Differential Effects of Bartonella henselae on Human and Feline Macro- and Micro-Vascular Endothelial Cells

Bartonella henselae, a zoonotic agent, induces tumors of endothelial cells (ECs), namely bacillary angiomatosis and peliosis in immunosuppressed humans but not in cats. In vitro studies on ECs represent to date the only way to explore the interactions between Bartonella henselae and vascular endothelium. However, no comparative study of the interactions between Bartonella henselae and human (incidental host) ECs vs feline (reservoir host) ECs has been carried out because of the absence of any available feline endothelial cell lines. To this purpose, we have developed nine feline EC lines which allowed comparing the effects of Bartonella strains on human and feline micro-vascular ECs representative of the infection development sites such as skin, versus macro-vascular ECs, such as umbilical vein. Our model revealed intrinsic differences between human (Human Skin Microvascular ECs –HSkMEC and Human Umbilical Vein ECs – iHUVEC) and feline ECs susceptibility to Bartonella henselae infection. While no effect was observed on the feline ECs upon Bartonella henselae infection, the human ones displayed accelerated angiogenesis and wound healing. Noticeable differences were demonstrated between human micro- and macro-vasculature derived ECs both in terms of pseudo-tube formation and healing. Interestingly, Bartonella henselae effects on human ECs were also elicited by soluble factors. Neither Bartonella henselae-infected Human Skin Microvascular ECs clinically involved in bacillary angiomatosis, nor feline ECs increased cAMP production, as opposed to HUVEC. Bartonella henselae could stimulate the activation of Vascular Endothelial Growth Factor Receptor-2 (VEGFR-2) in homologous cellular systems and trigger VEGF production by HSkMECs only, but not iHUVEC or any feline ECs tested. These results may explain the decreased pathogenic potential of Bartonella henselae infection for cats as compared to humans and strongly suggest that an autocrine secretion of VEGF by human skin endothelial cells might induce their growth and ultimately lead to bacillary angiomatosis formation.


Introduction
Since its discovery in 1992 [1], Bartonella henselae (B. henselae) is considered as the main agent of cat scratch disease (CSD) in humans and is also responsible for a growing number of diseases in patients [2]. This facultative intracellular bacterium is now known to be responsible for more serious diseases in both immunocompetent patients,e.g. endocarditis, and immunosuppressed patients, such as bacillary angiomatosis and peliosis [3], characterized by pseudotumoral proliferation of endothelial cells (ECs). These unusual vascular lesions occur mainly or exclusively in the skin, liver and spleen [3]. Cats are the main reservoir of this zoonotic bacterium [4]. However, as compared to humans, normal or immunosuppressed cats display high rates of sub-clinical infections and remain usually healthy, with only chronic bacteraemia [4,5,6]. In addition, in cats, B. henselae infection has not yet been associated with bacillary angiomatosis or peliosis [7,8].
Two genotypes (I and II) of B. henselae have been described on the basis of 16S rRNA sequence analysis [9]. Epidemiological studies strongly suggest that genotype I is more virulent in humans than genotype II [9,10,11,12,13]. In particular, only genotype I has been found associated to date to bacillary angiomatosis and peliosis [14], but this observation has to be confirmed by further studies.
The presence of B. henselae micro-colonies adjacent to proliferating endothelial cells has been histologically demonstrated, and suggested that Bartonella/ECs interactions might trigger a proangiogenic process, potentially leading to vascular lesions [15]. Due to the lack of any appropriate animal model, cultured ECs provide tools to study the interactions between B. henselae and the vascular endothelium.
These approaches have been useful for identifying B. henselae virulence factors. B. henselae adhesin A (BadA) (originally described as ''pilus'') [16] is important for pathogenicity [17], being involved in the adhesion to extracellular-matrix proteins and to ECs. It activates hypoxia-inducible factor-1 and pro-angiogenic cytokines secretion [18]. Recently, the VirB/VirD4 type IV secretion system and subsets of its translocated B. henselae effector proteins (BepA and BepG) were found to modulate the angiogenic activity of B. henselae [19,20]. Other studies have suggested that the process through which B. henselae triggers ECs proliferation involved released or secreted bacterial factors [21,22,23,24].
Host factors have also been found in vitro to play a role in B. henselae driven angiogenesis. VEGF (Vascular Endothelial Growth Factor) is known as the main angiogenic factor, which positively regulates migration, proliferation and survival of endothelial cells and has been shown to be over-secreted in the tumor micro-environment [25]. According to McCord et al [26], ECs infected by B. henselae Houston I may upregulate expression and production of pro-angiogenic proteins. Studies of VEGF expression in clinical samples [27] or in vitro [22,27,28], suggest a paracrine loop type of VEGF activity. Moreover, the anti-apoptotic activity of B. henselae BepA, in human umbilical vein endothelial cells (HUVEC), correlates with an important elevation of intracellular adenosine 39, 59-cyclic monophosphate (cAMP) level [29]. A more recent study demonstrated that B. henselae infection involves the intrinsic apoptotic pathway [30].
ECs are morphologically and functionally heterogeneous with major differences between those from the macro-versus microcirculation as documented for a variety of tissues [31,32,33]. Except rare cases where ECs of the microvasculature have been included in Bartonella infection in vitro experiments [26,28,30,34], studies are mostly based on the use of primary HUVEC or other macrovasculature-derived cells like the hybrid EA.Hy.926. These cells originate from a large vessel of the placenta and are very different from microvasculature-derived ECs [31,32,33,35] clinically involved in bacillary angiomatosis and peliosis.
In addition, primary ECs will not allow providing repeatable and reproducible data, as these cultures lead to highly activated cells, in limited amounts and for a short term. Cell lines, established in a controlled identical manner, represent the best alternative to overcome these problems.
No comparative studies on the interactions between B. henselae and human (incidental host) ECs versus feline (reservoir host) ECs have ever been undertaken, because of the absence of any available feline ECs. Hence, feline ECs lines were developed and the in vitro effects of infection by distinct Bartonella species/strains (B. henselae genotype I/II and B. tribocorum) on human versus feline ECs derived from the macro-and micro-vasculature were compared. By studying various parameters as in vitro angiogenesis, stimulation of wound-healing, induction of cAMP, production of VEGF and activation of VEGF Receptor-2 (VEGFR-2), we demonstrated the crucial significance of the tissue origin of ECs and the specificity of the relationship between various strains of B. henselae with human versus feline cell lines. Our work contributes to the understanding of the differential outcomes of B. henselae infection in humans versus cats.

Human endothelial cell lines
Two immortalized lines of human organospecific ECs previously developed [31] were used in this study: one isolated from macrovasculature (immortalized Human Umbilical Vein Endothe-lial Cells: iHUVEC) and one isolated from the skin microvasculature (Human Skin Microvascular Endothelial Cells: HSkMEC) [36].

Feline endothelial cell lines
Establishment. Four immature and not viable embryos were aseptically removed at the School of Veterinary Medicine, Alfort, France, in the reproductive diseases service by therapeutic caesarean from queen after 45 days of gestation, and organs or tissues biopsies were taken from each embryo. Tissues and organs were placed in RPMI medium (Gibco, BRL, Cergy Pontoise France) without fetal bovine serum (FBS), supplemented by antibiotics and stored at 4uC.
For intracellular vWF detection, a rabbit polyclonal anti-human vWF antibody (A0082-Dako) (20 mg/ml), was applied to the cells for 2 h at RT and washed twice with c-PBS containing 1% bovine serum albumin (c-PBS-BSA). The second antibody was a goat Fluorescein IsoThioCyanate (FITC)-anti-rabbit immunoglobulin antibody (20 mg/ml) (SBE, CliniSciences, Montrouge, France), and cells were further incubated for overnight at 4uC.
b-Angiogenesis assays. A 96-well plate (Falcon, BD Biosciences, Grenoble, France) was coated with Matrigel TM (BD Biosciences) mimicking the extracellular matrix (40 ml by well). The Matrigel TM was allowed to polymerize for 1 h at 37uC before cell seeding (8610 3 cells/well). ECs rearrangements and capillarylike structure formation were observed each hour until the 24 th hour and then every day until the 7 th day. It was photographed regularly under a microscope (Nikon TMS, Japan) Monkey kidney epithelial cells (Vero line, ATCC, CCL-81) were used as a non endothelial cell control.

Bacterial strains
Four strains of B. henselae were used, based on their species origin (human or feline) and genotype (I or II). The two strains isolated from human patients were the reference strain Houston-1 (genotype I (H1) / ATCC 49882) [1] and the genotype II Marseille strain (H2) kindly supplied by Jean Marc Rolain (Unité des rickettsies, Marseille, France). The two strains isolated from cats were a genotype I strain (F1/ Strain 297172) kindly supplied by Bruno Chomel (University of California, Davis, USA) and a genotype II B. henselae strain initially isolated in our laboratory (F2). B. tribocorum (Bt) (CIP 105476 T), isolated from rat and without known pathogenicity for human and cats was also used.
All Bartonella strains were cultured on sheep blood agar medium (BioMerieux, Craponne, France) for 5 to 7 days in humidified atmosphere at 35uC and 5% CO 2 .
To investigate the specificity of Bartonella-triggered angiogenesis, we used Escherichia coli (DH5a) (Invitrogen, France) as a negative control. This bacterium was grown at 37uC on Luria-Bertani (LB) agar or in LB broth (Sigma, France). Heat-inactivated B. henselae were used in order to check if dead bacteria were still able to stimulate ECs angiogenesis. The inactivation of heat-treated bacteria (56uC, 30 min) was checked on sheep blood agar medium.
Generation of Bartonella culture supernatants B. henselae strains and B. tribocorum were harvested from blood agar plates and suspended in Schneider medium 1x (Gibco, France). The bacterial cultures were gently shaken in Schneider medium for 18 h at 200 rpm and 37uC on an orbital shaker (Innova 4230, New Brunswick Scientific, Edison, NJ, USA). After 18 h of incubation, the suspensions were removed from the flasks and spun at 1000 g and 4uC for 10 minutes to form a soft pellet. The supernatants were removed and passed through a 0.22 mm filter to remove all bacteria. The bacteria-free supernatants were dialyzed with OptiMEM and concentrated in a Centrifugal filter (Amicon H Ultra-15 Millipore, France). Culture supernatant control medium was also generated through similar methods. Protein concentrations were determined by bicinchoninic acid assay (BC Assay Kit, Uptima, Interchim, France).

In vitro induction of angiogenesis
Twenty-four hours before infection of ECs (iHUVECimmortalized Human Umbilical Vein Endothelial Cells-, HSkMEC -Human Skin Microvascular Endothelial Cells-, FOmEC-Feline Umbilical vein Macrovascular Endothelial Cells and FSkMEC-Feline Skin Microvascular Endothelial Cells), antibiotics were removed from the culture medium. ECs (8610 3 cells/well) were infected with Bartonella at a multiplicity of infection (M.O.I.) of 50, 100 and 150 bacteria per cell or treated with bacteria culture supernatant (protein concentration: 250 mg/ml) and then were seeded on 96 well plates previously coated with Matrigel TM (BD Biosciences, Grenoble, France). Videos for ECs rearrangement and capillary-like structure formation were acquired and registered for image analysis (Zeiss axiovision program). The Zeiss axiovert 200 M, videomicroscope equipped for temperature, gas and humidity controls was used to acquire images in time lapse conditions and reconstitute the kinetics of the angiogenesis dynamic process. The pseudo-vessel formation was monitored continuously during the first 24 hours and then observed every day during the following 7 days.
Capillary-like structure formation was quantified by measuring the length of tubes and counting the branching points [39].

Healing test
iHUVEC, HSkMEC, FOmEC and FSkMEC were seeded in 24-well plates at a density of 10 5 cells/well and allowed to attach overnight in OptiMEM medium without antibiotics. When cells came to confluence, a wound was made in the center of the well using the extremity of a sterile pipette tip. Then, ECs were infected with bacteria (100 M.O.I.) for 24 h. The healing was observed and photographed under a Nikon TMS microscope. Healing distances (in mm) were measured at time 0 and 24 h after the wound.

Determination of intracellular cAMP level
After infection of ECs (iHUVEC, HSkMEC, FOmEC and FSkMEC) by Bartonella strains with 150 M.O.I.for 30 h in 24 well plates, cells were washed with PBS and lysed. Intracellular cAMP levels were determined by the EIA system (Biotrak, Amersham, Biosciences, France) as described by the manufacturer.

Measurement of VEGF levels in conditioned medium (ELISA)
VEGF concentration in culture medium was measured using a commercially available human VEGF ELISA Kit (R&D Systems, Minneapolis, MN, USA) according to the manufacturer's instructions. Briefly, cells were plated at a density of 5610 4 cells/0,5 ml/well in a 24 well plate and 12 h later cells were infected with Bartonella at a M.O.I. of 100. Conditioned medium was collected 72 h after infection, cellular debris and extracellular bacteria removed by centrifugation, and the medium was kept at 280uC until VEGF quantitation was undertaken. Cell number was determined immediately after medium recovery.

Statistical analysis
All experiments were performed at least three times (except VEGF production: twice) and gave comparable results. Differences between mean values of experimental and control groups were analyzed by Student's t-test. P-values less than 0.05 were considered to be statistically significant. The data were expressed as mean 6 standard deviation (SD).

Establishment of new feline endothelial cell lines
In order to compare the feline ECs with their human counterparts (i.e HSkMEC -Human Skin Microvascular ECand iHUVEC -immortalized Human Umbilical Vein EC -), two feline cell lines (FSkMEC -Feline Skin Microvascular ECand FOmEC-Feline Umbilical Macrovascular EC -) were selected among the nine established feline ECs lines which have been characterized for EC markers as vWf and ACE (data not shown).
Moreover, because ECs can retain in vitro their ability to undergo angiogenesis, feline ECs lines were tested for their potentiality to form capillary-like structures in the classical Matrigel TM assay. All feline ECs were capable of performing angiogenesis on Matrigel TM . Figure 1 shows angiogenesis observed with FSkMEC and FOmEC.

Angiogenic response of ECs to Bartonella infection
Capillary-like structures of uninfected feline ECs (Fig. 1) began to form within two hours after seeding on Matrigel TM as compared to uninfected human ECs, while occurring within approximately five hours for macrovasculature derived ECs (iHUVEC) and within 10 hours for microvasculature ECs (HSkMEC) (table 1). The network of pseudo-vessels persisted in human ECs, until 24 hours for macro-vasculature derived ECs and until 7 days for microvascular ECs. Of note this network disappeared only after 20 hours with feline ECs (table 1).
In parallel, we monitored the effects of Bartonella (B. henselae genotype I/II and B. tribocorum) infection on the four cell lines. Irrespective of the species of origin and genotypes of Bartonella, infected HSkMEC efficiently achieved angiogenesis in terms of tube formation, extended projections morphology, branching points and number of cells as compared to uninfected cells (Fig. 2 A). The influence of Bartonella infection on the formation of the capillary-like network by HSkMEC was dose-dependent (data not shown). Infection accelerated angiogenesis by few hours in Human Skin Microvascular cells (table 1).
Importantly, kinetics of angiogenesis of the FSkMEC was not changed upon Bartonella infection whatever strain used (Fig. 2B,   Fig. 3B and table 1). Angiogenesis of FSkMEC and FOmEC started 2 h after seeding on Matrigel TM . The destruction of the network was slightly faster in the infected feline cell lines (18 hours) as compared to uninfected cells (20 hours) (table 1). No effect was observed upon infection on the structure of the newly formed vessels (Fig. 2B and Fig. 3B). Importantly, as a negative control, no pro-angiogenic effect was observed, both on human and feline ECs with E. coli or with heat-inactivated B. henselae (data not shown).
The observed effects were quantified. As shown on Fig. 3A, Bartonella-infection increased capillary-like formation by HSkMEC by a factor 1.8 to 2.6 as compared to uninfected ECs (p,0.05) and the number of branching points by a factor of 1.78 to 3 (p,0.05) (Fig. 3 A). This network was observed until 7 days with these cells, and then disappeared (table 1).
Capillary-like formation was increased in iHUVEC by a 2 fold upon infection such as in HSkMEC. However, as opposed to HSkMEC, B. henselae infection did not speed this effect (table 1). The network was observed only until 24 hours, before resorbing (table 1).
Thus, the effect of Bartonella-induced angiogenesis was more potent in human ECs than feline ECs (where infection even accelerated network destruction). Moreover, capillary-like structures were markedly more durable in micro-versus macrovasculature ECs.  Angiogenic response of ECs to Bartonella culture supernatants In order to investigate the effects of Bartonella secretome, different Bartonella culture supernatants were generated and tested for their ability to induce angiogenesis in human and feline ECs. Bartonella culture supernatants did induce tube formation in human ECs. The pro-angiogenic responses obtained with Bartonella culture supernatants were accelerated by a factor 1.9 to 2.5 (p,0.05), as in Bartonella-infected human ECs (Fig. 3A). The resulting effect of Bartonella or their culture supernatants on the morphology of human ECs was similar, i.e. ECs appeared elongated with long membrane projections (data not shown).
In the system of feline endothelial cells, Bartonella culture supernatants, as bacteria themselves, did not influence angiogenesis ( Fig. 2B and Fig. 3B).

Wound-induced migration of human ECs by Bartonella infection
As ECs migration is an essential step for angiogenesis, the significance of the site of infection in term of endothelium susceptibility was examined. The endothelial cell reaction according to their tissue origin was followed upon Bartonella infection by measuring their migration speed to heal a mechanical wound ( Fig. 4A and 4B). In uninfected cells (at 24 h), woundhealing was at least twice faster in iHUVEC than in HSkMEC (4.560.7 and 1.2560.35 mm within 24 h, respectively with p = 0.015) (Fig. 4C). Infection speeded the wound-healing for both iHUVEC and HSkMEC. Remarkably, in iHUVEC, healing speed was approximately doubled, whereas in HSkMEC, it was increased by a factor of 4.3 up to 7.8 (p,0.05) (Fig. 4).
In feline ECs (FOmEC and FSkMEC) wound-induced migration was recorded and no stimulation by Bartonella infection was observed (Fig. 4A and Fig. 4C).

Effect of Bartonella infection on human and feline ECs on cAMP production
It was shown in the HUVEC model that B. henselae genotype I (reference strain Houston-1) triggered the production of cAMP [29]. As, according to these authors, anti apoptosis was mediated through increased cAMP levels, this induction was assessed in ECs upon infection by distinct Bartonella strains. In uninfected ECs of the macro-or the micro-vasculature, the levels of cAMP produced by the feline ECs (FOmEC [12.0667. 13 (Fig. 5). Moreover, cAMP production by human ECs of the macrovasculature (iHUVEC) was higher than microvasculature-derived ECs (HSkMEC) by a factor of 6.2 (203.57657.66 vs 32.6861.80 fmol respectively with p = 0.0004). In feline ECs, the low levels of production did not allow to detect any significant difference between ECs of the macro-and the micro-vasculatures (Fig. 5).
Upon infection, cAMP increased only in human macrovasculature ECs (iHUVEC) infected by strain H1 by a factor of 2.2 (reaching 405.1666.72 fmol) (p,0.05). Strains F1, F2 and Bt did not induce any change. Human Skin Microvasculature ECs were not affected by any of the strains.
Feline ECs from micro-or macro-vasculature ECs did not react to the bacterial infection (Fig. 5).

Bartonella increases VEGF production in human microvascular ECs
The involvement of VEGF in B. henselae effect on angiogenesis was investigated. We took into account: a) the origin of the bacterial strains (human versus feline); b) the bacterial genotype (I versus II); c) the ECs species (human versus feline), and d) their type (macro versus micro vessels derived-ECs). A differential production of VEGF was measured between non infected microvascular and macrovascular ECs both of human and feline origin as shown on Fig. 6. Human and feline microvascular ECs produced high levels of VEGF [717.88632.50 and 1623.19672.81 pg/ml respectively] as compared to human and feline macrovascular ECs [3.7563.12 and 6.8865.0 pg/ml respectively, p = 0.001] (Fig. 6).
An increase in VEGF production by B. henselae infection was induced only in human microvascular skin-derived cells (HSkMEC) (Fig. 6).
Interestingly, the four B. henselae strains significantly increased VEGF production by HSkMEC by a factor of 2.1 to 2.4 while B. tribocorum was less effective (factor of 1.3) (p,0.05). Feline microvascular cells production of VEGF was not significantly affected by the various strains.

Effects of B. henselae infection on VEGFR-2 activation
The level of activation of the VEGF receptor VEGFR-2 was estimated by monitoring VEGFR-2 phosphorylation in Westernblot. Figure 7 indicates that ECs (iHUVEC, HSkMEC, FOmEC and FSkMEC) infection by Bartonella was associated with an increased phosphorylation of VEGFR-2.
In addition, the level of the VEGFR-2 phosphorylated form was significantly higher by a 2 fold when ECs were infected by a strain originating from a homologous host (i.e. when human ECs were infected by B. henselae strains originating from humans and when feline ECs were infected by B. henselae strains originating from cats). In particular, VEGFR-2 phosphorylation induced in human ECs (HSkMEC and iHUVEC) by strain H1 was respectively 2.2 to 3.3 higher than that induced by F1. Conversely, VEGFR-2 phosphorylation induced in Feline ECs (FSkMEC and FOmEC) by H1 was lower than that induced by F1 respectively 0.95 to 0.6 ( Fig. 7A and 7B). B. tribocorum induced a comparable activation of VEGFR-2, whatever the origin of cell lines (Fig. 7A).

Discussion
B. henselae is a facultative intracellular pathogen associated with the induction of vasoproliferative tumors in humans (bacillary angiomatosis and peliosis) that are not observed in cats (host reservoir) [4,6,7,8].
Our original cellular models allowed us to reveal intrinsic differences in the angiogenesis kinetics and the wound-healing process between human micro-and macro-vasculature-derived ECs. Importantly, our results strongly suggest that effects of B. henselae on microvasculature-derived cells strongly differ from macrovasculature ones.
In human cells, Bartonella infection affected the angiogenic process at two different levels: a) a slight acceleration of the pseudovessel rearrangement was observed, only with microvasculature-derived ECs (Table 1) and b) the number of pseudo-vessels was increased in both micro-and macro-vasculature-derived ECs (figure 3).
One key point of our findings concerns feline ECs from both micro-and macro-vasculatures. Indeed, not only B. henselae had no visible pro-angiogenic effect in vitro, but the infection seemed to accelerate the network destruction of the feline cells. These results might recapitulate the clinical situation, i.e. the absence of bacillary angiomatosis in cats (even in FIV and/or FeLV infected cats) versus bacillary angiomatosis in humans and even in dogs as recently described [40].
Hence, our cellular models are relevant to understand the cellular and molecular bases of the pro-angiogenic potential of B. henselae and to explore the mechanisms underlying bacillary angiomatosis/peliosis. This model of infection shows that, independently of the bacterial species (B. henselae vs B. tribocorum), or of the B. henselae genotype (I or II), Bartonella tested here can induce in vitro angiogenesis. Interestingly, B. tribocorum has no known proangiogenic potential in vivo. Moreover we did not detect any obvious difference of activity between the different strains of Bartonella tested contrary to the work of Chang et al [30].
Furthermore, the stimulatory effect of B. henselae is not dependent upon attachment or penetration in the target cell suggesting that B. henselae may produce and secrete endothelial cell-stimulatory factor(s).
In addition, the inactivation of B. henselae by heat treatment abolished its vasoproliferative activity, suggesting that the angiogenic effect of B. henselae on ECs require live bacteria.
Previous studies dedicated to understand the interactions between primary HUVEC and B. henselae Houston 1 (genotype 1) have suggested a correlation between the anti-apoptotic activity of B. henselae and an increased cAMP production [29]. Remarkably, in our study, the stimulation of cAMP production was observed only when B. henselae Houston 1 was inoculated to   iHUVEC. Such effect was never obtained when skin-derived EC -HSkMEC -, which are directly involved in bacillary angiomatosis, were used, even when they were infected by Houston 1. This points to the fact that macrovasculature-derived ECs (iHUVEC/ HUVEC) deeply differ from microvasculature ECs [31,32,33,41,42]. Remarkably, in our study, cAMP production does not reflect ECs proliferation in the case of infected microvascular ECs. According to these results, microvascular skin endothelial cells-HSkMEC, clearly represent a better model for clinical bacillary angiomatosis.
Feline macro-and/or micro-vasculature ECs cAMP production was weak and not related to infection.
Such low level of cAMP production by human microvasculature ECs (HSkMEC) does not corroborate the pro-angiogenic effect of B. henselae interaction with human ECs. Nishihara et al [43], have shown that the effects of cAMP are highly cell typespecific. cAMP either promotes or antagonizes apoptosis in a cell type, environment-and stimulus-related manner. These, together with our results, illustrate that the model based on HUVEC which is not the clinical cell target, does not allow any definitive conclusion about the role of cAMP during B. henselaeanti-apoptotic activity.
VEGF, one of the main pro-angiogenic factors, plays a critical role in blood vessel formation [44] and pathological angiogenesis [45]. This factor seems also to be involved in Bartonella-induced angiogenesis. For instance, in verruga peruana (induced by B. baciliformis), the primary source of vascular endothelial growth factor (VEGF) is the epidermis [46] and endothelium of verruga peruana expresses VEGF receptors (VEGFR1 and VEGFR2) [46].
In addition, while HUVEC did not produce VEGF in response to B. henselae, infection promotes their proliferation [22]. Further study by Kempf et al showed the importance of VEGF in B. henselae-mediated ECs proliferation although it appeared not to be produced by ECs only. Overall, this strongly suggests that a paracrine loop could be involved in B. henselae VEGF action [27,28]. However, up to now, the majority of these studies have been carried out on macrovascular ECs (i.e. HUVEC). Interestingly, using our species-and organ-specific ECs model, we showed that B. henselae increased VEGF production by Human Skin Microvascular ECs but not by iHUVEC or feline macro-or micro-vascular ECs. These results, which might explain the decreased pathogenic potential of B. henselae infection for cat as compared to human, strongly suggests the involvement of an autocrine secretion of VEGF by skin ECs. This finding points out that the mechanism of B. henselae-mediated angiogenesis induction implies autocrine VEGF production and stimulation of the infected endothelium.
In parallel to VEGF production, the phosphorylation of VEGFR-2 was observed in ECs upon infection, mostly by homologous strains. This phenomenon is of high interest in the context of epidemiological human infection. As B. henselae isolates infecting a human have always a feline origin, it is tempting to speculate that an adaptive switch is taking place upon cat scratch in humans. Altogether, our results raise the hypothesis of hostdependent signaling in ECs upon bacterial infection as VEGFR-2 activation is more prominent in human ECs exposed to homologous strains. The same holds true for feline ECs. Because bacillary angiomatosis is not detected in cats infected by homologous strains, alternate mechanisms are likely taking place, Our results support the fact that VEGF signaling might be increased in human ECs through an autocrine production, which does not occur in feline ECs. Paracrine VEGF activation cannot be excluded; therefore, it will be interesting to further investigate whether VEGF can be released in the perivascular microenvironment by other cells, such as macrophages and polymorphonuclear cells. Our results on VEGF and VEGFR-2 are in apparent contradiction with those obtained by Scheidegger et al [47]. However, the concentration ranges and source of VEGF were far for being comparable.
Our study highlights the ECs organo-and species-specificity in their interactions with B. henselae. Indeed, our model further demonstrates the difference of reactivity displayed by human vs feline ECs on the one hand and by the macro-vs the microvasculature on the other hand. Additionally, local physiological microenvironment is most likely to play an important role. In this context, our cellular systems will allow studying the process of specific ECs infection by B. henselae. They will offer new possibilities to investigate bacterial and cellular factors which determine human vs cat reactivity, as well as the mechanisms involved in anti-apoptotic and/or pro-angiogenic effects, ultimately resulting to bacillary angiomatosis and peliosis only in humans.
In conclusion, according to the validated hypothesis stating the organospecificity of the endothelium [31,36,37,38], our work extends this concept to the species-specific endothelial cell phenotypes. In this scenario, ECs could act as a major reservoir for infectious pathogens. Moreover, at a second level the ECs from the same tissue but from different species display similar organospecificity but distinct responses to infection. This is clearly illustrated by the reactivity of the human skin-derived ECs compared to the feline skin-derived ECs. It will further be necessary to evidence the molecular mechanism by which feline ECs are recognized but not activated upon binding of B henselae, while human ECs are activated to grow and proceed to tumor-like state. The co-activation signaling should indicate differential gene induction and regulation processes.
Such information will provide new insights into the mechanisms of species-specific tolerance versus pathologic infections as observed in AIDS malignancies.