Dynamics of transposable element invasions with piRNA clusters

Abstract In mammals and in invertebrates the proliferation of a newly invading transposable element (TE) is thought to be stopped by a random insertion of one member of the invading TE family into a piRNA cluster. This view is known as the trap model. Here we explore the dynamics of TE invasions under the trap model using large-scale computer simulations. We found that piRNA clusters confer a substantial benefit, effectively preventing extinction of host populations from an uncontrollable proliferation of deleterious TEs. We show that TE invasions under the trap model consists of three distinct phases: first the TE rapidly amplifies within the population, next TE proliferation is stopped by segregating cluster insertions and finally the TE is permanently inactivated by fixation of a cluster insertion. Suppression by segregating cluster insertions is unstable and bursts of TE activity may yet occur. The transpositon rate and the population size mostly influence the length of the phases but not the amount of TEs accumulating during an invasion. Solely the size of piRNA clusters was identified as a major factor influencing TE abundance. Investigating the impact of different cluster architectures we found that a single non-recombining cluster (e.g. the somatic cluster flamenco in Drosophila) is more efficient in stopping invasions than clusters distributed over several chromosomes (e.g germline cluster in Drosophila). With the somatic architecture fewer TEs accumulate during an invasion and fewer cluster insertions are required to stop the TE. The inefficiency of the germline architecture stems from recombination among cluster sites which makes it necessary that each diploid carries, on the average, four cluster insertions, such that most individuals will end up with at least one cluster insertion. Surprisingly we found that negative selection in a model with piRNA clusters can lead to a novel equilibrium state, where TE copy numbers remain stable despite only some individuals in a population carrying a cluster insertion. Finally when applying our approach to real data from Drosophila melanogaster we found that the trap model reasonably well accounts for the abundance of germline TEs but not of somatic TEs. The abundance of somatic TEs, such as gypsy, is much lower than expected.