The formins FHOD1 and INF2 regulate inter- and intra-structural contractility of podosomes Free

Macrophages are highly invasive cells of the monocytic lineage. In order to fulfill their functions during immune surveillance, they have to cross tissue barriers and navigate through the dense meshwork of the extracellular matrix (ECM) (Hynes, 2009), which often involves proteolytic cleavage of ECM material (Sabeh et al., 2009). Accordingly, macrophages are able to form actin-rich podosomes at the cell–substrate interface that function as both sites of adhesion and hotspots of matrix degradation (Linder and Aepfelbacher, 2003; Linder and Wiesner, 2015; Murphy and Courtneidge, 2011). Podosomes are able to locally degrade the matrix through recruitment of matrix-lytic enzymes, in particular of the matrix metalloproteinase family (Linder, 2007). For efficient invasion, both podosome-localized ECM degradation and turnover of the podosome structure itself have to be spatiotemporally coordinated.

Classically, podosome architecture is seen as bipartite (Linder and Aepfelbacher, 2003), consisting of a core of F-actin and actin-associated proteins such as cortactin (Ochoa et al., 2000) or gelsolin (Chellaiah et al., 2000), and a surrounding ring structure of plaque proteins including vinculin (Zambonin-Zallone et al., 1989) and paxillin (Pfaff and Jurdic, 2001). The structure is anchored to the ventral plasma membrane by transmembrane matrix receptors such as integrins (Pfaff and Jurdic, 2001; Teti et al., 1989) and CD44 (Chabadel et al., 2007), with cytoskeletal linkage provided by proteins such as talin (Zambonin-Zallone et al., 1989) or kindlin-3 (also known as FERMT3) (Ussar et al., 2006). However, recent structural analyses have revealed a more complex picture of podosome architecture. In particular, plaque proteins were shown to be assembled not in a continuous ring, but in several clusters that surround the core structure (van den Dries et al., 2013b). In addition, the core structure, being dependent on Arp2/3 complex activity (Kaverina et al., 2003; Linder et al., 2000a), has been shown to be surrounded by a layer of unbranched filaments (Akisaka et al., 2008; Luxenburg et al., 2007).

Moreover, unbranched actin filaments were shown to connect individual podosomes into a higher-ordered network (Burgstaller and Gimona, 2005; Luxenburg et al., 2007) and to exert forces based on actomyosin contractility (Bhuwania et al., 2012), thus probably coordinating turnover of individual podosomes with the net movement of podosome groups. Actomyosin contractility at podosomes is regulated by the membrane-associated protein supervillin (Bhuwania et al., 2012). Interestingly, supervillin was shown to localize to yet another substructure of podosomes, a cap-like structure on top of the podosome core (Bhuwania et al., 2012), which only partially overlaps with the core structure. A similar localization on top of the actin core of macrophage podosomes has been described for the formin FMNL1 (Mersich et al., 2010), indicating the potential existence of further cap proteins and pointing to a likely function of this structure in the regulation of unbranched actin filaments and actomyosin contractility at podosomes.

Podosomes are highly dynamic structures, displaying several levels of external and internal dynamics. The lifetime of podosomes has been determined as 2–12 min, and podosomal actin can be exchanged three times during this period (Destaing et al., 2003). Moreover, besides de novo formation, podosomes can also be formed by fission from pre-existing mother podosomes, which is particularly evident in primary macrophages (Evans et al., 2003; Kopp et al., 2006). In addition, podosomes have been shown to undergo cycles of actomyosin-based internal stiffness (Labernadie et al., 2010), leading to protrusion of the podosome structure, as shown by deformation of pliable matrix (Labernadie et al., 2014), accompanied by oscillations of actin-based fluorescence (van den Dries et al., 2013a).

Several lines of evidence thus suggest the presence of unbranched actin filaments at podosomes: (i) the identification of connecting cables that mediate contact between individual podosomes (Burgstaller and Gimona, 2005; Luxenburg et al., 2007), (ii) the observed contractility of individual podosomes (Labernadie et al., 2014,, 2010; van den Dries et al., 2013a), which is probably based on the presence of unbranched actin filaments (Akisaka et al., 2008; Luxenburg et al., 2007), (iii) the presence of proteins such as myosin IIA (also known as MYH9) and supervillin (Bhuwania et al., 2012) that regulate actomyosin contractility, thus necessitating the presence of unbranched F-actin, (iv) the presence of the formin FMNL1, a regulator of unbranched F-actin, at podosomes (Mersich et al., 2010), and (v) as revealed by proteomic data, the presence of further formin isoforms in podosome-enriched cell fractions from macrophages (Cervero et al., 2012). Collectively, these findings indicate, in addition to Arp2/3 and associated proteins, the likely presence of regulators of unbranched filaments at podosomes.

Formins are a family of actin regulators defined by the presence of a formin homology (FH)-2 domain involved in actin binding (Schonichen and Geyer, 2010). The 15 mammalian formin isoforms form several subgroups, including among others the diaphanous formin (DIA, also known as DIAPH), formin-like protein (FMNL), dishevelled-associated activator of morphogenesis (DAAM), inverted formin (INF2) or FH1/FH2-domain-containing protein (FHOD) groups. Formins fulfill a variety of actin-associated functions such as nucleation, elongation, capping, severing or bundling of actin filaments, with the actual activities depending on the individual isoform (Schonichen and Geyer, 2010). However, in contrast to Arp2/3-complex-generated branched actin networks, formins are primarily associated with unbranched actin filaments (Chhabra and Higgs, 2007).

In the current study, we report the identification of the formins FHOD1 and INF2 as new podosome components. FHOD1 has emerged as a regulator of stress fibers (Koka et al., 2003), through regulating the dynamics of actin transverse arcs and dorsal fibers (Schulze et al., 2014). Notably, FHOD1 has recently been shown to regulate maturation of integrin-based adhesion sites (Iskratsch et al., 2013). In vitro, FHOD1 acts as capping and bundling protein for actin filaments (Schonichen et al., 2013), although the exact biochemical activity of FHOD proteins is currently unclear (Bechtold et al., 2014). In contrast to these more accessory activities of FHOD1, INF2 has been shown to be involved in polymerization and severing of actin filaments (Gurel et al., 2014). This unique set of abilities enables the formation of transient actin filaments that are, for example, involved in fission of mitochondria (Korobova et al., 2013,, 2014). Importantly, INF2 can be expressed in two splice variants, which mostly differ in the presence or absence of a C-terminal CaaX box motif (where C is the cysteine residue that is prenylated, a is any aliphatic amino acid, and the residue at X determines which enzyme acts on the protein) (Chhabra et al., 2009; Ramabhadran et al., 2011) which, upon prenylation, mediates localization to the ER (Chhabra et al., 2009). Mutations in INF2 have been linked to the disorders focal and segmental glomerulosclerosis (Brown et al., 2010) and Charcot–Marie–Tooth disease (Boyer et al., 2011).

Using primary human macrophages, we now find that FHOD1 and INF2 localize to different substructures of podosomes. FHOD1 surrounds the podosome core and is also present at podosome-connecting cables, whereas INF2 localizes at the podosome cap structure. Accordingly, both formins were found to regulate different aspects of podosome-associated contractility, with FHOD1 mediating actomyosin contractility between podosomes, and INF2 regulating contractile events at individual podosomes. FHOD1 and INF2 thus present as novel regulators of inter- and intra-structural contractility of macrophage podosomes.

Previous mass spectrometry analyses revealed the presence of several formins in podosome-enriched cell fractions of primary macrophages (Cervero et al., 2012), with the formins DAAM1, DIA1 (DIAPH1), DIA2 (DIAPH2), FHOD1, FMNL1, FMNL2 and INF2 being detected in at least one of three experiments performed in parallel. In order to screen for potential localization of these formins at macrophage podosomes, we first overexpressed respective EGFP fusions in primary human macrophages. As formins are mostly autoinhibited by intramolecular binding between their Dia inhibitory domain (DID) and Dia autoregulatory domain (DAD) domains (Fig. 1A) (Higgs, 2005; Kuhn and Geyer, 2014), both wild-type and constitutively active constructs were used, lacking either the C-terminal DAD domain or bearing respective mutations in their DID domain (Fig. 1A). In addition to the reported localization of FMNL1 (Mersich et al., 2010), we detected prominent enrichment of both FHOD1 and INF2 at podosomes. Western blots of macrophage lysates also confirmed that both proteins are expressed endogenously (Fig. 1B,C). Moreover, INF2 can be expressed in at least two isoforms, distinguished by the presence or absence of a C-terminal CaaX box. Prenylation of the cysteine residue within this motif mediates binding to the endoplasmic reticulum (Ramabhadran et al., 2011). Therefore, isoform-specific antibodies were used, revealing that only the non-CaaX-containing isoform was detectable in macrophage lysates (Fig. 1C).

Co-staining of macrophages expressing EGFP–FHOD1 or EGFP–INF2-nonCaaX with Alexa-Fluor-568–phalloidin or Alexa-Fluor-647–phalloidin to label podosome cores showed that each formin localizes to different regions at podosomes. EGFP–FHOD1 was present at a shell-like localization surrounding the F-actin-rich core (Fig. 1D–F; Fig. S1C–E) and partially colocalizing with proteins of the podosome ring structure such as vinculin or talin (not shown). This was especially evident in three-dimensional reconstructions of single podosomes (Fig. 1D″–F″) and also confirmed by fluorescence intensity measurements (Fig. 1G). In addition, FHOD1 was also present at podosome-connecting actin cables (Fig. 1F′; for visualization of cables, see also Fig. S1J). The latter phenotype was particularly striking in cells expressing non-autoinhibited EGFP–FHOD1ΔC, which in most cases led to a prominent formation of these cables (Fig. 1D–F; Fig. S1F–H) (note that these phenotypes represent endpoints of a continuum of possible phenotypes that depends on respective expression levels). By contrast, EGFP–INF2-nonCaaX localized to podosome cores (Fig. 1H–K), without prominent localization to connecting cables. Importantly, three-dimensional reconstructions revealed that INF2 only partially colocalized with podosomal F-actin, and was mostly present as a cap-like structure on top of the podosome core (Fig. 1H″–J″). Localization of FHOD1 and INF2 to these podosome substructures was also confirmed for endogenous proteins by using respective antibodies (Fig. S2A–H).

Furthermore, FMNL2, a formin previously found to be enriched at ventral membranes of macrophages (Cervero et al., 2012), did not localize to podosomes (Fig. S3A–F), supporting the specificity of the localization of FHOD1 and INF2. Also, addition of the formin inhibitor SMIFH2 (25 µM) to macrophage cultures led to strongly pronounced reduction of podosomes (Fig. S3H–J), pointing to the general importance of formin activity in podosome formation and/or maintenance. We conclude that FHOD1 and INF2-nonCaaX are expressed in primary macrophages, and localize to distinct substructures of podosomes. (Note: for convenience, henceforth ‘INF2’ refers to the INF2-nonCaaX splice variant, unless otherwise indicated.)

We next established siRNA-mediated knockdown for FHOD1 and INF2 to assess their impact on structural and functional parameters of podosomes, including number, size and lifetime, as well as extracellular matrix degradation. For both formins, two sets of individual siRNAs were established that led to efficient depletion (90–96% knockdown; Fig. 2A,F). (Note: for initial testing of FHOD1, a pool of 4 sequences was used, which included the 2 individual sequences.) Absence of endogenous formins was also confirmed on the single cell level, ensuring the absence also of residual formins at podosomes (Fig. S2I–N). Interestingly, depletion of either FHOD1 or INF2 resulted in a reciprocal increase of ∼75% of protein levels of the other formin (Fig. S3K), pointing to the existence of cross-regulatory mechanisms. Depletion of FHOD1 did not result in major changes in either the number of podosomes per cell (Fig. 2B), podosome size (Fig. 2C) or podosome lifetime (Fig. 2D), although detailed analyses showed increases in the number of cells containing only up to 50 podosomes (16.01±1.42%, compared with 7.02±0.6% in control cells; Fig. 2B) and in the number of podosomes persisting for more than 20 min (15.24±5.02%, compared with 2.78±0.04% of control cells; Fig. 2D). Fluorescence intensity measurements of FHOD1-depleted macrophages seeded on NHS–Rhodamine-labeled gelatin matrix showed no significant differences in matrix degradation to control cells (Fig. 2E). Comparable to FHOD1 knockdown, depletion of INF2 did not result in significant changes in the number of podosomes per cell (Fig. 2G). Strikingly, however, podosome size was increased, in both mean values (siRNA#1: 0.84±0.06 µm; siRNA#2: 0.88±0.07 µm, compared with 0.68±0.01 µm using control siRNA) and size distribution (Fig. 2H). This was accompanied by an increase of the subgroup of long-lived podosomes persisting for more than 20 min (Fig. 2I), and a significant decrease in gelatin matrix degradation (Fig. 2J). Interestingly, depletion of INF2 did not alter the cap-like localization of supervillin on top of podosome cores (Fig. S4), indicating that although INF2 localizes to the podosome cap, it is not necessary for the upkeep of this structure. Collectively, these analyses revealed only a minor impact of FHOD1 on the tested podosome parameters. By contrast, INF2 emerged as a strong negative regulator of podosome size, and as a positive regulator of podosomal matrix degradation.

The previous experiments showed an influence of INF2 on podosome size, by measuring podosome core diameters (Fig. 2H). However, this parameter does not distinguish between an overall increase in podosome size (i.e. a higher volume of podosome cores) and an altered shape of podosomes (higher diameter, but lower height, leading to unaltered podosome volume). Therefore, optical z-stacks of podosomes from INF2-depleted macrophages were obtained, and the height of podosome cores was measured. Indeed, the average height of podosomes in INF2-knockdown cells was also significantly increased (0.89±0.05 µm and 0.79±0.02 µm for INF2-specific siRNAs versus 0.69±0.01 µm for controls), with a more detailed analysis showing a clear overall shift to higher core height values (Fig. 3A). In reciprocal experiments, the podosome diameter was measured in macrophages overexpressing the non-autoinhibited form INF2-A149D (Fig. 3B). Overexpression of this mutant led to strongly decreased average size of podosomes (0.50±0.03 µm for INF2-A149D-expressing cells versus 0.95±0.03 µm for controls), which was based on a general shift of podosome sizes to lower values (Fig. 3C), and was also accompanied by a pronounced reduction of podosome height (0.74±0.03 µm vs 1.02±0.03 µm for controls; Fig. 3D). Furthermore, overexpression of INF2-A149D also strongly decreased overall number of podosomes per cell (Fig. 3E). Consistent with the unaltered diameter of podosomes upon FHOD1 depletion, podosome height was also not affected upon FHOD1 knockdown and height distribution closely followed control values (Fig. 2 and not shown). Collectively, these experiments indicated that INF2 is a negative regulator of podosome core volume and podosome number. It should be noted that, given the limits of optical resolution by confocal imaging, the indicated sizes for podosome diameter or height cannot be accurately measured and thus represent only approximate values. Also note that occasional variability in controls is based on the variability between different preparations of primary macrophages. Therefore, controls of individual experiments were always performed with cells from the same preparation, to preserve internal consistency.

As INF2 emerged as a regulator of podosome number and size, we next investigated whether it also influences podosome dynamics at the level of single podosomes. For this, we determined F-actin levels at podosomes in a confocal plane over time, by measuring Lifeact-GFP fluorescence. As reported earlier for podosomes of dendritic cells (van den Dries et al., 2013a), Lifeact-GFP-based fluorescence of individual podosomes underwent constant fluctuations, indicative of the periodic actomyosin-based contractions of podosomes in the z axis (Labernadie et al., 2010; Luxenburg et al., 2012), which results in varying F-actin intensities in a fixed plane of focus (Fig. 4A). Strikingly, these oscillations were dampened in INF2-depleted cells (Fig. 4B,C), leading to a significant decrease in the respective coefficients of variation (10.68±0.72% for INF2 #1 siRNA- and 11.51±0.82% for INF2 #2 siRNA-, vs 14.80±0.94% for control siRNA-treated cells) (Fig. 4D). Addition of the myosin II inhibitor blebbistatin (20 µM) to cells led to an even more pronounced decrease (6.41±0.28%), pointing to actomyosin-based contractility as the basis for podosome oscillations, consistent with previous results (Labernadie et al., 2010; van den Dries et al., 2013a).

These experiments indicated that INF2 regulates podosome contractions in the z axis, which manifests as periodic alterations of F-actin intensity in a plane of focus, pointing to a potential role of INF2 as a regulator of intra-podosomal contractility. However, this effect could also be achieved by INF2 influencing overall F-actin levels at podosomes, through regulating the turnover of actin itself. We therefore tested whether INF2 regulates actin turnover within podosomes, by using FRAP analysis of single podosomes in cells expressing GFP–β-actin. Measurements of normalized fluorescence intensities showed that bleaching resulted in a ∼60% decrease of GFP–β-actin-based fluorescence at single podosomes (n=3×30 podosomes, from 3 cells) in cells treated with control siRNA. Within 60 s post-bleaching, 89.8±2.1% of initial fluorescence values had been recovered, with a half time of recovery of 10.5 s±2.2 s (Fig. 4E–G). Importantly, these values were not significantly different in cells treated with INF2-specific siRNA (91.9±1.1% recovery of fluorescence levels after 60 s; half time of recovery 11.4±0.7 s) (Fig. 4F,G). The measured half times of recovery of 10 s are lower than earlier measurements of ∼30 s in osteoclasts (Destaing et al., 2003; Luxenburg et al., 2012) and dendritic cells (Gotz and Jessberger, 2013), but more similar to the 13 s value, as determined in another study using dendritic cells (Gawden-Bone et al., 2014). However, in all these previous studies, whole subcellular regions containing several podosomes were bleached, in contrast to bleaching of single podosomes in the current study. Collectively, this set of experiments showed that INF2 is a positive regulator of podosome oscillations, but has no apparent influence on the speed of actin turnover within podosomes.

Podosomes are dynamic structures that undergo constant turnover. In order to assess the potential impact of FHOD1 and INF2 also on podosome formation, cells treated with respective siRNA were analyzed using the podosome reformation assay. This assay is based on the disruption of podosomes by the Src tyrosine kinase inhibitor PP2, with subsequent washout of the drug to monitor podosome reformation (Cervero et al., 2013; Linder et al., 2000b). ImageJ-based analysis allows the quantification of a statistically relevant number of cells and their associated podosomes (Fig. 5A) (Cervero et al., 2013). Formin-depleted macrophages were treated with 25 µM PP2, leading to almost complete disruption of podosomes, and analyzed for podosome content at timepoints 30, 60, 90 and 120 min after washout of the drug. At all timepoints, knockdown of FHOD1 by two siRNAs did not lead to discernible alterations in podosome reformation compared with controls (Fig. 5B). By contrast, INF2-depleted cells showed a significant delay in podosome reformation, leading to lower numbers of podosomes per cell at the indicated timepoints, compared with controls (Fig. 5C). (Note: regular podosome numbers were eventually reached after a prolonged washout period of 3.5 h; not shown, see Fig. 2G.)

Primary human macrophages show donor-dependent variations. For direct comparability between FHOD1- and IFN2-based effects, experiments described in Fig. 5B and C were thus also conducted in parallel, using cells from the same preparations. These experiments confirmed a significant, although less pronounced, reduction of the number of reformed podosomes upon INF2 depletion (Fig. 5D). Moreover, podosome reformation experiments in cells treated with control siRNA (to mimic transfection) and the myosin II inhibitor blebbistatin (10 µM) showed enhanced numbers of podosomes per cell, compared with controls (Fig. 5E), pointing to a role of myosin-based contractility in the establishment of regular podosome numbers. Collectively, these experiments showed that depletion of INF2 resulted in a pronounced delay in podosome reformation.

In order to study the actual dynamics of podosome formation, macrophages treated with FHOD1- or INF2-specific siRNA were also analyzed in the podosome reformation assay using live cell imaging. Podosome cores were visualized by expression of Lifeact-GFP. In control cells, reformation of podosomes was mostly detectable at early timepoints of PP2 washout (1–5 min). Interestingly, two phases of podosome formation were observed, with a first phase consisting mostly of de novo formation of podosomes, followed by a phase in which podosome formation also occurred through fission of preexisting podosomes (Fig. 6A′). As podosomes show a typical lifetime of 2–12 min (Destaing et al., 2003), events of podosome dissolution were also detected during the course of these experiments. These observations are also in line with earlier reports showing the existence of podosome fission in macrophages (Evans et al., 2003; Kopp et al., 2006). To better visualize podosome formation over time, a temporal color code was given to time lapse videos using ImageJ software. All individual frames of a time lapse video were thus progressively colored along the spectrum, with final merges depicting all events of podosome formation in the respective period (Fig. 6A″).

For cells treated with FHOD1-specific siRNA, statistical analysis showed no significant changes in the overall events observed for podosome de novo formation, fission and dissolution (Fig. 6D), which is in line with results gained with fixed specimens (Fig. 5B). By contrast, macrophages depleted for INF2 showed a delay in the formation of podosomes, consistent with data from fixed cells (Fig. 6D, Fig. 5C). In line with these results, statistical analyses showed significantly decreased levels of de novo generated podosomes (31.42±9.14% for INF2 siRNA- vs 54.36±2.38% for control siRNA-treated cells), which was accompanied by a significant increase in the levels of fission-generated podosomes (68.59±5.68% for INF2 siRNA- vs 45.64±2.91% for control siRNA-treated cells). Respective time courses of cells treated with control siRNA (Fig. 6E), FHOD1-specific siRNA (Fig. 6F) or INF2-specific siRNA (Fig. 6G) showed that, in addition to the effects described above for fixed cells in 30 min intervals, the onset of both de novo formation and fission of podosomes are delayed in cells depleted for either FHOD1 or INF2, compared with controls. This was especially pronounced in INF2 depletion (Fig. 6G). Reduced de novo formation of podosomes can also be appreciated in the color-coded merges of INF2-knockdown cells, showing podosome formation mostly in clusters, indicative of podosome formation in the vicinity of preexisting structures, through fission (Fig. 6C″).

Collectively, these experiments show that podosome reformation occurs in two phases: first, predominantly by de novo formation, and second, by a phase also including fission of daughter podosomes from preexisting structures. Overall values for de novo formation and fission are not significantly influenced by FHOD1. However, a time course analysis revealed a delayed onset of both phases in case of FHOD1 depletion. By contrast, INF2 could be identified as a positive regulator of the de novo phase of podosome reformation, and also as a negative regulator of podosome fission.

FHOD1 shows a prominent localization at podosomes and podosome-connecting cables (Fig. 1D−F). However, the data so far showed no significant impact of FHOD1 depletion on podosome parameters. We therefore focused next on a potential role of FHOD1 at connecting cables. Notably, podosome-connecting cables also contain myosin, which is important for regulating actomyosin-based contractility between podosomes (Bhuwania et al., 2012). Consistent with previous results (Bhuwania et al., 2012), endogenous myosin IIA, the predominant myosin II isoform in macrophages (Maupin et al., 1994), was found to localize to podosome groups in macrophages expressing EGFP–FHOD1 constructs, in clusters surrounding individual podosomes, and also between podosomes (Fig. 7A–D). Note that connecting cables are hard to visualize using fluorescently labeled phalloidin (presence of connecting cables was checked by overexpression of Lifeact-GFP constructs; see also Fig. S1J). However, especially in cells expressing non-autoinhibited EGFP-FHOD1ΔC (∼50% of cells), F-actin-based connecting cables were more prominent, and their decoration by myosin IIA was clearly discernible (Fig. 7E–H). Moreover, podosome-connecting cables in these cells were often aligned in parallel, indicative of increased tension. This is in line with earlier results showing that increased actomyosin-based contractility results in a symmetry break in the uniform podosome pattern and parallel alignment of podosomes and podosome-connecting cables (Bhuwania et al., 2012).

These observations indicated that FHOD1 activity could regulate actomyosin-based contractility at podosome-connecting cables. To test this possibility, macrophages with established knockdown of FHOD1 were submitted to the podosome reformation assay, fixed at the indicated timepoints, stained for phosphorylated myosin light chain (p-MLC), a reporter of myosin contractility (Matsumura, 2005; Vicente-Manzanares et al., 2009), and p-MLC levels in the podosome-covered area of cells were measured. Strikingly, we found elevated levels of p-MLC during podosome reformation in control cells, compared with non-PP2-treated cells, indicative of increased actomyosin contractility. This was especially pronounced at the 90 min timepoint of reformation (Fig. 7L). By contrast, FHOD1-depleted cells showed no significant alteration in p-MLC levels during the whole reformation period (Fig. 7L), whereas cells overexpressing the EGFP-FHOD1ΔC construct showed enhanced p-MLC levels (Fig. 7M). We conclude that FHOD1 localizes together with myosin IIA around podosomes and at podosome-connecting cables, and that FHOD1 activity is necessary for the regulation of actomyosin contractility at and between podosomes, which probably helps to establish the regular pattern of podosome groups.

In order to assess the impact of FHOD1 and INF2 on podosome connecting cables in more detail, we imaged respective single and double knockdown cells, stained for F-actin by Alexa-Fluor-488–phallodin, using structured illumination microscopy (SIM). In cells treated with control siRNA, F-actin-rich podosome cores were clearly visible, whereas connecting cables appeared often only as faint structures that connect individual podosomes (Fig. 8A). By contrast, cells depleted for FHOD1 showed a more diffuse arrangement of cable-like structures surrounding and connecting podosome cores (Fig. 8B). Similar observations were made for knockdown of INF2 (Fig. 8C) or knockdown of both FHOD1 and INF2 (Fig. 8D). In some cases, we also observed an alteration of the radially symmetric shape of podosomes to a more dash-like appearance. To assess the impact of FHOD1 and INF2, respective specimens were grouped according to the severity of the phenotype in double-blinded experiments (Fig. 8E). Depletion of either formin led to a more pronounced phenotype compared with controls, with depletion of FHOD1 causing the most severe cases (Fig. 8E). Collectively, these data indicate that FHOD1 and INF2 are involved in the upkeep of regular podosome connecting cables, and that depletion of either formin results in phenotypic aberrancies.

In this study, we identify the formins INF2 and FHOD1 as novel components and regulators of macrophage podosomes. INF2 localizes to the podosome cap structure, regulating podosome de novo formation, size and oscillation, as well as matrix degradation, whereas FHOD1 localizes around podosome cores and at podosome-connecting cables, regulating podosome connectivity.

Recent research has revealed an increasing complexity of both podosome sub- and superstructures (Veillat et al., 2015). Unbranched actin filaments were found to surround the podosome core, connecting the top of the podosome to the ring of adhesion plaque proteins, thus providing additional anchoring to the plasma membrane (Luxenburg et al., 2007). Moreover, radial fibers emerging from the podosome cores connect individual podosomes into higher-ordered groups (Akisaka et al., 2008; Luxenburg et al., 2007). The identification of a cap structure on top of the podosome core is the most recent addition (Linder et al., 2011; Mersich et al., 2010), although the potential function of this substructure is currently unexplored.

Of note, the existence of several subsets of unbranched actin filaments at or between podosomes (Akisaka et al., 2008; Luxenburg et al., 2007) pointed to the presence of regulators also of unbranched actin filaments, in addition to the well-known influence of Arp2/3 complex (Kaverina et al., 2003; Linder et al., 2000a). We thus screened for the potential localization of formins at macrophage podosomes and detected prominent enrichment of both INF2 and FHOD1. In particular, we identify INF2 as a novel component of the podosome cap. Beside FMNL1 (Mersich et al., 2010) and supervillin (Bhuwania et al., 2012), INF2 is one of the few proteins currently identified at the podosome cap. Interestingly, INF2 does not seem to be involved in upkeep of the cap structure, as supervillin localization at the cap is not altered in INF2-knockdown cells.

Podosomes display several levels of dynamics, including of de novo formation, fission, fusion, lateral mobility and dissolution (Destaing et al., 2005; Evans et al., 2003; Linder et al., 2000b), with concomitant turnover of core actin (Destaing et al., 2003). In addition, podosomes undergo periodic oscillations, as demonstrated in macrophages (Labernadie et al., 2010) and dendritic cells (van den Dries et al., 2013a), and this phenomenon is coupled to their mechanosensing ability (Luxenburg et al., 2012). In macrophages, podosome cores show periodic cycles of stiffness, with a period of 40–60 s (Labernadie et al., 2010). This is based on cycles of both myosin II activity and actin polymerization. Moreover, the protrusive force of podosomes, which is in the 10²–10³ nN range, shows a similar periodicity, with cycles of 40–50 s (Labernadie et al., 2014). The current model thus pictures podosomes as a contractile system, constantly balancing forces generated by actin polymerization in the core and by the lateral actin cables that function as springs (Labernadie et al., 2014), ultimately enabling podosomes to function as mechanosensory devices (Linder and Wiesner, 2015). We now show that siRNA-mediated knockdown of INF2 leads to a ∼25% dampening of podosome oscillations, pointing to a role of INF2 in the mechanosensing ability of these structures. Involvement of the cap in podosomal substrate sensing is particularly novel, as no function of the cap structure has been identified so far. Furthermore, FRAP experiments showed that INF2 depletion has no detectable effect on actin turnover within podosomes, which seems to point to a more structural role of INF2. Still, it should be kept in mind that most of the actin filaments in podosomes are Arp2/3-generated, and major fluctuations as a result of INF2 depletion are thus not to be expected.

Furthermore, INF2 emerged as a crucial regulator of podosome size. Knockdown of INF2 led to an overall increase of ∼25% in podosome diameter. This was accompanied by a ∼20% increase in podosome height, thus corresponding to a true net increase in size (as opposed to the theoretical case of wider but more flat podosomes). Consistently, expression of a non-autoinhibited mutant, INF2-A149D, led to a reciprocal effect, resulting in a ∼50% reduction of podosome diameter and a ∼30% decrease of podosome height. Beside PAK4 (Gringel et al., 2006) and calpain (Chou et al., 2006), INF2 is thus one of the few identified regulators of podosome size, pointing to an important modulatory role of INF2 in podosome architecture. It is unlikely that INF2 at the cap directly inhibits the growth of Arp2/3-generated actin filaments, as formins track the barbed ends of actin filaments, and barbed ends within the core are expected to localize towards the plasma membrane (Linder, 2007). In an alternative scenario, INF2-regulated lateral cables, embedded into the cap, could act as a ‘corsage’ that enwraps the podosome, thus limiting growth of the core structure. Further analyses of podosome dynamics showed that INF2 depletion led to a delay in podosome reformation, which is based on reduced de novo formation of podosomes and is accompanied by increased fission of podosomes. This points to a common requirement of both Arp2/3-dependent polymerization of core material (Linder et al., 2000a) and INF2-dependent lateral fibers in the initial stages of podosome formation, and also seems to be in line with the proposed model of INF2-dependent cables limiting podosome core growth. It is also noteworthy that, although INF2 depletion leads to an upregulation of protein levels of FHOD1, and vice versa (Fig. S3K), FHOD1 is apparently not able to compensate for INF2 in these functions.

Considering the potential biochemical basis for these effects, two points should be mentioned. First, INF2 displays both actin polymerizing and depolymerizing activity (Chhabra and Higgs, 2006), and this twofold ability can lead to the formation of transient actin filaments, a mechanism that has been shown to be involved in mitochondrial fission (Korobova et al., 2013). Second, in order to achieve oscillation of podosomes, the lateral actin filaments have to be modeled as springs (Labernadie et al., 2014), and the constant polymerization and depolymerization of actin by INF2 might provide the necessary flexibility for this system. This seems to be counterintuitive, as we did not detect changes in actin turnover in FRAP experiments of INF2-knockdown cells. However, it should be kept in mind that the vast majority of podosomal F-actin is localized in the core and thus generated by Arp2/3 complex, whereas the lateral cables constitute only a minor amount of the podosomal F-actin pool.

The impact of INF2 on podosome formation, size and oscillation is probably linked to a regulation of podosome architecture itself. In contrast, the observed impact of INF2 on podosomal matrix degradation is probably more indirect. Upon INF2 depletion, most podosomes show normal lifetime, and also their number is unaltered under these conditions. The ∼30% decrease in matrix degradation is thus clearly not based on a reduction of available podosome structures, but more likely based on defects in the recruitment or release of matrix-lytic enzymes. Indeed, it has been speculated that the podosome cap could function as a hub for incoming vesicles (Linder et al., 2011). Moreover, INF2 activity has been implicated in the secretory pathway, and the importance of INF2-regulated short actin filaments was shown for vesicle trafficking during transcytosis in hepatocytes (Madrid et al., 2010), and also for targeting of Lck tyrosine kinase-containing vesicles to the plasma membrane in Jurkat T cells (Andres-Delgado et al., 2010). Clearly, the possibility of INF2-regulated vesicle transport at podosomes is a field that needs further investigation in the future. Collectively, our data identify INF2 as a novel important regulator of multiple aspects of podosome architecture, dynamics and function. INF2 is a positive regulator of podosome de novo formation, and a negative regulator of podosome size. It is important for the generation of podosome oscillations and thus for mechanosensing, and it is also involved in the matrix degrading ability of these structures.

In contrast to the multiple aspects influenced by INF2, the impact of FHOD1 on podosomes has been more elusive. FHOD1 localizes around podosomes and also at podosome-connecting cables, in both overexpressed and endogenous forms. As FHOD1 depletion did not influence a variety of tested parameters associated with individual podosomes, we focused on the potential role of FHOD1 on the regulation of podosome-connecting cables. These fibers, consisting of actomyosin, are contractile (Bhuwania et al., 2012), and thus ideally placed to mediate force propagation between podosomes and to regulate coherence within podosome groups (Proag et al., 2015). Indeed, we found marked differences in myosin activity, as detected by phospho-myosin light chain levels, at and between podosomes during podosome reformation. Control cells showed increasing pMLC levels during the reformation phase, which probably reflects the not-yet balanced forces between podosomes during the establishment of the equidistant podosome pattern. These results fit well with the observed preference of both myosin IIA (Verkhovsky and Borisy, 1993) and FHOD1 (Schulze et al., 2014) to bind to anti-parallel, i.e. potentially contractile, actin filaments. In contrast, FHOD1-depleted cells showed no alterations in pMLC levels, pointing to defects in actomyosin contractility. In a reciprocal experiment, expression of a non-autoinhibited form of FHOD1, lacking the C-terminal DAD domain, led to pronounced formation of connecting cables, their prominent decoration with myosin IIA, and to increased levels of p-MLC at podosomes. These results are also in line with current literature showing an impact of FHOD1 in the formation of actin cables in several cell types, as expression of a dominant active form of the Drosophila homolog Knittrig led to stress fiber formation in macrophages and endothelial cells (Lammel et al., 2014), and expression of non-autoinhibited FHOD1-V228E led to increased formation of stress fibers in U2OS osteosarcoma cells (Schulze et al., 2014).

Further analysis using superresolution microscopy showed disorganized connecting cables in cells depleted for FHOD1, and also for INF2. Although FHOD1 knockdown gave the severest phenotype, the apparent participation of INF2 might point to the cap structure as an organizer of these cables, as hypothesized earlier (Linder and Wiesner, 2015). It is currently unclear whether these disorganized filaments reflect a net increase in inter-podosomal F-actin, or whether, following formin depletion, connecting cables are more loosely bundled and thus more accessible for phalloidin staining. In fact, both scenarios could be envisioned, considering the currently known biochemical activities of FHOD1. In vitro, FHOD1 shows no actin polymerization or depolymerization activities, but acts as capping and bundling protein for actin filaments (Schonichen et al., 2013). The disordered connecting cables in FHOD1-knockdown cells could thus result from a lack of bundling. In addition, the ability of FHOD1 to inhibit the growth of parallel, i.e. non-contractile actin filaments (Schulze et al., 2014), might help to ensure that only anti-parallel bundles of actin are formed between podosomes. Depletion of FHOD1 might thus also lead to enhanced formation of parallel actin filaments between podosomes. Taken together, our results point to FHOD1 as a regulator of inter-podosomal contractility. FHOD1 regulates podosome-connecting cables, probably by bundling anti-parallel actin filaments, thus forming the basis for proper levels of actomyosin contractility, which ensures efficient force transduction between podosomes.

Collectively, our results identify the formins INF2 and FHOD1 as novel components of macrophage podosomes that regulate different aspects of podosome-associated contractility. INF2 emerged as a novel component of the recently identified podosome cap structure, and is involved in regulating podosome size, de novo formation, fission, and oscillations. Especially the latter fact points to a potential role of the cap structure and its components in the regulation of intrastructural contractility, and thus of mechanosensing by podosomes. In contrast, FHOD1 regulates actomyosin-based contractility of podosome-connecting cables, thus regulating the connectivity of podosomes in higher-ordered clusters. Podosomes thus present as one of the few currently identified structures, which depend on the concerted activity of both Arp2/3 complex and specific formins as their integral components, and might serve as a model system for the analysis of complex actin architectures in cells.

Human peripheral blood monocytes were isolated from buffy coats (kindly provided by Frank Bentzien, University Medical Center Eppendorf, Hamburg, Germany) and differentiated into macrophages, as described previously (Wiesner et al., 2013). Approval for the analysis of anonymized blood donations was obtained by the Ethical Committee of the Ärztekammer Hamburg (Germany). Cells were cultured in RPMI-1640 (containing 100 units/ml penicillin, 100 µg/ml streptomycin, 2 mM glutamine and 20% autologous serum) at 37°C, 5% CO₂ and 90% humidity. Monocytes were differentiated in culture for at least seven days, under addition of 20% human autologous serum. For transfection experiments, differentiated macrophages, at days 7–10 of culture, were transiently transfected using the Neon^® Transfection System (Thermo Fisher Scientific, Waltham, MA), an electroporation-based system, with standard settings (1000 V, 40 ms, 2 pulses) with siRNA (1 µM) and plasmids (0.5 µg/1×10⁵ cells), respectively.

EGFP–FHOD1 and EGFP–FHOD1ΔC [amino acids (aa) 1–1010] were kindly provided by Oliver Fackler (University Hospital Heidelberg, Germany) (Gasteier et al., 2005). EGFP–INF2-nonCaaX (aa 1–1240) and EGFP–INF2-nonCaaX A149D were kindly provided by Henry N. Higgs (Dartmouth Medical School) (Ramabhadran et al., 2011). EGFP–FMNL2B and EGFP–FMNL2BΔDAD were kindly provided by Klemens Rottner (Technische Universität Braunschweig, Germany). mRFP–SV was kindly provided by Elisabeth J. Luna (University of Massachusetts) (Fang et al., 2010). EGFP-actin was generated as described previously (Osiak et al., 2005). Lifeact-GFP was purchased from Ibidi (Martinsried, Germany). For knock down of FHOD1, an ON-TARGETplus SMARTpool (FHOD1 p) was used (Thermo Fisher Scientific), as well as two individual siRNAs present in the pool: #1, 5′-GCGCUUGAGUAUCGGACUU-3′, and #2, 5′-GAUACUACCUGGACACCGA-3′. siRNAs specific for human INF2 were purchased from Ambion (Life Technologies, Carlsbad, CA) with the following sequences: #1, 5′-CCAUGAAGGCUUUCCGGGA-3′; #2, 5′-CAUCCAACGUGAUGGUGAA-3′. Non-targeting siRNA No. 2 (Thermo Fisher Scientific) was used as negative control. Onset and persistence of respective knockdowns was checked by western blots of respective cell lysates.

F-actin was stained using Alexa-Fluor-488–, Alexa-Fluor-568– or Alexa-Fluor-647–phalloidin, as indicated (Molecular Probes, Eugene, OR). Mouse monoclonal anti-actin antibody MAB1501 (clone C4) was purchased from Millipore (Billerica, MA) and used at 1:5000 on western blots. FHOD1 was detected using mouse polyclonal antibody (pAb) FM3521 (ECM Bioscience, Versailles, KY) at 1:50 in immunofluorescence and mouse monoclonal antibody (mAb) D-6 (sc-365437; Santa Cruz Biotechnology, Dallas, TX) at 1:200 on western blots. Rabbit pAbs specific for INF2-nonCaaX and INF2-CaaX, and a general anti-INF2 polyclonal, were kindly provided by Henry N. Higgs (Dartmouth Medical School) (Ramabhadran et al., 2011) and were used at 1.5 ng/ml on western blots or at 1:50 in immunofluorescence. FMNL2 was detected using rabbit pAb 72105 from Abcam (Cambridge, UK) at 1:500 on western blots. Rabbit anti-myosin IIA pAb M8064 was purchased from Sigma-Aldrich (St Louis, MO), used at 1:100 in immunofluorescence, and a phospho-specific (pS20) anti-myosin II light chain rabbit pAb ab2480 from Abcam, used at 1:100 in immunofluorescence. Mouse anti-vinculin mAb V9264 was purchased from Sigma-Aldrich and used at 1:500 in immunofluorescence. All fluorochrome-coupled secondary antibodies (Alexa-Fluor-488 goat anti-mouse and Alexa-Fluor-568 goat anti-rabbit) were purchased from Molecular Probes (Eugene, OR), and were used at 1:200 in immunofluorescence, and all horseradish peroxidase (HRP)-coupled ones (HRP goat-anti mouse, HRP goat-anti-rabbit) from GE Healthcare (Chalfont St Giles, UK), and were used at 1:5000 on western blots. To stain the whole cytoplasm, HCS CellMask™ Red Stain (Thermo Fisher Scientific; 2 µg/ml) was used.

Cells were scraped from dishes in buffer [150 mM NaCl, 1% Triton X-100, 50 mM Tris-HCl, pH 8.0, with protease inhibitor cocktail (Roche, Basel, Switzerland)]. After 30 min on ice, lysates were centrifuged (10 min, 10,000 g, 4°C) and supernatants were examined by standard immunoblotting procedure, with actin as a loading control, using the above-mentioned primary antibodies, and HRP-coupled anti-mouse or anti-rabbit IgG as secondary antibodies. Protein bands were visualized by using Super Signal Pico kit (Pierce, Rockford, IL) and X Omat AR films (Kodak, Stuttgart, Germany).

Cells were fixed for 10 min in 3.7% formaldehyde and permeabilized for 10 min in (0.5% Triton X-100, PBS, pH 7.5). After staining the cells with antibodies or labelling reagents mentioned above, the coverslips were mounted in Mowiol (Calbiochem, Darmstadt, Germany) containing DABCO (25 mg/ml; Sigma-Aldrich) as anti-fading reagent.

Images of fixed samples were acquired with confocal laser-scanning microscopes (Leica DM IRE2 with a Leica TCS SP2 AOBS confocal point scanner equipped with an oil-immersion HCX PL APO 63× NA 1.4 λblue objective and 3× PMT detectors or Leica DMI 6000 with a Leica TCS SP5 AOBS confocal point scanner equipped with an oil-immersion HCX PL APO CS 63× NA 1.4 objective and 2× HyD, 2× PMT detectors). Acquisition and processing of images were performed with Leica confocal software (Leica, Wetzlar, Germany) and/or Volocity 6.1.1 software (Perkin Elmer, Waltham, MA) and Photoshop CS5 (Adobe, Dublin, Ireland). For analyzing podosome number and size, primary human macrophages were transfected with control siRNA and siRNA against FHOD1 and INF2 and seeded on glass coverslips at a density of 1×10⁵ cells. After 72 h of incubation, the cells were fixed, permeabilized and stained with Alexa-Fluor-488–phalloidin for highlighting podosome cores. Images of fixed samples were taken as mentioned above. Podosome numbers (3 donors, 100 cells per condition) were measured using an ImageJ (NIH, Bethesda, MD) macro (see Table S2). Podosome diameters were measured with ImageJ (3 donors, 8 cells per condition). The core height was measured using an adapted ‘Analyze Particle’ macro and an additional macro for ImageJ (see Table S2) to measure the intensity of individual ROIs. The core height for INF2-nonCaaX-A149D-expressing cells (3 donors, 80 podosomes from 8 cells) was measured using the line tool of Volocity 6.1.1 in xz mode. To detect pMLC levels at podosomes, pMLC-based fluorescence intensity of confocal micrographs of cells stained with pMLC antibody and labeled with Alexa-Fluor-568 anti-rabbit and Alexa-Fluor-488–phalloidin was measured using Volocity 6.1.1. The ROI tool was used to restrict the measurement to the podosome area. All data were processed using Excel 2013 (Microsoft, Redmond, WA) and GraphPad Prism 6 (La Jolla, CA).

Live-cell imaging was performed using a spinning disk microscope [Nikon Eclipse Ti with a UltraVIEW VoX system (Perkin Elmer)] and a small temperature- and CO₂-controllable environmental chamber combined with an objective lens heater [(Tokai Hit INU-F1, Japan), equipped with a Yokogawa CSU X1 spinning disk, an oil-immersion 60× Apo TIRF (corr.) objective, a 527 nm (W55) emission filter and a Hamamatsu EM-CCD C9100-50 camera], or the above mentioned confocal laser-scanning microscope (Leica DMI 6000). Acquisition and processing of images were performed with Leica confocal software and/or Volocity 6.1.1 software. To evaluate podosome lifetime and oscillations, cells from three individual donors were transfected with control siRNA and siRNA against FHOD1 and INF2. After 48 h, 1×10⁵ cells were re-transfected with siRNAs, as well as Lifeact-GFP and seeded on glass bottom live cell dishes. After further incubation for 24 h, cells were imaged for a period of 45 min with image acquisition rate of 4× min⁻¹. Podosome lifetime of FHOD1-knockdown cells was evaluated by tracking the podosomes with Imaris software (Bitplane, Zurich, Switzerland). Each track resembles the lifetime of an individual podosome. All track data were processed in Excel 2013 and GraphPad Prism 6. Podosome lifetime of INF2-knockdown cells was evaluated using Volocity (Perkin Elmer). xy-drift was corrected with the Rigid body transformation of the StackReg plugin in ImageJ. To analyze podosome oscillations, the intensities of five individual podosomes per cell for three different donors per condition were measured at every timepoint using ImageJ. All data were processed with Excel 2013 and GraphPad Prism 6. For FRAP experiments, cells were treated with INF2 siRNA#1 or control siRNA. After 48 h of incubation, 1×10⁵ cell were re-transfected with the individual siRNAs, as well as GFP–actin and seeded on glass bottom live cell dishes. After additional incubation for 24 h, cells were imaged using the Leica DMI 6000. Images of single cells were taken every 1.29 s. After collecting five pre-bleach images, two circular ROIs with individual podosomes were bleached for five timepoints using 78% of a 405 nm laser. Recovery images were taken for additional 60 s. FRAP analysis was performed with the FRAP profiler (ImageJ). The data was processed with Excel 2013 and GraphPad Prism 6. Fluorescence recovery was measured for 10 individual podosomes at each timepoint for three cells from three donors.

Gelatin (from swine; Roth, Karlsruhe, Germany) was fluorescently labeled with NHS–Rhodamine (Thermo Fisher Scientific), according to Chen et al. (1994). Coverslips were coated with labeled matrix solution, fixed in 0.5% glutaraldehyde (Roth) and washed with 70% ethanol, RPMI and culture medium. Cells were re-seeded on coated coverslips with a density of 1×10⁵ cells 72 h after siRNA transfection, fixed and permeabilized 6 h post seeding and stained with Alexa-Fluor-647–phalloidin. Values of matrix degradation were determined by a loss of fluorescence intensity relatively to non-degraded matrix, using ImageJ software. For comparability, laser intensity was not changed between measurements. For each value, 3×30 cells were evaluated. Values were analyzed using Excel 2013 and statistical analysis was performed with GraphPad Prism software. Differences between mean values were analyzed using Student's t-test.

This assay was performed and analyzed in control and knockdown, fixed and living cells 72 h after siRNA transfection as published previously (Cervero et al., 2013). For live cell imaging, siRNA transfected cells were re-transfected after 48 h with both siRNA and Lifeact-GFP and seeded on glass bottom live cell dishes for another 24 h before imaging. The podosome reformation process was further analyzed in live cell movies, counting the events of podosome de novo formation, cluster fission and dissolution in defined areas of control and knockdown cells. Numbers were processed in Excel 2013 and GraphPad Prism 6. For rainbow analysis, the movies were processed with ImageJ, adding a temporal color code for each timepoint (LUT: Rainbow Smooth).

Cells were fixed for 1 min at 37°C with 0.25% PFA/0.05% Triton X-100 in CB buffer (10 mM MES, 138 mM KCl, 3 mM MgCl, 2 mM EGTA) and for 30 min at room temperature with 4% PFA in CB buffer. Permeabilization was done with 0.5% Triton X-100 in CB buffer for 5 min at room temperature. Cells were washed once with CB/0.1 M Glycine for 10 min and twice with TBS-T (0.1% Tween) for 10 min before staining with 132 nM Alexa-Fluor-488–Phalloidin in TBS-T+5% NGS+5% NHS for 60 min. Samples were washed 3×5 min with TBS-T and 2×2 min with TBS. 250 µl 0.1 µm Tetraspec beads (Life Technologies, Darmstadt, Germany) diluted 1:2000 were added per sample for 15 min at room temperature. After washing 2×2 min with TBS coverslips were mounted in Mowiol containing DABCO (25 mg/ml; Sigma-Aldrich) as anti-fading reagent. Structured illumination imaging was performed using a Zeiss Elyra PS1 system. 3D-SIM data was acquired using a 63×1.4 NA oil objective. 405, 488, 561, 642 100 mW diode lasers were used to excite the fluorophores together with a BP 420-480+LP 750, BP 495-575+LP 750, BP 570-650+LP 750 or LP 655 excitation filter, respectively. For 3D-SIM imaging the recommended grating was present in the light path. The grating was modulated in five phases and five rotations, and multiple z-slices with an interval of 110 nm were recorded on an Andor iXon DU 885, 1002×1004 EMCCD camera. Raw images were reconstructed using the Zeiss Zen 2012 software.

We thank Frank Bentzien for buffy coats, Oliver Fackler for FHOD1 constructs, Henry N. Higgs for INF2 antibodies and constructs, Elizabeth Luna for mRFP–supervillin, Andrea Mordhorst for expert technical assistance, Klemens Rottner for FMNL2B constructs, the UKE microscope facility (umif) for help with microscopy and image analysis, Adriaan Houtsmuller for use of the SIM microscope, and Martin Aepfelbacher for continuous support. This work is part of the doctoral theses of L.P. and L.T.