Flower morphology in species and hybrids of the temperate Nymphaea subg. Nymphaea clade. (A) Sterile hybrid of Nymphaea leibergii × Nymphaea tetragona , first-day flower (Hellquist et al. 16531, NY266, Wekusko Lake, Manitoba); (B) sterile hybrid of Nymphaea leibergii × Nymphaea odorata , later-day flower (Borsch et al. 3333, NY510, Holeb Pond, Maine); (C) Nymphaea leibergii (Robson & Wiersema DR08-234, NY568, Minago River, Manitoba); (D) Nymphaea tetragona , first-day flower (Hellquist et al. 16525, NY265, Tramping Lake, Manitoba); (E) Nymphaea odorata subsp. odorata , later-day flower (Borsch & Wilde 3101, NY033, Sarasota Co., Florida); and (F) N. odorata subsp. odorata , first-day flower (Borsch et al. 3324; Vermont; not included in the sequence data sets of this study but illustrating the high infraspecific diversity in North America). Photo (C) by The Manitoba Museum, Diana Bizecki Robson (used with permission); all other photos by T. Borsch and C.B. Hellquist. 

Context in source publication

Context 1

... in” was set to 100 and 39.603 trees were sampled for calculating the consensus tree and the posterior probabilities (PP) of nodes. In the four runs of the rps4 – trnF data set likelihood values had con- verged after 20 000 generations or earlier; therefore, the “burn in” was set to 200 and 39.204 trees were sampled. Three of the primers (844, 873, and 878) yielded acceptable results, the remaining two (810 and (GGTA) 4 ) were excluded from further consideration either because of difficulties in interpreta- tion of the bands or excessive polymorphism (Appendix C provides ISSR genotype characteristics and the matrix). The unrooted neighbor-joining tree shown in Fig. 4 indicates the genetic similarity between the various individuals sampled. Apart from the isolated N. tetragona sample, the remaining individuals form two groups. One group includes all samples of N. leibergii and N. loriana and the two New England hybrids. The individuals of N. leibergii cluster into two geographical groups (Midwest and New England, respectively). All samples of N. loriana were depicted as having an identical ISSR banding pattern. The obvious variability among the two New England hybrids is noteworthy, in contrast to the uni- formity of the N. loriana samples and their clustering together with only N. leibergii . The second group comprises samples of N. odorata from the different localities. All the water-lily species mentioned here are members of the temperate Nymphaea subg. Nymphaea . Wiersema (1996), largely derived from Conard (1905), assigned the four known North American members of this subgenus to three different sections, sect. Nymphaea ( N. odorata ), sect. Xanthantha (Casp.) Wiersema ( N. mexicana ), and sect. Chamaenymphaea ( N. leibergii and N. tetragona ). Phylogenetic analysis (Borsch et al. 2007), as well as phylogeographic and evolutionary studies of the putative North American clade (Woods et al. 2005 a , 2005 b ) and of the Eurasian lineage (Volkova et al. 2010) have further elucidated the relationships among the members of subg. Nymphaea , although none of these studies included all of the taxa. The complete taxon coverage for the temperate clade of this study depicts N. leibergii and N. tetragona as sister species. This is congruently inferred for both plastid and nuclear sequence data (Fig. 4) with medium to high jackknife support and Bayesian posterior probabilities. Both species appear as monophyletic based on the inclusion of individuals from throughout their respective ranges into the molecular data set. For N. leibergii , which has been included for the first time into a phylogenetic analysis, this confirms its status as a well-defined North American endemic species in comparison with the circumboreal N. tetragona . Wiersema (1996) asserted this earlier based on comparative morphology. The Asian N. pygmaea , also included for the first time into a phylogenetic study, is inferred as sister to the N. leibergii – N. tetragona clade by nrITS (Figs. 4, 5) while rps4 – trnF data are inconclusive to its exact position in the small-flowered clade (Fig. 4). The slightly improved character sampling, by adding the rps4 – trnT spacer, compared with the phylogenetic analysis using trnT – trnF alone (Borsch et al. 2007), and the considerably increased taxon sampling has not significantly improved the understanding of the deeper nodes in the chloroplast topology of the temperate clade of Nymphaea ( Nymphaea subg. Nymphaea ). The Bayesian tree depicts a clade of N. mexicana and N. odorata , albeit without support, and an internal polytomy of the samples of N. odorata with a N. mexicana lineage, while the maximum parsimony tree (not shown) provides weak evidence for the monophyly of N. odorata evidenced in Borsch et al. (2007) and Woods et al. (2005 a ). A considerably larger proportion of the chloroplast genome needs therefore to be sequenced to capture enough hierarchical phylogenetic signal and resolve these nodes. The multilocus plastid phylogeny of Nymphaeales by Löhne et al. (2007) indicated that a considerable gain in resolution and support of the chloroplast tree can be achieved by this approach, but the taxon sampling was limited in that study. Interestingly, this study recovered a third haplotype for the Vermont populations of N. odorata (NY508 and NY512, both from Lake Memphremagog; a “C” in position 17 of the trnT–L spacer) as compared with the earlier detection of specific N. odorata subsp. odorata and N. odorata subsp. tuberosa haplotypes of the trnL intron (Woods et al. 2005 a ). This indicates that the evolutionary history of N. odorata in North America is probably far more complex than can be described by the recognition of just two subspecies. And by the same line of evidence, the infraspecific genetic diversity of this species appears to be distinctly higher than that of the boreal species ( N. leibergii , N. tetragona ), as is also evident from ITS data, chromosome numbers, and morphology (see Fig. 2; Wiersema 1997; Woods et al. 2005 a ). However, the new results in this study also suggest that further population sampling and extended chloroplast sequencing will improve our understanding of the evolution of temperate water lilies. For the nuclear ITS region, this study provides the first representatively sampled trees of subg. Nymphaea (Figs. 4, 5). The overall relationships appear well sup- ported with the dwarf-flowered water lilies (sect. Chamaenymphaea ) being sister to the Eurasian N. alba – N. candida clade. These two lineages are further resolved as sister to a well-supported New World clade of N. odorata s.l. and N. mexicana . Contrary to all other species, no differences were found among the ITS ribotypes of N. alba and N. candida . The latter was also observed by Volkova et al. (2010) sampling Karelian and Siberian populations, and may be explained by reticulate patterns in this Eurasian lineage. The sister-group relationship of N. mexicana and N. odorata is noteworthy because morphology suggests a first-branching position of N. mexicana (Borsch et al. 2007). Moreover, N. mexicana is the only warm-temperate to subtropical species of the Nymphaea subg. Nymphaea lineage, and therefore could have retained plesiomor- phic characters from tropical ancestors in a clade that has increas- ingly adapted towards more northern ranges with colder climates (Löhne et al. 2008). Further research will be needed to reliably reconstruct the chloroplast genome tree and to unravel the history of the nuclear genome, including potential ancient reticulation and genome duplication events. Sequence data unambiguously identify individuals of N. leibergii as the maternal parent and N. odorata as the paternal parent of N. loriana (Figs. 3, 4). The distribution of Nymphaea leibergii and N. odorata in northeastern North America is shown in Fig. 1. Two distinct regions of sympatry exist, one in northern New England and adjacent Quebec and a second in northern Ontario and southeastern Manitoba. The native range of N. odorata then extends southward from the area shown and N. leibergii has scattered populations further westward, so no other areas of sympatry are possible. Nymphaea odorata is more tolerant of open water and wave action than N. leibergii (J.H. Wiersema, C.B. Hellquist, and T. Borsch, personal observation, 2013), so even in the same body of water the two may be isolated. In the easternmost region, hybridization has been detected at both New England locations where the two species are known to coexist (Figs. 4, 5, and discussion below). Although hybridization may well occur in the second and larger westernmost region, no herbarium specimens of intermediate morphology are known from throughout this region that provide any evidence for this. Except for the extreme western part of this area, we have not conducted any field studies therein. While all known populations of the hybrid species N. loriana lie within the range of its maternal parent N. leibergii , of interest is the fact that these are some distance outside the range of the other parent, N. odorata . The two known Saskatchewan populations of N. loriana are separated by some 80 km, partly by an area known as the Pasquia Hills. Both sites are over 500 km removed from the nearest populations of N. odorata in southeastern Manitoba. A third population exists ca. 225 km east in central Manitoba, but still over 450 km from the current range of N. odorata . Because geographically distant populations of N. loriana are identical with respect to ISSR fingerprint and ITS sequence data presented here, a single origin may be postulated. ISSR data dif- ferentiate individuals of N. leibergii from Midwestern and New England populations (Fig. 6) while the species appears as clearly monophyletic. Genetic diversity within N. leibergii and also within N. tetragona appears to be low otherwise, and nuclear and plastid genomic regions sequenced here do not show any intraspecific variation in the two species. This is in striking contrast to N. odorata , where Woods et al. (2005 a , 2005 b ) found high levels of genetic diversity both in ISSR fingerprints and at the level of nuclear ITS sequences. This is confirmed in the present study, in which unique ribotypes for nearly all sequenced individuals of N. odorata were found. The genetic profile of the N. odorata × N. leibergii hybrids from New England is clearly distant from that of N. loriana (Fig. 6), further underscoring their independent origin as ...