Bartonella infections are prevalent in rodents despite efficient immune responses

We profiled the dynamics of a natural Bartonella isolate as it infected captive individuals of three rodent species that coexist in nature. We also tracked Bartonella-specific IgG antibody levels in these animals over 140 days after the first inoculation and 60 additional days following their re-exposure to the same strain. In nature, the mean longevity of the three rodent species ranges from 6.5 to 12 months [34, 46, 47]. Thus, the 200-day duration of this experiment likely allowed for a decent approximation of the infection dynamics that these rodents may experience in nature. Considering the high prevalence of Bartonella in these species of rodents in the Negev sand dunes, we tested the hypothesis that at least one of them exhibits a waning immune response, which could allow the pathogen to reinfect individuals that cleared prior infections. However, contrary to our hypothesis, we found a strong and long-lasting Bartonella-specific IgG antibody response, with a protective immunological memory in all the rodent species, which prevented infection upon re-exposure to the same Bartonella strain. In addition, two host individuals showed recurrent bacteremia during the first infection stage. Below, we discuss our findings of a comprehensive immune response, recurrent bacteremia, and species-specific differences in a broader disease ecology context and discuss future avenues of research for investigating the puzzle of limited-term Bartonella infections that are nonetheless pervasive. Altogether, insights from this study constitute an initial step toward a better understanding of the interplay between pathogen and host traits, and how the interplay of those traits influences epidemiological dynamics.

A comprehensive immune response against Bartonella

Longitudinal studies have shown that Bartonella species may be highly prevalent, and that the same strains can be repeatedly detected even after a nonbacteremic period [48,49,50,51], leading to the hypothesis that the host immune response against these species wanes [48, 52]. Contrary to this hypothesis, our results provide several lines of evidence suggesting that all three of the tested rodent species responded to the inoculation by mounting a strong, efficient, and long-lasting antibody response, which conferred protection and prevented bacteremia following reinoculation at day 140 d.p.i. First, all of the individuals that were included in the analyses mounted Bartonella-specific antibody responses within 10 days of the initial inoculation. Antibody levels then increased and approached local peak levels, which were maintained at relatively high levels or even increased during the remaining period until the repeat inoculation was performed. Importantly, Bartonella-specific antibody levels remained high long after the rodents managed to clear the infections. Second, the magnitude of the specific antibody response was positively correlated with the bacteremia load. Third, the specific antibody response increased in all rodents and was more rapid upon reinfection, suggesting immune memory and improved IgG antibody response upon re-exposure to the bacteria. Finally, none of the rodents that were reinoculated developed bacteremia or showed recurrent bacteremia, and we found evidence that Bartonella-specific IgG antibodies synthesized upon first inoculation efficiently cleared reinoculated Bartonella even in the individuals displaying recurrent bacteremia (see “Recurrent bacteremia” section). The absence of bacteremia did not appear to be a result of a low-quality or non-viable inoculum, as the second inoculum was prepared from the same Bartonella isolate, using the same procedure as for the first inoculum, and had a bacterial concentration that was similar to that of the first inoculum. Moreover, the bacterial dynamics displayed in the three control rodents, which were inoculated with the second inoculum, were similar to those of the rodents infected with the first inoculum. Thus, our results suggest that the observed specific antibody response most likely prevented Bartonella re-establishment in the rodents upon reinoculation.

These findings of the limited-term nature of infections with Bartonella krasnovii A2 strain, likely owing to the long-lasting Bartonella-specific IgG antibody response of its rodent hosts, align with observations of other Bartonella species (e.g., Bartonella grahamii, Bartonella taylorii, and Bartonella henselae) in a variety of reservoir hosts, including house mouse (Mus musculus), cotton rats (S. hispidus), and cats (F. catus), which illustrated similar in vivo bacterial dynamics, antibody kinetics, antibody-mediated clearance of bacteremia, and failures of reinfection [24, 27, 29, 30]. Our findings also add to experimental evidence showing that IgG antibodies activate the complement system and inhibit Bartonella adhesion to erythrocytes (reviewed in [53]). Taken together, this evidence suggests that phylogenetically distant reservoir hosts have similar strategies for clearing Bartonella infections. These strategies are based on the high turnover rate of erythrocytes and the development of IgG antibodies that prevent bacterial binding to host erythrocytes when Bartonella are periodically seeded from other niches (see below; [53]). Thus, the results of our experiment broaden the universal view of the interactions between Bartonella and their reservoir hosts and suggest that if Bartonella did not continue to evolve rapidly, they would likely be eliminated from natural communities.

Recurrent bacteremia

We observed recurrent bacteremia in one G. andersoni host and one G. pyramidum host. In these individuals, Bartonella cells reappeared in the bloodstream within 30–40 days of their disappearance despite the fact that both individuals mounted strong Bartonella-specific responses upon inoculation. There are several possible explanations for this pattern of recurrence. First, it is possible that the infection had never been cleared from the blood of these two rodents, but that its level decreased below detectable levels [48]. However, as infections in the other inoculated rodents in the current study, as well as in 20 G. andersoni that were inoculated with the same Bartonella strain in a previous study [25], never lasted for more than 70 days, this explanation seems implausible. Second, the recurrent bacteremia may have been a result of a waning immune response [52]; however, this is unlikely, as in addition to displaying comparable antibody levels to the other rodents, the two rodents in question did not develop secondary bacteremia upon reinoculation (once they had cleared the recurrent infections).

A third possible explanation is that, upon first inoculation of these two rodent individuals, some bacterial cells remained in—as yet unidentified—cellular niches in host tissues, where they persisted and replicated. Only later, after the rest of the bacteria were cleared, did these latent bacteria re-enter and recolonize the blood stream. The first recognized so-called primary niche in Bartonella species was endothelial cells, but additional cellular niches with similar roles were later proposed, including the dermis, lymph nodes, bone marrow, liver, spleen, and the kidney (reviewed in [54]). This hiding–seeding mechanism was proposed as an explanation for the 3- to 6-day interval of recurrent bacteremia that was detected in rodent models and in people infected with Bartonella quintana [26, 27, 53]. This mechanism, which allows subpopulations of bacteria to hide and reappear in different niches, could also contribute to the long duration of the IgG antibody response that was observed in our study. However, the “hidden niche” hypothesis alone cannot explain the longer intervals of recurrent bacteremia that were observed in the current study, i.e., 30–60 days of initial bacteremia, followed by 30–40 days of Bartonella-negative blood, and then 40–60 days of recurrent bacteremia (Fig. 1c, n). After such long infection intervals, we would expect that, upon their release from the hidden niche, these bacterial cells, which are similar to the cells in the repeated inoculation, would be revealed and immediately targeted by the host immune response. Alternatively, we propose that, in the hidden niche, some Bartonella organisms have evolved to escape the specific IgG antibodies (see the below section on antigen escape). This hiding–mutating–seeding scenario that is supported by both the observed bacterial and IgG antibody dynamics, and aligns with immunological evidence from other Bartonella species [53], may also be responsible for the long intervals between recurrent bacteremia that were observed in experimental infections of cats [55] and longitudinal field studies (e.g., [48, 50]). This hypothesis should be confirmed by comparing the genome sequences of Bartonella in the host’s blood during the peaks of initial and recurrent infections, and testing the cross-reactivity of the IgG response to these two bacterial sources.

Future studies should assess how common recurrent bacteremia is under natural conditions. In natural populations of cotton rats (S. hispidus) and deer mice (Peromyscus maniculatus), 8–15% of the hosts showed recurrent Bartonella infections [48, 50], similar rates to those observed in the current experiment (14%). However, since there is experimental evidence that recurrent bacteremia might be associated with intradermal inoculations, which resemble the vector-borne transmission route (current study; [28]), it is possible that, in the northwestern Negev Desert, flea transmission will even further amplify this phenomenon. Apart from assessing the commonality of this phenomenon, it is important to reveal the exact mechanism underlying recurrent bacteremia for a better understanding of host–pathogen interactions. The challenge of future longitudinal studies in animals and people will be to develop molecular techniques that differentiate between recurrent bacteremia and reinfection by the same strain. The distinction between these two processes, which was enabled here by our experimental set-up, is crucial for understanding pathogen population and community dynamics and for informing effective medical solutions against persistent infections [56].

Species-specific differences in antibody kinetics

Contrary to our hypothesis that at least one of the rodent species would exhibit a waning immune response, as mentioned above, we found that infection with B. krasnovii A2 strain elicited an efficient and protective immune response in all of the species tested. Yet, despite the similar microhabitats of G. gerbillus and G. pyramidum [34], and the similar body size of G. gerbillus and G. andersoni [57], our results indicate that the immune response of G. gerbillus may be more reactive than those of the two other species. This was demonstrated by the magnitude of the increases in antibody levels following reinfections compared to the primary responses, which were highest in G. gerbillus (Fig. 3a). In addition, the mean rate of antibody increase was greater (although not significantly different) in G. gerbillus than in G. pyramidum (Fig. 4). Finally, both the peak and overall bacterial loads of G. gerbillus individuals were significantly lower than those of the two other species. In accordance with these results, Bartonella prevalence was lower in G. gerbillus populations in the study region compared to populations of the other two species (see supporting data in [31]).

Higher resistance is often observed when hosts are locally adapted to their pathogens [58,59,60,61,62]. Thus, the greater resistance of G. gerbillus may indicate that this host is more adapted to Bartonella than the two other species. However, considering the sporadic temporal and spatial occurrence of G. gerbillus as compared to the steady occurrence of G. andersoni and G. pyramidum rodents in the natural environment [34], it is unlikely that G. gerbillus faces strong selection due to infections with B. krasnovii A2 strain. Instead, assuming that Bartonella is ahead of its hosts in the evolutionary arms race, the higher bacterial loads in G. andersoni and G. pyramidum compared to G. gerbillus may indicate that B. krasnovii A2 strain is more adapted to these more reliable hosts, and can therefore better hide and/or escape from their IgG antibodies [63, 64]. To test this hypothesis, future experiments should compare the bacterial dynamics and immune kinetics of different Bartonella strains that are either locally adapted or not adapted to each of these rodent species. The aim of these studies would be to elucidate the missing links between the long-term infection dynamics, immune kinetics, and history of coevolution between these pathogens and their hosts, which is crucial information for understanding patterns of epidemiological dynamics in natural communities.

Future directions to solve the Bartonella pervasiveness puzzle

While the unique strategy by which Bartonella persists for weeks within the protected niche of host erythrocytes is consistent with their high prevalence in reservoir hosts, our finding that there is a comprehensive host serological immune response with an efficient memory leaves the puzzle of Bartonella’s pervasiveness unresolved. Future laboratory experiments with food-deprived rodents, juvenile rodents, and reproductive female rodents under predation risk should be conducted to confirm our results in communities that better represent the states of these rodent populations in nature. These experiments could address whether a comprehensive immune response against Bartonella would also develop under more challenging conditions for hosts than the seminatural conditions provided in the current experiment. In parallel, it is important to follow the changes in infection status of, and Bartonella strain composition within, the same rodent individuals over monthly intervals in the field.

Considering that the studies proposed above may provide further support for the existence of a comprehensive immune response against Bartonella, we also suggest that other studies should be undertaken that focus on an alternative explanation for the puzzle of Bartonella’s pervasiveness, namely, the existence of genetic mechanisms that allow these pathogens to rapidly evade the well-adapted immune responses of their natural hosts. This alternative explanation is in line with the results of a longitudinal study of the dynamics of Bartonella observed in a natural population of cotton rats (S. hispidus), in which infections of the same individual hosts by Bartonella variants from different genogroups often followed one another [65, 66].

Antigenic variation—when pathogen populations evolve to alter surface features targeted by the host immune system—is one of the most widely used escape strategies that allows pathogens to reinfect hosts that have developed an immune response against the original strain (preceding antigenic changes; [67]). In Bartonella, at least three genetic mechanisms operate that could lead to rapid antigenic variation. First, contingency loci, which are hypermutable sites on specific genes, may undergo mutations that add or remove repeat units at high rates due to strand slippage during DNA replication. Although they have never been fully profiled in Bartonella, an elevated number of mononucleotide repeats in this genus, relative to other bacteria, has been noted [68]. Second, genome comparisons of virulence gene arrays in Bartonella revealed high rates of intragenomic recombination events that copy, delete, and hybridize the main versions of these genes with other nearby copies [69,70,71,72]. Finally, Bartonella share a domesticated prophage that acts as a gene transfer agent, packaging their DNA for transduction. Gene transfer agent-mediated recombination may accelerate antigenic variation, and virulence factor evolution through the exchange of DNA between co-infecting strains in the flea gut or host tissues [73]. Understanding what roles these and other genetic mechanisms for rapid evolution play in the spread and persistence of Bartonella may shed light on a universal mystery—the pervasive nature of limited-term pathogens despite efficient host immune responses.