13Evolutionary theory predicts that vertically transmitted symbionts are selected for low 14 virulence, as their fitness is directly correlated to that of their host. In contrast with this prediction, 15 the Wolbachia strain wMelPop drastically reduces its Drosophila melanogaster host lifespan at 16 high rearing temperatures. It is generally assumed that this feature is maintained because the 17 D. melanogaster-wMelPop symbiosis is usually not exposed to environmental conditions in which 18 the symbiont is virulent. To test this hypothesis, we submitted wMelPop-infected D. melanogaster 19 lines to 17 generations of experimental evolution at a high temperature, while enforcing late 20 reproduction. The fly survival was measured at different time points, as well as two traits that have 21 2 been proposed to be causally responsible for wMelPop virulence: its relative density and the mean 22 number of octomom copies present in its genome. We hypothesised that these conditions would 23 select for a reduced wMelPop virulence, a reduced wMelPop density, and a reduced octomom copy 24 number. Our results indicate that density, octomom copy number and virulence are correlated.
25However, contrary to our expectations, we could not detect any reduction in virulence during the 26 course of evolution. We discuss the significance of our results with respect to the evolutionary 27 causes of wMelPop virulence and propose that intra-host selection could explain this conundrum. 28 29 42correlated to that of their hosts, they are expected to be selected for low virulence (Ewald 1983). 43 3 This reasoning, just like the transmission/virulence trade-off hypothesis, rests solely on the 44inter-host level of selection, ignoring the potential effects of intra-host selection. Indeed, the most 45 competitive symbionts within a host may replicate faster and be more virulent than what would be 46 optimal from the standpoint of inter-host selection (Alizon et al. 2013). However, the causal link 47