SEARCH

SEARCH BY CITATION

Keywords:

  • collective behavior;
  • Dictyostelium;
  • oscillation;
  • quorum sensing;
  • synchronization

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Collective oscillations in microbial populations
  5. cAMP oscillations and waves in Dictyostelium
  6. Future prospect
  7. Acknowledgments
  8. References

From hormonal secretion to gene expression, multicellular dynamics are rich in oscillatory regulation. When organized in space and time, periodic cell–cell signaling can give rise to long-range coordination of gene expression and cell movement in tissues. Lack of synchrony of the oscillations on the other hand can serve as a source of initial divergence of cell fate in stem cells. How properties of individual cells can account for collective rhythmic behaviors at the tissue level remains elusive in most cases. Recently, studies in chemical reactions, synthetic gene circuits, yeast and social amoeba Dictyostelium have greatly enhanced our view of collective oscillations in cell populations. From these relatively simple systems, a unified view of how excitable and oscillatory regulations could be tuned and coupled to give rise to tissue-level oscillations is emerging. The review focuses on recent progress in cyclic adenosine monophosphate oscillations in Dictyostelium and highlights similarities and differences with other systems. We will see that the autonomy of single-cell level oscillations and different ways in which cells are coupled influence how group-level information can be encoded in collective oscillations.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Collective oscillations in microbial populations
  5. cAMP oscillations and waves in Dictyostelium
  6. Future prospect
  7. Acknowledgments
  8. References

Collective oscillations everywhere

There are countless examples of synchronized rhythmic activities exhibited by populations of cells. Electrophysiological activities of suprachiasmatic nuclei supports circadian rhythms (Yamaguchi et al. 2003). Action-potential waves of myocardiac cells generate rhythmic contraction of the heartbeat. Oscillatory secretion of insulin by pancreatic beta cells (MacDonald & Rorsman 2006) enhances its hormonal action (Schmitz et al. 2008). Expression of transcription factor Her1/7 is highly synchronized in presomitic mesoderm during somite segmentation (Horikawa et al. 2006; Riedel-Kruse et al. 2007) and appear as periodic traveling waves that determine precise timing and spacing of cell differentiation (Pourquié 2003). These periodic behaviors form the basis of coordinated cellular organization that requires precise repetition and ordering of events in a tissue. Loss of synchrony often has a detrimental effect such as segmentation defects (Horikawa et al. 2006; Riedel-Kruse et al. 2007) and death of circadian rhythms upon exposure to strong light (Ukai et al. 2007). Heart failure is often associated with the appearance of complex oscillatory patterns (Kim et al. 2007) that disrupt spatio-temporal coordination of heart contractions.

On the other hand, it is also becoming clear that asynchrony plays a critical role in increasing heterogeneity in cellular states. Cell–cell variability in Hes1 oscillations causes differences in how embryonic stem cells respond to the same inductive signal and is therefore associated with their early cell fate divergence (Kobayashi et al. 2009). In the case of oscillatory shuttling of NF-kappaB between the cytosol and the nucleus whose frequency coordinates differential gene expression (Ashall et al. 2009), synchrony appears to be prevented by a specific regulatory loop in the signaling pathway (Paszek et al. 2010). Degree of synchrony in rhythmic behaviors and how it is regulated are thus intricately tied to a key question in developmental biology; how tissues become organized in time and space and how diverse cell types appear depending on the presence or lack of coordination with other cells.

How collective rhythmic behaviors at the organismal level emerge from dynamics of cellular and subcellular levels remains elusive in most cases. This is a challenging problem as it requires one to bridge the large gap between multiple levels of organization; i.e., molecules to cells, cells to tissues. Collective rhythmic behaviors are known in a wide spectrum of systems from populations of fireflies (Buck & Buck 1968) and bees (Kastberger et al. 2008) to physical systems such as Josephson junctions (Wiesenfeld et al. 1996) and electro-chemical reactions (Kiss et al. 2002). Mathematical frameworks (Winfree 1980) have developed over the years that could guide us when extracting the essential ingredients and properties common to these phenomena. A key feature to remember when considering their roles in cell–cell signaling and gene regulation is that group-level information is stored in collective oscillations differently – either in the amplitude or the frequency depending on how they are realized. With only a few exceptions from physico-chemical systems (Kiss et al. 2002; Taylor et al. 2009), this, however, has remained a theoretical prediction waiting to be tested experimentally.

Recently, quantitative studies on synthetic circuits in bacteria, glycolytic oscillations in budding yeast and cyclic adenosine monophosphate (cAMP) oscillations in social amoeba Dictyostelium have greatly enhanced our view of collective oscillations in cell populations. From studies of these relatively simple systems, a unified view of how excitable and oscillatory regulations could be tuned and coupled to give rise to tissue-level oscillations is emerging. Because measurements can be made both at the single-cell and population level, and because quantitative analyses and manipulation are relatively easy, how density and other extracellular information is stored in the oscillations can be tested experimentally. In this article, we will first briefly describe the essential features required for oscillatory signaling by taking examples from synthetic circuits. We will then describe progress in chemical reactions, yeast and compare them to cAMP oscillations in Dictyostelium by reviewing early and recent studies. We highlight their similarities and differences and describe the distinct ways by which collective oscillations are realized and how they store group-level information.

Collective oscillations in microbial populations

  1. Top of page
  2. Abstract
  3. Introduction
  4. Collective oscillations in microbial populations
  5. cAMP oscillations and waves in Dictyostelium
  6. Future prospect
  7. Acknowledgments
  8. References

Generating oscillatory transcription by a synthetic circuit

Concentrations of signaling molecules may change periodically when the key component is regulated by a negative feedback loop. Suppose mRNA (X) encodes a kinase (Y) and that Y activates a transcription factor (Z) that inhibits transcription of X (Fig. 1a). If both activation of Z and inhibition of X follow simple first order kinetics, activation of X immediately turns to inhibit X through activation of Y. Such is the case for a simple harmonic oscillation of a pendulum swing. However, such linear oscillations are susceptible to friction so that amplitude decays quickly due to energy dissipation. This is certainly not the type of biochemical oscillations one typically finds in a cell. In order for oscillations to persist, the level of X needs to increase dramatically before Z starts to inhibit X. In other words there needs to be a delay between these two processes – one that acts to increase X and the other that works in the opposite direction – that creates a sufficient overshoot in both directions. This is certainly the case when the reaction follows Michaelis–Menten type kinetics with high Hill coefficients. Due to the all-or-none response in the reaction rate, activation of Z does not kick in until Y increases to a sufficient level. Likewise, Z will not be suppressed unless Y lowers to a value below a certain level. These types of nonlinearity in reaction rates realize a sufficient time delay is required for persistent oscillations.

image

Figure 1.  A negative feedback regulation combined with strong nonlinearity in reaction kinetics give rise to temporal oscillations. (a) A cyclic circuit composed of three nodes; X activates Y, Y activates Z, and Z represses X. (b) A simpler two-variable circuit with a negative feedback loop and auto-regulation. Autonomous oscillations in the circuit shown in (b); (c) time series of X and Y and (d) the trajectory of a time course in the XY plane. The ordinary differential equations dX/dt = −cXX + VX (X2 + S)/(X2 + aY + KX), dY/dt = −cYY + VYX2/(X2 + KY) are numerically solved. The first and the second terms of the equations represent degradation of the gene products and transcriptional activity, respectively. X, Y: concentrations of proteins. S: an inducer concentration (cX = 20, VX = 20, KX = 0.04, = 0.4, cY = 2, VY = 4, KY = 0.1, and S = 4.3 × 10−3). Blue and red lines are nullclines (Winfree 1980), dX/d= 0 and dY/d= 0, respectively. Following a vector field indicated by black arrows in (d), trajectories from any initial conditions converge to a closed cycle, i.e., limit cycle (black thick line) when the steady state represented by the intersection (blue filled circle) of the nullclines is unstable.

Download figure to PowerPoint

We should note that a simpler reaction with only two variables can give rise to persistent oscillations (Fig. 1c) provided that there is an auto-regulation. Figure 1b illustrates a case where X activates itself and Y, and Y negatively regulates X. In this case, inhibition of X by Y does not take place unless the level of activated Y becomes sufficiently high. Even when the level of activated Y is elevated, the level of activated X does not immediately come down due to the presence of auto-regulation (Fig. 1d). This is another way by which a time delay necessary for persistent oscillations can be brought about in cellular signaling.

An illuminating example of reaction kinetics that can give rise to oscillations is the synthetic transcription circuit “repressilator” (Elowitz & Leibler 2000). Here, transcription factors LacI, TetR and LambdaCI mutually repress expression of the other so that they form a relation akin to rock-paper-scissors (Fig. 2a). When the circuit is embedded in Escherichia coli together with a green fluorescent protein (GRP)-reporter that monitors activity of the tetracyclin repressor, fluorescence of individual bacterium oscillate in a cycle of approximately a few hours (Elowitz & Leibler 2000). By comparing Figures 1a and 2a, one would immediately see the logic behind the design of the circuit. Increase in X inhibits Y, and inhibition of Y acts to decrease X through Z. This forms a negative feedback loop necessary for oscillations to take place. The time delay can be brought about by nonlinearity of the transcriptional dynamics and time required for proteins to be translated. Such an artificial circuit helps one understand the logic behind the seemingly complex behavior of cells not to mention their importance in bioengineering and biomedical applications.

image

Figure 2.  Synthetic circuits that generate oscillatory gene expression follow a common scheme described in Fig. 1. (a) A cyclic circuit “repressilator” (Elowitz & Leibler 2000) composed of three promoter-gene pairs; TetR represses λcI, λcI represses lacI, and LacI represses tetR. A cascade of repression provides necessary nonlinearity and a negative feedback loop required periodic gene expression. (b) A circuit composed of a pair of promoter-gene sets (Stricker et al. 2008). LacI represses ntrC and the product of ntrC induces lacI expression. Product of ntrC also induces itself and adds the essential nonlinearity to the reaction kinetics. (c) Transcription of araC and lacI is driven by a hybrid promoter plac/ara-1 that is activated by AraC and repressed by LacI. (d) Time series and (e) a phase-plane trajectory of an excitable behavior exhibited by the equations in Fig. 1 (S = 4.3 × 10−4). A perturbation (orange arrow in e) that exceeds a threshold transiently elevates X and Y following the vector field (black arrows in e) before returning to the stable steady state (blue filled circle in e).

Download figure to PowerPoint

A more robust oscillator has been developed (Atkinson et al. 2003) that incorporates a circuit shown in Figure 2b. The system consists of the ntrC gene that encodes nitrogen regulator protein NRI that acts as a transcriptional activator. The expression of ntrC and lacI are driven by a ntrC promoter. The expression of ntrC is also inhibited by Lac repressor encoded by lacI thus forming a negative feedback loop. The benefit of this construct is that one could tune the level of repression by LacI by adding non-metabolizable analogue of lactose IPTG (isopropyl-beta-d-thiogalactopyranoside) in the medium. Hasty and others have also reported on a synthetic oscillatory circuit (Stricker et al. 2008) of similar nature that is also tunable. The system makes use of the plac/ara-1 promoter element (Lutz & Bujard 1997) to implement the necessary positive and negative regulation by AraC and LacI, respectively (Fig. 2c). When AraC binds to arabinose, it becomes a transcriptional activator. By altering concentrations of arabinose and IPTG, the strength of both of the feedback loops can be tuned in this system.

When strength of repression is sufficiently weak, the system no longer oscillates autonomously, but it can still exhibit excitability (Fig. 2d,e). Upon induction by IPTG, X rises rapidly and then falls to a steady state. This is followed by a slow decrease in Y, which keeps X inactivated for some time. This time window is called a refractory period (Keener & Sneyd 1998). Due to this property, X is re-activated only when external stimuli are applied at intervals longer than the refractory period. Excitability of action potential in neurons and cardiomyocytes is well known but not the only example. Differentiation of Bacillus subtilis to a state of competence is transient and driven by an excitable gene circuit (Süel et al. 2006). Here, expression of transcription factor ComK is positively regulated by itself and another transcription factor ComS. Because ComK negatively regulates ComS, and ComK is indirectly upregulated by ComS, the circuit contains a negative feedback loop. Due to stochasticity in gene expression, excitability of the circuit gives rise to random and transient differentiation. An illuminating aspect of Bacillus differentiation is that it can be made to switch between excitable and oscillatory states by changing the level of ComS or ComK by an additional copy of the gene under an inducible promoter (Süel et al. 2007). In mammalian NF-kappaB signaling, nuclear translocation of the transcription factor is observed only when the cytokine tumor necrosis factor-α (TNF-α) is repetitively applied at intervals longer than a certain time period (Ashall et al. 2009). The synthetic circuits and the real case examples described here demonstrate how a negative feedback loop in gene regulation gives rise to excitable and oscillatory response and how they are utilized in cells.

Collective oscillations by coupling synthetic circuits

The oscillations realized by the simple circuits in bacteria (Fig. 2a–c) do not persist at the population level. After a few rounds of rise and fall, oscillations appear to attenuate even though they continue at the individual cell level. Differences in phase will increase in time due to cell–cell variability so that after a few cycles, periodic gene expression in individual bacteria will be completely out of sync (Elowitz & Leibler 2000). This is also true in mammalian cells for the oscillatory response in NF-kappaB described above (Paszek et al. 2010) as well as for the periodic synthesis of p53 in response to stress signals (Lahav et al. 2004). Asynchronous oscillations of individual cells will cancel out by averaging, thus providing a false picture that cellular responses are only transiently oscillatory. Unless the frequency is finely tuned so that they are equal in all cells, some form of cell–cell communication must exist to synchronize timing of the oscillations.

Cell–cell signaling in Vibrio fischeri; a symbiotic marine bacterium that emits bioluminescence in the light organ of squid (Waters & Bassler 2005) is well studied and one that has been implemented in synthetic circuits (Basu et al. 2005; Danino et al. 2010). Regulation of bioluminescence is mediated by the small molecule N-acyl homoserine lactone (AHL) encoded by the luxI gene that induces itself (therefore called an “autoinducer”) and a set of enzymes and proteins such as luciferase to produce luminescence. AHL freely diffuses between the cells and penetrates cell membrane. As bacteria will only sense AHL above a certain concentration threshold, the signaling mechanism ensures that bioluminescence is produced only when bacteria are in the symbiotic state closely packed together in the light organ. This density sensing happens in an all-or-none fashion and is called “quorum sensing” (Waters & Bassler 2005).

It is natural to expect that by coupling cells harboring oscillatory circuits (Fig. 2a,b) by a quorum sensing mechanism, one could artificially establish synchronized oscillations in bacteria. Population-level oscillations is predicted from model simulations of a circuit that modifies repressilators (Fig. 2a) so that expression of lacI is dependent on AHL encoded by the luxI gene, and the luxI gene is negatively regulated by LacI (Garcia-Ojalvo et al. 2004). The first successful demonstration reported recently (Danino et al. 2010) instead built on the proven design of oscillatory circuit in E. coli (Fig. 2b). Here, luxI replaces ntrC as the autoregulatory activator gene, and lacI is replaced by aiiA, which encodes inhibitor of AHL (Fig. 3a). Because AHL freely diffuses between the cells, the positive feedback loop depends not only on the amount of AHL molecules synthesized within a cell but also on those secreted by others. Above a critical cell density, synchronized oscillations of gene expression emerge at the colony level. Although it remains to be tested whether the individual bacteria can oscillate without coupling, model simulations (Danino et al. 2010) suggest that they are autonomously oscillatory. The collective oscillations in such cases result from synchronization of individual oscillators as we will see in more detail below.

image

Figure 3.  Cell-density dependent transition to collective oscillations. (a) A synthetic system (Danino et al. 2010) that couples the circuit design in Figure 1 by extracellular signals. The luxI gene encodes secreting molecule N-acyl homoserine lactone (AHL) that can diffuse freely through the membrane to neighboring cells. LuxR is constitutively expressed. AHL is an auto-inducing molecule that binds to LuxR and promotes expression of luxI itself. The AHL-LuxR complex also induces expression of aiiA, which in turn inhibits luxI expression. This forms a negative feedback loop responsible for excitability and oscillations. The level of AHL-LuxR is probed by expression of GFP. Two types of transition to group-level oscillations; (b) Kuramoto-type and (c) dynamical quorum sensing (DQS)-type. In the Kuramoto-type transition, autonomously oscillating cells are desynchronized at low cell density (b, upper panel). Group-level behavior appears first as a partially synchronized oscillation of the autoinducer concentration at intermediate densities (b, middle panel) followed by full synchronization at high densities (b, lower panel). In DQS, cells are quiescent at low cell density (c, upper panel). Due to excitability, oscillations emerge by mutual induction above a critical cell density (c, middle and bottom panels). Computer simulations of 1000 cells with a simple circuit similar to (a) are shown. Time course of intracellular autoinducer concentrations for 10 representative cells (thin lines) and population average (thick blue line).

Download figure to PowerPoint

Two scenarios for collective oscillations: Hints from chemistry

As we have seen, synthetic gene-circuits have provided us with insights into how oscillatory dynamics could take place in a cell and in a group of cells. It is no less appropriate here to describe what chemistry has offered to this end. The well-known Belouzov-Zhabotinsky (BZ) reaction (Zhabotinsky 1991) involves bromination of malonic acid in the presence of catalyst such as ferroin. HBrO2 is made in this process and speeds up the reaction autocatalytically. As bromide ion increases, it starts to inhibit the autocatalytic reaction. Bromoderivatives of malonic acid thus returns to the original reduced form. The process repeats periodically and could be visualized by change in the color of ferroin. Depending on inline image concentration, BZ reaction can be either oscillatory or excitable (Field & Noyes 1974).

The large disparity between the BZ reaction and the oscillating cells is that the chemical reactions in most cases are studied in a continuously stirred solution where reaction takes place uniformly in space. In cells, a majority of events takes place inside the plasma membrane and organella, and limited species of molecules are exchanged extracellularly. A modified BZ reaction that takes into account the effect of compartmentization has been studied by Showalter and his colleagues (Taylor et al. 2009). Here, reaction takes place in the catalyst-loaded particles. By suspending the particles in solution including everything required for the BZ reaction except the catalyst ferroin, one can limit the crucial reaction to take place only in the particles while allowing exchange of reactants and subreactions to take place in the solution, such as exchange of autoinducer between intracellular and extracellular environment. Because ferroin only exists in the particle, HBrO2 is formed and so the particles change color from red to blue. This increases auto-catalytically before increased Br eventually drives the reaction back to the original reduced state.

The particle-based BZ reaction has elegantly demonstrated that there could be two ways by which synchronized oscillations could emerge from coupled reactions of many discrete substrates. When the solution including the particles is stirred at relatively low speed, the redox state of catalysts-loaded particles appears to be oscillating individually. Due to differences in phase and frequency, these oscillations cancel each other out when averaged so that as a sum, certain fractions of the particles effectively appear to be in the oxidized state at all times, i.e., no oscillations are observed at the level of population (thick blue line in Fig. 3b upper panel). When one increases the number of particles, the phase of the oscillations becomes synchronized, and as a result, one begins to observe oscillations at the population level (thick blue lines in Fig. 3b). This is the so-called Kuramoto-type transition (Winfree 1980; Kuramoto 1984). The hallmark of this type of transition is that the population-average of the amplitude, i.e., degree of synchronization, continuously increases (thick blue lines in Fig. 3b panels) with the coupling strength, e.g., number of particles, diffusion constant, stirring rate, etc. The gradual increase is due to the existence of an intermediate state where only a fraction of the elements are synchronized in phase and frequency (thin lines in Fig. 3b middle panel). Raising the coupling strength further promotes complete entrainment of all elements (Fig. 3b bottom panel). This is in fact what is observed in the particle-based BZ reaction when the number of particles is increased at a low stirring rate (Taylor et al. 2009).

Qualitatively different behavior is observed at high stirring rates. When the number of particles is small, particles are no longer in the oxidized state implying that the individual particles are not oscillating (Fig. 3c upper panel). As the number of particles is increased, oscillations appear at some critical density (Fig. 3c middle panel). As a result, the amplitude discontinuously increases in a switch-like manner from zero to a finite value above a critical number of cells (thick blue lines in Fig. 3c). This type of transition is recently coined a term “dynamical quorum-sensing” (DQS), because there is a density-dependent qualitative change in the behavior at the single-cell level; cells switch from a quiescent to an oscillatory state, and that change is mediated by an autoinducing signal of some form. The large amplitude at the onset of the DQS-type transition means that the cells are already highly synchronized (thin lines in Fig. 3c middle panel) when the group-level behavior first appears. This is in marked contrast to the Kuramoto-type transition where cells oscillate even before the transition and the average amplitude grows continuously (thick blue lines in Fig. 3b,c).

The distinction between the Kuramoto and the DQS-type transitions is that isolated cells are autonomously oscillatory in the Kuramoto-type, whereas in DQS they are merely excitable (Fig. 2d,e) (Tinsley et al. 2010). Below a threshold concentration of an autoinducer, some cells are randomly excited due to stochasticity of the signaling reaction. This transiently raises the concentration of the autoinducer. When the density is low, excitation is rare and isolated to individual cells. Above a critical number of cells, transient pulses of the autoinducer can accumulate and become large enough to mutually excite all of the cells at once (Fig. 3c middle panel), as we shall see in detail for Dictyostelium cAMP oscillations. As density is further increased, the frequency of the oscillations increases, because occurrence of the excitatory events increases (Fig. 3c bottom panel). In the engineered E. coli (Fig. 3a), both the amplitude and the frequency of the collective oscillations increase with the decay rate of extracellular AHL but remain relatively constant to changes in cell density (Danino et al. 2010). These features are difficult to interpret unless further investigations clarify the exact nature of the oscillations.

Collective oscillations in budding yeast

The concept of DQS was firstly put forward in yeast glycolysis (De Monte et al. 2007). When a population of budding yeast Saccharomyces cerevisiae is starved under limited glucose supply followed by cyanide treatment, the redox state of NADH begins to oscillate at a period of about 30 s which can be observed conveniently by its fluorescence (Hess & Boiteux 1971; Chance et al. 1973; Richard et al. 1993). The oscillations in yeast could be made to persist indefinitely by continuous feeding of the substrate and removal of secreted metabolites (Jacobsen et al. 1980; Danøet al. 1999). Since the oscillations persist at the population level, it is clear that something that is secreted extracellularly is mediating cell–cell communication. A candidate is acetaldehyde, which can cause a phase shift in the oscillations when applied exogenously (Richard et al. 1996). From biochemical assays, it is known that concentrations of other glycolytic intermediates such as adenosine triphosphate (ATP), adenosine monophosphate (AMP), and pyruvate oscillate at the same period although at different phases (Hess & Boiteux 1971; Madsen et al. 2005). There are several proposed mechanisms for glycolytic oscillations (Madsen et al. 2005). It is generally believed that inhibition of phosphofructosekinase (PFK) by ATP is the source of a negative feedback (Goldbeter 1996).

How do glycolytic oscillations emerge in a yeast cell population? It has been noted that glycolytic oscillations in Saccharomyces carlsbergensis disappear when cell density is lowered (Aldridge & Pye 1976). Some studies suggest that individual yeast cells are still able to oscillate even when group-level oscillations are absent (Chance et al. 1973; Aon et al. 2007). More recent observation indicates that under well controlled environments, single yeast cells in isolation are unable to oscillate (Poulsen et al. 2007). We should note, however, that the amplitude of yeast oscillations increases continuously with the square root of cell density. This is a characteristic feature of a Kuramoto-type transition. These observations can be reconciled if one assumes that oscillatory instability at the individual cell level increases with cell density and plays a key role at the onset of collective oscillations (De Monte et al. 2007). According to the definition of DQS-type transition based on the BZ reaction (see “Two scenarios for collective oscillations: Hints from chemistry” section), it appears that the glycolytic oscillation sits at the crossroad of the two forms of transition.

cAMP oscillations and waves in Dictyostelium

  1. Top of page
  2. Abstract
  3. Introduction
  4. Collective oscillations in microbial populations
  5. cAMP oscillations and waves in Dictyostelium
  6. Future prospect
  7. Acknowledgments
  8. References

History

A clear demonstration of DQS-type transition came from another model system: social amoebae Dictyostelium discoideum. Under starved conditions, cells aggregate synchronously to form a fruiting body (Kessin 2001). The periodic nature of cell aggregation became first evident by time-lapse filming of cells under a microscope (Gerisch 1965). This and subsequent studies revealed that a monolayer of aggregating cells exhibit patterns of dark and light bands that alternate in 5–10 min-cycle and propagate as waves at a speed of 2–4 × 102 μm/min (Alcantara & Monk 1974; Durston 1974). The light bands consist of elongated moving cells while the dark bands consist of relatively non-motile rounded cells (Alcantara & Monk 1974). It was realized that the waves appear to have generic properties of excitable media; waves of target and spiral form propagate directionally and annihilate when two fronts collide (Durston 1973, 1977). These properties were hence studied more quantitatively by use of well set up dark-field illumination (Gross et al. 1976) that allowed visualization of light scattering waves in high contrast (Siegert & Weijer 1989; Sawai et al. 2005) (Fig. 4a). With the use of automated imaging, the relative ease of visualization can be extended to high-throughput analyses of multiple samples (Sawai et al. 2007).

image

Figure 4.  Cyclic adenosine monophosphate (cAMP) dynamics in starved Dictyostelium cells. (a) Optical density waves observed under a dark-field illumination. Cells are placed on an agar substrate at 1 monolayor (6600 cells/mm2) cell density. Images were taken at 4, 5, and 6 h after starvation (panels from left to right). Frame-subtracted images shown in green are superimposed on the original grayscale images (8.3 mm × 8.3 mm); lighter gray represents a higher cell density, black regions are devoid of cells. Note that cells are aggregating in streams toward the centers of spiral waves. (b, c) Changes in cytosolic cAMP observed by means of fluorescence resonance energy transfer (FRET)-based sensor epac1-camps. Colors represent isolated cells in a perfusion chamber. The extracellular cAMP was added at the time indicated by the dashed line. Response is either in the form of damped or sustained oscillations depending on the concentration of extracellular cAMP (<100 nmol/L) (b) and (>100 nmol/L) (c), respectively. (d) Comparison of individual and averaged cAMP response. Cells are exposed to a constant flow of buffer with or without cAMP. Thin lines indicate response in five cells to 1 μmol/L extracellular cAMP applied at time zero. The red line in bold is the group average. Peaks after the initial spike are not clearly visible in the averaged signal.

Download figure to PowerPoint

By 1970s, it was shown that chemotactic response can be elicited by extracellular cAMP (Konijn et al. 1967; Bonner et al. 1969; Robertson et al. 1972). The close relationship between light scattering change and cAMP became clear from photometric measurements of continuously stirred cell suspension (Gerisch & Hess 1974; Gerisch & Wick 1975; Gerisch et al. 1979). These studies discovered that optical density changes periodically at about 7 min similar to the periodicity of waves observed in a monolayer of cells. The change in optical density in cell suspension is thought to reflect the degree of agglutination (Gerisch & Hess 1974; Wurster & Kurzenberger 1989). However, in a monolayer of cells on an agar substrate, waves of optical density are apparent well before aggregation begins. It was shown that cells become round when stimulated with cAMP and that this follows similar kinetics as the optical density change (Futrelle et al. 1982). Since round cells scatter less light, the optical density changes in a monolayer of cells likely reflect cell shape change in response to rise in extracellular cAMP (Futrelle et al. 1982).

The advantage of studying oscillations in cell suspension is that it makes assays for cAMP more feasible, since a large number of cells signaling in a uniform environment could be obtained. Using an assay based on competitive binding of cAMP and isotope-labeled cAMP to protein kinase A (PKA), it was shown that, in cell suspension, concentration of extracellular and intracellular cAMP oscillate and that they coincide with periodic changes in optical density. According to these studies, extracellular cAMP in a cell-suspension oscillate between 100 nmol/L to 10 μmol/L when the cell density is 2 × 108 cells/mL. The level of intracellular cAMP oscillates between 3 to 20 μmol/L in phase with extracellular cAMP oscillations (Gerisch & Wick 1975). In the absence of autonomous oscillations, the application of a pulse of 10 nmol/L extracellular cAMP induces the so-called “cAMP relay response”: i.e., a transient synthesis of cAMP upon stimulation with extracellular cAMP (Roos et al. 1975; Shaffer 1975). Similar perturbations applied to oscillating population shifts phase of the oscillations (Malchow et al. 1978; Gerisch et al. 1979).

It thus became evident that application of extracellular cAMP in a nanomolar range induces not only chemotactic response but also production and secretion of cAMP. If one extrapolates such behavior to cells on a solid substrate, the implication is quite clear: cAMP is secreted by the cells diffuse to the surrounding cells and elicits a further response. The whole process could be relayed between the cells and appear as propagating bands of optical density change. The final proof came from the works of Devreotes and his co-workers (Tomchik & Devreotes 1981; Devreotes et al. 1983). These authors demonstrated that optical density waves in a monolayer of cells matches closely with the concentration profile of cAMP measured by autoradiography. According to the estimate, the level of AMP in a monolayer of cells oscillates between 10 nmol/L and 1 μmol/L.

cAMP relay response in Dictyostelium

The amount of extracellular cAMP is determined by two processes: synthesis and secretion of intracellular cAMP and destruction of intracellular and extracellular cAMP by phosphodiesterases (PDEs). Upon binding of extracellular cAMP to the G-protein coupled receptor CAR1, cAMP is synthesized by membrane bound adenylyl cyclase ACA, which is transiently activated by PI3 kinase- and PKB-dependent pathways (Saran et al. 2002; Comer & Parent 2006; Cai et al. 2010; Charest et al. 2010). Before any of the molecular details were known, Steck, Devreotes and his co-workers revealed various aspects of the cAMP relay response by developing techniques to measure intracellular and secreted cAMP in cells under defined concentrations of exogenous cAMP (Devreotes & Steck 1979; Devreotes et al. 1979; Dinauer et al. 1980a,b,c; Theibert & Devreotes 1983). Their approach made use of a perfusion system that delivers cAMP stimuli of well-defined magnitude and duration and biochemical assays for cAMP. Although the measurements were made on cell populations, the series of experiments was a landmark. The studies made quantitative analyses on cell populations under various conditions to reveal the kinetics of cAMP synthesis, degradation and secretion.

One of the important properties that came to light from these and other studies is the excitable and adaptive nature of the cAMP relay response. The cAMP relay response is transient and appears as a pulse. The synthesis of cAMP and its secretion peaks at about 2 min after application of extracellular cAMP and attenuates within 3–10 min. Because perfusion was used to “clamp” the level of extracellular cAMP, the transient nature of the response is clearly not due to decrease in extracellular cAMP level by degradation. This is the so-called adaptation of cAMP relay response. In the adapted state, the dynamic range of the response curve shifts so that further increase in extracellular cAMP concentration is required to evoke cAMP relay response (Devreotes & Steck 1979). The other important feature is the process of the so-called “de-adaptation”; i.e., recovery of cAMP relay response from the adapted state. During this recovery period, cells show decreased responsiveness to the second stimulus. The magnitude of the response during the recovery process increased in first-order fashion, which take 3–4 min to recover the half-maximal response.

Despite these earlier efforts, questions remained regarding cAMP dynamics in Dictyostelium (Laub & Loomis 1998; Nanjundiah 1998; Goldbeter 2006; McMains et al. 2008). Are the cAMP oscillations observed in Dictyostelium a result of oscillations inherent at the single-cell level or are they population-level phenomenon? Our knowledge of cAMP relay response comes mainly from the combination of optical density measurement and biochemical assays for cAMP in cell suspension. Under such conditions, it is difficult to measure the amount of cAMP synthesized by the cells due to its degradation by extracellular phosphodiesterase secreted by the cells. Moreover, discriminating cAMP from cytosol and other intracellular pools is difficult if not impossible. When measuring cAMP relay response in population, eliminating endogenous cAMP synthesized by the cells is not easy even when perfusion is used to clamp the level of extracellular cAMP as we will describe in detail below. To circumvent these problems one would ideally like to measure cAMP relay response in single living cells in isolation.

Live-cell imaging of cAMP

In the early 1990s, Tsien and his co-workers developed the first FRET (fluorescence resonance energy transfer)-based sensor for cAMP named FlCRhR (pronounced “flicker”) (Adams et al. 1991). FRET is a quantum mechanical phenomenon whereby the excited-state energy of a fluorescence donor is directly transferred to a fluorescence acceptor in radiation-less manner when the donor and the acceptor are in close proximity and alignment. In FlCRhR, catalytic and regulatory subunits of PKA are chemically labeled with fluorescein and rhodamine, respectively (Adams et al. 1991). The regulatory subunit dissociates from the catalytic subunit upon binding of cAMP. Thus, an increase in cAMP concentration causes loss of FRET efficiency. Its use, however, was limited as purification and chemical labeling of the protein followed by microinjection to the cells were required.

To overcome these limitations, genetically encoded cAMP sensors were developed in which fluorescent proteins were fused to the regulatory subunit of PKA (Zaccolo et al. 2000; Zaccolo & Pozzan 2002). The first generation PKA-based sensors, were, however, slow-acting (Nikolaev et al. 2004). Also because the sensors consisted of two unique subunits each encoded by separate genes, variability in their expression level could hide the FRET signal in the background. To overcome these limitations, monomeric sensors based on a regulatory subunit of PKA, cyclic-nucleotide-gated channels (CNGC) or exchange protein directly activated by cAMP (Epac) have been developed (Nikolaev & Lohse 2006; Willoughby & Cooper 2008).

In Dictyostelium, Parent and her colleagues reported a cAMP sensor based on the cAMP-binding domain B from the mammalian PKA RII β-subunit (Bagorda et al. 2009). By live-cell imaging, they showed for the first time that there is rise in cytosolic cAMP upon application of extracellular cAMP. Due to its relatively high affinity to cAMP (EC50 < 600 nmol/L) signals obtained from the probe may have been saturated in some cases. Gregor et al. (2010) made use of a relatively low-affinity (EC50 ∼2.5 μmol/L) and fast responding (<2 s) cAMP sensor epac1-camps (Nikolaev et al. 2004) composed only of a single cAMP binding domain of human Epac. As described in detail below, epac1-camps allows detailed analysis of the cAMP relay response in Dictyostelium cells at both the single-cell and multi-cell levels.

cAMP relay response in single cells

By constitutively expressing epac1-camps under a strong promoter in Dictyostelium cells, it is now relatively easy to study cAMP relay response in single cells to well-defined concentrations and duration of extracellular cAMP stimuli (Gregor et al. 2010). When isolated cells are continuously exposed to the nanomolar level of extracellular cAMP by perfusion, cytosolic cAMP rises transiently within 1 min after addition of cAMP (Fig. 4b). This response attenuates during the next 15 min as it continues to oscillate at about 3–6 min periodicity. The cells therefore exhibit excitability to a supra-threshold level of extracellular cAMP and the response is not a single transient peak but of multiple peaks. At higher extracellular cAMP levels, the decay rate of the response slows down so that the response persists as long as extracellular cAMP is kept elevated (Fig. 4c). Although cells under such high extracellular cAMP concentrations appear autonomously oscillatory, the behavior depends strictly on the presence of extracellular cAMP. When extracellular cAMP is removed, cytosolic cAMP returns back to the basal level within 1 min (Gregor et al. 2010). Intracellular negative feedback loops (Fig. 1b) that quantitatively account for the excitatory and the oscillatory responses should be elucidated.

Although cAMP from intracellular pools other than the cytosol such as vesicles may also be secreted (Maeda & Gerisch 1977; Kriebel et al. 2008), a strong implication from the FRET observations is that the response observed in cytosolic cAMP is likely the main source of cAMP secreted as chemoattractant. The basal level of cytosolic cAMP estimated from FRET measurement is around 400 nmol/L to 1 μmol/L (Gregor et al. 2010). The peak of the cAMP response is around 5–20 μmol/L. Since these are close to concentrations of the total intracellular cAMP estimated from biochemical assays (Gerisch & Wick 1975), a large fraction of intracellular cAMP during this stage of Dictyostelium development must be in the cytosol. This is natural as ACA is predominantly found at the plasma membrane. A study that made use of electropermeabilization also indicates cytosolic origin of intracellular cAMP (Schoen et al. 1989). In intact cell populations, cytosolic cAMP is synchronized between the cells and appears in waves (Gregor et al. 2010). The pulse-shape of cytosolic cAMP changes show strong resemblance to those measured earlier for bulk intracellular and extracellular cAMP. It is therefore most probable that observed cytosolic cAMP is exported to serve as a chemoattractant. It is recently reported that exocytosis is not required for cAMP secretion (Zanchi et al. 2010) reinforcing the view that cytosolic cAMP is secreted by a transporter.

There are some differences between the cAMP relay response observed at the single cell-level (Gregor et al. 2010) and at the population-level. When cells are exposed continuously to 100 pmol/L to 10 μmol/L cAMP, the time required for the response to adapt is about 3–10 min in the population assays (Devreotes & Steck 1979) compared to more than 15 min when observed at the individual cell level (Gregor et al. 2010). The sustained oscillations observed at the single cell level upon exposure to extracellular cAMP above ∼100 nmol/L is not present in population-level assays. If we stimulate a group of cells and take their average response, we immediately see that the live cell imaging data are compatible with earlier observations. Only the first peak of the averaged response retains a comparable amplitude to that of the individual time course (Fig. 4d). Subsequent peaks are diminished when averaged because phases of the oscillations are not well synchronized between the cells (Fig. 4d). Thus, the decay rate estimated earlier was a sum of combined effects – the long time scale relaxation of the response at the individual cell level plus how fast phases of the persistent cAMP oscillations of individual cells desynchronize.

It is also worth noting that when well-isolated cells are continuously flushed with phosphate buffer, one sees that cytosolic cAMP remains at the basal level at almost all of the time. The occurrence of a transient pulse under such a condition is rare (Gregor et al. 2010). When extracellular cAMP is elevated to a picomolar range, we see more pulses, though still appearing randomly irrespective of the timing of stimulation. These observations strongly suggest that cAMP oscillations are properties that cannot be accounted for simply by oscillations of single cells. Single cells in isolation washed free of extracellular cAMP do not exhibit cAMP oscillation. The molecular mechanism behind such tight regulation of cAMP synthesis and secretion in Dictyostelium remains to be elucidated. Live-cell imaging of cAMP signaling should facilitate analysis of mutants defective in the adaptive pathway and clarify the essential negative feedback loop responsible for the oscillations.

The origin of collective cAMP oscillations

How collective oscillations appear in the population can also be studied under perfused conditions. Instead of applying cAMP exogenously, one could ask what happens when the number of cells is increased under constant flow of buffer without cAMP (Fig. 5a). It is known from earlier studies that the cells secrete cAMP constitutively at the rate of 0.34–1.4/min (Dinauer et al. 1980a; Van Haastert 1984). We can thus predict that adding cells will elevate the level of extracellular cAMP, and cells will react to it just as they did when stimulated in isolated conditions. It was shown that, in fact, cells in a flow chamber exhibit synchronized pulsing of cytosolic cAMP above a threshold density of cells (Fig. 5b) (Gregor et al. 2010).

image

Figure 5.  Transition to collective oscillations in Dictyostelium cell populations in a perfusion chamber. (a) Starved cells expressing the cyclic adenosine monophosphate (cAMP) probe epac1-camps are placed in a chamber (volume ∼0.25 mL) and fluorescence is observed under a microscope; buffer is well-mixed in the chamber and exchanged at a fixed flow rate of 2 mL/min. (b) Population averaged changes in cytosolic cAMP estimated from the ratio between fluorescence emission at 485 and 540 nm (y-axis). Cell densities are 1/4 monolayer (top panel) and 1/64 monolayer (bottom panel); the flow rate is 2 mL/min. (c) The occurrence of synchronized pulsing is plotted as a function of cell density ρ normalized by the decay rate k of extracellular cAMP, ρ/k is roughly proportional to extracellular cAMP concentration. Cells are quiescent below a critical value of ρ/k. (d) Model simulations of a cell population demonstrate build-up of stochastic pulsing by individual cells (dots) before synchronized pulses occur. Upper panel: time course of 1000 representative cells from a mode simulation is displayed along y-axis according to their cell-index; x-axis is time. The percentage of pulsing cells (black line, lower panel) increases as the extracellular cAMP concentration (red line) rises. This effectively promotes further firing by a larger fraction of cells. The chain reaction finally gives rise to synchronous firing of the entire population (7 and 16 min).

Download figure to PowerPoint

A qualitative change from a quiescent to oscillating state is observed at the level of individual cell level when collective oscillations emerge. This indicates that collective oscillations in Dictyostelium appear via DQS (Fig. 3c). Pulsing is sporadic with non-regular intervals when the ratio of cell density and flow rate is small. Its average frequency depends on this ratio until it plateaus around 5–6 min where pulsing becomes more regular in timing (Fig. 5b and c) (Gregor et al. 2010). It is expected that the extracellular cAMP level in the perfusion chamber rises as the number of cells increases. Conversely, it is diluted according to the flow rate. As we saw from the properties of individual cells, the sub-threshold level of extracellular cAMP elevates the chance of cells to pulse. These random excitations could begin to coincide because cAMP is secreted. The self-excitation driven by stochastic events could be examined in detail by a mathematical model (Lindner et al. 2004; Gregor et al. 2010). When there is a sufficient number of cells in a chamber, cells that are excited together in this way can elevate extracellular cAMP to a nanomolar range, which will evoke a global pulsing, i.e., almost all cells in the chamber fires (Fig. 5d). Therefore, collective oscillations in Dictyostelium do not appear by Kuramoto-type transition. Here, it is not the amplitude but the frequency that encodes the density information.

Future prospect

  1. Top of page
  2. Abstract
  3. Introduction
  4. Collective oscillations in microbial populations
  5. cAMP oscillations and waves in Dictyostelium
  6. Future prospect
  7. Acknowledgments
  8. References

As we saw from examples in synthetic circuits, yeast and Dictyostelium, information of cell population is encoded in the properties of the oscillations differently depending on whether individual cells exhibit oscillations or excitability. In the case of coupled oscillations in the synthetic circuit (Danino et al. 2010) and circadian rhythms (Ukai et al. 2007), cells appear to be able to oscillate independently of other cells. In such cases, synchronization of individual oscillations is the key to collective oscillations. The amplitude reflects the degree of synchrony and is able to encode density and other information regarding the extracellular environment. In yeast and Dictyostelium, oscillatory instability in single cells does not surface unless a sufficient number of cells is present in the vicinity. Critical concentration of secreting molecules (or cell density) is required at the onset of a synchronized pulse. Since the occurrence of such events increases as the concentration of signaling molecules increases, the frequency of collective oscillations encodes density information. In yeast, glycolytic oscillations at the level of individual cells and the population level seem to appear simultaneously at a critical cell density.

In Dictyostelium, oscillations at the level of population emerge first at low density before independent oscillatory instability at the single-cell level appears at a higher density. During aggregation, synchronized pulses of cAMP first occur sporadically and gradually increase in frequency to ∼5 min (Gregor et al. 2010). This is not only due to cell density change but also due to increase in CAR1, PDE and ACA that are auto-induced by the collective oscillations (Goldbeter 1996; Sawai et al. 2005). It would be interesting to see whether expression of these genes is frequency-dependent and whether the frequency encodes information vital for their survival such as genetic heterogeneity of the population or nutrient availability of the surrounding environment.

We should note that, as we saw in the case of the BZ reaction, the above two schemes can be realized by the same reaction depending on the autonomy of individual cells’ oscillation and coupling between the cells. It is tempting to speculate that many developing cells make use of the two types of transitions depending on the roles that transient pulses and oscillations play in the group. Appearance or loss of group-level behavior in systems where autonomy of oscillations has a significant role will naturally be following the Kuramoto-type transition. Relation between asynchrony and cell-fate determination may be of relevance in this regard. In embryonic stem cells, with the exception of neighboring daughter cells that are synchronized, Hes1 level fluctuates asynchronously (Kobayashi et al. 2009). This gives rise to a broad distribution in the level of Hes1 that can act as a source of divergent response to a common differentiation signal (Kobayashi et al. 2009; Kobayashi & Kageyama 2010). In Dictyostelium, asynchrony may also have a role in cell differentiation. The oscillations and waves of cAMP are observed late into development (Dormann et al. 1998). Cell differentiation that marks progression of distinct developmental stages is thought to be mediated by extracellular factors that regulate intracellular PDE (RegA) activity via receptor histidine kinases (Thomason et al. 1999, 2006; Anjard et al. 2009). In this light, it is interesting to note that cAMP oscillations become less synchronized in null mutants of regA (Gregor et al. 2010). Live cell imaging of cAMP and associated gene regulation during later stages of Dictyostelium development should be instrumental in exploring such possibilities.

As in the case of Dictyostelium, oscillations that arise from DQS may be advantageous for decision-making at the group-level, since it provides an all-or-none switch for a collective behavior. A density-dependent transition in cellular states is known as the community effect in animal development (Standley et al. 2001) and quorum sensing in bacterial populations (Waters & Bassler 2005). Although it is often assumed that the level of an inducing signal is steady in time, it is equally possible that the signaling molecules are oscillating in some cases. The two essential features – nonlinear kinetics and negative feedback loops – are prevalent in cell signaling. Live cell imaging and quantitative analyses in both time and space shall further clarify dynamic and cooperative properties of density-dependent phenomena. Knowing their common design principles and developing means to alter collectivity of the cells should also be instructive for future stem cell and tissue engineering.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Introduction
  4. Collective oscillations in microbial populations
  5. cAMP oscillations and waves in Dictyostelium
  6. Future prospect
  7. Acknowledgments
  8. References

This work was supported by Human Frontier Science Program (RGY 70/2008), JSPS Grant-in-Aid for Young Scientists (A) (22680024) and MEXT Grant-in-Aid for Scientific Research on Innovative Areas (23111506) to SS. KF is supported by Osaka University Life Science Young Independent Researcher Support Program (Japan Science and Technology Agency).

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Collective oscillations in microbial populations
  5. cAMP oscillations and waves in Dictyostelium
  6. Future prospect
  7. Acknowledgments
  8. References