Evidence of a leading role for VEGF in Bartonella henselae-induced endothelial cell proliferations


*For correspondence. E-mail volkhard.kempf@med.uni-tuebingen.de; Tel. (+49) 7071 298 1534; Fax (+49) 7071 293 435.


Bartonella henselae causes the vasculoproliferative disorders bacillary angiomatosis (BA) and bacillary peliosis (BP). The pathomechanisms of these tumorous proliferations are unknown. Our results suggest a novel bacterial two-step pathogenicity strategy, in which the pathogen triggers growth factor production for subsequent proliferation of its own host cells. In fact, B. henselae induces host cell production of the angiogenic factor vascular endothelial growth factor (VEGF), leading to proliferation of endothelial cells. The presence of B. henselae pili was associated with host cell VEGF production, as a Pil mutant of B. henselae was unable to induce VEGF production. In turn, VEGF-stimulated endothelial cells promoted the growth of B. henselae. Immunohistochemistry for VEGF in specimens from patients with BA or BP revealed increased VEGF expression in vivo. These findings suggest a novel bacteria-dependent mechanism of tumour growth.


Angiogenesis is the process of blood vessel formation in the adult. Physiological angiogenesis occurs almost exclusively in the female reproductive system. Pathological angiogenesis is a hallmark of tumour growth and metastasis, various cardiovascular and inflammatory diseases such as diabetic retinopathy, rheumatoid arthritis, psoriasis and others (Risau, 1997; Carmeliet and Jain, 2000). Vascular endothelial growth factor (VEGF) plays a fundamental role in triggering the process of angiogenesis (Yancopoulos et al., 2000) and is a highly specific and potent mitogen for endothelial cells (Leung et al., 1989). In addition, a central role for VEGF has been shown in the pathogenesis of the vasculoproliferative disorder Kaposi's sarcoma, a disease triggered by human herpesvirus 8 (HHV-8; Masood et al., 1997).

The Gram-negative bacterium Bartonella henselae is a fastidious, slow-growing and facultative intracellular microorganism that has been identified as an emerging pathogen causing a wide variety of disease syndromes, including cat scratch disease (CSD), bacillary angiomatosis (BA) and bacillary peliosis (BP) in humans (Anderson and Neuman, 1997). The domestic cat is a confirmed reservoir for B. henselae (Haimerl et al., 1999). CSD is usually a benign and self-limiting disease characterized by lymphadenopathy related to a cat scratch or cat exposure. Interestingly, in immunocompromised individuals, B. henselae causes tumorous proliferations of endothelial cells called BA (cutaneous) and BP (visceral). These vasculoproliferative disorders are likely to be induced by B. henselae, as the pathogen is detectable in these lesions and bacterial eradication by antibiotic treatment results in a complete regression of angiomatous tumours (Webster et al., 1992; Anderson and Neuman, 1997). The pathogenic process of this bacteria-triggered tumour induction is poorly understood. Recently, it has been described that B. henselae replicates within endothelial cells (Kempf et al., 2000). Moreover, a proliferative effect of B. henselae on endothelial cells has been shown in vitro (Conley et al., 1994; Maeno et al., 1999). However, the molecular basis of the events operating in BA and BP has not yet been established.

In this study, we analysed whether B. henselae triggers the production of VEGF by host cells. By means of a co-culture infection model, we found that B. henselae expressing pili induce VEGF in host cells. In turn, VEGF-stimulated endothelial cells proliferated and promoted the growth of B. henselae. Finally, immunohistochemistry for VEGF in a patient's specimen of BA or BP revealed strongly increased VEGF expression in vivo. Our findings suggest a novel bacteria-dependent mechanism of tumour growth.


Bartonella henselae triggers VEGF production of EA.hy 926 cells but not of human umbilical vein endothelial cells (HUVECs)

To investigate whether B. henselae can trigger VEGF production of host cells, we co-cultivated various human cells/cell lines with B. henselae and determined VEGF levels in culture supernatants. EA.hy 926 cells (Edgell et al., 1983) (Fig. 1A) and HeLa cells produced large amounts of VEGF upon B. henselae exposure, with a maximum peak after 72 h of co-cultivation [negative controls: 688.2 ± 41.5; positive controls (25 ng ml−1 phorbol-12myristate-13 acetate; PMA): 1848.8 ± 485.8, B. henselae-infected HeLa cells; 1461.8 ± 140.9 pg ml−1]. VEGF production was associated with the presence of intracellular B. henselae as shown by confocal laser scanning microscopy (CLSM) (Fig. 1B). In contrast, HUVECs produced no VEGF in response to B. henselae over 5 days [negative controls: not detectable (ND); positive controls (25 ng ml−1 PMA): ND; B. henselae-infected HUVECs: ND).

Figure 1.

VEGF production of EA.hy 926 cells co-cultured with B. henselae.

A. Determination of VEGF in cell culture supernatants by ELISA. EA.hy 926 cells were infected with B. henselae, and supernatants were taken after 1, 3 and 5 days of co-culture. Control cells were not infected or stimulated with phorbol-12-myristate-13 acetate (PMA; 25 ng ml−1).

B. Course of B. henselae infection in EA.hy 926 cell co-culture analysed by CLSM. Extracellular bacteria were labelled with FITC-conjugated antibodies (green signal), and intracellular bacteria were labelled with Cy-5-conjugated antibodies (blue signal). Filamentous actin was stained with TRITC-labelled phalloidin (red signal). Scale bar = 10 µm.

Production of interleukin-8 (IL-8), a further vasculoproliferative cytokine (Yoshida et al., 1997), was promoted in B. henselae-infected EA.hy 926 cells after 3 days of co-cultivation {negative control: 121.5 ± 12.6; positive control [50 ng ml−1 recombinant human tumour necrosis factor alpha (rh TNF-α)]: 1321.4 ± 18.9; B. henselae-infected EA.hy 926 cells: 372.0 ± 13.5 pg ml−1}. The increased VEGF levels in B. henselae-infected EA.hy 926 cells resulted from de novo protein synthesis of the host cells, as suppression of protein biosynthesis by cycloheximide (CHX; 20 µg ml−1) abolished VEGF production completely; suppression of actin polymerization by cytochalasin-D (Cy-D; 100 nM), which inhibits invasome formation (Dehio et al., 1997), but not invasion of singular B. henselae bacteria, resulted in a slight decrease in host cell VEGF production compared with controls, indicating that invasome formation is not significantly involved in VEGF production (VEGF levels 72 h after infection: B. henselae 168.2 ± 10.1; B. henselae + CHX: ND; B. henselae + Cy-D: 125.4 ± 8.2 pg ml−1).

Bartonella henselae-promoted VEGF production is a specific property of B. henselae

To investigate the specificity of B. henselae-triggered VEGF production, we tested several bacterial controls for their ability to modulate VEGF production. Infection with non-invasive Escherichia coli (HB 101), invasive E. coli pINV1914 (expressing invasin protein of Yersinia enterocolitica (Schulte et al., 2000), Listeria monocytogenes EGD serotype 1/2a or lipopolysaccharide (LPS) from E. coli did not induce the production of significant amounts of VEGF by EA.hy 926 cells, suggesting that B. henselae-promoted VEGF production depends on a factor specific for B. henselae(Fig. 2). Furthermore, heat-killed B. henselae induced ∼ 30% of the VEGF production obtained with viable B. henselae, suggesting a role for surface components of B. henselae (viable B. henselae: 81.3 ± 15.6; heat-killed B. henselae: 31.0 ± 4.2 pg ml−1).

Figure 2.

VEGF production of EA.hy 926 cells upon co-cultivation with E. coli LPS, E. coli HB101, E. coli HB101 pINV1914 and L. monocytogenes EG11. VEGF levels in cell culture supernatants were determined by ELISA after 3 days of co-culture. Control cells were not infected or stimulated with phorbol-12-myristate-13 acetate (PMA; 25 ng ml−1).

Expression of B. henselae pili is crucial for triggering host cell production of VEGF

To elucidate further bacterial factors triggering the process of VEGF production by host cells, we investigated the role of B. henselae pili. Expression of pili confers adhesion to and invasion of B. henselae into host cells (Batterman et al., 1995). For this purpose, Pil mutants were produced by extensive passaging on Columbia agar (see Experimental procedures). Expression of pili was assessed by colony morphology and subsequent electron microscopy. Infection of EA.hy 926 cells with B. henselae Pil mutants resulted in a dramatically decreased production of VEGF compared with Pil+B. henselae strains, suggesting a leading role for pili expression in triggering VEGF production (Fig. 3A). B. henselae Pil mutants showed a 90% diminished cell adhesion and invasion compared with B. henselae Pil+(Fig. 3B). From these results, we conclude that adhesion and invasion of B. henselae into host cells mediated by the expression of pili is a key mechanism in the induction of VEGF production by host cells.

Figure 3.

VEGF production of cells co-cultured with B. henselae Pil+ and B. henselae Pil mutants (30, 300, 900 and 1200 bacteria per cell).

A. VEGF levels in cell culture supernatants 3 days after co-culture determined by ELISA. Expression of pili was determined by transmission electron microscopy of Columbia agar-grown B. henselae. Scale bar = 0. 5 µm.

B. Adherence (30 min) of B. henselae Pil+ (wild strain) and B. henselae Pil mutants to, invasion (+2 h) into and intracellular presence (+72 h) in EA.hy 926 cells determined by gentamicin protection assays.

VEGF derived from B. henselae-infected host cells leads to proliferation of endothelial cells

To confirm that infection of EA.hy 926 cells by B. henselae leads to the production of biologically active VEGF, we stimulated HUVECs with conditioned media and quantified proliferation by [3H]-thymidine incorporation into DNA (for details, see Experimental procedures). For this purpose, supernatants from B. henselae-infected EA.hy 926 cells were sterile filtered (0.2 µm), diluted and added to HUVECs. Proliferation of HUVECs co-cultured with conditioned media was about 30–70 times higher than in controls (Fig. 4; not all data shown). The addition of VEGF-neutralizing antibodies significantly reduced the proliferation to ∼ 50% (Fig. 4A). Phase-contrast microscopy of HUVECs confirmed these observations (Fig. 4B). Therefore, although HUVECs do not produce VEGF, they respond to VEGF in conditioned media derived from B. henselae-infected EA.hy 926 cells with a strong mitogenic response.

Figure 4.

Proliferation of uninfected endothelial cells (HUVECs) treated with 0.2 µm of filtered conditioned medium from B. henselae-infected cell cultures.

A. [3H]-thymidine uptake of HUVECs treated with filtered conditioned medium from B. henselae-infected EA.hy 926 cell cultures (controls: 0.0, 0.1, 1.0, 10.0 pg ml−1 recombinant human VEGF; VEGF-neutralizing antibody concentration: 10.0 ng ml−1). SI, stimulatory index.

B. Morphology of HUVECs (phase-contrast microscopy) treated with recombinant human VEGF or filtered conditioned medium from B. henselae-infected EA.hy 926 cell cultures.

VEGF-stimulated proliferating endothelial cells promote the growth of B. henselae

Bartonella henselae replicates in endothelial cells (Kempf et al., 2000). Therefore, bacteria-triggered proliferation of endothelial cells might, in turn, promote the growth of Bartonella. To address this issue, we cultured B. henselae with endothelial cells in the presence of VEGF or conditioned media. In these experiments, we observed strongly increased growth rates of B. henselae co-cultured with HUVECs exposed to VEGF (stimulatory index 8.3, ∼ 75-fold increased growth rate) or conditioned media (stimulatory index 43.4, ∼ 150-fold increased growth rate) compared with controls (Fig. 5). These results demonstrate for the first time that proliferation of endothelial cells promotes growth conditions for B. henselae. Based on this observation, we suggest that the vasculoproliferative diseases BA and BP could thus be considered as creations of the bacterial habitat ensuring optimal living conditions.

Figure 5.

Number of intracellular B. henselae in proliferating HUVECs stimulated with VEGF or conditioned media.

A. [3H]-thymidine uptake and morphology of HUVECs (phase-contrast microscopy) treated with 0.2 µm of filtered conditioned medium from B. henselae-infected EA.hy 926 cultures (controls: 0.0, 10.0 pg ml−1 recombinant human VEGF). SI, stimulatory index.

B. Number of intracellular B. henselae (+2 h, +72 h) in HUVECs determined by gentamicin protection assays.

VEGF is expressed in tissue lesions of patients with BA or BP

To address the question as to whether VEGF levels are actually increased in B. henselae-induced tissue lesions of patients with BA/BP in vivo, we investigated the presence of VEGF in B. henselae-induced BA and BP by immunohistochemistry. For this purpose, patient specimens of BA (Schlupen et al., 1997) and BP that were positive for B. henselae DNA (Dauga et al., 1996) were analysed microscopically. Immunohistochemistry revealed strong signals for VEGF around vessels (BA) or in spindle cells and around vessels (BP) (Fig. 6). From these data, we can conclude that, as in Kaposi's sarcoma (KS) (Masood et al., 1997), VEGF plays an important role in the pathogenesis of BA and BP in vivo.

Figure 6.

Immunohistochemical detection of VEGF expression in patient specimens using a VEGF-specific antibody (red colour). Negative control (upper left), Kaposi's sarcoma (upper right), bacillary angiomatosis (lower left) or bacillary peliosis (lower right).


The major goal of this study was to determine the impact of VEGF in the pathogenesis of B. henselae-induced BA/BP (Anderson and Neuman, 1997). VEGF plays a fundamental role in pathological angiogenesis (Yancopoulos et al., 2000) and is a highly specific proliferative factor for endothelial cells (Leung et al., 1989). A central role for VEGF has been shown in the pathogenesis of the human herpesvirus 8 (HHV-8)-triggered vasculoproliferative disorder Kaposi's sarcoma (Masood et al., 1997). The most salient findings from our study are: (i) B. henselae triggers VEGF production; (ii) expression of B. henselae pili plays a significant role in the process of triggering VEGF production in host cells; (iii) conditioned media of B. henselae-infected host cells containing VEGF lead to proliferation of endothelial cells; (iv) proliferating endothelial cells promote growth of B. henselae; and (v) VEGF levels are increased in vivo as shown in patient specimens of BA and BP.

Bartonella henselae is a re-emerging human pathogen that has been associated with BA and BP, cat scratch disease (CSD), bacteraemia and endocarditis (Drancourt et al., 1996; Anderson and Neuman, 1997). Histological investigations of biopsy specimens from patients with BA revealed bacteria in close association with proliferating endothelial cells (Schneider et al., 1993; Monteil et al., 1994). This angiogenic process is likely to be induced by B. henselae, as bacterial eradication by antibiotic treatment results in a complete regression of angiomatous tumours (Webster et al., 1992). The ability to induce endothelial proliferations is a common feature of B. henselae, Bartonella quintana and Bartonella bacilliformis (Anderson and Neuman, 1997). This fascinating angiogenic potential is of special interest in terms of understanding bacteria-triggered tumour formation. Elucidation of tumorous diseases caused by bacteria might lead to new insights into microbial pathogenicity, tumour growth and pathological angiogenesis. However, only a few contradictory data have been published on the pathomechanisms operating in B. henselae-induced angiomatous diseases. Direct interaction of B. henselae with endothelial cells may result in increased endothelial cell proliferation (Conley et al., 1994; Palmari et al., 1996). The presumptive factor responsible for HUVEC proliferation is suggested to be a bacterial membrane protein (Conley et al., 1994), whereas others have provided evidence for a soluble factor secreted by B. henselae (Maeno et al., 1999). In addition, there is no evidence for VEGF production of B. henselae-infected HUVECs (Maeno et al., 1999).

In the previously described models, co-culture of HUVECs with B. henselae resulted in two- to threefold increased proliferation rates of endothelial cells. In contrast, in our model, proliferation rates increased 30- to 70-fold compared with controls. Because of the remarkable higher proliferation of endothelial cells, we assume that this mechanism might be of significance in understanding the pathomechanisms underlying BA/BP and most probably reflects a novel pathomechanism distinct from that reported previously. Our data are consistent with observations showing that B. bacilliformis-infected endothelial cells on their own do not participate effectively in angiogenesis (Verma et al., 2001). Rather, infected cells release signals that stimulate other endothelial cells to participate in capillary formation. Although most types of cells produce VEGF, endothelial cells in general transcribe the gene at very low levels in contrast to other cells (Barleon et al., 1994; Dvorak et al., 1999). This fact might explain why HUVECs, in contrast to EA.hy 926 and HeLa cells, do not respond with increased VEGF expression upon co-cultivation with B. henselae.

Induction of proinflammatory cytokines in bacterial infections that share vasculoproliferative effects (e.g. IL-8; Yoshida et al., 1997) have been shown for several other bacteria, such as Y. enterocolitica (Schulte and Autenrieth, 1998), Helicobacter pylori (Yamaoka et al., 2000) and Shigella flexneri (Philpott et al., 2000). In vitro stimulation of neutrophils showed that pneumococci induce VEGF secretion, suggesting a role as a mediator of increased vascular permeability in meningitis (van Der Flier et al., 2000). In contrast to B. henselae, all these pathogens do not induce vasculoproliferative diseases. The reasons for the exclusive role of Bartonella spp. in inducing endothelial proliferations remain unclear. A possible explanation might be that induction of VEGF is a Bartonella-specific process triggered by ‘silent’ intracellular bacteria that does not lead to cell death (unpublished observation), whereas pneumococci induce tissue damage and apoptosis (Zysk et al., 2000). In the present state, it cannot be totally excluded that Bartonella infections of host cells induce a mitogenic activation or lead to a stress response of epithelial cells, giving rise to a secondary production of vasculoproliferative cytokines. Unspecific stimuli such as ultrasound have been described as potent inductors of IL-8 and VEGF production in various cell types (Reher et al., 1999). It has been discussed that ultrasound-triggered VEGF production might be an unspecific process linked to mechanical alterations in the cytoskeleton. As it is known that B. henselae interacts with actin polymerization (Dehio et al., 1997), a similar process might also be involved in B. henselae-triggered VEGF production. However, pretreatment of B. henselae-infected EA.hy 926 cells with Cy-D, which inhibits actin polymerization, did not significantly influence the level of VEGF production, indicating that interaction of B. henselae with the actin cytoskeleton is probably not involved in the process of VEGF production. Moreover, bacterial controls, e.g. E. coli Hb 101, E. coli pINV1914 (expressing invasin protein of Y. enterocolitica), L. monocytogenes EGD serotype 1/2a or LPS derived from E. coli, were not able to induce detectable amounts of VEGF in the in vitro model described here (see Fig. 2). From these data, we conclude that induction of VEGF production in host cells is generally not a common phenomenon upon co-cultivation of host cells with bacteria or bacterial components. Experiments to elucidate this topic will possibly explain the mechanisms of B. henselae-triggered vasculoproliferative disorders.

The stages of angiogenesis, including endothelial cell migration and proliferation, lumen formation and the synthesis of matrix components, are tightly regulated and involve more than one angiogenic factor at each stage. Many other factors, such as fibroblast growth factors, angiopoetins, ephrins and interleukins, play important roles in the regulation of angiogenesis (Yoshida et al., 1997; Yancopoulos et al., 2000). Administration of VEGF-neutralizing antibodies reduced endothelial proliferation to ∼ 50%, suggesting that other additional factors might be involved in the process of B. henselae-triggered angiogenesis. In the past, several cytokines that share vasculoproliferative effects synergistic to VEGF, e.g. TNF-α, IL-8 and basic fibroblast growth factor (bFGF), have been described (Yoshida et al., 1997). IL-8, which we also found to be induced after B. henselae infection, might contribute partially to the vasculoproliferative effect of the conditioned media on HUVECs, which was not blocked after the administration of VEGF-neutralizing antibodies. The role of other vasculoproliferative cytokines, such as TNF-α, bFGF, angiopoetins or ephrins, in Bartonella infections has yet to be investigated, as it is known that these factors play essential roles in angiogenesis (Yancopoulos et al., 2000).

By triggering VEGF production in BA/BP, B. henselae might create its own habitat, as B. henselae growth is promoted by proliferating endothelial cells in which B. henselae replicates intracellularly (Kempf et al., 2000). A similar mechanism has been described for the plant pathogen Agrobacterium tumefaciens, which is phylogenetically closely related to B. henselae (Matthysse, 1987; Christie and Vogel, 2000). A. tumefaciens injects T-DNA into plant cells via its type IV secretion system, which produce growth hormones (auxins, cytokinines) and opines to trigger plant cell growth and the production of nitrogen sources in order to ensure good bacterial growth conditions. Whether the type IV secretion system of B. henselae (Schmiederer and Anderson, 2000) is involved in triggering host cell VEGF production is unknown.

Bartonella henselae attachment to, and entry into, human epithelial and endothelial cells has been reported previously (Batterman et al, 1995; Dehio et al., 1997; Kempf et al., 2000). Invasion of cultured epithelial cells by B. henselae may be preceded by cell adherence mediated by the expression of type 4-like pili (Batterman et al., 1995). We found that adhesion and invasion into host cells was 90% diminished in Pil mutants of B. henselae. As Pil mutants were unable to trigger VEGF production, we conclude that the initial host cell contact mediated by pili rather than intracellular replication of B. henselae is crucial in this process. In fact, 3 days after infection, B. henselae Pil mutants reached similar intracellular growth levels, although they did not trigger VEGF production (unpublished observations). This model is supported by infection experiments with heat-killed B. henselae, which also lead to the production of VEGF by host cells. Moreover, these data are consistent with the observation that adhesion and/or invasion of bacteria into cells triggers host cell production of cytokines or plant cell proliferation (Matthysse, 1987; Kagnoff and Eckmann, 1997).

Taken together, our results provide strong evidence for the first time that B. henselae causes endothelial cell proliferation by triggering VEGF production, suggesting a novel two-step strategy of bacterial pathogenicity in which B. henselae promotes its own habitat. The possible role of bacteria in the induction of other human tumour diseases also needs to be investigated.

Experimental procedures

Bacterial strains

Bartonella henselae strain Marseille (Drancourt et al., 1996) was grown on Columbia agar plates supplemented with 5% defibrinated sheep blood (Becton Dickinson). For infection experiments, bacterial stocks were thawed, washed, suspended in cell culture medium (see below) and adjusted to the appropriate concentration. The amount of actual inoculum was determined for each experiment by plating serial dilutions of the suspension and calculating the number of bacteria. L. monocytogenes strain EGD serotype 1/2a was grown in brain–heart infusion (BHI) medium (Autenrieth et al., 1992). Heat-killed B. henselae were generated by incubation of viable bacteria at 58°C for 30 min. Non-invasive E. coli HB101 and E. coli HB101 pINV1914 expressing the Y. enterocolitica invasin (Schulte et al., 2000) were grown in LB broth and diluted in cell culture medium. Lipopolysaccharide (LPS) derived from E. coli O55:B5 was purchased from Difco laboratories.

Cell lines and cell culture

EA.hy 926 cells (Edgell et al, 1983), which were derived by fusing HUVECs with the permanent human cell line A549 (lung carcinoma), were cultured in Clicks/RPMI-1640 medium (Biochrom) with 10% heat-inactivated fetal calf serum (FCS; Sigma), 2 mM glutamine, 1 mM sodium pyruvate, non-essential amino acids, 10 µg ml−1 streptomycin and 100 U ml−1 penicillin (Biochrom). HUVEC culture was performed in endothelial growth medium (Promocell) as described previously (Dehio et al, 1997; Kempf et al., 2000) with few modifications. For infection, cells were seeded in 24-well dishes (1.0 × 105 cells per well) on the day before the experiment. Media were removed 4 h before infection, and culture media without antibiotics to allow bacterial growth and without serum to avoid non-specific VEGF production were added. Bacteria were sedimented onto the cultured cells by centrifugation for 5 min at 1800 g. Phorbol-12-myristate-13-acetate (PMA, 25 ng ml−1; Sigma) was used as a positive control (Enholm et al., 1997). Supernatants were taken as indicated, centrifuged and frozen at −20°C. In some experiments, co-culture was performed in culture medium containing cycloheximide (CHX, 20 µg ml−1) or cytochalasin D (Cy-D, 100 nM) (Kempf et al., 2000).

VEGF-enzyme-linked immunosorbent assay (ELISA) and IL-8 ELISA

VEGF concentration in culture medium was measured using a human VEGF165-ELISA kit (Quantikine; R & D Systems) according to the manufacturer's instructions. IL-8 was determined by ELISA as described previously (Schulte et al., 2000). Briefly, ELISA microtitre plates (Nunc) were coated overnight with anti-human IL-8 monoclonal antibodies (mAbs) (G265-5; Pharmingen), and supernatants were added. After several washing steps, biotin-labelled anti-human IL-8 mAb (G265-8; Pharmingen) and, finally, an avidin–biotin–alkaline phosphatase complex (Strept ABC-AP kit; Dako) was added. For signal development, the wells were incubated with p-nitrophenylphosphate disodium (pNPP; Sigma), and the optical density (OD) was determined at wavelengths of 405 and 490 nm. IL-8 concentrations were calculated from the straight-line portion of standard curves with recombinant human IL-8 (Pharmingen).

Confocal laser scanning microscopy

For immunostaining, cells were seeded onto collagen-G-coated coverslips. Infection was stopped as indicated by three washes with PBS, and cells were fixed in 3.75% PBS-buffered paraformaldehyde solution. Rabbit polyclonal antibodies were raised against viable bacteria of B. henselae Marseille. Fluorescein isothiocyanate (FITC)-conjugated goat anti-rabbit IgG antibodies, Cy-5-conjugated goat anti-rabbit IgG Fab fragments and tetramethyl rhodamine isothiocyanate (TRITC)-labelled phalloidin were purchased from Dianova and Sigma. Differential staining of extra- and intracellular B. henselae and CLSM (Leica TCS NT scanning confocal microscope) was performed as described previously (Kempf et al., 2000). Briefly, the paraformaldehyde (3.75%, pH 7.4)-fixed cells were washed in PBS at the beginning and after each incubation step. For differential staining of extra- and intracellular B. henselae, fixed cells were incubated sequentially with rabbit anti-B. henselae antiserum (1 h), followed by FITC-conjugated goat anti-rabbit IgG antibodies (1 h). Then, cells were permeabilized by incubation with 0.1% Triton X-100 in PBS (15 min), washed and incubated with rabbit anti-B. henselae antiserum (1 h) followed by Cy5-conjugated goat anti-rabbit antibodies mixed with TRITC-labelled phalloidin (1 h). Finally, the slides were analysed with a Leitz DM RBE microscope. Three different fluorochromes could be detected simultaneously with three different photomultipliers and represented the green (FITC), red (TRITC or Cy-3) and blue (Cy-5) channels.The corresponding images were processed digitally with photoshop 7.0 (Adobe Systems).

Cell proliferation assay

HUVECs were cultured for 72 h in 96-well dishes at 5.0 × 103 cells 200 µl−1 for [3H]-thymidine uptake assays or in 24-well dishes at 2.5 × 104 cells ml−1 for phase-contrast microscopy in M-199 medium (Biochrom) containing various amounts of recombinant human VEGF (R & D Systems; 0.1 pg ml−1, 1.0 pg ml−1, 10.0 pg ml−1 and 100 pg ml−1) or filtered (0.2 µm) conditioned and 1:10 diluted medium from B. henselae-infected EA.hy 926 cells. For blocking VEGF activity, conditioned media were preincubated for 1 h with VEGF-neutralizing antibodies (10.0 µg ml−1; R & D systems). For [3H]-thymidine assays, cells were incubated with 0.5 µCi of [methyl-3H]-thymidine (ICN Biomedicals) for the last 24 h. Samples were collected with a cell harvester, and [3H]-thymidine uptake was measured in a liquid scintillation counter (Betaplate; Wallac). For phase-contrast microscopy, cells were fixed using 3.75% PBS-buffered paraformaldehyde (pH 7.4).

Generation of pili-negative B. henselae mutants and transmission electron microscopy

Pili-negative mutants (B. henselae Pil) were produced by passaging B. henselae on Columbia agar plates until the colony morphology changed from rough to smooth indicating loss of pili expression (Batterman et al., 1995; Anderson and Neuman, 1997). Smooth colonies were subcultivated for 35 passages on Columbia agar. Expression of pili was determined by transmission electron microscopy of Columbia agar-grown B. henselae using a Zeiss EM 902 transmission electron microscope (Kempf et al., 2000). Bacterial pellets were fixed for 2 h in 4% glutaraldehyde in a 0.05 M phosphate-buffered solution containing 0.15 M NaCl at pH 7.3 at room temperature. Post-fixation was based on 1% osmium tetroxide containing 1% potassium dichromate in 0.85% NaCl at pH 7.3 for 45 min. After embedding in glycide ether, the blocks containing cells were cut using an ultramicrotome (Ultracut; Reichert). Ultrathin sections (80 nm) were stained with 0.5% uranyl acetate for 10 min at 30°C and 2.7% lead citrate for 5 min (Ultrastainer; LKB) at 20°C. Grids were examined using a Zeiss EM 902 transmission electron microscope operating at 80 kV, at magnifications between ×2000 and ×500 000.

Bacterial adhesion and invasion assays

Bacterial adhesion was quantified 30 min after infection by osmotic lysis of host cells. Cell culture supernatants were removed gently, cells were washed extensively with culture medium, and osmotic lysis was performed to calculate the total amount of bacteria. For this purpose, 900 µl of sterile water was added. Additionally, cells were disrupted using a 1 ml syringe and a 0.40 mm × 20 mm needle. Cell lysates were resuspended with 10× PBS to overcome osmotic lysis, and bacterial numbers were determined by plating serial dilutions on Columbia agar.

Invasion (2 h) and intracellular presence (72 h) were determined by gentamicin kill assays (Mehock et al., 1998; Kempf et al., 2000). Cell culture supernatants were removed gently, cells were washed extensively with culture medium, and gentamicin (100 µg ml−1) was added for 3 h to kill extracellular bacteria. Then, cells were washed extensively to remove gentamicin, and osmotic lysis was performed as described above.

VEGF immunohistochemistry

For the detection of VEGF in patient specimens, histologically and polymerase chain reaction (PCR)-confirmed (Dauga et al., 1996) paraffin-embedded specimens of BA (Schlupen et al., 1997) and BP were stained with a VEGF-specific antibody (A-20; SantaCruz Biotechnology) using the alkaline phosphatase anti-alkaline phosphatase technique with haematoxilin counterstaining according to the manufacturer's instructions. Histologically and PCR-confirmed Kaposi's sarcoma was stained with a VEGF-specific antibody (A-20) as a positive control and with an irrelevant antibody (anti-Y. enterocolitica, polyclonal) as a negative control.

Statistical analysis

All experiments were performed at least three times and gave comparable results. Differences between mean values of experimental and control groups were analysed by Student's t-test. A P-value of < 0.05 was considered to be statistically significant.


We thank Jürgen Heesemann, Wolf-Dietrich Hardt, Rainer Haas, Sonja Behrendt, Guntram Grassl and Niclas Hitziger for their suggestions, and Hyan-Sook Trogisch and Elfriede Januschke for technical help. This work was supported by a grant from the Münchner Medizinische Wochenschrift and from the University of Tübingen (Fortuene Programm).