More than 500 bats were collected from 25 locations across Kenya (Figure 1). Sampling sites were chosen on the basis of available information about bat roosts and by using field observations of flying and foraging bats. The number of samples and the collection protocol were justified and approved by the National Museums of Kenya and the Kenyan Wildlife Service. Detailed information on the collection procedure has been described elsewhere (5). Briefly, bats were collected by using hand nets or collected manually in caves and human dwellings and were captured in mist nets around roosts or in locations of nocturnal foraging. Captured bats were anesthetized by an intramuscular injection of ketamine hydrochloride (0.05–0.1 mg/g bodyweight) and euthanized under sedation in compliance with the field protocol approved by CDC’s Institutional Animal Care and Use Committee. The bats were measured, and their sex and species were identified.

Species were identified phenotypically by using available field keys. Additionally, representative tissues of each collected species were submitted for confirmation to Guelph University (Guelph, Ontario, Canada), where partial sequences of the cytochrome oxidase gene were generated and compared with those available from the database of the Barcode of Life Data Systems (BOLD; www.barcodinglife.org). Because the taxonomy of African bats is under development and reference sequences for several species were unavailable in the BOLD database, the examined sequences were identified to the genus level only.

For microbiologic studies, selected organs and swabs were collected by using sterile plastic tubes. Serum was separated from the blood clots by centrifugation. All samples were transported on dry ice and stored at –80°C until testing.

Preliminary attempts to culture Bartonella spp. from bat blood showed that the technique developed for isolation of these bacteria from rodent blood (18) is appropriate, with some minor modifications, for processing bat samples. Specifically, we used agar plates supplemented by a 10% addition of rabbit blood. Blood from bats was resuspended in brain–heart infusion (BHI) broth supplemented with 5% amphotericin B. The ratio between blood and BHI was determined to be 1:4. However, this ratio was difficult to adhere to for samples with a limited amount of blood (obtained from small bats). Although dilution of tested blood reduces our ability to detect bacteria in cases with a low level of bacteremia, this approach allowed us to reduce the likelihood that bacterial and fungal contaminants would overgrow the fastidious and slow-growing Bartonella spp. bacteria. Bacterial colonies were presumptively identified as Bartonella spp. on the basis of their morphology and later were confirmed by PCR amplification and sequencing of a specific fragment of the gltA. Subcultures of Bartonella spp. colonies from the original agar plate were streaked onto secondary agar plates, also supplemented by a 10% addition of rabbit blood, and, in case of confluent and pure cultures, harvested and stored in 10% glycerol. Agar plates inoculated with bat blood were incubated at 35°C in an aerobic atmosphere of 5% carbon dioxide for ≤30 days postinoculation, and plates with subcultures were incubated with the same conditions until sufficient growth was observed, usually 5–7 days.

A heavy suspension of the microorganisms was heated for 10 min at 95°C followed by 1-min centrifugation at 8,000 rpm to precipitate the lysed cells. The supernant containing the genomic DNA was then moved to a clean centrifuge tube to be examined. PCR amplifications were performed in a 25-μL reaction mixture containing 5 μL 5× Green GoTaq PCR buffer (Promega, Madison, WI, USA), 0.4 μmol of each primer, 200 μM each dNTP, 1 U Taq DNA polymerase (Promega), and ≈20 ng of template DNA. Two oligonucleotides, BhCS781.p (5′-GGGGACCAGCTCATGGTGG-3′) and BhCS1137.n (5′-AATGCAAAAAGAACAGTAAACA-3′) were used as PCR primers to generate a 379-bp amplicon of the Bartonella gltA gene. Positive and negative controls were included with each PCR to evaluate the presence of appropriately sized amplicons and contamination, respectively. Each PCR was performed in a PTC 200 Peltier thermal cycler (MJ Research, Inc., Waltham, MA, USA) or in an Eppendorf Mastercycler Gradient (Eppendorf, Westbury, NY, USA). PCR products were separated by 1.5% agarose gel electrophoresis and visualized by ethidium bromide staining.

PCR products of correct size were purified with the QIAquick PCR Purification Kit (QIAGEN, Valencia, CA, USA) according to manufacturer’s instructions and sequenced in both directions by using an Applied Biosystems Model 3130 Genetic Analyzer (Applied Biosystems, Foster City, CA, USA). Sequencing reactions were performed in a PTC 200 Peltier Thermal cycler by using the same primers as the initial PCR with a concentration of 1.6 µM. Sequences were analyzed by using Lasergene (DNASTAR, Madison, WI, USA) sequence analysis software to determine consensus of sequences for the amplified region of the gltA gene. The Clustal V program within MegAlign (DNASTAR) was used to align and compare homologous Bartonella spp. gltA sequences obtained from bat samples and from the GenBank database.

Culturing from bat blood pellets, especially from small-sized bats, presented a challenge because of the limited sample volume and the potential for contamination with other bacteria and fungi, problems hard to avoid during field sampling. Consequently, of >500 processed blood samples, the presence or absence of Bartonella spp. cultures have been conclusively evaluated in samples from 331 bats of 13 species from 9 genera. Further estimation of the infection rate was determined exclusively on the basis of the specimens from these 331 animals. The time required for growth of Bartonella spp. colonies on agar greatly varied from 3 days, observed for several samples from Triaenops persicus bats to 28 days in one of the samples from Rousettus aegyptiacus bats.

Bartonella isolates were cultured from the blood of 30.2% (106/331) of the bats tested. All isolates were confirmed genetically by amplification and sequence of the gltA. Prevalence of bats positive for Bartonella spp. by culture was as follows: Eidolon helvum (straw-colored fruit bat), 23/88 (26.1%); R. aegyptiacus (Egyptian fruit bat), 22/105 (21.0%); Coleura afra (African sheath-tailed bat), 4/9 (44.4%); T. persicus (Persian trident bat), 7/8 (87.5%); Hipposideros commersoni (giant leaf-nosed bat), 1/4 (25.0%); and Miniopterus spp. (long-fingered bats), 49/87 (56.3%) (Table). Miniopterus spp. and T. persicus bats were at significantly higher risk (p<0.0001 and p<0.001, respectively) for being infected with Bartonella spp. when each was compared with all other bats, and R. aegyptiacus and Epomophorus spp. bats were at significantly lower risk for infection (p<0.01 for both) (Table).

Sequence analyses of DNA from the 94 Bartonella isolates obtained from bats revealed 58 gltA genotypes (unique sequence variants with ≥1 nucleotide difference) that represented 11 genogroups with a sequence identity value of >96% between genotypes within each group, as proposed by La Scola et al. (28), as a cutoff for this specific gene fragment for species definition within the Bartonella genus. The 11 novel gltA sequences representing each genogroup were submitted to GenBank and assigned accession nos. HM363764–363768 and HM545136–545141.

All 15 Bartonella spp. gltA sequences obtained from R. aegyptiacus bats were similar to each other (>96%) and distant from all sequences found in other bat species (<91% identity) and previously described Bartonella species (<84% identity). Six unique genetic variants identified in R. aegyptiacus bats were clustered in a monophyletic genogroup, marked as R in Figure 2.

The level of gltA sequence identity between 16 Bartonella genotypes identified in E. helvum bats varied greatly, ranging from 78.6%–100%. Eight genotypes (sequence variants) were clustered around 4 clades (genogroups), which are marked as E1, E2, E3, and EW in Figure 2. Intergroup identity values ranged from 78.6%–87.2%, much higher values than the sufficient criteria proposed for species demarcation within the Bartonella genus (28).

All 4 gltA genotypes of Bartonella spp. identified in C. afra bats differed slightly from one another but clustered tightly within genogroup C with a sequence identity range of 98.2%–99.7%, compared with a <90% identity between these sequences and Bartonella strains found in other bat species or other sources. Similarly, all 7 gltA sequences identified in T. persicus bats formed a monophyletic genogroup, marked as T in Figure 2, with the identity range of 98.2%–99.7% within the group and <90% identity to any Bartonella sequence outside the group. The only isolate obtained from H. commersoni bats also differed from all described Bartonella strains or sequences (identity <85%).

More complex phylogeny was observed when isolates obtained from bats of the genus Miniopterus were compared by using a similarity between the gltA sequences. In total, 51 Bartonella spp. gltA sequences, obtained from 3 identified Miniopterus spp. (M. africanus, M. minor, and M. natalensis) and from Miniopterus spp. bats that have not been identified at the species level, were analyzed. Among the 51 bats, 24 unique gltA genotypes represented at least 3 genogroups with a level of divergence ranging from 6.6%–18.5%, higher than has been recommended previously for differentiation of Bartonella spp (28). The first genogroup, M1, was composed of 30 sequences obtained from M. minor bats, 1 sequence from M. natalensis bats, and 1 sequence from an unidentified species within the genus Miniopterus. The second genogroup, M2, consisted of 5 sequences from M. minor bats, 2 sequences from M. africanus bats, 1 sequence from M. natalensis bats, and 2 sequences from an unidentified species of Miniopterus spp. bats. Although the second genogroup looked more diverse, an insufficient number of isolates were available to describe an association of specific Bartonella spp. lineages to certain Miniopterus spp. bats. In addition, 1 genotype, M3, from M. natalensis bats, was distant from both groups (identity <90%), but was relatively close (identity 94.2%) to the genotype identified in a whiskered bat (Myotis mystacinus) from the United Kingdom (11).