In this paper, we used experimental perturbation and kinetic simulation to test whether the two cell fate-determining processes induced by mitotic arrest, i.e. mitotic death and mitotic slippage, occur by pathways that are, or are not, mechanistically coupled and/or kinetically correlated. Our data of slippage kinetics when death was blocked, and death kinetics when slippage was blocked ( and ), argue against mechanistic coupling. A stronger and more quantitative test of the mechanistic independence is to ask if the kinetics of one pathway are altered by blocking or accelerating the other. We could make these direct comparisons only experimentally for death in the death-sensitive lines, and slippage in the death-resistant lines. Indeed, the kinetics of one pathway were unaffected by blocking the other in these cases ( and ). A direct comparison of kinetics with and without blocking the alternative pathway was more generally possible for the small subset of cells that undergo the less prevalent pathway when both are operational. Again, we observed no effect of blocking one pathway on the kinetics of the other, but this is a weaker test, since it includes only a subset of cells that might not be typical. In addition, we also observed that accelerating caspase activation had no effect on the kinetics of cyclin B1 proteolysis in single cells (), further substantiating the independence of the slippage and death pathways.
A subtler test of pathway independence is to ask if two pathways are kinetically correlated in individual cells. Kinetic correlation is likely if pathways share upstream events, but could occur for other reasons, for example if the kinetics of both pathways depend on some common external variables, such as cell size, protein synthesis rate etc. We approached this question using kinetic simulation, asking how well the rates of each pathway alone, measured when the other is blocked, predict cell behavior when both pathways are active. We observed a good fit of simulated to experimental data assuming uncorrelated kinetics (). A simulation of correlated kinetics fit the data less well in all the cell lines, though the distinction is not sufficiently large to completely rule out some degree of kinetic correlation. Overall, our observations strongly favor the independent pathway model in . We conclude that the large inter- and intra-cell line variability in response to anti-mitotic drugs is due to stochastic competition between two mechanistically and kinetically independent pathways, both of whose kinetics can be approximated by a gradual change in some parameter until a threshold is crossed, subsequently triggering an irreversible change that blocks the other pathway. In the exit pathway the gradual change is in Cyclin B concentration, in the death pathway its biochemistry is not yet known.
An interesting consequence of independent pathway competition is that fairly small changes in kinetics of either pathway can cause large changes in average cell fate. Average time for execution of the slippage and death pathways both varied ~2.5 fold across the 4 cell lines that we studied. However, this relatively small rate difference is enough that in a cell line where death is faster than slippage, >90% of cells die during mitotic arrest (HeLa), while in a cell line where slippage is faster than death, <10% of cells die (MCF7 and A549). Large variation in average cell fate despite relatively small variation in pathway may explain why the cell death response to anti-mitotic drugs varies strongly across cancer cell lines 
, and why individual cells within a line can exhibit different fates. It may contribute to response variation in anti-mitotics seen in the clinic 
, and might also help explain why clinical resistance seems to arise rapidly to current anti-cancer drugs, including anti-mitotics.
So what is the mechanistic source of the rate variation we observe? We know from experimental measurements that most proteins vary up to 2-fold in concentration between individual cells in clonal populations of human cancer cells in culture, presumably due to stochasticity in gene expression and protein degradation 
. Spencer et al argued that this kind of variation is sufficient to explain intra-line variation in response to TRAIL 
. In the case that we study, noise in protein degradation and protein phosphorylation are likely to be key factors in causing cell-to-cell variation, as mitosis is a hyper-phosphorylated state where transcription is silenced and translation significantly attenuated. Protein degradation rate is known to control slippage kinetics 
. It might also control the kinetics of apoptosis activation, though since our data favors uncorrelated kinetics for the two pathways, it would likely be proteolysis regulated in a different way. Cell-to-cell variation in proteolysis rates is an important topic for future research.
Many studies have revealed variation in sensitivity to apoptosis between cancer cell lines and investigated their origins 
, and it is easy to imagine that cancer clones are selected for decreased apoptosis sensitivity. Perhaps surprisingly, we observed just as much inter-cell line variation in slippage rates as we did in apoptosis rates (~2.5 fold in both cases), though we know of no reason drug-naïve cancer cells should be selected for altered slippage rates. Could this degree of variation in average rate between clonal cell lines be typical for complex cellular pathways? Our study, along with other recent work 
, shows that intra-line variation in pathway rates can cause significant variation in cell fate in response to drug, which presumably contributes to the difficulty of completely eradicating cancer using drug treatments.