• disease resistance;
  • epidemiology;
  • fungal pathogen;
  • host–pathogen interactions;
  • survival


  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  • 1
    The rust fungus Triphragmium ulmariae had a substantial effect on the survival of seedlings of Filipendula ulmaria with 89% dying over a 5-year period. Throughout the experiment, plants that survived the entire study period consistently showed very low disease severity when compared to that suffered by plants that died during its course.
  • 2
    The resistance of F. ulmaria to T. ulmariae was assessed through inoculation of progeny of six geographically separated populations of F. ulmariae with four bulk populations of T. ulmariae. Significant differences in resistance as expressed by differences in pathogen prevalence and severity were detected among the F. ulmaria populations. Within individual host populations, significant differences were also detected among open-pollinated family lines.
  • 3
    Although in some instances the susceptibility of host populations was greatest towards the pathogen collected from the same site, there was no consistent evidence for local adaptation.
  • 4
    An unplanned infection by Septoria ulmariae allowed assessment of resistance within F. ulmaria towards a single population of this pathogen. Again variation for resistance was detected among host populations and among family lines within some of the populations.


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

A range of pathological, ecological and forestry studies provide a coarse picture of the occurrence of resistance to a variety of diseases in natural plant-pathogen systems. In general these studies have shown that plants of a wide range of life history types – annuals, herbaceous perennials, and trees – show variability in their response to pathogen isolates or populations when assessed across a range of geographically spaced host populations. Although differences in resistance are perhaps most apparent in classical gene-for-gene systems involving annual or herbaceous hosts of biotrophic pathogens such as rusts and mildews (Senecio vulgarisErysiphe fischeri,Clarke 1997; Linum marginaleMelampsora liniBurdon 1997; Avena spp.–Puccinia coronata,Dinoor 1970), differences in resistance to a range of necrotrophic pathogens have also been shown (Plantago lanceolataPhomopsis subordinara,de Nooij & van Damme 1988). For many of these studies resistance was assessed under controlled conditions using genetically homogeneous pathogen lines. In forestry trials, however, resistance has been detected to a wide range of pathogens through the practical expedient of multisite provenance trials, where a sample of the geographical distribution of a range of tree species has been exposed to entire local pathogen populations (e.g. Pinus sylvestris–Phacidium infestans,Björkman 1948; Roll-Hansen 1989; Pinus taedaCronartium fusiforme,Kinloch & Stonecypher 1969; Populus spp.–Melampsora magnusiana, Gallo et al. 1985; Populus sp.–Hypoxylon mammatum, French & Manion 1975).

Both these survey approaches have shown broad correlations across species distributions between increased levels of disease resistance and abiotic (temperature, humidity, rainfall) or biotic factors (e.g. presence of alternate hosts) that are conducive to disease development. However, evidence is increasingly accumulating to indicate that the more important spatial scales of evolution in host–pathogen associations are those encompassing intermediate spatial scales varying from immediately adjoining populations, through individual metapopulations, to somewhat larger regional areas in which there is still some degree of gene flow between populations (Thrall & Burdon 1997; Burdon & Thrall 1999; Thompson 1999). In these situations, temporal and spatial asynchrony in pathogen epidemics (Ericson et al. 1999; Thrall et al. 2001), variations in epidemic amplitudes between host populations, and distance-dependent migration and gene-flow (Burdon & Thrall 1999) all contribute to a complex patchwork of varying levels of selective intensity within individual host–pathogen demes. These are summed across the metapopulations and regions as a whole to determine the evolutionary trajectory of the interaction.

Coupling information on the resistance of host species to their fungal pathogens to a picture of the numerical dynamics of given associations is essential to the development of an understanding of their longer-term coevolutionary dynamics. However, such fusions are currently restricted to just one or two systems (for example, the Linum–Melampsora interaction; Burdon & Thrall 1999). Here we combine assessment of the impact of the rust pathogen Triphragmium ulmariae on Filipendula ulmaria and the occurrence of resistance, to assess the potential for a coevolutionary interaction. We do this by assessing two hypotheses. First, that the impact of T. ulmariae on the fitness of its host is sufficient to cause death of heavily infected individuals. Although a loss of fitness resulting from reductions in seed production alone may be sufficient to ‘fuel’ a coevolutionary interaction, the ability to kill host individuals constitutes a more rigorous assessment of this question. Secondly, we hypothesize that if T. ulmariae has had a selective impact on its host, variation for resistance to T. ulmariae should occur within F. ulmaria such that different host populations from a single regional area (in this case coastal northern Sweden) differ in their response to different pathogen populations from the area.

To do this we examined a total of six populations, two from the metapopulation occurring on the Skeppsvik archipelago (Burdon et al. 1995); with the remaining four populations providing a geographical coverage of populations of F. ulmaria in northern Sweden. By incorporating two host and pathogen populations from the Skeppsvik archipelago we aimed to obtain information about variation in resistance and pathogenicity over relatively short distances, whereas inclusion of more distant populations was designed to provide a preliminary assessment of variation in resistance at a spatial scale likely to cover multiple individual metapopulations.

Materials and methods

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

THE HOST, filipendula ulmaria L. (ROSACEAE)

Filipendula ulmaria is a herbaceous, perennial dicotyledonous species typically found in mesic and moist habitats. In the study area in northern Sweden it is found as a dominant species in wet meadows, along rivers and drainage ditches, and on the upper part of the shore on many offshore islands. Plants reproduce by seed and by slow lateral spread of a rhizomatous rootstock. Establishment of seedlings occurs intermittently on river flood plains and on the open shore. Recruitment is largely restricted to years when new sites are available for colonization either through disturbance and/or silt deposition at inland sites or when prolonged low water tables at coastal sites coincide with seed germination (cf. Ericson 1981). Adult plants may produce large quantities of seed that float, thus aiding dispersal. The phenology of plants varies from year to year and among habitats.


Triphragmium ulmariae is an autoecious, macrocyclic rust pathogen that is restricted to F. ulmaria (Wilson & Henderson 1966). Because the host plant dies back to an underground rootstock during the winter, survival of the pathogen during this period is exclusively as teliospores. These germinate the following spring to initiate a process of sexual recombination and establish aecial infections characterized by bright orange eruptions on petioles and the abaxial veins of leaves. The aecial generation is followed by one or more uredinial (asexual) generations. Uredinia are small, circular and lemon-yellow in colour. Under generally unfavourable conditions, and towards the end of the summer, uredinia switch over to the production of brown–black teliospores.


In May 1991, 100 Filipendula ulmaria plants were established in an experimental garden at Umeå University from seed collected the previous summer on Nord Brunkögern, an island immediately outside the Skeppsvik study area of Burdon et al. (1995; Fig. 1). All plants were inoculated in July with a bulk sample of urediniospores of Triphragmium ulmariae collected from the same island. In September the plants were transplanted into a suitable shore meadow site on the island. At this time, nearby stands of F. ulmaria showed high disease prevalence. When transferred to the field, the position of each plant was marked with a plastic rod and plant height, number of leaves and the percentage leaf area infected by T. ulmariae were recorded. Plants were distributed evenly through a uniform area of shore meadow of approximately 1000 m2. In September of each subsequent year these parameters were again recorded. In the case of death of individuals the year of death was also recorded. These data were scored for all surviving plants until 1995 (5 years’ data) at which time monitoring was terminated, because only 11 plants were still alive.


Figure 1. Map of the northern part of Sweden showing the spatial distribution of the Filipendula ulmaria populations assessed for resistance to the rust pathogen Triphragmium ulmariae. The inset shows the distribution of islands in the Skeppsvik archipelago.

Download figure to PowerPoint

Statistical analysis

Analysis of the data was complicated by the increasing appearance of ‘missing values’ as the cumulative number of plants that had died increased with time. In particular, disease levels could not be measured on plants recorded as dead in any given year. As a consequence, initial analyses of plant survival were performed separately for each year to determine the effects of disease and growth parameters from the previous year and accumulated over all earlier years.

The data were analysed using a series of generalized linear models with binomial errors and logit link (Dobson 1990) in which the survival of seedlings in individual years was assessed by stepwise addition of log-transformed percentage disease in the previous year, cumulated log percentage disease since the start of the experiment, numbers of leaves in the previous year, height in the previous year, and change in the last two parameters between the current and previous year.

In the initial analyses, the only variables having a significant effect on survival were those measured in the previous year. Consequently a further stepwise generalized linear model with binomial errors and logit link (Dobson 1990) was fitted, in which data were pooled across all years and survival was modelled in relation to log-transformed percentage disease in the previous year, numbers of leaves in the previous year, and height in the previous year after removing the overall difference in survival between years.

EXPERIMENT 2 – RESISTANCE OF filipendula ulmaria TO triphragmium ulmariae

Identification and collection of F. ulmaria populations

To determine whether variation occurs within F. ulmaria in its response to the pathogen T. ulmariae, six populations of F. ulmaria were identified covering a broad geographical range of the species in northern Sweden. Two of these populations were located on islands of the Skeppsvik archipelago (islands S15 and S48, respectively, see Burdon et al. 1995; Fig. 1), whereas the remaining populations (Nabben [N], Önnesmark [Ö], Berghem [B] and Kukkola [Nb]) ranged from 363 km north of Skeppsvik to 67 km south (Fig. 1). The last four populations occurred at sites between 0 and 17 km from the coast. At each site, seed was collected in September 1995 from 10 to 15 plants and bagged separately. Seed was immediately sown outdoors in the experimental garden of Umeå University to ensure a normal winter treatment. Seedlings appeared in late May to early June 1996 and these were used to generate a series of 10 open-pollinated families from each of the six populations. At this time, seedlings of approximately equal size were individually potted up in a peat/sand/soil mix. A second selection of plants was carried out just prior to inoculation early in the summer of 1997 to ensure uniformity of size.

Sampling of T. ulmariae populations

In early summer 1997 (16th–27th June), four of the host populations (Skeppsvik islands S15 and S48, Berghem and Kukkola) from which seed had been collected were visited and, at each, a bulk collection of the rust pathogen T. ulmariae was made by collecting 100–150 aecia randomly throughout the plant population. Aeciospores were gently scraped into a glass vial with a scalpel and used the following day for inoculations.

Resistance assessment and experimental design

There are two broad strategies available for assessing disease resistance in natural populations, viz: (i) inoculation of host lines with individual pathogen lines known to be genetically uniform; or (ii) inoculation of host lines with a bulk sample of isolates representative of the pathogen population as a whole. Because we were interested in determining whether any resistance effective against the current pathogen populations occurs in this host–pathogen association, we chose to follow the latter strategy, using four bulked pathogen samples, one from each of the sites S15, S48, B and Nb. By assessing responses across host populations from widely separated localities (and hence most likely quite different local coevolutionary units) we maximized the possibilities of detecting variation, although with the inoculum strategy used we recognized that we would be unable to determine irrefutably the nature of the genetic control involved.

The resistance of F. ulmaria to T. ulmariae was examined in a large field experiment set out in the outdoor plant propagation area of the Department of Ecology and Environmental Science. The experiment was divided into four large areas, one for each pathogen, and each area was divided into a number of blocks (up to 40 in each, Fig. 2). Because there was insufficient space in each block for an offspring from each of the 60 parent plants (10 per population × 6 populations), an incomplete block design was used in each area in which the offspring for each of the host site parents were spread widely among the available blocks, so that no block contained more than one plant from each parent. On average, over all four pathogens, each parent was represented by an offspring plant in 58% of the blocks. There were 3960 offspring plants in the experiment, with total numbers of offspring for each host site ranging between 653 and 666. For each pathogen × host origin combination, the numbers of offspring plants from the 10 parent plants were approximately equal. Insufficient seedlings of each of the parental lines survived to provide complete replication across all pathogen treatments. As a consequence, fewer family lines were challenged with pathogen isolates B and Nb.


Figure 2. Design of field experiment to assess resistance to Triphragmium ulmariae in natural populations of Filipendula ulmaria. (a) Location of the four areas, one for each pathogen from populations S15, S48, Nb and B; (b) numbering and location of blocks within each of the four areas.

Download figure to PowerPoint

The plants in the blocks were separately inoculated with the bulk sample of the pathogen populations (diluted 10 times by weight with sterile talc) collected from S15, S48, B and Nb. Immediately following inoculation the plants were lightly misted with tap water and then covered with plastic to maintain a saturated atmosphere. The following day the plastic sheeting was removed. Plants were individually assessed 75–90 days later for the presence/absence of T. ulmariae (prevalence at the population level). At the same time percentage disease infection (severity) was scored by careful visual estimation of the percentage of leaf area infected with disease.

During the course of this experiment an opportunistic infection of Septoria ulmariae Oud occurred across all blocks. This pathogen is specific to F. ulmaria, forming small orbicular leaf spots with a dark purple border (Jorstad 1965). As a consequence, we took the opportunity to determine whether variation existed in the response of the six host populations to S. ulmariae by assessing the severity of natural infection on a 0–3 scale (0 = absent; 3 = extensive lesion formation).

The experiment enabled separation of three components of variation – block (primarily environmental since plants from each host population were widely spread through blocks), host origin (primarily genetic representing the overall genetic differences of populations from the different sites), and parent with host origin (genetic).

Statistical analysis

Because each experimental area was distinct, and the timing of inoculation and assessment varied between areas, the prevalence and severity of T. ulmariae for the four pathogen populations were analysed separately.

Because no block contained more than one plant from each parent, the prevalence data for T. ulmariae were analysed by fitting, separately for each pathogen population, generalized linear models with binomial error and logit link (Dobson 1990) to individual plant responses, resulting in analyses of deviance of binary data. Between-blocks variation was fitted first in the models, then variation among host sites, and finally variation among parents within host sites. This final component was broken into six sub-components, one for each host site. When fitting such a model to binary data there is, by definition, no extra-binomial variation and the residual mean deviance is typically less than 1. Consequently, the significance of each factor was tested by examining the significance of the deviance for the factor as a chi-squared variate on the degrees of freedom for that factor.

For each pathogen population, hierarchical analyses of variance were performed on percentage infection (severity) of T. ulmariae and on Septoria score. In these analyses, each parent was not represented in each block, hence a regression approach was used to handle this non-orthogonality. The terms fitted in these models were as described for the analyses of deviance. Based on examination of residuals from the analyses, percentage infection of T. ulmariae was log-transformed before analysis, whereas the Septoria score did not require transformation. Because differences between the six host populations were similar for each of the four pathogen populations, an overall analysis variance of the Septoria score was also performed, where between-areas variation was fitted first, then between-blocks variation within each area, then variation among host sites, and finally variation among parents within host sites.

Relationships between severity of T. ulmariae (log-transformed) and Septoria score were also investigated. Because both severity and Septoria score could significantly differ between blocks, host sites, and parents within host sites, the nature of the relationships could be dependent on the relationship of the mean responses for each combination of these factors. Consequently, relationships were also obtained between the residuals from the models after fitting these factors, to determine whether the levels of the two diseases in individual plants were related.


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


Over the course of the experiment the number of seedlings dying each year steadily rose from zero in the year of establishment (1991) to 44 in 1994, following which there was a decline in deaths in 1995 (Fig. 3a). Over the same period the mean level of infection of seedlings by T. ulmariae declined from a maximum of 12–13% of leaf area infected in 1991 and 1992 to c. 0.6% in 1995 (Fig. 3b). When seedlings are grouped on the basis of the year in which they died, highly significant differences (P < 0.001) were detected in 1991, 1992 and 1993 in the mean percentage infection suffered by the different groups (Fig. 4). Across all years, those plants that survived to the end of the experiment suffered consistently lower levels of rust infection. Moreover, among those that died, there was a consistent trend for the earlier dying groups to incur noticeably higher levels of disease than those groups surviving longer.


Figure 3. Trends in plant mortality and disease development in the interaction between Filipendula ulmaria and the rust pathogen Triphragmium ulmariae. (a) Yearly (shaded bars) and cumulative (solid line) plant mortality; (b) mean levels of disease infection through time.

Download figure to PowerPoint


Figure 4. Different levels of infection caused by Triphragmium ulmariae on different groups of seedlings of Filipendula ulmaria (classified according to the year of death) over the period 1991–95 inclusive. Measures of the statistical significance of differences in disease suffered by different seedling groups in each year were tested by anova.

Download figure to PowerPoint

In the first year of assessment, 1992, there was a significant effect of disease severity in the previous year on plant survival but no effect of plant size (Table 1). In following years, variables associated with plant growth (number of leaves and height in the previous year) became increasingly important as predictors of survival. In all cases, plant survival decreased as disease severity increased, and increased as height or number of leaves increased. Changes in height or number of leaves between the previous and current year and cumulated log percentage disease were not significant in any of the analyses. The latter is the result of consistently high correlations (> 0.8) between leaf area infected in the previous year and total cumulated leaf area infected.

Table 1.  Summary of analyses of deviance of survival of 100 Filipendula ulmaria seedlings over a 4-year period in relation to disease caused by Triphragmium ulmariae and size of seedlings. Only variables significant at the 5% level are shown. Significance occurs when deviance > 3.84, the 5% point of the chi-squared distribution with 1 d.f.
YearAnalyses including infection variables onlyAnalyses including infection and plant size variables
1992Ln% infected 1991  5.59Ln% infected 1991  5.59
1993Ln% infected 1992  7.61Height 1992 18.34
1994Ln% infected 1993 20.51No. leaves 1993 33.38
   Ln% infected 1993  6.35
1995Ln% infected 1994  4.40No. leaves 1994 22.95
 Ln% infected previous year 36.00No. leaves previous year 66.14
   Ln% infected previous year 11.77
   Height previous year  7.75

The best single predictor of survival in 1993 was height in 1992, and the best in 1994 and 1995 was number of leaves in 1993 and 1994, respectively. Disease severity in the previous year retained some predictive value in 1993 and 1994, but only significantly improved the relationship with the best plant growth variable in 1994 (Table 1). By 1995, plant size (as estimated by number of leaves) showed a highly significant positive relationship with survival, whereas disease severity was only weakly related to survival. This is perhaps not surprising given that only two of the 11 surviving plants showed infection severities > 5% in the previous 2 years, whereas seven individuals had been disease-free throughout the same period.

The overall model showed that, after the large difference in survival between years had been removed, number of leaves in the previous year was the most important predictor of survival, with a highly significant positive relationship. Disease severity in the previous year had a significant negative effect on survival, both in addition to number of leaves and by itself. Height in the previous year was the least important predictor of the three, but nevertheless had a significant additional positive effect on survival. Interactions of year with each of these three variables were also examined, but none was significant at the 5% level.

EXPERIMENT 2 – RESISTANCE OF filipendula ulmaria TO triphragmium ulmariae

Prevalence of T. ulmariae

Preliminary analyses of the prevalence of disease induced by pathogen S15 showed that disease prevalence in blocks 1–17 (mean range 30.5–66.0%) was distinctly different from prevalence in blocks 20–40 (mean range 85.8–100.0%). This was unlikely to have been a spatial effect, as blocks 37–40 were well separated from blocks 20–36 (Fig. 2), but was probably the result of timing of assessment, because individual plant assessments extended over 15 days. Consequently separate analyses were performed for each group. For pathogen B, prevalence was 100% throughout blocks 1–10, so no information was available to discriminate between host sites or parents within host sites for this variable, but prevalence for blocks 11–14 ranged from 22.9% to 58.4%. This difference was again probably caused by timing of assessment and, consequently only blocks 11–14 were included in the analyses.

There were highly significant differences in disease prevalence (percentage of individuals infected) among host populations for each pathogen, except blocks 11–14 of pathogen B (Table 2). The levels of prevalence presented for pathogen B were relatively low only because the means were over blocks 11–14 only. If means were taken over all 14 blocks they would range from approximately 0.76–0.85. Differences in disease prevalence varied across the host populations (Table 2), with plants from host population B showing the lowest prevalence for all pathogens. In contrast, host populations S15 and S48 consistently showed the highest disease prevalence scores or were not significantly different from the highest scores for each pathogen. Host population Nb had high prevalence to the pathogen from Nb, but had a lower than average response to the pathogens from sites S15 and S48.

Table 2.  Mean prevalence of disease (expressed as the proportion of plants that were diseased) caused by four populations of the rust pathogen Triphragmium ulmariae on Filipendula ulmaria collected from six sites in northern Sweden. Standard errors are provided in parentheses. See text for explanation of blocks
Host populationPathogen origin
S15 Blocks 1–17S15 Blocks 20–40S48B Blocks 11–14Nb
  • Based on chi-squared with 5 d.f.

S15    0.712 (0.049)    0.992 (0.005)     0.817 (0.021)0.453 (0.114)    0.961 (0.009)
S48    0.729 (0.048)    0.967 (0.010)     0.748 (0.024)0.428 (0.112)    0.969 (0.009)
B    0.269 (0.048)    0.854 (0.019)     0.477 (0.024)0.172 (0.086)    0.875 (0.016)
N    0.466 (0.054)    0.950 (0.012)     0.564 (0.024)0.289 (0.100)    0.976 (0.007)
Nb    0.401 (0.053)    0.886 (0.017)     0.600 (0.025)0.484 (0.112)    0.976 (0.007)
Ö    0.502 (0.054)    0.983 (0.007)     0.684 (0.024)0.484 (0.112)    0.985 (0.006)
Deviance 70.790 34.576116.3969.021 21.048
P-value< 0.001< 0.001 < 0.0010.108< 0.001
Residual mean deviance  1.217  0.326   0.8320.864  0.251

An estimate of genetic variability among different parental lines within each host population was obtained by splitting the among-parents within host population term in the analysis of deviance (54 d.f.) into six components, one for each host site (9 d.f. each). This analysis showed that within host populations there was significant genetic variation among family lines when these were challenged by some pathogen populations (Table 3). This was especially pronounced in host population B with respect to pathogen populations S15 and S48, and in host population Nb when challenged by pathogen populations S48 and B. Interestingly, among-parent variability was not significant in any combinations involving plant and pathogen populations from the same site (Table 3).

Table 3.  Results of an analysis of deviance testing for variability in the prevalence of disease caused by the rust pathogen Triphragmium ulmariae among half-sib families within each of six populations of the host plant Filipendula ulmaria.P-values corresponding to the deviance values are given in parentheses – these are distributed under the null hypothesis as χ2 variates with 9 d.f.; significant values are highlighted in bold. See text for explanation of blocks
Host populationPathogen origin
S15 Blocks 1–17S15 Blocks 20–40S48B Blocks 11–14Nb
S15 7.023 4.88026.404  14.87612.092
   (0.635)  (0.845)  (0.002)    (0.094)  (0.208)
S48 4.13911.02815.191  32.140 9.884
   (0.902)  (0.274)  (0.086)(< 0.001)  (0.360)
B23.03025.19722.072  12.21215.128
   (0.006)  (0.003)  (0.009)    (0.202)  (0.088)
N21.010 9.91212.186  13.61510.208
   (0.013)  (0.358)  (0.203)    (0.137)  (0.334)
Nb15.332 5.36819.296  17.290 6.150
   (0.082)  (0.801)  (0.023)    (0.044)  (0.725)
Ö17.872 6.550 15.461  28.096 5.774
   (0.037)  (0.684)  (0.079)(< 0.001)  (0.762)

Differences in among-parent genetic variability between the two groups of blocks, 1–17 and 20–40, resulting from challenge by pathogen population S15 (specifically with respect to host populations N and Ö), presumably reflect the consequences of disease assessment at different stages of pathogen development.

Severity of disease caused by T. ulmariae

Unlike disease prevalence, analyses of severity of disease caused by pathogen population S15 did not show a strong difference between blocks 1–17 (percentage infection range: 0.60–2.62%) and blocks 20–40 (1.34%–4.25%). However, blocks 20–40 still had generally higher values, consistent with the two groups of blocks having different timings of infection and assessment, as previously observed. Consequently, as for disease prevalence, separate analyses were performed for each group. For pathogen population B, disease severity was higher in blocks 1–10 (1.11%–2.02%) than in blocks 11–14 (0.37%–0.80%), again consistent with the two groups of blocks having different timings of infection and assessment. However unlike pathogen S15, a single analysis was performed over all blocks, as blocks 11–14 by themselves provided insufficient data to test differences between host populations properly (cf. disease prevalence analyses, Table 2) and differences between the two groups could be removed as part of the ‘blocks’ term in the analyses.

Differences in disease severity among host populations were much greater for assessments involving pathogen populations from S15 and S48 than for those from B and Nb (Table 4). Similarly to disease prevalence, plants from host population B were the most resistant, showing the lowest (or near lowest) percentage severity infection for all pathogens. On the other hand, plants from host populations S15 and S48 suffered consistently higher levels of disease in response to pathogens from the same two sites, but were neither especially susceptible nor resistant to pathogen populations B or Nb. The severity of disease occurring on plants from the other three populations (N, Nb and Ö), relative to that on host populations S15, S48 and B, varied according to the pathogen used.

Table 4.  Mean severity of disease caused by four populations of the rust pathogen Triphragmium ulmariae on six populations of its host Filipendula ulmariae. (a) Mean log-transformed disease severity percentages with standard errors; (b) mean raw disease severity percentages obtained by back-transforming log means
Host populationPathogen origin
S15 Blocks 1–17S15 Blocks 20–40S48BNb
  • Based on an F-test from the analysis of variance, with 5 d.f. for host site, and many d.f. for residual.

(a) Log-transformed disease severity means
S15  0.870 (0.066)  1.283 (0.050)  1.327 (0.044)0.737 (0.039)  1.001 (0.044)
S48  0.834 (0.066)  1.373 (0.050)  0.996 (0.045)0.707 (0.039)  0.839 (0.044)
B  0.253 (0.066)  0.714 (0.051)  0.477 (0.044)0.613 (0.039)  0.808 (0.044)
N  0.448 (0.066)  0.958 (0.050)  0.607 (0.044)0.610 (0.039)  1.047 (0.045)
Nb  0.383 (0.066)  0.917 (0.051)  0.656 (0.045)0.752 (0.039)  1.011 (0.044)
Ö  0.516 (0.066)  1.060 (0.050)  0.804 (0.044)0.758 (0.039)  1.145 (0.044)
Variance ratio   14.24   23.27   49.343.03  8.37
P-value< 0.001< 0.001< 0.0010.010< 0.001
(b) Raw disease severity means
S15  1.387  2.607  2.7701.090  1.721
S48  1.303  2.947  1.7071.028  1.314
B  0.288  1.042  0.6110.846  1.243
N  0.565  1.606  0.8350.840  1.849
Nb  0.467  1.502  0.9271.121  1.748
Ö  0.675  1.886  1.2341.134  2.142

As for the disease prevalence assessment, genetic variability in disease severity among parent lines within each host population was obtained by splitting the among-parents within host population term (54 d.f.) in the analyses of variance of log-transformed percentage infection into six components, one for each host site (9 d.f. each). There was significant genetic variability among parental lines for a number of host–pathogen combinations (Table 5), with some of these combinations being the same as those found to be significant for disease prevalence. However, unlike the assessment based on disease prevalence, among-parent genetic variation in disease severity was apparent in at least some host–pathogen combinations involving components from the same site (e.g. S48, Nb).

Table 5.  Significance values (P) for variability in the severity of disease caused by the rust pathogen Triphragmium ulmariae among half-sib families within each of six populations of the host plant Filipendula ulmaria. P-values were obtained by splitting the among-parents within-host population term (54 d.f.) in the analyses of variance of log-transformed severity data into six components, one for each host population (9 d.f. each). Significant values at the 5% level are highlighted in bold
Host populationPathogen origin
S15 Blocks 1–17S15 Blocks 20–40S48BNb
S150.0080.6880.0510.289    0.310
S480.6230.0080.0220.187    0.642
B0.3750.1370.0150.812    0.003
N0.0480.1170.5650.922    0.006
Nb0.4060.7360.0590.044< 0.001
Ö0.0150.8770.1340.014< 0.001


Geographical distances between every possible pairing of host and pathogen populations were calculated and compared with the adjusted disease prevalence and severity levels each pathogen population caused in each host–pathogen pair. The prevalence and severity data sets were first adjusted by logit and log transformation, respectively, to allow for differences in disease parameters resulting for differences in conditions among the different trials. The subsequent analysis failed to detect a significant relationship between distance and either of the disease parameters (disease presence and log percentage infection, respectively, adjusted for experimental and host origin effects vs. distance: r = −0.251 and r = −0.174; P > 0.10).

RESPONSE TO INFECTION BY septoria ulmariae

Disease caused by Septoria ulmariae was widespread, with no large areas of the experiment being generally more or less infected than others. The severity of Septoria disease was significantly higher on plants from host populations B, Nb and Ö, than on plants from populations S15, S48 and N (Table 6), suggesting differences in resistance of the six host populations to Septoria. Variability among-parents within each host population was examined by splitting the among-parent within host-population term in the overall analysis into six components, one for each host population. For two of the populations, N and Nb, a slightly different selection of 10 parents was used in some of the experiments, such that there were progeny from 11 parents from N, and from 13 parents from Nb. The variance ratios, degrees of freedom and P-values for significance of variability among parents within each host population are given in Table 6. Significant variation occurred among parents in two host populations with the higher level of Septoria infection (Nb and O) and two of the populations with the lower level of Septoria infection (S15 and S48).

Table 6.  Mean severity of disease caused by Septoria ulmariae (and standard errors) and variance ratios and significant values (P) for variability among half-sib families within each of six populations of Filipendula ulmaria. Significant values at the 5% level are highlighted in bold
Host populationMean severity of SeptoriaVariability among half-sib families
Variance ratiod.f.P-value
  • Based on an F-test from the analysis of variance, with 5 d.f. for host site, and many d.f. for residuals.

S15    0.965 (0.028)2.819    0.003
S48    0.900 (0.028)4.659< 0.001
B    1.419 (0.029)1.02 9    0.418
N    0.955 (0.028)1.0610    0.392
Nb    1.486 (0.029)2.4412    0.004
Ö    1.422 (0.028)8.11 9< 0.001
Variance ratio   96.21   
P-value< 0.001   


Correlations between severity of rust (T. ulmariae) and the S. ulmariae score were calculated for S. ulmariae scores vs. log T. ulmariae severity; S. ulmariae scores vs. T. ulmariae residuals (after fitting the full model); and S. ulmariae residuals vs. T. ulmaria residuals (obtained after fitting a full model to both data sets). Statistically significant correlations were obtained between the last two comparisons for both pathogen population S15 (blocks 1–17) and S48. However, although P < 0.001 in four cases, the actual correlation coefficients were not very high (ranging from 0.099 to 0.237) indicating that there is a slight tendency for association between the two diseases in some circumstances, rather than a predictive relationship.


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

A previous study of the interaction between Filipendula ulmaria and its rust pathogen Triphragmium ulmariae showed the complexity of the epidemiological patterns in an extensive metapopulation in the Skeppsvik archipelago (Burdon et al. 1995). Over a 4-year period, disease levels in the metapopulation as a whole remained relatively constant. In contrast, at the individual population level, disease incidence, prevalence and severity were all affected to varying degrees by host population size, and a range of physical environmental factors. This led to unpredictable dynamics in individual host populations, particularly those that were relatively small (< 600 individuals).

Although that study also investigated questions concerning the over-winter survival of the pathogen and among-population dispersal, no attempt was made to determine whether T. ulmariae was able to exert sufficient selective pressure to affect the fitness of host individuals, or what role genetic variation in host and pathogen populations had in determining the patterns observed. Here we have made a first step in bringing genetic considerations into our understanding of this interaction, by examining the fitness consequences of infection by T. ulmariae for the survival and growth of seedlings of F. ulmaria. We focused on this stage in the life-cycle because of its obvious importance in the establishment of new host populations. In addition, once plants become well established and begin to develop a rhizome system, the levels of disease severity commonly observed are likely to be insufficient to affect survival.


Survival among seedlings was significantly affected by both their disease status and growth parameters (height, number of leaves) (Table 1). Disease severity amongst the seedling population as a whole was high (12%) during the first 2 years of the experiment but declined from thereon as the number of plants dying rose rapidly (Fig. 2). Clearly the pathogen was increasing at differential rates, as by 1995 the surviving 11% of individuals incurred less than 1% leaf area loss caused by T. ulmariae. Indeed, disease severity on these plants was always significantly lower than that occurring on plants dying in the early years of growth (Fig. 3). This predictably lower level of disease severity is consistent with the idea that the original experimental population showed a broad range of levels of resistance to T. ulmariae and that the differential impact of the rust led to the more resistant host individuals being selectively favoured.

RESISTANCE IN f. ulmaria

There was considerable within- and among-population variation in the resistance of F. ulmaria to T. ulmariae. Resistance assessed either in terms of the percentage of individuals infected by particular pathogen populations (prevalence, Tables 2 and 3) or the severity of disease development (Tables 4 and 5) showed very good agreement with their overall response to individual pathogen populations. Thus significant differences in both infection and disease development occurred among the six host populations when challenged by pathogen populations S15, S48 and Nb, but the Berghem (B) population, showed no differences in prevalence and the smallest effects on severity (P = 0.01, Table 4).

Interestingly, when the data were analysed to account for the family structure among host individuals taken from individual sites (Tables 3 and 5), it was apparent that differences in resistance among populations were not entirely a reflection of a simple generalized difference expressed by all individuals. Rather, for both measures of resistance there was significant among-family within-population variation in response. In several cases this was apparent using both measures of resistance (e.g. resistance in host populations Nb and Ö to the Berghem pathogen population, and resistance among Berghem host family lines to the pathogen population from island 48 in the Skeppsvik archipelago). However, the two disease scores assess different aspects of the development of the interaction between Filipendula and Triphragmium (infection and growth (spore production, etc.), respectively) and it is therefore not surprising that there should be some differences between the distribution of among-family responses. Indeed, similar incomplete correlations between resistance measures assessed at different stages in the development of interactions between pathogens and their hosts are well known in a number of agricultural systems. Thus in the interaction between wheat and wheat stem rust, seedling infection types are a good but not infallible guide to resistance expressed later in plant development (McIntosh et al. 1995).

The opportunistic assessment of response to Septoria infection also found evidence for differences in resistance within and among the six Filipendula populations (Table 6). This result is perhaps not surprising, given that along with Triphragmium, Septoria is one of the two most common pathogens occurring in these and other Filipendula populations found growing in northern Sweden.


Evidence for genetic variation in disease resistance in natural plant populations is surprisingly sparse, given the emphasis on its use in agriculture. A few detailed studies have used a limited number of genetically pure pathogen isolates to assess the resistance structure of host populations grown under controlled conditions (Dinoor 1977; Burdon 1987; de Nooij & van Damme 1988; Parker 1988; Burdon & Jarosz 1992). Similarly, evidence for the occurrence of resistance within individual field populations can be gleaned from a number of experimental trials generally aimed at assessing links between host diversity and disease incidence (e.g. Schmid 1994; Morrison 1996). However, broader scale assessments have almost exclusively been the preserve of forestry provenance trials where representative families of a species from a range of sites have been grown in a single area and subject to natural infection. Such studies confirm the general pattern detected here – that disease resistance is a trait apparently under active selection in a wide range of environments and species (e.g. Hodson et al. 1986; Hunt et al. 1987; McDermott & Robinson 1989). Although several of these studies have simply demonstrated differences among provenances (often sourced from widely separated sites), many have also shown, as here, the existence of substantial within-population variation.

Because bulk samples of rust pathogen populations are likely to be composed of multiple pathotypes of differing virulence combinations (cf. Melampsora lini, Burdon & Jarosz 1992; Burdon et al. 1999; Melampsora epitea, Pei et al. 1996; Puccinia coronata, Dinoor 1977) we cannot use the current study to identify major resistance phenotypes, or interactions caused by particular host line–pathogen isolate combinations, or to make predictions about the mode of inheritance of the resistance detected. Furthermore, this approach also precludes comparison of our results with those of Burdon and colleagues (see Thrall et al. 2001; Burdon et al. 1999; for references) in the Linum marginaleMelampsora lini system, where individual lines of M. lini were used to investigate the detailed resistance structure of host populations. However, by using a bulk sample of the pathogen population, we are able to assess more directly the effective field susceptibility of a host population to its own or other pathogen populations.

The combination of within- and among-population patterns of disease resistance detected here is of particular interest in an understanding of the forces driving the various scales of interaction that together form a broad ‘geographical mosaic’ of coevolution (Thompson 1994, 1999). The two host populations from the Skeppsvik archipelago (S15 and S48, which are physically very close and environmentally very similar) showed the greatest susceptibility (incurring the highest values of both prevalence and severity) of all host populations to the pathogen populations collected from those sites (pathogens S15 & S48). However, similar patterns of a ‘better fit’ indicative of local adaptation (Parker 1985; Ebert 1994) between pathogen populations and their ‘home’ host populations rather than more distant ones, were not detected in other combinations.

Indeed, there were general similarities between the two host populations occurring in the Skeppsvik archipelago (S15 and S48) and their overall level of resistance to Triphragmium ulmariae. However, the possibility of some differences in resistance were indicated by the occurrence of within-family variation in one site (S48), while there was a general difference in the overall aggressiveness (as measured by the disease severity ratings) of the pathogen populations from both these two sites. It is not surprising that such differences may occur over the relatively short distances of little more than 1 km – population-level differences in both these characters have been regularly detected over distances as short as 300 m in the Linum–Melampsora host–pathogen system (Jarosz & Burdon 1991; Burdon & Jarosz 1992; Thrall et al. 2001). Furthermore, although dispersal of both host and pathogen in the Skeppsvik archipelago is assisted by water movement (Burdon et al. 1995), the two populations involved in this study were deliberately chosen to be isolated from direct water-assisted contact by their individual locations. These results, when coupled with those from the seedling fitness experiment, give strength to the idea that both epidemiological and genetic changes and differences occur within the Skeppsvik Filipendula–Triphragmium metapopulation system.


The two pathogens considered in this study represent examples of major differences in life history strategy. Triphragmium ulmariae is a biotrophic rust fungus that requires living host tissue for its continued growth and reproduction; in contrast, Septoria ulmariae is a necrotrophic pathogen that reproduces on host tissue killed by the release of toxins.

Disease development in the field is consistent with these two views. T. ulmariae builds up each spring from discrete aecial infections that result from the successful overwintering of teliospores. Ultimately this results in a limited number of foci that subsequently give rise to a spreading epidemic of uredinial infections. In the field this is typically manifest by a very patchy distribution of disease within and among host populations. In contrast, S. ulmariae re-establishes infection from spores widely distributed on dead plant material that is broadly spread through host populations. It is not surprising therefore that this disease typically shows far less spatial structure in incidence and severity. Elsewhere, one of us (Burdon et al. 1996) has suggested that such differences in the epidemiological dynamics of host–pathogen associations may favour the evolution of disease resistance based on qualitative (gene-for-gene) and quantitative (multiple minor genes) systems, respectively. As noted earlier, the use of a bulk-inoculum approach in the current study prevents an immediate assessment of this hypothesis. However, the Filipendula system may prove to be an effective one in which it can be tested.


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

This study was supported by the Swedish Natural Science Research Council.


  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  • Björkman, E. (1948) Studies on the biology of the Phacidium blight and its prevention. (Swedish with English summary). Medd. Statens Skogeforskningsinst., 37, 1136.
  • Burdon, J.J. (1987) Phenotypic and genetic patterns of resistance to the pathogen Phakopsora pachyrhizi in populations of Glycine canescens. Oecologia, 73, 257267.
  • Burdon, J.J. (1997) The evolution of gene-for-gene systems in natural pathosystems. The Gene-for-Gene Relationship in Plant–Parasite Interactions (eds I. R.Crute, E. B.Holub & J. J.Burdon), pp. 245262. CAB International, Wallingford, UK.
  • Burdon, J.J. & Jarosz, A.M. (1992) Temporal variation in the racial structure of flax rust (Melampsora lini) populations growing on natural stands of wild flax (Linum marginale): local versus metapopulation dynamics. Plant Pathology, 41, 165179.
  • Burdon, J.J. & Thrall, P.H. (1999) Spatial and temporal patterns in coevolving plant and pathogen associations. American Naturalist, 153, S15S33.
  • Burdon, J.J., Ericson, L. & Müller, W.J. (1995) Temporal and spatial changes in a metapopulation of the rust pathogen Triphragmium ulmariae and its host, Filipendula ulmaria. Journal of Ecology, 83, 979989.
  • Burdon, J.J., Thrall, P.H. & Brown, A.H.D. (1999) Resistance and virulence structure in two Linum marginale–Melampsora lini host–pathogen metapopulations with different mating systems. Evolution, 3, 70416.
  • Burdon, J.J., Wennström, A., Elmqvist, T. & Kirby, G.C. (1996) The role of race specific resistance in natural plant populations. Oikos, 76, 411416.
  • Clarke, D.D. (1997) The genetic structure of natural pathosystems. The Gene-for-Gene Relationship in Plant–Parasite Interactions (eds I. R.Crute, E. B.Holub & J. J.Burdon), pp. 231243. CAB International, Wallingford, UK.
  • Dinoor, A. (1970) Sources of oat crown rust resistance in hexaploid and tetraploid wild oats in Israel. Canadian Journal of Botany, 48, 153163.
  • Dinoor, A. (1977) Oat crown rust resistance in Israel. Annals of the New York Academy of Sciences, 287, 357366.
  • Dobson, A.J. (1990) An Introduction to Generalized Linear Models. Chapman & Hall, London.
  • Ebert, D. (1994) Virulence and local adaptation of a horizontally transmitted parasite. Science, 265, 10841086.
  • 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.
  • French, J.R. & Manion, P.D. (1975) Variability of host and pathogen in Hypoxylon canker of aspen. Canadian Journal of Botany, 53, 27402744.
  • Gallo, L.A., Stephan, B.R. & Krusche, D. (1985) Genetic variation of Melampsora leaf rust resistance in progenies of crossings between and within Populus tremula and P. tremuloides clones. Silvae Genetica, 34, 208214.
  • Hodson, A.C., French, D.W. & Jensen, R.A. (1986) Role of insects and diseases in a jack pine provenance study. Great Lakes Entomologist, 19, 239247.
  • Hunt, R.S., Ying, C.C. & Ashbee, D. (1987) Variation in damage among Pinus contorta provenances caused by the needle cast fungus Lophodermella concolor. Canadian Journal of Forestry Research, 17, 594597.
  • Jarosz, A.M. & Burdon, J.J. (1991) Host–pathogen interactions in natural populations of Linum marginale and Melampsora lini. II. Local and regional variation in patterns of resistance and racial structure. Evolution, 45, 16181627.
  • Jorstad, I. (1965) Septoria and septoroid fungi on dicotyledons in Norway. Skrifter utgitt av Det Norske Videnskaps-Akademi i Oslo I. Mat.-Naturv. Klasse. Ny Serie no. 22, pp. 1110. Oslo University Press, Oslo.
  • Kinloch, B.B. Jr & Stonecypher, R.W. (1969) Genetic variation in susceptibility to fusiform rust in seedlings from a wild population of loblolly pine. Phytopathology, 59, 12461255.
  • McDermott, J.M. & Robinson, R.A. (1989) Provenance variation for disease resistance in Pseudotsuga menziesii to the Swiss needle-cast pathogen, Phaeocryptopus gaeumannii. Canadian Journal of Forestry Research, 19, 244246.
  • McIntosh, R.A., Wellings, C.R. & Park, R.P. (1995) Wheat Rusts: an Atlas of Resistance Genes. CSIRO Publishing, Melbourne.
  • Morrison, J.A. (1996) Infection of Juncus dichotomus by the smut fungus Cintractia junci: an experimental field test of the effects of neighbouring plants, environment, and host plant genotype. Journal of Ecology, 84, 691702.
  • De Nooij, M.P. & Van Damme, J.M.M. (1988) Variation in host susceptibility among and within populations of Plantago lanceolata L. infected by the fungus Phomopsis subordinaria (Desm.) Trav. Oecologia, 75, 535538.
  • Parker, M.A. (1985) Local population differentiation for compatibility in an annual legume and its host-specific fungal pathogen. Evolution, 39, 713723.
  • Parker, M.A. (1988) Polymorphism for disease resistance in the annual legume Amphicarpaea bracteata. Heredity, 60, 2731.
  • Pei, M.H., Royle, D.J. & Hunter, T. (1996) Pathogenic specialization in Melampsora epitea var. epitea on Salix. Plant Pathology, 45, 679690.
  • Roll-Hansen, F. (1989) Phacidium infestans: a literature review. European Journal of Forest Pathology, 19, 237250.
  • Schmid, B. (1994) Effects of genetic diversity in experimental stands of Solidago altissima– evidence for the potential role of pathogens as selective agents in plant populations. Journal of Ecology, 82, 165175.
  • Thompson, J.N. (1994) The Coevolutionary Process. University of Chicago Press, Chicago.
  • Thompson, J.N. (1999) Specific hypotheses on the geographic mosaic of coevolution. American Naturalist, 153, S1S14.
  • Thrall, P.H. & Burdon, J.J. (1997) Host–pathogen dynamics in a metapopulation context: the ecological and evolutionary consequences of being spatial. Journal of Ecology, 85, 743753.
  • Thrall, P.H., Burdon, J.J. & Young, A. (2001) Variation in resistance and virulence among demes of a single host–pathogen metapopulation. Journal of Ecology, 89, 736748.
  • Wilson, M. & Henderson, D.M. (1966) British Rust Fungi. Cambridge University Press, London.