Deciphering multi-state mobility within single particle trajectories of proteins on the plasma membrane

Single particle tracking is a powerful technique often used in the study of dynamic mechanisms on the cell surface such as binding, confinement and trafficking. Experimental trajectories can be used to detect changes in the lateral mobility of individual molecules over time and space. Therefore, a potential problem in the analysis of single particle trajectories is to account for transitions between modes of mobility. Here we present two coupled statistical methods which characterize particle mobility that is temporally and spatially heterogeneous. The first method detects periods of drift diffusion or reduced mobility within single trajectories due to transient associations with other biomolecules. The second locates spatial domains which have higher or lower concentrations of these associating molecules. The trajectory is modeled as the outcome of a two-state Hidden Markov model parameterized by the diffusion coefficients and drift velocities of each state and the rates of transitions between them (which may change in space). Transitions between states arise from association and disassociation with a binding partner, either membrane-associated or cytosolic. These associations lead to either reduced Brownian diffusion or drift diffusion. An adapted Markov chain Monte Carlo algorithm was used to optimize parameters and simultaneously select the most favorable model of lateral mobility (transient reduced mobility or transient drift diffusion) and to locate spatial domains. Analysis of simulated particle tracks with a wide range of parameters successfully distinguished between the two models, gave accurate estimates for parameters and accurately located spatial domains.