A Translocated Effector Required for Bartonella Dissemination from Derma to Blood Safeguards Migratory Host Cells from Damage by Co-translocated Effectors

Numerous bacterial pathogens secrete multiple effectors to modulate host cellular functions. These effectors may interfere with each other to efficiently control the infection process. Bartonellae are Gram-negative, facultative intracellular bacteria using a VirB type IV secretion system to translocate a cocktail of Bartonella effector proteins (Beps) into host cells. Based on in vitro infection models we demonstrate here that BepE protects infected migratory cells from injurious effects triggered by BepC and is required for in vivo dissemination of bacteria from the dermal site of inoculation to blood. Human endothelial cells (HUVECs) infected with a ΔbepE mutant of B. henselae (Bhe) displayed a cell fragmentation phenotype resulting from Bep-dependent disturbance of rear edge detachment during migration. A ΔbepCE mutant did not show cell fragmentation, indicating that BepC is critical for triggering this deleterious phenotype. Complementation of ΔbepE with BepEBhe or its homologues from other Bartonella species abolished cell fragmentation. This cyto-protective activity is confined to the C-terminal Bartonella intracellular delivery (BID) domain of BepEBhe (BID2.EBhe). Ectopic expression of BID2.EBhe impeded the disruption of actin stress fibers by Rho Inhibitor 1, indicating that BepE restores normal cell migration via the RhoA signaling pathway, a major regulator of rear edge retraction. An intradermal (i.d.) model for B. tribocorum (Btr) infection in the rat reservoir host mimicking the natural route of infection by blood sucking arthropods allowed demonstrating a vital role for BepE in bacterial dissemination from derma to blood. While the Btr mutant ΔbepDE was abacteremic following i.d. inoculation, complementation with BepEBtr, BepEBhe or BIDs.EBhe restored bacteremia. Given that we observed a similar protective effect of BepEBhe on infected bone marrow-derived dendritic cells migrating through a monolayer of lymphatic endothelial cells we propose that infected dermal dendritic cells may be involved in disseminating Bartonella towards the blood stream in a BepE-dependent manner.


Introduction
Pathogenic bacteria have evolved a multitude of virulence factors in order to manipulate the host to evade immune responses and to reach their replicative niche -a safe compartment to proliferate that is also a prerequisite for transmissibility [1]. Translocation of bacterial effector proteins into host cells is one of the mechanisms to manipulate the host by interfering with its signaling pathways. A prominent example is CagA, a multifunctional effector protein of the Helicobacter pylori (Hpy) type IV secretion system (T4SS). CagA modulates both innate and adaptive immune responses of the host and assists Hpy to infect the gastric mucosa in about half of the world population for their lifetime [2,3]. Numerous effector proteins of Salmonella type III secretion systems (T3SS) SPI1 and SPI2 [1] and Shigella T3SS play a critical role in invasion of non-phagocytic intestinal cells, for further dissemination and modulation of the host inflammatory responses [4,5]. In addition to targeting the host cellular components, some bacteria have evolved effectors that regulate an activity of each other at a specific stage of the host invasion; like Legionella Dot/Icm ''metaeffector'' LubX mediates the degradation of SidH. Or this interplay may happen in an indirect fashion as for many cases of T4SS/T3SS effectors [6].
Bartonella species are fastidious, Gram-negative, facultative intracellular bacteria that are highly adapted to a distinct mammalian reservoir host [7,8,9,10]. Infections in the reservoir host range from asymptomatic or sub-clinical (for most animalspecific species) to clinical manifestations with low morbidity and limited mortality, such as human-specific B. quintana (Bqu) infections, or even to life-threatening disease, such as human infection by B. bacilliformis (Bba) [9,11]. Bartonellae transmission is mediated by blood-sucking arthropod vectors. The strategy involves replication of bacteria in the gut of the arthropod vector and excretion in the feces, with subsequent survival in the environment for several days [12]. The arthropods usually defecate when feeding on mammals and provide a source of local irritation that results in itching, followed by scratching and inoculation of Bartonella-containing feces into the derma of the skin [13]. Later, Bartonella is known to appear in the blood of the reservoir host, invades erythrocytes as immune-privileged niche and develops long-lasting persistent infections for more than a year for some species [11,14].
Bartonellae evolved two T4SSs (Trw and VirB) while adapting to a wide range of mammalian hosts [15]. Both of them are essential for the interaction with the host but at different stages of the infection cycle [14,15,16,17,18]. The Trw system seems to mediate host-specific adhesion of Bartonella to erythrocytes by binding to the cell surface with its manifold variants of pilus subunits [19,20], while the VirB system translocates a cocktail of evolutionarily related Bartonella effector proteins (Beps) into nucleated host cells. The Beps (named BepA-G in Bhe) evolved by duplication of an ancestral bep gene followed by their functional diversification and conservation of certain domains and motifs [15]. All Beps have at least one Bartonella intracellular delivery (BID) domain and a positively-charged tail in the C-terminus as a signal for translocation through the T4SS [21].
BepE is one of the effector proteins that is conserved within lineage 4, the largest clade within the Bartonella genus comprising 11 species, e.g. B. henselae, B. quintana, B. tribocorum and B. grahamii [15]. Bhe BepE (BepE Bhe ) consists of 464 amino acids (aa). The Nterminus of BepE Bhe , similarly to BepD Bhe and BepF Bhe , contains short repeated peptide sequences (EPLYA) with conserved putative tyrosine phosphorylation sites, similar to the EPIYA motif of Hpy effector CagA. BepE Bhe has two BID domains in the C-terminal part [21]. The domains and the motifs but not the spacing in between is well preserved in BepE homologues [15]. Mass spectrometric analysis of BepE Bhe pull-downs revealed several SH2 domain-containing eukaryotic signaling proteins that interact either with an individual phosphotyrosine of BepE Bhe within a Csk-like binding motif or two ITIM/ITSM (immunotyrosine inhibitory motif/immunotyrosine switch motif) tandems [22]. Conservation of these specific motifs of BepE Bhe and the described interaction partners suggest a molecular mimicry of ITIM-containing receptors by bacterial proteins and a potency to interfere with host signaling pathways.
In this study, we identified BepE as an essential bacterial factor for Bartonella reservoir host infection via the intradermal route resembling natural infection by arthropod vectors, but not for the intravenous route. This specific function was assigned to the Cterminal part of BepE Bhe including the two BID domains. The same BID domains were interfering with a prominent cell fragmentation phenotype in migrating endothelial cell induced by BepC and possibly other Beps as a secondary effect. Further, we show that Bartonella translocates an effector-fusion protein (Bla-BID) into dendritic cells and affects cell migration in the absence of BepE, suggesting that DCs may be involved in the BepEdependent dissemination of Bartonella to the blood.

Results
BepE Bhe is sufficient to abolish a cell fragmentation phenotype induced by the Bhe DbepDEF mutant A previous study on the interactome of Bartonella effector protein BepE Bhe revealed several SH2 domain-containing signaling proteins that interact with BepE Bhe upon phosphorylation of specific tyrosines within the motifs [22]. Based on these data, we hypothesized that BepE might be a factor impacting multiple cellular signaling pathways to promote the establishment of a successful Bartonella infection. To acquire first insights into the cellular phenotypes and uncover the molecular bases of BepE we used the well-established human umbilical vein endothelial cell (HUVEC) infection model for Bhe [23,24,25,26,27].
BepE, together with BepD and BepF, belongs to the class of the Beps harboring tyrosine-containing motifs in their N-termini. In order to reduce the complexity of potential redundant functional effects by any of these three tyrosinephosphorylated effectors we infected HUVECs with a Bhe mutant carrying an in-frame deletion of the chromosomal region encoding bepD, bepE and bepF (Bhe DbepDEF). The infection was studied by means of microscopy of either fixed samples or by time-lapse imaging over 72 h. Starting about 20-24 hours post infection (hpi) the Bhe DbepDEF mutant showed a drastic phenotype of cell fragmentation ( Fig. 1A and Movie S1) which was hardly detected in cells infected with wildtype bacteria at this early stage of infection (Fig. 1B, C), but became more apparent at later time points, beyond 48 hpi (Movie S2). Fragmenting cells were apparently normally moving forward but displayed pronounced difficulties in rear end retraction -such cells became more and more elongated and at some point the thin connection between a given cell body and the attached rear broke (Fig. 1A). The fragment left behind migrated on the substrate for a few hours and then came to a halt, while the cell body containing the nucleus got smaller with each fragmentation. Eventually, this process led also to a decreased number of cells within the sample ( Fig. 1A and B, Movie S1).
Infections with single effector-complemented Bhe DbepDEF revealed that the cell fragmentation phenotype was inhibited only by expression of BepE Bhe , while neither BepD Bhe nor BepF Bhe displayed a similar cytoprotective effect (Fig. 1B, D).

Author Summary
Cell migration, a fundamental feature of eukaryotic cells, plays a crucial role in mounting an effective immune response. However, several pathogens subvert the migratory properties of infected host cells to their benefit, such as using them as Trojan horses to disseminate within the host. Bartonella effector proteins (Beps) are bona fide virulence factors indispensable for the colonization of mammalian target cells. However, their multiple interferences with host cellular signaling processes might culminate in deleterious secondary effects that require additional effectors to maintain the host cell integrity. A striking example is BepE, which is shown here to preserve endothelial cells (ECs) from fragmentation and to inhibit the defects of dendritic cell (DCs) migration caused by BepC and possibly other Beps. Moreover, BepE is essential for Bartonella dissemination from the dermal site of inoculation to the blood stream where bacteria establish long-lasting intraerythrocytic bacteremia as a hallmark of infection in the mammalian reservoir host. Migration of Bartonella-infected DCs through a monolayer of lymphatic ECs was also found to be dependent of BepE, suggesting that BepE is required to preserve the migratory capability of DCs, a candidate cell type for systemic dissemination from the dermal site of inoculation. Quantification of cell fragmentation at 48 hpi was performed in semi-automated manner. Images were acquired in 96 well-plate format by MD ImageXpress Micro automated microscopes with 106 magnification. The number of fragmented cells (cells with thin and multipolar elongations) were defined by eye and counted manually. The percentage of fragmented cells is normalized to Bhe wild-type infection. In each condition triplicate wells with each 10 randomly picked fields were imaged and presented as mean +/2 SD. Statistical significance was determined using Student's t-test. P,0.05 was considered statistically significant. Data from one representative experiment (n = 3) are presented. (E) Protein levels of the BepD Bhe , BepE Bhe and BepF Bhe by plasmid overexpression in DbepDEF. The anti-Flag western blot was obtained from total lysate of corresponding Bhe strains. doi:10.1371/journal.ppat.1004187.g001 A single BID domain of BepE Bhe is sufficient to interfere with the fragmentation of HUVECs Provided that BepD and BepF could not abolish the cell fragmentation phenotype of the DbepDEF mutant we were interested to test whether single deletions of bepD, bepE or bepF would be sufficient to trigger cell fragmentation. Bhe DbepEinfected HUVECs showed indeed very similar morphological changes but the effect was less pronounced. Compared to wildtype infection, none of the other mutants could induce elevated fragmentation in a significant manner ( Fig. 2A). As expected BepE could also rescue the cell fragmentation led by Bhe DbepE (Fig. 2B and C, Movie S3).
We next delineated the functional domain responsible for the abrogation of cell fragmentation in BepE Bhe . As indicated earlier, BepE Bhe has two BID domains at its C-terminus plus a positively charged C-tail. These domains (BID1.E Bhe and BID2.E Bhe ) show high similarity (pair-wise aa identity of 53.3%) (Fig. S1A) and thus seem to be originated from a duplication event [15]. To understand which part of BepE Bhe was responsible for the effect observed during the infection of HUVECs, BIDs.E Bhe and BID2.E Bhe were expressed in Bhe DbepE as they both contain an intact C-term required for translocation ( Fig. 2B-E) and then used to infect HUVECs. The two BID domains (BIDs.E Bhe ) do overcome the Bhe DbepE-induced cell fragmentation phenotype. Even more, BID2.E Bhe is able to complement with almost the same efficiency as full length BepE.
In summary, these data indicate that the inhibition of cell fragmentation by BepE Bhe was mediated by the BID domains of BepE Bhe and more specifically, the BID2.E Bhe was sufficient.

Cell fragmentation induced by Bhe DbepDEF is inhibited by heterologous complementation with BepE Bhe homologues
BepE homologues from the Bartonella lineage 4 species display significant similarity in domain and motif composition (Fig. S1B). We thus tested whether BepE homologues from different Bartonella species, i.e. BepD Btr and BepE Btr from B. tribocorum, BepE Bqu from B. quintana and BepH Bgr from B. grahamii, can interfere with the cell fragmentation phenotype of the Bhe DbepDEF. BepE Btr , BepE Bqu , and BepH Bgr were indeed able to functionally replace BepE Bhe in Bhe DbepDEF background. Amongst all the homologues, BepD Btr has the least amino similarity to BepE Bhe (Fig. S1B) and was also the least potent in suppressing the cell fragmentation phenotype ( Fig. 3A and B). Interestingly, BepE Bqu from the human-specific Bqu was the most efficient in abrogating cell fragmentation. Considering that HUVECs are primary human endothelial cells this observation is suggestive for some level of host specificity in the activity of the BepE effector.

Deletion of BepC in the Bhe DbepE mutant background is sufficient to abolish the cell fragmentation phenotype
The cell fragmentation phenotype is triggered by the mutant strains Bhe DbepE and Bhe DbepDEF. In order to find the factor from Bhe leading to this severe phenotype we tested a set of available Bhe mutants. The effector-free mutant Bhe DbepA-G did not display cell fragmentation, indicating that the factor that triggers cell fragmentation in the absence of BepE must be another Bep ( Fig. 4A and B). Given that the Bhe DbepDEF mutant strain displays severe cell fragmentation indicates that that the cell fragmentation activity is confined to either BepA, BepB, BepC or BepG or a combination thereof. A double mutant Bhe DbepCE did not display cell fragmentation, indicating that BepC is essential for this activity. Although we cannot exclude that either BepA, BepB, or BepG may contribute to this phenotype, ectopic expression of mCherry-BepC in HUVEC resulted in cell fragmentation that could be reduced by co-expression of GFP-BepE (Movie S4), indicating that BepC can trigger cell fragmentation on its own.

BepE translocation and interference with cell fragmentation is T4SS-dependent
To confirm that inhibition of cell fragmentation observed upon infection with strains over-expressing BepE from plasmid was fully attributable to BepE effector function within the host cell and not to competitive inhibition with translocation of other Beps, we first confirmed VirB T4SS-dependency of BepE Bhe translocation into host cells. To this end we expressed Myc-BepE in Bhe wild-type or the VirB T4SS-translocation deficient Bhe DvirB4 mutant strain. Immunocytochemical staining with anti-Myc antibodies showed clearly that BepE Bhe is translocated in a VirB T4SS-dependent manner (Fig. 5A.) as previously demonstrated for BepD Bhe [21].
In order to avoid that BepE translocation interferes with BepC and possibly other Beps contributing to the cell fragmentation phenotype, we performed a mixed infection experiment with the strain Bhe DbepE (Fig. 5B) that triggers cell fragmentation and strain expressing BepE in an effector-free mutant background (Bhe DbepA-G-pBepE). The results showed that BepE translocated by one strain can suppress the cell fragmentation phenotype mediated by the effectors translocated by the other strain. As a negative control we showed that expression of BepE in a VirB T4SS-deficient background (Bhe DvirB4/bepA-G-pbepE) did not lead to suppression of the cell fragmentation phenotype triggered by the other strain (Fig. 5B, for Bhe DbepDEF see Fig. S2). These data demonstrate that BepE is translocated by the VirB T4SS system into the host cell and upon translocation shows a cytoprotective effect by interfering with the cell fragmentation.

Ectopic expression of BepE in HUVECs abrogates cell fragmentation
Next we expressed BepE ectopically in HUVECs in order to complement the data obtained for inhibition of cell fragmentation by VirB-translocated BepE. To this end we used a lentiviral transduction system to generate GFP, GFP-BepE Bhe , GFP-BIDs.E Bhe or GFP-BID2.E Bhe -expressing HUVECs. Lentiviral transduction resulted in mixed cultures of transduced HUVECs expressing the GFP-fusion protein and GFP-negative non-transduced cells. Such mixed cultures were infected with Bhe wild-type, Bhe DbepE (Fig. S3) and Bhe DbepDEF (Fig. 6A, B) or left uninfected. Microscopic analysis showed that non-transduced and GFPpositive cells, expressing plain GFP as control, resulted in much higher fraction of HUVECs displaying cell fragmentation than cells expressing GFP-BepE Bhe , GFP-BIDs.E Bhe or GFP-BID2.E Bhe ( Fig. 6A and Fig. S3). Expression of the full-length GFP-BepE Bhe fusion protein and its truncated derivatives in HUVECs were validated by anti-GFP western blot (Fig. S4). Considering that the cell fragmentation at late time points of the infection (48 h) eventually leads to the decrease of the cell number within the sample (Fig. 1B), we used flow cytometry for quantification of this phenotype. To do so, we monitored the ratio of transduced (GFPpositive) vs. non-transduced (GFP-negative) in both infected and uninfected HUVEC populations (Fig. 6B). We considered an increase in the ratio of GFP-positive cells in the infected sample to be indicative of a protective phenotype mediated by the GFPfusion protein, knowing that cells undergoing fragmentation would be lost during the experiment. These analyses confirmed that the cells expressing GFP-BepE Bhe , GFP-BID2.E Bhe and to a lesser extent GFP-BIDs.E Bhe were protected from fragmentation and BepE Bhe is recruited to the plasma membrane, localizes to cell-to-cell contacts and is enriched in the rear edge of migrating HUVECs In the previous section we showed by anti-Myc immunocytochemistry in combination with confocal microscopy the VirB T4SS-dependent translocation of Myc-BepE Bhe into HUVECs. Myc-BepE Bhe localized in a punctuate staining pattern to the plasma membrane, with some enrichment to cell contact areas (Fig. 5A). Due to that latter we performed a co-staining for Myc-BepE or Myc-BIDs.E Bhe additionally with VE-cadherin, a marker for adherence junctions. The stainings demonstrated co-localization of full length BepE or the BID domains with VE-cadherin ( Fig. 7A).
To test for the subcellular localization of GFP-BepE Bhe ectopically expressed in HUVECs, samples were analyzed by confocal microscopy and images were taken in the xyand the xzplanes. GFP-BepE Bhe clearly localized to plasma membrane with a clear exclusion of signal in the nucleus, while GFP alone localized to the cytoplasm and the nucleus (Fig. 7B). Similarly to the immunostaining of Myc-BepE Bhe , confocal analysis of ectopically expressed GFP-BepE Bhe revealed accumulation of BepE at cell-tocell contacts (Fig. S4, Movie S5).
Further, when GFP-BepE Bhe -expressing HUVECs were investigated in time-lapse, GFP signal was transiently accumulating and localizing to the rear edge of the cell. As displayed in Movie S6 or the series of images taken at 15 min intervals in Fig. 7C, the GFP signal was increasing at the rear sites of cell detachment, making it conceivable to assume that this transient accumulation of BepE Bhe may have a role in restoring rear end retraction during cell migration when the process is distorted in cells infected with Bhe mutants deficient of BepE.

Ectopic expression of BepE Bhe interferes with the disruption of stress fibers by Rho inhibitor I
RhoA is known to be involved in cell body contraction and rear release [28,29,30,31]. Since the fragmentation of HUVECs by infection with Bhe DbepDEF or DbepE mutant bacteria was associated with a distortion of rear edge detachment during cell migration, and because GFP-BepE Bhe transiently localizes to this rear edge during detachment, we became interested in testing the role of the RhoA signaling pathway for BepE activity.
The mixed population of GFP-BepE Bhe -expressing and nonexpressing HUVECs was treated with Rho inhibitor I (CTO4), which is known to inactivate RhoA, B and C by ADP-ribosylation. In response to intoxication by this inhibitor cells initially lose their stress fibers, then round up and eventually die [32,33,34]. Interestingly, HUVECs expressing GFP-BepE Bhe showed resistance to the Rho inhibitor I as shown by the selective preservation of stress fibers, while neighboring non-transduced cells identified by the lack of GFP signal displayed complete resolution of their stress fibers (Fig. 8A). These phenotypes were quantified using HUVECs expressing plain GFP as control sample (Fig. 8B). Interference of GFP-BepE Bhe with the activity of Rho inhibitor I was dose-dependent (Fig. 8B). Treatment of HUVECs expressing GFP-BepE Bhe mutants with the same Rho Inhibitor I showed that the second BID domain (BID2.E Bhe ) of BepE is sufficient to interfere with the activity of Rho inhibition (Fig. 8D).
These results support the interaction of BepE with the Rho signaling pathway. To address the question whether BepE acts downstream of ROCK, an effector of RhoA, GFP-BepE Bhetransduced HUVECs were exposed to the inhibitor Y27632 that is  In each condition triplicate wells with each 10 randomly picked microscopic fields were analyzed and represented as mean +/2 SD. Statistical significance was determined using Student's t-test. P,0.05 was considered statistically significant. Data from one representative experiment (n = 3) are presented. (C) Dose-dependent effect of Y27632 on HUVECs expressing GFP-BepE Bhe . Cells with actin stress fibers within GFP or GFP-BepE Bhetransduced or non-transduced HUVECs were quantified in semi-automated manner as described for (B). (D) Comparison of the potency of GFP-BepE Bhe and mutant derivatives to interfere with Rho inhibitor I. Lentivirally transduced HUVECs expressing GFP-fusions of BepE Bhe and the depicted mutant derivatives were treated with 0.5 mg/ml Rho inhibitor I. Specimen were stained and stress fiber-containing cells were quantified as described in (B). doi:10.1371/journal.ppat.1004187.g008 known to inhibit both ROCK1 and ROCK2 [35]. In contrast to the effect mediated by the Rho inhibitor I, Y27632-treated GFP-BepE Bhe HUVECs lost the stress fibers similar to the plain GFPexpressing control cells (Fig. 8C). Thus, BepE may have a direct or indirect effect on RhoA or even on ROCK, but not further downstream in this signaling pathway.
Btr DbepDE loses the ability to colonize rat blood in intradermal infection Observing that BepE interferes with deleterious secondary effects of other Beps on migrating cells during Bartonella infection in vitro we became interested to examine the role of BepE in the establishment of infection in vivo. Bhe causes intraerythrocytic bacteremia as hallmark of infection in its feline natural reservoir host only, which for ethical reasons is not an accessible experimental model. However, as we have noted functional conservation among BepE homologues (including Btr) in protecting the Bhe DbepDEF-mediated fragmentation of HUVECs we reasoned that we can switch to Btr as appropriate species for experimental infection of rats as its natural host. Initially we employed the well-described intravenous (i.v.) infection by Btr that leads to the onset of intraerythrocytic bacteremia at 5 days post infection (dpi) following colonization of the so called ''primary niche'' that is considered to include vascular endothelial cells [14,17]. Btr encodes the BepE Bhe homologues BepD Btr and BepE Btr (Fig. 9A) that both cause partial protection of HUVECs from Bhe DbepDEF-mediated cell fragmentation (see Fig. 3). We thus constructed an in-frame deletion of bepD and bepE (Btr DbepDE) and compared its course of infection to wild-type bacteria. Rats were injected with Btr wild-type or Btr DbepDE through the i.v. route and bacteremia was monitored over time as colony forming units (CFUs) per ml of blood. In this model, the Btr DbepDE strain did not show any significant difference in its capacity to cause long-term bacteremia with high bacterial burden (Fig. S5, Fig. 9C). Recently, an intradermal (i.d.) infection model of B. birtlesii (Bbi) has been introduced by Marignac et al [36], where the bacteria are inoculated in the derma on the ear pinnae of mice. This model more likely reflects the natural route of infection, mimicking the animal scratching an area bitten by the arthropod vector, thereby inoculating derma with infectious bacteria from the arthropod feces. We thus adapted i.d. inoculation to our rat model to test for a possible role of BepE at the very early dermal stage of natural infection (Fig. 9B, Fig. S6A). To this end we injected rats with Btr wild-type or Btr DbepDE bacteria through the i.d. route and monitored the development of bacteremia of infected animals. Btr wild-type-infected animals developed blood colonization between 7-9 days post infection (dpi), which occurs with 3-4 day delay as compared to the i.v. model (Fig. S6A). In sharp contrast, 12 out of 13 Btr DbepDEi.d. infected rats (from 3 different experiments) stayed abacteremic over 70 days (Fig. 9B). Only one Btr DbepDEinfected animal developed a delayed bacteremia starting on day 16 post infection (Fig. 9B, animal i.d. Btr DbepDE #4), which might result from the fact that the i.d. infection protocol can damage capillaries and small vessels in the dermis and thus inoculate bacteria directly into circulating blood similar to the i.v. inoculation protocol, yet at much smaller bacterial numbers.
Taken together, the data obtained in the rat infection model suggest that either one of the missing effectors in Btr DbepDE, BepD Btr or BepE Btr , or both of them, are of pivotal importance for Bartonella to colonize the reservoir host when inoculated in the derma reflecting natural arthropod-borne infection, while the mutant is as infectious as wild-type bacteria via the artificial i.v. infection route.  (Fig. 9C, Fig. S6B).
Taken together these data demonstrate for the i.d. rat model that complementation of the abacteremic Btr DbepDE mutant with either BepE Btr or its homologue BepE Bhe is sufficient to restore bacteremia. This finding strongly supports a functional conservation of BepE between different species of the genus Bartonella that is in line with our in vitro studies on the protection from cell fragmentation (see Fig. 3B).
The BID domains of BepE Bhe are sufficient to enable Btr to reach the blood Since BepE Bhe functionally complemented the DbepDE mutant for bacteremia establishment in the i.d. inoculation model we next tested to which domain this activity is confined and whether this might coincide to the BID domains as determined for the protection against cell fragmentation in vitro ( Fig. 2 and 6). To this end, mutant variants of BepE Bhe were ectopically expressed in the Btr DbepDE background and tested in the i.d. infection model. Two mutants of BepE Bhe were chosen, one with all five tyrosines exchanged to phenylalanine (Y 37 , 64 , 91 , 106 , 129 RF) and the second, a truncated version of BepE Bhe , with only the two BID domains and positively charged C-tail (BIDs.E Bhe ). The expression of the BepE Bhe mutant proteins was confirmed by western blotting (Fig.  S6C). At 16 dpi both complemented Btr DbepDE mutants triggered bacteremia at similar levels as the mutant complemented with fulllength pbepE Bhe or as Btr wild-type bacteria (Fig. 9D). Thus, the tyrosine-containing motifs of BepE Bhe , including its N-terminal end, are not required for Bartonella to reach its replicative niche in the reservoir host. The functional part of BepE for this infection phenotype can therefore be assigned to the BID domains.

Bartonella is able to translocate the effector protein within dendritic cells and affects their migration
In vivo experimental data from rat i.d. infections suggests a role for BepE in the establishment of infection at very early stages, prior to the invasion of the blood system. In addition, the in vitro model demonstrates that BepE is essential to inhibit fragmentation of infected HUVECs by interfering with the effect of other Beps. To better understand the in vivo relevance of BepE and based on our in vitro data, we decided to further explore the impact of BepE on cell migration for a cell type potentially relevant to the very early infection process. Assuming that the initial site of Bartonella infection is in derma, we speculate that innate immune cells represent the yet unidentified primary infection niche and consider resident dendritic cells (DCs) as a candidate cell type. These cells have high potency of phagocytosis and the property to migrate towards lymph nodes [37,38,39].
First we assessed the effector translocation by Bhe into the BMDCs (bone marrow-derived dendritic cells). The strain expressing Bla-BID, a beta-lactamase fusion protein with a bipartite translocation signal from Bhe BepD [21], was used to infect mouse BMDCs. b-lactamase activity was measured using CCF2, a FRET substrate that emits green fluorescence but emits   [40]. Starting at 12 hpi the Bla-BID translocation was observed within infected BMDCs (Fig. 10A).
To investigate the migration of Bartonella-infected DC towards the draining lymph nodes, we decided to use a trans-well migration assay [41], a tissue culture model system incorporating a 3D extracellular matrix and interstitial flow, where the infected BMDCs have to migrate first through the 3D extracellular collagen matrix and then a layer of iLECs (immortalized lymphatic endothelial cells), thus mimicking the entry of lymphatic system by an infected DC.
As shown in Fig. 10B, BMDCs infected prior with Bhe DbepDEF revealed less efficiency in 3D trans-well migration compared to uninfected or Bhe wild-type infection conditions. Similarly to HUVEC cells, the BMDC migration is also affected by Bhe DbepDEF and rescued by BepE. Thus, in vitro effector translocation and modulation of cell migration, together with the general relevance of DCs for infection processes, prioritizes these cells for future investigations of the in vivo functions of Bartonella effector proteins.

Discussion
Various bacterial pathogens secrete multiple effectors that act in concert to modulate different host cellular functions during the course of infection. Often these effectors may interfere with the activity of each other -either directly or indirectly -in order to orchestrate their multi-pronged interactions with the host in a spatially and temporally controlled manner. We have discovered a particularly striking example in Bartonella. We found that BepE acts via BID2.E Bhe on the RhoA signaling pathway, thereby alleviating deleterious secondary effects of BepC and possibly other Beps. These Beps have distinct functions within the host cell; however, in the absence of BepE, they result in an impaired cell migration and subsequent fragmentation of the infected ECs. Moreover, the rat model of Btr i.d. infection, which recapitulates the natural way of Bartonella infection by an arthropod, revealed the role of BepE and its BID domains on the dermal stage of infection, thus showing its essential role in establishing reservoir host infection.
Bartonella effectors are known to exhibit functional redundancy, i.e. the ability of structurally different elements to perform a similar function [42]. As an example, the F-actin-dependent uptake of Bhe via the invasome structure is triggered by distinct Bep-dependent pathways, either by BepC Bhe and BepF Bhe , or by BepG Bhe alone [27]. All Beps harbor at their C-terminus a bipartite T4SS signal composed of a BID domain and a positively-charged tail sequence [21], while only BepD Bhe , BepE Bhe and BepF Bhe contain multiple tyrosinephosphorylation motifs (ITIM/ITSM) in their N-terminal domains [22]. Assuming functional redundancy among this class of effectors we initially focused on the Bhe DbepDEF mutant in infection experiments in HUVECs, the latter representing a wellestablished in vitro model for Bhe infection [23,24,25,26,27]. HUVEC infection with Bhe DbepDEF resulted in a prominently impaired cell migration phenotype. Cell migration is a complex process that requires a series of repetitive but highly coordinated Rats (n = 5) were inoculated in the ear dermis with either Btr wild-type or Btr DbepDE. Blood was drawn at the indicated days post infection (dpi), diluted and plated on sheep blood supplemented Columbia agar plates (CBA) for counting of colony forming units (CFUs). (C) Complementation of the Btr DbepDE mutant with BepE is sufficient to restore bacteremia in rats infected by the i.d. route. Groups of rats (n$3) were infected with the indicated strains by the i.v. or i.d. route. Blood was drawn at 16 dpi and CFUs were recovered as described for B. The graph represents CFUs/ml of blood for individual animals (circles) and their cohort mean (line). Statistical significance was determined using Student's t-test. P,0.05 was considered statistically significant. (D) Heterologous complementation of Btr DbepDE with pBIDs.E Bhe is sufficient to rescue the abacteremia phenotype following infection by the i.d. route. The infections were performed as described for (C). Data represented for BIDs.E Bhe complementation were acquired in separate experiment from the other data shown. P,0.05 was considered statistically significant. doi:10.1371/journal.ppat.1004187.g009 BMDCs were pre-infected with MOI = 50 of the indicated bacterial strains. Infected cells were embedded in collagen and mounted in a trans-well migration system that was prior seeded with a confluent monolayer of iLECs (immortalized lymphatic endothelial cells). BMDCs that migrated though the iLECs were quantified after 24 h. The data normalized to uninfected condition. The bars represent the mean of triplicate samples +/2 SD. Statistical significance was determined using Student's t-test. P,0.05 was considered statistically significant. Data from one representative experiment (n = 3) are presented. doi:10.1371/journal.ppat.1004187.g010 processes [43,44,45,46,47,48,49]. The first step of forward protrusion of lamellipodia and filopodia involves extensive remodeling of the F-actin cytoskeleton. Next, attachment at the leading edge occurs via focal adhesion complexes. After the formation of new adhesions, cells undergo actomyosin-dependent contraction to pull the cell body forward. Coincidently, cell adhesions at the rear of the cell are released such that the cell can move forward [43,44,45,46,47,48,49]. Bhe DbepDEF-infected cells were deficient in rear end detachment and thus underwent fragmentation. Despite the similarity in domain architecture of BepD Bhe , BepE Bhe and BepF Bhe , only BepE Bhe and its homologues in other bartonellae were able to suppress cell fragmentation, which was shown both by bacterial mutant complementation and by ectopic expression of GFP-BepE Bhe within HUVECs.
The cell fragmentation phenotype is T4SS-dependent and strictly requires translocation of BepC Bhe , while we have not tested whether other Beps not encoded in the cell fragmentationinducing strain Bhe DbepDEF (i.e. BepA, BepB or BepG) may also contribute to the cell fragmentation phenotype. As already mentioned, BepC Bhe , BepF Bhe and BepG Bhe , are involved in invasome formation via massive F-actin rearrangements [24,27]. While this process enables Bartonella to be taken up efficiently as large aggregates it is tempting to speculate that the action of BepC and possibly BepG on cell cytoskeleton could affect rear end detachment during cell migration and thereby trigger the cell fragmentation phenotype. Although, by ectopic co-expression with BepC in HUVECs we demonstrated that BepE is reducing the stark influence on a cell cytoskeleton, future studies should address whether BepE acts as a ''metaeffector'' [50], which acts directly on BepC to modulate its function inside host cells, or rather indirectly tempering its activity at a different level.
GFP-tagged BepE Bhe was found to be transiently enriched at the rear edge of migrating cells. This localization is consistent with a potential role of BepE in modulating focal adhesion turnover at the rear edge of the cell as a possible mechanism to counteract the cell fragmentation phenotype triggered by other Bep(s). Focal adhesion kinase (FAK), GEFs for Rho and the Rho effectors ROCK and mDia are crucial for the regulation of the adhesion turnover and rear edge retraction [28,29]. mDia is a formin that produces F-actin filaments by nucleation and polymerization, while ROCK activates myosin to cross-link to F-actin for induction of actomyosin bundles and contractility [51,52]. Functional interferences with any of these components would disturb normal cell migration. Small GTPases are targeted by numerous bacterial effectors and several examples are known for pairs of effectors secreted by a given pathogen that have opposing effects on their regulation. Salmonella SopE and SptP have adopted GEF and GAP functions for CDC42 and RAC1, respectively, that allow to temporally switch between their active and inactive forms; first to induce membrane ruffle-mediated uptake of bacteria, which is followed by stabilization of the remodeled F-actin cytoskeleton [53]. Another example is the Legionella pneumophila effectors Drra/SidM and LepB that are known to regulate the Rab1 activity and dynamic membrane remodeling of the Legionellacontaining phagosome [54]. The Bep-dependent cell fragmentation phenotype, observed in the absence of BepE, might be related to the inhibition of RhoA. Inhibition of RhoA activity by its inhibitor, C3 (C. botulinum exoenzyme), leads to marked morphological changes of migrating monocytes, with frequent polarization and long tails trailed behind the cell body [30] -a phenotype reminiscent to an early stage of cell fragmentation. A counteraction of BepE on the RhoA signaling pathway is supported by the fact that ectopic expression of GFP-BepE Bhe interferes with the inhibitory effect of the C3-based Rho inhibitor 1, as shown by stress fiber preservation in GFP-BepE Bhe -expressing cells. However, BepE was not able to potentiate any stress fibers when using the ROCK inhibitor Y27632. Based on these observations, BepE could be a factor that directly (or indirectly) activates RhoA independently of Rho inhibitor 1-mediated ADP-ribosylation at asparagine 41 [33].
The VirB T4SSs of bartonellae and its multiple Bep effectors evolved as a toolbox facilitating the adaptation to specific mammalian reservoir hosts [15,55]. The prominent activity of BepE homologues in restoring cell migration raised our interest to decipher the in vivo function of this effector protein. Bartonellareservoir host interaction has been studied in a rat model developed by Schulein et al [14]. In this model, intravenously (i.v.) injected Btr wild-type are rapidly cleared from the blood stream, which remains sterile for at least three days. On day four, the bacteria re-appear in the blood where they adhere to and invade erythrocytes. Approximately every five days, a new wave of bacteria is seeded into the blood stream and invades erythrocytes, sustaining the bacteremia for about ten weeks. This is in accordance with other animal models, such as B. grahamii (Bgr) i.v. infection of mice [56]. The intimate interaction of bartonellae with ECs in vitro and the clinical descriptions of Bhe association to human vasculature, together with the disappearance of bacteria from blood for several days following i.v. infection lead to the proposition that vascular ECs represent, at least in part, the ''primary niche'' that is colonized initially and from where bacteria are seeded to the blood stream to invade erythrocytes. Upon i.v. infection the course of bacteremia was similar in cases of wild-type and Btr DbepDE mutant bacteria. In an i.d. infection model that reflects much better the initial dermal stage of Bartonella infection, wild-type bacteria still caused bacteremia, yet with a 4-5 day time delay compared to the i.v. model. In contrast, animals infected with the Btr DbepDE mutant remained abacteremic, while complementation with BepE restored bacteremia. This data suggests (i) that Bartonella presumably initially infects some cell type(s) in the derma, and (ii) that spreading the infection to the blood stream (possibly via the lymphatic system) is BepEdependent. Previous studies based on the i.v. infection model describe components of Btr VirB T4SS to be essential for colonization of the ''primary niche'', while those were dispensable for the subsequent stage of intraerythrocytic infection [17,19]. Addressing an earlier stage of vector-borne infection, our results in the i.d. model demonstrate that the VirB T4SS and specifically BepE are already required at the dermal stage of infection. This revisits our understanding of intracellular niches of Bartonella. We suggest replacing the old term ''primary niche'' by ''blood seeding niche'' and introducing ''dermal niche'' as a new term describing the preceding dermal stage of arthropod-borne infection.
The BepE-dependent phenotypes in vivo (progression from dermal to blood stage infection in the i.d. model) and in vitro (suppression of cell fragmentation and stress fiber disassembly by Rho inhibitor 1 activity) were consistently mediated by the BID domains of BepE Bhe . In fact, most phenotypes mediated by Beps of Bhe are attributed to a BID domain -further to the essential role of the most C-terminal BID domain as part of the bi-partite secretion signal for T4SS-dependent effector translocation [21] this domain has thus adopted diverse effector functions within host cells [15]. The anti-apoptotic activity of BepA Bhe is confined to its single BID domain [27], and the F-actin remodeling activity of BepF is mediated by its multiple BID domains. Surprisingly, the phosphotyrosine-containing motifs in the N-terminal part of BepE that were shown to recruit multiple SH2-domain containing signaling proteins of the host cell [22] revealed no essential role in the here described BepE-dependent in vitro and in vivo phenotypes.
However, we suggest that investigation of these motifs may be relevant in cells infected in the ''dermal niche'', e.g. to modulate immune responses. DCs represent one of the candidate cell types for Bartonella infection in this ''dermal niche''. In steady state skin DCs include immature epidermal DCs, Langerhans cells (LCs) and immature dermal DCs [57]. Those have a high phagocytic activity and constantly sample the environment as sentinels for foreign antigens (Ag) [58]. Ag uptake induces the activation/ maturation of DCs by starting a directional migration towards the draining lymph nodes and up-regulation of MHC class II/costimulatory molecules. Finally, in the lymph nodes DCs form an immune synapse with naive T cells and present Ags, and as a consequence initiate effective adaptive immune responses [59,60]. Considering that BepE is recruited to cell-to-cell contacts, the phosphotyrosine motifs and their proposed mimickry of immune inhibitory receptors [22] could be a valid tool to modulate signaling pathways within the immune synapse and thus altering the initiation of an adaptive immunity. On the way to lymph nodes DCs may serve as transporters of bacterial cargo. This strategy is considered to be used by other bacterial pathogens, such as Brucella abortus [61] or Bordetella bronchiseptica [39], to disseminate within their mammalian hosts. When Btr DbepDE infects the rat reservoir host we think that at the dermal stage fragmentation or impaired migration of yet unknown infected cells may occur that would prevent further dissemination of bacteria and colonization of the blood stream. This is further supported by our finding that migration of BMDCs in a trans-well migration assay is blocked by infection with Bhe DbepDEF, while expression of BepE Bhe in this mutant background restores it. Albeit it is difficult to compare directly the phenotypic changes in DC (altered cell migration) and HUVECs (fragmentation) upon Bartonella infection, yet considering dermal DCs as ''dermal niche'' for Bartonella represents a valid hypothesis to be investigated in vivo. Future work should also address how exactly the complex in vitro phenotype of cell fragmentation translates into the striking in vivo phenotype -lack of spreading from derma to blood.
In summary, this study builds on extending the ''primary niche'' of Bartonella infection that precedes blood stage infection in the mammalian reservoir host to a dermal stage reflecting natural infection by blood-sucking arthropods. Using this i.d. model we describe for the first time a specific role for a Bartonella effector protein in vivo, i.e. BepE is required to spread infection from the dermal site of inoculation to the blood, possibly via promoting normal migration of infected DCs by alleviating deleterious secondary effects mediated by other Beps targeting the RhoA pathway. As a result, this functional crosstalk between Beps could facilitate that Bartonella reaches its replicative niche in the blood, where it multiplies and persists within the immune-privileged intracellular compartment of erythrocytes, from where it is picked up by blood-sucking arthropods for transmission to naive hosts.

Conjugation of Bartonella-expression plasmids into Bartonella
Conjugation of plasmids from the E. coli dapmutant b2150 [26] into Bartonella spp. was done by triparental mating in the presence of helper plasmid pRK2013 as described previously by Dehio et al., 1998 [62].

DNA manipulations
Plasmids used in this study are listed in Table S1, primers are listed in Table S2. Generation of plasmids for expression of Myc-Beps or lentiviral expression, plasmids used for chromosomal integrations and in-frame deletions are detailed described in Materials and Methods S1.

Western blot analysis of protein expression in Bartonella
Western blot analysis of expressed protein levels in Bartonella was performed as described previously by Schulein et al., 2005 [21].

Infection of rats
Female Wistar rats were obtained at the age of 10 weeks from Harlan, RCC-Füllinsdorf. All animal studies were approved by the authors' institutional review boards.
After two weeks of adaptation the rats were infected with Btr. Bacterial strains were grown as described above, harvested in phosphate-buffered saline (PBS), and diluted to OD 595 = 1. Rats were anesthetized with a 2-3% Isoflurane/O 2 mixture and infected with 10 ml or 200 ml of the bacterial suspension in the dermis of the right ear or in the tail vein respectively. Blood samples were taken at the tail vein and immediately mixed with PBS containing 3.8% sodium-citrate to avoid coagulation. After freezing to 270uC and subsequent thawing, undiluted and diluted blood samples were plated on Columbia agar plates containing 5% defibrinated sheep blood (CBA plates) at 35uC and 5%. CFUs were counted after 7-10 days of growth.

Cell culture
Human umbilical vein endothelial cells (HUVEC) were isolated as described before [26] and cultured in EGM medium (Promocell) in a humidified atmosphere at 37uC and 5% CO2.
Mouse bone marrow-derived dendritic cells (BMDCs) were differentiated in vitro using standard protocol [63]. Briefly, bone marrow cells were flushed from the tibias and femurs of Balb/c mouse with culture medium composed of DMEM medium (Invitrogen Life Technologies) supplemented with 10% fetal calf serum (FCS, Invitrogen Life Technologies). After one centrifugation, BM cells were resuspended in Tris-ammonium chloride for 2 min to lyse RBC. After one more centrifugation, BM cells were cultured for 9-10 days at 1*10 6 cells/ml in culture medium supplemented with 200 ng/ml recombinant human Flt3L (produced by 40E1 hybridoma cells, kind gift from prof. A. Rolink). Cultures were incubated at 37uC in 5% CO 2 -humidified atmosphere.

Lentiviral transduction of HUVECs
Subconfluent (3*10 6 ) HEK 293T cells in 10 cm cell-culture dishes were transfected with a total of 5 mg of plasmid DNA following the FuGENE transfection protocol (FuGENE 6 Transfection Reagent, Roche). After 12 h, the cell culture media was replaced and the cells were kept in culture for virus production in the supernatant for additional 24 h. 7*10 4 HUVECs/well was seeded in gelatin coated 6-well plate 24 h before the viral infection. On day of infection the viral supernatant from transfected HEK 293T cells was filtered with 0.45 mm filter and transferred onto the monolayer of HUVECs, 2 ml of viral supernatant in presence of 0.5 mg/ml Polybrene (Sigma) was applied on each well. After the first harvesting of viral supernatant the HEK 293T cells were supplemented with a fresh medium. 24 h later the HUVECs were re-infected repeating the same procedure with followed replacement of the medium in 24 h.

Infection of HUVECs
HUVECs (passage 4-7) were plated on gelatin-coated glass slides in 24-well plates at 5*10 4 /ml using EGM or alternatively 3*10 3 /well seeded in 96-well plate. The next day cells were

Inhibitor treatment of transduced HUVECs
Mixed culture of non-transduced and lentivirally transduced HUVECs by pRO300, pRO301, pRO302, pRO303 or pRO304 were seeded in gelatin coated wells at 3,5*10 3 /100 ml using endothelial complete medium (PromoCell) in 96 well-plates in the morning. Cells were starved in endothelial medium with reduced serum (containing 206 reduced SupplementMix) for overnight. The next day the wells were treated with Rho Inhibitor I (CTO4, Cytoskeleton, Denver, CO) or Y27632 (Sigma) at different concentrations for 4 h or 30 m respectively. The experiment was stopped by fixation with 3.7% PFA, stained for F-actin and nucleus as described above and subjected to automated microscopy.

Image analysis and quantification of cell fragmentation or loss of stress fibers
Experiments performed in 96-well plates were subjected to automated microscopy, using MD ImageXpress Micro automated microscopes. In every well, 666 sites were imaged in 2-3 different wavelengths corresponding the applied cell staining. Images were visualized using MetaXpress software (MDC) and the number of cells per image was determined automatically by MetaXpress inbuild analysis modules (CountNuclei). The cells with the thin elongations or stress fiber-containing cells were defined and counted by eye. The percentage of elongated cells or stress fibercontaining cells was calculated using Microsoft Office Excel. In every condition randomly picked 10 fields of magnification 10 within the well and triplicates of the wells were analyzed.
For confocal laser microscopy, the stained samples were analyzed using a Zeis Point Scanning Confocal, LSM700 Upright microscope (Imaging Core Facility, Biozentrum, University of Basel, Switzerland). Z-stacks with 20-30 focal planes with a spacing of 0.1-0.3 mm were recorded and xz-and yz-planes were reconstructed using Zen 2010 software. Images were exported and finalized using MetaMorph, ImageJ and Adobe Photoshop.

Time-lapse microscopy
Time-lapse microscopy was carried out using an MD ImagXpress Micro automated microscope from Molecular devices (inside a controlled chamber held at 37uC and 5% CO 2 ). Images were recorded using a CoolSNAP ES digital CCD camera and processed using MetaXpress software. Briefly, HUVECs were infected with Bhe strains in 96-well format as described above. After incubating for 8 h, 2 to 4 points in each well were chosen for time-lapse microscopy. The points were imaged for 72 h with 10 min lapse between each imaging. Both, red and green, fluorescence were detected at 206 magnification, and the corresponding videos were processed in MetaXpress and compiled into QuickTime (Apple, Cupertino, CA) movies.

Quantification of cell survival by FACS analysis
Mixed population of non-transduced HUVECs with lentivirally transduced HUVECs by pRO300, pRO301, pRO302, pRO303 or pRO304 were seeded in gelatin coted 6 well-plates at 10 5 cell/ well the evening before experiment. Next day the infection was performed with MOI 200 as described above. After 48 hpi the cells were detached from the wells by using trypsin, resuspended in medium and analyzed for quantification of GFP-expressing cells by FACS (BD Biosciences). The data were analyzed using the FlowJo software.

Infection and Bla analysis of mouse BMDCs
Day 9-10 BMDCs were harvested from the flask by vigorous pipetting followed by washing twice with room temperature PBS without Ca 2+ or Mg 2+ (Invitrogen Life Technologies). 4*10 3 BMDCs were seeded within the wells of 96-well plate. Next day supernatant was replaced with the fresh M199 with Earls salts (M199, Gibco, Invitrogen Life Technologies) supplemented with 10% FCS (Invitrogen Life Technologies) and infected with a MOI of 50, 100, 300. At 3, 6, 9, 12, 24 and 48 h post infection the wells were loaded with 1 mM CCF2-AM (Invitrogen Life Technologies) in M199/10% FCS and incubated at room temperature for 60-120 min. The plate was imaged in two channels (blue and green) by Leica DM-IRBE inverted fluorescence microscope. Data were analyzed in MetaMorph and further processed in Microsoft Office Excel.

BMDC trans-well migration
Trans-well migration assay was performed as described previously [41]. Experiments were performed with 12 mm diameter, 8 mm pore cell-culture inserts (Millipore, Billerica, MA) in a modified Boyden chamber assay. One day before the experiment, iLECs were seeded onto the collagen-coated underside of the chamber at 10 5 cells/well and in parallel 2*10 5 mouse BMDCs were infected with Bartonella at MOI = 200. The next day, infected BMDCs were harvested and stained with DiD-Vybrant (Invitrogen Life Technologies), then seeded in 200 ml (1 mm thick) Matrigel (4.65 mg/ml, BD Biosciences, San Jose, CA) within the inserts. Basal medium was placed in the top and bottom chamber. After 15 h in a 35uC/5% CO 2 incubator, Matrigel containing non-migrated cells was removed and the inserts were fixed in 3.7% PFA. The number of migrated cells was counted by microscope on the underside of the chamber.