Molecular Detection of Bartonella spp. in China and St. Kitts

Bartonella are vector-borne hemotropic bacteria that infect a wide variety of hosts, including people. While there are PCR assays that can identify individual or groups of Bartonella, there is no reliable molecular method to simultaneously detect all species while maintaining genus specificity and sensitivity. By comparing highly conserved 16S rRNA sequences of the better-recognized Bartonella spp. on GenBank, we selected primers and probes for a genus-specific pan-Bartonella FRET-qPCR. Then, a gltA-based Bartonella PCR was established by selecting primers for a highly variable region of gltA, of which the sequenced amplicons could identify individual Bartonella spp. The pan-Bartonella FRET-qPCR did not detect negative controls (Brucella spp., Anaplasma spp., Rickettsia spp., Coxiella burnetii, and Wolbachia) but reliably detected as few as two copies of the positive control (Bartonella henselae) per reaction. There was complete agreement between the pan-Bartonella FRET-qPCR and the gltA-based Bartonella PCR in detecting Bartonella in convenience test samples from China and St. Kitts: cats (26%; 81/310), Ctenocephalides felis (20%; 12/60), cattle (24%; 23/98), and donkeys (4%; 1/20). Sequencing of the gltA-based Bartonella PCR products revealed B. henselae (70%; 57/81) and B. clarridgeiae (30%; 24/81) in cats and C. felis (67%; 8/12, and 33%; 4/12, respectively) and B. bovis in cattle (23.5%; 23/98) and donkeys (4.0%; 1/24). The pan-Bartonella FRET-qPCR and gltA-based Bartonella PCR we developed are highly sensitive and specific in detecting recognized Bartonella spp. in a single reaction. The pan-Bartonella FRET-qPCR is convenient requiring no gel electrophoresis and providing copy numbers, while the gltA-based Bartonella PCR reliably differentiates individual Bartonella species. The use of these PCRs should greatly facilitate large-scale surveillance studies and the diagnosis of infections in clinical samples.


Introduction
ere are now over 40 species and subspecies of Bartonella which are small, intracellular, vector-borne hemotropic Gram-negative bacteria. High prevalences of the organisms have been reported around the world in a wide range of insect vectors and domestic and wild animal hosts including rodents, felines, canines, ruminants, and even bats [1][2][3][4]. In China, Bartonella species have been detected in a wide range of animals [5][6][7][8] with at least 10 Bartonella species having been implicated in human diseases that range from selflimiting regional lymphadenitis to severe endocarditis [9][10][11].
Detecting Bartonella species in their vectors and mammalian hosts is often not easy as it is difficult and timeconsuming to culture them and serology may be unreliable. Although numerous molecular methods to detect Bartonella have been described [17][18][19][20][21][22][23][24][25][26][27][28], in most cases, a single PCR only detects individual or closely related species of Bartonella [29]. Although multiplex PCRs have been described to detect Bartonella, they require a wide range of primers and probes to detect all species. Furthermore, the PCRs have poor sensitivity (with up to 8 × 10 7 bacteria per g being undetectable in feces) [30,31] because of inhibitors in environmental and clinical samples. Recently, a highly specific and quantitative qPCR has been established to specifically amplify nuoG of the Bartonella genus and identified B. henselae and B. clarridgeiae in cats in Brazil [32].
We used the available 16S rRNA sequences of 20 Bartonella species in GenBank to develop and validate a highly sensitive and genus-specific pan-Bartonella FRET-qPCR. We also used available sequences for the citrate synthase gene (gltA) and considered a reliable tool for distinguishing genotypes [33], to develop a gltA-based Bartonella PCR to identify and differentiate the species. e development of these PCRs and their use on convenience samples from China and St. Kitts is described below.  (Figure 1). ese sequences were aligned using Clustal multiple sequence alignment in VectorNTI (Informax Inc., North Bethesda, MD, USA) to identify a highly conserved region of the 16S rRNA common to all the above Bartonella spp., but significantly different from the other species ( Figure 1). e primers and probes we developed were situated within the conserved region and synthesized by Integrated DNA Technologies (Coralville, IA, USA). e pan-Bartonella FRET-qPCR we established amplifies a 304 bp target with the positions of primers and probes shown in Figure 1: forward primer: 5′-AGCG-CACTCTTTTAGAGTGAGCGG-3′; reverse primer: 5′-CATGGCTGGATCAGGGTTGCC-3′; anchor probe: 5′-TCAGCCACACTGGGACTGAGACACG-(6-FAM)-3′; and reporter probe: 5′-(LCRed640)-CCAGACTCCTACGGGA GGCAGCA-phosphate-3′.

Sensitivity.
For quantitative standards, we used amplified DNA of B. henselae identified in this study. e amplification products were gel purified using the QIAquick Gel Extraction Kit (Qiagen, Valencia, CA, USA) and quantified using the PicoGreen DNA fluorescence assay (Molecular Probes, Eugene, OR, USA). e molarity of the DNA was estimated using the calculated molecular mass of the amplicons and dilutions made to give solutions containing 10,000, 1,000, 100, 10, and 1 gene copies/μl in T 10 E 0.1 buffer. ese were used as quantitative standards in the FRET-qPCR surveys to enable standard curves to be developed for the calculation of the gene copy numbers in positive samples. Twofold dilutions of 10 and 1 gene copies/μl solution were used to determine the minimal detection limit.

Development of a Citrate Synthase Gene-(gltA-) Based PCR to Differentiate Bartonella spp.
e amplicon of the pan-Bartonella FRET-qPCR we established was based on a region of the 16S rRNA that is highly conserved among different Bartonella spp. and is thus ideal for screening samples for the organisms. To differentiate Bartonella spp. that gave positive reactions in screening pan-Bartonella FRET-qPCR, we used a standard PCR that amplifies a relatively highly polymorphic region of the citrate synthase gene (gltA) (439 nucleotides for different Bartonella spp.) and sequenced the products (GenScript, Nanjing, Jiangsu, China). For the gltA-based Bartonella PCR, we designed, as above for the pan-Bartonella FRET-qPCR, a forward primer (5′-CGTAATGATCTYAGTTAYGCTGCAAA-3′) and a reverse primer (5′-AGAAGTGGATCATTTTGAATRTT-BARYTC-3′) that amplify all the 18 better-recognized Bartonella spp. available in GenBank ( Figure 2). e sensitivity and specificity of the gltA-based PCR were assessed as described above for the pan-Bartonella FRET-qPCR. e PCRs were performed in a LightCycler 480 ® II real-time PCR platform with 20 μl final volumes containing 10 μl extracted DNA, and 1 μM primers with the same PCR buffer as described above. ermal cycling of gltA-PCRs was performed the same as pan-Bartonella FRET-qPCR except removing the melting curve analysis and revising annealing temperature to 56°C. All the PCRs were run in duplicate.

Phylogenetic Analysis.
e retrieved Bartonella sequences of the present study and the reference sequences from GenBank were aligned using the MEGA 6.0 software    ( Figure 3). Based on these alignments, phylogenetic trees were constructed by the maximum likelihood and Bayesian inference method using MEGA 6.0. Bootstrap values were calculated using 500 replicates.

Testing of the Pan-Bartonella FRET-qPCR and Survey of Animals and Fleas for Bartonella spp. DNA.
e probes we designed for the pan-Bartonella FRET-qPCR were highly conserved, while the primers only had between zero and four nucleotide mismatches with the 18 representative Bartonella spp. we used in the study (Figure 1). In contrast, the primers and probes had 11,26,27,10,17,23,22, and 29 mismatches with A. phagocytophilum, A. marginale, Wolbachia, B. melitensis, B. chilestrain, R. felis, R. rickettsia, and C. burnetii, respectively (Figure 1). e pan-Bartonella FRET-qPCR was positive with B. henselae DNA, but no products were obtained with DNAs from Wolbachia, A. equi, A. marginale, B. melitensis, B. chilestrain, R. felis, R. rickettsia, and C. burnetii. Using the gel-purified PCR products as quantitative standards [20], we determined the detection limit of the pan-Bartonella FRET-qPCR was two copies of the Bartonella 16S rRNA per reaction.
To confirm accuracy and reproducibility of real-time PCR, the intra-assay precision was determined in three repeats within one LightCycler run. Interassay variation was investigated in three different experimental runs, while the quantitative standards (1,000, 100, 10, and 1 copies/reaction) were diluted in 1 × T 10 E 0.1 , and with the diluents containing lambda DNA (New England Biolabs, Inc.) at the concentration of 1 ng/μl. e calculation of test precision and test variability is based on the CP variation from the CP mean value. Test reproducibility for all investigated quantitative standards was lower in intratest experiments (<3.8%) and was higher in intertest experiments (<6.1%). e reproducibility dropped in lower copies of standards (1 and 10 copies per reaction) in comparison with the high copies of standards (100 copies and 1,000 copies).
Of the 310 cat blood samples we examined, 78 (25.2%) were positive in our pan-Bartonella FRET-qPCR. Cats from all the sites we studied were positive, mainly from Beijing (11.1%), Shanghai (5.6%), Guangdong (15.0%), Yangzhou (19.4%), and St. Kitts (39.7%) ( Table 1). e 60 fleas collected from the Chinese cats were identified as Ctenocephalides felis based on their morphology, and 20.0% (12/60) were positive in the pan-Bartonella FRET-qPCR, while only one donkey (5.0%) from China was positive. Horses, sheep, and donkeys from St. Kitts were negative, but cattle from the island were commonly positive (54.8%) in the pan-Bartonella FRET-qPCR, although cattle from China were negative.
In general, the copies detected in the pan-Bartonella FRET-qPCR were considerably higher in the positive blood samples we collected (10 6.20 /ml whole blood). Fleas positive for B. henselae generally had low copy numbers, but those positive for B. clarridgeiae had considerably higher copy numbers (10 7.01 vs. 10 2.88 ).

Discussion
As knowledge on the wide diversity and host ranges of Bartonella species increases, it becomes ever more important to develop reliable molecular assays that can detect all Bartonella species and thereby facilitate epidemiological and ecological studies. e pan-Bartonella FRET-qPCR we developed proved to be specific, not giving reactions with closely related organisms.
is was expected as, by systematically aligning the sequences of 18 better-recognized Bartonella spp. in GenBank and other species, we were able to identify a highly conserved and specific region of the 16S rRNA for the PCR (Figure 1). e pan-Bartonella FRET-qPCR was also very sensitive, detecting as few as two Bartonella 16S rRNA copies per reaction. Further advantages were that it was quantitative and relatively rapid to perform as gel electrophoresis was not needed to demonstrate amplicons.
Although the pan-Bartonella FRET-qPCR was very sensitive (2 copies per reaction) and specific, quantitative, and rapid (normally 1.5 hours including DNA extraction), because it was designed against a highly conserved region of the 16S rRNA, it did not enable us to determine the Bartonella spp. in the samples. is was possible, however, with the gltA-based Bartonella PCR which was as accurate and specific as the pan-Bartonella FRET-qPCR in identifying Bartonella spp. in our study. Scola et al. first reported that, of the commonly used genetic targets for Bartonella identification, only gltA and rpoB sequences provide sufficient discriminatory power and interspecies diversity to allow discrimination of Bartonella spp. [33]. We therefore used gltA as the target of our PCR and selected our primers to ensure that, although they only amplified Bartonella spp., they covered a 439 bp hypervariable area of the gene that enabled differentiation of the recognized and poorly characterized Bartonella spp. recorded in GenBank (Figure 2). Our gltA-based PCR clearly differentiated the control B. henselae and identified three common species in our survey, B. henselae, B. clarridgeiae, and B. bovis, which were the species expected to be present in the animals surveyed.
Applying our pan-Bartonella FRET-qPCR to a number of convenience samples available enabled us to rapidly test a variety of samples from different geographic areas for Bartonella. We obtained positive results with all the samples tested, mainly mammalian whole blood, and vector insects (fleas) which comprise the most commonly used samples to date in epidemiological studies. Although the gltA-based Bartonella PCR was more time-consuming, requiring gel electrophoresis to reveal amplicons, it identified all the samples that were positive in the pan-Bartonella FRET-qPCR.
Bartonella henselae and B. clarridgeiae were the species most commonly identified by the gltA-based Bartonella PCR in our study. ese are the organisms most commonly associated with cats and their fleas and have already been described to be prevalent in the West Indies [13] and China [40] where they are often associated with persistent subclinical bacteremia and transmitted by C. felis [2]. Although we found B. bovis at a high level in cattle from St. Kitts, we failed to identify the organism in cattle from China. We did, however, find B. bovis in a donkey in China, indicating the organism is present in the country. Bartonella bovis is a recently described species [41] that is generally the most prevalent of the three species that infect cattle, the others being Bartonella schoenbuchensis and Bartonella chomelii. e latter have mainly been described in Europe or in cattle originating from Europe [42], while B. bovis has been described from widely around the world with prevalences in cattle varying from 0% in Kenya and Japan [43] to up to 96% in the USA [44]. Although high levels of bacteremia can be found in cattle, persisting for up to seven months [45], most infections seem subclinical although cases of bovine endocarditis have been described [46]. ere are little data on B. bovis in equids although horses seropositive against the organism have been described and experimental infections generally cause no clinical signs with no bacteremia and a low-grade and short-duration antibody response [47]. e vector of B. bovis has not been determined but is suspected to be a blood-sucking arthropod [48]. e specificity of the 16S rRNA-based pan-Bartonella FRET-qPCR and the gltA-based Bartonella PCR we developed was based on the alignment of the available recognized Bartonella spp. sequences in GenBank. For completeness, the newly established PCRs should be tested on known positive control samples of DNA of all the recognized and poorly characterized Bartonella spp., but it is no simple matter for one laboratory to obtain all of these samples. Similarly, with the rapid growth in the recognized and suspected Bartonella species, it will be important for ongoing testing of the PCR to ensure it recognizes the newly recognized strains and species. Ideally, the established Bartonella PCRs in this work should be compared with other existing molecular approaches for their sensitivity and specificity and validated fully following the MIQE guidelines for publication of quantitative real-time PCR experiments [49].
In conclusion, in this report, we describe what we believe to be the first FRET-qPCR that specifically detects all the currently recognized Bartonella species. We also describe a gltA-based PCR that enables the differentiation of the recognized Bartonella spp. e pan-Bartonella FRET-qPCR enabled us to rapidly and quantitatively screen a variety of samples from a number of sources and different geographical areas for the presence of Bartonella spp., while the gltA-based PCR provided a convenient method to differentiate the species involved using only a single reaction.

Data Availability
Whole data including the nucleotide sequence used to support the findings of this study are included within the article.

Conflicts of Interest
e authors declare no conflicts of interest.