Multiple roles of filopodial dynamics in particle capture and phagocytosis and phenotypes of Cdc42 and Myo10 deletion

Macrophage filopodia, finger-like membrane protrusions, were first implicated in phagocytosis more than 100 years ago, but little is still known about the involvement of these actin-dependent structures in particle clearance. Using spinning disk confocal microscopy to image filopodial dynamics in mouse resident Lifeact-EGFP macrophages, we show that filopodia, or filopodia-like structures, support pathogen clearance by multiple means. Filopodia supported the phagocytic uptake of bacterial (Escherichia coli) particles by (i) capturing along the filopodial shaft and surfing toward the cell body, the most common mode of capture; (ii) capturing via the tip followed by retraction; (iii) combinations of surfing and retraction; or (iv) sweeping actions. In addition, filopodia supported the uptake of zymosan (Saccharomyces cerevisiae) particles by (i) providing fixation, (ii) capturing at the tip and filopodia-guided actin anterograde flow with phagocytic cup formation, and (iii) the rapid growth of new protrusions. To explore the role of filopodia-inducing Cdc42, we generated myeloid-restricted Cdc42 knock-out mice. Cdc42-deficient macrophages exhibited rapid phagocytic cup kinetics, but reduced particle clearance, which could be explained by the marked rounded-up morphology of these cells. Macrophages lacking Myo10, thought to act downstream of Cdc42, had normal morphology, motility, and phagocytic cup formation, but displayed markedly reduced filopodia formation. In conclusion, live-cell imaging revealed multiple mechanisms involving macrophage filopodia in particle capture and engulfment. Cdc42 is not critical for filopodia or phagocytic cup formation, but plays a key role in driving macrophage lamellipodial spreading.

Phagocytosis is a specialized form of endocytosis, which requires localized actin polymerization to engulf large particles (Ͼ ϳ0.5 m), and initially involves particle binding to phagocytic receptors, which triggers phagocytic cup formation through the activation of kinases, such as Syk, and Rho GTPases, such as the Cdc42 and Rac subfamilies (1). Although phagocytosis has been investigated in-depth, the involvement of filopodia (finger-like projections containing bundled actin filaments (2)) in actually capturing and clearing particles has not been conclusively demonstrated, except for the historical observations of Metschnikoff (3) and more recent work using brightfield microscopy (4 -7).
Young et al. (4) observed using time-lapse differential interference contrast (DIC) 2 imaging that Escherichia coli expressing invasin, a transmembrane protein of Yersinia pseudotuberculosis, could enter Hep-2 (HeLa-derived) cells via filopodia. Using another approach, coating of magnetic microbeads with invasin, Vonna et al. (5) found that the adhesion of beads to filopodial tips induced pulling toward the cell body. Similarly, Kress et al. (6) reported that filopodia act as "phagocytic tentacles" and pulled IgG-coated beads in an optical trap with discrete steps, suggesting that a motor protein was involved. More recently, Möller et al. (7) reported that macrophages can use filopodia to lift membrane-bound bacteria, thereby enabling a lamellipodial protrusion to engulf the target in a so-called hook-and-shovel mechanism.
In this study, we looked at phagocytosis from a new angle. We isolated resident peritoneal macrophages from transgenic Lifeact-EGFP mice (8) and used time-lapse spinning disk confocal microscopy to visualize the involvement of filopodial dynamics in the clearance of fungal and bacterial particles. Lifeact-EGFP mice ubiquitously express the F-actin-binding (17-amino acid) peptide Lifeact fused to the 238-amino acid fluorescent protein EGFP (9). The filopodiainducing proteins Cdc42 and Myo10 have both been implicated in phagocytosis (10,11). Using knock-out mouse models, we explored the roles of these proteins in macrophage phagocytosis.

Capture of zymosan particles by filopodial tentacles
Initially, we used superresolution-structured illumination microscopy (SR-SIM) to confirm that Lifeact-EGFP is expressed in resident peritoneal macrophages and localizes to F-actin (Fig. 1, A and B). Whereas Lifeact-EGFP and the highaffinity F-actin probe phalloidin, conjugated to Alexa Fluor 594, clearly co-localized, Lifeact-EGFP additionally accumulated in the nucleus. Similar to Lifeact-EGFP, we previously observed that EGFP alone also accumulated in the nuclei of resident peritoneal macrophages (isolated from LysM-EGFP (Lyz2 tm1.1Graf ) mice (12)). SR-SIM imaging provided striking detail of the actin architecture in macrophages (Fig. 1C). Instead of a dense branching network, typical for the lamellipodial membrane protrusions of fibroblast-like and epithelial cells (13,14), SR-SIM imaging showed nicely that macrophage lamellipodial membrane protrusions, as well as the cell body, are rich in arcs of actin bundles orientated parallel to the cell periphery. In addition, we observed that one or more prominent filopodial tentacles, or even a cluster of tentacles, typically extend from the cell periphery.
Using 2-channel spinning disk confocal microscopy, we found that the tips of filopodial tentacles were capable of capturing fluorescent zymosan particles (Fig. 2, A-C; supplemental Videos S1-S3). Once captured, the tentacle retracted, but concomitantly, it served as a bridge for the flow of actin and membrane toward and around the particle (Fig. 2, A and C; supplemental Videos S1 and S3). Time-lapse imaging revealed that new filopodial tentacles sprouted at active sites of tentacle-mediated capture and engulfment (Fig. 2, A-C; supplemental Videos S1-S3). Filopodial extensions could also be observed arising from closed phagocytic cups in snapshots of phagocytic events captured by fixation and scanning electron microscopy (Fig. 2D). While performing spinning disk confocal microscopy experiments, we invariably observed dynamic actin "hot spots" within the actin ring (corresponding to the phagocytic cup) surrounding ingested zymosan particles, and we speculated that these spots may correspond to transient filopodial tentacles emerging from the phagocytic cup. Instead, these hot spots, which accompany each uptake event, were variously located beneath, alongside, and above zymosan particles, and presumably serve as local pools of actin.
We assumed that zymosan uptake in our assays was triggered via the recognition of ␤-glucans by Dectin-1 (15), although both TLRs (Toll-like receptors) (16) and the MR (mannose receptor) (17,18) have been implicated in zymosan phagocytosis. To assess the contribution of these receptors to zymosan phagocytosis in our assays, we compared the phagocytic activity of wildtype (WT) versus Toll-like receptor-2/4 double knock-out (dKO), mannose-receptor knock-out, or Dectin-1 knock-out macrophages. Using zymosan particles labeled with the pHsensitive dye pHrodo, which becomes brighter at decreasing pH (Fig. 3, A and B), we could measure phagosome acidification, a readout for particle internalization. One caveat of this approach is that a subset of macrophages exhibits weak phagosome acidification (for example, see Fig. 3C). Nevertheless, using this approach, we could confirm that Dectin-1 is the primary recep- A, PCR genotyping of offspring derived from crossing heterozygous Lifeact-EGFP and wild-type (WT) mice. B, resident peritoneal macrophage isolated from a Lifeact-EGFP mouse. After fixation and permeabilization, the cell was stained with Alexa Fluor 594 (AF594)-conjugated phalloidin, an F-actin probe, and imaged by SR-SIM. The single channel images have been inverted. Scale bar, 10 m. C, SR-SIM (inverse) image of fixed and permeabilized mouse macrophages, stained with Alexa Fluor 488 (AF488)-conjugated phalloidin. The image, obtained via an oil-immersion ϫ63/NA 1.4 objective lens, is an overlay of 3 z-slices taken at 220 nm steps. Scale bar, 10 m.

Filopodia and phagocytosis
tor mediating both the binding and internalization of zymosan particles (Fig. 3D).

Multiple modes of filopodial tentacles in bacterial particle clearance
Next, we investigated the involvement of filopodial tentacles in the capture of E. coli particles (Fig. 4). Time-lapse spinning disk confocal microscopy revealed that filopodial tentacles were highly effective at clearing particles beyond the cell body. Once captured by the tips of filopodia, particles were retracted toward the cell body and internalized, or, alternatively, the sweeping of a tentacle was effective at bringing particles toward the cell body (Fig. 4, A and B; supplemental Videos S4 and S5). The growth of new filopodial tentacles, fed by a pool of actin, could be observed in time-lapse recordings ( Fig. 4A; supplemental Video S4). In addition to retraction and sweeping, the surfing of particles along the shaft of filopodial tentacles was frequently observed (Figs. 4C and 5, A and B; supplemental Videos S6 -S8). Particles surfed (moved) along filopodia with a velocity of 1-2 m/min (Fig.  5C). Combinations of surfing and retraction were also frequently observed (supplemental Video S9). As illustrated in Fig. 5D, surfing and combinations of surfing and retraction accounted for the majority of filopodia-mediated E. coli particle uptake events. We speculated that actin retrograde flow may drive the surfing of particles (see "Discussion"). Using kymograph analyses we found that the velocity of actin retrograde flow (treadmilling) in Lifeact-EFGP macrophages was in the range 2-4 m/min (Fig. 5E), consistent with previous reports (19,20).

Cdc42-deficient macrophages
To investigate the role of Cdc42 in macrophage filopodial tentacle formation and phagocytosis we generated myeloid-restricted Cdc42 (conditional) knock-out (cKO) mice (Fig. 6). Conditional deletion was confirmed by Western blot analysis of lysates from F4/80 ϩ cells (macrophages), purified by cell sorting (Fig. 6A). Spinning disk confocal microscopy of living cells (not shown in Fig. 6, but see Fig. 13D) and superresolution imaging of fixed cells stained with an F-actin probe, Alexa Fluor 488-conjugated phalloidin, showed that Cdc42-deficient macrophages had a severe cell spreading defect (Fig. 6B), reminiscent of Myo9b-deficient macrophages, which have overactive RhoA activity (21). Indeed, Cdc42-deficient macrophages were resistant to cell spreading, even in the presence of the Toll-like receptor 4 ligand lipopolysaccharide (LPS), used to activate cells, as shown in Fig. 6C and quantified in Fig. 6D, where the projected cell area is used as an index of spreading. Unstimulated Cdc42 cKO macrophages exhibited less filopodia per cell compared with WT cells (Fig. 6E). However, LPS stim-ulation for 90 min increased the number of filopodia per cell in both WT and Cdc42 cKO cells, such that there was no significance difference between activated (ϩLPS) WT and Cdc42  showing the fortuitous transfer of an E. coli particle from the filopodium of one cell to the filopodium of another cell. The particle is pulled toward the cell body by the first filopodium, then, following transfer, it appears to "surf" along the second filopodium.

Filopodia and phagocytosis
cKO cells (Fig. 6E). In addition, the median length of filopodia was shorter in unstimulated Cdc42 cKO macrophages, compared with WT, and LPS stimulation significantly increased filopodia length in both WT and Cdc42 cKO macrophages (Fig.  6F). We could not detect the putative filopodia-inducing Rho GTPases RhoJ, RhoQ, RhoD, or RhoF in mouse macrophages using RT-PCR and Western blot analyses (Fig. 7, A and B). Thus, Cdc42 is presumably the predominant filopodia-inducing Rho GTPase in macrophages (22), and, in Cdc42-deficient cells, the paucity of filopodia, but not cell spreading, is compensated by LPS stimulation.
Using time-lapse spinning disk confocal microscopy and fluorescent labeling of the cell membrane with Alexa Fluor 488conjugated anti-F4/80 antibodies, we imaged the engulfment of large (diameter, 5.19 m) mouse IgG-coated polystyrene beads by macrophages (Fig. 8). Most beads were taken up after passive contact with the cell membrane. The kinetics of phagocytic cup formation was measured by tracking XY, XZ, and YZ views of the cells (Fig. 8). Surprisingly, Cdc42-deficient macrophages were highly efficient at engulfing large particles (Fig. 8A), and the rate of cup closure was faster compared with WT cells (Fig.  8, B-D). However, the start of phagocytic cup formation may have been underestimated in Cdc42-deficient cells due to the highly ruffled surface of these rounded-up cells (Fig. 9A). In contrast, the initiation of phagocytic cup formation could be more easily detected in spread-out WT macrophages, especially when particles landed on lamellipodia (Fig. 8B). Nevertheless, whether first or second phagocytic half-cup kinetics were measured (Fig. 8D), Cdc42 cKO macrophages were faster at enveloping large beads. Notably, we occasionally observed that filopodia could haul in large particles before phagocytic cup formation was initiated (Fig. 9B), but these events were excluded from the above analyses in Fig.  8D.
Although Cdc42 cKO macrophages had faster phagocytic cup kinetics than WT cells, mutant cells engulfed less IgGcoated polystyrene beads and zymosan particles in end-point assays (Fig. 10A), presumably due to less frequent contacts (see "Discussion"). SR-SIM imaging of Cdc42-deficient macrophages, which were fixed while engulfing zymosan particles, revealed filopodia projecting from phagocytic cups (Fig. 10B).
Aside from phagocytosis (23), Cdc42 is implicated in chemotactic navigation (24 -26). Using real-time chemotaxis assays, we found that Cdc42-deficient macrophages had greatly impaired chemotactic navigation and modestly decreased median cell velocity (Fig. 10, C and D). Chemotaxis assays were performed in a complement C5a gradient (target concentration, 20 nM) and a uniform concentration of LPS (0.1 g/ml).
Myo10 is implicated in phagocytosis (11), especially in the engulfment of large particles, and it is thought to act downstream of Cdc42 in the generation of filopodia (27). We confirmed deletion of Myo10 in homozygous Myo10 reporter knock-out (Myo10 tm2/tm2 ) mice using Southern and Western blot analyses, as well as PCR analysis (Fig. 11A). 3 In contrast to Cdc42-deficient macrophages, Myo10 knock-out macrophages exhibited no defects in cell velocity or chemotactic navigation (Fig. 11B). Moreover, there were no impairments in phagocytic cup formation and the ingestion of large particles ( Fig. 11, C-E). Consistent with the time-lapse imaging data, no differences in the uptake of particles could be detected in end-point assays (Fig. 11F). To test whether Myo10-deficient macrophages have impaired filopodia formation we performed dynamic assays because filopodia cannot be unambiguously distinguished from retraction fibers arising from lamellipodial membrane retraction (Fig. 12A). We challenged macrophages with fluorescent E. coli particles, without prestimulation with LPS, and counted newly extended filopodia over a 20-min recording period (Fig.  12, B-D). Filopodia were formed by Myo10-deficient macrophages, but the total number per cell was greatly reduced compared with wild-type cells (Fig. 12D).
In another series of experiments, we compared the rates of filopodia formation in WT, Myo10 tm2/tm2 , and Cdc42 cKO macrophages during time-lapse recordings, with or without

Filopodia and phagocytosis
pretreatment with LPS (1 g/ml for 45 min) (Fig. 13, A-D). Consistent with the findings in Fig. 12, filopodia formation was markedly impaired in Myo10-deficient macrophages. Moreover, comparison of the images in Fig. 13, C (Myo10 tm2/tm2 ) and D (Cdc42 cKO), underscore the spreading defect of Cdc42deficient macrophages (see also Fig. 6), as well as the importance of distinguishing retraction fibers from filopodial protrusions, especially in the case of Myo10-deficient cells.

Discussion
Phagocytosis, the uptake of large particles, requires the force generated by actin polymerization to drive plasma membrane extensions around a target particle and form a phagocytic cup, followed by phagosome formation (28 -30). However, before a particle can be ingested it first needs to come in contact with a phagocyte. In 2010, Flannagan et al. (31) reported that particle-phagocytic receptor interactions were not passive, stochastic events, but, instead, macrophages continually probe the environment using Rac-dependent membrane protrusions. Filopodia, finger-like structures containing actin bundles, are highly suitable for probing the environment (2), and can be induced by "switching on" Cdc42 via Cdc42 guanine nucleotide exchange factors (22). As proof of principle, stud- ies using optical traps (tweezers) to present microbeads to macrophages showed that filopodia can pull beads (6), or, alternatively, beads can surf along filopodia toward the cell body (32). We asked what are the roles of filopodial dynamics in practice when primary macrophages are challenged with fungal or bacterial particles? In addition, we investigated the roles of the filopodia-inducing proteins Cdc42 and Myo10 in phagocytosis.

Filopodia and phagocytosis
We found that resident Lifeact-EGFP macrophages could capture moderately large (ϳ3 m diameter) zymosan particles beyond the cell body using filopodial tentacles. Particle binding triggered the flow of actin and membrane toward the particle, rather than simple retraction, such that filopodial tentacles widened and could support phagocytic cup formation, followed by internalization and new tentacle growth. A similar outward flow of the membrane, forming a pedestal, has been observed when a zymosan particle, initially held by a micropipette, is brought into contact with a neutrophil, held by a second micropipette (33). In contrast, no membrane protrusive effect was observed when neutrophils were presented with antibodyopsonized beads (33). Zymosan is a glucan prepared from yeast (Saccharomyces cerevisiae) cell wall and is recognized by TLR2/ TLR6 (Toll-like receptor 2/Toll-like receptor 6) heterodimers and Dectin-1 (34,35), and possibly also mannose receptors (36). Using macrophages isolated from TLR2/TLR4 dKO, Dectin-1 Ϫ/Ϫ and MR Ϫ/Ϫ mice, we confirmed that mouse macrophage zymosan binding and phagocytosis are primarily mediated by Dectin-1. We infer that Dectin-1, akin to Fc␥-receptors, transduces phagocytic signaling through an immunoreceptor tyrosine-based activation motif-like domain (37), and evokes anterograde actin flow and phagocytic cup formation when a filopodium makes contact with a zymosan particle (see schematic diagram, Fig. 14A).
Filopodial tentacles were highly efficient at fishing out E. coli particles beyond the cell body. We observed four main mechanisms (in order of frequency): capture and surfing along the filopodium, sweeping actions, combinations of surfing and retraction, and simple retraction (see schematic summary, Fig.  14B). In contrast to studies using optical traps (6, 32), we did not bring particles to macrophage protrusions, but, rather, macrophages were required to seek and capture targets. Notably, time-lapse imaging was important to distinguish the formation and dynamics of filopodia from the formation of retraction fibers. On the one hand, the relative contribution of surfing to particle uptake may have been inflated because retraction fibers were variably present at the start of time-lapse recordings. On the other hand, retraction fibers are structurally related to filopodia and similarly contain parallel, bundled actin (38). Furthermore, Svitkina et al. (38) demonstrated using time-lapse imaging that filopodia and retraction fibers are interconvertible structures. E. coli particles surfed along filopodia with velocities of 1-2 m/min, marginally slower than measured rates of retrograde flow (2-4 m/min). Assuming that the surfing of bound particles is driven by actin retrograde flow, a mechanism linking particles to actin filaments is required. Unconventional myosins could fulfill this role, along the lines nicely illustrated in Fig. 3 of the review article by Mermall et al. (39), or possibly the ERM (ezrin, radixin, and moesin) protein family could play a role (40).
In the pioneering study by Nobes and Hall (22), the injection of constitutively active Cdc42, together with inhibitors of Rac and Rho, into Swiss 3T3 (embryonic fibroblast-derived) cells was shown to induce filopodial tentacles remarkably similar to those we observed in macrophages, whereas this effect was much weaker in the presence of Rac activity. It is now known that Cdc42 stimulates actin nucleation by the actin-related protein-2/3 (Arp2/3) complex via WASP (Wiskott-Aldrich syndrome protein), which is predominantly expressed in hematopoietic cells (41). The Cdc42-WASP-Arp2/3 pathway is thought to be important for the initiation of filopodia formation, but, importantly, other proteins can induce filopodia, such as RIF (RhoF), mammalian Diaphanous-like (mDia) formins, and Myo10 (2). In relationship to this point, we could not detect filopodia-inducing Rho GTPases, other than Cdc42, in macro-
Remarkably, macrophages from myeloid-restricted Cdc42 knock-out mice had an aberrant rounded-up morphological phenotype, which was not rescued by activation with a chemoattractant (complement C5a) or the TLR4 ligand LPS (lipopolysaccharide). This overt phenotype suggests that, at least in the case of macrophages, the Cdc42-WASP pathway, rather than the Rac-WAVE (WASP family verprolin homologs) pathway, plays a predominant role in stimulating Arp2/3 activity and lamellipodial spreading. Less surprisingly, we found that chemotactic navigation and cell velocity were decreased in Cdc42 conditional knock-out macrophages, compared with WT cells, in line with observations in neutrophils (24) and dendritic cells (25), as well as earlier work using dominant-negative constructs (26). Indeed, Yang et al. (43) recently showed in ele-

Filopodia and phagocytosis
gant experiments that local activation of Cdc42, rather than Rac and Rho, chemotactically steer neutrophil-like cells. Moreover, loss of Cdc42 antagonism of RhoA, documented by Yang et al. (43), may explain the rounded-up phenotype we observed in Cdc42-deficient macrophages. We did not explore this possibility further, although, as proof of principle, Myo9b knock-out macrophages have both increased RhoA activity and a rounded up morphology, which is rescued by Rho inhibition (21).
Recently, knockdown of Cdc42 was reported to decrease the phagocytosis of large particles, giving rise to the notion that Cdc42 activity is required to take "big bites" (10). In contrast, we found that Cdc42-deficient macrophages generate phagocytic cups and engulf large particles (ϳ5.19 m diameter) at a faster rate than WT cells. The faster phagocytic cup kinetics is prob-ably due to the highly ruffled surface geometry of (rounded-up) mutant cells, which provides a rich local source of actin and membrane. Phagocytic cup formation upon a thin lamellipodial membrane protrusion in WT cells, in comparison, requires assimilation of actin and membrane over greater distances. Unstimulated Cdc42 conditional knock-out macrophages, imaged by SR-SIM, exhibited less filopodia than WT cells, as anticipated, and, in addition, the median filopodial length was significantly less in mutant cells. However, no significant differences in the number or lengths of filopodia, Cdc42 cKO versus WT macrophages, were observed after stimulating cells with LPS. Thus, LPS primes macrophages for the capture of bacteria by inducing filopodia formation (Fig. 14B), which is not critically dependent on Cdc42 activity.
Myo10 has been implicated in phagocytosis (11,44), but, surprisingly, we found that Myo10-deficient macrophages had no impairment in the engulfment of large IgG-opsonized beads, IgG-opsonized human red blood cells, 3 or zymosan particles. Moreover, Myo10 was redundant for cell spreading, motility, and chemotaxis. However, in further time-lapse imaging assays, we found that the frequency of filopodia formation was greatly reduced in Myo10 knock-out macrophages challenged with E. coli particles or LPS. Thus, Myo10 is important for the initiation of filopodia in macrophages. Importantly, as emphasized earlier, time-lapse imaging allowed filopodia to be easily distinguished from retraction fibers.
In conclusion, as partly illustrated in Fig. 14, we show that 1) zymosan engulfment is primarily mediated by Dectin-1, rather than mannose or Toll-like receptors, and in the case when a filopodium makes contact with a zymosan particle, Dectin-1 signaling triggers actin flow along the shaft toward the particle and phagocytic cup formation (Fig. 14A), 2) LPS, a component of Gram-negative bacteria such as E. coli, induces cell spreading and filopodia formation, the latter of which can help clear bacterial particles by various means: retraction, surfing, combinations of surfing and retraction, or sweeping (Fig. 14B), 3) both resting and LPS-induced cell spreading requires Cdc42, but not Myo10, whereas the Myo10 promotes filopodia formation in macrophages, and 4) Cdc42, but not Myo10, is important for cell motility and chemotactic navigation in a complement C5a gradient (Fig. 14C).

Mice
Lifeact-EGFP mice were generated as previously described by Riedl et al. (8). To generate myeloid-restricted Cdc42 knockout mice, Cdc42 (Cdc42 fl/fl ) mice, described by Wu et al. (45) and kindly supplied by Cord Brakebusch, were crossed with LysM-Cre mice (referred to as Lyz2 tm1(cre)Ifo mice by the Mouse Genome Informatics database), kindly supplied by Irmgard Förster (Heinrich-Heine-Universität, Düsseldorf). Heterozygous Myo10 reporter knock-out (Myo10 ϩ/tm2 ) mice were obtained from the Wellcome Trust Sanger Institute (Hinxton, Cambridge, United Kingdom). Phenotypic analysis of these mice is documented in Bachg et al. 3 TLR2 (Toll-like receptor 2) and TLR4 double mutant (TLR2/4 double knock-out) mice, described by Spiller et al. (46), were obtained from Carsten Figure 14. Schematic summary. A, zymosan phagocytosis in mouse macrophages is driven by Dectin-1 receptors. When a zymosan particle is engaged by a filopodium, Dectin-1 presumably signals via its immunoreceptor tyrosine-based activation motif (ITAM)-like domain to induce outward actin flow and phagocytic cup formation. Following ingestion, nascent filopodia grow upon the closed phagocytic cup. B, the TLR4 (Toll-like receptor 4) ligand LPS (lipopolysaccharide), derived from the cell wall of Gram-negative bacteria such as E. coli, induces cell spreading and filopodia formation within 60 min of application. Spreading is markedly impaired in Cdc42-deficient macrophages, whereas Myo10-deficient macrophages have normal spreading, but reduced filopodia formation. Myo10 harbors pleckstrin homology domains, which promote binding to PIP 3 (phosphatidylinositol (3,4,5)-trisphosphate)rich membranes. PIP 3 production is catalyzed by PI3K (phosphoinositide 3-kinase), activated by various pathways, including TLR4 signaling. Time-lapse imaging reveals multiple modes in which filopodia support particle capture and ingestion, including retraction of the tip-bound particle (red spot) toward the cell body, surfing of an attached particle along the shaft of a filopodium (presumably driven by actin retrograde flow), combinations of surfing and retraction, and sweeping actions. C, summary table. Both Myo10 and Cdc42 are redundant for phagocytic cup formation, whereas Cdc42, but not Myo10, is important for chemotactic navigation.
Kirschning (Universität Duisburg-Essen), and mannose receptor knock-out (MR Ϫ/Ϫ ) mice were provided by Sven Burgdorf (LIMES (Life & Medical Sciences) Institut, Bonn). Dectin-1deficient (Dectin-1 Ϫ/Ϫ ) mice were kindly provided by Gordon Brown (University of Aberdeen) and Laura E. Layland (Universitätsklinikum Bonn). Mice were bred in specific pathogen-free conditions and used in accordance with local animal experimentation guidelines. All procedures were approved by the University of Münster animal care and use committee.

Isolation and handling of resident peritoneal macrophages
Mice were killed by an overdose of isoflurane in air, and the peritoneal cavity was lavaged via a 24-guage plastic catheter (B. Braun, Melsungen, Germany) using 2 ϫ 4.5 ml of ice-cold Hank's balanced salt solution without Ca 2ϩ or Mg 2ϩ (Thermo Fisher Scientific). After centrifugation (300 ϫ g for 6.5 min), cells were resuspended in bicarbonate-free RPMI 1640 medium containing 20 mM Hepes (Biochrom AG, Berlin, Germany), and supplemented with 10% heat-inactivated fetal calf serum (FCS), 100 units/ml of penicillin, and 100 g/ml of streptomycin (pH 7.4). The cells were seeded into fibronectin-coated -slide I chambers or -slide chemotaxis chambers (Ibidi, Martinsried, Germany), and placed in a humidified incubator (37°C). After 2 h, -slide I chambers were filled with 1 ml of RPMI 1640 medium, including 10% FCS and antibiotics, and incubated overnight at 37°C with 5% CO 2 . Experiments were performed after switching back to bicarbonate-free RPMI 1640 medium containing 20 mM Hepes. Alternatively, cells were seeded onto fibronectin-coated glass coverslips or glass bottom dishes.

Two-dimensional chemotaxis assays
Cells obtained by peritoneal lavage of a single mouse were resuspended in 125-175 l of medium, and 10 l of the suspension was seeded into the narrow (1000 ϫ 2000 ϫ 70 m) channel of an uncoated (IbiTreat) -slide chemotaxis chamber (Ibidi). The narrow channel (observation area) connects two 40-l reservoirs. After 3 h, the chemotaxis chamber was filled with bicarbonate-free RPMI 1640 medium containing 10% FCS, 100 ng/ml of lipopolysaccharide and antibiotics. Next, 15 l of medium containing chemoattractant (C5a) and 0.003% Patent Blue V (blue dye) was drawn into one of the reservoirs. The final concentration of C5a was 20 nM. The observation area was imaged by phase-contrast microscopy via a ϫ10 (NA 0.3) objective lens. The blue dye served as a visual indicator of gradient formation, and we have previously confirmed that it does not affect cell migration. Images were captured every 2 min for 14 h, and cell migration tracks between 6 and 12 h were analyzed with ImageJ software (National Institutes of Health) using a manual tracking plug-in and the chemotaxis and migration tool from Ibidi. Twenty-five randomly selected cells were manually tracked in each chemotaxis experiment.

Spinning disk confocal microscopy
Two-or three-dimensional time-lapse imaging of living macrophages was performed using an UltraVIEW Vox threedimensional live cell imaging system (PerkinElmer Life Sciences, Rodgau, Germany) coupled to a Nikon Eclipse Ti inverse microscope. The system incorporated a Yokogawa (Japan) CSU-X1 spinning disk scanner, a Hamamatsu (Japan) C9100-50 EM-CCD camera (1000 ϫ 1000 pixels), and Volocity software. Cells were imaged via a Nikon ϫ60 (NA 1.49) oil immersion objective lens, and the temperature was maintained at 37°C using an Okolab all-in-one stage incubator (Okolab, Ottaviano, Italy). Z-stacks of Lifeact-EGFP macrophages or cells labeled with Alexa Fluor 488-conjugated anti-F4/80 antibodies were obtained using steps in the range 0.5 to 1 m. Focal drift was circumvented using the Nikon Perfect Focus System, which tracks the position of the coverslip in the z axis by reflecting infrared light (870 nm) from the coverslip and detecting it via a CCD (charge-coupled device) sensor.

DIC and fluorescence imaging
In selected experiments, time-lapse DIC and fluorescence microscopy was performed using a Zeiss inverted microscope (AxioObserver) equipped with a temperature-controlled incubator (incubator XL S, Zeiss) and controlled by AxioVision software (Zeiss). Images were captured via a ϫ63 (NA 1.4) oilimmersion objective lens and charge-coupled device camera (AxioCam MRm, Zeiss).

Phagocytosis assays
Time-lapse imaging of phagocytic events was performed after introducing particles to macrophages seeded in fibronectin-coated -slide I chambers, which have a channel volume of 100 l. Stock solutions of BioParticles (Thermo Fisher Scientific) were prepared at 20 mg/ml in Dulbecco's PBS and diluted 1:75 in live cell imaging medium, which consisted of bicarbonate-free RPMI 1640 medium containing 20 mM Hepes and 1 mM N-(2-mercaptopropionyl)glycine, a free radical scavenger. The suspension of BioParticles in live cell imaging medium was sonicated to disperse the particles, and 100 l was pipetted into a -slide I chamber before starting time-lapse imaging via spinning disk confocal microscopy. The following BioParticles were used: Alexa Fluor 594-conjugated E. coli (K-12 strain) particles (E23370; Thermo Fisher Scientific), Alexa Fluor 594-conjugated zymosan (polysaccharide prepared from the cell wall of S. cerevisiae (bakers' yeast), catalogue number Z23374; Thermo Fisher Scientific) and pHrodo-conjugated zymosan, prepared using unlabeled zymosan (Z2849; Thermo Fisher Scientific) and the succinimidyl ester of pHrodo red (P36600; Thermo Fisher Scientific), which has absorbance and emission maxima of 560 and 585 nm, respectively. In addition, phagocytosis assays were performed using mouse IgG-coated polystyrene particles (MsGPX-50-5; Spherotech), which had a mean diameter of 5.19 m. The 0.5% (w/v) stock solution was diluted 1:10 before use, and the beads were rendered red fluorescent by incubation with Alexa Fluor 594-conjugated goat anti-mouse IgG antibodies (ab150116; Abcam), followed by a wash step.

Scanning electron microscopy
Macrophages, seeded on fibronectin-coated glass coverslips, were fixed for 40 min with ice-cold Hepes-buffered (pH 7) medium containing 2% paraformaldehyde and 2.5% glutaraldehyde. As a pre-critical drying step, the cells were dehydrated by exposure, each time 15 min, to increasing concentrations of ethanol (30,50,70,90, and 100%). The transitional fluid ethanol was replaced with liquid CO 2 and the sample was brought to a critical point for CO 2 (31°C and 73.8 bar). Critical point drying (to remove CO 2 ) was achieved by subsequently decreasing the pressure at constant temperature (31°C). The dried sample was sputter coated with platinum and imaged using an Hitachi S5000 scanning electron microscope.

Superresolution structured illumination microscopy
Superresolution-structured illumination microscopy was performed using an Elyra S.1 inverted microscope system (Carl Zeiss MicroImaging, Germany), controlled by ZEN 2011 SP2 software (black edition; Zeiss). Cells, seeded on mouse fibronectin-coated coverslips or glass bottom dishes, were imaged via an oil-immersion Plan Apo ϫ63 (NA 1.4) objective lens and images were captured with an Andor iXon EM-CCD. The following lasers and filters (in parentheses) were used: 488 Filopodia and phagocytosis nm (BP 495-550 nm ϩ LP 750 nm) and 561 nm (BP 570 -620 nm ϩ LP 750 nm). Five grating positions and 5 phase shifts were used for each z-slice.

Statistics
Normality and homoscedasticity were tested using the Shapiro-Wilk and Levene tests, respectively. A one-way analysis of variance was used to test for statistical differences at the 0.05 level of significance. When the assumed conditions of normality and homogeneity of variance were not fulfilled, as in most cases, we used the non-parametric Mann-Whitney U test or Kruskal-Wallis one-way analysis of variance on ranks (at the 0.05 level of significance). Post hoc multiple comparisons were made using Dunn's method. Statistical analyses were performed using SigmaPlot (version 12) software (Systat Software), SPSS Statistics 22 (IBM Corporation), or Origin 2015 SR2 (OriginLab), and data are presented as box plots or mean Ϯ S.E.