Parallel signaling pathways regulate excitable dynamics differently to mediate pseudopod formation during eukaryotic chemotaxis

Chemotaxis to extracellular chemical signals plays important roles in various physiological phenomena, including neurogenesis, immune response and wound healing (Servant et al., 2000; Kalil and Dent, 2005; Mayor and Etienne-Manneville, 2016). Chemotactic cells, such as mammalian neutrophils and the social amoeba Dictyostelium discoideum, detect chemical gradients and migrate directionally by coordinating their anterior pseudopod formation and posterior tail contraction (Van Haastert and Devreotes, 2004; Hind et al., 2016). Evolutionally conserved molecular mechanisms are involved in the bias of cell motility directionally along chemoattractant gradients (Artemenko et al., 2014; Devreotes et al., 2017). In Dictyostelium cells, extracellular 3′,5′-cyclic adenosine monophosphate (cAMP) works as a chemoattractant and is detected by G protein-coupled receptors (GPCRs) and cognate heterotrimeric G proteins. This detection activates multiple signaling pathways including those mediated by (1) phosphoinositide 3-kinase, and phosphatase and tensin homolog (PI3K–PTEN), (2) phospholipase A2 (PLA2), (3) TorC2, phosphoinositide-dependent kinase 1 and protein kinase B family proteins (TorC2–PDK–PKB; note the PKB family is also called Akt), and (4) soluble guanylyl cyclase (sGC). Most of these parallel and compensatory pathways generate a polarized signal selectively at the front or back of a motile cell along the cAMP gradient (Cai and Devreotes, 2011; Artemenko et al., 2014; Devreotes et al., 2017). In the PI3K–PTEN pathway, a phosphatidylinositol 3,4,5-trisphosphate (PIP3)-enriched domain localizes to the pseudopod and regulates the polymerization or stabilization of F-actin (Kortholt et al., 2011; Devreotes et al., 2017). Similar localizations of sGC and PKB activities also act as intracellular cues for pseudopod formation (Veltman and Van Haastert, 2006; Kamimura et al., 2008). However, little is known about the relationship between the parallel signaling pathways and their unique functions in chemotaxis.

The spatiotemporal dynamics of PIP3 has been well characterized during the chemotactic motility shown by eukaryotic amoeboid cells such as Dictyostelium. A PIP3-enriched domain on the membrane is produced in a self-organized manner independently of cell motility and shows wave-like localization patterns (Asano et al., 2008; Arai et al., 2010; Gerisch et al., 2012). The PIP3-enriched domain spontaneously occurs in the absence of extracellular stimuli, and it has relatively constant biochemical activities and size, of a few microns in diameter. These spontaneous activities are linked to pseudopod formation and are biased along cAMP gradients, leading to a change from random to directional motility (Xiong et al., 2010; Shibata et al., 2013). The PIP3 dynamics has been explained by an intrinsic excitable system. In general, an excitable system shows an all-or-none response to a super-threshold stimulus so that a constant output that is independent of the input strength is assured (Huang et al., 2013; Nishikawa et al., 2014). Thus, the excitable behavior is able to amplify a small input and function as an internal biochemical compass for motile cells (Xiong et al., 2010; Shibata et al., 2013). Human neutrophils also show PIP3 enrichment into membrane domains in a manner consistent with an excitable system, suggesting that this a conserved property of eukaryotic chemotaxis (Tang et al., 2014). Another feature of excitability is characterized by the refractory period, which is the recovery time for the system to react to a second stimulus once its excitation starts. During the refractory state, the system cannot respond even to a stimulus beyond the threshold. Thus, excitability determines the responsiveness of a cell to the environmental stimulus. The PIP3 pathway in Dictyostelium cells has a refractory period of ∼60 s (Huang et al., 2013; Nishikawa et al., 2014), which is longer than the time scale at which cells respond to chemical gradients, giving ambiguity to the physiological meaning of the refractory period.

Among parallel pathways, the sGC pathway has been well documented to contribute to the chemotaxis of Dictyostelium cells (Veltman et al., 2008). A series of mutants in the sGC pathway have been shown to impair chemotaxis to various kinds of chemoattractants, including cAMP and folic acids, demonstrating the importance of the pathway in chemotactic signaling (Kuwayama et al., 1993, 1995). In the sGC pathway, extracellular cAMP induces intracellular cGMP elevation, which serves as a second messenger for tail contraction at the posterior of the cell via the regulation of myosin II (MyoII) (Soll et al., 2009). cGMP is synthesized by two different types of guanylyl cyclase, cytosolic sGC and membrane-embedded guanylyl cyclase A (GCA) (Roelofs et al., 2001; Roelofs and Van Haastert, 2002). sGC mainly contributes to guanylyl cyclase activity during the cell aggregation stage, at ∼5 h after starvation. Bioinformatic analysis has discovered four cGMP-binding candidate proteins, named GbpA–GbpD (Goldberg et al., 2002). All four proteins have cyclic nucleotide-binding motifs, but biochemical experiments have revealed that GbpC is a major cGMP effector (Bosgraaf et al., 2002). The degradation of cGMP is catalyzed by the two cGMP phosphodiesterases GbpA and GbpB (Bosgraaf et al., 2002). By contrast, GbpC serves as the intracellular high-affinity cGMP-binding molecule and is required for MyoII regulation, while GbpD regulates cell adhesion through its guanine nucleotide exchange factor (GEF) activity for Rap1, even though it has no cGMP-binding ability (Bosgraaf et al., 2005; Kortholt et al., 2006). In addition to the regulation of MyoII at the posterior tail, it has been suggested that the sGC pathway has additional function at the pseudopod (Veltman et al., 2005). sGC proteins localize to pseudopods through the N-terminus without the enzymatic domain during random and chemotactic migration, and cAMP stimulation induces transient recruitments of sGC to the membrane in a manner similar to the PIP3 dynamics (Veltman and Van Haastert, 2006). Cells expressing sGC lacking its N-terminus exhibit a high degree of turning, which brings about a loss of cell orientation, an effect that is also seen in pi3k1, pi3k2 double-knockout strains (Bosgraaf and Van Haastert, 2009). These observations suggest that the sGC pathway has characteristics of an excitable system; however, no direct evidence has been obtained experimentally.

Here, we found that sGC and PIP3 show pseudopod localizations at different frequencies in individual cells, and demonstrated that sGC localization shows behavior consistent with it being part of an excitable system including mediating an all-or-none response to cAMP stimulation, and showing a wave-like pattern formation and refractory period. The refractory period of the sGC-mediated signaling pathway was shorter than that of PIP3, and sGC localized to the pseudopod more frequently than PIP3. sGC localization depended on F-actin and was negatively regulated by cGMP and GbpC via F-actin dynamics, indicating that sGC localization and F-actin are incorporated into one excitable network. This regulation explains the short sGC refractory period as well as concomitant pseudopod formation with relatively high frequency. Overall, these results illustrate that multiple signaling pathways regulate pseudopod dynamics differently through their characteristic excitable properties for chemotactic motility.

To address how parallel pathways regulate cell migration, we observed the localization patterns of sGC and PIP3 simultaneously in individual cells. Since sGC localizes to pseudopods through its N-terminal region, we used a Halo-tagged sGC protein as a sGC probe (sGC_N–Halo) along with a previously reported PIP3 probe, the GFP-tagged pleckstrin homology domain of Dictyostelium PKBA (PH_PKB–GFP) (Meili et al., 1999; Veltman and Van Haastert, 2006; Asano et al., 2008). sGC_N–Halo was stained with tetramethylrhodamine (TMR) for multicolor imaging. In wild-type cells co-expressing sGC_N–Halo and PH_PKB–GFP, sGC_N signals were observed in all pseudopods, while PIP3 signals were observed in only half of the pseudopods (Fig. 1A; Fig. S1A). Statistical analysis revealed that the localization pattern was classified into three types: pseudopods with sGC_N localization alone (47.8%), pseudopods with sGC_N and PIP3 colocalization (50.2%) and pseudopods with PIP3 alone (2.0%) (n=452 pseudopods from 13 cells) (Fig. 1B). We analyzed several parameters of the pseudopod dynamics, including extended length, life-time, contribution to cell movement and appearance frequency, for each localization pattern by measuring the extended area of the spontaneously formed pseudopods (see Materials and Methods, Fig. S1B). While small extensions were observed in pseudopods with sGC alone, pseudopods with sGC and PIP3 colocalization were coupled with larger extensions (Fig. 1C,D). In fact, the elongation length of pseudopods with sGC and PIP3 colocalization (5.22±2.31 μm, mean±s.d., n=189 pseudopods) was higher than that of sGC alone (3.92±1.79 μm, n=147 pseudopods) (Fig. 1E). To see the sole effect of sGC localization on the pseudopod dynamics, we created sgcΔ cells, which express a mutant sGC lacking its N-terminal region (sGCΔN–Halo/sgcΔ) but that maintains the catalytic domain needed for cGMP synthesis. PIP3-alone pseudopods in these mutant cells exhibited smaller elongations than pseudopods where sGC and PIP3 were colocalized (4.28±2.16 μm, n=61 pseudopods, versus 5.22±2.31 μm, n=189). Wild-type cells treated with LY294002, a pharmacological PI3K inhibitor, exhibited only small elongations (2.82±1.08 μm, n=39 pseudopods) (Fig. 1E). Consistent with this, the life-time and the contribution to cell motility of pseudopods with both signals were higher than those with sGC alone (Fig. S1C,D). However, the appearance frequency was not significantly different (Fig. S1E). These genetic and pharmacological inhibitions also caused decreases in the velocity of random migration (Fig. S1F). These observations suggest that the two parallel pathways regulate pseudopod formation additively.

We next observed both signaling molecules in cells exposed to a cAMP gradients to see how their pathways regulate pseudopod formation during chemotaxis (Fig. 1F,G; Movie 1). Upon uniform cAMP stimulation, both signaling molecules exhibited a transient localization to the membrane independently of each other (Fig. S1G,H), demonstrating that the two pathways work in parallel. Under cAMP gradients, the localization patterns of both signaling molecules exhibited different characteristic dynamics (Fig. 1F,G; Movie 1). When the source of the chemoattractant gradient was continuously moved around the cell, sGC and PIP3 were able to follow the changes of the position similarly (Fig. 1F; Fig. S1I). By contrast, once the source position was fixed, the membrane localization dynamics of sGC and PIP3 was different (Fig. 1G). The stimulus first induced sGC and PIP3 localization to the same pseudopod, which was followed by a transient disappearance of both signals. Later sGC localization preceded PIP3 localization on the pseudopod, showing that sGC localized to the pseudopod more frequently than PIP3, consistent with the observations in randomly moving cells (Fig. 1A,B). These results demonstrate that both signaling pathways have their own unique dynamics in response to the same cAMP stimulation.

We next determined whether sGC localization has the features of excitability or not. An excitable system shows the same response to any perturbation over a certain threshold because the response is determined by an intrinsic excitable mechanism. Therefore, a pulse cAMP stimulus with a very short duration should trigger the same response in sGC localization to a stepwise cAMP stimulus if the responses are excitable. To determine whether this was the case, we used an experimental system in which caged-cAMP was photolyzed in a flow chamber, allowing us to control the timing and strength of the pulse stimuli by manipulating UV irradiation exposure times (Fig. S2A) (Beta et al., 2007). Upon a UV flash at a region upstream of the cells, the uncaged cAMP is applied to and removed from the cells within a few seconds because of the flow (Fig. S2B), which is short enough to test the excitability of sGC localization (Veltman et al., 2005). A C-terminal fusion of GFP to full-length sGC (sGC–GFP) was expressed in gcΔ cells lacking both sGC and GCA, and sGC–GFP localization was observed in response to the transient stimulation. Cytosolic sGC–GFP exhibited translocation to the membrane as a transient response to a pulse stimulus within 2 s of UV exposure as well as upon a step stimulus of 100 nM cAMP (Fig. 2A). The two transient responses had the same temporal changes and same peak amplitude (see Materials and Methods), indicating that the sGC localization dynamics is produced intrinsically by the system itself in a manner that is independent of the stimulus pattern. Furthermore, we examined the dose-dependent response of sGC localization to the membrane upon various UV light exposure times. As expected, the cells expanded the domain size of the sGC signals on the membrane with increasing levels of activated cAMP, although the domain amplitudes remained constant (Fig. 2B). To confirm that the uncaging by UV flash provided the expected cAMP concentrations precisely, we uniformly stimulated the cells with the defined cAMP concentrations and observed sGC–GFP translocation to the membrane. Similar to the results in Fig. 2B, the domain size on the membrane increased according to the cAMP concentration until 0.1 nM with constant domain amplitudes (Fig. S2C). These results reflect excitable features, such as stimulus-dependent domain size and constant amplitude, which were observed previously for PIP3 excitation (Nishikawa et al., 2014).

An excitable system often shows wave-like propagation (Sager, 1996; Khamviwath et al., 2013). In Dictyostelium cells, PIP3 and F-actin accumulations on the membrane are typical examples of traveling wave generation (Arai et al., 2010; Gerisch et al., 2011). When we mildly perturbed the actin cytoskeleton with a low concentration of latrunculin A, an actin polymerization inhibitor, we found newly polymerized F-actin showed a wave-like propagation on the bottom surface of the cells, as visualized by the fluorescent F-actin probe mRFP–LimEΔcoil (Fig. 2C) (Fischer et al., 2004). The same behavior was observed for sGC–GFP localization, which overlapped the F-actin signals (Fig. 2C; Movie 2). Furthermore, cAMP-elicited sGC localization to the membrane was significantly reduced in mutant cells lacking functional ArcB, a subunit of Arp2/3 complex, which is responsible for actin polymerization (Fig. 2D) (Langridge and Kay, 2007). Overall, these results imply that the excitable behavior of sGC relies on F-actin activity.

We determined whether sGC localization to the membrane exhibits a refractory period or not. gcΔ cells expressing full-length sGC–GFP were stimulated repeatedly with pulsed cAMP at different time intervals (Fig. 3A,B). Transient recruitment of sGC–GFP in response to repetitive stimuli at 21-s and 30-s intervals occurred normally. However, in response to repetitive stimuli at 12-s intervals, the first response was normal but the second and third responses were progressively weaker. These results indicate that the sGC localization has a refractory period of between 10 and 20 s. The existence of the all-or-none response, propagating wave and refractory period proves that the sGC signaling pathway is excitable.

Next, we compared the excitable properties of sGC and PIP3 localizations on the membrane. First, we directly evaluated their refractory times by observing cells that co-expressed both sGC_N–Halo and PH_PKB–GFP when stimulated repeatedly by means of UV irradiation at 21-s intervals. sGC localization exhibited a response repeatedly to all stimuli, whereas PIP3 did not (Fig. 3C). The refractory period of PIP3 localization was ∼30–60 s (Fig. 3D), consistent with previous reports (Huang et al., 2013; Nishikawa et al., 2014). sGC localized to the membrane in mutant cells lacking pi3k1 and pi3k2 genes upon repeated stimuli with a refractory period of 30 s, indicating that sGC responsiveness to cAMP stimulation is independent of PIP3 production (Fig. 3E). Thus, the excitable dynamics of sGC has a unique refractory period that is shorter than that of PIP3. Therefore, the sGC pathway can respond to stimulations of short intervals (about 10–20 s), but the PI3K–PTEN pathway cannot.

To gain molecular insights into the excitable behavior of the sGC pathway, we focused on the possible involvement of both the product of the pathway, cGMP, and its downstream signaling molecules, because excitable systems generally include a delayed negative-feedback mechanism to generate transient excitation (Cao et al., 2016). To reveal the effect of cGMP on sGC localization to the membrane, we first observed mutant sGC with the amino acid change Asp1106Ala, which lacks catalytic activity (sGCΔcat) and was previously expressed as a tagged protein with GFP (sGCΔcat–GFP) in gcΔ cells (Veltman and Van Haastert, 2006; Sato et al., 2009). It was reported that gcΔ cells expressing sGCΔcat fail to produce cGMP upon cAMP stimulation (Veltman and Van Haastert, 2006). Wild-type sGC–GFP (sGCwt–GFP) exhibited translocation to the membrane transiently upon cAMP stimulation, but the translocation terminated within 20 s. On the other hand, the catalytically dead mutant sGCΔcat–GFP showed prolonged and stronger membrane localization for several minutes (Fig. 4A), indicating that intracellular cGMP accelerates the dissociation of sGC from the membrane. As a control, we examined the response of sGCΔcat–GFP in wild-type cells (Fig. S3A). sGCΔcat–GFP showed translocation to the membrane upon cAMP stimulation in a similar manner to sGCwt–GFP. However, there were some differences in the behavior between sGCwt–GFP and sGCΔcat–GFP after 30 s of stimulation; much more sGCwt-GFP returned to the cytosol than sGCΔcat–GFP, probably because the overproduction of sGCwt–GFP increases the cGMP level in wild-type cells. We also examined the cAMP-mediated responses of wild-type and gcΔ cells that expressed a sGC with a deletion of the C-terminal catalytic domain (sGC_N–GFP; Fig. S3B). As expected from the data shown in Fig. 4A, sGC_N–GFP translocated to the membrane within 10 s in both cells, but the returning phase to the cytosol was slower in gcΔ cells. In addition, a much higher amount of sGCwt–GFP dissociated from the membrane in wild-type cells than in cells expressing sGC_N-GFP (compare Fig. S3A and S3B). When co-expressed in the same wild-type cells, sGCwt and sGC_N had almost the same dynamics upon cAMP stimulation, reaching equivalent cytosolic levels (Fig. S3C). These results further implicate intracellular cGMP in the dissociation of membrane sGC.

Next, we observed wild-type sGC localization in a series of mutant cells lacking GbpA and GbpB (gbpABΔ), GbpC (gbpCΔ) or GbpD (gbpDΔ). Among them, only gbpCΔ cells exhibited defects in sGC–GFP localization, in which sGC–GFP continued to localize on the membrane stably in a manner similar to sGCΔcat–GFP (Fig. 4B). The other mutants exhibited no obvious changes in sGC–GFP localization at resting state without cAMP stimulation (Fig. 4B). When stimulated with cAMP, gbpCΔ cells exhibited prolonged sGC–GFP translocation to the membrane, while wild-type cells exhibited transient responses (Fig. 4C). Thus, sGC localization to the membrane is regulated negatively by its product, cGMP, and a cGMP-dependent signaling molecule, GbpC, which ensures the transient response of sGC localization upon cAMP stimulation. In addition, mhcAΔ cells, which lack the myosin heavy chain, a gene required for cell contraction at the posterior side (Van Haastert and Devreotes, 2004), exhibited transient sGC localization on the membrane upon cAMP stimulation, although the activity was less efficient than in wild-type cells (Fig. 4C). These results demonstrate that cGMP-dependent GbpC signaling regulates negatively the lifetime of sGC accumulation on the membrane. Because sGC catalyzes cGMP production, this negative regulation by GbpC provides a delayed negative-feedback loop for the sGC excitation pathway.

The above results suggest that cGMP-dependent GbpC signaling suppresses sGC localization to the membrane and that this negative feedback regulates recovery to the basal state (i.e. the refractory period) during sGC excitation. In fact, sGC–GFP in gbpCΔ cells showed persistent rather than repetitive responses to 12-s and 21-s interval stimuli (Fig. 5A), which resembles the response to step stimuli (Fig. 4C), indicating that cells require GbpC to distinguish repeated pulsatile stimuli from continuous stimuli. Because sGC–GFP in gbpCΔ cells responded to repetitive pulse stimuli at 30-s intervals in a manner similar to sGC–GFP in wild-type cells (Figs 3A and 5A), we can conclude that sGC excitation can recover to basal state within 30 s in a GbpC-independent manner. Next, we observed sGC–GFP responses in a gbpAΔ cell line, in which the cGMP-specific phosphodiesterase gene is deleted, to see the effects of cGMP on the refractory period, because gbpAΔ cells are reported to show prolonged overproduction of cGMP upon cAMP stimulation (Lusche and Malchow, 2005). The gbpAΔ cells showed severely suppressed sGC–GFP responses to repeated stimuli at 21-s intervals, while wild-type cells responded normally (Figs 3A and 5B). That is, prolonged cGMP overproduction prolonged the refractory period. Thus, the cGMP-dependent GbpC signaling is required for shortening the refractory period and provides a mechanism for the fast recovery of the sGC pathway in order to respond to the subsequent cAMP stimulation.

To examine the effects of cGMP-dependent GbpC signaling on F-actin polymerization and pseudopod formation in response to cAMP stimulation, we observed mRFP–LimEΔcoil in gcΔ and gbpAΔ cells, which have defects in the production and degradation of cGMP, respectively (Fig. 5C; Fig. S4A and Movie 3). Wild-type cells showed transient F-actin polymerization at the entire cell surface at ∼10 s after continuous cAMP stimulation, which was followed by the formation of multiple pseudopods with F-actin polymerization for ∼1 min, as observed by mRFP–LimEΔcoil localization. When gbpAΔ cells were exposed to the same stimulus, subsequent responses were not observed over the observation period of several minutes, revealing that there was a long suppression of F-actin polymerization. The ectopic expression of GbpA in gbpAΔ cells rescued the delay of the secondary response (Fig. 5C, GbpA/gbpAΔ).

To further confirm these results, we observed the refractory period in a repetitive stimuli experiment. gbpAΔ cells significantly reduced the second response to a 21-s interval stimulus, while wild-type and rescued gbpAΔ cells revealed almost the same response to the first and second stimuli (Fig. S4B). On the other hand, both gcΔ and gbpCΔ cells exhibited enhanced F-actin polymerization (Fig. 5C; Movie 4). This persistent F-actin formation in gbpCΔ cells was also observed in a biochemical assay (Fig. 5D; Fig. S4C). GbpC has Ras-like Roc, RasGEF and an mitogen-activated protein kinase kinase kinase (MAPKKK) domain plus a cGMP-binding domain. It has been reported that cGMP binding to GbpC triggers intramolecular signal transduction, ultimately leading to kinase activation (Van Egmond et al., 2008). Therefore, we analyzed the involvement of this intramolecular signaling in GbpC-mediated F-actin regulation using deletion mutants for the MAPKK kinase or the cGMP-binding domains. We found neither mutant was able to rescue the phenotype of excessive F-actin formation upon cAMP stimulation, suggesting that GbpC requires kinase activation following cGMP binding for this regulation (Fig. 5D). These results are consistent with cGMP-dependent GbpC signaling suppressing F-actin polymerization and pseudopod formation. Thus, the negative feedback from cGMP/GbpC to sGC shortens the refractory period of sGC localization to the membrane, which provides a mechanism for high frequency and repetitive pseudopod formation upon cAMP stimulation.

Eukaryotic chemotaxis is mediated by parallel signaling pathways that have characteristic features of excitability for the generation of all-or-none signals to regulate pseudopod dynamics at the front of the cells (Cai and Devreotes, 2011; Devreotes et al., 2017). Previous reports have demonstrated the excitable behavior of PIP3 membrane localization during the chemotactic signaling of Dictyostelium cells and neutrophil cells (Huang et al., 2013; Nishikawa et al., 2014; Tang et al., 2014). Here, we demonstrated that sGC localization to the membrane was consistent with excitable behaviors, such as an all-or-none response (Fig. 2A,B; Fig. S2C), wave-like pattern formation (Fig. 2C) and refractory period (Fig. 3). These observations indicate the universality of excitable systems in chemotactic signaling pathways. Although both the PI3K–PTEN and sGC pathways have excitable properties, each pathway independently forms an excitable network, as shown by the following findings. First, sGC localization strongly relied on F-actin (Fig. 2D), but F-actin was not essential for PIP3 excitation on the membrane (Arai et al., 2010). Second, sGC responses upon pulse stimuli showed shorter refractory periods (10∼20 s) than PIP3 responses (30∼60 s) (Fig. 3C,D). Finally, the refractory period of sGC localization was independent of PIP3 production (Fig. 3E). These results indicate that each downstream pathway has unique characteristics for their excitable dynamics.

Our observations revealed that sGC localized to the pseudopod more frequently than PIP3 when cells moved along the chemoattractant gradient (Fig. 1F,G). Under constant chemical gradients, it took 10–30 s for sGC to re-localize to the pseudopod once the localization disappeared, which is consistent with the refractory periods seen for sGC responses to pulse stimuli (Fig. 3A). These results suggest that the refractory period for the localization of a signaling molecule corresponds to the frequency of the pseudopod formation. This idea was further supported by observations that showed that modulation of the intracellular cGMP concentration affected the refractory period of sGC responses and pseudopod formation. gbpAΔ cells lacking cGMP degradation activity showed prolonged refractory periods (‘refractoriness’), and prolonged suppression of F-actin polymerization and pseudopod formation (Fig. 5B,C; Fig. S4A) (Roelofs et al., 2001; Meima et al., 2002). The prolonged refractoriness has been reported also in gbpAΔ cells as a light-scattering response or Ca²⁺ influx upon a cAMP stimulus (Lusche and Malchow, 2005). Furthermore, gcΔ and gbpCΔ cells showed pseudopod formation more frequently than wild-type cells (Fig. 5C,D; Movie 4).

It has been reported that when stimulated with cAMP twice at 30-s intervals, cells synthesized cGMP upon the first stimulus but not upon the second, indicating some form of adaptation (Van Haastert and Van Der Heijden, 1983). That same report showed that the addition of a cAMP phosphodiesterase enabled cGMP production in response to both stimuli. These results suggest that cGMP could be produced repetitively when cAMP stimulation is short enough so as to avoid complete adaptation. Our optical-uncaging experiment was likely to have provided a transient stimulus to the cells before adaptation occurred, although the intracellular cGMP levels were not measured. Based on these observations, we suggest that the refractory period regulated by cGMP-dependent GbpC signaling is a physiologically important feature of chemotactic cell migration.

Although cAMP stimulus induces PIP3 and sGC excitation on the membrane independently (Fig. S1G,H), there may be crosstalk between these pathways. We showed that sGC localization strongly depends on F-actin dynamics (Fig. 2C). Previous studies have shown that F-actin is required for PI3K localization to the pseudopod, and thereby positive feedback comprising Ras, PI3K and F-actin is important for directional cell migration (Sasaki et al., 2004). These observations suggest there is positive feedback between the PI3K–PTEN and sGC pathways that is mediated through F-actin dynamics. This crosstalk could control pseudopod dynamics positively and cooperatively. We further found that pseudopods have different molecular fingerprints, in which most pseudopods showed sGC localization but only half showed both sGC and PIP3 localization (Fig. 1B). Thus, these two types of pseudopods differ. The localization of both sGC and PIP3 leads to more extensive pseudopod extension than the localization of either alone (Fig. 1E). Additionally, the life-times and contribution to cell movement are longer and bigger (Fig. S1C,D). Thus, these parallel pathways regulate the pseudopod elongation additively.

Our results are summarized diagrammatically in Fig. 5E. Chemoattractant stimulations trigger sGC translocation to F-actin on the membrane (Fig. 2C,D). At the same time, sGC catalyzes cGMP production upon chemoattractant stimulation (Roelofs and Van Haastert, 2002). The increased amount of cGMP binds to GbpC directly and induces its kinase activation through a series of intramolecular signals. Then GbpC initiates the destabilization of F-actin and in turn suppresses further pseudopod elongation and induces pseudopod retraction (Fig. 5C). In our model, cGMP-dependent GbpC signaling defines the refractoriness in sGC excitability, which is linked to suppression of the pseudopod. It has been previously reported that GbpC provides major intracellular cGMP-binding sites at a high affinity with a K_d of ∼10⁻⁹ M and slow dissociation rate with a halflife of ∼2 min (Van Haastert et al., 1982). These kinetics are not in accordance with our model, which assumes a much faster dissociation for cGMP of 20–30 s. To resolve this discrepancy, an unknown factor with low affinity and fast dissociation to cGMP could function downstream of cGMP-bound GbpC or it is possible that a threshold must be passed for cGMP-bound GbpC to disrupt F-actin stability. Future work will clarify how cGMP regulates F-actin suppression through GbpC.

Previous studies have also shown that cGMP and GbpC signaling activation induces pseudopod suppression at the rear end of cells via the regulation of myosin II (Bosgraaf et al., 2005; Veltman and Van Haastert, 2007). On the other hand, it is known that myosin II is not essential for the retraction of the anterior pseudopod observed in mhcAΔ cells (Iwadate and Yumura, 2008). We found myosin II was not essential for sGC excitation upon cAMP stimulation (Fig. 4C). Taken together, our results show that cGMP-dependent GbpC signaling could mediate anterior retraction of the pseudopod, with myosin II making only a minor contribution.

cGMP-dependent GbpC signaling constitutes a delayed negative-feedback in the sGC excitable pathway, which provides a mechanism for the fast cycling of repeated pseudopod formation via shortening of the refractory period. Multiple signaling pathways with different refractory periods can regulate pseudopods at different time scales to mediate chemotaxis. Such pseudopod regulation mechanisms may provide flexibility for motile cells to respond to complex environmental stimulations at various time scales.

Dictyostelium discoideum cells were grown at 22°C in HL5 medium supplemented with 6 ng/ml vitamin B12 and 100 ng/ml folic acid (Watts and Ashworth, 1970). Wild-type AX2 and AX3 cells were used in this study. The AX2 cell line was used for simultaneous imaging of sGC_N–Halo and PH_PKB–GFP, sGC–GFP and mRFP–LimEΔcoil. Because pi3k1Δ2Δ and arcB mutant cells were derived from the AX2 cell line, sGC–GFP responses in these strains were compared to those in AX2 cells. AX3 cells were used as controls when observing sGC–GFP and mRFP–LimEΔcoil localization, because the mutant strains gcΔ (sgcΔ/gcaΔ), gbpAΔ, gbpABΔ, gbpCΔ, gbpDΔ and mhcAΔ were derived from the AX3 cell line. All plasmids and strains are listed in Tables S1 and S2. When grown with selection, medium was supplemented with 10 μg/ml blasticidin S, 20 μg/ml G-418 or 50 μg/ml hygromycin B. Cells were starved for 5–7 h in development buffer consisting of 10 mM Na/KPO₄ pH 6.5, 2 mM MgSO₄ and 0.2 mM CaCl₂. For cAMP stimulation experiments, cells were treated with 4 mM caffeine for 30 min to inhibit de novo adenylyl cyclase activity and deplete extracellular cAMP. Cells were placed on glass-bottom dishes (IWAKI). sgcΔ cells were generated from AX2 cells by homologous recombination using a disruption cassette containing the blasticidin S resistance (BSR) gene. Two regions of the sgc sequence, from 607 to 1116 and 4194 to 4891, were amplified by using the full-length sgc gene as a template, and the BSR gene was inserted between them. The construct was linearized with BamHI and introduced into AX2 and sgcΔ cells, which were selected with 10 μg/ml blasticidin S. The cloned transformants were screened by PCR of their isolated genomic DNAs. The N-terminal fragment of sGC (sGC_N, amino acids 1–1019) and the N-terminal deletion fragment (sGCΔN, amino acids 877–2843) were amplified by PCR and cloned into pHK12-Halo7 to yield C-terminal Halo7-tagged proteins. The full-length gbpA gene was amplified by PCR and cloned into pDM358-eGFP to yield C-terminal eGFP-tagged protein.

Fluorescence images were obtained by using a confocal microscope (FV1000, Olympus) with a 60×/1.35 NA oil-immersion objective lens and software (Fluoview, Olympus). TMR and mRFP were excited by a 543 nm He-Ne laser, and GFP was excited by a 488 nm Ar laser. For pulse cAMP stimulation, cells were placed on a 4-cm round cover glass (Matsunami) with a flow chamber (FCS2, Bioptics). The inlet of the chamber was connected to a 10-ml glass syringe (Top) mounted in the syringe pumps (FP-1000, Melquest). The syringe was filled with development buffer containing 4 mM caffeine and 5 nM DEACM-caged cAMP (Biolog). Photolytic activation of caged cAMP was undertaken with a 405 nm mercury lamp and was carried out upstream of the cells, and activated cAMP was flowed through the system. To demonstrate the temporal change of cAMP concentration, 100 μM CMNB-caged fluorescein (Invitrogen) was used instead of caged cAMP (Fig. S2B). The flow rate was 300 μl/min. In repetitive pulse stimulus experiments, all exposures were carried out via a 1.5-s laser irradiation. To align the first pulse timing, data were normalized by the maximum value of cytosolic fluorescence intensity.

Differentiated cells without caffeine treatment were put on a glass-bottom dish and captured by the confocal microscope at 5-s intervals for at least 5 min. The pseudopod dynamics were analyzed by ImageJ according to the localization patterns of sGC and PIP3. The positions of the sGC and PIP3 signals were assigned from the fluorescence images to distinguish two types of localization: sGC alone, and the colocalization of sGC and PIP3. Morphological changes were extracted by differentiating between two consecutive binarized fluorescence images. In this study, we defined a pseudopod as an area that extended beyond 4 μm²/frame during growth. These extracted pseudopods were characterized by quantifying their size, appearance frequency, life-time and contribution to cell movement. The size was defined by the elongation length between the beginning and end points of a pseudopod during its total life cycle. The appearance frequency, or the average number of pseudopods per cell, was calculated by dividing the total number of pseudopods on each cell by the recorded time. The life-time is the length of time in which a positive change was maintained in each pseudopod's area. The contribution to cell movement was defined as the displacement of the centroid of a pseudopod during its life-time. For cell motility analysis, differentiated cells without caffeine treatment were observed with an Olympus IX-71 inverted microscope capable of producing phase-contrast optics. Movie files were recorded with a CCD camera (Digital Sight, Nikon) and software (NIS-Elements, Nikon). Cell behavior was recorded for 20 min at 5-s intervals, and images were analyzed using G-Count (G-Angstrom). The motility speed of each cell was calculated as the average speed for each short trajectory.

Membrane localization of sGC–GFP was analyzed with ImageJ. The fluorescence intensity profile of sGC–GFP on the membrane was obtained at every pixel along the cell periphery and processed with a 15-pixel moving average. In order to compensate for different sGC–GFP expression levels, profiles were normalized to the membrane fluorescence intensity seen in areas without sGC localization. To compare sGC localization under various conditions, we defined localization amplitude and domain size as follows. Localization amplitude represents the maximum value of the normalized intensity along the cell membrane. Domain size was quantified as the ratio of the region with the normalized value exceeding the threshold to the overall membrane area. The threshold was set as 1.2 in this report.

The spatiotemporal dynamics of fluorescence probes on cell membranes was analyzed and shown as 2D patterns with ImageJ. To monitor the pseudopod dynamics of moving cells, kymographs of the mRFP–LimEΔcoil localization in Fig. 5C were generated using an ImageJ plugin. The second response time was defined as the time that a cell showed a new mRFP–LimEΔcoil localization after its first response to 10 nM cAMP. Some cell lines showed a persistent response without transient suppression. The second response time of those cells was defined as 0. sGC_N–TMR and PH_PKB–GFP localizations on the bottom surface of the cell in Fig. 1C,D were analyzed via the ‘reslice’ function in ImageJ after 1.0 pixel mean filtering of movie.

As previously described (Kamimura et al., 2016), differentiated cells were incubated in phosphate magnesium buffer (2 mM MgSO₄ in 5 mM Na/KPO₄ buffer pH 6.5) with 3 mM caffeine at a density of 3×10⁷ cells/ml. The cell suspension was stimulated with cAMP at a final concentration of 10 nM. After stimulation, cells were lysed by adding equal volumes of 2× assay buffer consisting of 2% Triton X-100, 20 mM KCl, 20 mM EGTA, 20 mM imidazole and 0.1 mg/ml NaN₃, and incubated on ice for 10 min. The samples were centrifuged at 8000 g for 4 min to collect the pellet fraction followed by washing with 1× assay buffer and boiled in 2× SDS sample buffer. Finally, the samples were analyzed by PAGE and stained with Coomassie Brilliant Blue. Images were acquired with ImageQuant LAS4000 for the quantification of actin levels (40 kDa).

We thank Peter Van Haastert (Dept. Cell Biology, Univ. of Groningen, The Netherlands) for the plasmids expressing sGC–GFP, sGCΔcat–GFP and GbpC mutants, and the National BioResource Project (NBRP)-Nenkin for the AX3, gbpAΔ, gbpABΔ, gbpCΔ, gbpDΔ, mhcAΔ and mutated arcB cells. We appreciate the basic information from DictyBase. We also thank Peter Karagiannis for critically reading the manuscript, and thank all members of the Ueda laboratory for discussions and technical assistance.