SEARCH

SEARCH BY CITATION

Keywords:

  • co-evolution;
  • disease;
  • epidemiology;
  • longitudinal study;
  • metapopulation;
  • plant–pathogen;
  • spatial pattern

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results and discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

1. A long-term study (19 years) of a host–pathogen metapopulation involving 133–220 separate populations of the wild plant Filipendula ulmaria and its rust pathogen Triphragmium ulmariae shows marked changes in the occurrence (32–55% demes) and severity of disease and rates of extinction and re-establishment of individual populations (0.006–0.174 and 0.030–0.195 per annum, respectively) over time.

2. Modelling of the spatio-temporal dynamics of disease demonstrated year-to-year changes associated with a range of different environmental features, but also more consistent, longer-term patterns influenced by a complex suite of factors.

3. Both the level of disease and its spatial location varied through time and generated a changing pattern of selective pressure across the metapopulation.

4.Synthesis. Our results suggest that co-evolutionary hot spots and cold spots can be highly dynamic within metapopulations, thereby fuelling the co-evolutionary process even more than previously suspected.


Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results and discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

The dynamics of co-evolutionary systems are greatly affected by spatial scale and time. However, it took the development of metapopulation theory (Levins 1970; Gilpin & Hanski 1991; Hanski 1999) and the geographic mosaic theory of co-evolution (Thompson 1994, 2005) to provide a formal coherent theoretical framework describing the dynamics of co-evolutionary interactions. In this framework, organisms are dispersed in small, somewhat isolated and often ephemeral populations. These demes are distributed across a heterogeneous landscape where local selective pressures over the whole metapopulation or for any deme can differ over time. Isolation and interaction of specific life-history features of both players in the interaction lead to marked changes in genetic drift, migration, extinction and recolonization, and as a consequence, the evolutionary trajectory of the whole metapopulation may be quite distinct from that of its individual component demes.

Over the last decade, computer simulation studies and empirical assessments of interacting species have provided substantial corroborative support for the geographic mosaic theory of co-evolution. At the whole metapopulation scale, simulation models have reinforced the view that the evolutionary trajectory of interactions in the metapopulation as a whole may be quite different to that occurring in individual demes (Thrall & Burdon 1999; Gomulkiewicz et al. 2000). Empirical studies of insect–plant (Greya–HeucheraThompson & Merg, 2008; Eurosta–SolidagoCraig, Itami & Horner, 2007) and fungal–plant (Podosphaera–PlantagoLaine, 2006; Melampsora–LinumBurdon & Thrall, 2000 and Melampsora–HeptolinonSpringer, 2007) associations highlight the complexity and show that markedly different outcomes can occur in adjacent populations.

In fungal pathogen–plant associations, spatial variation in disease resistance from different patches within single populations to different demes within single metapopulations and beyond has been extensively documented (Laine et al. 2011). Despite this, the link between these outcomes and putative variation in selective pressures imposed by pathogens is very poorly documented. An important link in that chain is demonstrating how pathogen incidence and selection intensity vary in space and time over long time periods. Although a few data sets exist (Ericson, Burdon & Müller 1999; Laine & Hanski 2006), these are as yet either of insufficient duration or size as to follow the outcome of long-term dynamics or to assess Thompson’s view that the geographic mosaic of co-evolution is driven by the occurrence of ‘hot’ and ‘cold’ spots in which the intensity of interactions varies.

A long-term study of the interaction occurring between the host-specific rust pathogen Triphragmium ulmariae and its herbaceous perennial host plant Filipendula ulmaria provides an exception to this temporal shortcoming. Over a 19-year period, we have built up an extensive data set that records the incidence, prevalence and severity of disease in a large island metapopulation. Using the initial 11 years of this data set (1990–2000 inclusive), we demonstrated a high degree of spatial complexity of interactions occurring among populations (Smith, Ericson & Burdon 2003). Since that analysis, the period covered by this data set has extended (1990–2008), and during this time, the number of populations has also increased.

Here, we use this expanded data set to undertake a deeper assessment of the changing spatial and temporal dynamics of this metapopulation. Specifically, we wanted to know whether:

  • 1
     pathogen dynamics varies over time and how this affects the overall epidemiological pattern of interaction;
  • 2
     an observed doubling of the pathogen extinction and recolonization rate had a noticeable affect on disease dynamics;
  • 3
     the longer-time sequence of data provides a significant increase in the resolving power of our models; and
  • 4
     hot and cold spots of pathogen activity shift over time and space.

Materials and methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results and discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Study area

The pathogen–host plant metapopulation used in this study occurs on 70 islands (plus immediate mainland shore) of the Skeppsvik archipelago (covering 5.0 × 6.4 km) in the Gulf of Bothnia, northern Sweden (63° 44–48′N, 20° 34–40′E). Demes in the metapopulation showed a complex distribution mostly occurring on islands in four parallel glacially created ‘drumlin’ island chains with multiple populations often present on individual, larger islands (see Burdon, Ericson & Müller 1995, for a map of the islands). The location of each population was recorded by taking coordinates from detailed maps (Ekonomisk Karta över Sverige, Västerbottens län) of the area using position 7080/20K 6 h Tärnögen as the N/W 0/0 coordinate position.

Host–pathogen interaction

Filipendula ulmaria L. (Rosaceae) is a herbaceous, perennial dicotyledonous species that reproduces by seed and slow lateral spread of a rhizomatous rootstock. In the study area, recruitment is largely restricted to years when prolonged low water levels coincide with seed germination (Ericson 1981). Adult plants may produce large quantities of seed that float, thus aiding dispersal. In this system, the fungus Triphragmium ulmariae (DC.) Link (Sphaerophragmiaceae) is an autoecious, macrocyclic rust pathogen that is restricted to F. ulmaria (Wilson & Henderson 1966). The pathogen survives the winter exclusively as teliospores found on dead leaf and stem fragments, which are then dispersed by water movement during the autumn and winter. In the following spring, teliospores germinate producing basidiospores that infect host plants in the immediate vicinity, thereby initiating a process of sexual recombination and the development of aecial infections on petioles and abaxial veins of leaves. The aecial generation is followed by one or more uredial (asexual) generations that produce wind-dispersed urediospores. Towards the end of the summer, the pathogen’s life cycle is completed as uredia switch over to the production of teliospores.

Data collection

In 1990, 133 populations of F. ulmaria were identified on a total of 55 islands in the Skeppsvik archipelago and four areas of the immediately adjacent mainland. Up to seven distinct populations occurred on some of the larger islands, whereas single populations were found on many of the smaller ones. The size of each host population was estimated by counting numbers in representative areas and then adjusting these scores for the total area covered by the population. Each population was visited in mid-July when the disease incidence was determined by screening a sample of randomly chosen plants (minimum number = 50) for the presence of pustules. In very small populations (<80 plants), all individuals were examined; in larger ones, this was not practical, and a random sample determined by the numerical size and spatial distribution of the population was assessed (range of 50–200 per population). If disease was not found on these plants, a further intensive search determined whether the pathogen was present but at very low frequency.

Disease surveys were repeated yearly (1990–2008 inclusive) whereas host population size estimates were made in all years except 1991 and 1993. Over this 19-year period, we continued to monitor potential sites that were unoccupied throughout the archipelago. Due to isostatic rebound, land in the archipelago is rising at about 0.9 cm year−1 (Ekman 1996), resulting in the appearance of new islands and changing shore topographies. As a result of this and storm effects, the number of individual demes of F. ulmaria has increased over the 19-year survey period. New demes have established on both isolated parts of islands with existing populations as well as on previously un-colonized islands. As a consequence, additional host populations were identified such that by 2008, a total of 220 populations on 70 islands were being followed.

The scores of the presence or absence of disease on a per-plant basis were used to determine disease incidence (presence/absence) and prevalence (% plants infected) at the population level, whereas the severity of infection for the population as a whole was determined by taking the mean percentage of leaf area infected across the sampled plants. These data were used to generate an overall picture of the epidemiology of T. ulmariae over the 19-year period.

In addition to these surveys, we also collected information about physical variables that, on the basis of our understanding of this system, we believe influence the dispersal process. The drumlin lines of islands define four major island chains separated by deeper water channels or ‘water courses’. Populations were allocated to (i) individual island chains on the basis of their position, the depth of water between them, and islands in adjacent chains; and (ii) water courses by their aspect on individual islands, and hence, the water course to which they had greatest exposure. The shallower water and shorelines of the islands create lateral barriers to the dispersal of telia and teliospores on flotsam, thus increasing contagion among populations on the same island. Similarly, the arrangement of islands could enhance contagion among populations on the same island chain, or on populations on shorelines that are exposed to the same water course.

The predominant winds in the archipelago in autumn and winter flow, respectively, from south to north or from north to south. These winds, coupled with the inflow of the Sävarån River at the north-western end of the archipelago, water level fluctuations (amplitude 2.5 m) in connection with northern or southern storms, and differences in the onset of ice formation, also create a general environmental gradient in deposition of flotsam, and hence the dispersal of telia and teliospores on the flotsam. As a rough approximation of this process, we assigned populations to one of three weather zones. Finally, the islands in the outer part of the archipelago are surrounded by open waters, their shores being exposed to wave wash, especially storm waves in autumn through early winter. This contrasts with the sheltered conditions on islands in the inner part of the archipelago or on the immediately adjacent mainland. One consequence of the greater disturbance in the outermost part of the archipelago is that plant litter (as well as the telia) faces a greater likelihood of being removed.

Modelling

To investigate the spatio-temporal patterns of disease incidence in this extended data set, we followed the same procedure as used previously (for details, see Smith, Ericson & Burdon 2003). In essence, we identified several factors that may play a role in the persistence of the pathogen and then followed an approach that uses model building, model fitting using maximum likelihood and model selection using Akaike’s Information Criterion (AIC; Burnham & Anderson 1999) (for details, see Smith, Ericson & Burdon 2003 and Appendix S1 in Supporting Information). The factors assessed were chosen on the basis of our knowledge of the system and the likelihood that they may play a role in persistence of the pathogen.

Each model, denoted by a symbol and list of factors (Ψ(…)), was a complex ‘hypothesis’ about the process of contagion underlying the patterns of disease incidence. Seven simple one-factor models were the building blocks for more complex models:

  • Ψ(N): core-satellite metapopulations in which larger host populations were more likely to become infected;

  • Ψ(D): structured populations in which more severely infected populations were more likely to become re-infected;

  • Ψ(K): spatial metapopulations in which pathogen dispersal varied by distance;

  • Ψ(I): spatial metapopulations with hierarchical structure because of higher connectivity within an island;

  • Ψ(C): spatial metapopulations with geomorphologically defined hierarchical structure generating higher connectivity within the island chains;

  • Ψ(W): spatial metapopulations with geomorphologically defined hierarchical structure of watercourses defined by the drumlin lines; and finally,

  • Ψ(Z): spatial metapopulations with structure introduced by weather zones that describe the level of exposure to factors like increasing wave action.

The highest-ranked one-factor model, Ψ(N), was extended to include each one of the other factors. The best two-, three- and four-factor models were extended to develop a set of increasingly complex three-, four- and five-factor models, as well as all 6-factor models and the fully complicated model.

Results and discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results and discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Overall similarities in general host population and pathogen epidemic behaviour between the 1990–2000 and 2001–2008 periods mask a greater instability in both host and pathogen populations (Fig. 1a). Since 2000, the percentage of host populations infected has remained in the 40–55% band, but there was considerably greater year-to-year variability in the last 8 years compared with the first 11, with disease levels in 2002 and 2005 being considerably below the 8-year average (Fig. 1a). Host population sizes have shown similar drops.

image

Figure 1.  Ecology and epidemiology of the Triphragmium ulmariae–Filipendula ulmaria association over a 19-year period. (a) Mean size of individual host populations (squares) and % host populations in which the pathogen was present (triangles) over time; (b) rate of change in the number of pathogen populations that went extinct (squares), that were newly colonized (diamonds) and the total change (triangles) across years; (c) percentage of host populations in which the pathogen was present for varying periods of time.

Download figure to PowerPoint

One of the most noticeable differences between the first and second census periods was a marked increase in the rate of extinction and re-establishment of individual pathogen populations (Fig. 1b). Between 1990–2000 and 2001–2008, extinction rates jumped from 0.030 to 0.081 year−1 (P = 0.009), and re-establishment rates rose from 0.056 to 0.082 year−1 (P = 0.052). Not surprisingly, extinction rates recorded in any given year (n) correlated well with establishment rates in the following year (n + 1) (adj r2 = 0.481; P < 0.005), probably reflecting the increased number of disease-free host populations in which new pathogen appearances can be observed. Notwithstanding this, conditions favouring pathogen population extinction and recolonization are very different, although all factors ultimately relate back to the interaction of aspects of the pathogen’s life history with the particular environmental circumstances occurring in the archipelago. Thus, although host population size and site exposure are factors of importance for both extinction and recolonization, extinction events may often reflect a failure in the transition from aecial to uredial infections in dry summers. Recolonization, on the other hand, is strongly influenced by the deposition of telial-infected material on island shores, which in turn is driven by a range of environmental factors (especially water level and winter storm intensity) interacting with the spatial distribution of the islands.

The persistence of pathogen demes at any given site varied greatly. Although the largest fraction of sites were either continuously occupied or empty (disease-free, Fig. 1c), in line with the increased frequency of disease in the metapopulation as a whole, an increasing proportion of all sites were occupied for at least some years in the 2001–08 period. Sites that were either continuously occupied or empty were typically among the largest and smallest host populations, respectively.

As noted above, the frequency of host populations in which the pathogen was present varied over time (Fig. 1a). However, more subtly and reflecting the semi-independent behaviour of pathogen–host interactions in individual demes in this metapopulation, even when disease was present, disease severity fluctuated markedly between individual populations across the archipelago (Fig. 2a,c). Indeed, even on the one island (no. 15, Fig. 2b,c) different populations frequently showed diverse disease status with some demes being disease-free whereas other closely adjacent ones were moderately infected. Such seemingly unpredictable variation in disease severity with and among populations within and across seasons is a feature of all natural host plant–pathogen interactions studied to date (Ericson, Burdon & Müller 1999; Thrall, Burdon & Bock 2001; Laine & Hanski 2006). This reflects the patchy nature of pathogen survival and dispersal, and the semi-localized nature of the interaction between host, pathogen and environment in individual demes that generates different pathogen development dynamics.

image

Figure 2.  Temporal variation in disease severity patterns in individual host–pathogen demes. (a) Disease severity in 10 Filipendula ulmaria populations distributed across the Skeppsvik archipelago (marked in red on map); (b) disease severity in nine F. ulmaria populations occurring on a single island (no. 15; red arrow, map) in the Skeppsvik archipelago; (c) map of populations of F. ulmaria in the Skeppsvik archipelago showing islands.

Download figure to PowerPoint

Here, in focusing on levels of disease severity, we do not provide unequivocal evidence for co-evolutionary ‘hot spots’ (defined as regions in which reciprocal selection occurs; Gomulkiewicz et al. 2007). However, an earlier study of this pathogen–host association has already demonstrated that disease pressure leads to selective death of more susceptible individuals and that, within the Skeppsvik metapopulation, different host demes show different levels of resistance to T. ulmariae (Ericson, Burdon & Müller 2002). Given these observations, and the fact that the probability of death in that study was correlated with disease severity, we use here the risk of infection as an index for selection intensity.

Pathogen presence in the existing spatial mosaic of host populations was complex as both host and pathogen populations varied over time. Because we were particularly interested in temporal changes, we applied the modelling–fitting procedure to select a best overall model for shorter blocks of time, to evaluate what factors would have been important in 2-, 3-, 5- and 10-year studies, and hence whether there were marked differences that could reflect changing selection over time. Examination of these overlapping assessments demonstrated differences between the important factors operating during individual years and the longer-term behaviour of the metapopulation (Fig. 3). Over 2-year snapshots of the metapopulation, the shortest interval possible, simpler models were the most efficient in explaining pathogen dynamics but the identity of the most important factor changed from year-to-year. Core-satellite dynamics were favoured in all years [(Ψ(N,…)], but these were combined with different factors in different years: weather zone was important in 1997–1998 and 2000–2001 [Ψ(NZ)], islands in 1990–1991 [Ψ(NI)], severe disease in 2004–2005 [Ψ(ND)], islands and severe disease in 2006–2007 [Ψ(NDI)].

image

Figure 3.  The effectiveness of different models at explaining pathogen dynamics when assessed against epidemiological data of overlapping time periods of differing duration. The models of best fit are those with the smallest ΔAIC (Akaike Information Criterion) values and are identified by cells of increasing darkness (N = host population size; D = disease severity; I = connectivity between populations within islands; K = dispersal distance between populations; C = island chain; W = watercourses between island chains; Z = weather zone within the archipelago).

Download figure to PowerPoint

However, as has been seen in simulation models (Thrall & Burdon 1999) and hinted at in real-world host population resistance structures (Burdon & Thompson 1995), such snapshots make up only a small part of long-term co-evolutionary patterns. The extended analysis of the dynamics of the Triphragmium–Filipendula host–pathogen metapopulation made possible by the addition of a further 8 years of census data demonstrates the overall stability and resilience of this interaction in which multiple environmental and life-history factors play a role. Thus, for longer simulated studies, a small set of multifactor models was consistently among the best at approximating observed pathogen dynamics (Fig. 3, 5- and 10-year studies). The overall best model identified for the 1990–2000 census period (NDIKC) was also the best for the second census period (2001–2008) and for the entire 19-year data run. However, the analysis of a sequence of three overlapping 9-year runs (not shown) found that slightly different models best explained epidemiological patterns in the different periods. Surprisingly, although host population size, and pathogen extinction and recolonization rates showed greater stochasticity during the second census period (2001–2008), this appears to have had little effect on the factors influencing the metapopulation as a whole. This probably reflects the dominating impact that the fixed spatial structure of the island chains imposes both as a factor in its own right and also through its effect on the watercourses and general environmental gradient banding.

Stability in the Skeppsvik Triphragmium–Filipendula metapopulation is a balance of biological and geographical features. A perennial host, effective off-season survival of the pathogen, and the mild nature of their interaction characterize the biology. Infections are discrete, disease severity is generally low and, hence, the disease generally fails to kill individual plants, other than seedlings (Ericson, Burdon & Müller 2002). Against this biological overlay, the geography (island chains, watercourses) and a marked environmental gradient across the archipelago together provide strong forcing to the system. More variable and unpredictable host–pathogen metapopulations are likely when there is stronger endogenous instability because of host and pathogen life-history traits, and when features of the physical environment have relatively less long-term strength and predictability. Unfortunately, the distinctly different nature (e.g. physical structure of the environmental matrix; nature of the pathogens and their life histories) of the only other long-term host–pathogen associations (Silene–Microbotryum, Antonovics 2004; Plantago–Podosphaera, Laine & Hanski 2006; Soubeyrand et al. 2009) make comparisons inappropriate.

The models predict the risk of infection in a putative population as a measure of selection intensity and allow us to map and identify ‘selective hot spots’. Using the best overall model, NDIKC, we produced a 19-year mean selection intensity map. The intensity of selection ranged over several orders of magnitude with populations to the south and west showing significantly lower average selection intensity scores (Fig. 4a). The overall average can be contrasted with selection intensity maps generated by the best model from three specific snapshots (1990–1991, 2000–2001, 2007–2008). In 2000–2001 (Fig. 4c), there was a markedly higher probability of infection at the outer edges of the archipelago where exposure to the wave action was greater [Ψ(NZ)]. In 2007–2008 (Fig. 4d), selection intensity varied among populations on the different north–south running island chains [Ψ(NDC)]. In 1990–1991, the patterns were strongest within the populations on the same island [Ψ(NI)], with high local variation in selection intensity across the metapopulation (Fig. 4b).

image

Figure 4.  Temporal and spatial dynamics in selection intensity across the metapopulation (defined as the probability of infection in a new population at that location). In all instances, the bar to the right of the map provides a colour-coded log probability of the ratio of probabilities of infection of a new population in the year transition under consideration against the average value for the full 19-year period obtained from the overall model [Ψ(NDIKC), see Fig. 3 for abbreviations]; the black circles represent host demes. (a) Log mean selection intensity in the metapopulation plotted using the best overall model [Ψ(NDIKC)] averaged over all years. The hotspots in each individual year differed from the hotspots for the overall metapopulation trend. These hotspot maps were computed by taking the log ratio of the selection intensity from the best model for that year and the best model for all the years, Ψ(NDIKC). (b) The best model for the 1990–1991 season was based only within islands, Ψ(NI). (c) The best model for the 2000–2001 season was based on weather zones, Ψ(NZ), in which selection intensity was higher in more exposed areas; (d) The best model for the 2007–2008 study was based on disease severity and transmission within island chains, Ψ(NDC).

Download figure to PowerPoint

Finally, a comparison of average selection intensity for the first and last 5-year periods of the metapopulation survey shows a general increase in selection intensity (Fig. 5). The responses varied for some individual populations and areas of the metapopulation. Selection intensities showed little change or even declined in a few focal spots (Fig. 5, green), but many areas and populations showed large increases in selection intensity (Fig. 5, red and orange). Thus across these periods, selection intensity changed at different rates and in different directions in different populations. Hot spots became hotter at different rates and at least some hot spots became cooler!

image

Figure 5.  Temporal changes in selection intensity in the metapopulation illustrated by plotting the ratio of selection intensity for the first 5 (1990–1994) and the last 5 (2004–2008) years of the survey period using the best model [Ψ(NDIKC), see Fig. 3 for abbreviations]. The bar to the right of the map provides a colour-coded log probability of the selection intensity (increasing vertically); the black circles represent host demes.

Download figure to PowerPoint

The concept of hot and cold spots of selection and their spatial dynamic has generated considerable interest regarding the links between this patchy application of selection and the maintenance of variation, local adaptation (Lively 1999), clinal effects (Nuismer, Thompson & Gomulkiewicz 2003) and, when mediated by gene flow, evolutionary dynamics across metapopulations. Although the importance of the explicit spatial structure of hot and cold spots has been recognized (Nuismer, Thompson & Gomulkiewicz 2003), most studies to date have assumed a fixed identity of hot and cold spots over time; the intensity of a hot spot might vary over time, but they remain hot relative to cold spots. More importantly, there has been no specific consideration as to the impact on the strength, intensity and direction of co-evolution if hot spots move around the landscape over time. Certainly, data from a range of studies including shifts from mutualism to commensalism, and from commensalism to antagonism (Thompson & Fernandez 2006), and single-year population differences in systems with strong metapopulation dynamics (Laine 2006), can be interpreted as indicative that hot and cold spots may move around, but only Thompson (2005) has explicitly recognized that current models fail to address these complexities.

A particularly important feature of the current study is the demonstration that not only do hot and cold spots of pathogen severity (and hence potentially selective activity) occur, but, more importantly, that the intensity and spatial focus of those selective pressures is constantly shifting around the metapopulation as the relative importance of different environmental factors wax and wane. These changes have significant implications for the nature of selective pressures being applied and may well provide conditions favouring increased diversity among and within populations.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results and discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

The field component of this work was supported by the Swedish Research Council.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results and discussion
  6. Acknowledgements
  7. References
  8. Supporting Information
  • Antonovics, J. (2004) Long-term study of a plant-pathogen metapopulation. Metapopulation Ecology, Genetics, and Evolution (eds I.Hanski & O.E.Gaggiotti), pp. 471488. Academic Press, Amsterdam.
  • Burdon, J.J., Ericson, L. & Müller, W.J. (1995) Temporal and spatial relationships in a metapopulation of the rust pathogen Triphragmium ulmariae and its host, Filipendula ulmaria. Journal of Ecology, 82, 979989.
  • Burdon, J.J. & Thompson, J.N. (1995) Changed patterns of resistance in a population of Linum marginale attacked by the rust pathogen Melampsora lini. Journal of Ecology, 83, 199206.
  • Burdon, J.J. & Thrall, P.H. (2000) Coevolution at multiple spatial scales – from the population to the continent. Evolutionary Ecology, 14, 261281.
  • Burnham, K.P. & Anderson, D.A. (1999) Model Selection and Inference: A Practical Information-Theoretic Approach. Springer-Verlag, New York.
  • Craig, T.P., Itami, J.K. & Horner, J.D. (2007) Geographic variation in the evolution and coevolution of a tritrophic interaction. Evolution, 61, 11371152.
  • Ekman, M.A. (1996) Consistent map of the postglacial uplift of Fennoscandia. Terra Nova, 8, 158165.
  • Ericson, L. (1981) Aspects of the shore vegetation of the Gulf of Bothnia. Wahlenbergia, 7, 4560.
  • Ericson, L., Burdon, J.J. & Müller, W.J. (1999) Spatial and temporal dynamics of epidemics of the rust fungus Uromyces valerianae on populations of its host, Valeriana salina. Journal of Ecology, 87, 649658.
  • Ericson, L., Burdon, J.J. & Müller, W.J. (2002) The rust pathogen Triphragmium ulmariae as a selective force affecting its host, Filipendula ulmaria. Journal of Ecology, 90, 167178.
  • Gilpin, M. & Hanski, I. (eds) (1991) Metapopulation Dynamics: Empirical and Theoretical Investigations. Harcourt Brace Jovanovich, London.
  • Gomulkiewicz, R., Thompson, J.N., Holt, R.D., Nuismer, S.L. & Hochberg, M.E. (2000) Hot spots, cold spots, and the geographic mosaic theory of coevolution. American Naturalist, 156, 156174.
  • Gomulkiewicz, R., Drown, D.M., Dybdahl, M.F., Godsoe, W., Nuismer, S.L., Pepin, K.M., Ridenhour, B.J., Smith, C.I. & Yoder, J.B. (2007) Does and don’ts of testing the geographic mosaic theory of coevolution. Heredity, 98, 249258.
  • Hanski, I. (1999) Metapopulation Ecology. Oxford Univ. Press, Oxford.
  • Laine, A.-L. (2006) Evolution of host resistance: looking for coevolutionary hotspots at small spatial scales. Proceedings of the Royal Society of London, Series B: Biological Sciences, 273, 267273.
  • Laine, A.-L. & Hanski, I. (2006) Large-scale spatial dynamics of a specialist plant pathogen in a fragmented landscape. Journal of Ecology, 94, 217226.
  • Laine, A.-L., Burdon, J.J., Dodds, P.N. & Thrall, P.H. (2011) Spatial variation in disease resistance: from molecules to metapopulations. Journal of Ecology, 99, 96112.
  • Levins, R. (1970) Extinction. Lecture Notes in Mathematics, 2, 75107.
  • Lively, C. (1999) Migration, virulence, and the geographic mosaic of adaptation by parasites. American Naturalist, 153, S34S47.
  • Nuismer, S.L., Thompson, J.N. & Gomulkiewicz, R. (2003) Coevolution between hosts and parasites with partially overlapping geographic ranges. Journal of Evolutionary Biology, 16, 13371345.
  • Smith, D.L., Ericson, L. & Burdon, J.J. (2003) Epidemiological patterns at multiple spatial scales; an 11-year study of a Triphragmium ulmariae–Filipendula ulmaria metapopulation. Journal of Ecology, 91, 890903.
  • Soubeyrand, S., Laine, A.-L., Hanski, I. & Penttinen, A. (2009) Spatio-temporal structure of interactions in a host-pathogen metapopulation. American Naturalist, 174, 308320.
  • Springer, Y.P. (2007) Clinal resistance structure and pathogen local adaptation in a serpentine flax-flax rust interaction. Evolution, 61, 18121822.
  • Thompson, J.N. (1994) The Coevolutionary Process. Chicago Univ. Press, Chicago.
  • Thompson, J.N. (2005) The Geographic Mosaic of Coevolution. Chicago University Press, Chicago.
  • Thompson, J.N. & Fernandez, C.C. (2006) Temporal dynamics of antagonism and mutualism in a geographically variable plant-insect interaction. Ecology, 87, 103112.
  • Thompson, J.N. & Merg, K.F. (2008) Evolution of polyploidy and the diversification of plant-pollinator interactions. Ecology, 89, 21972206.
  • Thrall, P.H. & Burdon, J.J. (1999) The spatial scale of pathogen dispersal: consequences for disease dynamics and persistence. Evolutionary Ecology Research, 1, 681701.
  • Thrall, P.H., Burdon, J.J. & Bock, C.H. (2001) Short-term epidemic dynamics in the Cakile maritimaAlternaria brassicicola host-pathogen metapopulation association. Journal of Ecology, 89, 723735.
  • Wilson, M. & Henderson, D.M. (1966) British Rust Fungi. Cambridge University Press, London.

Supporting Information

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results and discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Appendix S1. Notation for the data and variables used in the models.

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials may be re-organized for online delivery, but are not copy-edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

FilenameFormatSizeDescription
JEC_1788_sm_AppS1.doc49KSupporting info item

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.