High-dimensional dynamics and stochasticity in mammalian cell fate commitment

Sui Huang
Vascular Biology Program and Department of Surgery
Children's Hospital, Boston


Development and tissue homeostasis in multicellular organism requires that cells switch between discrete phenotypes, such as proliferation, quiescence, or differentiation into a particular cell type, such as a stem cell, a liver cell or a nerve cell. If each such cell fate is defined by a gene expression profile consisting of ten thousands of genes, and if gene expression is subjected to random fluctuations, why are there, in a first place, discrete, mutually exclusive cell fates - and mot a vast continuum of cell phenotypes ? In fact, the commitment of stem or precursor cells to a particular cell fate is often stochastic, yet they choose among a limited, predefined repertoire of distinct fates, and reliably undergo all-or-none transitions between these discrete states. How do cells unite robustness with randomness ?

Traditional molecular biology would explain the cell type specific expression profile in which some genes are stably expressed, while others are stably repressed, with "epigenetic" chromatin modifications, such as covalent modifications of histones which stably determine the transcriptional activity of genes. But the enzymes performing these reversible molecular changes of histones themselves need to be regulated by upstream regulators which also have their own upstream regulators, etc. In other words, in strict epistemology terms, molecular biology in such cases, be it qualitative or quantitative, is not explanatory. It just "blames another level" in a chain of causality. A systems biology of cell regulation, other than embracing mathematical modeling of pathways and high-throughput characterization, will therefore also have to deal with a self-consistent, "closed" explanation for the existence of stable cell phenotypes.

In this "theory lunch" I will discuss various theoretical concepts that naturally explain the fundamental phenomenon of cell differentiation, and present results from experiments aimed at their validation. One idea is that the genome-wide gene regulatory network that controls cell behavior must impose constraints onto the global dynamics of the genomic regulatory network, such that only relatively few stable network states can arise, and that such high-dimensional "attractor states" in the gene expression state space represent the phenotypic cell fates. I will discuss experimental strategies and results based on dynamic gene expression profiling of the switch from the promyelocytic precursor state to the differentiated state of neutrophils, which suggest that the latter is indeed a high-dimensional attractor state in the gene expression state space of the genomic regulatory network.

In addition, when studying the changes of gene expression at the single-cell level rather than as averages over a cell population (as is the case with DNA microarrays) we find that a population of clonal, nominally identical cells is highly heterogenous, in that individual cells "switch " with respect to individual genes from the expression status of the precursor state to that of the differentiated state in a discrete but asynchronous and stochastic manner. This stochasticity with respect to multiple discrete switches explains the observed epigenetic heterogeneity of cell populations and is consistent with control by a high-dimensional, non-linear network.

Taken together, the data support a model in which a cell fate transition, such as differentiation, is driven by stochastic processes which are deterministically channeled by the state space structure of a high-dimensional system. This provides stability of phenotypes, while allowing flexibility for development, and also produces some stochastic dispersion within a clonal population. If time permits, I will discuss current work aimed at evaluating the consequences of such behavior for stem cell fate decision and mutation-less steps in tumorigenesis.


B de Bivort, S Huang, Y Bar-Yam, "Dynamics of cellular level function and regulation derived from murine expression array data", PNAS, 14 Dec 2004 Epub. PubMed.