OUP user menu

Molecular phylogeny of the genus Bartonella: what is the current knowledge?

Pierre Houpikian, Didier Raoult
DOI: http://dx.doi.org/10.1111/j.1574-6968.2001.tb10684.x 1-7 First published online: 1 June 2001


Species of the genus Bartonella are involved in an increasing variety of human diseases. In addition to the 14 currently recognized species, several Bartonella strains have been recovered from a wide range of wild and domestic mammals in Europe and America. Such a high diversity of geographic distributions, animal reservoirs, arthropod vectors and pathogenic properties makes clarification of our knowledge about the phylogeny of Bartonella species necessary. Phylogenetic data have been inferred mainly from 16S rDNA, 16S–23S rRNA intergenic spacer, citrate synthase and 60 kDa heat-shock protein gene sequences, which are available in GenBank. Comparison of phylogenetic organizations obtained from various genes allowed six statistically significant evolutionary clusters to be identified. Bartonella bacilliformis and Bartonella clarridgeiae appear to be divergent species. Bartonella henselae, Bartonella koehlerae and Bartonella quintana cluster together, as well as Bartonella vinsonii subsp. vinsonii and B. vinsonii subsp. berkhoffii. The fifth group includes bacteria isolated from various rodents that belong to native species from the New World and in the sixth, Bartonella tribocorum, Bartonella elizabethae and Bartonella grahamii are grouped with several strains associated with Old World indigenous rodents. The position of the other species could not be consistently determined. As some cat- or rodent-associated Bartonella appeared to cluster together, it has been suggested that these bacteria and their reservoir hosts may co-evolve. Lack of host specificity, however, seems to be frequent and may reflect the influence of vector specificity. Host or vector specificity may also explain the current geographic distribution of Bartonella species.

  • Phylogeny
  • Gene
  • 16S rRNA
  • Citrate synthase
  • gltA
  • groEL
  • Bartonella

1 Introduction

Members of the genus Bartonella are Gram-negative, oxidase-negative, fastidious bacteria belonging to the α2-subclass of Proteobacteria. Common features of Bartonella include transmission by an arthropod vector and survival within mammalian reservoir hosts [1]. During recent years, an increasing number of Bartonella species have been isolated and characterized, and the genus, extended by unification with the genera Rochalimaea and Grahamella, currently consists of 14 recognized species [2,3]. To date, seven of the 14 species have been implicated in human disease, and Bartonella are considered as emerging pathogens [4]. Bartonella bacilliformis is the causative agent of bartonellosis (Carrion's disease), a biphasic disease endemic to Andean valleys. Bartonella quintana and Bartonella henselae, etiologic agents of trench fever and cat-scratch disease respectively, have also been involved in endocarditis and in bacillary angiomatosis in immuno-compromised patients [4]. Bartonella clarridgeiae is suspected to be an additional agent of cat-scratch disease [5]. Bartonella elizabethae and Bartonella vinsonii subsp. berkhoffii have been shown to cause endocarditis [6,7], while B. vinsonii subsp. arupensis was first isolated from a febrile patient with valvular disease in the USA [8]. Bartonella grahamii has been implicated in cases of neuroretinitis [9]. The other species, B. vinsonii subsp. vinsonii, Bartonella taylorii, Bartonella doshiae, Bartonella talpae, Bartonella peromysci, Bartonella alsatica, Bartonella tribocorum, Bartonella koehlerae and Bartonella birtlesii were isolated from the blood of non-human mammals but have not yet been associated with human disease [3,1012]. Additionally, an increasing number of Bartonella strains have been recovered from a wide range of mammals, including rodents, cervids and cattle, in Europe and America [1,13,14]. Although partial, the genetic characterization of these isolates suggests that some of them may represent new Bartonella species [1517].

As the diversity of the genus Bartonella, including bacterial species and strains, animal hosts and vectors, geographic distribution, and pathogenic properties, appears to be far more extended and complex than initially thought, phylogenetic studies are urgently needed to clarify the taxonomic organization within the genus and the evolutionary relationships between Bartonella species. By identifying lineages of bacteria that share a common evolutionary lineage as well as common epidemiological or pathogenic characteristics, such studies may also help to recognize some of the determinants of the biology of these organisms. To date, phylogenetic data have been inferred from gene sequence comparisons, using different analytical methods. Most of the current knowledge about Bartonella taxonomy is supported by only three genes, namely the 16S rRNA-, the citrate synthase (gltA)- and the 60 kDa heat-shock protein (groEL)-encoding genes, which have been sequenced for most Bartonella species. Even with such limited data, interpretation of phylogenetic trees remains difficult: 16S rDNA-based phylogeny generally lacks statistical support at the interspecies level, and trees inferred from groEL and gltA gene sequences, although providing some statistically significant lineages, show discrepancies when different analytical methods are used [17,18]. The purpose of this brief review is to discuss the published data relating to Bartonella phylogeny and to compare the various architectures inferred from these different gene sequences in order to assess the reliability of the proposed reconstructions and to identify lineages that are consistently proposed. On the basis of such overall analyses, we discuss hypotheses that have been formulated about the natural history of the genus, especially that of co-evolution between the Bartonella and their mammalian reservoir hosts and that of divergent evolutionary process between Bartonella located in separated geographic areas [15,16].

2 Genotypic studies and phylogenetic trees

To date, comparisons of sequences from seven genes or genetic loci have been used for phylogenetic analyses of Bartonella species. These are the 16S rRNA-encoding gene, the 16S/23S intergenic spacer region (ITS), and five protein-encoding genes: the riboflavin synthetase (ribC), the cell division protein (ftsZ), the citrate synthetase (gltA), the 60 kDa heat-shock protein (groEL), which are involved in basic cellular functions, and more recently the 17 kDa antigen, an outer membrane protein eliciting a strong immune response in infected hosts [1720]. Comparison of 16S rRNA gene sequences is considered as being generally useful for establishing evolutionary relationships, because of its high information content, conservative nature and universal distribution [21]. Interest has focused on the protein-encoding genes because they appeared to be conserved at the species level but highly variable between different species, and also because sequences can be analyzed at the protein level as well as at the genomic level, for better reliability [17,18,22]. The use of hypervariable ITS sequence comparison has been controversial due to doubts concerning the reliability of such sequences as indicators of molecular evolution [22,23]. Nevertheless, the ability of a specific sequence to produce meaningful phylogenetic data can only be appreciated after comparison and reconstruction inferable, by assessing the statistical support of the resulting branching order. In practical terms, only four gene data sets are currently usable for phylogenetic studies: these are 16S rDNA, ITS, gltA, and groEL. Use of other genes is hampered by the lack of Bartonella sequences [19,20].

After gene sequence alignment, phylogenetic trees are constructed using three different analytical approaches: the parsimony method and two distance methods, neighbor-joining and maximum likelihood [24]. All 16S rDNA, ITS, gltA, and groEL sequences currently available in GenBank, including those from validated Bartonella species and those from partially characterized strains, have been used to infer the trees. The four parsimony trees are presented in Fig. 1. To investigate the stability of these trees, bootstrap values were obtained from 100 randomly generated sample trees. Bootstrap values are considered to be an approximation of a statistical confidence interval. Values >90% are considered to be statistically significant as they reflect a risk of <10% of accepting a wrong branching order [25,26].

Figure 1

Comparison of parsimony trees derived from 16S rRNA gene, ITS, groEL, and gltA for Bartonella species and strains. The gltA analysis was based on a 259 bp alignment of the gene, the ITS, groEL and 16S rDNA analysis on the complete sequences of the respective genes. The support for each branch, as determined from 100 bootstrap samples, is indicated by the value at the node. This analysis provided tree topology only, and the lengths of both vertical lines and horizontal lines are not significant. Boxes contain clusters that are consistently found with at least two genes. Cluster A: B. bacilliformis. Cluster B: B. clarridgeiae. Cluster C: B. quintana, B. henselae, and B. koehlerae. Cluster D: Nine strains associated with New World native rodents (Sigmodon sp., Oryzomys sp., Phylottis sp.) with one Rattus sp.-associated isolate from the USA. Cluster E: B. vinsonii subsp. vinsonii and B. vinsonii subsp. berkhoffii. Cluster F: B. elizabethae, B. grahamii, and B. tribocorum with seven strains associated with Old World native rodents (Rattus sp., Mus sp.).

While 16S rRNA-based analyses have played a major role in establishing the phylogenetic position of the genus Bartonella among Proteobacteria and also in defining the limits of this genus [2,3], they are considered insensitive for intragenus phylogeny, because levels of 16S rDNA sequence similarity are very high (97.5–99.9%) between Bartonella species [5,1012] and inferences are difficult to make with confidence from sequences that exhibit more than 97% similarity [21]. The branching orders of the trees derived from comparison of 16S rDNA genes have only moderate significance, as no branch, other than that containing B. bacilliformis, is supported by more than 90% of the bootstrap samples. Comparison of tree topology inferred using parsimony and distance methods has also revealed marked differences [17]. To date, only one study has assessed ITS-based phylogeny [27]. Intra-Bartonella phylogenies derived from gltA and groEL data have some statistical significance and resulted in the same topology with all used methods. The bootstrap values of the gltA trees were higher when using distance methods than when using parsimony, whereas GroEL trees showed greater support when using parsimony analysis [17,18]. Additionally, trees based on groEL and 17 kDa antigen amino acid sequences have been generated by using parsimony methods (the PROTPARS algorithm of PHYLIP) to overcome the potential bias resulting from synonymous base substitutions, but these received little statistical support [18,21,28].

3 Current phylogeny

Comparisons of phylogenetic information resulting from 16S rDNA-, ITS-, gltA- and groEL-based genotypic studies have resulted in six clusters being consistently identified (i.e. clusters which were statistically supported to at least two different genes, with all analytical methods) (Fig. 1). The unique, relatively deep-rooted divergence of B. bacilliformis (group A) is supported with 100% bootstrap values by 16S rDNA, ITS, gltA and groEL analyses, as well as by its particular fatty acid, protein and antigenic composition, growth requirements and colony morphology [4,17,18,29]. Levels of similarity for amino acid sequences of ftsZ proteins, as well as for nucleotide sequences of ribC genes, are low between B. bacilliformis and other Bartonella species, supporting its divergent position [19,20]. According to ITS and groEL analyses, B. clarridgeiae may also be a divergent species (group B). A close evolutionary relationship between B. clarridgeiae and cat-associated strains from the Netherlands and the USA is suggested by 16S rDNA-based phylogeny. As B. clarridgeiae and B. bacilliformis are two flagellated species, they have been suspected to share close relationships, but statistical support is lacking for such organization [15,18,27]. The third group (C), which contains B. quintana and B. henselae, is found with bootstrap values >90% in trees derived from gltA and groEL sequences. The high degree of conservation between ftsZ homologs, ribC proteins and 17 kDa antigens from B. henselae and B. quintana also indicates that these two species are closely related [19,20,28]. Comparison of 16S rDNA and gltA sequences suggests that B. koehlerae may cluster with this group, and more closely with B. henselae.

Group D, determined by gltA-based phylogeny only, included taxons associated with mice and rats caught in Peru and in the USA, respectively [1,1517]. This group can be divided into two subgroups, one including the mouse-associated bacteria and the other including the rat-associated species (Fig. 1). A strong association between B. vinsonii subsp. vinsonii and B. vinsonii subsp. berkhoffii is observed with ITS, groEL and gltA sequence studies (group E), but not with those of the 16S rDNA gene [30]. The position of this B. vinsonii cluster relative to other Bartonella species remains unclear. The sixth group (F) is obtained with 16S rRNA, ITS, gltA and groEL gene sequence comparisons. Like group D, it contains rodent-associated organisms. B. elizabethae, B. grahamii, B. tribocorum cluster into this clade, together with uncharacterized strains found in mice and rats from the USA, Peru and Europe [1,16,17]. Within group F, B. elizabethae clusters with isolates from Peruvian rats in a separate sub-clade supported by gltA analysis. The taxonomic position of other species and strains, mainly B. taylorii, B. doshiae and B. alsatica, could not be consistently determined on the basis of genetic sequence data available to date.

4 Host specificity and phylogeny

One of the main interests of phylogenetic studies is the possibility of eventually identifying lineages of bacteria that share common epidemiological or biological characteristics, as well as shared taxonomic origins. It can provide clues to help understand the natural history of Bartonella species, their relationships with mammalian hosts and arthropod vectors, and their geographical repartition. As Bartonella parasites a wide range of mammals, the existence of a specific association between one Bartonella species and a single mammalian host has been suggested. The notion that specific hosts are vulnerable to infection by only one species of bacterium, reflecting a co-evolution phenomenon, is supported by the fact that, to be a competent reservoir, an animal must be susceptible to a parasite in a way that allows development of a chronic infection [13,15,17]. Table 1 summarizes the identified vertebrate hosts, arthropod vectors and geographic distribution of the Bartonella species belonging to the major phylogenetic clusters that have been identified.

View this table:
Table 1

Characteristics of the major phylogenetic clusters obtained consistently with analysis of the 16S rDNA, ITS, gltA and groEL gene sequences for Bartonella species

ClusterGene providing significant support for cluster determinationTaxonsReservoir hostArthropod vectorPlace of isolationRef.
A16S rRNA, ITS, gltAB. bacilliformishumansandfly (Lutzomyia verrucarum)Peru, Ecuador, Colombia[1,4]
BITS, groELB. clarridgeiaecat?USA, France[5,37]
CgroEL, gltAB. quintanahumanbody louse (Pediculus humanis)worldwide[4]
B. henselaecatcat flea (Ctenocephalides felis)?worldwide[4,33,37]
B. koehleraecat?USA[11]
DgltABartonella sp. R-PHY1mouse (P. peruviana)?Peru[17]
Bartonella sp. SH8200GA, Bartonella sp. OP6399GA, Bartonella sp. RR11755TXrats (S. hispidus, O. palustris, R. rattus)?USA[15,16]
EITS, groEL, gltAB. vinsonii subsp. vinsoniivole (Microtus pennsylvanicus)rat flea (Nosopsyllus fasciatus), vole ear mite (Trombicula miroti)Canada[35]
B. vinsonii subsp. berkhoffiidogticks?USA, Europe[6,30]
F16S rRNA, ITS, groEL, gltAB. grahamiivole (C. glareolus)?UK, Netherlands[3,9]
B. elizabethaerat (R. norvegicus)fleas (Xenopsylla cheopis)?USA, Peru[1,7,13,16]
B. tribocorumrat (R. norvegicus)?France[10]
Bartonella sp. C5RATrat (R. norvegicus)?USA, Peru, Portugal[16,17]
Bartonella sp. MM5136CAmouse (M. musculus)?USA[16]
UnclassifiedB. doshiaevole (Microtus agrestis)?UK[3]
B. tayloriirat (Apodemus sp.)?UK[3]
B. alsaticarabbit (Oryctolagus cuniculus)?France[12]

Taxonomic data supporting the hypothesis of co-evolution can be mainly found among cat- and rodent-associated Bartonella strains. Thus, the cat-associated species B. henselae and B. koehlerae appear to be closely related to each other and B. clarridgeiae forms a cluster with other cat isolates (Fig. 1). Among rodent isolates, all bacteria recovered from Phylottis mice in Peru shared close relationships [17]. Isolates from rice rats (Oryzomys palustris) also appeared to be genetically very close [15]. In a study of 325 Rattus norvegicus rats in the USA and in Portugal, all 63 Bartonella recovered from these rats clustered into a well-defined clade [16]. According to 16S rDNA and ITS analyses, B. taylorii, a species isolated from mice (Apodemus sp.), may be closely related with other Apodemus-associated strains (Fig. 1) [22]. Similarly, a large-scale sampling of multiple species of Peromyscus sp. from a variety of sites in different physiographic regions of the south-eastern USA provided several Bartonella isolates that clustered together using gltA-based analyses [15]. A study of 128 cattle from California established that all 63 cattle isolates had the same profile by using PCR/RFLP analysis of the gltA gene [14]. These results, together with recent experimental evidence among certain rodent species, strongly suggest that some host specificity may occur within the genus Bartonella, although it is far from being absolute [31].

Infection of the same mammal species by several Bartonella which belong to distinct phylogenetic groups has been widely reported: in species of small wild mammals caught in the UK [32], in the cotton rat population from the south-eastern USA [15] and also in deer and elk in California and Oregon [14]. Conversely, the same Bartonella species can be isolated from different mammal species that inhabit the same geographical location. Although observed on only a few occasions, the observation that some Rattus rattus-associated strains from the USA were genotypically indistinguishable from bacteria isolated from the American cotton rat, Sigmodon hispidus, also suggests that species can cross from one rodent species to another. Indeed, the ubiquitous species R. rattus frequently occurs in rural environments, where it may come into contact with indigenous New World rodents [16]. Similarly, the identity between B. elizabethae- and R. norvegicus-associated strains demonstrated, for the first time, that rodent-associated Bartonella species could act as human pathogens [1] and suggested that the predominantly urban rat R. norvegicus could be a reservoir for human infection among inner-city residents living in poor social conditions [16]. Several other animal-associated Bartonella have shown their ability to infect humans, inducing clinical manifestations which are different from those observed in natural reservoir hosts: this is the case for B. vinsonii subsp. arupensis, which may have a rodent reservoir [8], for B. vinsonii subsp. berkhoffii, initially described in dogs [6], and for the cat-associated species B. henselae and B. clarridgeiae [4,5].

To explain such lack of host specificity, it has been speculated that vector specificity may be an important factor in the natural history, parasitic cycle and distribution of Bartonella species. Thus, as several different species of fleas are naturally found as ectoparasites on small mammals, infection by different Bartonella species may originate from different flea species, with each Bartonella exhibiting species specificity to its vector. Conversely, sharing of the same flea fauna by different mammals could result in infection of various mammals by the same Bartonella species [17]. Ticks have been also suspected to play a role in transmission of Bartonella from one mammal species to another [14,33]. Collection and analysis of arthropods caught on animals and humans and in the environment is necessary to determine if they can be infected with Bartonella species [14]. All vectors of Bartonella species will need to be identified before it can be determined if co-evolution of the organisms has occurred [34].

5 Geographic specificity and phylogeny

Within the genus Bartonella, gltA-based analyses suggest that bacteria associated with indigenous rodents from the New World (such as cotton rats S. hispidus, rice rats O. palustris and mice Phylottis peruviana) are phylogenetically distinct from species that have been recovered from Old World rodents (such as rats R. norvegicus and R. rattus, mice Mus musculus and voles Clethrionomys glareolus) (Table 1, groups D and F, respectively) [16]. Such a phylogenetic divergence between New World and Old World Bartonella species suggests that species that were separated as a result of the continental drift may have evolved independently of one-another. Consequently, Bartonella species may have a restricted geographical distribution in as far as their animal hosts and vectors have never been exported from one continent to another. B. grahamii, B. doshiae, B. talpae and B. taylorii for the Old World, as well as B. vinsonii subsp. vinsonii and B. vinsonii subsp. arupensis for the New World, may illustrate a situation of geographical segregation resulting from the restricted distribution of their specific mammalian reservoir hosts [3,8,32,34]. The strictly limited distribution of B. bacilliformis, however, might be related to the restricted geographical area of its sandfly vector, as no animal host was found in endemic areas and humans are regarded as the most probable reservoir for this species [1,11,17].

Conversely, dissemination of infected animals or vectors resulted in the widespread extension of their associated parasites. B. quintana, B. henselae and probably B. vinsonii subsp. berkhoffii, which are associated with ubiquist hosts (human, cat and dog, respectively), harbor a widespread extension (Table 1) [4,30,35]. While originally native from the Old World, rats from the genus Rattus were introduced into the New World during the Spanish Conquest and may have carried their own Bartonella strains [1,16]. However, data currently available regarding the geographical distribution of Bartonella should be interpreted carefully, as efforts to identify these organisms to the species level are not equally active in all regions of the world. The abundant information concerning Bartonella in the USA and Europe contrasts with only very limited data derived from Asia or Africa. Further characterization of the available Bartonella isolates and efforts to identify bacteria in other animals and locations, will help in the understanding of the geographical distribution of the genus.

6 Conclusion

Despite recent studies, the phylogeny of Bartonella remains confusing and it is likely that this may be exacerbated as new isolates of Bartonella continue to be made in large numbers from mammal populations [1417,22]. Although 16S rDNA-based phylogenetic studies give little information on evolutionary processes occurring within the genus Bartonella, studies using comparisons of ITS, gltA and groEL sequences have enabled six distinct evolutionary clades to be identified within the genus. As multiple gene data sets have been shown to be useful in determining the precise molecular relationships between genera and species [18], comparisons of the sequences of the ribC, the ftsZ and the 17 kDa antigen genes might be used to improve our understanding of the evolutionary process, but this will only be possible when such data become available for all Bartonella species. This will specifically help in the identification of lateral gene transfer, which occurs more commonly in intracellular pathogens [36].

Although still partial, the phylogenetic data available for Bartonella species suggest that a phenomenon of co-evolution occurred between the bacteria and their animal reservoir hosts. This host specificity may be partially offset by concomitant vector specificity. As a consequence of the restricted geographical distribution of their hosts and vectors, geographically distant Bartonella species may have evolved separately. It would be interesting to combine genetic classifications with phenotypic data sets (such as duration of experimental bacteremia [33]) to better understand the natural history of Bartonella species, their association with specific mammal hosts, and to confirm the apparent divisions identified within the genus.


The authors thank Alia Ben Kahla for her help in the computer-assisted construction of phylogenetic trees.


  1. [1].
  2. [2].
  3. [3].
  4. [4].
  5. [5].
  6. [6].
  7. [7].
  8. [8].
  9. [9].
  10. [10].
  11. [11].
  12. [12].
  13. [13].
  14. [14].
  15. [15].
  16. [16].
  17. [17].
  18. [18].
  19. [19].
  20. [20].
  21. [21].
  22. [22].
  23. [23].
  24. [24].
  25. [25].
  26. [26].
  27. [27].
  28. [28].
  29. [29].
  30. [30].
  31. [31].
  32. [32].
  33. [33].
  34. [34].
  35. [35].
  36. [36].
  37. [37].
View Abstract