Virulence constrains transmission even in the absence of a genetic trade-off

The virulence-transmission trade-off predicts that parasite fitness peaks at intermediate virulence. However, whether this relationship is driven by the environment or genetically determined and if it depends on transmission opportunities remains unclear. We tackled these issues using inbred lines of the macro-parasitic spider-mite Tetranychus urticae. When transmission was not possible during the infection period, we observed a hump-shaped relationship between virulence and parasite fitness, as predicted by theory. This was environmentally driven, as no genetic correlation between traits was detected. However, when transmission to uninfected hosts occurred during the infection period, virulence was positively, environmentally and genetically correlated with parasite fitness. Therefore, the virulence-transmission trade-off depends on within-host dynamics and on the timing of transmission, rather than on a genetic correlation. This fundamental correlation may thus be easier to manipulate than previously thought.


66
Spider mites have ambulatory and passive aerial dispersal, hence transmission can depend on 67 environmental factors such as wind (Smitley and Kennedy, 1985). They can transmit during the 68 infection period or overexploit the host plant before transmission occurs (Smitley and Kennedy, 69 1985). Transmission may, thus, be dependent on many factors such as within-host parasite

83
Genetic variation for virulence, R 0 and transmission

84
If the virulence-transmission trade-off is to be driven by genetic correlations among parasite traits, 85 genetic variance for these traits must be present in a parasite population. We thus measured the 86 variance for virulence, parasite fitness and transmission among the T. urticae inbred lines, using only; Fig. S3): we infected hosts (bean leaf patches) with 5, 10 or 20 females from each inbred 99 line for 4 days, then measured damage (virulence) and the number of females produced 10 days 100 later (R 0 ). In support of the trade-off hypothesis, we found a hump-shaped environmental 101 relationship (resulting from the residual co-variance of the model only) between virulence and R 0 .

102
This was shown by the model with a squared term between these traits having a lower DIC, as 103 compared to the model with only the linear term (DIC = 3379 and 3382, respectively). The model 104 with the lowest DIC included density, hence this factor affected trait correlations. This was 105 corroborated by the analysis of trait correlations for each density separately, as we found a 106 positive correlation at low density, no relationship at intermediate density, and a negative 107 correlation at high density (Fig. 1a, Table S3). This suggests that the relationship between these 108 traits is modulated by density dependence: beyond a certain level of virulence, within-host 109 competition prevents more daughters from becoming adult, such that R 0 is maximised at

260
of the models including density as a fixed factor or not. We also report the genetic and 261 environmental correlations when considering each density level separately (Table S3). In    Intermediate density; black = High density). The shape of the curve for each density (light grey = Low density; dark grey = 12 Intermediate density; black = High density) was calculated using a polynomial regression fitted with the geom_smooth function. 13 14 Figure S3. Experimental set-ups. Schematic representation of the experimental set-ups 15 and the traits that were measured in each experiment. Day 0 shows 5 spider mites on a 16 healthy 4cm 2 leaf patch (low density treatment). On day 4, virulence (mottled white areas 17 on leaf patches) was measured in experiments 1 and 2. R 0 (i.e. number of adult daughters) 18 was measured 14 days after mite installation. In experiment 2 transmission to an 19 uninfected leaf patch was possible from day 4 to day 14. 20 21 Table S1. The effect of the initial density of female spider mites on the patch on the 22 per capita production of adult daughters (R 0 ), virulence and transmission (i.e. "trait" in 23 the model column). Deviation information criterion (DIC) of the models (MCMCglmm 24 package) measured in the different experiments with density included or not as a fixed 25 factor. When applicable, best fit models are in bold.