Rap1 controls cell adhesion and cell motility through the regulation of myosin II

Rap1 has been implicated in the control of cell adhesion in a variety of cell types (Stork, 2003; Bos, 2005; Kinashi and Katagiri, 2005; Retta et al., 2006). In mammalian cells, Rap1 controls cell spreading by mediating the functions of integrins, by binding to and localizing the Rac exchange factors (guanine nucleotide exchange factors [GEFs]) VAV1 and TIAM1 to the sites of cell spreading, and by regulating cadherin-mediated cell–cell contacts (Bos et al., 2003; Arthur et al., 2004; Enserink et al., 2004; Fukuyama et al., 2005; Kinashi and Katagiri, 2005; Fukuhra et al., 2006). Rap1 is thought to control cell adhesion and the leading edge function of moving cells by interacting with and regulating adaptor proteins that help control the cytoskeleton, although there is no direct evidence of a regulatory role for Rap1. RIAM and lamellipodin are related members of the MRL family of adaptor proteins containing a Ras association (RA) Rap1-GTP–binding domain, a PI(3,4)P₂-binding pleckstrin homology (PH) domain, and a polyproline domain that interacts with Ena–vasodilator-stimulated phosphoprotein and profilin. RIAM1 induces cell spreading and lamellipodia formation and localizes to the sites of membrane protrusion (Lafuente et al., 2004), whereas lamellipodin controls leading edge formation and filamentous actin (F-actin) polymerization in PDGF-treated fibroblasts (Krause et al., 2004). Other Rap1 effectors include the following: affidin, which is associated with cell junctions and binds ZO-1, nectins, and profilin and, thus, may provide a link between Rap1 and the regulation of F-actin filaments (Asakura et al., 1999); RapL, which binds LFA-1 (lymphocyte function–associated antigen 1) and regulates LFA-1 localization in a Rap1-GTP–dependent manner (Fujita et al., 2005); ARAP3, which contains an Arf6 GTPase-activating protein, a RhoA GTPase-activating protein domain, and several PH domains and affects PDGF-induced lamellipodia formation (Krugmann et al., 2006); and PDZ-GEF, a Rap1-GTP–binding Rap1-GEF (Rebhun et al., 2000; Lee et al., 2002).

In Dictyostelium discoideum, Rap1 has been linked to cytoskeletal regulation, phagocytosis, and the response to osmotic stress (Rebstein et al., 1993, 1997; Kang et al., 2002). The overexpression of wild-type Rap1 leads to cell spreading and reduces the contraction of cells in response to treatment with azide. The reduction of Rap1 expression suppresses osmotic stress–induced cyclic guanosine monophosphate (cGMP) production and impairs cell viability, whereas the expression of constitutively active Rap1^G12V enhances cGMP production (Kang et al., 2002). Osmotic stress causes the activation of Rap1 with kinetics paralleling those of cGMP production. Rap1 was implicated in phagocytosis by an analysis of the putative Rap1 effector Phg2, a kinase with an RA domain that preferentially binds Rap1-GTP but not Ras-GTP (Gebbie et al., 2004). This phg2-null strain exhibits defects in particle binding during phagocytosis, cell substratum adhesion, and motility.

The mechanisms by which Rap1 controls cytoskeletal reorganizations in D. discoideum are not understood. To investigate this, we analyzed the role of Rap1 in cell adhesion and chemotaxis and found that Rap1 controls myosin II assembly and disassembly. We demonstrate that Rap1 is activated at the cell's cortex and preferentially at the leading edge of chemotaxing cells, the site at which Phg2 localizes. Cells expressing constitutively active Rap1 are unable to effectively regulate myosin II assembly and disassembly. We determined that Phg2 is required for myosin II phosphorylation in an in vitro assay. Our findings suggest a model in which localized recruitment to and activation of Phg2 at the leading edge by Rap1-GTP of chemotaxing cells are required for proper myosin II phosphorylation and disassembly at the newly formed pseudopod. We suggest that this leading edge activation of Rap1 regulates cell adhesion and, during chemotaxis, helps establish cell polarity by locally modulating myosin II assembly and disassembly.

D. discoideum has a single Ras subfamily member that is homologous to human Rap1 (Weeks et al., 2005). We investigated whether D. discoideum Rap1 plays a role in chemotaxis by examining kinetics of the chemoattractant (cAMP)-mediated activation of endogenous Rap1 and myc-tagged wild-type, constitutively active (Rap1^G12V), and dominant-negative D. discoideum Rap1 (Rap1^S17N) expressed in D. discoideum cells. We used a pull-down assay using the human RalGDS Rap1-GTP–binding domain, which shows high but not exclusive specificity for the GTP-bound form of Rap1 and only poorly binds Ras-GTP in crude extracts in this assay (Fig. 1 E; see Materials and methods; Herrmann et al., 1996; Franke et al., 1997). Fig. 1 A shows that endogenous Rap1 was rapidly activated in response to stimulation, with a peak at 5–10 s. Myc-Rap1 (Rap1^OE) expressed in wild-type cells displayed a similar activation profile (Fig. 1 A). There was a high level of myc-Rap1^G12V in the GTP-bound form in unstimulated cells, and the level did not change on stimulation. Myc-Rap1^S17N, which was expressed at approximately the same level as myc-Rap1^OE and myc-Rap1^G12V, did not exhibit binding before or after stimulation. The relative (normalized) level of endogenous Rap1-GTP (level of Rap1-GTP compared with total Rap1 as determined by Western blotting) was slightly reduced in cells expressing Rap1^S17N.

We investigated Rap1 localization using GFP-Rap1 and myc-Rap1. GFP-Rap1 exhibited the same activation kinetics as endogenous Rap1 or myc-Rap1, indicating that it was biologically active (unpublished data). GFP-Rap1 and myc-Rap1 localized predominantly to intracellular membranes, including the ER and endosomal membranes, which is similar to observations in mammalian cells (Fig. 2, A–C; Berger et al., 1994; Ohba et al., 2003; Bivona et al., 2004). In addition, a portion of GFP-Rap1 was present on the plasma membrane. This is more readily seen after removal of the soluble cytoplasmic fraction (Fig. 2 F). Total GFP-Rap1 showed a similar localization in chemotaxing cells and is absent from the domain immediately posterior to the leading edge, presumably because the endosomal membrane fraction is excluded by the pseudopodial F-actin cortex (Fig. 2 E). The localization of total Rap1 is not detectably altered on global (uniform) chemoattractant stimulation (Fig. 2 D). GFP-Rap1^G12V exhibited a similar subcellular localization (unpublished data).

We created a YFP-RalGDS reporter to examine the spatial-temporal activation of Rap1 (Bivona et al., 2004). We observed a low level of Rap1-GTP (localized YFP-RapBD) at the plasma membrane before chemoattractant stimulation, which is consistent with a basal level of Rap1-GTP in these cells. In response to chemoattractant stimulation, Rap1-GTP levels at the cell cortex rapidly increase within 3 s of stimulation and peak at ∼6–8 s (Fig. 1, B and C). These Rap1 kinetics are slower than those of Ras (Fig. 1 C; Sasaki et al., 2004). In chemotaxing cells, Rap1-GTP is found predominantly at the leading edge and extends weakly along the sides of cells (Fig. 1 D). This domain of Rap1-GTP is broader than that observed for Ras-GTP (Sasaki et al., 2004). No localization of the YFP-RapBD reporter is observed on vesicles and other intracellular membranes where the majority of Rap1 is found, suggesting that Rap1-GTP activation is restricted to the cortex.

Cells overexpressing wild-type Rap1 (Rap1^OE cells) are more flattened and spread than wild-type cells as previously described (Fig. 3 A; Rebstein et al., 1993, 1997). This phenotype is substantially enhanced in cells overexpressing Rap1^G12V and reduced in cells overexpressing Rap1^S17N (Fig. 3 A), suggesting that Rap1 may control cell-substratum attachment. This conclusion was confirmed using cell-substratum attachment assays, which measure the fraction of cells that detach from a membrane during agitation. Rap1-overexpressing cells exhibit a small increase in cell attachment (i.e., a decrease in cell detachment; Fig. 3 B and Fig. S1). The level of cell attachment is considerably greater for Rap1^G12V cells. We were unable to obtain a Rap1 knockout, suggesting that Rap1 is essential. To examine the phenotype of cells exhibiting reduced Rap1-GTP levels, we expressed Rap1^S17N in wild-type cells. Although expressing Rap1^S17N has little effect on endogenous Rap1 activation (Fig. 1 A), Rap1^S17N cells exhibit reduced attachment.

Unconventional myosin VII and talin A form a complex and are required for adhesion of cells to the substratum (Fig. 3 B; Niewohner et al., 1997; Tuxworth et al., 2005). We examined whether the overexpression of Rap1^G12V could bypass the talin A–null cell defects. Overexpressing Rap1^G12V in talin A–null cells results in cells with a strong substratum attachment similar to that of wild-type cells overexpressing Rap1^G12V. This suggests that either Rap1 lies downstream of myosin VII–talin A or Rap1 and myosin VII–talin A function in distinct pathways to control attachment and that Rap1^G12V can promote attachment independently of talin A.

Wild-type cells become polarized and chemotax to a micropipette emitting the chemoattractant cAMP with a high speed and chemotaxis index, an indicator of the directionality of movement (Fig. 3 C). Rap1^G12V cells move at <50% of the speed of wild-type cells and have a substantially decreased chemotaxis index resulting from the production of lateral pseudopodia (and an increased number of turns) and the decreased formation of a prominent leading edge in the direction of the chemoattractant gradient (Fig. 3 C and Fig. S2). Rap1^OE cells have a milder phenotype; Rap1^S17N has a speed and directionality similar to that of wild-type cells, although with a reduced number of turns and lateral pseudopodia (Fig. 3 C). Myc-Rap1^OE and myc-Rap1^G12V cells experience delayed multicellular development as a result of a delay in aggregation and morphogenesis. The aggregates are also smaller than those of wild-type cells (Fig. S3). In contrast, Rap1^S17N cells develop more rapidly than wild-type cells.

Wild-type cells rapidly round up and contract after treatment with azide (NaN₃) as described previously (Fig. 4 A; Pasternak et al., 1989; Kang et al., 2002). As shown in Fig. 4 A, cells overexpressing Rap1^G12V do not change shape after NaN₃ treatment, and, in agreement with a previous study, NaN₃-mediated contraction does not occur in cells lacking myosin II (mhcA⁻ cells; Pasternak et al., 1989), suggesting an essential role for myosin II in this response. Fig. 4 (A and B) indicates that mhcA⁻ (myosin II null) cells exhibit a flattened, spread shape and a high degree of substratum attachment compared with wild-type cells. The attachment of mhcA⁻ cells is further increased by overexpressing Rap1^G12V. These results suggest that Rap1 functions through myosin II and other pathways to control cell adhesion.

Myosin II assembly in D. discoideum is regulated by the differential phosphorylation of three Thr residues (Thr^1,823, Thr^1,833, and Thr^2,029) in the myosin II tail by a family of myosin II heavy chain kinases (MHCKs; Pagh et al., 1984; Egelhoff et al., 1993; De la Roche et al., 2002; Yumura et al., 2005). Unphosphorylated myosin II self-assembles, whereas myosin II is disassembled in vivo by phosphorylation by the MHCKs. Substitution of Thr^1,823, Thr^1,833, and Thr^2,029 with alanines (3XAla) leads to overassembled myosin II, whereas Asp substitutions (3XAsp) cause constitutively unassembled myosin II. As expected, mhcA⁻ cells expressing myosin II^3XAsp behave similarly to mhcA⁻ cells: they exhibited high attachment, are spread, and do not change shape upon NaN₃ treatment (Fig. 4, A and B). In contrast, mhcA⁻ cells expressing myosin II^3XAla are more rounded before NaN₃ treatment and do not round further upon NaN₃ treatment (Fig. 4 A), presumably because they are already contracted. These cells also exhibit less adhesion (Fig. 4 B).

Because of the similarity of phenotypes and properties of mhcA⁻ cells and cells expressing Rap1^G12V, we investigated the effect of Rap1 on the reorganization of the cytoskeleton in response to chemoattractant stimulation. In wild-type cells, chemoattractant stimulation results in a rapid, transient phosphorylation of myosin II, causing a small but reproducible drop in assembled myosin II in the Triton X-100 pellet after ∼5–10 s (Fig. 4 C). This response is followed by an ∼2.0–2.5-fold increase in cortical myosin II that peaks at ∼40 s (Liu and Newell, 1991; Steimle et al., 2001; Park et al., 2004). Myosin II assembly is not visibly affected by overexpressing wild-type Rap1. However, the overexpression of Rap1^G12V causes a major alteration in chemoattractant-stimulated changes in assembled myosin II. After the small decrease in cortical myosin II levels, there is only a slight and gradual rise in assembled myosin II that reaches a first plateau of ∼20% of the level of myosin II in wild-type cells at 40 s, the time when assembled myosin II levels peak in wild-type cells (Fig. 4 C). There is a secondary, slow rise in assembled myosin II after 60 s, reaching a level of ∼50–60% of that of assembled myosin II in wild-type cells at 120 s. The response in wild-type cells returns to near basal levels by 80 s.

We used GFP–myosin II expressed in mhcA⁻ cells, which complements mhcA⁻ cell defects, to examine the changes in cortically localized myosin II in resting cells and in response to chemoattractant stimulation (Levi et al., 2002; Bosgraaf et al., 2005). Unstimulated cells exhibit a basal level of GFP–myosin II at the cell cortex (Fig. 4, D and E). Upon chemoattractant stimulation, after a lag of a few seconds, there is a decrease in GFP– myosin II at the cortex with a minimum at ∼10 s followed by an increase to above basal levels with a broad peak centered at 25–30 s and a subsequent decrease to basal levels. Rap1^OE cells have a similar response.

As with the assay that quantifies myosin II in the Triton X-100–insoluble fraction, Rap1^G12V cells exhibit an abnormal GFP–myosin II response (Fig. 4, D and E). At 5 s after stimulation, the level of cortical GFP–myosin II decreases substantially. Although a minimum is reached at ∼10 s in wild-type cells, the level of cortical GFP–myosin II continues to decrease in Rap1^G12V cells with a low point at 15–20 s. Levels increase only slightly through 45 s. These data are consistent with a delayed and weak chemoattractant-mediated cortical accumulation of myosin II in cells overexpressing Rap1^G12V. Our findings suggest that Rap1 directly or indirectly negatively regulates myosin II assembly at the cell cortex.

We used GFP–myosin II to study the spatial localization of myosin II in unstimulated vegetative cells and in developed cells chemotaxing to a micropipette. Wild-type cells have some cortically localized GFP–myosin II, and some cells have small patches of myosin II (Fig. 4 F). In contrast, vegetative Rap1^G12V cells are highly polarized, with one pole exhibiting a broad domain of GFP–myosin II concentrated along the cortex. The other pole has a ruffled front similar to what is sometimes observed at the leading edge of chemotaxing cells. We found that although wild-type cells have a very low level of myosin II (endogenous myosin II or GFP–myosin II) along the bottom of cells (Fig. 4 G), Rap1^G12V cells have a highly polarized myosin II localization along one pole of the cell. In chemotaxing wild-type cells, myosin II localizes to the posterior and along the lateral sides toward the cell's posterior as reported previously (Clow and McNally, 1999). In Rap1^G12V cells, much of the myosin II is found in a highly localized region at the very posterior of cells. These findings suggest that Rap1 may be involved in spatially restricting myosin II away from the leading edge in polarized cells.

Wild-type cells exhibit a biphasic F-actin polymerization profile with a sharp peak at 5 s and a second, lower, and broader peak at ∼45–60 s that has been linked to pseudopod extension (Fig. 4 C; Hall et al., 1989). Overexpressing Rap1^G12V results in an ∼40% increase in the basal level of F-actin. The kinetics of the response are similar to those of wild-type cells, with the level of F-actin proportionally higher in Rap1^G12V cells as a result of the higher basal level of F-actin in unstimulated cells. The overexpression of wild-type Rap1 has little effect on the F-actin profile.

The Phg2 Ser/Thr kinase was identified in a genetic screen for genes required for phagocytosis (Gebbie et al., 2004) and was identified independently by us in a yeast two-hybrid screen for proteins that interact with the PH domain–containing protein PhdB. Phg2 has an N-terminal PI(4,5)P₂-binding domain, a Rap1-GTP–binding RA domain, and a Ser/Thr kinase domain and colocalizes with F-actin protrusions in vegetative cells (Gebbie et al., 2004; Blanc et al., 2005). We confirmed that the Phg2 RA domain preferentially binds Rap1-GTP over Ras-GTP (Gebbie et al., 2004; unpublished data).

We examined the subcellular localization of Phg2 using a GFP fusion (see Materials and methods). Fig. 5 A indicates that GFP-Phg2, which complements the phg2-null phenotypes (not depicted), exhibits a low level of cortical localization in unstimulated cells. Upon chemoattractant stimulation, GFP-Phg2 rapidly and transiently translocates to the cell cortex. The level of cortical Phg2 peaks at ∼10 s. Phg2 localizes to the leading edge of polarized chemotaxing cells, with the localization extending weakly toward the posterior (Fig. 5 B). Thus, Phg2 shows a temporal and spatial localization similar to that of Rap1-GTP. This contrasts with Ras-GTP and PH domain proteins (e.g., CRAC and Akt/PKB) that are highly restricted to the leading edge and are absent from the lateral sides of chemotaxing cells (Sasaki et al., 2004). To determine whether the RA domain is required for the leading edge localization of Phg2 in chemotaxing cells, we created GFP-Phg2ΔRA, a Phg2 construct lacking the RA domain. In chemotaxing cells, the localization of GFP-Phg2ΔRA was similar to that of GFP-Phg2, indicating that the RA domain is not required for Phg2 localization (Fig. 5 B). We also studied the possible mechanisms by which Phg2 is recruited to the cortex. The localization is unaffected in cells treated with LY294002 and in cells lacking PI3K1 and PI3K2 (pi3k1/2-null cells) nor is localization affected by pretreatment with the actin inhibitor latrunculin A (unpublished data). These results suggest that Phg2 localization is not controlled by the PI3K or F-actin pathways that mediate the leading edge localization of several signaling components involved in the formation of a stable leading edge (Sasaki and Firtel, 2006).

To investigate the function of Phg2, we created a phg2-null strain in our wild-type strain KAx-3 (see Materials and methods; unpublished data). phg2-null cells in either HL5 axenic growth medium or phosphate buffer are very flat and spread compared with wild-type cells (and are not multinucleated) and, like Rap1^G12V cells, often have large lamellipodia-like structures on one side of the cell (Rap1^G12V cells are shown as a comparison; Fig. 5 C). The cells often have numerous and more prominent filipodia-like structures compared with wild-type cells, which are not observed in Rap1^G12V cells. The phenotypes were complemented by expressing GFP-Phg2 but not GFP-Phg2ΔRA. Collectively with the localization of GFP-Phg2ΔRA, these data suggest that the RA domain and presumably Rap1-GTP are required for Phg2 function but not its localization. We created a second phg2-null strain in which the Bsr selection cassette was inserted in the opposite orientation. Both phg2-null strains had similar phenotypes that were complemented by expressing tagged wild-type Phg2 (unpublished data).

Previous studies determined that phg2-null cells exhibit defects in particle binding, cell attachment, and basal cell motility when assayed in HL5 axenic growth medium (Gebbie et al., 2004). We confirmed a decrease in the rate of phagocytosis in our phg2-null strain (unpublished data). Our phg2-null strain exhibited a very strong increase in cell attachment compared with our parental wild-type cells (Figs. 6 A and S1). These phenotypes were complemented by the expression of tagged Phg2 in the phg2-null strain. There are conflicting data concerning the cell adhesion of phg2-null cells. Gebbie et al. (2004) found that their phg2-null cells in a DH1 background exhibited ∼90% less adhesion compared with wild-type cells as assayed in a flow chamber. Using a plate-shaking assay, Kortholt et al. (2006) showed that the same strain had a mildly reduced adhesion. To confirm our data and to directly compare the two strains, we measured cell adhesion in the plate-shaking assay of Kortholt et al. (2006) (Fig. S1). These findings confirmed both our results and those of Kortholt et al. (2006) and demonstrated an intrinsic difference between the two phg2-null strains and parental backgrounds. Our phg2-null strain phenotypes are fully complemented by GFP-Phg2, indicating that the phenotypes result from the disruption of Phg2. Chemotaxing phg2-null (phg2^KAx-3−) cells exhibit strong chemotaxis phenotypes. The cells move slower than wild-type cells, have a broader leading edge than wild-type cells, and produce numerous dominant lateral pseudopodia and more pronounced smaller projections from the leading edge (Fig. 6, B–D). These phenotypes are complemented by expressing GFP- or GST-Phg2 in phg2-null cells.

As shown in Fig. 6 E and Fig. S4 , F-actin levels were ∼1.5-fold higher in unstimulated phg2-null cells compared with those in wild-type cells. Upon chemoattractant stimulation, phg2-null cells showed only a small first peak of F-actin polymerization relative to wild-type cells, which had a 2.2-fold increase. A small second peak was observed at ∼40 s. This overall reduced level of F-actin polymerization might account for the reduced speed of phg2-null cell chemotaxis. Unstimulated phg2-null cells had ∼30% more myosin II in the Triton X-100 pellet compared with wild-type cells (Fig. 6 E). Upon stimulation, phg2-null cells exhibited a small decrease in myosin II levels at ∼5 s and a subsequent peak at ∼40 s, which were changes similar to those in wild-type cells.

Our F-actin and myosin II assays used developed cells that are stimulated in suspension, whereas our attachment assays were performed with vegetative cells adhered to a substratum. Therefore, we examined whether the differences in the basal levels of F-actin and cortical myosin II between strains found in suspension assays are also representative of adherent cells. As depicted in Fig. S4, the data are similar to those obtained from developed cells in suspension.

To study the regulation of myosin II associated with the cortex in phg2-null cells, we examined the spatial-temporal localization of GFP–myosin II in vivo. Myosin II does not delocalize from the cortex in response to chemoattractant stimulation (unpublished data). As myosin II dissociation depends on myosin II phosphorylation (De la Roche et al., 2002), this observation suggested that Phg2 might regulate myosin II phosphorylation.

Gebbie et al. (2004) demonstrated that the expression of Phg2 in Escherichia coli led to the in vivo phosphorylation of bacterial proteins on Ser/Thr residues, indicating that Phg2 has Ser/Thr kinase activity. However, Phg2 purified from bacteria did not exhibit kinase activity under the conditions assayed. To examine Phg2 kinase activity, we expressed GFP- and GST-tagged Phg2 in phg2-null cells and found that both proteins complemented the Phg2 chemotaxis and attachment defects. Triton X-100 lysates of cells before and at various times after chemoattractant stimulation were fractionated into soluble (10K supernatant) and insoluble (cortical) fractions, which were assayed for potential Phg2 activity. For the soluble fraction, we purified Phg2 (GFP-Phg2 by immunoprecipitation and GST-Phg2 on glutathione–Sepharose beads) and assayed for kinase activity using myelin basic protein or histone 2B as substrates. No activity was observed (unpublished data). We assayed the insoluble/cortical fraction similarly and detected no activity on exogenous substrates (unpublished data). We studied Phg2 activity indirectly in the cortical fraction by assaying kinase activity against endogenous substrates by incubating the Triton X-100–insoluble pellet with Mg²⁺–γ- [³²P]ATP and comparing samples from wild-type and phg2-null cells. We observed a single [³²P]O₄-labeled band (of ∼250 kD) in samples from wild-type cells that was not observed in samples from phg2-null cells (Fig. 7 A; the remainder of the gel is not depicted). The band was barely detectable in samples from unstimulated cells, strong in samples from cells stimulated for 10 and 20 s, and barely detectable in samples from later time points. As the band is not observed in phg2-null cells, we suggest that Phg2 is required for phosphorylation of the ∼250-kD protein in this assay. The kinase activity is restored in phg2-null cells expressing wild-type Phg2 but not GFP-Phg2ΔRA, suggesting that the RA domain is required for Phg2 activity and that Phg2 activity may be activated by Rap1-GTP. Our assay does not distinguish between Phg2 directly phosphorylating the ∼250-kD protein or being required for the activity of another kinase that directly phosphorylates myosin II.

Coomassie blue–stained gels showed that the ∼250-kD phosphorylated protein comigrated with the myosin II band. To examine the possibility that the phosphorylated band is myosin II, we repeated the assay in mhcA⁻ cells and found that the band was absent (Fig. 7 B). To distinguish between the protein being myosin II and it requiring myosin II for its cortical localization, we expressed GFP–myosin II in mhcA⁻ cells. The phosphorylated band was then present, and its mobility corresponded to GFP–myosin II (Fig. 7 B), indicating that the phosphorylated protein is myosin II. This activity was unaffected by pretreatment with LY294002 or latrunculin A, which is consistent with our findings that Phg2 cortical localization is unaffected by the treatment of these drugs and indicating that these pathways were not dependent on the PI3K pathway or F-actin polymerization (Fig. 7 C).

We then assayed mhcA⁻ cells expressing myosin II^3XAla or myosin II^3XAsp to examine whether Thr^1,823, Thr^1,833, or Thr^2,029 might be the sites of myosin II phosphorylation. Fig. 7 D illustrates that neither of these substituted myosin II proteins are phosphorylated. Although these findings are consistent with the sites of Phg2-dependent myosin II phosphorylation being Thr^1,823, Thr^1,833, or Thr^2,029, we cannot exclude the possibility that Thr^1,823, Thr^1,833, or Thr^2,029 are not the sites of phosphorylation and that the actual sites of phosphorylation are unavailable in myosin II^3XAla and myosin II^3XAsp in our assay for an unknown reason. We also investigated myosin II phosphorylation in cells lacking three of the four MHCKs (mhck-A/-B/-C–null cells), which exhibit an extensive myosin II overassembly (Yumura et al., 2005). Fig. 7 E indicates that myosin II phosphorylation is highly reduced in this strain, suggesting that the myosin II phosphorylation pathway requires MHCK. The overexpression of MHCK-A in phg2-null cells suppresses the lack of myosin II phosphorylation in our assay (Fig. 7 F). The kinase activity is found in unstimulated as well as stimulated cells, with the activity increasing after stimulation. We find that as previously reported for wild-type cells (Steimle et al., 2001), MHCK-A localizes to the leading edge of chemotaxing phg2-null cells, suggesting that the loss of myosin II kinase activity in phg2-null cells is not caused by the inability to recruit MHCK-A (unpublished data). Phg2 has a Rap1-GTP–binding RA domain and, therefore, is expected to lie downstream from Rap1. Cells expressing Rap1G12V have an elevated level of myosin II phosphorylation that is also extended, which is consistent with increased and extended activation mediated by the activated form of Rap1 (Fig. 7 G).

We demonstrated that Rap1 is dynamically activated in response to chemoattractant stimulation with a Rap1-GTP GFP reporter transiently localizing to the cell cortex with a peak at 5–10 s. In chemotaxing cells, activated Rap1 is found predominantly at the leading edge of chemotaxing cells with a distribution that weakly extends along the cortex toward the posterior, whereas total Rap1 is found predominantly on membrane vesicles and along the plasma membrane, as in mammalian cells. The spatial distribution of Rap1-GTP extends more laterally than that of Ras-GTP, and the kinetics of activation are slower, suggesting that the sites of function of Rap1 and Ras overlap but are not coincident.

Rap1^G12V cells are very flat and adherent to the substratum, which is consistent with findings in mammalian cells, indicating the involvement of Rap1 in controlling cell shape and attachment. These cells chemotax slowly and produce numerous lateral pseudopodia, probably in part because they cannot regulate their attachment to the substratum and restrict the activation of pseudopod formation to the side of the cell closest to the chemoattractant source. Rap1^S17N cells have a modest reduction in attachment. The Rap1^S17N phenotypes are weak, presumably because Rap1^S17N does not effectively inhibit the Rap1 GEFs, as there is little change in the activation of endogenous Rap1 in these cells.

Our studies suggest that one mechanism by which Rap1 controls cell spreading and attachment in D. discoideum acts at least partially through the regulation of myosin II assembly. Rap1^G12V cells exhibit weak chemoattractant-mediated myosin II assembly, which is consistent with an inability to control cell contraction. In response to NaN₃ treatment, wild-type but not Rap1^G12V cells contract, which is consistent with previous studies (Pasternak et al., 1989; Kang et al., 2002). These phenotypes are shared by cells lacking myosin II (mhcA⁻ cells) or mhcA⁻ cells expressing myosin II^3XAsp. These strains exhibit a spread, flat cell shape and a high substratum attachment. In Rap1^G12V cells, GFP–myosin II dissociates from the cell cortex in response to chemoattractant stimulation but undergoes only a slight and greatly delayed reassociation with the cortex. Rap1^G12V cells exhibit an abnormal and deeply reduced chemoattractant-mediated myosin II assembly with slow kinetics. These findings suggest that Rap1 negatively controls myosin II assembly.

MHCKs are the major kinases that phosphorylate myosin II on Thr^1,823, Thr^1,833, and Thr^2,029 and result in myosin II disassembly (Egelhoff et al., 1993; De la Roche et al., 2002). MHCK-A is recruited to the leading edge by binding to F-actin, which activates the kinase (Egelhoff et al., 2005), thereby disassembling myosin II at the site of leading edge protrusion. Our findings that Rap1 is activated at the leading edge and that the expression of Rap1^G12V causes extended myosin II disassembly suggest that Rap1 may be a key upstream spatial regulator of myosin II assembly. We suggest that Rap1 may prevent myosin II assembly at the leading edge and may indirectly (passively) promote its assembly at the posterior of cells. Myosin II assembly and phosphorylation are regulated by cGMP in D. discoideum (Bosgraaf et al., 2002, 2005; Bosgraaf and van Haastert, 2006). Cells lacking the cGMP-binding protein GbpC exhibit a spatial-temporal profile of GFP–myosin II cortical disassembly/assembly similar to that of Rap1^G12V cells. The profile of myosin II in the Triton X-100–insoluble pellet is similar to that of cells lacking both guanylyl cyclases (Bosgraaf et al., 2002, 2005). Thus, cells expressing Rap1^G12V exhibit phenotypes similar to those lacking cGMP or the cGMP effector GbpC, suggesting that Rap1 may be a negative regulator of the cGMP pathway. Just before submission of this manuscript, a study by Kortholt et al. (2006) demonstrated that GpbD is a Rap1 GEF and that its overexpression leads to increased cell adhesion and chemotaxis defects similar to those of the Rap1^G12V cell presented here. They present evidence that Rap1 lies upstream of Phg2.

Our model by which Rap1 may control anterior and posterior functions may be analogous to the proposed mechanism by which anterior and posterior functions are differentially regulated by Gαi and Gα12/13 in neutrophils (Wong et al., 2006). Gαi promotes the leading edge function and inhibits posterior function (activation of RhoA and assembly of myosin II), whereas Gα12/13 controls posterior functions by mediating RhoA activation. The loss of Gαi function increases RhoA activity, whereas the inhibition of RhoA function increases leading edge function, indicating that the regulation of anterior and posterior functions is balanced. In D. discoideum, anterior and posterior functions have different response kinetics and are spatially restricted in polarized cells (Van Haastert and Devreotes, 2004; Sasaki and Firtel, 2006). Chemotaxing Rap1^G12V cells often have a flat, extended posterior region that does not lift from the substratum, suggesting that Rap1 plays a part in controlling the posterior function of moving cells. In addition, the localization of myosin II is altered in unstimulated Rap1^G12V compared with wild-type cells. Although myosin II has a general localization along the cortex except at sites of pseudopod extension with an enrichment at the posterior in wild-type cells (Clow and McNally, 1999), resting Rap1^G12V cells have a polarized cytoskeletal organization with myosin II present in a single, extended domain along approximately one third of the cortex opposite a domain that has the appearance of a projecting lamellipod. In chemotaxing Rap1^G12V cells, myosin II is found almost exclusively in a small cap/domain at the posterior of cells, as though myosin II sites are restricted from other areas of the cell through the constitutive activation of Rap1. This lack of visible myosin II along the cell's lateral sides may contribute to a possible loss of myosin II–mediated cortical tension and the presence of lateral pseudopodia.

We demonstrated that Phg2 is a key regulator of the cell's cytoskeleton and cell attachment. phg2-null cells are highly adhesive, which is a finding that contrasts that of Gebbie et al. (2004). We discovered that under two different assay conditions, their phg2-null strain has increased adhesion in phosphate buffer, suggesting that differences may be caused by assay conditions. As phg2-null cells exhibit high adhesion, we suggest that Phg2 negatively regulates cell adhesion.

Phg2 is required for proper chemotaxis. Chemotaxing phg2-null cells move slowly and have a broad leading edge with numerous small protrusions. The broad leading edge structure (more of a lamellipod than a pseudopod) and protrusions are presumably related to the broad lamellipod and filopodia-like protrusions exhibited by vegetative phg2-null cells. These cells have reduced random motility (Gebbie et al., 2004; unpublished data). Phg2 leading edge localization does not require the RA domain, but the ability to complement the null phenotypes does, indicating that the genetic function of Phg2 is dependent on Phg2's ability to bind Rap1-GTP. These findings demonstrate that Phg2 and, thus, its putative regulator Rap1 play key roles in controlling leading edge formation and in suppressing these lateral and adventitious protrusions. phg2-null cells have a deeply reduced F-actin response, which, along with the high attachment of these cells, may account for their reduced chemotactic speed.

A further understanding of the possible mechanism by which Phg2 controls the cytoskeleton was obtained by examining the physiological and biochemical phenotypes of phg2-null cells. phg2-null cells show an elevated basal level of myosin II and F-actin in vegetative and developmentally competent cells, which are normal kinetics of chemoattractant-mediated myosin II assembly, but show a substantially reduced F-actin polymerization response. Motile amoeboid cells (e.g., neutrophils and D. discoideum) spatially restrict F-actin polymerization and leading edge protrusion from myosin II assembly to the opposite poles of polarized cells (Van Haastert and Devreotes, 2004; Sasaki and Firtel, 2006; Wong et al., 2006). Myosin II thick filaments are partially responsible for cortical tension or rigidity. To protrude a pseudopod, cells must disassemble the myosin II thick filaments at that site. Our imaging data suggest that Phg2 is required for the disassembly of myosin II thick filaments; in the absence of Phg2, cells accumulate elevated levels of myosin II in the Triton X-100–insoluble fraction. We revealed that Phg2 is required for in vitro myosin II phosphorylation in a cortical fraction in response to chemoattractant stimulation. As this phosphorylation is thought to be responsible for myosin II disassembly from the cortex (De la Roche et al., 2002), our combined data suggest that Phg2 is required for efficient cortical myosin II phosphorylation and disassembly. We suggest that the defect in delocalizing myosin II from the cortex that we observe in phg2-null cells may be responsible for the restricted chemoattractant-mediated F-actin response. In Rap1^G12V cells, myosin II disassembly is extended, and there is minimal reaccumulation over a 2-min time frame.

The initial responses observed after a global stimulation recapitulate the changes that occur at the leading edge of chemotaxing cells. Like other leading edge responses, Rap1 is activated rapidly at the cortex, and Phg2 translocates to the cortex in response to global stimulation. Rap1 is also activated at the leading edge, and Phg2 accumulates at this site. We suggest that Phg2 is a new component of the rapidly activated leading edge regulatory network that, in this case, controls myosin II phosphorylation and disassembly at this site. The reduced myosin II phosphorylation in mhck-A⁻, -B⁻, or -C⁻ cells and constitutive activity in MHCK-A^OE cells suggest that this network may regulate MHCK-A (or other MHCKs), which localizes to the leading edge. The higher basal level of myosin II in phg2-null cells is consistent with this hypothesis. Phg2 may play other roles, as phg2-null cells have many cellular protrusions and extended lamellipodia in resting and chemotaxing cells, phenotypes that are not observed in cells lacking myosin II. phg2-null cells exhibit a normal myosin II response when one examines total myosin II that associates with the Triton X-100–insoluble fraction and are able to contract upon NaN₃ treatment. As such, we expect that phg2-null cells can assemble and disassemble myosin II in other domains of the cell. In conclusion, Rap1 is a major regulator of the cytoskeleton in resting cells and cells stimulated by chemoattractant or responding to osmotic stress, and it functions, at least in part, to control myosin II through Phg2. Therefore, Rap1 may be part of the signaling cascade by which leading edge and posterior functions of the cell interact to control cell polarization and coordinate directional movement.

We obtained latrunculin A, LY294002, aprotinin, and leupeptin from Sigma-Aldrich, anti-myc antibodies from Santa Cruz Biotechnology, Inc., glutathione–Sepharose beads from GE Healthcare, and γ-[³²P]ATP from MP Biomedicals. Anti-Rap1 antibodies have been described previously (Kang et al., 2002). 13-mm nitrocellulose filters (0.22 μm) used for the adhesion assay were obtained from Millipore.

The full coding sequence of the rap1 cDNA was generated by RT-PCR and was subcloned into the EcoRI–XhoI site of pSP72. An EcoRI restriction site in the 5′ end region of the rap1 gene was removed by substitution of cytosine at nucleotide 18 with thymine without changing the coding of amino acids. For GFP- or myc-Rap1 fusion proteins, the EcoRI–XhoI fragments of pSP72 were subcloned into the expression vector EXP-4(+) containing either a GFP or myc fragment.

The mutants Rap1^G12V and Rap1^S17N were generated using the QuikChange Site-Directed Mutagenesis kit (Stratagene). Because the D. discoideum rap1 gene product contains two additional N-terminal amino acids compared with the human rap1 and ras gene products, amino acids were numbered according to the consensus alignment of Ras proteins to facilitate comparison (Rebstein et al., 1997). GST-RalGDS–Ras-binding domain (RBD) was a gift from J.L. Bos (University Medical Center Utrecht, Utrecht, Netherlands). For RBD-RalGDS-YFP fusion protein expression, the RBD of RalGDS was amplified by PCR using GST-RB-RalGDS as a template, was ligated to a YFP fragment, and was cloned into the expression vector EXP-4(+). All clones were confirmed by DNA sequencing.

A phg2 knockout construct was made by inserting the blasticidin resistance cassette into a BamHI site created at nucleotide 595 of the phg2 cDNA and was used for a gene replacement in the KAx-3 parental strains. Randomly selected clones were screened for a gene disruption by PCR, which was then confirmed by Southern blot analysis and RT-PCR. These studies confirmed that the knockout strains did not express Phg2 transcripts 5′ or 3′ to the site of insertion.

A full-length phg2 sequence was obtained by PCR using a series of primers and was cloned into pBluescript, sequenced, and subcloned into EXP-4(+) plasmids for GFP or GST fusion protein expression. To create a GFP-Phg2 construct lacking the RA domain (GFP-Phg2ΔRA), two pairs of primers were used to amplify the sequence 5′ and 3′ to that encoding the RA domain, creating an XbaI site at the junction site. This created an in-frame deletion of the RA domain. This construct was then subcloned into GFP-fused pEXP4 plasmid. The GFP-Phg2ΔRA coding region was confirmed by sequencing.

The mhca-null mutant strain in the KAx-3 background and MhcA-GFP plasmids were provided by J.A. Spudich (Stanford University, Palo Alto, CA). The myosin VII– and talin A–null cells were obtained from M.A. Titus (University of Minnesota, Minneapolis, MN), and DH1-10 and phg2-null cells in the DH1-10 background were obtained from P. Cosson (Universite de Geneve, Geneva, Switzerland). The myosin heavy chain phosphorylation mutant constructs 3XAsp and 3XAla were gifts from T.T. Egelhoff (Case Western Reserve University, Cleveland, OH).

The Rap1-GTP–binding domain of mammalian RalGDS was expressed in E. coli as a GST fusion protein as described previously (Franke et al., 1997). The purified GST-RalGDS-RBD was used for the detection of activated Rap1. Log-phase vegetative cells were washed twice and resuspended at a density of 5 × 10⁶ cells/ml in Na/K phosphate buffer and pulsed for 6 h with 30 nM cAMP every 6 min. The cells were collected and resuspended at a density of 2 × 10⁷ cells/ml and were treated with 1 mM caffeine for 30 min. The 300-μl aliquots were stimulated with 15 μM cAMP and lysed by mixing with an equal volume of 2× lysis buffer (100 mM Tris, pH 7.5, 300 mM NaCl, 50 mM MgCl₂, 20% glycerol, 1% NP-40, 2 mM DTT, 2 mM vandate, aprotinin, and 5 μg/ml leupeptin) at the indicated times. The lysates were centrifuged for 10 min, and the supernatants were incubated with 10 μg GST-RBD on glutathione–Sepharose beads at 4°C for 30 min in the presence of 1 mg/ml BSA. The beads were washed three times and subjected to SDS-PAGE and Western blot analysis with either an anti-myc mAb or an anti-Rap1 pAb. For control of the input amount of total Rap1 proteins, 40 μl of the cells were taken right after lysis of the cells and analyzed.

Analyses of chemotaxis toward cAMP were performed as described previously (Park et al., 2004; Sasaki et al., 2004). The images of chemotaxing cells were taken at time-lapse intervals of 6 s for 30 min using an inverted microscope (TE300; Nikon) with a plan Fluor ELWD 40× NA 0.6 lens and a camera (CoolSnap HQ; Roper Scientific). The frames were captured using MetaMorph software (Molecular Devices), and the data were analyzed using DIAS software (Soll Technologies; Wessels et al., 1998). Cells for analysis were randomly chosen with the requirement that they move for at least 15 min without touching another cell and were examined. Differential interference contrast images were hand traced (10 frames/min for 15–20 min). At least five randomly chosen cells were analyzed from each of at least three independent experiments performed on separate days.

The quantitation of membrane or cortical localization of GFP fusion proteins was performed as described previously (Sasaki et al., 2004) with slight modification. Fluorescence images were obtained using a confocal microscope (DM IRE2; Leica) with HCX plan Apo NA 1.40 100× or 63× objective lenses (oil CS; Leica) and a camera (EM-CCD or ORCA-ER; Hamamatsu Photonics). Images were captured using SimplePCI software (Compix Inc., Imaging Systems) and were analyzed using ImageJ software (National Institutes of Health). Differential interference contrast images were taken with the same equipment but using an ORCA-285 camera (Hamamatsu Photonics).

Adhesion assays were performed as described previously (Sun et al., 2003) with minor modifications. Log-phase growing cells on the plates were washed with Na/K phosphate buffer and resuspended at a density of 2 × 10⁶ cells/ml. The amount of 4 × 10⁵ cells in 200 μl were plated onto 13-mm circular nitrocellulose filters (Millipore). After 30 min, unattached cells were removed by dipping filters into Na/K phosphate buffer. Filters were transferred to microcentrifuge tubes filled with 800 μl Na/K phosphate buffer and were vortexed for 1 min with a mixer (model 5432; Eppendorf) simultaneously. 150 μl of the detached cells from the filters were plated onto a 30-mm Petri plate with a hole covered by a 0.17-mm glass coverslip, and an additional coverslip was placed on top. The cells were photographed and counted (detached cell number). To determine the total cell number, 200 μl of the cells was transferred into microcentrifuge tubes filled with 600 μl Na/K phosphate buffer and counted. Cell adhesion is presented as a percentage of detached cells compared with total cells. This experiment was repeated three times or more, each time with four filters for each strain.

5 × 10⁶ cells/ml in Na/K phosphate buffer were pulsed and stimulated with cAMP as described in the PKB/Akt activity assay (Meili et al., 1999). After stimulation with cAMP, the cells were lysed by mixing with an equal volume of 2× lysis buffer. The cytoskeleton fractions were isolated by centrifugation of the lysates at 4°C for 10 min. The pellets were washed once with 1× lysis buffer and then with kinase buffer (25 mM MOPS, pH 7.4, 25 mM β-glycerophosphate, 20 mM MgCl₂, and 1 mM DTT). The pellets were resuspended and incubated in 50 μl of kinase buffer containing 10 μCi γ-[³²P]ATP and 5 μM ATP for 20 min at 22°C. The reaction was stopped by the addition of 25 μl of 4× sample buffer followed by SDS-PAGE. The gels were stained with Coomassie blue, dried on a Hoefer slab gel drying apparatus, and exposed to film.

F-actin polymerization and myosin II assembly were assayed as described previously (Park et al., 2004). Fluid-phase endosomes were labeled using 70,000-M_r TRITC-dextran (Sigma-Aldrich). Log-phase cells expressing GFP-Rap1 proteins were incubated with 0.75 mg/ml TRITC-dextran for 1 h and were washed twice with Na/K phosphate buffer. The cells were observed using a confocal microscope (DM IRE2; Leica). Sodium azide treatment of cells was performed as described previously (Pasternak et al., 1989). Vegetative cells were washed twice with MES buffer (20 mM MES, pH 6.8, 2 mM MgSO₄, and 0.2 mM CaCl₂) and oxygenated by bubbling for 20 min. The cells were allowed to adhere to glass coverslips for 10 min and were inverted onto a plate with a hole covered by a glass coverslip holding 150 μl MES buffer with or without 25 μM NaN₃. After 30 min of incubation, the cells were photographed.

Fig. S1 shows that the phg2-null strain created in the KAx-3 background exhibits strong attachment both in the plate-shaking method described in Kortholt et al. (2006) and in the cell attachment assay described in Materials and methods. Fig. S2 shows an increased number of turns and more frequent extrusion of lateral pseudopodia in cells expressing Rap1^G12V. Fig. S3 shows that the multicellular development of Rap1^OE and Rap1^G12V cells is delayed. In contrast, Rap1^S17N cells develop more rapidly than wild-type cells. Fig. S4 shows basal levels of F-actin and assembled myosin II in vegetative cells or in aggregation-competent cells.

We would like to thank Tom Egelhoff for his helpful suggestions and for providing mhck-A/-B/-C–null cells and mhca-null cells expressing 3XAla–myosin II. We also thank Peter van Haastert for stimulating discussions, Pierre Cosson for the phg2-null strain and discussions of his group's results, and Jim Spudich for the anti–myosin II antibody. We thank the other members of the Firtel laboratory for advice and assistance.

This work was supported by grants from the U.S. Public Health Service to R.A. Firtel and grants from the Canadian Institutes of Health Research to G. Weeks.