ForC, a novel type of formin family protein lacking an FH1 domain, is involved in multicellular development in Dictyostelium discoideum

Proper spatial and temporal regulation of cytoskeletal function is essential for such eukaryotic cell activities as mitosis, endocytosis,exocytosis, cell migration and morphogenesis. To better understand the molecular basis for cell motion and the underlying regulation of the cytoskeletal system, we are using the soil amoeba Dictyostelium discoideum as a model system.

Dictyostelium discoideum has a relatively simple cytoskeleton;nevertheless, many of its movements appear similar to those observed in higher eukaryotes. In rich medium, they proliferate as a unicellular organism and carry out cytokinesis that looks morphologically very similar to that of vertebrate cells in culture. When starved, the cells aggregate to form multicellular structures called fruiting bodies, which consist of spores and stalks that hold sori above the substrate. During this process, the cells first migrate to an aggregation center in a fashion similar to leucocytes. The resultant aggregates behave as a multicellular entity and undergo programmed cell differentiation and morphogenesis to yield a fruiting body. In this way, Dictyostelium provides a model system with which to investigate how individual cells behave within a multicellular system and how multicellular morphogenesis is regulated. In addition, Dictyostelium is highly amenable to genetic manipulation, including gene disruption and introduction of exogenous genes. And since its genome is haploid, it is possible to see an effect of a mutation even when it is recessive.

Formin family proteins are thought to play crucial roles in the regulation of cytoskeletal function (Tanaka,2000; Wasserman,1998). They are found in a wide variety of eukaryotic cells, from unicellular organisms and fungi to higher plant and animal cells. Many of the formin proteins were isolated genetically on the basis of mutations that affect cytoskeletal function. For example, budding yeast Bni1(Kohno et al., 1996) and Bnr1(Imamura et al., 1997),fission yeast Cdc12 (Imamura et al.,1997), Asperugius nidanas SepA(Harris et al., 1997),nematode Cyk-1 (Swan et al.,1998), and fruit fly diaphanous(Castrillon and Wasserman,1994) and cappuccino(Emmons et al., 1995) were all discovered through mutations that affected cytokinesis. Of these, Bni1(Jansen et al., 1996;Zahner et al., 1996), Bnr1 and cappuccino are also known to be involved in the establishment of cell polarity. In the fission yeast, however, establishment of cell polarity is mediated by another formin protein, For3(Feierbach and Chang, 2001). In addition, mutation of mouse formin, the first formin isoform identified,results in limb deformity and renal agenesis(Jackson-Grusby et al., 1992;Woychik et al., 1990);mutation of DFNA1(hDia1), a human homologue of diaphanous, results in nonsyndromic deafness caused by a defect in actin organization in the hair cells of the inner ear(Lynch et al., 1997), and a mutation in DIA(hDia2), another human homologue of diaphanous, results in premature ovarian failure(Bione et al., 1998).

Formin proteins are characterized by the presence of three FH (formin homology) domains (FH1, FH2 and FH3)(Tanaka, 2000;Wasserman, 1998). The FH1 domain consists of multiple poly-proline stretches and is located at the middle of the protein. Many formin proteins are known to interact with profilin, an actin-monomer-binding protein, via the FH1 domain(Evangelista et al., 1997;Holt and Koffer, 2001;Imamura et al., 1997;Wasserman, 1998;Watanabe et al., 1997). In addition, some formin proteins interact with the Src homology 3 (SH3) domain or WW domain through the FH1 domain (Holt and Koffer, 2001). The FH2 domain is a highly conserved region that spans about 130 amino acid residues, and is located near the C-terminus(Tanaka, 2000;Wasserman, 1998). Recent truncation analysis of Bni1 indicated that the FH2 domain alone is able to nucleate polymerization of actin filaments in vitro(Pruyne et al., 2002). The FH3 domain is less well conserved than the other two FH domains, is located near the N-terminus and is thought to be important for determining intracellular localization of formin family proteins(Kato et al., 2001;Petersen et al., 1998).

These biochemical properties of the FH1 and FH2 domains, as well as the phenotypes related to formin mutations, implicate formin proteins in the regulation of the actin cytoskeleton. Consistent with this view, a variety of mutations affecting one or more formin proteins, or their overproduction, all result in actin cytoskeletal disorganization(Castrillon and Wasserman,1994; Chang et al.,1997; Evangelista et al.,1997; Swan et al.,1998; Watanabe et al.,1997; Watanabe et al.,1999). In addition, a growing number of studies, including analyses of phenotype and protein localization, suggest that formin proteins are also involved in regulating microtubule function(Giansanti et al., 1998;Lee et al., 1999;Miller et al., 1999;Palazzo et al., 2001).

Several formin proteins have been shown to bind Rho-type small GTPases. This places formin proteins at a critical position, where they can receive signals from Rho and organize the actin and/or microtubule cytoskeleton in response to that signal. This prompted us to examine the functions of formin proteins using Dictyostelium discoideum as a genetic model with which to study cell motility. Our aim was to establish a general model of cytoskeletal regulation in eukaryotic cells.

Standard methods were used for DNA manipulation(Sambrook et al., 1989). The sequences of the entire coding regions of forA, forB and forC were determined mainly by inverse PCR using genomic DNA of wild-type Dictyostelium Ax2 cells. For each PCR, the sequences of several clones were determined, and their consensus was taken as the sequence of each gene.

Entire genomic DNA of forC was obtained by PCR and cloned into the pGEM-T cloning vector (Promega). The 2.4 kb SalI-EcoRV fragment of the forC ORF was then replaced with the Blasticidin resistance gene cassette (Adachi et al.,1994). The resultant disruption construct was digested with SpeI and NcoI, and used to transform Ax2 cells. Successful disruption was determined with PCR using primers 5′-ATGAAAATTAGAGTTGAATTAATAAATGG-3′, and 5′-GCTCGTTTTACCATATCATTTG-3′.

Wild-type Dictyostelium (strain Ax2) and ΔforCcells were cultured in HL5 medium(Sussman, 1987) supplemented with 60 μg/ml each of penicillin and streptomycin (+PS) at 20°C. Blasticidin selection was performed by adding 10 μg/ml Blasticidin to HL5+PS. Transformants with pBIG-based plasmids were maintained in HL5+PS supplemented with 15 μg/ml G418. For suspension cultures, cells were shaken in conical flasks at ∼140 rpm. Dictyostelium development was carried out either on MES agar plates(Peterson et al., 1995) or on Klebsiella aerogenes on SM/5 agar plates(Sussman, 1987).

Ax2 cells were allowed to develop on MES agar plates, during which cells were collected from each 100 mm plate every 4 hours. RNA was extracted from the cells using TriZol reagent (Gibco Invitrogen), and was used for synthesis of first strand cDNA using reverse transcriptase (ReverTra Ace; Toyobo) with Oligo dT primer(5′-CCAGTGAGCAGAGTGACGAGGACTCGAGCTCAAGCTTTTTTTTTTTTTTTTT-3′),after which 1% of the first strand cDNA was used for standard PCR using primers specific for both sides of the intron of forC(5′-ACAACAATCTCAACAAACTCC-3′ and 5′-ACAAGCCAACAGTACGGTATC-3′). The PCR products were subjected to agarose gel electrophoresis.

Genomic DNA encoding ForC was amplified by PCR using a pair of oligonucleotides (5′-GGATCCAATGAAAATTAGAGTTGAATTAATAAATGG-3′ and 5′-GAGCTCTTAAAATGCTCGTTTTACCATATC-3′) that add BamHI and SacI sites at either end of the PCR product, enabling it to be subcloned into pBIG (Ruppel et al.,1994) or pBIG-GFP (Nagasaki et al., 2001). Subsequent expression of ForC or GFP-ForC was driven by the actin 15 promoter.

Development of Dictyostelium was observed with a dissection microscope (SZX 12; Olympus, Tokyo, Japan). A fluorescence microscope (IX50;Olympus) equipped with a 100× oil immersion objective lens(Plan-NEOFLUOAR; Carl Zeiss, Thornwood, NY) and the appropriate sets of filters for GFP or rhodamine was used to observe cells expressing GFP fusion proteins. Images were obtained using a cooled CCD camera (C5985; Hamamatsu Photonics, Hamamatsu, Japan) coupled to an image analysis system (ARGAS-20,Hamamatsu Photonics) and recorded using NIH Image (National Institutes of Health, Bethesda, MD). A microscope (IX70; Olympus) equipped with a 60×oil immersion objective lens (U-planApo; Olympus) connected to a real-time confocal system (CSU10; Yokogawa, Tokyo, Japan) equipped with argon-krypton laser was employed for confocal microscopy. Images were obtained using a chilled CCD camera (Orca; Hamamatsu Photonics) and analyzed using IP lab(Scanalytics, Fairfax, VA).

For fluorescence microscopic observation, cells were transferred to a plastic Petri dish with a glass coverslip at the bottom and allowed to adhere to the bottom for about 30 minutes. Live cells were observed in MES buffer (20 mM MES, pH 6.8, 0.2 mM CaCl₂, 2 mM MgSO₄). Thereafter,the cells were fixed by incubation in fix solution (3.7% formaldehyde, 20 mM MES pH 6.8, 2 mM MgSO4, 1 mM EGTA) for 4 minutes at 20°C. Observation was then carried out in 16.7 mM K-phosphate buffer. F-actin was stained by incubating fixed cells in buffer containing rhodamine —phalloidin for 10 minutes, after which they were washed with K-phosphate buffer and observed. Micrographs were pseudocolored by Adobe Photoshop 5.5 (Adobe Systems Inc.).

In order to identify genes that encode formin family proteins in Dictyostelium discoideum, we performed a Blast search against the database of the Japanese Dictyostelium cDNA project using the S. pombe Cdc12 amino acid sequence as a query. We found that two different cDNAs, FCL-AB11 and SLB408, could potentially code for formin proteins, and cloned the entire coding regions of the two genes using colony hybridization,inverse PCR and 5′ and 3′ RACE. From their predicted amino acid sequences, we determined that both genes encode typical formin proteins and named the genes forA and forB, respectively. In order to compare their amino acid sequences with other known formin family proteins, we performed multiple sequence alignment using clustalW 1.8 and determined their FH2 and FH3 domains. A domain situated between FH2 and FH3 and containing multiple poly-proline stretches was designated as FH1. forA encodes a polypeptide of 1219 amino acids; its FH1, FH2 and FH3 domains are located between amino acid residues 650-765, 904-1039 and 245-461, respectively. forB encodes a polypeptide of 1128 amino acid residues; its FH1, FH2 and FH3 domains are between residues 532-612, 766-916 and 120-229,respectively.

By using DNA constructs to knock out each gene, we generated disruption mutants (ΔforA and ΔforB) by homologous recombination, but neither ΔforA nor ΔforBshowed any mutation-related phenotype (data not shown). Even a double-knockout mutant lacking both forA and forB showed no detectable phenotype, at least in our assays that include growth on substrate and in suspension, and development of fruiting bodies (data not shown). This observation led us to speculate that Dictyostelium might express other formin proteins, and we performed another Blast search. This time, in addition to the data from the cDNA project, we included data from the Dictyostelium genomic DNA sequencing project. With this search, we found there to be at least nine genes that could potentially encode formin proteins (Fig. 1A).

Among the various formin genes within the genome of Dictyostelium discoideum, we focused our attention on one that we named forCbecause it apparently lacks an FH1 domain, though it clearly has FH2 and FH3 domains (Fig. 1B). The gene encoding ForC was discovered as a partial sequence in the Japanese cDNA library (clone SSC675). We cloned the entire coding region by inverse PCR, and found the resultant predicted amino acid sequence to consist of 1158 amino acids. Multiple amino acid alignment with other formin proteins revealed that ForC has an FH2 domain between amino acid residues 756 and 893 (black boxes inFig. 1B and C) and an FH3 domain between residues 117 and 312. The results of a Blast search indicated that the FH2 domain of ForC is most similar to that of fruit fly cappuccino, with 33% amino acid identity, and the FH3 domain is most similar to the human FHOS FH3 domain, with 27% identity. Consistent with an earlier report by Peterson et al.(Peterson et al., 1995), the FH3 domain of ForC contains three highly conserved regions(Fig. 1D, solid underlines),although we noticed that the third is 11 amino acid residues longer on the N-terminal side than was proposed by those investigators(Fig. 1D, dashed underline).

All formin proteins discovered so far have an FH 1 domain located between the FH2 and FH3 domains. FH1 is a highly proline-rich domain containing several poly-proline stretches, each of which contains up to 13 continuous prolines (Bione et al., 1998;Emmons et al., 1995). ForC, by contrast, has no poly-proline stretches, either between or outside the FH3 and FH2 domains (Fig. 1B), and thus lacks an apparent FH1 domain.

The FH1 domains of formin proteins are known to bind various proteins. In particular, many formin isoforms bind the actin monomer binding protein,profilin, via their FH1 poly-proline domains(Holt and Koffer, 2001). Likewise, profilin is known to bind poly-proline domains in the Ena/VASP, ERM and WASP families of proteins. So far, all known profilin-binding sequences contain a common motif, XPPPPP, where X=G, L, I, S or A(Holt and Koffer, 2001). ForC,however, does not possess this sequence. The only amino acid sequences with continuous prolines in ForC are HPP and TPP. In neither case is the proline stretch long enough to match the consensus sequence for profilin binding;moreover, the residues before these proline pairs do not match the known profilin binding motif. That this region of ForC is in fact not a profilin-binding site was then confirmed using yeast two-hybrid assays. Dictyostelium has two genes that encode profilin, pfyA and pfyB (Haugwitz et al.,1994). As predicted, we detected no interactions between ForC and either PfyA or PfyB. By contrast, in a control experiment, we demonstrated interaction of ForB, which has typical profilin-binding motifs, with both PfyA and PfyB (data not shown).

In order to better understand the in vivo function of ForC, we made a forC knockout mutant in which approximately 70% of the forCORF was replaced with a Blasticidin S resistance gene cassette(Fig. 2A). Wild-type Ax2 cells were transformed with the linearized DNA fragment, and individual Blasticidin S-resistant colonies were analyzed for disruption of forC using genomic PCR (Fig. 2B). We obtained six independent clones that lacked the forC gene. These cells were viable and grew normally in the HL5 medium both on substrates and in suspension culture (data not shown), suggesting that ForC is not essential for cytokinesis. Furthermore, detailed observation of cytokinesis ofΔ forC cells on substrate failed to detect any morphological and temporal abnormalities (data not shown). ΔforC cells grew at normal rates on lawns of food bacteria Klebsiella aerogenes as well(data not shown). That the growth rates of ΔforC cells were not impaired either in nutrient media or on lawns of bacteria suggests that ForC does not play essential roles in macropinocytosis or phagocytosis.

In contrast, when the cells were placed on bacterial lawns and allowed to go through their developmental program, they all formed aberrant fruiting bodies (Fig. 3A, right panel). The cells were rescued from this developmental defect by expression of exogenous forC driven by the constitutively active actin 15 promoter(Fig. 3D, middle), which confirmed that the developmental defect in these clones was caused by the absence of forC.

We then allowed the wild-type and mutant cells to develop on MES agar plates and observed their development more closely. When Dictyostelium cells are starved, they first migrate up a cAMP gradient towards an aggregation center, after which further development transforms the aggregates into tipped mounds. ΔforC cells migrated normally towards chemotactic centers(Fig. 3B, 10 hours), suggesting that the individual mutant cells can move in a directional fashion. The aggregated mutant cells formed mounds (Fig. 3B, 10 hours) and subsequently formed tipped mounds. The difference between the wild-type and the mutant strains became apparent only after this tipped mound stage: wild-type cells started culmination, but the mutant cells did not (Fig. 3B,20 hours). The morphological changes in the mutant strain gave one the impression that it could not generate enough `force' to raise tall stalks and then lift the sori along the stalks. The mutant strain was able to make stubby stalk-like structures, but they were much shorter and thicker than those in the wild-type cells. Moreover, the sori were not lifted and remained at the base of the stalk-like structures (Fig. 3B, 42 hours). These stalk-like structures were stained with calcofluor (data not shown).

To determine whether the morphologically aberrant ΔforCfruiting bodies contained viable spores, we treated them with 0.6% Triton-X for 15 minutes, which has been shown to selectively lyse unsporulated or undifferentiated cells (Ennis et al.,2000). When wild-type and ΔforC fruiting bodies were treated with Triton-X, washed, resuspended in HL5 growth medium and observed the following day, we found that both stains produced detergent-resistant spores (data not shown). As a negative control, myosin II-null cells, which also cease development at the tipped aggregate stage, did not yield any viable spores (data not shown). Formation of viable spores and calcofluor-positive stalk-like structures by ΔforC suggests that the cellular differentiation and maturation of spore and stalk cells proceeds normally even though the morphological changes do not.

When Dictyostelium cells are starved on unbuffered agar plates,they form slugs following aggregation that migrate towards a light source(Sussman, 1987). When we placed wild-type and ΔforC mutant cells under slug-forming conditions, wild-type cells aggregated and formed tipped mounds and then slugs that migrated around until they eventually formed fruiting bodies.Δ forC cells also aggregated normally on unbuffered plates, but they remained as tipped mounds and did not form slugs(Fig. 3C).

Taken together, these results demonstrate that defects present inΔ forC cells make them unable to proceed through the proper morphological changes after the tipped mound stage, either towards culmination or slug formation.

In order to investigate the pattern of forC expression during development, we collected whole RNA from cells cultured on MES agar plates every 4 hours and performed RT-PCR using primers designed to amplify a fragment of the forC ORF. We found a low level of forCexpression during vegetative growth, and the level remained low until the aggregation stage. Expression of forC then significantly increased following mound formation and remained high through culmination, after which it declined during the final stage of fruiting body formation(Fig. 4). The period of high forC expression is consistent with the general sequence of events during which the defects caused by the ΔforC mutation became apparent, and strongly supports our conclusion that forC plays a key role during these multicellular stages.

Many mutations related to cytoskeletal components are known to affect the developmental morphogenesis of Dictyostelium(Noegel and Schleicher, 2000). In some mutants, proper function can be restored through synergetic effects elicited by mixing the defective mutants with wild-type cells(Tsujioka et al., 1999;Witke et al., 1992). To test whether adding wild-type cells would rescue the developmental function ofΔ forC cells, we allowed ΔforC cells to develop on MES agar plates after mixing them with wild-type cells at various ratios(Fig. 5). WhenΔ forC and wild-type cells were mixed at a ratio of 1:4, the overall shape of the fruiting bodies was normal, but unlike cultures of pure wild-type cells, there were small cell masses at the bottoms of the stalks(Fig. 5b). When the two strains were mixed at a 2:3 ratio, the stalks appeared normal, and the sori were of normal size, but the majority of the sori were not lifted all the way to the top of the stalks; they remained about halfway up the stalk(Fig. 5c), and beneath them were usually additional cell masses. When mixed at a 3:2 ratio, the overall shape was similar to that seen with the 2:3 ratio, but larger masses of cells remained at the bottom of the stalks, and the shape of the sori was more severely deformed (Fig. 5d). When ΔforC and wild-type cells were mixed at a 4:1 ratio, there were still stalks, but the stalks were shorter than in the above cases, and there were large cell masses that were probably unlifted sori at the bottom(Fig. 5e). Without the added wild-type cells, ΔforC cells formed stalk-like structures that were much shorter than those formed in the presence of added wild-type cells(Fig. 5f). This graded response indicates that the morphological defects in ΔforC development were not rescued through a synergetic effect elicited by mixingΔ forC cells with wild-type cells.

We made a chimeric gfp-forC gene by fusing gfp to the 5′-end of forC, and placed it downstream of the actin 15 promoter, which drives high levels of expression during the vegetative phase into the middle of the developmental phase(Knecht et al., 1986). Expression of GFP-ForC in ΔforC cells rescued their development, indicating this fusion protein functions in a way very similar to the native protein (Fig. 3D,right). When we initially observed living cells under a fluorescence microscope, GFP-ForC was seen throughout the cytoplasm, and no strong localization to any distinct component was observed(Fig. 6A). However, when we fixed the cells and extracted the cytoplasmic proteins, we found that GFP-ForC was localized to the crowns (Fig. 6Ba,b), which are macropinocytotic cups rich in F-actin. Staining GFP-ForC-expressing cells with rhodamine-phalloidin revealed that GFP-ForC does indeed colocalize with F-actin at the crowns(Fig. 6B). Furthermore,flattening live cells by overlaying them with a sheet of agarose made GFP-ForC present at the crowns detectable even without fixation(Fig. 6C).

The localization of GFP-ForC at crowns led us to suspect thatΔ forC cells may have defects related to the functions of the actin cytoskeleton. However, rhodamine-phalloidin staining failed to detect any noticeable differences in actin structures between ΔforCand wild-type cells in the vegetative phase (data not shown).

In order to determine which domain within ForC determines its localization in vivo, we expressed various truncated forms as GFP fusion proteins(Fig. 7A) and observed their distribution. GFP-ForC-1-633, a GFP-fused N-terminal half of the molecule, was distributed within cells exactly as GFP-ForC was — i.e., pan-cytoplasmic localization detectable in live cells and co-localization with F-actin at the crowns in fixed cells (Fig. 7Ba, live data not shown). Thus, the targeting sequence of GFP-ForC must reside in the N-terminal half of the molecule. GFP-ForC-1-468,which was truncated at amino acid residue 468 to remove the potential FH1 domain from GFP-ForC-1-633, was distributed in the same way(Fig. 7Bc,d). Interestingly,GFP-ForC-1-323 was detected at the crowns even in live cells without fixation,though there was still pan-cytoplasmic localization of the GFP-fused protein(Fig. 7C). Apparently,localization of GFP-ForC in the crown was enhanced by this truncation. By contrast, GFP-ForC-ΔFH3, which lacks N-terminal amino acids 1-312, was not detected at the crowns even after fixation(Fig. 7Bb). Thus, the sequence that targets ForC to the crowns must reside between amino acid residues 1 and 323 (i.e. within a region extending from the first methionine to the end of the FH3 domain).

None of the truncation mutants were functional: none rescued the development of the forC knockout mutant, and none disturbed either growth or development when expressed in wild-type cells (data not shown). Because crowns are structures responsible for macropinocytosis, we expected that overproduction of GFP-ForC-1-323 might perturb macropinocytosis by causing mislocalization of endogenous proteins. This does not appear to be the case, however, as assayed by measuring the rates of rhodamine-dextran uptake(data not shown).

Because GFP-ForC-1-323 could be detected at macropinocytotic cups without fixation, we were able to carry out time-lapse observation of Dictyostelium cells expressing GFP-ForC-1-323 using confocal microscopy (Fig. 8A). The GFP signal was detected at the edges of the ruffling membrane of macropinocytotic cups, enabling us to visualize their engulfing of the medium. In analogous fashion, we observed the GFP signal at the phagocytotic cups surrounding yeast cells (Fig. 8B). Finally, when cells expressing GFP-ForC-1-323 touched neighboring cells, a GFP signal was detected at the site where the cell protrusion touched the neighboring cell(Fig. 8C). There was no increase in fluorescence intensity at the corresponding site on the touched cell (Fig. 8C).

Since ForC probably works during the multicellular stages, we next tried to determine the intracellular localization of GFP-ForC-1-323 within multicellular structures. In order to reduce out-of-focus background fluorescence and to identify individual cells, we mixed wild-type cells harboring GFP-ForC-1-323 with those carrying the vector plasmid pBIG at a ratio of about 1:10 and allowed them to develop on agar plates. Culminating fruiting bodies were picked with tweezers, placed on coverslips and observed with a confocal microscope. Fibrillar fluorescent signals were detected in cells expressing GFP-ForC-1-323, but not in those expressing GFP alone(Fig. 9). We were able to identify boundaries of cells expressing GFP-ForC-1-323 when they were surrounded by nonfluorescent cells, and the fluorescent fibrillar structures were positioned along these cell boundaries. We speculate that these fibrillar structures are cortical actin structures at the sites of firm contacts between individual cells that constitute the multicellular structures.

The Dictyostelium genome contains at least nine formin genes or pseudogenes. Of these, four (forA, forB, forC and forD)appear in the cDNA database, and we have confirmed expression of forIby RT-PCR (C.K. and T.Q.P.U., unpublished). This makes it certain that at least five formin genes are expressed. Expression of the remaining four genes has not yet been verified, but each has a long uninterrupted ORF, and we have no evidence to suggest any are pseudogenes. Why are there so many genes that encode formin proteins in Dictyostelium?

One reason may be the presence of multiple cell types in the Dictyostelium life cycle: vegetative cells and starved cells that first differentiate into prestalk and prespore cells and then respectively into mature stalk and spore cells. There is also a relatively poorly characterized pathway to zygote formation(Urushihara, 1996). Each formin gene may be expressed in a particular cell type(s) during the life cycle of this organism, as was the case with forC. A second reason that Dictyostelium may express so many formin proteins is that different isoforms might have different and specific functions within each cell type. In the fission yeast, for instance, cdc12 is specifically required for the assembly of actin contractile rings, while for3 is required for organization of the actin cable(Feierbach and Chang,2001),

Nevertheless, one has to acknowledge that the repertoire of cell differentiation and cell architectures exhibited by Dictyosteliumduring its life cycle must be simpler than those of higher animal cells. Therefore, the large number of formin genes present in Dictyosteliummust be at least in part attributable to redundancy. The finding that a double mutant lacking both forA and forB showed no related phenotype suggests that there is at least one functionally redundant formin gene.

To our knowledge, ForC is the first formin family protein that does not possess an obvious proline-rich FH1 domain, though it clearly has both the FH2 and FH3 domains. The interaction of FH1 with profilin has been demonstrated for a number of formin proteins using biochemical and yeast two-hybrid assays(Chang et al., 1997;Evangelista et al., 1997;Imamura et al., 1997;Watanabe et al., 1997) and, in some cases, genetic interaction that supports this binding has also been observed (Chang et al., 1997;Evangelista et al., 1997;Imamura et al., 1997). Because the interaction with profilin via the FH1 domain has been observed in a wide variety of cells and organisms from yeast to mammals, it seemed a ubiquitous characteristic of formin proteins. Nevertheless, the absence of the FH1 domain suggests that ForC does not bind to profilin, and results of our yeast two-hybrid assays support this conclusion.

That GFP-ForC rescued ΔforC cells from their developmental defect suggests that the intracellular distribution of GFP-ForC reflects the distribution of native ForC. We first detected GFP-ForC in vegetative cells,even though ForC probably does not play an essential role in these cells; it was localized at the crowns and was detected only after fixation, which reduced background fluorescence by removing cytoplasmic GFP-ForC. Crowns are circular ruffles observed in Dictyostelium cells growing in liquid medium, and are the sites of macropinocytosis for fluid-phase uptake(Hacker et al., 1997). They are highly dynamic structures, with high concentrations of actin filaments. The localization of GFP-ForC at crowns suggests that the function of ForC is related to the actin cytoskeleton.

Macropinocytosis shares features with phagocytosis, and proteins known to be present at the crowns are also present at phagocytotic cups(Furukawa and Fechheimer,1994; Hacker et al.,1997). Likewise, ForC appears to localize at phagocytotic cups, as suggested by our detection of GFP-ForC-1-323 at the leading edges of membrane ruffles in the phagocytotic cups of live cells. Analogous to the presence of ForC at crowns and phagocytotic cups in Dictyostelium is the presence of mouse p140mDia at the phagocytic cups engulfing fibronectin-coated beads in Swiss 3T3 cells (Watanabe et al.,1997).

The localization of GFP-ForC at crowns was observable without fixation in cells subjected to agarose overlay. This might be due to flattening of the cytoplasm and the resultant reduction in background fluorescence derived from cytoplasmic GFP-ForC. Alternatively, the mechanical stress of the cell deformation caused by the agarose overlay might have enhanced the accumulation of GFP-ForC at the crowns. Because detection of GFP-ForC at crowns in the absence of agarose overlay was difficult using confocal microscopy (data not shown), we prefer the latter explanation. It has been reported that physical stress caused by agarose overlay enhances cortical localization of myosin II through dephosphorylation of threonine residues in the heavy chain(Neujahr et al., 1997). It may be that the same or an analogous stress-induced pathway is involved in enhanced translocation of ForC to the crowns.

Truncation analysis of GFP-ForC showed that the FH3 domain is important for targeting ForC to the crowns. FH3-dependent intracellular localization has also been observed with other formin proteins and appears to be a general feature of the FH3 domain (Kato et al.,2001; Petersen et al.,1998).

In a complementary experiment, GFP-ForC-1-323, which is truncated immediately after the FH3 domain, was detected at crowns without fixation or agarose overlay, suggesting that its affinity for the crowns is greater than that of the intact protein. Similarly, fission yeast Fus1 seems to have a stronger affinity for the presumptive FH3-binding site than the full length Fus1, as overexpression of Fus1-FH3-GFP perturbs the functions of other formin proteins, as well as Fus1 itself, probably by masking their localization sites(Petersen et al., 1998). We suggest that the FH3 domain contains a targeting sequence and that, in the native molecule, its affinity for the crowns is modulated by a regulatory domain within the same molecule. In ForC, this hypothetical regulatory domain must reside within a region extending from residue 323 to 468, as GFP-ForC-1-468 retained the same affinity for the crowns as the intact protein.

The stronger affinity of GFP-ForC-1-323 for its localization site enabled us to use it as a probe to examine the dynamic behavior of ForC in live cells. In this way, the motion of GFP-ForC-1-323 at the crowns and phagocytotic cups was visualized in vegetative cells. More interestingly, we found that when a cell touches another cell, GFP-ForC-1-323 accumulates at the site of attachment. Interpretation of this observation requires caution, since localization of native ForC and that of GFP-ForC-1-323 may differ. However,because localization of GFP-ForC-1-323 at crowns in live vegetative cells and that of GFP-ForC in fixed cells agreed with each other, and also because we were unable to detect localization of GFP-ForC-1-323 elsewhere, we believe this localization at the cell-cell attachment site is real. We speculate that ForC is recruited to sites of cell-cell attachment within multicellular aggregates, where it contributes to the formation of a firm `liner ` structure for efficient cell-cell adhesion through reorganization of the actin cytoskeleton. Analogous phenomena have been observed in fibroblasts, where activated mDia1 localizes at focal contact sites and mediates rearrangement of focal adhesion (Ishizaki et al.,2001).

RT-PCR analysis revealed there to be a low level of forC mRNA expression during the vegetative phase and the early developmental phase. However, the lack of any detectable ΔforC cell-specific phenotype suggests that ForC does not play an essential role during these phases. The phenotype of the forC knock out mutant (i.e. aberrantly shaped fruiting bodies with viable spores and the inability to form slugs)became apparent only after the tipped aggregate stage. A number of mutants affecting the actin cytoskeleton also show developmental defects similar to the ΔforC mutant. For instance, a double mutant lacking the actin crosslinking proteins, gelation factor and α-actinin, is unable to develop much beyond the mound stage, even though spore differentiation occurs normally (Witke et al., 1992). Cells lacking TalB, one of the two Dictyostelium homologues of talin, also stop at the mound stage, again despite normal spore differentiation (Tsujioka et al.,1999). Myosin II null mutants also arrest at the mound stage,though in this case viable spores are not formed(De Lozanne and Spudich, 1987;Knecht and Loomis, 1987). The phenotype of these mutants suggest that the culmination stage, which involves sorting differentiated cells within aggregates and movement of a multicellular mass of prespore cells up into the air along stalk cells, requires development of strong, coordinated motive forces that depend on the acto-myosin cytoskeleton. The similarity between the phenotype of ΔforCcells and other actin cytoskeletal mutants, as well as the intracellular localization of GFP-ForC-1-323, support the idea that ForC function is related to the actin cytoskeleton.

Unlike the case of the gelation factor/α-actinin double mutant and the TalB mutant (Tsujioka et al.,1999; Witke et al.,1992), culmination in ΔforC cells could not be rescued by mixing them with wild-type cells. The lack of a synergy effect suggests that ΔforC cells were sorted out of wild-type cells within aggregates. It may be that the actin cytoskeleton ofΔ forC is more severely disrupted than that of other actin-related mutants. Alternatively, ForC may be specifically involved in cell-cell contacts, and the synergistic coordination with neighboring wild-type cells in heterologous aggregates may be impaired, even though the general integrity of the actin cytoskeleton is intact. Of these two hypotheses, we favor the latter since vegetative cells, which do not adhere to one another, do not express high levels of ForC, and vegetativeΔ forC cells showed no mutation-related phenotype. This idea is also supported by the fact that, in multicellular forms, GFP-ForC-1-323 was detected at the edges of cells, which are the sites for cell-cell adhesion. This hypothesis is reminiscent of the finding by Riveline et al., who reported that in fibroblasts a locally applied mechanical force induces formation of focal contacts via a Rho-mDia pathway(Riveline et al., 2001). They speculated that this response is mediated by activated mDia1(Ishizaki et al., 2001), which induces FH2-dependent rearrangement of focal adhesions. Three conserved lysine residues in the FH2 domain of mDia1 are required for this activity(Ishizaki et al., 2001), and two of these lysine residues are conserved in the ForC FH2 domain. Moreover,as agarose overlay seems to enhance the translocation of GFP-ForC to the crowns, the localization of ForC seems to be controlled by physical stress. We therefore suggest that during multicellular processes of Dictyostelium, mechanical stress exerted by attachment to other cells leads to ForC-dependent reorganization of the local actin cytoskeleton and a strengthening of cell-cell contacts.

How might ForC achieve this effect? Several studies suggest that formin family proteins accelerate polymerization of actin filaments in vivo(Evangelista et al., 2002;Watanabe et al., 1999). In those cases, polymerization was dependent on the activities of the FH1 domain and profilin. Very recently, Bni1, a yeast formin, was found to promote nucleation of unbranched actin filaments in vitro(Pruyne et al., 2002;Sagot et al., 2002). Particularly noteworthy was that its FH2 domain is sufficient for the nucleation activity in vitro (Pruyne et al., 2002), although the profilin binding to FH1 domain enhances the acitivity to assemble actin structures in vivo(Pruyne et al., 2002;Sagot et al., 2002). Since ForC lacks a typical FH1 domain but still retains the FH2 domain, ForC may exert the actin nucleation activity that is independent from profilin in vivo. More study will be necessary to fully elucidate the function of ForC.

We thank H. Urushihara and the Dictyostelium cDNA project in Japan for the gift of cDNA clones, Dictyostelium genome project for allowing us to access the sequence information and J. Chuai for technical assistance. We also thank the New Energy and Industrial Technology Development Organization for the fellowship to C.K. during the initial phase of this study.