Bacteria in the genus Bartonella are found in a wide variety of mammalian hosts (2, 8, 18, 20, 21) and are thought to be transmitted by arthropod vectors, including fleas, ticks, and possibly mites (4, 7, 9, 14, 16); humans serve as accidental hosts. Of the at least 20 named Bartonella species, 10 have been shown to cause disease in humans, including Carrion's disease (13), cat scratch disease (7, 14, 25), endocarditis (6, 11), and recently a febrile illness in humans from Thailand (caused by Bartonella tamiae) (17). Because of their wide distribution and potential for frequent contact with humans, many Bartonella species are considered potential emerging pathogens (1, 26, 28).

A whole-genome scan was performed on complete genomic sequences from B. henselae and B. quintana and shotgun sequences from B. bacilliformis available in GenBank. Each subsequence of 16, 17, 18, and 19 nucleotides present in published Bartonella genomes was compared with subsequences from other genomes present in GenBank, including genomes for bacteria that could infect human blood and tissues and potential mammalian hosts and arthropod vectors for bartonellae. The number of base changes necessary to convert each Bartonella subsequence to the closest subsequence in the background collection was calculated to identify potential primers with a reduced probability of hybridizing to and amplifying nontarget DNA.

The nuoG, gltA, and ftsZ primer sets were used to amplify reference DNAs from 11 Bartonella species, chosen for their distant phylogenetic relationships, under conditions optimized for each primer set. The amplification results differed considerably between primer sets and species of Bartonella being amplified (Table ​(Table2)2) and are as follows: the nuoG primer set performed best (amplifying first, with the lowest threshold cycle [C_T] value) on 3 of the 11 tested species, the gltA set performed best on 7 of the 11 species, and the ftsZ set performed best on 1 of the 11 species. Although the gltA primer set performed best for the highest number of reference species, only the nuoG set successfully amplified all 11 species.

Analysis of a nuoG-derived phylogeny showed strong statistical support for the following clades: B. henselae and Bartonella koehlerae; the species found in Rattus and related hosts, including Bartonella elizabethae, B. rattimassiliensis, and B. tribocorum; 3 Bartonella vinsonii subspecies (Bartonella vinsonii subsp. arupensis, B. vinsonii subsp. vinsonii, and B. vinsonii subsp. berkhoffii); and two strains of B. tamiae, described to occur in febrile Thai patients (type strain Th239 and strain Th307) (Fig. ​(Fig.1).1). All of these species groups share high genetic similarity within their respective clades, suggesting that the nuoG primer set provides better phylogenetic estimation with closely related species. Bartonella bovis was placed extremely distant to the other Bartonella species, with strong statistical support; conversely, B. bacilliformis was placed more centrally within the phylogeny than is seen with other genes, though this placement did not have strong statistical support. These placements, which are different from those generated with multiple concatenated Bartonella sequences (Fig. ​(Fig.1)1) (17), are likely due to the genetic rearrangements and horizontal gene transfer events that commonly occur in Bartonella (3, 12, 19, 23). Because of this, care should be taken when interpreting phylogenies based solely on nuoG sequences. A much more reliable approach is to include the nuoG sequence as one of many concatenated sequences to be used for phylogenetic analysis.