Transforming the understanding, mathematical analysis and characterisation of cell cycle kinetics across the biosciences

Characterising cell cycle dynamics is fundamental to our understanding of biology including in cancer, immunology, developmental biology, toxicology, within-host infectious disease dynamics, ageing science and tissue engineering. In the past decade, genetically encoded cell cycle probes have revolutionised proliferation assays in cells and model organisms - allowing live imaging of cell cycle dynamics. RLM’s group previously developed two widely adopted biosensor systems based around Fucci (Fluorescent Ubiquitination-based Cell Cycle Indicator). We have now developed the most sophisticated probeset to date (Quiescence ubiquitination based cell cycle indicator - Qucci) allowing the discrimination of G0, G1, S and G2/M cell cycle phases in cells and mice. However, the impact of these models is hindered because no unbiased mathematically sound methodology exists to extract cell cycle phase lengths, cell cycle times or their distributions from a population of cells using any existing assay. Existing methods of cell cycle analysis rely on either: 1) live imaging of complete cell cycles or the transitions between cell cycle phases; 2) quantification of cell numbers over time to calculate a doubling time or; 3) derivation of the cell cycle time by pulse labelling of S-phase to calculate its length followed by upscaling to the entire cell cycle (classical Thymidine analogue studies). We have implemented method 1 and 2 and extended method 3 to Fucci/Qucci based live imaging data and performed extensive mathematical modelling. Our modelling reveals that; method (1) suffers from both spatial and temporal censoring (unintentional exclusion of some cell subpopulations with particular characteristics). For example, in populations of moving cells, subpopulations with longer cycle times are less likely to remain in the field of view and divide during the imaging period; method (2) suffers from incorrect assumptions on the distribution of cell cycle lengths. Our modelling has shown that populations of cells simulated with the correct distribution will grow more slowly than those with an assumed underlying exponentially distributed cell cycle time but with the same mean. Consequently, this will lead to mischaracterisations when cell cycle parameters are inferred from growth curves; method (3) is flawed because it employs the assumption that the number of cells in each cell cycle phase should be proportional to the length of that phase, which our modelling and experiments have revealed to be incorrect. We now understand the functional form of the underlying distribution of cell cycle times, which will allow us to correct both for spatial and temporal censoring bias and to account for the non-uniformity in the distribution of cells throughout the cell cycle. In this project we will develop the mathematical framework required to correctly calculate both inter-division times and the times of different phases of the cell cycle using any one of these methods. With a focus on live imaging methods, including Fucci/Qucci, we will deliver an open source software package for Imagej/Fiji designed for non-specialist end users. Our software will incorporate pre-trained neural networks able to classify the cell cycle phases and to automatically extract cell inter-division/cell phase times from time-lapse data generated from legacy Fucci models and Qucci. Using our mathematical framework, the tools we develop will be able to characterise cell cycle times and distributions for a given dataset. Delivering these analysis tools with the ‘gold standard’ Qucci models will have a transformative impact on the biosciences.