Cytoskeletal remodeling of myeloid cells after their engulfment of PS- versus IgG-coated targets
While the remodeling of the cytoskeleton that facilitates phagocytosis has been well studied, the longer-term consequences of phagocytosis on cell dynamics have been largely neglected. We initiated our studies by evaluating the long-term responses of macrophages to target particles of identical size, shape, and stiffness coated with either 5% PS (complemented with other phospholipids) to trigger efferocytosis8, or with IgG to instigate Fc-mediated phagocytosis. We challenged bone marrow-derived macrophages (BMDM) with the particles and allowed an initial 10 min period of uptake, ensuring a similar number of targets were phagocytosed in both cases. Unbound particles were removed and the macrophages incubated for an additional 30 min before analysis. Scanning electron microscopy revealed clear differences in the morphology of the BMDM after uptake of PS- versus IgG-coated targets (Fig. 1a). BMDM appeared to contract and round up after ingestion of IgG-coated particles. In stark contrast, BMDM that had ingested PS-coated beads retained their spread morphology and broad dorsal ruffles (Fig. 1a). Contraction was quantified by measuring the area of the cells in contact with the coverglass, monitoring the distribution of F-actin by confocal microscopy using fluorescent phalloidin. These estimates confirmed the differential behavior of the cells following ingestion of the two types of beads (Fig. 1b).
More detailed analysis of the distribution of phalloidin-stained cells also revealed that prominent F-actin-dense foci, which we subsequently identified as podosomes (see below), were entirely disassembled in cells that had ingested IgG-coated particles (Fig. 1c-d). Dynamic membrane ruffling, determined measuring the surveillance area of RAW 264.7 cells expressing actin-GFP by video recording (see Video 1, S Fig. 2 for details), was also depressed in cells that had ingested IgG-opsonized targets (Fig. 1e). Membrane ruffles formed by macrophages can fold back on the cell body and fuse, resulting in the formation of macropinosomes that entrap the surrounding fluid medium2. We measured the uptake of labeled 70 kDa dextran, which is internalized preferentially by macropinocytosis, as an alternate means of assessing ruffling activity. Convincingly, we found that dextran uptake decreased considerably in BMDM that had ingested IgG-opsonized targets, but not PS-targets (Fig. 1f-g). Together, these findings imply that Fc-mediated phagocytosis –but not efferocytosis– had a marked impact on actin remodelling and membrane dynamics.
To determine if this feature was shared by other types of phagocytes, we also investigated the effect of IgG-opsonized particle uptake in human polymorphonuclear neutrophils (PMN). These cells display high motility, a process that is dependent on actin. We used video microscopy to track the motility of PMN before and after phagocytosis of IgG-opsonized red blood cells (RBC) (Fig. 1h-i, Video 2). As shown in Fig. 1h and quantified in Fig. 1i, PMN the chemokinesis stimulated by N-formylmethionyl-leucyl-phenylalanine (fMLP) was virtually eliminated following engulfment of IgG-opsonized red blood cells (RBC). Therefore, a mechanism that arrests the surveillance programs upon ingestion of IgG-opsonized targets is conserved in various phagocytes.
Phagocytosis of IgG-opsonized particles causes cortical thickening associated with increased activation of Rho
As described above, phagocytosis of particles via Fcg receptors –but not PS receptors– induces major cytoskeletal rearrangements, a response that is proportional to the number of targets ingested (S Fig. 1). The observed changes included a pronounced thickening of the submembranous F-actin (Fig. 2a-b). Cortical F-actin dynamics are largely controlled by RhoA, which stimulates processive polymerization of actin filaments mediated by formins9. Accordingly, treatment of the macrophages with CN0310, a bacterial toxin-derived activator of Rho caused marked cortical thickening (not illustrated) that resembled the effects of Fcg-mediated phagocytosis. Considering these similarities, we measured biochemically the activation of Rho following phagocytosis of IgG-coated particles. Because some of the agents used to manipulate and assess the activity of Rho interact with the A, B and/or C isoforms, we hereafter refer to their target as Rho(A-C). The effectiveness of the G-LISA assay used was validated using CN03, which induced a ≈ 5-fold stimulation of the GTPase. In good agreement with the microscopic observations of cortical F-actin, Rho(A-C) was activated nearly 3-fold following Fcg receptor-mediated phagocytosis (Fig. 2c).
While documenting its increased activity, the G-LISA assays do not reveal the site(s) where Rho is stimulated. Spatial information was obtained using a genetically-encoded biosensor of active Rho(A-C), namely dTomato-2xrGBD11, that contains the G protein-binding domain (GBD) of rhotekin, a Rho(A-C) effector. The biosensor was preferentially accumulated at the cortex of cells treated with CN03 and at the cortex of macrophages following phagocytosis of multiple IgG-coated particles. The redistribution of the dTomato-2xrGBD probe was prevented by CT04, a Rho(A-C) antagonist derived from Clostridium botulinum C3 transferase, validating the specificity of the probe (Fig. 2d-e). These observations suggested that the activation of Rho(A-C) at or near the plasma membrane contributed to the cortical F-actin thickening described earlier.
The gain in Rho activity associated with phagocytosis was accompanied by a decrease in the activity of Rac12, whether assessed by G-LISA or using the PAK(PBD)-YFP biosensor13 (Fig. 2f-h); the extension of ruffles where PAK(PBD)-YFP accumulates preferentially, was inhibited following phagocytosis of three or more IgG-coated particles. In contrast, we saw no appreciable difference in the overall activity of Cdc42 following phagocytosis (Fig. 2i). To determine if the shift in the balance of GTPase activity towards Rho accounted for the cytoskeletal changes induced by phagocytosis, we quantified F-actin density at the cortex in RAW cells in the presence and absence of Rho(A-C) inhibitors. As in primary BMDM, RAW cells showed a clear increase in cortical F-actin post-phagocytosis of IgG-opsonized RBC to a level comparable to that observed following activation of Rho(A-C) with CN03 (Fig. 2j). Conversely, inhibition of Rho(A-C) with CT04 prevented thickening of the cortex (Fig. 2j).
Rho is anticipated to increase actin polymerization at the cortex by activating the formin family of nucleators that recruit actin monomers to growing linear filaments via their FH2 domains14. RhoA, B, and/or C can activate mDia1, mDia2, Daam1, FMNL2, and FMNL31515, though the relative contribution of individual formins to cortical F-actin is not clear. We therefore opted to broadly inhibit the formin family using the FH2 inhibitor (SMIFH2). As with inhibition of Rho itself, we found that inhibiting the formins also prevented thickening of the cortex in RAW cells that had ingested IgG-opsonized targets (Fig. 2j). Taken together, these results implicate the gain in Rho(A-C) and formin activities in the cortical F-actin remodelling observed in macrophages containing phagosomes taken up via Fc receptors.
The loss of podosomes associated with IgG-opsonized target engulfment is dependent on Rho/formin activation
In addition to cortex thickening, we previously noted that RAW cells that had ingested several IgG-opsonized targets lost the ventral F-actin puncta, reminiscent of podosomes, that are present in untreated cells. Immunostaining for the integrin-associated adaptor vinculin, confirmed that these structures were indeed podosomes (Fig. 3a, ). Like the gain in cortical F-actin, the loss of podosomes was dependent on Rho(A-C) activation: inhibition by CT04 was sufficient to preserve the podosomes in RAW cells following ingestion of IgG-opsonized targets (Fig. 3b). Reciprocally, activating of Rho(A-C) alone, using CN03, eliminated the podosomes in otherwise untreated cells (Fig. 3b,). As was the case for the thickening of cortical F-actin, we found that formins were involved in the loss of podosomes, since inhibition with SMIFH2 prevented their disassembly following phagocytosis (Fig. 3b). Taken together, these data implicated Rho/formins in the loss podosomes that accompanies the gain of cortical F-actin.
Interestingly, when immunostaining for vinculin we noted that loss of podosomes was associated with a gain in focal adhesion-like structures (Fig. 3a). Because the formation and signaling from focal adhesions are dependent on myosin II16, we suspected that the rounding and retraction of the cells following phagocytosis may involve this contractile protein. It is noteworthy that RhoA increases the activity of myosin II through its activation of Rho-associated kinase (ROCK), which phosphorylates myosin light chain. The involvement of these proteins can be tested experimentally, because ROCK and myosin II are potently inhibited by Y-27632 and blebbistatin, respectively. Neither of these inhibitors affected the binding or internalization of phagocytic targets (S Fig. 3). Using these pharmacological agents we found that ROCK and myosin were in fact dispensable for the Rho-mediated loss of podosomes after Fc-mediated phagocytosis (S Fig. 3a-f).,
ERM proteins are necessary to enact the changes in cell morphology associated with the engulfment of IgG-opsonized targets
It is remarkable that phagocytosis, which normally happens at or near the dorsal surface of the cells caused the loss of ventral podosomes. How this remote effect was exerted was not clear. It has been reported that changes in membrane tension can cause podosome disassembly17, and the thickening or cortical actin that follows phagocytosis suggested that a similar process may be involved18. We therefore measured membrane tension by fluorescence lifetime imaging microscopy (FLIM) using the recently described Flipper probe19. As shown in (Fig. 4a), membrane tension indeed increased as a result of phagocytosis, a change that persisted long after particle engulfment was completed (i.e. for at least 30 min). The tension change was mimicked by application of CN03, suggesting that cytoskeletal contraction due to stimulation of Rho(A-C) is responsible in both instances.
We next considered how the compaction of the F-actin cytoskeleton exerted by Rho(A-C) activation is transmitted to the membrane. Such transmission would necessitate firm attachment of linear actin filaments to the membrane. The adaptor proteins ezrin, radixin and/or moesin (ERM) may fulfill such a role. ERM proteins are recruited by phosphatidylinositol-(4, 5)-bisphosphate to the cytosolic leaflet of the plasma membrane, where they can then bind to multiple transmembrane proteins. Phosphorylation of ERM proteins near their C-terminus then facilitates their interaction with F-actin 20,21. We observed an increase in ERM phosphorylation following phagocytosis (Fig. 4c), supporting the idea that these proteins may contribute to the conveyance of tension to the membrane.
The possible role of ERM proteins in membrane tension and podosome disassembly was then assessed directly. Owing to their structural similarities, the three ERM proteins are believed to have overlapping functions22. Because RAW macrophages were found to express all three proteins, we were compelled to generate a triple ERM KO cell line using CRISPR-Cas9 editing. The elimination of the transcripts and resulting proteins was validated by PCR and immunoblotting, respectively (Fig. 4d-e). Moreover, using single-particle tracking, we observed that CD44, a transmembrane protein that is anchored to the actin skeleton via ERM, displays a large increase in mobility in the KO cells23. As in a recent report24, we did not observe gross morphological differences between wildtype and ERM KO cells in normal culture medium. When incubated in minimal medium (HBSS), we did note a fraction of the ERM-KO cells underwent spontaneous blebbing, which was never observed in wildtype cells (Fig. 4f and S Video 3). Membrane blebbing became more pronounced upon activating Rho(A-C) with CN03 (S Fig. 4a-c), supporting the idea that a decreased attachment of the membrane to the cortex increases blebbing when the cytoskeleton contracts 25.
We proceeded to test whether ERM proteins are required for the cytoskeletal changes that follow Fc-mediated phagocytosis. Remarkably, we found that, unlike the wildtype cells, triple-KO ERM cells retained their podosomes after ingesting similar numbers of IgG-opsonized targets, despite undergoing a comparable increase in Rho activity (Fig. 4g-h). We also found that the ERM KO cells continued to ruffle (Fig. 4i, Video 4), performed macropinocytosis (Fig. 4k), and remained more spread (Fig. 4j) after phagocytosis than their wildtype counterparts. These data suggest that the concerted activation and function of Rho, formins, and ERM proteins are necessary to coordinate the phagocytosis-induced changes.
The NADPH oxidase and associated ROS mediate the activation of Rho that renders the cells quiescent after phagocytosis
How does Rho(A-C) become activated downstream of Fc-receptor mediated phagocytosis? We initially considered the possibility that GEF-H1, a major Rho guanine nucleotide exchange factor (GEF) could be released into the cytosol following phagocytosis. GEF-H1 is made inactive by its association with microtubules and disassembly of the latter could account for its release and activation. However, we found that the microtubule cytoskeleton remained intact in the fed macrophages as determined by immunostaining and cytoskeletal fractionation (Fig. 5a-b) suggesting this was not a likely mechanism of Rho activation.
A distinguishing feature between efferocytosis and Fc-mediated phagocytosis is the magnitude of the accompanying respiratory burst, which reflects the activity of the NADPH oxidase26,27. Upon incubating primary BMDM with phagocytic targets in the presence of nitroblue tetrazolium (NBT)28, formazan –a measure of reactive oxygen species (ROS)– was evident only in cells containing IgG-opsonized particles, but not PS-coated ones (Fig. 5c). The inhibitor GSK279503929 prevented the response, confirming that the ROS were produced by NOX2, the predominant NADPH oxidase of phagocytes (Fig. 5d-e). Of note, a number of studies have implicated ROS in regulating the activity of Rho proteins by oxidation of their GEFs and GAPs or of the GTPases themselves30,31. The phagosome, like other membrane bilayers, is permeable to ROS so it seemed possible that oxidation of cytosolic determinants of Rho(A-C) activity could account for the morphological changes associated with Fc-mediated phagocytosis. Supporting this idea, inhibition of NOX2 with GSK2795039 prevented the rounding of the cells upon phagocytosis (Fig. 5g). Moreover, diphenyleneiodonium (DPI), another potent inhibitor of the NADPH oxidase complex, prevented the arrest of chemokinesis in PMN following Fc-mediated phagocytosis (compare Fig. 5i to Fig. 1i, Video 2 to Video 5). Lastly, podosomes persisted after phagocytosis when macrophages were treated with DPI (Fig. 5j-k).
These data suggest a direct connection between the respiratory burst and the activation of Rho. We indeed noted that the targeting of the Rho biosensor to the cortex that follows phagocytosis in otherwise untreated cells was absent when the NADPH oxidase was inhibited (Fig. 5l-m). We also found that the levels of active Rho were not elevated when the NADPH oxidase was inhibited by either DPI or by GSK2795039 (Fig. 5n). Taken together, these data suggest that Fc-receptor triggered ROS production, mediated by NOX2, leads to Rho activation, causing changes in the shape and ability of phagocytes to migrate and survey their environment.
Activation of the NADPH oxidase by pathogens and MAMPs drives podosome disassembly
In the preceding experiments we used non-biological targets to specifically contrast inflammatory vs. non-inflammatory responses. We next wondered how encounters with pathogens or soluble MAMPs would influence the surveillance activity of macrophages. Phagocytosis of pathogens generally triggers a respiratory burst32, as does signaling from TLR433. We first challenged BMDM with heat-killed Candida albicans, an opportunistic pathogen that is bound and phagocytosed by receptors like the mannose receptor and Dectin-1 that recognize components of the fungal cell wall. After 40 min, we fixed the macrophages and subsequently stained for podosomal markers. As we had previously observed with IgG-opsonized targets, we found a remarkable change in the cell shape associated with the loss of podosomes and a concomitant gain in focal adhesion-like structures, illustrated in Fig. 6a and quantified in Fig. 6b. In addition, thickening of the cortical F-actin cytoskeleton was evident (Fig. 6c). Therefore, the observations made using Fc-coated targets were recapitulated when stimulating fungal recognition receptors, which similarly activate the NADPH oxidase.
The oxidase is also made active by LPS. Interestingly, exposure of cells to LPS had been previously shown to cause podosome disassembly34, which we were able to replicate (Fig. 6d-e). While the mechanism of podosome loss in response to LPS is not entirely clear, it had previously been suggested that this could involve proteinases, notably the a-converting enzyme ADAM17. Alternatively, the cytoskeletal changes may be secondary to the activation of NOX2. We therefore investigated the role of i) the NAPDH oxidase and ii) ERM proteins in the disassembly of podosomes associated with TLR4 signaling. We found that inhibiting NOX2 caused the macrophages to retain their podosomes in LPS-treated cells, albeit with dysregulation of their positioning within the cells (Fig. 6d-e). ERM proteins were also entirely necessary for the podosomes to disassemble in response to LPS (Fig. 6d-e).
Taken together, these findings demonstrate that diverse signaling pathways downstream of pattern recognition receptors and Fc-receptors, which converge on the production of ROS via the NADPH oxidase, cause myeloid cells to change their phenotype from a motile to a quiescent state.