A Translocated Bacterial Protein Protects Vascular Endothelial Cells from Apoptosis

The modulation of host cell apoptosis by bacterial pathogens is of critical importance for the outcome of the infection process. The capacity of Bartonella henselae and B. quintana to cause vascular tumor formation in immunocompromised patients is linked to the inhibition of vascular endothelial cell (EC) apoptosis. Here, we show that translocation of BepA, a type IV secretion (T4S) substrate, is necessary and sufficient to inhibit EC apoptosis. Ectopic expression in ECs allowed mapping of the anti-apoptotic activity of BepA to the Bep intracellular delivery domain, which, as part of the signal for T4S, is conserved in other T4S substrates. The anti-apoptotic activity appeared to be limited to BepA orthologs of B. henselae and B. quintana and correlated with (i) protein localization to the host cell plasma membrane, (ii) elevated levels of intracellular cyclic adenosine monophosphate (cAMP), and (iii) increased expression of cAMP-responsive genes. The pharmacological elevation of cAMP levels protected ECs from apoptosis, indicating that BepA mediates anti-apoptosis by heightening cAMP levels by a plasma membrane–associated mechanism. Finally, we demonstrate that BepA mediates protection of ECs against apoptosis triggered by cytotoxic T lymphocytes, suggesting a physiological context in which the anti-apoptotic activity of BepA contributes to tumor formation in the chronically infected vascular endothelium.


Introduction
Bacterial pathogens have developed various strategies to subvert host cell functions to their benefit. In particular, intracellular bacteria have adapted mechanisms to modulate the apoptotic pathway of their host cells [1]. The resulting induction or inhibition of apoptosis is often crucial for a successful infection of the host. Pathogen-induced apoptosis can serve to eliminate key immune cells or to evade other host defenses [1]. Several bacteria elicit an inflammatory process by inducing a specific form of apoptotic cell death, which at the place of infection leads to the disruption of tissue barriers and thus may secure efficient microbial spread in the host [2,3]. In contrast, inhibition of apoptosis may be essential for intracellular pathogens to establish chronic infection. Pathogen-triggered anti-apoptosis of infected host cells facilitates a slow microbial replication process and enables persistence in the infected host. For example, the obligate intracellular pathogen Chlamydia pneumoniae degrades-by an unknown effector mechanism-pro-apoptotic BH-3 host cell proteins [4]. Rickettsia rickettsii protects invaded host cells from apoptosis by activating a nuclear factor kappa B (NFjB)-dependent survival pathway [5]. For many human pathogens, the activation and inhibition of the apoptotic machinery of the infected host cell thus has a central role during the infection process [1].
The bacterial effectors known to modulate apoptosis are mostly pro-apoptotic. The only described anti-apoptotic effector is the outer membrane protein PorB of Neisseria meningitidis [6]. PorB has been shown to interact with mitochondria, where it is thought to block apoptosis by preventing mitochondrial depolarization and cytochrome C release. Interestingly, opposite effects were reported for the PorB ortholog of N. gonorrhoea, which provokes a proapoptotic effect upon interaction with mitochondria [7,8].
Bartonellae are facultative intracellular pathogens associated with the formation of vasoproliferative tumors in humans (e.g., bacillary angiomatosis, bacillary peliosis, and verruga peruana) [9]. These vascular lesions consist of an increased number of endothelial cells (ECs), which are colonized by extracellular and intracellular bacteria. Although at least nine Bartonella species are known to infect humans, vascular proliferation is mainly caused by the species Bartonella henselae (Bh), B. quintana (Bq), and B. bacilliformis [9]. Upon EC infection in vitro, Bh and Bq are able to stimulate proliferation and inhibition of apoptosis [9][10][11]. The antiapoptotic activity of bartonellae is considered to contribute Editor: Pascale Cossart, Institut Pasteur, France synergistically to an unrelated mitogenic activity and results in the formation of vascular tumors [10,11].
In Bh, inhibition of apoptosis is dependent on a functional VirB/VirD4 system [12,13]. This type IV secretion (T4S) system is a major virulence determinant for Bartonella-EC interaction [12][13][14]. T4S systems are versatile transporters ancestrally related to bacterial conjugation machines. Different human pathogens have recruited such conjugation machineries to inject macromolecular effectors across the bacterial and host cell membranes directly into the host cell cytosol, where they alter various cellular processes [15]. Bh carries a pathogenicity island that encodes next to the virB/ virD4 locus seven putative T4S substrates, the Bartonellatranslocated effector proteins Bh BepA-G [13]. Deletion of the pathogenicity island regions encoding bepA-G abolishes the anti-apoptotic activity and all other VirB/VirD4-dependent EC phenotypes [13]. However, the contribution of individual Bh Beps to anti-apoptosis and the other cellular phenotypes is unknown. All seven Bh Beps carry in their Cterminus at least one conserved region, called the Bep intracellular delivery (BID) domain [13]. The positively charged C-terminal tail sequence together with the proximal BID domain constitute a bipartite signal for T4S. Based on a Cre-recombinase assay for translocation (CRAfT), this T4S signal was shown to be functional for VirB/VirD4-dependent transfer of four different Bh Beps (Bh BepB, Bh BepC, Bh BepD, and Bh BepF) [13]. Evidence for translocation of Bh BepA, Bh BepE, and Bh BepG is still missing.
Bh BepA, Bh BepB, and Bh BepC are paralogous proteins. A Bh BepA ortholog is found in the animal pathogen B. tribocorum (Bt) [16]. Remarkably, all BepA homologs encode in their N-terminal part a conserved FIC (filamentation induced by cyclic adenosine monophosphate [cAMP]) domain, which was proposed to be involved in bacterial cell division [17], while the putative effector function within human cells is unknown.
In this paper, we identify BepA as anti-apoptotic effector of the vasoproliferative bartonellae Bh and Bq. We show that this effector is translocated by the VirB/VirD4 system into ECs. Further, we show that the anti-apoptotic activity is confined to the BID domain of Bh BepA, which also mediates localization to the EC plasma membrane. We demonstrate that Bh BepA provokes an elevation of intracellular cAMP and the upregulation of cAMP-responsive genes, suggesting that plasma membrane-associated Bh BepA triggers cAMP production and signaling, resulting in the abrogation of apoptotic processes. Accordingly, we were able to mimic the anti-apoptotic effect of BepA by artificially increasing the intracellular cAMP level in ECs. Finally, we demonstrate that BepA can inhibit cytotoxic T lymphocyte (CTL)-mediated apoptosis of ECs.

Bh BepA Mediates Anti-Apoptosis in ECs
The anti-apoptotic activity of Bh on human umbilical vein endothelial cells (HUVECs) requires a functional VirB/VirD4 system and at least one of the putatively secreted effectors Bh BepA-G [12,13]. To identify the effector(s) required for inhibition of apoptosis, we expressed them individually in the substrate-free bepA-G mutant and assayed for anti-apoptotic activity. To this end, HUVECs were infected with the different Bh strains and apoptosis was induced by exposure to actinomycin D. Morphological examination ( Figure 1A), measuring of caspase-3/-7 activity ( Figure 1B), and flow cytometric analysis of Annexin V-and propidium iodide (PI)-stained cells ( Figure 1C) were used to monitor apoptosis. DbepA-G did not display anti-apoptotic activity as reported previously [13]. Strikingly, the expression of Bh BepA (by plasmid pBh BepA) restored the anti-apoptotic activity to wild-type level (Figure 1), whereas no other Bh Bep interfered with actinomycin D-triggered apoptosis (unpublished data). Consistently, strain DbepA carrying an in-frame deletion of bepA did not protect HUVECs from apoptosis, whereas complementation of this mutant with pBh BepA restored the anti-apoptotic activity ( Figure 1). To test for functionality of the VirB/VirD4 system in the DbepA mutant, we monitored the T4S-dependent process of invasome formation [12,18], and furthermore quantified the translocation efficiency of a reporter substrate (encoded by pRS51) using CRAfT [13]. Both assays confirmed that the VirB/VirD4 T4S system is functional in the DbepA strain (unpublished data).
Taken together, these results provide evidence that Bh requires the putative T4S substrate Bh BepA to protect HUVECs from apoptosis.

Bh BepA Is a VirB/VirD4 Substrate Translocated into ECs
CRAfT was used to demonstrate functionality of most of the C-terminal bipartite T4S signals of Bh Beps [13], while this assay was negative for the putative T4S substrate Bh BepA (unpublished data). Here, we adapted the calmodulindependent adenylate cyclase (Cya) reporter assay [19,20] as an alternative to test for translocation of Bh BepA into HUVECs. Translocation of a Cya reporter fusion is demonstrated by the increase in the intracellular cAMP level. An Nterminal FLAG-tag epitope was fused to Cya to determine the stability of fusion proteins by immunoblot analysis (Figure 2A and 2C). As positive control, the T4S signal of Bh BepD (Bh BepD 352-534 ) [13] was fused to the FLAG-Cya reporter

Synopsis
The capacity of infected host cells to die by apoptosis (programmed cell death) is critical for controlling pathogen replication and survival. Bacterial pathogens have thus developed strategies to inhibit host cell apoptosis, allowing them to preserve their cellular habitat during chronic infection. For instance, the capacity of the facultative intracellular pathogen Bartonella henselae to trigger tumor formation as a consequence of chronic infection of the human vasculature is linked to the inhibition of endothelial cell apoptosis. This study describes the identification and functional characterization of the anti-apoptotic bacterial effector protein BepA of B. henselae, which is shown to be sufficient to inhibit endothelial cell apoptosis, i.e., as triggered by activated cytotoxic T lymphocytes. Upon translocation into endothelial cells via a bacterial type IV secretion system, BepA localizes to the plasma membrane, where it triggers the production of second messenger cyclic adenosine monophosphate in quantities effective for blocking apoptosis. Strikingly, the capacity of BepA to mediate membrane localization, cyclic adenosine monophosphate production, and the resulting inhibition of apoptosis is confined to a conserved domain that originally evolved in bacteria as a signal for type IV secretion. This study thus highlights the convergent evolution of an anti-apoptotic effector protein of purely bacterial origin. (pFLAG-Cya-Bh BepD 352-534 ). To test for translocation of Bh BepA, either full-length protein (Bh BepA 1-544 ) or the Cterminal domain harboring the putative T4S signal (Bh BepA 305-544 ) was fused to FLAG-Cya. These reporter constructs were expressed either in Bh wild-type or in the DvirB4 mutant lacking a functional VirB/VirD4 system. HUVECs infected with wild-type-expressing FLAG-Cya did not display any significant increase of the cAMP level compared with wild-type without the pFLAG-Cya plasmid. In contrast, FLAG-Cya fused to either Bh BepD 352-534 , Bh BepA 1-544 , or Bh BepA 305-544 resulted in an approximately 10-fold increase of intracellular cAMP levels in HUVECs ( Figure 2B). This effect was dependent on a functional VirB/VirD4 system. In summary, we show that BepA is a VirB/VirD4 substrate that harbors a functional C-terminal T4S signal.
The BID Domain of Bh BepA Is Sufficient to Inhibit Apoptosis Next, we determined whether ectopic expression of Bh BepA in HUVECs is sufficient to mediate anti-apoptosis. Nterminal green fluorescent protein (GFP) fusions to fulllength Bh BepA or fragments thereof were constructed in an appropriate eukaryotic expression vector. As control we included the expression vector pGFP encoding just GFP ( Figure 3A). Then, 24 h after transfection of HUVECs with the different GFP fusion constructs, cells were either exposed to actinomycin D (apoptosis induction) or left untreated (noninduced control) for another 12 h. Then, cells were stained with Annexin V and PI, and the proportion of apoptotic cells among the GFP-positive cell population was quantified by flow cytometry.
HUVECs transfected with pGFP and treated with actinomycin D displayed an apoptotic rate of 13% ( Figure 3B). Fusion of full-length Bh BepA to the C-terminus of GFP (pGFP-Bh BepA 1-544 ) reduced the apoptotic population almost 4-fold to a level similar to that of the non-induced control. The same anti-apoptotic activity was observed when only the C-terminal bipartite T4S signal of Bh BepA was fused, thus missing the first 304 N-terminal amino acids encoding the FIC domain (pGFP-Bh BepA 305-544 ). In contrast, fusion of the FIC domain (pGFP-Bh BepA 1-304 ) did not result in inhibition of apoptosis. Further analysis showed that fusion of the BID domain (142 amino acids [aa]) to GFP (pGFP-Bh BepA 305-446 ) was sufficient to inhibit apoptosis, whereas fusion of the positively charged C-terminal tail plus only part of the BID domain (pGFP-Bh BepA 403-544 ) did not result in anti-apoptosis ( Figure 3B). Rather, expression of the latter construct had a pro-apoptotic effect, as indicated by Figure 2. Bh BepA Is a Genuine VirB/VirD4 T4S Substrate That Is Translocated into ECs (A) The bars indicate the parts of Bh BepA or Bh BepD that were fused to Cya. These reporter fusions were used to monitor translocation via the VirB/VirD4 system. All constructs contain an N-terminal FLAG epitope for immunological detection of the encoded fusion protein.
(B) Quantification of the amount of intracellular cAMP in HUVECs infected for 20 h with the indicated bacterial strains (MOI ¼ 300). Isogenic strains with a functional (wild-type) or non-functional (DvirB4) VirB/VirD4 T4S system were used to express the different Cya reporter constructs. Mean and SD are shown for one representative out of three independent replica experiments. (C) Steady-state FLAG-Cya fusion protein levels of the indicated Bh strains grown on IPTG-containing medium. doi:10.1371/journal.ppat.0020115.g002 the increased apoptotic cell population in the untreated sample. Taken together, these data demonstrate that ectopic expression of the BID domain as part of the bipartite T4S signal of Bh BepA in ECs is sufficient to mediate protection against apoptosis.

Anti-Apoptotic Bh BepA Constructs Associate with the Plasma Membrane
To test for the subcellular localization of Bh BepA in host cells, the generated GFP fusions ( Figure 3A) were ectopically expressed in HEK293T cells for 24 h, followed by immunocytochemical staining for the cell surface with Texas Redconjugated wheat germ agglutinin (WGA). Samples were analyzed by confocal microscopy by taking images in the xyplane. To better distinguish between cytoplasmic and membrane-associated localization, we also captured images in an xz-plane (indicated in Figure 3C by dashed lines). Ectopically expressed GFP localized to the cytoplasm of HEK293T cells, whereas the GFP-Bh BepA 1-544 fusion localized to the plasma membrane ( Figure 3C). Interestingly, all fusion proteins with anti-apoptotic activity localized to the plasma membrane, while those that did not confer protection localized primarily to the cytoplasm.
Fractionation of post-nuclear extracts of the transfected HEK293T cells by ultracentrifugation was used as an independent biochemical assay for analyzing the membrane or cytosolic localization of ectopically expressed GFP-BepA fusions. The obtained data ( Figure 3D) were in good agreement with the microscopic analysis presented in Figure 3C, except that construct pGFP-Bh BepA 403-544 displayed both cytosolic and membrane localization.
Taken together, these data suggest that association with the plasma membrane is critical for the anti-apoptotic activity of Bh BepA.

Only BepA Orthologs from Vasoproliferative Bartonella Species Display Anti-Apoptotic Activity
To assess whether anti-apoptosis is a general feature of BepA homologs, we tested the orthologs encoded by other Bartonella species for this activity. Bq was previously shown to cause vascular tumor formation, whereas Bt was never associated with vascular lesions [9]. Consistent with these observations, only Bh and Bq, but not Bt wild-type, were able to block actinomycin D-triggered caspase-3/-7 activation in a T4S-dependent manner ( Figure 4A). In the genome sequence of Bq, the bepA gene is annotated as a pseudogene [21]. Closer inspection of the sequence revealed that because of an internal stop codon and a downstream-located start codon in frame, this bepA locus is split into two open reading frames (Bq bepA1 and Bq bepA2). Bq bepA1 encodes a FIC domain, and Bq bepA2 encodes a BID domain and the positively charged Cterminal tail, the latter representing a putative T4S signal. Comparison of the amino acid sequences of Bq BepA1 and Bq BepA2 with Bh BepA revealed high similarity (59% and 63%, respectively). We cloned Bt bepA and Bq bepA2 into expression plasmids, which were introduced into DbepA-G ( Figure 4B). As controls we included the paralogs Bh bepB and Bh bepC, which in the initial screening of Bh Beps were found to lack antiapoptotic activity ( Figure 4B and unpublished data). HUVECs were infected for 24 h with the different isogenic strains, and subsequently apoptosis was induced by actinomycin D. Of the tested BepA homologs, only Bh BepA and Bq BepA2 inhibited actinomycin D-triggered caspase-3/-7 activation ( Figure 4C). These findings revealed that Bq bepA2 encodes an antiapoptotic effector. To assign Bq BepA2 as a novel T4S substrate, we translocated a FLAG-Cya-BepA2 fusion from Bh into HUVECs by the Cya reporter assay (unpublished data). These results show that the anti-apoptotic activity of BepA is conserved among two human pathogens with vasoproliferative capacity (Bh and Bq), but not in Bt, which has not been associated with vasoproliferation. Moreover, these data confirm the localization of the anti-apoptotic activity to the C-terminal region of BepA that composes a functional T4S signal [13].

Inhibition of Apoptosis Correlates with an Increased Intracellular cAMP Level
To identify potential survival pathways mediated by antiapoptotic BepA, we analyzed Affymetrix GeneChip data obtained for the transcriptome of HUVECs infected with Bh wild-type versus the DvirB4 mutant impaired in T4S (M. Dehio, M. Schmid, M. Quebatte, and C. Dehio, unpublished data). These data revealed a T4S-dependent upregulation of the NFjB-and cAMP-dependent CREM/CREB regulons in HUVECs. Both signaling pathways have been described to mediate, among other functions, protection against apoptosis [22][23][24].
We tested whether these pathways were activated by antiapoptotic BepA homologs. We monitored activation of the NFjB pathway in infected HUVECs by measuring the release of interleukin 8 (IL-8). Bh wild-type triggered an increased IL-8 release compared with DbepA-G or the uninfected control. In contrast, none of the BepA homologs expressed in DbepA-G induced increased IL-8 secretion ( Figure 5A). This finding indicates that the survival mechanism triggered by antiapoptotic BepA homologs is independent from the activation of the NFjB pathway triggered by wild-type bacteria. Activation of the cAMP-dependent CREM/CREB pathway was assayed by quantitative real-time PCR of two cAMPinducible genes, namely pde4B and crem [25,26]. Only strains translocating anti-apoptotic BepA homologs were found to induce the expression of these genes in a statistically significant manner ( Figure 5B). By quantifying intracellular cAMP upon infection of HUVECs by the various isogenic Bh strains, we further demonstrated that anti-apoptotic BepA homologs significantly increase the intracellular cAMP level ( Figure 5C).

A Rise in Intracellular cAMP Results in Protection of ECs against Apoptosis
Since translocation of anti-apoptotic BepA homologs resulted in a rise of the intracellular cAMP level, we further studied the role of cAMP in anti-apoptosis. The intracellular cAMP level is regulated by adenylate cyclases (ACs) generating cAMP, and phosphodiesterases degrading cAMP [27,28]. To trigger a physiological rise of the cAMP level in ECs, we activated ACs with forskolin and in parallel inhibited cAMP degradation by adding the phosphodiesterase-inhibiting drug 3-isobityl-1-methylxanthine (IBMX) [29,30]. HUVECs were infected with the substrate-free DbepA-G strain and with the BepA-expressing strain DbepA-G/pBh BepA in the absence and presence of forskolin/IBMX. Apoptosis was induced by exposure to actinomycin D and monitored by measuring caspase-3/-7 activity. By adding forskolin/IBMX we were able to reduce caspase-3/-7 activity in uninfected HUVECs and cells infected with the DbepA-G strain to a level similar to that found in cells infected with DbepA-G/pBh BepA. Interestingly, the addition of forskolin/IBMX did not have an additive protection effect in DbepA-G/pBh BepA-infected HUVECs ( Figure 6A). The same results were obtained by adding the cell-permeable cAMP analog dibutyryl cAMP to the culture medium ( Figure 6B). From these data we conclude that a moderately increased intracellular cAMP level, as triggered by Bh BepA, is sufficient to protect ECs against apoptosis.

The Role of Bh BepA in Evading CTL-Mediated Apoptosis
During colonization of the human endothelium, intracellular Bh are hidden from antibody-and complementmediated immune responses [9]. However, the infected ECs may be killed by CTLs [31]. To study a putative role of Bh BepA in protecting ECs from CTL-mediated apoptosis, we used a major histocompatibility complex (MHC) class Irestricted CTL clone that kills HLA-A2-positive HUVECs upon extracellular loading with a Y-chromosome-encoded, male-specific minor histocompatibility Ag peptide with the sequence FIDSYICQV (SMCY peptide) [32]. To specifically measure the rate of EC apoptosis without interference by apoptotic CTLs, we pre-labeled the EC population with the fluorogenic dye CSFE. Pre-labeled HUVECs were left uninfected (control), or infected with wild-type, DbepA-G, or DbepA-G/pBh BepA. Then, 24 h after infection, HUVECs were either pre-loaded for 30 min with the CTL-stimulating SMCY peptide (10 À5 M) or not pre-loaded. After washing, CTLs were added in an effector-to-target cell ratio of 5:1. At different time points (0 h, 2.5 h, and 6 h), Annexin V and PI staining was performed to quantify the amount of apoptotic HUVECs (CSFE-positive, Annexin V-positive, and PI-negative) ( Figure  7A and 7B). SMCY-treated HUVECs, which were either uninfected or pre-infected with the BepA-deficient DbepA-G mutant strain, displayed a time-dependent increase in the apoptotic population in comparison with respective control cells without pre-loading of the CTL-activating SMCY peptide. In sharp contrast, SMCY-treated HUVECs preinfected with wild-type or the BepA-expressing strain DbepA-G/pBh BepA did not display any significant increase of apoptosis over control cells without pre-loading with the SMCY peptide. These data indicate that BepA is capable of protecting ECs against the MHC class I-restricted apoptotic activity of CTLs. Consistently, we noticed that BepA-expressing bacteria, as well as the ectopic expression of a GFP-Bh BepA fusion protein alone, protected ECs against cell death triggered by a different CTL clone, which was activated by phytohemagglutinin in an MHC class I-independent manner (unpublished data). Together, these data indicate that translocated Bh BepA can effectively protect the chronically infected vascular endothelium against CTL-mediated cell death.

Discussion
The modulation of host cell apoptosis is a recurrent theme in bacterial pathogenesis [1]. Research in this area focused initially on the pro-apoptotic mechanisms triggered by pathogens that typically cause acute infections (e.g., Shigella, Salmonella, and Yersinia species). Some pro-apoptotic effectors and their targeted cellular pathways have been studied in molecular detail [33,34]. More recently, pathogen-triggered anti-apoptosis was recognized as an important virulence trait of bacteria that predominately cause chronic infection (e.g., Bartonella, Brucella, Chlamydia, Helicobacter, Mycobacterium, and Rickettsia species) and thus need to protect their cellular habitats by suppressing host-triggered apoptosis [35]. However, the bacterial effectors, and to a large extent also the cellular pathways involved in mediating pathogen-induced anti-apoptosis, remain poorly defined. In the case of Bh and Bq, the formation of vascular tumors in immunocompromised patients was shown to be linked to the inhibition of apoptosis of infected ECs [10]. In this report, we identified the anti-apoptotic factor of these vasoproliferative bartonellae. Deletion of the Bh bepA gene resulted in the complete loss of the anti-apoptotic activity of Bh, whereas expression of Bh bepA in trans restored the activity to wild-type level. Bh BepA was previously described as a putative substrate of the T4S system VirB/VirD4 [13]; we used the Cya reporter assay to demonstrate that Bh BepA is indeed translocated into ECs in a T4S-dependent manner. Bh BepA is thus a genuine T4S effector that inhibits apoptosis upon translocation into HUVECs. It is worth noting that we have been unable to show translocation of Bh BepA by CRAfT, an assay that we previously used to demonstrate translocation of several other Bh Beps [13]. Nuclear import of the Cre reporter protein fusion is a prerequisite for a positive readout by CRAfT [13], suggesting that the negative readout obtained for Bh BepA fusions could result from protein recruitment to an intracellular localization that interferes with nuclear import. Indeed, ectopic expression of full-length Bh BepA fused to GFP revealed a prominent localization of the fusion protein to the plasma membrane, in contrast to the cytosolic localization of GFP alone. Unlike ectopically expressed GFP, the GFP-Bh BepA fusion also conferred protection against EC apoptosis. These data demonstrate that Bh BepA is not only required but also sufficient for inhibiting EC apoptosis,   and that this anti-apoptotic effector localizes to the plasma membrane. Moreover, ectopic expression of various parts of Bh BepA fused to GFP revealed a strict correlation of the capacity to mediate anti-apoptosis with localization to the plasma membrane (as opposed to the primarily cytoplasmic localization of fusions that did not cause anti-apoptosis). This finding indicates that plasma membrane targeting may be important for the anti-apoptotic activity of Bh BepA.
In search of the cellular pathway by which Bh BepA mediates anti-apoptosis in ECs, we first tested whether an NFjB-dependent survival pathway is involved. The antiapoptotic activity of R. rickettsii in ECs was reported to depend on NFjB activation [5], and we have previously shown that Bh activates NFjB in a VirB/VirD4-and Bep-dependent manner [13]. Here, we demonstrated that Bh BepA alone does not trigger the NFjB-dependent secretion of IL-8; even so, it fully protects against apoptosis. Thus, NFjB does not seem critical for Bh BepA-mediated anti-apoptosis. Affymetrix GeneChip experiments indicated that Bh also triggers a cAMP signaling pathway in a T4S-dependent manner (M. Dehio, M. Schmid, M. Quebatte, and C. Dehio, unpublished data). Here, we have shown that the anti-apoptotic activity of Bh BepA in HUVECs correlates with elevation of the intracellular cAMP level and results in the upregulation of cAMP-stimulated gene expression. Recently, it emerged that the regulation of apoptosis is an important facet of cAMP signal transduction [30,[36][37][38]. Moderately elevated cAMP levels were reported to protect several cell types against apoptosis, while the survival mechanisms differed from cell type to cell type and were considered to require the activation of protein kinase A, extracellular signal regulated, mitogen-activated protein kinase, or guanine nucleotide exchange factor signaling pathways, which subsequently resulted in the expression of anti-apoptotic genes [23,36,37,[39][40][41][42][43][44]]. In our system, the specific downstream signaling pathway mediating cAMPdependent anti-apoptotis remains elusive: efforts to block putatively involved signaling molecules (i.e., protein kinase A) failed because of apparent cytotoxic effects of effective inhibitor concentrations during the extended time frame of the apoptosis assay.
Importantly, a physiological rise in the intracellular cAMP level has been reported to fully protect ECs against apoptosis [45]. Consistent with this, we observed that increased cAMP levels caused by the combined action of the AC-activating drug forskolin and the phosphodiesterase-inhibiting drug IBMX resulted in a complete suppression of actinomycin Dinduced apoptosis in HUVECs. Forskolin/IBMX-treated cells displayed an increase in the expression of cAMP-regulated genes similar to that observed upon infection with antiapoptotic Bh strains. cAMP is produced in eukaryotic cells by the family of membrane-anchored ACs. ACs are activated by heterotrimeric G proteins that are regulated by G proteincoupled receptors [46]. Since the anti-apoptotic activity of Bh BepA in HUVECs is strictly associated with the plasma membrane localization of this effector, concomitant with an increase in the cAMP level, we propose that Bh BepA may trigger anti-apoptosis by interacting either with ACs directly, or with plasma membrane-associated heterotrimeric G proteins or G protein-coupled receptors that regulate AC activity.
Ectopic expression of GFP fusions with different parts of Bh BepA confined the anti-apoptotic activity to a region of 142 aa, which corresponds to the BID domain. This conserved domain was previously shown to be present in a least one copy in all Beps of Bh and Bq [13,14,21]. The BID domain plus the C-terminal positively charged tail sequence of the Beps was shown to constitute a bipartite translocation signal for T4S [13]. Interestingly, in contrast to the BID domain of Bh BepA, the conserved BID domains of the paralogs Bh BepB and Bh BepC have not been associated with anti-apoptosis. This indicates that subsequent to the expansion of this paralogous protein family by gene duplication, the Bh BepA BID domain acquired, in addition to its crucial function as a signal for T4S, the capacity to mediate anti-apoptosis. How this rather short domain may mediate these unrelated activities is presently unknown. It may be speculated that the BID domain represents a basic fold that mediates a protein-protein interaction with the T4S machinery that is crucial for protein translocation, and that in Bh BepA this basic fold is adapted to also mediate specific interaction with the plasma membrane-associated, cAMP-generating signaling cascade of ECs.
Each of the paralogs Bh BepA, Bh BepB, and Bh BepC carries in the N-terminal region one copy of the FIC domain. While this conserved domain of unknown function might be important for the effector function of Bh BepB and Bh BepC, our results demonstrate that the FIC domain is dispensable for the anti-apoptotic activity of Bh BepA. Interestingly, an internal stop codon and downstream-located start codon within the bepA ortholog of Bq split this locus into two separate open reading frames (Bq bepA1 and Bq bepA2). Bq BepA1 has the FIC domain but lacks a T4S signal, and thus cannot be translocated into ECs. In contrast, Bq BepA2 consists only of the T4S signal, which we have shown mediates both T4S-dependent translocation and anti-apoptosis in ECs. It thus appears plausible that the functional diversification of BepA into an anti-apoptotic effector occurred in the common ancestor of the closely related species Bh and Bq [21], and that following speciation the acquisition of additional mutations resulted in the elimination of the dispensable N-terminal FIC domain in Bq.
Interestingly, unlike Bh BepA and Bq BepA2, the BepA ortholog of Bt does not mediate any anti-apoptotic activity, nor any measurable activation of the cAMP pathway. This observation is in agreement with the previously reported finding that, among the bartonellae, the anti-apoptotic activity is limited to the vasoproliferative species Bh and Bq [10].
To asses if the here identified anti-apoptotic activity of Bh BepA might be able to protect the cellular habitat of Bh during the infection process, we used a co-culture system of HUVECs with human CTLs. CTLs execute cell-mediated immunity, an immune mechanism probably involved in the elimination of Bh infection [9,12,47]. Cell-mediated immunity particularly serves as a defense mechanism against microbes that survive and replicate inside infected host cells. Upon recognition of MHC class I-displaying microbial peptides, CTLs are activated to kill their target cells by the release of perforin and granzymes [31,48]. Perforin forms pores in the target cell membrane and assists the delivery of pan-caspaseactivating granzymes into the cytoplasm of the target cell [44,49,50]. In our study, we used a CTL clone that is activated in an MHC class I-restricted manner by extracellular loading of HLA-A2-positive HUVECs with an HLA-A2-specific peptide (SMCY). Such activated CTLs kill their target cells in a perforin-dependent manner [32], primarily by triggering granzyme-dependent apoptosis [50,51]. Bh BepA-expressing strains indeed inhibited CTL-triggered EC apoptosis, indicating that the biological function of Bh BepA expression in vivo might be to protect the integrity of its colonized cellular niche. The Bh BepA-mediated resistance of ECs to CTLdependent cell death points towards an important role of BepA in escaping cell-mediated immunity and thus in protecting the integrity of the chronically infected vasculature, which is a prerequisite for vascular proliferation. The establishment of an animal model for Bartonella-triggered vasoproliferation now appears an urgent need for studying the precise contribution of BepA-mediated anti-apoptosis in the process of vascular tumor formation.
In summary, the anti-apoptotic T4S effectors Bh BepA and Bq BepA2 characterized in this study represent striking examples of the evolution of new pathogenic traits in bacteria. The delineation of their anti-apoptotic activity to the conserved BID domain, and their proposed role in maintaining the cellular habitat by mediating anti-apoptosis via specific interaction with the plasma membrane-associated, cAMP-generating signaling cascade of ECs, should pave the way for future studies to elucidate the molecular and structural basis of Bartonella-mediated anti-apoptosis in the vascular endothelium and the role of this pathological process in vasoproliferative tumor growth.

Materials and Methods
Bacterial strains and growth conditions. The bacterial strains used in this study are listed in Table 1. Bartonella spp. were grown on Columbia agar plates containing 5% defibrinated sheep blood (CBA plates) at 35 8C and 5% CO 2 for 2-4 d. Strain RSE247, a spontaneous streptomycin-resistant strain of Bh ATCC 49882 T [12], served as wildtype in this study. When indicated, media were supplemented with 30 lg/ml kanamycin, 100 lg/ml streptomycin, 12.5 lg/ml gentamicin, and/or 500 lM isopropyl b-D-thiogalactosidase (IPTG). Escherichia coli strains were cultivated in Luria-Bertani liquid medium, or after addition of agar on plates, at 37 8C overnight. When indicated, media were supplemented with 50 lg/ml kanamycin, 200 lg/ml ampicillin, 25 lg/ml gentamicin, 500 lM IPTG, and/or 1 mM diaminopimelic acid.
DNA manipulation. Plasmids used in this study are listed in Table  1. Primers are listed in Table 2.
Plasmids for in-frame deletions and complementation of deletion mutants were constructed as follows: pMS5 used for creating the DbepA mutant was already described before [13]. The use of pMS5 for gene replacement in RSE247 resulted in the DbepA mutant MSE154. Plasmid pRS63, used for generating a DvirB4 in-frame deletion mutant in Bq, was constructed as follows. The BamHI insert of pRS14 [16] was replaced by a 1,318-bp BamHI fragment of the Bq virB locus containing a 2,112-bp in-frame deletion in virB4. This fragment was constructed by megaprime PCR from two PCR products. Product 1 (0.75 kb) was amplified with primers prRS226 and prRS227 and contained the first 115 bp of the virB4 gene and upstream sequences. Product 2 (0.55 kb) was amplified with primers prRS228 and prRS229 and contained the last 120 bp of the virB4 gene and downstream sequences. Megapriming and PCR amplification with primers prRS226 and prRS229 were performed as described [16].
Plasmids for complementation of DbepA mutant and DbepA-G mutant were constructed as follows: For constructing vectors expressing N-terminal FLAG-tagged Bep proteins, the plasmid pPG100 was used. The fragments were generated by PCR amplification from chromosomal DNA of RSE247 and Bt, using oligonucleotide primers prPG92/93, prPG95/129, prPG97/130, or pMS27/28, and were inserted by the flanking NdeI site into the corresponding site of pPG100, resulting in pPG101 (encoding Bh BepA), pMS006 (encoding Bh BepB), pMS007 (encoding Bh BepC), and pMS011 (encoding Bt BepA). To construct plasmid pMS100, carrying a multiple cloning site, prGS01 and prGS02 were annealed together and the resulting 45-bp fragment was inserted into pPG100 using the NdeI site. Using oligonucleotide primers pMS102 and prMS103, a fragment of 0.85 kb was amplified using chromosomal DNA of Bq as template. Then, using SalI/XmaI sites, this fragment was inserted into the corresponding site of pMS100, giving rise to pMS106. Plasmid pMS105 was constructed as follows: using flanking SalI/XmaI sites, the 1.65-kb PCR fragment Figure 7. Bh BepA Protects ECs against CTL-Mediated Cell Death Confluent HUVEC monolayers fluorescently labeled with the cell-tracking dye CFSE were infected with the indicated bacterial strains for 24 h or left uninfected (control). Then cells were incubated in the presence (þ SMCY) or absence (À SMCY) of SMCY peptides, followed by washing. CTLs were added in an effector-to-target cell ratio of 5:1 for the indicated period, and after washing and Annexin V-and PI-staining, the apoptotic cell population of HUVECs was determined by flow cytometry analysis (CFSE-positive, Annexin V-positive, and PI-negative). (A) Dot plots are shown for representative samples. (B) Summary graph representing HUVEC apoptosis during CTL co-culture. The mean and SD of one of two independent experiments performed in duplicates is shown. doi:10.1371/journal.ppat.0020115.g007 generated by prMS37/38 (chromosomal DNA of RSE247 serves as a template) was inserted into the corresponding sites in pRS40, yielding pMS13. Using flanking SalI/XmaI sites, the 1.65-kb fragment was shuttled into the SalI/XmaI sites of pMS100. Using oligonucleotide primers prRS167/180 and prRS187/188, fragments of 0.73-kb size were amplified using chromosomal DNA of RSE247 and of Bt, respectively, as template. Using flanking SalI/XmaI sites, the fragments were inserted into the corresponding site of pRS40, giving rise to pRS48 and RS55, respectively. C-terminal sequences of the different Bep proteins were shuttled into pMS100 using the flanking SalI/XmaI sites. The plasmids resulting from this step are given below, as well as the range of amino acids of a given Bep protein fused to the N-terminal FLAG: pMS100-A (shuttled from pRS48, aa 305-544 of Bh BepA), pMS100-B (shuttled from pRS49, aa 303-542 of Bh BepB), pMS100-C (shuttled from pRS50, aa 292-532), and pMS100-D (shuttled from pRS51, aa 352-534 of Bh BepD). Plasmids for expression of Cya-Bep fusion proteins were constructed as follows: To construct pMS400, the cya of pMS111 was PCRamplified with oligonucleotide primers prMS78/90 introducing a start and a stop codon to the resulting cya fragment. Using flanking NdeI/ XmaI sites, the 1.23-kb fragment was inserted in the corresponding sites of pMS100. Plasmid pMS401 was derived by insertion of a 1.2-kb NdeI/StuI fragment of pMS400, which includes the cya gene without stop codon, into the corresponding sites of pMS100-A. The NdeI/StuI fragment of pMS400 was further inserted in the corresponding site of pMS100-D and pMS105, giving rise to pMS404 and pMS405, respectively. To construct pMS406, Bq BepA2 was PCR-amplified from chromosomal DNA of Bq with oligonucleotide primers prMS103/104. Using flanking PmeI/XmaI sites, the 0.86-kb fragment was inserted in the StuI/XmaI sites of pMS400.
Plasmids for ectopic expression of GFP-BepA fusion proteins were constructed as follows: To construct eGFP-Bep fusion proteins (see Figure 4A), pWAY21 (Molecular Motion Lab, http://momotion.cns.montana.edu), a CMV-driven EGFP for C-terminal fusion, was used as basic vector. The plasmids resulting from this step and the oligonucleotide primers used (with incorporated XmaI/XbaI sites used for cloning to the corresponding sites of pWAY21) are given below, as well as the range of amino acids of a The integrity of all constructs was confirmed by sequence analysis and Western blotting using anti-FLAG M2 antibodies (Sigma-Aldrich, http://www.sigmaaldrich.com).
Generation of deletion mutants. pRS63 was used to generate the DvirB4 mutant RSE569 in the RSE356 background as previously described for Bt [16]. The spontaneous streptomycin-resistant mutant RSE356 was obtained by selection of B. quintana JK-31 on 100 mg/l streptomycin.
Cell lines and cell culture. HUVECs were isolated as described [18]. HUVECs and the human embryonic kidney cell line HEK293T were cultured as described before [12]. The stable transfected endothelial cell line Ea.hy296/pRS56-c#B1 was cultured as reported [13].
HEK293T cell transfection. Subconfluent (2.5 million cells) HEK293T cells in 8-cm cell culture dishes were transfected with a total of 2.5 lg DNA following the protocol ''Calcium phosphatemediated transfection of eukaryotic cells with plasmid DNAs'' [52]. After 12 h, the cell culture medium was replaced and the cells kept in culture for an additional 24 h before fixation for immunocytochemical staining or harvesting for subcellular localization of the transfected constructs.
Subcellular fractionation. Subcellular fractionation was performed according to standard protocols [53,54]. Briefly, confluent monolayers were washed three times with cold homogenization buffer (10 mM triethanolamine, 10 mM acetic acid, 1 mM EDTA, 250 mM sucrose [pH 7.4]), and cells were harvested by scraping and homogenized with a syringe and a 22-gauge 1/4 inch needle. The post-nuclear supernatant was obtained by centrifugation at 240g (15 min). Membranes were separated from cytosol by ultracentrifugation of the post-nuclear supernatant at 100,000g (30 min).
Caspase activity assay. The infection of HUVECs and the determination of caspase-3/-7 activity (MOI ¼ 300) were carried out as described [12].
cya assay. cAMP was assayed after 20 h contact between bacteria (MOI ¼ 300) and HUVECs in 24-well plates. HUVECs were washed one time in pre-warmed PBS and lysed in denaturizing conditions as described previously [19]. cAMP was assayed by an EIA system (Biotrak; Amersham, http://www.amershambiosciences.com). Total cell proteins were assayed by the method of Bradford [55] (Bradford Reagent, Sigma-Aldrich).
CRAfT. The infection of the stably transfected Ea.hy926/pRS56-c#B1 cell line and the quantification of GFP-positive cells (percent positive events) by flow cytometry were carried out as described [13].
Immunoblot analysis. To monitor the steady-state level of Cya fusion proteins, bacteria were grown on IPTG-containing medium for 2 d. Cells were harvested and processed as described previously [13], except that anti-FLAG M2 (Sigma-Aldrich) was used as the primary antibody.
The stability of the different GFP fusion proteins used for ectopic expression in HUVECs was monitored after transient transfection of HEK293T cells. Cells were transfected by FuGENE 6 (Roche, http:// www.roche.com) following the manufacturer's instructions. After 24 h expression, cells were washed twice with PBS and harvested in 100 ll of sample buffer. Samples were then further processed as described before [13] and probed with anti-GFP (Roche) antibody.
Annexin V assay. Twenty-four hours after infection or transfection of HUVECs, apoptosis was induced for 12-24 h. Cells were then collected by mild trypsinization and briefly centrifuged together with the culture supernatant. The cell pellet was washed, resuspended, and stained with PI (1 ug/ml) and Annexin V AlexaFluor488 (Molecular Probes, http://probes.invitrogen.com) or Annexin V APC (Alexis, http://www.axxora.com). The total apoptotic population was determined by analyzing Annexin V-positive and PI-negative/positive cells with a FACSCalibur flow cytometer (BD Biosciences, http://www. bdbiosciences.com). Prior to determining the apoptotic population, transfected cells were gated by their positive GFP signal.