Hedging Your Bets
by Merry Youle
Bacteria that are born genetically equal aren't necessarily the same. The same genome, residing in cells side-by-side in the same medium in the same flask, does not guarantee the same phenotype. One example that comes to mind is the persisters in E. coli populations – the small number of cells that spontaneously stop growing. If the population is hit by β‑lactam antibiotics, those cells escape death. Similarly, under lab conditions that trigger genetic competence in B. subtilis, only a small fraction of the cells make the switch to competence.
Click here to download a movie from the Elowitz lab showing different behaviors within a clonal population of Bacillus subtilis in relatively homogeneous conditions. Here, the cells express a green fluorescent protein when growing normally, but turn on a red fluorescent protein when differentiating into a transient competent state. Some others differentiate into resilient spores (white objects). Source
B. subtilis cells growing in a rich growth medium offer yet another example. Here genetically identical cells comprise two distinct types. Most are flagellated and actively swimming about as individuals, while a minority have no flagella and form long chains. The game is different in cells in the stationary phase where virtually all of the cells are found in long chains, bound together by an abundant matrix. Losick, Kolter, and colleagues have been working with this system for some years (for earlier papers, click here and here) seeking to determine how such bimodal cell populations are established and maintained in growing cultures.
Others have approached this question from a theoretical point of view asking what sort of a regulatory "circuit" would it take to produce robust bistability. In other words, how might a cell, using a simple system employing but a few regulatory genes, be able to operate in either of two distinct states. In 2000, Gardner et al. started with a model for a bistable network (Figure 1) and then attempted to create such a circuit in E. coli. (A simple system this may be, but conveying it in words is not so simple. I suggest you wrap your mind around the diagrams first.)
Figure 1. A model of a simple toggle switch. Redrawn from Source
In this model we have two genes, A and B. Promoter A controls the transcription of Gene A and hence the synthesis of Protein A. Likewise, Promoter B controls transcription of Gene B and the synthesis of Protein B.
When Protein B is present in the cell, it binds to and blocks Promoter A, thus repressing expression of Gene A. As a result, when Protein B is present, Protein A is not made and soon the cells will be depleted of it. Such cells are said to be "locked" in the stable B State. Cells could instead be in an alternative state, the A State. Here, they would contain Protein A, which blocks the synthesis of Protein B. These two states are mutually exclusive. Either a cell transcribes Gene A but not Gene B, or vice versa. What I'm calling a "gene" here might encode an entire operon or a regulatory factor that affects numerous operons. Therefore this would be a potential mechanism whereby some cells could have their motility genes turned off and their matrix genes turned on, or vice versa.
Figure 2. An inducer can flip this toggle switch. Redrawn from Source
In the case of a growing culture of B. subtilis, an individual cell in either state can give rise to a mixed population. This means that we also need some way for a cell to switch states. One possibility is by a stochastic mechanism triggered by the random fluctuations inherent in gene expression in any cell. The rate of transcription of a gene will vary somewhat by chance – sometimes faster, sometimes slower depending on what else might be going on. If the bistability is robust, the cell will not flip randomly between states due to such fluctuation. However, bistability comes in all gradations. Suppose it takes 50 molecules of Protein A to maintain a cell in the A State. If the cell normally contains several hundred copies, minor fluctuations in its synthesis are unlikely to kick the cell into the B State. However if the cell typically has only 60 copies, some cells would likely fall below the threshold occasionally.
Figure 3. Protein A (SinR) represses the matrix genes and also gene B. Protein B (SlrR) represses gene A and also combines with Protein A to repress the motility genes.
Alternatively the cell could make use of a programmed mechanism. These are more predictable and can convert an entire population to the same state. Here the trigger might come in the form of a specific inducer synthesized in response to an environmental signal (Figure 2). Suppose that, when starved, cells in the B State make Inducer A. Inducer A binds to Protein B, thus preventing it from doing its usual job of repressing Gene A. This would turn on synthesis of Protein A – and switch the cell to the A State. Moreover, this would happen in virtually every cell in the population given the same environmental signal. Gardner et al. set to work and constructed this sort of a "genetic toggle switch" in Escherichia coli. Their synthetic switch worked.
In a recent paper, the Losick and Kolter labs present data indicating that B. subtilis uses a stochastic toggle switch during exponential growth to maintain a bimodal population in which most cells are motile but a few are sessile and chained together within an extracellular matrix. In a flask of exponentially growing B. subtilis, the individual motile cells contain a protein (we'll call it Protein A) that turns off the genes required to make the biofilm matrix. Protein A also blocks the synthesis of another important regulatory protein that we'll call Protein B. If Protein B were present, it would turn off the motility genes. The majority of the cells in our flask make Protein A, have little Protein B present, and therefore are motile, individual, and do not synthesize matrix components.
Figure 4. A culture of wild type B. subtilis growing in LB medium showing a mixture of individual and chained cells. (upper panel) Phase contrast. (lower panel) Red = membrane dye FM4-64; green = Gfp fusion indicating the low Protein B state with motility genes on. Source
Although most cells in the population are individual and motile, some are found in chains. Why is there any chaining? In the authors' words: We propose that the switch is activated stochastically by noise in the expression of these same components. In other words, given even a brief pulse of Protein B synthesis, the system flips over to the high Protein B, low Protein A state where the motility genes are turned off and matrix genes are turned on. This apparently does happen, but only rarely, since most of the population is motile. However, if the population is transferred to a poor growth medium, essentially all the cells shift to biofilm formation. This response to environmental stress is orchestrated by the programmed production of an inducer that turns on the synthesis of Protein B.
The mechanism by which Protein B promotes chaining involves a protein-protein interaction – a new twist on our basic model. Protein B interacts with Protein A to form the A•B heterodimer. This heterodimer does two things: (1) It ties up the Protein A that is present so that the matrix genes are expressed (and so that still more Protein B is made); (2) It turns off the transcription of the motility genes.
This B. subtilis story is but one example of what appears to be bacterial bet-hedging. In a 2008 review article about this tactic, the authors note: Under challenging conditions, the production of offspring with variable phenotypes ensures that at least one offspring will be appropriate (fit) under a given situation. This is a risk-spreading or bet-hedging strategy… and it's a good bet that more such stories are yet to come.
Reference
Chai Y, Norman T, Kolter R, Losick R (2010). An epigenetic switch governing daughter cell separation in Bacillus subtilis. Genes & development, 24 (8), 754−765. PMID 20351052
Wow, Merry. Simply wow. That was a tour de force (or however it is spelled, my "culture" is associated with 2059 tubes). I was just talking about this kind of thing in class. I tip my hat, literally.
I would add this, Merry. I have long been interested in "unstable phenotypes," which we have all seen (sectored colonies, etc). Sometimes these are due to well understood mechanisms, but often note.
The "bet hedging" model is lovely, and reminds me of this story from a few years ago:
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC280332/
Doug Bartlett was able to show that an IS element would hop in and out of an EPS gene regularly (or irregularly?). When in the EPS gene, the bacterium was not "sticky" and did not attach to surfaces. When the IS element precisely excised, it lead to synthesis of the EPS, and a "sticky" phenotype allowing it to adhere to surfaces. Thus—another "metastable bimodal switch" that would allow both colonization of surfaces and dispersal to new sites?
I've been thinking about this for a few days, and the thing that I am having problems with is the use of the term "epigenetics". I think bistability and oscillatory gene regulation are fascinating (they appeal to the math dork in me), but I always associate epigenetics with things like methylation patterns that allow for changes in gene expression levels to be passed from one generation to the next, rather than simply "aspects of phenotype that are not directly from genotype" — or am I interpreting this incorrectly? This seems like a way to genetically program a discrete probability distribution. In other words, based on the ratios of regulator required for either state, you can adjust the probability of each state (so in a different strains, the ratio of motile to sessile might be 3:1, or 10:1, or 1:1, under the same conditions, just because they have different regulatory variants). I guess that what I don't see is why this is "epigenetics"? Is it just my ignorance here (freely admitted!)?
Hi, Paul,
It is very pleasing to hear that you (and perhaps others) think about items in our posts for days afterwords!
As to "epigenetics," I appreciate your point of view. That is why I did NOT use the term in the post itself, even though it was in the title of the featured paper. Perhaps the authors chose it to emphasize that the state is inherited and can be stable for many generations.
We may encounter more terminology issues similar to this in the future because "genetics" came to be gene focused, and any mechanism of inheritance other than or in addition to a nucleotide sequence is both suspect and in need of a term to classify it.
Lastly, your point about ways to "genetically program a discrete probability distribution" is an important one. Stochastic events meet natural selection.