AFLPs
The DNA quality, the reproducibility of AFLP chromatograms, the different patterns observed in E. canadensis compared to E. densa and L. major and the similar geographical separation of Otamangakau and Taupo South populations both in E. canadensis and in L. major, indicate that AFLPs are appropriate for the study of genetic variation in these clonal species. If the observed genetic patterns were due to artefacts of the AFLP technique, similar results would have been expected in the three species, considering the similarly low levels of genetic variation.
Number of introductions
Low levels of genetic diversity were found in all three species, which suggest one introduction or multiple introductions of similar genotypes for each of the three species in the northern part of the North Island, NZ. High levels of genetic diversity in invasive species may be the result of multiple introductions [1, 17]. However it is not possible to exclude multiple introductions even in cases with low genetic diversity, as the introduced genotypes might be highly similar to begin with. The genetic variation pattern in the native range is an important reference point of the natural variability within an invasive species and can support the hypotheses of single vs. multiple introductions. The situation in invasive species can, however, be more complicated than this, as they may be derived from areas in the introduced rather than the native range, which can be expected to give rise to different patterns. Such a high level of genetic similarity as that observed in this study was unexpected and is an important finding for further genetic research in these and other invasive species. A geographically wider set of samples, including many native and introduced populations, would need to be analysed to conclusively address the number of introductions in NZ. In the case of L. major, genetic variation within and between populations in the native range in South Africa was studied by Triest [18] using isozymes. Most of the populations were unisexual and monoclonal. Low variation levels were recorded also between populations, indicating predominant vegetative propagation. Evidence of sexual reproduction was provided by the presence of one heterozygotic clone in an otherwise homozygotic, bisexual population. The polymorphism between this clone and the most spread one was 9.5%, which, interestingly, is comparable with the level of polymorphism found in our study (9.8%). However, in the NZ populations this level of diversity was based on the whole sample set. The two most different genotypes had a polymorphism of 5.6%, and it decreased to 3.5% and 2.1% if differences between these clones and the most widespread one were considered. Our study shows lower levels of polymorphism than between the two different genotypes in the native range; however some more samples from South Africa should be analysed to estimate the variability in the native range. In addition, the different markers applied in the two studies (isozyme vs. AFLPs) are not totally comparable because the amount and the parts of the genome analysed in the two studies are very different and AFLPs are designed to detect higher levels of intraspecific polymorphism than isozymes.
Genetic diversity was studied in introduced populations of E. densa in Oregon (USA) and Chile using RAPDs by Carter and Sytsma [19]. Very little variation was found in these populations, and surprisingly the same genotypes were found in Oregon and Chile, suggesting low genetic diversity in the native source populations and/or similar introduction histories in the new ranges. Clones were genetically very similar to each other and the very short genetic distances between strains, as indicated by the UPGMA tree suggest a situation siMilar to that in the present study even though the molecular techniques used are different. The study of Carter and Systma [19] shows that multiple introductions of E. densa cannot be ruled out in the North Island, NZ. A stepping-stone model of colonization could, however, explain the similar genetic patterns in Oregon and Chile and hide higher levels of genetic diversity in the native range. In a study using allozymes, Kadono et al. [20] found genetically uniform populations of E. densa throughout Japan, but as in the case of isozymes, the technique yields a limited resolution compared to RAPDs and AFLPs.
To the best of our knowledge this is the first study of the genetic variation pattern of E. canadensis populations. In the congeneric invasive species Elodea nuttallii (Planchon) St John, both multiclonal [21] and monoclonal populations [20] were found in introduced ranges. In the Moder catchment (France), 2.7% AFLP polymorphism was detected in populations of E. nuttallii [22]. A single introduction was hypothesised and the low level of variation that did exist was attributed to somatic mutations. Interestingly, two different mutants were found in two of the ten investigated populations and, like in our study, the UPGMA tree showed geographical structure in the distribution of genetic variation. Multiple introductions in NZ from Tasmania were discussed by Thomson [15] for E. canadensis. If this is correct, genetic diversity had to be very limited in the source populations, suggesting a stepping-stone model of invasion in NZ. We do not know the genetic pattern in Tasmania and how many source populations were involved. We can only move the discussion about the number of introductions from NZ to Tasmania. The differences with the Danish genotype suggest that higher levels of DNA variation are present within the species, but E. canadensis is invasive also in Europe and different evolutionary patterns could explain the divergence of allopatric gene pools.
Not knowing the genetic variation of these species in their native range, we cannot draw conclusions about the number of introductions to NZ (or Tasmania). Two possible scenarios emerge, however, from this study: multiple introductions of genetically similar genotypes, or one single founding event (consisting either of one genotype or a set of genetically similar genotypes) and evolution in situ by somatic mutations. The first scenario would entail either low genetic diversity in the native range or a stepping-stone model of colonization from other areas of the introduced range, as appears to be the case for E. canadensis. The second scenario, involving evolution of genetic diversity in NZ by somatic mutations is consistent with our results. A third model, which parsimoniously combines the previous two, limiting both the number of dispersal events and the number of mutations to have occurred in NZ, is that of ancestral polymorphisms retained in the three species [23].
Sources of genetic variation
Several results of this study indicate that evolutionary processes occurred after these species were introduced in the northern part of NZ's North Island. The measures of population differentiation show that the time elapsed since introduction is reflected in the genetic structure of the populations. Elodea canadensis was the first of the three species to be introduced in 1868 and its populations have higher Fst values and Nei's genetic distances than the later-established populations of E. densa and L. major. Also, the pattern of E. canadensis differs from that of the other species in showing more monophyletic groups, a continuum of genetic differences between individuals and populations, and differentiation between the geographically close populations of Taupo South and Otamangakau from the downstream locations. Strikingly the populations of Taupo South and Otamangakau were found to be a monophyletic group also in L. major, which was introduced in NZ as recently as in 1950. This can be explained either by recent "co-dispersal" of the two species (after 1950), or by independent dispersal histories shaped by the same geographic/physical constrains. In both cases multiple independent introductions appear very unlikely, as those would have blurred any common pattern. Considering that only one sex is present, or dominant, in each species, the absence of seeds and evident outcrossing events (recorded in the spectrum of pairwise genetic distances), somatic mutations seem to be a plausible source of genetic diversity in these clonal species.
The spectrum of pairwise genetic distances has previously been utilized to study genetic diversity in populations of clonal outcrossing plant species, as it may be helpful in distinguishing between genetic variation produced by somatic mutations and genetic variation due to outcrossing. The spectrum of clonal outcrossing populations of the marine submerged Posidonia oceanica (L.) Del. [24] shows a bimodal distribution, with a large peak at zero pairwise distance (α peak) indicating clonal reproduction, followed by decreasing frequencies in pairwise distances, due to somatic mutations. The second peak of the spectrum (β peak) is the mode of an almost normal distribution of frequencies in pairwise distances, which has been attributed to genetic diversity produced by outcrossing events. The same spectrum was obtained in outcrossing populations of the clonal tree Populus tremuloides Michaux [25]. Compared to these studies, the spectra of L. major and E. densa populations show the same pattern that has been considered indicative of clonal reproduction and somatic mutations. Contrastingly, the spectrum of E. canadensis populations does not fit with this model. The normal distribution of pairwise differences points to outcrossing genotypes, whereas the limited range of polymorphism, which is comparable to that of E. densa, points to somatic mutations. The comparison with a genotype from Europe shows higher numbers of polymorphic fragments with all NZ genotypes. In addition if the spectrum was the result of outcrossing events between the NZ genotypes, outcrossing would be more frequent than clonal propagation, as the frequency of "zero" pairwise differences is very low, and seeds would be common in the populations. Seeds have, however, never been observed in E. canadensis in the northern part of NZ's Northern Island [14]. However, the lack of β peak is not a conclusive evidence of absence of outcrossing. Sexual reproduction between clonal stands with the same or highly similar genotypes is not supposed to produce a β peak. Pairwise difference frequencies are also affected by somatic mutation rates as well as by population sizes at the time of introduction and at present, and by genetic diversity at the time of introduction. Several factors might interact in E. canadensis's spectrum, such as a pool of genetically similar/closely related founders, possible mutation-drift disequilibria and a longer introduction time.
Genetic diversity levels are different in the three species, in relation to the initial gene pools, possible ancestral polymorphisms, different somatic mutation rates and establishment times. Egeria densa, which in the North Island of NZ has the typical spectrum of a clonally reproducing species that has accumulated somatic mutations, has more genetic diversity than the other macrophytes. This makes E. densa more adaptable. In invasive populations of the clonal aquatic species Hydrilla verticillata (L. f.) Royle, Albrecht et al. [26] demonstrated that herbicide resistance evolved by somatic mutations.
Dispersal
In the absence of evident sexual reproduction, the distribution of genetic diversity in the populations documented in this study is found to be due to the dispersal of vegetative propagules, moderated by the presence of possible geographic/ecological barriers. The analysis of the geographic distribution of DNA polymorphic fragments showed more frequent dispersal events between neighbouring populations, even though long-distance dispersal was also common. Multiple introductions in lakes and rivers appeared also to be a recurring phenomenon, as well as down- and up-stream dispersal. The co-occurrence of more invasive species in the same localities and similarly complex dispersal routes in all three species suggest that human dispersal had a major role in the distribution of the genetic diversity. Johnstone et al. [27] ruled out bird-mediated dispersal because the distribution patterns of these species in NZ were not random in nature, but linked to fishing and boating activities. Considering that the populations sampled in this study were from a wide range of different habitats [28], dispersal opportunities combined with an empty niche [29] rather than suitable environmental conditions, explain the distribution of these species in the northern part of NZ's North Island, as also found by de Winton et al. [12]. As discussed by Howard-Williams [30], NZ has no native canopy-forming submerged aquatic plants. Some kind of barrier seems, however, to limit recruitment and dispersal from Otamangakau and Taupo South from/to downstream populations in E. canadensis and L. major. Surprisingly, in L. major, these populations are isolated also from Lake Rotoaira, which is geographically close and connected by waterways to Otamangakau. A wider set of samples and populations covering also the southern part of the North Island and the South Island, combined with historic records of first introduction date in each location would help better understand the relationships between populations and reconstruct the introduction history of these species in NZ with higher resolution. It would also greatly enhance interpretation of nodes of dispersal vs. recruitment, providing a valuable tool for the local management of these aquatic weeds.