Previous studies aiming to characterize the course of hemotropic Bartonella
infections have employed blood culture on petri dishes (to determine CFUs) and/or light microscopy or ultrastructural analysis of RBCs 32930
. These established techniques revealed a typically prolonged course of bacteremia in association with intraerythrocytic bacterial inclusions. In this study, we have dissected the in vivo course of Bartonella
erythrocyte parasitism in considerable detail by using an appropriate animal model in combination with a novel approach for tracking blood infections. We have used a GFP-expressing strain of the rat-adapted species Btr
) in an experimental rat infection model. Importantly, the expression of GFP did not alter bacterial infectivity (unpublished results) but enabled us to trace individual bacteria within the blood of infected rats by means of flow cytometry. As validated by confocal microscopy and conventional blood cultures, this powerful fluorescence-based in vivo detection method for bacterial infections allowed a precise quantification of both the number of infected erythrocytes and the number of bacteria colonizing individual erythrocytes. Moreover, in vivo whole blood biotinylation before or after Btr
infection allowed (by staining with fluorescent streptavidin conjugates) to trace the aging erythrocyte population for studying the course of intraerythrocytic replication and the resulting effect on the life span of infected erythrocytes. A model of erythrocyte colonization by Bartonella
as deduced from these extensive studies is depicted in .
Model of erythrocyte parasitism by Bartonella spp.
After intravenous inoculation, Btr
was rapidly cleared (within a few hours) from circulating blood, which thereafter remained sterile for 3–4 d. This eclipse phase indicates the colonization of a primary niche by Btr
before the onset of hemotropic infection. Given the marked tropism of bartonellae for vascular endothelial cells (1215
; for a review, see reference 13
), we could speculate that this cell type may represent the primary niche without discounting the possibility that other cell types or organs may also contribute. Eventually, the competence for the erythrocyte invasion process that follows has to be acquired by an adaptation process (i.e., via transcriptional reprogramming) during infection of this primary niche.
The bacteremic phase of Btr–gfp synchronously began at 4–5 d.p.i. with a single bacteria found in association with circulating RBCs. In vivo biotinylation experiments allowed us to demonstrate without ambiguity that infected and subsequently invaded cells are mature erythrocytes. This finding is remarkable considering that in contrast to erythroid precursor cells, the mature erythrocyte is devoid of any endocytic activity, thus leaving the active part of cell entry to the invading bacterium.
To test for the time point of erythrocyte invasion, we have used in parallel two complementary assays based on the impermeability of the intact erythrocyte membrane to either antibodies (differential immunocytochemical staining) or the antibiotic gentamicin (gentamicin protection assay). Both assays indicate an extracellular localization of erythrocyte-associated bacteria at the onset of bacteremia. However, by differential immunocytochemical staining, the invasion process was completed within 2 d, while in gentamicin protection assays bacteria required an additional day before becoming gentamicin protected. Based on ultrastructural analysis, a pore-like structure has been described for B. henselae
–invaded cat erythrocytes 3
. Moreover, it has been suggested that the partially characterized lipophilic erythrocyte membrane–deforming protein deformin secreted by B. bacilliformis
and B. henselae
can insert itself as a pore-like structure into the erythrocyte membrane 3233
. This presumable pore may account for the observed differential permeability for antibodies and gentamicin during erythrocyte invasion.
Invasion into mature erythrocytes is immediately proceeded by intracellular replication. TEM indicated that the replicating bacteria are surrounded by a presumable vacuolar membrane. A similar observation has been reported for the invasion of human erythrocytes by B. bacilliformis 34
. Interestingly, bacterial replication slowed after several days, reaching a plateau at approximately eight intracellular bacteria on average per infected erythrocyte. This plateau was maintained for the remaining life span of the infected erythrocytes. Cessation of bacterial replication may result from the deprivation of either essential nutrients or growth factors or may indicate an active mechanism of growth control (i.e., by quorum sensing; for a review, see reference 35
). Strikingly, whole blood biotinylation in a rat devoid of any reinfection wave revealed that during the period of invasion and intracellular replication, the clearance of the infected erythrocytes was moderately increased compared with uninfected erythrocytes. However, after cessation of bacterial replication, clearance of infected erythrocytes was indistinguishable from that of uninfected erythrocytes. To our knowledge, we provide the first example of a pathogen capable of persistently colonizing the limited intracellular spaces of erythrocytes (until the infected cells are cleared by normal turnover), which is clearly distinguished from the typically hemolytic infection cycle of other erythrocytic parasites, such as the malaria parasite Plasmodium
(for a review, see reference 36
). Bartonellae thus persist for weeks within the immunologically privileged intracellular niche of erythrocytes, thereby increasing their chances for transmission by blood-sucking arthropods. This unique pathogenic strategy certainly contributes to the remarkable epidemiological success of bartonellae in their reservoir hosts with prevalences of bacteremia typically ranging from 15 to 95% (for reviews, see references 1
). While most known bartonellae may behave similarly to Btr
in their respective reservoir hosts, the human-adapted species B. bacilliformis
appears to provide an exception, as it can trigger massive hemolysis after intraerythrocytic infection 1112
. A contact-dependent hemolytic activity has been identified recently in B. bacilliformis 17
, which may account for this striking difference in the course of erythrocyte infection.
After the synchronous onset of erythrocyte-associated bacteremia on day 4 or 5 p.i. (<106
afflicted cells/ml of blood), the number of infected erythrocytes decreased on subsequent days due to erythrocyte turnover. However, in the majority of infected animals we noticed recurrent erythrocyte infection waves of variable amplitude that had not been detected before by any other method. The recurrence of these reinfection waves typically ranged from 3 to 6 d, which is in accordance with the variable time intervals of the feverish relapses in human trench fever (five day fever) caused by B. quintana 14
. B. quintana
bacteremia in humans may have an asymptomatic course as well 1415
, like erythrocyte infection waves observed in our rat model, which were not associated with fever nor any other obvious clinical manifestations (data not shown). Experimental infection of cats as the reservoir host of B. henselae
indicated that, depending on the bacterial strain used for inoculation, the clinical outcome of bacteremia ranged from no symptoms to fever in conjunction with other manifestations 282937
. While the principle process of intraerythrocytic Bartonella
infection may occur similarly in any infected reservoir host, the severity of clinical presentations may inversely reflect the level of adaptation of the pathogen to its specific reservoir host, eventually allowing Bartonella
to cause an extended, high-titer bacteremia with little or no harm.
While the synchronous “lytic” erythrocytic cycle of Plasmodium liberates parasites, causing the subsequent erythrocyte infection wave, the recurrent infection waves by Btr–gfp should be of a different origin, considering the nonhemolytic erythrocyte infection process by these bacteria. We propose that the recurrent erythrocyte infection waves are seeded from the same primary niche as the initial infection wave (). The observed recurrence may result from a lytic infection cycle that lasts ~5 d in this primary niche, liberating bacteria that coincidentally infect erythrocytes as well as reinfect this primary niche ().
It will be intriguing to identify the yet elusive primary niche of bartonellae in the reservoir hosts and to study how the recurrent intraerythrocytic infection waves are seeded. A molecular understanding of the mechanisms of forced entry into mature erythrocytes and of growth control allowing the persistent intraerythrocytic colonization will be both scientifically compelling and important for developing strategies to control or prevent human infection by bartonellae as important emerging pathogens.