There is a considerable amount of noise within our cells, owing to the variable expression of many genes. A new theoretical model that incorporates noise has been applied to understand how multipotent cells make fate decisions in a clonal population of mouse haematopoietic progenitor cells; it concludes that the unique expression profiles of subpopulations of cells within a lineage lead to a distinct preference for choosing one of two alternative developmental fates.

One type of mammalian haematopoietic progenitor cell has the choice of differentiating into either the erythroid or myeloid lineage. This clonal population is highly heterogeneous with respect to the expression of the stem cell-surface marker Sca-1 (also known as Ly6a ), levels of which can vary 1,000-fold between individual cells. What factors govern this heterogeneity? And does it affect the differentiation potential of these cells?

To answer the first question, Chang and colleagues sought to describe in mathematical terms the heterogeneous expression pattern of Sca-1 across cells. When cells from three different points in the Sca-1 bell-shaped expression range (low, medium or high) were separated out and left to develop in culture, each fraction retained its starting expression level for 9 days before gradually returning towards the distribution seen in the parental population. This behaviour could not be described by a simple reversion towards the population mean through a build-up of transcriptional noise. Instead, the authors propose the existence of many stable expression states within cell subpopulations — that is, cells would transit between discrete and temporarily stable expression states.

To answer the second question, the authors examined whether the metastable states that the cells found themselves in could be somehow linked to their future fate. Indeed, cells in the fraction of the population with lower Sca-1 expression were more likely to differentiate into the erythroid lineage when exposed to a differentiation stimulus, and the preference for lineage choice correlated with a 200-fold increase in expression of Gata1 , which encodes a lineage-specific transcription factor. Conversely, cells that developed preferentially into the myeloid lineage had high mRNA and protein levels of the transcription factor PU.1, which is known to antagonise GATA1 in lineage determination.

What makes this observed behaviour all the more remarkable is that the three separate fractions differ not only in the expression of Sca-1 but also across approximately 4,000 'background' genes, all of which return to a population average after a few days in culture. Thus, cells experience a vast range of apparently randomly generated heterogeneous expression, some of which contributes to priming cells to commit to a particular fate.

Although descriptions of this type have a theoretical importance for understanding the nature of intercellular variability, they might also have a practical relevance in directing lineage choices for therapeutic applications.