FIGURE LEGENDS
Figure 1. Conceptual diagram outlining the ‘contact contingency’ hypothesis. Hypothetical fig wasp relationships and predicted status of RI inducing Wolbachia according to variation in ecological contact and evolutionary time since speciation. We predict Wolbachia infection to occur only in community III where species 1 & 2 should harbour unrelated strains. Sister species 3 & 4 are not in ecological contact as they form separate communities I and II, while sister-species 5 & 6 in community IV, despite ecological contact, have had sufficient evolutionary time for alternative (less costly) RI mechanisms to evolve.
Figure 2. Stylised schematic showing a fig in cross section. Five layers of ovules are used in our model (white and grey) and no oviposition occurs in the central lumen (black). Ovule length (and embryo relative fitness, ω) decreases towards the fig wall (green) where larvae are at greater risk of parasitism. We use a descriptive model to contrast inclusive fitness (W) between foundress wasps that do not experience cytoplasmic incompatibility (wasp 1, blue) and those that do (wasp 2, orange). Here, in a toy example, each foundress has 10 eggs (open circles represent viable hybrid eggs with decreased fitness while closed circles are non-hybrids with full fitness) and we limit oviposition to two eggs per layer. While CI wasps lay fewer eggs (hybrids are lost to CI) they do not fill valuable oviposition sites with hybrids of decreased fitness. Here, the CI wasp gets an inclusive fitness of 3.8 for its seven remaining eggs and the noninfected wasp gets 3.1 for a full complement of 10 eggs (i.e., by multiplying egg fitness by oviposition fitness then summing). Inclusive fitness is therefore greater in wasp 2 despite this fecundity loss, as it lays a higher number of high fitness eggs in premium oviposition sites. This example would represent one pixel on the heat maps displayed in Figure 5. Please see text for further details.
Figure 3. Wolbachia strains mapped along the pollinating wasp phylogeny. Strain type is indicated by the different colours, with uninfected individuals in black. For each wasp clade the Ficushost is given.
Figure 4. Wolbachia strain prediction accuracy by wolPredictor modelling the ‘contact contingency’ hypothesis across 253 fig wasp pollinator samples at species clustering thresholds of 2-50 for run name pleio\sout4 . Accurate positive assignations (orange) are shown above the zero-line whilst accurate negative assignations (blue) are shown as positive values below the zero-line.
Figure 5. Heat map evaluation of the ‘oviposition trade-off’ hypothesis. Comparative inclusive fitness values of fig wasp foundresses across relative conspecific-heterospecific fitness space at different population-level frequencies of conspecific mating (between 5-95%) under alternate scenarios of CI-induced egg mortality (i.e., ‘CI ’ vs. ‘no CI ’). Redder tones (i.e., above zero) indicate relative conspecific-heterospecific fitness where foundress inclusive fitness is higher under CI-induced mortality due to preferential oviposition of higher fitness conspecific offspring despite trade-offs with fecundity reduction. NB in order to explore all relative fitness space, heatmaps indicate regions where heterospecific fitness is greater than conspecific fitness, which will generally be an unrealistic scenario.
Figure S1. Single phylogenies for (a) coxA , (b) fbpA , (c)ftsZ , d) gatB , (e) hcpa and (f) wspincluding sequences generated for this study and all accessions from the MLST data base, (g) a wsp single gene phylogeny for sequences generated in this study and (h) a phylogeny derived from the five MLST genes for sequences from this study only. Colour coding depicts Wolbachia strains red: wspC1, blue: wspC2, purple: wspC3, green wspC5, yellow: wspC6_1 and orange: wsp6_2.