Stochastic variation in the initial phase of bacterial infection predicts the probability of survival in D. melanogaster

Abstract

A central problem in infection biology is understanding why two individuals exposed to identical infections have different outcomes. We have developed an experimental model where genetically identical, co-housed Drosophila given identical systemic infections experience different outcomes, with some individuals succumbing to acute infection while others control the pathogen as an asymptomatic persistent infection. We found that differences in bacterial burden at the time of death did not explain the two outcomes of infection. Inter-individual variation in survival stems from variation in within-host bacterial growth, which is determined by the immune response. We developed a model that captures bacterial growth dynamics and identifies key factors that predict the infection outcome: the rate of bacterial proliferation and the time required for the host to establish an effective immunological control. Our results provide a framework for studying the individual host-pathogen parameters governing the progression of infection and lead ultimately to life or death.

Common structuring principles of the Drosophila melanogaster microbiome on a continental scale and between host and substrate

“…Kuraishi et al, 2011). Furthermore, in a systemic infection, pathogens can reach high titers of millions of bacteria before the fly dies (Galac and Lazzaro, 2011;Duneau et al, 2017). Finally, flies that survive systemic infections often continue to carry chronic infections with pathogen loads that can be in the range of tens of thousands of live bacteria with little effect on their lifespan (Chambers et al, 2014(Chambers et al, , 2019Duneau et al, 2017).…”

Natural variation in the contribution of microbial density to inducible immune dynamics

“…The parasites to which hosts are exposed, the frequency of exposure, and the costs and benefits associated with mounting defences against those parasites, are all expected to contribute to the evolution of inducible immune responses (Cressler et al, ; Frank, ; Hamilton et al, ; Mayer, Mora, Rivoire, & Walczak, ). The dynamics of immunity during the acute infection phase are particularly important for host fitness, as modest levels of variation in early responses could lead to drastic variation in host mortality and parasite persistence (Duneau et al, ). However, the ecological, microbial, and temporal factors to which these dynamics are sensitive have not been well established, providing few avenues to quantify and understand natural variation in inducible immune responses.…”

“…It may be that IMD pathway‐mediated immunopathology is equivalent to approximately one order of magnitude of bacterial virulence per unit time, thus producing no net effect on host survival. It is more likely, however, that survival against Bt is determined long before the inducible immune response becomes relevant, relying instead on variation in alternative arms of the immune system like melanization or phagocytosis to create differences in Bt‐induced mortality rates among individuals and populations (Duneau et al, ). Finally, AMP expression was sensitive to P. luminescens density, but there was very little variation in sensitivity among populations just as there was negligible variation in resistance and mortality rates (Figure ).…”

“…Supporting this idea, knocking down imd reduces defensin-1 expression by around three orders of magnitude at the highest bacterial densities (Figure 4d), reduces sensitivity, and reduces resistance by at least one order of magnitude (Figure 4b,c) but does not ultimately impact survival or even the distribution of time to death in insects inoculated with an LD50 dose of Bt (Figure 4a). It may be that IMD response becomes relevant, relying instead on variation in alternative arms of the immune system like melanization or phagocytosis to create differences in Bt-induced mortality rates among individuals and populations (Duneau et al, 2017). Finally, AMP expression was sensitive to P. luminescens density, but there was very little variation in sensitivity among populations just as there was negligible variation in resistance and mortality rates ( Figure S1).…”

“…, RNAi-mediated knockdown of imd expression (4-fold reduction relative to MalE-injected individuals) in Snavely T. castaneum did not significantly affect bacterial density in infected individuals even as dose was a highly significant predictor of Bt density (linear model, IMD vs.MalE, effect of treatment: estimate = 10.7, st. error = 0.78, t = 1.322, p = .191; effect of dose: estimate = 2.44, st. error = 0.26, t = 9.19, p < .00001). However, we noticed that bacterial density followed a noncontinuous distribution of low and high groups within and among doses(Figure 4c) consistent with bifurcating infection outcomes(Duneau et al, 2017) and complicating statistical interpretation. Therefore, we performed another infection experiment with a lower dose and earlier sampling to avoid the bimodal distribution(Figure 4b), and in this experiment IMD-RNAi individuals had a significantly higher bacterial load than MalE-RNAi individuals (linear model, estimate = 18-fold increase in Bt density, st. error = 1.88, t = 4.556, p = .000378).…”