PtdIns3P and Rac direct the assembly of the NADPH oxidase on a novel, pre-phagosomal compartment during FcR-mediated phagocytosis in primary mouse neutrophils

Neutrophils are a critical component of our innate immune system, primarily responsible for combating bacterial and fungal infections.¹  One of the key weapons used by neutrophils to kill pathogens is the formation of reactive oxygen species (ROS) by the nicotinamide adenine dinucleotide phosphate (NADPH) oxidase.²  The active neutrophil NADPH oxidase complex is composed of at least 6 proteins: gp91^phox, p22^phox, p40^phox, p67^phox, p47^phox, and Rac1/2.^3,4  The genes encoding these proteins can be mutated in an immune deficiency disorder called chronic granulomatous disorder.⁵  gp91^phox, and p22^phox form a tightly bound, intrinsic membrane protein called cytochrome b₅₅₈, which is predominantly found in neutrophil granules and the plasma membrane.⁶  In resting neutrophils, p40^phox, p67^phox, and p47^phox are cytosolic and GDP-Rac1/2 is partitioned between a cytosolic complex with RhoGDI and neutrophil membranes.^7-9  Activation of neutrophil cell-surface receptors triggers a signaling network that coordinates assembly of soluble phox proteins around cytochrome b₅₅₈ and GTP-Rac1/2 in a specified membrane location, to form an active enzyme complex.^3,4,10,11  This active complex catalyses the electrogenic transfer of electrons from NADPH on one side of the membrane, to molecular oxygen on the other side, to generate the superoxide anion, which is then converted to other ROS depending on the location and context of activation.

A substantial body of work has begun to define important components of the regulatory network governing oxidase assembly and activation but several features are still unclear. Most work indicates that the core contacts that stimulate NADPH oxidase activity are between cytochrome b₅₅₈, p67^phox, and GTP-Rac1/2.^4,12,13  The delivery of GTP-Rac1/2 to sites of oxidase assembly is catalyzed by guanine nucleotide exchange factors (GEFs), the precise family member being dependent on the type of receptor involved.¹⁴  The delivery of p67^phox is thought to be driven by the combined actions of GTP-Rac1/2, p40^phox and p47^phox.⁴  Recombinant p40^phox and p67^phox bind tightly to one another, and these 2 proteins co-exist as a complex in neutrophil cytosol.^9,15  The phospholipid PtdIns3P is generated in phagosomal membranes^16,17  and binds with high specificity and affinity to the PX domain of p40^phox,^15,18,19  and this interaction is important for p67^phox activity on both mouse and human phagosomes.^20,21  p40^phox/p67^phox also readily binds recombinant p47^phox to form a heterotrimer,⁴  but the extent to which p47^phox exists in a soluble complex with p40^phox/p67^phox in unstimulated neutrophils, or the extent to which this trimer only forms on stimulation is debated.^4,9,15  It is generally agreed, however, that protein kinase C-mediated phosphorylation of p47^phox C terminus relieves an autoinhibitory constraint that allows binding of p47^phox to p22^phox and that, through promoting association of p67^phox with cytochrome b₅₅₈, this is an important trigger in oxidase assembly.⁴  Several further phosphorylation events on phox proteins have also been identified, but their precise roles in the activation process are still uncertain.¹¹ 

Many different receptors and signal transduction systems activate the neutrophil NADPH oxidase, but there are still substantial gaps in our knowledge of how they connect to the GEF, protein kinase, and lipid messenger generating pathways coordinating oxidase activation described in the previous two paragraphs. Such receptors include protein tyrosine kinase-coupled receptors for opsonins, which support phagocytosis of particulate stimuli (eg, β₂-integrin receptors for the complement fragment iC3b or low affinity Fc-receptors for immunoglobulin G (IgG) [FcγRs]).²²  In this context, the oxidase is generally perceived to deliver ROS into the phagosome lumen, where it participates in microbe destruction. However, β₂-integrins and FcγRs can also stimulate extracellular ROS formation in response to neutrophil adhesion and spreading on extracellular matrix²³  or immobilized immune complexes,²⁴  respectively. Extracellular ROS formation is also stimulated by Gi-coupled receptors for chemoattractants (eg, bacterial peptide fMLP).²⁵  Further, in most physiological settings, many of these different receptors operate in the context of costimulation by each other and other receptors for chemokines, cytokines, inflammatory lipids, and complex pathogen ligands that are also found at the inflammatory site. In particular, several cytokines are poor activators of the oxidase in isolation but can dramatically potentiate activation by other ligands (eg, granulocyte macrophage colony-stimulating factor or tumor necrosis factorα).²⁶ 

Among the major obstacles to achieving a more detailed understanding of the pathways linking oxidase activation to different receptor systems in neutrophils are their short lifespan in vitro and refractory nature to standard protein expression protocols. We have tried to overcome some of these obstacles using viral transduction of mouse bone marrow progenitors and repopulation of the hematopoietic cell compartment in radiation chimeras to express several different fluorescent protein reporters of oxidase assembly in fully differentiated neutrophils. We have then used live and fixed cell fluorescence imaging in the context of different genetically engineered backgrounds to investigate the mechanism and spatiotemporal kinetics of oxidase assembly in response to several different agonists, with a particular focus on the role of Rac1/2 and PtdIns3P in directing the localization and function of p40^phox/p67^phox.

p40^phox−/−,²⁷  p40^{phoxR58A/R58A},²⁸  Rab27a^−/− mice,²⁹  and mice permitting conditional Rac1 knockdown on a wild-type (WT) and Rac2^−/− background,³⁰  were described previously. Full details can be found in the supplemental Methods (available on the Blood Web site; see the Supplemental Materials link at the top of the online article).

Human and mouse neutrophils were isolated as described previously.²⁶  Nitroblue tetrazolium (NBT) assays were essentially as previously described,³¹  with full details described in supplemental Methods. All peripheral blood was drawn from healthy volunteers with written informed consent in accordance with the Declaration of Helsinki and Cambridge Research Ethics Committee approval.

For chemiluminescence ROS assays, neutrophils were preincubated with wortmannin or dimethyl sulfoxide vehicle (0.1%) for 10 minutes before stimulation. Rate kinetics of intracellular and extracellular³²  and total ROS²⁸  production were measured using a luminol-based assay in 96-well plates (Berthold Technologies) essentially as described previously.^28,32  Briefly, 5 × 10⁵ cells were incubated with either: luminol (150μM)/superoxide dismutase (SOD; 375 U/mL; intracellular); isoluminol (150μM)/horseradish peroxidase (HRP; 18.75 U/mL; extracellular) or luminol/HRP (total) for 10 minutes at 37°C. Cells were then added manually to immobilized immune complexes or particles (final ratio 1:20 Staphylococcus aureus, 1:5 zymosan, 1:4 beads, or 20 μL of IgG-opsonized–sheep red blood cells [SRBCs]) prepared as described in the supplemental Methods, and measurement started immediately. Light emission was recorded by a Berthold Mircolumat Plus luminomter (Berthold Technologies). Data output is relative light units per second (RLU/s) or total RLU integrated over 20 minutes.

Where indicated, neutrophils were incubated with 80μM dynasore (for 10 minutes at 37°C) before addition of phagocytic stimuli. Neutrophils (5 × 10⁴) were incubated in suspension at 37°C with either 1 × 10⁶ serum-opsonized S aureus (for 7 minutes), 2.5 × 10⁵ zymosan (for 15 minutes), 8 μL of IgG-SRBCs (for 12 minutes), or 5 × 10⁴ beads (for 20 minutes), prepared as described in the supplemental Methods. For S aureus assays, samples were cytospun onto glass coverslips. For all other conditions, samples were allowed to adhere onto glass coverslips for 3 minutes at 37°C, with nonphagocytosed SRBCs hypotonically lysed beforehand. Cells were fixed in 4% paraformaldehyde and processed as previously described.³³  Fixed cells were stained with antibodies against p67^phox (Millipore); and CD14, EEA1, neutrophil elastase, or lactoferrin (Santa Cruz Biotechnology), as indicated for 1 hour at room temperature, followed by appropriate AlexaFluor488/568 secondary antibodies (Invitrogen). Coverslips were mounted on glass microslides with Aqua-Polymount antifading solution (Poly-Science), and fluorescence visualized by a Zeiss LSM510 META point-scanning confocal microscope with integrated camera, using a Plan/Apochromat 63×/1.4 oil objective. Cytosolic p67^phox levels and phagosomal accumulation was quantitated using LSM 510 Image browser software Version 4.2.

Fluorescent fusion protein reporters were expressed in fully differentiated mouse neutrophils using viral transduction of mouse bone marrow progenitors and repopulation of the hematopoietic cell compartment in irradiated recipient mice. Fetal liver (FL) cells were harvested from E14.5-day-old embryos, cultured, and transduced in the presence of cell-specific cytokine cocktails as described in the supplemental Methods. Seventeen hours after transduction, cells were washed in Hanks balanced salt solution, resuspended in phosphate-buffered saline/10% fetal bovine serum (FBS) and 1-2 × 10⁶ FL cells injected (300 μL) into 8-week-old recipient mice, which had been lethally irradiated (using 2 doses of 5 Gy from a ¹³⁷Cs source, separated by 3 hours) the day before injection. A small proportion of cells were kept in culture for flow cytometric analysis of green fluorescent protein (GFP) expression 72 hours after transduction.

Neutrophils were isolated and primed (or left unprimed) as described and kept on ice until use. Neutrophils and IgG-SRBCs (∼ 1:20 ratio) were briefly centrifuged (for 1 minute at 200 × g at 4°C), gently mixed, and immediately added to imaging chambers containing prewarmed phosphate-buffered saline ++ on FBS-blocked coverslips. Serum-opsonized S aureus were added at a similar ratio; however, neutrophils were added to either FBS-blocked or noncoated borosilicate glass coverslips in imaging chambers in a volume of 1.5 mL, allowed to settle for 5 minutes at 37°C and S aureus added in a volume of 500 μL. Detailed live imaging protocols can be found in the supplemental Methods. For widefield imaging, GFP- and differential interference contrast (DIC)–images were recorded at an exposure time of 750 milliseconds, and 4-sesond intervals. The length of each video was 15 minutes. Images were recorded using an Olympus CellR epifluorescence microscope equipped with an Olympus 1X2-UCB camera, using a Plan/Apochromat 60×/1.4 oil objective. Quantification of mean fluorescence intensity (MFI) around the cup and/or phagosome was performed using ImageJ software Version 1.38x, as described in the supplemental Methods.

For 3-dimensional (3D) live imaging, a series of z-stacks were recorded at 0.5 μm intervals. To encompass the whole cell, a set of 20-22 z-stacks was acquired for each time frame. Typically, exposure was 50-100 milliseconds depending on intensity of intracellular fluorescence. The time-interval between each frame was minimum 2.78 seconds. Images were acquired on a spinning Nipkow disk Nikon-Eclipse confocal microscope equipped with an ultrasensitive EM-CCD camera (Andor iXon-897) allowing very fast image aquisition, using a Plan/Apo 100× oil objective (NA1.4). 3D reconstruction was performed using Volocity software Version 5.4.1. AVI videos of 3D-reconstructed, unprocessed z-stacks were created using Volocity and ImageJ software.

We first visualized subcellular sites of oxidase activity in primary mouse neutrophils using an assay based on O₂⁻-mediated conversion of NBT to insoluble formazan salts. Phagocytosis of serum-opsonized S aureus induced substantial formazan deposition in both cytokine-primed (Figure 1ii) and unprimed (supplemental Figure 1) cells. These formazan deposits were always tightly localized to S aureus-containing phagosomes with no significant deposition seen in other subcellular locations. In contrast, phagocytosis of IgG-SRBCs elicited formazan deposition in both the phagosome itself and in adjacent puncta (primed cells are shown in Figure 1iii; no significant oxidase responses were seen in response to IgG-targets in unprimed cells; supplemental Figure 1). A similar pattern of formazan deposition was seen in response to phagocytosis of IgG-latex beads but not unopsonized beads (Figure 1v-vi). Further, extra-phagosomal formazan deposits were seen during phagocytosis of IgG-zymosan but not serum-opsonized–zymosan, despite similar overall levels of oxidase activation (Figure 1vii-viii). A large number of small, formazan deposits were also induced in response to immobilized IgG-bovine serum albumin (BSA) immune complexes (ie, in the absence of phagocytosis; Figure 1iv).

ROS responses to serum-S aureus are mediated by β2-integrin recognition of complement fragments,²⁸  those to non-IgG–opsonized zymosan are dependent on the Dectin-1 receptor³⁴  and those to IgG-beads, IgG-SRBCs, and IgG-BSA are totally dependent on the Fcγ-chain (reference³⁵  and data not shown; DNS). Therefore, these results suggest the oxidase can assemble on intracellular, non-phagosomal membranes specifically in response to IgG-FcγR stimulation.

We initially visualized membrane accumulation of endogenous oxidase subunits using immunofluorescence in fixed mouse neutrophils. In response to phagocytosis of S aureus, p67^phox (Figure 2Ai), p47^phox (DNS), and p40^phox (DNS) clearly accumulated around the perimeter of phagosomes. These accumulations were typically heterogeneous but no extra-phagosomal concentrations were observed. In response to phagocytosis of IgG-beads, endogenous p67^phox accumulated on the phagosome itself and also on membrane structures in close juxtaposition to the phagosome, and these were similar in size, shape, and position to the equivalent formazan depostions described above (Figure 2Aii).

To measure assembly kinetics of oxidase subunits in live cells, we generated mouse bone marrow chimeras heterologously expressing GFP-tagged versions of p40^phox (GFP-p40^phox) in a p40^phox−/− background, or p67^phox (p67^phox-GFP) in a WT background (no p67^phox−/− mice are available; supplemental Figure 2). An N-terminal GFP-tag for p40^phox and a C-terminal GFP-tag for p67^phox were chosen on the basis that these constructs support significant oxidase activity in p40^phox-deficient neutrophils (DNS) and p67^phox-deficient K562 cells.³⁶  GFP-p40^phox was expressed at levels equivalent to endogenous p40^phox in WT cells, and p67^phox-GFP was expressed at levels similar to endogenous p67^phox (supplemental Figure 2). Neutrophils were isolated from these mice, and movements of GFP-tagged proteins during phagocytosis of serum-S aureus and IgG-SRBCs analyzed by wide-field and confocal, fluorescence imaging. Examples of 3D video sequences assembled using a high speed, z-stacking confocal microscope describing the membrane accumulation of p67^phox-GFP during phagocytosis of serum-S aureus or IgG-SRBCs are shown in supplemental Videos 1 and 2, respectively, with selected images from these videos shown in Figure 2B and C. Examples of wide-field imaging of p67^phox-GFP during phagocytosis of serum-S aureus or IgG-SRBCs are also shown in Figures 3 and 4, respectively; examples of GFP-p40^phox images are shown in supplemental Figure 3.

Phagocytosis of serum-S aureus in primed neutrophils was extremely rapid (t ≤ 6 seconds from first point of neutrophil contact), and hence phagocytic cups were poorly discerned, usually defined by only a single frame or missed entirely. Within these limitations, p67^phox-GFP was found in the cytosol of resting neutrophils and typically accumulated on phagosomes within the first 3 frames that they were observed within the cell (4-16 seconds). Phagosomal accumulation of p67^phox-GFP was unevenly distributed around the circumference of individual phagosomes, consistent with images acquired by immunofluorescence of endogenous phox proteins. Further, retention of p67^phox-GFP on individual phagosomes varied substantially (78-150 seconds). Similar patterns of p67^phox-GFP accumulation were seen in unprimed neutrophils (an example is shown in supplemental Video 3), although the time between particle engulfment and phagosomal accumulation of p67^phox-GFP was highly variable (4-75 seconds) but on average slower than in primed cells (mean 26 seconds). We never observed any membrane accumulation of p67^phox–GFP or GFP-p40^phox, in primed or unprimed cells, that did not appear to be directly associated with the S aureus phagosome.

Phagocytosis of IgG-SRBCs was characteristically slower than for serum-S aureus (t ≥ 10 seconds) and clear phases of initial contact, emergence of a phagocytic cup, and phagosome closure could be defined. p67^phox-GFP first accumulated on a large tubulovesicular structure at the base of the emerging phagocytic cup (marked by an arrow in Figure 2Ci). This structure appeared to grow in size and complexity during cup development and, in a number of examples, appeared to fuse with the developed phagosome (Figure 2Ciii). p67^phox-GFP also accumulated on the phagosome itself but these accumulations were highly variable (lasting between 48-260 seconds) and were typically less intense than on the pre-phagosomal structure.

Previous studies have defined a role for PtdIns3P-p40^phox interaction in phagocytosis-elicited oxidase activation.^20,21  We sought to establish the relative kinetics of oxidase subunit assembly and PtdIns3P formation during neutrophil phagocytosis by generating bone marrow chimeras expressing GFP-iPX in a WT genetic background. GFP-iPX was found in both neutrophil cytosol and on vesicular compartments that colocalized with the endosomal marker EEA1 (DNS). During phagocytosis of S aureus, GFP-iPX accumulated on phagosomes soon after they separated from the plasma membrane (supplemental Video 4; Figure 2Bii), consistent with previous studies using cultured cells.^16,17  We sought to more precisely define the relative kinetics of PtdIns3P and oxidase subunit accumulation during phagocytosis by co-expressing p67^phox-GFP and mCherry-iPX in the same neutrophils. We obtained mouse bone marrow chimeras with good levels of co-expression of these reporters but systematically observed that cells with measurable levels of mCherry-iPX expression exhibited poor membrane accumulations of p67^phox-GFP (supplemental Figure 4A) and poor formazan deposition in NBT assays (DNS), suggesting the iPX-reporter sequesters PtdIns3P in structures relevant to oxidase assembly. However, in a small number of examples we did manage to image both mCherry-iPX and p67^phox-GFP accumulation during S aureus phagocytosis; in these cases p67^phox-GFP accumulation slightly preceded that of mCherry-iPX (supplemental Video 5).

GFP-iPX also accumulated dramatically on IgG-SRBC phagosomes (supplemental Video 6; Figure 2Cii), but whether it also accumulated on the same pre-phagosomal structures as the oxidase was less clear. Pre-existing endosomal accumulations of this reporter made it very difficult to identify the appearance of relatively small structures and, to an even greater extent than was evident during S aureus phagocytosis, we were unable to localize both iPX and tagged oxidase subunits in the same cell (because expression of iPX severely blocked oxidase activity and membrane accumulation of p67^phox-GFP; supplemental Figure 4B and DNS). However, the emergence of a GFP-iPX–decorated structure with the characteristic timing, position, and morphology of analogous accumulations of p67^phox-GFP could usually be discerned, before extensive accumulation on the phagosome itself (marked by arrows in Figure 2Cii and supplemental Video 6).

We investigated the role of PtdIns3P binding to p40^phox in regulating the assembly of an active oxidase during phagocytosis of S aureus using primed (Figure 3) and unprimed (supplemental Figure 1) neutrophils isolated from mice carrying a mutation in the PX domain of p40^phox that prevents interaction with PtdIns3P (p40^{phoxR58A/R58A}).²⁸  Priming agents themselves did not stimulate ROS production, and the R58A mutation caused a similar, 30%-40% decrease in ROS formation in both primed (Figure 3A) and unprimed cells (supplemental Figure 1). Further, phagosomal accumulation of endogenous p67^phox (Figure 3B) and of exogenously expressed p67^phox-GFP (Figure 3C) were all similarly reduced by approximately 30% in primed p40^{phoxR58A/R58A} neutrophils compared with WT cells.

We also investigated the role of Rac binding to p67^phox during S aureus phagocytosis using neutrophils isolated from mice lacking Rac2 (Rac2KO) and/or supporting inducible deletion of Rac1 (Rac1KD). Loss of Rac2 or Rac1 alone did not reduce ROS formation or phagosomal accumulation of p67^phox in response to S aureus and, indeed, loss of Rac1 characteristically generated increased responses (Figure 3A-B). Loss of both Rac2 and Rac1 however, prevented measurable ROS formation or p67^phox translocation, indicating Rac1 and 2 play an essential but redundant role in oxidase activation in this context (phagocytosis of serum-S aureus was unaffected by loss of both Rac1 and 2; supplemental Figure 5). Over 80% deletion of Rac1 in a Rac2^−/− background was required to see a significant reduction in S aureus oxidase responses (supplemental Figure 5), indicating only a small proportion of the neutrophil's total complement of Rac proteins is required to fully support this response. An absolute requirement for direct Rac binding to p67^phox was also confirmed by the inability of p67^phox-H69E-GFP, a mutant incapable of binding Rac1/2 (supplemental Figure 6), to translocate to S aureus phagosomes (Figure 3C; expression of p67^phox-H69E-GFP was comparable to p67^phox-GFP; supplemental Figure 2).

IgG-SRBC–stimulated ROS formation assessed by either NBT reduction (Figure 4A) or chemiluminescence (Figure 4B) was reduced by ≥ 85% in p40^{phoxR58A/R58A} neutrophils. Membrane accumulation of p67^phox-GFP in response to IgG-SRBCs was also severely reduced in p40^{phoxR58A/R58A} neutrophils, with no discernable concentration on either pre-phagosomal compartments or the phagosome itself (Figure 4C).

ROS formation in response to IgG-SRBCs was significantly increased by loss of Rac1 (Figure 4B) but completely blocked by loss of Rac2 (Figure 4A-B). Further, p67^phox-H69E-GFP was unable to translocate to any membrane structure during IgG-SRBC phagocytosis (Figure 4C), indicating Rac2 binding to p67^phox is essential for oxidase assembly and activation under these conditions.

FcγRs are known to undergo rapid endocytosis upon antibody engagement³⁵  and therefore we sought to establish if receptor endocytosis was required for pre-phagosomal accumulation of the oxidase in response to IgG targets. The dynamin inhibitor dynasore³⁷  blocked FcR-mediated endocytosis of transferrin (supplemental Figure 7) and soluble antibody (DNS), but did not significantly block intracellular formazan deposition in response to IgG-beads (supplemental Figure 7), IgG-SRBCs (supplemental Figure 7), or immobilized-IgG-BSA (DNS). Further, p67^phox accumulation on extra-phagosomal structures was unaffected by dynasore in response to phagocytosis of IgG-beads (supplemental Figure 7). Together, these results indicate FcγR-endocytosis is unlikely to be a trigger for oxidase assembly under these conditions.

We attempted to define the source(s) of pre-phagosomal membrane on which the oxidase assembles during FcR-mediated phagocytosis by searching for colocalization with known markers of neutrophil membrane compartments. IgG-beads were chosen for most of these studies because, unlike rabbit IgG-opsonized SRBCs, they could be opsonized with mouse antibodies that were not recognized by the anti–rabbit secondary antibodies used for immunofluorescence localization of endogenous p67^phox and granule markers.³⁸  We did not observe good colocalization of p67^phox-positive extra-phagosomal structures with CD14 (secretory vesicles) or EEA1 (early endosomes). We did, however, observe a small degree of colocalization of p67^phox with lactoferrin (specific granules) and elastase (azurophil granules; Figure 5). Our interpretation of these results is that the membrane compartment on which p67^phox assembles is probably derived, in part, from granule fusion but may contain a mixture of active ROS/proteases that reduce the sensitivity of epitope detection.

Rab27a has previously been implicated in neutrophil granule secretion^39,40  and in targeting the oxidase to phagosomes in dendritic cells.⁴¹  We therefore measured ROS responses to serum-S aureus and IgG-SRBCs in neutrophils lacking Rab27a (Fig 6). Using an enhanced chemiluminescence assay designed to distinguish between intra- and extracellular ROS formation, we found that loss of Rab27a caused significant decreases in intracellular but not extracellular ROS responses to IgG-SRBCs, while both intracellular and extracellular ROS responses to S aureus were unaffected (Figure 6A). Similarly, intracellular formazan deposition in response to S aureus was unaffected (Figure 6B). In contrast, formazan deposition in response to IgG-SRBCs, normally restricted to the base of the phagosome and the phagosome itself, was much more extensively distributed throughout the cell in Rab27a^−/− neutrophils (Figure 6B). Therefore, we postulate that Rab27a directs the assembly of pre-phagosomal structures and/or their delivery to phagosomes containing IgG-opsonized targets.

We also investigated whether the oxidase assembles on extra-phagosomal structures during serum-S aureus and IgG-target phagocytosis in human neutrophils. In contrast to our observations in mouse neutrophils, we observed a small number of formazan-positive intracellular structures in human cells that were not exposed to particles and a significant increase in extra-phagosomal formazan deposits on phagocytosis of both serum-opsonized S aureus (Figure 7A, marked by arrows) and zymosan (DNS). Intriguingly, these extra-phagosomal formazan deposits were abolished by depleting the human opsonizing serum of IgG (Figure 7A), and this coincided with a specific reduction in intracellular, but not extracellular, ROS formation (Figure 7B; zymosan DNS; IgG depletion did not significantly reduce levels of phagocytosis; Figure 7Aii). We have previously shown mouse neutrophil phagocytosis and ROS formation in response to mouse serum-opsonized S aureus are independent of FcγR-signaling²⁸  and, consistent with this, we saw no effect of first depleting mouse serum of IgG before opsonizing S aureus or zymosan in these assays (DNS). We also observed substantial extra-phagosomal formazan deposits in human neutrophils phagocytosing IgG-SRBCs and IgG-beads (Figure 7Ci-ii), and endogenous p67^phox also accumulated in extra-phagosomal compartments in response to IgG-beads (DNS). Taken together these results indicate human neutrophils also assemble active oxidase complexes in extra-phagosomal structures specifically in response to FcγR-stimulation.

Formazan-deposition assays, immunofluorescence of endogenous oxidase subunits, and live fluorescent imaging of GFP-tagged subunits, all point to the rapid and direct assembly of the oxidase complex on S aureus-containing phagosomes in mature mouse neutrophils. This would be expected on the basis of the well-described function of the oxidase in bacterial phagosomes to participate in bacterial killing.²  Surprisingly however, the same approaches indicated the oxidase can assemble on non-phagosomal structures in response to IgG-targets and, in the case of FcγR-mediated phagocytosis, these structures develop at the base of the forming phagosome at sites of granule delivery, inviting questions concerning their mechanism of formation and function.

Using conditional deletion of Rac1 in a Rac2^+/+ or Rac2^−/− genetic background, we found that Rac2 and/or Rac1 were required for measurable ROS formation in response to both serum-S aureus and IgG-SRBCs. This is consistent with the view that GTP-Rac1/2 forms the critical complex with p67^phox and cytochrome b₅₅₈ that allows efficient electron transfer between NADPH and molecular oxygen. In vitro assays with recombinant proteins suggest both Rac1 and 2 isoforms can participate in active oxidase assembly but most previous work had suggested that Rac2 plays the major role in neutrophil oxidase activation.^42,43  While our results confirm an absolute requirement for Rac2 in oxidase responses to IgG-SRBCs, they revealed a surprising redundancy between Rac1 and 2 in the ROS response to serum-S aureus, supporting previous findings that Rac2 is not essential for serum-zymosan ROS responses.⁴⁴  Further, the effects of Rac 1 and/or Rac2 deletion on oxidase activation were paralleled by their effects on membrane accumulation of p67^phox, suggesting the interaction between these proteins is also important in determining the overall affinity of p67^phox for the relevant membrane/protein location. The molecular basis for this difference in isoform usage is unknown, but Rac2 has been shown to have a preference for less charged membrane surfaces.⁴⁵ 

We also investigated the role of PtdIns3P binding to the PX domain of p40^phox in the membrane recruitment and activation of the oxidase during phagocytosis of both serum-S aureus and IgG-SRBCs. We confirmed previous studies conducted in mouse neutrophils indicating a partial role for the PtdIns3P-p40^phox interaction in determining levels of ROS formation in response to phagocytosis of S aureus.²⁰  We have now extended these studies to show a more major role for this interaction in ROS responses to IgG-SRBCs. We also show a close correlation in both responses between the impacts of PtdIns3P-p40^phox interaction on ROS formation and membrane accumulation of p67^phox. These observations were also consistent with the effects of heterologous expression of a PtdIns3P-binding GFP-iPX domain, which severely reduced oxidase activation and p67^phox membrane accumulation, particularly in response to IgG-targets. We also note that the relatively modest decreases in ROS responses to serum-S aureus seen in p40^{phoxR58A/R58A} mouse neutrophils were much less than equivalent inhibitions reported in analogously mutated p40^{phoxR105Q/−} human neutrophils,²¹  possibly reflecting intrinsic differences in the relative balance of pathways regulating oxidase assembly in mouse versus human neutrophils. However, we also found that, unlike the mouse, human neutrophil ROS responses to phagocytosis of serum-S aureus possessed a significant IgG-component and were associated with more numerous, IgG-dependent, extra-phagosomal formazan deposits. It is therefore possible that the greater dependence of human neutrophil ROS responses on PtdIns3P-interaction with p40^phox is because of the greater involvement of FcγR-driven oxidase assembly in the assays conducted.

Previous studies in mouse neutrophils, together with work presented here, indicates that the PtdIns3P-p40^phox interaction makes a differential contribution to ROS responses elicited by different agonists, having little impact on extracellular ROS responses to fMLP, phorbol 12-myristate 13-acetate (PMA),²⁰  and either serum or IgG-opsonized zymosan (DNS), a partial impact on intracellular ROS responses to fMLP (DNS), PMA and S aureus²⁰ , and a major impact on ROS responses to IgG-targets. Thus, it appears that PtdIns3P binding to p40^phox is not a core requirement for neutrophil oxidase activity but a regulatory input guiding oxidase assembly in only certain contexts of activation. The simplest hypothesis is that among the balance of mutual interactions that govern the assembly of oxidase subunits, PtdIns3P binding to p40^phox makes a significant contribution to driving increased local concentration of p67^phox on some PtdIns3P-containing membranes, presumably dictated by the concentrations and affinities of all the relevant interactions. Consistent with this, we observed that loss of Rac2 results in ROS responses to S aureus becoming much more sensitive to the PI3K inhibitor wortmannin (supplemental Figure 8). Thus it is possible that when less Rac is available, p67^phox translocation becomes more dependent on PtdIns3P-binding to p40^phox. Of relevance to this idea are previous studies showing that phagocytic/endocytic structures accumulate PtdIns3P and lose Rac proteins on severance from the plasma membrane,⁴⁶  thus sustained oxidase activity on these structures may increasingly depend on the PtdIns3P-p40^phox interaction. Indeed a recent study concluded that in PLB-985 cells, PtdIns3P played a more important role in retaining p40^phox/p67^phox on the phagosome than initial recruitment to the cup and phagosome.³²  This idea is also consistent with our observation that p67^phox-GFP accumulates on S aureus phagosomes before GFP-iPX. However, Tian and colleagues also concluded that p40^phox was more important in oxidase activation than recruitment.³²  We cannot rule out a role for p40^phox in oxidase activation beyond recruitment of p67^phox, it is simply that our data do not demand it.

We attempted to identify the origin of the pre-phagosomal membranes on which the oxidase assembles in response to IgG-target phagocytosis. Due to their size, shape, granularity, and relative nuclear to cytoplasmic ratios, fully differentiated neutrophils present a challenge to accurately assess the colocalization of markers for different membrane compartments by standard immunofluorescence techniques. We were able to show significant colocalization between markers for specific and azurophil granules (lactoferrin and elastase) and p67^phox on pre-phagosomal structures but a more extensive analysis is likely to require further techniques, such as imaging granule membrane and content markers by electron microscopy and live cell fluorescence. However, the simplest hypothesis that encompasses the morphology and kinetics of both formazan and phox protein accumulation during IgG-SRBC phagocytosis is that the oxidase first assembles on granule membranes, which then fuse with each other, further membranes, and finally the phagosome.

Several previous studies have suggested the oxidase may assemble on neutrophil intracellular membranes and granules in response to PMA.^47-49  Recent work has also identified delivery of cytochrome b₅₅₈-containing vesicles to dendritic cell phagosomes via a Rab27a-dependent mechanism.⁴¹  Further, Rab27a is important in the secretion of myeloperoxidase-enriched azurophil granules in neutrophils.⁴⁰  Consistent with these findings, we observed a partial role for Rab27a in the delivery of ROS and peroxidases to common compartments during IgG-SRBCs, but not S aureus, phagocytosis. However, this effect is not simply due to a failure to deliver myeloperoxidase to IgG-SRBC phagosomes, because loss of Rab27a also created more numerous, extra-phagosomal formazan deposits, suggesting Rab27a is directly involved in oxidase trafficking to IgG-mediated phagosomes.

Clearly further work needs to be done to gain a clearer view of how the oxidase assembles on pre-phagosomal, internal membranes in both mouse and human neutrophils, in particular how relevant intracellular signals, including GTP-Rac and PtdIns3P are formed at these locations and how fusion events regulate their fate. We do not yet understand whether pre-phagosomal oxidase accumulation during IgG-SRBC phagocytosis merely reflects differences in extent and timing of granule fusion versus S aureus phagocytosis or reflects a more profound, functional difference. In this regard, recent studies in dendritic cells and macrophages have uncovered a role for the oxidase in resisting phagosomal acidification and allowing appropriate proteolysis and antigen presentation to occur.^41,50  Antigen cross-presentation via dendritic cell phagosomes is dependent on the nature of the opsonin, and FcγR-directed delivery, but not complement-receptor directed delivery, is Rab27a-dependent.³⁹  Given the IgG and partial Rab27a dependency of oxidase assembly on the extra-phagosomal structures described here, this could suggest a role for these structures in antigen presentation downstream of FcγR signaling. It will therefore be interesting to investigate whether sites of extra-phagosomal oxidase assembly in neutrophils represent compartments with distinct processing capabilities, such as extracting antigenic peptides from antibody-opsonized prey.

We thank Mary Dinauer and Chris Ellson for helpful discussions and for critical reading of the manuscript. We also thank Simon Andrews, John Coadwell, John Ferguson, Thomas Henley, Elena Vigorito, and Keith Boyle for technical assistance and helpful discussions. Finally, we thank the staff of the small animal unit (SAU) and the small animal barrier unit (SABU) at the Babraham Institute for assistance with care of animals.

This work is supported by the Biotechnology and Biological Sciences Research Council (BBSRC), Medical Research Council (MRC; G0600840), and Wellcome Trust (WT085889MA). T.C. is a UCB BBSRC-Collaborative Awards in Science and Engineering (CASE) award student.

Wellcome Trust

Contribution: K.E.A. and T.A.M.C. designed and performed experiments, analyzed and interpreted data, and wrote the manuscript; K.D., R.B.H., and S.W. performed experiments; T.T., K.G., O.R., M.C.S., and V.L.J.T. provided valuable reagents; and L.R.S. and P.T.H. designed experiments, analyzed and interpreted data, and wrote the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Phillip Hawkins, Inositide Laboratory, Babraham Institute, Babraham Research Campus, Cambridge CB2 4AT, United Kingdom; e-mail: phillip.hawkins@bbsrc.ac.uk.