Environmental sensing, information transfer, and cellular decision-making
Graphical abstract
Introduction
Life for single cells is stochastic [1]. Cells sense fluctuating signals with biochemical networks that are themselves stochastic and history-dependent [2], and yet living organisms are able to flourish in nearly all environments. Understanding how cells prosper despite stochasticity and environmental variability is the focus of a relatively new area of systems biology, that of cellular decision-making 3, 4. By a cellular decision we mean the process by which a cell makes a ‘choice’ of phenotype from a range of possible phenotypes in response to or in anticipation of extracellular change. Such choices could include new gene expression, changes in cell morphology, intracellular re-arrangements, movement, or the option not to change phenotype at all.
In addition to the stochasticity of signal transduction, cells locally sense signals that fluctuate both in time and across space, whereas often it is the successful identification of broader environmental changes that is important in enabling an effective response [5]. Even bacteria appear to be able to solve this kind of inference problem, using local signals to identify, for example, that they are in the human gut and thereby anticipate likely future events 6, 7.
Here we review the theoretical approaches developed so far to understand cellular decision-making. Motivated by the surge of interest in biochemical stochasticity generated by the theoretical work of McAdams and Arkin in 1997 [8], we ask if theory is now poised to have a similar effect on the experimental study of decision-making in single cells.
Section snippets
Dose–response and information theory
Most theorists have focused on applying ideas from information theory, often inspired by neuroscience [9]. In systems biology, the experimental confirmation that gene expression is stochastic 10, 11 and the related discovery that genetically identical cells can vary significantly in their response to the same stimulus 12, 13, 14 implies that dose–response, or ‘input–output’, relationships are also often substantially stochastic. Information theory, through mutual information, provides an
Updating inferences over time
We have argued that higher mutual information between an input and output implies better inference of the input from the output. Mutual information has so far typically been quantified at individual time points or at steady-state but cells may well update their inferences over time as they sense and learn more about their environment [35]. Sharpening inferences over time in this fashion is referred to as sequential inference in the statistics literature. Any posterior distribution given all the
Optimality theory
Cells not only make inferences about the state of the environment but also make decisions [43], such as whether to differentiate or not, which affect the fitness of the cell. Mutual information is a generic approach implying a certain cost or ‘scoring’ for evaluating inferences (Box 1), and therefore does not suggest a cost function related to a specific cellular decision. Nevertheless it is possible to show (equating fitness with the expected long term growth rate and under certain additional
Conclusions
What insights does the above theory provide for experimental studies? Perhaps the most successful use of information theory to date has been the investigation of development in Drosophila. Nearly all the information needed to describe the patterning of the fly embryo is established early in development by the gap genes [26••]. We do not yet know whether this result is fundamental or if other organisms can reach maturity by establishing positional information later. In signal transduction, the
Acknowledgements
We thank Ilya Nemenman, Pieter Rein ten Wolde, Margartis Voliotis, and especially Gasper Tkacik for helpful discussions.
References (58)
- et al.
Cellular decision making and biological noise: from microbes to mammals
Cell
(2011) - et al.
Signaling to ERK from ErbB1 and PKC: feedback, heterogeneity and gating
J Biol Chem
(2013) - et al.
Probing the limits to positional information
Cell
(2007) Information processing by biochemical networks: a dynamic approach
J R Soc Interface
(2011)- et al.
Functional roles for noise in genetic circuits
Nature
(2010) - et al.
Origins of regulated cell-to-cell variability
Nat Rev Mol Cell Biol
(2012) - et al.
Strategies for cellular decision-making
Mol Syst Biol
(2009) - et al.
Noisy information processing through transcriptional regulation
Proc Natl Acad Sci USA
(2007) - et al.
Predictive behavior within microbial genetic networks
Science
(2008) - et al.
Adaptive prediction of environmental changes by microorganisms
Nature
(2009)