Non-canonical activity of the podosomal formin FMNL1γ supports immune cell migration

Unwarranted macrophage activation and infiltration of tissues has been implicated in a number of diseases, including diabetes, rheumatoid arthritis (RA), organ transplant rejection, atherosclerosis and cancer progression (Cookson, 1971; Coussens and Werb, 2002; Davies et al., 2013; Lech et al., 2012; Mantovani et al., 2013; Sebbag et al., 1997; Steinman and Cohn, 1973; Wasowska, 2010; Weisberg et al., 2003; Xu et al., 2003). Macrophages are capable of both amoeboid and mesenchymal migratory phenotypes (Van Goethem et al., 2010). For mesenchymal migration to extravascular tissues, macrophages employ adhesive structures termed podosomes, which are transmembrane junctions that structurally connect the extracellular matrix with the actin cytoskeleton via integrin receptors (Marchisio et al., 1988). Podosomes serve not only as anchor points to support adhesion and migration but are important for localized release of metalloproteinases that degrade extracellular matrix (Cougoule et al., 2010). Macrophage podosomes have been mostly studied in two-dimensional (2D) culture but have also been observed in 3D environments (Cougoule et al., 2010; Cougoule et al., 2012).

Macrophage podosomes comprise more than 200 different proteins (Cervero et al., 2012). The organization of podosomes includes a concentric circular arrangement of proteins around a central actin core with integrins at the membrane interface. In 3D observations, podosomes seem larger and not nearly as organized, appearing as foci of podosomal proteins near the tips of cellular projections (Van Goethem et al., 2011). Visualized with fluorescent 2D microscopy, the actin core appears as a unique entity; however, higher resolution imaging suggests that the core pillar of actin may form from the elements of the existing cortical actin (Luxenburg et al., 2007). The central core of thick actin fibers projects perpendicularly 1–2 µm from the adhesive surface as a pillar and is surrounded, primarily at the basal plane, by the ring of associated proteins that are largely excluded from the densest actin at the center (Chen, 1989; Van Goethem et al., 2011; Duong and Rodan, 2000; Luxenburg et al., 2007; Marchisio et al., 1988). Proteomic screens reveal that podosomes have many proteins in common with traditional adhesion structures. These include proteins involved in structural interactions with actin, signaling proteins and proteins acting as scaffolds by forming multiple protein–protein interactive sites (Cervero et al., 2012). Often the proteins involved in podosomes are related, but are not identical, to focal adhesion proteins; however, the purpose of these distinctions has not been elucidated. Some of these proteins are leukocyte-specific variants, such as Vav1, Pyk2 or leupaxin, which could be potential causes of differences with focal adhesions (Cervero et al., 2012; Duong and Rodan, 2000; Gao and Blystone, 2009). An important question in podosome biology, relevant to macrophage migration, is the origin of the actin at the podosome core. It is unknown if existing filaments are harnessed, if de novo formation is involved or if both occur in some combination for assembly and disassembly. Notably, multiple actin-modifying proteins have been located in the podosome and demonstrated to contribute to normal function, including WASp, Arp2/3 and various formins (Linder et al., 1999).

We have previously demonstrated that the formin FMNL1 localizes to the actin core of macrophage podosomes (Mersich et al., 2010). Formins are a family of actin-binding proteins that function as homodimers, dimerizing in a head-to-tail fashion at a conserved formin homology 2 (FH2) domain, and have been demonstrated to nucleate, elongate, cap, bundle and/or sever actin filaments (Copeland et al., 2004; Goode and Eck, 2007; Pruyne et al., 2002; Wallar and Alberts, 2003). Interestingly, FMNL1 has been reported to be capable of performing each of these actin filament regulating events in vitro (Esue et al., 2008; Harris et al., 2004; Harris et al., 2006). Three isoforms of FMNL1 have been identified in macrophages, including FMNL1α, FMNL1β and FMNL1γ; these isoforms deviate in sequence near the C-terminus (Han et al., 2013; Yayoshi-Yamamoto et al., 2000). In this report, we have developed the first murine knockout model of the formin FMNL1. Null mutations of Dia1, Dia2, DAAM1, FHOD3 and FMN1 have resulted in varying degrees of severity, ranging from non-viability to cellular dysfunction (Eisenmann et al., 2007; Kan-O et al., 2012; Li et al., 2011; Watanabe et al., 2013; Zhou et al., 2009). We determined that global depletion of FMNL1 is embryonic lethal; however, targeted depletion of FMNL1 from macrophages provided a model to investigate the functional contributions of FMNL1 in primary macrophages both in vitro and in vivo. We report here that the γ isoform of FMNL1 contributes to immune function by modulating podosome stability and subsequent macrophage migration in vitro and in vivo. Further, FMNL1γ does not require barbed-end actin-binding activity for these functions, suggesting an atypical mechanism of formin action.

Conditional FMNL1 knockout mice were generated using a targeting vector for the FMNL1 gene with 5′ and 3′ homology arms exceeding 5000 bases (Fig. 1A). This vector contained a β-galactosidase reporter cassette and a neomycin selection cassette between exons 3 and 4, flanked with FRT sites. Exons 4, 5 and 6 were flanked with loxP sites, with the 5′ loxP site located downstream of the FRT site. Insertion of the reporter and selection cassettes caused a frame shift, leading to an untranslatable transcript for the FMNL1 protein, thus causing a null mutation. The reporter and selection cassettes were removed through FLP-mediated recombination to alleviate the frame shift, leaving exons 4, 5 and 6 floxed, allowing for proper FMNL1 expression (Fig. 1B, second lane from the left, Cre absent). FMNL1 floxed (fFMNL1) mice bred with mice expressing Cre recombinase produces a frame shift mutation in the FMNL1 allele by excising exons 4, 5 and 6, resulting in targeted deletion of FMNL1 (Fig. 1B, second lane from the left, Cre present).

Crossing fFMNL1(+/−) / EIIa Cre (+/−) (heterozygous) mice failed to produce progeny of the fFMNL1(+/+) / EIIa Cre (+/+) (homozygous) genotype despite normal distribution of other genotypes. This was observed in all litters, totaling 28 pups from seven heterozygous crosses, thus complete gene deletion is deemed embryonically lethal at a 95% confidence interval (Adams et al., 2013). Necropsies provided no indication of staging for gestational failure of homozygous deletions. These results suggest an additional developmental role for FMNL1 that we did not explore further here.

fFMNL1 mice were then crossed with LysMcre mice to generate animals devoid of FMNL1 selectively in cells of the myeloid lineage. fFMNL1(+/−) / LysMcre(+/−) produced viable fFMNL1(+/+) / LysMcre(+/+) progeny at expected Mendelian ratios (Fig. 1C,D). Genotyping, as described in Materials and Methods, showed distinct homozygosity and/or heterozygosity of animals for FMNL1 and Cre (Fig. 1C). The fFMNL1(+/+) / LysMcre(+/+) mice, exhibiting no FMNL1 were used as experimental knockout (KO) animals, since fFMNL1(+/−) / LysMcre(+/−) mice or fFMNL1(+/+) / LysMcre(+/−) mice still expressed some measurable level of FMNL1 protein in comparison to fFMNL1(−/−) and LysMcre(−/−) (WT) mice (WT 100%, heterozygous/heterozygous 55.3%, homozygous/heterozygous 5.24%, KO 0.0%, Fig. 1D). Depletion of the FMNL1 protein for each genotype was verified by western blot analysis of lysates from purified murine bone-marrow-derived macrophages (BMDMS) (Fig. 1D). Macrophages from fFMNL1(+/+) / LysMcre(+/+) exhibited a complete loss of FMNL1 protein [Fig. 1D, (+/+)/(+/+)]. Expression of additional formin proteins within macrophages (Dia 1, Dia 2, FHOD1, FHOD3) was also analyzed using western blot analysis, demonstrating that depletion of FMNL1 did not affect expression levels of other macrophage formins (Fig. 1D).

Initial characterization experiments were performed on littermate mice that had been generated by crossing fFMNL1(+/−) / LysMcre(+/−) mice. KO mice [fFMNL1 (+/+) / LysMcre (+/+)] showed no deviation in gross phenotype, weight or sex distribution from WT littermates [fFMNL1(−/−) / LysMcre (−/−)] (data not shown). Additionally, KO mice were viable, fecund and were aged to 2 years alongside WT littermates with no detectable health abnormalities. Peripheral blood analysis was performed on WT and KO mice, (Fig. S1) and fFMNL1(+/−) / LysMcre(+/−) and fFMNL1(+/+) / LysMcre(+/−) littermates (data not shown), with no significant changes observed. Notably, peripheral blood counts of monocytes did not differ, and numbers were within normal ranges for all genotypes. Tissue morphology was analyzed between WT and KO mice using hematoxylin and eosin (H&E) staining of various tissues. No differences in tissue morphology were apparent between WT or KO tissue samples, including lung and liver (Fig. S1), as well as kidney, bone, skin, small intestine and large intestine (not shown), where residential macrophages can be located. Interestingly, although the overall tissue morphology was unaffected, immunofluorescent microscopy of tissue sections stained for F4/80 (also known as ADGRE1) revealed a 38.6% reduction in residential macrophages (Kupffer cells) in FMNL1 KO liver sections in comparison to in FMNL1 WT animals (Fig. 1E; Fig. S4A). This suggests that FMNL1 is important for establishment and/or maintenance of normal populations of hepatic tissue macrophages.

To determine whether the loss of FMNL1 affects macrophage immune migratory function, we induced experimental peritonitis with thioglycollate instillation in FMNL1 WT and KO mice (Ezekowitz, 1985). F4/80-positive macrophages that had been recruited into the peritoneal cavity were quantified, and we observed a 20.0% reduction in macrophages from FMNL1 KO mice compared to in WT mice (Fig. 1F; Fig. S4B). This suggested that FMNL1 contributes to efficient macrophage immune migration in vivo.

To evaluate whether other formins are involved in macrophage migration, we utilized the pan-formin inhibitor SMIFH2, as previously described (Miller and Blystone, 2015a). WT, fFMNL1(+/−) / LysMcre(+/−), fFMNL1(+/+) / LysMcre(+/−) and KO BMDMs were treated with 30 μM SMIFH2 and subjected to in vitro Transwell macrophage migration assays. Macrophages were challenged to transit a filter barrier in response to serum in the basal chamber. FMNL1 KO macrophages were much less efficient in their migration than either WT or heterozygous macrophages. While in vivo inflammatory migration of KO macrophages was reduced by 20%, the in vitro migration of KO macrophages was decreased by 50% in the Transwell migration assay, compared to wild type. Interestingly, all genotypes showed a reduction in migration between vehicle (DMSO) and SMIFH2-treated macrophages (Fig. 2A). However, WT, fFMNL1(+/−) / LysMcre(+/−) and fFMNL1(+/+) / LysMcre(+/−) macrophages, all of which express some FMNL1 (Fig. 1D) displayed a 52.2% reduction in migration when treated with SMIFH2. Migration of FMNL1-deficient macrophages, which was already half of that exhibited by the other genotypes, was suppressed an additional 14% by the inclusion of SMIFH2 (Fig. 2A). Contrasting with the differences in migration, adherent macrophage cultures from WT and KO mice revealed no variation in adhesion between these groups after 96 h (Fig. S3E). These results suggest a potential role for formin FH2 activity in macrophage migration, but this does not appear to be specific to FMNL1. Further, these data suggest that FMNL1 has a greater role in macrophage migration than other macrophage formins.

To further characterize the extent to which FMNL1 is involved in macrophage locomotion, we evaluated WT and KO macrophage random motility in a 2D environment. For this, we recorded videos of WT and KO macrophages that had been plated on glass-bottomed dishes in the presence of serum for 6 h. We observed no significant differences in either macrophage velocity or distance traveled (Fig. 2C), suggesting that FMNL1 may be important for directional movement – such as chemotaxis or in 3D environments – but that it is unnecessary for random motility. Surprisingly, we found no loss of phagocytic function in FMNL1 KO macrophages when non-specific, CR3-mediated or Fc-receptor-mediated routes of entry were examined, in contrast with other reports (Naj et al., 2013; Seth et al., 2006), which could be potentially explained by differences between FMNL1 KO and suppression using RNA interference (Fig. S2).

We have previously demonstrated that SMIFH2-mediated inhibition of formins causes a significant loss in the number of macrophage podosomes (Miller and Blystone, 2015a). We have also shown that siRNA-mediated silencing of FMNL1 significantly reduces the number of podosomes formed in macrophages (Mersich et al., 2010). Therefore, we evaluated whether knock out of FMNL1 impacts macrophage podosomes. WT and KO macrophages were fixed, permeabilized and stained for actin and podosome-associated proteins as previously described (Miller and Blystone, 2015b). Macrophages that formed podosomes were quantified as previously described (Mersich et al., 2010; Miller and Blystone, 2015a). FMNL1 KO macrophages exhibited a 47.8% decrease in the number of macrophages forming podosomes in comparison to WT (Fig. 2B; Fig. S4D). These data correlate well with previous studies in our laboratory using siRNA-mediated silencing of FMNL1 in human macrophages. Surprisingly, in remaining podosomes, we found no differences between WT and KO macrophages in height or width of the dense actin core of podosomes (Fig. 2D). This contrasts with our earlier findings and may be explained by the incomplete loss of FMNL1 in siRNA experiments or by differences between human peripheral blood monocytes and murine BMDMs.

Previous studies report that siRNA-mediated silencing of FMNL1 in prostate cancer cells (PC3) results in increased cell spreading (Vega et al., 2011). We quantified cell surface area between WT and FMNL1 KO macrophages using a membrane-permeable live-cell dye, and live-cell microscopy was used to capture images. FMNL1 KO macrophages showed a 12.6% increase in cell surface area in comparison to WT (Fig. S3C). Both WT and KO macrophages exhibited a polarized phenotype in culture, with an apparent leading edge and retracting tail (Fig. S4C).

Diaphanous formins are regulated by activated (GTP-bound) GTPases, and it has been suggested that FMNL1 is regulated specifically by the Rho GTPases Rac1, Cdc42 and RhoA (Seth et al., 2006, Wang et al., 2015; Yayoshi-Yamamoto et al., 2000). To determine if these GTPases regulate FMNL1, activated levels of Rac1, Cdc42 and RhoA were assessed in FMNL1 KO and WT macrophages by comparing GST–rhotekin-binding domain (RBD) and/or p21-activated kinase-binding domain (PBD) pull-downs with total GTPase expression as determined by western blot analysis. Interestingly, increases of 34.5%, 42.0% and 59.4% in the amount of activated Rac1, Cdc42 and RhoA, respectively, were observed in FMNL1 KO macrophages (Fig. S3). These data suggest that FMNL1 is a target for these activated GTPases as increased activity levels of these small effector proteins may reflect compensatory efforts in the absence of FMNL1.

To confirm that KO of FMNL1 disrupts macrophage migration, FMNL1 KO macrophages were reconstituted with GFP or GFP-tagged FMNL1 as previously described (Miller and Blystone, 2015b). We constructed lentivirus expressing GFP-tagged FMNL1β or FMNL1γ isoforms for reconstitution using cDNA from primary human macrophages. Fluorescent microscopy showed that GFP and GFP–FMNL1β distribute nonspecifically in the cytoplasm of FMNL1 KO macrophages, while GFP–FMNL1γ localized discretely to podosomes, as reported using staining of endogenous FMNL1 (Fig. 3A) (Mersich et al., 2010). Expression of tagged full-length FMNL1 isoforms was verified by western blotting of transduced FMNL1 KO macrophages, with expression of FMNL1γ being slightly less than that of FMNL1β, and expression of both being below that of endogenous levels in WT macrophages (Miller and Blystone, 2015b). Transduced macrophages, along with WT and untreated FMNL1 KO macrophages, were subjected to migration assays. Interestingly, the loss of macrophage migration seen in FMNL1 KO cells was fully rescued only by the GFP–FMNL1γ isoform (Fig. 3B). KO macrophages expressing GFP or GFP–FMNL1β showed decreased migration compared to WT, consistent with untreated FMNL1 KO macrophages, although limited but nonsignificant rescue was seen with FMNL1β. These data demonstrate that not only is proper macrophage migration dependent on the formin FMNL1, but it is also specific to the γ isoform. We also observed that reconstitution of KO macrophages with FMNL1 yielded an increased number of podosome-forming cells, compared to GFP transduction alone (Fig. 3C). Both FMNL1β and FMNL1γ expression increased podosome numbers in macrophage culture; however, only expression of FMNL1γ rescued podosome number significantly to be equivalent to that of WT macrophages (Figs 3C and 2B). These findings suggest that FMNL1γ localizes to podosomes and is necessary for their formation and/or maintenance.

After determining which isoform of FMNL1 localizes to podosomes and is required for normal macrophage migration, we wanted to gain insight into the mechanism of FMNL1 action at macrophage podosomes. We mutated two residues within the FH2 domain of FMNL1γ, Ile720 and Lys871, that have been predicted to be critical for proper actin binding at F-actin barbed-ends (Harris et al., 2006). Produced as a GST fusion protein, this actin-binding mutant (ABM) exhibited a loss of ability to precipitate actin from cell lysates. We inserted the FMNL1γ ABM into the lentiviral vector to produce GFP-tagged FMNL1γ ABM. FMNL1 KO macrophages were reconstituted with GFP, GFP-FMNL1γ or GFP-FMNL1γ ABM and, together with WT and untreated FMNL1 KO macrophages, challenged in migration assays. Remarkably, macrophage migration was fully rescued by the GFP-FMNL1γ ABM to levels equivalent to those with GFP-FMNL1γ (Fig. 3B). These observations suggest that barbed-end binding of the FH2 domain of FMNL1 is not necessary for proper macrophage migration.

We also analyzed macrophages transduced with GFP-FMNL1γ ABM using fluorescent confocal microscopy. We observed that GFP–FMNL1γ ABM restored podosome numbers in KO macrophages and localized to podosomes in a manner similar to GFP–FMNL1γ (Fig. 3A).

Using a conditional murine model for the genetic deletion of FMNL1, we report that FMNL1 serves a critical role in embryonic development and contributes to macrophage podosome formation, adhesion and migration both in vitro and in vivo. FMNL1 accounts for the majority of formin-dependent migration activity in macrophages.

FMNL1 floxed mice crossed with EIIa Cre mice demonstrated that global depletion of FMNL1 is lethal. Recent studies have implicated FMNL1 in myofibril maintenance, suggesting that FMNL1 may be a necessary formin for embryonic heart development (Rosado et al., 2014). Targeting LysMcre crosses were performed to generate mice containing FMNL1 knockout macrophages. Complete blood counts were surprisingly normal, indicating that FMNL1 is unlikely to serve a critical role in hematopoiesis. In addition to FMNL1, we have previously reported that other formins, such as Dia1 and FHOD1, are expressed in macrophages (Krainer et al., 2013), yet we observed no compensatory changes in the expression of other formins when FMNL1 was depleted from macrophages here. Strikingly, the tissue distribution of residential macrophages from knockout mice was significantly reduced compared to in WT, most notably in the liver. These results suggest that loss of FMNL1 in macrophages may reduce their ability to maintain strong adhesions with the sinusoidal lumen, as these cells are under constant mechanical stress from persistent fluid flow through the sinusoids (Mersich et al., 2010; Miller and Blystone, 2015a). Alternatively, genetic loss of FMNL1 may retard the movement of macrophages into tissues for establishment or maintenance of resident populations. The diminished population of Kupffer cells did not appear to have a detrimental impact on the unchallenged animal, indicating that a normal density of such cells may not be essential for survival under naïve conditions.

Knockout of FMNL1 significantly decreases macrophage migration across a porous barrier. However, doubly heterozygous macrophages showed no reduction in migration compared to WT macrophages, despite a partial loss of FMNL1 protein in macrophages from these heterozygous animals. This indicates that haploinsufficiency does not occur with the loss of one functional FMNL1 allele and that low levels of FMNL1 are sufficient to perform the necessary actin-regulating events in macrophages.

Directed macrophage migration is facilitated by the presence of signaling and environmental cues, sometimes presented in a gradient (Dong et al., 2003; Linder et al., 1999). In the absence of chemoattractant gradients, macrophages exhibit random motility, which does not rely on strong stable adhesions that occur at podosomes, as demonstrated previously in WASp (−/−) macrophages that are incapable of forming podosomes (Linder et al., 1999). Rather, actin is regulated at cellular locations during random motility, including at the lamellipodium, lamellum, stress fibers and filopodia, all of which have been shown to have associated formins in cell types other than macrophages (Chesarone et al., 2010; Kuün and Geyer, 2014; Rottner and Stradal, 2011). Directional in vitro and in vivo migration assays showed a clear requirement for FMNL1; however, no differences were observed in random motility studies between WT and KO macrophages. These data suggest that FMNL1 is specifically important for modulating actin at strong adhesion sites in macrophages (podosomes), yet is not necessary for regulating actin at other sites during random motility.

Two independent studies have suggested that FMNL1 is required for different types of macrophage phagocytosis (Naj et al., 2013; Seth et al., 2006). FMNL1 KO macrophages did not differ from WT in their ability to perform non-specific, CR3-mediated or FcγR-mediated phagocytosis. It is possible that differences in macrophage preparation or, more likely, the RNA interference approach used in other studies produced unique effects that are not seen in KO macrophages. Our results only support a role for FMNL1 in macrophage podosome-dependent migration.

The number of podosomes observed was significantly reduced in adherent KO macrophages in comparison with WT. This observation is in agreement with previous studies in our laboratory in which knock down of FMNL1 in human macrophages with siRNA also resulted in a loss of podosomes (Miller and Blystone, 2015b). Loss of podosomes has been observed in macrophages isolated from Wiskott-Aldrich syndrome patients, which are deficient in WASp protein expression, and these macrophages also exhibit reduced directed migration (Linder et al., 1999). These data suggest that FMNL1 is required for both normal dynamics of podosomes and proper podosome function.

Diaphanous-related formins are regulated through interactions of their GTPase-binding domain with active GTPases (Wasserman, 1998). GTPases spatially and temporally regulate formins, and influence formin localization, and comprehending these processes is extremely important for understanding how formins regulate the actin cytoskeleton. In this study, we demonstrate that FMNL1 KO macrophages have increased activation of RhoA, Rac1 and Cdc42 compared to WT. These observations support previous studies indicating that FMNL1 is fairly promiscuous in the GTPases with which it interacts (Gomez et al., 2007; Seth et al., 2006; Wang et al., 2015; Yayoshi-Yamamoto et al., 2000). Coincidentally, we also observed an increase in surface area in macrophages that lacked FMNL1 expression in comparison to controls. This has also been demonstrated in previous studies of PC3 cells following siRNA-mediated silencing of FMNL1, supporting our observation (Vega et al., 2011). It is possible that KO macrophages attempt to compensate for the loss of actin cytoskeleton regulation by increasing the number of active GTPases associated with FMNL1. It is possible that FMNL1 has other roles unrelated to podosomes that require unique GTPases, but any non-podosomal localization of FMNL1 escaped our detection capability.

We demonstrate that reconstitution of KO macrophages with the full-length FMNL1γ isoform specifically rescues directional macrophage migration across a porous barrier. Additionally, we observed that the FMNL1γ isoform localizes to KO macrophage podosomes and rescues their podosome population. These observations strongly suggest that macrophage migration and actin dynamics at podosomes are regulated specifically by the FMNL1γ isoform and, furthermore, that either localization to podosomes or a specific FMNL1 behavior is solely due to the tail-end splice variation of this isoform. Two residues within the FH2 domain of FMNL1, Ile720 and Lys871, have been demonstrated to be critical residues for binding of actin by the FH2 domain (Colón-Franco et al., 2011). To determine whether actin filament barbed-end binding by the FH2 domain of FMNL1γ is important for actin modulation at macrophage podosomes, a mutated full-length FMNL1γ construct, lacking the barbed-end actin binding capability, was introduced to rescue KO macrophages. FMNL1γ lacking these FH2 motifs fully reconstituted FMNL1γ activity for both macrophage migration and podosome formation to levels seen in WT macrophages. This suggests that F-actin barbed-end binding by the FH2 domain of FMNL1γ is dispensable for its function and localization in macrophage podosomes. Combined with the rescue by the γ isoform, the longest variant of FMNL1, we suggest that an atypical formin behavior modulates podosome actin dynamics, possibly via actin regulatory activity within its C-terminal WH2-like motif.

Our data and other reports suggest three possibilities for FMNL1γ modulation of the actin cytoskeleton at macrophage podosomes: (1) WH2-like motifs of FMNL1γ are necessary, (2) electrostatic interactions of the outside of the FH2 domain in FMNL1γ are required, or (3) both of these interactions are working in tandem. Studies on similar WH2-like motifs in FMNL3 have reported that these regions have the capacity to bind to barbed ends of actin filaments and slow the barbed-end elongation in monomeric and dimeric form, even when FH2 binding to actin is abolished (Heimsath and Higgs, 2012). Thus, FMNL1γ may be important for F-actin barbed-end association at macrophage podosomes via its WH2-like motifs to regulate barbed-end elongation. In vitro biochemical studies of FMNL1 have proposed a novel mechanism of F-actin interaction that is utilized for bundling filaments and is not disrupted when conserved essential barbed-end associating residues of the FH2 domain are mutated (Harris et al., 2006). Therefore, FMNL1γ could be required for bundling actin filaments at the macrophage podosome. This seems feasible since FMNL1 is located at the dense actin core where bundling would occur. Furthermore, we have demonstrated that partial reduction of FMNL1 using RNAi causes any remaining podosomes to become wider in diameter, suggesting that this loss of FMNL1 bundling causes the tightly bundled core filaments to disperse, although this was not seen previously in KO macrophages (Mersich et al., 2010). WH2-like motifs are reported to be important for actin filament severing, including the WH2-like motifs in FMNL3, which are similar to those found in FMNL1 (Heimsath and Higgs, 2012; Ramabhadran et al., 2012; Vaillant et al., 2008). Importantly, mutations of the FH2 domain residue associated with barbed-end binding are not reported to significantly impact severing capacity, unlike mutation of the WH2-like motif (Heimsathand Higgs, 2012). Actin filament severing seems to be the most likely function of FMNL1γ at macrophage podosomes since FMNL1 has been reported to function primarily as a severing protein (Harris et al., 2004). Therefore, we hypothesize that FMNL1γ associates with the sides of actin filaments via electrostatic bonds with residues located on the ‘outside’ of the FH2 domain and utilizes its WH2-like motifs to promote severing of actin filaments. This correlates with observations implicating WASp and Arp2/3 in podosome actin assembly and in the potential reorganization of the existing cortical actin cytoskeleton.

This is the first demonstration, of which we are aware, of a cellular formin function that can be rescued by an FH2 domain actin-binding mutant. This study also reaffirms that the reported unique ability of FMNL1 to bind to actin filament sides in vitro is also likely to occur in vivo. Our findings provide avenues for future work, including the study of the dynamics of FMNL1 activity in podosomal actin structures as well as the spatial roles of the γ isoform in both FMNL1 localization and its interaction(s) with actin.

All murine experiments were performed under Institutional Animal Care and Use Committee approved protocols. Conditional FMNL1 knockout mice were generated using standard methods (Hall et al., 2009). Briefly, homologous recombination occurred in C57/BL6 ES cells that contained an FMNL1-targeting replacement vector comprised of a 5′ homologous arm of 5525 bp containing FMNL1 exons 2 and 3 and a 3′ homologous arm of 5956 bp containing FMNL1 exons 4–9, flanking the neomycin selection cassette (Fig. 1A). Exons 4, 5 and 6 were flanked with loxP sites, and the reporter and selection cassettes were flanked with short flippase recognition target (FRT) sites. The β-galactosidase reporter cassette and neomycin selection cassette were removed by utilizing flippase-mediated (FLP) recombination through breeding heterozygous FMNL1 floxed mice with FLP-expressing mice [B6.Cg-Tg(Pgk1-FLPo)10Sykr/J]. Viable mice with inserted loxP sites flanking exons 4, 5 and 6 of the FMNL1 gene (floxed) were backcrossed eight times with C57BL/6 mice before beginning crosses with Cre-recombinase-expressing mice. Homozygous FMNL1 floxed mice were crossed with mice expressing Cre recombinase under the adenovirus Ella promoter [B6.FVB-Tg(Ella-cre)C5379Lmgd/J] for widespread deletion of FMNL1. Additionally, FMNL1 floxed mice were crossed with mice expressing Cre recombinase under the Lyz2 promoter (B6.129P2-Lyz2^tm1(cre)Ifo/J), commonly called LysMCre mice, to generate viable mice with myeloid cell-specific depletion of FMNL1. All mice used were generated on a C57BL/6 background. Genotypes of experimental animals were determined through PCR on toe DNA and also verified by PCR on purified BMDM cultures. Genotyping for Cre recombinase expression (Lyz2 and EIIa) was performed in accordance with the protocol and primers recommended by The Jackson Laboratory (Bar Harbor, ME). Floxed FMNL1 mice were genotyped using four PCR primers. Wild-type alleles were amplified using forward primer Fmnl1-wtF2 (5′-GCTTACATGGGCATTCCTGTATGC-3′) and reverse primer Fmnl1-wtR2 (5′-AGGGTGGGTGCCTACCTACATGC-3′). FMNL1 floxed alleles and Cre-mediated excision events of exons 4, 5 and 6 were determined using forward primer CSD-flpF (5′-CGCATAACGATACCACGATATCAACAAG-3′) and reverse primer CSD-flpR2 (5′- CCTGAACTCAGGAGTGTCAGC-3′).

Immunofluorescent microscopy of resident macrophage populations in the liver (Kupffer cells) was performed as described previously (Watson et al., 2015). Dissected liver sections were snap-frozen in optimal cutting temperature (OCT) medium (Sakura Finetek USA, Torrance, CA) and stored at −80°C. Liver sections were cut using a cryostat (Leica Microsystems CM1900, Wetzlar, Germany) at 6 μm, adhered to glass slides and stored at −80°C. Before staining, sections were fixed and permeabilized using a [50:50] mixture of ice-cold acetone and methanol for 20 min at −20°C. Sections were air dried and blocked in 10% goat serum in PBS overnight at 4°C, followed by overnight incubation with 10% goat serum containing an anti-F4/80 antibody (1:2000; cat. no. 123115, BioLegend, San Diego, CA) at 4°C. Tissue sections were incubated in 10% goat serum containing TRITC-conjugated secondary (1:500; Jackson ImmunoResearch) at room temperature for 2 h, followed by a 5 min incubation with DAPI (0.0001 μg/ml) (Invitrogen, Carlsbad, CA) in PBS. Anti-fading mounting solution was used to cover sections with glass coverslips. Microscopy was performed on a Nikon Eclipse E800 fluorescent microscope equipped with a Hamamatsu ORCA-ER digital camera. Images were analyzed using NIS-Elements software to quantify F4/80-positive cells. Absolute numbers of F4/80 cells within a ×60-magnified field were counted.

Purified cultures of murine BMDMs were subjected to migration assays as previously described (Miller and Blystone, 2015a,b). Macrophages were washed and resuspended in Dulbecco's modified Eagle's medium (DMEM) with penicillin-streptomycin (serum free). For some studies, macrophages were treated with either vehicle control, DMSO (Sigma-Aldrich) or the formin inhibitor SMIFH2 (Sigma-Aldrich). SMIFH2 was diluted in complete medium to yield a final concentration of 30 μM in cell suspensions that were used for the migration assay. Macrophages were counted and used in the Transwell migration assay at 10⁵ macrophages per Transwell insert following the manufacturer's protocol. Complete medium was added to the lower chamber, and cells were allowed to migrate for 16 h at 37°C. Medium was removed from the upper well of the inserts, and insert membranes were fixed, dried and stained using a differential staining kit. The inserts were dried overnight and mounted onto glass slides. Transmitted light microscopy was performed on the insert membranes to manually count all cells within five random ×60-magnified fields per sample.

Inflammatory macrophages were harvested after inducing peritonitis using methods adapted from Ezekowitz (1985). Peritonitis was induced by injecting 1.0 ml of 4% thioglycollate into the peritoneal cavity. At 72 h post injection, cells were harvested from the peritoneal cavity, counted and plated onto glass coverslips in 24-well plates at 2×10⁶ cells per well in 1 ml DMEM with 10% FBS, 1% penicillin-streptomycin and 10 mM L-glutamine supplement. Harvested peritoneal cells were adhered to glass coverslips via centrifugation at 1250 g with a slow braking for 20 min at 37°C and then fixed and permeabilized as described above. Macrophages were stained with anti-F4/80 and TRITC-conjugated secondary antibodies to identify macrophages, and DAPI staining to determine total cell population. At least 2000 cells were counted per sample, and the percent macrophages is reported.

Macrophages were prepared for microscopy as described previously (Miller and Blystone, 2015b). Macrophages were fixed, permeabilized and stained with Rhodamine–phalloidin (Cytoskeleton, Denver, CO) at 1:10,000 in PBS for F-actin visualization. Macrophages adhered to coverslips were mounted on glass slides using an anti-fade solution. Images were captured using a Nikon Eclipse E800 fluorescent microscope fashioned with a Hamamatsu ORCA-ER digital camera, and analyzed using NIS-Elements software for quantification of macrophages exhibiting podosomes and for measuring podosome diameter (Mersich et al., 2010). Podosome height was imaged using a Nikon Ti microscope equipped with a Perkin Elmer spinning disk laser (Perkin Elmer) and quantified with Volocity Quantitation software.

Macrophages were plated at 3.0×10⁴ cells per dish onto 35-mm glass-bottomed dishes (Mat-Tek) in complete medium as described above and incubated at 37°C under 5% CO₂ for 24 h. Macrophage motility was analyzed over the course of 6 h, capturing bright-field images every 10 min using a Leica AF6000 LX deconvolution microscope (Leica) equipped with a LUCA R EMCCD camera (Andor, South Windsor, CT). Macrophage motility analysis was performed for the distance traveled and velocity using ImageJ software.

Human macrophages were derived from peripheral blood monocytes as described previously (Miller and Blystone, 2015a). Whole-cell RNA was extracted and purified from primary human macrophages with the RNeasy Mini Kit (Qiagen, Hilden, Germany). Primary human macrophage cDNA was generated from RNA using the SuperScript III First-Strand Synthesis System for reverse-transcriptase (RT)-PCR (Life Technologies). Full-length FMNL1 was cloned into a previously generated pEGFP vector from primary human macrophage cDNA (Chandoke et al., 2004). FMNL1α (accession: NM_005892), FMNL1β (accession: BC001710) and FMNL1γ (accession: FJ534522) were cloned from primary human macrophage cDNA via PCR. A NotI site was introduced into each isoform following the stop codon at the C-terminus via PCR mutagenesis and was used in conjunction with a BsiWI site for transfer to pEGFP-FMNL1. The N-terminal XbaI site and C-terminal BamHI site were used to transfer FMNL1β and FMNL1γ into pLVX-AcGFP-C1 (Clontech). All FMNL1 constructs were verified by DNA sequencing.

Two substitution mutations, I720A and K871A, were generated to eliminate actin binding by PCR ‘sewing’ into the FMNL1γ isoform. For the I720A mutation, primers used for the first reaction were forward primer 5′-CCGAATGCCACTCTTGAACTGGGTGGC-3′ and reverse primer 5′-CGCAGGGTGGCGGCCAAGTTCTTGGCCCGG-3′. The second reaction primers were forward primer 5′-CGGGCCAAGAACTTGGCCGCCACCCTGCG-3′ and reverse primer 5′-CGAGGCTGATCAGCGGGTTTAAACG-3′. The combining reaction primers were forward primer 5′-CCGAATGCCACTCTTGAACTGGGTGGC-3′ and reverse primer 5′-CGAGGCTGATCAGCGGGTTTAAACG-3′. For the K871A mutation, the primers used for the first reaction were forward primer 5′-CCGAATGCCACTCTTGAACTGGGTGGC-3′ and reverse primer 5′-GCTTGCGATCAGTCGAGGCCATCTCCAACAGC-3′. The second reaction primers were 5′-GCTGTTGGAGATGGCCTCGACTGATCG-3′ and reverse primer 5′-CGAGGCTGATCAGCGGGTTTAAACG-3′. The combining reaction primers were forward primer 5′-CCGAATGCCACTCTTGAACTGGGTGGC-3′ and reverse primer 5′-CGAGGCTGATCAGCGGGTTTAAACG-3′. Nucleotides in bold indicate introduced changes to the sequence. The PCR product containing both substitution mutations was inserted into the pLVX-AcGFP-FMNL1γ lentiviral vector between SfiI and NotI sites. The GFP–FMNL1γ actin-binding mutant (ABM) was transformed into DH10β competent cells (Life Technologies), and mutations were confirmed by DNA sequencing. Constructs encoding GST fusion proteins of FMNL1γ and the ABM were prepared by performing PCR. For solubility, these constructs encode FMNL1 beginning with the FH1 domain and continuing through to the C-terminus. Fusion proteins were grown in Escherichia coli using standard techniques and harvested with glutathione–agarose. Agarose-bound FMNL1γ and FMNL1γ ABM were incubated with lysate from MDA-MB-231 cells, washed six times in PBS and precipitates were subjected to SDS-PAGE and western blotting to demonstrate loss of actin binding in the ABM mutant (Fig. S3D).

Macrophages were infected with lentivirus coding for either GFP, GFP–FMNL1β, GFP–FMNL1γ or GFP-FMNL1γ ABM as described previously (Miller and Blystone, 2015b). Infected macrophages were subjected to migration assays and analyzed as described above. In parallel, podosomes of infected macrophages were analyzed using fluorescent microscopy (Miller and Blystone, 2015b). Modifications to the previously described methods include staining with Rhodamine–phalloidin at 1:10,000 for actin visualization, and using a SP5 scanning confocal microscope (Leica, Bannockburn, IL) equipped with hybrid photomultiplier tube detectors. Image analysis was performed using the Leica Application Suite Advanced Fluorescence (LAS AF) software. For podosome rescue experiments, virally infected BMDMs were differentiated with recombinant murine granulocyte macrophage colony-stimulating factor (GM-CSF) (1:2000; R&D Systems, Minneapolis, MN).

Macrophages (WT, n=69; KO, n=70) were treated with vital cell dye Calcein-AM, as described by Denholm and Stankus (1995), and plated onto 35-mm glass-bottomed dishes at 10⁶ cells per dish in 2 ml complete medium. Dishes were incubated at 37°C under 5% CO₂ overnight. Medium was replaced with 2 ml fresh complete medium and live fluorescent cell images were taken using a Nikon Ti microscope equipped with a Perkin Elmer spinning disk laser. Volocity Quantitation software was utilized to quantify surface area of the macrophages.

Macrophages were plated at 2.5×10⁴ cells per well in 1 ml complete medium and allowed to adhere for 24 h as described above. Medium was replaced with 500 μl of fresh complete medium, and 10 μl of 0.4% Trypan Blue (Sigma-Aldrich) in PBS was added to each well, and macrophages were incubated at 37°C under 5% CO₂ for 48 h (Mikhael and Evens, 1975). Macrophages were fixed and mounted on glass slides using immersion oil. Differential interference contrast (DIC) microscopy was performed on a Nikon Eclipse E800 fluorescent microscope equipped with a Hamamatsu ORCA-ER digital camera. Images were analyzed using NIS-Elements software.

Macrophages were plated as described above. Medium was replaced with 500 μl of fresh serum-free medium, and cells were serum-starved at 37°C under 5% CO₂ for 1 h. After serum starvation, 4 μg of opsonized E. coli BioParticles conjugated to Alexa Fluor 488 (Molecular Probes, Eugene, OR) was added to each well, and macrophages were incubated at 37°C under 5% CO₂ for 3 h following the manufacturer's protocol. Non-internalized bioparticle fluorescence was quenched by aspirating medium and incubating macrophages with 300 μl of 0.4% Trypan Blue in PBS per well for 1 min at room temperature. Macrophages were fixed, permeabilized, stained with Rhodamine-phalloidin for F-actin visualization, mounted on glass slides with anti-fade solution and analyzed as described above.

Macrophages were plated as described above. Medium was replaced with 500 μl of fresh complete medium. 25 μl of latex beads coated with FITC-labeled rabbit IgG (Cayman Chemical, Ann Arbor, MI) was added to each well, and cells were incubated at 37°C under 5% CO₂ for 24 h following the manufacturer's protocol. Non-internalized fluorescent beads were quenched as described above. Macrophages were fixed, permeabilized, stained with Rhodamine–phalloidin for F-actin visualization and mounted on glass slides with anti-fade solution as described above. Images were captured and analyzed as described above.