Volume 221, Issue 4 p. 2261-2272
Full paper
Free Access

Comparison of host susceptibilities to native and exotic pathogens provides evidence for pathogen-imposed selection in forest trees

Jules S. Freeman

Corresponding Author

Jules S. Freeman

School of Natural Sciences and ARC Training Centre for Forest Value, University of Tasmania, Private Bag 55, Hobart, 7001 Tas, Australia

Author for correspondence:

Jules S. Freeman

Tel: +61 437907375

Email: [email protected]

Search for more papers by this author
Matthew G. Hamilton

Matthew G. Hamilton

School of Natural Sciences and ARC Training Centre for Forest Value, University of Tasmania, Private Bag 55, Hobart, 7001 Tas, Australia

Search for more papers by this author
David J. Lee

David J. Lee

Forest Industries Research Centre, University of the Sunshine Coast, Locked Bag 4, Maroochydore DC, 4558 Qld, Australia

Search for more papers by this author
Geoff S. Pegg

Geoff S. Pegg

Department of Agriculture, Fisheries and Forestry, Ecosciences Precinct, GPO Box 267, Brisbane, 4001 Qld, Australia

Search for more papers by this author
Jeremy T. Brawner

Jeremy T. Brawner

Forest Industries Research Centre, University of the Sunshine Coast, Locked Bag 4, Maroochydore DC, 4558 Qld, Australia

Search for more papers by this author
Paul A. Tilyard

Paul A. Tilyard

School of Natural Sciences and ARC Training Centre for Forest Value, University of Tasmania, Private Bag 55, Hobart, 7001 Tas, Australia

Search for more papers by this author
Brad M. Potts

Brad M. Potts

School of Natural Sciences and ARC Training Centre for Forest Value, University of Tasmania, Private Bag 55, Hobart, 7001 Tas, Australia

Search for more papers by this author
First published: 22 October 2018
Citations: 17

Summary

  • The extent to which spatial structuring of host resistance in wild plant populations reflects direct pathogen-imposed selection is a subject of debate. To examine this issue, genetic susceptibilities to an exotic and a coevolved native fungal pathogen were compared using two Australian host tree species.
  • Damage to common host germplasm of Corymbia citriodora ssp. variegata (CCV) and Eucalyptus globulus, caused by recently introduced (Austropuccinia psidii) and native (Quambalaria pitereka and Teratosphaeria sp.) pathogens was evaluated in common-garden experiments.
  • There was significant additive genetic variation within host species for susceptibility to both the exotic and native pathogens. However, susceptibility to A. psidii was not genetically correlated with susceptibility to either native pathogen, providing support for pathogen-specific rather than general mechanisms of resistance.
  • Population differentiation (QST) for susceptibility to the native pathogens was greater than neutral expectations (molecular FST), arguing for divergent selection. Coupled with lower native, but not exotic, pathogen susceptibility in host populations from areas climatically more prone to fungal proliferation, these findings suggest that pathogen-imposed selection has contributed directly to a geographic mosaic of host resistance to native pathogens.

Introduction

The extent to which spatial structuring of host resistance in wild plant populations reflects direct enemy-imposed selection is a subject of ongoing debate (Parker, 1991; Thompson & Burdon, 1992; Laine et al., 2011; O'Reilly-Wapstra et al., 2014; Carmona et al., 2015; Toor & Best, 2016). Plants constantly interact with a multitude of animals, bacteria, viruses and fungi (Walters, 2011). These interactions may have a genetic basis, occur at many levels, from individual plants to entire species, and vary through time and space (Whitham et al., 2012; Burdon & Thrall, 2014). It is argued that differences in such interactions throughout the native range of a species will produce complex ‘mosaics of coevolution’, resulting in geographically structured differences in coevolved traits between host and pathogen populations (Thompson, 1999). While there are well-cited examples of coevolution in the wild (Soubeyrand et al., 2009) and agricultural systems (Flor, 1971), providing evidence that specific biotic interactions have shaped the evolutionary trajectory of even one of a species pair is challenging (Smith et al., 2011; Desprez-Loustau et al., 2016; Stenlid & Oliva, 2016), particularly when this involves long-lived hosts such as forest trees.

Consistent with coevolutionary theory, most studies assessing host resistance in the wild report spatial variation (reviewed by Salvaudon et al., 2008; Laine et al., 2011). However, such genetic differentiation could also occur through nonadaptive processes, such as random drift arising from bottlenecks and founder effects (Thompson & Burdon, 1992). It could also arise indirectly, through a correlated response to selection by other biotic or abiotic factors (Close & McArthur, 2002; Leimu & Koricheva, 2006; O'Reilly-Wapstra et al., 2014). In the case of pathogens, for example, there are general resistance mechanisms (van der Hoorn & Kamoun, 2008; Cook et al., 2015) that would make it difficult to differentiate pairwise selection as a result of a specific pathogen from diffuse (multispecies) selection (Leimu & Koricheva, 2006). We apply a novel approach to help unravel these possibilities. Forest systems world-wide are increasingly being impacted by exotic pathogens (Stenlid & Oliva, 2016). By comparing the genetic architectures of host resistance to native (coevolved) and exotic pathogens, as recently outlined by Perry et al. (2016), we provide evidence that the evolutionary trajectories of a temperate (Eucalyptus globulus Labill.) and a subtropical (Corymbia citriodora ssp. variegata Bean and McDonald; hereafter CCV) eucalypt species have been shaped by pathogen-imposed selection.

Eucalypts belong to the Myrtaceae family and comprise the genera Eucalyptus, Corymbia and Angophora (Grattapaglia et al., 2012). They are foundation species which dominate most Australian forest ecosystems and many have become economically important plantation species (Doughty, 2000). In two host species, we quantified variation in susceptibility to both the recently introduced pathogen Austropuccinia psidii (G. Winter) Beenken (myrtle rust or guava rust; formerly Puccinia psidii Winter – Beenken, 2017) and the most significant native pathogens of the host species, Teratosphaeria spp. (Syd and P. Syd) Crous (Teratosphaeria leaf disease) in E. globulus and Quambalaria pitereka (J. Walker and Bertus) J.A. Simpson (Quambalaria shoot blight) in CCV. Host resistance to native pathogens was assessed following natural infection in field trials and, in the case of CCV, artificial inoculations.

We first establish that there is significant additive genetic variation and population divergence in host resistance within each pathosystem. We then hypothesize that, if specific pathogen-imposed selection has played a significant role in the evolution of population divergence in host susceptibility to native pathogens, then there should be no genetic correlation between susceptibilities to native and exotic pathogens; population differentiation in host susceptibility to the pathogen should exceed that expected through drift (i.e. QST > FST; Leinonen et al., 2013) for the native but not the exotic pathogen; and variation in home-site climate variables related to disease risk should be significantly correlated with host population susceptibility to the native (White et al., 2007; Perry et al., 2016) but not to the exotic pathogen.

Materials and Methods

Study systems

Hosts

Eucalyptus globulus is native to Tasmania and coastal regions of mainland southeastern Australia (Jones et al., 2013) and has become an important plantation species in temperate regions world-wide (Doughty, 2000). Within its native range in southeastern Australia, E. globulus has been variously classified as a species or subspecies of four closely related taxa with core distributions that are geographically and morphologically distinct (Jones et al., 2013). There is significant quantitative genetic variation across the geographic range of E. globulus in virtually all traits studied, which has been used to partition the native populations of E. globulus into 13 geographic races (Dutkowski & Potts, 1999).

Corymbia citriodora ssp. variegata is the most important taxa for hardwood plantations in subtropical Australia (Lee, 2007). It is one of a complex of four closely related spotted gum taxa (C. citriodora ssp. variegata, ssp. citriodora, C. henryi and Cmaculata) that occur as a latitudinal replacement series on the east coast of Australia (Shepherd et al., 2012). CCV has a wide natural distribution, spanning coastal and subcoastal regions in northern New South Wales and southeastern Queensland. Commercial plantations have more recently been derived from blight-tolerant provenances, such as the Woondum population that was a focus for this study (Johnson et al., 2009; Lee et al., 2010).

Pathogens

Exotic pathogen

Austropuccinia psidii (Phylum Basidiomycota, Order Pucciniales, family Sphaerophragmiaceae) is a pathogen of global significance. Native to South America, it is now spreading rapidly world-wide, with multiple biotypes identified (Stewart et al., 2017). The pathogen has a remarkably broad host range within the Myrtaceae, a predominantly southern hemisphere plant family with many economically important species, including the eucalypts (Giblin & Carnegie, 2014). It affects young, actively growing leaves, shoots, flower buds and fruits, with repeated infection leading to death of highly susceptible plants (Coutinho et al., 1998; Pegg et al., 2014a, 2017; Carnegie et al., 2016). The potential risk to Australia has long been recognized, owing to its rich Myrtaceous flora (Glen et al., 2007). Following the first detection in Australia in 2010, it spread rapidly across the east coast from Victoria to Northern Queensland within a year. It was reported on the island of Tasmania by early 2015 and most recently west to the Northern Territory.

A wide range of Eucalyptus and Corymbia species are susceptible to A. psidii, as evidenced by controlled inoculation and field infection (Zauza et al., 2010; Pegg et al., 2014a; Potts et al., 2016).

Native pathogens

Quambalaria pitereka (Phylum Basidiomycota, Order Microstromatales, Family Quambalariaceae; Simpson, 2000) causes the disease Quambalaria shoot blight in Corymbia, Blakella and Angophora in Australia (Pegg et al., 2011b) and China. It infects foliage and juvenile stems, which can lead to losses in leaf area, and negatively impacts stem form (as a result of loss of apical dominance) and growth (Johnson et al., 2009; Pegg et al., 2009). Native to the east coast of Australia (Old, 1990), Q. pitereka has also been introduced into Western Australia, where it has been reported from plantations of C. maculata as well as native forests (C. calophylla) (Paap et al., 2008). In China, it has been reported on C. citriodora ssp. citriodora (Zhou et al., 2007) and the other three spotted gum taxa (J. T. Brawner, pers. obs.; 2017 Guangxi). Teratospheria leaf disease (TLD, formerly Mycosphaerella leaf disease), caused by Teratospheria cryptica and Teratospheria nubilosa (Phylum Ascomycota, Order Capnodiales, Family Teratosphaeriaceae), is the most significant foliar pathogen in E. globulus plantations in temperate regions of the world (Park et al., 2000; Mohammed et al., 2003). TLD lesions reduce photosynthetic capacity and cause leaf necrosis and defoliation, which can be highly detrimental to tree growth and form (Mohammed et al., 2003).

Genetic material, raising seedlings and trial designs

In both E. globulus and CCV, susceptibility to A. psidii was assessed using artificial inoculation of open-pollinated families derived from wild populations in Australia. Most families used had been previously assessed for native disease damage following natural infection of field trials and, in the case of CCV, the same plants that were inoculated with A. psidii were also artificially inoculated with Q. pitereka. Artificial inoculations and assessment of disease symptoms were performed at the Ecosciences Precinct, Brisbane (Queensland, Australia).

Austropuccinia psidii

Susceptibility of E. globulus to A. psidii was assessed using 191 open-pollinated families from 13 native populations, representing the races of Dutkowski & Potts (1999). The families were initially grown in glasshouse facilities at the Queensland Department of Agriculture and Fisheries in Gympie, Queensland, following the procedure outlined in Lee et al. (2015). Plants were arranged in a randomized incomplete block design with 20 replicates, each comprising four incomplete blocks of 40 plants in a five × eight arrangement. Randomization occurred at the family level, and each family was represented once per replicate, except where poor germination or mortality limited the number of available seedlings per family. Seedlings were transferred to the screening facility in Brisbane for rust inoculation in two batches, comprising 10 replicates each, at 2 and 3 months of age. At this stage, 79% of the families were represented in both batches. The first batch was screened in October/November 2013, when plants were 3 months old. Owing to a lack of active growth (flushing) of the second batch, and thus poor infection of the plants in the initial inoculation (at 3.5 months of age), the plants were cut back to remove any diseased tissue, keeping approximately two leaves on the main stem to produce a flush of new leaves for inoculation. Plants were then re-inoculated in January 2014 (at 6 months of age), when most plants had two pairs of fully expanded new leaves, and data from this second inoculation were analysed. When A. psidii inoculations are successful, there is high repeatability of damage scores obtained before and after cutting back seedlings (Butler et al., 2016). Overall, there were 2597 flushing seedlings assessed for A. psidii damage, with nine to 38 families represented per population and a mean of 13.7 seedlings per family.

Susceptibility of CCV to A. psidii was assessed in 125 families from the Woondum provenance from the coastal region of southeastern Queensland. The experimental design and growth of seedlings were essentially the same as described for E. globulus, except that 125 families were represented in each replicate, with five incomplete blocks of 25 plants. The first batch of 10 replicates was screened in June 2013, when plants were 3.5 months old, and the second batch was screened in December 2013 when plants were 10 months old and most had susceptible new growth. Both batches were successfully infected following a single inoculation.

Quambalaria pitereka

Following the assessment of susceptibility of CCV to A. psidii, the seedlings were cut back and allowed to regrow as described earlier for E. globulus. In this case, seedlings had to be cut back twice to ensure the majority of plants had uniform new growth for inoculation (see later) with Q. pitereka in January and February 2014.

The susceptibility of CCV to Q. pitereka was also quantified following natural infection in a field trial (25.76°S, 152.68°E; termed Bakers in Brawner et al., 2011), which included the same 125 families that were used in the artificial inoculations (as well as two additional families from the same provenance). The field trial was established with 5-month-old seedlings (King, 2004), with seven replicates of each family represented in line plots of four trees each. The trial also contained seed lots from other C. variegata provenances, but only the Woondum families were used in this study. A severe outbreak of Q. pitereka was first noted 4 months after planting, and in April 2000, 9 months after planting, the percentage of crown damage on each tree was visually scored using a six-point scale (1, 0%; 2, 1–10%; 3, 11–25%; 4, 26–50%; 5, 51–75%; 6, 76–100%). These data formed part of the study of Brawner et al. (2011).

Teratosphaeria leaf disease

The susceptibility of E. globulus to TLD was assessed in 247 open-pollinated families from 13 populations (as described earlier), based on natural infections in five common-garden field trials planted between 2005 and 2008. The severity of TLD damage was assessed as the percentage leaf area necrosis on the juvenile canopy in spring 2007, when plants were 1–2 yr old, with the exception of one trial assessed in spring 2010 (at 3 yr of age). A quantitative genetic analysis of these TLD data was reported in Hamilton et al. (2013). Of the families we assessed for rust damage, 81% were also assessed for TLD, and of those assessed for TLD, 96% were also assessed for rust.

Inoculum, inoculation and assessment of disease symptoms

Austropuccinia psidii

The procedures for collecting inoculum, rust inoculation and assessment of disease severity largely followed Pegg et al. (2014b). In brief, frozen urediniospores from the A. psidii biotype present in Australia were used for inoculation. These urediniospores were derived from a pustule isolate of A. psidii, collected from a susceptible host, Rhodamnia sessiliflora, growing in Brisbane.

Seedlings were assessed at 20 d (CCV) and 25 d after inoculation (E. globulus) for the severity of infection on new shoots and leaves using a disease rating scale modified from Junghans et al. (2003). This scale was as follows: 1, no symptoms, or minor yellow flecking evident; 2, presence of a hypersensitive reaction with flecking or necrosis; 3, small pustules, < 0.8 mm diameter, with one or two uredinia; 4, medium-sized pustules, 0.8–1.6 mm diameter with about 12 uredinia; 5, large pustules, > 1.6 mm diameter, with 20 or more uredinia on leaves, petioles and/or shoots (Pegg et al., 2014b). Only seedlings that were actively growing and had new shoots and leaves at inoculation were assessed.

Quambalaria pitereka

Selected isolates of Q. pitereka were obtained from the Brisbane Plant Pathology (BRIP) storage collection (BRIP samples 48 368, 48 387, 48 343, 48 424) and cultured onto potato dextrose agar for 2–3 wk in the dark at 25°C. To ensure the effects of storage had not impacted pathogenicity, each isolate was re-inoculated onto CCV seedlings, and new cultures of each isolate were established. A suspension containing an equal concentration of spores from all four isolates was used for artificial inoculation following the general procedure outlined in Pegg et al. (2009). Seedlings were assessed 20 d after inoculation using a 1–5 scale similar to that used for A. psidii.

Data analysis

Eucalyptus globulus pathosystems

Genetic analyses of the A. psidii data collected from E. globulus focused on the estimation of heritabilities (h2), degree of population differentiation (QST) (Latta, 1998; Leinonen et al., 2013) and the extent to which susceptibility to A. psidii was genetically correlated (rg) with the susceptibility to TLD reported in Hamilton et al. (2013). Following Hamilton et al. (2013), variance components for estimating narrow-sense heritability and QST were obtained from restricted maximum likelihood (REML) mixed-model analyses implemented with ASReml 4.0 (Gilmour et al., 2014), and fitting the following linear model:
urn:x-wiley:0028646X:media:nph15557:nph15557-math-0002
where Y is the pathogen damage score, μ is the mean, REP is the fixed replicate effect, POPULATION is the random population effect, IBLK(REP) is the random incomplete block within replicate effect, TREE is the within-population additive genetic effect, and RESIDUAL is the error. This model was fitted separately for each screening batch and the combined data. The tree term allowed the estimation of the pooled additive genetic variance within populations (urn:x-wiley:0028646X:media:nph15557:nph15557-math-0003). This analysis used a multigeneration pedigree file to define the additive relationship matrix for parents and open-pollinated offspring assuming a selfing rate of 0.3 (Dutkowski et al., 2001; Gilmour et al., 2014).
The narrow-sense open-pollinated heritability (urn:x-wiley:0028646X:media:nph15557:nph15557-math-0004) and QST were estimated as:
urn:x-wiley:0028646X:media:nph15557:nph15557-math-0005
urn:x-wiley:0028646X:media:nph15557:nph15557-math-0006
where urn:x-wiley:0028646X:media:nph15557:nph15557-math-0007 is the population variance and urn:x-wiley:0028646X:media:nph15557:nph15557-math-0008 is the residual variance.

The significance of the population and additive genetic variances was tested with a ‘one-tailed’ likelihood ratio test (Gilmour et al., 2014). To test whether the urn:x-wiley:0028646X:media:nph15557:nph15557-math-0009 estimates differed between batches, a bivariate extension of this linear model (with each batch treated as different variables) was fitted, allowing for covariation among populations and additive genetic effects (Gilmour et al., 2014). The common heritability that achieved the maximum log-likelihood value was identified iteratively by varying the common additive to residual variance ratio, and comparing the log likelihood with that of the unconstrained model where independent variances were estimated for each batch. This comparison was done using a two-tailed likelihood ratio test (LRT) with one degree of freedom (Gilmour et al., 2014). Following Hamilton et al. (2013), ‘two-tailed’ likelihood ratio tests (Gilmour et al., 2014) were used to test the significance of the difference of the QST estimate for A. psidii damage from a previously published estimate of FST (0.09) which was based on eight microsatellite markers and used a similar group of E. globulus populations (Steane et al., 2006), and from previously published estimates of QST for TLD damage (Hamilton et al., 2013).

To estimate the population and additive genetic correlations between exotic (A. psidii) and native (TLD) pathogen damage, the A. psidii inoculation data were combined with previously analysed TLD data (Hamilton et al., 2013). This was done by treating TLD damage at the various field sites and the inoculation experiment as separate traits linked through a common pedigree. The correlations were estimated using bivariate analyses, fitting the model terms detailed in Hamilton et al. (2013) for the TLD damage and terms described in the described linear model for A. psidii damage, and allowing for covariation among random population and additive genetic effects. These random genetic effects were common to both the field trial and inoculation data. Standard errors of parameters were estimated from the average information matrix, using a standard truncated Taylor series approximation (Gilmour et al., 2014). Population means for plotting spatial trends and for path analyses (see later) were predicted by treating population as a fixed effect in univariate analyses with the linear models described.

Corymbia citriodora ssp. variegata pathosystems

The analysis of the CCV inoculation data followed that described for E. globulus, except that the population term was excluded from the model, as only a single population of CCV was studied. In addition, as the families were planted as four-tree line plots in the field trial, a random plot term was fitted for the analysis of Quambalaria field damage. For each pathogen, a bivariate model was used to estimate genetic correlations among the different inoculation batches, and the significance of these correlations deviating from one was tested using a LRT. The bivariate model was then extended to the multivariate level to estimate all the genetic correlations among the A. psidii damage scores and the Quambalaria field and inoculation damage scores. This was a five-trait analysis, as the two glasshouse inoculation batches were treated as separate traits for both pathogens. As the same CCV plants were sequentially screened with the exotic and native inocula, it was necessary to allow for covariation among their residuals and incomplete block terms in the multivariate model. As the additive genetic correlation between the two batches was not significantly different from one in the bivariate analysis of Quambalaria, this correlation was fixed to one in the multivariate analysis, and other paired correlations involving these two batches were constrained to be equal. To test the homogeneity of the genetic correlations across field and inoculation studies, a further constraint was applied, which required the correlations of both batches of rust screening with the field and inoculation screenings with Quambalaria to be equal. The difference in log likelihoods between the two models was then tested with a LRT.

Path analysis

The relationships of population home-site geographic and climate variables with population variation in disease susceptibility were modelled in each pathosystem using the path analysis framework of proc calis of Sas (v.9.4). Susceptibility was linked to variation in the key climate variables expected to affect disease risk: precipitation and temperature. For consistency with Hamilton et al. (2013), the climatic data used for the E. globulus populations was derived from the SILO Data Drill (https://legacy.longpaddock.qld.gov.au/silo/datadrill/index.php) and modelling was undertaken with population mean annual precipitation (mm) and mean daily maximum temperature (°C) based on data from 1908 to 2007. For CCV, modelling used 1976–2005 mean annual precipitation (RANN, mm) and mean annual temperature (TANN, °C) derived from BIOCLIM climate surfaces (Anuclim version 6.1). For the native pathosystems, we used published population data on susceptibility from field trials showing the greatest population differentiation (E. globulus/TLD – trial TEM06 (Hamilton et al., 2013); CCV/Quambalaria – mean tip damage from trial 451c (Brawner et al., 2011)). For the exotic pathogens, we used population means for E. globulus (Fig. 1), and for Corymbia we used the average disease severity ratings published in Pegg et al. (2014b) for nine spotted gum populations (which encompassed C. henryi and two subspecies of C. citriodora), five of which were CCV.

Details are in the caption following the image
(a, b) Host population variation in susceptibility to the native pathogen Teratosphaeria species (TLD) (a) and the exotic rust Austropuccinia psidii (b) across the native range of Eucalyptus globulus in southeastern Australia, including Tasmania and the Bass Strait Islands. Least-squares mean damage for each population is shown; the bigger the triangle or circle, the greater is the host tree disease damage above or below the overall mean, respectively. Small circles and triangles represent values close to the overall mean. The damage from Teratosphaeria infection is shown for the field trial exhibiting the greatest population differentiation (Qst) in damage following natural infection at 1 yr of age (Temma06; Table 1; Hamilton et al., 2013). The rust damage was assessed following artificial inoculation of seedlings, mainly from the same open-pollinated E. globulus families.

Results

Eucalyptus globulus

Significant genetic variation in damage as a result of A. psidii was observed both within (i.e. urn:x-wiley:0028646X:media:nph15557:nph15557-math-0010 > 0; Table 1) and among (LRT (QST = 0), P < 0.001) populations of E. globulus. As similar amounts of additive and population variation were evident in the two different batches (data not shown) and high additive genetic (0.93 ± 0.05; LRT from 1, P = 0.070) and population-level (0.84 ± 0.14; LRT from 1, P = 0.007) correlations were found across batches, batch data were combined into a single trait. The heritability of A. psidii damage across the artificial screenings was 0.65, nearly two-fold higher than that of even the highest value obtained following natural TLD field infection of comparable germplasm (Table 1). By contrast, the quantitative inbreeding coefficient (QST) among populations for TLD damage was, on average (0.14), over two-fold greater than that for A. psidii damage (0.06 ± 0.03; Table 2). In the field trial with the highest TLD damage (Temma06), where population differences were most expressed, the QST for TLD was significantly (P < 0.01) greater than that observed for the rust (Table 2) and significantly (P < 0.05; Hamilton et al., 2013) greater than neutral expectation based on the published FST from neutral molecular markers of 0.09. The QST for A. psidii damage was less than and not significantly different (P = 0.417) from FST. The genetic correlation between A. psidii and TLD damage was not significantly different from zero for any of the field trials at either the population or additive genetic levels (Table 2).

Table 1. Narrow-sense heritability (urn:x-wiley:0028646X:media:nph15557:nph15557-math-0011) estimates for Austropuccinia psidii and Teratospheria leaf disease (TLD) damage in Eucalyptus globulus
Pathogen and assessmenta urn:x-wiley:0028646X:media:nph15557:nph15557-math-0012 SE P b
A. psidii
 Batch 1 0.63 0.09 < 0.001
Batch 2 0.70 0.10 < 0.001
Combinedc 0.65 0.07 < 0.001
TLDd site
Tog05 0.22 0.05 < 0.001
SR05 0.17 0.04 < 0.001
Temma06 0.26 0.05 < 0.001
SR06 0.13 0.03 < 0.001
GC08 0.35 0.06 < 0.001
Average 0.23
  • a Screening batch for A. psidii artificial inoculations and trial site for TLD natural infections.
  • b Significance of the additive genetic variance from zero.
  • c Combined data with the difference between batches included in the modelled replicate effect.
  • d Results are from Hamilton et al. (2013).
Table 2. Quantitative inbreeding coefficients (Qst) for Teratospheria leaf disease (TLD) and Austropuccinia psidii (rust) damage in Eucalyptus globulus, and correlations with susceptibility to A. psidii damage for TLD in each field trial at the population (rp) and (ra) additive levels
Pathogen and trial n fams a n Q ST SE P (QST TLD = QST rust) Population correlations (rp) Additive genetic correlations (ra)
r p SE P (rp = 0) r g SE P (ra = 0)
TLD
Tog05 140 2295 0.05 (0.03) 0.655 −0.36 (0.41) 0.417 −0.10 (0.17) 0.610
SR05 146 2727 0.12 (0.06) 0.173 −0.38 (0.34) 0.313 0.08 (0.16) 0.624
Temma06 124 2771 0.25 (0.09) 0.005 0.06 (0.35) 0.888 0.21 (0.15) 0.182
SR06 140 2863 0.12 (0.06) 0.168 0.29 (0.35) 0.431 0.16 (0.17) 0.354
GC08 141 2236 0.16 (0.07) 0.057 0.49 (0.30) 0.164 −0.04 (0.15) 0.806
Average 0.14 0.02 0.06
Rust 189 2597 0.06 (0.03)
  • a nfams indicates the number of families, and n the number of individuals, used for TLD damage estimates in each field trial.

At the population level, the nonsignificant correlation between A. psidii and TLD damage was a result of their different patterns of spatial variation across the native range of E. globulus (Fig. 1a,b). Specifically, there was a distinct latitudinal cline in TLD damage (Fig. 1a) with the most susceptible population found in southern Tasmania and resistance increasing northwards into mainland Australia (see also Hamilton et al., 2013). These trends are associated with changes in climate, with the path analysis showing that populations originating from warmer and wetter environments are less susceptible to TLD (Fig. 2a). This is consistent with an expectation of an increased disease risk in these environments. By contrast, the only significant path coefficient detected for A. psidii susceptibility was with precipitation (Fig. 2b), but in this case the association was positive. This association was mainly driven by differences in the proportion of plants in disease class 1 (no symptoms or mild necrotic flecking; r = −0.73, P < 0.004), suggesting that provenances from wetter regions have a greater proportion of plants with symptoms indicative of host leaf penetration by the pathogen. This result is inconsistent with pathogen-induced selection having shaped population divergence in susceptibility, as the populations originating from the wetter areas (i.e. higher disease risk) were more susceptible. Although A. psidii susceptibility decreased northwards on the east coast of Tasmania, this trend did not extend to mainland populations as it did with TLD (Fig. 1). Indeed, the western Victorian populations, which were the most resistant to TLD, were the most susceptible to A. psidii.

Details are in the caption following the image
Path diagram depicting the relative effect of home-site climate on host susceptibility in the different pathosystems in this study: (a) Eucalyptus globulus/Teratosphaeria; (b) Eucalyptus globulus/Austropuccinia psidii; (c) Corymbia citriodora ssp. variegata (CCV)/Quambalaria pitereka; (d) Corymbia citriodora ssp. variegate/Austropuccinia psidii. To standardize interpretation, response data were adjusted such that increasing values reflected increasing provenance susceptibility. Although all paths were fitted in the model, only significant (P < 0.05) paths and their standardized coefficients reflecting the relative effect size and their significance were plotted. R2 refers to the proportion variation in susceptibility explained by the full model. Significance levels of path coefficients are: *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Corymbia citriodora ssp. variegata (CCV)

Significant additive genetic variation was observed within the Woondum population of CCV for susceptibility to A. psidii (i.e. urn:x-wiley:0028646X:media:nph15557:nph15557-math-0013 > 0) and, as with E. globulus, heritability estimates were high (Table 3). Similar degrees of additive genetic variation were evident in the two batches (data not shown); however, the different batches were treated as separate traits, as bivariate analysis showed that the additive genetic correlation between batches was lower than for E. globulus and highly significantly different from 1 (0.77 ± 0.08; LRT from 1, P < 0.001).

Table 3. Narrow-sense heritability (urn:x-wiley:0028646X:media:nph15557:nph15557-math-0014) estimates for Austropuccinia psidii and Quambalaria pitereka damage in the Woondum population of Corymbia citriodora ssp. variegata (no. of families = 125)
Pathogen and assessmenta urn:x-wiley:0028646X:media:nph15557:nph15557-math-0015 SE P b
A. psidii
 Inoculation
Batch 1 0.63 0.10 <0.001
Batch 2 0.59 0.10 <0.001
Q. pitereka
 Inoculation
Batch 1 0.08 0.07 0.113
Batch 2 0.08 0.09 0.196
Combinedc 0.12 0.05 <0.001
Natural infection 0.42 0.06 <0.001
  • a Screening batch for artificial inoculations, and natural infection vs artificial inoculation in the case of Q. pitereka.
  • b Significance of the additive genetic variance estimate from zero.
  • c Combined data with the difference between batches included in the modelled replicate effect.

When the same CCV plants were artificially inoculated with Q. pitereka, the heritability of damage estimates (urn:x-wiley:0028646X:media:nph15557:nph15557-math-0016 = 0.08–0.12; Table 3) was markedly lower than that observed for A. psidii (urn:x-wiley:0028646X:media:nph15557:nph15557-math-0017 = 0.59–0.63; Table 3). It was also lower than observed following other inoculations of CCV with Q. pitereka (urn:x-wiley:0028646X:media:nph15557:nph15557-math-0018 = 0.33; Supporting information Table S1; Notes S1). In a bivariate analysis, the genetic correlation between batches was effectively 1 (estimate was at the boundary of the parameter space) and the different batches were therefore treated as a single trait. While the individual batch heritability estimates were not significantly greater than zero, the pooled estimate was (P < 0.001). Despite the low heritability observed under artificial inoculation, the same families exhibited a high heritability for Q. pitereka damage under natural infection in the field (urn:x-wiley:0028646X:media:nph15557:nph15557-math-0019 = 0.42 ± 0.06; Table 3), and the estimates from natural and artificial inoculation were highly correlated at the additive genetic level (ra = 0.72 ± 0.15; LRT from 0, P < 0.001; Table 4).

Table 4. Genetic correlations of Quambalaria pitereka and Austropuccinia psidii damage in the Woondum population of Corymbia citriodora ssp. variegata (no. of families = 125)
Q. pitereka a.i.a A. psidii batch1 A. psidii batch 2
A. psidii batch 1 0.21 (0.18)c
A. psidii batch 2 0.13 (0.19)c 0.77 (0.08)***
Q. pitereka f.t.b 0.71 (0.15)*** −0.21 (0.11)c −0.20 (0.12)c
  • a Artifical inoculation.
  • b Natural infection (crown damage) in field trials.
  • c not significantly different from zero at the 0.05 level; ***, P < 0.001.

As with E. globulus, no genetic association between susceptibilities to the native and introduced pathogens was detected in CCV. This result was consistent across natural and artificial inoculations, as well as across the different CCV batches screened for A. psidii (Table 4) and another smaller-scale confirmation inoculation of CCV (Notes S1). Although there was a slightly positive correlation of susceptibility with the different pathogens under artificial inoculation, this was not observed with the natural infection of Q. pitereka, and the LRT constraining the four native/exotic genetic correlations to be the same indicated that they were not significantly different (pooled ra = −0.13 ± 0.10; LRT for homogeneity with 4 df, P = 0.190).

Integration and reanalysis of the published data on population variation in C. citriodora revealed that the population differences in susceptibility were significantly associated with home-site climate for the native (Fig. 2c) but not the exotic (Fig. 2d) pathogen. The associations were somewhat consistent with those observed in E. globulus. In both native pathosystems, increased population susceptibility was associated with decreasing home-site precipitation. However, in contrast to the relationship between temperature and susceptibility observed in the temperate E. globulus native pathosystem, the populations originating from warmer sites within the subtropical CCV range (i.e. sites further north, or further inland) tended to be more susceptible to Q. pitereka damage. No significant associations between climate and population susceptibility to the exotic pathogen A. psidii were detected (Fig. 2d).

Discussion

Three key results relevant to our initial hypotheses emerged. First, despite significant additive genetic variation for susceptibility within host populations, host susceptibilities to the native and exotic pathogens are genetically uncorrelated, suggesting that pathogen-specific defence mechanisms are involved. Second, host populations are more differentiated in their susceptibility to native than to exotic pathogens in cases exceeding that expected through drift for the native, but not the exotic pathogen. Third, where testable, population differentiation in native and exotic pathogen susceptibilities is also uncorrelated, and susceptibility generally decreases to the native, but not the exotic, pathogen where climate increased disease risk. Together, these findings provide strong evidence that pathogen-imposed selection has shaped the natural distribution of host resistance to native pathogens in these tree species.

Significant genetic variation was detected within populations for susceptibility to the native and exotic pathogens in both host species. In the case of the native pathogens, this finding is consistent with previous reports for TLD in E. globulus (Costa e Silva et al., 2013; Hamilton et al., 2013; Balmelli et al., 2014) and Q. pitereka in CCV (Brawner et al., 2011; Pegg et al., 2014b). Such genetic variation in host susceptibility is common in native pathosystems in which spatial and temporal fluctuations in selection pressure from pathogens are likely to maintain variation in host resistance (Burdon et al., 2006; Laine et al., 2011). Indeed, for the native pathogens in this study, seasonal variation in epidemics related to climatic differences within and between years has been documented (Pinkard et al., 2010) and is accentuated by host developmental and ontogenetic variability in susceptibility (Park, 1988a,b; de Little et al., 2008; Hunter et al., 2009). Variation in disease risk and host susceptibility has also been documented over finer spatial scales than our sampling (Wilkinson, 2008; Pinkard et al., 2010), which may contribute to the genetic variability within populations. By contrast, native hosts often lack significant resistance to exotic pathogens (Burdon et al., 2013). Thus, the significant genetic variation in susceptibility to the exotic A. psidii evident in both hosts (present study, Pegg et al., 2014b) is somewhat unexpected, as they are unlikely to have been exposed to this pathogen in their recent evolutionary history (Tobias et al., 2016).

Consistent with our findings, variation in susceptibility to A. psidii has been reported in numerous Australian Myrtaceae species (Morin et al., 2012). As well as the existence of genetic variation for susceptibility in a range of taxa, the nature of the resistance response to the exotic A. psidii is intriguing, as in some cases it appears to involve a highly specific response that is normally indicative of pathogen recognition (Thumma et al., 2013; Tobias et al., 2016; Hsieh et al., 2017). This raises the following questions: what genetic mechanisms confer variation in susceptibility to the exotic A. psidii in the Australian flora, and is this variation related to susceptibility to native pathogens? Past studies employing artificial inoculation in eucalypt (Butler et al., 2016; Potts et al., 2016) and other taxa (Morin et al., 2012) have shown that resistant genotypes can display no symptoms/mild necrotic flecking, suggesting a lack of host penetration or cell invasion, as well as symptoms consistent with a hypersensitive response, implying that the pathogen has entered the host cells and been recognized by the host (Jones & Dangl, 2006).

Among the potential explanations for specific resistance to A. psidii in the Australian flora, Tobias et al. (2016) argue that the surveillance of host integrity (Cook et al., 2015), as proposed by the guard or decoy models (van der Hoorn & Kamoun, 2008), is likely to play an important role. Under this hypothesis, resistance to A. psidii may be attributable to a common Myrtaceae ‘effector hub’, which, when modified, leads to host recognition and response. This implies exotic pathogens may produce similar ‘invasion patterns’ (Cook et al., 2015) to those produced by coevolved pathogens. Thus, selection by coevolved pathogens could account for the variation in resistance to A. psidii in Australian Myrtaceae. The absence of significant genetic correlations between host susceptibility to the native pathogens in this study and susceptibility to the exotic A. psidii would argue against the common effector hub hypothesis. In the case of E. globulus, this is consistent with the observation that none of the quantitative trait loci (QTLs) identified for A. psidii resistance were co-located with QTLs for TLD (Freeman et al., 2008; Butler et al., 2016). Together, the lack of significant genetic correlations, or common QTLs in the case of E. globulus, suggest that the genetic mechanisms influencing susceptibility to the native and exotic pathogens within each host are largely independent, and therefore selection by the native pathogens studied has not substantially impacted resistance to A. psidii.

Our findings also support the contention that host population divergence in resistance to the native pathogens is not a result of diffuse (multispecies) selection arising from a pathogen general resistance mechanism (Leimu & Koricheva, 2006) or variation in other foliar traits that indirectly impact general pathogen susceptibility. However, our native pathogens are quite phylogenetically distant from A. psidii (Teratospheria is in a different phylum and Quambalaria a different order, but the same phylum). Thus, in terms of diffuse selection, we cannot dismiss the possibility that the variation in susceptibility of our host species to A. psidii is more similar to its variation in susceptibility to phylogenetically closer native pathogens than those in our study.

If diversifying pathogen-imposed selection has shaped the evolution of wild host species, we would expect population differentiation (measured by the quantitative inbreeding coefficient, QST) in host susceptibility to exceed that expected through drift (i.e. QST > FST; Leinonen et al., 2013) for the native but not the exotic pathogens. This was the case in E. globulus, where population differentiation in susceptibility to the native pathogen was, on average, more than two-fold greater than that to the exotic pathogen (Table 2); in the trial with the highest TLD damage, this significantly exceeded neutral marker FST (average 0.09; Steane et al., 2006; Hamilton et al., 2013). Similar trends are likely in CCV. Using genetic parameters from Brawner et al. (2011) and Pegg et al. (2011a), we derived QST values for the susceptibility of CCV to the native pathogen Q. pitereka (Notes S2). Regardless of the manner in which host susceptibility was assessed, the average QST value (Table S1) was (0.17) about twice the average neutral marker FST values reported for CCV (0.07–0.09; Ochieng et al., 2010; Dillon et al., 2012), providing additional evidence that pathogen-imposed selection has shaped population divergence in susceptibility to the native pathogen.

Our final line of evidence for pathogen-imposed selection is the population-level associations between susceptibility to the native pathogens and climatic gradients related to disease risk in both host species. In each case, population-level susceptibility is significantly associated with home-site temperature and precipitation (Fig. 2), and populations from environments with greater climatic suitability for disease outbreaks are generally less susceptible to the native pathogens. Such trends have been reported previously in E. globulus (Hamilton et al., 2013) and other forest trees (Ades et al., 1992; Perry et al., 2016). In E. globulus, physiological tolerance (Park, 1988b) and disease risk modelling (Pinkard et al., 2010) suggest that the probability of infection by TLD will increase with increasing temperature and precipitation in the near-coastal region of southeastern Australia, which E. globulus inhabits. Therefore, the negative relationships of susceptibility to TLD with home-site temperature and precipitation, as well as modelled bioclimatic risk (Pinkard et al., 2010), are consistent with historical pathogen-imposed selection influencing population-level susceptibility to TLD, as proposed by Hamilton et al. (2013).

Population-level susceptibility of CCV to Q. pitereka is also significantly associated with climatic and geographic factors related to disease risk. Past studies have reported provenance-level variation in susceptibility with mean annual rainfall (Dickinson et al., 2004), latitude (Johnson et al., 2009) and in coastal vs inland provenances (Self et al., 2002; Johnson et al., 2009). Our analysis highlights a strong relationship with precipitation and an association between susceptibility and temperature. Although climatic risk modelling has not been undertaken for this pathosystem, the requirements for Q. pitereka conidia germination are similar to those of A. psidii (Pegg et al., 2009). The modelled climate envelope for A. psidii is mostly confined to near coastal regions within the natural range of CCV, as inland regions are too hot and dry (Kriticos et al., 2013). In this subtropical region, high rather than low temperatures (as in the case of the temperate E. globulus) are a limiting factor. The decreased susceptibility in wetter and cooler populations (Fig. 2) is therefore consistent with increased pathogen-imposed selection.

By contrast, host susceptibility to the exotic A. psidii did not decrease in provenances from areas with a greater disease risk. In fact, the opposite association with rainfall was observed in E. globulus, whereby susceptibility to A. psidii tended to be higher in populations originating from wetter areas, where disease risk is probably greater (Fig. 2). Such population differences in susceptibility could reflect an indirect response associated with adaptation of functional traits to rainfall gradients across the range of E. globulus. Indeed, increased susceptibility to A. psidii in populations originating from wetter areas, where disease risk would be expected to be higher, was also reported in E. cloeziana in Queensland (Lee et al., 2015), consistent with preadaptation to precipitation gradients influencing susceptibility at the population level. Such preadaptation could impact susceptibility to specific pathogens and involve constitutive morphological, anatomical or chemical traits (Niinemets, 2001; Smith et al., 2006, 2017; Close et al., 2007).

A number of factors may be confounded in our comparison of susceptibilities to native and exotic pathogens. These include the comparison of results from artificial inoculation with a single strain of A. psidii with natural infections by the native pathogens, which may reflect multiple strains and multiple species in the case of TLD. Hence, findings such as the higher heritability estimates for susceptibility to A. psidii than to the native pathogens in both host species, for example, require further verification. The lower heritability estimates for the native pathogens may reflect an increased error variance in field-based estimates as a result of factors such as host escapes from infection, variability in the pathogen and greater assessment error. However, such factors are unlikely to affect the relative differences in QST, as this is a ratio of genetic variances. In addition, the heritability for field-based infection by Q. pitereka (hop2 = 0.42) was actually higher than for artificial inoculation (hop2 = 0.08–0.12), although the latter may have been reduced by the fact plants were inoculated with A. psidii before inoculation with Q. piterka, as well as issues with variation in the vigour of reshooting plants. Nonetheless, the high correlation between estimates of CCV susceptibility to Q. pitereka from natural infections and artificial inoculation, and the fact that neither was genetically correlated with susceptibility to A. psidii reinforce our findings (Table 4). Similarly, the stability of TLD damage across five different trial sites (Hamilton et al., 2013) and the fact that none of these damage estimates was genetically correlated with susceptibility to A. psidii at the additive or population level provide further evidence that these key findings are robust.

In conclusion, comparison of the genetic architectures of susceptibility to native and exotic pathogens provides multiple lines of evidence for direct pathogen-imposed selection. The independence of susceptibility to native and exotic pathogens indicates that different resistance mechanisms are involved, and suggests that a general pathogen recognition/resistance mechanism does not play a significant role. The greater population divergence in native disease susceptibility is consistent with divergent selection, which quantitative genetic independence argues is pathogen-specific. At the population level, reduced susceptibility in areas climatically more suitable to disease outbreaks was evident for the native but not the exotic pathogen, arguing against an indirect abiotic basis to population divergence in susceptibility to the native pathogens. Taken together, our findings provide strong evidence that historic pathogen-imposed selection has directly shaped the evolutionary trajectory of these forest-tree gene pools and contributed to a geographic mosaic of host resistance to native pathogens.

Acknowledgements

This research was supported by the Australian Government's Collaborative Research Network involving the University of the Sunshine Coast, Griffith University and the University of Tasmania, the Plant Biosecurity CRC and Australian Government's Cooperative Research Centres Program. Analysis of the E. globulus data and writing was further supported by an Australian Research Council Linkage grant (LP140100506) held in partnership with the Southern Tree Breeding Association. We thank René Vaillancourt, Jakob Butler, Peter Ades, Josquin Tibbits, Simon Southerton, Bala Thumma and Karanjeet Sandhu for discussion.

    Author contributions

    JTB and all other authors contributed to the experimental design and critical review and approved the final manuscript. PAT managed seed collections and prepared figures. DJL organized seedling production. GSP performed the controlled inoculations and disease assessments. MGH and BMP analysed the data. JSF and BMP drafted the manuscript.