The F-actin-binding RapGEF GflB is required for efficient macropinocytosis in Dictyostelium Free

Macropinocytosis is a highly conserved endocytic process dependent on the actin cytoskeleton (Bloomfield and Kay, 2016; Swanson, 2008). It involves non-specific endocytosis of solute molecules, nutrients and antigens. It is also a pathologically important process; several pathogens invade host cells via macropinocytosis (Mercer and Helenius, 2012), and cancer cells can use this pathway to scavenge extracellular nutrients (Commisso et al., 2013; Kamphorst et al., 2015). Although it has biological importance, macropinocytosis is not well understood.

A social amoeba, Dictyostelium discoideum, is a professional phagocyte that engulfs bacteria as a food source. Nevertheless, because cultivation on bacteria is unsuitable for biochemical and molecular biological analyses, most laboratories use axenic strains carrying mutations that dramatically upregulate macropinocytosis, which allow them to grow in axenic liquid culture (Sussman and Sussman, 1967). Most large membrane protrusions in such mutant cells are crowns (macropinocytic cups) in the liquid culture (Veltman et al., 2014). Recently, Bloomfield et al. revealed that the loss of the RasGAP protein NF1 is responsible for this phenotype (i.e. NF1 mutations cause increased fluid-phase uptake) (Bloomfield et al., 2015). Because most mechanisms and signaling pathways of macropinocytosis in Dictyostelium are also conserved in mammalian cells (Maniak, 2002), Dictyostelium is a useful model for studying macropinocytosis. Accumulating evidence has suggested that Ras and its effectors, class-I phosphoinositide 3-kinases (PI3Ks), regulate early events of crown formation (Bar-Sagi and Feramisco, 1986; Hoeller et al., 2013; Sasaki et al., 2007). PI3K phosphorylates the membrane lipid phosphatidylinositol 4,5-bisphosphate (PIP2) to produce phosphatidylinositol 3,4,5-trisphosphate (PIP3) (Hoeller et al., 2013). PIP2 and PIP3 are ether-linked plasmanyl inositols in Dictyostelium (Clark et al., 2014). PIP3 is a strong inducer of membrane protrusion (Funamoto et al., 2001; Huang et al., 2003; Parente et al., 1998). It activates Rho GTPases that in turn activate the Scar/WAVE complex and Wiskott–Aldrich syndrome protein (WASP), which act as drivers of actin polymerization and concomitant membrane ruffling by enhancing the actin nucleation by the Arp2/3 complex (Pollitt and Insall, 2009; Rivero and Somesh, 2002). Additionally, Diaphanous-related formin G, regulated by Ras, is also important for actin assembly in crowns and phagocytic cups (Junemann et al., 2016). Recently, it was proposed that self-organizing patches of active Ras and PIP3 form cup-shaped plasma membrane structures (Veltman et al., 2016).

Rap1 belongs to the Ras subfamily of small GTPases (Mitin et al., 2005; Wennerberg et al., 2005). Small GTPases of the Ras superfamily have two states, a GTP-binding state (‘on’) and a GDP-binding state (‘off’). Guanine nucleotide exchange factors (GEFs) catalyze the exchange of GDP with GTP, thereby activating Ras GTPases, whereas GTPase-activating proteins (GAPs) stimulate GTPase activity and inactivate them. Rap1 is present in mammalian cells as well as in D. discoideum (Kortholt and van Haastert, 2008) and is important for substrate and cell–cell adhesion (Bos et al., 2003; Kortholt et al., 2006), chemotaxis (Jeon et al., 2007a,b), cytokinesis (Dao et al., 2009; Plak et al., 2014), phagocytosis (Chung et al., 2008; Seastone et al., 1999) and other processes that involve the regulation of actin cytoskeletal dynamics (Kang et al., 2002; Schmölders et al., 2017). In Dictyostelium, Rap1 may also have important roles during macropinocytosis because active Rap1 localizes to crowns (Plak et al., 2014), and cells overexpressing WT and constitutively active Rap1 showed reduced rates of macropinocytosis (Seastone et al., 1999). However, the role of Rap1 during macropinocytosis remains unclear. Studies of sequence homology have revealed the existence of a large number of putative regulatory proteins of Ras subfamily GTPases: 25 RasGEFs, and 16 RasGAPs or RapGAPs in Dictyostelium (Kortholt and van Haastert, 2008; Wilkins et al., 2005). Because these numbers are larger than those of Ras GTPases themselves (12 Ras GTPases and three Rap GTPases), target GTPases for these regulators would be expected to overlap and each regulator might exert its function in distinct cellular processes.

The GEF-like protein B (GflB) of D. discoideum is a recently identified Rap1GEF involved in chemotaxis (Liu et al., 2016; Senoo et al., 2016). During chemotaxis, GflB acts as an effector of Gα2 subunits: active Gα2 releases an auto-inhibition of GflB and in turn GflB exerts GEF activity towards Rap1 (Liu et al., 2016). In addition to the RasGEF domains (RasGEFN and RasGEFC), GflB also has an N-terminal lipid-binding and inactive Rho GAP domains that can interact with several Rac GTPases, including Rac1, RacB, RacE and RacL, regardless of its activation status (Liu et al., 2016; Senoo et al., 2016). These domains may contribute to the correct localization of GflB. Here, we studied GflB functions in vegetative cells and revealed that GflB is involved in macropinocytosis, phagocytosis and cytokinesis. In gflB-knockout (KO) cells, crown dynamics was impaired, causing defects in cell morphology and macropinocytosis. Additionally, we observed that a short region of GflB at the extreme N-terminal end of GflB, showing high sequence similarity to villin headpiece (VHP) domains, directly interacts with filamentous (F)-actin and contributes to the correct subcellular localization of GflB.

Dictyostelium GflB has domains with sequence homology to the RhoGAP, RasGEF N-terminal domain (RasGEFN), and RasGEF catalytic domain (RasGEFC) (Wilkins et al., 2005) (Fig. 1A). GflB also carries a domain motif, the nenkyrin domain (NKD), at its C-terminus showing weak similarity to the novel protein nenkyrin (dictybase ID: DDB_G0282471, our unpublished data). We also identified a domain enriched in basic amino acids (depicted as B in Fig. 1A).

To assess GflB function, we first disrupted the gflB gene (see Materials and Methods). There were 9 clones out of 68 transformants examined that were confirmed as gflB-KO strains by PCR. All KO strains showed severely decreased growth rate and a more flattened and polarized cell shape in plate culture compared with the parental strain AX2 (Fig. 1B,C). The KO strain HA201 was used as the representative gflB-KO strain in further experiments shown below (Fig. 1B). Next, we constructed fusion proteins with GFP linked to the N-terminus of GflB (GFP–GflB) and overproduced it both in AX2 and gflB-KO strains. Both strains were more rounded in plate culture than AX2 cells (Figs 1B,C; 8B,C). Because overproducing GFP–GflB rescued the morphological defect in gflB-KO cells, it is likely that GFP–GflB is functional. The fact that both loss and overexpression of gflB resulted in abnormal cell shape suggests that GflB has an essential role in regulating cell morphology.

Next, to examine whether loss and overproduction of GflB influence basal cell motility, we observed the motility of vegetative cells on plates using ImageJ software 1.47v (National Institutes of Health, Fig. 1D and Movies 1–3). The average velocity of cell migration in gflB-KO cells was comparable to that in AX2 cells, but it increased in AX2 cells overproducing GFP–GflB [hereafter denoted AX2 (GFP–GflB)] (Fig. 1E; Fig. S1). The directionality of cell migration was comparable in AX2 (GFP–GflB) and was reduced in gflB-KO cells relative to AX2 cells (Fig. 1F and Fig. S1). We also examined the actin-driven large membrane protrusions including crowns and pseudopods (arrowheads in Fig. 1D) in gflB-KO and AX2 (GFP–GflB) cells. Crowns (macropinocytic cups) are protrusions created by a hollow ring of actin polymerization, whereas pseudopods are those created by a solid block of actin polymerization. In this experiment, they could not be distinguished from each other. The rates of protrusion formation were lower in both gflB-KO and AX2 (GFP–GflB) cells than in AX2 cells (Fig. 1G). Nevertheless, the number of protrusions per cell was higher in gflB-KO cells and reduced in AX2 (GFP–GflB) cells compared with that in AX2 cells (Fig. 1H). The lifetimes of protrusions in gflB-KO cells were longer than in AX2 cells (Fig. 1I), thus accounting for the increased number of protrusions despite a reduced rate of formation. In gflB-KO cells, ∼10% of protrusions kept protruding over 3 min, a duration rarely observed in AX2 cells. In contrast, the lifetimes of protrusions in AX2 (GFP–GflB) cells were shorter than in AX2 cells. This shorter lifetime and reduced formation rate are consistent with the lower number of protrusions in AX2 (GFP–GflB) cells than in the parent strain.

AX2 cells are axenic mutants, which dramatically upregulate macropinocytosis (Sussman and Sussman, 1967). It was reported that most of these protrusions in AX2 cells are crowns and not pseudopods in liquid growth culture (Veltman et al., 2014). To examine macropinocytosis alone, we added tetramethylrhodamine isothiocyanate (TRITC)–dextran into the growth medium and observed the living AX2 and gflB-KO cells using confocal microscopy (Fig. 2A; Movies 4–6). Almost all of these membrane protrusions were crowns in both AX2 and gflB-KO cells (Fig. 2A). However, some defects were observed in gflB-KO cells. First, although some crowns were newly formed before former protrusions at the same site had been completely retracted in both cells, such repetitive crown formation was much more frequent in gflB-KO cells (Fig. 2A, bottom, see Movie 6, and 2B). Repetitive crown formation itself may not be a defect because it is reported that new crowns predominantly form by splitting from existing ones (Veltman et al., 2016). In addition, both time from the initiation of extension to the closure of cups (Fig. 2C) and particularly that from cup closure to the completion of retraction of cup (Fig. 2D), were longer in gflB-KO cells than in AX2 cells. In some crowns of gflB-KO cells, after the macropinosomes had entered into the cell bodies, the protrusions persisted and remained elongated for a long time (Fig. 2A, bottom; Movie 6). Next, we examined PIP3 localization in AX2 and gflB-KO cells by using the PIP3 biosensor PHcrac (the CRAC PH domain) conjugated to GFP (Parente et al., 1998) (Fig. 2E; Movies 7 and 8), because PIP3 is strongly associated with macropinocytosis (Posor et al., 2013; Veltman et al., 2014). There were no apparent differences in the localization pattern of PHcrac–GFP during macropinocytosis between AX2 and gflB-KO cells: PHcrac–GFP localized throughout the peripheral regions of cups and early pinosomes. The only difference was that the PIP3-marked membrane ruffles tended to be slightly larger in gflB-KO cells than in AX2 cells prior to the closure into pinosomes (Fig. 2F). Collectively, these data suggest that GflB is important for macropinocytosis, mainly in its completion.

Because gflB-KO cells showed defects in crown dynamics, we examined the rates of macropinocytosis in gflB-KO cells. The uptake rates of TRITC–dextran in suspension were reduced in gflB-KO cells (Fig. 3A). We also examined the rates of phagocytosis and revealed that the rates of TRITC-labeled yeasts in suspension were also reduced in gflB-KO cells (Fig. 3B). These reduced rates in gflB-KO cells can be explained by defects in crown or phagocytic cup dynamics mentioned above (Fig. 2).

GflB is involved in multiple actin cytoskeleton-based cellular processes. These processes depend on precise local changes in actin polymerization and depolymerization; hence, we analyzed the subcellular distribution of F-actin in gflB-KO cells by fixing and staining for F-actin with Alexa Fluor^® 488–phalloidin. The gflB-KO cells were more intensely stained than AX2 cells during both interphase and cell division (Fig. 3C–E), suggesting that the amount of F-actin increased in gflB-KO cells. Second, we analyzed the subcellular distribution of F-actin biochemically by fractionation. Cells were lysed with 1% Nonidet P-40 and fractionated with a low-speed centrifugation (Fig. 3F). The 1% Nonidet P-40-insoluble fraction consists mainly of cytoskeletal ghosts derived from the cortical actomyosin meshwork (Kuczmarski et al., 1991). The total cell lysate (T), soluble fraction (S) and insoluble fraction (P) were separated with SDS-PAGE and actin bands quantified via Coomassie Brilliant Blue (CBB) staining (Fig. 3F) and densitometry (Fig. 3G,H). In gflB-KO cells, the total amount of actin was comparable to that in AX2 cells, whereas the amount of actin in the soluble fraction was reduced and the amount in the insoluble fraction increased (Fig. 3G). Thus, the amount of F-actin in ghosts was higher in gflB-KO cells (Fig. 3H). This increase in cortical F-actin in gflB-KO cells was reversed in cells overproducing GFP-GflB (Fig. 3F–H). Taken together, microscopy and biochemical results indicate that the amount of F-actin is elevated in gflB-KO cells.

Because Rap1 is also involved in cytokinesis (Plak et al., 2014), we examined whether an alternation in GflB expression affects this process by using time-lapse video analyses (Fig. S2; Movies 9–13). Most gflB-KO cells appeared to initiate cell division normally; cells became rounded, then elongated and formed a cleavage furrow at the equatorial region. However, 24.4% of gflB-KO cells examined (8 of 33) exhibited failed cytokinesis at abscission (the final detachment of daughter cells, Fig. S2B; Movie 10). In contrast, only 7.3% of AX2 cells (4 of 55) failed at abscission (indicating higher successful cytokinesis, Fig. S2A; Movie 9). Furthermore, careful observation of cellular movement during cytokinesis revealed that whereas two AX2 daughter cells migrated in opposite directions, gflB-KO daughter cells moved asymmetrically and were unable to part efficiently (Fig. S2B; Movie 10). Although cells successfully divided, gflB-KO daughter cells moved more randomly than did AX2 daughter cells (Fig. S2F). In addition, 12% of gflB-KO cells became incompletely spherical at initiation of cell division as judged by phase-contrast microscopy (Fig. S2C; Movie 11). These defects can be explained by pseudopod defects (Pollitt and Insall, 2008). Similar to gflB-KO cells, AX2 (GFP–GflB) cells (7 of 39, 17.9%) exhibited failed cytokinesis at abscission (Fig. S2D, Movie 12). However, the directionality of AX2 (GFP–GflB) daughter cells that successfully divided was comparable to AX2 daughter cells (Fig. S2F).

In spite of the higher frequency of cytokinesis failure in gflB-KO cells, the cells were not multinucleated in plate culture (Fig. 1B). It is known that so-called type II cytokinesis mutants (Adachi, 2001) such as myosin II-null cells (mhcA^–) can divide on substrate by a process called traction-mediated cytofission (TMCF) (De Lozanne and Spudich, 1987; Fukui et al., 1990), which is uncoupled from the cell cycle and also reported in mammalian cells (Choudhary et al., 2013). We examined whether gflB-KO cells divide by TMCF during interphase, and found that multinucleated cells generated by failure of cytokinesis frequently divided by TMCF (Fig. S2E, Movie 13) and became mono-nucleated. In contrast, gflB-KO cells formed multinucleated cells and scarcely grew in suspension culture, similar to myosin II-KO cells (Senoo et al., 2016). This similarity of phenotype suggests that GflB is involved in cellular processes regulated by myosin II.

Next, we studied localization of GflB. Vegetative AX2 (GFP–GflB) and gflB-KO (GFP-GflB) cells were fixed and F-actin was stained with Rhodamine–phalloidin. In interphase cells, GFP–GflB strongly localized to crowns where F-actin was enriched (arrowheads in Fig. 4A). GFP-GflB also weakly localized to the cell cortex and filopodia where F-actin was present, but the amount of F-actin did not correspond to that of GFP-GflB (filopodia are indicated by arrows in Fig. 4A). During macropinocytosis, GFP–GflB strongly localized to crowns and remained there (i.e. at macropinosomes) for a short time after the cups were enclosed (Fig. 4B, Movie 14). During phagocytosis of heat-killed yeast, GFP–GflB strongly localized to phagocytic cups with F-actin (arrowheads in Fig. 4C). During cytokinesis, GFP–GflB mainly colocalized with F-actin at filopodia (arrows in Fig. 4D) or lamellipodia (arrowheads in Fig. 4D) of daughter cells. In particular, GFP–GflB strongly localized with these structures at the lateral regions of daughter cells. Additionally, very weak GFP–GflB signals were found at the cleavage furrow where F-actin was also weakly stained. Thus, GflB localized to the actin cortex throughout the cell cycle of vegetative cells, suggesting that GflB locally regulates F-actin function.

We next constructed multiple GFP-fused fragments of GflB (Fig. 5A) and overproduced them in AX2 cells to determine regions necessary for proper GflB localization. In interphase cells, GFP-N (consisting of N-terminal amino acids 1–700), GFP–ΔGEF (missing GEF amino acids 1107–1204), GFP–CΔR5 (consisting of only N-terminal amino acids 1–175 and missing the basic region ‘B’), GFP–F5R6 (missing N-terminal amino acids 1–128 but containing ‘B’), and GFP–ΔNKD (missing C-terminal amino acids 1262–1601 corresponding to the NKD) were found in crowns and/or the cortical region, whereas GFP–ΔN (missing N-terminal amino acids 1–644, thus also missing ‘B’) was diffusedly present in the cytoplasm (Fig. 5B). Therefore, it is likely that the N-terminal region is necessary for localization of GflB to cortical regions, including crowns. Two patterns of cortical localization were observed. Like intact GFP–GflB, GFP–ΔGEF and GFP–CΔR5 localized to regions where F-actin was strongly stained. On the other hand, GFP–N, GFP–F5R6 and GFP–ΔNKD localized weakly over the entire cortical region including crowns, and strongly to certain membrane regions that were free of F-actin and did not contain substantial intact GFP–GflB in AX2 cells. During cytokinesis, GFP–CΔR5 localized to the lateral regions of daughter cells and weakly to the cleavage furrow like intact GFP–GflB (Fig. 5C). In contrast, GFP–F5R6 and GFP–ΔNKD localized weakly over the entire cortical region as in interphase cells, and strongly to the cleavage furrow and the intracellular bridge (Fig. 5C). These data suggest that the short N-terminal domain of GflB (CΔR5) recruits GflB to cortical actin structures, including crowns, while the domain including the basic region (F5R6) potentially recruits the protein to membrane cortex, probably through an interaction with lipids enriched in the furrow. Furthermore, the NKD determines localization to actin-rich regions, probably by modulating these two domains.

We then identified the specific GflB domain responsible for the interaction with the actin cytoskeleton. First, 1% Nonidet P-40 soluble (S) or insoluble (P) fractions were prepared as in Fig. 3F from AX2 cells expressing various GFP-fused GflB fragments (Fig. 5A). The distribution of GFP-fused fragments in both fractions was detected by immunoblotting with anti-GFP antibody (Fig. 5D). Intact GFP–GflB and GFP–CΔR5 were present at comparable levels to actin itself in both cytoskeletal and cytosolic fractions (Fig. 3H), while little GFP–ΔN was found in the cytoskeletal fraction. These results suggest that GflB associates with the actin cortex through CΔR5 (within N-terminal amino acids 1–175). On the other hand, high levels of GFP–N and GFP–F5R6 proteins were present in the pellets, unlike actin and intact GFP–GflB.

Second, to examine whether localization of GflB at the cortical region was dependent on interaction with the actin cytoskeleton, transformants were treated with the actin polymerization inhibitor latrunculin A (LatA) (Yarmola et al., 2000) and the change in GflB subcellular distribution was assessed (Fig. 6). Cells producing intact GFP–GflB, GFP–CΔR5 or GFP–F5R6 were treated with 10 µM LatA, incubated for 10 min, fixed and stained for F-actin with Rhodamine–phalloidin and observed by confocal microscopy (Fig. 6A). Upon drug treatment, GFP–GflB and GFP–CΔR5 disappeared from the cortical region and crowns, while GFP–F5R6 remained. To visualize cortical localization more clearly, GFP–CΔR5 and GFP–F5R6 in living cells were subjected to time-lapse observations (Fig. 6B; Movies 15–18). Both GFP-fused fragments localized to the cortical region in vehicle (DMSO)-treated control experiments. The GFP–CΔR5 movie showed a sequence in which GFP–CΔR5 is recruited to the cortex and then gathers in a macropinocytic cup (Fig. 6B, upper ‘DMSO’ panels; Movie 15). After 2 µM LatA treatment, however, localization of GFP–CΔR5 to the cortical region gradually decreased and finally disappeared, suggesting that CΔR5 localization to the cortical region is dependent on the actin cytoskeleton. On the other hand, GFP–F5R6 remained at the cortical region even after the LatA treatment, suggesting that F5R6 localization is independent of the actin cytoskeleton. Collectively, these biochemical and microscopic results suggest that GflB is localized to the cortical region via two distinct interactions, (1) an interaction with the actin cytoskeleton dependent on the N-terminal domain, and (2) one not with the actin cytoskeleton but with specific regions of membrane cortex dependent on the basic region ‘B’.

To examine whether GflB interacts directly with F-actin, we conducted in vitro pulldown assays. We first attempted to purify bacterial recombinant GflB protein in the GST-fused form, but GST–GflB was poorly expressed. GST–CΔR5 could be purified, but was precipitated by high-speed centrifugation, so we tested a much shorter N-terminal fragment, CΔR10 (Fig. 7A). Similar to GFP–GflB and GFP–CΔR5 (Figs 4A and 5B), GFP–CΔR10 localized to the F-actin-rich region (Fig. 7B) and GST–CΔR10 could be purified as a soluble protein. Using GST–CΔR10, we conducted an F-actin pulldown assay using actin from rabbit skeletal muscle (Fig. 7C,D). Indeed, GST–CΔR10, but not GST alone, co-sedimented with F-actin both in 25 mM KCl (Fig. 7C) and 150 mM KCl (Fig. 7D), indicating that the N-terminal 47 amino acids (CΔR10) of GflB directly interacts with F-actin. Compared with well-known actin-binding motifs, we found that CΔR10 has high similarity to the VHP domains (Fig. 7E). Although VHP domains are found at the extreme C-terminal end of larger core domains in cytoskeletal proteins (Vardar et al., 1999), CΔR10 is located at the extreme N-terminal end of GflB. Furthermore, CΔR10 only has a part of the C-terminus subdomain of the VHP domain; it lacks the N-terminus subdomain and a short N-terminal sequence of the C-terminus subdomain. Therefore, we call this region the VHP-like domain. It has also been reported that surface charge distribution is important for interaction with F-actin (Vardar et al., 2002; Vermeulen et al., 2004), and CΔR10 is actually rich in lysine residues.

Finally, to investigate the domains of GflB essential for its biological functions, the GFP-fused GflB fragments shown in Fig. 8A were expressed in gflB-KO cells through plasmid transformation, and changes in cell morphology and fluid-phase uptake were assessed. Among these fragments, GEF+NKD (amino acids 783–1601 consisting of the RasGEFN, RasGEF and NKD domains) and ΔGEF partially rescued the morphological phenotype (Fig. 8B,C), but ΔNKD or NKD (amino acids 1252–1601 corresponding to NKD) did not. As for the rate of fluid-phase uptake, GEF+NKD almost fully rescued, ΔGEF and NKD only partially rescued, and ΔNKD did not rescue the defects (Fig. 8D). These data suggest that both the GEF and NKD domains, particularly the NKD, are important for GflB function.

Dictyostelium GflB is a multidomain protein containing an inactive RhoGAP and RasGEF domains, which regulates Rap1 during chemotaxis (Liu et al., 2016; Senoo et al., 2016). In this study, we focused on GflB functions in the vegetative cells and revealed that GflB is a broad-spectrum regulator of cellular processes dependent on the actin cytoskeleton, including macropinocytosis (Figs 2 and 3A), phagocytosis (Fig. 3B) and cytokinesis (Fig. S2). Both depletion and overexpression of the gflB gene affected the number and lifetime of membrane protrusions, and thereby affected cellular morphology and cell migration (Fig. 1). Similar morphological defects in GflB-overproducing cells growing in plate culture were also reported by Senoo et al. (Senoo et al., 2016).

Because, in a liquid medium, most membrane protrusions of axenic strains are crowns, we focused on macropinocytosis. Both extension and retraction, and particularly retraction, of crowns were impaired in gflB-KO cells, suggesting that GflB would be important for macropinocytosis, mainly in its completion (Fig. 2A–D). Thus, it is reasonable that gflB-KO cells exhibited a reduced rate of macropinocytosis (Fig. 3A). In gflB-KO cells, the rate of phagocytosis was also reduced (Fig. 3B). This could be because the dynamics of phagocytic cup formation resembles that of crown formation at the molecular level (Bloomfield and Kay, 2016; Swanson, 2008). Because Rap1 has been reported to regulate phagocytosis (Chung et al., 2008; Seastone et al., 1999), GflB possibly regulates phagocytosis by modulating Rap1 activity. However, Rap1 is believed to mostly function in cell–particle adhesion during phagocytosis (Chung et al., 2008), which is specific to phagocytosis but not to macropinocytosis. Meanwhile, some studies have showed that Rap1 is involved in macropinocytosis (Plak et al., 2014; Seastone et al., 1999). Therefore, Rap1 could be involved in cup extension, closure and/or retraction during macropinocytosis downstream of GflB. Accordingly, the production of GflB lacking the GEF domain (ΔGEF) only partially rescued the defect of macropinocytosis in gflB-KO cells. In addition to macropinocytosis and phagocysosis, cytokinesis is also impaired in gflB-KO cells (Fig. S3) (Senoo et al., 2016), in which reorganization of the actin cytoskeleton is also essential and involvement of Rap1 has been reported (Dao et al., 2009; Plak et al., 2014).

It has been suggested that GflB regulates the balance between Rap1 and Ras during chemotaxis because gflB depletion causes broader Ras activation at the leading edge (Liu et al., 2016; Senoo et al., 2016). In addition, because the Ras–PIP3 and Ras–ForG axes are reported to be essential for macropinocytosis (Bar-Sagi and Feramisco, 1986; Hoeller et al., 2013; Junemann et al., 2016; Sasaki et al., 2007; Veltman et al., 2016), Ras GTPases may also be candidates for downstream molecules of GflB. Recently, it was reported that loss of NF1 RasGAP results in overactivation of Ras signals, which in turn results in larger Ras–PIP3 patches and increased macropinocytosis in Dictyostelium cells (Bloomfield et al., 2015; Veltman et al., 2016). In this study, we showed that PIP3 patches were in fact slightly larger in gflB-KO cells than in AX2 cells (Fig. 2E,F), suggesting that the GflB–Ras axis has only a little, if any, contribution to the formation of the PIP3 patch. Therefore, it is possible that GflB also regulates the balance between Rap1 and Ras during macropinocytosis, and unbalanced Rap and Ras could not only affect cup size but also entire cup dynamics including cup extension and retraction, in which gflB-KO cells were defective. Because whether GflB directly regulates Ras activity or the activity of Rap1 affects Ras activity remains unclear, further examinations are required to reveal whether and how GflB works through Rap1 and/or some Ras during macropinocytosis.

The GflB–Rap1 axis also regulates myosin II assembly and disassembly during chemotaxis, which regulates cortical tension and promotes pseudopod retraction (Jeon et al., 2007a,b; Liu et al., 2016). Several observations suggest that GflB also regulates myosin II assembly in vegetative cells. First, gflB-KO cells were flattened compared with AX2 cells, a morphological phenotype akin to myosin II heavy chain-null (mhcA⁻) cells and mhcA⁻ cells with constitutively disassembling 3×Asp-myosin II (Fig. 1B,C) (De Lozanne and Spudich, 1987; Jeon et al., 2007b; Stites et al., 1998). By contrast, GflB-overproducing cells were more rounded than were AX2 cells, akin to mhcA⁻ cells with constitutively assembling 3×Ala-myosin II (Fig. 1B,C) (De Lozanne and Spudich, 1987; Jeon et al., 2007b; Stites et al., 1998). Second, gflB-KO cells divided by TMCF in plate culture (Fig. S2E), and were multinucleated and scarcely grew in suspension culture (Senoo et al., 2016), both of which are features of mhcA⁻ cells (De Lozanne and Spudich, 1987; Fukui et al., 1990). Third, the F5R6 fragment of GflB localized at the cleavage furrow (Fig. 5C) where myosin II also accumulates (Yumura and Fukui, 1985). Because myosin II also has important roles in crown dynamics (Shu et al., 2005; Swanson, 2008), it would be interesting to study myosin II localization during macropinocytosis in gflB-KO cells.

Most GEFs are peripheral membrane proteins that function in multiple protein signaling cascades when attached to the cell membrane. In this study, we demonstrate that GFP–GflB localized to the actin cortex, including crowns (Figs 4–6), where GflB appears to modulate cytoskeletal dynamics through a direct interaction of the VHP-like domain (CΔR10) with F-actin (Fig. 7). In addition, the basic region B (F5R6) localized to cell cortex independently of F-actin and appeared to interact with membrane lipids (Figs 5 and 6). Thus, GflB potentially associates with F-actin and membrane cortex through distinct domains. Localization of the basic region to membrane cortex was also shown in several recent studies, although it was suggested that cortical localization of GflB is regulated by interaction with Rac (Liu et al., 2016; Senoo et al., 2016). The F5R6 fragment strongly accumulated in the cortex of the cleavage furrow during cytokinesis, suggesting that in this specific region of the cortex, GflB interacts with both F-actin and membrane lipids. Because both F-actin and GFP-GflB were present but not abundant in the cleavage furrow, where GflB appeared to act in cytokinesis, GflB may exert highly localized functions when attached to specific lipid molecules or membrane proteins accumulated in small regions of the membrane cortex (e.g. cleavage furrow) and then interact with downstream small GTPases. Although the basic region of GflB was shown to interact with phosphatidylserine (PS) in a lipid dot assay (Liu et al., 2016; Senoo et al., 2016), we propose here that PIP2 could be a candidate for such a specific lipid molecule based on the following observations: GFP–F5R6 localized to the cleavage furrow (Fig. 5C), and GFP–GflB_1–644 (also including the B domain) localized to the rear end of chemotaxing cells (Senoo et al., 2016) where PIP2 is enriched (Field et al., 2005).

The NKD is also required for cellular localization because the ΔNKD fragment did not show proper cellular localization (Fig. 5C). Despite carrying the ‘dominant’ VHP-like actin-binding domain (ABD), GFP-N lacking the NKD showed similar cellular localization to GFP–F5R6, suggesting that the NKD confers ABD dominance over the membrane-binding basic domain for determining cellular localization. Accordingly, here, we propose a local membrane activation model for GflB action (Fig. 8E). GflB localizes to the actin cortex through the interaction of the short N-terminal ABD. This region predominates over the membrane-binding domain, including the B region, for determining cellular localization. Thus, in regions of the membrane cortex where specific lipid molecules such as PIP2 (or possibly PS) are absent, GflB cannot bind to the membrane cortex or associate with Rap1 for activation, even if F-actin and GflB are abundant (the left half of Fig. 8E). In contrast, at actin-driven membrane protrusions or cleavage furrows where such lipid molecules are present, GflB interacts with the membrane cortex as well as the actin cortex via the membrane-binding domain (F5R6) to interact with and activate Rap1. Interaction of GflB with Rac may also be involved in this process. Activation of Rap1 regulates macropinocytosis, phagocytosis and cytokinesis possibly by modulating Ras (the right half of Fig. 8E). The potential involvement of Gα2–GTP, which activates GflB during chemotaxis (Liu et al., 2016), in these cellular processes is currently unclear.

Finally, the functions as well as localization of the NKD are important, at least during macropinocytosis (Fig. 8), because the fragments lacking the NKD hardly rescued the reduced rates of fluid-phase uptake. Future studies should focus on NKD functions during macropinocytosis or other cellular processes.

All the strains used in this study were derived from D. discoideum AX2 (our laboratory stock) (Watts and Ashworth, 1970). The parental AX2 and mutant strains were axenically grown in HL5 medium (Sussman and Sussman, 1967) containing Proteose Peptone No. 3 (BD Biosciences, San Jose, CA) at 22°C. For suspension culture, cells were shaken at 150 rpm. Penicillin G sodium and streptomycin sulfate (Thermo Fisher Scientific, San Jose, CA) were always added to the medium at concentrations of 6 U/ml and 6 μg/ml, respectively. For marker selection, blasticidin S (Funakoshi, Tokyo, Japan) and/or G418 (Thermo Fisher Scientific) were used at final concentrations of 4–10 μg/ml and 5–10 μg/ml, respectively.

The plasmid for disrupting the gflB gene, pUCgflBCΔR2EcoRVBsr-8, was produced through several steps. Briefly, a 5′ portion of the gflB gene (nucleotides 1–3183) was amplified by PCR from genomic DNA and cloned into a pUC118 vector (TaKaRa, Shiga, Japan). Then, the bsr cassette from pBsR503 (Puta and Zeng, 1998) was inserted at an EcoRV site located in the RhoGAP domain. For the production of GFP-fused GflB, gflB cDNA was first amplified by RT-PCR from mRNA in vegetative cells using forward and reverse primers with 5′-ATGGATCCA-3′ (a BamHI site is underlined) and 5′-TGAGCTCTTTA-3′ (a SacI site is underlined and a termination codon is bolded) extensions, respectively. Then, the amplified DNA was cut with BamHI and SacI, and the fragment was inserted into a shuttle vector pGFP-411-2p10 derived from pGFP-kinII (Oishi et al., 2000) that can express gfp-fused cDNA under the Act15 promoter to produce pGFP-GflBFL-8. For construction of gflB fragments encoding the indicated amino acid spans of GflB in the figures (Figs 5A, 7A and 8A), DNA was amplified by PCR from pGFP-GflBFL-8. The amplified fragment cDNAs were cut and inserted into pGFP-411-2p10 as described above. The deletion in pGFP-GflBFL-8 was constructed in a pHSG298 vector (TaKaRa) by digesting the gflB cDNA with MscI and SnaBI and re-ligating. For GST-fused constructs, the fragments were cloned between BamHI and SacI sites of pGEX-6P-3BBS4, in which NotI-BamHI of pGEX-6P-3 (Clontech, Palo Alto, CA) was replaced by BamHI-BglII-SacI. The sequences of all cloned fragments were verified by sequencing. The PHcrac-GFP construct, crac PH domain-GFP/pBIGΔBam (NBRP-ID: G90307), was a kind gift of NBRP nenkin (http://nenkin.nbrp.jp/).

The procedures for Dictyostelium transformation were as previously described (Adachi et al., 1994). Briefly, AX2 cells were transformed by electroporation with 2 µg of plasmid DNA for expressing one of the GFP-fused constructs, while gflBKO cells were transformed using 5 µg of plasmid by electroporation.

pUCgflBCΔR2EcoRVBsr-8 was linearized with SalI and introduced into AX2 cells as described above. For each transformant, gene disruption was confirmed by PCR with two sets of primers directly after minimum generations without making spore stocks. As it was difficult to maintain original phenotypes of gflB-KO cells and rapidly growing cells appeared during subculture, the KO cells were grown from the spore stock for the minimum number of generations were used for any experiment including transformation by plasmids. For the same reasons, transformants of the gflB-KO cells were used for experiments before making spore stocks. The gflB-KO20 strain was assigned strain code HA201.

Dictyostelium cells were grown axenically on glass-bottom dishes (non-coated, Matsunami Glass, Osaka, Japan) as described above. In general, cells were fixed with picric-acid–formaldehyde (de Hostos et al., 1991) for 15 min followed by 70% ethanol for 5 min at room temperature. To observe yeast uptake, Dictyostelium cells were incubated in Sorensen's buffer (SB, 17 mM Na-K-phosphate pH 6.0) at 22°C for 30 min, then 5×10⁶ TRITC-labeled yeast cells per ml were added and the mixture incubated for an additional 13 min before fixation. For the LatA treatment assay, cells were incubated with 5 µM LatA (Enzo Life Sciences, Farmingdale, NY) or 0.5% DMSO (GC grade, Wako Pure Chemical Industries, Ltd., Osaka, Japan) for 5 min and fixed. F-actin was stained with 50 nM Rhodamine–phalloidin (Thermo Fisher Scientific) or Alexa Fluor^® 488 phalloidin (Thermo Fisher Scientific) for 30 min at 37°C. To visualize nuclei, the cells were counterstained with 0.1 μg/ml 4′, 6-diamidino-2-phenylindole (DAPI, Sigma-Aldrich, St Louis, MO) at 37°C for 30 min. To stain microtubules, the cells were incubated with rat anti-tubulin antibody (1:500; Harlan Sera-Lab Ltd., Leicestershire, UK) at 37°C for an hour, followed by Alexa Fluor^® 568-conjugated goat anti-rat-IgG antibody (1:500, Thermo Fisher Scientific) for an hour.

The images were obtained with confocal laser-scanning microscopes (FV500 and FV1200, Olympus, Tokyo, Japan) equipped with a 100×1.40 NA or 60×1.35 NA oil-immersion lens, and Fluoview or FV10-ASW software (Olympus). Images were obtained at 0.25 µm intervals in the z-axis. z-stack images and time-lapse movies were constructed with Fluoview or FV10-ASW software. For the LatA treatment assay, cells were incubated in 17 mM potassium phosphate buffer (KPB) for more than an hour before observation. Then, the cells were treated with 2 µM LatA or 0.2% DMSO and observed every 20 s for 20 min. Adobe Photoshop CS4 (Adobe Systems, San Jose, CA) was used for image processing and Adobe Illustrator CS6 (Adobe Systems) was used for final figure preparation.

Cells were grown on glass-bottom dishes (non-coated) at 22°C for 2–3 days and observed with an IX-70 inverted microscope (Olympus) equipped with a 20×0.70 NA lens and AQUACOSMOS software (Hamamatsu Photonics Co., Shizuoka, Japan). Images were obtained every 30 s for 15 min. For analyzing cellular movement, the manual tracking tool and chemotaxis and migration plug-in (Ibidi, Martinsried, Germany) of ImageJ software was used. The cell center was determined by eye.

Cells were grown on glass-bottom dishes at 22°C for 3 days and observed with a time-lapse video system connected to an IX-70 inverted microscopy (Olympus) equipped with a 40×0.75 NA lens. Observations were made at 22°C under temperature control. Movies were processed with Adobe Premier version 6.0 (Adobe Systems).

Fluid-phase uptake was measured using a modification of the previously published protocol (Hacker et al., 1997). Samples (5 ml) of 5×10⁶ AX2 or gflB-KO cells/ml were shaken in 30-ml Erlenmeyer flasks at 100 rpm. TRITC–dextran of Mr 70103 (Sigma-Aldrich) was added at a final concentration of 2 mg/ml. From 5 min after TRITC–dextran addition, samples of 500 µl were withdrawn at intervals and added to 100 µl Trypan Blue solution to quench extracellular fluorescence [2 mg/ml of Trypan Blue (Merck Millipore, Guyancourt, France)] prepared according to Hed (Hed, 1986) and passed through 3MM filter paper and a 0.45 µm pore size Millipore filter. Samples were then centrifuged for 3 min at 1300 g. The cell pellets were washed once in 500 µl SB and resuspended in 500 µl SB, and then the relative fluorescence intensity was immediately measured in an F-2500 spectrometer (Hitachi High-Technologies, Co., Tokyo, Japan) at 574 nm emission and 544 nm excitation.

Uptake of TRITC-labeled yeast was measured as described previously (Maniak et al., 1995) with minor modifications. Heat-killed yeast (YSC2, Sigma-Aldrich) were labeled with TRITC (Sigma-Aldrich) and added at 8×10⁷ cells/ml to 2×10⁶ cells/ml AX2 or gflB-KO cells. The mixture (10 ml) was shaken in a 30 ml Erlenmeyer flask at 100 rpm. At 5 min after yeast addition, samples of 1 ml were withdrawn at intervals and added to 200 µl of Trypan Blue solution to quench non-ingested yeasts. Samples were then centrifuged for 3 min at 1300 g. The cell pellet was resuspended in 1 ml of SB and relative fluorescence intensity was immediately measured in the F-2500 fluorescence spectrometer at 574 nm emission from 544 nm excitation.

Cytoskeletal fractions were isolated as insoluble proteins in 1% Nonidet P-40 (Nacalai Tesque, Inc., Kyoto, Japan). About 10⁷ cells at logarithmic growth phase were collected and washed in buffer [20 mM Tris-HCl pH 7.6, 150 mM NaCl, 2 mM EGTA, 100 µg/ml leupeptin, 20 µg/ml pepstatin, 20 µg/ml chymostatin (Peptide Institute, Inc., Osaka, Japan), 1 mM PMSF (Wako Pure Chemical Co., Osaka, Japan)], and lysed with 600 µl lysis buffer (wash buffer containing 1% Nonidet P-40). A 200 µl volume of lysate was sampled to measure the total fraction. The remaining lysate was centrifuged at 8300 g for 10 min at 4°C. Supernatants were collected and the pellet resuspended in 400 µl lysis buffer. The amount of total protein was determined by using the Bradford method (Bradford, 1976). The total fraction was separated at 50 µg per gel lane by SDS-PAGE. The gels were stained with Quick-CBB (Wako Pure Chemical Co.).

After SDS-PAGE, separated proteins were transferred to PVDF membrane by electrophoresis using the XCell II Blot Module (Thermo Fisher Scientific). Blots were blocked for an hour with 0.5% skimmed milk in Tris-buffered saline with Tween 20 (TBS-T) and then incubated in this same buffer plus a rabbit polyclonal anti-GFP antibody (1:4000) (Sato et al., 2009). Blots were developed with ECL-Plus reagents (GE Healthcare, Chalfont St Giles, England). Enhanced chemiluminescence signals were captured by an image analyzer equipped with a cooled charge-coupled-device camera (LAS-1000plus; GE Healthcare).

The GST-fused constructs were expressed in Escherichia coli BL21 strain. The cells were grown in 2xYT medium (16 g/l Bacto tryptone, 10 g/l yeast extract, 5 g/l NaCl) containing 50 μg/ml ampicillin. Cells were induced at an optical density at 660 nm (OD₆₆₀) of 0.5 with 0.5 mM isopropyl-1-thio-beta-D-galactopyranoside (IPTG, Wako Pure Chemical Co.) and incubated for a further 5 h at 25°C. After protein production, the cells were harvested by centrifugation (5 min, 10,000 g, 4°C), washed in PBS, and resuspended in PBS containing 5 mM dithiothreitol (DTT, molecular biology grade, Wako Pure Chemical Co.) and 2 mM EDTA. To prevent protein degradation, 1 mM PMSF and protease inhibitor cocktail [1 mg/ml each of chymostatin, aprotinin, leupeptin, pepstatin A, antipain, and 1 mM benzamidine (Sigma-Aldrich)] were added. Cells were lysed by sonication and then 1% Triton X-100 was added. After 30 min incubation, lysates were cleared by centrifugation (10 min, 12,000 g at 4°C), and GST-fused proteins were purified using glutathione-Sepharose (GSH)™ 4B (GE Healthcare). For excluding DnaK, GSH-binding GST-fused protein was incubated with 20 mM MgSO₄ and 4 mM ATP (Roche, Basel, Switzerland). Proteins were eluted from GSH as GST-fused proteins by 50 mM L-glutathione reduced (Sigma-Aldrich) in Tris-HCl buffer (pH 8.0).

Sample proteins at 5 µM and an equal concentration of F-actin from rabbit skeletal muscle (>95% pure, Cytoskeleton Inc., Denver, CO) were incubated in actin dilution buffer (20 mM MOPS-NaOH [pH7.4], 2 mM MgCl₂) containing 25 or 150 mM KCl for 2 h at 22°C. The mixture was centrifuged at 436,000 g (TLA-100, Beckman Instruments Inc., Fullerton, CA) for 10 min at 4°C. The supernatant was collected, and the pellet resuspended in an equal volume of reaction buffer. The pellet samples were resolved by SDS-PAGE and stained with Quick-CBB.

We thank Dr Yoichi Noda for the helpful discussion and Enago (www.enago.jp) for the English language review. We would also like to acknowledge National BioResource Project (NBRP) by MHLW Japan for the PHcrac-GFP plasmid.