TE copy numbers and distribution: Australia vs. Old World colonisations
Comparing this study with previous works that analysed original (from Argentina) and colonising (from the Iberian Peninsula and Australia) populations, some differences are apparent. While the original populations always demonstrate low occupancy per chromosomal site, the colonising populations (Australian and Spanish) have some sites with high occupancy. When colonising populations are compared, Old World populations have more highly occupied Osvaldo sites (eight) and a higher occupancy rate of some sites (up to 63% occupancy) than the Australian populations (three sites with less than 33% occupancy). The mean copy number is lower in Australian than in Old World populations, placing Australia in an intermediate position between Argentina and the Old World. Data from Isis are limited to the Australian populations, but interestingly, the mean copy number is similar in Australian and Old World laboratory stocks.
The overall picture that emerges from these data suggests a global increase in the mean copy number per site of Osvaldo during the colonisation process. This may be explained either by the drift effect following the founder event acting on many low occupied sites (turning them into high occupied sites) or by increases of site-directed transposition rates in colonisation, or both. According to previous work [16, 17, 24, 25], sites demonstrating high insertion frequencies are most likely due to a founder effect that occurred during the colonisation process, and low insertion sites are probably the result of new transposition events.
The Australian and Spanish populations have similar insertion-site distribution patterns but some of the aforementioned differences deserve more detailed analysis. First, high insertion frequency sites are generally located in different chromosomal positions in the two colonisations, indicating two independent colonisation events from Argentina. Second, the number of highly occupied sites is greater in the Spanish than in the Australian colonising populations. These differences could be related to specific differences in the two colonising processes. While Old World colonisation appears to be caused by a low number of founders , the number of D. buzzatii individuals introduced during the Australian colonisation was certainly large initially [8, 9] but fell dramatically following the Opuntia biological control program. Therefore, we can imagine a scenario where D. buzzatii spread rapidly through the Opuntia area until 200,000 hectares of prickley-pear had been largely destroyed by biological control . Since D. buzzatii is restricted to the Opuntia distribution area , a parallel drastic decrease of the D. buzzatii population size should have occurred. This rapid reduction in the number of coloniser survivors could explain, at least in part, why the number of highly occupied sites in Osvaldo is lower in the Australian colonisation than in the Old World one.
The results showed here are in accordance with previous comparisons using allozyme variability [19, 28], microsatellites [10, 29] and mitochondrial DNA , where a decrease in variation was observed in Australian populations, probably due to a bottleneck followed by a population expansion after the initial founder event. Though the founder effect hypothesis is widely accepted by all authors working on Australian D. buzzatii populations, there are discrepancies in relation to the bottleneck magnitude depending on the markers used. Whereas nuclear markers such as microsatellites  and allozymes  suggest a moderate bottleneck, mitochondrial DNA  data point to a very low number of founders. The mtDNA was considered a particularly valuable marker for studies of recent population history owing to its maternal inheritance, lack of recombination and rapid evolution (relative to nuclear genes) [32, 33]. The fact that data concerning TE distribution tend to be more in agreement with mitochondrial data, and that records of the cactus control program implied a strict quarantine of associated insects, suggest that the most likely explanation is that only a few individuals founded the Australian populations. Therefore, it is expected that the magnitude of the founder effect was stronger in Australian populations than in Old World populations
On the other hand, during colonisation selection could also play un important role for the elimination of deleterious TE insertions. For example, homozygosity can increase during the colonisation process, if a few colonisers are involved in the process, increasing the strength of selection against deleterious insertions as observed in selfing plant species [34, 35] Additionally, selection could contribute to the adaptation to new environments as seen during the expansion of D. melanogaster from Africa [36, 37]. Recent results demonstrate a role for some TEs in the adaptation of D. melanogaster colonising populations to the out of Africa temperate climate . It is not impossible that some insertions are maintained because of their positive effects in terms of adaptation to the new environments encountered by colonisers. The bottleneck suffered by Australian populations could have produced gene frequency shifts that have been proposed as responsible for driving populations to new adaptive peaks by selection. But, in the present study, the presence of high insertion sites is more likely to be due to the drift effect associated to the colonisation process because the Australian D. buzzatii colonisation time of approximately 840 generations (70 years multiplied by 12 generations per year) is probably insufficient to detect selection effects unequivocally. This hypothesis is supported by results in previous studies where the Osvaldo DNA altered sequence and its flanking genomic sequences in high frequency sites were identical in all Old World colonising populations , a result highly improbable under the selective hypothesis Thus, when considering all results together we can conclude that selection does not seem to play a major role, compared to demography, in the distribution of transposable elements in the Australian colonising populations.
TE dynamics in the colonisation processes
Colonisation has been suggested as one of the processes responsible for awakening transposable elements in D. simulans . During colonisation, individuals are exposed to stress attributable to new environmental conditions and population regimes. Environmental stresses can induce changes in chromatin structure  and promote TE mobilisation as observed in plants  and other organisms . D. buzzatii, a cactophilic species, has specific resource requirements that imply intense natural selection for survival in extreme environments. Therefore, these species are less able to avoid environmental stress by habitat selection .
Considering a very simple model of colonisation, transposition rates in colonisers can be estimated. We assume that sites originating from Argentina during colonisation are found at high frequency in colonising populations and that new transpositions correspond to low frequency sites. The estimations begin with the historical data, which suggest that Australian colonisation of D. buzzatii occurred approximately 70 years ago [31, 43] and assume a mean of 12 generations per year. Taking the pooled data from the two Australian populations (155 genomes), we estimate 840 generations of colonisation (70 years multiplied by 12 generations), 93 new sites and a mean copy number for Osvaldo insertions of 1.75. We estimate a minimum value of transposition rate per site per generation in colonisers, for Osvaldo, of 4.08 × 10-4 (93/155 × 1/840 × 1.75) and for Isis of 3.62 × 10-4. These values have the same order of magnitude as those calculated for Osvaldo (1.28 × 10-4) in previous studies with natural populations from the Old World . No differences in transposition rates were observed between the two colonisations, or between the two transposable elements considered in this study.
We cannot completely disregard local transposition events in some coloniser populations. It is well known that transpositional bursts have occasionally been found in Drosophila melanogaster laboratory stocks , other Drosophila species  and in plants subjected to different environmental stresses . However, transposition bursts are difficult to see in nature in real time because TE copies involved in these events are later silenced owing to their deleterious effect to the host. The majority of these mechanisms include host factors as methylation , deletions , silencing by RNA interference  or random mutations, excisions and purifying selection .
In the case of D. buzzatii colonisation processes, we can accept that the populations analysed have not suffered large transposition rate changes. Another possibility is to imagine that even if colonisation induced local transposition it could be unnoticed owing to the time elapsed, especially if the TE eliminations are quick, as observed in the Helena element in genomes of D. melanogaster and D. simulans . Old World colonising populations of D. buzzatii contain TEs where active and inactive copies coexist  and the low occupancy sites observed could be the result of a unique transposition or very few master copies present in the genome. However, the possibility that some low occupied sites harbored copies that have diminished their frequency in populations because of drift or selection effects against insertions with a deleterious effect cannot be disregarded. A colonising population from the Old World (Carboneras) demonstrated a decrease in the insertion frequency of many highly occupied sites (2B2a, 2F4a). In other studies concerning D. subobscura colonising species, the existence of common low frequency sites of bilbo and gypsy elements was attributed to a decrease in insertion frequency in some sites in coloniser populations . However, we hypothesize that high occupancy sites correspond to sites that have increased their frequency by drift associated to the colonisation process. Arguments in favor of this hypothesis are that sites of both elements, found at a low frequency in the original populations, are now at a high frequency in populations resulting from the two colonisations. However, a common highly occupied site (2B2a) of Osvaldo and Isis elements was observed in the Australian populations. This site could be a transposition hot spot induced by the colonisation process as it is present at a very low frequency in some original populations. Moreover, the two elements are inserted in the same chromosomal band, confirming previous results where Isis retrotransposon appeared to have preferential insertions inside Osvaldo sequences .
Comparisons of TE copy numbers between chromosome X and autosomes, after elimination of high insertion sites, demonstrated that Isis appears to be controlled by purifying selection in Australian populations, whereas the selection effect is not so evident for Osvaldo. However, the comparison is highly significant when the two Australian populations are pooled, which may suggest a stronger selection intensity against Osvaldo insertions in these populations. These differences may be due to a stronger selection effect in the Australian colonisation than the Old World one. Alternatively, the differences observed between the two elements could be due to the existence of differential transposition and regulation mechanisms. Another explanation may be that selection effects on the Osvaldo element could go unnoticed if transposition events occurred only a few generations ago. A third possibility is that selection acts more strongly on the Isis element (for example, if the element is inserted inside genes or regulatory regions) than on Osvaldo and, provided the Australian colonisation is recent, its effects are only detected in the element under the most selection pressure.
Comparisons of the proportion of elements between autosomes demonstrated large differences due to an over-representation of Osvaldo on chromosome 2 (58-63% from the total) and of Isis on chromosome 3 (37-44% from the total). D. buzzatii Australian populations have one inversion on chromosome 2. Inversions [23, 51], inversion break-points and nearby regions [52–54] are considered by some authors as sites of accumulation of TEs. These regions are characterised by a reduction in recombination rates that diminishes the probability of deleterious ectopic exchanges. Correlations between highly occupied sites of Osvaldo and the J arrangement produce negative values; only two sites were significant: one located outside and another inside the J inversion. The most likely explanation is that most correlations observed are a consequence of the founder effect. This would explain why significant negative correlations are observed even in sites located inside the inversion. A similar interpretation was evident in a study concerning colonising D. subobscura populations, where the founder hypothesis was favored by the fact that all correlations between sites and arrangements were significant only in the colonising populations . Moreover, there were positive associations between chromosomal arrangements and highly occupied sites located outside of inversions. This result is not unexpected if we take into account that the effective population size is reduced during the first stages of colonisation. If the TE abundance is regulated by the effect of ectopic recombination between elements , the existence of insertions in a homozygous state could reduce ectopic exchange events [35, 55] outside inversions, leading to a relaxation of selection pressure.