Conceived and designed the experiments: TK PWM DAP. Performed the experiments: TK PWM. Analyzed the data: TK PWM DAP. Contributed reagents/materials/analysis tools: TK PWM DAP. Wrote the paper: TK PWM DAP.
The authors have declared that no competing interests exist.
Adaptation in eukaryotes is generally assumed to be mutation-limited because of small effective population sizes. This view is difficult to reconcile, however, with the observation that adaptation to anthropogenic changes, such as the introduction of pesticides, can occur very rapidly. Here we investigate adaptation at a key insecticide resistance locus (
Adaptation in eukaryotes is often assumed to be limited by the waiting time for adaptive mutations. This is because effective population sizes are relatively small, typically on the order of only a few million reproducing individuals or less. It should therefore take hundreds or even thousands of generations until a particular new mutation emerges. However, several striking examples of rapid adaptation appear inconsistent with this view. Here we investigate a showpiece case for rapid adaptation, the evolution of pesticide resistance in the classical genetic organism
The speed of adaptation in eukaryotes is commonly assumed to be limited by the waiting-time for an appropriate adaptive mutation. This notion is based on estimates of the population parameter Θ = 4
However, adaptation to anthropogenic changes such as the evolution of insecticide resistance has been observed to occur very rapidly and often involves complex alleles
In order to understand the population parameters that allow for rapid adaptation in eukaryotes, we study here a well-documented example: the evolution of pesticide resistance in
Acetylcholinesterase (AChE), a key neuronal signalling enzyme, is the major target of the most commonly used insecticides, organophosphates (OPs) and carbamates (CMs)
In
Here we collect data and provide quantitative arguments (both analytical and simulation-based) that the observed signatures of adaptation at
We collected
We detected resistant mutations at the first three sites (I161V, G265A, F330Y) but did not find the resistant mutation at the fourth site (G368A). We estimated that ∼40% of the strains contain resistant mutations in the modern NA and AUS populations of
Alleles containing mutations I161V, G265A and F330Y are numbered 1, 2 and 3 respectively. Sizes of the circles correspond to the number of identical sequences representing each haplotype; tick marks along a branch indicate the number of mutations between two neighbouring haplotypes. Sensitive haplotypes are labelled with capital letters and resistant haplotypes with lowercase letters. Note that our sample is enriched for resistant haplotypes. Resistant NA alleles containing a single mutation (all at the first site) appear to have arisen on the common out-of-Africa haplotype L, with one specific L-related allele (labelled p) present at the highest frequency. The resistant AUS alleles also cluster together. AUS resistant alleles containing a single resistant mutation in the first or second site appear to have arisen either on the background of the common out-of-Africa sensitive haplotype L, or on the background of the specifically AUS haplotype N. The alleles containing two mutations in NA (first plus second or first plus third sites) are all related to the sensitive L haplotype and the common resistant allele (labelled p) containing the mutation in the first site. The 3-mutation alleles are present both in NA and AUS populations (v and w) and are the most closely related to the sensitive L haplotype. There are two resistant alleles containing single mutations in the first and the second site that we detected in AF. One of these is very similar to the AUS alleles containing the second mutation and is likely a migrant from out-of-Africa back to AF. The other appears to have arisen
The table shows segregating sites within the 1.5-kb region of
In all cases the NA and AUS resistant alleles show no signs of having predated the spread of
In summary, the sequence analysis of the resistant and sensitive alleles reveals two signatures of the adaptive evolution of pesticide resistance at the
Below we consider a simple scenario of a single locus in a panmictic population of effective size
In
After the onset of pesticide application, resistant alleles become advantageous (
Let us now consider
The probability of successful adaptation from
This equation implies that selection must already be very strong for a single 1-mutation allele to arise and to become prevalent in less than 1500 generations (
We have established that under this simple model if Θ is 0.01, the adaptation at
We can imagine two scenarios that would generate this observation. In the first, the so called hard sweep scenario, a single adaptive mutation arises in frequency in the population and eventually ends up on different haplotypes due to recombination or mutation events that take place in its vicinity during the sweep. In the other, an example of the so-called soft sweep scenario, several independent adaptive mutations take place on different haplotypes and increase in frequency simultaneously.
Theoretical investigations under simple scenarios by Pennings and Hermisson
What is the probability that a mutated or recombined haplotype also reaches at least frequency
Let us assume that the second haplotype becomes established at time
Mutations establishing after
Here
In conclusion, under this simple scenario, our empirical observations at
Note that if Θ were much higher, for example on the order of one or larger, then all of our observations are expected. Soft sweeps would be commonplace because many more mutations enter the population in every generation and can increase in frequency simultaneously thereby generating multiple haplotypes containing the same adaptive mutations
We have shown above that under very simple population scenarios the pattern of adaptive evolution at
To investigate quantitatively the potential impact of such effects we conducted extensive simulations of adaptation at
In
(A) Frequency trajectories of resistant haplotypes from a typical simulation of a single population with Θ = 0.01 during the first 1500 generations after pesticides are applied. The selection scenario is
Interestingly our simulations show that if Θ∼1, then most of the observed signatures of soft sweeps are generated by multiple
Our data and analysis strongly suggest that the patterns of adaptation observed at
Such a large value of
In sharp contrast, adaptation at
(A) The population history similar to that inferred by Thornton and Andolfatto for
The short-term
It is reasonable that
The possibility that adaptation at single sites in
The number of sweeps (hard or soft) might also in general be lower than the number of adaptive substitutions if complex adaptations requiring multiple substitutions are common. Indeed, in our simulations of evolution at
Note that all of these expectations hold especially well for strong selection because it operates over shorter time scales and is therefore less sensitive to recurrent but infrequent bottlenecks
Most of the current statistical approaches for the study of adaptation rely on the expected signatures of hard sweeps
Recurrent boom-bust cycles are a general feature in population dynamics of most studied organisms. Adaptation and recurrent selective sweeps reducing the long-term but not the short-term
We sequenced 1450 bp encompassing exons 2 through 4 of
Ace1F:
Ace1R:
Ace2F:
Ace2R:
Ace3F:
Ace3R:
PCR products were then sequenced. Of the 68 sequenced strains, 26/68 (∼40%) have a single or multiple resistant mutations. Mutations at I161V, G265A and F330Y were identified in isolation and in combination in multiple populations, while G368A was never observed. We then used PASA
161-F:
161-R:
265-F:
265-R:
330-F:
330-R:
The 161 primer pair amplifies more effectively in the presence of the mutation I161V. The 265 primer pair is specific to G265A and the 330 primer pair is specific to G330Y. The annealing temperatures required for allele specific priming used for 161, 265 and 330 were 61.5°C, 59.5°C and 60.6°C respectively. As positive and negative controls we performed PASA on strains in which the resistant sites had been previously characterized. We sequenced 37 strains from 8 populations that had amplified with one or more of the allele-specific primers. 31/37 (84%) of these strains contained resistant mutations. The incorrect classification of the 6 strains is likely due to the addition of excess template to these PCR reactions resulting in non-specific priming. In total, we sequenced the
The most parsimonious haplotype network was constructed using TCS 1.21
Measures of Θ
Our simulation models the population frequency dynamics of haplotypes at the 1.5 kb-long sequenced
Haplotypes are classified by their particular adaptive allele configuration at the three adaptive sites. We describe this configuration in terms of a vector
We use an infinite alleles model for new haplotypes,
Mutations at adaptive sites and recombination events where the recombination breakpoint lies between two adaptive sites can generate new haplotypes with different adaptive-allele configuration (
The evolution of haplotype frequencies is simulated in terms of a Wright-Fisher model with directional selection,
We group resistant haplotypes into three classes according to the number of resistance-conferring mutations they bear: 1m haplotypes have one resistant allele (100,010,001), 2m haplotypes have two (011,101,110), and 3m haplotypes have all three resistant alleles (111). For simplicity, we assume that all haplotypes in the same class have equal selection coefficients
The key simulation parameters are the selection scenario defined by the selection coefficients
Simulation runs start with one single sensitive haplotype present in all subpopulations at 100% frequency. Before pesticide application commences, mutation-selection equilibrium of resistant haplotypes is established within a burn-in period of 1000 generations. This fully suffices to establish equilibrium due to the strong purifying selection against all resistant haplotypes prior to pesticide application (
A random number of mutation events is drawn from a Poisson distribution with mean
A random number of recombination events is drawn from a Poisson distribution with mean
The numbers of migrating individuals to each other subpopulation are drawn from a Poisson distribution with mean
All haplotype frequencies are evolved one generation according to the above-described binomial sampling procedure.
During a simulation run we analyze whether resistant haplotypes emerged and whether soft sweep signatures among 1m haplotypes were observed. We define 1m resistance by at least one of the three 1m adaptive-allele configurations (001, 010, or 100) ever being present in more than 10% of the population during the run. Accordingly, 3m resistance is defined by the complex 3-mutation allele (111) ever present in at least 10% of the population. A soft sweep signature (ss) is ascertained if at any time during the run two independently drawn alleles have greater than 10% probability to bear the same 1m configuration on different haplotypes. The statistics
A crucial assumption of our simulation is the applicability of an infinite alleles model,
The simulation was implemented in C++. Runs were performed on the Bio-X2 cluster at Stanford University. All source code is available from the authors upon request.
Structure of the
(0.04 MB PDF)
Description of
(0.05 MB PDF)
Segregating sites at the
(0.07 MB PDF)
Dynamics of resistance adaptation for different population parameters.
(0.06 MB PDF)
Change of mutation configuration due to mutation or recombination.
(0.04 MB PDF)
Origin of soft sweep signatures.
(0.11 MB PDF)
We thank members of the Petrov lab, Hunter Fraser, Ward Watt, Marcus Feldman, Joanna Kelley, Graham Coop, Molly Przeworski, Joachim Hermisson, Richard Lewontin, Daniel Fisher, John Novembre, Georgii Bazykin, two anonymous reviewers, and the participants of the Aspen 2010 Biophysics Conference “Populations, Evolution and Physics” for helpful comments and discussion.