Title:

Concentration oscillations in single cells : the roles of intracellular noise and intercellular coupling

Concentration oscillations are a ubiquitous characteristic of intracellular dynamics. The period of these oscillations can vary from few seconds to many hours, well known examples being calcium oscillations (seconds to minutes), glycolytic oscillations (minutes) and circadian rhythms (1 day). Considerable advances into understanding the mechanisms and functionality of concentration oscillations have been made since glycolytic oscillations were observed in the late 1950s, and mathematical methods have been an essential part of this process. With increasing ability to experimentally measure oscillations in single cells as well as in cell ensembles, the gold standard of modelling is to provide tools that can elucidate how the systemwide dynamics in complex organisms emerge from a system of single cells. Both abstract and detailed mechanistic models are complementary in the insight they can bring, and for networks of coupled cells considerations such as intrinsic intracellular noise, cellular heterogeneity and coupling strength are all expected to play a part. Here, we investigate separately the potential roles played by intrinsic noise arising from finite numbers of interacting molecules and by coupling among cellular oscillators. Regarding the former, it is well known that internal or molecular noise induces concentration oscillations in chemical systems whose deterministic models exhibit damped oscillations. We show, using the linearnoise approximation of the chemical master equation, that noise can also induce oscillations in biochemical systems whose deterministic descriptions admit no damped oscillations, i.e., systems with a stable node. This nonintuitive phenomenon is remarkable since, unlike noiseinduced oscillations in systems with damped deterministic oscillations, it cannot be explained by noise excitation of the deterministic resonant frequency of the system. We here prove the following general properties of stablenode noiseinduced oscillations for systems with two species: (i) the upper bound of their frequency is given by the geometric mean of the real eigenvalues of the Jacobian of the system, (ii) the upper bound of the Qfactor of the oscillations is inversely proportional to the distance between the real eigenvalues of the Jacobian, and (iii) these oscillations are not necessarily exhibited by all interacting chemical species in the system. The existence and properties of stablenode oscillations are verified by stochastic simulations of the Brusselator, a cascade Brusselator reaction system, and two other simple chemical systems involving autocatalysis and trimerization. We also show that external noise induces stable node oscillations with different properties than those stimulated by internal noise. Having demonstrated and explored this nonintuitive effect of noise, we extend the work to investigate the phenomenon of noise induced oscillations in cellular reaction systems characterised by the ‘bursty’ production of one or more species. Experiments have shown that proteins are typically translated in sharp bursts and similar bursty phenomena have been observed for protein import into subcellular compartments. We investigate the effect of such burstiness on the stochastic properties of downstream pathways by considering two identical systems with equal mean input rates, except in one system molecules (e.g., proteins) are input one at a time and in the other molecules are input in bursts according to some probability distribution. We find that the stochastic behaviour falls in three categories: (i) both systems display or do not display noiseinduced oscillations; (ii) the nonbursty input system displays noiseinduced oscillations whereas the bursty input system does not; (iii) the reverse of (ii). We derive necessary conditions for these three cases to classify pathways involving autocatalysis, trimerization and genetic feedback loops. Our results suggest that single cell rhythms can be controlled by regulation of burstiness in protein production. We go on to investigate roles played by intercellular coupling in whole organlevel oscillations with an experimental analysis of circadian rhythms in Arabidopsis thaliana †. Circadian clocks in animals are known to be tightly coupled among the cells of some tissues, and this coupling profoundly affects cellular rhythmicity. However, research on the clock in plant cells has largely ignored intercellular coupling. Our research group used luciferase reporter gene imaging to monitor circadian rhythms in leaves of Arabidopsis thaliana plants, with both a lower resolution, high throughput method and a highresolution (cellular level), lower throughput method. Leaves were grown and imaged in a variety of light conditions to test the relative importance of intercellular coupling and cellular coupling to the environmental signal. We analysed the high throughput data and described the circadian phase by the timing of peak expression. Leaves grown for three weeks without entrainment reproducibly showed spatiotemporal waves of gene expression, consistent with intercellular coupling. A range of patterns was observed among the leaves, rather than a unique spatiotemporal pattern, although some patterns were more often observed. All of the measured leaves exposed to lightdark entrainment cycles had fully synchronised rhythms, which could desynchronise rather quickly when placed in a nonentraining environment (i.e., constant light conditions). After four days in constant light some of these leaves were as desynchronised as leaves grown without any rhythmic input, as described by the phase coherence across the leaf. The same fast transition was observed in the reverse experimental scenario, i.e., applying lightdark cycles to leaves grown in constant light resulted in full synchronisation within two to four days. From these results we conclude that singlecell circadian oscillators were coupled far more strongly to external lightdark cycles than to the other cellular oscillators. Leaves did not spontaneously completely desynchronise, which is consistent with a presence of intercellular coupling among heterogeneous clocks. We also developed a methodology, based on the notion of two functional spatial scales of expression across the leaf, to analyse the highresolution microscope data and determine whether there is a difference in the phase of circadian expression between vein cells and mesophyll cells in the leaf. The result from a single leaf suggests that the global phase wave dominates the phase behaviour but that there are small delays in the veins compared to their nearby mesophyll cells. This result can be validated by applying the methodology developed here to new microscope leaf data which is currently being collected in the research group. † This work was performed as a collaboration between David Toner (DT) and Benedicte Wenden (BW). BW designed and carried out the experiments, DT performed the data analysis and led on data visualisation.
