Dictyostelium DdCP224 is a microtubule-associated protein and a permanent centrosomal resident involved in centrosome duplication

From an evolutionary perspective, the Dictyostelium centrosome is an intriguing organelle. Like its counterpart in yeast, the spindle pole body, it exhibits a compact, layered structure that lacks the centrioles typical of metazoan cells, but as in the latter, the core is surrounded by an electron-dense, amorphous matrix. With vertebrate centrosomes it shares a cytoplasmic localization in interphase, but it inserts itself into an opening of the nuclear envelope during mitosis like the yeast spindle pole body, which is a permanent resident of the nuclear envelope. This centrosome with its hybrid features resides in a cell type that resembles mammalian cells in terms of behavioral repertoire and motile properties, making Dictyostelium amoebae an important model system for the analysis of centrosome structure, function, and evolution.

The Dictyostelium centrosome is located in close proximity to the nucleus (Roos, 1975) to which it is tightly connected via a fibrous linkage (Omura and Fukui, 1985). It consists of a box-shaped core structure with three major layers surrounded by a corona, which is composed of regularly spaced, dense nodules embedded in an amorphous matrix (Moens, 1976; Roos, 1975). The nodules seem to be the sites of microtubule nucleation since they contain γ-tubulin, and since all interphase microtubules emanate from these nodules (Euteneuer et al., 1998). Thus the corona seems to be the functional equivalent of the pericentriolar matrix of animal cells. Recently, the mode of centrosome duplication in Dictyostelium was elucidated in detail by a combination of electron microscopic analysis of fixed cells and observation of living cells tranformed with γ-tubulin-GFP, which possess green fluorescing centrosomes (Ueda et al., 1999). In contrast to animal cells and yeast, the entire duplication process occurs during mitosis. Centrosome duplication starts in prophase with an enlargement of the three-layered core. In late prophase, the corona dissociates and the interphase microtubules are lost. The central layer disappears at the transition to prometaphase, and the two outer layers peel apart and become the mitotic centrosomes, or spindle pole plaques. Microtubules are nucleated at the formerly inner surfaces of the two layers to form a nascent spindle that starts to separate the spindle pole plaques. During the separation process the edges of the plaques bend away from the nucleus until, in telophase, each plaque folds back onto itself. As a result of this folding process the microtubule-nucleating surface turns into the outside of the new daughter centrosome while the former outer surface becomes buried inside (Ueda et al., 1999). These morphological changes during centrosome duplication are unique among eukarytic cells.

At present, relatively little is known about the centrosomal components that orchestrate this intriguing duplication process and carry out all other centrosomal activities. So far, molecular analyses were based on the expected similarity of Dictyostelium centrosomal proteins to their homologues in other species. Thus, Dictyostelium γ-tubulin was cloned by a PCR approach using degenerated primers (Euteneuer et al., 1998). Recently, four more centrosomal proteins were identified by database analysis (Gräf et al., 2000) of the Dictyostelium cDNA project (Morio et al., 1998) and genome project (http://genome.imb-jena.de/Dictyostelium). They include the Dictyostelium homologues of centrin (Schiebel and Bornens, 1995), Spc97p, Spc98p (Knop and Schiebel, 1997) and human Nek2, a centrosomal NIMA-related kinase involved in centrosome duplication (Fry et al., 1998a,b).

However, novel or weakly conserved centrosomal proteins cannot be uncovered by database searches. An alternative aproach is immunoscreening of DNA libraries with specific monoclonal antibodies (mAbs), as successfully performed in yeast (Donaldson and Kilmartin, 1996; Kilmartin et al., 1993; Wigge et al., 1998). Following the development of a protocol for the isolation for Dictyostelium centrosomes in high quantity and purity (Gräf et al., 1998), 14 new monoclonal antibodies against Dictyostelium centrosomes were generated (Gräf et al., 1999). One of these antibodies was used here to clone a 224-kDa centrosomal component which we have named DdCP224 (Dictyostelium discoideumcentrosomal protein). Sequence comparison revealed that this protein belongs to a family of weakly conserved MAPs including, among others, human TOGp (Charrasse et al., 1995), and yeast Stu2p (Wang and Huffaker, 1997). The analysis of DdCP224 in wild-type cells and GFP mutants provides insights into the centrosome and microtubule-binding properties of this protein. Moreover, unexpected and potentially significant differences to other members of this protein family are revealed, and a novel function in centrosome duplication is suggested.

Vegetative Dictyostelium discoideum amebae and hybridoma cells were cultivated as described previously (Gräf et al., 1999). DdCP224-GFP mutants were grown at 21°C in HL-5c medium containing 4 μg/ml blasticidin S (ICN Biomedicals, Eschwege, Germany) or 5 μg/ml G418 (Sigma, Deisenhofen, Germany), respectively.

Approximately 2×10⁹Dictyostelium cells yielding ∼100 mg of total cytosolic protein were used for enrichment of DdCP224 and its GFP versions. All steps were performed at 4°C or on ice except chromatography which was done at room temperature and all solutions contained a protease inhibitor cocktail (1 mM Pefabloc SC, 25 μg/ml leupeptin, 10 μg/ml tosyl-arginine-methyl ester, 10 μg/ml soybean trypsin inhibitor, 1 μg/ml aprotinin, 1 μg/ml pepstatin, 1 mM benzamidine, all from Biomol, Hamburg, or Sigma, Deisenhofen, Germany). Cells (prepared as described by Gräf et al., 1998) were suspended in 10 ml SPLB (20 mM Na-Pipes (pH 6.9), 30 mM NaCl, 2 mM MgCl₂, 10% sucrose, 1 mM ATP, 1 mM DTT) and lysis was achieved by filtration through 5 μm polycarbonate filters (Nuclepore, Corning Costar, Bodenheim, Germany). The lysate was cleared by centrifugation with 20000 rpm for 15 minutes and the supernatant was loaded onto a SP-Sepharose cation exchange column (Amersham-Pharmacia, Freiburg, Germany). After washing with WB (20 mM Na-Pipes, pH 6.9, 30 mM NaCl, 2 mM EGTA 4 mM MgCl₂, 1 mM DTT), DdCP224 was eluted with 0.5 ml 100 mM Na-Pipes (pH 6.9), 150 mM NaCl, 2 mM EGTA, 4 mM MgCl₂, 1 mM DTT.

For microtubule binding, the eluate was diluted with an equal volume of 4 mM MgCl₂/2 mM EGTA. The diluted protein sample was cleared by centrifugation at 260000 g for 10 minutes (Beckman TLA100.3 rotor). Porcine brain tubulin (Mandelkow et al., 1985) was polymerised for 15 minutes at 37°C by addition of 1 mM GTP and, after 5 minutes, of 24 μM taxol. 25 μl (3 mg) of microtubules were mixed with 500 μl of the cleared protein sample (∼0.3 mg of protein), supplemented with 24 μM taxol and incubated for 5 minutes at 25°C. Microtubules and associated proteins were sedimented through a sucrose cushion (100 μl of 40% sucrose, 100 mM Na-Pipes, 2 mM MgCl₂, 24 μM taxol) at 260000 g for 10 minutes at 25°C (Beckman TLA100.3 rotor). Microtubule pellets were resuspended in 50 μl of 0.5× urea sample buffer (4.5 M urea, 5% SDS, 2.5% β-mercaptoethanol, 125 mM Tris/HCl, pH 6.8). Proteins in the supernatants were precipitated with TCA (Bollag et al., 1996) and dissolved in 50 μl of 0.5× urea sample buffer.

1. ml of cytosolic extract prepared as described above was diluted with WB to reduce the sucrose concentration to 7% and loaded onto a sucrose step gradient of 0.5 ml each of 10%, 15%, 20%, 25%, 30%, 35% and 40% sucrose in WB including protease inhibitors (see above). The gradient was centrifuged for 4 hours at 234000 g (Beckman SW50.1 rotor) and was fractionated from the bottom in 0.3 ml steps. The proteins in each fraction were precipitated with TCA and dissolved in 30 μl 0.5× urea sample buffer.

μg of double-stranded cDNA with EcoRI/NotI adaptors were prepared from mRNA of vegetative myosin null mutants (Manstein et al., 1989; QuickPrep mRNA micro and TimeSaver cDNA Synthesis Kit from Amersham-Pharmacia, Freiburg, Germany) and size-fractionated on a 1.5% agarose gel. An agarose block including all cDNAs longer than 4.5 kb was excised. The cDNA was recovered by a glass milk procedure (JetSorb, Genomed, Bad Oeynhausen, Germany), ligated into λZAPII-EcoRI arms and packaged in vitro (GigapackIII, Stratagene, Amsterdam, Netherlands). The titer of the primary library was 1×10⁵ pfu. Immunoscreeing of ∼3×10⁵ pfu of the amplified cDNA library with the mAb 4/148 (Gräf et al., 1999) lead to the isolation of 6 positive clones. 4 clones contained a ∼6.3 kb insert and the two other clones were false positives since they encoded EF1α. Thus, only one of the 6.3 kb clones was sequenced completely on both strands (MWG Biotech, Ebersberg, Germany).

GFP constructs with DdCP224 were created in vectors based on the C-terminal GFP fusion vector pB15-GFP (Ueda et al., 1997), a derviative of pDXA-3C (Manstein et al., 1995). The SacI-site downstream of the S65T-GFP stop codon was destroyed by PCR-based point mutation resulting in pB15GFPXSac. Fusion protein expression is driven by the constitutive actin 15 promoter. A further cloning vector, p1ABsr8, was constructed from pB15GFPXSac by replacing the G418 resistance cassette by the blasticidin resistance cassette from pUCBsrΔBam (XbaI-XhoI fragment; Adachi et al., 1994). All three constructs described below were cloned into pB15GFPXSac and p1ABsr8 after BamHI/SacI double digestion. The polypeptide chain of all DdCP224 mutants is preceded by the vector-derived peptide MDGTEL and the DdCP224 and GFP sequences are separated by a glycine and a serine encoded by the BamHI restriction site.

For expression of the N-terminal part of DdCP224 (ΔC-GFP), the cDNA encoding the N-terminal half of DdCP224 (amino acids 1-813) was amplified by PCR with a sense SacI-primer/linker and an antisense BamHI-primer/linker. For the construction of expression vectors for full-length DdCP224 and its C-terminal part with GFP (DdCP224-GFP and ΔN-GFP) the last 78 bp of coding sequence and the 3’-untranslated region of the original DdCP224 clone in pBluescript were excised by cleavage with SpeI and SacI and replaced by a 86mer custom-synthesized linker substituting the excised coding sequence and introducing a BamHI recognition site at the 3’-end. In an independent step, the first 543 bp of coding sequence were amplified by PCR using a sense SacI-primer/linker and an antisense primer downstream from the NcoI restriction site. Then the SacI/NcoI 5’-fragment of the original DdCP224 clone (745 bp of cDNA including the 5’-untranslated sequence) was replaced by the PCR product. The insert of this plasmid contained only the complete coding sequence of DdCP224. Similarly, in the case of the ΔN-GFP expression vector, 304 bp of coding sequence starting at base position 2660 were amplified by PCR using a sense SacI-primer/linker and a suitable antisense primer downstream from the SphI restriction site. The first 2938 bp including the 5’-unstranslated region were deleted by SacI/SphI double digestion and replaced by the PCR product.

All PCR-generated sequences of DdCP224 were verified by DNA sequencing. Plasmids were transformed into AX2 cells by using electroporation or the calcium phosphate method (Mann et al., 1998; Nellen et al., 1987). All data shown in the figures and tables of this paper were obtained with the blasticidin-resistant mutants; the G418-resistant strains showed the same phenotypes and were only used for comparison.

SDS polyacrylamide electrophoresis, immunoblotting, silver staining, indirect immunofluorescence microscopy and confocal light microscopy were performed as described previously (Gräf et al., 1998, 1999; Ueda et al., 1997).

Three of the 14 mAbs against purified Dictyostelium centrosomes obtained recently (2/165, 4/95 and 4/148) reacted with the same ∼200 kDa protein band in western blots of isolated centrosome preparations (Gräf et al., 1999). Further immunoblot analysis revealed the antigen to be ∼5-fold more abundant in cytosolic than in nucleus/centrosome extracts (Fig. 1). Since gel electrophoresis and immunoblotting showed that the ∼200 kDa protein was a minor protein component, we expected the corresponding cDNA to be highly underrepresented in conventional cDNA libraries. Indeed, initial attempts to clone the corresponding cDNA by immunoscreening of a random-primed Dictyostelium cDNA library with the three mAbs failed. Thus we generated a size-fractionated cDNA library containing mainly cDNAs longer than 4.5 kb. These cDNAs comprise only a minor amount of the total cDNA, hence the cDNA encoding the ∼200 kDa protein should be highly enriched in this library. Indeed, six positive clones could be isolated by immunoscreening with mAb 4/148. All positives were also recognized by mAbs 2/165 and 4/95 confirming that all three mAbs are directed to the same antigen. Complete sequencing of one of these clones yielded 6325 bp of cDNA sequence (EMBL accession no. AJ012088) with one complete open reading frame encoding a basic protein (pI = 8.06) of 2015 amino acids with a calculated molecular mass of 224,126 Da, which was designated DdCP224. No functional sequence motifs known from other proteins could be detected, and unlike many other centrosomal proteins it apparently does not contain long coiled-coil regions.

However, employing the coilscan program at a window size of 21 amino acids, two short coiled-coil regions (amino acid position 1211-1234 and 1989-2015) were predicted (Lupas et al., 1991). These stretches are probably too short to promote stable interactions based on a coiled-coil.

Amino acid sequence comparison with EMBL data library entries revealed a number of weakly homologous proteins. The closest homologues are the human TOG protein, the Drosophila Msps protein and Xenopus XMAP215, whose complete sequence was published during the revision process of this work (Charrasse et al., 1995; Cullen et al., 1999; Tournebize et al., 2000). All three proteins share ∼40% similarity and ∼30% identity with the DdCP224 sequence (Fig. 2A). The homology group also includes Caenorhabditis elegans ZYG-9 (Matthews et al., 1998), Saccharomyces cerevisiae Stu2p (Wang and Huffaker, 1997), and Schizosaccharomyces pombe p93^dis1 (Nabeshima et al., 1995). All these proteins bind to microtubules and appear to have centrosomal functions, at least during mitosis, but they exhibit only weak amino acid similarity to each other and differ markedly in length (Fig. 2A). Thus the two yeast proteins have a size of only ∼100 kDa and are homologous to the N-terminal half only of the Dictyostelium and human proteins. ZYG-9 has a size of 150 kDa and, compared to TOGp, the Msps protein and DdCP224, it possesses only two conserved ∼250 amino acid sequences (Fig. 2B,C). They were called region 1 and 2 (Matthews et al., 1998). Region 2, which is missing in the yeasts, is located more N-terminal in ZYG-9 than in DdCP224 and TOGp. Region 1 (designated region 1a and 1b in Fig. 2B) is clearly duplicated in ZYG-9, and at a lower level of conservation, both regions may occur in more than one copy in the other species as well (Nakaseko et al., 1996; Cullen et al., 1999). Recently, Cullen et al. (1999) suggested that even four tandemly arranged copies of region 1 exist in the N-terminal half of the Msps and TOG protein sequence, whereas region 2 is duplicated in both animal proteins. However, in the case of DdCP224 both sequence motifs are not clearly repeated. With respect to region 1, only the short sequence KKILADVNVM (amino acid position 318-327) is clearly repeated more N-terminally where it reads KKILADINPM (amino acid position 53-62). Yet, there is no experimental evidence for the biological function of the repeated motifs in any of these proteins.

The dynamics of DdCP224 localization through the cell cycle were studied by confocal microscopy. Cells were double-stained with anti-DdCP224 mAbs and anti-γ-tubulin as a centrosomal marker (Fig. 3). During interphase, DdCP224 was localized to the centrosomal corona which appears as a doughnut-shaped structure since it surrounds the unstained core of the Dictyostelium centrosome (Fig. 3A’). This staining pattern was similar to that of γ-tubulin antibodies (Fig. 3A). However, DdCP224 seems to reside more in the periphery of the corona compared to γ-tubulin because the diameter of the structure stained with anti-DdCP224 was always bigger than the anti-γ-tubulin labeling at the same centrosome (Fig. 3A’’). Weak DdCP224 staining could sometimes be detected along interphase microtubules as well (see below). the interphase microtubule cytoskeleton together with dissociation of the corona in prophase (Kitanishi-Yumura and Fukui, 1987; Ueda et al., 1999). This implies that DdCP224 is redistributed from the corona to the two outer layers of the core structure which start to form the spindle poles in prometaphase. DdCP224 was also present at the corona of isolated centrosomes (Fig. 4A) which lack microtubules (Gräf et al., 1998), strongly suggesting that its localization at the centrosome is not dependent on the presence of an intact

During mitosis, strong centrosomal staining persisted without any detectable change in intensity (Fig. 3B’-G’). In prophase, the doughnut-shaped structure stained with anti-DdCP224 and anti-γ-tubulin became larger and more oval, but DdCP224 was still located farther outside than γ-tubulin (Fig. 3B-B’’). Later on, both antigens colocalized almost exactly at the spindle poles (Fig. 3D’’). In addition to centrosomal labeling, anti-DdCP224 mAbs stained the region where kinetochores reside in metaphase cells (Fig. 3D’) in a manner similar to the MPM2 antibody (Fig. 3E’; Engle et al., 1988), a good centrosomal and kinetochore marker in mitotic Dictyostelium cells (Ueda et al., 1999). Spindle microtubules are also stained, especially in the midbody region in anaphase and telophase (Fig. 3D’-G’).

DdCP224 was found at the centrosome throughout the entire cell cycle, irrespective of the dramatic changes in centrosomal morphology during mitosis, which includes breakdown of microtubule system. When cells were treated with the microtubule depolymerizing drug thiabendazole (Fig. 4C-E; Kitanishi et al., 1984) for 3 hours, long microtubules emanating from the centrosome were absent (Fig. 4C’,D’), but the centrosomal presence of DdCP224 was unchanged (Fig. 4C-E’). In contrast, kinetochores in metaphase-like mitotic figures were unlabeled with DdCP224 antibodies but still showed labeling by the MPM2 antibody (Fig. 4D,E). Since mitotic cells are unable to form proper spindles in the presence of thiabendazole (Kitanishi et al., 1984; Kitanishi-Yumura et al., 1985) DdCP224 localization in the kinetochore region requires either its transport along microtubules towards the spindle midzone or the presence of microtubules at the kinetochore.

The localization of DdCP224 throughout the cell cycle is similar to that of yeast Stu2p, the only other protein of this family that is also permanently located at the microtubule-organizing center (Wang and Huffaker, 1997). As DdCP224, Stu2p binds weakly along microtubules in interphase and mitosis as well. Since Stu2p is only half as long as DdCP224 and corresponds to the N-terminal half only, we wondered about the function of the C-terminal half of DdCP224. We generated three types of Dictyostelium mutants expressing the full-length protein (DdCP224-GFP), the N-terminal half (ΔC-GFP, amino acids 1-813) and the C-terminal half (ΔN-GFP, amino acids 809 to 2016) of DdCP224 as a C-terminal fusion to GFP. The site separating the two halves of DdCP224 was chosen based on sequence alignments with Stu2p and p93^dis1 using the GAP and PILEUP programs of the GCG package, which placed the C-termini of the Stu2p and p93^dis1 at approximately the same position. The transformation plasmids insert randomly into the genome and expression is driven by the actin 15 promoter with wild-type DdCP224 as background. All three GFP-mutants were viable in axenic shaking culture and were able to form fruiting bodies immunoblots stained with anti-GFP and anti-DdCP224 mAbs (Fig. 5). The electrophoretic mobility of DdCP224-GFP, ΔC-GFP and ΔN-GFP in SDS-gels was as predicted for GFP-fusion proteins with calculated molecular masses of 251, 127 and 162 kDa, respectively. The epitope for the anti-DdCP224 mAbs is located in the C-terminal half of DdCP224, since ΔC-GFP could not be labeled by these mAbs (data shown only for 4/95 in Fig. 5). Thus the expression levels of the fusion proteins can be estimated from protein band intensities only in the case of DdCP224-GFP and ΔN-GFP. Cytosolic extracts of DdCP224-GFP cells contain equal amounts of fusion protein and spores. No phenotypic differences were observed between several independent transformants. Therefore, all the data reported here are based on one representative mutant of each type.

The expression of the expected fusion protein was confirmed by (B-E). and endogenous DdCP224, whereas ΔN-GFP cells show approximately 10-fold overexpression of the fusion protein.

As predicted, DdCP224-GFP cells exhibited strong GFP fluorescence at the centrosome in interphase and mitosis (Figs 6A’-B’, 9). Fluorescence along interphase microtubules was also clearly visible (Figs 6A’, 9A’,B’). Hence, the labeling pattern was indistinguishable from that of endogenous DdCP224 visualized by indirect immunofluorescence in untransformed cells (Fig. 3A’). In contrast to full-length DdCP224, the GFP localization pattern of the two truncated mutants was quite unexpected. ΔC-GFP was not localized at either microtubules or centrosomes; rather, it was distributed uniformly over the cytosol (Fig. 6C-C’’), and there was no difference between interphase and mitosis (not shown). ΔN-GFP mutants, on the other hand, displayed strong GFP fluorescence at the centrosome and no detectable labeling of microtubules in interphase (Fig. 6D’). During mitosis, spindle microtubules were labeled in addition to the centrosomes and the kinetochore region (Fig. 6E’,F’). These results show that the centrosome targeting domain of DdCP224 resides in the C-terminal half of the molecule, which is missing in the spindle pole body (SPB) component Stu2p, and they confirm that centrosomal localization and microtubule binding are separable functions. However, microtubule binding evidently requires the interplay of both parts of the protein.

DdCP224 binds to microtubules in vitro and behaves as a monomeric protein

The microtubule-binding properties of DdCP224 were studied in more detail in vitro. DdCP224 and all three GFP-mutants were enriched from cytosolic Dictyostelium extracts by cation exchange chromatography. Further purification attempts were hampered by problems with proteolysis and low protein yields, but this partially purified preparation was found sufficient to test for binding to microtubules in a spin-down assay. In untransformed AX2 cells, most of the cytosolic DdCP224 co-sedimented with taxol-stabilized pig brain microtubules (Fig. 7), whereas almost no DdCP224 was detected in the pellet in the absence of microtubules. Co-sedimentation experiments with the three GFP fusion proteins were fully consistent with the microscopic observations. Most of the DdCP224-GFP fusion protein could be sedimented with microtubules, but ΔC-GFP and ΔN-GFP exhibited no detectable microtubule binding. These findings show that DdCP224 binds to microtubules, but we cannot judge from this result whether DdCP224 interacts directly with microtubules or via another associated protein.

To determine whether DdCP224 can self-associate or form a complex with other proteins, Dictyostelium cytosolic extracts were loaded onto sucrose density gradients. DdCP224 fractionated almost exactly as catalase (232 kDa) and thus behaved like a monomeric protein in these gradients (Fig. 8). Fractionation was unaffected by the presence of 0.1% Triton X-100 (data not shown).

DdCP224-GFP mutants displayed informative defects in centrosome duplication and cytokinesis (Table 1). In axenic shaking culture, only ∼25% of the transformants contained one nucleus and one centrosome (Fig. 6A), whereas more than 50% had more than one centrosome per nucleus (Fig. 9) These supernumerary centrosomes were usually not linked to the nucleus but seemed to be functional in microtubule nucleation since they were the center of a microtubule aster (Fig. 9A’,B’).

Furthermore, labeling of supernumerary centrosomes with antibodies against γ-tubulin and the 350-kDa centrosomal antigen (Kalt and Schliwa, 1996) was indistinguishable from that of nucleus-associated centrosomes (data not shown), and the doughnut-like appearance of anti-DdCP224-labeled supernumerary centrosomes in confocal microscopy suggests that they possess a normal corona as well (Fig. 9F-F’’’). Unusual centrosomal shapes such as large elongated, dumbbell-, or kidney-shaped centrosomes at the nucleus were also observed (Fig. 9C,D,E). These aberrant shapes may be the result of incomplete centrosome duplication. Moreover, 50% of all cells possessed more than one nucleus, and giant cells with 6 and more nuclei were quite frequent. Moreover, unequal cytokinesis is suggested by an abundance of tri- and hexanucleate cells. Taken together, these observations indicate that DdCP224-GFP cells have a cytokinesis defect as well (Fig. 9A). Since multinuclear cells exhibited supernumerary centrosomes with about the same frequency as mononuclear cells (Table 1), the centrosome duplication defect and the cytokinesis defect seem to be independent of each other.

As DdCP224-GFP cells, approximately 50% of all ΔC-GFP cells had two or more nuclei (Fig. 6C), but immunofluorescence analysis with the mAb 4/148 revealed that only a small fraction of these mutant cells contained supernumerary centrosomes GFP fluorescence intensity (not shown). In contrast, ΔN-GFP mutants showed no centrosome duplication or cytokinesis defect at all (Table 1). Supernumerary centrosomes were not observed and less than 30% of all cells contained more than one nucleus, and hardly any cell more than two. These cells therefore were essentially indistinguishable from the untransformed AX2 strain.

Surprisingly, the strikingly frequent centrosomal defects of DdCP224-GFP mutants and the cytokinesis defects of both DdCP224-GFP and ΔC-GFP cells did not affect the growth rates of these cell lines. As the normal-appearing ΔN-GFP cells, these two mutants had maximum doubling times between 10 and 11 hours in axenic shaking culture at 21°C. Untransformed cells divide only slightly faster under these growth conditions (doubling time 9-10 hours; data not shown). Thus the extra The occurrence of more than one nucleus did not correlate with centrosomes apparently did not interfere significantly with mitosis. We frequently observed mitotic cells in meta- or anaphase with normal spindles and supernumerary centrosomes in the cytosol not associated with the chromosomes or spindles (Fig. 9G’,H’). Dictyostelium undergoes closed mitosis, so the nuclear envelope might protect the mitotic spindle from interference by these supernumerary, cytosolic centrosomes. The cytokinesis defect does not seem to be severe, and it apparently can be overcome in a subsequent cell cycle. Fig. 9J shows a once trinucleate cell in telophase undergoing cytokinesis that appears to lead to 6 daughter cells. Multiple fission events such as these might explain the good viability of DdCP224-GFP cells despite their centrosomal and cytokinesis defects.

Using mAbs raised against isolated Dictyostelium centrosomes we have obtained the complete cDNA sequence of a protein we have named DdCP224. It is a new member of a family of microtubule-binding proteins consisting of human TOGp, the Drosophila Msps protein, Xenopus XMAP215, C. elegans ZYG-9, S. cerevisiae Stu2p and S. pombe p93^dis1 (Charrasse et al., 1995; Cullen et al., 1999; Matthews et al., 1998; Nabeshima et al., 1995; Tournebize et al., 2000; Wang and Huffaker, 1997). The primary structures of these proteins are only weakly conserved and show some striking differences in domain organization. These likely reflect unique cell biological properties of these proteins. Among the sequences published so far, DdCP224 is the first non-animal member of this family which displays sequence similarity over the entire sequence to the human, Xenopus and Drosophila proteins and which has approximately the same size. Surprisingly, DdCP224 is more closely related to the Drosophila protein than ZYG-9, the other invertebrate member of this protein family.

Curiously, the cell biological properties of DdCP224 shows the closest similarity to Stu2p (Wang and Huffaker, 1997) whose amino acid sequence is the most distantly related. Whereas all seven proteins localize to spindle microtubules and spindle poles during mitosis (Charrasse et al., 1998; Cullen et al., 1999; Gard et al., 1995; Matthews et al., 1998; Nabeshima et al., 1995; Wang and Huffaker, 1997), only DdCP224 and Stu2p are generally localized at the centrosome (SPB in yeasts) in interphase as well. In both cases, the presence at the centrosome is unaffected by microtubule depolymerising drugs, suggesting that microtubules are not required for centrosomal localization. The presence of DdCP224 at isolated centrosomes and the association of the ΔN-GFP protein with centrosomes but not microtubules (see below) are consistent with this conclusion. Thus DdCP224 is a genuine centrosomal component, unlike the DdCP224-like proteins in animals which have other localizations in interphase. For example, TOGp colocalizes with markers of the endoplasmic reticulum (Charrasse et al., 1998), while ZYG-9 is distributed in the cytoplasm (Matthews et al., 1998). Less is known about the cellular distribution of XMAP215 (Gard et al., 1995). The Msps protein in cellularized Drosophila embryos is also absent from the centrosome during interphase, but it localizes to the centrosome also in interphase during the first 13 cell cycles of the syncycial embryo (Cullen et al., 1999). But these early cell cycles are unique: They are extremely fast since there are no gap phases, DNA replication checkpoints or cell divisions. Thus the permanent centrosomal presence of the Msps protein during these early cell cycles might be due to the very short interphase which lasts only ∼10 minutes. The two yeast proteins Stu2p and p93^dis1 were localized along microtubules in interphase by fluorescence microscopy just as DdCP224, suggesting that they bind laterally along microtubules. In the case of Stu2p, this lateral association could be confirmed by microtubule binding experiments in vitro (Wang and Huffaker, 1997). Taken together, the members of this protein family exhibit considerable variety in their localization patterns.

The strikingly similar localization patterns of DdCP224 and Stu2p raises the question of the role of the C-terminal half of DdCP224, since Stu2p corresponds to the N-terminal half of DdCP224 only. Therefore, we generated the ΔC-GFP mutant, which is the topological equivalent of the yeast proteins, its ΔN-GFP counterpart, and the full-length mutant, DdCP224-GFP, for reference. DdCP224-GFP behaved as expected, i.e. its major localization was at the centrosome during the entire cell cycle as well as along interphase-microtubules. Unexpectedly, ΔN-GFP localized to centrosomes in interphase and mitosis to the same extent as full-length DdCP224-GFP. Thus the centrosomal targeting domain of DdCP224 resides in its C-terminal half. The binding of Stu2p to the SPB must therefore be achieved by a different mechanism and might involve different protein binding partners. One possible candidate is the SPB component Spc72p which also serves as the docking site for yeast γ-tubulin complexes (Tub4p-complexes) at the SPB (Knop and Schiebel, 1998). Chen et al. (1998) have shown that Stu2p and Spc72p bind to each other and form a cytosolic complex. At the SPB, Spc72p likely anchors both the Tub4p complex and Stu2p. No homologue for Spc72p has so far been found in higher organisms or in Dictyostelium. Thus one might speculate that a Spc72p homologue does not exist in Dictyostelium because anchoring of DdCP224 is accomplished by its C-terminal domain, which serves the function of a Spc72p-like protein. Furthermore, unlike Stu2p, cytosolic DdCP224 behaves as a monomer in sucrose density gradients suggesting that there is no Spc72p-like binding partner of DdCP224. The lack of sequence similarity between the DdCP224 C-terminal half and Spc72p does not exclude the possibility that they are functionally equivalent, at least in part.

The issue of microtuble binding seems to be more complicated. In interphase, the full-length GFP fusion protein (DdCP224-GFP) is localized along microtubules, whereas ΔN-GFP is present at centrosomes only. During mitosis both constructs were detected at the mitotic spindle and at the spindle poles. This suggests that the N-terminal half is required for binding along cytoplasmic microtubules but not involved in binding to spindle microtubules. Hence binding to cytoplasmic and spindle microtubules, respectively, might be independent activities of DdCP224. However, the N-terminal half alone does not seem to be sufficient for binding to cytoplasmic microtubules because ΔC-GFP was present neither at microtubules nor at the centrosome. The topologically equivalent yeast protein, on the other hand, has to provide the binding sites for both localizations. Consequently, microtubule binding appears to require interaction between the N- and C-terminal half of DdCP224. The microscopic observations were confirmed by in vitro microtubule binding assays where wild-type DdCP224 and DdCP224-GFP but neither of the two truncated GFP fusion proteins cosediment with taxol-stabilized pig brain microtubles. Since DdCP224 could not be purified to homogeneity we cannot exclude that microtubule binding is accomplished via a DdCP224 associated protein. However, since the vast majority of cytosolic DdCP224 seems to be monomeric it is likely that DdCP224 binds directly to microtubules.

Sequence comparison does not reveal the microtubule binding determinants of DdCP224. TOGp possesses a motif close to its C terminus that is similar to the microtubule binding repeats of MAP2, MAP4 and tau (Charrasse et al., 1998), but DdCP224 and both yeast proteins do not contain such a consensus sequence. The microtubule binding sites of p93^dis1 and Stu2p could be mapped to a ∼100 amino acid region starting approximately at amino acid position 550 (Nakaseko et al., 1996; Wang and Huffaker, 1997). In Stu2p, this region includes two imperfect repeats that both contribute to microtubule binding, but the corresponding region in p93^dis1 does not contain these repeats and shows no striking similarity to Stu2p. DdCP224 does not exhibit any similarity in this region to either of the two yeast proteins. Thus the sequence motifs or domains that contribute to microtubule binding have yet to be identified.

The functions of DdCP224-like proteins at microtubules are not known. It is conceivable that these functions vary in different organisms as suggested by their considerable sequence divergence. However, particularly in the case of XMAP215, the microtubule binding properties have been thoroughly studied in vitro (Andersen, 1998; Gard and Kirschner, 1987; Tournebize et al., 2000; Vasquez et al., 1994, 1999). These studies suggest that XMAP215 binds along the entire length of microtubules, and that it mainly functions as a stimulator of microtubule plus-end assembly and turnover without blocking catastrophes. Since XMAP215 seems to be regulated by CDK1/cyclin B (Vasquez et al., 1999; Charrasse et al., 2000) it may play a role in promoting the increased microtubule dynamics at the transition from interphase to mitosis (Desai and Mitchison, 1997).

The analysis of the DdCP224-GFP and ΔC-GFP mutants provides strong evidence for an important role of DdCP224 in centrosome duplication. Such a function has not been discussed for any other member of this protein family so far. More than 50% of all DdCP224-GFP cells contained supernumerary centrosomes not associated with the nucleus. Furthermore, these mutants had a cytokinesis defect as well, since about 50% were multinuclear, often with more than two nuclei per cell. However, the two defects do not seem to be directly linked since supernumerary centrosomes did not occur at a higher frequency in multinuclear than in mononuclear cells. It is unlikely that these defects are caused by integration of the transformation vector into another gene also involved in centrosome duplication or cytokinesis, mainly for two reasons: first, the mutant phenotypes were observed in five independent DdCP224-GFP and three independent ΔC-GFP mutants, respectively; and second, ΔN-GFP mutants based on the same GFP-transformation plasmid possess normal phenotypes. Moreover, the presence of GFP at the centrosome can also be excluded as a reason for these defects since ΔN-GFP cells and γ-tubulin-GFP cells (Ueda et al., 1997) appear normal. Thus the centrosomal and cytokinesis defects likely are caused by the expression of supernumerary copies of DdCP224-GFP and ΔC-GFP, respectively, in a wild-type background of DdCP224. The GFP forms, which are slightly bulkier due to the GFP-tag, may compete with the endogenous DdCP224 and slightly impair its function. The defects observed here then imply that DdCP224 plays a role in centrosome duplication. Interestingly, ΔN-GFP does not impair the function of wild-type DdCP224 even though it shows correct centrosomal localization. Conceivably, due to the N-terminal deletion, it may be unable to interact with microtubules or other potential protein partners involved in centrosome duplication or cytokinesis. Since DdCP224 is both a centrosomal and a microtubule binding protein, defects in its crosstalk with microtubules might cause the centrosome duplication deficiency. Microtubules are not required for centrosome duplication per se but rather for completion of the centrosome cycle, which includes separation of the duplicated centrosomes. This conclusion is supported by earlier experiments with thiabendazole-treated Dictyostelium cells (Kitanishi et al., 1984; Kitanishi-Yumura et al., 1985) which have defects similar to DdCP224-GFP cells.

Spindle microtubules interact not only with the centrosome but also with kinetochores. DdCP224 and C. elegans ZYG-9 were both detected at the kinetochore region in metaphase, in addition to their localization at the spindle poles (Matthews et al., 1998). Moreover, in several dis1 mutants of S. pombe, loss of functional p93^dis1 inhibits sister chromatid separation without blocking spindle elongation (Nabeshima et al., 1995, 1998). Similarly, functionally compromised DdCP224-GFP could interfere with sister chromatid separation in Dictyostelium, but the spindle poles may still be separated by the elongating spindle. After breakdown of the spindle, the chromosomes would then be associated with one pole only and the second centrosome would be set free. However, in the case of dis1 mutants, viable cells with supernumerary SPBs have not been reported, and the occurence of large elongated, dumbbell- and kidney-shaped centrosomes in DdCP224-GFP cells cannot be explained solely by a defect in chromatid separation. Furthermore, the nuclear size as a measure for ploidy of DdCP224-GFP cells with supernumerary centrosomes is usually not increased (data not shown). The Drosophila msps mutant, which displays a strongly reduced expression of the Msps protein, exhibits chromosome segregation defects as well (Cullen et al., 1999). However, in this case chromosome segregation seems to be disturbed by a disruption of the mitotic spindle. This is characterized by the appearance of one or more small additional bipolar or monopolar spindles, so-called mini spindles. Frequently, spindles are totally disorganized and contain no more distinct spindle poles at all. Thus, Cullen et al. (1999) concluded that one of the functions of the Msps protein might be an involvement in microtubule bundling which holds the mitotic spindle together. Supernumerary centrosomes as in DdCP224-GFP cells were not described and there is so far no evidence for the presence of any centrosomal marker protein at the mini-spindle poles of mutant Drosophila embroys. By contrast, three lines of evidence indicate that supernumerary centrosomes of the DdCP224-GFP mutant closely resemble normal centrosomes. First, they contain DdCP224, γ-tubulin and the 350-kDa antigen (Kalt and Schliwa, 1996); second, confocal microscopy suggests that they possess a normal corona; and third, they are capable of microtubule organization in interphase. Interestingly, the phenotype of the msps mutant is quite reminiscent of the C. elegans zyg-9 mutant embyos showing disorganized spindles and numerous cytoplasmic clusters of short microtubules during meiosis (Kemphues et al., 1986). Taken together, within this protein family, DdCP224-GFP is the first mutant exhibiting supernumerary centrosomes and no significant defects in spindle integrity and mitotic progression which are characteristic for all functionally defective mutants of DdCP224-related proteins in the other species.

The role of DdCP224 in cytokinesis is more difficult to explain. There is no evidence for an association of DdCP224 with proteins directly involved in cytokinesis. However, in late telophase, the spindle midbody to which DdCP224-GFP is localized comes in close contact with the cytokinetic constriction zone. This might allow DdCP224 to fulfill a late function in cytokinesis. ΔC-GFP cells have a similar cytokinesis defect but in these cells no GFP labeling at any structure involved in mitosis is observed. Thus the role of DdCP224 in cytokinesis may be quite indirect.

This work shows DdCP224 to be a multifacetted and multifunctional molecule involved in serveral aspects of microtubule and centrosome biology. The roles of the members of the interesting protein family to which it belongs may have changed in the course of evolution, resulting in a highly adaptive and therefore relatively poorly conserved molecule with nevertheless essential roles in centrosome function and mitosis.

We thank Nicole Brusis for expert technical assistance and Timo Zimmermann for his help at the confocal microscope, Ursula Euteneuer for helpful discussion and John V. Kilmartin for generously providing us with the YL1/2 antibody. This work was supported by the Deutsche Forschungsgemeinschaft (SFB184) and the Fonds der Chemischen Industrie.

Montross, W. T., Ji, H. and McCrea, P. D. (2000). A -catenin/engrailed chimera selectively suppresses Wnt signaling. J. Cell Sci. 113, 1759-1770.

The authors apologise for an error that occurred in Fig. 1A and its legend.

In Fig. 1A, the chimeric -Engrailed construct should not have been depicted or described as metabolically stabilized (-catenin’s N-terminal domain did not contain stabilizing point mutations). This does not alter the results presented, the discussion or the conclusions. If a stabilized form of -Engrailed were generated and compared with the non-stabilized form employed in this study, it would be expected to display a greater potency in blocking canonical Wnt signals at equivalent injected doses.

The error appeared in both the print and the pdf versions of this article. The correct Fig. 1 is shown below.