Elucidating transmission dynamics and host-parasite-vector relationships for rodent-borne Bartonella spp. in Madagascar

Abstract

Bartonella spp. are erythrocytic bacteria transmitted via arthropod vectors, which infect a broad range of vertebrate hosts, including humans. We investigated transmission dynamics and host-parasite-vector relationships for potentially zoonotic Bartonella spp. in invasive Rattus rattus hosts and associated arthropod ectoparasites in Madagascar. We identified five distinct species of Bartonella (B. elizabethae 1, B. elizabethae 2, B. phoceensis 1, B. rattimassiliensis 1, and B. tribocorum 1) infecting R. rattus rodents and their ectoparasites. We fit standard epidemiological models to species-specific age-prevalence data for the four Bartonella spp. with sufficient data, thus quantifying age-structured force of infection. Known zoonotic agents, B. elizabethae 1 and 2, were best described by models exhibiting high forces of infection in early age class individuals and allowing for recovery from infection, while B. phoceensis 1 and B. rattimassiliensis 1 were best fit by models of lifelong infection without recovery and substantially lower forces of infection. Nested sequences of B. elizabethae 1 and 2 were recovered from rodent hosts and their Synopsyllus fonquerniei and Xenopsylla cheopsis fleas, with a particularly high prevalence in the outdoor-dwelling, highland-endemic S. fonquerniei. These findings expand on force of infection analyses to elucidate the ecological niche of the zoonotic Bartonella elizabethae complex in Madagascar, hinting at a potential vector role for S. fonquerniei. Our analyses underscore the uniqueness of such ecologies for Bartonella species, which pose a variable range of potential zoonotic threats.

Keywords: Bartonella spp.; Force of infection; Madagascar; Rattus rattus; Synopsyllus fonquerniei.

Fig. 1

A. Vegetation map of Madagascar (green = forest; yellow = grassland; orange = desert) with 800 m elevation range limit of S. fonquerniei, the endemic Malagasy rat flea, highlighted in red. Ankazobe (navy) and Ranomafana (fuschia) sites are marked as diamonds. B. Age-for-weight relationships for captured R. rattus as estimated from the von Bertalanffy equation (Ricker, 1979): W(t) = W(1 − e^−k(t−t0)) in which W represents the highest weight class rodent in the data subset (males: 175 g, females: 152 g). Data from Ankazobe sites are depicted in navy (outside sampling sites = circles, inside sampling sites = pluses) and from Ranomafana in fuchsia.

Fig. 2

Age-prevalence and force of infection (respectively) for sampled Rattus rattus infected with B. elizabethae 1 (A,B), B. elizabethae 2 (C,D), B. phoceensis 1 (E,F), and B. rattimassiliensis 1 (G,H). In the age-prevalence charts (A,C,E,G), open circles signify age-stratified prevalence from the data binned over 15-day intervals, and circle size correlates to sample size within each bin. The blue line represents the expected proportion infected in each age class from the best fit model for each species (SIS with 3 age classes for B. elizabethae 1 and 2; SI with a constant force of infection for B. phoceensis 1; and SI with 3 age classes for B. rattimassiliensis 1). Pink shading encompasses the 95% confidence interval as determined via partial profile likelihood, and faint background lines depict predicted prevalence from the more relaxed version of the model allowing for deviations in age-specific FOI by sampling site (navy solid = Ankazobe Outside, navy dashed = Ankazobe Inside; fuchsia solid = Ranomafana Outside; fuchsia solid = Ranomfana Inside). Confidence intervals for site-specific FOIs are listed in Table S4 (not shown in figure for ease of viewing). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 3

Maximum likelihood phylogeny of representative Bartonella genotypes obtained from nuoG gene sequencing of rodent (red) and arthropod ectoparasite (blue) samples in our dataset (outgroup: Brucella abortus) (RAxML, GTR + G + I model, partitioned by codon position, with 1000 bootstrap replicates) (Stamatakis, 2006). All bootstrap values are shown on corresponding nodes. Branch lengths are scaled by nucleotide substitutions per site. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 4

Venn Diagram of flea co-infestation for 54 R. rattus rats from our dataset, for which ectoparasites were isolated and identified. Numbers within each cell give the raw number and corresponding percent (%) of these 54 rats found infested with each combination of ectoparasites. Circle size scales with percentage.