Comparative Analyses of Plastid Sequences between Native and Introduced Populations of Aquatic Weeds Elodea canadensis and E. nuttallii

Non-indigenous species (NIS) are species living outside their historic or native range. Invasive NIS often cause severe environmental impacts, and may have large economical and social consequences. Elodea (Hydrocharitaceae) is a New World genus with at least five submerged aquatic angiosperm species living in fresh water environments. Our aim was to survey the geographical distribution of cpDNA haplotypes within the native and introduced ranges of invasive aquatic weeds Elodea canadensis and E. nuttallii and to reconstruct the spreading histories of these invasive species. In order to reveal informative chloroplast (cp) genome regions for phylogeographic analyses, we compared the plastid sequences of native and introduced individuals of E. canadensis. In total, we found 235 variable sites (186 SNPs, 47 indels and two inversions) between the two plastid sequences consisting of 112,193 bp and developed primers flanking the most variable genomic areas. These 29 primer pairs were used to compare the level and pattern of intraspecific variation within E. canadensis to interspecific variation between E. canadensis and E. nuttallii. Nine potentially informative primer pairs were used to analyze the phylogeographic structure of both Elodea species, based on 70 E. canadensis and 25 E. nuttallii individuals covering native and introduced distributions. On the whole, the level of variation between the two Elodea species was 53% higher than that within E. canadensis. In our phylogeographic analysis, only a single haplotype was found in the introduced range in both species. These haplotypes H1 (E. canadensis) and A (E. nuttallii) were also widespread in the native range, covering the majority of native populations analyzed. Therefore, we were not able to identify either the geographic origin of the introduced populations or test the hypothesis of single versus multiple introductions. The divergence between E. canadensis haplotypes was surprisingly high, and future research may clarify mechanisms that structure native E. canadensis populations.


Introduction
Non-indigenous species (NIS) are species living outside their historic or native range [1]. Invasive NIS often cause severe environmental impacts [2][3][4], and may have large economical and social consequences [2,3,5,6]. Human activities, such as agriculture, aquaculture, recreation and transportation promote both intentional and accidental spread of NIS [7] and the rate of invasions by NIS is accelerating [8][9][10]. Freshwater systems are considered particularly sensitive to invasions for several reasons [11]. Aquatic environments are homogeneous on a large spatial scale and, therefore, aquatic plants can survive and establish more easily outside their native geographic range. Furthermore, aquatic environments are more difficult to monitor, and early detection of submerged species is seldom possible. In addition, water is an effective vector of propagules allowing dispersal over long distances [12,13].
In order to control the dispersal and effects of invasive species it is important to know the origin of invasive taxa, the size of the introduction, the level of genetic variation compared to the native range, and whether multiple introductions have occurred.
Molecular genetic methods have provided new tools to answer these questions [14][15][16]. Molecular markers may also help in early detection of invaders and, therefore, in preventing new invasions [17]. The loss of genetic variation due to the founder effect is a frequent feature in introduced populations compared to the diversity found across the native range [5,[18][19][20]. This combined with multiple introductions and adaptation to a new environment may result in rapid genetic changes in introduced populations [20,21], enhancing genetic differentiation between native and introduced populations [22] and increasing genetic diversity in individual populations in the invaded region [23,24].
The gene order and content of cp genomes are generally highly conserved [25] and the substitution rate in cpDNA is much lower than that in the nuclear DNA in plants [26,27]. Maternally inherited cpDNA is transmitted only through seeds and, therefore, it has less potential for gene flow than nuclear genes, which are spread also by pollen dispersal. Consequently, genetic variation in the cp genome is often more geographically structured than that in the nuclear genome, making chloroplasts valuable sources of genetic markers for intraspecific phylogeographic analyses [28][29][30][31]. Despite limited variation, cpDNA data have been successfully Table 1. List of Elodea canadensis and E. nuttallii specimens used in this study and associated haplotypes based on the observed sequence differences within nine combined chloroplast genome regions (total 4072 bp). analyzed and used to trace the origin and biogeographical history of many invasive plants [15,17,[32][33][34][35][36][37][38]. Several loci useful for phylogeographic studies have been discovered and exploited since the pioneering work in 1991 [39]; however, many polymorphic loci still remain undetected. The non-coding cp genome regions widely used in phylogenetic and phylogeographic studies might actually be among the least variable ones, whereas the most variable regions are rarely exploited [40,41]. The number of completely sequenced cp genomes is growing rapidly [28], and lower sequencing costs enable genome-wide intraspecific comparative studies to help detect the most variable genomic regions. Elodea (Hydrocharitaceae) is a New World genus with at least five submerged aquatic angiosperm species living in fresh water environments. The native range of E. canadensis, E. nuttallii and E. bifoliata is temperate North America, while E. potamogeton and E. callitrichoides are native to South America [32]. In the present paper, we analyzed the phylogeographical cpDNA variation of two invasive species E. canadensis and E. nuttallii, which have been introduced to other continents [42]. Elodea canadensis was brought to Europe in 1836, first to the UK [43], and at present it is widespread in the whole Europe. It was introduced to New Zealand in 1868 [44], to Australia in 1931 [45], and it has been considered a noxious weed also in many regions of Asia and Africa [46]. Elodea nuttallii was reported in Europe in 1939, first in Belgium [47]. It has not yet been found in the northernmost Europe, but it is likely to spread to new areas with a high risk of being invasive [47,48]. The species was introduced to Japan in 1961 [49] and to China in 1980 [50], while it has not yet been found in Australia or New Zealand. Both Elodea species were reputedly brought to Europe as aquarium plants or with timber [42]. Once in Europe, water currents and birds have spread these plants locally, while botanists and botanic gardens were responsible for their long distance dispersal [42,51]. Vegetative reproduction by fragmentation or specialized buds dominates in both native and introduced populations, and in introduced populations sexual reproduction is extremely rare or totally absent [52,53]. Both species grow aggressively and form dense stands making the recreational use of lakes difficult. These mass occurrences may change the balance of lake and river ecosystems in many ways, e.g. by outcompeting native species, changing the pH and nutrient levels and reducing the oxygen concentrations of water column [46,47,54].
Both E. canadensis and E. nuttallii show a wide range of morphological variation and they are difficult to identify [33,55]. These species have been successfully discriminated by sequencing the nuclear ITS region [56]. However, ITS has proved to be problematic for several reasons, such as the existence of extensive sequence variation in several plant genomes [57]. Therefore, additional molecular identification tools would be valuable in monitoring risky species and in early detection of new invasions. In this study we compared the plastid sequences of native and introduced individuals of E. canadensis. We determined the distribution and location of the most variable regions between the two plastid sequences, and developed PCR-based molecular markers covering these regions. We used these markers to reveal the cpDNA variation in E. canadensis and E. nuttallii within native and introduced ranges, and tested the hypothesis that the genetic patterns of E. canadensis and E. nuttallii are more uniform in the introduced range due to a single introduction event. The aims of this study were to 1) survey unexplored areas of the cp genome suitable for phylogeographic analyses and for species discrimination between E. canadensis and E. nuttallii, 2) provide detailed genetic data of cpDNA variation within the native and introduced ranges of E. canadensis and E. nuttallii to test the hypothesis that the genetic patterns are more uniform in the introduced range due to a single introduction event, and 3) survey the geographical distribution of the cpDNA haplotypes across the invasions in order to reconstruct the spreading histories of E. canadensis and E. nuttallii.

Plant materials
The plant material for E. canadensis cpDNA extraction and cp genome sequencing was obtained from a native population in Minnesota, the United States of America (N47u14.9789, W95u14.7509) (Site number Ec14 in Table 1 and Figure 1). For sending the fresh plant material from the United States to Finland, a phytosanitary certificate was obtained from the United States Department of Agriculture. Leaf material for phylogeographic analyses was collected from 24 E. canadensis populations: seven from North America, 13 from Europe, three from New Zealand and one from Australia (Table 1, Figures 1 and 2). Leaf material was also collected from 13 E. nuttallii populations: ten from North America and three from Europe (Table 1, Figures 1 and 2). One to three samples were collected from each population. No specific permits were required for collecting the plant material and the field studies did not involve endangered or protected species. Moreover, herbarium specimens collected from the native range and stored at three herbaria (see Table 1) were included in the analyses. Nine herbarium specimens represented E. canadensis (five from the United States of America and four from Canada), while  Table 1, and shapes, shades and patterns correspond to the haplotypes in Figure 3. doi:10.1371/journal.pone.0058073.g001 12 represented E. nuttallii (all from the United States of America). Species identification was verified by sequencing the nuclear ITS region of each specimen.

Sequencing and primer development
The treatment of fresh E. canadensis plant material and the cpDNA extraction were performed as in Huotari and Korpelainen [58], using the extraction method based on a differential chloroplast lysis using a non-ionic detergent, Triton X-100 [59].  Table 1, and shapes and shades correspond to the haplotypes in    Figure 1) (GenBank accession number JQ310743) sequenced earlier [58]. The software Tablet [60] was used to check for the quality of contigs and to assure the order of contigs in relation to each other. Altogether we obtained 75 contigs related to the cp genome of E. canadensis. The contigs consisted of 3318 reads (3.4% of total reads) and were on average 1510 bp long. The contigs had a collective length of 112,193 bp, when only one copy of IRs was included; leaving 18,158 bp (13.9%) missing from the total genome. Several of the remaining gaps were too large to be closed via PCR and, therefore, all further comparative analyses of plastid sequences between the native and introduced populations of E. canadensis are based on this 112,193 bp-sequence of good quality (GenBank accession number for alignment will be provided).
In order to detect variable sites, the plastid sequences of E. canadensis collected from the native and introduced range were compared using Gap4 [61] and manually aligned using BioEdit [62] (Alignment S1). The total number and distribution of all SNPs, indels and inversions were recorded. In addition, indels consisting of mono-or dinucleotide SSRs were surveyed. We designed specific primers for the flanking sequences outside the most variable regions using Primer3 (http://frodo.wi.mit.edu/ primer3/). The goal was to choose both coding and non-coding regions informative in species identification and valuable in phylogeographic analyses. Altogether, 29 primer pairs were developed containing 78 SNPs, one inversion and 24 indels, of which 15 were mono-or dinucleotide SSRs (see the total list of primer sequences and polymorphisms in Table 2). All these primers were located within single-copy regions (LSC and SSC) of the cp genome. The average coverage of single-copy regions in the pyrosequencing was 4.8. In our analyses, we only accepted polymorphic sites with 4-fold coverage or more, verified using Tablet [60], as true differences between the two plastid sequences. Additionally, variable sites with a lower coverage verified by direct sequencing with primers developed were taken into account.

Molecular methods
For the phylogeographic analyses, genomic DNA was extracted from dried leaf tissue of 70 E. canadensis samples collected from 16 native and 17 introduced populations, and 25 E. nuttallii samples collected from 22 native and 3 introduced populations. The DNA was extracted using a commercial kit (E.Z.N.A. Plant DNA Mini Kit, Omega Bio-Tek) following the manufacturer9s protocol. All these samples were analyzed using nine, potentially highly variable markers comprising 4072 bp (see Table 2). Additionally, we used all the designed 29 primer pairs comprising 11,464 bp ( Table 2) to compare the level of intraspecific variation between the two E. canadensis plastid sequences (Ec14, hereafter USA and Ec17, hereafter FIN in Table 1) to that of interspecific variation between E. canadensis and E. nuttallii (En1, Table 1) plastid sequences. All PCR amplifications were performed in a total volume of 20 ml containing about 5 ng genomic DNA, 13 ml ddH 2 O, 16 reaction buffer including 1.5 mM MgCl 2 , 1 mM dNTP mix, 0.6 ml Dynazyme II DNA Polymerase (Finnzymes) (2 U/ml) and 5 pmols of both primers. The thermal cycler was programmed for 5 min denaturation at 94uC, followed by 35 cycles of denaturation at 94uC for 30 s, annealing at 50-60uC for 30 s, and elongation at 72uC for 1 min, and a final 10 min extension at 72uC. The amplification products were separated on 1.2% agarose gel, and clear bands were excised and purified with a commercial kit (E.Z.N.A Gel Extraction Kit, Omega Bio-Tek). Purified products were sent for DNA sequencing (Macrogen Inc., South Korea). Alignments of sequences were adjusted manually using BioEdit [62] and trimmed to the same length.

Data analysis
The sequences obtained using 29 cpDNA markers comprising 11,464 bp ( Table 2) were compared using BioEdit [62] to detect the level of variation between native and introduced individuals of E. canadensis, and between E. canadensis and E. nuttallii. In addition, all collected E. canadensis and E. nuttallii specimens were characterized for cpDNA haplotypes based on the nine selected cpDNA regions comprising 4072 bp (Table 2). These regions were chosen to represent both coding and non-coding cp genome regions, and to contain SNP-, SSR-and indel-variation based on the initial comparison between the two E. canadensis plastid sequences USA and FIN. The geographical distributions of the cpDNA haplotypes were analyzed in native and introduced ranges of E. canadensis and E.nuttallii. Haplotype networks from the sequence data were constructed using the network building software TCS 1.2.1 [63], which uses statistical parsimony and the genealogical reconstruction algorithms of Templeton et al. [64]. We treated indels as a fifth state [65,66], and coded each indel as a single mutational event [67]. One E. canadensis haplotype (H1) could not be connected by TCS and is linked to other haplotypes with a dashed line in Figure 3. This haplotype was connected manually with the observed mutational differences (this is not intended to connect the network but used to demonstrate minimum distances between unconnected haplotypes).

Results
The 112,193 bp of cpDNA sequence compared between native (USA) and introduced (FIN) E. canadensis plastid sequences contained altogether 235 variable sites (186 SNPs, 47 indels and two inversions) ( Table 3). Most variation occurred within the non-coding regions of LSC (Figure 4). The substitution rate within the single-copy regions was 0.00181 per nucleotide, while the rate in the inverted repeat region was 0.00119. The observed ratio (R) of transitions (Tn) to transversions (Tv) was 0.74. Of the 186 SNPs detected, 79 were Tn and 107 were Tv (Table 3). In all, 109 SNPs were located in non-coding regions while 77 SNPs were found within coding regions. Of these, 42 SNPs were nonsynonymous and 35 were synonymous substitutions (Table 3, Figure 4).
A total of 47 indels were identified between the USA and FIN plastid sequences, of which four were located within coding regions (rpoC1, rpl23, ycf1 and ycf2) ( Table 3) and four in introns (trnG-GCC, ycf3, petB and rpl16). The three longest indels were 141 bp (between ycf2 and trnL-CAA), 38 bp (between ndhC and trnV-UAC) and 23 bp (between petN and psbM) in length, while the rest of the indels were 10 bp or less in length. Altogether, 28 indels located in LSC were mono-or dinucleotide repeat regions (SSRs) showing length polymorphism. Of them, 25 SSRs were mononucleotide repeats (46A, 196T, 16C, 16G) with an average length   (Table 4). Two structures were interpreted as minute inversions, typical for non-coding regions of cpDNA [68,69]. The flanking sequence outside the inverted regions was explored, and a hairpin structure was found from the inversion located in rps16 intron, consisting of a stem of eight bp and a loop of three inverted bp (GTA/TAC). Another potential inversion (CG/GC) was found in ycf1 gene, but it did not form a hairpin. 29 cpDNA markers comprising 11,464 bp ( Table 2) revealed 102 variable sites (77 SNPs, 24 indels and one inversion) between the two E. canadensis plastid sequences (USA and FIN), 216 variable sites (191 SNPs, 24 indels and one inversion) between E. canadensis (FIN) and E. nuttallii (En1) plastid sequences and 226 variable sites (205 SNPs, 18 indels and two inversions) between E. canadensis (USA) and E. nuttallii (En1) plastid sequences ( Table 2). We discovered on average 61% more SNPs between the two Elodea species than between the two E. canadensis plastid sequences, while the number of indels was 10% larger within E. canadensis comparison. Both inversions found between E. canadensis and E. nuttallii formed a hairpin structure similar to the one found between the two E. canadensis cp genome sequences. One inversion was the same as between E. canadensis cp genomes (GTA/TAC) located in rps16 intron, and the other inversion was located in noncoding region between petN and psbM genes, consisting of a stem of 12 bp and a loop of three inverted bp (TAT/ATA). On the whole, the level of variation between the two Elodea species was 53% higher than that within E. canadensis (Table 2).
In the phylogeographic analysis using nine cpDNA markers comprising 4072 bp (Table 2) we found three E. canadensis and four E. nuttallii haplotypes (Figures 1 and 3) (Alignment S2). The haplotype network constructed indicates that E. canadensis haplotype H1 is very divergent from haplotypes H2 and H3, as it could not be connected within the limits of parsimony (95%) (Figure 3). Haplotypes H1 and H2 were separated by 48 mutations including eight indels, while haplotypes H2 and H3 were separated by just one mutation (Figure 3). In E. nuttallii, haplotypes A and B, separated by two mutations, were divergent from haplotypes C and D, which differ from each other by two mutations (Figure 3).
Only a single haplotype was found in the introduced range in each of the two species. These haplotypes H1 (E. canadensis) and A (E. nuttallii) were also widespread in the native range, covering the majority of native populations analyzed (H1 88% and A 86%) (Figure 1). The largest divergence between E. canadensis haplotypes was about four times larger than the largest divergence between the E. nuttallii haplotypes.

Plastid sequence polymorphisms
Most intraspecific studies of cpDNA variation have utilized a limited number of markers from a few selected gene regions. Only few studies have examined intraspecific variation in the whole cp genome sequence [70][71][72]. We were now able to investigate variation in an 112,193 bp (86.1%, only one IR included) region of two E. canadensis plastid sequences and found a total of 235 variable sites. In general, the mutation rate of SNPs (often about 10 28 -10 29 ) is low compared to SSRs (about 10 24 ). On average, one SNP can be expected every 500-1000 bp in coding regions and every 200-500 bp in non-coding regions [73]. The evolutionary rate of cpDNA is only half of that in nuclear DNA in plants [26] and, therefore, also the incidence of SNPs in cp genome is expected to be lower. We found on average one SNP every 863 bp in coding and every 420 bp in non-coding regions. The total frequency was one SNP every 603 bp. Based on these figures, the frequency of SNPs between E. canadensis cp genomes is higher than expected. The SNP rate in the IR region was five times lower than that in the single-copy regions. This result is consistent with a previous report indicating that the synonymous substitution rate of IR region is roughly five times lower than that of the single copy regions [74].
In our earlier study [58] we found a total of 83 mono-and 41 dinucleotide repeats of 8 bp or greater in a survey of the whole E. canadensis (FIN) cp genome. In this study, we identified polymorphic SSRs between E. canadensis USA and FIN plastid sequences and found 25 mono-and 3 dinucleotide repeats showing length polymorphism (Table 4). All polymorphic SSRs were located in LSC, whereas in the previous survey, 77.2% of SSRs were found in LSC [58]. Nine polymorphic SSRs discovered in this survey were shorter than 8 bp in both plastid sequences, and therefore they were excluded from the previous search for longer repeat regions. We also found two inversions as a part of a hairpin loop. These kinds of small inversions flanked by inverted repeats are ubiquitous in the cp genomes of angiosperms [69], and have been found between species and genera [68,69] as well as within species, for example in Lolium [71] and Abies [75].

Chloroplast DNA phylogeography
The number of haplotypes in an invasive range is a function of many factors, including the number of introductions, the size of each introduction, the population structuring in the native range, and subsequent drift and selection pressures that occur after the introduction [17]. In our phylogeographic analysis, only a single haplotype was found in the introduced range in both species (Figure 2). These haplotypes H1 (E. canadensis) and A (E. nuttallii) were also widespread in the native range, covering the majority of native populations analyzed (Figure 1). Therefore, we were not able to identify either the geographic origin of the introduced populations or test the hypothesis of single versus multiple introductions. The result could indicate one introduction event, but multiple introductions of the same haplotype are just as possible. The introduced populations of E. canadensis have been previously studied using AFLP and SSR markers. The analysis using AFLP markers detected low levels of genetic diversity among E. canadensis populations and suggested one introduction event or multiple introductions of similar genotypes to New Zealand [76]. An analysis performed using SSR markers revealed a moderate level of variation in introduced populations of E. canadensis [77,78]. However, the mutation rate of SSR markers is high [73], and these markers were developed in order to discover variation patterns in the introduced E. canadensis populations. Therefore, it is not surprising that the markers were polymorphic and revealed multiclonal populations even though only one cpDNA haplotype was found from the introduced range.
Despite the limited variation, the data presented here provides interesting information of the genetic patterns in the native and introduced ranges of E. canadensis and E. nuttallii. The cpDNA haplotype homogeneities at introduced range are most likely attributed to bottleneck followed by a single introduction event or few introductions of similar haplotypes. This pattern is further supported by the vegetative regeneration, combined to the relatively recent and fast expansions of both species in their introduced ranges. On the other hand, the divergence between E. canadensis haplotypes at the native range was surprisingly high, as haplotype H1 differed from haplotypes H2 and H3 by 40 SNPs and eight indels. In comparison, the largest difference between E. nuttallii haplotypes was 11 SNPs and one indel. The diverged haplotypes were also geographically clustered in both species, although, a more detailed sampling would be necessary to get further support for the results presented here (Figure 1). Overall, the level of variation between the E. canadensis plastid sequences (USA and FIN), was constantly higher than what has been discovered in other intraspecific comparisons in angiosperms [70][71][72]. On the other hand, the level of interspecific variation between E. canadensis and E. nuttallii was still 53% higher than intraspecific variation within E. canadensis based on 11,464 bp of cpDNA. The large divergence between E. canadensis haplotypes let us hypothesize the evolutionary event behind this phenomenon.
The divergent haplotypes could be explained by chloroplast capture, where the cytoplasm of one species is replaced by that of another species through hybridization or introgression, as reported to occur in natural ecosystems [79,80]. Chloroplast capture can occur in species with sympatric distribution and reproductive compatibility, and the process is facilitated by weak reproductive barriers between species [81]. There is some support for naturally occurring hybrids between E. canadensis and E. nuttallii [42]. However, most of the variable sites found between E. canadensis plastid sequences were not detected in the E. nuttallii plastid sequence, indicating that hybridization between these species could not explain the large divergence between E. canadensis haplotypes. Interestingly, the distributions of E. canadensis and E. bifoliata would enable a hybrid zone allowing the chloroplast capture event involving the largely parapatric E. bifoliata. However, an extensive survey of genetic characteristics of E. bifoliata would be needed to firmly retrace the possibility of chloroplast capture.
Another reason for this high haplotype divergence in E. canadensis might be species mis-identification, but this is unlikely. We used the ITS region for species identification in all specimens and did not find any contradictory results. This region differs  between E. canadensis and E. nuttallii by nine indels and 27 point mutations [56]. Therefore, it may be expected to differ also between E. canadensis and E. bifoliata, which are morphologically more differentiated. The ITS regions of morphologically similar Hydrilla verticillata (GenBank accession number AY870353.1) and Egeria densa (GenBank accession number AY870360.1) are clearly different from those of E. canadensis and E. nuttallii leaving no risk of confusion. In addition, neither E. canadensis nor E. nuttallii are usually confounded with other Elodea species based on morphology. On the other hand, we found several variable sites between E. nuttalli haplotypes as well, and the level of variation between E. canadensis and E. nuttallii plastid sequences was still 53% higher than that between the two E. canadensis plastid sequences. In addition, while ITS sequences of E. nuttallii have been found to be highly homologous, those of E. canadensis include several polymorphic sites [56]. Therefore, the possibility of a higher level of naturally occurring variation in both nuclear and cp genomes of E. canadensis cannot be ruled out based on our results.

Chloroplast genome as a source of genetic markers
Elodea canadensis and E. nuttallii are morphologically very variable and, therefore, difficult to discriminate [33,47,55]. Also in this study, 38% of all sampled populations, were misidentified when using morphological characteristics. Even though the nuclear ITS region has been reported as a reliable identification tool to discriminate E. canadensis and E. nuttallii [56], additional molecular markers for species identification are useful especially for monitoring purposes and for early detection of new invasions. The cpDNA markers developed in this study are valuable tools for species identification between E. canadensis and E. nuttallii. Of these markers, loci cp2367A (rps11, IGS (rps11 -rpl36) differed by at least 20 point mutations and 2381A (IGS (ycf1 -ndhF), ndhF) by at least 40 point mutations between the two Elodea species (Table 2). In addition, loci cp1110A (IGS (ndhC -trnV-UAC), cp1872A (IGS (ycf4 -cemA) and cp2328A (IGS (trnF-GAA -ndhJ) differed by at least ten point mutations ( Table 2). Most variable loci between the two E. canadensis cpDNA sequences were cp1345B (IGS (atpI -rps2), rps2), cp1872A and cp2381A ( Table 2). All the cpDNA markers developed in this study are potentially variable novel markers applicable to population level, phylogenetic and phylogeographic research in Elodea species and most likely among other angiosperms as well.

Supporting Information
Alignment S1 Alignment of the two Elodea canadensis chloroplast genomes (USA and FIN) in nexus format. GenBank accession number for FIN sequence is JQ310743. (NEX) Alignment S2 Chloroplast DNA sequence polymorphisms of three Elodea canadensis haplotypes (H1-H3) and four E. nuttallii haplotypes (A-D) revealed using nine cpDNA loci. For primer information and genomic regions amplified, see Table 2. (TXT)