Register for Mechanobiology 2016!


Through analysis of a chemotaxis mutant obtained from a genetic screen in Dictyostelium discoideum, we have identified a new gene involved in regulating cell migration and have named it costars (cosA). The 82 amino acid Costars protein sequence appears highly conserved among diverse species, and significantly resembles the C-terminal region of the striated muscle activator of Rho signaling (STARS), a mammalian protein that regulates the serum response factor transcriptional activity through actin binding and Rho GTPase activation. The cosA-null (cosA) cells formed smooth plaques on bacterial lawns, produced abnormally small fruiting bodies when developed on the non-nutrient agar and displayed reduced migration towards the cAMP source in chemotactic assays. Analysis of cell motion in cAMP gradients revealed decreased speed but wild-type-like directional persistence of cosA cells, suggesting a defect in the cellular machinery for motility rather than for chemotactic orientation. Consistent with this notion, cosA cells exhibited changes in the actin cytoskeleton, showing aberrant distribution of F-actin in fluorescence cell staining and an increased amount of cytoskeleton-associated actin. Excessive pseudopod formation was also noted in cosA cells facing chemoattractant gradients. Expressing cosA or its human counterpart mCostars eliminated abnormalities of cosA cells. Together, our results highlight a role for Costars in modulating actin dynamics and cell motility.


In multicellular organisms, cell motility is a fundamental mechanism for various physiological and pathological processes (Gillitzer and Goebeler, 2001; Koizumi et al., 2007; Lecaudey and Gilmour, 2006; Moser et al., 2004). Events such as embryonic development, immune responses, wound healing and cancer metastasis involve directional migration, where cells respond to gradients of diffusible signaling molecules by chemotaxis and move towards the attractant source. Effective chemotactic movement requires coordination of a number of distinct steps; the cell needs to recognize a chemical gradient, produce appropriate intracellular signals through signal transduction, establish polarity by asymmetrical distribution of effectors, dynamically rearrange cytoskeletal components and, finally, generate motion. Owing to the complexity of the event and the underlying signaling network, eukaryotic chemotaxis still defies full description at the molecular level.

Dictyostelium discoideum is an excellent model for studying cell motility and chemotaxis. Naturally living single D. discoideum amoeboid cells hunt bacteria for food by chemotaxing towards bacteria-secreted folate. Upon food depletion, starving cells enter a developmental program where individual amoebae aggregate to form multicellular structures. The initial aggregation process is organized by cAMP secreted from cells at aggregation centers; surrounding cells respond by moving chemotactically towards the signaling cells and relaying the signal to cells farther from the center. The resulting multicellular aggregates undergo further differentiation and morphogenesis – which also involves chemotactic cell movements – to form mounds, then slugs and, eventually, fruiting bodies (Annesley and Fisher, 2009). D. discoideum possesses some unique advantages for an experimental system. Their cells are easy to culture, either xenically on bacteria or axenically in simple media. Molecular genetic tools and cell biological approaches are available to manipulate the expression of specific proteins and perform phenotypic analyses (Gerisch et al., 1995; Howard et al., 1988; Janetopoulos et al., 2001; Kuspa and Loomis, 1992b; Manstein et al., 1989; Nellen et al., 1984). An annotated genome database ( facilitates both forward and reverse genetic studies (Eichinger et al., 2005; Fey et al., 2009). In particular, D. discoideum can exist as a haploid and, hence, disrupting a single copy of a specific gene would reveal the phenotype of the mutation, which sets the groundwork for powerful genetic screening.

Progress has been made in elucidating the molecular mechanisms and signaling pathways associated with D. discoideum chemotaxis. A cell-surface G-protein-coupled receptor (GPCR) detects the chemoattractant cAMP, triggering multiple transient responses, such as cGMP accumulation, adenylate cyclase activation and actin polymerization (Klein et al., 1987; Kumagai et al., 1989; Saxe et al., 1988; Wu et al., 1995). Some elicited events participate in mediating directional sensing. For example, upon receptor and G-protein excitation, D. discoideum phosphoinositide 3-kinase (PI3K) is activated and recruited from cytoplasm to plasma membrane at the leading edge; by contrast, the phosphatase and tensin homologue deleted on chromosome 10 (PTEN) dissociates from the membrane at the cell front, while it remains bound at the back (Funamoto et al., 2002; Iijima and Devreotes, 2002; Janetopoulos et al., 2004). The reciprocal localization of PI3K and PTEN confines phosphatidylinositol (3,4,5)-trisphosphate [PtdIns(3,4,5)P3] to the newly formed leading edge, helping to generate and/or keep a steep intracellular anterior–posterior PtdIns(3,4,5)P3 gradient (Huang et al., 2003; Iijima and Devreotes, 2002). The localized PtdIns(3,4,5)P3 accumulation acts as a docking site for proteins with a pleckstrin homology (PH) domain or lysine-rich PtdIns(3,4,5)P3-binding motif (Insall et al., 1994; Meili et al., 1999; Myers et al., 2005). Other proteins, such as myosin II, move toward the rear of chemotaxing cells (Steimle et al., 2001). These asymmetric protein distributions establish the cell polarity, serving to increase the efficiency of chemotaxis. For directional migration, signaling pathways eventually converge upon the motility machinery, which involves the remodeling of cytoskeleton. At the leading edge, Rac-dependent signaling leads to the activation of the adaptor proteins Scar and WASP, stimulating the actin-nucleating activity of the Arp2/3 complex to produce a network of branched actin filaments for pseudopod extension (Bear et al., 1998; Chung et al., 2000; Insall et al., 2001; Myers et al., 2005). The suppression of lateral pseudopods and the contraction of the trailing edge are also important for efficient migration. The suppression of lateral pseudopod formation requires PTEN function (Wessels et al., 2007). Rear-end retraction depends on myosin assembly, the regulation of which involves PakA, cGMP-modulated mechanisms, and myosin heavy-chain and light-chain phosphorylation (Bosgraaf et al., 2002; Bosgraaf et al., 2005; Chung et al., 2001; Egelhoff et al., 2005; Yumura et al., 2005). Despite current understanding, there are still missing links in the molecular pathways that lead to chemotaxis.

By taking advantage of D. discoideum as a model system, we searched for new molecular players involved in regulating directional cell migration by a ‘forward genetics’ approach. A Transwell-based enrichment-screening scheme for isolating chemotaxis mutants was established. Through analysis of one mutant, we uncovered a new gene, which we named costars (cosA). We present evidence linking Costars function to regulation of the actin cytoskeleton, pseudopod numbers and cell motility.


Identification of the costars gene and Costars-like sequences

To study chemotaxis mechanisms, we launched a genetic screen in D. discoideum for chemotaxis-defective mutants (Fig. 1A). A library of insertional mutants was created by restriction enzyme-mediated integration (REMI) mutagenesis (Kuspa and Loomis, 1992a). To enrich for mutants that are unable to chemotax, pools of REMI transformants were developed to the aggregation stage and placed in the upper chamber of a Transwell device together with cAMP in the lower chamber; cells remaining in the upper chamber after a period of incubation were collected. Individual mutants obtained from this enrichment step were subsequently subjected to two chemotaxis tests, i.e. an agar-cutting assay and a trough assay; clones behaving abnormally in both assays were collected as candidates for chemotaxis mutants.

Fig. 1.

Identification of costars. (A) Summary of the screen for chemotaxis mutants. (B) Developmental phenotype of the T21#14 mutant. Photographs were taken after 5 days or 36 hours of development on bacterial lawn or non-nutrient agar, respectively. Wt, Ax2 cells. Scale bars: 2 mm (bacterial lawn) or 200 μm (non-nutrient plate). (C) The T21#14 genotype. Arrows represent ORFs. pYW3, REMI plasmid; Bsr, blasticidin resistance expression cassette; E, EcoRV sites cut and ligated in cloning the REMI-flanking genomic fragments. (D) Schematic alignment of D. discoideum Costars and mammalian STARS proteins. Hs, Homo sapiens; Mm, Mus musculus. # aa, number of aa in each protein; %Idt and %Sml, percentages of identity and similarity, respectively, in sequences comparing STARS and Costars. (E) RT-PCR analysis of cosA expression in total RNA isolated from Ax2 (Wt), the original REMI mutant (T21#14) and two independent re-knockout cosA clones (Reko6 and Reko7). Expression of actin was used as internal control. (F) Rescue of the developmental defect through expression of Costars or mCostars in re-knockout cosA cells. Photographs were taken after 5 days or 36 hours of development on bacterial lawn or non-nutrient agar, respectively. Scale bars: 2 mm (bacterial lawn) or 500 μm (non-nutrient plate). (G) Expression of the cosA gene during development. Total lysates prepared from wild-type cells at vegetative (Veg), pre-aggregation (Pre), aggregation (Agg), mound (Mnd), slug (Slg) and fruiting body (Fb) stages were subjected to western blotting using a polyclonal anti-Costars antiserum. Actin was used as a loading control.

One of the collected mutants was the original REMI mutant, T21#14, which produced smooth plaques on bacterial lawns (Fig. 1B), appearing unable to aggregate or undergo further morphogenesis. On non-nutrient agars, T21#14 cells formed smaller fruiting bodies compared with the wild-type Ax2 strain. We recovered a genomic fragment containing the REMI insertion in T21#14 and sequence analysis revealed that the insertion was located 50 bp upstream of DDB0202716, a 249-bp uncharacterized open reading frame (ORF) (Fig. 1C). Using BLAST searches, we found the predicted 82-amino-acid (aa) product significantly resembled the C-terminal region of the striated muscle activator of Rho signaling (STARS) (Fig. 1D), which is a mammalian actin-binding Rho-activating protein (Arai et al., 2002). We hence named the ORF costars (cosA) for its homology to the C-terminal region of STARS.

Searches in the NCBI database revealed that Costars is a highly conserved protein and homologous sequences exist among diverse species, ranging from D. discoideum to human. A multiple sequence alignment revealed the high degree of homology among Costars-like proteins (supplementary material Fig. S1A). A phylogenetic tree was constructed (supplementary material Fig. S1B) and we found that even homologs appearing distant in the tree share remarkable homology. For example, sequences of D. discoideum Costars and the human gene equivalent (C6ORF115) – hereafter referred to as human mCostars – are 69% identical and 83% similar (see accession no. NP_067066.1 for mammalian Costars).

We transformed into Ax2 cells the linearized recovered plasmid to regenerate the T21#14 genotype by homologous recombination. Independent cosA ‘re-knockout’ clones were identified and their developmental morphology recapitulated the phenotype of T21#14 (Fig. 1B), indicating that the REMI insertion indeed caused the observed defect. Because the REMI insertion was located upstream and did not disrupt the cosA ORF, we checked whether cosA expression is affected in the mutants. Reverse transcription (RT)-PCR analysis failed to detect the cosA transcript in RNA samples isolated from T21#14 or re-knockout clones (Fig. 1E), indicating that they were cosA-null (cosA) cells. We constructed a D. discoideum vector for expressing green fluorescent protein (GFP)-tagged Costars under the control of the promoter of an actin gene (act15), and transformed it into cosA cells. Results showed that GFP–Costars expression restored the normal developmental morphology in cosA cells both on bacterial lawns and non-nutrient agar (Fig. 1F). The observation identified cosA as the gene whose dysfunction was responsible for the defects of mutant cells. Prompted by the striking similarity between D. discoideum Costars and mCostars, we tested whether human mCostars complements the cosA mutant. The results showed that the act15 promoter-driven expression of mCostars in cosA cells indeed restored the normal developmental morphology, demonstrating the conservation of Costars-like proteins in function.

We also examined Costars levels during D. discoideum development using western analysis. The results showed that Costars expression was developmentally regulated, with peak levels appearing at early stages of development, corresponding to the time cells become highly chemotactic and aggregation-competent (Fig. 1G).

Reduced migration of cosA cells towards the chemoattractant source

To further characterize the phenotype of cosA cells, a semi-quantitative small-drop chemotaxis assay was employed to test the response of cosA cells to a range of chemoattractant concentrations. Wild-type cells showed directed movement towards the cAMP source, the concentration of which ranged from 10−9 to 10−5 M, with the greatest movement in response to 10−7 M cAMP. By contrast, although cosA cells did display a chemotaxis response, the efficiencies were below 20% of those of the wild-type cells (Fig. 2A). Expressing cosA or mCostars restored the wild-type-like cAMP chemotaxis response in cosA cells.

To more closely observe the cell behavior in chemotaxis, we subjected cells developed in suspension for 5 hours to a micropipette chemotaxis assay combined with time-lapse video microscopy. When exposed to the chemoattractant gradient generated from a micropipette releasing cAMP, wild-type cells showed coordinated movement towards the micropipette tip and many cells reached the micropipette tip within 20 minutes (Fig. 2B). However, under the same conditions, cosA cells moved much more sluggishly; few cells arrived at the micropipette tip within 20 minutes. This reduced migration towards the cAMP source was corrected by expressing Costars or human mCostars, indicating that loss of Costars function was responsible for the defect.

Fig. 2.

Decreased migration of cosA cells towards the cAMP source. (A) Small-drop chemotaxis assay. Cells developed for 5 hours were placed in small drops adjacent to cAMP drops of different concentrations. Drops showing asymmetrical distribution of cells within 30 minutes were scored positive. The chemotaxis response was represented by the percentage of positive drops. Results are from 4 independent experiments (mean ± s.d.). (B) Micropipette chemotaxis assay. At time 0, a femtotip filled with 100 μM cAMP was positioned to stimulate the cells that were developed for 5 hours. Microscopic images were captured at 10-second intervals. Photographs at 0 and 20 minutes are shown. Right panels: magnified cropped images from the photographs at 20 minutes to better reveal the morphology of migrating cells. Scale bars: 100 μm.

Developmental differentiation and chemoattractant-induced signaling in cosA cells

We investigated the mechanism underlying the cosA phenotype by first checking the developmental differentiation of cosA cells. During the early stage of starvation-induced development, D. discoideum cells differentiate and express developmentally regulated genes to acquire competence for sensing and relaying cAMP signals. We used RT-PCR to examine the induction of two representative genes, carA and acaA, which encode the cAMP receptor cAR1 and the aggregation-stage adenylyl cyclase ACA, respectively. Expression of these two genes is required for responding to and relaying the cAMP signal. The expression of a developmentally regulated cell adhesion molecule, contact site A protein (gp80), was examined by western blotting. Results showed similar temporal expression patterns of carA, acaA and gp80 in wild-type and cosA cells (Fig. 3A), indicating that cosA cells entered the developmental program and differentiated in a manner similar to that of the wild-type cells.

We next examined the chemoattractant-induced signaling events in cosA cells. To investigate the function of the GPCR signaling module, we used activation of one downstream effector, guanylyl cyclase, as a read-out. The accumulation of cGMP in response to cAMP stimulation was examined, and the responses were similar in wild-type, cosA and the rescued cells (Fig. 3B), suggesting that the G-protein receptor coupling remained intact in cosA cells. The data positioned Costars action downstream of G-protein activation.

We then investigated whether mutant cells can determine the direction of a concentration gradient. The best-studied indicator of directional sensing is the localized accumulation of PtdIns(3,4,5)P3 at the cell front; PH domain-containing proteins can be directed to PtdIns(3,4,5)P3-rich membrane locations (Huang et al., 2003; Iijima and Devreotes, 2002; Parent et al., 1998). We employed GFP–PHCRAC, a GFP fused to the PH domain from cytosolic regulator of adenylyl cyclase (CRAC) (Insall et al., 1994), as a PtdIns(3,4,5)P3 probe and examined wild-type and cosA cells that express GFP–PHCRAC for directional sensing. In the micropipette cAMP chemotaxis assay, both wild-type and cosA cells displayed localized membrane distribution of GFP–PHCRAC on the side facing the cAMP source (Fig. 3C). When the kinetics of GFP–PHCRAC translocation was checked in cells receiving a uniform cAMP stimulus, similar temporal patterns of membrane localization of GFP–PHCRAC were observed in wild-type and cosA cells, with the most intense membrane signals occurring at ~4 to 8 seconds (Fig. 3D). The data suggest that the PI3K pathway for directional sensing remains intact without Costars function.

Fig. 3.

Chemoattractant-induced events in cosA cells. (A) Analysis of developmental gene expression. Cells were developed in suspension and sampled at different time points. RT-PCR and western blotting were performed on RNA samples and total-cell lysates, respectively. (B) Chemoattractant-induced cGMP accumulation. Aggregative-stage cells were stimulated with 1 μM cAMP and samples were collected at indicated time points to determine the cGMP content. (C) GFP–PHCRAC localization in a cAMP gradient. Cells expressing GFP–PHCRAC were developed for 5 hours and subjected to the micropipette cAMP chemotaxis assay. Images were taken under a confocal fluorescence microscope. *, position of micropipette tip. (D) Translocation of GFP–PHCRAC to the membrane in response to uniform cAMP stimulation. Aggregation-competent cells were uniformly stimulated by 1 μM cAMP at time 0. Confocal fluorescence microscopic images were captured at the indicated time points.

Table 1.

Analysis of cell behavior during migration

Cell motility defect of cosA cells

To explore the migration defect of cosA cells during chemotaxis, we performed computer-assisted motion analysis on individual cells. Cells moving in spatial cAMP gradients were traced using recorded micrograph images (Fig. 2B), and their velocity and directionality were analyzed using imaging software. Comparing centroid tracks of migrating cells, we found that tracks of cosA cells were shorter than those of wild-type or rescued cells (Fig. 4A). Analysis of motility parameters showed a significantly reduced mean velocity in cosA cells; nevertheless, the directional persistence and chemotaxis indices were similar among these strains (Table 1). Therefore, cosA cells can probably sense the direction of a chemoattractant gradient, but cannot respond by moving as efficiently as the wild-type cells.

To determine whether cosA cells generally move more slowly than wild-type cells, randomly migrating vegetative-stage cells were recorded and the micrograph images were analyzed. Similar to the above results, cosA cells appeared to have shorter migration tracks than the wild-type or rescued cells (Fig. 4B). Computed data demonstrate a significant decrease in the mean velocity of cosA cells compared with that of wild-type or rescued cells, and expressing Costars or mCostars in cosA cells markedly increased their migration velocity (Table 1). These results suggest that cells lacking Costars function suffer a defect in the cellular machinery mediating cell motility rather than chemotactic orientation.

Fig. 4.

Reduced motility of cosA cells. (A) Directional migration in cAMP gradients. Microscopic images of developed cells during their migration in the micropipette chemotaxis assay were analyzed using Metamorph imaging software. Ten representative migration tracks of the cell centroid are shown for each strain. Red circle, position of micropipette tip. (B) Random migration in the absence of chemoattractant gradients. Microscopic images of randomly moving vegetative cells in HL5 medium were analyzed. Ten representative migration tracks of the cell centroid are shown, with the start point normalized to the origin in the plot; 20 pixels correspond to 28.17 μm.

Fig. 5.

Analyses of interaction between Costars and actin. (A) In vitro F-actin co-sedimentation assay. Non-muscle actin (Cytoskeleton) was mixed with recombinant GST or GST–Costars purified from E. coli, or commercially available α-actinin purified from rabbit skeletal muscle (Cytoskeleton). After incubation to allow actin polymerization, the mixtures were centrifuged to sediment the F-actin. Proteins in supernatant and pellet fractions were separated by SDS-PAGE and stained with Coomassie Blue. (B) Distribution of Costars in Triton-soluble (S) and insoluble (P) fractions. Total lysates (TL) were prepared from vegetative (0 hours) or aggregation-competent (5 hours) cells in a 1% Triton X-100-containing buffer. After centrifugation, Triton-soluble (S) and insoluble (P) fractions were resolved by 15% SDS-PAGE and subjected to western blotting using polyclonal anti-actin and anti-Costars antibodies. Samples were of equal cell equivalents (8.4×105 cells/lane). (C) Fluorescence cell staining. Vegetative (0 hours) or aggregation-competent (5 hours) Ax2 cells prepared by development in suspension were fixed and subjected to immunofluorescence staining for Costars and concomitant staining with TRITC-conjugated phalloidin (for F-actin) or Alexa-Fluor-594–dnase-I (for G-actin), respectively. Scale bars: 10 μm.

Analysis of the interaction between Costars and actin

Prompted by the similarity between Costars and the actin-binding region of STARS (Arai et al., 2002), we explored the physical interaction between Costars and actin. The glutathione transferase (GST)–Costars fusion protein expressed in Escherichia coli was purified and mixed with commercially available non-muscle actin to perform an in vitro actin co-sedimentation assay. In this assay, the F-actin-binding protein α-actinin co-fractionated with the F-actin pellet, while GST stayed in the supernatant after centrifugation (Fig. 5A). Although GST–Costars predominantly appeared in the supernatant in the absence of actin, a noticeable amount of GST–Costars appeared in the pellet in the presence of actin. The results suggest that Costars directly associates with F-actin in vitro, although probably with low stoichiometry or affinity. We examined the distribution of Costars in Triton X-100-soluble and -insoluble (i.e. cytoskeleton) fractions by western blotting and found that Costars was predominantly in the soluble fraction (Fig. 5B), indicating that Costars was not associated with the cytoskeleton under the experimental conditions used. To check for co-localization of Costars and actin in cells, immunofluorescence staining for Costars was performed with concurrent fluorescence staining for F-actin [using tetramethyl rhodamine isothiocyanate (TRITC)-conjugated phalloidin] or G-actin (using Alexa-Fluor-conjugated DNase I) on vegetative- and aggregative-stage cells. Under the confocal microscope, the Costars signal was observed predominantly around the periphery of vegetative cells, whereas in aggregative-stage cells the signal showed an asymmetrical distribution concentrated in lamellipodium-like structures (Fig. 5C). In the merged view, very limited co-localization was noted between Costars and G- or F-actin; even in the pseudopod of a representative aggregative-stage cell, where a patch of F-actin signal also appeared, there was little overlap of signals. Despite multiple attempts, we were not able to detect Costars–actin interactions in cell lysates by co-immunoprecipitation (data not shown). Collectively, our results suggest that, although Costars can directly bind to F-actin in vitro, it does not associate tightly with either G- or F-actin within the cellular context. However, given the interaction demonstrated in vitro, we cannot rule out the possibility that Costars interacts with actin only under specific cellular conditions or very transiently during dynamic actin changes.

Role of Costars in regulating actin organization and pseudopod formation

We next explored the role of Costars in actin regulation. Actin organization was examined using TRITC-phalloidin cell staining and fluorescence microscopy. When randomly moving vegetative cells were fixed and examined, most wild-type cells were polarized and showed a localized patch of F-actin at one end. By contrast, cosA cells were mostly non-polarized, staining for F-actin was stronger within the cells and signal was more uniformly distributed along the cell periphery (Fig. 6A). To examine the F-actin pattern in chemotaxing cells, cells that were developed under submerged conditions to aggregating streams were fixed in their streams and stained. Streaming wild-type cells were well elongated and presented one F-actin patch per cell at the leading edge. The ‘stream-like’ structures formed by cosA cells were short, with cells very loosely packed; individual cell in these structures were more amoeboid in shape, often showing multiple F-actin patches along the cell contour. The aberrant morphology and F-actin pattern of cosA cells could be corrected by expressing GFP–Costars or GFP–mCostars, suggesting the involvement of Costars in regulating actin organization.

Fig. 6.

Altered actin cytoskeleton and excessive pseudopod formation in cosA cells. (A) F-actin distribution. Cells at the vegetative stage (Veg) or developed under submerged conditions to the aggregation stage (streaming) were fixed, stained with TRITC-conjugated phalloidin and examined under a fluorescence microscope. Scale bars: 20 μm. (B) In vivo actin polymerization assay. Cells developed for 5 hours in suspension were stimulated by 100 μM cAMP at time 0. F-actin in each sample was labeled by TRITC-phalloidin and measured by fluorescence spectroscopy. Shown are the means ± s.d. of relative F-actin contents (normalized to measurements from corresponding unstimulated cells) from three independent experiments. (C) F-actin content analysis. Total lysates (TL) and corresponding Triton-insoluble fractions (TIF) were resolved on SDS-PAGE gels. Proteins were stained with Coomassie Blue and actin bands were quantified using densitometry. Gel images from representative experiments are shown, with F-actin ratios (actin in TIF:actin in equivalent TL) and fold change indicated. (D) Pseudopod analysis. Cells developed under submerged conditions were examined for pseudopod numbers. For each strain, 40–70 cells were counted in each of three independent experiments. Error bars are + s.d. (B–D) Wt, Ax2 cells.

We tested whether Costars modulates dynamic actin rearrangement by examining the in vivo actin polymerization response to uniform cAMP stimulation. Relative F-actin contents, measured by normalizing TRITC-phalloidin staining of samples taken at different time points to the staining level of unstimulated cells, were plotted against time after stimulation. A similar biphasic response was obtained in cosA cells as in wild-type and rescued cells (Fig. 6B), suggesting that Costars function is not required for cAMP-induced actin polymerization. However, we noticed that the fluorescence readings of unstimulated (time 0) samples were consistently higher in cosA cells than wild-type or rescued cells (data not shown), raising the possibility that loss of Costars results in increased cellular F-actin content.

To determine the percentage of cellular actin distributed in F-actin, we isolated the Triton X-100-insoluble fraction (TIF) as the representative pool of cellular F-actin (McRobbie and Newell, 1983). Total lysates and TIF samples were resolved by SDS-PAGE and the Coomassie Blue-stained actin bands in these samples were compared using densitometry. The concentration of actin monomers in TIF relative to that in the corresponding total lysate sample was determined as the ‘F-actin ratio’ using the densitometrical data.

The results revealed an increased F-actin ratio in cosA cells compared with wild-type or rescued cells (Fig. 6C, Table 2). The absolute concentration of total actin monomers, estimated by comparing stained bands of actin and bovine serum albumin (BSA) standards and expressed in relation to cell number (μg/106 cells), was similar in different strains (Table 2), excluding the possibility of misregulated actin synthesis in cosA cells. These data suggest that Costars serves a negative function in actin polymerization or has a positive regulatory role in F-actin depolymerization.

Actin polymerization is pivotal for pseudopod formation. Because TRITC-phalloidin cell staining revealed multiple F-actin patches in cosA cells, we were prompted to check the pseudopods in mutant cells. Cells in aggregating streams were prepared by submerged development and fixed for fluorescence staining – to help better define the cell contour – for both F- and G-actin. Cells were examined under a fluorescence microscope and grouped according to their pseudopod numbers. The results showed that, although most wild-type cells put out one single pseudopod towards their chemotactic direction, remarkably higher percentages of cosA cells exhibited two or more pseudopods (Fig. 6D). Expressing Costars or mCostars in cosA cells suppressed the formation of excessive pseudopods. The results suggest that Costars has a role in suppressing superfluous pseudopod formation during D. discoideum chemotaxis, perhaps through the regulation of dynamic actin rearrangement.

Table 2.

Quantitation of actin monomers in cells

Transient enrichment of Costars at the pseudopods during cell migration

To understand how Costars may regulate pseudopod formation and actin dynamics in chemotaxis, we examined the localization of Costars during chemotactic cell movement. When aggregation-competent cells that expressed GFP–Costars were subjected to the micropipette chemotaxis assay and fluorescence video microscopy, transient enrichment of GFP–Costars to the pseudopod extending towards the chemoattractant source was noted as cells were moving up the gradient (Fig. 7A). Compared with the signal of translocated GFP–PHCRAC (Fig. 3C), the GFP–Costars signal enriched in pseudopods appeared as a broader band, implying that Costars is not associating with the plasma membrane during cell movement. The kinetics of GFP–Costars translocation was examined in cells uniformly stimulated with cAMP. Similar to the membrane translocation of GFP–PHCRAC (Fig. 3D), the enrichment of GFP–Costars signal along the cell rim was transient, appearing by 4 seconds and disappearing ~12 seconds after cAMP addition (Fig. 7B). Although the mechanism underlying this observation is presently unclear, this dynamic distribution of Costars might have a role in coordinating actin cytoskeletal rearrangement during cell migration.


Although the REMI clone that lacked expression of Costars came through our forward genetics screening as a candidate for ‘chemotaxis’ mutant, our analyses have clarified that the defect is not related to directionality but, rather, to cell motility. The cellular machinery for motility is regulated by various upstream signaling mechanisms, including the GPCR-mediated signal transduction. In cosA cells, the intact cAMP-induced cGMP production and actin polymerization responses confirm the presence of a functional GPCR module and position the point of Costars action downstream of G-protein activation. Given the increased F-actin:total actin ratio with an unchanged amount of total actin, and the excessive pseudopod formation in cosA cells, we suggest that Costars serves as a regulator of actin dynamics and cell migration. Together with the observations that unstimulated cosA cells displayed stronger intracellular F-actin signals and chemotaxing cosA cells exhibited multiple pseudopods filled with F-actin, we speculate that Costars functions as a negative regulator for F-actin formation, either by impeding actin polymerization or promoting actin depolymerization.

Fig. 7.

Recruitment of Costars to the pseudopods upon chemoattractant stimulation. (A) Localization of GFP–Costars in cells facing a cAMP gradient. The cells that expressed GFP–Costars were developed in suspension for 5 hours and subjected to the micropipette chemotaxis assay. Microscopic images were captured every 4 seconds under a confocal fluorescence microscope. Shown are serial images of two representative cells. *, position of micropipette tip. Scale bars: 10 μm. (B) GFP–Costars localization in uniform cAMP stimulation. Aggregation-competent cells that expressed GFP–Costars were prepared by development in suspension and uniformly stimulated by 1 μM cAMP at time 0. Confocal fluorescence microscopic images captured at the indicated time points are shown. Scale bar: 10 μm.

How does Costars exert its function on actin dynamics? Given the resemblance of Costars to the actin-binding region of STARS, we originally hypothesized that Costars physically interacts with F- and/or G-actin and regulates the dynamic polymerization-depolymerization cycle. However, despite our multiple approaches, the interaction between Costars and actin remains elusive. On the one hand, Costars can indeed directly bind to F-actin, as evidenced by in vitro co-sedimentation. On the other hand, the amount of Costars co-sedimented with F-actin was too little to support a stoichiometric interaction. Furthermore, we have not been able to convincingly demonstrate any Costars–actin interaction in the cellular context. This is in striking contrast to the characteristics of STARS. A previous study has shown direct and stoichiometric association of STARS with F-actin (Arai et al., 2002). It should be noted that the C-terminal 142 aa of STARS are required for actin binding, whereas Costars has only 82 aa; perhaps the additional aa in STARS help to confer strong F-actin binding activity.

STARS functions in a RhoA-dependent manner, yet whether Costars regulates actin through a similar mechanism involving small GTPases remains to be tested. In the D. discoideum genome, multiple Rac-related GTPases exist, but not a typical Rho or Cdc42 (Vlahou and Rivero, 2006). Of the more characterized Rac-related GTPases, several are linked to chemotaxis-related functions. Rac1A controls filopodia formation and cell motility (Dumontier et al., 2000); RacB regulates actin polymerization and chemotaxis, and is controlled by the PI3K pathway (Park et al., 2004); RacC stimulates WASP-dependent F-actin assembly, and is required for PI3K activation and proper chemotaxis (Han et al., 2006); and RacG induces filopodia formation and actin polymerization, and regulates chemotaxis (Somesh et al., 2006). It is of interest to check whether Costars function depends on any of these D. discoideum Rac-related GTPases.

STARS and Costars may affect actin dynamics differently. STARS promotes F-actin formation in the presence of Rho activity, which depletes the G-actin pool and releases myocardin-related transcription factors (MRTFs) from the inhibitory influence of G-actin, allowing the nuclear import of MRTFs and stimulation of serum response factor activity (Kuwahara et al., 2005). In this scenario, loss of STARS function would be expected to cause decreased F-actin and increased G-actin. However, we observed higher levels of F-actin polymerization in cosA cells. Therefore, Costars appears to represent an actin regulatory protein distinct from STARS.

Pseudopod extension during cell migration requires intricate regulation of actin polymerization and/or depolymerization. Consistent with a function of Costars in regulating actin dynamics, cosA cells also show a pseudopod phenotype, displaying excessive pseudopods in chemoattractant gradients. Because extracellular signals may elicit signal transduction through multiple pathways to activate actin rearrangement, mutations affecting components of these pathways can result in abnormal pseudopod dynamics. Interestingly, like the cosA cells, several D. discoideum mutants – including the pten-null cells, the cytoplasmic cAMP phosphodiesterase regA null mutant and the adenylyl cyclase aca null mutant – also display excessive lateral pseudopod formation and low migration efficiencies in chemotaxis (Iijima and Devreotes, 2002; Stepanovic et al., 2005; Wessels et al., 2000). Whether Costars functionally interacts with these effector proteins in signaling pathways requires further investigation. Recently, a ‘pseudopod-centered’ model of eukaryotic migration and chemotaxis has been proposed (Insall, 2010). This model emphasizes the constant splitting of old pseudopods to form new ones, regardless of the extracellular signals, and hypothesizes that chemotactic signals act by altering the rate of pseudopod growth or biasing the position at which new pseudopods are generated. In support of this model, a quantitative analysis of pseudopod generation in migrating cells has demonstrated that moving cells often produce new pseudopods from bifurcation of existing ones rather than forming lateral pseudopods (Andrew and Insall, 2007). It remains to be determined whether the deranged actin abnormal in cosA cells affects the rate or efficiency of pseudopod bifurcation and, therefore, hinders efficient cell migration.

Given the transient enrichment of Costars to the anterior pseudopod in response to cAMP stimulation, we suspect that Costars is recruited to sites where actin polymerization actively happens, which helps to quench the polymerization reaction and to maintain an optimal cellular F-actin ratio for proper motility. Because we could not demonstrate close interaction between F-actin and Costars within cells, it is conceivable that Costars achieves its actin-regulating function by cooperating with other cellular proteins. Finding Costars-interacting proteins should help further elucidate the mechanism of action of Costars.

Materials and Methods

D. discodeum growth, development and transformation

D. discoideum cells were grown at 22°C in HL5, or on SM nutrient agar plates with Klebsiella aerogenes (Sussman, 1987). Typically, development was done by plating cells at 1.5×106 cells/cm2 onto non-nutrient agar, i.e. 1.5% agar in Development Buffer (DB; 5 mM Na2HPO4, 5 mM KH2PO4, 2 mM MgSO4, 0.2 mM CaCl2), or cells were grown in suspension at 2×107 cells/ml in DB with 90 nM cAMP added every 6 minutes (Devreotes et al., 1987). A submerged development method (Bear et al., 1998) was used to prepare cells in aggregating streams (which typically appear after 12 hours). Transformation was done in E buffer (10 mM K2HPO4/KH2PO4, 50 mM sucrose, pH 6.1) by electroporation using a Bio-Rad Gene Pulser II set at 900 V, 3 μF. After overnight recovery in HL5, cells were selected in 5 μg/ml blasticidin S (Cayla, Toulouse, France) or 10 μg/ml G418 (Sigma-Aldrich) for transformants.

Plasmids, antibodies and reagents

The integrating plasmid pYW3 for REMI was constructed by ligating the XbaI-HindIII bsr-expression cassette from pPTGal-BsrΔBgl (a gift from W.-T. Chang, National Cheng-Kung University, Taiwan) into the XbaI- and HindIII-digested pUC18. The cosA-expressing plasmid pTX-GFP-Costars was constructed by amplifying a full-length cosA (DDB0202716) DNA fragment from Ax2 genomic DNA using the primers T21#14-S1 (5′-GAGAGCTCATGGACGTTGATCACGAAGTTAAG-3′) and T21#14-A1 (5′-TTATTCGTCTTTAAGTAAAATGACATC-3′), subcloning the PCR product first into the yTA vector (YEASTERN), and moving the SacI-XhoI cosA fragment into SacI- and SalI-digested pTX-GFP (Levi et al., 2000). The mCostars-expressing plasmid pTX-GFP-mCostars was constructed by amplifying a full-length mCostars cDNA fragment from human kidney cDNA (Clontech) using primers HSPC280-S1 (5′-ACGAGAGCTCATGAATGTGGATCACGAGGTTAAC-3′) and HSPC280-A1 (5′-AAGGCTCGAGTTAATCTTGCAGTAATATAATGTCAAC-3′), subcloning the 246-bp product into yTA, and moving the SacI-XhoI mCostars fragment into SacI- and XhoI-digested pTX-GFP. For expressing a glutathione transferase (GST)–Costars fusion protein in E. coli, a ~250-bp cosA fragment amplified using primers T21#14-S2 (5′-CCACGAATTCCATGGACGTTGATCACGAAGT-3′) and T21#14-A2 (5′-CCGCTCGAGTTATTCGTCTTTAAGTAAAATGACATC-3′) was digested with EcoRI and XhoI, and ligated into EcoRI- and XhoI-digested pGEX-5X-3 (Amersham), resulting in pGEX-Costars. The polyclonal antibody against Costars was generated as follows: E. coli BL21(DE3) (Novagen) cells transformed with pGEX-Costars were induced to express GST–Costars by the addition of 1 mM isopropyl-1-thio-β-d-galactopyranoside, and lysates were prepared by sonication. GST–Costars was then purified using glutathione sepharose beads (Amersham) following GST pull-down procedures as suggested by the manufacturer. Purified GST–Costars was mixed with TiterMax Gold adjuvant (CytRx) and used to immunize female New Zealand white rabbits. Commercial antibodies used were: goat polyclonal anti-actin (I-19) (Santa Cruz Biotechnology); monoclonal anti-α-tubulin and horseradish peroxidase (HRP)-conjugated rabbit anti-goat IgG (Sigma-Aldrich); HRP-conjugated donkey anti-rabbit IgG and HRP-conjugated sheep anti-mouse IgG (GE Healthcare); and fluorescein isothiocyanate (FITC)-conjugated donkey anti-rabbit IgG (Jackson ImmunoResearch). The mouse monoclonal anti-gp80 antibody was a kind gift from C.-H. Siu (University of Toronto, Toronto, Canada). TRITC-phalloidin and Alexa-Fluor-594–DNase-I were purchased from Sigma-Aldrich and Molecular Probes, respectively.

REMI and Transwell enrichment of chemotaxis mutants

For REMI transformation, electroporation was performed on 0.8 ml samples of Ax2 cells (4×107 cells/ml in E buffer) mixed with 25 μg BamH I-linearized pYW3 DNA and 4 U DpnII (Adachi et al., 1994). After recovering overnight, cells were transferred onto 96-well plates and selected in HL5 containing 5 μg/ml blasticidin S. Transformants from independent transformation reactions were kept as separate pools. To enrich for chemotaxis mutants, pools of REMI transformants were developed for 5 hours and resuspended to 106 cells/ml; 105 cells were loaded into the top chamber of the Transwell device (5 μm pore size, Corning Life Sciences) set in a 24-well plate with 500 μl of 10−5 M cAMP in the lower chamber. After incubation for 3 hours at 22°C, cells that remained in the top chamber were collected by first wiping away cells on the lower surface of the Transwell membrane with cotton swabs, then cutting out the membrane with a surgical blade and vortexing the membrane in 10 ml of HL5 containing 5 μg/ml blasticidin S. The mixture was subsequently distributed onto a 96-well plate; single colonies appeared in ~7–10 days.

Chemotaxis assays

The agar-cutting assay was performed as previously described (Kuwayama et al., 1993), with modifications. Clones were inoculated as well-separated spots on SM agar with a lawn of K. aerogenes and grown until plaques 2–3 cm in diameter were formed. Two wedge-shaped agar blocks were excised from the edge of each plaque and placed upside-down on the surface of DB agar plates containing no or 10−6 M cAMP, respectively. The dispersion of cells was observed after 1 hour of incubation. Only clones showing very poor dispersion were collected for the subsequent trough assay. The trough assay was derived from a previously described under-agarose chemotaxis assay (Laevsky and Knecht, 2001). To prepare the assay plate, a 10-ml aliquot of 0.6% agarose (ultraPURE LMP, Invitrogen) solution in DB was allowed to harden in a 9 cm plastic Petri dish for 1 hour, then – by using a razor blade set and light suction from a Pasteur pipette – three 2 mm × 39 mm troughs spaced 5 mm apart were produced in the agarose bed. A 40 μl aliquot of 10−5 M cAMP was added to the left trough and allowed to stand for 30 minutes to develop a spatial gradient; 40 μl of buffer was added to the right trough as a control. A 60 μl aliquot of 1×107 cells/ml suspension of cells developed for 5 hours was subsequently loaded into the middle trough and the plate was further incubated for 40–60 minutes. Chemotaxis response was determined by observing if there were cells moving away from the trough edge in the direction of the cAMP source trough. The small-drop assay was performed as previously described (Konijn and Van Haastert, 1987). In the micropipette chemotaxis assay, a 50 μl aliquot of 4×105 cells/ml suspension of cells developed for 5 hours was spotted on a 3.5 cm culture dish and allowed to stand for 10 minutes before the dish was filled with 5 ml of phosphate buffer (pH 6.5). A femtotip (Eppendorf AG) filled with 100 μM cAMP was positioned in the center of the field, releasing cAMP with a pressure of 40 hPa. Cells were observed under an inverted microscope (DMIRBE, Leica) with a 10× objective (N PLAN, NA 0.25) and a 10× optivar setting. Images were captured every 10 seconds for 20 minutes using a CCD camera (CoolSNAP, Photometrics) and Metamorph imaging software (Molecular Devices).

Identification and re-knockout of the cosA gene

The cosA gene was identified by cloning a ~6-kb sequence of EcoRV genomic DNA flanking the REMI insertion in the T21#14 mutant following procedures as previously described (Kuspa and Loomis, 1992b). Sequencing analysis of the recovered plasmid (pT21#14) revealed the cosA 249-bp ORF (DDB0202716) 50 bp downstream of the REMI insertion site. The cosA re-knockout strain was generated by transforming EcoRV-linearized pT21#14 into Ax2 cells and selecting for blasticidin-S-resistant clones; the knockout genotype was checked by PCR and Southern blotting.

RT-PCR analysis

Total RNA was isolated from cells at different developmental stages using Trizol (Gibco BRL) according to the manufacturer's protocol and used as template in reverse transcription (RT)-PCR amplifications. Gene-specific primers used were: T21#14-S1 (5′-GAGAGCTCATGGACGTTGATCACGAAGTTAAG-3′) and T21#14-A1 (5′-TTATTCGTCTTTAAGTAAAATGACATC-3′) for cosA; ACA-senA (5′-TTGGCAATGAGAAAAGCATGGG-3′) and ACA-antiA (5′-GCATTCTAGAGGCGGTATTGG-3′) for acaA; carA1-senA (5′-GGATTGGTGTAAGTTTCACTGG-3′) and carA1-antiA (5′-CCATATCGGAACTACATTGCAC-3′) for carA; and actin15-1 (5′-ATGGTTGGTATGGGTCAAAAG-3′) and actin15-2 (5′-GAAACATTTTCTGTGAACAATTG-3′) for actin.

PHCRAC domain and Costars localization in living D. discoideum cells

Cells expressing GFP–PHCRAC or GFP–Costars were developed for 5 hours and subjected to the micropipette chemotaxis assay or uniform cAMP stimulation. For uniform stimulation, a 50-μl aliquot of cell suspension (2×105 cells/ml) was spotted on a 3.5-cm FluoroDish (World Precision Instruments), incubated for 10 minutes and subsequently stimulated by adding 3 ml of 1 μM cAMP at time 0. Cells were observed using an inverted confocal laser-scanning microscope (TCS SP5, Leica) with a 100× HCX PL APO objective (NA 1.4). Images were captured every 4 seconds using a CCD camera (CoolSNAP, Photometrics).

Migration analysis

Chemotactic migration of cells developed for 5 hours in the micropipette chemotaxis assay and random migration of vegetative cells (in HL5) placed on 35 mm dishes were observed under an inverted microscope (DMIRBE, Leica) with a 10× objective (N PLAN, NA 0.25) and a 10× optivar setting. Images were captured every 15 seconds for 20 minutes by a CCD camera (CoolSNAP, Photometrics). Three independent experiments were performed for each analysis. Metamorph imaging software (Molecular Devices) was used to trace individual cells and take measurements for calculating motility parameters [i.e. mean velocity, directional persistence and chemotaxis index (CI)]. A total of 30 cells were traced for each strain. Mean velocity is the displacement of the cell centroid along the total path per unit time. Directional persistence is the ratio of the net path taken by the cell over the entire video divided by the total path taken measured in intervals. CI is the cosine value of the angle between the line connecting the start- and end-point of a moving cell and the line connecting the start-point of the cell and the chemoattractant source (Futrelle et al., 1982); a CI value of 1 indicates that the cell moves directly towards, while a value of −1 indicates that the cell moves directly away, from the chemoattractant source.

Fluorescence staining of fixed D. discoideum cells

For simultaneous detection of Costars and F- or G-actin, cells were fixed in 0.5% paraformaldehyde/PBS, washed in PBS (pH 7.4) and permeabilized in 0.2% Triton/PBS. After three washes in PBS and blocking in 10% FBS/PBS, cells were incubated with a polyclonal anti-Costars antiserum and, subsequently, with FITC-conjugated secondary antibodies diluted in PBS containing either 80 nM TRITC-phalloidin or 150 nM Alexa-Fluor-594–DNase-I. After final washing, stained cells were observed under a confocal microscope (TCS SP5, Leica) with a 100× objective (HCX PL APO, NA 1.4) and images were captured with a CCD camera (CoolSNAP, Photometrics). For observation of F-actin distribution, fixed cells stained with TRITC-phalloidin were observed under a microscope (Axioskop 2, Carl Zeiss) with a Plan-NEOFLUAR 40× objective (NA 0.75) and images were captured with a CCD camera (AxioCam MRm, Carl Zeiss).

Actin assays

The in vitro F-actin co-sedimentation assay was performed following the manufacturer's instructions (BK013, Cytoskeleton) using GST–Costars, GST–mCostars and GST expressed and purified from E. coli as the test proteins. After centrifugation, equal volumes of the pellet and supernatant fractions were analyzed by SDS-PAGE and stained with Coomassie Blue. The in vivo actin polymerization assay was done essentially as previously described (Insall et al., 1996). To determine the F-actin:total actin ratio, total lysates (TL) prepared in 1% Triton X-100-containing buffer were fractionated by low-speed centrifugation as previously described (McRobbie and Newell, 1983). TL and the TIF (regarded as the F-actin fraction) were resolved on a SDS-PAGE gel. Coomassie Blue-stained gels were digitized and actin bands were quantified by densitometry using ImageQuant software (GE Healthcare, UK). To estimate the absolute concentration of actin monomers in TIF or total actin, samples were resolved on the same gel together with a series of BSA standards; after quantification of Coomassie Blue-stained actin and BSA bands, the concentration of actin in each sample was calculated using the standard curve constructed by the BSA data.

Miscellaneous methods

Cyclic AMP-induced cGMP accumulation response was assayed as previously described (Mato et al., 1977) and the cGMP content in samples was measured using an isotope dilution assay kit (TRK 500, Amersham). Preparation of the TIF (cytoskeletal) and the soluble fraction from cell lysates in a 1% Triton X-100-containing buffer was performed as previously described (McRobbie and Newell, 1983). For pseudopod analysis, cells were developed under submerged conditions (Bear et al., 1998). When aggregating streams formed in the wild-type strain, cells of all strains were fixed and stained for G-actin by Alexa-Fluor-488–DNase-I and for F-actin by TRITC-phalloidin; the staining helped to define the cell contours and the presence of pseudopods. Images were captured using fluorescence microscopy (Axioskop 2, Carl Zeiss) and individual cells were examined for pseudopod numbers.


We thank Peter N. Devreotes at Johns Hopkins University, Baltimore, MD, for providing plasmids. This work was supported by grants NHRI-EX92-9230SI, NHRI-EX93-9230SI and NHRI-EX94-9230SI from the National Health Research Institutes, Taiwan; grants NSC97-2320-B-010-021 and NSC 98-2320-B-010-023-MY3 from the National Science Council; and a grant (Aim for the Top University Plan) from the Ministry of Education, Taiwan.


  • Accepted June 21, 2010.


View Abstract