Aggregating cells of Dictyostelium discoideum are able to release cyclic AMP periodically. The oscillations of cAMP generation are associated with changes in adenylate cyclase activity. Cyclic AMP receptors on the cell surface are functionally coupled to the oscillating system as evidenced by phase shifts that are induced by small pulses of extracellular cAMP. An important element of the oscillating system is the signal processing from surface receptors to the adenylate cyclase. This pathway exhibits adaptation resulting in the suppression of responses to constant, elevated concentrations of cAMP. The signal input for adenylate cyclase activation is, therefore, a change in the extracellular cAMP concentration with time.
Oscillations in the absence of detectable changes of intra- or extracellular cAMP concentrations suggest the possibility that there is a metabolic network in D. discoideum cells that undergoes oscillations without coupling to adenylate cyclase. Cyclic GMP concentrations oscillate with a slight phase difference in advance of that of cAMP, suggesting that the two nucleotide cyclases might not be activated by the same mechanism. Elevation of extracellular calcium exerts an inhibitory effect on the accumulation of cAMP and on the second of the two cGMP peaks.
Cyclic AMP is the chemotactic agent of aggregating cells of Dictyostelium discoideum. Time-lapse films made by Arthur Arndt (1937) revealed, more than 40 years ago, that the streaming of cells towards aggregation centres is often periodic, and that excitation can spread in waves from a centre towards the periphery of an aggregation territory. As emphasized by Shaffer (1962), this propagation of excitation waves is due to a relay system in which the cells act as amplifiers of the signals that originate in the centre. Wave patterns are formed in layers of preaggregation cells by self-organization (i.e. by interaction of randomly mixed cells) and often assume spiral shapes (Gerisch, 1965). The speed of propagation and period of the oscillations in a layer of densely packed cells is shown in Fig. 1. From the speed of propagation a minimal delay of about 15 s between signal input and output has been calculated (Cohen & Robertson, 1971).
The waves visualize the pattern in space and time of cellular activities which are under oscillatory control. The spiral lines, which connect cells of identical phase, become visible in the aggregation territories because the cells change shape and location during their chemotactic response (Gerisch, Kuczka & Heunert, 1963; Alcantara & Monk, 1974). The evolution of spatio-temporal order indicates that the activities of the individual cells are coupled to each other. The following experiments were attempted to show that cAMP is a coupling factor and that the waves reflect the spatio-temporal pattern of adenylate cyclase activities.
For the biochemical analysis of the oscillating system it is important to minimize phase differences between cells (i.e. to eliminate spatial heterogeneity but to preserve temporal pattern). This was performed by studying spontaneous and cAMP-induced activities in stirred cell suspensions (Gerisch & Hess, 1974). Fig. 2 illustrates the developmental stage where autonomous oscillations of adenylate cyclase have been observed. For a period before and also after this stage, the cells proved to be excitabale in the sense that they are able to release cAMP in response to small pulses of cAMP. It is assumed that the changes in cells stimulated by an extraneous cAMP pulse resemble those occurring during a period of limit cycle oscillations (Goldbeter & Segel, 1977). The sequence of events following stimulation can thus be considered as the analogue of a cycle of sustained oscillations.
There are two responses to cAMP stimuli that can be easily registered in cell suspensions: light-scattering changes (Gerisch & Hess, 1974) and transient increases in the extracellular proton concentration (Malchow et al. 1978a, b). Either one of these responses has been routinely used by us as an indicator for the responsiveness of the cells.
BIOCHEMISTRY OF EXCITATION
Cyclic AMP binds to receptors on the cell surface (Malchow & Gerisch, 1974). Within the first few seconds after cAMP application two changes are observed: an increased influx of extracellular 45Ca (Wick, Malchow & Gerisch, 1978) and an increase of the cellular cGMP concentration (Mato et al. 1977; Wurster et al. 1977). The behaviour of cAMP during the early phase of a response has proven to be more variable. In some of our experiments, no or only small increases in cellular cAMP occurred during this early phase; in others an appreciable increase was observed.
Fig. 3 A shows that the early changes may be followed within 1 min (or 2 min at 12 °C) by a second cGMP peak and a peak of cAMP. Whereas the first cGMP peak is correlated with chemotactic stimulation (Mato & Konijn, 1977; Wurster et al. 1977; Wurster, Bozzaro & Gerisch, 1978), the second cGMP peak could be involved in pignal propagation. This is definitely the case for the cAMP peak.
Stimulation of D. discoideum cells resulted in the activation of adenylate as well as guanylate cyclase (Roos & Gerisch, 1976; Mato & Malchow, 1978) (Fig. 4). Therefore, the rise of the cyclic nucleotide concentrations can be explained by activation of the enzymes that synthesize these compounds. The cAMP synthesized in response to cAMP stimuli was released into the extracellular space where it was rapidly hydrolysed by extracellular and cell surface phosphodiesterases (Gerisch et al. 1977a). The release was specific for cAMP and probably other cyclic nucleotides, as was shown by the absence of an increased rate of 5′-AMP extrusion from stimulated cells (Fig. 5).
A knowledge of the temporal relationship between fluctuating stimulant concentrations and responses is essential for the mathematical analysis of oscillations in D. discoideum (Sperb, 1979). Analysis of this relationship is complicated by the short lifetime of cAMP in the extracellular space. By using high concentrations of stimulant or, alternatively, by using a slowly hydrolysed analogue of cAMP it is possible to keep the stimulant concentration for a reasonable time at a virtually constant high level. In all cases the responses were transient even under these conditions, indicating that the were not time-independent functions of stimulant concentrations. The responses rather depend on changes of stimulant concentrations with time. This has been demonstrated for the light-scattering response (Gerisch et al. 1975a), for adenylate cyclase activation (Gerisch et al. 1977b) and, its consequences, the increase of the intracellular cAMP concentration (Rossier et al. 1978) and the rate of cAMP release into the medium (Gerisch et al. 1977 b). The exact relationship of the latter response to fluctuating extracellular cAMP concentrations has been recently examined by Devreotes, Derstine & Steck (1979) and Devreotes & Steck (1979). These authors elegantly demonstrated, by the use of a perfusion technique, that the duration of cAMP release can be extended to more than 30 min provided that the cells were stimulated by continuously increasing cAMP concentrations. The efflux of cAMP from the cells stopped soon after the rise of the stimulant concentration was discontinued.
The mechanism of the adaptation demonstrated by these experiments remains to be clarified. The alternatives are changes at the receptor level on one hand and alterations of subsequent steps of signal processing on the other. ‘Desensitization’ (a decrease of either the number of overt receptors or their affinity) refers to a change of the first category. A decrease in the number of available cAMP binding sites on the surface of cells exposed to 10−5 to 10−3M CAMP has been demonstrated by Klein & Juliani (1977). This effect progressed slowly as compared to the adaptation which is relevant for oscillations. The latter should be manifested within 1 min after cAMP addition, and it should require not more than 1–2 μM cAMP (i.e. extracellular concentrations which are actually reached during spontaneous oscillations (Gerisch & Wick, 1975)).
The cAMP released from stimulated cells competes for the radioactive cAMP that is. added for the testing of cAMP-binding to surface receptors. This complicates the measurement of affinity changes. Using a protocol which attempted to minimize this artifact (Fig. 6) we found after cAMP stimulation a transient decline in the affinity constant of the cAMP-receptor complexes by a factor of 2. This value refers to a range of cAMP concentrations between 10 and 120 nM, and thus mainly concerns the low affinity branch of the biphasic binding kinetics as they have been measured by Green & Newell (1975).
The presumed change in affinity of the cAMP receptors appears not to be extensive enough to explain the complete cessation of cAMP-synthesis and release in adapted cells. A subsequent step in signal processing seems therefore also to be affected. The relevant change in adapted cells is certainly not depletion of the substrate of adenylate cyclase, since the total ATP content is not detectably reduced after stimulation (Fig.7).
MODULATION OF THE RESPONSE BY CALCIUM
Calcium had a clear-cut effect on the responses when it was added shortly before stimulation (Fig. 3). The experiment shown in Fig. 3 has been performed in a salt solution before and after the addition of 0·1 mM CaCl2. When calcium was added shortly before cAMP was applied - and only then - the second cGMP and cAMP peaks were suppressed (Fig. 3 B). It is not clear whether calcium acts directly on nucleotide cyclases. Guanylate cyclase of D. discoideum was not inhibited by 1 mM-CaCl2 (Mato, Roos & Wurster, 1978) and adenylate cyclase was only half-inhibited by 0·1 M CaCl2 (Klein, 1976). Furthermore, the first sharp increase of cGMP was virtually unchanged in cells previously supplemented with calcium.
PERIODIC CELLULAR ACTIVITIES ARE LINKED TO ADENYLATE CYCLASE OSCILLATIONS
The investigation of synchronized periodic activities in agitated cell suspensions began with the discovery of rhythmically generated spikes of decreased light scattering (Gerisch & Hess, 1974). These spikes were produced in cell suspensions at the same developmental stage as propagated waves in cell layers. Measurement of intra- and extracellular cAMP concentrations revealed that the spontaneous light-scattering changes were related to the periodic generation of chemotactic stimuli (Gerisch & Wick, 1975). At the beginning of a light-scattering spike cAMP was accumulated within the cells before it was released, after a delay of about 0·5 min, into the extracellular medium where it was rapidly hydrolysed. Simultaneously with cAMP accumulation the number of vesicles increased within the cells, and declined thereafter (Maeda & Gerisch, 1977). This can be interpreted as indicating that cAMP is released by exocytosis, but as yet no evidence is available which shows that the vesicles which change in number contain cAMP.
The cAMP oscillations are associated with periodic changes in adenylate cyclase activity (Roos, Scheidegger & Gerisch, 1977). However, it is not only the cAMP concentration which oscillates, but also that of cGMP (Wurster et al. 1977). Moreover, the cGMP peaks slightly precede those of cAMP (Fig. 8). This raises the question of whether guanylate cyclase is more directly linked to the oscillating system than adenylate cyclase, which might be secondarily activated by some event accompanying the cGMP increase.
Fig. 8 shows an experiment in which elimination of extracellular calcium by EGTA reduced the amplitude of light scattering, as it has been previously shown for EDTA (Gerisch et al. 1975b). Nevertheless, the peaks of intracellular cGMP and cAMP were of essentially unchanged amplitude. The transport of cAMP into the extracellular space was also unaffected. After, however, calcium had been returned, the cyclic nucleotide concentrations remained at the basal levels, similar to the suppression of the second cGMP and cAMP peaks in excitable cells stimulated by cAMP (Fig. 3B).
The blockage of oscillations by a sudden increase in the extracellular calcium concentration can be hypothetically explained by two assumptions. (1) Cytoplasmic calcium mediates a reaction which inhibits guanylate and adenylate cyclases. This does not necessarily mean a direct action on the respective enzymes. (2) The cytoplasmic calcium concentration is determined by the rate of calcium influx and the activity of calcium pumps that remove calcium from the cytoplasmic space. If the pumps adapt their activity only slowly in response to increasing calcium concentrations, the cytoplasmic calcium concentration would transiently rise to an elevated level, and would decline after adaptation of the pumps.
SIGNAL INPUT FROM CELL SURFACE RECEPTORS TO THE OSCILLATOR
Small pulses of cAMP shift the phase of the oscillations. The direction and extent of the shifts depend on the phase in which the pulses are applied (Malchow et al. 1978a). Only during their own cAMP production are the cells insensitive to extraneous pulses (Fig. 9). These results show that at all phases of an oscillatory cycle (apart from the signalling phase) the receptors can be activated by extracellular cAMP.
It is nevertheless likely but not definitely established that the affinity of the receptors for cAMP changes during the oscillations. Oscillatory changes of cAMP binding have been reported by Klein, Brachet & Darmon (1977) and by King & Frazier (1977). Because affinity changes can be simulated by the periodic release of cAMP, which competes with the labelled cAMP added, the latter authors have studied oscillations in cells whose cAMP production was blocked by dithiothreitol. The relationship of the oscillations observed under these conditions to those discussed in this paper is uncertain. The oscillations observed by King and Frazier were of shorter period (2 min) and apparently ran with almost undamped amplitude in the absence of cAMP. They also appeared to be associated with periodic modification of surface sites by epicellular protein kinase. The periodically phosphorylated sites were suggested to be cAMP receptors. In contrast, 32P incorporation, from extracellular ATP into intact cells, was extremely low, if not absent, under our conditions and remained so after stimulation with cAMP (Rahmsdorf, Malchow & Gerisch, 1978). Cell surface phosphorylation is therefore unlikely to be a candidate for receptor modification during oscillations as they have been studied in our laboratory.
IS ADENYLATE CYCLASE ACTIVATION INDISPENSABLE FOR OSCILLATIONS TO OCCUR?
Models for oscillations in D. discoideum have been designed in which adenylate cyclase plays a central role (Goldbeter, 1975; Cohen, 1977). Such models are supported, in principle, by the finding that pulses of cAMP cause phase-shifts, indicating that cAMP receptors are coupled to the oscillating pathway, probably through guanylate or adenylate cyclase activation. The importance of extracellular cAMP pulses for the maintainance of sustained oscillations is also supported by experiments in which continuous fluxes of cAMP were applied in order to establish steady, elevated extracellular cAMP concentrations. These fluxes irreversibly abolished light-scattering oscillations, probably because the cells could no longer detect and produce their own cAMP pulses (Gerisch & Hess, 1974). It is possible, however, that cAMP only couples the oscillations of the individual cells to each other and thus acts as a synchronizer in cell suspensions, without being essential for oscillations at the single cell level.
As discussed above, periodic light-scattering changes are clearly correlated with oscillations of intra- and extracellular cAMP concentrations. This is only true, however, for oscillations in which light scattering decreases in the form of spikes. After some time this type of oscillation is often converted into a sinusoidal pattern in which the spike-like depressions are replaced by smoothly curved increases of light scattering (Fig. 10). Sometimes small peaks of cAMP concentrations were still observed in the phases in which the spikes were to be expected, but these peaks became progressively smaller from one period to the next (Fig. 11). In other cases neither intra- nor extracellular changes of cAMP concentrations could be detected during sinusoidal oscillations (Fig. 12). Similar results have been obtained by determining adenylate cyclase activities which either slightly increased, in the appropriate phase, or remained at basal levels without any obvious oscillations. These data leave the following possibilities open. (1) Rudimentary cAMP oscillations are present during sinusoidal oscillations which the cells can detect and use for the synchronization of their activities. Our assays are, however, less precise than the recognition system of the cells. This possibility is improbable because sinusoidal oscillations proved to be fairly resistant against disturbing cAMP pulses, suggesting that during these oscillations the cells are relatively insensitive to cAMP. (2) Sinusoidal oscillations occur in the absence of any cAMP changes. If this were true, the question would be whether there is a synchronizing agent other than cAMP. Sinusoidal oscillations are, however, slightly damped. Thus it is not impossible that they continue, after initial synchronisation, without any coupling between the cells.
FUNCTIONAL CONNEXION OF THE OSCILLATING SYSTEM TO MEMBRANE DIFFERENTIATION
In one particular strain of D. discoideum (Nc-4, including its axenically growing derivative Ax-2), the periodically generated cAMP pulses are important for cell differentiation from the growth phase to the aggregation competent stage. Small ^pulses of cAMP stimulated the differentiation, whereas a continuous influx of cAMP of the same average rate inhibited (Gerisch et al. 1975c). The inhibitory effect of steady concentrations was particularly strong in the case of a slowly hydrolysable cAMP analogue, adenosine-3′,5′-cyclic phosphorothioate (Rossier et al. 1978). The effect of cAMP pulses was particularly clear-cut in the case of certain non-aggregating mutants (Darmon et al., 1975). The importance of oscillations for cell differentiation is underlined by results obtained with a phosphodiesterase-negative mutant (Darmon, Barra & Bracket, 1978). In this mutant no oscillations were observed and aggregation competence was not acquired. However, the mutant exhibited oscillations and development proceeded when purified phosphodiesterase was added to the medium.
The reason for the inefficiency of steady cAMP concentrations is apparently the adaptation of the response system. Fluctuation of cAMP production – not necessarily periodically – is, therefore, a prerequisite for optimal stimulation of development. The intracellular factors which connect the control of development to the activation of receptors at the cell surface are unknown. Calcium, cGMP and cAMP are certainly candidates. Soluble cAMP or cGMP binding proteins specific for either cAMP or cGMP are present in the cells and could mediate effects of intracellular cyclic nucleotides on development. D. discoideum has been reported to possess cAMP-dependent developmentally regulated protein kinases (Sampson, 1977). The protein kinases tested in our laboratory were neither activated by cAMP nor cGMP and did not clearly co-fractionate with the binding proteins for these cyclic nucleotides (Rahmsdorf & Gerisch, 1978).
The stimulation of cell development by periodic cAMP pulses implies that the control of membrane proteins occurs through the activation of membrane receptors. The membrane components controlled by cAMP pulses include the constituents of the cAMP signal system: adenylate cyclase, cAMP-phosphodiesterase and cell surface cAMP receptors (Roos, Malchow & Gerisch, 1977a). In addition, the expression of a specific glycoprotein on the cell surface is enhanced by cAMP pulses (Gerisch et al. 1975c). This glycoprotein has been suggested to be involved in a specific type of cell adhesion which is characteristic of cells in the aggregation stage (Müller & Gerisch, 1978).
Our work was supported by the Swiss Nationalfonds, the Deutsche Forschungs-gemeinschaft and the Stiftung Volkswagenwerk.