Two Components of a Secreted Cell Number-counting Factor Bind to Cells and Have Opposing Effects on cAMP Signal Transduction in Dictyostelium*

A secreted 450-kDa complex of proteins called counting factor (CF) is part of a negative feedback loop that regulates the size of the groups formed by developing Dictyostelium cells. Two components of CF are countin and CF50. Both recombinant countin and recombinant CF50 decrease group size in Dictyostelium. countin- cells have a decreased cAMP-stimulated cAMP pulse, whereas recombinant countin potentiates the cAMP pulse. We find that cf50- cells have an increased cAMP pulse, whereas recombinant CF50 decreases the cAMP pulse, suggesting that countin and CF50 have opposite effects on cAMP signal transduction. In addition, countin and CF50 have opposite effects on cAMP-stimulated Erk2 activation. However, like recombinant countin, recombinant CF50 increases cell motility. We previously found that cells bind recombinant countin with a Hill coefficient of ∼2, a KH of 60 pm, and ∼53 sites/cell. We find here that cells also bind 125I-recombinant CF50, with a Hill coefficient of ∼2, a KH of ∼15 ng/ml (490 pm), and ∼56 sites/cell. Countin and CF50 require each other's presence to affect group size, but the presence of countin is not necessary for CF50 to bind to cells, and CF50 is not necessary for countin to bind to cells. Our working hypothesis is that a signal transduction pathway activated by countin binding to cells modulates a signal transduction pathway activated by CF50 binding to cells and vice versa and that these two pathways can be distinguished by their effects on cAMP signal transduction.

A secreted 450-kDa complex of proteins called counting factor (CF) is part of a negative feedback loop that regulates the size of the groups formed by developing Dictyostelium cells. Two components of CF are countin and CF50. Both recombinant countin and recombinant CF50 decrease group size in Dictyostelium. countin ؊ cells have a decreased cAMP-stimulated cAMP pulse, whereas recombinant countin potentiates the cAMP pulse. We find that cf50 ؊ cells have an increased cAMP pulse, whereas recombinant CF50 decreases the cAMP pulse, suggesting that countin and CF50 have opposite effects on cAMP signal transduction. In addition, countin and CF50 have opposite effects on cAMP-stimulated Erk2 activation. However, like recombinant countin, recombinant CF50 increases cell motility. We previously found that cells bind recombinant countin with a Hill coefficient of ϳ2, a K H of 60 pM, and ϳ53 sites/cell. We find here that cells also bind 125 I-recombinant CF50, with a Hill coefficient of ϳ2, a K H of ϳ15 ng/ml (490 pM), and ϳ56 sites/cell. Countin and CF50 require each other's presence to affect group size, but the presence of countin is not necessary for CF50 to bind to cells, and CF50 is not necessary for countin to bind to cells. Our working hypothesis is that a signal transduction pathway activated by countin binding to cells modulates a signal transduction pathway activated by CF50 binding to cells and vice versa and that these two pathways can be distinguished by their effects on cAMP signal transduction.
Little is known about how organisms regulate the size of a multicellular structure (see Refs. 1-5 for a review). A simple model for this sort of regulation is provided by the eukaryote Dictyostelium discoideum, where starving cells form fruiting bodies that in our laboratory strains contain typically 2 ϫ 10 4 cells. A fruiting body is a thin stalk supporting a mass of spores, and for optimal spore dispersal this structure needs to be as large as possible. However, since an excessively large fruiting body will collapse, Dictyostelium cells regulate the size of these structures. Dictyostelium grows as individual amoebae that eat bacteria on soil surfaces and increase in number by fission (for reviews, see Refs. 6 and 7). When the cells overgrow the bacteria and starve, they use relayed pulses of cAMP as a chemoattractant to aggregate in dendritic streams. One way in which Dictyostelium cells regulate the number of cells in a fruiting body is to have the streams break up into groups of ϳ2 ϫ 10 4 cells if there are too many cells in a stream (8 -10).
The cells appear to sense the number of cells in a stream by secreting and sensing counting factor (CF), 1 an ϳ450 kDa complex of polypeptides (11)(12)(13). When there is a large number of cells in a stream, as indicated to the cells by a high concentration of CF, the cells increase their random motility and decrease their cell-cell adhesion (14,15). This disrupts the stream, allowing breaks to form. Disrupting the genes encoding countin, CF45, or CF50, three components of CF, causes cells to secrete virtually undetectable levels of CF activity (13,16,17). The streams formed by developing countin Ϫ , cf45-1 Ϫ , or cf50 Ϫ cells do not break up, and as a result the streams coalesce into huge groups that form huge fruiting bodies that either collapse or fall over (13,16,17). Adding recombinant countin to countin Ϫ cells causes the countin Ϫ cells to form normally sized fruiting bodies, indicating that exogenous recombinant countin rescues the countin Ϫ phenotype (18). Similarly, exogenous recombinant CF50 rescues the phenotype of cf50 Ϫ cells, and exogenous CF45-1 rescues the phenotype of cf45-1 Ϫ cells (16,17). Adding either recombinant countin, recombinant CF45-1, or recombinant CF50 to developing wild-type cells causes the wild-type cells to form smaller groups (16 -18). These data suggested that CF is part of a negative feedback loop that limits group size.
The relayed pulses of cAMP that the cells chemotax toward in order to form multicellular aggregates regulate both the expression of adhesion proteins and cell motility in Dictyostelium (19 -23). When the cells sense a pulse of cAMP, they activate adenylyl cyclase to generate a relayed cAMP pulse. A rapid cGMP pulse is also generated in response to a cAMP pulse (24,25). Cells move toward the source of the cAMP pulse by phosphorylating the heavy chain of the myosin II molecules at the front of the cell, causing them to depolymerize in that region, polymerizing actin at the front to form a pseudopod, and then activating myosin at the back of the cell to effectively squeeze the cytoplasm into the pseudopod (for reviews, see Refs. 26 -28). Increasing cGMP in Dictyostelium reduces the formation of lateral pseudopodia and increases myosin II heavy chain phosphorylation and chemotaxis, whereas decreasing the cGMP pulse decreases myosin II heavy chain phosphorylation and chemotaxis (29). CF represses the cAMP-stimulated cGMP * Spectroscopic facilities were provided by the Keck Center for Computational Biology and the Lucille P. Markey Charitable Trust. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
¶ pulse (30). smlA Ϫ cells have an attenuated cGMP pulse, whereas countin Ϫ cells have an increased cGMP pulse compared with parental cells. Treatment of wild-type cells with anti-countin antibodies increases the cGMP pulse, and treatment of cells with recombinant countin decreases the cGMP pulse (18,30). The effect of CF on the cAMP-stimulated cGMP pulse requires several hours of exposure to CF. smlA Ϫ cells have a low GTP␥S-stimulated guanylyl cyclase activity, whereas countin Ϫ cells have a high activity. This suggested that CF decreases guanylyl cyclase activity. CF does not appear to regulate cGMP phosphodiesterase activity (30). In addition, CF affects group size in streamer F cells, which lack the cGMP phosphodiesterase (23). This suggested that the cGMP phosphodiesterase is not a key component of the CF signal transduction pathway. Together, the data suggested that CF inhibits the cAMP-stimulated cGMP pulse through a slow signal transduction pathway that regulates guanylyl cyclase.
In contrast to its effect on the cGMP pulse, CF appears to potentiate the cAMP-stimulated cAMP pulse (30). smlA Ϫ cells have a large 2Ј-deoxy-cAMP-stimulated cAMP pulse compared with parental cells, whereas countin Ϫ cells have a smaller pulse. A 1-min treatment of wild-type cells with either partially purified CF or recombinant countin increased the size of the cAMP pulse (18,30). Countin regulates GTP␥S-stimulated adenylyl cyclase without affecting the basal or Mn 2ϩ -stimulated activities, and this regulation can be seen with a 1-min exposure to recombinant countin (18). Together, the data suggest that CF and recombinant countin potentiates the cAMP-stimulated cAMP pulse through a fast signal transduction pathway that regulates adenylyl cyclase.
Although countin and CF50 both appear to negatively regulate group size, these two components of CF appear to have unique properties. Starved countin Ϫ cells show a normal initial differentiation into cells expressing CP2 (a marker for an initial population of prestalk cells that gives rise to the first set of cells expressing other markers such as ecmA (31,32). Starved countin Ϫ cells also show a normal initial differentiation into cells expressing SP70 (a prespore protein that later appears on spore coats (31). However, starved cf50 Ϫ cells differentiate into an abnormally low percentage of CP2-positive cells and an abnormally high percentage of SP70-positive cells, and this abnormal differentiation can be partially rescued by adding recombinant CF50 to the cf50 Ϫ cells (16).
This suggested that countin and CF50 have different effects on the initial differentiation and hinted that they may activate different signal transduction pathways. There are two key tests of the hypothesis that countin and CF50 activate different signal transduction pathways. The first is that countin and CF50 should have different effects on some aspect of signal transduction in addition to an outcome such as cell type differentiation. The second is that countin and CF50 should both bind to cells and specifically bind to cells in each other's absence (indicating that one is not inhibiting or promoting the binding of the other). In this report, we find that countin and CF50 have clearly different effects on the cAMPstimulated cAMP pulse as well as Erk2 activation, validating the first test. Countin and CF50 both bind to cells and bind in each other's absence, validating the second test. This then suggests that the two proteins activate different signal transduction pathways.

MATERIALS AND METHODS
Cell Culture, Cell Motility, and Group Number Assays-Cell culture was done following Brock et al. (13). The strains used were the wild types Ax2 and Ax4, cf50 Ϫ (clone HDB17-4) (16), and cells with a disruption of the ctnA gene (strain HDB2B/4; referred to in this and previous work as countin Ϫ cells). For motility assays, Ax4 wild-type cells growing in HL5 at a density of 2 ϫ 10 6 cells/ml were collected by centrifugation, resuspended in PBM (20 mM KH 2 PO 4 , 0.01 mM CaCl 2 , 1 mM MgCl 2 , pH 6.1, with KOH), recentrifuged, and resuspended twice, with the final resuspension at a concentration of 2 ϫ 10 5 cells/ml in PBM or PBM containing 150 ng/ml recombinant CF50. 200 l of cells were placed in the well of a 155409 8-well chambered coverglass (Nalge, Naperville, IL) for starvation. Following Tang et al. (15), cells were videotaped 6 h after being placed in the wells. Recombinant countin was prepared following Gao et al. (18), recombinant CF50 was prepared following Brock et al. (16), and recombinant CF45-1 was prepared following Brock et al. (17). Group number assays were done following Brock et al. (12).
cAMP and cGMP Assays-Starvation of cells and measurements of cGMP accumulation in response to a pulse of cAMP and cAMP accumulation in response to a pulse of 2Ј-deoxy-cAMP (a functional cAMP analog that does not interfere with the subsequent cAMP assay) were assayed as described by Tang et al. (30). Where indicated, Ax4 cells were starved as described in Tang et al. (30), and recombinant CF50 was added to 200 ng/ml 2 h after starvation. Erk Activation-Cells were starved at 1 ϫ 10 7 cells/ml in PBM in shaking culture. After 6 h, 2 ml of cells were collected by centrifugation and were resuspended in 1 ml of PM (33), and the cells were shaken continuously thereafter. One minute after resuspending the cells, 10 l of a 10 M stock solution of cAMP was added, and 100-l samples were then collected at various times after stimulation and mixed with 100 l of 2ϫ Laemmli sample buffer containing 75 mM NaF, 60 mM Na 4 P 2 O 7 , 2 mM dithiothreitol, 2 mM sodium orthovanadate (prepared following Gordon (34)), and 2ϫ Complete protease inhibitors (Roche Applied Science). The samples were heated to 100°C for 3 min and stored at Ϫ20°C. Western blotting and staining to detect phosphorylated Erk followed Schenk et al. (35) with the exception that 12% SDS-polyacrylamide gels (Bio-Rad) and Immobilon-P membranes (Millipore Corp., Bedford, MA) were used.
Iodination of Recombinant CF50 and Binding Assays-The dialysis of recombinant CF50 against PBK (6.15 mM K 2 HPO 4 , 3.85 mM KH 2 PO 4 , pH 7.0) and subsequent iodination were done as previously described for the production of 125 I-countin (18). The ability of recombinant CF50 or 125 I-recombinant CF50 to affect group size was assayed in submerged culture following Brock et al. (12). For binding studies, cells were grown as described above and starved in shaking culture in PBM at 5 ϫ 10 6 cells/ml. At 0 h (for vegetative cells) or at 6 h, cells were collected by

FIG. 3. Countin inhibits and CF50
potentiates cAMP-stimulated activation of Erk2. A, 6-h starved wild-type, countin Ϫ cells, and Ax4 wild-type cells exposed to 200 ng/ml recombinant countin for 4 h before stimulation were stimulated with cAMP, and aliquots were solubilized at the indicated times after stimulation. Western blots of the solubilized samples were stained for activated Erk2, and densitometry was used to assess staining intensity. Scans of a representative experiment are shown. B, 6-h starved wild-type, cf50 Ϫ cells and wild-type cells exposed to 200 ng/ml recombinant CF50 for 4 h before stimulation were similarly stimulated with cAMP and assayed for Erk2 stimulation. The samples shown in A were electrophoresed on the same gel, and the samples shown in B were electrophoresed on a different gel; and the blotting and staining were done separately. The differences in the wild-type values between A and B are thus due to differences in exposure of the x-ray film to the chemiluminescent stain. A and B show representative scans from three independent assays. WT, wild type. centrifugation for 5 min at 350 ϫ g; the vegetative cells but not the 6-h cells were resuspended in PBM and recentrifuged. The collected cells were resuspended to a concentration of 5 ϫ 10 7 cells/ml in B buffer (PBM containing 10 g/ml of bovine serum albumin (New England Biolabs, Beverly, MA)). Binding assays were performed by mixing a 20-l mixture containing varying amounts of 125 I-recombinant CF50 (ranging from 5.6 to 0.165 ng/reaction) in B buffer with either 50 l of the cells or 50 l of B buffer in a siliconized Eppendorf tube (Online Products for Science, Petaluma, CA). The time course experiments contained 1.65 ng of 125 I-labeled recombinant CF50 per reaction. These reactions were then gently mixed, incubated in a 21°C water bath for 10 min unless otherwise indicated, and processed exactly as in Gao et al. (18) with the exception that the sucrose cushions were made in siliconized Eppendorf tubes, and the cells were collected by centrifugation through the cushions at 12,000 ϫ g. The ability of different concentrations of unlabeled recombinant CF50 to compete for binding with 125 I-recombinant CF50 was measured using 0.56 ng of 125 I-recombinant CF50 per tube; duplicate incubations were done with cells or buffer alone, and the counts from the reactions without cells were subtracted from the counts from the reactions with cells. Binding of 125 I-recombinant countin to cells was performed as described in Gao et al. (18). Nonlinear regressions to fit binding data to a standard one-site binding model or to a model with cooperative binding and F-tests to determine whether there was cooperative binding were done with the Prism software package (GraphPad Software, San Diego, CA).
Analytical Ultracentrifugation-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. 100, 50, and 25 g/ml recombinant CF50 in 10 mM KH 2 PO 4 /K 2 HPO 4 , pH 6.8, 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, 14,000, or 18,000 rpm. Analysis of data was accomplished using software provided by Beckman Instruments.

CF50 Decreases Both the cGMP and the cAMP Pulses-Coun-
tin decreases the cAMP-stimulated cGMP pulse and increases the cAMP-stimulated cAMP pulse (30). Since countin Ϫ and cf50 Ϫ cells have some differences (16), we examined cAMP signal transduction in cf50 Ϫ cells. Compared with parental Ax2 cells, cf50 Ϫ cells have an increased cAMP-stimulated cGMP pulse (Fig. 1A), similar to that of countin Ϫ cells (30). Compared with Ax2, the Ax4 wild-type strain shows much stronger cAMPstimulated cGMP and cAMP pulses. 2 Treatment of Ax4 cells with 200 ng/ml recombinant CF50 decreased the size of the cAMP-stimulated cGMP pulse (Fig. 1B). The increased cGMP pulse in cf50 Ϫ cells and the decreased cGMP pulse in cells exposed to increased levels of CF50 together indicate that CF50, like countin, decreases the cAMP-stimulated cGMP pulse.
Stimulation of cf50 Ϫ cells with the cAMP analogue 2Ј-deoxy-cAMP caused an increased cAMP pulse compared with parental Ax2 cells ( Fig. 2A). Conversely, treatment of Ax4 cells with 200 ng/ml recombinant CF50 caused a decrease in the 2Ј-deoxy-cAMP-stimulated cAMP pulse (Fig. 2B). Together, the data indicate that CF50 decreases the cAMP-stimulated cAMP pulse. This is in contrast to countin, which increases the cAMP pulse.
Countin and CF50 Have Opposite Effects on Erk2 Activation-Inside aggregating Dictyostelium cells, the pulse of cAMP activates the Erk2 mitogen-activated protein kinase (36 -38).
To determine whether the altered cAMP-stimulated cAMP pulse sizes affect downstream targets of cAMP signal transduction, we examined the cAMP-stimulated activation of Erk2. As previously observed, a pulse of cAMP transiently activates Erk2 in 6-h starved wild-type cells, with a peak extending from 15 to 90 s after stimulation (Fig. 3A). A similar stimulation of starved countin Ϫ cells consistently resulted in a greater Erk2 activation, with strongly activated Erk2 still present 2 min after the cAMP pulse (Fig. 3A). The addition of recombinant countin to wild-type cells consistently decreased the cAMPstimulated Erk2 activation (Fig. 3A). Although the basal (t ϭ 0) 2 L. Tang and R. Gomer, unpublished observations. level of Erk2 activation in wild-type cells and cells starved with recombinant countin were variable, the basal level of Erk2 activation in countin Ϫ cells was constantly higher. Conversely, cAMP stimulation of 6-h starved cf50 Ϫ cells uniformly resulted in a relatively smaller pulse of Erk2 activation that was prolonged compared with wild-type cells (Fig. 3B). The addition of recombinant CF50 to starving wild-type cells consistently increased the cAMP-simulated activation of Erk2 (Fig. 3B). The basal level of Erk2 activation was variable in cf50 Ϫ cells and cells treated with recombinant CF50. Together, the data suggest that countin inhibits whereas CF50 potentiates cAMPstimulated Erk2 activation.
Like Recombinant Countin, Recombinant CF50 Increases Cell Motility-We previously observed that both countin Ϫ and cf50 Ϫ cells tend to move more slowly than parental cells (15,16). The addition of recombinant countin to starving wild-type cells causes an increase in cell motility (18). To determine whether recombinant CF50 also affects cell motility, we measured the motility of Ax4 cells after 6 h of starvation in the absence or presence of recombinant CF50. An increased motility was observed in cells exposed to recombinant CF50 (Fig. 4). These data indicate that, like recombinant countin, recombinant CF50 causes an increase in cell motility.
The Effect of a Combination of Countin and CF50 on Group Size Is Nonlinear-We previously found that recombinant CF50 decreases group size in wild-type cells and cf50 Ϫ cells but had little effect on countin Ϫ cells, suggesting that the effect of CF50 on group size is strongly potentiated by the presence of countin (16). Recombinant countin affects group size in wildtype and countin Ϫ cells (18). To determine whether countin similarly requires the presence of CF50, we added recombinant countin to wild-type and cf50 Ϫ cells. As previously observed, recombinant countin increased group number (Fig. 5), and we saw a corresponding decrease in group size in populations of wild-type cells. The recombinant countin, however, had no significant effect on cf50 Ϫ cells. Together, the data suggest that countin and CF50 require each other's presence to have a strong effect on group size.
Recombinant countin, recombinant CF50, and recombinant CF45-1 all decrease group size and increase cell motility (16 -18). 3 However, countin and CF50 have opposite effects on the cAMP-stimulated cAMP pulse, and altering the size of the cAMP pulse affects group size (30). This suggested the possibility that the effects of recombinant countin and recombinant CF50 might not be additive. To examine the effects of the presence of recombinant countin on recombinant CF50, wildtype cells (which secrete moderate amounts of all of the proteins present in CF) were starved in the presence of different combinations of the proteins. As previously observed, all three proteins increased the number of groups formed by starving cells. For instance, 100 ng/ml recombinant countin increased the group number by ϳ30% (Fig. 6). The addition of 100 ng/ml recombinant CF45-1 to this much recombinant countin caused the group number to increase by only 22%, whereas the addition of 100 ng/ml recombinant CF50 to 100 ng/ml recombinant countin caused the group number to increase by roughly 38%. At 30 ng/ml, both countin and CF50 increased group number by 3  more than 20% (Fig. 6). The combination of 30 ng/ml countin and 30 ng/ml of CF50, however, caused the group number to increase by only ϳ14% (Fig. 6). The difference between CF50 alone and the combination of CF50 and countin was significant, with p Ͻ 0.005. This suggests that under some conditions where countin and CF50 are both present, the proteins do not act additively or synergistically but rather might actually oppose each other.
Examining the effect of recombinant CF45 on group number indicated that, as previously observed (17), this protein increases group number. At 100 ng/ml, the addition of a 100 ng/ml concentration of either recombinant countin or recombinant CF50 decreased the activity, whereas the addition of both proteins had no effect (Fig. 6). The addition of 30 ng/ml either countin or CF50 to 100 ng/ml recombinant CF45-1 appeared to slightly potentiate the activity but was not statistically significant. Some of the strongest effects on group number were seen with 100:30:30, 30:100:30, and 30:30:100 ng/ml mixtures of recombinant CF50/CF45/countin. Together, the data suggest that the effects of recombinant countin and recombinant CF50 in the presence of recombinant CF45 also seemed to be nonlinear.
Like Recombinant Countin, Recombinant CF50 Binds to Cells-Since recombinant CF50 and recombinant countin have opposite effects on the cAMP-stimulated cAMP pulse, we investigated the possibility that there might be a separate signal transduction pathway for CF50. To determine whether cells bind CF50, we labeled recombinant CF50 with 125 I and used the 125 I-recombinant CF50 as a ligand for binding studies. After iodination, we found that the molar ratio of incorporated 125 I to recombinant CF50 was ϳ11. Fig. 7A is an autoradiogram of the iodinated protein electrophoresed on a SDS-polyacrylamide gel, which was stopped and immediately exposed to film when the dye front was roughly 1.5 cm from the bottom of the gel. On these gels, free 125 I and 125 I-tyrosine run at the dye front. The labeled recombinant CF50 appeared to migrate at its predicted molecular mass of 30.5 kDa, and there was no detectable degradation. There was very little free 125 I or 125 I-tyrosine present in the purified 125 I-recombinant CF50. To check whether the iodination reaction had an effect on the activity of recombinant CF50, proteins were assayed for CF activity by developing cells in submerged culture in the presence of different concentrations of 125 I-labeled or unlabeled recombinant CF50. As shown in Fig. 7B, 125 I-labeled recombinant CF50 was able to increase group number approximately as well as unlabeled recombinant CF50. This suggested that the 125 I-recombinant CF50 retained bioactivity.
To establish steady state conditions to carry out more complex binding assays, we examined the time course of 125 Irecombinant CF50 binding. For this, vegetative or 6-h starved wild-type and cf50 Ϫ cells were allowed to bind to a fixed concentration of 125 I-recombinant CF50. Fig. 8A shows that 125 Irecombinant CF50 bound to intact vegetative wild-type and cf50cells rapidly, with the maximal binding to cf50 Ϫ cells at 10 min. The binding to vegetative wild-type cells was slightly higher at 30 min than at 10 min, but by a small and statistically insignificant (p Ͼ 0.35) amount. The binding of 125 Irecombinant CF50 to 6-h developing wild-type and cf50 Ϫ cells showed similar kinetics (Fig. 8B), with wild-type cells apparently saturating at 10 min. However, for 6-h starved cf50 Ϫ cells, there was more binding at 30 min than at 10 min (p Ͻ 0.025).
Since the binding in Fig. 8  showed a slight sigmoidal curve, and nonlinear regression fit a model with cooperative binding significantly better than a standard binding model with no cooperative binding (p Ͻ 0.022; F-test). This statistic takes into account the fact that the cooperative binding model has more degrees of freedom (available on the World Wide Web at www.graphpad.com/curvefit/ 2_models_1_dataset.htm). The predicted K H (the equivalent of a K D when there is cooperative binding) was 11 ng/ml, the Hill coefficient was 1.8, and the maximal number of binding sites (B max ) was 0.021 ng/ml (Fig. 9A). Using a molecular mass of 30.5 kDa for the recombinant CF50, this corresponds to a K H of 3.9 ϫ 10Ϫ10 M and 41 binding sites/cell (Table I). There appeared to be less binding to vegetative cf50 Ϫ cells, and as with the wild-type cells, there was a sigmoidal binding curve that was fit significantly better with a cooperative binding model (p Ͻ 0.32; F-test). The K H was 19 ng/ml, the Hill coefficient was 2.1, and there appeared to be 31 sites/cell. Thus vegetative cf50 Ϫ cells appear to have somewhat fewer binding sites and a higher K H than parental cells. The K H values and Hill coefficients were not statistically different from each other, but the numbers of binding sites were significantly different (p Ͻ 0.005; t test).
As with the binding to vegetative cells, the binding of 125 Irecombinant CF50 to 6-h starved cells showed apparently sigmoidal curves (Fig. 9B) and were fit significantly better with models having cooperative binding (p Ͻ 0.034, Ax2; p Ͻ 0.011, cf50 Ϫ ; F-test). For 6-h starved Ax2 cells, the K H was 15 ng/ml, the Hill coefficient was 1.9, and there appeared to be 65 sites/ cell. For 6-h starved cf50 Ϫ cells, the K H was 12 ng/ml, the Hill coefficient was 2.8, and there appeared to be 20 sites/cell. As for the vegetative cells, the K H values and Hill coefficients were not statistically different from each other, but the numbers of binding sites were significantly different (p Ͻ 0.005; t test).
The observation of a Hill coefficient of roughly 2 for the binding of recombinant CF50 to cells could be due to either a dimerization of the protein or a cooperativity of the cell surface binding sites. To distinguish between these possibilities, we performed ultracentrifugation on the recombinant CF50 and found that at 25, 50, and 100 g/ml the native molecular mass was ϳ35.6 kDa. Since the predicted mass of the protein is 30.5 kDa, this suggests that, even at relatively high concentrations, recombinant CF50 is a monomer. Assuming that iodination of CF50 does not cause it to form dimers, this would then suggest that the cell surface binding sites have cooperative binding.
Because the iodination of recombinant CF50 could conceiv-ably affect its ability to bind cells, we examined the binding of unlabeled recombinant CF50 to cells by measuring its ability to compete with 125 I-recombinant CF50 for binding to cells. As shown in Fig. 10, the binding of 125 I-recombinant CF50 was competed with until the unlabeled recombinant CF50 concentration reached ϳ43 ng/ml. At higher concentrations of unlabeled recombinant CF50, the binding of the labeled CF50 remained at a constant plateau. This plateau value represents the nonspecific binding (39). The approximate concentration where half of the specific binding was competed for by the unlabeled recombinant CF50 was ϳ17 ng/ml, indicating that the K D for binding of unlabeled recombinant CF50 to cells is ϳ17 ng/ml. Since this is in rough agreement with the K H for binding of 125 I-recombinant CF50 to cells (Table I), the data suggest that the iodination of CF50 does not significantly affect its ability to bind to cells.
Countin and CF50 Do Not Need Each Other to Bind to Cells-Because countin and CF50 are part of the same complex and the effect of CF50 on group size appears to be greatly potentiated by the presence of countin (16), the possibility exists that the binding of CF50 to cells requires the presence of countin. To test this, we examined the binding of 125 Irecombinant CF50 to countin Ϫ cells and parental Ax4 cells (Fig. 11). There were significantly better fits with models incorporating cooperative binding. The binding of 125 Irecombinant CF50 to Ax4 cells was similar to the binding to Ax2 cells at 0 and 6 h, although the vegetative Ax4 cells appeared to have somewhat more binding sites for CF50 than the vegetative Ax2 cells (Fig. 11 and Table I). The binding of 125 I-recombinant CF50 to countin Ϫ cells was less than the binding to the parental Ax4 cells at both 0 and 6 h ( Fig. 11 and Table I). However, the binding of 125 I-recombinant CF50 to the countin Ϫ cells was not significantly different from its binding to cf50 Ϫ cells (Table I).
Conversely, the possibility exists that the binding of countin to cells requires the presence of CF50. To test this, we examined the binding of 125 I-recombinant countin to cf50 Ϫ cells. We previously observed that 125 I-recombinant countin bound to both vegetative and 6-h starved countin Ϫ cells with a K H of ϳ10 ng/ml, a Hill coefficient of 2.3, and ϳ53 sites/cell (18). We found that 125 I-recombinant countin bound to vegetative Ax2 cells with a K H of ϳ17 ng/ml, a Hill coefficient of 2.3 and ϳ27 sites/cell ( Fig. 12A and Table II). At 6 h, there appeared to be somewhat more binding sites/cell ( Fig. 12B and Table II). 125 I-Recombinant countin also bound to vegetative and 6-h starved cf50 Ϫ cells ( Fig. 12 and Table II). The K H values were roughly the same as for the parental cells, and the Hill coefficients were slightly lower. For vegetative cf50 Ϫ cells, the number of binding sites/cell was slightly less than for vegetative parental cells, and for 6-h starved cells, the number of binding sites was significantly less (p Ͻ 0.005; t test) for cf50 Ϫ cells than parental cells ( Fig. 12 and Table II). DISCUSSION Dictyostelium cells appear to regulate the size of groups and fruiting bodies in part by secreting and sensing CF. Computer simulations indicated that by sensing the levels of a single factor, cells could sense the number of cells in a stream and then use this information to modulate aspects of chemotaxis, motility, and adhesion in order to regulate stream breakup and thus group size (14). We have found that two components of CF appear to have opposite effects on the cAMP-induced cAMP pulse, a key aspect of chemotaxis.
The streamer F mutant has a defective cGMP phosphodiesterase, resulting in an abnormally large cAMP-stimulated cGMP pulse (40 -42). Streamer F cells form large unbroken streams, which then form huge groups, although the large  streams and huge groups mutation appears to be unrelated to the defect in the cGMP phosphodiesterase (41). Disrupting the genes encoding the cGMP phosphodiesterases in Dictyostelium results in abnormally large cAMP-stimulated cGMP pulses without altering group size (29). In addition, disrupting a guanylyl cyclase substantially reduces the cAMP-stimulated cGMP pulse and reduces the ability of the cells to chemotax at low cell densities but does not affect fruiting body size when the cells are plated at high cell density (43). Together, this strongly suggests that altering the cAMP-stimulated cGMP pulse alone does not strongly affect group size, and thus the effects of CF on group size are not mediated primarily by CFs effect on the cAMP-stimulated cGMP pulse. Increasing the size of the cAMP pulses while cells were forming streams increased motility and decreased the size of groups, whereas adding cAMP phosphodiesterase to cells to decrease the size of cAMP pulses decreased motility and increased the size of groups (30). However, altering the size of the cAMP pulse did not affect cell-cell adhesion, suggesting that CF regulates adhesion by a pathway that does not involve the cAMP-stimulated cAMP pulse (30). countin Ϫ cells have a small cAMP-stimulated cAMP pulse (30), whereas cf50 Ϫ cells have a large cAMP pulse. Recombinant countin and recombinant CF50 also have opposite effects on the size of the cAMP-stimulated cAMP pulse. Since both countin Ϫ and cf50 Ϫ cells form large groups, and treatment of wild-type cells with either recombinant protein causes them to form small groups, our data suggest that regulating the size of the cAMP pulse is not the major pathway whereby CF regulates group size.
Assays with countin Ϫ cells, cf50 Ϫ cells, and cells treated with recombinant countin or recombinant CF50 suggested that countin and CF50 also have opposite effects on cAMP-stimulated Erk2 activation. As above, since both countin and CF50 have similar effects on group size, this suggests that Erk2 is not a major component of the CF signal transduction pathway.
There is no obvious difference between the effect of countin and CF50 on motility, adhesion, or group size and no obvious effect of countin or CF50 on aggregation up to the point where streams break up. Therefore, the different effects of countin and CF0 on the cAMP pulse and Erk2 activation do not appear to be physiologically significant with respect to the initial process of aggregation, adhesion, motility, or group size. Countin does not affect the initial differentiation of cells into CP2positive or SP70-positive cells, so the effects of countin on the cAMP pulse and Erk2 stimulation do not appear to be physiologically significant with respect to this cell type differentiation. However, CF50 does affect this differentiation, so there is a formal possibility, albeit slight, that the effect of CF50 on the cAMP pulse and/or Erk2 activation has a physiological effect on differentiation.
Activation of Erk2 by cAMP in Dictyostelium requires the cAMP receptor but does not require the associated G␣2, G␤, or cAMP-stimulated Ca 2ϩ influx, and Erk2 is required in turn for cAMP to be able to activate adenylyl cyclase (36 -38). Since countin potentiates the cAMP pulse but represses Erk2 activation, and CF50 represses the cAMP pulse but increases Erk2 activation, there does not seem to be a positive relationship between the effect of CF on Erk2 and the effect of CF on the cAMP pulse or vice versa.
Observations of countin Ϫ or cf50 Ϫ transformants as well as cells treated with recombinant countin or recombinant CF50 show that both countin and CF50 increase motility, whereas the two proteins have opposite effects on the cAMP-stimulated cAMP pulse. This suggests that at least one of the proteins regulates motility by a pathway that does not involve the effect of that protein on the cAMP-stimulated cAMP pulse.
Disrupting either countin or cf50 increases glucose levels in cells, and exogenous glucose negates the effects of either recombinant countin or recombinant CF50 on group size (17,44). Exogenous glucose increases the cAMP-stimulated cAMP pulse, whereas altering the size of the external cAMP pulse with cAMP phosphodiesterase or exogenous cAMP pulses does not affect glucose levels, suggesting that glucose lies upstream of the cAMP pulses in the CF signal transduction pathway (44). This suggests that countin and CF50 have the same effect on glucose, and the effect of recombinant CF50 on the cAMP pulse size is by a pathway that appears to bypass the effect of CF50 on glucose levels.
Whereas recombinant CF50 strongly affects the size and number of groups formed by wild-type and cf50 Ϫ cells, recombinant CF50 had a poor effect on the size and number of groups formed by countin Ϫ cells (16). Recombinant CF45-1 also has a relatively weak effect on the size and number of groups formed by countin Ϫ and cf50 Ϫ cells (17). This suggested that the presence of countin might potentiate the ability of recombinant CF45-1 and recombinant CF50 to affect group size. We found here that the presence of CF50 strongly potentiates the ability of recombinant countin to affect group size. However, when wild-type cells (which secrete a "normal" amount of CF45-1, CF50, and countin) were developed in the presence of recombinant CF45-1, CF50, and countin, in some cases the three proteins appeared to counteract each other. Together, the data suggest that countin and CF50 need each other's presence to strongly decrease group size, but that when the levels increase too far, the CF components begin to counteract each other to some extent. In the submerged assay, the recombinant CF50 showed high activity at ϳ6 ng/ml, whereas in filter pad assays, recombinant CF50 was most active at ϳ200 ng/ml (16). In the latter assay, the cells were at a much higher density, so there may be increased endocytosis or degradation per unit volume of the exogenous CF50. This would then mean that a greater starting concentration of recombinant CF50 is needed for a biological effect in filter pad assays than in submerged culture. 125 I-Recombinant countin binding reached a maximum at 3 min in the absence of added conditioned medium (18). The binding of 125 I-recombinant CF50 at 3 min was lower than that at 10 min, suggesting that despite the slight deleterious effect of iodination on recombinant countin, the binding of 125 I-recombinant CF50 to cells is slower than the binding of recombinant countin.
Although there are about 4 ϫ 10 4 receptors on Dictyostelium cells for the glycoprotein cell density-sensing factor CMF and an equal number of receptors for the chemoattractant cAMP (45)(46)(47), there appear to be only ϳ53 receptors for countin (18) and ϳ56 receptors for CF50. Similarly low numbers of receptors have been observed in many other systems (18). For instance, there are ϳ22 IL-2 receptors per cell on human monocytes (48), ϳ40 -100 interferon ␣ 2B receptors on human urothelial cells (49), ϳ74 receptors for the glycoprotein granulocyte-macrophage colony-stimulating factor on HL-60 cells, and even lower numbers of receptors on other granulocytemacrophage colony-stimulating factor-sensitive cells, with as few as eight receptors per cell on human monocytes (50,51). In all of these systems, the receptors have been shown to be functional, so the low numbers of countin and CF50 receptors  11. CF50 can bind to cells in the absence of countin. A, the binding of 125 I-recombinant CF50 to vegetative Ax4 parental and countin Ϫ cells was measured as in Fig. 9A. B, the binding of 125 I-recombinant CF50 to 6-h starved Ax4 parental and countin Ϫ cells was measured as in Fig. 9B. In A and B, values are means Ϯ S.E. from six separate experiments. The absence of an error bar indicates that the error bars were smaller than the plot symbol. WT, wild type. are consistent with known signal transduction pathways.
We previously observed that adding conditioned medium from countin Ϫ cells affected the binding of countin to cells (18), suggesting that other components of CF might affect the binding of countin to cells. We observed here that the absence of CF50 or countin decreases the number of binding sites for 125 I-recombinant CF50 on cells. This suggests that CF potentiates the number of CF50 receptors. There are many examples of ligands increasing the number of their receptors on cells; for instance, in Dictyostelium, pulses of cAMP cause an increase in the number of cAMP receptors on cells (52,53).
Countin and CF50 both bind to cells, and there is a striking similarity in the number of receptors/cell, the Hill coefficients, and the K H values for the binding of the two proteins. This could be a coincidence, or the two proteins could interact with a common receptor, conceivably by binding to a third component of CF, which is what then actually binds to the receptor. If there was a single receptor it would have to be effectively two different receptors, sensitive not only to the presence of the ligand but to its composition; this hypothetical single receptor would cause a decrease in group size, an increase in the cAMP pulse, and no change in initial cell type choice in response to countin but a decrease in group size, an increase in the cAMP pulse, and a change in initial cell-type choice in response to CF50.
Together, our observations suggest that countin and CF50 bind to cells in the presence or absence of each other. We envision that there are effectively two different receptors and two different initial signal transduction pathways. Some activation of either pathway is required for the other pathway to be able to affect group size, and this modulation appears to be nonlinear. The pathways have different effects on the cAMP pulse, Erk2 activation, and initial cell type choice, but have similar effects on the cGMP pulse, motility, adhesion, and group size. The existence of what appear to be two different receptors and two different signal transduction pathways for two components of CF suggests that the multiple subunits of this eukaryotic cell number-counting factor are present not just to make the factor large and thus diffuse more slowly but also may help to regulate group size.