RasGEF-containing proteins GbpC and GbpD have differential effects on cell polarity and chemotaxis in Dictyostelium

Dictyostelium discoideum live as unicellular amoebae that feed on bacteria in the soil. Upon starvation, cells start to secrete cAMP that serves as a chemoattractant. After aggregation is complete, a multicellular slug is formed that eventually develops into a fruiting body composed of dead stalk cells and a slime droplet containing viable spores. When Dictyostelium cells are placed in a chemoattractant gradient, they become polarised and start to move in the direction of the gradient. Polarisation improves the chemotactic efficiency of the cells and many mutants that have a chemotaxis defect also display a decreased ability to maintain polarity (Bosgraaf et al., 2002a; Khurana et al., 2002; Wessels et al., 1988; Zhang et al., 2003). However, mutant cells that make fewer lateral pseudopodia and consequently appear hyperpolar also aggregate less efficiently compared to the wild type (Zhang et al., 2002). This indicates that a balance exists between the spatial confinement of events leading to polarisation on one hand, and the tendency to make lateral protrusions that allows direction changes on the other.

Actin and conventional myosin (myosin II) play an important role in the cytoskeletal rearrangements that underlie chemotactic movement. In particular, actin polymerisation is crucial for the formation of new protrusions whereas myosin II plays an important role in the maintenance of cell polarity and retraction of pseudopodia and the uropod (Clow and McNally, 1999; Koehl and McNally, 2002; Laevsky and Knecht, 2003; Wessels et al., 1988). Protruding pseudopodia are major sites of actin polymerisation and branching (Condeelis et al., 1988). This is probably mediated by the Arp2/3 complex and by a number of small GTPases (reviewed by Lee et al., 2001).

In contrast to actin, myosin II is mainly localised at the cortex of the posterior half in translocating cells and in retracting pseudopodia (Moores et al., 1996). Studies on myosin heavy chain knockout cells have revealed that myosin is required for cortical tension, efficient chemotaxis and cell polarisation (Egelhoff et al., 1996; Wessels et al., 1988). Myosin is regulated primarily by phosphorylation of the regulatory light chain (RLC) and of the heavy chain (MHC). Phosphorylation of the RLC increases the ATPase activity of myosin II (Griffith et al., 1987), whereas MHC phosphorylation regulates the formation of myosin filaments (Egelhoff et al., 1993; Stites et al., 1998). Chemoattractant stimulation not only leads to the transient phosphorylation of RLC and MHC (Berlot et al., 1985), but also induces the translocation of myosin filaments to the cortex (Liu and Newell, 1991). Recently it was found that the translocation event requires the presence of F-actin (Levi et al., 2002). Furthermore, another recent study indicates that actin polymerisation depends on the phosphorylation state of myosin (Heid et al., 2004). These findings indicate that actin and myosin cooperate in an intricate signal transduction pathway that is poorly understood.

A number of studies have implicated cGMP as an important mediator of myosin II signalling (reviewed by van Haastert and Kuwayama, 1997). Recently, we identified two guanylyl cyclases and four putative cGMP binding proteins in Dictyostelium (Roelofs et al., 2001a; Roelofs et al., 2001b; Goldberg et al., 2002). Two of the cyclic nucleotide binding proteins (GbpA and GbpB) encode non-canonical phosphodiesterases (Bosgraaf et al., 2002a; Bosgraaf et al., 2002b; Meima et al., 2002). The other two proteins (GbpC and GbpD) are presumably the main targets of cGMP, as no other potential cGMP binding candidates could be identified in the nearly completed genome database (http://www.dictybase.org/). GbpC is the high affinity cGMP binding component in Dictyostelium, as disruption of the gene caused the loss of all high affinity cGMP binding capacity of cell lysates (Bosgraaf et al., 2002a). GbpC has a unique domain topology and contains small GTPase, kinase, RasGEF, GRAM and two cyclic nucleotide binding domains, respectively. GbpD is homologous to the C-terminal half of GbpC and also contains RasGEF, GRAM and two cyclic nucleotide binding domains. Mutants in which both gbpC and gbpD are disrupted display severely impaired chemotaxis, which is caused in part by the strongly diminished myosin II translocation to the cortex that was observed in this mutant (Bosgraaf et al., 2002a). Furthermore, MHC phosphorylation and especially RLC phosphorylation was also reduced in gbpC^–/gbpD^– cells. The same aberrancies were found in a mutant that is unable to produce cGMP owing to the disruption of two guanylyl cyclases (Bosgraaf et al., 2002a). This indicates that a cGMP-mediated pathway acts through GbpC and/or GbpD, leading to the phosphorylation and translocation of myosin II.

Here we report the effect of separate disruption of the gbpC and the gbpD genes. The results show that the phenotype of gbpC^– cells is essentially identical to the previously described phenotype of the double gbpC^–/gbpD^– mutant, both exhibiting a cGMP-dependent, myosin-mediated defect in the suppression of lateral pseudopodia and reduced chemotaxis. Disruption of gbpD leads to highly polar elongated cells that exhibit enhanced cell speed, which is accompanied by a decreased substrate adhesion. Furthermore, cells overexpressing GbpD extend many substrate-attached pseudopodia and are impaired in chemotaxis. This phenotype is similar to that of cells overexpressing Rap1 (Rebstein et al., 1993), which may well be a target of GbpD. These data indicate that GbpD induces the formation of substrate-attached pseudopodia that leads to increased adhesion and a concomitant decrease in polarity. Thus, GbpC and GbpD independently affect chemotactic motility in opposing manners.

DH1 (a uracil auxotroph wild-type Dictyostelium), gca^–/sgc^– (a guanylyl cyclase-null cell line) (Roelofs and Van Haastert, 2002), and the mutant cell lines described below were grown on HG5 medium (14.3 g/l pepton, 7.15 g/l yeast extract, 10 g/l glucose, 1.36 g/l Na₂HPO₄.12H₂O and 0.49 g/l KH₂PO₄) unless indicated otherwise. To select for transformants with one of the extrachromosomal plasmids described below, HG5 was supplemented with 10 μg/ml Geneticin (Gibco) and/or Hygromycin B (Invitrogen).

The gbpC or gbpD gene was disrupted using gene-specific plasmids that were previously used to make the double gbpC/gbpD deletion mutant (Bosgraaf et al., 2002a). Potential knockouts of gbpC or gbpD were screened by PCR and confirmed by Southern blot analysis.

The gbpD gene was cloned in five parts by PCR using the Expand Long Template kit (Roche). Each PCR fragment was first separately cloned in pGemTeasy (Promega) and sequenced. The PCR fragments were joined by making use of the restriction sites that are underlined in the primers below. The 5′ part of the gene that encodes the rasGEF domain contains introns and was obtained using reverse-transcriptase PCR with AX3 RNA as a template. Two PCR fragments were obtained with the primer pairs 5′-TGGATCCGAAAAAATGACAGATTACCCATTC-3′ (BamHI site bold) and 5′-GTTATCTAGACATCGATGATCTACTTGGA-3′; and 5′-TGGATCCAAAAATGCATCGATGAAAATGGGTAAAG-3′ with 5′-GTTATCTAGACTGCAGTGGATGATGATGTCTCTTCTCTAATC-3′. The C-terminal half of the protein was obtained using three PCR fragments with AX3 genomic DNA as a template using the primer pairs: 5′-TGGATCCAAAAATGCTGCAGTCATTAAAAGAATCAAATGG-3′ with 5′-GTTATCTAGATTCGAACTACCACCACCATTGCC-3′; 5′-TGGATCCAAAAATGAGTTCGAATGGTGCGGGTGGTGG-3′ with 5′-TGCTAGCACTAACTGATTGTTGTA-3′; and 5′-TGCTAGCTCAACAACACCACCTGTAC-3′ with 5′-GTCTAGATTAAGTTAAATTACGGACAC-3′ (XbaI site bold). Finally, the complete gbpD gene was released with BamHI and XbaI, and cloned between the actin 15 promoter and the actin 8 terminator of MB74 using its BglII and SpeI sites. MB74 is an extrachromosomal shuttle plasmid for Dictyostelium derived from MB12NEO (Linskens et al., 1999). Approximately 1-4 μg plasmid was electroporated to the cells as described above.

To fuse the C-terminal part of GbpD with GFP, the same six primers were used as mentioned above, except that the primer 5′-GTTATCTAGAAGTTAAATTACGGACACGT-3′ (XbaI site bold) was used as the reverse primer fragment at the 3′ end of the gene to omit the stop codon. The truncated gene was released using the BamHI and XbaI sites and ligated into the BglII and SpeI sites of MB74GFP. The MB74GFP plasmid is similar to MB74 but contains the S65T GFP gene behind the SpeI site. The final fusion protein consists of a methionine followed by amino acids 585 to 1312 of GbpD, two serines and the complete S65T GFP protein.

For the fusion of the C-terminal part of GbpC with GFP, two PCR products were obtained using genomic AX3 DNA as a template with the primer pairs 5′-GAGATCTAAAAATGACCTCGAGTAATTTCTTTGGTAATGGTTC-3′ (BglII site bold) with 5′-GTTATCTAGAAGTCGACGTTTGAGTACGACCTAAATAAG-3′; and 5′-GAGATCTAAAAATGACGTCGACTTCACCATTGAATGAAGG-3′ with 5′-GTCTAGAAGCATAAAGTTGTGATTCTCT-3′ (XbaI site bold). The fragments were assembled in MB74GFP. The final protein consists of a methionine followed by amino acids 2005 to 2631 of GbpC, two serines and the complete S65T GFP protein.

LimEΔcoil-GFP was made to detect filamentous actin. This was accomplished by amplification of the fragment encoding the first 145 amino acids of LimE using the primers 5′-AAAGATCTAAAAAATGTCTGCTTCAGTTAAATG-3′ (BglII site bold) and 5′-TTACTAGTACCAGTTGGTTGACCATC-3′ (SpeI site bold) using Reverse transcriptase PCR with AX3 RNA as a template. The fragment was ligated in a modified MB74GFP vector that contained a hygromycin-instead of neomycin-resistance cassette. The hygromycin resistance cassette (kind gift of J. Williams, School of Life Sciences, Wellcome Trust Biocentre, University of Dundee, UK) is comprised of an actin 15 promoter followed by the HPH hygromycin resistance gene and the cabA terminator 5′-TAAATAAAATAAATAAATTGT-3′. The final LimEΔcoil-GFP fusion protein consists of amino acids 1 to 145 of LimE, a threonine and serine residue and the complete S65T GFP protein.

The myosin heavy chain-GFP overexpressor construct was a kind gift from T. Egelhoff. Cells expressing this construct were starved in six-well plates (Nunc) under 2 ml PB (10 mM Na/K phosphate, pH 6.5) until the onset of aggregation. Cells were washed off and transferred to a custom-made flow chamber that has been described previously (Potma et al., 2001). After adherence for 10 minutes, cells were stimulated with 10^–6 M cAMP. Confocal images were taken every 5 seconds with a confocal laser-scanning microscope (Zeiss ConfoCor 2-LSM 510 combination set up). Images were analysed with the QuimP software package version 1 (http://www.garching-innovation.de/e_akterf/e_softwa/e_softtxt/quimp/QuimP-guide/) (Dormann et al., 2002).

Actin polymerisation assay was carried out as described (Zigmond et al., 1997). Briefly, 5-hour starved and pulsed cells were incubated in PM (2 mM MgCl₂ in 10 mM Na/K phosphate, pH 6.5) with 2 mM caffeine at 3 ×10⁷ cells/ml. After 20 minutes, 1.1 ml cell suspension was stimulated at t=0 with 200 nM cAMP. At the indicated time points, 100 μl samples were taken and transferred to 1 ml AB (20 mM KPO₄, 10 mM PIPES, 5 mM EGTA, 2 mM MgCl₂, 3.7% formaldehyde, 0.1% Triton X-100, 0.25 μ M TRITC-Phalloidin, pH 6.8). After shaking for 1 hour at room temperature, samples were centrifuged for 10 minutes at 14,000 g and the resulting pellet was resuspended in 1 ml methanol. After shaking overnight, the amount of F-actin was determined by measuring the fluorescence (excitation 540 nm, emission 565 nm).

To quantify the strength of cell adhesion to the surface, we used a previously published protocol (Fey et al., 2002) with a few modifications. Cells were grown on HG5 in six-well plates from Nunc to maximum 60% confluency. The medium was replaced by 2 ml PB and the plates were incubated for 1 hour. Then, the plates were rotated on a rotary shaker at 150 rpm and 150 μl samples were taken at the indicated time points. The wells were replenished with fresh PB. After the final time point, the remaining cells were detached thoroughly by repeated pipetting. After vortexing for 15 seconds, the number of cells in the samples was determined in quadruple using a haemocytometer. The number of cells removed at previous time points was added to the amount of loose cells and divided by the total number of cells on the plate to yield the percentage of loose cells.

To determine the amount of intracellular cAMP and cGMP, 5 ×10⁸ cells were washed once with 50 ml PB and centrifuged at 450 g for 3 minutes. The cell pellet was resuspended in PB to a volume of 400 μl and an equal volume of 3.5% (v/v) perchloric acid was added. The samples were centrifuged for 5 minutes at 14,000 g, the supernatant was neutralised with 200 μl KHCO₃ (50% saturated solution), centrifuged again for 5 minutes at 14,000 g, and the supernatant was collected. The amount of cAMP and cGMP was determined using the cAMP and cGMP assay systems from Amersham. The assays contained 50 μl supernatant derived from 2.5 ×10⁷ cells. In the absence of cAMP or cGMP the assays yield 0±0.0097 pmol cAMP/assay and 0±0.0076 pmol cGMP/assay, respectively (n=6). This implies that the sensitivity of the assay is 0.024 pmol cAMP/10⁷ cells and 0.019 pmol cGMP/10⁷ cells. The data obtained for the cell extracts are in pmol/10⁷ cells (n=4): wild type, 0.20±0.05 cAMP and 0.89±0.19 cGMP; gbpD^–/GbpD^OE, 0.23±0.03 cAMP and 0.89±0.10 cGMP; gbpD^–/GbpB^OE/GbpD^OE, 0.004±0.031 cAMP and 1.01±0.09 cGMP; gca^–/sgc^–/GbpD^OE, 0.18±0.02 cAMP and 0.03±0.05 cGMP; gca^–/sgc^–/GbpB^OE/GbpD^OE, –0.004±0.016 cAMP and –0.01±0.04 cGMP. The data show that guanylyl cyclase-null cells have normal cAMP levels, but no detectable cGMP (detection limit 0.19 pmol/10⁷ cells), whereas cells overexpressing GbpB have normal cGMP levels and no detectable cAMP (detection limit 0.024 pmol/10⁷ cells).

To initiate development, growth phase cells were washed in buffered salts solution (BSS) containing 20 mM KCl, 2.5 mM MgCl₂ and 20 mM KH₂PO₄ (pH 6.4), and dispersed on a black filter pad saturated with BSS at a density of 5 ×10⁶ cells per cm². Amoebae were washed off the filter pads at the ripple stage, which in dense cultures represents the onset of aggregation, the time at which Dictyostelium amoebae attain their highest average velocity (Varnum and Soll, 1984). In this case, the onset of aggregation was observed between 8 and 11 hours, depending on the particular strain.

The motile behaviour of cells in buffer and in spatial gradients of cAMP was analysed using computer-assisted methods previously described (Soll, 1995; Soll et al., 2001). Briefly, cells were dispersed on the bridge of a Plexiglas gradient chamber, in which one of the two troughs bordering the bridge contained BSS and the other trough contained BSS plus 10^–6 M cAMP. Cells were video recorded through a 25 × objective with bright field optics for a 10 minute period following an initial 5 minute incubation period necessary for establishing a steep gradient of cAMP. Images were digitised at a rate of 15 frames per minute. 2D-DIAS software then automatically outlined the cell perimeters and converted them to beta-spline replacement images from which the position of the cell centroid was determined. Chemotactic index, speed and change of direction were computed from the centroid position as described (Soll, 1995; Soll et al., 2001).

Cells were starved for 1 hour in starvation buffer (40 mM HEPES/NaOH, pH 7.0), washed and resuspended at a density of 2 ×10⁸ cells/ml in starvation buffer supplemented with 0.25 M sucrose, 0.5 mM EDTA, 0.5 mM EGTA and one protease inhibitor cocktail tablet (mini Complete-tablets, EDTA-free from Roche) per 10 ml buffer. Cells were immediately lysed by passage through a 0.45 μm Nuclepore filter. The lysate was centrifuged for 2 minutes at 14,000 g and the supernatant was centrifuged again for 4 minutes at 14,000 g. 100 μl of the newly obtained supernatant was added to 100 μl assay buffer containing 40 mM HEPES/NaOH pH 7.0, 0.5 mM EGTA, 6 mM MgCl₂, 10 mM dithiothreitol, 10nM 8-Br-cAMP, 100 nM [³H]cGMP. This mixture was incubated on ice for 20 minutes followed by filtration of the mixture over nitrocellulose filters (pore size 0.45 μm) and the filters were washed twice with 3 ml ice-cold 40 mM HEPES/NaOH pH 7.0. After drying, 3 ml scintillation liquid was added and the radioactivity associated with the filters was determined using a liquid scintillation counter. To determine non-specific binding 500 μM cGMP was included in the assay mixture. For measurement of the binding of 5 nM [³H]cGMP (Fig. 6B), the same procedure was applied, with a few modifications: the precleared lysate was centrifuged for 60 minutes at 48,000 g; assay buffer contained 100 mM PB, 10 mM dithiothreitol, 6 mM MgCl₂ and 5 nM [³H]cGMP; 5 μM unlabelled cGMP was used to determine non-specific binding and 50 mM PB was used to wash the filters.

To study the role of GbpC and GbpD, we have created cell lines in which either the gbpC or gbpD gene is disrupted. This was accomplished by homologous recombination in the Dictyostelium uracil auxotroph DH1, using pyr5/6 as a selection marker for gbpD and Bsr for gbpC. A double knockout was obtained by inactivating the gbpC gene in the gbpD knockout, as described (Bosgraaf et al., 2002a). All knockout cells grew with a normal rate in axenic medium and on bacterial lawns, and developed normally, although gbpC^– mutant cells made somewhat smaller fruiting bodies (data not shown).

We subjected starved gbpC^– and gbpD^– single knockout cells to a chemotaxis analysis using the DIAS software package. When placed in a spatial gradient of cAMP, gbpC^– cells failed to become elongated and moved much more slowly and less persistently than wild-type cells (see Table 1 and Fig. 1A). Moreover, the chemotaxis index, which is defined as the movement towards the cAMP source divided by the total cell movement, was greatly diminished. The polarity of a cell is reflected in the parameter `roundness'; a perfect circle would have a roundness of 100 whereas a straight line would have a roundness of 1. As reported in Table 1, gbpC^– cells were significantly rounder than wild-type cells. These characteristics closely resemble the phenotype of the gbpC^–/gbpD^– cells that was described before (Bosgraaf et al., 2002a).

As opposed to the gbpC^– and gbpC^–/gbpD^– double knockout cells, gbpD^– cells appeared very polar and moved with increased speed and efficiency towards the cAMP source (see Table 1 and Fig. 1A). The increased polarity is reflected in a decreased roundness. It was also observed that the gbpD^– cells often wiggled for a while with only the uropod attached to the substratum. This was accompanied by a looser adherence to the substratum of gbpD^– cells in general (see below). In fact, the decreased substrate adhesion is presumably the primary cause of the increased cell speed and chemotaxis performance. Although the tracks of some gbpD^– cells undergoing chemotaxis suggest that the cells might turn by making lateral pseudopodia or bifurcation of the anterior pseudopod, separate images of the cells taken at different time points clearly indicate that the cells almost never made lateral protrusions (Fig. 1B). Rather, the cells turned by making a new anterior pseudopodium to a slightly different direction. Indeed, as shown in Table 1, gbpD^– cells made on average 0.46 lateral pseudopodia per 10 minutes, a significant decrease compared to wild-type cells that made 1.33 lateral pseudopodia in 10 minutes. Furthermore, the frequency of anterior pseudopod bifurcation, which often leads to a direction change, was lowered in gbpD^– cells (1.76 per 10 minutes) when compared to wild-type cells in which 4.88 pseudopodia bifurcated per 10 minutes (Table 1). These findings indicate that gbpD^– cells displayed enhanced chemotaxis due to an increased integrity of the leading pseudopod and a reduced tendency to produce lateral pseudopodia.

In order to rescue the phenotype of the gbpD^– mutants, the gbpD gene was overexpressed using an extrachromosomal expression plasmid. Surprisingly, the resulting gbpD^–/GbpD^OE cells were large, flattened and contained many filopodial protrusions and lamellipodia at the substrate surface (described in detail below). Furthermore, this mutant attached more strongly to the surface than wild-type cells (see below). The gbpD^–/GbpD^OE cells were unable to aggregate and this could not be rescued by synergistic development with wild-type cells (not shown). The gbpD^–/GbpD^OE cells displayed very poor chemotaxis when subjected to a cAMP gradient (Fig. 1A, Table 1). The cells produced many pseudopodia that were often bifurcated or filopod-like. Nevertheless, the cells were able to move directionally, albeit with decreased speed. The reduction of cell speed is probably caused by the increased substrate adhesion and the greater number of pseudopodia produced. Although the cells were able to move persistently, the direction of movement was almost random, an observation that was confirmed by the calculated chemotaxis index that was only 0.06 for this mutant. This lack of chemotaxis was presumably caused by the large amount of lateral pseudopodia that the cells formed (four times more than seen in the wild type) and the extreme extent of pseudopod bifurcation (three times more than that in the wild type), which hindered the formation of a dominant anterior pseudopod (Table 1). We observed that about 10% of the cells had a much higher speed and displayed clear positive chemotaxis. As expression levels of ectopic genes are generally very heterogeneous in Dictyostelium, we believe that these cells had very low expression levels of gbpD. When these cells were excluded from the chemotaxis analysis, the chemotaxis index dropped to –0.01 and the percentage of cells that moved in the direction of the gradient was only 55%, very close to random movement, which would result in 50% of the cells moving towards the chemoattractant and 50% away. These data illustrate that overexpression of GbpD caused a near total loss of the ability to undergo chemotaxis.

As mentioned before, gbpD^– cells seemed less well attached to the substratum than wild-type cells, whereas gbpD^–/GbpD^OE cells appeared very adhesive. Therefore the adhesion capability of these mutants was quantified by shaking petri dishes seeded with cells and counting the number of cells that detached from the substratum at different time points. The gbpD^– cells detached much faster from the substratum when compared to wild-type cells (Fig. 2), whereas gbpD^–/GbpD^OE cells were much more resistant to detachment. These data suggest that GbpD plays a role in attachment of the cells to the substratum.

As gbpD^–/GbpD^OE cells produced many filopodial extensions, we were interested in the actin distribution in this cell line. To visualise filamentous actin in living cells, the first 145 amino acids of the F-actin binding protein LimE were fused to GFP (LimEΔcoil-GFP). This part of the protein lacks the C-terminal coiled coil domain but retains the ability to bind F-actin (Schneider et al., 2003). When gbpD^– cells were transformed with this construct, fluorescence was typically enhanced in leading edges of starved cells (Fig. 3). Furthermore, when a confocal image was taken just above the glass surface, F-actin-rich foci could be detected, as described before (Uchida and Yumura, 2004). Thus, the distribution of F-actin appears normal in gbpD^–cells. In contrast, when gbpD^–/GbpD^OE mutants were transformed with the same construct, many GFP-labelled protrusions were visible in these cells (Fig. 3). A cross section of a cell revealed that they are more flattened than the parental strain. Measurements of 19 different cells revealed that the contact area is 80% larger than in gbpD^– cells whereas the volume has decreased slightly (Table 2). As actin foci transmit force to the substratum (Uchida and Yumura, 2004) and overexpression of GbpD protein caused a much stronger attachment to the substrate, it might be expected that the amount of actin foci would also be higher in gbpD^–/GbpD^OE cells. However, the number of actin foci in gbpD^–/GbpD^OE cells was not significantly higher than in gbpD^– cells (Table 2). On the other hand, the number of protrusions that made surface contact was significantly higher in gbpD^–/GbpD^OE cells compared to the parental gbpD^– cells. These findings indicate that overexpression of GbpD protein causes an increase in the amount of substrate-attached protrusions, which results in a flattened cell morphology and stronger substrate adhesion.

The mutants were subjected to a phalloidin binding assay to measure cAMP-induced changes in the amount of filamentous actin. For this experiment, the cells were kept in suspension and pulsed with exogenous cAMP for 4 hours. A typical experiment is shown in Fig. 4. In wild-type cells, a twofold increase in the amount of F-actin was observed that reached its maximum 4-8 seconds after cAMP stimulation, whereas basal levels were recovered after 40-60 seconds. The actin response of both the gbpD^– and the gbpC^– cells displayed no obvious aberrancies. As the gbpD^–/GbpD^OE cells produce many pseudopodia containing F-actin when attached to a surface, it might be expected that the total F-actin content would be elevated before stimulation in these cells. Surprisingly, we did not observe a significant difference in basal F-actin levels in this or any of the mutants in our experiments (data not shown). However, the gbpD^–/GbpD^OE cells did show a somewhat aberrant F-actin response upon cAMP-stimulation: maximum F-actin levels were lowered to around 70% of that of the wild type. Moreover, pre-stimulus levels were not reached until 90 seconds after stimulation in this mutant, compared to 45 seconds in wild-type cells. We conclude that overexpression of GbpD protein mildly affects the F-actin response, whereas disruption of gbpC or gbpD has no obvious effect on the actin response.

Stimulation with extracellular cAMP is known to induce myosin II translocation to the cortex, a process that depends on intracellular cGMP (Bosgraaf et al., 2002a). Therefore the cAMP-induced myosin II translocation was investigated in gbpC^– and gbpD^– cells. This was accomplished by cAMP stimulation of Gbp mutant cells overexpressing GFP-myosin II and taking confocal images at 5-second intervals, as described (Levi et al., 2002). We observed that in wild-type cells, a small but significant portion of the myosin-GFP was present at the cell membrane prior to stimulation (Fig. 5A). This was not overcome by prolonged incubation with caffeine that inhibits the cAMP production of the cells. Within 10 seconds after stimulation with exogenous cAMP, most of this membrane-associated myosin was released into the cytosol (Fig. 5B). This depletion was followed by a rapid and enhanced retranslocation of myosin-GFP to the membrane which was maximal about 30 seconds after stimulation with cAMP. Myosin was subsequently released again from the membrane around 50 seconds after cAMP addition. During the first 60 seconds after cAMP stimulation, cell movement halted. After this period cells started to move again, and myosin-GFP was primarily found in the cytosol and at the membrane in the typical `C-shape' located at the posterior of the cell (data not shown).

In contrast to wild-type cells, no cAMP-induced myosin II translocation could be detected in gbpC^– cells during the first 80 seconds after stimulation (Fig. 5B). Instead, a prolonged depletion of myosin-GFP from the membrane was observed, that recovered to pre-stimulus levels after 50 seconds. These data correspond well with our previous finding that cAMP-induced myosin II translocation is abolished in gbpC^–/gbpD^– cells. The gbpD^– cells were also transfected with myosin-GFP. When these cells were stimulated with cAMP, myosin-GFP translocation was similar to that observed in wild-type cells (Fig. 5B); myosin-GFP dissociated from the membrane during the first 15 seconds and re-associated with the membrane immediately thereafter for a period of about 45 seconds. Possibly the initial release of myosin II from the membrane is somewhat stronger in the gbpD^– cell line than in wild-type cells. We conclude that GbpC is necessary for cAMP-induced myosin translocation to the cortex whereas disruption of the gbpD gene does not influence this process.

GbpD contains a RasGEF domain and two cyclic nucleotide binding domains. These domains are also present in human Epac and PDZ-GEF, but only Epac is stimulated by cAMP whereas the RasGEF activity of PDZ-GEF is not regulated by cAMP or cGMP (de Rooij et al., 1998; Kuiperij et al., 2003; Rehmann et al., 2003). To test whether GbpD is regulated by cyclic nucleotides, the binding properties of this protein were studied by overexpressing GbpD. It has previously been determined that GbpC specifically binds cGMP with an affinity of 4 nM (Bosgraaf et al., 2002a). As such high affinity binding activity obscures the measurement of lower affinity components in cell lysates, the gbpC^–/gbpD^– mutant was used as a background strain to overexpress GbpD protein. Under the conditions we tested, we could not measure any significant increase in cAMP or cGMP binding activity in the lysates of gbpC^–/gbpD^–/GbpD^OE cells when compared to the control cell line (Fig. 6A). We did, however observe a significant decrease in the amount of cAMP binding in cells overexpressing GbpD protein. cAMP binding to lysates of wild-type cells is predominantly to the regulatory subunit of PKA (Mutzel et al., 1988). Therefore, the decreased cAMP binding capacity in lysates of GbpD-overexpressing cells could be caused by a decreased expression of PKA. Because of this, and as overexpression of GbpD protein causes a severe change of cell morphology and behaviour, the C-terminal half of GbpD fused with GFP at its C-terminus was also expressed. This construct, ΔNGbpD-GFP, contains both cyclic nucleotide binding domains but does not contain the rasGEF domain. The corresponding truncated GbpC-GFP fusion protein, which is homologous to GbpD, retains its ability to bind cGMP (Fig. 6B). The ΔNGbpD-GFP fusion protein could be expressed at high levels without affecting the morphology of the cells (data not shown). However, we did not observe any significant increase in cAMP or cGMP binding activity compared to the background strain (Fig. 6A). This result was unaffected by clearing the cell lysate using high-speed centrifugation or changing the assay buffer to phosphate buffer (data not shown). Because the ΔNGbpD-GFP fusion protein was partly localised at the membrane, the membrane fraction of the cell lysates was also investigated for increased cyclic nucleotide binding. Also in this case, the expression of the ΔNGbpD-GFP fusion protein did not result in a measurable increase in cyclic nucleotide binding capacity (data not shown). We also tried to immunopurify the ΔNGbpD-GFP fusion protein using a pulldown assay with GFP antibody. Again, no increase in binding capacity was found in ΔNGbpD-GFP expressing cells compared to control cells (data not shown). Thus, at present we do not have any evidence that GbpD binds either cAMP or cGMP in vitro.

Although we do not have any indication that GbpD binds cyclic nucleotides in vitro, the increased adhesion strength of cells overexpressing GbpD might depend on cyclic nucleotides. To test this, mutants with altered cyclic nucleotide levels were transformed with the GbpD overexpression construct. In the gca^–/sgc^– cell line, the two known guanylyl cyclases, guanylyl cyclase A and soluble guanylyl cyclase, have been disrupted and these cells no longer have detectable cGMP (see Materials and Methods) (Roelofs and Van Haastert, 2002). When GbpD was overexpressed in this mutant, the same increase in attachment was found as in gbpD^–/GbpD^OE cells (Fig. 7). Thus, we conclude that the role of GbpD in attachment of the cells to the substratum is not cGMP dependent. To test whether cAMP is needed for GbpD function, GbpB was overexpressed in the GbpD-overexpressing mutants. GbpB is a very active phosphodiesterase that predominantly hydrolyses cAMP, and it was previously shown that overexpression of this enzyme results in very low cAMP levels (Bosgraaf et al., 2002b). Overexpression of GbpB in gbpD^–/GbpD^OE cells led to a very strong decline of the cAMP concentration to undetectable levels (below 0.024 pmol/10⁷ cells compared to 0.23±0.03 pmol/10⁷ cells in the parental strain; see Materials and Methods). The adhesion strength of the gbpD^–/GbpD^OE/GbpB^OE mutant was virtually identical to that of gbpD^–/GbpD^OE cells as judged by the amount of cells that detached from the surface after shaking the plate for 1 hour (Fig. 7). Thus the GbpD-induced increase in adhesion strength does not depend on the presence of cAMP either. However, it might be possible that the cyclic nucleotide binding domain(s) of GbpD does not discriminate between cAMP and cGMP. In that case, the presence of either cyclic nucleotide could suffice to activate the protein. Therefore, GbpD was also overexpressed in gca^–/sgc^–/GbpB^OE cells, in which the level of either cyclic nucleotide is below the detection limit (see Materials and Methods). In this strain, GbpD overexpression also resulted in dramatically increased adhesion strength, although to slightly lower levels than in wild-type cells (Fig. 7). These findings strongly suggest that GbpD can function in the absence of both cGMP and cAMP, as has been observed for human PDZ-GEF.

Deletion of gbpC yields cells that display reduced chemotaxis due to slower movement and reduced polarity. This phenotype closely matches that of gbpC^–/gbpD^– cells that has been reported (Bosgraaf et al., 2002a). In contrast to gbpC^– cells, gbpD^– cells displayed enhanced chemotaxis caused by a reduced tendency to make lateral pseudopodia and a higher speed. The latter is presumably caused by the reduced substrate adhesion that was observed. Cells overexpressing GbpD protein were very flattened and produced an unusual amount of lateral pseudopodia that often bifurcated. These cells attached more strongly to the substrate, which led to a decreased cell speed. Moreover, cells were no longer capable of aggregation, and chemotaxis was severely impaired owing to a lack of directionality and polarity. These findings indicate that GbpC is needed for polarity and efficient chemotaxis, whereas GbpD induces the formation of substrate attached pseudopodia, thereby negatively influencing cell speed and polarity.

Although the chemotaxis index of gbpC^– cells was close to zero, this mutant is still capable of aggregation on agar in a relatively normal way. One possible explanation is that when plated at high density, cells probably need little chemotaxis in order to form aggregates. Another reason could be that gbpC^– cells display better chemotaxis when placed on a different surface; we observed that gbpC^– cells have improved, but still aberrant chemotaxis when placed on a glass surface instead of the Perspex surface that was used for the experiments described in this paper. The chemotaxis index increased to about 0.3 in this case (L.B. and P.J.M.v.H., unpublished). In contrast to this, GbpD overexpressing cells behaved identical on glass and Perspex surfaces, displaying highly impaired chemotaxis (L.B. and P.J.M.v.H., unpublished). In accordance with this observation, this mutant was unable to aggregate or form fruiting bodies.

The faster cell movement that was observed in starved gbpD^– cells was presumably caused by a weaker adhesion to the substratum. Moreover, overexpression of GbpD protein caused a dramatic increase in the adhesion strength to the substratum and a concurrent decrease in cell speed. Recently it was shown that actin foci are located close to the substratum and transmit traction force to the substratum, indicating that these are the sites of cell substrate anchoring (Uchida and Yumura, 2004). We therefore expected that overexpression of GbpD protein would lead to a higher number of actin foci. We found that although the amount of actin foci is higher in cells overexpressing GbpD protein, it is not significantly different from the values of the parent strain. This could be due to the large spread that was found in the amount of foci per cell. However, if the number of foci per cell is divided by the contact area (which might be a more correct measure), the values are virtually identical in gbpD^–/GbpD^OE and gbpD^– cells (0.61 and 0.62 respectively). We propose that the significantly increased number of surface attached protrusions in gbpD^–/GbpD^OE cells causes the enhanced surface adhesion and flattening of the cells.

We found that in gbpC^– cells, cAMP-induced myosin-GFP translocation to the cortex was lost, whereas gbpD^– mutants displayed no aberrancies. It is noteworthy that in gbpC^– cells moving in buffer, myosin-GFP was still localised in the posterior half of the cell (data not shown). Therefore, there must be a GbpC-independent mechanism that mediates myosin II translocation to the posterior half of unstimulated cells, although cAMP-induced myosin translocation is dependent on GbpC.

The use of GFP-labelled myosin was recently used to study cAMP-regulated myosin behaviour (Levi et al., 2002). In that study, the myosin-GFP translocation was not preceded by a depletion from the membrane and was stronger and more prolonged. These differences are possibly caused by the distinct treatment of the cells; in our study, cells were starved on solid support under phosphate buffer until the ripple stage was reached, whereas Levi and co-workers applied periodic cAMP pulses to cells that were kept in suspension by shaking (Levi et al., 2002).

As in gbpC^–/gbpD^– cells (Bosgraaf et al., 2002a), cAMP-induced MHC phosphorylation was slightly decreased in gbpC single knockout cells (L.B., P.J.M.v.H. and J. Smith, unpublished). As diminished heavy chain phosphorylation should lead to increased filament formation, this cannot account for the observed decrease in cAMP-induced MHC translocation in gbpC^– cells. Myosin II translocation is also known to be dependent on the presence of filamentous actin (Levi et al., 2002), therefore GbpC could mediate the cAMP-induced binding of myosin to F-actin.

There are at least two other proteins that also participate in myosin II translocation to the cortex. One is RegA (Wessels et al., 2000), a cAMP-selective phosphodiesterase that also serves as a response regulator (Thomason et al., 1999). The other is PAKa, a serine/threonine protein kinase that colocalises with myosin II in migrating cells (Chung and Firtel, 1999; Muller-Taubenberger et al., 2002). As the mechanisms by which these proteins influence myosin II translocation are currently unknown, it remains to be determined whether they interact with GbpC.

As gbpD^–/GbpD^OE cells produce many lamellipodia and pseudopodia, we expected a higher level of filamentous actin in unstimulated cells. However, this was not observed. A possible explanation is that the cells were kept in suspension during the actin polymerisation assay. The cells probably become less flattened and possibly lose their pseudopodia under these circumstances. Furthermore, F-actin was enriched primarily at the surface contact area in cells overexpressing GbpD protein. At 2 μm above the surface, F-actin distribution appeared normal. As suspended cells do not experience any surface, cells overexpressing GbpD protein could behave relatively normally when kept in suspension. The cAMP-induced actin polymerisation of gbpD overexpressing cells was altered compared to wild-type cells; recovery to basal levels was significantly delayed. This indicates that GbpD acts on a signal transduction pathway that influences actin polymerisation.

Although GbpD contains two cyclic nucleotide binding domains, we could not measure any increase of cAMP or cGMP binding activity in cell lysates of cells overexpressing GbpD protein. The same result was obtained when we overexpressed the GFP protein fused to the C-terminal half of GbpD, containing only the cyclic nucleotide binding domains and the GRAM domain. Lysates of cells overexpressing the corresponding C-terminal part of GbpC fused to GFP retained their cGMP binding capacity. Furthermore, the fluorescent signal observed in ΔNGbpC-GFP and ΔNGbpD-GFP expressing cells was comparable, indicating that the protein expression levels were approximately the same. The cGMP-binding capacity of this GbpC-overexpressor strain as determined by Scatchard plots is approximately 0.8 pmol/mg protein (data not shown). Assuming the same protein expression level of GbpD, we can calculate the maximal possible dissociation constant for cAMP and cGMP that is still consistent with the data. The s.e.m. represents the maximal binding activity, which is 0.125 pmol/mg for cGMP and 0.045 pmol/mg for cAMP (ΔN-GbpD-GFP expressing cells; error bars in Fig. 6A). As 50 nM label was used in these experiments, the dissociation constant of GbpD must be higher than 270 nM for cGMP and higher than 840 nM for cAMP. In wild-type cells the concentration of cGMP and cAMP reaches levels up to 2 μM during stimulation, so it cannot be totally excluded that GbpD is regulated by cyclic nucleotides during peak concentrations. However, the striking flattened phenotype and the strong substrate adhesion were not diminished when GbpD was overexpressed in cell lines that have undetectable cAMP and/or cGMP concentrations (see Materials and Methods section). This indicates that GbpD can act without cyclic nucleotides, and we conclude that GbpD is not strongly regulated by intracellular cAMP or cGMP. In this respect, it is noteworthy that a family of proteins has been identified in mammalian cells that also contain RasGEF and cyclic nucleotide domains (Kuiperij et al., 2003). Two of these, Epac1 and Epac-2, bind cAMP with physiological affinity and specifically activate Rap in a cAMP-dependent manner (de Rooij et al., 1998; Rehmann et al., 2003). However, two other proteins of this family, PDZ-GEF 1 and 2, display guanine nucleotide exchange activity but are not activated by cAMP or 8Br-cGMP (Kuiperij et al., 2003). Furthermore, the purified cyclic nucleotide binding domains of PDZ-GEF 1 do not bind cAMP, cGMP or a number of other small molecules (de Rooij et al., 1999; Kuiperij et al., 2003). Thus, like GbpD, the PDZ-GEFs also seem to function without cyclic nucleotides. Remarkably, these proteins specifically activate Rap1, which is an important regulator of integrin-mediated cell adhesion (Bos et al., 2003). We have shown in this article that GbpD also influences cell adhesion. Moreover, as overexpression of the Dictyostelium Rap1 homologue induces the same flattened cell morphology as GbpD overexpression (Rebstein et al., 1993), Rap1 may well be a target of GbpD.

In summary, we conclude that GbpC and GbpD have differential functions in Dictyostelium chemotaxis and probably act via different signal transduction pathways; GbpC is regulated by cGMP and is needed for chemoattractant-induced myosin translocation and cell polarisation. On the other hand, GbpD does not depend on intracellular cyclic nucleotides and induces surface attached lateral pseudopodia, which provokes depolarisation and enhanced substrate adhesion leading to reduced cell motility. Thus, these two proteins independently affect chemotactic motility in opposing manners. The exact temporal and spatial regulation of GbpC and GbpD proteins might play an important role in the chemotactic movement of Dictyostelium cells.

We thank Janet Smith for stimulating discussions on the gbpC knockout data. We thank Thomas Egelhoff and Jeffrey Williams for providing the plasmids encoding myosin-GFP and hygromycin resistant gene, respectively.