Information processing in mammalian cells
Signal transduction is the process by which signals impinging on a cell (for us, usually chemical signals) are converted into cellular behaviors, often initiated by programs of gene transcription. Historically, signalling has been thought of as some kind of pipe that conveys information from the outside of a cell to the inside, with the essential decision making taking place at the level of genetic regulatory networks. However, the molecular networks underlying signalling are extraordinarily complex, with enormous internal state, spatial and temporal localization, multi-protein complexes and interlinked positive and negative feedback loops. Instead of being just a pipe, it seems that signal transduction networks are undertaking complex information processing. How do we characterize this?
Broadly speaking, we take two approaches. First, from the “inside out”, we study particular molecular mechanisms found in signalling networks. A major interest is in protein post-translational modification (PTM). Many signalling proteins are modified in multiple ways on multiple sites, leading to a potentially enormous range of combinatorial molecular states. We are developing biophysical methods (mass spectrometry and nuclear magnetic resonance spectroscopy) for accurately measuring these states, at least for proteins with small (< 10) numbers of sites [Prabhakaran et al. 2011]. We also use what we call “systems biochemistry” to understand how multiple forward and reverse enzymes collectively regulate the distribution of PTM states. The combinatorial complexity in PTM is a particular challenge for mathematical modelling. Simulations require parameters to be specified in advance, especially the number of modification sites, making it hard to see the biological wood for the molecular trees. We have been developing analytical methods for overcoming this, at least in the steady state, [Thomson and Gunawardena 2009a, 2009b; Manrai and Gunawardena, 2008]. These methods rise above the “parameter problem” and have proved to be widely applicable to cellular mechanisms other than PTM.
The second approach is from the “outside in”. This arises from a different perspective, which suggests that the complexity found inside cellular networks cannot be fully explained by just studying the networks. Instead, the internal complexity reflects the external complexity of the environments in which those networks were evolved. We are developing new types of microfluidic devices to subject cells to complex external environments, using various forms of microscopy to assay their downstream behaviours at a single-cell level and exploiting mathematical models of the signalling networks to predict what we should see. We have developed flexible computational infrastructures (“little b” and, more recently, “Proteus”) to support this kind of modelling, [Mallavarapu et al. 2009]. Instead of studying networks by pulling them to pieces, as in the first approach, we try to ask them more complex questions, so that their answers are more informative about why they are wired up as they are.
One of the most awkward problems in studying signalling is that each cell is an individual entity and may respond to a signal differently from its neighbours, even in a clonal population. This cell-to-cell variation is obscured by population averages like Western blots. We try and use the variation to glean more information about the networks but cell culture is such an artificial environment that the variation is hard to interpret biologically. We are now moving towards more physiological contexts, including embryonic stem cells.
Signalling in eukaryotic cells is a fascinating area in which ideas from mathematics, physics and engineering are starting to provide new conceptual insights into biology. There are no shortage of hard questions for those who enjoy prospecting in unexplored territory.