Monoclonal cells grown under identical conditions display strikingly heterogeneous behavior. Some of this heterogeneity may be caused by the inherent probablistic nature of biochemical reactions inside cells. The majority of cell-to-cell variability, particularly in complex cellular activities and phenotypic states, is however not random, but forms quasi-deterministic patterns that are regulated. 

The reason for this is that cells do not operate in isolation, but create heterogeneous social contexts, to which they adapt their phenotypic behaviour . This is true for single-cell organisms as well as for cells from multicellular organisms. The effect of this type of cell-to-cell variability on shaping the phenotypic spectrum of single cells has major consequences for how we study cellular processes and interpret molecular mechanisms and activities in single cells . It also shows that basic social properties of mammalian cells can be studied in in vitro experimental systems using cells grown in culture.

Screen Shot 2013-04-15 at 09.39.27

Explaining and using regulated cell-to-cell variability. This example explains how cell-to-cell variability in infection probability by a simple virus (SV40) in a population of tissue culture cells occurs. a. Histogram of single-cell SV40 infection probabilities (P(infected); x-axis), as measured in 2.6 million individual HeLa cells. Neither median nor mean probabilities (indicated by dashed lines) are a good representative of the measured single-cell probabilities. the coefficient of variation (CoV), a common dimensionless noise measure calculated by dividing the mean by the standard deviation, is 0.78 for the probability of infection. b. The observed cell-to-cell variability can be largely explained by a single predictor, in this case cell size (nucleus area). Importantly, in this example, SV40 does not induce an increase in cell size. the local average probability of infection (y-axis, blue line and circles) is plotted against nucleus area (x-axis). Note that the y-axis in this graph is the x-axis of the graph in a and that the infection probability distribution of a is plotted as a function of nucleus area (cell size). Dark and light grey regions indicate, respectively, the ×0.5 and ×1 local standard deviations of infection probabilities. A fitted Hill function (red line, adjusted to different minima and maxima) reveals a Hill coefficient of >1, which indicates strong switch-like behaviour of SV40 infection probability in HeLa cells as a function of cell size. the average remaining CoV is 0.31. c. When considering more predictors, the noise can be further reduced. the same single-cell SV40 probabilities of infection are now plotted as a function of both nucleus area and local cell density. the average CoV is further reduced to 0.21. the colour corresponds to the probability of infection. d. Bayesian network structure inference performed on single-cell data reveals part of the underlying molecular network that determines the regulated cell-to-cell variability observed in SV40 infection, including glycosphingolipid monosialotetrahexosylganglioside (GM1) levels at the plasma membrane and activated focal adhesion kinase (FAK). the displayed circuitry contains a coherent feedforward loop, which displays synergy to cancel out intrinsic noise while amplifying regulated cell-to-cell variability.

 

Current questions:

  • What are the molecular mechanisms by which single cells adapt their membrane lipid composition, the internalization, recycling and degradation of membranes, and their signalling capacity to their social context ?
  • Do multivariate measurements of the social context of single cells allow predictive quantitative modelling of single-cell activities over time?
  • Does a generalizable principle exist in how a phenotypic spectrum of single cells emerges from social behaviour?
  • Can we create a bottom-up phenomenological modelling framework that predicts and simulates cell-to-cell variability of cell growth, position in the cell cycle, and cell shape, as well as the feedbacks between these properties?
  • Can such models be used to obtain quantitative insights in how the spectrum of cell-to-cell variability in transcription, molecular activities, and phenotypic states emerges during growth of a cell population and to predict these during perturbations?

 

Current lab members involved:

Prisca Liberali

Mathieu Frechin

Gabriele Gut

 

Some relevant publications from the lab: