Heme Binding Proteins of Bartonella henselae Are Required when Undergoing Oxidative Stress During Cell and Flea Invasion

Bartonella are hemotropic bacteria responsible for emerging zoonoses. These heme auxotroph alphaproteobacteria must import heme for their growth, since they cannot synthesize it. To import exogenous heme, Bartonella genomes encode for a complete heme uptake system enabling transportation of this compound into the cytoplasm and degrading it to release iron. In addition, these bacteria encode for four or five outer membrane heme binding proteins (Hbps). The structural genes of these highly homologous proteins are expressed differently depending on oxygen, temperature and heme concentrations. These proteins were hypothesized as being involved in various cellular processes according to their ability to bind heme and their regulation profile. In this report, we investigated the roles of the four Hbps of Bartonella henselae, responsible for cat scratch disease. We show that Hbps can bind heme in vitro. They are able to enhance the efficiency of heme uptake when co-expressed with a heme transporter in Escherichia coli. Using B. henselae Hbp knockdown mutants, we show that these proteins are involved in defense against the oxidative stress, colonization of human endothelial cell and survival in the flea.


Introduction
B. henselae is now recognized as one of the most common zoonoses acquired from animal companions in industrialized countries [1]. The bacterium causes cat scratch disease as well as a number of other syndromes associated with tumoral proliferation of endothelial cells [2]. Most Bartonellae species appear to share a similar natural cycle that involves arthropod transmission, followed by exploitation of a mammalian host. Each Bartonella species appears to be highly adapted to one or several reservoir hosts in which it causes long-lasting intra-erythrocytic bacteremia as a hallmark of infection. Before colonizing erythrocytes, the bacteria need to replicate and become competent in a primary niche [3] such as endothelial cells, although other nucleated cells might constitute part of the primary niche [4]. Bacterial persistence in erythrocytes is an original strategy, and is considered an adaptation to the mode of transmission by bloodsucking arthropod vectors. The invasion of erythrocytes might also be a strategy for Bartonellae species to obtain heme, an absolute requirement for growth [5]. Not all of the already sequenced Bartonellae genomes contain heme biosynthesis genes [6]. Moreover, these genomes do not encode for siderophore biosynthesis or a complete iron Fe 3+ transport system. Only genes sharing strong homology with all compounds of an Fe 2+ transport system, already characterized in Yersinia pestis [7] and Photorhabdus luminescens [8], are present in Bartonellae genomes. Moreover, Bartonellae genomes encode for a complete heme transport system shown to be active in Bartonella quintana [9].
Analysis of Bartonellae genomes sequenced to date clearly shows the absence of numerous genes proven to be required for E. coli in order to face oxidative stress [10]. Genes coding for polypeptides involved in degradation of hydrogen peroxide, like catalase and peroxidase [11], methyl sufoxide reduction (MsrA and MsrB) [12] and oxidative stress response regulation (SoxR, OxyR) [13], [14] are not contained in Bartonellae genomes. Based on the above information relating to Bartonellae genome contents, it can be hypothesized that Bartonellae are highly sensitive to oxidative stress. However, hydrogen peroxide challenges performed with B. bacilliformis clearly show that this bacterium can efficiently face exposure to 1 mM H 2 O 2 for 30 min [15]. Moreover, the lifestyle of Bartonellae suggests that these bacteria must face oxidative stress after the blood meal of the arthropod vector [16]. Bartonellae genomes are small in size, 1.2 to 2 M bases. In spite of their reduced size, they encode for three to five heme binding proteins (Hbps) [17]. Comparison of B. quintana Hbps (HbpA, HbpB, HbpC and HbpD) shows that three of these proteins are similar in size and peptide sequences. The HbpA, HbpC and HbpD sequences of these approximately 30 kDa polypeptides are nearly 53% identical. HbpB can be distinguished from other Hbps, as this polypeptide is about 11 kDa bigger than the other Hbps and contains a central part not present in other Hbps from B. quintana [17].
To survive and multiply, Bartonellae are forced to encounter various environments in their hosts and vectors. In mammals, free heme is rare [18] and the blood oxygen concentration is low (5% compared to 21% in the atmosphere) [19]. In contrast, in arthropod vectors, toxic heme level is found in the gut and the oxygen concentration is high [16]. The regulation pattern of hbp gene expression was thus investigated in B. quintana [19]. Under all conditions tested, hbpA was more strongly expressed than other hbp genes [19]. Based on their regulatory pattern, hbp genes were divided into two groups. The first contained hbpB and hbpC, over expressed under conditions that mimick the gut arthropod environment (high heme concentration and low temperature, high O 2 concentration). The transcription level of hbpB and hbpC was higher at high heme concentrations (2.5 to 5 mM) [19]. Those authors suggested that HbpB and HbpC play a critical role in the arthropod gut [19]. The transcription of hbpA, hbpD and hbpE was increased at low heme concentrations (0.05 mM) at 37uC. The authors suggested that HbpA, HbpD and HbpE are required when the free heme concentration is low, such as in blood circulation in the mammalian host. However, it was recently shown that transcription of B. henselae hbpA is also significantly increased at 28uC, suggesting that HbpA could protect B. henselae from heme toxicity in the arthropod gut [20]. The function of HbpA was first investigated in B. henselae. Controversial results excluded a direct role for HbpA in the heme uptake process [21], [22]. Various reports demonstrated that HbpA of B. henselae, which shares homology with Opa from Neisseria meningitidis [23], plays a role in the endothelial cell adhesion process [24], [25]. HbpB of B. tribocorum was shown to be required for establishing long-term bacteremia in a rat model, but its role has not been elucidated [26]. HpbC of B. henselae was recently identified as a heme detoxifying protein [20]. Finally, it was suggested that HbpD of B. henselae is required for survival in endothelial cells [27].
In this report, we investigated the activity of four Hbps (HbpA, B, C, D) of B. henselae using both homologous and heterologous (E. coli) models. In E. coli, the ability to bind Congo red and heme was investigated in vivo and in vitro respectively. In B. henselae, hbp knockdowns were checked for their growth ability, oxidative stress response and capacity to invade and survive within endothelial cells. Moreover, these mutants were tested for their ability to develop within the B. henselae arthropod vector Ctenocephalides felis.

Ethics Statement
Animals were handled in strict accordance with good animal practice as defined by the relevant European (European standards of welfare for animals in research), and/or local animal welfare bodies. Animal work performed at the Ecole Nationale vétérinaire d'Alfort (ENVA/ANSES) was reviewed and approved by the institute's ethics committee on September 2011(agreement nu:14/ 06/2011-1). Animal work performed at the Ecole nationale vétérinaire de Toulouse was reviewed and approved by the INRA Toulouse/ENVT ethics committees (agreement nu: MP/01/22/ 06/09) for 3 years.

Bacterial Strains and Plasmids
Bacterial strains and plasmids used in this study are listed in Table 1.

Media and Growth Conditions
Bovine hemoglobin (Hb) and 2, 29dipyridyl (Dip) were obtained from Sigma-Aldrich France (38297 Saint-Quentin Fallavier-France). Heme was dissolved immediately before use in 0.02 M NaOH. Hb was dissolved in 100 mM NaCl. Heme and Hb solutions were filter-sterilized with 0.20 mm Millipore filters for bacterial growth experiments. E. coli strains were grown on LB medium (Sigma), M63 minimal medium, aerobically at 37uC [28] or on Congo red plates. M63 medium was supplemented with 0.4% glycerol (Gly) as carbon source. Solid media and soft agar respectively contained 1.5% or 0.7% Difco agar. Congo red plates consisted of solid BHI medium (Difco) supplemented with Congo red dye (0.02% final concentration). Iron-depleted medium was obtained with the addition of Dip at an 80 mM final concentration. Antibiotics were added to the following final concentrations (mg ml 21 ): ampicillin (Amp), 50; kanamycin (Km), 50; and spectinomycin (Spc), 50. Arabinose (Ara) was added to a final concentration of 0.02%, 0.2% or 0.4% for induction of the P ara promoter. B. henselae was grown on a Columbia blood agar (CBA) plate containing 5% defibrinated sheep blood (Biomérieux; ref 43041; 5, rue des Aqueducs BP 10 -69290 Craponne, France) or in Schneider's medium (Gibco; Route de L'Orme des Merisiers Immeuble Discovery -Zone Technologique 91190 Saint Aubin, France) supplemented with 10% fetal calf serum [29] at 35uC under a 5% CO 2 atmosphere. For flea infection assays, Bartonella strains were collected after 5 days of growth on CBA plates and suspended in PBS buffer. The bacterial suspension was diluted with PBS to obtain a cell density of 1.98610 8 bacteria/ml. The survival of bacteria in PBS buffer was not significantly decreased after 24 h storage at room temperature.

Congo Red Binding Assay
Tested strains containing pBAD derivatives expressing Hbps were grown overnight at 37uC in LB medium containing 50 mg/ ml ampicillin. Two ml of LB medium containing 50 mg/ml ampicillin were inoculated to an OD 600 of 0.05 with overnight culture and grown at 37uC. Expression was induced at an OD 600 of about 0.6 for 2 h by adding arabinose (0.4% final concentration). Cultures were diluted, plated on Congo red plates and incubated at 37uC for 24 h.

E. coli Heme-dependent Growth Assays
Tested strains were grown for 18 h in M63 medium without iron with 0.4% glycerol as carbon source, and in the presence of 0.02% arabinose. Cultures were checked for OD at 600 nm and adjusted to OD 600 = 1. A 100 ml sample of the bacterial suspension was mixed with 4 ml of soft agar. The mixture was poured onto M63 plates containing 0.4% glycerol, 0.02% arabinose and 80 mM Dip (M63D). Wells (5 mm in diameter) were cut in the agar and filled with 100 ml of 50 mM, 10 mM, 5 mM, or 1 mM of filter-sterilized Hb solution. Growth around the wells was recorded after 2-day incubation at 37uC. All experiments were performed in triplicate.

Physiological Characterization of hbp Knockdown Strains
To evaluate the effect of hbp knockdown on growth of B. henselae, tested strains were grown both in liquid Schneider's medium and on CBA plates. B. henselae pNS2Trc and B. henselae pNS2Trc::hbps AS were collected after 5 days of growth on CBA plates and suspended in Schneider's medium or phosphate buffered saline (PBS). For growth in Schneider's medium, the OD 600 of the bacterial suspension was adjusted to 0.05. Five ml samples of this suspension were poured into 6-well plates and grown at 35uC in the presence of 5% CO 2 . OD 600 was checked at days 2, 4, 5 and 7 after inoculation. For growth on CBA plates, serial dilutions of bacterial suspension in PBS were plated on CBA plates and colony size was evaluated after 6 and 10 days of growth at 35uC in the presence of 5% CO 2 . All experiments were performed in triplicate.

H 2 O 2 Challenge
B. henselae pNS2Trc and B. henselae pNS2Trc::hbps AS were grown on CBA plates for five days at 35uC under a 5% CO 2 atmosphere. Bacteria collected from one plate were suspended and washed twice in PBS buffer. The cell suspension was then diluted to OD 600 0.5. Before H 2 O 2 challenge, several dilutions of the tested cell suspension were spread on CBA plates (T 0 ). For the challenge assay, bacteria were incubated 30 min in PBS buffer in the presence of 1 mM and 10 mM H 2 O 2 at 35uC under a 5% CO 2 atmosphere. After exposure to H 2 O 2 , bacteria were washed twice in PBS buffer and several dilutions plated onto CBA plates (T 1 ). After 15-day incubation at 35uC under a 5% CO 2 atmosphere, colonies were counted. Survival rate was expressed by (T 1 /T 0 ) X100%. All experiments were performed in triplicate.

Endothelial Cell Culture and Invasion Assay
Endothelial cell line Ea.hy 926 resulting from a fusion of HUVEC and lung carcinoma cell line A549 (ATCC; reference CRL 2922) were cultured in DMEM medium (Gibco) supplemented with 10% fetal bovine serum decomplemented by heating 30 min at 56uC before use. When required, kanamycin was added at a 50 mg/ml final concentration. Cells were incubated at 37uC in humidified 5% CO 2 and cultured every 7 days using 0.025% trypsin and 1 mM EDTA in Hanks' balanced salt solution [30]. Endothelial cells were seeded in 24well plates at a density of 10 4 cells/well. After 6 days, cell number was estimated as 1.5610 5 /well. Bartonella strains harvested from CBA plates after 5-day growth at 35uC under a 5% CO 2 atmosphere were washed twice in modified DMEM buffer and then resuspended in the same medium. Bacterial number was adjusted to 3610 4 /ml (1OD was estimated at 6.6610 8 /ml). For cell invasion assays, medium in the well was removed and 0.5 ml of modified DMEM containing 1.5610 4 colony-forming units (CFUs) of the B. henselae tested strains were added to the well to obtain 0.1 multiplicity of infection (m.o.i). A low m.o.i. was used since we considered a cell excess as more suitable to visualize the full invading ability of bacteria. After remaining for 1 h at 37uC, 1.5 ml of modified DMEM medium was supplemented in the well. The number of bacteria was controlled by plating several dilutions on CBA plates, and CFUs were determined after 15 days of growth (T 0 ). Mixtures were incubated at 37uC at 5% CO 2 for 24 h. After 24 h, the number of viable bacteria was determined by plating serial dilutions of mixtures on CBA plates, and CFUs were determined after 15 days growth (T 24 ). After 24 h, bacterial viability was nearly 100%. The intracellular bacterial population was quantified by the gentamicin protection assay as described by Mehock [31]. Briefly, DMEM with gentamicin (final concentration 250 mg/ml) was added to the mixture assay and removed after 2 h at 37uC in the presence of 5% CO 2 . A control B. henselae bacterial suspension showed no survival after 2 h exposure to gentamicin (250 mg/ml). After removing modified DMEM medium containing gentamicin, endothelial cells were then washed three times with modified DMEM medium to remove residual antibiotic. Endothelial cells were then collected after 4 min incubation with 400 ml trypsin at 37uC. After centrifuging at 12,000 rpm, cells were suspended in 1 ml of sterile water and disrupted using a 1 ml syringe and a 0.4 mm620 mm needle and 5 pushes [32]. Microscopic controls revealed that, after 5 pushes, all cells were lysed. Cell lysates were supplemented with 100 ml 106PBS to overcome osmotic lysis. The number of viable bacteria was determined by plating lysates on CBA plates. After 15 days of incubation at 35uC under a 5% CO 2 atmosphere, colonies were counted (T I ). The invasion rate was expressed as (T I /T 0 )x100%. Each assay was performed in double wells and all experiments were performed in triplicate.

Survival Assay in Endothelial Cells
To perform survival assays of B. henselae pNS2Trc and B. henselae pNS2Trc::hbps AS in endothelial cells, the monolayer contained in mixtures challenged for gentamicin killing was washed three times and incubated in modified DMEM medium at 37uC and 5% CO 2 for 24 and 48 h. Cells were treated as described above and the number of viable bacteria was determined by plating the lysates on CBA plates. After 15-day incubation at 35uC under a 5% CO 2 atmosphere, colonies were counted (T S24 or T S48 ). The survival rate was expressed as (T S24 /T I or T S48 /T I )x100%. Each assay was performed in double wells and all experiments were performed in triplicate.

Flea Maintenance and Supply
Strain Ctenocephalides felis (C. felis) (Siphonaptera: Pulicidae) originating from a wild strain harvested from a cat has been maintained on cat under laboratory conditions since 1990. A new generation is obtained every 2 to 4 weeks. Fleas were controlled to be PCR negative for B. henselae using primers bh2390fo, and bh2390re ( Table 2).
Feeding of C. felis with B. henselae pNS2Trc-or B. henselae pNS2Trc::hbps AS -infected Blood Dog blood used in all experiments was obtained from 3 beagles (15 ml obtained from each dog) from the Ectoparasite Laboratory of the National Veterinary School in Toulouse, France. The absence of Bartonella spp. in the blood of these dogs was confirmed by PCR using primers bh2390fo, and bh2390re ( Table 2). Lithium heparin-coated vacutainer tubes (Venosafe, Terumo Europe) were used to draw blood by venipuncture. Blood functional complement was deactivated by maintaining blood samples at room temperature for 2 h after the blood test and before storing them at 4uC. Blood samples were stored less than 48 h at 4uC. When required, kanamycin was added to blood at a 50 mg/ml final concentration. Kanamycin was previously determined to have no effect on C. felis feeding, viability or egg production. A total of 500 unfed fleas (males and females aged between 8-10 days) were placed in a plexiglas box in contact with a glass feeder closed at the bottom by a parafilm membrane. To stimulate flea blood-feeding, a constant temperature (38.5uC) was maintained by a water-jacket circulation system through the glass feeder. For blood infection, 500 mL of bacterial suspension at a concentration of approximately 1.98610 8 bacteria/ml in PBS were added to 5 ml of blood. Viability of Bartonella in blood was about 100% after 2-h incubation. Blood was complemented by bacterial suspension for the first two days of feeding. Then, fleas were fed with uninfected dog blood for the next 8 days. Every 24 h, the glass feeder was cleaned, a new parafilm membrane was stretched and blood was changed. At the same time, flea feces were collected. All samples were stored at 220uC until PCR analysis. Ethanol 70% was added to flea feces samples.

Genetic Techniques
E. coli cells were transformed by the calcium chloride method [33]. Bartonella cells were transformed by electroporation as previously described [34].

DNA Manipulations
A small-scale plasmid DNA preparation was performed using a QIAprep Spin Miniprep kit (Qiagen; QIAGEN France 3 av du Canada LP 809 91974 Courtaboeuf). Restriction, modification, and ligation were carried out according to the manufacturer's recommendations. DNA fragments were amplified in a Hybaid PCR thermocycler using Phusion DNA polymerase (Finnzymes). Nucleotide sequencing was performed by Eurofins MWG Operon. Purification of DNA fragments from the PCR reaction, restriction reaction or agarose gels was performed using the Macherey-Nagel NucleoSpinH Extract II kit.

Construction of a Recombinant Vector Expressing Hbps of B. henselae
Complete B. henselae hbpA, hbpB, hbpC and hbpD genes with a Cterminal Histag (6 His) were amplified by PCR from B. henselae chromosomal DNA using primers HbpABhamont and HbpABhaval, HbpBBhamont and HbpBBhaval, HbpCBhamont and HbpCBhaval or HbpDBhamont and HbpDBhaval, respectively ( Table 2). The fragments amplified (890 bp for hbpA, 881 bp for hbpC and 875 bp for hbpD) were purified and digested with NheI and KpnI. The hbpB fragment (1235 bp) was purified and digested with NheI and Hind III. Then, fragments were ligated with pBAD24 plasmid digested with NheI and KpnI or NheI and Hind III. Ligation mixtures were introduced into CaCl 2 -competent E. coli strain XL1 Blue cells and transformants were selected on LB plates containing ampicillin. Clones were screened by the PCR method with corresponding primers. Six PCR-positive clones were then sequenced.

Construction of a Knockdown Vector for Decreasing the Amount of Hbp in B. henselae
The entire coding region of hbpA, hbpB, hbpC and hbpD was amplified by PCR from the B. henselae chromosomal DNA using primers hbpAantisensamt and hbpAantisensavl, hbpBantisensamt and hbpBantisensavl, hbpCantisensamt and hbpCantisensavl or hbpDantisensamt and hbpDantisensavl, respectively ( Table 2). The PCR product (846 bp for hbpA, 1,182 bp for hbpB, 837 bp for hbpC and 831 bp for hbpD) was purified, digested with BamHI and XbaI and then ligated with plasmid pNS2Trc digested with BamHI and XbaI. Ligation mixtures were introduced into CaCl 2competent E. coli strain XL1 Blue cells. Transformants were screened by PCR with corresponding primers. Six PCR-positive clones were then sequenced. Detection of B. henselae DNA from Flea Feces DNA of B. henselae was detected by amplification of a 1,052 bp fragment containing B. henselae bh02390 gene https://www. genoscope.cns.fr/agc/microscope/home/index.php using primers bh2390fo, and bh2390re ( Table 2). DNA of C. felis was detected by amplification of a fragment of C. felis 18S rDNA using primers Cf18Sf and Cf18Sr (Table 2) [35]. Amplifications were performed with at least 20 ng of DNA extract for flea feces. Each reaction was conducted in a total volume of 20 mL with 0.5 mM of each primer, 200 mM of each dNTP, 4 mL of 56 PCR buffer G/C and 0.02U/ ml of Taq DNA polymerase (PhusionH High-Fidelity DNA Polymerase, Thermo scientific). The PCR program was as follows: an initial denaturation step for 30 s at 98uC, followed by 40 cycles of denaturation for 10 s at 98uC, annealing for 30 s at 55uC and extension for 1 min at 72uC, and a final extension step at 72uC for 7 min.

Expression and Purification of Recombinant His-tagged Hbps
Strain XL1-Blue pBAD24::hbpA, XL1-Blue pBAD24::hbpB, XL1-Blue pBAD24::hbpC and XL1-Blue pBAD24::hbpD were grown overnight at 37uC in LB medium containing 50 mg/ml ampicillin. Then, 200 ml of LB medium containing 50 mg/ml ampicillin was were inoculated to an OD 600 of 0.05 with the overnight culture and grown at 37uC. Expression was induced at an OD 600 of about 0.6 for 2 h by adding arabinose (0.4% final concentration). Bacteria were harvested by centrifugation for

Heme Binding Assay in vitro
For heme blotting, proteins were separated on standard 12.5% SDS gels followed by electrophoretic transfer to nitrocellulose membranes according to the protocol of Vargas [36]. Briefly, samples were mixed with loading buffer to which no DTT was added and samples were not boiled before electrophoresis. 1 mg of HbpA, HbpB, HbpC and HbpD or 3 mg of BSA was separated on 12.5% SDS-PAGE. One gel was stained with Coomassie brilliant Blue R. Another gel was transferred to nitrocellulose using the general methods of Towbin et al. [37]. The heme binding blot was done according to the protocol of Carroll et al. [22]. Briefly, the resulting blots were rinsed with Tris-buffered saline containing 0.1% Tween 20 (TBST; 10 mM pH 8.0 Tris-HCl containing 150 mM NaCl and 0.1% Tween 20) three times for 30 min and subsequently probed for 1.5 h with TBS containing heme (10 26 M) at room temperature. Nitrocellulose was washed three times for 30 min with TBS-Tween 20 (0.1%) at room temperature. Heme was visualized via it intrinsic peroxidase activity [38] using enhanced chemiluminescence (ECL) reagents (Amersham Pharmacia, Piscataway, N.J). Hbp bands were visualized by exposing the blot to autoradiographic film (Labscientific, Livingston, N.J.).

Antibody Preparation
200 ml of an emulsion containing purified HbpB (10 mg) and ISA 61 VG adjuvant (Seppic) (120 ml), completed with NaCl 0.9%, were inoculated twice (at a 1-month interval) subcutaneously into 4 weeks old C57B6 mice (Charles River) (Ethics Committee Anses/ENVA/UPEC agreement nu:14/06/2011-1). Two weeks after the second inoculation, 200 ml blood samples were collected every 3 weeks via retro-orbital bleeding. Blood samples were centrifuged twice (3,600 rpm 5 min) to obtain serum which was stored at 220uC. Before use, non-specific antibodies were removed by incubating the immune serum with E. coli cell extract for 1 h at 4uC and centrifugation for 10 min at 8,000 rpm. The supernatant was then used as serum.

Protein Analysis by Electrophoresis
Proteins were analyzed by 12% sodium dodecyl sulfate polyacrylamide gel (SDS-PAGE) electrophoresis [39], followed by Coomassie Blue staining.

Immunoblot Analysis
Sodium dodecyl sulfate/polyacrylamide gel electrophoresis (SDS-PAGE) and immunoblotting for detecting decreased expression of HbpA in B. henselae were performed as follows. Tested strain biomass was collected after 5 days of growth on CBA plates, suspended in PBS buffer and centrifuged. Bacterial pellets calculated to contain 40 mg of proteins were suspended in loading buffer and heated for 5 min at 100uC. Proteins were separated by 12% SDS-PAGE and transferred to a nitrocellulose membrane (Hybond-C Extra, GE Healthcare; VWR International S.A.S Le Périgares -Bât B 201, Rue Carnot 94126 Fontenay-sous-Bois cedex) according to Towbin et al [37]. Non-specific binding sites were blocked with 5% skim milk in TBS-Tween 20(0.05%). The immunoblot was probed with polyclonal mouse sera raised against recombinant HbpB (1:200), followed by a 1:1,000 dilution of a rabbit anti-mouse IgG horseradish peroxidase[HRP]-conjugated secondary antibody (cell signaling; Ozyme, 10 avenue Ampère, BP 128, 78053 St Quentin en Yvelines France). Visualization of the HRP signal was performed using enhanced chemiluminescence (ECL) reagents (Amersham Pharmacia, Piscataway, N.J). HbpA bands were visualized by exposing the blot to autoradiographic film (Labscientific, Livingston, NJ, USA).

Protein Assay
The concentration of the protein was determined using the BC assay protein quantitation kit (Interchim).

Statistical Analysis
Statistical analysis was performed using GraphPad Prism 5 software for Windows. Statistical significance of the data was ascertained by use of Student's t test. A value of P,0.05 was considered significant.

Hbps are Able to Bind Congo Red when Expressed in E. coli
To determine that Hbps of B. henselae can bind heme, we first checked for Congo red binding activity, since it had been shown that Hbps can also bind Congo red [40]. Moreover, it had already been shown that expressing a Congo red binding protein on the outer membrane of E. coli conferred a Congo red binding phenotype [41]. We transformed E. coli XL1-Blue with plasmids pBAD24, pBAD24::hbpA, pBAD24::hbpB, pBAD24::hbpC and pBAD24::hbpD to determine whether hbp genes can produce a Congo red binding phenotype. Strain XL1-Blue with plasmid pBAD24 did not confer a Congo red phenotype when grown on the plate with 0.2% arabinose (Fig. lB). However, strains expressing HbpA, B, C and D formed a red colony on the plate with 0.2% arabinose (Fig. 1 C, D, E, F). Without arabinose, neither XL1-Blue pBAD24 nor XL1-Blue pBAD24::hbpA, XL1-Blue pBAD24::hbpB, XL1-Blue pBAD24::hbpC or XL1-Blue pBAD24::hbpD, conferred a Congo red binding phenotype (data not shown). Taken together, these data showed that Hbps can bind Congo red and confirmed that Hbps were exposed on the cell surface when expressed in E. coli.

Recombinant HbpA, HbpB, HbpC and HbpD can Specifically Bind Heme in vitro
To produce and purify recombinant HbpA, B, C, D from E. coli, we amplified their structural gene using B. henselae chromosomal DNA as template and primers allowing addition of a six His-tag at the C-terminal of the protein. The fragments were cloned into the pBAD24 plasmid as described in Materials and methods. Plasmids pBAD24::hbpA, pBAD24::hbpB, pBAD24::hbpC and pBAD24::hbpD were introduced into strain XL1-Blue. To check for quantities of Hbps in E. coli strain XL1-Blue, SDS gel electrophoresis (SDS-PAGE) was used to compare protein extracts of strain XL1-Blue pBAD24::hbps his and XL1-Blue pBAD24. A supplementary visible band was observed on SDS page gel for XL1-Blue derivatives harboring plasmid pBAD24::hbpA, pBAD24::hbpB, pBAD24::hbpC or pBAD24::hbpD when grown in the presence of 0.4% arabinose (Fig. 1A). Recombinant Hbp proteins were purified by Ni-agarose affinity chromatography as described in Materials and methods. The size of purified Hbp proteins corresponded well to those predicted from their sequences.
To test whether pure Hbps can specifically bind heme in vitro, a standard method already used for detecting heme binding of cytochrome C was used [36]. Pure HbpA, HbpB, HbpC, HbpD and BSA were separated on two SDS-polyacrylamide gels. One gel was then stained with Coomassie brilliant Blue R ( Fig. 2A). Another gel was transferred to a nitrocellulose filter to perform heme blotting and ECL detection (Fig. 2B). Pure Hbps were able to bind heme added at 10 26 M concentration (Fig. 2B). In contrast, under our assay conditions, BSA was unable to bind heme (Fig. 2B). The latter result underlines the specificity of the proteins that bind heme. Taken as a whole, we conclude that when purified, all Hbps can bind heme in vitro.

Hbps Increase the Efficiency of Heme Uptake Mediated by HemR from Serratia marcescens in E. coli
Conflicting results concerning the heme transport activity of HbpA encouraged us to first check for the ability of Hbps to transport heme using E. coli hemA mutant complementation. Our results demonstrated that Hbps from B. henselae did not act as heme transporters (Table 3). These results are at variance with those obtained for Pap31 (HbpA) from B. henselae [21], but are in agreement with those obtained for HbpA of B. quintana [9] and Bartonella birtlesii (Biville F, unpublished data). Heme binding proteins do not act as heme porins, but their ability to bind heme can increase heme concentration on the bacterial surface. This local increase of heme concentration can facilitate its uptake by the bacteria. We checked whether Hbps could modulate the efficiency of the heme uptake process. We first verified the effect of Hbps on the activity of HutA of B. henselae expressed in E. coli. HutA from B. quintana was shown to transport heme and, consequently, to restore growth of an E. coli hemA mutant. [9]. Experiments were performed in the presence of high heme concentrations in liquid medium. The weak growth restoration may have been the consequence of mutations enhancing outer membrane permeability [42]. These complementation assays failed for an E. coli K-12 hemA mutant expressing HutA of B. birtlessii when grown on heme-supplemented solid medium [43]. To avoid factual results concerning the heme transportation activity of HutA from B. henselae, we expressed it in an E. coli entF mutant that cannot grow in the presence of an iron chelator [44]. When a heme transporter is expressed in such strain, addition of heme to the medium restores growth. In order to be used as an iron source, the amount of heme required is 100 times greater than that required for its use as a heme source. Such a complementation assay had already been used to check the heme transport activity of HasR from Serratia marcescens [45] and also for characterization of heme-degrading enzymes [46], [47]. No growth was observed in the E. coli entF strain expressing HutA from B. henselae when grown in irondepleted medium in the presence of hemoglobin (data not shown).
To check for the effect of Hbps from B. henselae upon the heme uptake process, plasmids harboring hbps genes were introduced into the E. coli entF mutant [44] expressing a HemR heme transporter from S. marcescens [48]. Strains obtained were tested for growth on iron-depleted medium in the presence of hemoglobin added at different concentrations. As seen in table 3, HemR alone led to growth around the well containing hemoglobin at 50 mM in the presence of 80 mM dip. Strains expressing hbpA, hbpB, hbpC and hbpD were able to grow at lower concentrations of Hb ( Table 3), suggesting that HbpA, HbpB, HbpC and HbpD increase the efficiency of the HemR-mediated heme uptake process. HbpB and HbpD were more efficient than HbpA and HbpC (Table 3). To further investigate the efficiency of heme uptake mediated by Hbps and HemR, we grew the bacteria on a minimal iron-depleted plate with 100 mM Dip in the presence of different concentrations of Hb. For the control strain only expressing HemR and strains co-expressing HbpA or HbpC and HemR, no growth was observed around the well whatever the concentration of Hb added (data not shown). The strain coexpressing HbpD and HemR showed growth only around the well containing 50 mM Hb. The strain co-expressing HbpB and HemR was able to grow around the well whatever the concentration of Hb (data not shown). Taken together, we concluded that HbpB and HbpD were more efficient than others at increasing heme uptake when expressed in E. coli, and thus they might play an important role at low heme concentrations. Similar results were obtained by introducing plasmids harboring hbps genes in an E. coli entF mutant expressing HasR from S. marcescens (data not shown). Binding of heme by Hbps increased its concentration around the bacteria and consequently facilitated its uptake.

Hbp Activity is Important for B. henselae Growth
Since Hbps can modulate heme uptake efficiency when expressed in E. coli, a similar activity was hypothesized in B. henselae. In that bacterium, heme serves both as heme and as an iron source [47]. Thus, abolishing or decreasing synthesis of the different Hbps in B. henselae could potentially affect its growth capacity. Knockout of HbpA in B. henselae was hypothesized as being lethal to the bacteria [20]. A preliminary unsuccessful assay in our lab to disrupt hbpA of B. henselae was in agreement with this hypothesis. To investigate the function of the four Hbps of B. henselae using the same genetic tool, we chose the knockdown method that had been successfully used for B. henselae [47], [49], [50]. We cloned hbps of B. henselae oriented in the reverse direction such that the anti-sense strand was transcribed in plasmid pNSTrc [49]. Plasmids pNS2Trc, pNS2Trc::hbpA AS, pNS2Trc::hbpB AS pNS2Trc::hbpC AS and pNS2Trc::hbpD AS were introduced into B. henselae using electroporation. We first checked for the knockdown effect on the Hbp expression level in B. henselae using a western blot experiment. Multiple sequence alignment of Hbps of B. henselae revealed a high degree of amino acid sequence conservation at its N-terminal and C-terminal parts even for HbpB, which is longer than other Hbps. According to its size, HbpB can easily be distinguished from other Hbps. Mouse anti-serum against HbpB was prepared and tested for recognition of purified Hbps. Our results demonstrated that anti-HbpB antibody can recognize wellpurified HbpB and HbpA. For HbpD and HbpC, recognition by anti-HbpB antibodies was lower or absent (data not shown). Detection of HbpB expression in B. henselae was investigated using anti-HbpB anti-serum. We failed to detect HbpB in B. henselae using the western blot method. This result is in good agreement with those of proteomic analysis of outer membrane fractions of B. henselae and B. quintana [51], [52], [53], [54], [55], [56], which showed that HbpB was not detectable using this method. Since HbpA is the most abundant Hbp in B. henselae [52], [51], we checked its level in B. henselae pNS2Trc and B. henselae pNS2Trc::hbpA AS using mouse antibody directed against HbpB. As seen in Fig. 3, the level of HbpA was lower in strain B. henselae pNS2Trc::hbpA AS than in strain B. henselae pNS2Trc.
The strains obtained were tested for growth on both blood agar plates and in Schneider's medium as described in Materials and methods. After 6 and 10 days of growth on blood agar plates, colonies produced by strain B. henselae harboring pNS2Trc::hbpA AS , pNS2Trc::hbpB AS , pNS2Trc::hbpC AS or pNS2Trc::hbpD AS were much smaller than those formed by B. henselae pNS2Trc ( Table 4). The B. henselae strain containing pNS2Trc::hbps AS also grew more slowly than B. henselae pNS2Trc in Schneider's medium (Fig. 4). These results indicated that a decrease in the amount of Hbps slowed growth of B. henselae.

Knockdown of Hbps Increased B. henselae Sensitivity to Hydrogen Peroxide
Analyses of Bartonellae genomes demonstrated that numerous genes involved in the oxidative stress response were not present. However, it was shown that Bartonella bacilliformis was able undergo oxidative stress generated by exposure to H 2 O 2 [15]. It was suggested that one potential role for Hbps was to bind heme at the surface of the bacteria and provide an antioxidant barrier via heme intrinsic peroxidase activity [57]. To examine this hypothesis, we tested the effect of hbp knockdown on the ability of B. henselae to face 30 min exposure to 1 mM and 10 mM hydrogen peroxide. After exposure to 1 mM H 2 O 2, survival was about 50% for both control strain B. henselae pNS2Trc and B. henselae pNS2Trc::hbpC AS. There existed slightly decreased survival capacity for B. henselae pNS2Trc::hbpA AS , B. henselae pNS2Trc::hbpB AS and B. henselae pNS2Trc::hbpD AS (data not shown). After exposure to 10 mM H 2 O 2 , survival was about 25% for control strain B. henselae pNS2Trc. With strain B. henselae pNS2Trc::hbpA AS , B. henselae pNS2Trc::hbpB AS and B. henselae pNS2Trc::hbpC AS, sensitivity to hydrogen peroxide increased about 3-4-fold (Fig. 5). Decreasing HbpD levels more sharply increased B. henselae sensitivity to hydrogen peroxide (Fig. 5). These results indicated that lowering the Hbp level in B. henselae significantly increased its sensitivity to H 2 O 2 .

Effect of Hbp Knockdown on B. henselae Capacity to Invade Endothelial Cells
Within cells, bacterial infection was shown to induce ROS production [58]. Thus, decreasing the ability to undergo oxidative stress is expected to decrease the ability of B. henselae to survive in endothelial cells. This promoted us to check the effect of Hbp knockdown upon the capacity of B. henselae to invade human endothelial cells and to survive within them.
For survival in the endothelial cell assay, mixtures were grown for 24 h or 48 h after the gentamicin killing assay. After 24 h in endothelial cell, cell lysates were spread on the blood plate to check viable bacterial number. Surprisingly, no bacteria were visible on the blood plate after 2-week incubation. To overcome this problem, we first grew the mixture in Schneider's liquid medium overnight, sustaining primary isolation of B. henselae [29] before plating it on blood plates. Colony count after 2-week incubation showed a decrease in viable bacteria for all strains tested, though it has been claimed that B. henselae is able to replicate inside endothelial cell through bacterial rRNA replication [59]. Survival rates for B. henselae pNS2Trc::hbps AS were much lower than those of control B. henselae pNS2Trc (Fig. 7), after 24 h or 48 h growth in endothelial cells. To check for an effect of overnight growth in Schneider's medium, about 600 B. henselae pNS2Trc or B. henselae pNS2Trc::hbps AS bacteria were grown overnight in that medium and plated on blood plates for enumeration. The increase in bacterial numbers was calculated for all tested strains. The increase was about 20-30% for both B. henselae pNS2Trc and B. henselae pNS2Trc::hbps AS . The differing survival rates of B. henselae pNS2Trc and B. henselae pNS2Trc::hbps AS were not related to a growth defect in Schneider's medium. We conclude that Hbps of B. henselae play an important role in survival within endothelial cells.

Hbps are Involved in Multiplication of B. henselae in C. felis
It was demonstrated that B. henselae-infected cat fleas (C. felis) can transmit B. henselae to cats [60] and that B. henselae can replicate in the gut of the cat flea [61]. Inside the arthropod gut, bacteria confront oxidative stress after each blood meal [16]. Since Hbps of B. henselae play a protective role against H 2 O 2 -produced oxidative stress, this prompted us to check for an Hbp knockdown effect on B. henselae multiplication in fleas. After feeding fleas for 2 days with blood containing the bacteria, fleas were fed with blood without bacteria for another 8 days. For control strain B. henselae pNS2Trc, we were able to detect B. henselae DNA in the feces from day 1 to day 10 (Table 5). For strains with a decreased amount of HbpA, HbpB, HbpC or HbpD, no bacterial DNA could be detected after day 6. To exclude the possibility that this result was due to small amounts or to the quality of DNA from flea feces, we amplified flea 18S rDNA from day-7-to-10 flea feces samples using primers Cf18Sf and Cf18Sr. Flea 18S rDNA was detected in all of the day-7-to-10 flea feces samples (data not shown). This suggests that Hbps play an important role in multiplication of B. henselae in fleas.

Discussion
In this report, we investigated the role of B. henselae Hbps in heme utilization, the oxidative stress response, cell colonization and survival within arthropod vector C. felis. Previous data had  shown that recombinant HbpA of B. quintana was able to bind heme in vitro, but did not confer a heme binding phenotype in vivo when expressed in E. coli [22]. Later it was claimed that HbpB of B. quintana did not bind heme [19]. Recently, it was shown that hbpC of B. henselae, when expressed in E. coli, confers a heme binding phenotype in vivo [20].
Our results clearly show that expression of all Hbps from B. henselae in E. coli confers a Congo red binding phenotype (Fig. 1), thus suggesting that Hbps have a surface location when expressed in E. coli. We also demonstrate that, in vitro, all purified Hbps specifically bind heme. To characterize the physiological importance of Hbps, we first investigated their effect on the heme uptake process. Pap31 (HbpA) of B. henselae was claimed to act as a heme porin when expressed in E. coli [21], but conflicting data about this activity were published for HbpA of B. quintana [57]. Moreover, we failed to visualize any heme porin activity for Hbps of B. henselae when expressed in E. coli (data not shown). Based on the above results, we hypothesized that the Hbp family of Bartonella could act as a heme reservoir, thus rendering it available under hemelimited conditions. This hypothesis is in good agreement with the absence of genes encoding for heme and iron storage proteins in Bartonellae genomes. Such heme storage activity might enhance the efficiency of the heme uptake process. We examined the effect of Hbps on the activity of heterologous heme transporters HemR and HasR of S. marcescens [48], [62]. All Hbps increased heme uptake efficiency mediated by HemR and HasR. Efficiency at low heme concentrations was better with HbpB and HbpD than with HbpA and HbpC. Such differences in efficacy could be related to the differing levels of Hbps in E. coli. This is the case for HbpB, but not for HbpD (Fig. 1). This result suggests that HbpD might be active when the heme concentration is low. This conclusion is in good agreement with the increase in hbpD expression when B. quintana was grown in the presence of low heme concentrations [19]. How heme is transferred from heme binding proteins to HemR and HasR heme transporters remains unknown.
During mammal and flea invasion, Bartonellae must face microenvironmental shifts, stress and the host immune defense. For example, it was shown that reactive oxygen species (ROS: and OH 2 ) production is an important immune defense mechanism for mammal hosts and arthropod vectors against pathogenic bacteria [16], [63]. Recently, it was shown that ROS (H 2 O 2 ) levels in midgut were higher (over 10 mM) in Y. pestis infected fleas. Antioxidant treatment prior to infection decreased ROS levels and resulted in higher Yersinia pestis loads [64]. An OxyR Y. pestis mutant showed reduced growth in fleas early after infection [64]. ROS are potentially toxic for both the host cell and pathogenic bacteria [16], [65], [66], [67]. Host cells are protected from oxidative damage by enzymes that detoxify ROS, such as SOD, catalase, glutathione peroxidase (Gpx) and thioredoxin peroxidase, that detoxify H 2 O 2 [68]. In arthropod vectors, the adaptive response to ROS has also been thoroughly investigated [16]. B. henselae replicates in the gut of the cat flea and is able to survive several days in flea feces [61]. It was shown that the hematophagous vector has a substantial need for huge amounts of blood at each meal; digestion of hemoglobin within the gut of the vector releases large quantities of heme, which has potential prooxidant and cytotoxic effects if not bound to proteins [69], [70]. To successfully replicate in the cat flea gut, B. henselae must confront toxic ROS and repair damage. Many homologues of genes involved in the oxidative stress response in E. coli are not present in the B. henselae genome. The absence of homologues of these genes suggests that Bartonellae must possess uncharacterized  About 500 fleas were first feed with blood containing 500 ml bacteria (2610 8 /ml) for 2 days and then fed uninfected blood for 8 days. Flea feces were collected every day. DNA was extracted from flea feces and PCR was performed as described in Materials and methods. doi:10.1371/journal.pone.0048408.t005 mechanisms in response to oxidative stress. One actor in this oxidative stress response was recently identified as being a hemedegrading enzyme enabling release of iron from heme [47]. In this report, we demonstrate that all Hbps are required to efficiently undergo exposure to hydrogen peroxide. A role for the intrinsic peroxidase activity of heme bound to Hbps was hypothesized for B. quintana [19] and was shown to be the case for Porphyromonas gingivalis [71]. The competence of Hbps in detoxifying H 2 O 2 might also be required for efficient invasion and survival in endothelial cells, major target cell types for bacterial colonization in the reservoir host(s) as well as in the infected host [72]. Bacterial infection was shown to induce ROS production in endothelial cells [58]. As a consequence, the weaker survival capacity in endothelial cells, related to knockdown of all Hbps, might be related to lower resistance to oxidative stress. For HbpA and HbpD, we also show a weakening in the endothelial cell invasion process when their level decreases. For HbpA, this effect might be related to its Opa domain, which has been shown to be involved in entry into the host epithelial cell in Neisseria gonorrhoeae [73]. Indeed, previous data also showed that Pap31 (HbpA) of B. henselae was able to bind endothelial cells in a dose-dependent manner, and binding was inhibited by anti-Pap31 antibodies [24]. However, HbpD involvement in the endothelial cell invasion process is striking, and can be explained by greater sensitivity to oxidative stress produced by endothelial cells, thus leading to more rapid killing of bacteria. This hypothesis is supported by the fact that the mutant of HbpD cannot delay lysosomal fusion, leading to significantly reduced viability within endothelial cells [27].
Clearance of bacteria in the flea feces can be explained by a decrease in survival within the arthropod. Our results clearly demonstrate more rapid clearance of B. henselae in the flea feces when invasion assays are performed using Hbp knockdown mutants. In addition, this more rapid clearance of bacteria in the flea feces can be attributed to decreased ability to confront oxidative stress. Thus, the ROS-detoxifying activity of Hbps also plays an important role during colonization of fleas. Taken together, we reveal the functions of Hbps in heme utilization, the oxidative stress response, cell colonization and flea transmission. The ability of Hbps to bind heme provides competence in destroying ROS, thus constituting an important immune defense system for host cell and arthropod vectors. Functional identification of Hbp families in cell interactions and flea transmission should help to develop strategies for fighting infection and transmission. It will be interesting in the future to elucidate antioxidant mechanisms used by other vector-borne pathogens.