Distinct activities of Bartonella henselae type IV secretion effector proteins modulate capillary-like sprout formation

Authors


*E-Mail christoph.dehio@unibas.ch; Tel. (+41) 61 267 2140; Fax (+41) 61 267 2118.

Summary

The zoonotic pathogen Bartonella henselae (Bh) can lead to vasoproliferative tumour lesions in the skin and inner organs known as bacillary angiomatosis and bacillary peliosis. The knowledge on the molecular and cellular mechanisms involved in this pathogen-triggered angiogenic process is confined by the lack of a suitable animal model and a physiologically relevant cell culture model of angiogenesis. Here we employed a three-dimensional in vitro angiogenesis assay of collagen gel-embedded endothelial cell (EC) spheroids to study the angiogenic properties of Bh. Spheroids generated from Bh-infected ECs displayed a high capacity to form sprouts, which represent capillary-like projections into the collagen gel. The VirB/VirD4 type IV secretion system and a subset of its translocated Bartonella effector proteins (Beps) were found to profoundly modulate this Bh-induced sprouting activity. BepA, known to protect ECs from apoptosis, strongly promoted sprout formation. In contrast, BepG, triggering cytoskeletal rearrangements, potently inhibited sprouting. Hence, the here established in vitro model of Bartonella- induced angiogenesis revealed distinct and opposing activities of type IV secretion system effector proteins, which together with a VirB/VirD4-independent effect may control the angiogenic activity of Bh during chronic infection of the vasculature.

Introduction

The feline-adapted zoonotic pathogen Bartonella henselae (Bh) and the human-specific pathogens Bartonella bacilliformis (Bb) and Bartonella quintana (Bq) share the remarkable capacity to cause vasoproliferative tumour lesions in humans. In case of Bh these lesions are known as bacillary angiomatosis/peliosis (BA/BP) and are predominately observed in immuno-compromised patients (Dehio, 2003; Dehio, 2005). BA manifests as reddish-brown papules on the skin – an accumulation of immature blood vessels with misshapen endothelial cells (ECs), a mixed inflammatory infiltrate and numerous bacteria associated with proliferating ECs (Chian et al., 2002). BP is characterized by vascular proliferation in liver and spleen, which results in the formation of blood-filled cysts. Similar to BA, bacteria are found in association with proliferating ECs (Tappero et al., 1993). These bacteria-triggered vascular proliferations are reminiscent of tumour angiogenesis, the pathological process of the formation of new capillaries out of pre-existing vessels. Angiogenesis is a highly orchestrated multistep process involving EC activation, the degradation of extracellular matrix and basement membranes, EC proliferation, EC migration and invasion into the surrounding matrix, and finally the formation of tubular structures to build immature vessels later stabilized by tight interactions with smooth muscle cells and pericytes (Carmeliet, 2003). Unlike normal vessels, the vessels triggered by tumour angiogenesis are not properly matured and will regress upon deprival of the pro-angiogenic signals released by the tumour (e.g. overproduction of vascular endothelial growth factor, VEGF). Likewise, Bartonella-triggered vessels do not mature properly, and eradication of the bacteria colonizing these newly formed vessels by antibiotic treatment leads to vessel regression (Koehler and Tappero, 1993). Thus, the Bartonella-induced vascular proliferations are benign and depend on the continuous presence of bacteria (Dehio, 2004), which apparently produce angiogenic factors that maintain the angiogenic process.

Most of the current knowledge on Bartonella–EC interaction is derived from two-dimensional cell culture infection assays with Bh. The VirB/VirD4 type IV secretion system (T4SS) and the thereby translocated Bartonella effector proteins (Beps) of Bh mediate most of the cellular phenotypes associated with infection of human umbilical vein endothelial cells (HUVECs), such as: (i) cytoskeletal actin rearrangements leading to the internalization of a large bacterial aggregate by a unique structure called the invasome, (ii) NF-κB-dependent pro-inflammatory activation and (iii) inhibition of apoptosis (Schmid et al., 2004; Schulein et al., 2005). BepG was recently reported to trigger invasome formation (Rhomberg et al., 2009). The capacity of Bh to prevent EC apoptosis is dependent on BepA and could contribute indirectly to vasoproliferative growth by enhancing cell survival (Kirby and Nekorchuk, 2002; Schmid et al., 2006). Another prominent phenotype of Bh infection of ECs is a direct mitogenic stimulation, which is VirB/VirD4/Bep-independent and is even counterbalanced by the VirB/VirD4 system at a high multiplicity of infection (moi) (Schmid et al., 2004). The activation of a pro-inflammatory response in ECs, which occurs both in a VirB/VirD4-dependent (Schmid et al., 2004) and -independent manner (Fuhrmann et al., 2001), may lead to the recruitment of monocytes, which upon activation by Bh infection release pro-angiogenic factors, such as VEGF, that promote EC proliferation in a paracrine manner (Kempf et al., 2001; Resto-Ruiz et al., 2002). The non-fimbrial adhesin BadA (Bartonella adhesin A) is considered to trigger the secretion of VEGF by activating hypoxia-inducible factor-1 (Riess et al., 2004). In summary, the current model for Bartonella-triggered vascular tumour formation proposes for Bh-infected ECs a direct stimulation of proliferation, inhibition of apoptosis and the activation of a paracrine loop of pro-angiogenic factors such as VEGF released from infected monocytic cells (Dehio, 2005).

A number of rodent animal models have been used to study different aspects of the Bartonella infection cycle and the general immune response to infection (Arvand et al., 2001; Koesling et al., 2001; Schulein et al., 2001; Schulein and Dehio, 2002; Resto-Ruiz et al., 2003). With the exception of one report (Velho et al., 2002), no publication indicated the appearance of vasoproliferative lesions in these animal models. Due to the absence of a suitable animal model, appropriate in vitro models of angiogenesis may represent an alternative to study the molecular basis of Bartonella-triggered vasoproliferation. Three-dimensional cell culture model offer an option to the complexity of live tissue experiments, providing a defined set-up where physiological cell–cell and cell–extracellular matrix interactions are better mimicked than in conventional two-dimensional cultures (Pampaloni et al., 2007). Bh was shown to be able to promote survival of EC cords in type I collagen matrix as well as matrix invasion, survival and tubular differentiation of single embedded cells (Kirby, 2004). The in vitro cord formation assay measures the ability of cells to form a web-like network of interconnected cells (cord) and is believed to model a late morphogenic step of new vessel formation (Goodwin, 2007), whereas formation of capillary networks originating from single gel-embedded cells might rather represent a model to study the process of vasculogenesis – the de novo formation of a vessel (Davis et al., 2002). Neither of these models does really reflect sprouting angiogenesis originating from the confluent endothelium of the pre-existing vessel. Systems aiming to mimic a vessel in vitro are based on focal delivery of spheroidal aggregates of EC from which sprouting can occur. Embedded in matrices such as type I collagen and stimulated with growth factors, for example, VEGF, those spheroids display radial sprouting. Cross sections of collagen gels showed numerous capillary-like structures forming a true lumen. Beyond a critical spheroid density, capillary sprouts from neighbouring spheroids will even form a complex anastomosing capillary-like network (Korff and Augustin, 1999).

Here we adapted this three-dimensional in vitro angiogenesis assay of collagen gel-embedded HUVEC spheroids to study the angiogenic properties of Bh involved in pathogen-triggered vasoproliferation. Bh wild type was found to activate HUVEC spheroids and to trigger radial outgrowths. Different VirB/VirD4 T4SS effectors of Bh individually expressed in trans in the effector-less mutant ΔbepA–G had distinct and opposing effects on spheroid sprouting, suggesting that these effectors play a modulatory role contributing to the regulation of the Bh angiogenic activity in the course of human infection.

Results

The Bh VirB/VirD4/Bep system modulates in vitro angiogenesis of collagen gel-embedded HUVEC spheroids

To establish an in vitro assay for studying the angiogenic activity of Bh we adapted the in-gel-based three-dimensional in vitro angiogenesis model described by Korff and Augustin (1998). A defined number of HUVECs (400 cells) cultured as a hanging drop in methocel-containing EGM medium for 24 h were found to form a spheroid of similar size, regardless whether cells were not infected or pre-infected for 24 h with an moi of 300 bacteria per cell. Noteworthy, compared with the typical regular spheroid morphology formed by HUVECs that were uninfected or pre-infected by the isogenic deletion mutants ΔvirB4 (deficient for a functional T4SS) (Schmid et al., 2004) or ΔbepA–G (lacking the entire set of translocated effector proteins) (Schulein et al., 2005), HUVEC pre-infection with wild type bacteria (a spontaneous streptomycin-resistant variant of the well-characterized clinical isolate and typing strain ATCC49882T) (Schmid et al., 2004) resulted in spheroids with a less regular boundary and less compact morphology. This is illustrated in Fig. 1A by bright field imaging or maximum intensity z-projections of confocal microscopy stacks of spheroids stained with the EC marker CD31. To study the ability of Bh to manipulate sprouting, these spheroids were embedded in collagen gels and after 24 h the outgrowth of sprouts representing capillary-like structures (Korff and Augustin, 1999) was evaluated. Quantification of in vitro angiogenesis was assessed by measuring the cumulative sprout length (CSL) of individual spheroids (Fig. 1G, determined as the mean ± SD of 10 spheroids per experimental group). Spheroids of uninfected HUVECs embedded in collagen gel showed a very low level of spontaneous sprouting (Fig. 1B and G), but were readily responsive to exogenous addition of VEGF (25 ng ml−1), resulting in an approximately fivefold increased CSL (Fig. 1F and G). Compared with uninfected control spheroids, we observed a higher sprouting activity of spheroids formed from HUVECs pre-infected with Bh wild type (Fig. 1C and G), displaying an increase in CSL up to levels comparable to the stimulation by VEGF. Compared with wild type, the sprouting activity was reduced about twofold in spheroids of HUVECs pre-infected with the isogenic deletion mutants ΔvirB4 (Fig. 1D and G) or ΔbepA–G (Fig. 1E and G). Hence, Bh wild type showed a clear VirB/VirD4- and Bep-dependent angiogenic activity on top of a basal level of VirB/VirD4/Bep-independent sprout formation. Similar results were obtained when bacteria were added at the stage of spheroid formation (data not shown). In contrast, the addition of bacteria onto the collagen gel immediately after embedding of spheroids had no stimulatory effect on sprouting, probably as the non-motile Bh cannot sufficiently interact with the spheroids embedded in the collagen matrix (data not shown). Strikingly, spheroids of wild type infected HUVECs displayed similar, but distinct sprout morphology in comparison with those induced by VEGF (Fig. 2 and Fig. S1). Despite showing rather similar average sprout numbers per spheroid (Fig. 2C), sprouts triggered by Bh wild type were generally shorter than those induced by VEGF (Fig. 2D). This was reflected by the migration distance of the most distal nuclei from the spheroid surface (Fig. 2F). Furthermore, whereas sprouts from spheroids stimulated by VEGF frequently consisted of two nuclei migrating into the collagen gel, those projected from spheroids of wild-type infected HUVECs typically contained only a single nucleus (Fig. 2E).

Figure 1.

The Bh VirB/VirD4/Bep system modulates sprouting angiogenesis of collagen gel-embedded EC spheroids.
A. The HUVECs where infected with an moi of 300 with the indicated strains 24 h before a defined number of cells was suspended in culture medium containing methylcellulose and grown overnight as hanging drops. Spheroids were whole-mount stained with a mouse anti-human CD31-specific monoclonal antibody and detected with an anti-mouse secondary antibody conjugated to AlexaFluor488. Bright field image and maximum intensity z-projections of stacks acquired by confocal spinning disc microscopy are shown.
B–G. Spheroids out of uninfected HUVECs or HUVECs infected with the indicated strains were embedded in collagen and control uninfected spheroids were stimulated with VEGF (25 ng ml−1).
G. Quantification of in vitro angiogenesis of collagen-embedded spheroids after 24 h by calculating the cumulative length of sprouts from individual spheroids. Shown are means/SD of 10 spheroids per experimental group. Differences between experimental groups were analysed by Student's t-test. *P < 0.05, **P < 0.01 were considered statistically significant.
B–F. Bright field images of 1-day-old collagen-embedded HUVEC spheroids. Shown are representative images demonstrating strong capillary-like sprouting of spheroids upon stimulation with VEGF (F), as well as VirB/D4 T4SS-dependent sprouting of spheroids out of Bh wild type-infected HUVECs (C). The bar represents 100 μm.

Figure 2.

Bh wild type induces sprout outgrowths distinguishable from those induced by VEGF or BepA. HUVECs were preloaded with the live dye CFSE before infection. Spheroids from uninfected HUVECs or HUVECs infected with the indicated strains were embedded in collagen and control spheroids were stimulated with VEGF (25 ng ml−1). After 24 h collagen gels were incubated with the DNA stain DAPI.
A. Stacks were acquired by confocal microscopy. Shown are maximum intensity z-projections.
B. Quantification of in vitro angiogenesis of collagen-embedded spheroids after 24 h by calculating the cumulative length of sprouts from individual spheroids.
C. Average number of sprouts per spheroid.
D. Quantification of sprouts longer than 100 μm.
E. Number of nuclei counted per sprout of a spheroid.
F. Quantification of the most distal nuclei that migrated more than 50 μm away from the spheroid surface. Shown are means ± SD of 10 spheroids (B–D), respectively, five spheroids per experimental condition (E and F). Differences between experimental groups were analysed by Student's t-test (*P < 0.05, **P < 0.01).

Pro- and anti-angiogenic activities of Beps

To investigate the role of translocated Bep effectors in VirB/VirD4-mediated angiogenesis, we performed assays using isogenic strains of the effector-less mutant ΔbepA–G expressing individual bep genes (bepA to bepG) on a plasmid in trans. Spheroids were generated and embedded into collagen gels. 24 h after embedding, sprouting was quantified by measuring the CSL (Fig. 3A). Expression of BepA, which is involved in protection of HUVECs from apoptosis (Schmid et al., 2006), showed a strong sprout-promoting activity (Fig. 3A and F), with a threefold increase in CSL compared with the ΔbepA–G mutant (Fig. 3A and E) and a twofold increase compared with wild type (Fig. 3A and D). We also observed a moderate increase of sprouting for the ΔbepA–G mutant expressing bepD in trans, with a 1.5-fold increased CSL in comparison with the ΔbepA–G mutant (Fig. 3A and I). BepD is known to become tyrosine-phosphorylated upon translocation into host cells (Schulein et al., 2005) and possibly interferes with host cell signalling processes (Pulliainen and Dehio, 2009; Selbach et al., 2009). The ΔbepA–G mutant expressing bepG in trans displayed a marked reduction in CSL of about sixfold in comparison with the ΔbepA–G mutant and almost ninefold compared with wild type (Fig. 3A and L). BepG is involved in actin rearrangements leading to invasome formation (Rhomberg et al., 2009). No statistically significant differences in CSL were observed for the ΔbepA–G mutant strains expressing in trans either bepB, bepC, bepE or bepF (Fig. 3A, G, H, J and K), for which effector function is the topic of current research.

Figure 3.

Modulation of spheroid sprout formation by Bh mutant strains translocating individual Bep effectors. Spheroids out of uninfected HUVECs or HUVECs infected with Bh wild type or the effector-less mutant ΔbepA–G trans-complemented with single-bep genes were embedded in collagen and control was stimulated with VEGF (25 ng ml−1).
A. Quantification of in vitro angiogenesis of collagen-embedded spheroids after 24 h by calculating the cumulative length of sprouts from individual spheroids. Shown are means/SD of 10 spheroids per experimental group. Differences between experimental groups were analysed by Student's t-test. *P < 0.05, **P < 0.01 were considered statistically significant.
B–L. Corresponding representative bright field images of 1-day-old gel-embedded spheroids.

Sprout outgrowth induced by BepA was found to be very similar to sprouting triggered by VEGF (Fig. 2 and Fig. S1), resulting in a comparable frequency of long sprouts (Fig. 2D), number of nuclei per sprout (Fig. 2E) and migration distance of distal nuclei (Fig. 2F).

Confirmative evidence for the data obtained with ΔbepA–G mutant strains expressing individual Bep effectors in trans was obtained by analysing a set of isogenic mutants carrying in-frame deletions of genes encoding individual Bep effectors (Fig. 4). Compared with wild type, the ΔbepA mutant showed a threefold-reduced CSL (Fig. 4A and G). We measured a slight decrease of 1.3-fold in CSL for the ΔbepD mutant (Fig. 4A and I), whereas the ΔbepG stimulated a statistically significant 1.3-fold increase of the CSL in comparison with wild type (Fig. 4A and K). Loss of bepG, though, did not result in CSL levels as triggered by BepA in the ΔbepB–G mutant (Fig. 4A and F), suggesting that additional factors might be involved in inhibition of sprouting. Deletion of bepC (Fig. 4A and H), bepE (data not shown) or bepF (Fig. 4A and J) had no significant effect on CSL compared with wild type. In conclusion, BepA and BepD showed distinct pro- and BepG a marked anti-angiogenic activity in the established three-dimensional in vitro angiogenesis assay.

Figure 4.

Modulation of spheroid sprout formation by Bh mutant strains deficient for the translocation of individual Bep effectors. Spheroids out of uninfected HUVECs or HUVECs infected with Bh wild type, the effector-less mutant ΔbepA–G or the indicated in-frame deletion mutants of individual bep genes were embedded in collagen and control was stimulated with VEGF (25 ng ml−1).
A. Quantification of in vitro angiogenesis of collagen-embedded spheroids after 24 h. The cumulative length of sprouts of individual spheroids was calculated. Shown are means/SD of 10 spheroids per experimental group. Differences between experimental groups were analysed by Student's t-test. *P < 0.05, **P < 0.01 were considered statistically significant.
B–K. Corresponding representative bright field images of 1-day-old gel-embedded spheroids.

Most of the single-gene-deletion mutants (i.e. ΔbepA, ΔbepD, ΔbepE and ΔbepF, bright field images, Fig. 4G, I and J) also stimulated the altered sprout morphology as illustrated for wild type (Fig. 2). In contrast, the sprouts stimulated by ΔbepC and ΔbepG deletion mutants had the typical sprout appearance as previously described for VEGF stimulation of uninfected spheroids (bright field images, Fig. 4H and K). Sprouts induced by the expression of individual Bep effectors in the ΔbepA–G mutant background showed as well a morphology comparable to those promoted by VEGF (Fig. 2, illustrated for BepA, bright field images, Fig. 3F–L). The slightly altered sprout morphology triggered by wild type bacteria thus appears to depend on the translocation of multiple Bep effectors, including at least BepC and BepG.

The C-terminal part of BepA is sufficient to promote sprouting

BepA is sufficient to mediate protection of ECs against apoptosis. Moreover, deletion of its N-terminal FIC domain [filamentation induced by cyclic AMP (cAMP), a domain of unknown molecular function] demonstrated that the BID (Bep intracellular delivery) domain and the proximal positively charged C-terminal tail, which together constitute a functional signal for effector translocation (Schulein et al., 2005), are enough for the full anti-apoptotic activity. In accordance with this finding, also the Bq orthologue BepA2, which due to an internal stop codon lacks an FIC domain, is inhibiting EC apoptosis (Schmid et al., 2006). To elucidate whether the prominent sprout-promoting activity of BepA is also independent of the FIC domain, we compared the ability of full-length BepA, a truncated version harbouring only the BID domain plus positively charged C-terminal tail (tBepA encoded by plasmid ptbepA), and the full-length Bq orthologue BepA2 (encoded by plasmid Bq pbepA2) to promote sprouting of spheroids. BepC and a truncated version (tBepC encoded by plasmid ptbepC) were used as negative controls. Spheroids formed form HUVECs infected with the ΔbepA–G mutant complemented in trans with the indicated plasmids were embedded in collagen and sprout formation was quantified after 24 h (Fig. 5A). Similar to full-length BepA, the truncated version of BepA, as well as the Bq orthologue BepA2, resulted in an increased CSL, whereas infection with ΔbepA–G mutant strains expressing either full-length BepC or truncated BepC in trans had no effect on sprouting.

Figure 5.

Ectopic expression of BepA promotes sprouting and BepA BID domain plus C-tail is sufficient to promote sprouting of infected HUVECs. Spheroids out of uninfected, respectively, untransfected HUVECs, HUVECs infected with the indicated strains or transfected with the indicated constructs were embedded in collagen and control was stimulated with VEGF (25 ng ml−1).
A and B. Quantification of in vitro angiogenesis of collagen-embedded spheroids after 24 h. The cumulative length of sprouts of individual spheroids was calculated. Shown are means/SD of 10 spheroids per experimental group. Differences between experimental groups were analysed by Student's t-test. *P < 0.05, **P < 0.01 were considered statistically significant.
A. N-terminally truncated BepA (tBepA) harbouring BID domain plus C-tail (sufficient for anti-apoptosis) is enough to promote sprouting of spheroids. Moreover the Bartonella quintana homologue BepA2, encoded by the downstream ORF of the split bepA locus containing only the BID domain and C-tail, also strongly promoted sprouting.
C–F. Representative bright field images of 1-day-old gel-embedded spheroids for the ectopic expression of GFP–BepA fusion. HUVECs transfected with gfp-BepA (plMS21; B and F) show increased capillary-like sprouting compared with untransfected cells or cells transfected with control gfp (pWay21; B and E).

Ectopic expression of full-length BepA fused to GFP was previously shown to be sufficient to protect ECs from apoptosis (Schmid et al., 2006). To test whether BepA alone is also enough to promote sprouting in the established three-dimensional angiogenesis assay in the absence of any additional bacterial factor, we transfected HUVECs prior to spheroid formation for 48 h with plasmids encoding GFP (gfp) or a GFP–BepA fusion (gfp-bepA), or left cells untransfected for this period (see Fig. S2 for expression of gfp and gfp–bepA at 48 h post transfection). Spheroids formed by those cell populations were embedded in collagen gels and in vitro angiogenesis was quantified 24 h later by measuring the CSL (Fig. 5B). Ectopic expression of the GFP–BepA fusion (Fig. 5B and F) lead to a statistically significant increase of CSL of about twofold compared with spheroids formed from either GFP-expressing or untransfected HUVECs (Fig. 5B, C and E). Conclusively, paralleling what was previously described for the anti-apoptotic activity (Schmid et al., 2006), the BID domain plus C-terminal tail of BepA was sufficient to promote sprouting.

BepG interferes with sprout formation from spheroids

Invasome-mediated entry of Bh into ECs (Dehio et al., 1997) depends on the VirB/VirD4 T4SS (Schmid et al., 2004) and thereby translocated effector(s) (Schulein et al., 2005). A ΔbepA–G mutant expressing BepG was recently found to promote the actin cytoskeletal rearrangements characteristic for the invasome structure, whereas no other single Bep elicits a similar actin remodelling (Rhomberg et al., 2009). To test the impact of this F-actin modulating effector on spheroid sprouting, we infected HUVECs with BepG-expressing strains of either the ΔbepA–G mutant background (not expressing BepA) or the ΔbepB–G mutant background (expressing BepA). As negative control we used the ΔbepA–G and ΔbepB–G mutant backgrounds expressing BepC, which alone does not interfere with spheroid sprouting. Spheroids formed from such infected ECs were embedded in collagen and in vitro angiogenesis was quantified after 24 h (Fig. 6A). Overexpression of BepG in the ΔbepA–G mutant led to a marked reduction of CSL in comparison with the ΔbepA–G mutant (Fig. 6A and G), while BepC did not alter the basal VirB/VirD4-independent sprouting activity (Fig. 6A and F). Similarly, when BepG was overexpressed in the ΔbepB–G background, this effector reduced the BepA-induced increase in CSL more than threefold (Fig. 6A and J), whereas BepC had no influence (Fig. 6A and I). Thus, BepG antagonizes the BepA-triggered sprouting activity, but also generally interfered with VirB/VirD4-independent sprouting in the in-gel-based three-dimensional in vitro angiogenesis model.

Figure 6.

BepG is interfering with sprouting of spheroids. Spheroids out of uninfected HUVECs or HUVECs infected with the indicated strains were embedded in collagen and control was stimulated with VEGF (25 ng ml−1).
A. Quantification of in vitro angiogenesis of collagen-embedded spheroids after 24 h. The cumulative length of sprouts of individual spheroids was calculated. Shown are means/SD of 10 spheroids per experimental group. Differences between experimental groups were analysed by Student's t-test. *P < 0.05, **P < 0.01 were considered statistically significant.
B–J. Corresponding representative bright field images of 1-day-old gel-embedded spheroids.

Discussion

In this study we have addressed the angiogenic potential of Bh by employing a three-dimensional in vitro sprouting angiogenesis assay of collagen-embedded HUVEC spheroids, which allows the quantitative assessment of both pro- and anti-angiogenic factors (Korff and Augustin, 1999). We could show that Bh triggers an angiogenic response, which is manifested by an increase in CSL up to a similar range as induced by the potent angiogenic factor VEGF. Formation of these sprouts was in part dependent on a functional VirB/VirD4 T4SS. Importantly, we observed that three out of the seven known VirB/VirD4-translocated Bep effectors had distinct and in part opposing activities in this assay. BepA showed a marked and BepD a relatively moderate pro-angiogenic effect, while BepG displayed a prominent anti-angiogenic activity.

BepA is known to inhibit apoptosis in HUVECs via a rise in the cytosolic concentration of the second messenger cAMP (Schmid et al., 2006). As reported for the cAMP-dependent anti-apoptotic activity, an N-terminally truncated version of BepA harbouring only the BID domain and the C-terminal tail sequence was sufficient to promote sprouting. Likewise, the ectopic expression of a GFP–BepA fusion protein alone was sufficient to trigger sprout formation, indicating that the same protein domains and effector mechanisms could be involved in anti-apoptosis and sprout induction. Noteworthy, BepB and BepC do not inhibit EC cell death, despite representing highly conserved paralogues of BepA (Schmid et al., 2006). Likewise, neither BepB nor BepC alone was able to promote sprouting. Yet it remains to be demonstrated whether indeed the cAMP-mediated anti-apoptotic activity of BepA is directly linked to its strong angiogenic stimulus.

BepD displayed a significant, although much weaker, pro-angiogenic activity than BepA. Upon translocation in ECs, this effector becomes phosphorylated on specific tyrosine residues (Schulein et al., 2005) that may serve as specific docking sites for host cell signalling proteins bearing phosphotyrosine binding or Src-homology 2 domains (Pulliainen and Dehio, 2009; Selbach et al., 2009). This signalling complex could then modulate angiogenesis. Interestingly, the effectors BepE and BepF that bear similar tyrosine phosphorylation motifs did not significantly affect sprout formation, which indicates the specificity of the pro-angiogenic signalling events triggered by BepD. A better understanding of BepD effector function in general might shed light as well on the pathways involved in sprout induction.

BepG was the only effector that demonstrated a marked anti-angiogenic activity. BepG was previously shown to be able to promote actin rearrangements required for invasome-mediated invasion, a specific mode of Bh entry into ECs (Dehio et al., 1997). This F-actin-dependent BepG-triggered sequence of events appears to inhibit the uptake of individual bacteria via conventional endocytosis (Rhomberg et al., 2009). Notably, invasome formation is associated with a reduction in EC motility (Dehio et al., 1997), which may represent a link of the BepG-dependent processes of invasome-medicated invasion and the inhibition of in vitro angiogenesis as reported here. It is unknown how BepG interferes with endocytosis and migration in ECs; however, these activities may relate to the colocalization of BepG with F-actin (Rhomberg et al., 2009).

Further to the pro- and anti-angiogenic activities of individual Bep effectors, we observed that the interaction of several effectors can result in modified sprout morphology. In contrast to sprouts stimulated by VEGF or BepA alone, sprouts induced by Bh wild type were shorter and usually comprised only one nucleus, which indicates a reduced capacity of cell migration into the collagen gel. The altered sprouting phenotype seemed to require multiple Bep proteins, including BepG, which by itself inhibits sprout formation, and BepC, which by itself does not affect sprout formation, and probably in addition at least one sprout-inducing effector, such as BepA.

It is unclear whether Bartonella-induced vasoproliferation represents a cause or rather a consequence of bacterial colonization of the human host. Bacterial factors may directly target the angiogenic switch in order to trigger proliferation as part of a dedicated infection strategy that expands the specific host cell habitat (Kempf et al., 2002). However, vasoproliferation may as well represent a ‘biological accident’ induced by the complex physiological changes resulting from chronic vascular colonization. It is likely that a yet unidentified immunological parameter on the host side influences the different clinical manifestations of Bartonella infection, as BA and BP lesions triggered by Bh or Bq infections are almost exclusively observed in patients with a severe immuno-suppression; even so, chronic infections by these pathogens also occur in immuno-competent individuals. Bb on the contrary regularly causes VP lesions in immuno-competent humans (Chian et al., 2002). Based on these clinical observations, it is tempting to speculate that all three vasoproliferative Bartonella species encode vascular colonization factors that, either directly or indirectly, contribute to a potent pro-angiogenic response, while Bh and Bq may in addition encode anti-angiogenic factors to compensate this pro-angiogenic activity, at least in individuals with an intact immune response. The Bep effector proteins translocated by the VirB/VirD4 T4SS of Bh and Bq represent strong candidates for such compensating modulators of the angiogenic response. First, the VirB/VirD4/Bep system was horizontally acquired and evolved as host adaptation system in the ‘modern’Bartonella lineage comprising Bh and Bq, while it is absent from the ancestral lineage solely represented by Bb (Saenz et al., 2007; Dehio, 2008). Second, the VirB/VirD4 system and its translocated effector can suppress a VirB/VirD4-independent strong proliferative response of ECs (Schulein et al., 2005; Schmid et al., 2006). Third, the here observed prominent but adverse effects of BepA and BepG on spheroid sprouting point towards a modulatory role of these effectors on angiogenesis.

Taken together, we provide in this study a novel model of Bh-triggered angiogenesis and an initial characterization of the distinct activities of VirB/VirD4-translocated Bep effectors in the triggered angiogenic response. The adapted collagen gel-embedded HUVEC spheroid model represents a validated quantitative in vitro assay of sprouting angiogenesis that in further studies will allow elucidating the molecular basis of the described pro- and anti-angiogenic activities of BepA, BepD and BepG. Noteworthy, a recent report demonstrated that HUVEC spheroids as used here for in vitro sprouting angiogenesis can also be used in vivo (Alajati et al., 2008). Using this novel murine xenotransplantation model with HUVEC spheroids pre-infected with Bh may allow in future studies to establish the badly needed animal model to study Bartonella-triggered angiogenesis in vivo.

Experimental procedures

Reagents

Recombinant human VEGF-A165 was produced in Sf9 insect cells (Tumor Biology Centre, University of Freiburg). Antibiotics and carboxymethylcellulose (methocel, 4000 centipoises) were from Sigma (http://www.sigmaaldrich.com). Collagen type I was isolated from rat tail tendons (Augustin, 2004). CellTrace CFSE was purchased from Molecular Probes (http://www.invitrogen.com). The monoclonal mouse anti-human CD31 antibody (clone JC70A) was purchased from DAKO (http://www.dako.com). Monoclonal mouse anti-human VE-cadherin (clone 55-7H1) was obtained from BD Pharmingen (http://www.bdbiosciences.com). Secondary AlexaFluor 488-conjugated goat anti-mouse antibody was from Molecular Probes (http://www.invitrogen.com). TRITC-phalloidin from Sigma (http://www.sigmaaldrich.com) was used to stain F-actin and DAPI from Roche (https://www.roche-applied-science.com) to stain double-stranded DNA.

Bacterial strains and growth conditions

The bacterial strains used in this study are listed in Table S1. Bartonella spp. were grown on Columbia agar plates containing 5% defibrinated sheep blood in a humified atmosphere at 35°C and 5% CO2 for 2–3 days. Strain RSE247, a spontaneous streptomycin-resistant strain of ATCC 49882T (Schmid et al., 2004), served as wild type in this study. Media were supplemented with 30 μg ml−1 kanamycin, 100 μg ml−1 streptomycin, 12.5 μg ml−1 gentamicin, and/or 500 μM isopropyl β-D-thiogalactosidase (IPTG, http://www.applichem.de) when indicated.

Genetic manipulation of Bartonella strains

Plasmids were introduced into Bartonella henseae RSE247 by conjugation from Escherichia coli using two- or three-parental matings (Dehio and Meyer, 1997; Dehio et al., 1998). Plasmids used in this study are listed in the Table S1. Oligonucleotide primers used in this study for plasmid construction are listed in Table S2. In-frame deletion mutants of B. henselae RSE247 were generated by a two-step gene replacement procedure as described (Schulein and Dehio, 2002; Schmid et al., 2004). The basic mutagenesis vector pTR1000 was described before (Schulein et al., 2005). All mutagenesis plasmids harbour a cassette with the flanking regions of the in-frame deletion in the gene(s) of interest. This cassette was generated from two PCR fragments amplified from chromosomal DNA of RSE247 as template, either by megaprime PCR or by conventional cloning.

Construction of the in-frame mutant ΔbepA has been described previously (Schmid et al., 2004).

pTR1069 used for generating a ΔbepC in-frame mutant was constructed as follows. Oligonucleotide primers prTR082 and prTR083 amplified fragment 1 (846 bp) and prTR084 and prTR085 amplified fragment 2 (753 bp, containing 256 bp of 3′ end of bepD). Both fragments were combined by megaprime PCR with oligonucleotide primers prTR082 and prTR085, resulting in a fragment of 1.59 kb carrying an in-frame deletion of 1.58 kb in bepC. By using flanking BamHI sites, the fragment was inserted into the corresponding site of pTR1000, yielding pTR1069. The use of pTR1069 for gene replacement in RSE247 resulted in the ΔbepC mutant TRB288.

pPG161 used for generating a ΔbepD in-frame mutant was constructed as follows. Oligonucleotide primers prTR041 and prTR051 amplified fragment 1 (853 bp, containing 321 bp of 5′ end of bepC) and prTR042 and prTR044 amplified fragment 2 (860 bp, containing 473 bp of 3′ end of bepE). Both fragments were combined by megaprime PCR with oligonucleotide primers prTR041 and prTR044, resulting in a fragment of 1.67 kb carrying an in-frame deletion of 1.54 kb in bepD. By using flanking BamHI sites, the fragment was inserted into the corresponding site of pTR1000, yielding pPG161. The use of pPG161 for gene replacement in RSE247 resulted in the ΔbepD mutant PGC80.

pPG162 used for generating a ΔbepE in-frame mutant was constructed as follows. Oligonucleotide primers prPG134 and prPG135 amplified fragment 1 (898 bp, containing 497 bp of 5′ end of bepD) and prPG136 and pr137 amplified fragment 2 (906 bp, containing 522 bp of 3′ end of bepF). Both fragments were combined by megaprime PCR with oligonucleotide primers prPG134 and prPG137, resulting in a fragment of 1.75 kb carrying an in-frame deletion of 1.33 kb in bepE. By using flanking BamHI sites, the fragment was inserted into the corresponding site of pTR1000, yielding pPG162. The use of pPG162 for gene replacement in RSE247 resulted in the ΔbepE mutant PGD20.

pTR1075 used for generating a ΔbepF in-frame mutant was constructed as follows. Oligonucleotide primers prTR053 and prTR054 amplified fragment 1 (736 bp, containing 356 bp of 5′ end of bepE) and prTR055 and prTR056 amplified fragment 2 (665 bp, containing 345 bp of 3′ end of bepG). Both fragments were combined by megaprime PCR with oligonucleotide primers prTR053 and prTR056, resulting in a fragment of 1.38 kb carrying an in-frame deletion of 2.45 kb in bepF. By using flanking XbaI sites, the fragment was inserted into the corresponding site of pTR1000, yielding pTR1075. The use of pTR1075 for gene replacement in RSE247 resulted in the ΔbepF mutant TRB222.

pTR1078 used for generating a ΔbepG in-frame mutant was constructed as follows. Oligonucleotide primers prTR057 and prTR058 amplified fragment 1 (713 kb, containing 339 bp of 5′ end of bepF) and prTR059 and prTR060 amplified fragment 2 (625 kb). Both fragments were combined by megaprime PCR with oligonucleotide primers prTR057 and prTR060, resulting in a fragment of 1.26 kb carrying an in-frame deletion of 2.93 kb in bepG. By using flanking XbaI sites, the fragment was inserted into the corresponding site of pTR1000, yielding pTR1078. The use of pTR1078 for gene replacement in RSE247 resulted in the ΔbepG mutant TRB223.

Cell culture

The HUVECs were isolated as described before (Dehio et al., 1997) and cultured in EGM medium (http://www.promocell.com) in a humified atmosphere at 37°C and 5% CO2.

Infection assay

The HUVECs (passages 4–7) were plated in gelatine-coated well plates at 6 × 104 ml−1 using EGM. The next day cells were washed twice with M199 with Earls salts (M199, Gibco, http://www.invitrogen.com) supplemented with 10% foetal calf serum (FCS, http://www.invitrogen.com) and unless stated differently infected with an moi of 300 bacteria per cell in M199/10% FCS/500 μM IPTG and incubated for 24 h.

HUVEC transfection

The HUVECs were transfected using the HUVEC Nucleofector Kit – OLD from Amaxa (Cat. No. VPB-1492, Amaxa, http://www.amaxa.com) following the manufacturer's guidelines. After transfection, cells were seeded into gelatine-coated six-well plates.

Generation of EC spheroids

The EC spheroids of defined cell number were generated as described previously (Korff and Augustin, 1998) with some modifications. In brief, uninfected EC monolayers or EC monolayers infected with the indicated Bh strains were trypsinized and suspended in EGM containing 20% methocel (Weber et al., 2008). Spheroids were generated by pipetting 400 ECs (in 25 μl) on quadratic Greiner Petri dishes (http://www.huberlaborworld.ch) and grown overnight as hanging drops in a humified atmosphere at 35°C and 5% CO2. Under these conditions, all suspended cells contribute to the formation of a single spheroid of defined cell number per drop.

Spheroid whole-mount staining

Staining of spheroids was performed in 2 ml Eppendorf tubes as done previously (Scharpfenecker et al., 2005). Spheroids were harvested and fixed for 1 h in PFA-fixans (0.02 mM CaCl2, 1.36% w/v saccharose, 4% PFA in PBS). The fixative was removed and spheroids were washed three times with 1 ml PBS. Permeabilization was done with 150 μl 0.1% Triton X-100 in PBS for 10 min. Spheroids were washed three times with 1 ml PBS and blocked with 300 μl 3% FCS in PBS. Spheroids were then incubated for 1 h with a mouse anti-human-CD31 antibody (1:100) diluted in 150 μl blocking solution. After three washing steps with 1 ml PBS, spheroids were then exposed to AlexaFluor 488-conjugated goat anti-mouse antibody (1:200) and DAPI (1:1000, 1 mg ml−1 stock) in 150 μl blocking solution. After 1 h, spheroids were washed tree times with 1 ml PBS and supernatants were removed. Then, one drop of mowiol (http://www.sigmaaldrich.com) was added and spheroids were transferred to glass slides and sealed with coverslips. Short pieces of human hair or parafilm were used as spacers in order not to destroy the 3D structure of the spheroids. Z-stacks were acquired using a spinning disc confocal microscope (Olympus IX81, Andor Revolution XD) and analysed using the ImageJ software (http://rsb.info.nih.gov/ij/).

Spheroid-based in vitro angiogenesis assay

The in vitro angiogenesis assay was performed as described (Weber et al., 2008). The generated spheroids were harvested and suspended in the methocel solution containing 20% FCS. Subsequently, the ice-cold collagen solution (rat tail type I collagen in 0.1% acidic acid) was neutralized by adding 10% 10-fold Medium 199 and approximately 10% 0.2 N NaOH to adjust the pH to 7.4. EC spheroid/methocel solution was mixed 1:1 with the neutralized collagen solution and 1 ml of the mixed solution containing approximately 50 EC spheroids were pipetted into individual wells of a pre-warmed 24-well plate to allow polymerization in the incubator in a humified atmosphere at 35°C, 5% CO2. After 30 min VEGF was added to an end concentration of 25 ng ml−1 by pipetting 100 μl of a 10-fold concentrated working dilution on top of the polymerized gel. Plates were incubated in a humified atmosphere at 35°C, 5% CO2 for 24 h and fixed by adding 3.7% paraformaldehyde. Pictures were taken with a Leica DM IRBE inverted microscope using a MicroMAX camera (Princeton Instruments with MetaMorph software) using the 20× objective. In vitro angiogenesis was digitally quantified by measuring the cumulative length of the sprouts (CSL) that had grown out of each spheroid using the MetaMorph software analysing 10 spheroids per experimental group and experiment.

In-gel staining of sprouting spheroids

The HUVEC monolayers were preloaded with CFSE following the manufacturer's protocol and infected with indicated Bh strains (moi 300) or left uninfected in M199/10% FCS/500 μM IPTG for 24 h. As for the spheroid angiogenesis assay EC spheroids were generated, harvested, suspended in methocel 20% FCS and mixed 1:1 with neutralized collagen. Two millilitres of the mix was distributed over the wells of an 8-well μ-slide (http://www.ibidi.de) and was allowed to polymerize. Fifty microlitres of a 10-fold concentrated working dilution of VEGF was pipetted on top of the polymerized gel. The μ-slides were incubated in a humified atmosphere at 35°C, 5% CO2 for 24 h and fixed by adding 3.7% paraformaldehyde. Gels were washed tree times with 150 μl PBS and blocked with 200 μl PBS 5% bovine serum albumin for 2 h before being incubated with DAPI in 100 μl PBS-bovine serum albumin overnight at 4°C. Gels were washed tree times with PBS and analysed by confocal spinning disc microscopy (Olympus IX81, Andor Revolution XD). 3D reconstruction of in-gel sprouting spheroids was done using image stacks and the ImageJ software.

Statistical analysis

Ten spheroids were analysed per experimental condition, except for five spheroids in nuclei migration determination. Experiments were repeated at least two times. Statistical significance was determined using Student's t-test. *P < 0.05, **P < 0.01 were considered statistically significant.

Acknowledgements

We are grateful to Arto Pulliainen for critically reading of the manuscript. We thank C. Mistl and O. Siedentopf for technical assistance. The work was supported by Grant 3100A0-109925/1 from the Swiss National Science Foundation to C.D., Grant 55005501 from the Howard Hughes Medical Institute to C.D. and a PhD fellowship from the Misrock Foundation to F.S. We acknowledge the donation of human umbilical cords from the University Women's Hospital Basel.

Ancillary