Brucella β 1,2 Cyclic Glucan Is an Activator of Human and Mouse Dendritic Cells

Bacterial cyclic glucans are glucose polymers that concentrate within the periplasm of alpha-proteobacteria. These molecules are necessary to maintain the homeostasis of the cell envelope by contributing to the osmolarity of Gram negative bacteria. Here, we demonstrate that Brucella β 1,2 cyclic glucans are potent activators of human and mouse dendritic cells. Dendritic cells activation by Brucella β 1,2 cyclic glucans requires TLR4, MyD88 and TRIF, but not CD14. The Brucella cyclic glucans showed neither toxicity nor immunogenicity compared to LPS and triggered antigen-specific CD8+ T cell responses in vivo. These cyclic glucans also enhanced antigen-specific CD4+ and CD8+ T cell responses including cross-presentation by different human DC subsets. Brucella β 1,2 cyclic glucans increased the memory CD4+ T cell responses of blood mononuclear cells exposed to recombinant fusion proteins composed of anti-CD40 antibody and antigens from both hepatitis C virus and Mycobacterium tuberculosis. Thus cyclic glucans represent a new class of adjuvants, which might contribute to the development of effective antimicrobial therapies.


Introduction
Cyclic glucans are intrinsic components of the envelopes of Gram negative bacteria such as Agrobacterium, Rhizobium, Ralstonia solanacearum, Xanthomonas campestris, Rhodobacter sphaeroides and Brucella spp. [1]. Brucellae are intracellular pathogens of mammals, including humans [2]. The pathogenesis of brucellosis is linked to the ability of Brucella to survive and replicate inside host cells [3] through the expression of several effector molecules [2]. In particular, the periplasmic cyclic glucan is required for B. abortus intracellular trafficking [4][5][6] through the recruitment of the raft protein flotilin-1 at the site of the Brucella-containing vacuole [6]. This polysaccharide is built of a cyclic backbone of 17 to 25 glucose units in b-1,2 linkages (CbG). CbG are abundant as they represent 1-5% of the bacteria dry weight. If the CbG content of a single bacterium is released inside a Brucella-containing vacuole, its concentration in the vacuole can reach 10 mM. The release of CbG in mM concentration upon bacterial killing through immune mechanisms [6] might alter the host immune responses.
The discovery of vaccination is one of most important medical discoveries in the history and a turning point in the war between microbes and humans [15,16]. The goal of vaccination is to induce long-lasting protective immunity from infection and prevent infectious diseases. Vaccines operate through the activation of antigen-presenting cells such as DC that eventually stimulate antigen-specific T and B lymphocytes. Unlike attenuated live vaccines, killed whole organisms or subunit vaccines generally require the addition of adjuvants to be effective. Adjuvants promote and enhance immune responses to vaccine components [15,16]. It is now clear that adjuvants activate DC [15]. Adjuvants derived from microorganisms stimulate DC directly, leading to the up-regulation of cytokines, MHC class II, and co-stimulatory molecules and to their migration to the T cell area of lymph nodes. These pathogen-associated molecular patterns (PAMP) activate pattern-recognition receptors (PRR), which act as microbial sensors and are expressed by DC and other leukocytes [13,14]. Most of PAMP used as vaccine adjuvants, like CpG oligonucleotides and monophosphoryl lipid A (MPLA), are agonists of Tolllike receptors (TLR) [17,18]. For infectious as well as for noninfectious diseases, TLR activation have been used in both established and experimental vaccines [17,18].
Bacterial components are often potent immune activators trough commonly associated with toxicity, for example, bacterial DNA with immunostimulatory CpG motifs that bind TLR-9 are potent cellular adjuvants. Overall, several hundred natural and synthetic compounds have been identified to have adjuvant activity such as microbial products, mineral salts, emulsions, microparticles, and liposomes. Although many are more potent than alum, the almost universal human vaccine adjuvant, toxicity is the limiting step for their use in humans. Consequently there is a major unmet need for safer and more effective adjuvants suitable for human use [15][16][17].
We show here that Brucella CbG is neither toxic nor immunogenic when compared to LPS. It is a potent activator of DC thereby triggering antigen-specific CD8 + T cell responses in vivo. Brucella CbG enhance antigen-specific CD4 + and CD8 + T cell responses including cross-presentation by different human DC subsets.

Brucella cgs-mutants are poor inducers of DC maturation
Wild type B. abortus triggers limited activation of mouse bone marrow-derived dendritic cells (BMDC) [19]. Infection of BMDC with B. abortus cgs-(cyclic glucan synthase) mutant failed to activate BMDC as measured by the production of TNF-a and IL-12 ( Figure S1A and S1B). Likewise these infected DC displayed a low expression of immune co-stimulatory molecules such as CD80, CD83 and CD86 (not shown). This pilot findings led us hypothesize that CbG indeed acts as a DC activation molecule. To this end we had to ensure that the CbG preparations would not be contaminated by the potent DC activators LPS and lipid A. Whereas CbG are highly soluble in water, B. abortus and B. melitensis LPS partition in the phenol phase of the classical Westphal hot water-phenol extraction procedure. Thus, this extraction method was applied twice to a CbG water extract previously digested with nucleases and proteinase K. The identity of the CbG was established by several methods, including 13 C-NMR, and the absence of LPS tested by both conventional analytical methods (SDS-PAGE, inability to elicit anti-LPS antibodies, and Kdo analysis). MALDI-TOF analysis further showed both the spectra expected from Brucella CbG and the absence of molecular species signalling like Brucella lipid A ( Figure  S2A and S2B).

Cyclic glucan activate murine DC
Mouse BMDC were incubated with synthetic methyl-b-cyclodextrin (MbCD) and cyclic glucans purified from Brucella melitensis, Brucella abortus and Ralstonia solanacearum. Brucella CbG consists of a cyclic backbone of 17-25 glucose residues linked in b-(1R2) associated with O-succinyl modifications [20,21]. Ralstonia cyclic a-glucan (CaG) is composed of 12 glucoses linked in eleven b-(1R2) plus one a-(1R6) linkages [20]. MbCD consists of 7 Omethyl substituted glucoses in b-(1R4) linkages. Both B. melitensis and B. abortus CbG induced DC to express levels of CD80, CD86, CD40 and MHC II molecules, comparable to those elicited by E. coli LPS ( Figure 1A). The two Brucella CbG induced the secretion of high levels of pro-inflammatory cytokines such as TNF-a and IL-12 (Figures 1B). The induction was dose-dependent ( Figure  S3A and S3B). These findings contrast with the poor DCactivating ability of Brucella LPS [2]. When compared to Brucella abortus CbG, the synthetic MbCD did not stimulate DC (**p,0.01) and the Ralstonia CaG hardly induced (**p,0.01) the production of TNF-a and IL-12 ( Figure 1B). Thus Brucella melitensis CbG is a potent activator of mouse DC.
We then compared DC activation induced by B. melitensis CbG to that induced by linear ß1-3 glucans (curdlan), which bind to Dectin-1 and activate DC in a MyD88/TRIF-independent manner [22,23]. First, several monoclonal antibodies specific for Dectin-1 that can inhibit curdlan-mediated activation failed to inhibit CbG-mediated DC activation (not shown). While double MyD88/TRIF 2/2 BMDC did not respond to CbG (**p,0.01), they secreted high levels of IL-12 in response to curdlan ( Figure 2C). These results indicate that CbG and curdlan use

Author Summary
Vaccination is one of the key strategies to fight against infectious diseases though numerous diseases remain without appropriate vaccines. The challenge is to generate potent vaccines capable of inducing long-lasting immunity in humans. Successful vaccines include adjuvants that enhance and appropriately skew the immune response to given antigens. The development of new adjuvants for human vaccines has become an expanding field of research. Here we show that bacterial cyclic b-glucans can be used to enhance cellular immunity by activation of dendritic cells, from both mice and humans. In particular, Cyclic-b glucans enhance the in vitro memory CD4+ T cell responses of patients suffering from hepatitis C and tuberculosis. Thus cyclic-b glucans are new adjuvants, which might be used in vaccines.
different signalling pathways and that Dectin-1 is not the receptor for CbG.
Altogether, these data show that CbG signalling is dependent on TLR4 but does not use CD14 as a co-receptor.
Brucella CbG is neither toxic nor immunogenic LPS displays a toxicity that precludes its use as a vaccine adjuvant. Previous studies in cell cultures revealed that CbG, when compared to MbCD was not cytotoxic even at very high concentrations [6]. In addition, the Limulus Ameobocyte Lysate (LAL) test showed that CbG preparations did not contain significant endotoxin levels. To assess CbG toxicity in mice LD50 (Lethal dose 50%) was determined by injecting increasing amounts with death being recorded at 12 h, 24 h, 48 h, 72 h postinjection. Results showed that more than 500 mg of Brucella CbG were required to kill 50% of mice, when compared to 65 mg of E. coli LPS.
To determine the immunogenicity of CbG, Balb/C mice were injected with by PBS, E. coli LPS, B. melitensis LPS or B. melitensis CbG. Primary and secondary antibody responses were analyzed and total immuoglobulin levels quantified by ELISA. Unlike E. coli LPS or B. melitensis LPS, B. melitensis CbG did not induce the generation of specific antibodies ( Figure S4A). We also analyzed CbG-mediated ability to induce pro-inflammatory cytokines in the sera of immunized mice by Cytokine Bead Array (CBA) assay ( Figure S4B). C57Bl/6 mice were injected with PBS, CbG, monophosphoryl-lipid A (MPLA), LPS or Poly I/C. After 6 h, 24 h and 72 h of immunization, mice were bled and cytokine levels were measured. At 6 h post-immunization CbG, MPLA and PolyI/C did not induce pro-inflammatory cytokine secretion in contrast to E. coli LPS ( Figure S4B). Altogether, these data show that Brucella CbG is neither toxic nor immunogenic in mice.

Brucella CbG increases antigen-specific CD8 + T cell responses in vivo
Given its ability to activate DC and its low toxicity profile we wondered whether administration of CbG to mice would enhance antigen-driven cellular immune responses in vivo. We used transgenic mice (OT-I Rag 2/+ ) that express a CD8 + T cell population specific for ovalbumin (OVA) as well as a congenic Ly5.1/Ly5.2 mouse model with 2 allelic forms of CD45. We transferred CD8 + Ly5.2 CFSE + OT-I T cells into C57Bl/6 Ly5.1 mice, and immunized them subcutaneously with either OVA alone, or a mix of OVA with CbG or OVA with Poly I:C or OVA with MPLA [15]. At day 3 post-immunization, OVA-specific OT-I T cell proliferation of all immunized mice was detected in the draining lymph nodes (Figure 3). OVA-specific T cells from mice immunized with Poly I:C, MPLA and CbG showed an upregulation of CD25 and a down-regulation of CD62L, which correlated with T cell migration from lymph nodes to the sites of infection. OVA-specific T cells from mice treated with CbG showed a stronger down-regulation of CD62L expression than those treated with Poly I:C or MPLA (Figure 3).
At day 6 post-immunization ( Figure S5), OVA-specific T cells from CbG + OVA-immunized mice displayed stronger antigenspecific OT-I T cell proliferation and activation than those from mice immunized with OVA alone. Comparable responses were detected using the three adjuvants ( Figure S5). We concluded that CbG is able to enhance antigen-specific CD8 + T cell responses in vivo.
To study CbG capacity to induce local inflammation, mice were immunized either with MPL, LPS, CbG or Poly I/C by skin intradermal injection. At 48 h post-treatment, both the adjuvanttreated and untreated ears were collected for histological analysis of cutaneous inflammation ( Figure S6). The skin of adjuvanttreated mice revealed marked increase in ear thickness accompanied by inflammatory cell infiltration ( Figure S6). The inflammation was observed in all mice immunized with different adjuvants. Animals developed acute diffuse dermatitis characterized by heavy neutrophilic infiltration associated with hyperemia of dermal capillaries with neutrophilic margination and well-developed edema of the dermis ( Figure S6). Similarly to other known adjuvants, CbG is capable of inducing local inflammation.

Brucella CbG activates human DC subsets
We next analysed whether CbG was able to activate different human DC subsets (Table 1). We included myeloid DC isolated from blood, dermis and epidermis [28,29] as well as DC generated in vitro by culturing monocytes in the presence of GM-CSF with either IL-4 or IFNa. In response to CbG, all tested DC showed increased expression of HLA-DR, CD40, CD86, and CD83 and increased secretion of IL-12, IL-6 and TNF-a (Figures 4 and 5). Notably, CbG did not activate pDC as measured by cell surface phenotype and secretion of IFN and pro-inflammatory cytokines such as TNF-a and IL-12p70.
To eventually distinguish the effects of LPS from those of CbG, blood mDC from five donors were activated for 6 h with either LPS or CbG and the early transcriptional responses was assessed using microarray profiling. 562 transcripts were significantly modulated in CbG-treated mDC as compared to media controls ( Figure 4A). These genes displayed a similar expression profile in LPS-treated mDC, albeit with lesser global intensity as measured by the molecular distance to media. Statistical comparison (Ttest, p-value 0.01, no correction) between CbG-treated and LPStreated mDC yielded 133 differentially regulated transcripts (data not shown), highlighting the similarities and the differences between the two stimuli. Ingenuity Pathway Analysis (IPA) identified DC maturation as the most significantly represented canonical pathway among genes over-expressed in CbG-treated mDC ( Figure 4B), and PDFG signalling as the most represented pathway among genes under-expressed in CbG-treated mDC ( Figure 4C). Furthermore, CbG-treated mDC displayed increased transcription of the co-stimulatory molecules CD40, CD80, CD86, CD70 and 4-1BBL, but decreased transcription of the Th2 co-stimulatory molecule OX40-L ( Figure 4D). The most significantly over-expressed transcripts in CbG-treated mDC are represented as a network ( Figure 4E) depicting a strong proinflammatory response network. Overall, the global transcriptional  changes that CbG elicit in mDC support the concept that this molecule might enhance DC-dependent T cell responses.
CbG activated blood mDC elicit CD8 + T cell priming We focused our attention to the effects of CbG on human blood mDC. Like E. coli LPS, CbG increased the surface expression of CD40, CD83, CD86 and MHC II ( Figure 5A). Furthermore, CbG-treated mDC secreted high levels of IL-6, TNF-a and IL-12 (p40) ( Figure 5B) and were efficient at inducing allogeneic naïve CD4 + and CD8 + T cell proliferation ( Figure 5C). E. coli LPSactivated DC were slightly more potent than CbG-activated DC at inducing the proliferation of naïve allogeneic T cells ( Figure 5C).
DC loaded with heat inactivated influenza virus can crosspresent the Flu MP antigen to CD8+T cells as 0.8-1.38% of cocultured CD8 + T cells are specific for Flu-MP as assessed using peptide-MHC Class I tetramers. Activation of the mDC with both CbG and E. coli LPS resulted in a considerably increased MPspecific CD8 + T cell response ( Figure 5D). This indicates that CbG can enhance secondary CD8 + T cell responses.
To assess whether CbG can enhance the priming of naïve CD8 + T cells by DC, we chose the melanoma-derived antigens MART-1 and gp100. The experiments illustrated in Figures 6A and 6B have respectively been performed with human skin CD1a + and CD14 + DC [28]. Priming of naïve CD8 + T cells requires activation of the DC through CD40 and addition of IL-2 and IL-7 to the cultures. Further activation of DC with LPS or CbG did not enhance the response to the relatively abundant MART-1 T cells. However, CbG enhanced the priming to the less frequent gp100-specific naïve T cells ( Figure 6).
Naïve CD8 + T cells were also cultured for seven days with allogeneic mDC that were activated or not with either LPS or CbG ( Figure 7A). CbG-activated mDC were also capable of inducing naïve CD8 + T cells to express high levels of IFNc and granzyme B in CD8 + T cells when compared to unactivated mDC ( Figure 7A).
Thus, CbG-activated mDC are able to activate CD8 + T cells.
Brucella CbG increases CD4 + T memory responses in PBMC from HCV cured and TB patients To characterize the CD4 + T cells exposed to CbG activated mDC, CD4 + T cells were cultured with allogeneic blood mDC activated or not with either LPS or CbG for seven days ( Figure 7B). The cultured T cells were then restimulated with PMA/ Ionomycin and stained for intracellular IFN-c, IL-13, and IL-17. Both LPS and B. melitensis CbG-activated mDC polarized CD4 + T cells into IFN-c-expressing Th1 cells ( Figure 7B). In addition, both E. coli LPS and CbG-activated DC induced a minor subpopulation of naïve CD4 + T cells to differentiate into IL-13 + CD4 + T cells. Naïve CD4 + T cells co-cultured with either E. coli LPS-activated or CbG-activated mDC did not express IL-17 ( Figure 7B). Thus, CbG-activated mDC induce Th1 responses.
DC targeting approach consists in delivering antigens directly to DC in vivo using chimeric proteins composed of an anti-DC receptor antibody coupled to a selected antigen [30]. The selection of the appropriate adjuvant is a critical parameter for the induction of the desired type of immune response. PBMC from cured chronic HCV infected patients were cultured for ten days with monocyte-derived DC and recombinant humanized anti-CD40 or anti-DCIR fused to the HCVNS3 antigen with or without CbG or Poly-IC. Antigen-specific responses were measured by exposing the cultured cells to specific peptides clusters and stained for intracytoplasmic IFNc expression. Low IFN-c levels were observed in PBMC cultured with DC and control IgG4 with or without the adjuvants ( Figure 8A). Moreover DC targeting by anti-CD40 and anti-DCIR enhanced IFN-c production by CD4 + T cells. Potent memory CD4 + T ( Figure 8A), but not CD8 + T cell (not shown) responses were observed after DC-PBMC coculture in these conditions.
Few CD4 + T cell responses could be induced when anti-CD40 targeted-DC were treated with Poly I:C, as demonstrated by the presence of some CD3 + CD4 + IFN-c + T cell (1.21% of parents cells, Figure 8A). However, CbG-treated DC using anti-CD40 targeting induced a dramatic increase of CD3 + CD4 + INF-c + cells (3.53% of parent cells compared to 0.85% without stimulation, Figure 8A). When anti-DCIR vaccine targeting was used, we could not observe any difference in CD4 + INF-c + T cell population with or without stimulation ( Figure 8A).
We next studied CbG effect in DC targeting experiments with PBMC from acute TB patients. For this, we used humanized anti-CD40 or anti-DCIR antibodies coupled to Ag85BD41-ESAT6-Rv1980D24 Mycobacterium tuberculosis antigens developed by the ANRS (French agency of research against AIDS and viral hepatitis). DC targeting with control IgG4 and the specific peptides in the presence of adjuvants triggered IFN-c production by CD4 + T cells ( Figure 8B). This induction was increased in the presence of CbG upon anti-CD40 and anti-DCIR targeting. The stimulation by Poly I:C slightly enhanced IFN-c production only upon anti-DCIR targeting ( Figure 8B).
Taken together, these data show that Brucella CbG increases CD4 + T memory responses after DC targeting of PBMC in HCV cured and acute TB patients.

Discussion
Vaccination represents the most effective strategy to combat infectious diseases. Vaccines are composed of antigens and adjuvants, which activate antigen-presenting cells which then triggers the activation, differentiation and expansion of antigenspecific T and B lymphocytes [15,[31][32][33]. The number of approved adjuvants effective in humans is very limited since they are mostly based on alum and on emulsions. Ideally, adjuvants should elicit a selected immune response (i.e. cellular or humoral immunity depending on the requirements for protection), be safe (sufficiently immunogenic, without excessive inflammation) and cost-effective [16,18,31]. The benefits of adjuvant incorporation into any vaccine formulation should significantly outweigh the risks of adverse reactions. Unfortunately, potent adjuvant action is often correlated with increased toxicity, as exemplified by Freund's complete adjuvant or LPS. Thus, one of the major challenges in human adjuvant development is to identify compounds that enhance vaccine antigen induced responses with maximum tolerability and safety [31]. In particular, there is a high demand for adjuvants that stimulate cellular immunity [15,16,18]. Here, we demonstrate that Brucella CbG is a non-immunogenic and nontoxic molecule capable of triggering the activation of cellular responses in vivo. Moreover, Brucella CbG dramatically increases specific memory CD4 + T cell responses of human PBMC induced by DC-targeting fusion proteins expressing either HCV antigens or Mycobacterium tuberculosis antigens. Based on these data, we propose that Brucella CbG is a candidate adjuvant that might be used in humans.
There has been an interest to identify TLR4 agonists with a dampened toxicity. A recent example of lipid A analog is MPLA known to activate immune cells with similar properties of the LPS but less toxic and non immunogenic [34,35]. Our results show that CbG-induced DC maturation is dependent on TLR4 as well as Myd88 and TRIF adaptor molecules. While CbG activates all human mDC subsets, it does not activate pDC, which is consistent with their lack of TLR4 [24][25][26]. However, what differentiates CbG from MPLA and Ploy I:C is that CbG induces an early immune response (Figure 3). This might have a significant impact on the quality of the immune response to vaccines [36].
We previously showed that Brucella CbG is capable of interacting with lipid rafts and modulate their organization [6]. Lipid rafts are plasma membrane microdomains enriched in cholesterol and sphingomyelin that are involved in intracellular signalling and membrane transport. In particular, lipid rafts are involved in the regulation and activation of several important immune receptor complexes such as the TLR4 complex [37]. [38]. It is thus possible that Brucella CbG triggers TLR4-dependent signalling through its effect on cholesterol in lipid rafts as it has been suggested for alum [15]. Indeed, alum recognition may occur indirectly through the release of endogenous uric acid. Recently, it was shown that monosodium urate crystals activate DC by interacting with the cell membrane, which leads to plasma membrane lipid sorting probably via interaction with cholesterol [15].
Herein, we observed that in contrast to E. coli LPS [27], Brucella CbG-induced BMDC maturation was independent of the GPIanchored protein CD14. Actually, the synthetic lipid A compound CRX-527 does not require CD14 to engage MyD88-dependent and TRIF/IRF3-dependent pathways downstream TLR4 [39]. Furthermore, the uropathogenic E. coli (UPEC) triggers innate responses during urinary tract infection in a TLR4-dependent and CD14-independent manner both in mice and humans [40]. These results clearly indicate that CD14 is not required for TLR4dependent cell activation. The detailed signalling pathway involved in CbG-dependent cell activation will require further studies including the discovery of putative co-receptors associated to TLR4. Until now, we know that neither dectin-1 nor TLR2 contribute to the activation. Additional studies are planned to determine whether the CD14-independent CbG-dependent DC activation will have any therapeutic value.
CbG displays interesting properties such as water solubility, limited toxicity and lack of immunogenicity together with a potent DC activation capacity that can trigger CD4 + and CD8 + T cell responses. Therefore, CbG might constitute a new class of adjuvants for future vaccines.

Brucella CbG extraction, purification and characterization
CbG were obtained from B. melitensis 16 M or B. abortus 2308 grown and inactivated as described before [42]. For CbG extraction and purification, the protocol described before [6] was used. Briefly, a CbG-rich crude fraction was first obtained by ethanol precipitation of a hot water extract of killed bacteria, and freed from nucleic acids or proteins by digestion with DNase and RNase proteinase K. To remove LPS, the fluid was extracted with a volume of phenol (in contrast to most LPS, Brucella LPS and lipid A partition into the phenol phase [42] at 70uC for 30 min, the mixture chilled and centrifuged (80006 g, 0uC, 15 min), and the aqueous phase collected and re-extracted again with phenol under the same conditions. The new aqueous phase was dialyzed, clarified by brief centrifugation and freeze-dried. The identity of CbG was demonstrated by 13 C-NMR spectroscopy and highperformance TLC, and the absence of Brucella LPS or other contaminants was demonstrated by UV-spectrophotometry, SDS-PAGE, gel immunoprecipitation, and 3-deoxy-D-manno-2-octulosonic acid analyses [42]. To further demonstrate the absence of lipid A, purified CbG were analyzed by MALDI-TOF mass spectrometry as described before [45]. Briefly, 5 mg of lyophilized Brucella's CbG were resuspended in 100 ml of chloroformmethanol-water (3:1.5:0.25 [vol/vol/vol]) and 1 ml aliquot of was deposited on the target and covered with the same amount of 2,5dihydroxybenzoic acid matrix (Sigma) dissolved in chloroformmethanol-water (3:1.5:0.25 [vol/vol/vol]). Different ratios between the samples and dihydroxybenzoic acid were used. Alternatively, matrix solution was prepared by dissolving 1 mg of 2,5-dihydroxybenzoic acid with 0.1 ml of CH 3 CN/H 2 O (3:2, vol/vol). Analyses were performed in reflector modes and in both the positive and the negative ion modes on a Bruker Autoflex II MALDI-TOF mass spectrometer (Bruker Daltonics, Inc.). A peptide calibration standard (Bruker Daltonics) was used to calibrate the MALDI-TOF. Spectra were recorded between 1900 and 3900 Da. Each spectrum was an average of 500 shots. Mark the absence of the ion species at m/z 2073, 2145, 2173 that are characteristic of Brucella lipid A [46].
Cyclic glucan toxicity and immunogenicity assays LAL (Limulus ameobocyte lysate) test was used for the detection and quantification of bacterial endotoxins. Briefly, the samples were incubated with the circulating blood of horseshoe crab (the LAL) and a synthetic color producing substrate to detect endotoxins. LAL contains enzymes that are activated in a series of reactions in the presence of endotoxins. This assay is quantitative and the color intensity developed upon addition of the sample to the LAL is proportional to the amount of endotoxin in the sample and can be calculate from a standard curve. Immunogenicity of purified CbG was tested in mice and rabbits following described protocols [47]. LD50 (Lethal dose 50%) was measured in CD-1 mice injected with appropriate amounts of reagents. Animal death was recorded at 12 h, 24 h, 48 h, and 72 h post-injection. To determine the immunogenicity of B. melitensis CbG, Balb/C mice were immunized by either PBS or E. coli LPS, (10 mg/mouse) or B. melitensis LPS (10 mg/mouse) or B. melitensis CbG (10 mg/mouse). The primary antibody response was measured 21 days after immunization. Mice were boosted 45 days with the same molecules (5 mg/mouse) to measure the secondary antibody response. Antibody responses were determined by an indirect ELISA.

mRNA extraction and hybridization
After culture, human mDC were lysed in RLT buffer and stored at 280uC until further processing. Total RNA was extracted using the mirVana miRNA Isolation Kit, from Ambion. Following RNA extraction, RNA concentration was measured using a Nanodrop 1000 (Nanodrop Technologies, Wilmington, DE) and the RIN was measured with an Agilent 2100 Bioanalyzer (Agilent, Palo Alto, CA) for quality control purposes. All samples with RIN values greater than seven were retained for further processing. 250 ng of total RNA were amplified and labelled with the Illumina TotalPrep-96 RNA amplification kit (Ambion, Austin, TX). 750 ng of amplified labelled RNA were hybridized overnight to Illumina HT12 v4 Beadchip arrays (Illumina, San Diego, CA). Following hybridisation, each chip was washed, blocked, stained and scanned on an Illumina iScan following the manufacturer's protocol. The dataset described in this manuscript is deposited in the NCBI Gene Expression Omnibus (GEO, http://www.ncbi. nlm.nih.gov/geo, GEO Series accession number GSE32023).

Microarray analysis
Transcripts present in at least one of the samples as defined by significant chip detection value were used as a starting list. Nonparametric test was applied between CbG-treated mDC and medium-treated mDC from 5 donors, false discovery rate: 0.01, with Benjamini-Hochberg correction. Pathway analysis was conducted with Ingenuity Pathway Analysis (IPA) software, Ingenuity Systems, Inc, Redwood City, CA. The molecular distance to medium (MDTM) quantifies the global transcriptional perturbation of a group of transcripts in a specific sample as compared to its reference control(s). It is calculated as follows. For an arbitrary list of transcripts normalized to their reference control(s), the absolute fold changes greater or equal to 2 for a specific sample are summed up.

Cytokine measurement
Murine IL-12 and TNF-a were quantified in culture supernatants of stimulated DC by sandwich enzyme-linked immunosorbent assays (ELISA) according to the manufacturer's protocol (Abcys). Human cytokine (IL-6, TNF-a, and IL-12p40) were determined using the BeadLyte cytokine assay kit (Upstate, MA). To determine the amounts of cytokines produced in the sera, a CBA assay (BD Biosciences) that detects IL-6, IL-10, MCP-1, IFN-c, TNF-a and IL-12p70 in a single sample was used. The sera from immunized mice were incubated with a mixture of specific capture antibodies that were coupled to the beads containing specific amounts of PE fluorescence intensity. The 6 different fluorescence intensities of PE were detected by flow cytometry and cytokine concentration in the samples was quantified from a standard curve according to manufacturer's protocol. 10.000 events were analyzed by flow cytometry Canto II (BDBiosciences) and the data were analyzed using the FlowJo software.

Cytometric Bead Array (CBA) assay
Human CD4 + and CD8 + T cell responses Blood mDC were co-cultured with CFSE-labeled allogeneic naïve CD4 + T and CD8 + T cells. The expression of intracellular cytokines and granzyme B were measured after 6 h of cell stimulation by PMA and Ionomycine, in the presence of Brefeldin A. Blood mDC from HLA-A0201 + healthy donors were loaded with a multiplicity of infection (MOI) of 0.2 of heat-inactivated influenza virus (PR8) for 2 h at 37uC. Autologous CD8 + T cells were mixed and cultured for 7 days in the presence of 20 units/ml IL-2. Cells were then stained with anti-CD8 antibody and tetramer (HLA-A*0201-Flu M1 58-66 ). MART-1-specific CD8 + T cell responses were measured after co-culturing with CD1a + and CD14 + skin DC loaded with 10 mM 15-mer MART-1 peptidecontaining the immunodominant epitope MART-1 26-35 (27L) or with gp100 peptide epitope for 10 days. Skin DC were purified as previously described [28]. To activate individual DC subsets we used 0.25 mM of either E. coli LPS or CbG.
Autologous IFNa DC loaded with vaccine candidates (for Mycobacterium: ESAT6 protein, for HCV: antigens from NS3 Helicase helB construct) were co-cultured with PBMC from HCV or TB patients and incubated for 10 days. T-cell specific responses elicited by vaccine candidate loaded-DC were assessed by restimulating PBMC with peptide clusters

Adoptive transfer of OT-I T cells and immunization
OT-I transgenic cells that express TCR specific for an H-2K b restricted CD8 + T cell epitope from OVA were used. Lymph nodes from OT-I Ly5.1 mice were harvested and digested with collagenase type I (Sigma) at 37uC for 30 min. CD8 + T cells were then negatively sorted by using mouse CD8 negative isolation kit (Dynal). CD8 + T cells were labeled with 10 mM CFSE (Invitrogen) and transferred intravenously (i.v.) into naive congenic C57Bl/6 Ly5.2 recipient mice. At 24 h, recipient mice were immunized subcutaneously (s.c.) either with 30 mg OVA (EndoGrade) alone in endotoxin free PBS or 30 mg OVA mixed with 200 mg of CbG or 30 mg OVA mixed with 50 mg poly I:C (Sigma) or 30 mg OVA mixed with 20 mg MPLA (InvivoGen).

Histology
For histological assessment of cutaneous inflammation, intradermal injection of mouse ear skin was performed. C56BL/6 mice were immunized in the right ear with either: 20 mg of Monophosphoryl Lipid A (MPL) in 10 ml of PBS, 200 mg of CbG in 10 ml of PBS, 10 mg of LPS in 10 ml of PBS, 50 mg of Poly I/C in 10 ml of PBS. PBS was injected in the left ear as negative control. The animals were sacrificed 48 h later and both the adjuvant-treated and untreated ears were collected for further determination. Ears were fixed in 10% neutral buffered formalin for 48 h, embedded in paraffin, then sectioned at 4 mm at two levels separated by 1 mm interval and stained with hematoxylin and eosin. Normal skin histology was evaluated in PBS-inoculated contralateral ear for each mouse.

Statistical analysis
All experiments were carried out at least 3 independent times and all the results correspond to the means 6 standard errors. Statistical analysis was done using two-tailed unpaired Student's t test. Significance was defined when P values were ,0.05.

Supplementary material
See legends of Figures S1, S2, S3, S4, S5 and S6. Figure 8. B. melitensis CbG potentiates CD4 + T memory responses in HCV and TB patients. (A) CbG induces CD3 + CD4 + IFNc + T cells after anti-CD40 HCV-NS3HelB DC-targeting of PBMC in HCV cured patients. HCV antigens from NS3 Helicase HelB construct were delivered to IFNa DC through anti-CD40 or DCIR-Ig4 humanized recombinant antibodies. IFNa DC were targeted with anti-CD40-NS3HelB (5 nM), anti-DCIR-NS3HelB (5 nM) or Ig4 control antibody (5 nM) in the presence of Poly I:C (25 mg/ml) or CbG (10 mg/ml) before co-culture for 10 days with PBMC from chronic HCV-infected patients cured after IFNa-Ribavarin therapy. Cells were stimulated for 6 h with peptide clusters (10 mM; 10 peptides of 15-mers) covering HCV NS3 HelB constructs. PBMC were stained for measuring the frequency of peptide-specific CD3 + CD4 + IFNc + T cells and analyzed by flow cytometry. (B) CbG increases CD4 + T cell memory responses induced by anti-CD40 or anti-DCIR coupled to Ag85BD41-ESAT6-Rv1980D24 antigens in PBMC from TB patients. IFNa DC were loaded on with IgG 4 humanized recombinant mAb Ag85BD41-ESAT6-Rv1980D24 (1 pM) either in the presence of medium versus Poly I:C (2.5 mg/ml) or CbG (1 mg/ml). After 10 days of co-culture with PBMC from acute TB patients, cells were re-stimulated with peptides at 2.5 mM covering the entire ESAT6 protein (22 peptides of 15-16 mers) for 6 h and the magnitude of the immune recall response in terms of the percentage of CD3 + CD4 + IFNc + T cells was analyzed by flow cytometry. doi:10.1371/journal.ppat.1002983.g008