Infection of Trientalis europaea by the systemic smut fungus Urocystis trientalis: disease incidence, transmission and effects on performance of host ramets


Jesús Piqueras. (Fax +46-8-162268; e-mail:


1 The effects of fungus infection on a natural population of the pseudoannual plant Trientalis europaea were followed for 4 years.

2 The incidence of the disease was low, showed little temporal variation during the period of study and was not affected by ramet size. Disease reduced flowering, fruiting, stolon length and the number and size of daughter tubers, all of which were positively correlated with ramet size. The year–disease interaction was not significant, except for flowering, suggesting little variation in the aggressiveness of the pathogen.

3 Disease reduced survival of ramets to the end of the growing season, although the effect varied with ramet size, and decreased tuber survival both by reduction of tuber size and by reduction of the overwintering ability of tubers of a given size.

4 For two of the three annual transitions the size of the offspring ramets was affected negatively by infection in the previous year.

5 Disease transmission occurred along the stolons of only 31% of the diseased ramets. The probability of disease being shown in the following year decreased with stolon length.

6 Although disease had a detrimental effect on ramet fitness, the low level of incidence and the stability of the clone dynamics in simulation models suggest only a minor role of the disease in population regulation in this species.


Pathogenic fungi that reduce individual fitness may play an important role in the regulation of population growth (Burdon 1987a; Augspurger 1988). The impact of disease on a population of a particular plant species may depend on both the incidence of the disease and on its effect on plant performance. Studies of pathogen–host interactions suggest that there is substantial variation between species in the incidence and patterns of temporal variation in pathogen attack (Carlsson & Elmqvist 1992; Jarosz & Burdon 1992; Johansson 1993; Wennström & Ericson 1994; Wennström 1994; Ericson & Wennström 1997; Ericson et al. 1999). In a comparative study of systemic fungal diseases, Wennström (1994) showed that most species that are characterized by low infection rates exhibit small temporal fluctuations, whereas species with high infection rates display considerable between-year variation in the level of disease. There may also be considerable temporal variation in disease incidence among individual plants within a population, such that some may experience consistently high levels of disease while others remain healthy for long periods (Wennström & Ericson 1994; Ingvarsson & Ericson 1998). This can be explained by differences in the selective action of disease vectors (Carlsson-Granér et al. 1998; Ingvarsson & Ericson 1998), environmental factors (Jerling 1988; Wennström & Ericson 1990; García-Guzmán et al. 1996), reduced efficiency of disease spread (Alexander 1984) or individual variation in genetic and phenotypic characteristics (Alexander et al. 1984; Burdon 1987a, 1987b; Thrall & Jarosz 1994). Of the phenotypic traits analysed, plant size has been demonstrated to be important for susceptibility to disease (Johansson 1993; Wennström & Ericson 1994; Folgarait et al. 1995) and can also be decisive in the outcome of the pathogen attack. Fungal pathogens can alter host fitness by affecting survival, reproduction, competitive ability, defence against herbivores and growth (Burdon 1987a; Augspurger 1988; Jarosz & Davelos 1995). However, the effects of the disease may differ between host-plants of different size and, as most features of ramet performance and survival are largely size-dependent (Alexander & Burdon 1984; Wennström & Ericson 1991; Johansson 1994), may be confounded by initial differences in size between healthy and diseased individuals or by pathogen-induced changes of plant size. Additionally, changes in plant size can alter the ability to compete with neighbours and thus affect the current and future size and structure of the plant population (Paul & Ayres 1986; Wennström & Ericson 1991),

The dynamics of disease caused in the pseudoannual clonal plant Trientalis europaea by the specific systemic smut fungus Urocystis trientalis was investigated. A 4-year field study of the incidence and effects of the disease and its vegetative disease transmission was used to address two main questions: (i) How do incidence and transmission of disease vary between years and among ramets within a clonal population? and (ii) How does the pathogen affect clonal performance, sexual reproduction, survival and growth of the host ramet?

Materials and methods

The study species

The host, Trientalis europaea L. (Primulaceae), is a clonal understorey herb that is widely distributed in boreal woodlands of the northern hemisphere and is characterized by a pseudoannual life cycle (sensuWarming 1918). The mother plant dies back at the end of the growing season (September in the south of Sweden) after having produced tubers that will give rise to new ramets in the following May. As the stolons connecting the mother ramet with newly formed daughter tubers decay at the same time, the future ramets become physiologically independent immediately after their formation. Underground stolons are initiated in July and single tubers develop at their tips during August. If several stolons are produced the longest stolon almost invariably bears the largest tuber (hereafter referred to as ‘main stolon’ and ‘main tuber’, respectively; cf. Piqueras & Klimeš 1998). Flowering occurs between May and June and, usually, a single flower is produced by each shoot. Seed set is low (6.1 ± 2.5 seeds per fruit; mean ± SD; n = 102). Seedling recruitment is rare in established populations and seems to occur only under very favourable moisture conditions combined with disturbances of the forest ground (Hiirsalmi 1969; Eriksson & Ehrlén 1992)

The pathogen Urocystis trientalis (Berk. and Br.) B. Lindeb. (Ustilaginales) is an obligate systemic pathogen restricted to T. europaea. Disease symptoms are first observed in June as a white conidial spore stage on the lower surface of the leaves, and is followed in the late summer by a teliospore stage characterized by black pustules on leaves and stalks. The disease is not seed-transmitted but systemic infection of the vegetative offspring may occur through stolons (Wennström & Ericson 1990). There is no evidence that the smut U. trientalis can live asymptomatically in host-plants (Ericson & Wennström 1997). Teliospores are dispersed and infection takes place in autumn (Woronin 1882, in Wennström & Ericson 1990). Urocystis trientalis is common throughout Sweden, especially in coastal localities (Lindeberg 1959; Carlsson et al. 1990).

Demographic study

Demography and clonal spread of T. europaea were studied in a mesic spruce forest about 45 km south-south-west of Stockholm, Sweden (58°59′N, 17°33′E) from 1991 to 1994. The dominant tree species were Picea abies and Pinus sylvestris and the field layer consisted mainly of Vaccinium myrtillus, V. vitis-idaea, Maianthemum bifolium and Linnaea borealis. In June 1991, 23 permanent plots (200 × 200 cm in size) were established between 2 and 10 m apart in a representative area of the forest. Eight points were marked in a regular pattern in each plot. A radius of 20 cm around each point was searched and, if any were present, the position of the T. europaea ramet closest to the centre was mapped (at least six individuals per plot). This procedure ensured a representative sample of the population while allowing sufficient distance between the selected ramets for their underground parts to be distinguished.

In June 1991, leaf area (sum of the area of all leaves), number of flowers and disease symptoms (presence of conidial spore stage) were recorded for each ramet. The area of a single leaf was assessed as the product of leaf length and leaf width. The estimated leaf area correlated well with the dry weight of the ramet (above-ground vegetative structures, r = 0.86, P < 0.001, n = 12; J. Piqueras, unpublished data) and was used as measure of ramet size. Fruiting frequency and seed set were recorded in the first week of August. At the beginning of September, careful excavation allowed exposure of stolons and their daughter tubers. Stolon length and daughter tuber size (distance between the connection with the stolon and the base of the shoot meristem that would develop the following year) were measured using dividers and a ruler. The positions of the daughter tubers were mapped and marked with small, green, plastic wires and the soil was then replaced. Ramets of T. europaea live for less than 1 year, and a ramet was therefore considered to have survived the growing season if it had produced at least one daughter tuber (ramet survival). Overwinter survival was recorded for those tubers that developed into a new ramet in the following spring (tuber survival). In June 1992, the size and disease symptoms on the ramets rising from the surviving tubers were recorded. These measurements (i.e. ramet size, flowering, presence of disease symptoms, fruiting, seed set, stolon length, tuber size and survival) were repeated for descended ramets arising from surviving tubers during 1992–94. A final census, recording only disease symptoms, was made in June 1995.

Additional diseased ramets

Due to the low frequency of diseased ramets in the demographic study (31 out of 565 ramets surveyed from 1991 to 1994, see the Results), data were supplemented by measurements of all diseased ramets located into the plots during systematic searches of the plots in June in 1991, 1992, 1993 and 1994. Each of these selectively chosen ramets (n = 207) was recorded in the same way as the ramets of the demography study during the year of selection and for survival, presence of the disease symptoms and subsequent ramet size in the following spring.

Statistical analyses

Differences among years in the proportion of diseased ramets were analysed using contingency tables with data from the demographic study. Except for the effect of disease on ramet size in the following year, all subsequent analyses were performed with data pooled over all 4 years combined with data from selectively chosen ramets. Because different additional ramets were included each year, it was not possible to use repeated analyses correction for time autocorrelation in the studied variables. The Durbin–Watson test (Neter et al. 1996) was applied to detect the presence of autocorrelated error terms for continuous variables. Variables with binary response were analysed by logistic regression (logistic regression module in Statistica 1994) after the use of plots of residuals against time, to detect patterns of autocorrelation. The effect of ramet size on the probability of a ramet being diseased was analysed by logistic regression (binary response, 0 = healthy ramet, 1 = diseased ramet). The effects of disease, year and ramet size on tuber size, stolon length and seed set were analysed by analysis of covariance (ancova). As the number of diseased ramets with more than one tuber was low (9.2%, n = 238), analyses were only performed on the main stolon and the main daughter tuber. The effects of disease, year and ramet size for flowering frequency (non-flowering ramets = 0, flowering ramets = 1), fruiting frequency (flowering ramets that did not produce fruit = 0, flowering ramets that produce fruit = 1) and tuber production (ramets producing one daughter tuber = 1, ramets producing two or more daughter tubers = 2) were analysed by logistic regression after binary categorization of the variables. This procedure was motivated by the low number of ramets producing more than one flower (2.7%, n = 803) or more than two tubers (3.2%, n = 803). The effects of disease, year and ramet size on ramet survival (ramet producing no daughter tuber = 0, ramet producing daughter tubers = 1), and disease, year and tuber size on tuber survival, were analysed by logistic regression. Ramet size was entered as a continuous variable, disease as a binary variable (as above) and year as a dummy variable (categorical variable coded as a set of binary variables; Hosmer & Lemeshow 1989; McCullagh & Nelder 1989). The fitness of the model was calculated using likelihood ratio statistics, whereas the significance of the effects was estimated with the Wald test (Hosmer & Lemeshow 1989). The odds ratio in logistic regression is the multiplicative factor that describes the change in the probability of an event (e.g. of survival) when the independent variable increases by one unit. Values above 1 show that the frequency of the event increases, while values between 0 and 1 indicate that it decreases. The effects of disease and ramet size on the ramet size in the subsequent year were analysed by ancova. As a ramet can generate several ramets in the following year, these analyses were performed on the ramet growing from the main daughter tuber. Finally, the effects of stolon length, year and ramet size on the probability of disease transmission (healthy ramet descended from a diseased ramet = 0, diseased ramet descended from a diseased ramet = 1) through stolons in the following year were analysed by logistic regression. Leaf area and length of stolons and tubers were log-transformed before calculations.


In this population of T. europaea, the proportion of ramets that showed symptoms of infection by U. trientalis was rather low (Fig. 1) and did not vary significantly between years (χ2 = 1.59, d.f. = 3, P = 0.660) or between plots (χ2 = 29.32, d.f. = 22, P = 0.136). Of 147 ramets originally selected for the demographic study, 97 were still represented by through descendant in June 1995. Nine of these showed symptoms of disease in a single year and only two for 2 consecutive years: none showed disease symptoms for more than 2 years. Of all diseased ramets recorded in the plots between 1992 and 1994, 79% (n = 178) were newly infected (i.e. they descended from a healthy ramet). The risk of suffering the disease was not correlated with ramet size [–LogLikelihood = 0.139, χ2 = 0.279, d.f. = 1, P = 0.596, n = 803 (238 diseased), logistic regression].

Figure 1.

Temporal patterns of the disease caused by the fungus Urocystis trientalis on its host Trientalis europaea from 1991 to 1994. (a) Proportion of ramets diseased (data from the demographic study alone, n = 565). (b) Proportion of ramets surviving the growing season (white bars = healthy ramets, n = 565; black bars = diseased ramets, n = 238). (c) Proportion of tubers surviving winter (white bars = tubers from healthy ramets, n = 527; black bars = tubers from diseased ramets, n = 173). (d) Proportion of diseased ramets that survived that showed disease symptoms in the following year (n = 132). (b), (c) and (d) combine data from the demographic study and additional diseased ramets.

The Durbin–Watson test (for a level of significance = 0.05) and graphic diagnosis of residuals for binary data did not suggest autocorrelation for any of the studied variables over time. The fungal disease had a significant negative effect on the size of the main daughter tuber (3.7 ± 2.4 vs. 4.8 ± 2.2 mm; mean ± SD for diseased, n = 173, and healthy, n = 527 individuals), the length of the main stolon (43.1 ± 46.9 vs. 78.6 ± 79.9 mm; Table 1a) and on flowering frequency, fruiting frequency and tuber production (1.14 ± 0.4 vs. 1.32 ± 0.56 daughter tubers per ramet; Table 1b). Only flowering frequency showed a significant interaction between year and disease (Table 1). None of the disease–size interactions in ancovas and logistic regressions was significant (P > 0.1); with the exception of fruiting frequency, all variables studied showed a highly positive correlation with ramet size (Table 1).

Table 1.  Effects of disease caused by the fungus Urocystis trientalis on ramet performance of its host Trientalis europaea. (a) Effects of year, ramet size (ln leaf area in mm2) and disease on the continuous variables tuber size (ln length in mm), stolon length (ln length in mm) and seed set were analysed by ancova. (b) Effects of year, ramet size (ln leaf area in mm2) and disease (diseased ramets = 0, healthy ramets = 1) on the binary variables flowering frequency (0 = non-flowering ramet, 1 = flowering ramet), fruiting frequency (0 = fruiting ramet, 1 = non-fruiting ramet) and tuber production (1 = one daughter tuber, 2 = two or more daughter tubers) were analysed by logistic regression. Disease–size interaction was included in a first run of the models but in no case was significant (P > 0.1) and was therefore removed from the models presented
Tuber size
 Disease149.32 < 0.001 
 Size691500.89 < 0.0010.648
Stolon length
 Year37.91 < 0.001 
 Disease173.48 < 0.001 
 Size1244.87 < 0.0010.511
Seed set
 Effectd.f.Wald-χ2POdds ratio (95% CI)
Flowering frequency
 Disease120.85 < 0.0010.58 (0.47–0.73)
 Size1106.57 < 0.0013.85 (2.98–4.97)
n = 535 ramets (268 flowered), –LogLikelihood = 136.02, χ2 = 272.05, d.f. = 8, P < 0.001
Fruiting frequency
 Disease111.10 < 0.0010.44 (0.28–0.71)
n = 268 flowering ramets (105 with fruits), –LogLikelihood = 11.53, χ2 = 23.06, d.f. = 8, P = 0.003
Tuber production
 Disease111.08 < 0.0010.63 (0.49–0.82)
 Size147.81 < 0.0012.44 (1.92–3.09)
n = 700 (530 with 1 tuber, 170 with ≥ 2 tubers), –LogLikelihood = 61.51, χ2 = 123.02, d.f. = 8, P < 0.001

Diseased ramets experienced lower survival than healthy ramets (72.6% vs. 93.6%; Fig. 1). There was a significant interaction between disease and ramet size (Table 2). This meant that survival of healthy and diseased ramets could not be compared over the whole range of sizes, and therefore the model was rerun replacing ramet size and its interaction with values for ramet size only for the individuals in each particular group in turn (cf. Sokal & Rohlf 1995). The log likelihood was the same as before, but now there were separate parameters for the logistic regression on size within each group. Diseased and healthy ramets differed significantly in the slopes of their regression lines; in both, survival increased with ramet size, but this was only significant for healthy individuals (Table 2). Main daughter tubers originating from diseased ramets had a lower survival than those originated from healthy ramets (diseased = 76.3%, healthy = 90.1%; Fig. 1), even when the effect of tuber size was taken into account (Table 2).

Table 2.  Effects of disease on host survival. (a) Effects of year, ramet size (ln leaf area in mm2) and disease (diseased ramets = 0, healthy ramets = 1) on the probability of ramet survival were analysed by logistic regression. Because the interaction disease–size was significant, parameters were estimated separately for healthy and diseased ramets, excluding any year effect (see text for explanation). (b) Effects of year, tuber size (ln length in mm) and disease on the probability of tuber survival analysed by logistic regression. Disease–tuber size interaction was included in a first run of the model but it was not significant (P > 0.1) and was therefore removed from the model presented. Parameters of the model were estimated without a year effect
Effectd.f. Wald-χ2P
Year3 4.750.190
Disease1 2.210.136
Year–disease3 1.440.695
Size1 33.31 < 0.001
Disease–size1 12.50 < 0.001
Size (healthy)0.720.1050.42 < 0.001
Size (diseased)
(b) Tuber survival
d.f. Wald-χ2P
Year3 3.680.297
Disease1 6.580.010
Year–disease3 2.560.464
Size (tuber)1 44.23 < 0.001
Size (tuber)1.140.1744.17 < 0.001

Ramet size in 1992 and 1994 was significantly affected by fungal infection in the previous year (Table 3). On average, the reduction in ramet size was 23.9% (SD = 26.6, n = 69; ln-transformed leaf area). Thus, in 1992 a ramet was smaller if it descended from a 1991 ramet that was diseased (379 ± 531 mm2, n = 36 vs. 1019 ± 741 mm2, n = 125); equivalent values in 1994 were 235 ± 306 (n = 33) and 342 ± 369 mm2 (n = 127). In 1993, however, no difference in ramet size was found between the descendants of diseased and healthy ramets (Table 3). To discern whether this difference among years in the effect of disease on ramet size could be related to differences in tuber size in the previous year, the size of those tubers that survived the winter and raised ramets in the following spring was compared. Such tubers were smaller if they arose from diseased compared with healthy ramets in both years when there was a disease effect (4.0 ± 2.4 mm, n = 36 vs. 5.8 ± 1.7, n = 125, t = 6.34, P < 0.001, and 3.1 ± 1.7 mm, n = 33 vs. 4.3 ± 2.3, n = 127, t = 2.41, P = 0.016, for t-tests for 1991–92 and 1993–94 survival) but not for the year when there was no effect (5.2 ± 2.6 mm, n = 35 vs. 5.4 ± 1.8, n = 130, t = 1.41, P = 0.159, for 1992–93).

Table 3.  Effects on ramet size of disease and ramet size (Ln leaf area in mm2) in the previous year analysed by ancova
Disease135.43 < 0.001 
Size (1991)148.72 < 0.0010.485
Size (1992)177.23 < 0.0010.567
Disease111.25 < 0.001 
Size (1993)184.21 < 0.0010.590

Overall, 31% (n = 132) of the descendants from diseased ramets showed symptoms of disease in the following year. This probability of disease transmission through stolons did not vary between years (Wald-χ2 = 3.57, d.f. = 3, P = 0.311; Fig. 1) but decreased with stolon length (Wald-χ2 = 19.54, d.f. = 1, P < 0.001; Fig. 2). For easier interpretation of the results, this analysis was performed on non-transformed stolon length. On average, a 1-mm of increase in stolon length decreased the risk of infection 0.93 times (odds ratio, 95% CI = 0.91–0.95), with the more pronounced decrease in the range 0–50 mm. Although stolon length was positively correlated with ramet size (r2 = 0.32, F = 336.98, P < 0.001, n = 700, linear regression), the model showed ramet size did not have a significant effect on disease transmission through stolons (Wald-χ2 = 0.22, d.f. = 1, P = 0.632; Fig. 2). Large ramets with short stolons experienced a large probability of re-infection, whereas small ramets with relatively long stolons tended to escape from the disease (Fig. 2).

Figure 2.

Disease transmission along stolons in Trientalis europaea. Effects of stolon length (mm) and ramet size (ln leaf area in mm2) on the probability of disease transmission along stolons. Dots at the top of the 3-dimensional plot represent diseased ramets that survive and show disease symptoms in the following year. Dots at the bottom represent ramets that escape from disease. The surface representing the risk of disease transmission as a function of stolon length and ramet size was fitted to the observed values in the logistic regression model [–LogLikelihood = 23.48, χ2 = 49.96, d.f. = 2, P < 0.001, n = 132 (41 diseased)].


The disease caused by U. trientalis in its host T. europaea in this study showed a low incidence and smaller temporal fluctuations than in previous field experiments with the same pathogen–host interaction (Ericson & Wennström 1997). These differences in the level of incidence might be explained by regional variation (García-Guzmán et al. 1996) or by difference in age between the populations (Carlsson et al. 1990), while the temporal stability reported here could be related to differences in herbivore pressure. Indeed, when vertebrate herbivores were excluded from experimental plots of T. europaea levels of the disease did not display a significant variation between years (Ericson & Wennström 1997), and vertebrate herbivore pressure at the field site was probably too low to affect the levels of the disease (Piqueras 1999). The results of the demographic study also showed that the disease did not persist through the descendant ramets for more than 2 years. Thus, diseased ramets either died or became healthy within a few generations. This contrasts with the results of a study on the systemic smut U. carcinoides, in which host-plants (Actaea spicata) remained diseased for 8 years, possibly because the effects of the disease on the host ramet performance (Wennström & Ericson 1994) were not severe.

Fungal disease caused by U. trientalis had a negative effect on almost all stages of the life cycle of T. europaea, reducing the probability of flowering and fruiting, as well as lowering ramet survival, tuber production, stolon length, tuber size and tuber survival. All variables studied, except fruiting frequency, were positively correlated with ramet size. This suggests, as reported in other studies (Hiirsalmi 1969; Anderson & Beare 1983), that fruit set is limited by other factors, probably pollen supply, in T. europaea.

Although ramet survival decreased as a result of disease, the difference in the regression lines between diseased and healthy ramets indicated that ramet size was less decisive for the survival of diseased ramets. Small plants are typically much more vulnerable to mortality factors (Watkinson 1997) and it is commonly assumed that vigorous individuals are better able to survive pathogen disease. Although this assumption is supported by studies on plant–pathogen interactions in the early recruitment phase (Augspurger 1984; Paul & Ayres 1986; Burdon 1987a), detailed epidemiological studies that consider the effect of size on survival are scarce.

Although tuber survival was relatively high both for healthy and diseased individuals over the 4 years of study, disease significantly reduced tuber survival. Reduction of tuber size in diseased ramets had a negative effect on tuber survival, but the consistent negative effect of the disease on tuber survival after correction for size suggests that tubers from diseased ramets experienced a significant decrease in their overwintering ability. In contrast, vertebrate herbivory reduced tuber survival in T. europaea, but only as consequence of size reduction (Piqueras 1999).

Of all the variables studied, only flowering frequency showed a significant interaction between years and disease, suggesting that the aggressiveness of the disease in this population of T. europaea varied little. Temporal variation in the effects of fungal disease has been rarely considered, although Thrall & Jarosz (1994) found a significant between-year variation in mortality of Silene alba caused by Ustilago violacea.

Disease affected subsequent ramet size, but not in all years. Interestingly, the lack of effect in one year seemed to be a combination of the effects of the disease in tuber size and tuber mortality. In autumn 1992, tubers from diseased ramets were smaller than those produced from healthy ramets (P = 0.004, d.f. = 187, t-test) but the smallest diseased tubers showed increased mortality. Thus, by the following spring, ramets originated by diseased ramets differed only marginally from those originated by healthy ramets. This reflects the complex relationships between ramet size, survival and disease, and suggests that it will be difficult to assess the effects of pathogens on population structure from analysis of the effects on single fitness traits.

Disease transmission did not necessarily occur along stolons of T. europaea. This has been observed in other host–systemic smut systems (Wennström & Ericson 1990, 1992; Frantzen 1994b; García-Guzmán & Burdon 1997) and suggests that escape from disease may occur as a result of clonal growth. One possible explanation might be that the fungal mycelium grows only slowly within stolons of T. europaea, and the first 50 mm of the stolon appear to be crucial in determining growing away from the disease. Cytological studies of the systemic rust Puccinia suaveolens revealed that the growth rate of the systemic mycelium within roots of Cirsium arvense was much lower than in other parts of the plant (Menzies 1953). The results here show that the probability of disease transmission was correlated with stolon length but not with ramet size. Thus, it can be hypothesized that factors that promote lengthening of stolons may prevent disease transmission in this species. Favourable light conditions have been found to have a positive effect on stolon length in T. europaea in natural populations (Hiirsalmi 1969; J. Piqueras, personal observation) and Wennström & Ericson (1990) found that the probability of daughter tubers from infected parent plants of T. europaea developing into diseased ramets was lower under high light conditions. The almost complete absence of the disease in open areas of the forest (J. Piqueras, personal observation; cf. Wennström & Ericson 1990) may be due to these effects of light.

Alternatively, the incomplete transmission of disease might be caused by a rather short dispersal range for teliospores above ground. However, most of the diseased ramets (79%) were descendants of healthy ramets in the previous year, indicating that the teliospores are in fact the main mechanisms of infection and maintenance of the disease in the population. This seems to be in accordance with an epidemiological study of Puccinia punctiformis on its host Cirsium arvense, which showed that systemically infected shoots were the primary sources of uredinospores causing local infection of neighbouring shoots (Frantzen 1994a). It was also shown in this study that disease escape was due to the dispersal of teliospores over short distances compared with the spread of ramets from their mother. It is therefore possible that the ability of T. europaea to escape from disease, either by growing away from the mycelium or by avoiding infection by teliospores, might depend on the pattern of teliospore deposition and the distance at which daughter tubers are placed. Although stolon length was severely reduced in diseased ramets of T. europaea, escape may still be possible if the daughter tubers are placed at more than a critical distance from the mother ramet.

It has been suggested that systemic fungal pathogens characterized by high virulence may select patterns for clonal growth that allow the host-plant to escape from disease, for instance by increasing the distance between ramets or by fragmentation (Wennström & Ericson 1992). Pseudoannual plants can adopt both strategies and differ from most perennials in that the mother ramet dies back at the end of the growing season. This means that, in order to survive vegetatively, the fungus has to grow along the stolons and infect the next generation of ramets as they are formed. It is difficult to assess whether systemic pathogen infection may have played an important role in the origin of the pseudoannual life cycle. The number of species with this growth habit is not large (Klimešet al. 1997) and knowledge about the occurrence of fungal disease in this group is limited. However, many pseudoannuals occur in wet habitats (Warming 1918), where the presence of fungal pathogens can be important (Johansson 1993; Jerling & Berglund 1994).

Although the disease caused by U. trientalis had a negative effect on most of the fitness traits of T. europaea ramets, it is not easy to assess its effect on long-term regulation of the population. The disease showed a low incidence, low temporal fluctuation and its effects on ramet performance, although significant, were not dramatic. Reduced flowering and fruiting in particular will have a rather marginal effect in established populations of this species, where seedling recruitment is rare. Simulation models based on demographic data from the same population (Piqueras & Klimeš 1998) revealed that the dynamics of clonal fragments (aggregations of ramets derived from a common parent ramet) was very stable to changes in survival probabilities. The model also suggested that the production of secondary daughter tubers had very little effect on population dynamics. However, the same model predicted that reduction in size of the main daughter tuber (one of the effects of the disease) had a highly detrimental effect on the persistence and number of ramets of the clonal fragment.


I thank Lenn Jerling, Patrik Dinnétz, Graciela García-Guzmán, Bernhard Schmid, Andrew Jarosz, Lindsay Haddon and an anonymous referee for discussions, comments and many valuable suggestions on earlier versions of the manuscript.

Received 8 December 1998 revision accepted 8 June 1999