The present study has revealed a broadly heterogeneous genetic structure of M. persicae at a global scale as evidenced by high allelic differentiation and relatively high FST values between certain populations, and the partitioning of genetic variation by the Bayesian clustering analysis. The observed genetic variation can be attributed to mode of reproduction, host-plant adaptation, differences between regions and dispersal.
Genetic diversity and reproductive mode
A moderate to high genic diversity was found at the intrapopulation level, as well as cases of both single and multi locus deviations from HW equilibrium which was associated with heterozygote deficiency. The populations on peach, which are expected to contain only or mostly cyclical parthenogenetic genotypes, showed HW deviations in one or both of M40 and myz9 loci. In the populations from herbaceous hosts, which probably consist of a mixture of obligate/functional and cyclical parthenogens, some heterozygote deficiency was observed in M35, M49 and M63 loci. Heterozygote deficiency in microsatellite loci appears to be common among aphid species (S. avenae, France , R. padi, France ) including M. persicae (France [7, 21], Australia  and Greece ) and it has been recorded in both sexual and asexual populations. However, some studies have found less deviation from HW equilibrium in sexual populations (M. persicae, Australia  and M. persicae, Greece ). In populations found on secondary hosts, HW deviation might be expected due to the presence of asexual lineages. Asexuality with strong clonal selection is likely to cause deviations from HW equilibrium in polymorphic loci, such as microsatellite markers, via hitch-hiking and evolution in clonal lineages. The direct effects of local clonal propagation were mediated in the current study by removing clone duplicates. In some cases, e.g., hI lineages of Rhopalosiphum padi (L.) in France  and M. persicae in Victoria Australia , a heterozygote excess has been found in asexual lineages. In R. padi this excess is attributed either to ancient loss of sexuality and the consequence of accumulated mutations or to a hybrid origin. In other cases, however, asexual lineages showed heterozygote deficiency and heterozygosity levels close to that of their sexual counterparts (hII R. padi lineages in France  and M. persicae in France ). This has been associated with a recent loss of sexuality and the time has not been sufficient to allow accumulation of mutations in asexual lineages. It has also been suggested that gene flow between sexual and asexual functional parthenogens producing males may be sufficient to prevent differences in heterozygosity accumulating between reproductive modes. Previous studies have discussed reasons why R. padi  and M. persicae  sexual populations show homozygous excess (selection, clonal expansion, Wahlund effect, inbreeding and other population effects). All of these studies concluded that null alleles were not responsible for the effects as these would have been detected during the scoring process. Wahlund effect of sampling from distinct gene pools in the same population may contribute to the homozygote excess at least in some populations that have been examined here. In support of this, the Bayesian analysis showed that some populations contain members of more than one genetic cluster.
Partitioning of genetic variation – host-plant and region
In general, the high FST values obtained in pairwise population comparisons and the estimated overall value (0.086) are among the highest reported in an aphid species using microsatellite markers [25, 26, 28] and in the same order as those reported in previous studies for M. persicae populations from Europe [7, 15] and Australia .
The Bayesian clustering and admixture analysis partitioned the genetic variation into three clusters, European (1), tobacco (2) and Australasian populations (3). Clusters 1 and 3 correspond to the generalist M. persicae persicae while Cluster 2 corresponds to the tobacco-adapted subspecies M. persicae nicotianae. Cluster members are spread over all continents and in most of the countries from which populations have been examined. These results support the hypothesis that the globalization of agriculture will have an immediate impact on the evolution of pest populations. Previous studies have provided more direct evidence of this through the spread of obligate/functional parthenogenetic genotypes (see detailed discussion below). In addition to anthropogenic activity, M. persicae populations will be influenced by natural mating and biological processes according to geographical region (Australasia vs. Southern Europe) and to tobacco adaptation, i.e., nicotianae vs. persicae (tobacco vs. other crops, tobacco or peach in tobacco regions vs. peach in non tobacco regions). In general, the proportion of membership for the tobacco Cluster 2 was greater in tobacco-growing areas. In addition, the genetic distance and the FST analyses supported the separation of the tobacco aphid populations as well as the regional population structure of persicae. It is worth noting that the separation between the peach population from eastern central Greece and the equivalent from northern Greece, as revealed by FST and Bayesian analyses (membership coefficient of CGP to European persicae: 0.55), was not as strong as observed in a previous paper (membership coefficient of the equivalent peach population to persicae cluster: 0.78, ). A possible explanation is that the two peach samples from Greece examined here (NGP and CGP) were a mixture of genotypes of both subspecies at a different ratio according to the region. An influx of the tobacco aphids (Cluster 2) into the peach orchards of eastern central Greece associated with Cluster 1 and the converse in peach orchards in northern Greece could be an explanation.
Our results suggest that certain alleles are associated with the genotypes of M. persicae feeding on tobacco. This can be considered as defining a tobacco adapted aphid 'genome' perhaps encoding a series of important enzyme variants for this specialisation as a result of continuous selection on this plant. This phenomenon is associated with the taxon M. persicae nicotianae and according to our results it is widespread and appears to have moved into countries where tobacco is not cultivated (UK and Slovenia) or the cultivation is limited (Canada). It is not surprising to find nicotianae on other crops, since it is able to colonize and reproduce on various herbaceous hosts within the vast host range of M. persicae s.l. [16, 29]. It is likely that the source of the UK nicotianae genotypes is Europe. Studies using the European suction trap network  have shown that M. persicae s.l. in Europe can migrate over southern England. Evidence that supports the continental origin of UK nicotianae is the identification in this study of a red nicotianae genotype found in the UK and also a tobacco region in southern Greece (and in Bulgaria, Fenton unpublished data). We also noticed that none of the M. persicae genotypes sampled historically in the UK (e.g., C, D, E, I, J, L in Fenton et al. [21, 31], are amongst the tobacco Cluster 2 aphids. This suggests that nicotianae has arrived rather recently in UK. The ability of the tobacco aphid to colonise new territories, even if its optimal host is not present, has interesting evolutionary implications. In addition to marker differences, it also differs physically from persicae. Generally a red nicotianae colour morph predominates in various parts of the world  and red colour populations have been associated with a complete or partial loss of sexuality [3, 32]. The red colour morph might have ecological advantages such as absorption of solar radiation  and lower choice selection by parasitoids . The red form present on tobacco plants in North America may be more resistant to organophosphorus insecticides than the green form . It has also been found that the red form of M. persicae s.l. mostly has an A1,3 autosomal translocation, which is linked to the E4-based resistance mechanism, whereas the same translocation only occasionally appears in the green form [11, 36]. Lastly, parthenogenetic lineages of the red form of M. persicae s.l. have shown better performance on tobacco plants than green ones . It has been hypothesised that adaptation to tobacco arose as a single evolutionary event in sexual populations, probably in East Asia where nicotianae was first reported as a pest [13, 38]. The tobacco-adapted population then established as permanently asexual populations in various regions. In some temperate regions the availability of peach favoured the return to a yearly sexual generation. The Bayesian clustering and admixture analyses in the present study revealed a genetic similarity of the nicotianae genotypes which strengthens the hypothesis that the adaptation to tobacco was a single evolutionary event.
The UK population and a potential link with Australasia
The UK population contained only 23 genotypes. Despite this it was the most diverse by every measurement. Approximately one third of the UK lineages belonged to each of the three Clusters. Given that peach is not openly cultivated in the UK, recent asexual populations appear to mostly develop from successive waves of colonising clones . Surprisingly, unlike the rest of the European populations, elements of the UK population were most like the Australasian population. Taking into account the admixture clustering plot, it seems that the UK is a good candidate as a source of exchange with the gene pool of the Australasian (mainly New Zealand) aphids. The earliest introductions of exogenous aphids to Australia and New Zealand were likely to be associated with settlers from Europe, especially from the UK. In support of this, previous studies revealed many common microsatellite alleles between M. persicae genotypes from Australia  and Europe . Europe does also seem to be the origin of other non-indigenous Australasian aphid pest species such as Elatobium abietinum (Walker) (Hemiptera: Aphididae) . In the present study, parameters were similar or higher when the New Zealand population were compared with European populations. Moderate or high genetic diversity has also been reported in previous New Zealand  and Australian studies [4, 6]. Theoretical and empirical work suggests a general pattern of loss of genetic diversity during colonization [42, 43]; this is because emigrant populations are serially bottlenecked [44, 45]. The substantial genetic variation observed in New Zealand and Australian M. persicae suggests that the species has not been bottlenecked and this could be attributed to sexual reproduction [41, 46] and the time that it has been there. In Australia the species was first recorded in 1910 , although it is believed to have existed there since at least 1893. This period is adequate for the mutation of new microsatellite length alleles in asexual lineages  and for sexual reproduction to give rise to diversified genotypes. In the aphid Schizaphis graminum (Rondani) (Hemiptera: Aphididae), which was introduced in the USA in the 1880s, sexual reproduction was considered as the main reason for the high diversity observed . The UK M. persicae population is believed to lack holocyclic forms and therefore restoration of the full sexual cycle through mutation or though breeding between clones with partial loss of sexual reproduction (functional parthenogenesis)  may have been required. Nevertheless, introduction of sexual clones cannot be excluded. Another factor is possible multiple introductions of M. persicae in New Zealand. Van Toor et al  reported that clones NZ2 and NZ3 appeared to be introduced as they contained many unique alleles when compared to the remaining NZ population. Additional support for the existence of multiple introductions was found when NZ3 was recognised as being a common asexual M. persicae clone found in Scotland (clone D in Fenton et al. ). Exchange of genotypes between New Zealand and Australia could also occur as demonstrated for Sitobion genotypes . The two Australian lineages examined in the current study had 17 of the 21 alleles recorded in common with the New Zealand population suggesting a recent common origin.
The present study revealed some genotypes that were sampled many miles apart in different countries some of which had been identified before, e.g., Clones B (UK and Turkey) and D (UK and New Zealand) [31, 41], and others we identified for the first time in the current study e.g., Clone M in UK and Slovenia; a genotype found in France and Greece, another in UK, Greece and Bulgaria and another in southern Greece and Slovenia. These studies have suggested that widespread clones appear to occur as a result of selection for insecticide resistance in agriculture [9, 31, 41, 52]. We have examined a relatively small number of individuals in the M. persicae population, yet we have detected these clones. This suggests that the number of successful insecticide resistant genotypes is still relatively limited, despite the possibility of resistance genes combining into more and more genotypes in sexual populations every year. In addition to the spread of resistant clones, it has also been found that asexual tobacco aphid lineages have spread between neighbouring countries such as Greece and Italy  and in the current study a widespread nicotianae lineage was found in southern Greece and Slovenia and another in Greece, UK and Bulgaria. We also report here that a distinct tobacco lineage has been found in Greece and Chile. A previous study found only one microsatellite genotype of the tobacco aphid in Chile and it seems likely that the lineage we have identified is the same as that reported by Fuentes-Contreras et al. . These studies suggest an old world origin of southern American nicotianae as the subspecies exhibits genotypic variation in Europe , but none in Chile . During the last decade several studies have revealed that the spread of certain genotypes over distant geographical areas is a common phenomenon among aphid species (e.g., Sitobion , S. avenae [25, 28, 53, 54]) including M. persicae [4, 8, 31, 41]. The rapid spread of the M. persicae s.l. lineages in different countries and continents should be attributed mostly to human transport and commerce. While winged aphids may be transported very rapidly over great distances by low level-jet streams  other studies have found that particular genotypes remain localised . The widespread lineages reported in the current work probably represent asexual genotypes reproducing parthenogenetically all-year-round. This trait enables them to spread because their reproduction will not be altered by temperature, day length or the requirement for peaches to complete their life cycle. These clones might represent 'general-purpose genotypes'  with broad ecological tolerance, which predominate in fluctuating environments through selection, although anthropogenic activities, e.g., insecticide selection pressure might also be involved [9, 52].