stepping stones and evolutionary rescue

I promised to discuss “evolutionary rescue” – the ability of populations to adapt to environmental change so that they do not go extinct. Like many terms in evolutionary biology this should not be taken anthropomorphically – there is no rescuer. However it is relevant in an era of anthropogenic extinctions and evolution of antibiotic resistance (where extinction, or at least population size reduction, is the desideratum). The concept gets the experimental evolution treatment in this recent Nature paper in which 1255 replicate E. coli B populations were evolved under varying regimes of exposure to the antibiotic rifampicin.

Lindsey et al. examined the effect on population survival of varying the rate of change of the environmental driver. Populations were exposed to concentrations of rifampicin that increased at different rates: suddenly, gradually or moderately. Amusingly and aptly the authors cite the Reverend William Dallinger’s 19th century experiment in which “minute septic organisms” (protists) were evolved to resist elevated temperatures that individuals from the original population could not survive. In the new study sudden increases in rifampicin also frequently caused population extinction, but populations were more resilient when changes occurred gradually. What makes this paper particularly interesting is Lindsey et al.’s examinations of the reasons for this.

Broadly there are two reasons why population survival is increased under more gradual change:

1. Under less severe pressure the population is larger. So, in a given time, the probability of sampling beneficial mutations is higher. (Note that beneficial mutations are also less likely to be lost under drift in larger populations – something not discussed in the paper).

2. Transient moderate conditions may also open up new evolutionary paths through “stepping stone” mutations.

Point number 2 is not as obvious as it appears. As the researchers argue it requires the presence of genetic and gene-environment sign epistasis. Sign epistasis describes the situation in which the direction of the effect of an allele changes depending on the background. For example, imagine two loci: a and b, with alleles (A or a) and (B or b), respectively (and think haploid). We have an instance of sign epistasis if allele A reduces fitness on a b background, but it increases fitness on a B background. What this means is that, when starting with the ab genotype, an AB genotype is inaccessible assuming one-step mutations only (that is if B behaves in the same way and is costly on an a background). In fitness landscape terms, we have to cross a valley to get to a peak = ruggedness. If this is the situation at a high antibiotic concentration then higher fitness is inaccessible. But crossing the valley is possible if the effect of an allele changes direction depending on the environment = gene-environment sign epistasis. For example, there would need to be an environment in which A increases fitness on a b background (opposite to the above description). When intermediate antibiotic concentrations provide this environment they open up new paths to adaptation – and increase the probability of population survival. If my explanation is confusing examine figure 3 in the paper.

I said Lindsey et al. gave the experimental evolution treatment to evolutionary rescue. Accordingly the authors do what is only possible within this paradigm: they reconstruct ancestral genotypes, and combinations thereof, and they expose them to varied antibiotic concentrations. Data thus collected showed that, in at least some gradually changing lineages, sign epistasis occurred (with the caveat that genotype information was limited to a single gene). Stepping stone mutations were identified that were deleterious at high rifampicin concentrations but beneficial at intermediate concentrations, while combinations of these alleles demonstrated enhanced fitness at higher concentrations (although not always at the maximal concentration, figure 4). Of course the two explanations for evolutionary rescue listed above are quite likely to operate simultaneously as the authors allow.

With respect to reversing environmental damage (mentioned in my last post) these results are interesting but not conclusive. For example, it is possible that evolution to resist environmental change will result in higher fitness even when that change is reversed (as implied by the idea of a fitness valley). But since we have acknowledged the existence of gene-environment epistasis it is also possible that the fitness of resistance alleles is lower in absence of the selective pressure (this is shown in the paper’s figure 3a where the AB genotype has low fitness in environment x). In the case of antibiotic resistance it may be more reasonable to assume that resistance carries a cost in the absence of a drug. If it does and if susceptible and resistant strains co-exist, it is wise to reduce the duration of treatment contra medical orthodoxy.

Finally it is worth noting that extinction can occur from intrinsic (mutational) causes such as Muller’s Ratchet (in small populations) and deterministic mutation accumulation (in any population). In a sense we have here a special case of the problem of induction in that any extant population is descended from an extinction-resistant lineage – but we also know that extinction is very common in the history of life and that we live in an era of unprecedented environmental change.

a drop in the ocean

Human-caused increases in atmospheric carbon dioxide have consequences besides climate change. For example, they are causing the ocean’s pH to drop with unwelcome consequences as corals struggle to accrete their calcium carbonate skeletons. What makes this process interesting from a scientific perspective is that a large time lag is built into this process as carbon dioxide gradually dissolves in the ocean. Alas this is not the only instance.

A review in TREE discusses the consequences of time lags in abstract terms. The authors argue that lags should be afforded greater attention for their role in regime shifts. Regime shifts occur when one ecological equilibrium gives way to another under the influence of some driving factor. One example of this is the degradation of Caribbean coral reefs under pressure from over-fishing, ocean acidification and other causes (I recently read this excellent book on the subject). A decline in coral cover has occurred over extended periods (punctuated by catastrophic events such as hurricanes). When changes occur gradually – over human generations – it can be difficult to perceive them as people in each generation base their expectations on personal experience or recent data – an effect which is called “the shifting baseline” (from a 1995 TREE article by Daniel Pauly – for an accessible account see this voice-overed slide show). Also, it can be difficult to assess the causes of gradual change.

Setting human understanding to one side, lags have real-world consequences because, by definition, they entail that a system is in a non-equilibrium state for a time. The focal article shows how this can impede ecosystem recovery more than equilibrium-only models predict. (Note also that they do not entail anything about the acceleration of a system just its velocity so sudden changes may occur – just later). More optimistically, they offer an opportunity to intervene to prevent regime shifts that are otherwise fated to occur. To use the phrase of the article, certain systems are likely “living on borrowed time”.

From an evolutionary perspective this offers one interesting possibility which is that adaptation can also occur during a lag period – leading to “evolutionary rescue”. This will be the topic of my next post (just as soon as I’ve read a recent article)! For now suffice it to say that adaptation is itself a non-equlibrium process and a brief survey of your environment, physical and social, suggests that, when lags are taken into account, equilibria might be the exception rather than the rule.

Correction: above I say that lags relate to velocity of change not its acceleration. This is probably inaccurate. In a complex system the rate of change and the change in that rate could both change. In a sense a lag means that both have changed relative to what we’d expect in the absence of a lag.

complex processes

Inferring evolutionary dynamics is difficult because there is not a one-to-one relationship between genotype and phenotype. One phenotype may be the product of different genotypes. Conversely, the expression of a given genotype depends on the environment. This leads Michael Travisano and Ruth Shaw to guarded pessimism in this open access article published in February’s Evolution. The authors argue that more work is needed that directly examines process – something their own research programmes in experimental evolution demonstrate very well.

It is possible to generalise this. The ubiquity of genome-level data (driven by the low cost of high-throughput sequencing) is worrying if it leads to an emphasis on descriptions of genome-level phenomena without attention to their evolutionary causes. If this were chemistry we might end up counting the molecular masses of elements thinking that these are fundamental chemical properties rather than being contingent on isotope proportions. Fortunately a body of evolutionary theory has built up allowing us to interpret genome data although recent strident criticism of the ENCODE project suggests that consistency with theory is not always achieved. This dispute centres on what it means to impute “function” to a genome region. From an evolutionary perspective sequence conservation is the best guide, but, given the complexity of evolution, it is far from perfect and it is precisely this complexity with which we should engage if we want to understand genome structure and function.

This blog is named after an example of complexity in evolutionary dynamics and I’d like it to prioritise our developing understanding of processes – which is increasingly enhanced through experimental evolution as well as via the development of theory and application of the comparative method.

This week’s PLoS Biology brings a rapid answer to Travisano’s and Shaw’s call in a paper by Matthew Herron and Michael Doebeli that explores the fate of mutations in a system undergoing diversifying (and frequency-dependent) selection. This kind of selection leads to divergent genetic changes between subpopulations and here it leads to the evolution of two ecotypes that exploit the two available carbon sources differently (see this summary in the same journal).

This is a really nice example of evolutionary complexity on several levels. First, parallel and divergent changes were observed in populations indicating a role for chance and for necessity in this system – and suggesting partial degeneracy in the genotype-phenotype map. Second, the role of history was indicated by the manner in which mutations spread within the two ecotypes. Some mutations inferred to be from one ecotype would only increase in frequency after mutations from the other ecotype had risen to intermediate frequencies. The last point is critical because it also indicates that ecological interactions were responsible: overall the best explanation of the data supported a role for clonal interference within ecotypes, but also for frequency-dependent selection between them.

soft sweep

Welcome. Aim: to highlight recent literature. Focus: evolution and genetics/genomics particularly experimental evolution.

Why the name? There is plenty of good stuff out there so there is no need to wait for the next good thing…