Pathogen dynamics in a crop canopy and their evolution under changing climate

Authors


E-mail: sukumar.chakraborty@csiro.au

Abstract

Canopy-level interactions have been largely ignored in epidemiological models and their applications in defining disease risks under climate change, although these interactions are important for disease management. This paper uses anthracnose of Stylosanthes scabra as a case study and reviews research on dynamics of the pathogen (Colletotrichum gloeosporioides) at the canopy level and pathogen evolution under changing climate. It argues that linking of pathogen dynamics, crop growth and climate models is essential in predicting disease risks under climate change. A plant functional-structural model was used to couple S. scabra growth and architecture with disease under ambient and elevated CO2. A level of induced resistance in plants with enlarged canopy determined anthracnose severity at elevated CO2. Moisture-related microclimatic variables determined infection at ambient but not at elevated CO2. At high CO2 increased disease level from raised pathogen fecundity in enlarged canopy accelerated pathogen evolution after 25 sequential infection cycles. Modelling of pathogen dynamics under climate change currently suffers from a paucity of quantitative data, mismatch of scales in coupling climate and disease models, and model uncertainties. Further experimental research on interactions of biotic and abiotic factors on plant diseases under climate change and validation of models are essential prior to their use in climate-change prediction. Understanding and anticipating trends in host–pathogen evolution under climate change will improve the durability of resistance and lay the foundation for increased crop adaptation through pre-emptive plant breeding.

Introduction

Plant disease epidemics result from interactions of host and pathogen populations in a conducive environment, the so called ‘disease triangle’, which is a basic tenet of plant pathology. In reality, this dynamic host–pathogen interaction is played out within a plant canopy with its own complex structure and microclimate, which mediates the external environment. Crop canopy, simply defined as the three-dimensional (3D) space occupied by a growing crop, changes in density with changing leaf area to influence microclimate, including radiation interception, temperature, moisture and wind (Zadoks & Schein, 1979). For the pathogen, processes including dispersal, infection and production of secondary inoculum are repeated during each infection cycle when host canopy microclimate and external environment are favourable. For the host plant, canopy architecture modifies microclimate and influences dispersal of aboveground pathogens and disease development. Changing size and characteristics of a growing plant canopy influence epidemic progression by increasing the density of susceptible host tissue, which in turn increases the basic reproduction ratio defined as the total number of daughter lesions relative to the number of mother lesions (Ferrandino, 2008).

Changing atmospheric composition and climate also modify the plant canopy. At elevated CO2 a ‘fertilization effect’ stimulates plant growth and productivity and changes plant morphology, canopy architecture, anatomy, physiology, chemical composition and gene expression profile (Matros et al., 2006). Morphological changes at elevated CO2, including increased plant height, number of branches and tillers, and number, thickness and area of leaves (Allen, 1990), all lead to an enlarged canopy. These and other changes in canopy size, density and architecture influence canopy microclimate (e.g. Sauer et al., 2007). A disease epidemic itself can change the density and architecture of canopies, thereby influencing water and radiation balance (Carretero et al., 2010), whilst disease-induced changes in host canopy act as a feedback loop to influence epidemic progression. Although many have considered the physiology, biochemistry and molecular biology of host–pathogen interactions (Serrago et al., 2009), very few studies have integrated canopy characteristics to analyse and understand the dynamics of epidemics (Pangga, 2002; Carretero et al., 2010) or to modify canopy architecture for disease control (Ando et al., 2007).

The effect of changing atmospheric composition and climate on individual pathosystems can be positive, negative or neutral (Coakley et al., 1999; Garrett et al., 2006; Chakraborty et al., 2008) as a result of the highly specific nature of host–pathogen interactions. However, the number of case studies has been small, with even fewer studies under realistic field conditions. A summary of these studies shows that the effect on diseases stems from altered stages and rates of pathogen development, modified host resistance and changed host–pathogen interactions (e.g. Pangga et al., 2004a; Hibberd et al., 1996; Matros et al., 2006; Melloy et al., 2010). Climate change affects many aspects of disease development at different levels of integration in a pathosystem and the effect can be complex and unpredictable (Jeger & Pautasso, 2008). Therefore, it is essential to examine hypotheses at various levels of spatial and temporal integration, as extrapolation between levels may not be easy.

New information is emerging on changed pathogen dynamics in response to changing climate and atmospheric composition at large spatial and temporal (decadal) scales. These include fluctuations in the incidence of two wheat pathogens with changing rainfall, temperature and SO2 emission (Bearchell et al., 2005); and a domination of the wheat stripe rust population in the eastern USA since the year 2000 by more aggressive strains better adapted to rising temperature (Milus et al., 2009). Surveys in Europe have shown an increased prevalence of the wheat and barley head blight pathogen Fusarium graminearum, with higher temperature optima, over Fusarium culmorum and Microdochium nivale as the temperature warms up (Isebaert et al., 2009). Of the three pathogens, only Fusarium species produce mycotoxins such as deoxynivalenol, which is harmful to human and animal health if high levels are produced in grain and grain products. Therefore, the changing composition and dynamics in the head blight pathogen have serious implications for food safety and human and animal health. Altered frequency of species/strain as a result of climate and the consequent level/type of mycotoxin contamination are also known for Aspergillus flavus, A. parasiticus and A. nomius, producing different forms of aflatoxin; and A. carbonarius producing ochratoxin A (Paterson & Lima, 2010). These species differ in their temperature optima for mycotoxin production and growth. These findings are useful to increase preparedness for impending changes, but on-farm forecasting and management of diseases requiring information at enterprise and/or farm levels require detailed understanding of the complexities of host–pathogen interactions at the canopy level when dealing with climate-change effects. One aim of this paper is to summarize host–pathogen–canopy microclimate dynamics from the limited literature.

Climate change can lead to the emergence of pre-existing or introduced pathogens as major disease agents (Anderson et al., 2004), sometimes inflicting severe loss to crop production. Models linking host–pathogen biology and climate are often used to predict outbreaks and biosecurity risks (Desprez-Loustau et al., 2007). Even in the absence of climate change, the introduction of exotic pathogens and the evolution and subsequent dispersal of new virulent pathogen races, such as the wheat stem rust pathotype Ug99 (Pretorius et al., 2000), are among the biggest threats to international agriculture and global food security (Chakraborty & Newton, 2011). Ug99 is virulent on the rust resistance gene Sr31 that has been effective worldwide for over 30 years. It has the potential to threaten global wheat production; even a 10% loss of yield could cost US$ 1–2 billion in Asia alone (Duveiller et al., 2007). The emergence of this race has led to the formation of the Borlaug Global Rust Initiative (BGRI, http://www.globalrust.org), with an internationally coordinated programme to develop resistant cultivars. Despite modelling efforts (Leonard & Czochor, 1980), predicting the evolution of new rust races has not been possible, and increased uncertainty under a changing climate will make this task even more difficult. There are already examples in the literature of rapid genetic adaptation to various biotic and abiotic stress under elevated CO2, including that of a plant pathogen (Chakraborty & Datta, 2003).

Dynamics and evolution of plant pathogens have been neglected areas of research in dealing with climate change. Improved understanding of how future climates may influence pathogen dynamics within plant canopy and host–pathogen evolution is essential for future disease management. This paper uses anthracnose of Stylosanthes as a case study and reviews experimental and modelling research on pathogen dynamics at the canopy level and pathogen evolution under changing CO2 and climate to help raise awareness of the need to intensify research.

Pathogen dynamics at the canopy level

Pathogen dynamics can change as a direct result of enlargement of plant/crop canopy at elevated CO2, trapping many more spores and offering more infection sites with a microclimate of high relative humidity without the desiccating effect of sunshine and wind and changed rainfall interception, making it conducive to the development of many diseases. These effects may be predicted by incorporating changes in canopy and other characteristics under climate change into existing models of pathogen dynamics within a crop canopy. Some other changes result from modifications in the biology of host and pathogens, which impact on host resistance and pathogen aggressiveness. Change in pathogen biology at elevated CO2 was first demonstrated in the powdery mildew pathogen Blumeria graminis f. sp. hordei, which grew faster inside barley tissue and produced more conidia (Hibberd et al., 1996). Since then, many studies have demonstrated altered pathogen biomass or fecundity, whilst others show no change at elevated CO2 (Table 1). The information in Table 1 does not include reports of increase or decrease in disease severity or data from the large number of in vitro studies on various pathogens under altered CO2, O3, UV-B or physical climatic factors (Manning & Tiedemann, 1995). Changes resulting from altered host–pathogen biology are harder to predict because of the highly specific nature of host–pathogen interactions and the changes in pathogen dynamics that are intricately linked to changes in the host plant. Important changes in host plants under elevated CO2 and O3 that impact on plant diseases are summarized in Eastburn et al. (2011). Also, a paucity of knowledge makes it difficult to extrapolate, therefore more experimental research under realistic field conditions is needed to improve understanding.

Table 1.   Changes in pathogen biology under changing atmospheric CO2 composition or temperature
PathogenHostEffectReference
Podosphaera (Sphaerotheca) pannosaRoseHigh CO2 reduced sporulationVolk, 1931 [cited in Manning & Tiedemann, 1995]
Cladosporium fulvumTomatoHigh CO2 enhanced sporulationVolk, 1931 [cited in Manning & Tiedemann, 1995]
Blumeria graminisBarleyHigh CO2 reduced germination and penetration; increased growth rate and fecundityHibberd et al. (1996)
Puccinia reconditaWheatNo difference between CO2 in leaf area covered by pustulesTiedemann & Firsching (2000)
Phytophthora parasiticaTomatoHigh CO2 increased pathogen biomass in host tissueJwa & Walling (2001)
Puccinia striiformisWheatHigh CO2 did not influence fecundity in either resistant or susceptible varietyChakraborty et al. (2010)
Colletotrichum gloeosporioidesStylosanthesHigh CO2 delayed germtube growth and appressoria; no change in latent period but increased fecundityChakraborty et al. (2000)
Fusarium pseudograminearumWheatHigh CO2 increased pathogen biomass in host tissueMelloy et al. (2010)
Maravalia cryptostegiaeRubber vineHigh CO2 extended latent period, reduced pustules per leaf but increased fecundityS. Chakraborty, M. Weinert & J. Brown, unpublished data
Barley yellow dwarf virusOatsHigh CO2 increased persistence of infected plants to alter epidemiologyMalstrom & Field (1997)
Xanthomonas campestris pv. pelargoniiGeraniumHigh CO2 reduced pathogen numbers in host tissueJiao et al. (1999)
Puccinia striiformisWheatIncreasing temperature shortening latent periodMilus et al.(2009)

Modelling plant growth and architecture

Complex process-based models computing plant growth and development at leaf or organ level to project climate-change impacts from field to regional or global scale have been developed for wheat (Tubiello & Ewert, 2002) and rice (Matthews & Wassmann, 2003). These models simulate the canopy as a homogeneous medium and do not reflect canopy structure and development in space and time or changes resulting from interactions between plants and environment. Canopy architecture and dynamics are realistically expressed using Lindenmayer systems or L-systems, where architecture is defined by its constituent plant parts, topology and geometry in 3D (Room et al., 1996; Prusinkiewicz, 2004). The integration of physiology with these architecture models produces functional-structural plant models, or ‘virtual plants’, representing both function and structure of plants in 3D (Hanan & Prusinkiewicz, 2008). Among examples of the use of ‘virtual plants’ in climate change, Chen et al. (1997) modelled the effects of elevated CO2 on development, light interception and photosynthesis by combining 3D poplar tree and radiative transfer models; Drouet & Pagès (2007) modelled growth, architecture and carbon nitrogen allocation using the Unified Modelling Language approach; and Pachepsky et al. (2004) modelled the effects of temperature on soybean growth and development using an open parametric L-system model.

Modelling pathogen dispersal

Wind dispersal of fungal spores in the canopy has been modelled using empirical or mechanistic (physical) models. Empirical models are generally consistent with more mechanistic models of dispersal and can accurately characterize observed gradients (Madden et al., 2007). For short-range wind dispersal, physical Lagrangian models, such as stochastic simulation models, have been used within a plant canopy or close to the ground (Aylor, 2002). Lagrangian scale models simulate the paths of individual air parcels or particles as a pseudo-random walk using turbulence statistics for the air flow, and simulate the trajectories of spores moving in plant canopies in response to turbulent motions of the atmosphere (Aylor, 2002). Eulerian advection-diffusion models or gradient transfer (K-theory) models simulate dispersal over distances far enough from the source such that turbulent eddies are small compared with the vertical width of the plume or diffusion cloud (Fitt & McCartney, 1986). These models cannot be applied in crops where local sources and sinks cause the concentration gradient to change rapidly (Fitt & McCartney, 1986).

Splash dispersal is difficult to model because of the complexity of the splash process, the effects of secondary splash and the difficulties in up-scaling mechanisms of simple splash events to the canopy level (Huber et al., 2006). Interaction of rain-splash dispersal and stem extension influence vertical movement in the canopy (Pielaat et al., 2002). Saint-Jean et al. (2004) developed a framework for modelling splash of water droplets in a 3D plant canopy using Monte Carlo integration. This approach, which considers the propagation, interception and transport of droplets, may offer insights on canopy structure effects on splash dispersal (Huber et al., 2006). Another approach has been the implementation of empirical models predicting splash in drop, which is a 3D rainfall interception model (Bassette & Bussiere, 2008).

Modelling disease development

Disease development has been modelled using various approaches, including growth curve analysis, analytical models using linked differential equations, and mechanistic simulation models (Madden et al., 2007). Often, models have been used to predict or forecast plant diseases for management interventions and/or to discover causal relationships. However, most of these models ignore canopy-level interactions.

In one study, a 3D ‘virtual plant’ model was modified to incorporate disease development within canopy structure using L-systems (Wilson & Chakraborty, 1998). A recent development has been the coupling of plant disease models with 3D architectural models. Examples include the linking of a septoria leaf blotch model with a wheat architectural model (Robert et al., 2008) and a powdery mildew model with a grapevine architectural model (Calonnec et al., 2008).

Modelling microclimate

Leaf wetness, temperature, wind and, to a lesser extent, radiation are important microclimatic variables that influence plant disease epidemics. Leaf wetness, formed by rain, irrigation, dew and guttation is difficult to characterize experimentally or to define physically because of canopy heterogeneity. Complex microclimatic models have been used to estimate dew or leaf wetness using energy balance equations. Physical leaf wetness simulation models developed for grapes (Dalla Marta et al., 2005) and adapted to other crops (Magarey et al., 2006), although accurate, can be too complex, site-specific and models are difficult to parameterize. In another approach the crop is divided into layers and microclimate is simulated by solving the energy balance in each layer (Norman, 1982).

More recently, microclimatic variables such as radiation (Chelle, 2005) have been integrated in 3D plant architecture models. Luquet et al. (2003) simulated temperature variability in a cotton crop using an energy balance model at leaf level based on a 3D description. The use of ‘phylloclimate’, which is the microclimate of individual plant organs (Chelle, 2005) and the drop rainfall distribution model (Bassette & Bussiere, 2005), are among other examples of canopy microclimate modelling. Despite these recent advances in 3D plant architecture models, links between canopy structure, microclimate and pathogen life cycles are still lacking.

Only a few models have considered climate-change effects on canopy microclimate. A Crop Micrometeorological Simulation System was used to simulate elevated CO2 influence on maize microclimate and physiology (Anda & Kocsis, 2008). The canopy response to short exposures to elevated CO2 was simulated by combining gas exchange at the single-leaf level with canopy structure and light penetration (Reynolds et al., 1992). An energy balance model was used to predict rises of 1·6–2·0°C in rice paddy water by 2081–2100, which could promote pests and diseases (Ohta & Kimura, 2007).

Case study: modelling Colletotrichum gloeosporioides dynamics within the plant canopy under elevated CO2

The tropical pasture legume S. scabra cv. Fitzroy was grown under ambient (350 p.p.m.) and elevated (700 p.p.m.) CO2 in a controlled-environment facility (CEF) for 13 weeks; plants were digitized weekly using a 3D sonic digitizer to determine rules of plant morphogenesis (Pangga, 2002) and modelled using L-systems by adapting an existing ‘virtual Fitzroy’ model (Wilson et al., 1999). Increased probabilities of branching and shoot development rates along the primary and secondary branches; higher secondary and tertiary internode lengths; and reduced axillary bud dormancy of secondary shoots were among changes at elevated CO2. Fitzroy plants at elevated CO2 had 35, 26, 25 and 44% higher shoot length, node number, leaf area and shoot biomass, respectively, than those at ambient CO2. The overall effect of elevated CO2 on plant height, internode length, number of leaves and growth habit can be clearly seen from the virtual Fitzroy model taken as a snapshot of 97-day-old plants (Fig. 1a). In summary, elevated CO2 increased leaf area and plant biomass as expected. However, such differences in plant growth and canopy size may already occur within the range of environments from tropical savannah to forest understorey where S. scabra is currently grown.

Figure 1.

 Simulated Stylosanthes scabra cv. Fitzroy ‘virtual plants’ at elevated (700 p.p.m.) and ambient (350 p.p.m.) CO2 (a) 14 weeks after planting and (b) incorporating anthracnose lesions at 14 weeks old (green = 0–1 lesions; blue = 2–5 lesions; red = >5 lesions).

The number of anthracnose disease lesions on Fitzroy was obtained from a separate CEF study, where plants grown at elevated and ambient CO2 for 12 weeks were artificially inoculated with C. gloeosporioides. After inoculation, the plants were placed in dew chambers inside their respective CO2 growth rooms for 48 h and incubated for another 10 days before disease assessment. Induced resistance at elevated CO2 was observed from the significant reduction in susceptible- and resistant-type lesions per leaf and infection efficiency as compared to ambient CO2 (Pangga et al., 2004a). Multiple linear regression analysis was used to derive relationships between lesion number and the number of nodes at the primary, secondary and tertiary branch levels, and equations were incorporated in models for ambient and elevated CO2. As young leaves are more susceptible than older leaves (Chakraborty et al., 1988), the distribution of lesions on primary, secondary and tertiary branches largely reflected leaf age. Elevated CO2 induces a level of resistance to anthracnose (Pangga et al., 2004a). This effect is clearly evident from a reduced number of lesions at elevated CO2 (Fig. 1b).

In a separate study, the dynamics of C. gloeosporioides under Fitzroy canopy microclimate at ambient and elevated CO2 was studied in the field during 1997–1999 (Pangga, 2002). Batches of Fitzroy grown at the two CO2 levels in the CEF for 12–14 weeks were exposed to C. gloeosporioides inoculum in the field for 48 h on different occasions. After exposure, plants were transported to a glasshouse at ambient CO2 and placed inside a dew chamber for 48 h at 25 ± 5°C and outside the dew chamber in the same glasshouse for another 10 days before disease and other assessments were made. There was no significant difference in lesion number per leaf between the two CO2 levels. However, lesions per leaf were almost always lower in plants from elevated CO2 than in those from ambient CO2. The induced resistance previously observed in elevated-CO2 plants did not persist after plants were removed from the high-CO2 environment and exposed to pathogen inoculum in the field; whereas, the enlarged canopy of elevated-CO2 plants trapped many more pathogen spores to produce twice as many lesions per plant (Pangga et al., 2004a).

In the same study, microclimatic variables, including temperature, relative humidity, radiation and wind speed, were monitored by placing electronic sensors within the canopy. Correlation and regression analyses were used to examine the relationships between anthracnose lesions on plants grown at elevated and ambient CO2, weather and inoculum variables (Tables 2 & 3). At both CO2 levels, lesion number per plant was positively correlated with mean hourly relative humidity (MHRH) and negatively correlated with solar radiation (SOLRAD). These findings are similar to previous research showing positive effects of moisture-related variables on this disease (Chakraborty & Billard, 1995) and negative effects of light intensity on C. graminicola causing anthracnose of maize (Hammerschmidt & Nicholson, 1977). By contrast, none of the weather variables significantly influenced lesions per leaf for plants grown at either CO2 level (Table 2).

Table 2.   Significant Pearson’s correlation coefficients between weather, source of inoculum and lesions per plant and per leaf [ln (n + 1)] caused by Colletotrichum gloeosporioides on Stylosanthes scabra cv. Fitzroy grown at elevated and ambient CO2 for 12–14 weeks and exposed to field conditions on eight separate occasions in 1997 and 1999
VariableLesions per plantLesions per leaf
Ambient CO2 (350 p.p.m.)Elevated CO2 (700 p.p.m.)Ambient CO2 (350 p.p.m.)Elevated CO2 (700 p.p.m.)
  1. Only significant correlations are shown: ns, not significant at > 0·05; *significant at ≤ 0·05.

  2. BKGND: disease severity of background plants; MHRH: mean hourly relative humidity; RINT: rain intensity; SOLRAD: solar radiation; AGE: days after planting.

BKGNDns0·71*nsns
MHRH0·82*0·71*nsns
SOLRAD−0·72*−0·73*nsns
RINT0·79*nsnsns
AGEnsnsns−0·76*
Table 3.   Quantitative relationship between lesion number and microclimatic variables following exposure of Stylosanthes scabra cv. Fitzroy plants from 350 and 700 p.p.m. CO2 to Colletotrichum gloeosporioides inoculum in the field on eight occasions
Dependent variableCO2 levelRegression equationAdjusted R2
  1. *Significant at ≤ 0·001.

  2. MHRH: mean hourly relative humidity; WSRH95: wind speed when relative humidity >95%; BKGND: disease severity of background plants used as the source of inoculum; SOLRAD: solar radiation; HRTEMP: hours of mean hourly air temperature between 25 and 30°C.

  3. aThe standard error of the estimated partial regression coefficient in parenthesis.

Lesion number per plant350 p.p.m.Y = −0·27 (±3·38)a + 0·35 (±0·04) MHRH + 1·30 (±0·23) WSRH950·94*
700 p.p.m.Y = 0·24 (±0·02) BKGND − 0·0003 (±0·00006) SOLRAD + 0·26 (±0·06) HRTEMP0·97*

Results from stepwise multiple linear regression analysis showed that different microclimatic factors influenced C. gloeosporioides dispersal and infection of plants from ambient and elevated CO2 (Table 3), indicating a definite influence of plant canopy. Interestingly, in this and other studies, moisture-related variables influenced dispersal (Pangga et al., 2004b) and infection by C. gloeosporioides (Chakraborty et al., 2000) at ambient CO2, but not at elevated CO2 (Table 3). One possible explanation is that canopy moisture level may have remained sufficiently high in high-CO2 plants, even when atmospheric relative humidity became low. At elevated CO2 lesion number increased with decreasing solar radiation and increasing hours of mean hourly air temperature between 25 and 30°C. Anthracnose development under elevated CO2 results from a balance between increased canopy size that traps more spores and enhanced host resistance, which interacts with the canopy microclimate differently from ambient CO2 (Pangga, 2002; Pangga et al., 2004a). These modified interactions, together with changes in the host–pathogen interaction, such as increased fecundity and delayed and reduced infection at elevated CO2 (Chakraborty et al., 2000), can potentially influence pathogen evolution.

Accelerated pathogen evolution

An enlarged canopy and increased fecundity of some pathogens at elevated CO2 can have important implications from epidemiological and evolutionary perspectives. High relative humidity, reduced wind flow and reduced penetration of solar radiation within enlarged and dense crop canopies are potentially more conducive to disease development by many fungal and bacterial plant pathogens. However, splash dispersal of some pathogens may be impeded inside a dense canopy. A large canopy also contains many more infection sites and can physically trap many more pathogen propagules (Pangga et al., 2004a). If a combination of these canopy-related changes increase the number of infections, this leads to an increase in pathogen population size with implications for the amount of inoculum arising from an infected crop. The population size gets a further boost when fecundity increases under elevated CO2, as seen with some pathogens (Table 1). With polycyclic epidemics, even a modest increase in fecundity can lead to very large populations.

Mutation, selection and other forces can act on the large pathogen population to potentially accelerate evolution, leading to more aggressive/virulent races. Studies have shown genetic adaptation in pathogens under elevated CO2 (Chakraborty & Datta, 2003), elevated temperature (Gijzen et al., 1996) and altered precipitation (Travers et al., 2007). In a separate study, two C. gloeosporioides strains of different aggressiveness were artificially inoculated over 25 sequential infection cycles on two S. scabra cultivars with different level of resistance grown at ambient (350 p.p.m.) and elevated (700 p.p.m.) CO2. Pathogen aggressiveness increased at both CO2 concentrations on both cultivars but the genetic fingerprint changed for only some strain–cultivar combination after the 25 cycles. Although fecundity of the aggressive strain increased significantly over the less aggressive strain at elevated CO2, the study ignored the influence of changing fecundity by inoculating an equal number of spores of each strain for each infection cycle (Chakraborty & Datta, 2003).

However, the evolution of plant pathogens is not a simple matter of population size, and reproductive fitness and aggressiveness of a pathogen strain are not always correlated (Zhan et al., 2002). Often, necrotrophs with broad host range may reach high levels of pathogenicity on some hosts, whilst maintaining population size on the sympatric host. For instance, Phytophthora ramorum has high fecundity on the tolerant Pacific Madrone (Arbutus menziesii), but only has limited fecundity on oaks, which it devastates by sudden oak death epidemics (Rizzo & Garbelotto, 2003). Local host–pathogen adaptation is difficult to detect when gene flow obscures evolution (Parker & Gilbert, 2004). This can happen in many fungal pathogens with large population size and long-distance dispersal (Aylor, 2002). The wheat stem rust pathotype Ug99 and its variants (Pretorius et al., 2000) are recent examples of new races travelling large geographical distances to infect previously resistant cultivars without local host–pathogen adaptation. Predicting pathogen evolution is a challenging problem and many of the theoretical models on the evolution of virulence have considered only the simplest case of infection of a host by a single pathogen strain (Zhan et al., 2002). McDonald & Linde (2002) proposed combining estimates of gene flow, effective population size and reproduction/mating system to place pathogens into different categories of evolutionary potential.

The host plays a key role in the evolution of virulence controlling the rate of evolution, and nowhere is this more obvious than in agricultural systems. In natural ecosystems, pathogens may adopt different strategies, combining virulence and other fitness traits to cope with host and other environmental elements, but many maintain intermediate levels of overall pathogenicity (Lenski & May, 1994). In agriculture, new cultivars developed through selection and breeding and deployed over large areas put pathogens under constant selection pressure. This alters both dynamics and evolution of virulence in agriculture, which has been termed ‘man-guided’ evolution in cereal rusts (Johnson, 1961), but rust pathogens do not respond to all rust resistance genes in the same way. In the wheat stem rust pathogen P. graminis f. sp. tritici, genes corresponding to resistance genes Sr5, Sr15 and Sr21 have very high mutation rates, whereas those matching genes like Sr13, Sr24 and Sr27 rarely mutated (reviewed by Leach et al., 2001). What drives this durability of some resistance genes is not well understood.

Plant pathogenic fungi are highly plastic organisms generating variation via a myriad of mechanisms including mutation, selection, sexual and parasexual reproduction. Selection for increased aggressiveness in fungi can happen after only a few asexual cycles (Zhan et al., 2002) of biotrophic (Newton & McGurk, 1991) or necrotrophic pathogen (Wang et al., 2008). Recent literature indicates horizontal transfer of genes between species as a novel mechanism that generates variation in fungi. Horizontal transfers most likely occur when both fungi colonize the same host, either as a saprotroph or a pathogen, and examples include the transfer of virulence genes between wheat pathogens Pyrenophora tritici-repentis and Stagonospora nodorum (Friesen et al., 2006) and the transfer of mobile pathogenicity chromosomes between Fusarium species (Ma et al., 2010). These transfers can have disastrous economic consequences and the transfer of key pathogenicity genes from S. nodorum has allowed P. tritici-repentis, with enhanced survival in stubble under no-till agriculture, to become the major pathogen of wheat in Australia.

Gaps in knowledge and the way forward

There is overall paucity of knowledge in all areas of research on climate change and plant diseases, and pathogen dynamics and evolution are not exceptions. Of the handful of experimental studies on the influence of atmospheric composition or temperature on host–pathogen interactions, the majority have ignored canopy-level interactions, host–pathogen adaptation or evolution. Given the lack of data, modelling studies have and will continue to offer insights into potential impacts of climate change, mostly on the changing geographical distribution of pathogens. Experimental approaches are essential to understanding the mechanisms of host–pathogen interactions, offering answers to specific questions and hypothesis-testing to enhance adaptation of agriculture to climate change. Modelling studies, with their ability to handle large number of interacting host, pathogen and environmental/climatic variables, offer a different perspective, where extrapolation of results from experimental studies are difficult. Coordinated and sustained research effort using both approaches is needed to fill knowledge-gap in all areas of plant pathology research dealing with a changing climate, and the dynamics and evolution of plant pathogens are no exceptions.

Baseline quantitative information

Not surprisingly, the need for experimental studies on host–pathogen biology has been recognized by most commentators and authors dealing with plant diseases under changing climate (Coakley et al., 1999; Garrett et al., 2006; Chakraborty et al., 2008; Jeger & Pautasso, 2008). Economic impact is the currency needed to communicate key messages from the plant protection community to the industry and policy makers, and new baseline information on potential climate-change impacts or new adaptation strategies needs to be translated into economic significance. Data on the economic significance of climate change to plant diseases are lacking, except for commentaries linking plant diseases to food security (Mahmuti et al., 2009; Chakraborty & Newton, 2011) or the influence of diseases under rising temperature on wheat yield in the Eastern Gangetic Plains (Sharma et al., 2007).

Detailed knowledge of changes in pathogen life stages and disease cycles is the key to developing process-based models that can be applied to fine-tune and evaluate disease management strategies. For instance, changes in farming systems as a result of altered geographical distribution of hosts and diseases may bring together pathogens and new alternative hosts. Knowledge of altered biological and epidemiological processes will allow, among other things, assessment of climate-change impact on diseases (Goudriaan & Zadoks, 1995) and efficacy of management strategies, and may point to new opportunities and options for disease management (Chakraborty et al., 2010).

Data on the influence of single factors on pathogen dynamics are limited in their ability to predict the consequences of climate change on plant diseases (Ziska & Runion, 2007). Studies of host–pathogen dynamics have to embrace interaction and feedback at the ecosystem level. For example, changing species composition and size structure of forest pathogens can influence CO2 flux and heat transfer, creating climate feedbacks (Ayres & Lombardero, 2000). Interacting abiotic factors, such as soil nutrients and air pollutants, and biotic factors, such as diseases, insects and weeds, should be a part of overall assessments (Tubiello & Ewert, 2002; Perarnaud et al., 2005). Nevertheless, detailed understanding of physiological and/or molecular mechanisms under specific combinations of selected elements of the physical environment, such as drought or heat stress, may be useful to address specific aims, such the development of adapted plant cultivars, and growth chambers may be adequate for these studies. Once developed, their performance and durability have to be confirmed under more realistic field conditions, such as free to air CO2 enrichment (FACE) facilities. On the other hand, studies to understand and manipulate interactions at the farming-system level between many interacting biotic and biotic factors can only be done at FACE or other similar field facilities.

Pathogen populations will naturally adapt to changing climate in both natural and agricultural systems, but the host population will adapt to keep pace only in natural ecosystems, although the selection pressure resulting from climate change can decrease genetic diversity and reduce the plant’s ability to resist pest and disease outbreaks and extreme climates (Jump & Peñuelas, 2005). In agriculture, where improved cultivars of genetically uniform annual crops are grown over large tracts of land, new cultivars will have to be developed to increase their adaptation. Knowledge of pathogen evolution to anticipate the appearance of new pathotypes under a changing climate will allow the international plant breeding community to respond accordingly. Instances of rapid evolution of virulent pathogen races with novel virulence are generally well documented in biotrophic pathogens with a ‘gene-for-gene’ specificity with the host plant (Leach et al., 2001) and experimental studies are needed to determine whether climate change will significantly alter evolutionary response of these pathogens. Increasing fecundity with progressive infection cycles and accelerated evolution under elevated CO2 has been shown in a necrotrophic pathogen (Chakraborty & Datta, 2003). Understanding and anticipating evolutionary processes in host–pathogen interactions under climate change will be among the most fruitful areas of research and can change the magnitude of climate change impacts on plant diseases (Garrett et al., 2006).

Uncertainty, scale and approaches

Although modern climate models provide credible quantitative estimates of future climate change at continental scales and above (Pachauri & Reisinger, 2007), uncertainties remain in regional-level predictions. Cumulative uncertainty resulting from different socioeconomic assumptions and concentrations of greenhouse gases enlarges as the simulations are downscaled from global to local scale (Bergant et al., 2006) and uncertainty in climate models has a larger impact on yield than crop-model uncertainty. Multiple climate models (Lobell et al., 2006) or multiple simulations from a single model (Challinor et al., 2005) have been used to reduce uncertainty in some crop-yield predictions. A comparison of different disease models (Schaafsma & Hooker, 2007) to quantify and understand uncertainties can help to discover the domain over which a validated model may be properly used.

To develop credible and useful climate-change predictions, plant disease models themselves must capture a thorough quantitative understanding of disease epidemiology and their reliability in disease forecasting must be proven through rigorous testing and validation (Shaw, 2009). Consequently, predictions of climate-change impacts on plant diseases for various emission scenarios (Pachauri & Reisinger, 2007) must align with observed historical changes when applied retrospectively. However, long-term datasets on climate and plant diseases without the confounding effects of management or biological factors, essential for such comparative studies, are lacking (Jeger & Pautasso, 2008; Shaw, 2009).

Nonlinearity of temperature response curves for pathogens also leads to substantial differences in growth at constant versus fluctuating temperatures (Scherm & van Bruggen, 1994). Thus, inaccurate predictions can be made if low resolution data averages of Global Circulation Model outputs are used in disease models. One solution is to fit nonlinear models directly or to integrate the rate function at small time intervals (Xu, 1996).

Canopy-level interactions are relevant to many epidemiological processes and disease management and coupling of climate projections to paddock/canopy-level processes pose significant challenges. The problem of scale incompatibility is not unique to plant diseases and for instance, more than one approach is used to address this issue in crop models (Baron et al., 2005). Another approach is by ‘integration’ models, such as the combination of canopy radiation interception and energy budgets to compute microclimate temperature from regional climatic data and simulate the effect on insect development within the canopy (Pincebourde et al., 2007). ‘Virtual plants’ offer an interface to couple canopy-level interactions between crop growth and disease models to examine pathogen dynamics under current and future climate scenarios (Pangga, 2002). For trees and other plants with large canopies, virtual plants offer new opportunities to identify disease ‘hot spots’ within a canopy (Wilson & Chakraborty, 1998) to target and improve application of management options such as fungicides (Dorr et al., 2008). Linked 3D plant architecture and plant disease models (Calonnec et al., 2008; Robert et al., 2008), models incorporating splash dispersal of water in 3D canopies (Saint-Jean et al., 2004; Bassette & Bussiere, 2005) and improvements in simulating microclimate at the plant organ level (Chelle, 2005) have further enhanced the usefulness of virtual plant models for canopy-level interactions. However, significant challenges including the dispersal of pathogen propagules and the linking of canopy microclimate still remain.

As a tool, virtual plants can potentially serve many areas of plant pathology research and training (Wilson & Chakraborty, 1998). They offer a one-step tool to integrate growth and development of host and pathogen in a physical environment to help analyse, understand and forecast plant diseases. With the focus on the understanding of host–pathogen interaction at the canopy level, virtual plants can help improve pest control measures, such as better targeted fungicide sprays to optimize applications. One of the obvious uses can be in explaining and communicating complex interactions to students and other target audiences. As an on-farm decision-support tool, visual outputs may be simpler than the numeric outputs of traditional models.

Acknowledgements

Matthew Weinert, Paul Melloy, Rosanna Powell and Ross Perrott provided technical and other assistance. Co-investment from the Cooperative Research Centre for Tropical Plant Pathology, Cooperative Research Centre for National Plant Biosecurity and CSIRO Plant Industry in this research is gratefully acknowledged.

Ancillary