Oxidized phagosomal NOX2 complex is replenished from lysosomes

ABSTRACT In dendritic cells, the NADPH oxidase 2 complex (NOX2) is recruited to the phagosomal membrane during antigen uptake. NOX2 produces reactive oxygen species (ROS) in the lumen of the phagosome that kill ingested pathogens, delay antigen breakdown and alter the peptide repertoire for presentation to T cells. How the integral membrane component of NOX2, cytochrome b558 (which comprises CYBB and CYBA), traffics to phagosomes is incompletely understood. In this study, we show in dendritic cells derived from human blood-isolated monocytes that cytochrome b558 is initially recruited to the phagosome from the plasma membrane during phagosome formation. Cytochrome b558 also traffics from a lysosomal pool to phagosomes and this is required to replenish oxidatively damaged NOX2. We identified syntaxin-7, SNAP23 and VAMP8 as the soluble N-ethylmaleimide-sensitive factor attachment protein receptor (SNARE) proteins mediating this process. Our data describe a key mechanism of how dendritic cells sustain ROS production after antigen uptake that is required to initiate T cell responses. Highlighted Article: In human dendritic cells, the membrane component of the NADPH oxidase NOX2 complex is initially recruited to phagosomes from the plasma membrane, and oxidized NOX2 complex subunits are replenished from a lysosomal pool.


INTRODUCTION
Phagocytosis is an essential process by which immune cells clear microbial pathogens as well as apoptotic and necrotic (tumor) cells. Both the plasma membrane and intracellular vesicles contribute to phagosome formation, as intracellular vesicles fuse with the plasma membrane at the cup of the nascent phagosome (Canton and Grinstein, 2014;Fairn and Grinstein, 2012;Lu and Zhou, 2012). The protein and lipid compositions of phagosomes progressively change during phagosome formation and during the subsequent maturation steps, when early phagosomes convert into late phagosomes and then into lysosomes (Canton and Grinstein, 2014;Fairn and Grinstein, 2012;Lu and Zhou, 2012). A key player in the processing of ingested pathogens by phagocytes of the immune system is the NADPH oxidase NOX2 complex. NOX2 consists of the three cytosolic proteins: p47 phox (NCF1), p40 phox (NCF4) and p67 phox (NCF2), and the two integral membrane proteins p22 phox (CYBA) and gp91 phox (CYBB), which together are called cytochrome b 558 (Bedard and Krause, 2007;Groemping and Rittinger, 2005;Nauseef, 2008). Cytochrome b 558 traffics in complex, and p22 phox and gp91 phox need to form a stable heterodimer before exit from the endoplasmic reticulum (Casbon et al., 2009;Nauseef, 2008;Zhu et al., 2006). NOX2 produces reactive oxygen species (ROS) that facilitate the killing of the ingested pathogen by neutrophils (Bedard and Krause, 2007;Segal, 2005) and dendritic cells (Vulcano et al., 2004). In dendritic cells, NOX2 also modulates the antigenic repertoire for presentation to both CD4+ and CD8+ T cells (Allan et al., 2014;Bedard and Krause, 2007;Dingjan et al., 2016;Hoffmann et al., 2012;Jancic et al., 2007;Mantegazza et al., 2008;Matsue et al., 2003;Rybicka et al., 2012;Savina et al., 2006;Vulcano et al., 2004). NOX2 localizes to the plasma membrane, to early and recycling endosomes, and to antigen-containing endosomes and phagosomes (Bedard and Krause, 2007;Borregaard et al., 1983;Casbon et al., 2009;Nauseef, 2008). In macrophages and dendritic cells, cytochrome b 558 is recruited to phagosomes from vesicles of a recycling endosomal and lysosomal nature (Casbon et al., 2009;Jancic et al., 2007;Matheoud et al., 2013). How these vesicles are delivered to the phagosome is a field of intense study, and although several players in this process have been identified, such as the small GTPase Rab27a (Jancic et al., 2007) and the Ca 2+ -sensing protein synaptotagmin-11 (Arango Duque et al., 2013), many questions remain. One key open question is which soluble N-ethylmaleimidesensitive factor attachment protein receptor (SNARE) proteins catalyze the delivery of vesicles containing cytochrome b 558 to phagosomes.
SNARE proteins drive all intracellular membrane fusion reactions in eukaryotic cells (except mitochondrial fusion). Cognate SNAREs present in the donor and acceptor membranes engage and assemble into a four-α-helix coiled-coil bundle. Each helix of this bundle is contributed by a different SNARE motif belonging to separate subfamiliestermed Qa-, Qb-, Qc-and R-SNAREs, respectively (Hong, 2005;Jahn and Scheller, 2006). The SNAREs that catalyze the membrane fusion events underlying phagosome formation and maturation are not well characterized. In the mouse RAW264.7 macrophage cell line, one study has shown a clear switch from syntaxin (stx)12 to stx7 (both Qa) during uptake of sheep red blood cells (Collins et al., 2002). There, stx12 readily disappeared within 10 min after phagosome formation, and this was accompanied by an increase in the levels of stx7. This result suggests a role for stx12 in early, and stx7 in late, phagosomal fusion events, in accordance with their described roles in endosomal maturation (Antonin et al., 2000;Collins et al., 2002;McBride et al., 1999;Mills et al., 2001;Mullock et al., 2000;Nakamura et al., 2000;Prekeris et al., 1998;Pryor et al., 2004;Sun et al., 2003;Ward et al., 2000). For stx7, this agrees well with several quantitative proteomics studies in which stx7 increases in late (>60 min after uptake) compared to early (<30 min) phagosomes for different cell types and phagocytic cargoes (Dill et al., 2015;Goyette et al., 2012;Rogers and Foster, 2007). An exclusive role for stx12 in early phagosomal fusion events is much less clear, as stx12 can be detected on late phagosomes by proteomic analyses (Boulais et al., 2010;Buschow et al., 2012;Jutras et al., 2008;Lee et al., 2010) and western blotting (Fratti et al., 2003;Sakurai et al., 2012). In addition to stx7 and stx12, proteomics studies have identified dozens of other SNARE proteins present on phagosomes (Boulais et al., 2010;Buschow et al., 2012;Dill et al., 2015;Goyette et al., 2012;Jutras et al., 2008;Lee et al., 2010;Rogers and Foster, 2007), but the roles of all these SNAREs remain largely unexplored. The recruitment of some SNAREs, including VAMP8 (R-SNARE) and SNAP23 (both Qb and Qc motifs), can even occur in several phases during phagosome maturation (Rogers and Foster, 2007). All these findings indicate a complex cascade of phagosomal fusion events, with many different SNAREs involved at multiple trafficking steps during phagosomal maturation.
Recently, it has been shown that phagosomal recruitment of cytochrome b 558 is mediated by the R-SNARE VAMP8 (also called endobrevin), and that the intraphagosomal pathogen Leishmania suppresses an immune response by selectively cleaving VAMP8 in order to block NOX2 assembly (Matheoud et al., 2013). In this study, we have identified the Q-SNAREs mediating cytochrome b 558 recruitment to phagosomes containing zymosan ( particles derived from Saccharomyces cerevisiae) in dendritic cells derived from human blood-isolated monocytes. Through a combination of immunofluorescence microscopy, flow cytometry and siRNA knockdown, we demonstrate that gp91 phox has already been recruited from the plasma membrane to the phagocytic cup, sometime during formation of the phagosome. Concurrently, gp91 phox traffics to phagosomes from an intracellular pool residing in LAMP1-positive compartments. This serves to replenish oxidatively damaged NOX2 at the phagosome. The phagosomal recruitment of gp91 phox depends on the SNARE proteins VAMP8, SNAP23 and stx7 that can interact with each other and form a stable complex. siRNA-mediated knockdown of these SNAREs results in reduced recruitment of gp91 phox to the phagosome and impairs ROS production. Our findings contribute to our understanding of how dendritic cells are able to sustain the production of ROS for hours after antigen uptake, which is required for antigen processing and presentation Hoffmann et al., 2012;Jancic et al., 2007;Mantegazza et al., 2008;Matsue et al., 2003;Rybicka et al., 2012;Savina et al., 2006).

RESULTS
Gp91 phox is recruited to phagosomes from the plasma membrane and from late-endosomes/lysosomes In immune cells, gp91 phox is present on the plasma membrane and in intracellular compartments (Bedard and Krause, 2007;Borregaard et al., 1983;Casbon et al., 2009;Nauseef, 2008). Activation by immune stimuli can result in enrichment of gp91 phox on the plasma membrane (Bedard and Krause, 2007;Borregaard et al., 1983), raising the possibility that gp91 phox is recruited to nascent phagosomes from the plasma membrane. In order to determine the source of phagosomal gp91 phox in dendritic cells, we first determined which fraction of gp91 phox resided on the plasma membrane. Dendritic cells were derived from monocytes from the blood of healthy volunteers and incubated with zymosan particles for 1 h. The presence of gp91 phox on the plasma membrane was measured by flow cytometry using an antibody recognizing an extracellular epitope and no permeabilization. This value was compared to that for the total cellular pool of gp91 phox with permeabilization ( Fig. 1A). Around 10% of the total level of gp91 phox was present on the plasma membrane in resting dendritic cells, and this percentage increased to approximately 22% after zymosan stimulation (Fig. 1A). However, we observed a small (∼15%) but significant decrease in total cellular levels of gp91 phox upon zymosan stimulation (Fig. 1A,B), suggesting that gp91 phox is degraded following zymosan uptake. The decrease of cellular gp91 phox could be inhibited by addition of α-tocopherol, a free radical scavenger, during zymosan uptake (Fig. 1B), indicating that the degradation was due to oxidation of gp91 phox .
We then investigated whether gp91 phox is recruited to the phagosome from the plasma membrane during the formation of the phagocytic cup. By pre-incubating the cells with an antibody against gp91 phox before zymosan addition, we determined whether plasma membrane gp91 phox is taken up together with the zymosan particles during phagocytosis. Most phagosomes were positive for gp91 phox that had originated from the plasma membrane at 15 min after uptake, which is the earliest time point at which we obtained a sufficient number of phagosomes for analysis ( Fig. 1C, surface recruited). We compared this to the total pool of gp91 phox on phagosomes, which we visualized by immunostaining in the presence of detergent (Fig. 1C, total). To quantify the presence of gp91 phox on phagosomes, phagosomal membranes were first selected based on morphology (Fig. S1A). Then, we calculated the mean fluorescence intensity of the gp91 phox signal at the phagosomal membrane and normalized this by division over the mean fluorescence of the imaged area of the cell to correct for differences in staining efficiencies and varying expression levels among cells and donors. These normalized intensity values allow estimation of phagosomal enrichment for gp91 phox in an unbiased manner. Normalized intensity values above 1 indicate enrichment on the phagosomal membrane, whereas values below 1 indicate reduction. The phagosomal enrichment was calculated as the percentage of phagosomes with normalized intensity values above 1. Interestingly, the signal of plasma membrane-originated gp91 phox gradually decreased over time after uptake, whereas the signal of total gp91 phox on the phagosomes remained stable (Fig. 1D). This decrease of plasma membrane-originated gp91 phox signal was caused by sequestration of antibody-labeled gp91 phox from the phagosomes and not by degradation or dissociation of the antibody, as experiments with a control antibody (directed against the zymosan particles) showed no decrease in signal ( Fig. 1E;  Fig. S1B). To test whether gp91 phox was replenished at the membrane from intracellular compartments, the same preincubation experiment was performed with dendritic cells transfected with siRNA against VAMP8 (VAMP8 siRNA; 85% knockdown efficiency; Fig. 1F). VAMP8 has been shown to mediate gp91 phox recruitment to phagosomes in murine dendritic cells derived from bone marrow (Matheoud et al., 2013). After 60 min of incubation with zymosan, the phagosomal enrichment of plasma membrane-originated gp91 phox was equal for cells expressing non-targeting and VAMP8 siRNA (Fig. 1G). However, the total level of gp91 phox on phagosomes was only increased in non-targeting siRNA samples and not with VAMP8 siRNA (Fig. 1G). These results not only confirm that VAMP8 mediates the recruitment of gp91 phox to phagosomes (Matheoud et al., 2013) but also support the conclusions that gp91 phox is initially recruited to phagosomes from the plasma membrane and that there is a subsequent turnover of gp91 phox through replenishment from intracellular compartments.
Cytochrome b 558 is known to be recruited to phagosomes from vesicles of an endosomal/lysosomal nature (Arango Duque et al., 2013;Casbon et al., 2009;Jancic et al., 2007;Matheoud et al., 2013), and we tested whether gp91 phox was replenished from these compartments in our dendritic cells. Gp91 phox clearly overlapped with the late-endosomal/lysosomal marker LAMP1 in unstimulated cells ( Fig. 2A), similar to that reported previously for murine dendritic cells and human islet cells (Jancic et al., 2007;Li et al., 2012). However, LAMP1 seemed largely absent from gp91 phox -positive phagosomes, indicating that LAMP1 and gp91 phox localization might be mutually exclusive ( Fig. 2A). We calculated Pearson correlation coefficients in order to quantify the overlap of LAMP1 with gp91 phox (Fig. 2B). Phagosomes were, again, first selected based on morphology and then the Pearson correlation coefficients of LAMP1 with gp91 phox were calculated (Fig. S1A). These were compared with the Pearson correlation coefficients calculated for entire cells in the absence of zymosan. We used Pearson correlation coefficients rather than Mander's overlap coefficients as Pearson correlation coefficients are not influenced by changes in signal intensity (Adler and Parmryd, 2010;Dunn et al., 2011). Interestingly, the Pearson coefficient for LAMP1 with gp91 phox dropped from ∼0.7 in cells without zymosan to ∼0.3 with zymosan. To further investigate (A) Representative flow cytometry histograms (left) and mean fluorescence intensity bar graphs (right) of dendritic cells stained for surface gp91 phox (without detergent) or total gp91 phox (with detergent) in the presence or absence of zymosan. Gp91 phox levels were normalized to those in permeabilized untreated cells (n=6 donors; one-way ANOVA with Tukey's test). (B) Total cellular levels of gp91 phox following zymosan stimulation both in the presence and absence of α-tocopherol (α-Toc; n=3 donors; two-way ANOVA with Bonferroni's test). (C) Confocal images of dendritic cells that had been pre-incubated with an antibody against gp91 phox (recruited to the surface; green in merged image) or with immunostaining against total gp91 phox (total) after 15, 30, 60 and 120 min of stimulation with zymosan (magenta). Yellow arrowheads, gp91 phox -positive phagosomes. BF, brightfield. Scale bars: 10 µm. (D) Quantification of the phagosomal enrichment of gp91 phox shown in panel C (n=3 donors; two-way ANOVA with Bonferroni's test). (E) Comparison of surface-enriched gp91 phox (from panel D) with the presence of a control antibody directed against the zymosan particles (normalized to values at t=15 min; n=3 donors; two-way ANOVA with Bonferroni's test; representative images are shown in Fig. S1B). (F) Representative western blots and quantification of siRNA-mediated knockdown of VAMP8 relative to that with a non-targeting control siRNA (NT) (n=9 donors; two-sided paired Student's t-test). KD, knockdown; GAPDH, loading control. (G) Quantification of phagosomal enrichment of surface-recruited and total gp91 phox after 60 min of zymosan uptake upon siRNA knockdown of VAMP8 (KD) by immunofluorescence analysis (individual donors shown; two-way ANOVA with Bonferroni's test). Results show means±s.e.m.
the distribution of gp91 phox and LAMP1 on phagosomes, we performed timecourse pulse-chase experiments. Dendritic cells were pulsed with zymosan at 4°C (at which zymosan will bind to the cells but is not taken up) and subsequently washed and chased at 37°C (a temperature at which phagocytosis occurs). We observed a transition over time with the content of gp91 phox on phagosomes gradually decreasing and phagosomal LAMP1 gradually increasing, and this transition occurred via an intermediate phase at which the phagosomes were positive for both proteins (Fig. 2C,D). We could not detect a pool of gp91 phox at the Golgi network (Fig. 2E,F), indicating that gp91 phox was not recruited to phagosomes from the Golgi. These results indicate that gp91 phox traffics from LAMP1-rich late-endosomes/lysosomes to phagosomes containing no or low levels of LAMP1. (B) Pearson correlation coefficients between gp91 phox and LAMP1 channels from A (individual donors shown; two-sided unpaired Student's t-test). (C) Confocal images of dendritic cells with a pulse of zymosan at 4°C and chased for 15, 30, 60 or 120 min at 37°C that were then immunostained for gp91 phox (green in merge) and LAMP1 (magenta). Yellow arrowheads, LAMP1-positive phagosomes; magenta arrowheads, gp91 phox -positive phagosomes; cyan arrowheads, gp91 phox and LAMP1 double-positive phagosomes. Scale bar: 10 µm. (D) Quantification of the phagosomal enrichment of gp91 phox and LAMP1 shown in panel C. (E) As panel A but cells were stained for gp91 phox (green) and TGN38 (magenta). Scale bar: 10 µm. (F) Pearson correlation coefficients between gp91 phox and TGN38 channels from E (individual donors shown). (G) Confocal images of dendritic cells that had been transfected with the GFP-tagged PH domain of TAPP2 [binds to PI(3,4)P 2 ; cyan in merges], the PX domain of p40 phox [PI(3)P] or the N-terminus of MCOLN1 [PI(3,5)P 2 ] immunostained for gp91 phox (yellow) and LAMP1 (magenta). Pink arrowheads, phagosomes positive for both the described phosphoinositide (PI) and gp91 phox ; yellow arrowheads, phagosomes positive for both the described phosphoinositide and LAMP1. Scale bar: 10 µm. (H) Quantification of phagosomal enrichment of gp91 phox or LAMP1 on phosphoinositide-positive phagosomes from panel G (n=3 donors; one-way ANOVA with Tukey's test). Results show means±s.e.m.
For complete assembly and activity of NOX2, not only does cytochrome b 558 need to be present on the phagosome but also the 3-phosphoinositide lipids phosphatidylinositol (3,4)-bisphosphate [PI(3,4)P 2 ] and/or phosphatidylinositol (3)-phosphate [PI(3)P], which interact with p47 phox and p40 phox , respectively, two cytosolic subunits of NOX2 (Anderson et al., 2010;Ellson et al., 2001;Groemping and Rittinger, 2005;Kanai et al., 2001). We tested for the presence of these phosphoinositide lipids on phagosomes by transfecting our dendritic cells with GFP fused to the PH domain of TAPP2 (also known as PLEKHA2) or to the PX domain of p40 phox , which specifically recognize PI(3,4)P 2 and PI (3)P, respectively (Baranov et al., 2016;Kanai et al., 2001;Marshall et al., 2002). We also tested for the presence of late-endosomal/ lysosomal phosphatidylinositol (3,5)-bisphosphate [PI(3,5)P 2 ] using the N-terminal sequence of MCOLN1 fused to GFP (Baranov et al., 2016;Li et al., 2013). After pulsing the transfected cells with zymosan, we immunolabeled the cells for gp91 phox and LAMP1. PI(3,4)P 2 and PI(3)P were mostly present on gp91 phox -enriched phagosomes, but to a lesser extent or not present on LAMP1-enriched phagosomes (Fig. 2G,H). In contrast, PI(3,5) P 2 was present on phagosomes containing gp91 phox or LAMP1. These results show that phagosomes containing gp91 phox also contain the 3-phosphoinositides required for NOX2 activity. Taken together, our data show a turnover of NOX2 on phagosomes in which gp91 phox is initially recruited from the plasma membrane during formation of the nascent phagosome and, later, oxidatively damaged gp91 phox (see Fig. 1B) is replenished from LAMP1positive compartments of lysosomal nature. Our data further show that LAMP1-enriched phagosomes do not contain high levels of gp91 phox , nor do they contain sufficient levels of the 3-phosphoinositide species required for NOX2 activity.
Gp91 phox colocalizes with stx7, stx8, SNAP23, Vti1b and VAMP8 on phagosomes Next, we addressed the question of which SNAREs mediate the phagosomal recruitment of gp91 phox from late-endosomal/ lysosomal compartments. We focused on the potential roles of the Q-SNARE proteins SNAP23, Vti1b (Qb), stx7, stx8 (Qc) and stx12 in phagosomal cytochrome b 558 recruitment. We selected these Q-SNAREs for three reasons. First, these SNAREs interact with VAMP8 (Antonin et al., 2000;Murray et al., 2005a,b;Offenhäuser et al., 2011;Sakurai et al., 2012;Wade et al., 2001), which is the R-SNARE involved in phagosomal recruitment of cytochrome b 558 (Matheoud et al., 2013). Second, a recent study with phagosomes purified from human dendritic cells (i.e. the cell type studied here) identified the Q-SNAREs stx7, stx8 and stx12 on phagosomes (Buschow et al., 2012). The presence of SNAP23, stx7, stx12 and Vti1b on phagosomes has been confirmed by immunofluorescence, cellular fractionation and overexpression of tagged fusion proteins (Cai et al., 2011;Cebrian et al., 2011;Collins et al., 2002;Fratti et al., 2003;Nair-Gupta et al., 2014;Sakurai et al., 2012). Third, levels of cytochrome b 558 on phagosomes correlate with phagosomal SNAP23 in a macrophage cell line (Sakurai et al., 2012), and gp91 phox trafficking and ROS production could be inhibited by targeting SNAP23 in primary human neutrophils (Uriarte et al., 2011), suggesting a role for this SNARE in the phagosomal recruitment of NOX2 in dendritic cells. We started by investigating the localization of gp91 phox relative to that of VAMP8 and to the aforementioned candidate Q-SNARE proteins. As a negative control, we also immunolabeled cells for the ER and Golgi SNARE stx5 (Qa) (Xu et al., 2002;Zhang and Hong, 2001). The colocalization of gp91 phox with SNAP23, stx7, VAMP8, Vti1b and stx8 could be clearly observed, both in unstimulated cells and on zymosan-containing phagosomes ( Fig. 3A; Fig. S2A). To quantify this colocalization, we performed three unbiased analyses. First, we calculated Pearson correlation coefficients between the channels of interest on the phagosomal membrane (Fig. S1A). These Pearson correlation coefficients were compared with the Pearson correlation coefficients calculated for entire cells in the absence of zymosan. As a positive control for the maximum overlap observable in our samples, we immunolabeled cells with a primary antibody against gp91 phox and a combination of secondary antibodies labeled with two different fluorophores. The maximum Pearson coefficient observable in our samples calculated with our positive control was 0.92 ( Fig. 3B; Fig. S2B). In contrast to our previous observations with LAMP1 (Fig. 2B), the Pearson correlation coefficients measured for the SNARE proteins with gp91 phox were similar between the absence and presence of zymosan ( Fig. 3B; Fig. S2B). Pearson coefficients were between 0.5 and 0.7 for gp91 phox with Vti1b, stx7, VAMP8, stx8, SNAP23 and stx12 regardless of the presence of zymosan. Colocalization of Stx5 gave a Pearson correlation coefficient below 0.5. The differences in Pearson correlation coefficients for gp91 phox with Vti1b, stx7, VAMP8, stx8 and SNAP23 were all found to be statistically significant compared to gp91 phox with stx5 (Fig. 3C). Furthermore, the Pearson correlation coefficients for gp91 phox with Vti1b and VAMP8 were also significantly higher than the Pearson correlation coefficients for gp91 phox with SNAP23, stx7 and stx12. Thus, although we did observe some localization of stx5 at gp91 phox -positive phagosomes, the localization of Vti1b, stx7, VAMP8, stx8, SNAP23 and stx12 at these phagosomal membranes was more prevalent.
For the third analysis, we calculated the phagosomal enrichment factors (Fig. S1A). Note that this analysis only provides a rough estimate for the enrichment of proteins on phagosomes as the fluorescence on the phagosomal membrane is divided over the average fluorescence over the imaged cell area. We calculated the enrichment factors for phagosomes containing both gp91 phox and the SNARE (double positive), and for phagosomes containing only one of these proteins (Fig. 3E). The results of this analysis were in reasonable agreement with those from the Pearson and regression analyses, with the overlap with gp91 phox higher for Vti1b, stx7, VAMP8, stx8, SNAP23 and stx12 than for stx5. We also used phagosomal enrichment factors to quantify the kinetics of gp91 phox and SNARE recruitment to phagosomes (Fig. 3F). As we expected based on the phagosomal recruitment of gp91 phox from the plasma membrane shown by our data (Fig. 1C,D) and on the pulse-chase timecourse experiments with LAMP1 (Fig. 2C,D), gp91 phox was already enriched in the majority of the phagosomes within 15 min after stimulation (the earliest time point at which we obtained a sufficient number of phagosomes for analysis). The phagosomal enrichment of gp91 phox was stable or increased somewhat up to 60 min after uptake and decreased afterwards. Similarly, SNAP23 and Vti1b remained present at comparable levels over the entire 2 h timecourse of the experiment. The phagosomal enrichment of stx7, VAMP8 and stx8 gradually increased over time, indicating that these SNAREs were recruited from intracellular compartments to the phagosome. In contrast, the phagosomal enrichment of stx12 decreased over time, indicating that it was removed from the phagosomal membrane.
siRNA knockdown of stx7, SNAP23 and VAMP8 decreases phagosomal NOX2 complex recruitment and activity In order to investigate which SNAREs are involved in the phagosomal recruitment of gp91 phox from lysosomal compartments, we performed siRNA-mediated silencing of stx7, stx12 and SNAP23 ( Fig. 4A; average knockdown efficiencies of 51% for stx7, 56% for stx12 and 67% for SNAP23). As a positive control, we also transfected dendritic cells with siRNA against gp91 phox (92% knockdown efficiency; Fig. 4A) and against VAMP8 (85% knockdown efficiency; Fig. 1F), which is the R-SNARE required for phagosomal recruitment of gp91 phox (Fig. 1G) (Matheoud et al., 2013). We were unable to obtain substantial knockdown (<25%) of Vti1b and stx8 in dendritic cells. In unstimulated cells, we did not see any obvious rearrangement of gp91 phox upon knockdown of the SNAREs, and it still located to the plasma membrane and intracellular pools (Fig. S4A); however, due to the limited resolution of our microscope, we cannot exclude that ( part of ) the intracellular pool of gp91 phox mislocated to other intracellular compartments upon SNARE knockdown. For each condition, we evaluated the enrichment of gp91 phox to the phagosome 1 h after uptake of zymosan particles labeled with Alexa-Fluor-633. Each condition was compared to a non-targeting siRNA-transfected condition (Fig. 4B). Compared to results with the non-targeting siRNA, we observed significant loss of phagosomal gp91 phox recruitment upon knockdown of VAMP8, stx7 and SNAP23 (Fig. 4C). In contrast, we did not observe loss of gp91 phox recruitment upon stx12 knockdown (Fig. 4C). The phagosomal enrichment of LAMP1 and of the early endosomal marker EEA1 were unaffected upon VAMP8, stx7 and SNAP23 knockdown ( Fig. 4D; Fig. S4B), indicating that the reduced phagosomal recruitment of gp91 phox was not due to a general maturation defect. These results support a role for SNAP23 and stx7 (but not stx12) in the recruitment of gp91 phox to phagosomes.
For the complete assembly of NOX2, the cytosolic subunits p47 phox , p40 phox and p67 phox need to bind to gp91 phox (Bedard and Krause, 2007;Groemping and Rittinger, 2005;Nauseef, 2008). Indeed, recruitment of p67 phox to phagosomes was reduced upon knockdown of gp91 phox (Fig. 4E,F), confirming that gp91 phox is required for NOX2 assembly. In line with this, knockdown of stx7 (i. e. one of the SNAREs required for phagosomal recruitment of gp91 phox ) resulted in decreased recruitment of both gp91 phox and p67 phox to phagosomes (Fig. 4E,F). To further investigate the potential roles of SNAREs in the assembly of functional NOX2, we measured the effects of SNARE knockdown on phagosomal ROS production. We measured intraphagosomal ROS with the ROS probe OxyBURST Green H 2 DCFDA (2′,7′-dichlorodihydrofluorescein diacetate), which we covalently coupled to zymosan particles. Oxidation of the non-fluorescent H 2 DCF by ROS results in the formation of the highly fluorescent dichlorofluorescein, thereby providing a direct means for detecting ROS in phagosomes of dendritic cells by confocal microscopy (Bass et al., 1983;Hempel et al., 1999) (Fig. 4G). Upon knockdown of VAMP8, stx7 and SNAP23, the OxyBURST fluorescence was decreased compared to that with the non-targeting siRNA (Fig. 4H). We did not observe a significant decrease in ROS production upon stx12 knockdown (Fig. 4H). These findings further support the involvement of stx7, SNAP23 and VAMP8 in gp91 phox recruitment to the phagosome. These SNAREs could function in membrane trafficking steps upstream of the recruitment of gp91 phox to the phagosome. However, as VAMP8, stx7 and SNAP23 are all located at the phagosomal membrane, it is also conceivable that they directly mediate the delivery of gp91 phox to phagosomes, with VAMP8 (R-SNARE) complexing with stx7 (Qa) and SNAP23 (Qb and Qc). A SNARE complex comprising VAMP8, stx7 and SNAP23 has not been described in the literature and we next determined whether these SNAREs can form a complex.
SNAP23, stx7 and VAMP8 can form a SNARE complex VAMP8 is well known to complex with stx7 (Antonin et al., 2000(Antonin et al., , 2002Pryor et al., 2004;Wade et al., 2001) and with SNAP23 (Paumet et al., 2000;Wang et al., 2004). The interaction of stx7 with SNAP23 has not been described, but stx7 can be immunoprecipitated with overexpressed SNAP23 in J774 macrophages (Sakurai et al., 2012). To investigate whether stx7 complexes with endogenous SNAP23 in dendritic cells, dendritic cell lysates were incubated with antibodies against stx7, and the immunoprecipitates were subjected to western blot analysis for SNAP23 (Fig. 5A). We observed the presence of SNAP23 in the immunoprecipitates, and this presence increased >10-fold when we blocked NSF-mediated SNARE disassembly with N-ethylmaleimide (NEM). The presence of zymosan increased the interaction of SNAP23 with stx7, but not significantly. This experiment demonstrates that SNAP23 interacts with stx7 in vivo; we then addressed whether these two SNAREs could form a ternary SNARE complex with VAMP8. We combined recombinantly expressed (in Escherichia coli) and purified full-length SNAP23, stx7 and VAMP8 in a stoichiometric ratio. After overnight incubation, this mixture was subjected to SDS-PAGE. A four-helix bundle SNARE complex is very stable and cannot be denatured by SDS at room temperature, but it does disassemble upon heating the sample to 95°C. We observed clear SNARE complex formation (i.e. multiple bands running at high molecular masses) provided that all three SNARE proteins were present, whereas no complexes could be observed with binary mixtures of only two of the SNARE proteins present (Fig. 5B). This experiment demonstrates that SNAP23, VAMP8 and stx7 are capable of forming a SNARE complex in vitro.

DISCUSSION
An essential step for the function of NOX2 is the recruitment of its transmembrane component cytochrome b 558 to phagosomes. In this study, we found that gp91 phox , which traffics together with p22 phox (Casbon et al., 2009;Nauseef, 2008;Zhu et al., 2006), is initially internalized together with the zymosan phagocytic cargo from the plasma membrane (Fig. 6). During or after formation of the phagosome, an intracellular pool of gp91 phox has already been trafficked to phagosomes from LAMP1-rich compartments of a lateendosomal/lysosomal nature (Fig. 6), as has been reported previously (Jancic et al., 2007;Matheoud et al., 2013). When phagosomes convert into LAMP1-rich lysosomes, they contain no or only low levels of gp91 phox and the 3-phosphoinositides PI(3,4) P 2 and PI(3)P required for NOX2 activity. Our data show a decrease of cellular levels of gp91 phox following zymosan uptake, and this decrease can be blocked by the radical scavenger α-tocopherol; the trafficking from intracellular compartments is likely to serve to replenish gp91 phox affected by oxidative damage. Thus, the turnover of the phagosome allows a sustained production of ROS by NOX2 during phagosome maturation. This is likely to be of particular importance for dendritic cell function, where, and in contrast to the more transient oxidative burst in neutrophils and macrophages, sustained ROS production is essential for antigen processing and presentation Hoffmann et al., 2012;Jancic et al., 2007;Mantegazza et al., 2008;Matsue et al., 2003;Rybicka et al., 2012;Savina et al., 2006).
We identified a role for stx7, SNAP23 and VAMP8 in trafficking of cytochrome b 558 to the antigen-containing phagosome. VAMP8 and SNAP23 have been shown previously to be involved in this process (Matheoud et al., 2013;Sakurai et al., 2012;Uriarte et al., Fig. 4. siRNA knockdown of Stx7, SNAP23 and VAMP8 reduces phagosomal recruitment of cytochrome b 558 and ROS production. (A) Representative western blots and quantification of siRNA knockdown (KD) of stx7 (51% knockdown efficiency), stx12 (56%), SNAP23 (67%) and gp91 phox (92%) relative to levels with a non-targeting siRNA control (NT). GAPDH, loading control (n=8 donors; two-sided paired Student's t-test). (B) Confocal images of dendritic cells pulsed with Alexa-Fluor-633-conjugated zymosan particles (magenta in merge) and then immunolabeled for gp91 phox (green) with knockdown of VAMP8, stx7, stx12 or SNAP23. Yellow arrowheads, phagosomes enriched for gp91 phox ; red arrowheads, phagosomes negative for gp91 phox . BF, brightfield. Scale bars: 10 µm. (C) Quantification of the phagosomal enrichment of gp91 phox from panel B (individual donors shown; one-way ANOVA with Dunnett's test; representative confocal images of gp91 phox KD are shown in panel F). ns, not significant. (D) Quantification of the phagosomal enrichment of EEA1 and LAMP1 (individual donors shown; representative confocal images are shown in Fig. S4B). (E) Quantification of the enrichment of gp91 phox and p67 phox to zymosan-containing phagosomes in dendritic cells with knockdown of gp91 phox or stx7 (n=3 donors; one-way ANOVA with Tukey's test). (F) Representative confocal images for data presented in E. Yellow arrowheads, phagosomes enriched for gp91 phox (green) and p67 phox (magenta); red arrowheads, phagosomes negative for gp91 phox and p67 phox . BF, brightfield. Scale bar: 10 µm. (G) Confocal images of zymosan-pulsed dendritic cells with VAMP8, stx7, stx12 or SNAP23 knockdown. Zymosan particles were labeled with ROS-sensitive OxyBURST that becomes fluorescent upon oxidation (shown in rainbow lookup table). Scale bar: 10 µm. (H) Intraphagosomal ROS production calculated from the fluorescence intensities shown in panel G (normalized to those with non-targeting siRNA; one-way ANOVA with Dunnett's test; individual donors shown). Results show mean±s.e.m. 2011), and our data confirm this. Although SNAP23 has long been considered to be a plasma membrane SNARE (Chamberlain and Gould, 2002;Hong, 2005;Kawanishi et al., 2000;Pagan et al., 2003;Paumet et al., 2000;Tellam et al., 1997;Volchuk et al., 1996;Wang et al., 2004), it is increasingly clear that it also has intracellular functions in many different cell types (Aikawa et al., 2006;Boström et al., 2010;Guo et al., 1998;Martín-Martín et al., 2000;Mollinedo et al., 2006;Nair-Gupta et al., 2014;Sakurai et al., 2012;St-Denis et al., 1999;Takuma et al., 2002;Wang et al., 1997). The localization of SNAP23 to phagosomes was only discovered recently, and it was implied to mediate phagosomal recruitment of gp91 phox and MHC class I (Nair-Gupta et al., 2014;Sakurai et al., 2012). Stx7 is widely considered to be a late-endosomal SNARE that mediates fusion of late-endosomes and lysosomes (Antonin et al., 2000;Mullock et al., 2000;Nakamura et al., 2000;Pryor et al., 2004;Ward et al., 2000). Our findings are in line with this given that cytochrome b 558 is recruited from intracellular compartments of a lysosomal nature to phagosomes. As our data show that SNAP23, stx7 and VAMP8 all localize to phagosomes with gp91 phox and that they can form a complex together, it seems plausible that these SNAREs directly catalyze the fusion of gp91 phox -containing vesicles with phagosomes. However, it could also be that these SNAREs catalyze upstream fusion steps needed for cytochrome b 558 trafficking to the phagosome. Accordingly, we cannot exclude the involvement of other SNAREs, including stx8 and Vti1b, due to insufficient knockdown levels. Finally, it may well be that the trafficking of cytochrome b 558 differs among cell types or stimuli, especially considering that zymosan results in much more pronounced NOX2 activity compared to other immune stimuli (Gantner et al., 2003). Stimulus-dependent assembly of NOX2 occurs in primary mouse neutrophils, where NOX2 assembles within seconds on phagosomes bearing serum-opsonized Staphylococcus aureus, as opposed to phagosomes containing immunoglobulin G-bound targets, where the NOX2 complex assembles on a tubulovesicular compartment at the base of the emerging phagosome (Anderson et al., 2010).
Possibly, the vesicles trafficking cytochrome b 558 to phagosomes also contain late-endosomal/lysosomal cargo molecules such as the V-ATPase and/or lysosomal cathepsin proteases. This could be the case since the V-ATPase and LAMP1 are recruited to phagosomes in a SNAP23-dependent manner, similar to cytochrome b 558 in a mouse macrophage cell line (Sakurai et al., 2012). In macrophages, V-ATPase subunit a3 is recruited from late-endosomal/lysosomal compartments containing LAMP2 and stx7, and this occurs early during phagosome formation (Sun-Wada et al., 2009), similar to the recruitment of gp91 phox observed in our study. However, we found that LAMP1 and gp91 phox do not colocalize on phagosomes, and we did not detect a defect of LAMP1 trafficking upon SNARE knockdown. Moreover, both LAMP1 and the V-ATPase are recruited to phagosomes later than gp91 phox in human dendritic cells (Mantegazza et al., 2008), and LAMP1 and cathepsins are recruited later than the V-ATPase in murine macrophages (Sun-Wada et al., 2009;Tsang et al., 2000;Yates et al., 2007), making it unlikely that cytochrome b 558 -trafficking vesicles also contain these molecules. Another interesting question is how cytochrome b 558 trafficking to the phagosomal membrane is regulated. Key proteins for this process include the small GTPase Rab27a in murine dendritic cells (Jancic et al., 2007) and the Ca 2+ -sensor synaptotagmin-11 in murine macrophages (Arango Duque et al., 2013). The recruitment of gp91 phox may also be regulated by phosphorylation of SNAP23, as it has recently been shown that recruitment of MHC class I to phagosomes is promoted by IKK2dependent phosphorylation of SNAP23 in murine dendritic cells (Nair-Gupta et al., 2014). Similar to NOX2 Fig. 5. Complex formation of stx7, SNAP23 and VAMP8. (A) Immunoprecipitation (IP) of stx7 (α-stx7) from zymosan-pulsed dendritic cells with or without NEM and probed for SNAP23 (α-SNAP23). Shown are a representative western blot and the quantification of the band intensities normalized to control conditions (without zymosan and NEM) (mean±s.e.m. of five donors; one-way ANOVA with Tukey's test). WCL, whole-cell lysate. First lane shows only antibody in buffer without lysate. Results are mean±s.e.m. (B) In vitro complex formation of purified full-length VAMP8 (blue arrowhead), SNAP23 (green) and stx7 (red) analyzed by SDS-PAGE and Coomassie staining. Ternary SNARE complexes (multiple bands between ∼250 and 37 kDa; pink arrowheads) are SDS resistant at 20°C but disassemble at 95°C. A representative gel from three experiments is shown. Jancic et al., 2007;Mantegazza et al., 2008;Savina et al., 2006), the presence of MHC class I at the antigen-containing phagosome is important for cross-presentation to cytotoxic T lymphocytes by dendritic cells (Burgdorf et al., 2008;Joffre et al., 2012). Our findings contribute to the emerging concept that SNAP23 is a key regulator for dendritic cell function. Targeting SNAP23 and other SNAREs responsible for phagosomal maturation might be a novel strategy to combat autoimmune diseases, infection and cancer.
For microscope sample preparation, 50,000 dendritic cells were plated on 12-mm-diameter glass coverslips in the presence of unlabeled or Alexa-Fluor-633-labeled zymosan particles in a particle-to-cell ratio of 5:1 in serumfree RPMI medium with 1% AA and 2 mM UltraGlutamine, and incubated at 37°C under 5% CO 2 . For the pulse-chase experiments, cells were incubated for 30 min with zymosan at 4°C, subsequently washed and incubated for 15, 30, 60 or 120 min at 37°C under 5% CO 2 . The cells were washed with PBS and fixed in 4% paraformaldehyde (PFA) in PBS for 15 min at room temperature. Cells were permeabilized with 0.1% (v/v) saponin and blocked with confocal laser scanning microscopy (CLSM) buffer [PBS, 20 mM glycine and 3% (w/v) BSA] for 30 min. For immunostaining, the cells were incubated with primary antibodies (1:200) in CLSM with 0.1% saponin for 1 h at room temperature or overnight at 4°C. Subsequently, cells were washed with PBS and incubated with secondary antibodies (1:400) for 1 h at room temperature. Cells were embedded in 68% glycerol with DAPI.

Preincubation with antibodies against gp91 phox
Cells on coverslips were incubated with or without anti-gp91 phox antibody (1:400) for 30 min at 4°C. Cells were washed and stimulated with a Alexa-Fluor-633-labeled zymosan particles (5 particles/cell) at 37°C under 5% CO 2 . Subsequently, the cells were fixed, blocked and permeabilized as described above. Samples without anti-gp91 phox preincubation were stained with anti-gp91 phox antibody (1:200) in CLSM with 0.1% saponin and incubated overnight. This was followed by staining and mounting as described above. For the control antibody experiments, zymosan conjugated to FITC (Baranov et al., 2016) was incubated with Alexa-Fluor-647-labeled anti-FITC antibody (1:200) for 30 min at 4°C, extensively washed and resuspended in RPMI medium. Cells were incubated with these particles and stained with Alexa-Fluor-568-labeled secondary antibody as described above.

Phosphoinositide transfection
For colocalization experiments with phosphoinositides, cells were transfected with plasmids (10 pg/cell) encoding the PH domain of TAPP2, the PX domain of p40 phox or the N-terminus of MCOLN1, as described previously (Baranov et al., 2016). At 7 h post transfection, cells were stimulated with unlabeled zymosan (5 particles/cell) in serum-free RPMI for 1 h at 37°C, and fixed and stained as described before.

Microscopy and colocalization analysis
All samples were imaged with a Leica SP8 confocal laser scanning microscope with a 63×1.20 NA water immersion objective (Leica HC PL APO 63×/1.20 W CORR CS2). All images were evaluated with Fiji (ImageJ 1.49 s). Colocalization and intensity levels were evaluated by first selecting a phagosome based on morphology. The areas corresponding to phagosomal membranes were then selected, and finally colocalization of proteins was calculated using the Pearson correlation coefficient. All images were evaluated independently. At least 25 phagosomes for at least three independent experiments each were analyzed.

Flow cytometry
Cells were cultured with Alexa-Fluor-633-labeled zymosan (5 particles/ cell) with or without 500 µM α-tocopherol in serum-free RPMI with 1% AA and 2 mM UltraGlutamine, and incubated for 30, 60 or 120 min. Cells were then washed with PBS supplemented with 0.5% BSA (in phosphatebuffered albumin PBA) and fixed in 4% PFA in PBS for 4 min on ice. Cells were blocked with PBA supplemented with 2% human serum, or blocked and permeabilized with 0.5% (v/v) saponin in PBA for 10 min on ice. For staining, cells were incubated with anti-gp91 phox antibody or isotype control (1:200) in PBA with or without 0.5% saponin for 30 min at 4°C. Subsequently, the cells were washed with PBS and incubated with goatanti-mouse IgG (H+L) Alexa-Fluor-488 (1:400) in PBA with or without 0.5% saponin for 30 min at 4°C. After incubation, the fluorescence intensity of Alexa-Fluor-488 was measured using flow cytometry (excitation, 488 nm; emission, 530/30 nm; FACSCalibur, BD Biosciences).

Immunoprecipitation
Cells were incubated with or without zymosan (50 particles/cell) for 45 min at 37°C. Cells were washed twice with PBS and incubated with or without NEM at a final concentration of 2 mM for 15 min at 37°C. NEM was quenched by washing once with 4 mM DTT. Cells were lysed in immunoprecipitation buffer (IP) buffer (20 mM Tris-HCl pH 7.6, 137 mM NaCl, 1% IGEPAL, 2 mM EDTA). Lysates were precleared and incubated with 1 µg of mouse IgG1 anti-stx7 antibody for 30 min on ice, followed by incubation overnight with protein-G-Sepharose at 4°C. Samples were washed with IP buffer, and protein was eluted by boiling in SDS sample buffer. IP samples and 4% of total lysates were analyzed by western blotting for SNAP23 and stx7.

SNARE complex formation
Fragments of rat full-length VAMP8, rat full-length stx7 and rat full-length SNAP23 were purified as described previously (Antonin et al., 2002;Stein et al., 2009). For assembly of the SNARE complexes, 10 mM of the three protein fragments was incubated overnight at room temperature. Samples were analyzed by SDS-PAGE with colloidal Coomassie Blue (ThermoFisher).

Statistical analysis
Sample sizes represent the number of individual donors or phagosomes as indicated in the figures. Data were analyzed using one-way ANOVA (following post-hoc Tukey or Dunnett's tests), two-way ANOVA (following post-hoc Bonferroni's test), regression analysis or Student's (un-)paired two-sided t-tests. A value of P<0.05 was considered statistically significant for all statistical analyses (*P<0.05; **P<0.01; ***P<0.001).