The different components of a multisubunit cell number-counting factor have both unique and overlapping functions

Little is known about the regulation of tissue size, despite it being a significant problem in both agriculture and medicine (Day and Lawrence, 2000; Gomer, 2001; Haldane, 1928; Potter and Xu, 2001). In the model system Dictyostelium discoideum, starving cells use relayed pulses of cAMP as a chemoattractant to aggregate in dendritic streams. On agar plates the streams of our laboratory strains break up into groups of a remarkably uniform size of 2×10⁴ cells/group. Other strains similarly form even sized groups of up to 10⁵ cells (Shaffer, 1957). Each group then develops into a fruiting body consisting of a 1-2 mm tall stalk supporting a mass of spore cells. The spores can then be dispersed by the wind to start new colonies of cells. The group formation appears to function to limit the size of fruiting bodies, since fruiting bodies with more than these numbers of cells tend to collapse or fall over (Brock and Gomer, 1999).

Using shotgun antisense transformation for mutagenesis, we found a Dictyostelium transformant we named smlAas that formed very small groups and fruiting bodies (Spann et al., 1996). The smlAas cells formed streams with normal morphology and size which then broke up into a large number of small groups. Disrupting the smlA gene using homologous recombination resulted in cells with a phenotype identical to that of smlAas (Spann et al., 1996). SmlA is a cytosolic protein with no strong similarity to any known protein. It appears to regulate some aspect of secretion, as the smlA^– and smlAas cells secrete a soluble factor which when added to early developing wild-type cells causes them to form small groups (Brock et al., 1996).

One way in which cells can theoretically sense the size of a tissue or group of cells is if all the cells in the group secrete a diffusible factor. As the number of cells in the group increases, the concentration of the factor will increase, and thus the cells themselves or a different set of cells can sense the size of the group (Brock et al., 1996; Clarke and Gomer, 1995; Gomer, 2001). We hypothesized that the factor secreted by the smlA^– cells might be an example of such a factor. Using conventional protein chromatography and the ability of this factor to cause wild-type cells to form small groups as a bioassay, we partially purified the factor and found that it appeared to be a ∼450 kDa complex of polypeptides we named counting factor (CF). We found that CF was also secreted by wild-type cells, although to a lesser extent than by smlA^– cells. This suggested that the smlA^– cells oversecrete a factor that the wild-type cells use to sense the number of cells in a group, so that when there is a small number of smlA^– cells in a group, the higher concentration of CF causes these cells to sense that there is a much larger number of cells (Brock and Gomer, 1999).

A SDS-polyacrylamide gel of CF showed that there are at least 5 different polypeptides in the complex (Brock and Gomer, 1999). We obtained N-terminal amino acid sequence for 3 of the proteins; another 2 were blocked. We isolated the cDNA encoding a 40 kDa component, which we named countin. The derived amino acid sequence of the countin protein has some similarity to amoebapores, a superfamily of proteins that bind to membranes (Zhai and Saier, 2000). A prediction of the ‘secreted factor being used to count cells’ model is that if the cells secrete little or no factor, huge groups will form. To test the hypothesis that countin is part of such a factor, we disrupted the countin gene using homologous recombination. The countin^– cells did not secrete any measurable CF activity according to the bioassay, and the streams of countin^– cells did not break into groups, resulting in the streams coalescing into huge mounds of cells which then formed huge fruiting bodies (Brock and Gomer, 1999). Addition of anti-countin antibodies to developing wild-type cells caused them to form large groups, again indicating that countin is used by wild-type cells for group size regulation. A secreted protein with ∼40% identity to countin, countin2, also regulates group size (Okuwa et al., 2001).

To try to understand what could cause a stream of cells to break up into groups, we wrote a simple computer simulation of cells moving in a stream while secreting and sensing a diffusible factor (Roisin-Bouffay et al., 2000). This simulation indicated that if a secreted factor negatively regulates cell-cell adhesion, and/or increases random motility, the presence of too many cells in a stream results in a high level of the factor and thereby cause the stream to dissipate and subsequently break up. The subsequent lower concentrations of the factor amongst the dispersed cells will cause them to recoalesce, and the simulation indicated that they would coalesce into separate groups rather than in a continuous stream (Roisin-Bouffay et al., 2000). During early Dictyostelium development, there are two main cell-cell adhesion molecules: gp24 and gp80 (Loomis, 1988; Siu et al., 1987; Siu and Kamboj, 1990). Overexpression of gp80 caused the formation of unbroken streams and large aggregates, while blocking gp80 binding activity with monoclonal antibodies resulted in broken streams and many small aggregates (Kamboj et al., 1990; Siu and Kamboj, 1990). Because altering cell-cell adhesion can alter group size in Dictyostelium, we examined whether countin^– or smlA^– cells have altered cell-cell adhesion. We found that countin^– cells have abnormally high cell-cell adhesion, while smlA^– cells have low adhesion (Roisin-Bouffay et al., 2000). Addition of highly purified preparations of CF to wild-type cells decreased cell-cell adhesion within 30 minutes. Decreasing adhesion with exogenous anti-gp24 antibodies also decreased group size. The simulations indicated that if a secreted factor regulates both adhesion and motility, the group size becomes more uniform (Tang et al., 2002). We found that CF increases cell motility by increasing actin polymerization and the levels of the actin-crosslinking protein ABP-120, and by decreasing levels of myosin polymerization (Tang et al., 2002). Motility and the expression of the adhesion molecules is regulated by the relayed pulses of cAMP, and we found that CF indeed regulates cAMP signal transduction (Tang et al., 2001). These results suggested that CF regulates cell-cell adhesion and motility, and this in turn can regulate the size of a group of cells.

All of the above work was done with either countin^– cells or exogenous anti-countin antibodies to inhibit countin function. However, CF contains several other proteins in addition to countin. To help understand why cell-number counting in Dictyostelium is mediated by a large complex of polypeptides, we have characterized the function of CF50, another protein component of CF.

To isolate fragments of cDNA, PCR was done with primers that matched the genomic sequence from the Dictyostelium sequencing project, using vegetative and developmental cDNA libraries as templates. PCR was also used to obtain fragments of genomic DNA, using Ax4 genomic DNA as a template. The DNA was ligated into pCR 2.1 (Invitrogen, San Diego, CA) and sequenced at Lonestar Labs (Houston, TX). Possible O-glycosylation sites were searched for using the DictyOGlyc 1.1 server http://www.cbs.dtu.dk/services/DictyOGlyc/ (Gupta et al., 1999). To generate a gene disruption construct, PCR was done on Ax4 genomic DNA with the primers CGATAATCATCCGCGGGAGGGATGTATCCTTAGC and GGTAAAGTCACCAGGGACTTGTCTAGAGCATGC, yielding a 437 bp fragment of the 5′ side of a gene designated ctnB that will be referred to as cf50. This was digested with Sac II and Xba I, and ligated into the same sites in pBluescript SK+ (Stratagene, La Jolla, CA) to generate p50L. PCR was then done with GCCAATGTAAGCTTGGCAGCAGAACAAGCTGG and CATGTGATATACACGAATTACGGGCCCAATGCG on Ax4 genomic DNA to generate a 771 bp fragment of the 3′ side of cf50. This was digested with HindIII and ApaI and ligated into the same sites in p50L to generate p50LR. The 1.4 kb XbaI-HindIII blasticidin resistance cassette from pUCBsrΔBam (Adachi et al., 1994; Sutoh, 1993) was then ligated into the same sites in p50LR to generate p50KO. This was digested with SacII and ApaI and the insert was purified by gel electrophoresis and a Geneclean II kit (Qbiogene, Inc., Carlsbad, CA). This was used to transform Dictyostelium Ax2 cells following the method of Shaulsky et al. (Shaulsky et al., 1996). PCR was used to identify cf50^– clones with a disruption of cf50. The specific clone used in this study is HDB17-4 and is referred to as cf50^–.

Cell culture was done according to Brock and Gomer (Brock and Gomer, 1999). The disruption of strain ctnA (also known as countin) used in this study was HDB2B/4 and will be referred to as countin^–. The smlA disruption strain used was HDB7YA and will be referred to as smlA^–. Slices of agar with adhering fruiting bodies were turned sideways and photographed on Kodak Ektachrome 160T with a light blue filter using a Nikon Microphot with a 4× lens stopped down with a 1 mm pinhole in front of the lens. Synergy assays were done according to Brock et al. (Brock et al., 1996). Production of conditioned starvation medium (CM), exposure of cells to conditioned medium, and exposure of cells to 1:500 anti-countin antibodies or 1:500 anti-CF50 antibodies was done according to Brock and Gomer (Brock and Gomer, 1999). Group size assays were done following the method of Brock et al. (Brock et al., 1996). To make conditioned growth medium, cells were inoculated into growth medium at 1×10⁶ cells/ml. After 20 hours, the medium was clarified by centrifugation at 2100 g for 3 minutes, and the supernatant was further clarified by centrifugation at 12,000 g for 10 minutes. The supernatant was mixed with SDS sample buffer and heated to 100°C for 3 minutes for western blots stained with anti-CF50. For western blots stained with anti-countin antibodies, the supernatant was concentrated by a factor of 10 using an Ultrafree Biomax10 kDa cutoff spin filter (Millipore, Bedford, MA) and then treated as above. The procedure of Wood et al. (Wood et al., 1996) was used to determine the percentages of CP2-positive and SP70-positive cells at low cell density. For each cell type, cells were grown in shaking culture, and (where indicated) recombinant CF50 was added to the cells to 0.2 μg/ml at the time indicated. The cells were then collected by centrifugation, washed in starvation buffer and starved in duplicate wells of an 8 well slide at low cell density in monolayer culture in the presence of a 1:10 dilution of Ax4 conditioned medium to supply CMF and thus allow CP2 and SP70 expression at low cell density. After 6 hours, cAMP was added to induce the expression of prestalk and prespore genes. 12 hours later, the cells were fixed and one well was stained for CP2 while the other was stained for SP70. The total number of cells per well and the number of positive cells were then counted. To examine differentiation in aggregates, cells were starved on filter pads (Brock et al., 1996) and were dissociated after 18 hours. Approximately 2500 cells in 200 μl of PBM were placed in the well of an 8 well slide. After allowing the cells to settle for 10 minutes, the cells were fixed and stained following Wood et al. (Wood et al., 1996).

RNA was isolated using a RNeasy mini kit (Qiagen, Valencia, CA) following the manufacturer’s directions with the exception that 2×10⁷ cells were used. Northern blot analysis was as described previously (Brock et al., 1996).

To make recombinant CF50, a PCR reaction was done using a vegetative cDNA library and the primers GGCAGCCATATGGAATGTGCCATTGATTTCTC and GGATCCTCGAGTTAAATGGATGATCCACTTCC to generate a fragment of the cf50 coding region corresponding to the entire polypeptide of the secreted protein, with a NdeI site on one end and a XhoI site on the other. To make a recombinant fragment of CF50 lacking the C-terminal serine-glycine rich region (which is very similar to a region in a 45 kDa component of CF; D. A. B., R. D. H. and R. H. G., unpublished), a PCR reaction was carried out using the first primer above and GCCGGATCCTCGAGTTAATAAAAATCTAAATCAACAC. After digestion with NdeI and XhoI, these were ligated into the corresponding sites of pET-15b (Novagen). The recombinant proteins were expressed in bacteria. The recombinant CF50 was found in inclusion bodies, so these were solubilized in 8 M urea and the recombinant CF50 was purified using a B-Per 6x His spin purification kit (Pierce, Rockford, IL) following the manufacturer’s directions. To refold denatured recombinant CF50, a final concentration of 100 μg/ml protein was dialyzed in 20 mM Tris-HCl, pH 7.5, 0.1 mM DTT at 4°C for 24 hours with three changes of buffer. The protein was then dialyzed in the same buffer without DTT at 4°C for 48 hours with four changes of buffer, and stored as aliquots at –80°C. The purified protein was quantitated in two ways. First, dilution curves of known amounts of bovine serum albumin and various dilutions of recombinant CF50 were electrophoresed side-by-side on SDS-polyacrylamide gels and stained with Coomassie Blue, and the lanes were scanned. Second, a Bio-Rad protein assay (Bio-Rad, Hercules, CA) was performed. The two methods gave similar results. Both constructs generated proteins that had poly-histidine tags at the N terminus, a thrombin cleavage site, followed by the first amino acid of the secreted form of CF50 (Fig. 1, first vertical arrow). The first construct generated a protein that terminated after the last amino acid of the predicted sequence. The second construct generated a protein that terminated after the last amino acid before the serine/glycine-rich motif (Fig. 1, second vertical arrow). Proteins were assayed for lysozyme activity using the method of Monchois et al. (Monchois et al., 2001) using hen’s egg white lysozyme (Sigma) as a standard.

Circular dichroism measurements were performed using a model 62ADS spectrometer (Aviv, Lakewood, NJ) with a 2 mm × 10 mm dual pathlength quartz cuvette (Starna Cells, Atascadero, CA) at 24°C. Data was collected from 260 to 190 nm at 1 nm intervals, 1 nm bandwidth, and a 1 second integration time. For these measurements, recombinant CF50 was diluted to a final concentration of 20 μg/ml in PBM.

Sedimentation equilibrium experiments were performed on a Beckman XL-A (Beckman, Palo Alto, CA) analytical ultracentrifuge with a four-position An60Ti rotor and double-sector centerpieces at 25°C. 1.0, 0.5, and 0.25 mg/ml of recombinant CF50 were examined in a six-channel centerpiece unit in which three channels on one side contained the different concentrations of protein and the three channels on the other side contained buffer. Samples were centrifuged at 10,000 rpm, 14,000 rpm, or 18,000 rpm. Analysis of data was accomplished using software provided by Beckman Instruments.

For antibody production, the recombinant truncated CF50 was additionally purified by SDS-polyacrylamide gel electrophoresis (Gomer, 1987). Immunization of a rabbit with the recombinant truncated CF50 and affinity purification of the antibodies was done by Bethyl Laboratories (Montgomery, TX). For western blots, 10⁶ cells in 80 μl, or 80 μl of CM, was mixed with 20 μl of 5× Laemmli sample buffer and heated to 100°C for 3 minutes. 15 μl was electrophoresed on a 12% polyacrylamide Tris-HCl gel (Bio-Rad, Hercules, CA). Protein was transferred to Immobilon P and stained as described for the ubiquitin western blots by Lindsey et al. (Lindsey et al., 1998) with the exception that filters were not boiled, were blocked overnight at 4°C in PBS/5% BSA/1% Tween-20/1% Nonidet P-40/0.1% SDS, and the primary antibody was used at a 1:3000 dilution. Staining of western blots with anti-countin antibodies and size-exclusion gel chromatography was performed as described in Brock and Gomer (Brock and Gomer, 1999). For immunofluorescence, 200 μl of cells at a density of 5×10⁵ cells/ml were placed in the well of a Lab-Tek eight-well slide (Nalge, Naperville, IL). After 30 minutes, the liquid was gently removed from the settled cells and replaced with 2% formaldehyde, 0.2% glutaraldehyde, 0.002% Triton X-100 in PBM. After 7 minutes, this was removed and the cells were washed with two changes of PBM and then incubated for 15 minutes in several changes of 4 mg/ml sodium borohydride. The cells were then stained with 10 μg/ml of affinity-purified anti-CF50 antibody (Gomer, 1987). Cells were examined with a Nikon Microphot Fx with a 1.4 NA 60× lens or a Deltavision (Applied Precision, Issaquah, WA) deconvolution microscope with a Zeiss 1.4 NA 100× lens. Sieving gel chromatography was done as described by Brock and Gomer (Brock and Gomer, 1999).

Adhesion was measured (Roisin-Bouffay et al., 2000) using cells developing on filter pads and the cells were then dissociated before doing the assay following (Tang et al., 2002). To measure motility, cells were starved in PBM as described above, diluted to 3×10⁵ cells/ml in PBM and 400 μl was placed in the well of a Lab-Tek 155411 8 well chambered coverglass (Nalge, Naperville, IL). After 30 minutes, 250 μl of liquid above the cells was removed from the well. Cells were videotaped for 10 minutes, 5 to 6 hours after being placed in the well, using the method of Yuen et al. (Yuen et al., 1995) with the exception that a 20× objective was used.

Dictyostelium cells use CF to sense the number of cells in an aggregation stream (Brock et al., 1996). We previously found that the factor is a complex of polypeptides, and that disruption of countin, the gene encoding the 40 kDa component of CF, essentially abolishes the activity of the factor. To elucidate why CF contains multiple polypeptides, we examined a second component of the complex. We previously determined the amino-terminal sequence of CF50, the 50 kDa component of the semi-purified factor (Brock and Gomer, 1999). A search of the Dictyostelium genomic sequence database revealed an open reading frame encoding a methionine followed by a predicted secretion signal, then an exact match to the sequence of the N terminus of purified CF50 (underline, Fig. 1), followed by an open reading frame, with an AATAAA polyadenylation signal 4 nucleotides after the stop codon (double underline, Fig. 1). Additional AATAAA sequences were observed beginning at 27, 33 and 37 nucleotides (nt) past the stop codon. Using PCR primers matching portions of the genomic sequence, cf50 cDNAs were isolated from a vegetative and a developmental cDNA library. The sequence of both cDNAs matched the sequence of the genomic DNA between nucleotides 87 and 743. A search of Dictyostelium cDNA databases revealed sequences with exact matches to nucleotides 28 to 268, and 390 to 83 nt past the stop codon shown in Fig. 1. These cDNAs showed a poly(A) sequence beginning at 84 nt past the stop codon.

The first 24 amino acids of the predicted CF50 protein appear to be a signal sequence (typically a lysine followed by a hydrophobic region), and are missing from the purified native protein (Fig. 1). This suggests that CF50 is a secreted protein. The predicted molecular mass of the 279 amino acid protein starting from the observed N terminus is 28.2 kDa. There are two potential N-linked glycosylation sites (shaded boxes, Fig. 1) and no significant O-linked glycosylation sites. We previously observed that other proteins secreted by Dictyostelium cells are extensively glycosylated, resulting in polypeptide backbones that are ∼70% of the total mass of the protein (Brock and Gomer, 1999; Jain and Gomer, 1994). The first 200 amino acids of CF50 show a 34% identity and a 51% similarity to the Entamoeba histolytica lysozyme II precursor, and the last 70 amino acids contain 13 copies of (A/N)SGS motifs. There are no concentrations of charged amino acids, only 3 positively charged amino acids, and the predicted pI is 3.5. There are no large regions of hydrophobicity. Recombinant CF50 (which as described below has CF bioactivity) appeared to contain a mixture of α helix, β sheet and random coil as determined by circular dichroism (data not shown), and ultracentrifugation indicated a molecular mass of 35 kDa, suggesting that the recombinant protein behaves as a monomer. Lysozyme activity assays indicated that recombinant CF50 has more than 10⁴-fold lower specific enzymatic activity than hen’s egg-white lysozyme (Table 1), suggesting that despite the sequence similarity, CF50 is not a lysozyme.

To help understand the function of CF50, we examined the effect of disrupting its expression. A northern blot showed the presence of a 0.9 kb cf50 mRNA in Ax2 vegetative cells (Fig. 2A). This RNA was absent in cf50^– cells, suggesting that these cells lack the cf50 mRNA. Western blots of total cell protein from vegetative or 6-hour starved cells showed a 50 kDa band staining with anti-CF50 in wild-type, smlA^– and countin^– cells; this band was absent from cf50^– cells (Fig. 2B and data not shown). Compared to parental cells, there appeared to be more CF50 protein in countin^– and smlA^– cells. Western blots of conditioned starvation medium indicated that CF50 is secreted by developing wild-type, smlA^– and countin^– cells, but not from cf50^– cells (Fig. 2C). Compared to the amount of CF50 in wild-type CM, there appeared to be higher levels of CF50 in the CM from both countin^– and smlA^– cells. As previously described, there was a higher levels of countin protein in smlA^– CM than in wild type CM, and no countin in countin^– CM (Brock and Gomer, 1999). We observed variable but low levels of countin protein in cf50^– CM, with bands at ∼35 and ∼30 kDa reacting with the anti-countin antibodies (Fig. 2D and data not shown). Sequencing of tryptic peptides from bands purified from crude preparations of CF indicate that there can exist degradation products of countin at these molecular masses. In support of the idea that countin is degraded in cf50^– CM, we observed that cf50^– cells occasionally revert to a phenotype very similar to wild type. Each time this happened, we observed that concomitantly the countin protein in the CM was no longer being degraded. Western blots of conditioned growth medium indicated that wild-type, smlA^– and countin^– cells but not cf50^– cells secrete CF50, with higher levels in growth media conditioned by countin^– and smlA^– cells (Fig. 2E). When similar samples were stained for countin, this protein was observed in growth medium from wild-type, smlA^–, and cf50^– cells but not countin^– cells (Fig. 2F). Compared to media from wild-type cells, growth media conditioned by cf50^– cells contained somewhat more countin, while media from smlA^– cells contained very high levels. The above results suggest that cf50^– cells lack the cf50 mRNA and the CF50 protein, that vegetative as well as developing wild-type and smlA^– cells secrete countin and CF50, and that countin is degraded in cf50^– CM.

The cf50^– cells grew normally on plates with bacteria as well as in shaking culture. When the cells starved, they formed normal aggregation ‘spiders’ which either coalesced without breaking up, or broke up infrequently, both resulting in the formation of very large groups and fruiting bodies (Fig. 3). The large groups and fruiting bodies were observed when the cf50^– cells were allowed to develop on filter pads, on water agar, on SM/5 plates with or without bacteria, and a variety of other conditions (data not shown). Although cf50^– and countin^– cells share a similar phenotype of abnormally large aggregates and fruiting bodies, the developing structures of cf50^– are much more aberrant. Along with irregularly-shaped, large aggregates, extremely long, very extended slug or finger-like structures can be seen which even appear to fruit horizontally (data not shown). Together, the data suggest that like countin^– cells, cf50^– cells form large groups and fruiting bodies, but that cf50^– cells have additional defects such as extended structures.

To determine if the phenotype of the cf50^– cells is due to a defect in a signal transduction pathway or is instead due to the lack of a secreted factor, we first performed synergy assays. As we previously observed (Brock et al., 1996), when wild-type cells were mixed with 10% smlA^– cells, they formed 1.7±0.2-fold more groups than wild-type cells alone. When cf50^– cells were mixed with 10% smlA^– cells, they formed 2.9±1.1-fold more groups than cf50^– cells alone. This suggested that the cf50^– cells were able to change their group size in response to an extracellular signal provided by the smlA^– cells. To determine if this signal is soluble, cf50^– cells were starved in the presence of buffer or various conditioned media. As previously observed, smlA^– conditioned medium (CM) caused Ax2 cells to form a larger number of smaller groups (Brock et al., 1996). smlA^– CM also caused cf50^– cells to form larger number of smaller groups, suggesting that the factor is indeed soluble (Fig. 4). We also measured the CF activity in conditioned starvation medium from cf50^– cells. In a low cell density assay, wild-type cells formed 52±10 groups in the presence of PBM, 228±8 groups in the presence of smlA^– CM, 73±4 groups in the presence of cf50^– CM, and 75±1 groups in the presence of countin^– CM. This suggests that smlA^– CM causes wild-type cells to form a large number of groups compared to buffer alone, whereas countin^– or cf50^– CM cause a much smaller increase in group number. Together, the data suggest that cf50^– cells can respond to a secreted factor, which regulates group size, and that decreased levels of a secreted factor causes the large group size in cf50^– cells.

A simple explanation for the above results is that CF50 is a necessary or at least important component of CF. To test this hypothesis, recombinant CF50 was added to starving cells. A dilution curve of recombinant CF50 showed that for both parental Ax2 and cf50^– cells, the optimal concentration of recombinant CF50 for increasing the number of groups was ∼200 ng/ml. At this concentration the cf50^– cells formed fruiting bodies that were indistinguishable in size and morphology from untreated wild-type cells (Fig. 5 and data not shown). Under these conditions 0.2 μg/ml recombinant CF50 caused a 46±9% (mean ± s.e.m., n=3) increase in group number for Ax2 cells, an 89±18% increase in group number for cf50^– cells, and a 24±6% increase for countin^– cells (Fig. 5). When recombinant CF50 was added to cf50^– cells starving in shaking culture, the countin levels in the CM increased and there were less countin degradation products. Together, the data indicate that addition of recombinant CF50 rescues the defects observed in cf50^– cells.

In the group number assay shown in Fig. 5, 30 μl of cells at 5×10⁶ cells/ml were spotted onto a filter. For untreated wild-type cells there were ∼150 groups (Fig. 5) and thus ∼1000 cells/ group. This number is much smaller than the number of cells per group when the cells are on agar, and we attribute this partially to CF and other factors being trapped by the filter. When 1 μl of either Ax2 or cf50^– cells at 1×10⁷ cells/ml were spotted onto filters, both cell types formed 1 group per spot, and thus roughly 10⁴ cells/group. Since the small spots have a greater ratio of periphery to surface area, we hypothesize that under these conditions there is less trapping of secreted factors and group sizes approach those seen on agar.

We previously observed that when wild-type cells were starved on filters soaked with anti-countin antibodies, the number of groups decreased and the group size increased (Brock and Gomer, 1999). To determine the effect of depleting countin and CF50 simultaneously, we starved Ax2 parental and cf50^– cells on filters for 2 hours and then transferred the filters with cells to pads soaked with buffer, preimmune or immune anti-countin antibodies. There was no significant difference in the number of aggregates formed on buffer as opposed to preimmune sera. Anti-countin antibodies decreased the number of groups formed by cf50^– cells, and anti-CF50 antibodies decreased the number of groups formed by wild-type cells but not cf50^– cells (Table 2). The anti-CF50 antibodies also caused a slight reduction in the number of groups formed by countin^– cells. The data suggest that there is some residual CF activity secreted by cf50^– cells, and that depleting extracellular countin can decrease this activity. The data also indicate that anti-CF50 antibodies are able to deplete CF activity from the extracellular medium, and that countin^– cells secrete a small amount of CF activity that can be neutralized with anti-CF50 antibodies.

To elucidate the developmental regulation of cf50 expression, a northern blot of RNA from different developmental stages was probed with a section of the cf50 gene. As shown in Fig. 6A, cf50 mRNA is present in vegetative cells and throughout development, with somewhat higher levels from 2 to 8 hours of development. A western blot of protein from cells at various stages of development showed that CF50 was present in cells at all stages of development, with a slight increase in amount between 5 and 7.5 hours of development (Fig. 6B). Interestingly, in vegetative cells and up to 7.5 hours, there was a 53 kDa band that stained with anti-CF50 (Fig. 6B). Together, the data suggest that CF50 is present throughout development and that like the secreted signal CMF (Jain et al., 1997; Yuen et al., 1991), its synthesis involves a slightly larger precursor form.

Immunofluorescence with anti-CF50 antibodies showed a punctate distribution in wild-type, countin^– and smlA^– cells during both growth and development (Fig. 7 and data not shown). The punctate staining was absent during growth and development in cf50^– cells (Fig. 7 and data not shown). There was somewhat brighter staining with anti-CF50 antibodies of the smlA^– and countin^– cells compared to wild type, similar to the higher levels of CF50 in these transformants observed on western blots (Fig. 2). The punctate distribution of CF50 is similar to that observed previously when cells were stained with anti-countin antibodies (Brock and Gomer, 1999), and suggests that before it is secreted, CF50 is located in vesicles.

We previously identified CF50 as a protein present in the CF preparation. To determine whether all or just some of the secreted CF50 is in the 450 kDa CF complex, CM was size-fractionated by gel chromatography, and the fractions were stained for CF50 by western blotting. As shown in Fig. 8, CF50 protein from Ax4 CM appears in a single peak around 450 kDa, in the same fractions where the countin peak appears (Brock and Gomer, 1999). Similar chromatography of CM from smlA^– cells also shows a peak at ∼450 kDa, and in addition a band at 30 kDa that is present in both the 450 kDa fraction as well as a fraction representing material with molecular mass less than 66 kDa. Very long exposures did not reveal any of the 30 kDa protein in the sieve column fractions of Ax4 CM, suggesting that the 30 kDa protein that binds the anti-CF50 antibodies is specific to the smlA^– CM. These results suggest that in smlA^– CM some of the CF50 is broken down.

Although both countin and CF50 are found in the fraction with CF activity after purification through several columns and a non-denaturing gel, there is a possibility that the two proteins are in different 450 kDa complexes. To test the hypothesis that the two proteins do not co-associate, we determined whether deletion of CF50 would affect the apparent molecular weight of the complex containing countin, because if the two proteins are in different complexes one would expect that deletion of CF50 would not affect the size of the complex containing countin. As shown in Fig. 8B, the countin in wild-type CM elutes at approximately 450 kDa, whereas the countin in cf50^– CM elutes at a lower molecular weight. As mentioned before, there appears to be increased degradation of countin in the cf50^– CM. The observation that CF50 causes the apparent molecular weight of the complex containing countin to decrease disproves the hypothesis that CF50 and countin are in different complexes, thus suggesting that countin and CF50 are part of the same complex.

Computer simulations predicted that a higher cell-cell adhesion and/or a lower random cell motility would keep a stream intact, and thus by preventing stream breakup lead to larger groups (Roisin-Bouffay et al., 2000). We previously observed that countin^– cells have a higher cell-cell adhesion than their parental cells. Compared to wild-type parental, cf50^– cells had 5.9±1.5% higher adhesion at 2 hours of development, 3.0±1.3% higher at 4 hours, and 2.2±0.9% higher at 6 hours (means ± s.e.m. from 5 separate experiments). These experiments were done on cells that were forming streams on filter pads. The streams were morphologically similar up to about 8 hours of development, and had not begun to coalesce. Thus whereas during later development the cf50^– cells form huge groups and cells in the interior might be starved for oxygen, in these assays the cells were in structures of a similar size. In addition to having abnormally high cell-cell adhesion, countin^– cells have an abnormally low cell motility during the time wild-type streams are breaking up (Tang et al., 2002). cf50^– cells also have significantly decreased cell motility at 6 hours after starvation (Fig. 9). These assays were done with cells developing in submerged monolayer culture, so, as above, these differences are not due to factors such as a lack of oxygen. Together, the data indicate that disruption of cf50 has qualitatively the same effect on cell-cell adhesion and cell motility as disruption of countin.

We previously observed that countin^– cells show a normal initial differentiation into CP2-positive prestalk and SP70-positive prespore cells (Brock and Gomer, 1999). CP2 is a protein which is a marker for an initial population of prestalk cells that gives rise to the first set of cells expressing other markers such as ecmA (Clay et al., 1995; Gomer et al., 1986), while SP70 is expressed in a subset of prespore cells and later appears on spore coats (Gomer et al., 1986). A similar assay showed that an abnormally low percentage of cf50^– cells initially differentiate into CP2-positive cells while, compared to parental, a higher percent differentiate into SP70-positive cells (Table 3 and data not shown). These assays were done with cells developing at very low cell density in submerged monolayer culture, so these differences are not due to differences in group size. The initial decision to differentiate into a CP2-positive, SP70-positive or null cell is made at the time cells starve (Gomer and Firtel, 1987). Because we observed that wild-type cells secrete CF50 into the growth medium, we determined if the abnormal initial cell-type differentiation of cf50^– cells is due to a lack of extracellular CF50 just before the cells starve. Cells growing in shaking culture were treated with recombinant CF50 for various times, starved, and allowed to differentiate at low cell density. Recombinant CF50 had little effect on Ax2 or countin^– cells when added for 1, 3, or 6 hours before starvation (Table 3 and data not shown). However, a 1-hour exposure of vegetative cf50^– cells to recombinant CF50 caused a partial rescue of the abnormal differentiation, while a 6-hour exposure caused the percentages of cf50^– cells differentiating into CP2-positive and SP70-positive cells to be similar to the percentages seen with wild-type cells. Adding recombinant CF50 to growing cells for 8 hours, and then washing and starving the cells caused a 2.1±0.7 (mean ± s.e.m., n=4) increase in the number of groups formed by cf50^– cell, but no significant change in the number of groups formed by Ax2 cells. To determine if the rescue of the abnormal differentiation of cf50^– cells by exogenous CF50 was due to exogenous CF50 affecting something during vegetative growth or whether it was being sequestered by the cells and then affecting something during later development, we examined the effect of starving the CF50-treated cells in the presence of anti-CF50 antibodies. As shown in Table 4, adding anti-CF50 antibodies reversed the effect of the recombinant CF50 on the differentiation of cf50^– cells. These data suggest that during vegetative growth cf50^– cells can apparently sequester exogenous CF50 and then when starved release the CF50, which then causes them to alter the percentages of cell types that they would differentiate into if starved.

To determine if cf50^– cells developing in aggregates also have an abnormal initial cell-type differentiation, we starved cells on filter pads and then dissociated them at 18 hours, when there is maximal accumulation of CP2 and SP70. As shown in Table 5, cf50^– cells developing under conditions where they form aggregates also have an abnormal cell-type differentiation. This abnormal differentiation was observed for cells developing in spots of either 10⁴ cells in a 1 μl spot or 5 ×10⁵ cells in a 50 μl spot. When the cells were grown in the presence of recombinant CF50 and then starved, there was a slight increase in the percentage of CP2-positive cells, albeit much lower than the increase seen when these cells were starved at low cell density. The data thus suggest that cf50^– cells developing in aggregates have an abnormal cell-type differentiation.

One of the ways cells could theoretically sense the number of cells in a group or tissue is by secreting and simultaneously sensing a characteristic diffusible factor (Gomer, 2001). We identified such a factor in Dictyostelium, and found that it is a relatively large complex of polypepetides (Brock and Gomer, 1999). We have found here that disrupting the expression of cf50 has essentially the same effect as disruption of countin with respect to group size, adhesion and motility, but that unlike the effect of disrupting countin, disrupting cf50 affects the initial cell-type choice.

Native CF50 is a 50 kDa molecule, whereas the polypeptide backbone is only 28 kDa. Since many secreted proteins in Dictyostelium are glycosylated, one possibility is that the 22 kDa of non-polypeptide mass is post-translational glycosylation. Our observation that bacterially synthesized CF50 polypeptide backbone has CF activity when added to wild-type or cf50^– cells suggests that the functional domain of CF50 is in this backbone. countin^– cells did not exhibit a strong response to the recombinant CF50, possibly because they already secrete large amounts of CF50 (Fig. 2C). For a different secreted polypeptide factor, CMF, we found that the presence of post-translational glycosylation appears to protect the protein from proteases (Jain and Gomer, 1994; Yuen et al., 1991). Thus a possible explanation for the existence of the CF50 glycosylation is that it protects CF50 from degradation. We previously observed that highly purified preparations of CF could cause a 3-fold increase in group number, and that a 1.5-fold increase could occur at concentration as low as 0.3 μg/ml (Brock and Gomer, 1999). The recombinant CF50 occasionally will cause a 2-fold increase in group number, but a 1.5-fold increase tends to be the norm and the lowest concentration this occurs is at ∼0.1 μg/ml; Fig. 5 and data not shown). Assuming that the CF50 is roughly one-fifth of the CF, the specific activity of CF50 is roughly that of CF, although recombinant CF50 does not seem to increase group number to the extent that CF can.

Both countin^– and cf50^– cells produce conditioned medium that causes a slight increase in group number compared to buffer alone, suggesting that there is some residual CF activity secreted in cells lacking just countin or just CF50. The antibody depletion experiments indicated that cf50^– cells, which form large groups, will form even larger groups when countin is depleted from the extracellular environment with anti-countin antibodies, and countin^– cells similarly form slightly larger groups when CF50 is depleted with antibodies (Table 2). Thus both the CF production assays and antibody depletion experiments suggest that neither CF50 nor countin is the sole effector molecule in the CF complex, but that both molecules can separately affect group size.

During early development, countin^– cells have a considerably higher cell-cell adhesion than parental cells. We found that cf50^– cells have only a slightly higher adhesion than parental cells. This higher adhesion is probably not sufficient to strongly affect group size. However, the cf50^– cells have a significantly lower motility than parental cells, and most importantly far fewer highly motile cells than wild type (Fig. 9). Our computer simulations suggest that streams break when strongly motile cells break adhesive bonds and tear away from other cells (Roisin-Bouffay et al., 2000; Tang et al., 2002). The highly motile cells thus have the greatest influence on stream breaking, and we hypothesize that the greatly reduced numbers of these cells seen in cf50^– is the reason cf50^– cells form streams that rarely break and thus form large groups.

Vegetative countin^– cells have high cell-cell adhesion and high levels of the adhesion molecule gp24 compared to wild-type cells, while vegetative smlA^– cells have lower levels (Roisin-Bouffay et al., 2000) (Roisin-Bouffay and Gomer, unpublished). Vegetative cells contain smlA and countin proteins (Brock et al., 1996; Brock and Gomer, 1999), and we found here that vegetative cells also contain CF50. If we assume that vegetative cells can respond to CF, our observation that countin and CF50 are secreted by vegetative cells suggests that the differences in gp24 levels and cell-cell adhesion in vegetative smlA^–, wild-type and countin^– cells could be due to low CF activity being secreted by vegetative countin^– cells and high levels of CF being secreted by smlA^– cells. This could then result in different levels of gp24 and different levels of cell-cell adhesion in vegetative wild-type cells compared to smlA^– or countin^– cells.

Wild-type cells all appear to have the same amount and distribution of CF50 (Fig. 7), so the altered cell-type differentiation of cf50^– cells (Tables 3, 4, and 5) does not appear to be due to CF50 levels or distribution specifying a cell fate. This then suggests that CF50 somehow affects a process that is however heterogeneous in the cells. Although recombinant CF50 can rescue the abnormal initial cell-type differentiation when added to vegetative cf50^– cells that are subsequently starved at low cell density, anti-CF50 antibodies added during starvation can block this rescue (Table 4). In addition, when the CF50-treated cells are starved at high cell density, there is very little change in the initial cell-type differentiation (Table 5). Together, the data suggest that the recombinant CF50 added to growing cf50^– cells affects initial differentiation at a step after growth. One possibility is that the cf50^– cells are able to sequester the added CF50 and then release it during development, and that this released CF50 can rescue the altered initial cell-type differentiation of cf50^– cells when they are starved at low density but is degraded when the cells are starved at high cell density. Recombinant CF50 affects group size when added to vegetative cells that are then washed and starved at high cell density. Under these same conditions there is little effect on initial cell-type choice. A possible explanation for this is that during growth CF50 affects the expression of genes involved in adhesion and/or motility, and the altered levels of the associated proteins persist into development and affect stream breakup.

It is unclear why recombinant CF50 affects group size in Ax4 cells, which secrete CF50 as part of the CF complex. One possibility is that there are limiting amounts of CF50 in CM, and the recombinant CF50 allows CF complexes to form and or become active. However, western blots of gel filtration column fractions of whole conditioned medium from wild-type cells indicate that all of the detectable countin and CF50 are both present in a roughly 450 kDa complex, suggesting that there are no incomplete complexes. In addition, the removal of CF50 causes much of the countin present in CM to elute at a lower molecular weight (Fig. 8B). However, we cannot exclude the possibility that some fraction of countin was present in ∼450 kDa complexes that lack CF50, and that exogenous CF50 increased the activity of this subset of complexes. Another possibility is that the exogenous CF50 (and thus the CF50 in the CF complex) binds to and activates a cell surface receptor.

There are several examples of signals that are present in complexes. One example is a ternary complex of Twisted gastrulation (TSG)/Short gastrulation (SOG or chordin)/Bone morphogenetic protein (BMP) (Chang et al., 2001; Ross et al., 2001; Scott et al., 2001). TSG and SOG act synergistically to repress signaling by BMP and TSG appears to enhance a proteolytic cleavage of SOG. Insulin-like growth factors, which can regulate the size of developing tissues, are found in ternary complexes with two specific binding proteins (Boisclair et al., 2001). Another example is TGFβ, which is stored in platelets and secreted as a complex with TGFβ masking protein, a 440 kDa complex of polypeptides. TGFβ forms a homodimer, and the masking protein contains 4 to 6 subunits (Nakamura et al., 1986; Okada et al., 1989). For all three of these examples, one protein in the complex can act as a signal, and other proteins regulate its activity. Another signal that is normally in a complex is thrombospondin, a trimeric ∼450 kDa secreted extracellular matrix protein with binding sites for a variety of factors and receptors (Bornstein et al., 2000; Chen et al., 2000). Like CF, thrombospondin inhibits cell adhesion and increases cell motility, allowing cells to move and thus reshape structures (Mansfield et al., 1990; Murphy-Ullrich, 2001). CF appears to be somewhat different from these examples of multi-subunit factors insofar as having subunits that have similar as well as distinct effects on cells. This then suggests the possibility that for unknown reasons there may be different receptors for different components of the cell-number counting factor.

We thank the international Dictyostelium genome sequencing project and the Japanese cDNA sequencing project for generating the sequences that allowed us to easily determine the sequence of CF50. We thank Darrell Pilling for advice, Richard Atkinson for assistance with the deconvolution microscopy, Heather Seidle and Ron Parry for help with column chromatography, and Jeff Nichols for assistance with the circular dichroism measurements. Spectroscopic facilities utilized were provided by the Keck Center for Computational Biology and the Lucille P. Markey Charitable Trust. R. H. G. is an Investigator of the Howard Hughes Medical Institute.