Evidence of Bacteroides fragilis Protection from Bartonella henselae-Induced Damage

Bartonella henselae is able to internalize endothelial progenitor cells (EPCs), which are resistant to the infection of other common pathogens. Bacteroides fragilis is a gram-negative anaerobe belonging to the gut microflora. It protects from experimental colitis induced by Helicobacter hepaticus through the polysaccharide A (PSA). The aim of our study was to establish: 1) whether B. fragilis colonization could protect from B. henselae infection; if this event may have beneficial effects on EPCs, vascular system and tissues. Our in vitro results establish for the first time that B. fragilis can internalize EPCs and competes with B. henselae during coinfection. We observed a marked activation of the inflammatory response by Real-time PCR and ELISA in coinfected cells compared to B. henselae-infected cells (63 vs 23 up-regulated genes), and after EPCs infection with mutant B. fragilis ΔPSA (≅90% up-regulated genes) compared to B. fragilis. Interestingly, in a mouse model of coinfection, morphological and ultrastructural analyses by hematoxylin-eosin staining and electron microscopy on murine tissues revealed that damages induced by B. henselae can be prevented in the coinfection with B. fragilis but not with its mutant B. fragilis ΔPSA. Moreover, immunohistochemistry analysis with anti-Bartonella showed that the number of positive cells per field decreased of at least 50% in the liver (20±4 vs 50±8), aorta (5±1 vs 10±2) and spleen (25±3 vs 40±6) sections of mice coinfected compared to mice infected only with B. henselae. This decrease was less evident in the coinfection with ΔPSA strain (35±6 in the liver, 5±1 in the aorta and 30±5 in the spleen). Finally, B. fragilis colonization was also able to restore the EPC decrease observed in mice infected with B. henselae (0.65 vs 0.06 media). Thus, our data establish that B. fragilis colonization is able to prevent B. henselae damages through PSA.


Introduction
Intracellular pathogens, including many different bacteria, live and replicate either within endosomal compartments or in the cytosol of diverse host cells such as macrophages, dendritic cells, neutrophils, fibroblasts, epithelial or endothelial cells or erythrocytes [1,2].
Bartonella henselae belongs to a group of gram-negative facultative intracellular bacteria [3]. It is the causative agent of cat scratch disease, a human infection usually characterized by persistent regional lymphadenopathy [4,5]. The bacterial pathogenicity and clinical manifestations depend on the immune status of the infected host [2,[6][7][8]. Noteworthy, B. henselae is able to adhere and invade EPCs [9] HSCs (hematopoietic stem cells) [10] other than human erythrocytes [11,12]. These circulating progenitors had been previously shown to be resistant to the infection of other intracellular pathogens [13]. Moreover, these cells have been suggested to transport B. henselae to peripheral tissues, particularly to the endothelium of microcirculation where vasoproliferative disorders occur [2,9].
On the other hand, Bacteroides fragilis is a gram-negative anaerobe bacterium and an integral component of the gut microflora of most mammals [14][15][16][17][18]. This intestinal commensal is capable of mediating powerful effects on the host immune system, which are mostly mediated by the capsular polysaccharide A (PSA), an immunomodulatory molecule that mediates T celldependent immune responses [17][18][19]. Indeed, it was shown to protect animals from experimental colitis induced by Helicobacter hepaticus [20] and this beneficial activity was demonstrated to require the PSA [20]. This microbial molecule can exert its effects through the immune system within both the intestinal and the systemic compartments [21][22][23][24].
Thus, in order to gain further insights in our understanding of how commensal organisms, such as B. fragilis, shape host immunity in other models, we investigated whether B. fragilis colonization could protect animal from B. henselae infection and exert beneficial effects on EPCs and vascular system.

B. fragilis Internalizes Progenitor Cells in vitro and Competes with B. henselae during Coinfection
To evaluate the possible capacity of either B. fragilis or the mutant strain defective in the ability to produce PSA (B. fragilis DPSA) to adhere and invade progenitor cells, we infected in vitro cultured EPCs with both these bacterial strains at a multiplicity of infection (MOI) of 100. As control of infection we used B. henselae at 100 MOI, which has been demonstrated to elicit low cytostatic effects [9,25]. The presence of bacteria was revealed after 24 h from infection by confocal microscopy, as shown in Figure 1A. Immmunofluorescence staining with specific antibodies showed that both B. fragilis and its mutant DPSA could internalize EPCs as previously documented for another intracellular pathogen, B. henselae [9]. Longer times of infection (48 h and 72 h) as well as higher MOIs showed similar results (see Figure S1-S2). Thus, we have used a MOI of 100 and 24 h of infection for further experiments.
We also carried out coinfection experiments with B. fragilis, or B. fragilis DPSA, and B. henselae on in vitro cultured EPCs ( Figure 1B). A MOI of 100 was used for all bacteria strains. Interestingly, in the coinfection with B. fragilis a prevalence of Bacteroides was observed whereas in the coinfection with B. fragilis DPSA the presence of Bartonella was more evident ( Figure 1B). These findings suggest B. fragilis, but not B. fragilis DPSA, prevents -or at some extent impairs -the internalization of B. henselae. Cells were observed after 24 h, 48 h and 72 h ( Figure S2).
Overall, these findings clearly demonstrate that both B. fragilis and B. fragilis DPSA are able to internalize EPCs independently of PSA. However, only B. fragilis but not B. fragilis DPSA is able to partially prevent Bartonella internalization.

Analysis of Inflammatory Chemokines and Cytokines
To analyze gene expression changes occurring in cultured EPCs infected by B. fragilis and B. henselae, and to assess the effect of DPSA on the transcription of inflammation-related genes, we used RT 2 Profiler PCR Arrays to simultaneously profile 84 inflammatory genes encoding chemokines, cytokines and their receptors, as recently performed by Sumegi J. et al. [26]. The degree of differential expression for all the inflammatory genes within all analyzed samples is schematized in Figure 2A.
Surprisingly, even though in B. henselae-infected EPCs about 30 (out of 84) inflammatory genes were differentially expressed (DE) compared to uninfected cells, the highest percent of DE genes was detected in B. fragilis DPSA-infected cells ( Figure S3). Particularly, in these cells 78 inflammatory chemokine and cytokine encoding genes (about 93% of all analyzed genes) were DE ( Figure S3). Of note, 75 out of 78 DE genes were significantly up-regulated. On the opposite, infection with the wild-type strain of B. fragilis (i.e. with intact PSA) led to a decreased number of up-regulated inflammatory genes (43 vs 75) and a marked reduction in the expression levels of some crucial genes ( Figure 2B).
In addition, we also evaluated the expression of inflammatory genes during coinfection with B. henselae-B. fragilis and B. henselae-B. fragilis DPSA compared to the infection with B. henselae alone. We observed an increased number of DE genes during both coinfections even though the extent of differential expression of inflammatory genes was more pronounced in B. henselae-B. fragilis DPSA cells. In particular, a direct comparison between the expression levels in the coinfected samples revealed that B. fragilis exerts a significant down-regulation of pro-inflammatory chemokines (shown in Figure 2C) simultaneously providing a positive modulation of anti-inflammatory genes such as IL-10 and Interferon-alpha 2 (IFNA2) ( Figure 2D).
We also analyzed an array of secreted cytokines involved in the inflammation response by enzyme-linked immunosorbent assay (ELISA) ( Figure 2E-2F and Figure S4) on the supernatants of the same infected cells. For most of analyzed cytokines we observed a similar trend of differential expression as observed in the analysis of gene expression levels, even though the extent of DE was different. In particular, similarly to the results of RT 2 Profiler PCR Arrays analysis, in coinfected samples the wild-type strain of B. fragilis exerts a significant down-modulation of pro-inflammatory interleukins ( Figure 2E) and, on the opposite, a positive increase of IL-10, shown to repress in vivo the expression of TNFa and Granulocyte macrophage colony-stimulating factor (GM-CSF) ( Figure 2F).
Overall, these data indicate a marked activation of the inflammatory response in B. fragilis DPSA-infected cells, reflecting both the high ability of such intracellular bacteria to invade EPCs and possibly the more rapid cell division compared to B. henselae. Moreover, such data confirm the role of PSA in modulating the inflammatory response, by both repressing and inducing several chemokine/cytokine-encoding genes, during in vitro infection of cultured EPCs.

Murine Infection and Detection of Bacteria in Tissues from Infected Mice
We have developed a mouse model of coinfection to investigate whether B. fragilis colonization could protect animal from B. henselae infection. In order to determine whether PSA is essential for protection against pathogenic effects of B. henselae, mice were inoculated with either B. fragilis or its mutant B. fragilis DPSA [20]. Three groups of five mice each (n = 5) were infected with 10 9 CFU of either B. henselae or B. fragilis or B. fragilis DPSA respectively, whereas two groups of animals were coinfected with B. henselae and alternatively B. fragilis or B. fragilis DPSA (see Figure 3A). No animals died after infection and the health was evaluated once a day for the full duration of the experiment. After 36 days, animals were euthanized in order to assess both the bacterial infection and B. fragilis protection by histopathology and transmission electron microscopy (TEM) of infected tissues (see paragraphs below).
The presence of the above-mentioned bacteria was documented by TEM and immunofluorescence analysis, which was also used to evaluate the colocalization (fluorescein-rhodamine) of both bacteria in coinfected mice. Bacterial internalization was observed in the endothelial and subendothelial layers around centrilobular veins and veins of the portal triads. There is no evidence of the presence of bacteria within the hepatocytes or between the hepatic laminae ( Figure 3B-3C).
Paraffin-embedded liver samples obtained from infected animals were also examined for the detection of DNA of both bacterial species by PCR. Nested-PCR for the gltA gene (encoding for citrate synthase) (353 bp) and frdA gene (encoding for fumarate reductase) (354 bp) revealed the presence of B. henselae and B. fragilis DNA respectively for at least 36 days in the tissues of all mice examined ( Figure 3D). No bacterial DNA was isolated from control mice (data not shown).

B. fragilis Protects from the Pathogenic Effects of B. henselae Infection in Mice
Morphological analysis by hematoxylin-eosin. The morphological study performed by hematoxylin-eosin staining showed the presence of several alterations in both aorta and liver of mice infected with B. henselae compared to the control uninfected tissues. In the liver, we observed hydropic degeneration, nuclear pycnosis, hepatocytes with granular cytoplasm, activation of Kupffer cells and a granulomatous inflammatory reaction, as described in the literature ( Figure 4) [27,28]. These alterations were localized predominantly in the perivascular region, in the form of histioid aggregates with a mild lymphocytic infiltration ( Figure 4A). Interestingly, the granulomatous infiltrates were less invasive in the group coinfected with B. henselae and B. fragilis DPSA and almost absent in the group of mice coinfected with B. henselae and B. fragilis. This was confirmed by the cytopathic effect and the inflammatory reaction ( Figure 4A). Uninfected control, B. fragilis and B. fragilis DPSA groups were negative for all the described aspects. In the aorta, we documented minimal lesions in B. henselae infected mice. Particularly, focal endothelial swelling, with aspects of hydropic degeneration and hyalinization of basal membranes were observed ( Figure 4B). These changes tend to disappear in the group of mice infected with B. henselae and B. fragilis DPSA andmore evident -in the aortas of B. henselae and B. fragilis infected mice ( Figure 4B). In the spleen of B. henselae infected mice we observed an expansion of the white and red pulp, with a consequent activation of the reticulo-endothelial system (see Figure S5). In mice coinfected with B. henselae and B. fragilis DPSA a reduced inflammatory reaction was observed, with a decreased reactivity of the reticulo-endothelial system. A dramatic reduction in the inflammatory reaction was even more evident in B. henselae and B. fragilis infected mice.
Immunohistochemistry analysis. To confirm the morphological observations we performed immunohistochemistry assays on the liver, aorta and spleen with an anti-Bartonella specific antibody. We determined the relative count of positive cells per field (HPF) in all analyzed sections (20 fields per each section).  (2064 6 HPF) ( Figure 5A). Few granulomatous lesions were observed, mainly concentrated in the parenchyma near the limiting pericentral laminae and, in the area adjacent to the vessels. Focal positivity in perisinusoidal areas was present. Infected cells were also detected in the granulomatous formations.
In the aorta, the percentage of B. henselae infected endothelial cells was minimal, with a count of positive cells of 1062 6 HPF   The spleen of mice infected with B. henselae had a cellular positivity to anti-Bartonella antibody of 4066 cells 6HPF, mainly in the red pulp. Focal positivity in the follicular centers was also evident ( Figure S5). The trend of positivity reduction in both groups of coinfected mice was similar to that observed for liver and aorta tissues ( Figure S5).
The decreased damage and the presence of inflammatory response in coinfected mice, particularly with B. henselae and B. fragilis, suggested us a possible pathogenic mechanism related to bacteria competition within infected cells. This assertion is corroborated by the presence of both bacteria in the cells and by a smaller cytopathic effect that attains a less inflammatory component. An intermediate state of cell damage between the groups infected with only B. henselae strain and those coinfected is represented by the group infected with the mutant B. fragilis DPSA. In this case a reduction of the damage is achieved compared to the B. henselale infected group, although to a lesser extent compared to the group of coinfected mice.
Ultrastructural analysis. The ultrastructural analysis (TEM) performed on the liver of B. fragilis-infected mice ( Figure 6I) revealed a regular architecture of the hepatic tissue with no evidence of any inflammatory and immune response, as in the uninfected mice ( Figure 6A). In contrast, the observation by TEM of livers from B. henselae-infected mice revealed the presence of massive granulomas, due to a B. henselae-induced chronic inflammatory state ( Figure 6B). Such granulomas are characterized by the presence of neutrophils and few bacteria ( Figure 6B and S6C), indicating a still ongoing immune response.
Surprisingly, there was any pronounced inflammatory response in the liver samples of mice coinfected with B. henselae and the wild type strain of B. fragilis ( Figure 6E). Within these samples, the ultrastructural analysis revealed only few macrophages in the portal triads and around the centrilobular veins ( Figure S6D  In all analyzed mice we did not observe the massive presence of bacteria, neither of invasomes, but few individual bacteria were visible in the space of Disse or in the granulomas ( Figure S6A-E I ).
The same ultrastructure investigations were then conducted on aorta's sections isolated from control ( Figure 6C) and infected mice (Figure 6 panels D, G, H, K and L). We did not observe the presence of granulomatous formations in any of the samples analyzed. The only exception is the presence of white cells, not organized to form a granuloma, observed in the adventitia of the aorta of B. henselae-infected mice ( Figure 6D). Aorta, in all analyzed samples, has not significant morphological alterations, except for a henselae and B. fragilis, respectively wild type and DPSA strains ( Figure 6G-6H). These bulges may be regarded as a reaction of the organ to bacteria passage, in particular of B. henselae, indeed, they have not been found in the aortas of mice infected with either B. fragilis and B. fragilis DPSA (Figure 6K-6L). Even in the aorta, we did not observe a massive number of bacteria and of invasomes, but they were found only sporadically ( Figure S6F-J I ).
EPC number in infected mice. At 36 days after infection, before mice sacrifice, blood was collected in order to establish whether the EPC number was influenced by infection in all the studied groups. As shown in Figure 7, the mean number of EPCs, measured as Sca-I/Flk-I positive cells, dramatically decreased of about 90% in mice infected with B. henselae (0.06) compared to the uninfected group, whose EPC number was considered as 100% (1.0). Means were calculated by standard error method. This number was partially restored in mice coinfected with B. fragilis or B. fragilis DPSA and B. henselae (0.65 and 0.42 means respectively).
Thoroughly, these results indicate that B. fragilis, but not B. fragilis DPSA, is able to prevent the decrease of EPC induced by B. henselae infection.

Discussion
The results of this study establish that: 1) B. fragilis is able to adhere and invade EPCs; 2) B. fragilis competes with B. henselae during in vitro coinfection assays; 3) the immunomodulatory properties of B. fragilis depend on PSA; 4) damages induced by B. henselae in liver and aorta of infected mice can be prevented by colonization with B. fragilis but not with its mutant B. fragilis DPSA; 5) EPC impairment prompted by B. henselae can be counteracted by B. fragilis colonization.
Our in vitro studies indicate that B. fragilis is able to adhere and invade EPCs and it competes with B. henselae during coinfection. Furthermore, we observed a marked activation of inflammatory response in B. fragilis-infected cells compared to B. henselae-infected, as determined by Real time PCR and ELISA of pro-inflammatory chemokines and cytokines. In the mouse model of coinfection, morphological and ultrastructural analyses showed that the damage induced by B. henselae in liver and aorta can be prevented in the coinfection with B. fragilis but not with its mutant B. fragilis DPSA. Specifically, the number of B. henselae-positive cells per field counted by immunohistochemistry analysis decreased of at least 50% in the liver, spleen and aorta sections of mice coinfected with both bacteria compared to mice infected only with B. henselae. This reduction was less evident in the tissue from mice coinfected with B. fragilis DPSA suggesting a possible role of the PSA also in this model. Finally, B. fragilis colonization was also able to restore the EPC number decrease observed in mice infected with B. henselae.
The findings of this complex study were achieved by performing a great bulk of experiments in vitro as well as in vivo. We have previously demonstrated that B. henselae internalizes EPCs [9,29], also documenting the gene expression changes occurring during infection [9,30]. A further study indicated that EPCs can be also a potential reservoirs of Kaposi's sarcoma-associated herpesvirus [31], thus extending our findings to other microorganisms. Noteworthy, this is the first study demonstrating that also B. fragilis can adhere to and invade EPCs as shown by confocal observations. Moreover, its internalization is not fully dependent on PSA since also the mutant DPSA strain exhibits such ability. This important finding has been achieved both in vitro on EPCs by confocal microscopy (Figure 1) and in vivo on infected mice tissues analyzed by ultrastructural analysis (Figure 6 and S6). To our knowledge, TEM examination has been used only in a few in vitro studies to explore cell internalization of B. fragilis but never for B. fragilis DPSA [32,33]. A previous study showing the ability of B. fragilis PSA to stimulate an anti-inflammatory background in the intestine suggested that PSA may be critical for the establishment of a commensal intestinal flora [34]. Indeed, colonization experiments indicated that B. fragilis DPSA could not colonize the colonic epithelium as efficiently as the wild type strain [34]. Such different capability in tissue colonization has suggested that PSA may function as an adherence factor for B. fragilis, or reflect the immunomodulatory properties of PSA [18]. The protective role of B. fragilis, recently shown to be sufficient to correct some immune system deficiencies in the absence of bacterial colonization, and also to prevent intestinal inflammation in animal models of colitis [19,20].
Our results corroborate the latter hypothesis. Indeed, on one side, both strains are able to adhere and invade cells, independently of the presence of PSA; on the other side, we confirmed the PSA immunomodulation ability through the analysis of inflammatory chemokine/cytokine during in vitro infection of EPCs.
In our previous study [9], we also showed the detrimental effects of B. henselae on EPC number as well as the deregulation of inflammatory gene networks after in vitro infection [30]. Here, we document in vivo the Bartonella-induced damage in the tissues of infected mice, also showing for the first time a dramatic decrease in EPC number compared to uninfected control mice. Interestingly, B. fragilis colonization was also able to restore the EPC number decrease occurring in mice infected with B. henselae (0.65 vs 0.06 means).
The absence of a marked inflammatory response evidenced by TEM in mice coinfected with B. fragilis and the presence of granulomas in the liver as well as of cells of the immune system in the adventitia of mice infected with only B. henselae are all signs of the protective effect of the PSA molecule of Bacteroides. This effect was also evidenced by a weaker inflammatory response in mice coinfected with B. fragilis DPSA. This finding suggests that even other molecules could be involved in the process of immunoregulation and exert a protective effect.
Moreover, a previous paper has shown that toxins of B. fragilis modify the actin cytoskeleton inducing disruption of intercellular junctions [35]. However, we have seen cellular junctions perfectly arranged between the liver cells (data not shown); this result indirectly demonstrates the inability of these bacteria to invade the liver cells, as observed in our TEM data where bacteria are seen only in endothelial cells or in the space of Disse.
In coinfected mice we also documented a decreased presence of granulomas in liver and minimal changes in aorta, thus demonstrating both the tissue damage reduction and a lower presence of B. henselae within the cells, with a consequent decrease in its cytopathic effect. In few granulomas we found a lower extension and an intracellular colocalization at the immunofluorescence with B. fragilis. According to colocalization experiments, we can assess that the expression of granulomas is due to a protective effect of B. fragilis that competes with B. henselae for adhesion sites, leading to a lower cytopathic effect, reduced necrosis and impaired release of pro-inflammatory molecules, resulting in a global protective effect. We can therefore conclude that the reduced amount of infected cells in coinfected mice leads to impaired cellular suffering, occurring in the few cells where both bacteria colocalize and where B. henselae-induced cytotoxic effect is exhibited. The lower cytotoxic effect is likely to be a direct consequence of the reduced expression of several inflammatory molecules and, in turn, an impaired inflammatory response (immunohistochemical data). Such process is usually related with the pathogenicity of a bacterial agent. The increase of the cell damage induced by B. henselae is shown in our model by an inflammatory infiltrate that is more conspicuous in the liver.
Taking together our data strongly indicate that B. fragilis protects from B. henselae induced damages, including vascular ones. This relevant finding may provide a basis for the development of novel therapies against such infections.

Bacterial Strains and Growth Conditions
The B. henselae strain ATCC49882 (LGC Promochem) was stored at 280uC in tryptone soya broth USP (TSB) (Oxoid) until use, and grown on Columbia agar supplemented with 5% defibrinated sheep blood (Oxoid) in a humidified atmosphere at 37uC and 5% CO 2 for seven days. B. fragilis NCTC9343 and its mutant B. fragilis DPSA were kindly provided by Professor Dennis L. Kasper. The strains were stored at 280uC in brain heart infusion broth (Oxoid) until use, and grown anaerobically onto Schaedler agar with 5% sheep blood at 37uC for 48 h.

EPC Isolation and Cultivation
Human EPCs were isolated from total peripheral blood mononuclear cells (PBMCs) as previously described in detail [9,36,37]. Written informed consent was obtained from donors according to the Declaration of Helsinki. The cultures were stopped at the established time and EPCs were washed three times with phosphate buffer solution (PBS) to remove non-adherent bacteria. For selective removal of extracellular bacteria, 100 mg/ml gentamicin (Schering-Plough) was added. To determine CFU [9], the cells were washed 2 h with PBS after addition of gentamicin, lysed by the addition of ice-cold distilled water, and incubated for 59 on ice. The MOI used for each experiment was confirmed by plating serial dilutions.

Confocal Immunofluorescence Microscopy
Cultured cells were fixed with 3% paraformaldehyde for 209 at RT and permeabilized with 0.1%Triton-X100. Cells were then incubated 609 at RT with specific antibodies. Anti-B. henselae primary antibody (Novus Biologicals), diluted 1/100, was then revealed with secondary antibodies conjugated to Alexa Fluor 488 (1:500; Invitrogen, Carlsbad, CA) for 309 at RT. Anti-B. fragilis Lipopolysaccharides (LPS) primary antibody (Santa Cruz Biotechnology, INC.) was diluted 1/250 and revealed with secondary antibody conjugated with tetramethylrhodamine isothiocyanate, diluted 1/20 (TRITC, Dako). Microscopy was performed using Zeiss LSM 510 confocal microscope equipped with a planapochromat x40 (NA 1.4) oil immersion objective and cropped to higher magnification. Fluorescence emissions were acquired in multi-track mode using LP 505 filter. To minimize noise, each image was averaged three times.

Measurement of Cytokines in Cell Culture Supernatants
Cell-free supernatants were collected from infected EPCs and assayed for levels of cytokines and chemokines by commercial enzyme-linked immunosorbent assay (ELISA) kits according to the manufacturer's instructions (SABiosciences A Qiagen Company). The concentrations of cytokines (pg/ml) IL1A, IL 1B, L2, IL4,  IL6, IL8, IL10, IL12, IL17A, IFNc, TNFa and GM-CSF were measured using an ELISA reader (Tecan). Results are presented as mean error of the values 6 standard mean. The Student's t test was used for the comparison of means, and regression analysis was used to evaluate any correlation between different cytokine levels. A P value of ,0.05 was considered significant.

RNA Isolation and Quantitative Real-Time PCR Assays
Total RNA was isolated from in vitro cultured EPCs infected with B. henselae, B. fragilis, B. fragilis DPSA and uninfected, using standard TRIZOL (Invitrogen) protocol as previously described [37,38]. Integrity and concentration were evaluated by gel electrophoresis and Nanodrop. The RNA quality of each sample was good, with 260/280 ratio in the range between 1.8 and 2.0. All purified RNAs were then converted into cDNA using the RT 2 First Strand Kit (SABiosciences, Frederick, USA). Resulting cDNAs were used for the quantitative Real-Time PCR assays performed on a specific RT 2 Profiler TM PCR Arrays (SABiosciences) [26].
In particular, samples were analyzed on the 96-well plate Human Inflammatory Cytokines & Receptors RT 2 Profiler PCR Arrays (PAHS-011). Such arrays provide a mix between the largescale ability of microarrays to analyze gene expression and the quantitative sensitivity of Real-Time PCR assay, with highly reproducible results.
A total of 89 genes (including 5 house-keeping genes) was analyzed on this PCR-array for each sample. Briefly, cDNAs were independently added to the RT 2 SYBR Green qPCR Master Mix (SABiosciences, Frederick, USA) and then spotted on 6 Human Inflammatory Cytokines & Receptors PCR-arrays. All the experimental procedure was performed according to the manufacturer's instructions for the 7900HT Real Time PCR system (Applied Biosystems). Melting curves were generated after the amplification of 40 cycles.
Relative gene expression was measured by using the 2 2DDCt method. Values of threshold cycle (Ct) were converted to 2 2Ct in order to be proportional to the amount of transcripts in the samples. Ct values $35 cycles were considered undetectable. In order to compare the samples among them, 2 2DCt were calculated by normalizing expression data by the mean of housekeeping genes (as specified for PCR-Arrays). For comparing gene expression data derived from the different experimental conditions, we calculated 2 2DDCt values, obtained by normalizing experimental data by reference data (uninfected EPCs) [26]. Data derived from such analysis were also checked by using an Excel sheet downloaded from the manufacturer's website (http://www. sabiosciences.com/pcrarraydataanalysis.php). For each PCR reaction, the Excel sheet allows data normalization correcting all Ct values by the average Ct values of constantly expressed housekeeping genes present on the array. In our experiments we observed that among the standard housekeeping genes, Hypoxanthine-guanine phosphoribosyltransferase (HPRT) was highly variable among samples, thus it was not considered for further normalization of data. Differentially expressed genes were defined as those with at least 2-fold change variation.

Ethics Statement
The mice were fed with laboratory food pellets and tap water ad libitum, and were bred and housed under specific pathogen free conditions at CEINGE-Advanced Biotechnologies, Naples, Italy. All animal experiments were carried out according to institutional guidelines. All efforts were made to minimize animal suffering and to reduce the number of mice used, in accordance with the European Communities Council Directive of November 24, 1986 (86/609/EEC). The protocol of the study has been reviewed and approved by Ethical Animal Care and Use Committee ''Federico II'' University of Naples, Prot. 2012/0085083.

Murine Strains and Infection
Specific pathogens free (SPF) C57BL/6J, female mice 8 weeks of age, were purchased from Charles River Laboratory. B. fragilis or B. fragilis DPSA was given to animals in concentration 10 9 CFU in 0.1 ml of PBS per mouse by oral administration. Mouse enteric colonization was initiated 7 days prior to B. henselae infection, and the bacterial suspension was given 4 times on alternate days. B. henselae was prepared as described above and thawed on ice prior to infection. Mice were infected with B. henselae via intraperitoneal route at dose of 10 9 CFU/mouse suspended in TSB. Control mice were injected with the same medium in the absence of bacteria. Following infection, mice were observed daily for signs of illness. At indicated time after infection (36 days) blood was withdrawn by cardiac puncture before sacrifice and collected in heparinized tubes. At the times of euthanasia, liver, spleen and aorta were excised and fixed for molecular and microscopic analysis.

Detection of B. henselae and B. fragilis DNA by PCR
To determine the presence of bacterial DNA in mouse tissue, we developed a nested-PCR assay to amplify a fragment of the B. henselae gltA gene (outer: 674 bp; inner: 353 bp) [28] and B. fragilis frdA gene (outer: 599 bp; inner: 354 bp). Slices of infected liver tissue (20 mm thick), from paraffin-embedded organs, were provided by the Comparative Pathology Facility of CEINGE-Advanced Biotechnologies. DNA was isolated by using the QIAamp tissue extraction kit (Qiagen, Hilden, Germany) according to the manufacturer's protocol.
PCR was performed with AmpliTaq Gold polymerase (Applied Biosystems) in a volume of 50 ml in two sequential tubes. Following initial denaturation for 49 at 95uC, amplification with outer primers was performed with 30 cycles of 19 at 94uC, 19 at 70uC, and 29 at 72uC. Amplification with inner primers consisted of 30 cycles of 19 at 94uC, 109 at 30uC, and 19 at 72uC. The final extension step was extended to 109 in both rounds of PCR. Several negative controls were included to exclude cross-contamination [28]. The expected amplicons were visualized by electrophoresis in an agarose gel and by staining with ethidium bromide. The expected amplicons were detected by using two primer pairs: OuterBh R/L (674 bp), InnerBh R/L (353 bp); OuterBf R/L (599 bp), InnerBf R/L (353 bp) (Supporting information Table S1); due to the small number of bacterial cells in the analyzed sample, the Outer amplicons are not detectable by electrophoresis. The assay confirmed the presence of B. henselae, B. fragilis and B. fragilis DPSA in infected and coinfected tissue samples. PCR was repeated twice from two different DNA liver samples.

Immunohistochemistry
All samples (liver, spleen and aorta) for immunohistochemistry were fixed in formaldehyde (4%), embedded in paraffin and sliced into 5 mm thick sections were cut; one slide was stained with haematoxylin-eosin to verify the presence of a sufficient amount of bacterial cells and the remaining were used for IHC analyses. After de-paraffinization, slides containing two serial sections were processed for heat-induced epitope retrieval. When cooled, slides were washed in PBS and then incubated for 609 at 37uC with the primary antibody. Arterial segments, livers and spleens were stained with a mouse monoclonal antibody anti-B. henselae diluted 1/100 (Novus Biologicals). After a further wash in PBS, sections were incubated with the secondary antibody biotinylated for 159 at 37uC. Epitopes recognized by the primary antibody were detected by an avidin-biotin-peroxidase method [39]. A count of positive cells per field (at 2066 magnification) was made and 20 fields per each section were analyzed. The relative means and standard deviation (6SD) were calculated.

Immunofluorescence
A number of staining procedures were available for localizing antigen-antibody reaction in tissue by fluorescence. Combination of either fluorescein-and rhodamine-conjugated antibodies in a single solution using the direct staining method will simultaneously detect two antigens in the same section. Liver tissues were collected, preserved and fixed as for immunofluorescence staining, were incubated with the two antisera in two different times (twostep method). Briefly, sections were deparaffinized and then applied normal serum. Samples were incubated with B. henselae primary antibody (Novus Biologicals), diluted 1/100, and washed in TRIS-HCl buffer. Samples were incubated with secondary antibody conjugated with tetramethylrhodamine isothiocyanate, diluted 1/20 (TRITC, Dako) and then washed. Sections were incubated with B. fragilis LPS primary antibody (Santa Cruz Biotechnology, INC.), diluted 1/250 washed in TRIS-HCL buffer, and incubated with secondary antibody conjugated with fluorescein isothiocyanate diluted 1/20 (FITC, Dako) and then washed. Antibodies incubation time was 1 h for each step. The observation was performed with a Confocal Microscope LSM 510 (Zeiss) by three independent observers.

Transmission Electron Microscopy (TEM)
The aorta and liver tissues were cut into 1 mm 3 blocks and fixed in 1.5% glutaraldehyde in 0,067 M cacodylate buffer at pH 7.4 [40]. The samples were left in fixative for 3 h at 4uC and then washed in 0,134 M cacodylate buffer, pH 7.4, at 4uC. Then they were post-fixed in 1% osmium tetroxide in 0,067 M cacodylate buffer, pH 7.4, at 4uC for 1 h and after washed in 0,134 M cacodylate buffer, pH 7.4, at 4uC. Fixed tissues were dehydrated in ascending series of ethyl alcohol and then embedded in epon. Semi-thin (1,5 mm) sections of TEM embedded liver and aorta samples were cut with a glass knife for light microscopic observations. Sections were stained with 1% toluidine blue solution prepared in 1% sodium tetraborate buffer. Ultra-thin (50-80 nm) section were cut and stained first with uranyl acetate 3% in 50% ethyl alcohol and then with 2,6% lead citrate. These sections, loaded on 200 mesh grids, were observed in a JEM-1011 (JEOL) transmission electron microscope at 100 kV.

EPC Number in Infected Mice
A single blood aliquot was collected by cardiac puncture from the ventricle under anaesthesia, with Tribromoethanol 240 mg/ Kg, using appropriate size needles and collected into heparincoated tubes. A volume of 100 ml blood was incubated for 309 in the dark with Fluorescein isothiocyanate (FITC)-conjugated monoclonal antibody against mouse (Ly-6A/E) Phosphatidylinositol-anchored protein (Sca-1) (BD Pharmingen) and Phycoerythrin (PE)-conjugated antibody against mouse Fetal liver kinase 1 (Flk-1) (BD Pharmingen). Isotype-identical IgG antibodies were used as controls (BD Pharmingen) [41][42][43]. After incubation, erythrocytes were lysed using FACS lysing solution (Becton-Dickinson), washed with PBS and analysed on a FACScan (Becton-Dickinson) using CellQuest software. Each analysis included approximately 10000 events, referred to the total number of white cells.