Sender-receiver systems and applying information theory for quantitative synthetic biology

Abstract

Sender-receiver (S-R) systems abound in biology, with communication systems sending information in various forms. Information theory provides a quantitative basis for analysing these processes and is being applied to study natural genetic, enzymatic and neural networks. Recent advances in synthetic biology are providing us with a wealth of artificial S-R systems, giving us quantitative control over networks with a finite number of well-characterised components. Combining the two approaches can help to predict how to maximise signalling robustness, and will allow us to make increasingly complex biological computers. Ultimately, pushing the boundaries of synthetic biology will require moving beyond engineering the flow of information and towards building more sophisticated circuits that interpret biological meaning.

Graphical abstract

Figure 1

Sender–Receiver (S–R) systems occur at all levels of biology. (a) The schematic visualises these layers of communication as a multi-level jigsaw puzzle, working between factors such as proteins, cells and organs. In principle, Shannon's information theory can be used to quantify the information flow in all such S–R systems. The Proteins image, Alpha-Amanitin–RNA polymerase II complex, is licensed under Public domain via Wikimedia Commons. (b) Inputs from the sender (S) vary in terms of dose (e.g. chemical concentration), type of biomolecule (e.g. AHL, volatile Aldehyde, Dopamine, or even DNA fragments) and the rate of production. The channel is the medium of information transfer. The channel capacity C (measured in bits per second) is modulated by the equation shown; bandwidth B is the range of frequency allowed by the channel (the change in concentration of molecules; Hz) and S and N are signal and noise respectively. The receiver (R) mediates signal reception via cognate receivers like cell surface receptors. A modulation system like a cell signalling pathway links the signal to the interpreter (e.g. a responsive promoter for gene expression) resulting in extraction of the ‘meaning’ in the signal. The outputs, such as gene expression, are measured relative to space, time and input dose responses.

Figure 2

An information channel within a single protein. (a) A recent study [3^••] visualised the communication channel within the Fyn SH2 domain, linking a peptide binding site (blue), and an SH2/SH3 connecting loop for Fyn kinase (green). The change in mutual information upon binding is measured between each pair of residues (white nodes). Adjacent residue pairs with significant changes in mutual information are represented as black or red lines. The largest changes, shown in red, are observed between residues forming a connected path from the peptide binding region to the connecting loop region. Thus, information theory reveals the major communication path. (b) and (c) are two structural views, highlighting the positions of the residues involved in the binding region and loop region (b) or the communication channel (c). The peptide (including the phosphorylated tyrosine) is in dark purple (top), the peptide binding site is in blue and the connecting loop is in green (bottom). Images kindly provided by Dr. Jesper Ferkinghoff-Borg and Dr. Joost Schymkowitz.

Figure 3

An systematic ‘network atlas’ approach finds all 3-node receiver networks that respond to a sender morphogen gradient by making a central stripe [37^••]. (a) ∼3000 transcription activator-repressor networks were explored computationally, using 100,000 parameter sets, to see which would form stripes in response to a monotonic gradient of arabinose (ara). The green node (GFP) had to be OFF at high and low concentrations of arabinose (input to the red node) and ON at middle concentrations. The resulting 109 solutions (grey nodes) are organised by relative complexity, with four ‘stalactites’ indicating the minimal mechanisms for stripe formation (large circles). These mechanisms are all incoherent feedforward loops (I1,I2,I3,I4) and can be reduced even further to an archetypal 2-node network, I_zero (I₀). (b) All minimal networks were constructed synthetically in E. coli. Lawns of bacteria on Petri dishes were tested against morphogen gradients from central paper disks (senders containing arabinose; white circles). All networks successfully exhibited stripe behaviour (green GFP rings). Importantly, the networks use distinct mechanisms and stripe-forming dynamics (i.e. they cannot be interconverted into each other merely by altering rate constants, etc.). The approach demonstrates the stripe forming capability of the entire incoherent network family.