Male and female Silene latifolia plants differ in per-contact risk of infection by a sexually transmitted disease

Authors

  • Oliver Kaltz,

    Corresponding author
    1. Experimentelle Ökologie, Eidgenössische Technische Hochschule (ETH) Zürich, ETHZ-NW, 8092 Zürich, Switzerland
    Search for more papers by this author
  • Jacqui A. Shykoff

    1. Laboratoire d’Evolution et Systématique, CNRS URA 2154, Université de Paris-Sud (XI), Bâtiment 362, 91405 Orsay Cedex, France; and
    Search for more papers by this author

and present address: Oliver Kaltz, Department of Biology, McGill University, 1205 Docteur Penfield Avenue, Montreal, Quebec H3A 1B1, Canada (tel. 514-398-6459, fax 514-398-5069, e-mailoliver.kaltz@esv.u-psud.fr).

Summary

  • 1 Behavioural, physiological or immunological constraints often render one sex more susceptible to parasites, thereby potentially generating sex-specific trade-offs between traits associated with infection risk and other life-history characters.
  • 2 The fungal pathogen Microbotryum violaceum systemically infects the dioecious plant Silene latifolia when pollinators deposit fungal spores on the flowers of healthy plants. Male plants produce many short-lived flowers, whereas females produce few flowers that remain connected with the plant after fertilization. We investigated how variation in flower production and flower longevity affects the infection risk for males and females.
  • 3 In glasshouse experiments, we varied the number of flowers inoculated (4 vs. 16 per plant) with spores and the time until these flowers were removed (1 or 2 days for both sexes, 14 days for females only). We also measured the longevity of male flowers receiving simulated visits, with or without spores, to test for an abscission response to visitation and/or contamination. In a field survey, we measured male and female disease prevalence in 17 natural populations.
  • 4 Varying the number of inoculated flowers did not affect infection probability, but females retaining inoculated flowers for 14 days became diseased more often (20.0%) than did plants with flowers removed within 2 days (7.3%).
  • 5 Males that had dropped more inoculated flowers prematurely were more likely to remain uninfected. Spore-bearing visits shortened male flower longevity (38.4 ± 2.8 h) relative to non-spore visits (47.9 ± 5.2 h).
  • 6 Female field disease prevalence (19.7 ± 3.5%) was higher than that of males (14.3 ± 2.6%), especially in populations with a high disease incidence.
  • 7 Continuing physical connection during fruit ripening appears to increase invasion time and thus the per-contact infection risk in females. This is consistent with higher female field prevalences, although other explanations, unrelated to disease transmission, are possible. These results illustrate how interactions between plant reproductive behaviour and pollinator activity may affect disease spread. Female mating behaviour may evolve towards lower attractiveness to pollinators to minimize infectious contacts, while males can afford to be more promiscuous with an attractive, but disposable, floral display.

Introduction

In many organisms, males and females differ in one or other of development, morphology, physiology, life history and behaviour. Such sexual dimorphism is often considered to be the direct or indirect result of intra- and intersexual selection (Darwin 1871; Andersson 1994; Short & Balaban 1994) for traits that maximize the acquisition of mates or mating opportunities.

However, complications may arise if sexual selection produces conflicts with other aspects of the life history of an organism, particularly if such conflicts are different for males and females. Constraints or trade-offs may explain the sexual dimorphism in disease susceptibility that is found in certain host–parasite systems in vertebrates (Poulin 1996; Schalk & Forbes 1997; Møller et al. 1998) and invertebrates (Wedekind & Jakobsen 1998). In vertebrates, susceptibility to parasites may be higher in males due to the presence of secondary sexual characteristics associated with negative effects on the immune system, produced as a result of stronger sexual selection on males (Alexander & Stimson 1988; Zuk 1990; Folstad & Karter 1992; Zuk & McKean 1996). However, susceptibility bias is not always the result of an impaired immune system, and not always towards males. In humans, for example, women may become infected with sexually transmitted diseases more easily than men because of the retention of the ejaculate in their reproductive organs (Hook & Handsfield 1990; Ewald 1994). Furthermore, differences between the sexes in their behaviour or ecology may lead to higher parasite encounter, and thus higher infection rates, of one or other sex (Bundy 1988; Zuk & McKean 1996; Schalk & Forbes 1997).

In both animals and plants, sexual dimorphism in susceptibility to disease may influence epidemiology as well as the evolutionary trajectories of traits involved in sexual selection and/or host–parasite interactions. Recent theoretical work has suggested that optimal mating strategies in the presence of sexually transmitted disease may not necessarily be those which minimize the risk of infection and disease spread in a population (Guldbrandtsen 1997; Thrall et al. 1997). Although monogamy may reduce infection risk (Loehle 1995), limited promiscuity can generate similar or even higher reproductive success, despite the higher chance of becoming infected (Thrall et al. 1997). Moreover, if males and females differ in their probabilities of encountering disease and becoming infected thereafter, varying constraints and trade-offs between selection for minimizing disease risk and selection for maximizing reproductive success may lead to different optimal mating strategies in the two sexes.

We investigated differences in infection probability for males and females of the dioecious plant Silene latifolia (Caryophyllaceae) to the anther-smut disease, Microbotryum violaceum. This fungal pathogen sporulates in the flowers of infected hosts, thereby sterilizing the host, and is mainly transmitted by pollinating insects (Roche et al. 1995), thus having the features of a sexually transmitted disease (Alexander & Antonovics 1988; Thrall et al. 1993; Lockheart et al. 1996).

Sexual activity, i.e. flowering behaviour, influences the probability of disease contact and subsequent infection in S. latifolia. Plants producing more flowers, flowering earlier or longer in the season (Alexander & Antonovics 1988; Alexander 1989; Thrall & Jarosz 1994b; Biere & Antonovics 1996; Biere & Honders 1996b), or presenting more receptive surface (Elmqvist et al. 1993) are more likely to become infected, partly because some of these traits are negatively correlated with biochemical resistance (Biere & Antonovics 1996), but also because such plants receive more visits (Shykoff & Bucheli 1995) and thus become contaminated with fungal spores more often. Hence, evolution of floral traits and breeding systems in such cases may be determined by the interaction with this pathogen as well as by sexual selection (Elmqvist et al. 1993; Skogsmyr 1993; Thrall et al. 1993; Meagher 1994; Shykoff et al. 1997).

Because male plants of S. latifolia produce more flowers (e.g. Gross & Soule 1981; Delph & Meagher 1995; Carroll & Delph 1996) with more nectar sugar (Shykoff & Bucheli 1995; Shykoff & Kaltz 1998; but see Biere & Honders 1996a) and receive more pollinator visits than do females (Shykoff & Bucheli 1995; Altizer et al. 1998), exposure, and hence infection, is likely to be more frequent in males than in females. Indeed, male-biased infection rates have been documented (Alexander 1989; Thrall & Jarosz 1994b; Alexander & Antonovics 1995; Biere & Antonovics 1996; Biere & Honders 1998), but are not universal. Sometimes there is no sex difference in disease prevalence (Zillig 1921; Alexander 1990; Thrall & Jarosz 1994b), and a recent survey across several natural populations of S. latifolia and its close relative S. dioica even found, on average, higher disease prevalence in females (Shykoff et al. 1996; see also Lee 1981; Alexander & Antonovics 1988; Alexander 1989). This may result from a higher per-contact infection risk (Alexander 1989; Shykoff et al. 1996) because females retain physical connections with fertilized flowers during fruit development, whereas male flowers drop within a short time. Thus fungal spores deposited on female flowers may have more time to initiate an infection than spores deposited on male flowers.

We tested the effects of frequency of infectious contacts and per-contact infection risk on infection rates in males and females in two glasshouse experiments. In the first, we inoculated a variable number of flowers, which were removed from the plant after a range of times. In the second, we measured flower longevity following artificial visits that did or did not deposit fungal spores, to test whether males potentially reduce the per-contact infection risk by dropping visited and/or contaminated flowers. In addition, we measured disease prevalence of males and females in 17 natural populations of S. latifolia.

Methods

Study organisms

Silene latifolia[=S. alba (Miller) Krause; =Melandrium album (Miller) Garcke] is a short-lived perennial of the family Caryophyllaceae. The life span of flowers on male plants in the field is 1.5 ± 0.5 days (Shykoff et al. 1996). Unpollinated flowers of females remain accessible for several days (Alexander 1987), but start wilting and developing into fruits within 12 h of pollination, generally by noctuid moths (Shykoff & Bucheli 1995; Altizer et al. 1998), which can also deposit up to 106 fungal spores per flower (Alexander 1990; Antonovics & Alexander 1992; Bucheli & Shykoff 2000).

After deposition on a new host, germination and meiosis of teliospores of Microbotryum violaceum[(Pers.) Deml & Oberw.; =Ustilago violacea (Pers.); Ustilaginales: Basidiomycetes] give rise to a promycelium, and haploid sporidia of opposite mating types must conjugate to initiate the growth of dikaryotic infection hyphae (Day & Garber 1988; Hood & Antonovics 1998). Initiation of infection hyphae occurs within 24 h of inoculation of flowers with teliospore suspensions (O. Kaltz, unpublished data), but the timing in vivo is not known.

Experiment a

Experimental design

We manipulated the number of flowers receiving spores (4 or 16 per plant), and the time between inoculation and removal of these flowers (≈ 24 h and 48 h for both sexes, 14 days for females). These timings are likely to reflect natural exposure times in male and female flowers, and the 14-day maximum ensured that fruit set did not reduce further flower production below the 16 needed for possible inoculation. Seeds were collected during 1996 from 16 natural populations (including 14 described in Fig. 1 and Table 1 in Kaltz et al. 1999). Following germination and growth to flowering stage, plants were arbitrarily assigned to treatments. From five of the source populations (populations a, d, i, n and o, Table 1 of Kaltz et al. 1999), infected plants provided teliospores from seven fungal strains. A mixture of these strains was freshly prepared for each series of inoculation by collecting teliospores in a microcentrifuge tube.

A toothpick was dipped into the inoculum and then inserted into the corolla tube of an freshly opened flower. We tried to ensure that both the sexual organs (stamens and anthers in males, ovary and stigma in females) and the inner side of the corolla received teliospores (≈ 7 × 105 spores per flower), with minimal damage (particularly to the stamens in male flowers). Unfertilized female flowers live for only a few days and females were therefore always pollinated before inoculation by rubbing an open male flower over the stigma, to allow the 14-day treatment. Inoculated flowers were then misted with water from a vaporizer to provide high-humidity conditions which facilitate fungal development (Alexander et al. 1993) and marked with cotton thread around the pedicel. Flowers or maturing fruits were subsequently removed at the base of the flower using scissors. For males, whose natural flower life span can be variable, we recorded whether any of the inoculated flowers had already fallen and how many were still present at the time of experimental removal (1 or 2 days).

We inoculated 252 male and 242 female plants during 37 sessions between May and early July 1997, carried out between 21.00 and 02.00. In each session, we alternated inoculation of male and female plants, with up to four plants per sex and treatment combination. Males were distributed equally over the four treatment combinations. In order to obtain approximately equal numbers of females receiving long (14-day) and short (1 and 2-day) exposure times, we replaced two 1-day treatments with 14-day treatments in half of the sessions. Some plants received inoculations in more than one session. Inoculated plants were kept in the glasshouse and surveyed regularly for disease onset (production of spore-bearing flowers). Plants had stopped flowering by December 1997, when they were placed outside until February 1998 for vernalization. They were then brought to flower again in the greenhouse and monitored until June 1998.

Data analysis

Some males dropped inoculated flowers within under 24 h. We tested whether the proportion of prematurely dropped flowers was a function of inoculation date and/or mean glasshouse temperature (calculated from daily minimum and maximum temperatures) on the day(s) following a given session. For this multiple regression, individual flowers were treated as independent observations. We further compared the mean proportion of dropped flowers in plants that remained healthy or became diseased with a t-test, and used logistic regression to test the effect of the proportion of prematurely lost flowers on the probability of infection.

As the experimental design was not fully factorial, we tested the effects of flower treatments and sex on plant status (healthy or diseased) in three separate analyses of deviance. Analysis of deviance is based on a maximum likelihood approach and logistic regression to account for the binomial error structure of binary response variables (Crawley 1993). First, we considered only 1- and 2-day flower removal treatments in a factorial model with number of flowers inoculated, time of removal and plant sex as explanatory variables. Second, using only female plants, we included the number of flowers inoculated and all three removal times in a factorial model. Third, we pooled data for male and female plants across all treatments to analyse the overall effects of sex on plant status. We found that inoculation date affected infection probability (see Results) and therefore included this factor (in Julian days) as a covariate in all analyses. For plants with flowers inoculated on several dates we calculated a weighted mean inoculation date.

In all analyses, numbers, dates and temperatures were ln-transformed, and proportions arcsine-transformed. For the analyses of deviance, inoculation date was always fitted first to the models, whereas the other terms were fitted in accordance with SAS-type II model procedures (Schmid & Dolt 1994). Logistic regressions and analyses of deviance were performed with GLIM 3.77 (Baker 1987), correlations and t-tests with JMP 3.0 (SAS 1994).

Experiment b

Experimental design

Male plants, taken from the same source population as in experiment A, were randomly assigned to spore (n = 27) and non-spore (n = 28) treatments (hereafter referred to as ‘plant treatments’). Two newly produced flowers from neighbouring terminal branches were chosen at random on each plant for two ‘flower treatments’. Pollinator visitation was mimicked by gently inserting a fine paint brush (Pelikan™, no. 2) into the corolla tube of one flower (paintbrush flower treatment), while the other was not touched (control flower treatment). For spore treatment plants, the paintbrush was dipped into the spore mixture (see above) and ≈ 106 spores were transferred per flower; non-spore treatment plants received spore-free paintbrush visits. We separated spore/non-spore treatments at the level of the plant in case spore deposition affects flowers on a plant other than those directly contaminated. Alternating spore and non-spore treatments, replicate 1 of this experiment was established between 21.00 and 02.00, on 24–25 April 1998, and the same treatment was reapplied to each plant on 28–29 April 1998 (replicate 2). Experimental flowers were marked and misted as described above, and monitored at ≈ 2-h intervals during daylight hours (flowers are rarely dropped during the night).

Data analysis

Flower survival probability was not constant over time and therefore we used a maximum likelihood regression approach and a Weinbull survival function (Crawley 1993) to account for non-normality and non-linearity in the data. A separate survival analysis for each replicate tested effects of plant treatment (spore vs. non-spore), flower treatment (paintbrush vs. control) and their interaction on flower longevity. Orthogonal contrasts were used to compare flower longevity between spore and non-spore plants for each flower treatment. We further investigated whether the response of individual plants was consistent between replicates, by testing the correlation between the two replicates in mean flower longevity (averaged over the two flowers per plant) and in the difference between the two flower treatments on a given plant. Both analyses were carried out as repeated measurement analyses, with plant treatment as the explanatory variable. Analyses were performed with the SAS (SAS 1988) and JMP (SAS 1994) statistical packages.

Field survey

Between June and September 1996, we made single estimates of disease prevalence in each of 17 natural S. latifolia populations in northern Switzerland, France and Germany, some of which were the source of plant and fungal material for this and other studies (see Delmotte et al. 1999; Kaltz et al. 1999). We examined all flowering plants and determined the proportion of each sex that showed disease symptoms.

Data analysis

In an analysis of deviance, we tested the effects of plant sex and plant population on the weighted proportions of the number of diseased plants. In analogy to an analysis of variance on proportion data, population and sex were tested over the residual deviance (quasi-F-tests). This analysis was carried out in GLIM (Baker 1987). We further tested whether over- or underinfection of females relative to males varied with sampling date and disease frequency or density in a population. Using the difference between the (arcsine-transformed) proportions of diseased female and male plants, we performed multiple regressions with sampling date (Julian days, ln-transformed) and disease frequency (= overall proportion of diseased plants in a population, arcsine-transformed) or disease density (= overall number of diseased plants, ln-transformed) as explanatory variables. Note that estimates of density do not account for variation in plant spacing. Analyses were carried out in JMP (SAS 1994), using type-III sums of squares.

Results

Experiment a

Of the 494 inoculated plants, 19 had produced spore-bearing flowers by December 1997, and another 22 did so after vernalization. For plants that became diseased during 1997, mean latency (± SE), i.e. time from first inoculation until production of diseased flowers, was 111.7 ± 13.2 days (median: 88 days). Eight had still not yet flowered by June 1998, and these, together with 12 plants that died during 1997 and 67 that died during overwintering, were excluded from analysis because of potentially long latency.

Overall, we found a significant positive effect of inoculation date on infection probability (logistic regression: χ2 = 11.64, d.f. = 1, P = 0.0006) that was similar for males and females (sex × date interaction: χ2 = 0.64, d.f. = 1, n.s.). Plants were more likely to become diseased if inoculated later in the season [mean inoculation date (± SE) of diseased vs. healthy plants: 59.00 ± 2.60 vs. 50.11 ± 0.83; t-test: t = 3.36, d.f. = 405, P = 0.0006]. There was no significant correlation between session date and mean temperature (Pearson correlation coefficient: r = −0.01, n = 37, n.s.), and mean temperature on the 2 days following inoculation did not significantly affect infection probability (logistic regression: χ2 = 0.05, d.f. = 1, n.s.). However, daily maximum temperatures decreased markedly in early June [mean max. temperatures (± SE) before and after day 72: 29.75 ± 0.53 vs. 25.88 ± 0.62, t = 4.76, d.f. = 64, P < 0.0001]. Plants inoculated after this date had a significantly higher infection probability than plants inoculated before this date (χ2 test: χ2 = 5.19, d.f. = 1, P = 0.0227).

Of the inoculated male flowers, 13% were dropped within 24 h, and a further 36% fell off before the end of the second day (female plants lost no inoculated flowers). Inoculation date was negatively correlated with the proportion of inoculated male flowers dropped in less than 1 day (Pearson correlation coefficient, r =−0.60, P = 0.0001, n = 37), and dropped in less than 2 days (r = −0.41, P = 0.0216, n = 31 for some dates we did not record the fate of flowers for the second day). The (mean) daily temperatures following the inoculation dates were positively correlated with premature flower loss in less than 1 day (r = 0.48, P = 0.0024, n = 37) and less than 2 days (r = 0.48, P = 0.0058, n = 31). Males that remained healthy were more likely to have lost a higher proportion of inoculated flowers than had males that became diseased [proportion of dropped flowers (± SE) of healthy vs. diseased males: 0.31 ± 0.03 vs. 0.15 ± 0.06; t-test: t = 1.85, d.f. = 209, P = 0.0651]. However, the correlation of flower loss with inoculation date largely obscured any effect on infection probability in a multiple logistic regression (χ2 = 2.29, d.f. = 1, P = 0.1300) after fitting inoculation date. Fourteen male plants from the 2-day treatment had lost all inoculated flowers within 48 h, and were therefore considered as 1-day treatment plants in the following analyses.

In the 1- and 2-day treatments, only 23 of 317 plants produced diseased flowers (Fig. 1). Infection probability increased with inoculation date (χ2 = 11.67, d.f. = 1, P = 0.0006), but was not strongly affected by either of the two treatment types or by sex. Although the treatments had somewhat different effects in males and females (Fig. 1), neither sex × flower number × flower removal treatment interaction (χ2 = 2.49, d.f. = 1, P = 0.1146) nor any other model term was significant (χ2 < 1.28, all d.f. = 1, P ≥ 0.2579).

Figure 1.

Effect of time of flower removal on proportion of male and female Silene latifolia plants becoming diseased with Microbotryum violaceum. The number below each column indicates the number of flowers per plant inoculated (4 or 16) and that above the number of diseased plants.

Considering all treatment combinations, for female plants only, infection probability again tended to increase with inoculation date (χ2 = 3.70, d.f. = 1, P = 0.075). While neither the flower number (χ2 = 0.5, d.f. = 1, n.s.) nor its interaction with the removal treatment (χ2 = 1.80, d.f. = 2, n.s.) had significant effects on infection probability, infection was more likely when inoculated female flowers were retained for longer periods (effect of removal treatment: χ2 = 9.20, d.f. = 2, P = 0.0101). Twenty per cent of females developed disease symptoms when flowers and maturing fruits were retained until 14 days after inoculation, compared with 2.6% and 9% for 1- and 2-day treatments, respectively (Fig. 1). Aggregating 1- and 2-day removal treatments into a single treatment class did not significantly increase residual deviance (χ2 = 1.53, d.f. = 1, n.s.), and this combined treatment class produced significantly fewer infected plants than did the 14-day removal treatment (χ2 = 7.67, d.f. = 1, P = 0.0056).

The third analysis showed that, overall, after fitting inoculation date, there was a marginally significant difference (χ2 = 3.66, d.f. = 1, P = 0.0558) between the proportion of diseased plants in females and males (0.128 vs. 0.076). However, comparing the treatments which most closely represent the natural flower longevity and thus the time span of exposure to the fungus (14 days for females, 1–2 days for males), the proportion of diseased females was almost three times as high than that of males (χ2 = 9.60, d.f. = 1, P = 0.0019).

Experiment b

Replicates 1 and 2 of this experiment produced similar results, with significant effects of both flower and plant treatments on flower longevity (Table 1). Overall, insertion of the paintbrush into the corolla tube significantly reduced flower longevity, as did spore treatment (Fig. 2). Depositing fungal spores into the corolla tube reduced flower life span by ≈ 20% compared with flowers that received paintbrush visits without spores (Fig. 2); this was significant in both replicates (orthogonal contrast, Table 1). Control flowers were dropped earlier if the other flower on the plant received spores (Fig. 2), although this difference was only significant in replicate 1 (orthogonal contrast, Table 1).

Table 1.  Analysis of deviance testing effects of plant treatment (spore vs. non-spore), flower treatment (paintbrush vs. control) on flower longevity in two consecutive runs of the experiment (replicate 1 and 2)
Sourced.f.Deviance (replicate 1)Deviance (replicate 2)
  • ***

    P < 0.001;

  • **

    P < 0.01;

  • *

    P < 0.05.

Flower treatment  1 23.68***  7.59**
Plant treatment  1 12.45***  4.30*
Plant × flower treatment  1  0.54  0.70
Orthogonal contrasts:   
 Spore vs. non-spore (paintbrush flowers)  1  7.67**  4.24*
 Spore vs. non-spore (control flowers)  1  6.37*  0.77
Residual106119.21117.91
Figure 2.

Effect of paintbrush simulation of insect visitation with or without depositing Microbotryum violaceum spores into the corolla tube on mean longevity of male Silene latifolia flowers (± SE). Control flowers remained untouched. The same plants received the same treatment in each replicate run.

Repeated measure analysis revealed a significant positive correlation between mean flower longevity in replicates 1 and 2 (r = 0.43, d.f. = 53, P = 0.0012; Fig. 3a). Overall, flower longevity was significantly shorter in replicate 2 (F1,53 = 60.09, P < 0.0001) and the correlation in mean flower longevity tended to differ between spore and non-spore plants (F1,53 = 3.44, P < 0.0691). Furthermore, we found a weak, but significant positive correlation in the flower dropping response (the difference in longevity between control vs. paintbrush flowers) between replicates 1 and 2 (r = 0.28, d.f. = 53, P = 0.0378). Plants that dropped paintbrush flowers earlier than control flowers in replicate 1 were therefore likely to do so again when replicate 2 was run 4 days later (Fig. 3b). However, this response was less marked in replicate 2 (F1,53 = 18.48, P < 0.0001; 65%, rather than 78% of plants dropped the paintbrush flower first).

Figure 3.

Correlations between two replicate runs of the experiment for (a) male Silene latifolia flower longevity (averaged over the control and paintbrush flowers on each plant), and (b) flower dropping response (longevity difference between control and paintbrush flower on each plant). Linear regression lines are shown.

Field survey

Across populations, the proportion of flowering plants (healthy and diseased plants combined) that were female ranged from 43% to 67%, but there was no significant difference from a 1 : 1 male–female ratio (t-test on the arcsine-transformed proportion of female plants: t = 1.28, d.f. = 16, n.s.). There was no significant effect of sampling date on sex ratio across populations (F1,15 = 1.33, n.s.), and variation in sex ratio did not significantly influence the difference between disease prevalence in females and males (F1,15 = 0.25, n.s.; see below), suggesting that general sex differences in phenology had little effect on our estimates of sex differences in disease prevalence.

Mean disease prevalence across all populations was 0.171 ± 0.030 (median = 0.126, Table 2). There were significant effects of population (F16,16 = 17.24, P < 0.0001) and plant sex (F1,16 = 22.14, P = 0.0002) on the proportion of diseased plants, with higher infection rates in females (mean proportion infected = 0.197 ± 0.035) than males (0.143 ± 0.026). In 14 of the 17 populations, disease prevalence was higher in females by up to twofold (Table 2). Multiple regressions showed significant effects of sampling date and overall disease prevalence on the difference between females and males. Female overinfection decreased over the season (effect of sampling date in the model with disease density: F1,14 = 6.30, P = 0.0250), but increased with increasing total number (F1,14 = 14.88, P = 0.0017; Fig. 4) or proportion (F1,14 = 6.76, P = 0.0210) of diseased plants in a population. Interactions between sampling date and measures of disease incidence were not significant (F values < 1) and were therefore removed from the initial full models.

Table 2.  Proportion of male and female Silene latifolia plants diseased with Microbotryum violaceum in natural populations in the bordering region of France, Switzerland and Germany (except for Orsay, near Paris) during 1996, based on counting healthy and diseased flowering plants in each population
  Proportion of diseased plants (number diseased/total)
PopulationDateFemalesMales
Oetwil10/050.110 (8/73)0.083 (5/60)
Blotzheim02/060.058 (16/278)0.019 (5/268)
St. Louis02/060.256 (11/43)0.156 (7/45)
Weiningen02/060.361 (35/97)0.211 (24/114)
St.Louis-Chaussée11/060.079 (18/229)0.030 (6/197)
Marthalen14/060.193 (23/119)0.099 (16/161)
Rosenau16/060.313 (30/96)0.172 (20/116)
Friedlingen21/060.111 (9/81)0.092 (10/109)
Dietikon21/060.258 (25/97)0.113 (9/80)
Gerstheim24/060.105 (11/105)0.120 (9/75)
Merisheim30/060.111 (6/54)0.146 (6/41)
Orsay01/070.236 (21/89)0.227 (20/88)
Neerach08/070.412 (28/68)0.275 (14/51)
Magstatt26/070.067 (6/89)0.067 (6/90)
Waltenheim30/080.042 (3/71)0.086 (7/81)
Ottmarsheim07/090.102 (14/137)0.084 (8/95)
Märkt07/090.537 (36/67)0.455 (15/33)
Figure 4.

Difference between female and male (arcsine-transformed) proportions of diseased plants plotted against disease density in each population, measured in 17 natural populations of Silene latifolia infected with Microbotryum violaceum.

Discussion

Flower longevity and per-contact risk of infection

Infection probability was affected by experimentally varying the life span of inoculated S. latifolia flowers and thus the duration of exposure to M. violaceum (experiment A, Fig. 1). Establishment on a new host requires several potentially time-consuming developmental steps: newly arrived teliospores must germinate, undergo meiosis and, after conjugation of haploid sporidia, infection hyphae must grow from the flower into the plant. This could explain the marked increase in infection rates of females that retained inoculated flowers for 14 rather than 1 or 2 days. Similarly, relatively low infection success has been reported in other studies using floral inoculation (Altizer et al. 1998; P. van der Putten, personal communication), suggesting that time constraints pose a general barrier to infection. Hot and dry weather further reduces infection probability (e.g. Thrall & Jarosz 1994a; Altizer et al. 1998; Bucheli & Shykoff 2000), presumably because such conditions are unfavourable to early fungal development (Day et al. 1981; Day & Garber 1988). Male plants in our experiment also lost inoculated flowers earlier at higher temperatures, implying an additional, indirect influence of temperature via the time constraint on infection.

Moreover, males seem to actively regulate flower longevity (experiment B), with paintbrush flowers dropped earlier than untouched controls. Although the paintbrush might cause flowers to drop simply because they were damaged, this would not explain the even shorter longevity of flowers treated with spore-bearing paintbrushes, or the difference between control flowers from spore and non-spore plants (Fig. 2).

Flower longevity can be an adaptive trait, balancing the benefits of pollen dispersal with flower maintenance costs (Ashman & Schoen 1994; Schoen & Ashman 1995; Charnov 1996), and the processes of wilting and abscission of flowers are often induced by insect visitation (Nyman 1993; Ladley & Kelly 1995; Clayton & Aizen 1996). In the Geraldton waxflower, infection of petals with a pathogen causes flower abscission (Taylor et al. 1998), and it is therefore possible that such a response, induced by the presence of fungal spores, could have evolved in S. latifolia as a means to reduce infection risk. The findings that plant individuals varied in both flower longevity and flower dropping response (experiment B, Fig. 3a,b), and that loss of inoculated flowers had occurred more often in males that remained healthy (experiment A), are as would be expected in such a case. However, the latter effect was confounded with glasshouse temperature variation, and infection rates in 1- and 2-day flower removal treatments did not differ between males and females, even though females did not drop inoculated flowers. Further experiments are underway to clarify the relationship between flower dropping response and probability of infection, and whether there is genetic variation in dropping response.

Per-contact infection rates, disease epidemiology and field prevalence

The need to develop seeds appears to put females at a higher per-contact infection risk than males, and could explain higher female disease prevalence in the field (see also Lee 1981; Alexander 1989; Shykoff et al. 1996).

However, the number of infectious contacts in our experiment (4 or 16 flowers inoculated on each sex) may not adequately reflect natural frequencies of spore deposition. Male plants produce several hundreds of flowers (e.g. Carroll & Delph 1996), tend to flower earlier and longer in the season (Thrall & Jarosz 1994a; Biere & Antonovics 1996; Purrington & Schmitt 1998) and are more rewarding to pollinators (Shykoff & Bucheli 1995; Shykoff & Kaltz 1998), and are therefore likely to have much higher pollinator visitation frequencies. The resulting increase in spore contacts may outweigh the lower male per-contact risk and lead to higher male than female infection rates (Alexander 1989; Thrall & Jarosz 1994a; Alexander & Antonovics 1995; Biere & Antonovics 1996; Biere & Honders 1998). Infection probability usually increases with higher flower production (Alexander & Antonovics 1988, 1995; Thrall & Jarosz 1994b; Biere & Antonovics 1996), presumably because spore deposition is more frequent, and the absence of such a response here suggests that either the difference in the number of flowers inoculated was insufficient to have an effect (but see Alexander 1989), or that the statistical power was too low to detect it (because relatively few plants became infected).

It should also be noted that single incidence surveys cannot detect whether factors other than transmission dynamics influence disease prevalence. For example, the fact that healthy females can live longer than healthy males (Correns 1928; Lovett Doust et al. 1987) could lead to higher male than female disease prevalence, because healthy males would be under-represented. Alternatively, an infection bias could result from differential patterns of flowering (e.g. between-year flowering propensity) or mortality of diseased plants. Differences in mortality between healthy and diseased plants do occur in some winters (Alexander & Antonovics 1995), but to date there is no evidence for any differences between the sexes (Alexander & Maltby 1990; Thrall & Jarosz 1994a; O. Kaltz, unpublished data).

We found female overinfection to be less pronounced in populations sampled later in the season. This may reflect changing transmission dynamics over the season or sex differences in the phenology of healthy and diseased plants. For example, females with a heavy fruit set may produce few, if any, flowers later in the season, so that no new infections would be seen. Nevertheless, even after controlling statistically for this effect of sampling date, female overinfection increased with increasing density of diseased plants, with females equally or less often infected than males at the lowest densities (Fig. 4). Biere & Honders (1998) found a similar pattern in a study in which disease transmission and not just prevalence was measured. These observations suggest that the relative contribution of the two components of infection risk, disease encounter frequency and per-contact infection probability, can vary with population disease levels. At low disease prevalence, infectious contacts are likely to be rare: in such situations, producing fewer targets (i.e. flowers) may keep spore deposition below critical levels (Roche et al. 1995), thereby balancing their higher per-contact risk to the point at which the more often exposed males become similarly or even more often infected. At higher disease levels, contamination with fungal spores may be so frequent (for both sexes) that females pay the full price of prolonged exposure while maturing fruits.

At high overall disease incidence, female-biased disease prevalence could affect population structure and dynamics. Shortage of available healthy females could reduce the population reproductive output, especially because, even if they do not become infected, contamination with fungal spores reduces seed production in otherwise healthy females (Alexander 1987; Carlsson-Granér et al. 1998; Marr 1998). These effects may influence not only the genetic diversity available for the co-evolutionary interaction with this pathogen (Kaltz et al. 1999), but also, because S. latifolia is not a strong competitor, the maintenance of the population as a whole in the course of succession (Antonovics et al. 1994; Alexander & Antonovics 1995).

Evolutionary implications: ‘safe sex ’ only for males?

Sexually transmitted diseases may select for changes in host mating behaviour and investment in attractive signals in order to minimize infection risk (Loehle 1995; Able 1996; Thrall et al. 1997). In plants, selection mediated by pollinator-borne pathogens may generate lower optima of allocation to floral attractiveness, thereby balancing the chance of reproductive success against the risk of disease encounter (Elmqvist et al. 1993; Thrall et al. 1993; Meagher 1994; Biere & Antonovics 1996; Shykoff et al. 1997) or against the waste of pollen deposited on flowers of diseased plants (Skogsmyr 1993). Hence, S. latifolia may evolve to produce fewer, smaller or less rewarding flowers in the presence of M. violaceum.

Our results suggest that shortened flower life span is an additional means to reduce (per-contact) infection risk. However, dropping contaminated flowers appears to be an option only available to males because females must develop flowers into fruits. Hence, this may represent a situation in which different trade-offs between minimizing disease risk and maximizing reproductive success result in different evolutionary trajectories of male and female mating strategies (Thrall et al. 1997). Males may thus be able to afford to be promiscuous by having an attractive, but disposable, floral display. Female plants, however, can only minimize infection risk by reducing the number of contacts, e.g. by reducing their attractiveness to pollinators. This may contribute to differences in the relative strength of sexual vs. natural selection in males and females and explain, for example, why female plants of S. dioica produced smaller flowers in populations infected with M. violaceum compared with disease-free populations, whereas males did not (Elmqvist et al. 1993).

Nonetheless, where the balance between natural and sexual selection lies, and whether it differs between males and females, may be difficult to determine. Although heritable variation is known for floral traits in S. latifolia (Meagher 1992; Biere & Antonovics 1996; Shykoff 1997), genetic correlations of different sign and magnitude between these traits can render the outcome of directional selection less predictable (Meagher 1994). Furthermore, genetic correlations for floral traits exist between the sexes (Meagher 1994; Biere & Antonovics 1996; Shykoff 1997). Hence, selection for reduced floral display in females may be counteracted by selection in the opposite direction in males, especially if low-risk character states reduce reproductive success in the absence of disease (Biere & Antonovics 1996). Finally, mating strategies do not necessarily evolve to minimize infection risk if exposure to disease is frequent (Thrall et al. 1997). More annual-like life-history strategies, such as early and intensive flowering, may grant high reproductive success even in the face of infection risk (Biere & Antonovics 1996), and, if plants become truly annual, even lead to pathogen extinction (Thrall et al. 1993). Indeed, selection for avoiding or reducing infection risk may then be too weak to change floral traits, and ignoring, rather than avoiding, the risk of infection may represent the evolutionary stable strategy.

Conclusions

In animals, males are often more susceptible than females to parasites (Poulin 1996; Schalk & Forbes 1997; Møller et al. 1998; Wedekind & Jakobsen 1998). Here, in contrast, per-contact infection risk to the fungal pathogen M. violaceum was higher in female than male S. latifolia plants. This is consistent with higher female field prevalences, although patterns of disease prevalence do not necessarily reflect patterns of disease transmission. This study indicates, however, that considering ‘behavioural’ components of disease transmission, such as flower longevity, may further help understanding of how interactions between plant traits and pollinator behaviour affect disease spread in this system (Real et al. 1992; Skogsmyr 1993). More detailed partitioning of factors contributing to the ‘transmission parameter’ can also be important in applied fields. In fact, the time constraint imposed by female reproductive biology, which facilitates the infection described here, is similar to that proposed for higher female risk of infection with sexually transmitted parasites in humans (Hook & Handsfield 1990; Ewald 1994). From the evolutionary perspective, our results illustrate how the presence of disease may generate conflicts between natural and sexual selection that are differently perceived and solved by males and females.

Acknowledgements

We thank Lindsay Haddon, Anders Pape Møller, Massimo Pigliucci and Bitty Roy and three anonymous referees for comments on earlier versions of the manuscript. Philip Egli gave helpful statistical advice. This work was supported by Swiss National Science Foundation Grant 31-33638.92 to J.A.S.

Received 20 January 2000 revision accepted 22 August 2000

Ancillary