish P. angustifolia from its advanced backcross hybrids. P. angustifolia samples collected in the hybrid zone may include advanced backcross hybrids that are genetically very similar to P. angustifolia. Backcross hybrids with P. fremontii are rare (Martinsen et al. 2001).

In addition to evaluating host genetic effects on pathogen community structure at the tree-level in the wild, we also evaluated stand-level environmental factors potentially affecting pathogens. Among stands in the narrow hybrid zone, environmental variation is minimal. Since stands are similar in size (1.8–2.6 ha2), basal area (29.03 m2 ha−1 ± 4.07 standard error), climate and soils (Schweitzer et al. 2011), these factors were not included as covariates in models. One potentially important covariate that we included in hybrid zone analyses was canopy openness. By influencing relative humidity and leaf wetness, canopy openness should in part determine the conductivity for foliar pathogen infection (Agrios 2005).

We measured canopy openness in each hybrid zone stand using hemispherical (aka fisheye) photography (Frazer et al. 2001). Photographs were taken at dawn using a Canon EOS 1 Mark II Digital camera with a Sigma EX DG Fisheye 8 mm 1 : 3.5 lens (Sigma Corporation of America, NY, USA). The camera was mounted on a 1 m tripod to ensure that all photographs were taken at a fixed height. In each stand, we captured photographs from 12 haphazardly selected locations. We then used Gap Light Analyzer software to convert the images into black-and-white and calculate the percentage of white (i.e. open canopy) in each image (Jarčuška 2008). The 12 values were averaged for each stand.

While overall Populus density (i.e. both parental species and their hybrids) did not differ among sites (Schweitzer et al. 2011), a second covariate included in hybrid zone analyses was the density of the susceptible host, P. angustifolia, which did differ among sites. A greater density of susceptible host species should increase the probability of successful infection by airborne inoculum (Burdon & Chilvers 1982). Hybrid zone stands vary in the density of each Populus species and hybrids: stands in closer proximity to the pure P. angustifolia zone are more heavily dominated by P. angustifolia, and vice versa for stands in closer proximity to the P. fremontii zone, with P. × hinckleyana reaching its greatest abundance in stands located near the middle of the hybrid zone (Wimp et al. 2004; Schweitzer et al. 2011). This naturally occurring variation in the density of host species, but not overall host tree density, allowed us to evaluate how the density of the most susceptible Populus host species affects pathogen community structure. We used the proportion of stand-level P. angustifolia alleles as a measure of the local density of the susceptible host within hybrid zone stands. This was previously calculated by genotyping 20 trees per stand (haphazardly selected and at least 30 m from another sampled tree) at 48 amplified fragment length polymorphism (AFLP) loci (Wimp et al. 2004).

In contrast to the hybrid zone, environmental conditions must vary considerably among stands found throughout the watershed. We used elevation as a covariate proxy for abiotic environmental variation along this 1000 m gradient. Stands throughout the watershed also vary biotically, in that some are composed entirely of P. fremontii hosts or P. angustifolia hosts, whereas others contain a mixture of both host species and their interspecific hybrid. So as in our hybrid zone analysis, we include the proportion of P. angustifolia alleles in models. For stands in the pure P. fremontii zone, this value was zero; for stands in the pure P. angustifolia zone, this value was one.

Statistical analyses

All statistical analyses were conducted in r 2.8.1 (R Development Core Team 2008). We used permutational multivariate analysis of variance using distance matrices (permanova, vegan package) (McArdle 2001) to test how genetic differences within and among Populus species and hybrids affect pathogen community structure in the common garden. We also used permanova to test how genetic differences among Populus species and hybrids affect pathogen community structure at two different spatial scales in the wild: the hybrid zone and the watershed. Our community matrices consisted of columns of pathogen severities, one for each species, excluding unknown pathogen species. We transformed severity data to ensure that all pathogen species contributed equally to community analyses. We arcsin-transformed proportional (0–1) P. populi severity data (Zar 1996) and fourth-root transformed Mycosphaerella and D. populi per cent damage data to eliminate high-scoring variables while preserving the weights (Clarke 1993).

For the common garden permanova analysis, genotypes were nested with species and hybrids, and both were fixed effects. For the hybrid zone and watershed analyses, we included stand-level biotic and abiotic environmental covariates in our models. For the hybrid zone analysis, we included the density of P. angustifolia (proportion of P. angustifolia alleles), the interaction between Populus species and hybrids and P. angustifolia density, and canopy openness. The interaction term tests whether pathogen communities on different hosts respond differently to P. angustifolia density. Elevation was not included in the model because it did not vary strongly among stands in the small (13km) hybrid zone. For the watershed analysis, we included elevation (m), the density of P. angustifolia, and the interaction between Populus species and hybrids and P. angustifolia density. To test the significance of each fixed effect, we used F-tests based on sequential sums of squares from permutations of the raw data (McArdle 2001). To visualize community results, we used two-dimensional representations of pathogen communities where the x- and y-coordinates are based on non-metric multidimensional scaling (NMDS) analysis as in Keith, Bailey & Whitham (2010). We also calculated principal component scores using the same community matrix to visualize the relationship between the density of P. angustifolia and pathogen community structure.

In addition to our analysis of pathogen communities, we also evaluated pathogen species individually. We used analysis of covariance to test how genetic differences within and among Populus species and hybrids affect symptom severities for individual pathogens in the common garden, and how genetic differences among Populus species and hybrids affect symptom severities for individual pathogens in the wild. Genotype was nested within host species in the common garden models. For the hybrid zone and watershed models, we used the same set of factors as in our community analysis. The significance of each factor was tested using F-tests.

Finally, to evaluate and compare the strength of plant genotype effects for each cottonwood species and hybrids in the common garden, we calculated the variation in pathogen community structure explained by host genotype for each host species separately. This measure is called the broad-sense community heritability (H2 = VarGenotype/[VarGenotype + VarError]) (Shuster et al. 2006). To calculate VarGenotype and VarError, we used permanova. The significance of genotype was tested using F-tests based on sequential sums of squares from permutations of the raw data. We also estimated individual heritability values for each pathogen species and each host species using analysis of variance to calculate VarGenotype and VarError.

Results

Pathogen communities in the common garden

We found support for our genetics hypotheses at both inter- and intraspecific levels. Genetic differences among Populus species and their F1 hybrid explained 27% (< 0.001) of the variation in pathogen community structure (Table 1). We also found that genetic differences within Populus species and their hybrid affected pathogen communities: host genotype explained an additional 33% (P < 0.001) of the variance in community structure (Table 1). The observed differences in pathogen communities found among species, and genotypes within species, are visually depicted in the NMDS plot (Fig. 2). Vector analysis of this two-dimensional solution revealed that symptom severity of P. populi is strongly associated with axis 1 (i.e. a nearly horizontal vector), whereas symptom severities of Mycosphaerella and D. populi are strongly associated with axis 2 (i.e. nearly vertical vectors). Broad-sense heritability analyses revealed that the magnitude of plant genotypic effects on the community ranged from H2 = 0.32 for P. angustifolia, to 0.46 for P. × hinckleyana, and to 0.93 for P. fremontii (Table 2). We also found significant broad-sense heritabilities for individual pathogen species on plant species and hybrids in six of the nine analyses (Table 2).

Table 1. Model results for pathogen community structure in the hybrid zone, throughout the watershed, and in the common garden
 Community Drepanopeziza populi Mycosphaerella Phyllactinia populi
d.f. F R 2 P d.f. F R 2 P d.f. F R 2 P d.f. F R 2 P
Common garden
Host species or hybrid2290.27< 0.0012140.12< 0.0012500.38< 0.0012120.14< 0.001
Host genotype (intraspecific)174.20.33< 0.001176.40.49< 0.001174.630< 0.001173.60.36< 0.001
Hybrid zone
Host species or hybrid2130.087< 0.00122.50.0180.08042750.37< 0.00121.30.0110.27
Populus angustifolia density1380.13< 0.0011400.14< 0.001180.020.005211.60.00680.211
Host × P. ang density230.02040.01723.80.0270.0252100.051< 0.00122.70.0240.069
Stand openness10.610.00210.5910.1900.6610.980.00260.3210.320.00440.32
Watershed
Host species or hybrid2220.103< 0.00123.40.0180.0342730.26< 0.00126.10.0330.0025
P. angustifolia density1200.046< 0.0011190.05< 0.00117.60.013< 0.00112.90.00780.089
Host × P. ang density29.90.045< 0.00124.60.0240.0112110.11< 0.00126.60.0360.0015
Elevation18.50.0190.3711.300.261310.019< 0.00111.50.0040.23
Table 2. Common garden results showing the proportion of variation in the structure of pathogen communities and in individual pathogen attack severities, explained by genotype for different Populus species and their hybrid (broad-sense heritability, H 2)
Host speciesCommunity Drepanopeziza populi Mycospharella Phyllactinia populi
d.f. F H 2 P d.f. F H 2 P d.f. F H 2 P d.f. F H 2 P
P. fremontii 3540.93< 0.00133.20.440.05730.750.150.5434300.99< 0.001
P. × hinckleyana 54.10.460.00650.880.150.5157.40.61< 0.00154.30.470.0062
P. angustifolia 92.60.320.00297.40.57< 0.00194.90.47< 0.00191.20.170.35
Figure 2.

Two-dimensional representation of pathogen communities found on 10 Populus angustifolia genotypes, 4 P. fremontii genotypes and 6 P. × hinckleyana genotypes in the common garden. Coordinates are based on global, non-metric multidimensional scaling analysis. Error lines are standard error of the mean. Vector analysis revealed that symptom severity of Phyllactinia populi is strongly associated with axis 1 (i.e. a nearly horizontal vector), whereas symptom severities of Mycosphaerella and Drepanopeziza populi are strongly associated with axis 2 (i.e. nearly vertical vectors).

Within the pathogen community, we found that symptom severities of individual pathogens differed within and among Populus species and their hybrids (Table 1 and Fig. 3). In agreement with our previous work (Busby, Aime & Newcombe 2012), symptom severities of all pathogens were greatest on P. angustifolia. Given that D. populi, Mycosphaerella and P. populi were the only pathogens found, and that they were present on almost all plant genotypes, our results show that pathogen communities varied primarily in the abundance of individual pathogens rather than in species composition.

Figure 3.

Pathogen severity in the common garden for Drepanopeziza populi (a, d) Mycosphaerella (b, e) and Phyllactinia populi (c, f) for Populus angustifolia, P. fremontii and P. × hinckleyana. Left panels show differences between species (different letters indicate significant differences, P < 0.01); right panels show differences between genotypes within species (= 4 for P. fremontii,= 6 for P. × hinckleyana and = 10 for P. angustifolia). Error lines are standard error of the mean.

Pathogen communities in the wild

Genetic differences among Populus species and their interspecific hybrids also significantly affected pathogen communities in the hybrid zone and the larger watershed, explaining 9% and 10% of the variation in pathogen community structure, respectively (Table 1). Community differences were most strongly influenced by D. populi, which had the highest loading value for the first principal component (Fig. 4a). The overall pattern of weaker plant genetic effects in the wild than in the common garden was also found for individual pathogens (Table 1). The extreme case was P. populi: genetic differences among Populus species and their hybrids had no significant effect on symptom severity in the hybrid zone.

Figure 4.

Relationship between the density of stand-level Populus angustifolia alleles and pathogen community structure (measured by the first principal component) for P. angustifolia, P. fremontii and P. × hinckleyana in hybrid zone stands (Panel a). Drepanopeziza populi was the species with the highest loading value for PC1. Panels b, c and d show the relationship between the density of stand-level P. angustifolia alleles and pathogen severity for each pathogen: D. populi, Mycosphaerella and Phyllactinia populi, respectively. Error lines are standard error of the mean.

In the hybrid zone, P. angustifolia density was the most influential factor for pathogen community structure, explaining 13% of the variation in pathogen communities (Table 1). Pathogen communities in stands dominated by the susceptible species (P. angustifolia) varied significantly among their Populus species and hybrid hosts (Fig. 4a). In contrast, pathogen communities in stands dominated by the resistant species (P. fremontii) did not differ among host species (Fig. 4a). The relationship between P. angustifolia density and pathogen community structure appeared to be driven by particular pathogens on different Populus host species responding differentially to P. angustifolia density (i.e. significant interaction terms) (Table 1). For example, symptom severities for D. populi and Mycosphaerella on P. angustifolia were positively correlated with density, whereas symptom severity for D. populi only on P. fremontii and P. × hinkleyana was positively correlated with density (Table 1, Fig. 4). Symptom severity for P. populi was not influenced by the density of P. angustifolia for any host species (Table 1, Fig. 4d). Stand openness was not significant for the community or individual pathogen symptom severities (Table 1).

At the watershed scale, genetic differences among Populus species and their interspecific hybrids most strongly affected pathogen community structure, explaining 10% of the variation (Table 1). The density of P. angustifolia and the interaction between Populus species and hybrids and P. angustifolia density were also significant, each explaining 5% of the variation (Table 1). Elevation did not significantly affect the pathogen community, though it was associated with symptom severity for Mycosphaerella (Table 1).

Overall, in agreement with our common garden results, pathogen communities varied mostly in the abundance of pathogens rather than in species composition and symptom severities of D. populi and Mycosphaerella were greatest on P. angustifolia (Fig. 5). Unlike results from the common garden, the biotrophic pathogen, P. populi, was found in low abundance on all hosts and was completely absent from P. fremontii in pure stands (Fig. 5).

Figure 5.

Pathogenic severity for Drepanopeziza populi, Mycosphaerella and Phyllactinia populi in natural stands along the Weber River. Severity for D. populi and Mycosphaerella is percent leaf area infected; severity for P. populi is the proportion of shoots infected. Grey bars mark the boundaries of the hybrid zone. Error lines are standard error of the mean.

Discussion

Genetic basis of pathogen community structure

Both in a common garden and in the wild, we found evidence that pathogen communities differ among Populus species and their F1 hybrids. While we speculated that non-host resistance could drive these differences, results of this study, and our previous work on species of Mycosphaerella in this hybrid system (Busby, Aime & Newcombe 2012), reveal a high degree of pathogen sharing. All pathogens infected both Populus species and their hybrids. Therefore, non-host resistance was apparently not a major factor shaping pathogen communities. Plant genetic variation resulting from hybridization may have enabled pathogens to overcome non-host resistance and shift from a native to a non-native host. Floate & Whitham (1993) have argued that genetic intermediates in this Populus hybrid system facilitate host shifting for insect herbivores and pathogens. Our results appear to be consistent with this mechanism, though we did not directly test the ‘hybrid-bridge’ hypothesis.

In the common garden, we also found that intraspecific variation within Populus species and their hybrids contributes to pathogen community structure. The differences in pathogen communities that we observed among genotypes were mostly in the severities of damage inflicted by constituent pathogens rather than in composition. Because quantitative genetic resistance is thought to affect species abundances rather than composition (Geiger & Heun 1989), we speculate that quantitative genetic resistance is involved in shaping these pathogen communities. This is consistent with other studies showing that quantitative resistance strongly affects necrotrophic pathogens, like D. populi and Mycosphaerella (e.g. Newcombe & Bradshaw 1996). In fact, symptom severities of D. populi and Mycosphaerella were positively correlated in their co-occurrence on the same tree genotypes (R2 = 0.41, < 0.001). Significant broad-sense heritability of this fungal leaf pathogen community suggests the possibility that plants evolved generalized resistance to these co-occurring pathogens (Leimu & Koricheva 2006). A second, non-mutually exclusive hypothesis is that D. populi or Mycosphaerella is directly or indirectly facilitating the other.

While we speculate that quantitative resistance is important for structuring pathogen communities in this Populus hybrid system, we cannot rule out a role for major resistance genes. The absence of P. populi on particular genotypes within P. fremontii in the common garden suggests the presence of a major gene for resistance to this pathogen (Newcombe 1996). Many trees in the wild were also uninfected by P. populi. However, we suspect that hot, dry conditions were not conducive to powdery mildew infection in the wild. In contrast, we found high levels of infection in the closed canopy common garden environment.

It is also important to emphasize that individual trees can exhibit within plant variation in traits that are known to affect pathogens and other organisms. For example, fungi associated with antipathogen defence can be locally distributed within trees (Arnold et al. 2003; Raghavendra & Newcombe 2013). Furthermore, using many of the same genotypes as those studied herein, Holeski et al. (2012) demonstrated significant within plant variation in phytochemistry (juvenile vs. mature zones) and induced defence following herbivore damage. Because within plant variation can vary among genotypes and be heritable, such predictable within plant variation is considered as part of a plant's ‘multivariate defence phenotype’ (Holeski et al. 2012). Such variation probably contributes to patterns of community specificity in which individual species and whole communities of organisms (see next section) are generally associated with individual plant genotypes or classes of genotypes exhibiting similar traits (review by Whitham et al. 2012). However, given that the emphasis of our study is on evolutionary ecology, the individual tree genotype is the unit of selection that we focused upon. We minimized within plant variation by sampling only in lower, juvenile portions of the canopy. A permanova test revealed that branch effects explained a significant, though small percentage of the variation (1–2%) in pathogen community structure for different Populus species and hybrids (data not shown). Thus, overall, pathogen communities sampled in different parts of the lower, juvenile canopy zone were similar within trees.

Contrasts of plant pathogen community with communities of other organisms

Studies in the same Populus hybrid system have also found that Populus species and their hybrids support distinct arthropod communities (Wimp et al. 2005), understorey plant communities (Adams et al. 2011), and unidentified communities of soil microbes (Schweitzer et al. 2008) and endophytes (Bailey et al. 2005). For arthropods, studies in Populus and other plant systems have further shown that intraspecific plant genotypic variation strongly affects community structure (Johnson & Agrawal 2005; Crutsinger, Cadotte & Sanders 2009; Keith, Bailey & Whitham 2010). Plants that are more genetically similar are also more similar in their phytochemistry, which in turn are more similar in their arthropod communities (Bangert et al. 2005; Barbour et al. 2009; Zytynska et al. 2011, 2012; Ferrier et al. 2012).

Although constitutive, preformed defence is effective in limiting the number of pathogens attacking a plant, this form of defence is breached by a small number of active pathogen species specific to that host plant. For these specific pathogens, constitutive defence is unimportant when compared to the induced defence responses that have been the subject of much research (Heath 2000; Veronese et al. 2003). Accordingly, we found no correlation between previously reported levels of condensed tannins and phenolic glycosides for the plant genotypes evaluated and damage caused by individual pathogens or pathogen community structure in the common garden (data not shown).

The magnitude of plant genotypic effects on pathogens that we observed (H2 = 0.32–0.93) is broadly similar to that reported for arthropods [e.g. math formula = 0.65 (Keith, Bailey & Whitham 2010), math formula = 0.7 (Crutsinger, Cadotte & Sanders 2009), math formula = 0.41 (Johnson & Agrawal 2005)]. However, an explicit comparison between our results and previous work in the same Populus study system reveals that genotypic variation in P. angustifolia explained twice as much variation in arthropod communities (65%, Keith, Bailey & Whitham 2010) as pathogen communities (32%). The pattern of stronger plant genotype effects on arthropod than fungal communities is consistent with the results of a meta-analysis (Bailey et al. 2009). This may appear counterintuitive, since fungi interact more intimately with their hosts than arthropods, and thus may be expected to display stronger genetic effects than arthropods. However, fungal dependence on above- and below-ground moisture may be one reason why studies have not found stronger plant genetic effects on fungal communities. In the case of fungal leaf pathogens, 12–72 h of continual leaf wetness is required for infection (Agrios 2005), indicating that environmental factors will likely play a strong role in shaping pathogen communities. Alternatively, variation in biotic factors within the common garden (e.g. fungal endophytes conferring disease resistance) could be strongly affecting pathogen communities, and these are known in Populus (Raghavendra & Newcombe 2013). Lastly, the lack of an equivalent systematic treatment for microbes relative to arthropods could obscure patterns that might otherwise be more apparent. Future research explicitly evaluating and comparing genetic mechanisms will be poised to address the broader implications of plant genetic variation for the ecology and evolution of different communities.

Host density effects

Increasing host density increases the probability of successful infection by airborne inoculum and thus pathogen severity (Burdon & Chilvers 1982). Experimentally constructed communities dominated by susceptible host species are thus characterized by greater disease severity than communities dominated by resistant host species (Mitchell, Tilman & Groth 2002). We found evidence that the density of the susceptible host species can also affect the structure of pathogen communities in the wild. In the hybrid zone, where overall Populus density was similar among sites, pathogen communities responded strongly to the density of the most susceptible host, P. angustifolia. This was not simply a matter of symptom severities for all pathogens in the community increasing. Only one of the three pathogens in the community, D. populi, responded uniformly positively (i.e. greater severity) on all host species to the density of the susceptible host. Symptom severity for Mycosphaerella increased with P. angustifolia density for P. angustifolia only. And the powdery mildew fungus P. populi did not respond at all to the density of P. angustifolia. The overall outcome was similar pathogen communities on different Populus host species in stands dominated by the resistant species (P. fremontii), but divergent pathogen communities on different Populus host species in stands dominated by the susceptible species (P. angustifolia). This result highlights how individual pathogens in a community can respond differentially not only to genetic differences between hosts or to the environment, but also to the density of the most susceptible host.

We also found evidence that the density of the most susceptible host species can affect pathogen community richness at the watershed scale. Pure stands of the resistant host, P. fremontii, were characterized by lower pathogen species richness (two pathogen species) than hybrid zone stands that included the susceptible host and hybrids (three pathogen species). Evaluating all three species of Mycosphaerella, our previous study found only two pathogen species on P. fremontii in the pure zone (D. populi and Mycosphaerella sp. 1), while four species were found on P. fremontii in the hybrid zone (D. populi and M. angustifoliorum, M. sp. 1, P. populi) (Busby, Aime & Newcombe 2012). Thus, three of the five total pathogen species in this Populus system were missing from pure P. fremontii stands, suggesting that stands composed of the resistant host species are not able to support populations of these pathogens.

Conclusions and genetic scaling

Studies documenting the importance of genetic variation in foundation species for associated communities are mounting (Hersch-Green, Turley & Johnson 2011). A review of 75 different communities associated with 28 focal plant genera (15 plant families world-wide) found intraspecific plant genetic effects on communities of endophytes, mycorrhizal fungi, pathogens, epiphytic and terrestrial plants, soil microbes, and terrestrial invertebrates (Whitham et al. 2012). While pathogen communities have not been a focus in this literature, a Eucalyptus globulus common garden study reported significant intraspecific genetic effects on foliar communities composed of insect herbivores and unidentified pathogens (Barbour et al. 2009). Our study, focusing exclusively on foliar pathogens, supports the importance of intraspecific genetic effects on this important community.

Some studies have argued that environmental heterogeneity swamps plant genetic effects on associated communities in the wild (Johnson & Agrawal 2005; Tack et al. 2010) and that findings of genetic effects in common gardens are suspect. However, our results from both a common garden and the wild argue that plant genetic effects can be important for pathogen communities even in natural landscapes. These findings are in agreement with studies of the epiphyte and invertebrate communities associated with bromeliads and the tree, Brosimum alicastrum, in a Neotropical rain forest (e.g. Zytynska et al. 2011). They also found significant genetic effects on these communities in the wild.

The next big step for community genetics will be to explicitly evaluate how genetic effects on associated communities scale in natural landscapes (Bangert et al. 2008). For example, how does genetic scaling compare and contrast with other scaling metrics that are commonly used by ecologists such as fractals, power laws and metabolism (Brown et al. 2002)? Although the genetic basis of these scalers is often assumed, the direct incorporation of genetics will provide a stronger evolutionary basis to these arguments. Future research utilizing reciprocal transplant study designs in combination with studies in the wild show great potential to address these issues.

Acknowledgements

We thank G Gilbert, M Saunders, J Bailey and two anonymous reviewers for comments on an earlier draft of this manuscript; J Lamit, M Lau and R Guevara for advice on analysis; and A Le and D Willett for assistance with fieldwork. We are grateful to the Ogden Nature Center for hosting this research and R Adams for helping with canopy openness data. This research was supported by the DOE Global Change Education Program, the Heinz Environmental Fellowship, the Stanford Field Studies Program and an NSF FIBR grant.

Ancillary