Lysosome-Dependent Activation of Human Dendritic Cells by the Vaccine Adjuvant QS-21

The adjuvant properties of the saponin QS-21 have been known for decades. It is a component of the Adjuvant System AS01 that is used in several vaccine candidates. QS-21 strongly potentiates both cellular and humoral immune responses to purified antigens, yet how it activates immune cells is largely unknown. Here, we report that QS-21 directly activated human monocyte-derived dendritic cells (moDCs) and promoted a pro-inflammatory transcriptional program. Cholesterol-dependent QS-21 endocytosis followed by lysosomal destabilization and Syk kinase activation were prerequisites for this response. Cathepsin B, a lysosomal cysteine protease, was essential for moDC activation in vitro and contributed to the adjuvant effects of QS-21 in vivo. Collectively, these findings provide new insights into the pathways involved in the direct activation of antigen-presenting cells by a clinically relevant QS-21 formulation.

responses (12). The activation of different pattern recognition receptors can lead to the production of various cytokines and thus alter the adaptive response. It is therefore crucial to understand the early pathways activated by adjuvants as this step can have profound effects on the quality of the adaptive response. The mechanisms by which saponins activate antigen-presenting cells are still poorly understood. Both QuilA and QS-21 activate the NLRP3 inflammasome in vitro (13,14). Furthermore, ISCOMATRIX, an adjuvant containing different QuilA fractions also activates the inflammasome in vitro and in vivo (15). However, the antigenspecific responses elicited by QS-21-containing vaccines were found to be largely independent of the NLRP3 inflammasome pathway, suggesting that other mechanisms must be involved in its adjuvanticity (14,15).
Here, we show that QS-21 directly activated human monocytederived dendritic cells (moDC) and identified specific signaling pathways that lead to acute transcriptional activation. We found that QS-21 was endocytosed in a cholesterol-dependent manner and accumulated in lysosomes. We further demonstrate that lysosomal destabilization and cathepsin B activity were central for the response of moDCs to QS-21. We also show that absence of cathepsin B decreased antigen-specific CD4 and CD8 T cell responses in a murine vaccination model.

Mice
Six-week-old C57BL/6 mice were purchased from Harlan Horst. Cathepsin B-deficient mice (Ctsb −/− ) were previously described (17). These mice were housed and bred under specific pathogenfree conditions. Animal studies were approved by the institutional animal care and local committee for animal welfare.

Mice immunization
Six-week-old female mice received injections containing 4 µg of HbsAg, 1 µg of ovalbumin (OVA) and 1 µg of QS-21 in the gastrocnemius muscles of both hind limbs in a volume of 20 µl per muscle. OVA was obtained from Calbiochem and confirmed to be endotoxin-free.
I counts in lysates were measured with a gamma counter, as described in Ref. (18) (except that pronase was replaced by 1% trypsin). Uptake was saturable with 14 C-QS-21 concentration around 10 µg/ml. Accumulation was linear with time for at least 6 h indicating lack of degradation and effective retention, which was further confirmed by the negligible loss after a 2-h chase. 14 C-QS-21 integrity was confirmed by TLC/phosphorimager.

subcellular Fractionation
After 14 C-QS-21 accumulation for 4 h and chase overnight, cells were surface stripped with trypsin or not, homogenized, and submitted to differential sedimentation to yield postnuclear particles as described in Ref. (19). These were mixed with 15% Percoll to form a self-generating gradient. Ten fractions of 1 ml were collected from the bottom and analyzed for radioactivity (stripped: recovery, 80% of homogenate; non-stripped, 50%), beta-hexosaminidase activity (stripped or not; >80%), and Western blotting for Na + /K + -ATPase (non-stripped cells).

Microarray analysis
A total of 10 6 moDCs from four donors were stimulated with QS-21 for 4 h in complete medium. Cells were washed with PBS, scraped, and lysed in 1 ml Tripure reagent (Roche Applied Science). RNA was extracted with chloroform and purified with the RNeasy Minikit (Qiagen) following manufacturer's instructions, with DNase treatment of the column to remove possible genomic DNA contamination. RNA was concentrated by ethanol precipitation followed by resuspension in RNase-free water. RNA was quantified with a NanoDrop spectrophotometer. Biotinylated aRNA was generated from 100 ng total RNA using a Genechip 3′IVT express kit (Affymetrix, Santa Clara, CA, USA). Samples (10 µg biotinylated and fragmented aRNA) were hybridized on a HG-U133A 2.0 Expression chip (Affymetrix) for 16 h. Arrays were washed according to manufacturer's recommendation and scanned using the Genechip scanner 3000 Affymetrix. Fold changes [QS-21 or 3-O-desacyl-4′-monophosphoryl lipid A (MPL; GSK Vaccines) vs. medium] were calculated for each donor and a heatmap of top differentially expressed genes (FC > 5 or <0.5) was generated with MeV software. A volcano plot was generated with the average fold change and p-values for each gene using GraphPad Prism software. A scatter plot of fold changes vs. medium of QS-21 (x axis) and MPL (y axis) was generated using the GraphPad Prism software.

elisa
For cytokine measurements, 2 × 10 5 moDCs were seeded in 96-well plates and allowed to rest for 2 h. Cells were treated with the appropriate inhibitors and stimulated. The supernatants were collected after 24 h and ELISAs (hIL-6, hIL-8, hTNF) were carried out following manufacturer's instructions (R&D Systems). For Ag-specific antibody measurements, the HBs protein was used as the coating antigen and goat anti-mouse IgG (GAM-SouthernBiotech or Jackson ImmunoResearch) was used as the coating antibody for the standard curve. Plates were blocked at room temperature for 1 h (PBS-1% BSA) and serial dilutions of serum samples and the IgG standard (SouthernBiotech) were added and incubated for 2 h at room temperature, followed by addition of biotin-conjugated anti-mouse IgG (Jackson Research or SouthernBiotech) and streptavidin-HRP. TMB was used as a substrate and plates were read at 450 nm on a microplate reader.

Virus Production and Transduction
pGIPZ vector-based short hairpin RNA (shRNA) targeting Syk (three constructs), cathepsin B (five constructs), and GAPDH as a control were purchased from Dharmacon. The following constructs exhibiting the highest knockdown capacity (verified by qPCR) for each target were selected (Syk: TCTATGATGTTCTTATCCT, cathepsin B: AACTTGACAGGGTGAAGCT). These vectors were transiently transfected into the human embryonic kidney HEK293T cells in combination with the packaging vector psPAX2 (Addgene plasmid 12259) and the VSV-G encoding plasmid pMD2.G as previously described (21). The supernatant was harvested 48 h after transfection and viruses were concentrated by ultracentrifugation and resuspension in Opti-MEM medium (Thermo Fisher). Viral titers were determined by transduction of 293 T cells with serial dilutions of the concentrated virus stock followed by flow cytometric analysis of GFP expression. SIV virion-like particles (SIV-VLPs) were produced by cotransfecting HEK293T cells with pMD2.G and pSIV3+ (22). moDCs were transduced as previously described (22,23). Briefly, CD14 + monocytes were purified as described above, and seeded in 6-well plates in complete medium containing 800 U/ml GM-CSF, 200 U/ ml IL-4, 1 µg/ml polybrene, and 100 µl of supernatant containing SIV-VLPs. After 1 h, lentiviral particles were added at a MOI of 1 and the plates were centrifuged for 1 h at 900 g at room temperature. On day 3, 1 ml of medium containing 2,400 U GM-CSF and 600 U IL-4 was added to each well. The cells were recovered on day 6, and the transduction efficiency (GFP expression) was measured by flow cytometry (typically, between 80 and 90% of cells were transduced).

Qs-21 Promotes the activation and Maturation of human moDcs
QS-21 has been shown to activate the inflammasome in mouse macrophages and dendritic cells (14). In order to explore the signaling pathways involved in the immunostimulatory properties of saponins in human cells, monocyte-derived DCs (moDCs) were stimulated with increasing concentrations of QS-21 formulated into cholesterol-containing liposomes (hereafter referred to as "QS-21"), as in the clinical AS01 formulation. We observed dose-dependent induction of TNF, IL-6, and IL-8 and selected 10 µg/ml for following experiments ( Figure 1A). The activation of moDCs by QS-21 or MPL, a TLR4 ligand, was analyzed by flow cytometry for the surface expression of the class II major histocompatibility complex (MHCII) molecule, HLA-DR, and the costimulatory receptor, CD86. Both proteins were upregulated similarly by either QS-21-or MPL-stimulated moDCs, indicating that QS-21 promotes moDCs activation ( Figure 1B). Furthermore, this effect was dependent on the QS-21 component as the liposomes alone did not induce upregulation of these markers ( Figure 1C). Next, moDCs from four donors were stimulated with either QS-21 or MPL for 4 h and global gene expression was analyzed by microarray. PBStreated cells were used as the baseline to which both QS-21 and MPL treatments were compared. MPL was used as a control for its ability to uniquely activate TLR4 and to better discriminate QS-21-specific pathways.
In order to visualize changes in gene expression induced by QS-21, average microarray data were visualized on a "volcano plot" (Figure 1D). Overall, QS-21 induced significant upregulation of 102 probe sets and downregulation of 43 (Table S1 in Supplementary Material). Comparison of transcript expression profiles between QS-21 and MPL with a scatter plot identified co-regulated genes and genes specifically regulated by either QS-21 or MPL at this time point (Figure 1E). Genes upregulated by both stimulants included cytokine or chemokine genes such as IL6, IL1A, IL1B, CXCL2, CXCL3, or CCL20, demonstrating (h) moDCs were stimulated with QS-21 (10 µg/ml) or MPL (1 µg/ml) for the indicated durations, and IL-8, IL-1β, IL-6, and TNF transcript abundance was measured by qPCR. Data are from one donor representative of at least three independent experiments. (i) moDCs (n ≥ 5) were treated with Z-VAD-fmk (10 µM) and stimulated with QS-21 (10 µg/ml) or MPL (1 µg/ml), and IL-8, TNF, and IL-1β in the supernatant were measured by ELISA. (J) moDCs (n ≥ 5) were treated with recombinant IL-1 receptor antagonist (Anakinra-10 µg/ml) and stimulated with QS-21 (10 µg/ml) or IL-1β (10 ng/ml) and IL-8 was measured by ELISA. Statistical significance was determined by two-way ANOVA followed by Tukey's multiple comparisons test. a common inflammatory core response as previously reported for other adjuvants (24). Top upregulated probe sets identified as specific to QS-21 (FCQS-21 > 2 and FCMPL < 2) at this time point included RGS1, RGS2, NR4A2, DUSP1, and EGR1, which are primary response genes that are rapidly induced following cell activation with inflammatory stimuli (25)(26)(27)(28). Clustering analysis of the top up-or downregulated genes for individual donors (FC vs. medium > 2 or FC vs. medium < 0.5) showed that individuals were clustered by stimulation (QS-21 vs. MPL) rather than by donor, indicating that the difference between MPL and QS-21 stimulation is greater than variations between different donors ( Figure 1F). Pathway enrichment analysis with the InnateDB resource (29) identified pathways either shared by QS-21 and MPL, such as cytokine, Nod-like receptor (NLR), and G-protein-couple receptor signaling (Figure 1G), or specific to MPL such as TLR and IL-1 signaling, as expected (30). Pathways specific to QS-21 included AP-1 and ATF2 transcription factor networks, which mediate gene regulation in response to cytokines, stress signals, or infectious agents (31). Given that the QS-21-specific genes identified at this time point were known early response genes, we focused on the early expression kinetics of several upregulated cytokines shared by MPL and QS-21 by qPCR. Expression of these genes upon stimulation with QS-21 or MPL showed different kinetics, suggesting that slower or more complex pathways were triggered by QS-21 ( Figure 1H). Finally, since saponins including QS-21 can activate the inflammasome and promote the release of IL-1β, we examined whether caspase activation or IL-1β signaling were required for cytokine production. Treatment of moDCs cells with the pan-caspase inhibitor Z-VAD-FMK had no effect on IL-8 or TNF production and, as expected, decreased IL-1β levels following QS-21 stimulation ( Figure 1I). Furthermore, a recombinant IL-1 receptor antagonist had no effect on IL-8 production by cells stimulated with QS-21, but significantly decreased the response to IL-1β ( Figure 1J). These results strongly suggest that cytokine production by QS-21-stimulated moDCs does not depend on inflammasome activation.

Membrane cholesterol is required for the endocytosis of Qs-21 and activation of immune cells
Next, we analyzed the mode of capture and subcellular localization of QS-21. In order to have a sufficient number of cells, phorbol 12-myristate 13-acetate (PMA)-differentiated monocytic THP-1 cells were used for both biochemical and fractionation experiments. Cells were first incubated at 4°C with the fluorescent analog BODIPY-QS-21 (formulated into cholesterol-containing liposomes), which arrests endocytosis and limits passive diffusion. At this temperature, the fluorescence signal was exclusively detected at the cell surface. In contrast, after incubation at 37°C, QS-21 became detectable in distinct cytoplasmic, mostly perinuclear puncta (Figure 2A).
To separately measure surface binding vs. intracellular uptake, we validated a surface stripping assay after incubation with 14 C-QS-21 (also formulated into cholesterol-containing liposomes) at 4°C. Surface digestion with trypsin or treatment with heparin sulfate virtually abolished cellular uptake, indicating that surface binding depended on plasma membrane proteins and electrostatic interactions (Figure 2B). To further exclude passive diffusion as mechanism of intracellular uptake, the ATP content was depleted by preincubation with azide/deoxyglucose and intracellular (trypsin-resistant) uptake of 14 C-QS-21 was measured at 37°C. Similarly to incubation at 4°C, energy depletion abolished the intracellular uptake of QS-21 at 37°C, demonstrating an active process requiring cellular energy and resulting in its accumulation in endocytic structures ( Figure 2C). Given that saponins interact with cholesterol (8), we next examined the role of membrane cholesterol in the endocytosis of 14 C-QS-21. Minimal depletion of membrane cholesterol (<10% loss) by the addition of 2.5 mM (low-dose) methyl-βcyclodextrin (MβCD), followed by extensive washing to eliminate residual MβCD, inhibited 14 C-QS-21 uptake by ~50%, contrasting with the absence of effect on the intracellular uptake of either caveolae/lipid-raft-mediated endocytosis ( Figure 2D). Given that QS-21 endocytosis involves cholesterol, the effect of membrane cholesterol extraction on the response of moDCs to QS-21 was then investigated. Pretreatment of moDCs with MβCD inhibited their response to QS-21 but not to MPL, indicating that membrane cholesterol was a QS-21-specific requirement (Figures 2E,F).
These results show that QS-21 was actively endocytosed in a cholesterol-dependent manner and that membrane cholesterol was crucial for the response to QS-21.

Qs-21 Mediates syk Kinase activation That is essential for response of moDcs
Membrane alterations, and more specifically membrane cholesterol remodeling, can induce activation of the Syk kinase independently of receptor ligation (32). We therefore investigated the role of this kinase in the response to QS-21. Dendritic cells were activated and Syk phosphorylation was measured by Western blotting. When cells were stimulated with QS-21, Syk phosphorylation increased on tyrosine 323 and 352, two known sites of Syk autophosphorylation (33), to a comparable extent as in response to MPL or zymosan, a known inducer of Syk phosphorylation (Figures 3A,B). Blockade of Syk activity by BAY 61-3606 (34) inhibited response to QS-21 in moDCs, as seen by the reduction of TNF, IL-6, and IL-8 secretion (Figures 3C-E). Since pharmacological compounds can lack specificity, Syk expression was also inhibited by shRNA interference. CD14 + monocytes were transduced during their differentiation into moDCs with lentiviral particles encoding Syk or GAPDH shRNAs, as previously described (22). Syk mRNA knockdown, measured by quantitative PCR, was greater than 50% when compared to cells expressing an shRNA specific to GAPDH ( Figure 3F). Syk knockdown strongly decreased both IL-6 and TNF mRNA expression induced by QS-21, confirming the results obtained with the inhibitor (Figures 3G,H). Finally, the effect of Syk inhibition on NF-κB activation was investigated by observing p65 nuclear translocation. Syk inhibition by BAY 61-3606 strongly decreased QS-21-induced p65 translocation to the nucleus, indicating that Syk activity is required for both NF-κB activation and cytokine production ( Figure 3I). The punctate pattern of cells incubated with BODIPY-QS-21 (Figure 2A) suggested that QS-21 may be confined to a specific subcellular compartment or structure. We therefore established the subcellular distribution of 14 C-QS-21 by density gradient fractionation in THP-1 cells after 4 h of pulse and overnight chase. Equilibration of postnuclear particles in Percoll gradients fully resolved the plasma membrane (Western blotting for Na + / K + -ATPase, buoyant fractions) from lysosomes (N-acetyl-βhexosaminidase activity, dense fractions) (Figure 4A). In the absence of surface stripping by trypsin, the density distribution of QS-21 was bimodal between these two positions in the gradient, indicating two pools at the plasma membrane and lysosomes, respectively. When cell surface proteins and associated material were stripped by trypsin digestion, the intracellular pool of QS-21 perfectly co-distributed with dense lysosomes. Colocalization of BODIPY-QS-21 with LysoTracker Red confirmed that internalized QS-21 was concentrated in lysosomes ( Figure 4B).
Since saponins can interact with cell membranes and induce pore formation (8), we next examined the effect of QS-21 on lysosomal integrity. AO is a lysosomotropic dye that emits red fluorescence when concentrated under acidic conditions. moDCs preincubated with AO showed a rapid, yet partial decrease in red fluorescence upon stimulation with QS-21, indicating loss of lysosomal acidification or membrane integrity (Figure 4C). To investigate whether this second possibility could be due to saponin-mediated pore formation, moDCs were incubated with After stimulation with QS-21, tracers were detected in the cytosol (80% of cells for the small LY and 60% for the two dextrans) ( Figure 4D). Furthermore, in 40% of cells, LY fluorescence showed only a diffuse signal with loss of puncta, while both dextrans were still visible in lysosomes. QS-21 therefore partially permeabilized lysosomal membranes with a selectivity in pore size, resulting in widespread cytosolic diffusion of the small tracer but only partial diffusion of the two dextran polymers.
To determine whether low lysosomal pH was a crucial factor for the response to QS-21, lysosomal acidification was blocked by inhibition of the vacuolar ATPase by bafilomycin A1 (BafA1), or by incubation with the weak base ammonium chloride (NH4Cl). Incubation of moDCs with either agents lessened both QS-21-induced IL-6 and TNF mRNA transcription ( Figure 5A) and cytokine release ( Figure 5B). No such effect was observed upon stimulation with MPL, excluding a defect in IL-6 secretion. BafA1 treatment also inhibited Syk Y352 phosphorylation induced by QS-21 but not by zymosan, strongly suggesting that QS-21 induced Syk phosphorylation takes place downstream of lysosomal destabilization ( Figure 5C). Finally, pretreatment of cells with BafA1 had no effect on QS-21-induced relocation of either LY or dextrans from the lysosomes to the cytosol, indicating that lysosome pore formation occurred even if acidification was inhibited (Figure 5D). These results suggest that acidic pH in lysosomes and lysosomal membrane permeabilization by QS-21 are necessary for the activation of moDCs.
lysosomal cathepsin B Mediates the response to Qs-21 in human Dendritic cells Some lysosomal cathepsins are reported to activate NF-κBmediated transcriptional events when released into the cytosol (35,36). To screen for involvement of member(s) of this protease family in QS-21-mediated activation of moDCs, cells were  Only the two first inhibitors significantly decreased QS-21-mediated TNF production, pointing to a specific involvement of cathepsin B in this response ( Figure 6A). Cathepsin B inhibition with either CA-074 Me or Z-FA-FMK also inhibited QS-21-induced TNF and IL-6 mRNA expression with no effect on MPL-induced gene expression, indicating that the effect was transcriptional and specific to QS-21 ( Figure 6B). Finally, shRNA-mediated knockdown of cathepsin B strongly decreased the expression of both TNF and IL-6 mRNAs, confirming that cathepsin B expression is essential for QS-21-mediated cytokine production in human DCs (Figures 6C,D). The frequency of cytokine (IL-2, TNF, and IFN-γ)-producing Ag-specific T cells, measured by flow cytometry following ex vivo restimulation with antigenic peptides, was significantly lower in cathepsin B-deficient mice than in their wild-type counterparts for both HBsAg-specific CD4 and CD8 T cells (Figure 7B). A strong decrease in IFN-γ and IL-2 secretion was also observed in the supernatants of cathepsin B-deficient splenocytes in the same experimental setting (Figure 7C). However, despite these marked defects in both CD4 and CD8 Ag-specific T cells responses, antibody titers were not significantly reduced ( Figure 7D).

DiscUssiOn
QuilA saponins, and QS-21 in particular, are well-known adjuvants but, excluding NLRP3 inflammasome activation, little is known about the molecular and cellular mechanisms leading to their adjuvant effect. Here, we demonstrate that cholesteroldependent endocytosis is required to induce human moDC activation and that QS-21 accumulates in lysosomes and causes lysosomal membrane permeabilization. Cell activation depends on the activity of the Syk kinase and of cathepsin B (Figure 7E). Finally, we established that cathepsin B participates in the adjuvant properties of QS-21 on both CD4 and CD8 antigen-specific-T cell responses in vivo. QS-21 is a saponin with specific affinity for cell membrane cholesterol that can induce pore formation therein (8). Co-formulation with cholesterol is required to avoid toxicity when used as injectable (10). In this study, we used QS-21 formulated in liposomes containing cholesterol and thereby devoid of any measurable lytic activity. Membrane cholesterol in the target cells was required for both cell entry and activation, suggesting that either QS-21 is transferred from liposomal cholesterol to the cell membrane-associated cholesterol or that the whole liposome containing QS-21 is endocytosed via a cholesterol-dependent mechanism. Partial depletion of membrane cholesterol did not impact the endocytosis of either transferrin (TfR) or cholera toxin B (CTxB), arguing against clathrin-dependent and caveolin/lipid raft-dependent mechanisms for endocytosis of liposomal QS-21 (37). Cell entry could instead be mediated through cholesteroldependent macropinocytosis as shown for several viruses (38)(39)(40)(41)(42). Macropinocytosis of viruses can be receptor mediated (42), which could explain the surface protein dependence we observed for QS-21. Irrespective of the mechanism of entry, QS-21 was eventually transferred to and concentrated in dense lysosomes where it promoted lysosomal destabilization and the potential release of lysosomal content. Because particles of up to 40 kDa could leak out of lysosomes upon incubation with QS-21, it is likely that pores were formed in the lysosomal membrane. Several mechanisms could explain why QS-21 could form pores in lysosomal membranes but not in plasma membranes. On the one hand, due to the lysosomal accumulation and lengthy retention of QS-21, cholesterol may be preferentially extracted by QS-21 in these organelles. Reduced lysosomal membrane cholesterol has been linked to increased permeability to positive ions leading to osmotic imbalance and lysosomal destabilization (43)(44)(45). Lysosome destabilization by listeriolysin and lysosomotropic detergents is dependent on acid-driven conformational changes occurring in this organelle (46,47). This does not seem to be the mechanism for QS-21, as inhibition of lysosome acidification did not affect QS-21-mediated pore formation. Interestingly, inhibition of lysosome acidification also inhibits the response to QS-21 in human PBMCs. Indeed, treatment with chloroquine inhibited QS-21-induced production of IL-6 by PBMCs, although no effect of chloroquine administration was observed on T cell or antibody responses in healthy adults immunized with AS01-adjuvanted antigens (48). QS-21-mediated pore formation resulted in the translocation of macromolecules from lysosomes to the cytosol and induced activation marker upregulation and cytokine transcription. This complex activation pathway is most likely the reason for the different expression kinetics observed by PCR for cytokines induced by QS-21 when compared to MPL. Furthermore, the genes upregulated specifically by QS-21 identified in the microarray included EGR1, RGS1, NR4A2, and DUSP1 that are known TLR4-induced primary response genes. These genes are expressed as rapidly as 15 min following TLR4 stimulation and their expression returns to baseline after 2-3 h (25)(26)(27)(28). It is therefore likely that after 4 h, the time point chosen for the microarray analysis, the expression of these genes, had returned to baseline after stimulation with MPL but not yet with QS-21.
Other groups have reported that lysosomal permeabilization by compounds including Alum and QuilA activates the inflammasome, although pro-IL-1β expression required priming with a TLR-ligand (15,49,50). In contrast, here we show that QS-21 directly promoted the expression of pro-IL-1β and other cytokines. This discrepancy between our study and previous work may be due to the cell type studied or to the formulation. Indeed, a recent study has shown that stimulation with unformulated (i.e., not formulated with liposomes) QS-21 does not lead to the direct activation of murine bone marrow-derived dendritic cells or macrophages (14). It is possible that formulation of QS-21 with cholesterol-containing liposomes allows for better lysosomal targeting and activation of alternative signaling pathways. Cell type specificity has also been observed for saponin-dependent translocation of endocytosed antigens, as it occurs in human moDCs but not in human monocytes or macrophages (6). Human dendritic cells derived from monocytes in vitro have an increased capacity for lysosomal proteolysis when compared to other human or mouse dendritic cells and express high levels of lysosomal cathepsins B, D, L, and S (51). This distinctive characteristic may explain why saponins could promote both antigen translocation in and direct activation of human moDCs.
We have indeed shown that cathepsin B expression and activity are critical for the response of moDCs to QS-21. Furthermore, lysosomal cysteine proteases, such as cathepsins B and L, are involved in inflammasome activation downstream of lysosomal destabilization following stimulation with saponins (14,15). The exact mechanism by which cathepsin B could promote the transcription of pro-inflammatory cytokines remains elusive, although a role for cathepsins in NF-κB activation has already been described. The synthetic double-stranded RNA, poly IC, can induce lysosomal destabilization and cathepsin D release into the cytosol, where it mediates the cleavage of caspase-8, which is important for increased NF-κB activity (36). In our experiments however, cathepsin B did not seem to cleave caspase-8, as no cleavage fragments were detectable in cells stimulated with QS-21 (data not shown). Cathepsin B has nevertheless multiple other potential lysosomal and cytoplasmic targets.
In vivo, cathepsin B was not involved in Ag-specific antibody responses promoted by QS-21 or innate cell recruitment to the draining lymph node, yet cathepsin B-deficient mice showed decreased HBsAg-specific CD4 and CD8 T cell responses. This discrepancy could be due to a general defect in antigen presentation. Indeed, cathepsin B activity has been linked to antigen processing, which could affect antigen presentation on both class I and class II major histocompatibility complexes (52)(53)(54)(55)(56). Interestingly, ISCOMATRIX was shown to induce antigen translocation from lysosomes to the cytosol, thereby facilitating proteasome-independent cross-priming (6). We have also shown that QS-21 promotes pore formation in lysosomal membranes allowing the release of macromolecules of up to 40 kDa into the cytosol. Cathepsin B could partially degrade protein antigens in the lysosome and the fragments could be released into the cytosol by QS-21 for cross-presentation. It is, however, unlikely that this mechanism contributes to lessened QS-21-elicited Ag-specific CD8 T cell responses observed in cathepsin B-deficient mice, since QS-21 and antigens display different localizations and pharmacokinetic properties after immunization (16).
We have also identified Syk as a key signaling molecule for the response of moDCs to QS-21. Indeed, Syk was phosphorylated following QS-21 stimulation and Syk knockdown or pharmacological inhibition blocked NF-κB activation and cytokine production. Syk is a tyrosine kinase generally associated with ITAM-motif containing receptors (57), and can also play a role in lysosomal function in B cells. For example, B cell receptor cross-linking causes Syk-dependent changes in lysosomal pH, which lead to apoptosis (58). However, in our system, Syk was found to act downstream of lysosomal function, as bafilomycin A1 blocked QS-21-mediated Syk phosphorylation. Syk may therefore play a role of lysosomal integrity sensor and autophosphorylate following lysosomal membrane damage, a possibility that has already been suggested (59). Since Syk kinase can be activated by plasma membrane lipid alterations, it is possible that a similar mechanism could occur at the lysosomal membrane (32,60). However, neither Syk relocation to nor phosphorylation at the lysosomal membrane could be detected by confocal microscopy in our system (data not shown). Nevertheless, Syk could promote cathepsin B proteolytic activity as has been shown in other models (61). In summary, we describe a novel lysosome-dependent pathway ( Figure 7G) that contributes to the immunostimulatory properties of a clinically approved saponin-based vaccine adjuvant. This knowledge may help the rational development of adjuvants which could be instrumental to the success of future vaccination strategies.

eThics sTaTeMenT
Ethics committee (Erasme hospital) (comité d' éthique de la faculté de médecine, 808 route de Lennik, B-1070 Brussels, Belgium). Animal studies were approved by the local animal welfare committee (commission d' éthique en expérimentation animale du Biopôle ULB-Charleroi). Buffy coats were obtained from local blood donations by the Red Cross.
aUThOr cOnTriBUTiOns IW conducted most of the experiments. SD, FN, and SW contributed to some experiments. ST provided technical help for the