Usually a contrast is made between bet hedging and adaptive plasticity. What makes adaptive plasticity different from hedging is that organisms respond adaptively to environmental cues. But what if organisms respond to environmental cues in ways that do not increase mean fitness, but do increase geometric mean fitness? In other words, what happens if bet hedging occurs in response to uninformative cues? This is explored in this paper accepted last month in the journal Evolution.
The study consists constructs and makes use of an artificial system in which organisms use an cue to generate a phenotype that enhances fitness in one environment, but decreases it in another. Critically the provision of the cue is orthogonal to the environment and is therefore not predictive of it. This trade-off is critical for the evolution of diversified bet hedging.
The study uses Saccharomyces cerevisiae and makes use of the gene URA3 which codes for orotidine 5’- phosphate decarboxylase (Ura3p). Ura3p synthesises uracil from precursors, but if the pro-toxin 5-fluororotic acid (5-FOA) is present in the environment, it will convert this to the toxin 5-fluorouracil (5FU) killing the cell. The trade-off was generated via two phenotypes/genotypes:
- P1: URA3+ yeast
- P2: URA3– yeast
And these were grown on two environments:
- E1: SC-ura: a media that lacks uracil
- E2: SC+5-FOA: a media containing the pro-toxin
A strong trade-off was demonstrated via a concave fitness function with P1 strongly favoured on E1 and P2, on E2. In order to create bet hedging based on a cue, the authors chose estradiol. Estradiol doesn’t regulate any genes in S. cerevisiae, but they constructed a new phenotype/genotype where it does:
- plastic yeast: URA3 under the control of the transcription factor (TF) Z4EV
When estradiol is added, the TF, usually excluded from the nucleus, is able to enter and bring about expression of URA3. Spotting estradiol at the centre of E1 plates resulted in plastic yeast having high fitness at the centre, while spotting onto E2 plates gave the highest fitness at the periphery.
The authors constructed an varying environment by growing the different yeasts on alternating E1 and E2 plates spotted with estradiol. Between plates, yeast cells were collected using glass beads rolled on the plate (often used as a neat method for spreading plates). GFP-labelled (green) P1 or P2 was competed against mCherry-labelled (red) plastic yeast. (Label swaps were also used in the study).
The authors found that the plastic yeast could rapidly outcompete the P1 and P2 yeast in this regime, presumably owing to the near zero fitness of P1 on E2 (and P2 and E1). Having zero in a series brings the geometric mean of the series to zero. To show the other requirement for bet hedging: that there is a fitness cost, they looked at the selection coefficients of plastic versus the other strains in their best environments. Against P2 on E2 plastic yeast were significantly less fit. Against P1 on E1, they were also less fit although this didn’t reach significance.
When it evolves, diversified bet hedging should have phenotype frequencies correlated with the probability of (or proportion of time spent in) each environment and the fitness of the phenotypes in each. The authors manipulated the response to estradiol with a third phenotype/genotype:
- plastic ∆/∆ yeast: deleted ABC transporters PDR5 and SNQ2
These transporters remove estradiol from the cell and their deletion is expected to increase intracellular estradiol concentration. As a result of this change, URA3 expression at a given [estradiol] should be dialled up in plastic ∆/∆ yeast. As expected, these yeast occupied more space on E1 plates, and less on E2 plates, than the original plastic yeast.
This is an interesting study although clearly it is limited by being an artificial system. The authors refer to it as “constructive proof”of cue-based hedging and its tune-ability. In particular they consider the implications of environmentally cued bet hedging for the detection of bet hedging. Detecting hedging is already a difficult task because a fitness trade-off between phenotypes in different environments and a matching frequency of environments must be demonstrated. With cue-based hedging we have the added difficulty that the usual experimental practice is to hold everything else constant (e.g., in a chemostat) except the variable of interest. Bet hedging may disappear in this case.
The Origins of Sensing
As the authors indicate, it is usually assumed that molecular stochasticity within the cell is exploited to generate bet hedging. Sensing the environment is a costly, and therefore disfavoured, addition. However, estimating costs is difficult and once sensing emerges it might be co-opted to generate diversity if doing so has small or no costs. Sensing can be quite rudimentary – depending on protein conformation alone for example.
If you can “wire” sensing to responding even with uninformative cues this has implications for the evolution of sensing and responding. Adaptive phenotypic plasticity is usually considered as the context for this, but it may not be.