The composition of bacterial communities in blood-sucking arthropods can shift dramatically across time and space (Jones et al., 2010). These shifts are linked in part to variable arthropod characteristics (for example, arthropod species, life stage and engorgement level; Pidiyar et al., 2004; Moreno et al., 2006; Heise et al., 2010). However, blood-sucking arthropods acquire a portion of their microbial communities from vertebrate animals during feeding and additional microbes from the environment during free-living stages (fleas and mosquitos) or host-questing periods (ticks). Therefore, it is likely that both vertebrate host-related (for example, age, gender and reproductive status) and environmental (for example, temperature and habitat type) factors will also influence the bacterial communities of blood-sucking arthropods. However, there is a paucity of studies on the potential role of the vertebrate host and environment on bacterial composition within blood-sucking arthropods, and thus the relative impacts of vertebrate host-related, arthropod-related and environmental factors on bacterial community composition in blood-sucking arthropods are not well understood.

We used 16S rRNA pyrosequencing to characterize the bacterial community composition in 66 fleas (Orchopeas leucopus and Ctenophthalmus pseudagyrtes) and 132 ticks (Dermacentor variabilis and Ixodes scapularis) collected from 193 rodents (Peromyscus leucopus and Microtus ochrogaster) in southern Indiana, USA (Supplementary Material and Methods).

Most sequences (92%) were assigned to 103 phylotypes, representing six presumptive phyla (Actinobacteria, Bacteroidetes, Cyanobacteria, Firmicutes, Proteobacteria and Tenericutes) and 30 genera (Supplementary Figure S2). Similar to other studies on blood-sucking arthropods, we found diverse bacterial communities where most phylotypes were sparsely represented or occurred in only a few individual arthropods (Reed and Hafner, 2002; Pidiyar et al., 2004; Jones et al., 2010; Andreotti et al., 2011). For D. variabilis and I. scapularis, bacteria similar to Francisella spp. and Rickettsia spp., respectively, were most abundant. The high prevalence of Arsenophonus 1 and Francisella 1 and 2 in D. variabilis and of Rickettsia 1 in I. scapularis, and their similarity to previously characterized endosymbionts (Supplementary Table S2), suggests that they are stable residents of these tick species. Bartonella spp. were the most frequently identified bacteria in the two flea species. In particular, Bartonella 1, a phylotype with high resemblance to the vertebrate (including human) pathogen, B. grahamii (Kerkhoff et al., 1999; Breitschwerdt and Kordick, 2000), infected the two flea species at high frequencies. Brevibacillus spp., the second most common sequences in all four arthropod species, may also have an important role in bacterial communities of arthropods, and thus deserve additional study.

To the best of our knowledge, this is the first study that simultaneously quantified the effects of vertebrate host-related, arthropod-related and environmental factors on the bacterial community composition of blood-sucking arthropods. Only the effect of arthropod-related variables was significant and accounted for 18.6% of the variation in the bacterial composition (Supplementary Table S3). In particular, the arthropod group (fleas vs ticks) had a major effect on bacterial composition (F=31.9, P<0.0005). The major clusters of bacterial communities in our ordinations corresponded to (i) O. leucopus and C. pseudagyrtes fleas, (ii) I. scapularis and (iii) D. variabilis ticks (Figure 1). The unexplained variance in bacterial community composition could be a result of other factors not quantified in this study, such as interspecific interactions between bacteria, or a result of stochastic events.

Figure 1
figure 1

Non-metric multidimensional scaling ordination of (a) all 198 arthropods based on Bray–Curtis similarities in abundances of the 103 bacterial phylotypes and (b) a subset of 146 arthropods based on Bray–Curtis similarities in abundances of five bacterial phylotypes belonging to the commonest arthropod-specific genera (Rickettsia 1, Francisella 1, Francisella 2, Bartonella 1 and Bartonella 2; Supplementary Table S2 and Supplementary Figure S2). Closer points represent higher similarity in bacterial community composition than points that are further apart. Each point represents an individual arthropod. The division of arthropod to species (green squares for D. variabilis, black circles for I. scapularis, blue triangles for C. pseudagyrtes, and red X’s for O. leucopus) was the best explanatory variable of the bacterial community composition within ticks but not fleas. Oval shapes surround bacterial communities in individual arthropods of the same species or two species that were not distinguished based on their bacterial communities.

Independent analyses for fleas and ticks significantly improved the explanatory power of the tested variables for ticks (39.7%) but not for fleas (20.4%), for which none of the tested variables were significant. Arthropod species had a major effect on bacterial composition in ticks (F=55.8, P<0.0005). The Rickettsia 1 and Francisella 1 phylotypes were primarily responsible for differences across tick species, as the former was found mainly in I. scapularis and the latter was found only in D. variabilis. Arthropod life stage also significantly affected the bacterial community composition of ticks (F=9.8, P<0.0005). The Francisella 1 phylotype was 37 times more abundant in nymphs than in larvae and was the primary determinant of differences across life stages. Life-stage differences in blood engorgement (for example, nymphs consume larger and more blood meals than larvae) may explain those differences, as blood ingestion often induces bacteria multiplication in the arthropod gut (Heise et al., 2010). Alternatively, Francisella bacteria may accumulate with transmission from one developmental stage of the tick to the next.

The lower diversity of bacterial species in ticks (Supplementary Table S2) and their higher sensitivity to arthropod-related variables compared to fleas may be a result of the dominance of endosymbionts in ticks, which are likely to be highly dependent on the arthropod and may potentially exclude competing bacteria (Lively et al., 2005). In addition, free-living flea larvae (unlike ticks) may acquire bacteria including Comamonas, Pelomonas, Ralstonia (Supplementary Figure S2) from soil and transtadially transmit these to adult fleas (in addition to acquisition of bacteria from blood meals).

The most intriguing results of this study are that the vertebrate host and environmental characteristics do not seem to affect bacterial community composition in fleas and ticks. Fleas and ticks were collected over a range of environmental conditions and sites, but none of those conditions significantly affected the bacterial communities of the arthropods. Geographic differences have been detected in other tick-associated bacterial communities (Wielinga et al., 2006; Clay et al., 2008; van Overbeek et al., 2008). While we collected ticks from rodents, in these other studies ticks were sampled from the environment. Rodents could have transported ectoparasites and their associated bacteria among sites or may have provided more stable conditions and buffered potential environmental variability. However, none of the seven host-related factors tested had a significant impact on bacterial community composition. The minor effects of host traits on the microbial community composition may reflect the fact that vertically-transmitted endosymbionts are the dominant members of these communities, at least in ticks. The nature of association between fleas and their dominant bacterial association with Bartonella deserve further study.

Taken together, our study suggests that bacterial communities in blood-sucking arthropods are composed of both pathogens and endosymbionts, where the exact species composition is determined largely by arthropod traits.