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.
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: p47phox (NCF1), p40phox (NCF4) and p67phox (NCF2), and the two integral membrane proteins p22phox (CYBA) and gp91phox (CYBB), which together are called cytochrome b558 (Bedard and Krause, 2007; Groemping and Rittinger, 2005; Nauseef, 2008). Cytochrome b558 traffics in complex, and p22phox and gp91phox 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 b558 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 Ca2+-sensing protein synaptotagmin-11 (Arango Duque et al., 2013), many questions remain. One key open question is which soluble N-ethylmaleimide-sensitive factor attachment protein receptor (SNARE) proteins catalyze the delivery of vesicles containing cytochrome b558 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 subfamilies – termed 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 b558 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 b558 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 gp91phox has already been recruited from the plasma membrane to the phagocytic cup, sometime during formation of the phagosome. Concurrently, gp91phox 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 gp91phox 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 gp91phox 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 (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).
Gp91phox is recruited to phagosomes from the plasma membrane and from late-endosomes/lysosomes
In immune cells, gp91phox 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 gp91phox on the plasma membrane (Bedard and Krause, 2007; Borregaard et al., 1983), raising the possibility that gp91phox is recruited to nascent phagosomes from the plasma membrane. In order to determine the source of phagosomal gp91phox in dendritic cells, we first determined which fraction of gp91phox 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 gp91phox 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 gp91phox with permeabilization (Fig. 1A). Around 10% of the total level of gp91phox 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 gp91phox upon zymosan stimulation (Fig. 1A,B), suggesting that gp91phox is degraded following zymosan uptake. The decrease of cellular gp91phox 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 gp91phox.
We then investigated whether gp91phox 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 gp91phox before zymosan addition, we determined whether plasma membrane gp91phox is taken up together with the zymosan particles during phagocytosis. Most phagosomes were positive for gp91phox 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 gp91phox on phagosomes, which we visualized by immunostaining in the presence of detergent (Fig. 1C, total). To quantify the presence of gp91phox on phagosomes, phagosomal membranes were first selected based on morphology (Fig. S1A). Then, we calculated the mean fluorescence intensity of the gp91phox 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 gp91phox 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 gp91phox gradually decreased over time after uptake, whereas the signal of total gp91phox on the phagosomes remained stable (Fig. 1D). This decrease of plasma membrane-originated gp91phox signal was caused by sequestration of antibody-labeled gp91phox 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 gp91phox 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 gp91phox 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 gp91phox was equal for cells expressing non-targeting and VAMP8 siRNA (Fig. 1G). However, the total level of gp91phox 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 gp91phox to phagosomes (Matheoud et al., 2013) but also support the conclusions that gp91phox is initially recruited to phagosomes from the plasma membrane and that there is a subsequent turnover of gp91phox through replenishment from intracellular compartments.
Cytochrome b558 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 gp91phox was replenished from these compartments in our dendritic cells. Gp91phox 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 gp91phox-positive phagosomes, indicating that LAMP1 and gp91phox localization might be mutually exclusive (Fig. 2A). We calculated Pearson correlation coefficients in order to quantify the overlap of LAMP1 with gp91phox (Fig. 2B). Phagosomes were, again, first selected based on morphology and then the Pearson correlation coefficients of LAMP1 with gp91phox 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 gp91phox dropped from ∼0.7 in cells without zymosan to ∼0.3 with zymosan. To further investigate the distribution of gp91phox 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 gp91phox 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 gp91phox at the Golgi network (Fig. 2E,F), indicating that gp91phox was not recruited to phagosomes from the Golgi. These results indicate that gp91phox traffics from LAMP1-rich late-endosomes/lysosomes to phagosomes containing no or low levels of LAMP1.
For complete assembly and activity of NOX2, not only does cytochrome b558 need to be present on the phagosome but also the 3-phosphoinositide lipids phosphatidylinositol (3,4)-bisphosphate [PI(3,4)P2] and/or phosphatidylinositol (3)-phosphate [PI(3)P], which interact with p47phox and p40phox, 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 p40phox, which specifically recognize PI(3,4)P2 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)P2] 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 gp91phox and LAMP1. PI(3,4)P2 and PI(3)P were mostly present on gp91phox-enriched phagosomes, but to a lesser extent or not present on LAMP1-enriched phagosomes (Fig. 2G,H). In contrast, PI(3,5)P2 was present on phagosomes containing gp91phox or LAMP1. These results show that phagosomes containing gp91phox also contain the 3-phosphoinositides required for NOX2 activity. Taken together, our data show a turnover of NOX2 on phagosomes in which gp91phox is initially recruited from the plasma membrane during formation of the nascent phagosome and, later, oxidatively damaged gp91phox (see Fig. 1B) is replenished from LAMP1-positive compartments of lysosomal nature. Our data further show that LAMP1-enriched phagosomes do not contain high levels of gp91phox, nor do they contain sufficient levels of the 3-phosphoinositide species required for NOX2 activity.
Gp91phox colocalizes with stx7, stx8, SNAP23, Vti1b and VAMP8 on phagosomes
Next, we addressed the question of which SNAREs mediate the phagosomal recruitment of gp91phox 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 b558 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 b558 (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 b558 on phagosomes correlate with phagosomal SNAP23 in a macrophage cell line (Sakurai et al., 2012), and gp91phox 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 gp91phox 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 gp91phox 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 gp91phox 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 gp91phox were similar between the absence and presence of zymosan (Fig. 3B; Fig. S2B). Pearson coefficients were between 0.5 and 0.7 for gp91phox 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 gp91phox with Vti1b, stx7, VAMP8, stx8 and SNAP23 were all found to be statistically significant compared to gp91phox with stx5 (Fig. 3C). Furthermore, the Pearson correlation coefficients for gp91phox with Vti1b and VAMP8 were also significantly higher than the Pearson correlation coefficients for gp91phox with SNAP23, stx7 and stx12. Thus, although we did observe some localization of stx5 at gp91phox-positive phagosomes, the localization of Vti1b, stx7, VAMP8, stx8, SNAP23 and stx12 at these phagosomal membranes was more prevalent.
For the second analysis, we correlated the average fluorescence intensity of each of the SNAREs at the phagosomal membranes with that of gp91phox for all individual phagosomes and normalized these by dividing them by the mean fluorescence of the imaged area of the cell (Fig. S1A). We then performed linear regression analysis. Our positive control of gp91phox immunostaining with two differently labeled secondary antibodies showed almost perfect positive correlation (Fig. 3D; β=0.95; R2=0.96). The correlations of most of the SNAREs with gp91phox were in reasonable agreement with the Pearson correlation coefficients. We observed correlation with the SNAREs Vti1b (β=0.59; R2=0.39), stx7 (β=0.35; R2=0.17), VAMP8 (β=0.38; R2=0.35), stx12 (β=0.46; R2=0.26) and stx8 (β=0.52; R2=0.29; Fig. 3D; Fig. S3). No or only weak correlation was observed with stx5 (β=0.06; R2=0.05). Interestingly, SNAP23 showed the strongest correlation with gp91phox of all SNAREs tested (β=0.80; R2=0.57), supporting the possibility that cytochrome b558 delivery to phagosomes is mediated by SNAP23, as suggested previously (Sakurai et al., 2012; Uriarte et al., 2011).
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 gp91phox 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 gp91phox higher for Vti1b, stx7, VAMP8, stx8, SNAP23 and stx12 than for stx5. We also used phagosomal enrichment factors to quantify the kinetics of gp91phox and SNARE recruitment to phagosomes (Fig. 3F). As we expected based on the phagosomal recruitment of gp91phox from the plasma membrane shown by our data (Fig. 1C,D) and on the pulse–chase timecourse experiments with LAMP1 (Fig. 2C,D), gp91phox 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 gp91phox 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 gp91phox 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 gp91phox (92% knockdown efficiency; Fig. 4A) and against VAMP8 (85% knockdown efficiency; Fig. 1F), which is the R-SNARE required for phagosomal recruitment of gp91phox (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 gp91phox 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 gp91phox mislocated to other intracellular compartments upon SNARE knockdown. For each condition, we evaluated the enrichment of gp91phox 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 gp91phox recruitment upon knockdown of VAMP8, stx7 and SNAP23 (Fig. 4C). In contrast, we did not observe loss of gp91phox 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 gp91phox was not due to a general maturation defect. These results support a role for SNAP23 and stx7 (but not stx12) in the recruitment of gp91phox to phagosomes.
For the complete assembly of NOX2, the cytosolic subunits p47phox, p40phox and p67phox need to bind to gp91phox (Bedard and Krause, 2007; Groemping and Rittinger, 2005; Nauseef, 2008). Indeed, recruitment of p67phox to phagosomes was reduced upon knockdown of gp91phox (Fig. 4E,F), confirming that gp91phox is required for NOX2 assembly. In line with this, knockdown of stx7 (i.e. one of the SNAREs required for phagosomal recruitment of gp91phox) resulted in decreased recruitment of both gp91phox and p67phox 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 H2DCFDA (2′,7′-dichlorodihydrofluorescein diacetate), which we covalently coupled to zymosan particles. Oxidation of the non-fluorescent H2DCF 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 gp91phox recruitment to the phagosome. These SNAREs could function in membrane trafficking steps upstream of the recruitment of gp91phox 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 gp91phox 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, 2002; Pryor 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.
An essential step for the function of NOX2 is the recruitment of its transmembrane component cytochrome b558 to phagosomes. In this study, we found that gp91phox, which traffics together with p22phox (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 gp91phox has already been trafficked to phagosomes from LAMP1-rich compartments of a late-endosomal/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 gp91phox and the 3-phosphoinositides PI(3,4)P2 and PI(3)P required for NOX2 activity. Our data show a decrease of cellular levels of gp91phox 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 gp91phox 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 (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).
We identified a role for stx7, SNAP23 and VAMP8 in trafficking of cytochrome b558 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., 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 gp91phox 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 b558 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 gp91phox and that they can form a complex together, it seems plausible that these SNAREs directly catalyze the fusion of gp91phox-containing vesicles with phagosomes. However, it could also be that these SNAREs catalyze upstream fusion steps needed for cytochrome b558 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 b558 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 b558 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 b558 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 gp91phox observed in our study. However, we found that LAMP1 and gp91phox 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 gp91phox 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 b558-trafficking vesicles also contain these molecules. Another interesting question is how cytochrome b558 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 Ca2+-sensor synaptotagmin-11 in murine macrophages (Arango Duque et al., 2013). The recruitment of gp91phox 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 IKK2-dependent phosphorylation of SNAP23 in murine dendritic cells (Nair-Gupta et al., 2014). Similar to NOX2 (Dingjan et al., 2016; 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.
MATERIALS AND METHODS
Primary cultures of human monocyte-derived dendritic cells were generated from peripheral blood monocytes (PBMCs) obtained from buffy coats of healthy individuals (informed consent obtained and approved by RadboudUMC ethical committee) according to institutional guidelines and as described previously (de Vries et al., 2005). Monocytes were differentiated into dendritic cells by culturing for 6 days at 37°C under 5% CO2 in complete RPMI-1640 (Gibco, ThermoFisher) supplemented with 10% fetal bovine serum, 2 mM UltraGlutamine (Lonza), 1% antibiotic-antimytotic (AA, Gibco, ThermoFisher), 300 U/ml IL-4 and 450 U/ml granulocyte-macrophage colony-stimulating factor (GM-CSF).
Primary antibodies used were: mouse-IgG1 anti-gp91phox (D162-3, MBL), rabbit serum anti-SNAP23 (111202, Synaptic Systems), rabbit serum anti-stx8 (110083, Synaptic Systems), rabbit serum anti-stx7 (110072, Synaptic Systems), mouse-IgG1 anti-stx7 (sc-514157, Santa Cruz), rabbit serum anti-stx12 (299022, Synaptic Systems), rabbit serum anti-VAMP8 (104302, Synaptic Systems), rabbit serum anti-Vti1b (164002, Synaptic Systems), rabbit serum anti-stx5 (110053, Synaptic Systems), rabbit serum anti-LAMP1 (L1418, Sigma), rabbit serum anti-TGN38 (sc-27680, Santa Cruz), mouse-IgG1 anti-EEA1 (610456, BD Biosciences), rabbit serum anti-p67phox (07-002, Merck Millipore), rabbit monoclonal-IgG anti-GAPDH (2118, Cell Signaling Technology), mouse-IgG1 anti-FITC (200-602-037, Jackson ImmunoResearch Laboratories) and mouse-IgG1 isotype control (400102, Biolegend) antibodies. The following secondary antibodies were used for immunofluorescence: goat anti-mouse IgG (H+L) Alexa-Fluor-488 or -568-conjugated (A-11029 and A-11031, ThermoFisher) and goat anti-rabbit IgG (H+L) Alexa-Fluor-568 or -647-conjugated (A-11036 and A-21245, ThermoFisher) antibodies. For immunoblotting, we used the secondary antibodies goat anti-rabbit or anti-mouse IgG (H+L) IRDye-800CW-conjugated (926-32211, Li-Cor) antibodies, and for flow cytometry, goat anti-mouse IgG (H+L) Alexa-Fluor-488-conjugated antibodies were used.
siRNA knockdown assays
Dendritic cells were transfected with siRNA (all ThermoFisher) against gp91phox (5ʹ-CCGAGUCAAUAAUUCUGAUCCUUAU-3ʹ), syntaxin-7 (mix of 3 siRNAs: 5ʹ-GCAGCUGUCAAGGGCAGCAGAUUAU-3ʹ, 5ʹ-GGAGUUGCGAUUAUCAGUCUCAUCA-3ʹ, 5ʹ-GAGAAUCUUCUAUCAGGCAACUUGA-3ʹ), SNAP23 (mix of 3 siRNAs: 5ʹ-GACACCAACAGAGAUCGUAUUGAUA-3ʹ, 5ʹ-GGAUAAUCUGUCAUCAGAAGAAAUU-3ʹ, 5ʹ-CAUAGGCAAUGAGAUUGAUGCUCAA-3ʹ), VAMP8 (mix of 3 siRNAs: 5ʹ-GAGGAAAUGAUCGUGUGCGGAACCU-3ʹ, 5ʹ-GAGGUGGAGGGAGUUAAGAAUAUUA-3ʹ, 5ʹ-CGACAUCGCAGAAGGUGGCUCGAAA-3ʹ) or syntaxin-12 (mix of 3 siRNAs: 5ʹ-GCAACAGUUACAACACUCCACAAAU-3ʹ, 5ʹ-UCACUGAGCAGGAUUUGGAACUUAU-3ʹ, 5ʹ-ACAGUUACAGCGAGCUGCUUACUAU-3ʹ) or with irrelevant ON-TARGET plus Non-Targeting (NT) siRNA#1 (Dharmacon), as described previously (Dingjan et al., 2016).
Knockdown efficiency was analyzed by SDS-PAGE (20 µg/lane; 12% acrylamide). Proteins were transferred to PVDF membranes, blocked with 3% milk and 1% BSA in PBS (137 mM NaCl, 2.7 mM KCl, pH 7.4) and incubated with primary antibodies (1:500 dilution). Protein expression was visualized using IRDye-800- (Li-Cor) labeled secondary antibody (1:5000 dilution). Western blots were imaged using the Odyssey CLx Infrared Imaging System (Li-Cor), and processed with ImageStudio (Li-Cor).
Microscopy sample preparation
Alexa-Fluor-633-labeled zymosan particles were produced by rehydrating 10 mg zymosan (Z4250, Sigma) in 900 µl of 0.2 M Na2CO3/NaHCO3 pH 9.0. 50 µl of this suspension was subsequently added to a vial containing 220 nmol Alexa-Fluor-633 C5-maleimide (A-20342, ThermoFisher). Unbound dye was removed by vigorous washing with PBS.
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 serum-free RPMI medium with 1% AA and 2 mM UltraGlutamine, and incubated at 37°C under 5% CO2. 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% CO2. 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 gp91phox
Cells on coverslips were incubated with or without anti-gp91phox 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% CO2. Subsequently, the cells were fixed, blocked and permeabilized as described above. Samples without anti-gp91phox preincubation were stained with anti-gp91phox 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.
For colocalization experiments with phosphoinositides, cells were transfected with plasmids (10 pg/cell) encoding the PH domain of TAPP2, the PX domain of p40phox 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.
Measurement of phagosomal ROS production
Amine-reactive OxyBURST-Green H2DCFDA (3 mg; D-2935, ThermoFisher) was incubated with 10 mg zymosan for 1 h at room temperature. Subsequently, OxyBURST was activated with 1.5 M hydroxylamine. Free dye and hydroxylamine were removed with a Nanosep 300K Omega filter tube (OD300C34, Pall corporation). Cells on coverslips were cultured with OxyBURST-zymosan (4 particles/cell) for 1 h at 37°C under 5% CO2, fixed with 4% PFA and mounted as described above.
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.
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 phosphate-buffered 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-gp91phox 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 goat-anti-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).
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).
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).
I.D., M.t.B. and G.v.d.B. designed the experiments and wrote the paper. P.T.A.L. and L.v.d.B. performed colocalization experiments. M.V.B. contributed to the phosphoinositide experiments. P.H. purified SNAREs. M.t.B. performed the immunoprecipitation. I.D. performed all other experiments. All authors contributed to writing of the manuscript.
G.v.d.B. is funded by a Hypatia fellowship from the Radboud University Medical Center (Radboud Universitair Medisch Centrum) and is the recipient of a Starting Grant from the European Research Council (ERC) under the European Union's Seventh Framework Programme (grant agreement number 336479), a Career Development Award from the Human Frontier Science Program and a grant from the Netherlands Organization for Scientific Research (Nederlandse Organisatie voor Wetenschappelijk Onderzoek; NWO ) (NWO-ALW VIDI 864.14.001). This work was supported by the NWO Gravitation Programme 2013 (ICI-024.002.009). Deposited in PMC for immediate release.
The authors declare no competing or financial interests.