Stochastic expression of proteins in a single cell

Every thought about how variable the expression of a particular gene is across an entire cell population?

That's what the Weissman lab described in a manuscript in a recent issue of Nature. Anytime you want to take on such a project - take my advice, you turn to yeast. The yeast field has created a library of strains, each containing a copy of GFP (Green fluorescent protein) fused to a particular gene within the genome. If you measure the fluorescence, you can quantify the level of protein expression. The next trick is to use flow cytometry to rapidly measure the brightness individual cells in a population. Brightness per cell = GFP per cell = expression of the tagged gene, per cell.

Now you can record how a particular gene is expressed, in terms of protein levels, on a cell to cell level. In addition one can identify how cells alter their protein levels when exposed to various conditions. Note that this type of experiment has been done at the mRNA level using microarrays, yet until now no one has published any account of how to perform these measurements at the protein level.

So what did they find?

- For cells grown in rich media, 30% of their genes have elevated protein expression, 10% have decreased protein expression when compared to cells grown in minimal medium. Proteins with increaed levels in rich media, are those involved in cell division and cell wall biosynthesis. Proteins with increaed levels in minimal media, are biosynthetic genes (if the environment doesn't have aminoacids and nucleotides, you've got to make them yourself.)
- Importantly, the researchers asked whether protein levels reflected mRNA levels, or whether other post-translational events (such as protein degradation) played significant roles. To the relief of many microarray manufactures, changes in protein levels largely correlated with changes in mRNA fir most genes. There were some exceptions (like ribosomal RNA processing enzymes and enzymes involved in the Kreb cycle).

But now comes the interesting part, proteins involved in core functions (i.e. ribosomal proteins, initiators of translcription, protein synthesis, protein degradation) have low variability in protein levels on a cell-to-cell basis. The output from these genes had low noise. Strangely, Golgi components have low noise as well, but mitochondrial and peroxisome proteins have high noise. Overall it seems like mRNA production is the greatest determinant of noise:

...variation most likely originates from the stochastic production and destruction of mRNA molecules. Indeed, the magnitude of the variation observed here (CV 30% for low-medium abundant proteins) is entirely consistent with that expected if protein variation results from Poisson noise owing to small mRNA numbers (1-2 per cell) and is mitigated by a filtering effect that arises because proteins are typically far longer-lived than their messages.


... high noise is likely to be due, at least in part, to the introduction of a slow step into the production of mRNA, making the process more prone to bursts.

Furthermore, noisy genes are regulated at the transcriptional level by similar transcription factors and chromatin remodeling enzymes. Stable genes tend to be regulated by another group of transcription factors. Stable genes are also less likely to be affected by fluctuating mRNA numbers. This is best achieved by have increased numbers of messages that turnover rapidly.

The authors point out that noise (or the variability of expression) is an important consideration in how genes are regulated. Cells may want certain genes, such a those that respond to environmental stress, to have "noisy outputs".

for some proteins that permit cells to respond to environmental perturbations, excursions from the mean at the single-cell level might benefit populations. In the short term, such deviations might facilitate a cell's initial response to environmental variation. More generally, the capacity to vary might permit a population to sample multiple phenotypic states to maximize the chances of some, but not all, cells' survival in an adverse environment.

Other genes, such as those involved in ribosomal maintenance and cell cycle regulation, need to have stable levels in order to ensure cellular homeostasis.

The takehome message, forget about the average, the generation of variability or stability may be a crucial to cellular life.

John R. S. Newman, Sina Ghaemmaghami, Jan Ihmels, David K. Breslow, Matthew Noble, Joseph L. DeRisi and Jonathan S. Weissman
Single-cell proteomic analysis of S. cerevisiae reveals the architecture of biological noise
Nature (2006) 441:840-846

Cross posted at ScienceSampler.

More like this

This seems to confirm and expand some work that has been done in bacteria (Mary Lidstrom, principally, recently) and some work done with theoretical models of lambda gene expression. Txnal regulation is noisier and more responsive, RNA stabilization and tlnal regulation results in more stable and predictable protein expression - GFP studies always worry me a bit because of the properties of the fluorophore, but it's been used so much, I'm sure they know what they are doing.

By Paul Orwin (not verified) on 30 Jun 2006 #permalink


In a living cell, gene expression-the transcription of DNA to messenger RNA followed by translation to protein-occurs stochastically, as a consequence of the low copy number of DNA and mRNA molecules involved. These stochastic events of protein production are difficult to observe directly with measurements on large ensembles of cells owing to lack of synchronization among cells. Thanks a lot...