Genome size and phenotypic variation of Nymphaea ( Nymphaeaceae ) species from Eastern Europe and temperate Asia

Despite long-term research, the aquatic genus Nymphaea still possesses major taxonomic challenges. High phenotypic plasticity and possible interspecific hybridization often make it impossible to identify individual specimens. The main aim of this study was to assess phenotypic variation in Nymphaea taxa sampled over a wide area of Eastern Europe and temperate Asia. Samples were identified based on species-specific genome sizes and diagnostic morphological characters for each taxon were then selected. A total of 353 specimens from 32 populations in Poland, Russia and Ukraine were studied, with nine biometric traits being examined. Although some specimens morphologically matched N. ×borealis (a hybrid between N. alba and N. candida) according to published determination keys, only one hybrid individual was revealed based on genome size data. Other specimens with intermediate morphology possessed genome size corresponding to N. alba, N. candida or N. tetragona. This indicates that natural hybridization between N. alba and N. candida is not as frequent as previously suggested. Our results also revealed a considerably higher variation in the studied morphological traits (especially the quantitative ones) in N. alba and N. candida than reported in the literature. A determination key for the investigated Nymphaea species is provided, based on taxonomically-informative morphological characters identified in our study.


Introduction
The Nymphaeaceae is a family of hydrophytes that occur in water reservoirs worldwide.Nymphaea L. (water lily) is one of the six genera belonging to the Nymphaeaceae.Approximately 50 representatives of the genus are known [1].Several varieties, chromosome races and hybrids have been reported for many of them [2].
This study focuses on four taxa of Nymphaea that belong to the Eurasian Nymphaea clade [3], namely: N. alba L., N. candida J. Presl, N. tetragona Georgi, and N. ×borealis Camus, the hybrid between N. alba and N. candida.Each of three aforementioned species has its well-defined distribution range [4].Nymphaea alba occurs almost throughout Europe with the exception of northern areas and a considerable part of the Iberian Peninsula, and also in the Caucasus, western part of Asia and North Africa [5][6][7].Nymphaea candida is a Euro-Siberian species with the center of its occurrence in northern, central and eastern parts of Europe, from where it extends up to western Siberia and central Asia.The species reaches the southern limit of its continuous range in Poland [5][6][7].Nymphaea tetragona occurs in northern and eastern parts of the Scandinavian Peninsula, northern and eastern parts of Asia, northern part of India and in North America [5,8].All three species also differ by habitat preferences.Representatives of N. alba occur in meso-and eutrophic waters, with a pH reaction ranging from weakly acidic to alkaline (pH 5.5-8.3)whereas N. candida prefers poorer, meso-and oligotrophic waters, with a pH ranging from weakly acidic to neutral (pH 5.5-7.1)[9,10].Although N. alba often forms phytocoenoses in eutrophic waters and N. candida grows better in more nutrient-poor waters, with a narrower range of requirements [11], ecological requirements of both species overlap and they can be found in both aquatic habitat types [11,12].Nymphaea tetragona is found in lakes, river backwaters and small oxbow lakes in thickets and silty bottom in the lowland and montane zones [5].
All three Nymphaea species are closely related [8].They have similar morphological traits and their distribution ranges overlap to some extent.Taxonomically informative characters include the number of stigma rays, stigma color, the shape of inner filaments, the size of the flower and pollen sculpture [9,10,[13][14][15][16][17][18][19][20].However, individuals cannot always be identified using these characters, especially when a specimen presents some transitional characters or even a combination of diagnostic characters of different taxa.It may be especially difficult to determine species in areas where distribution ranges of N. alba and N. candida overlap and the hybrid N. ×borealis can be formed [21,22].
Nymphaea ×borealis has just been noted in the XIXth century [21] and it is included in many determination keys of vascular plants [9,10,[14][15][16][17].Over the time, the hybrid has been reported from several geographic regions [12,13,23,24].On the other hand, some authors, bearing in mind a great plasticity of N. alba and N. candida, refrain from identification of N. ×borealis and report only Nymphaea sp.[25] or "intermediate forms" [26,27] to avoid incorrect determination.Based on morphological and biological investigations of N. ×borealis to date, its morphological traits are a combination of traits characteristic for both parental species [13,20,26,27].The hybrid is also characterized by decreased pollen production and lower fertility [28,29].Although the hybridization between N. alba and N. candida was convincingly confirmed by genetic studies [18,20], the majority of N. ×borealis records is solely based on morphological identification.Recently published cytogenetic studies of Nymphaea taxa conducted in the Czech Republic [20] proved that only 1.8% of investigated specimens represented N. ×borealis.This result is incongruent with previous reports on extensive hybridization between N. alba and N. candida [24].
In front of the controversial delimitation of Nymphaea taxa due to their high phenotypic plasticity and vague determination of hybrids [30], we performed a detailed study of morphological variation of N. alba, N. candida, their putative hybrid N. ×borealis, and N. tetragona, sampled over a wide area of Eastern Europe and temperate Asia.Specifically, our aims were: (i) to identify the sampled specimens based on morphological characters provided in the published determination keys (a subjective identification); (ii) to delimit the taxonomic categories based on differences in relative genome size estimated by DNA flow cytometry (an objective identification), (iii) to compare the results of both approaches and discuss potential sources of incongruence, (iv) to assess the frequency of the interspecific hybrid N. ×borealis, based on intermediate genome size values, and finally (v) to select morphological traits having the greatest discrimination power in morphometric analyses and therefore presenting the highest taxonomic value.

Morphometric analysis
In total, 353 specimens from 32 localities in Poland, Russia and Ukraine (Tab. 1, Fig. 1) were included in the study.Depending on a population size two to thirty one individuals per population were studied.Nine traits, including four quantitative and five qualitative (Tab.2), were scored or measured directly in the field on randomly chosen individuals.One randomly chosen part of plant (e.g.outer stamen, stigma, leaf etc.) per each investigated individual was measured.Measurements were conducted on fresh plants because some parameters, including length of outer petals, leaf length and maximum leaf width, can change after dehydration.In the field, analyzed individuals were tentatively ascribed to N. alba, N. candida, N. ×borealis and N. tetragona using morphological characters indicated in the published determination keys [5,13,14,17,19,20,24,26,27,31,32] (also Volkova unpublished).

Statistical analyses
Before statistical analyses each specimen was ascribed to a particular taxon based on estimated relative genome size values.It has been shown previously that all four taxa under study have distinct sizes of holoploid genome, allowing their reliable identification [8,20].
Each a priori determined specimen was treated as an operational taxonomic unit (OTU) in accordance with the methods used in numerical taxonomy [36].Prior to the statistical analyses, the distribution normality of quantitative variables was verified using the Lilliefors test.For multivariate analyses the data were standardized to avoid the effect of differences between measurement scales.
Principal component analysis (PCA) was conducted using all quantitative characters, based on the correlation matrix [37].The analysis was performed in order to get insight into the overall pattern of phenotypic variation.The analysis yielded a reduced set of variables (traits), which were most strongly correlated with the principal components.Factors were chosen according to the scree test [38].Principal coordinate analysis (PCoA) performed with Gower's similarity coefficient [39] was also used to assess morphological differentiation of the taxa.All quantitative and qualitative characters were included in PCoA.
After using Levene's test to assess the equality of variance, one-way analysis of variance (ANOVA) followed by the Tukey's HSD test for unequal sample frequencies was conducted to assess interspecific differences between means of quantitative characters.
Standard discriminant analyses were used to determine which characters best discriminate the studied species and to check the correctness of specimens' assignment into particular species defined on the basis of sizes of holoploid genome.The matrix of quantitative traits was subjected to a forward stepwise analysis.Discriminatory power was expressed by the Wilks' lambda statistic.Classificatory discriminant analysis was then applied to classify the samples [40].This procedure yields the percentage of correctly classified individuals into the a priori set taxonomic categories (i.e.Nymphaea species delimited based on genome size values).For data analyses and statistical calculations STATISTICA 9.1 and MVSP 3.2 were used [41].
Based on different multivariate statistical analyses a determination key for the investigated species was prepared.The values represent 10% and 90% percentiles, and the values in parentheses are minima and maxima.Nymphaea ×borealis was not included due to the low number of samples (only a single specimen was ascribed to this nothotaxon based on genome size data).

Results
FCM analyses revealed four non-overlapping groups of relative genome sizes, corresponding to individual Nymphaea taxa.Nymphaea alba, N. candida, N. tetragona and N. ×borealis were recorded at TaxonMORPH -tentative determination based on morphological data; TaxonFCM -determination based on relative genome size.
Tab. 1 (continued)  Four quantitative characters with the highest factor loadings on the first principal component (r > 0.60) were detected in PCA.The first two components accounted for 93.33% of total variance (Tab.3).The first axis explained 82.95% of the variation and the second axis 10.39%.The four groups representing particular taxa were not clearly separated on the scatterplot (Fig. 3) and the specimens expressed a rather continuous range of morphological variation regardless of their taxonomic identity.The first PCA axis was most highly influenced by: leaf length, maximum leaf width and length of outer petals (Fig. 4) while the number of stigma rays was most strongly correlated with the second PCA axis.
Variability ranges of quantitative characters for particular taxa are presented in Tab. 4. The results of the one-way ANOVA revealed significant (P < 0.001) differences in all quantitative characters.Values of F-statistics obtained from ANOVA are given in Tab. 4. The most important quantitative traits were: length of outer petals, number of stigma rays and maximum leaf width.
The variation of qualitative characters for each taxon is shown in Fig. 5. Linear filaments of inner stamens were most frequently recorded in N. alba (97.5% of individuals), whereas filaments were usually lanceolate (86.7%) or rarely linear (13.3%)The leaf vein leading to the lobe tip was bent along its length in the majority of N. candida specimens (77.7%) while it was straight in 44.1% of N. tetragona specimens.For N. alba this character was completely uninformative.In 40.7% of N. alba individuals the vein was bent, in 20.3% it was bent in the first third of its length and in 39.0% cases it was straight.The only N. ×borealis specimen detected in this study had respectively: lanceolate stamen filaments, yellow stigma, a short, hemispherical central stigma projection, tetrahedral cup base and the leaf vein leading to the lobe tip bent along its entire length.Discriminant analysis (DA) confirmed the diagnostic value of two quantitative characters indicated already in the PCA, namely the length of outer petals and the number of stigma rays (Tab.5).Leaf length and maximum leaf width also had some, though less, discrimination power (Tab.5).In the CDA, the chi-square test for all canonical roots for the data matrix confirmed their statistical significance.Standardized coefficients of the discriminant function for canonical variables are presented in Tab. 5.The scatterplot of the canonical variables shows three distinct, though slightly overlapping, groups formed by the taxa (Fig. 6).The first canonical discriminant function clearly separates N. tetragona from N. alba and N. candida.
A principal coordinate analysis (PCoA) scatterplot using both qualitative and quantitative characters did not clearly separate the groups of individuals representing particular taxa (Fig. 7).Individuals representing N. tetragona formed the most distinct group.In contrast, specimens of N. candida and N. alba overlapped to a great extent.Nevertheless, the recognized taxonomic groups are better separated as compared to PCA and DA scatterplots (Fig. 3, Fig. 6), which proves the usefulness of qualitative characteristics for taxonomic decision-making.Nymphaea ×borealis merges with the N. alba group.

Discussion
Many authors have reported that N. alba, N. candida, N. tetragona and N. ×borealis can be distinguished by quantitative characters, such as length of outer petals, maximum leaf width and leaf length [10,13,19,20,26].Additionally, the number of stigma rays has been recognized as one of the most important diagnostic characters in the majority of published studies [9,10,13,14,16,19,20,24,26,31].Our morphometric results confirmed the taxonomic value of the following characters: length of outer petals, maximum leaf width and number of stigma rays.All these traits, however, can vary considerably what is evidenced by classifying method of discriminant analysis (Tab.6), where partial incongruence between molecular and morphological taxa identification was revealed.Distinguishing of N. alba and N. candida is particularly challenging and hardly possible solely based on quantitative characters.
Although leaf length was previously regarded as taxonomically important, our results did not confirm its diagnostic   value in the species determination.Qualitative traits examined by us that proved to be helpful in identification are (Fig. 5): shape of inner stamen filament, shape of central stigma projection, color of stigma and shape of cup base.
The last character however is more variable that previously assumed [13,17,19,24,26,27,31].Except typical and round cup bases we recorded also the tetrahedral ones within N. alba individuals (Fig. 5).The variation of cup bases was also noticed by Kabátová et al. [20], indicating that this character should be used with caution as well.The taxonomic value of the shape of the first main leaf vein turned out to be quite low.According to Kabátová et al. [20] this trait is useful for identification of N. alba and N. candida, however it cannot distinguish N. tetragona, which has the same shape of the first main leaf vein as N. alba.
Our study revealed a higher morphological variation of individuals of N. alba, N. candida and N. tetragona than reported in the literature.The values of particular traits obtained in this study are similar to those published previously [9,10,13,14,16,19,20,24,26,31], however, their ranges are wider (see Tab. 4, Fig. 5).In all examined species, individuals with narrower and wider leaves were recorded, as well as those with shorter and longer outer petals.Considerable phenotypic variation of quantitative traits was confirmed by classificatory discriminant analysis (Tab.6).
In general, N. alba has the greatest size of both vegetative and generative parts, including the leaf length and width, length of outer petals and the number of stigma rays.In contrast, N. tetragona has the smallest size of above mentioned parts and the lowest number of stigma rays (Tab.4).The plant size, however, should be taken with caution, because growing conditions may strongly influence the morphological appearance of Nymphaea individuals [29,42,43].This was also observed in our study: individuals of Nymphaea alba from the humic lake (population No. 103) were visibly smaller and resembled N. candida.The opposite situation was also noticed: individuals of N. candida occurring in eutrophic lakes (populations No. 104, 105, 109) developed larger leaves and flowers than are usually recorded for this species.Under such conditions proper identification was only possible based on qualitative morphological traits.
All qualitative and quantitative characters of hybridogenous N. candida [8] are intermediate between its parental species N. alba and N. tetragona (Tab.4, Fig. 5).Qualitative traits shared by N. candida and N. tetragona include: red-or orange-colored stigma and a long conically-shaped stigma projection.The shape of the inner stamen filament and the shape of cup base are the only investigated qualitative traits that differentiate N. candida from N. tetragona.Oval-shaped inner stamens and tetrahedrally-shaped cup base with a rib are diagnostic for N. tetragona (Fig. 5).
The sole hybrid individual of N. ×borealis detected during this study combined morphological traits of both parental species, what is congruent with previous observations [13,20,26,27].The shape of inner stamen filaments and the cup base shape were typical for N. candida, while the shape of central stigma projection resembled N. alba.Quantitative traits were intermediate between the values typical for the parental species except for the maximum leaf width, which exceeded even that of N. alba (353 mm).
Phenotypic similarities between N. alba, N. candida and N. tetragona [8] seems to favor hybridization.Ejankowski and Małysz [24] state that due to the absence of barriers preventing crossing in Nymphaea, hybrids may in time become more widespread than parental species.This hypothesis, however, was not confirmed by Kabátová et al. [20], who, based on genome size investigations, found that N. ×borealis occurred definitely much less frequently than previously assumed and only 11 of 625 specimens from the Czech Republic examined by these authors were of hybrid origin.Our results also show that the hybrid is not as frequent as previously thought.Although based on morphological data, hybrid origin was suspected in 32.3% of investigated specimens (see Tab. 1), all of them turned out to be N. alba or N. candida based on the genome size values.Only one cytometrically-proven N. ×borealis individual was revealed among 353 specimens investigated.Therefore, it can be concluded that morphologically intermediate forms are largely a result of phenotypic plasticity rather than products of interspecific hybridization.

Fig. 1
Fig. 1 Map with the localities of investigated Nymphaea populations.Numbers correspond with populations described in Tab. 1.

Fig. 3 Fig. 4 4
Fig. 3 Scatterplot presenting the result of principal component analysis (PCA) of samples representing N. candida, N. alba, N. tetragona and N. ×borealis delimited based on distinct genome sizes.

2 Fig. 6
Fig.6 Scatterplot presenting the results of canonical discriminant analysis of specimens representing the three investigated Nymphaea species delimited based on distinct genome sizes.
14, 19, 12 and one locality, respectively.Mean relative fluorescence intensities ±SD and ranges of variation (setting fluorescence intensity of in N. candida.Nymphaea tetragona is characterized by a lanceolate (48.5%) or round (51.5%)List of quantitative and qualitative characters measured and scored in morphometric analysis.Description, corresponding abbreviation and a measure unit or scale are given for each character.
Fig. 2 Scatterplot of relative genome sizes of 353 individuals of water lilies as estimated by DAPI flow cytometry.Four nonoverlapping groups were detected, corresponding to N. alba (118 specimens), N. candida (166 specimens), N. tetragona (68 specimens) and N. ×borealis (one specimen).Internal reference standard (Lycopersicon esculentum cv.Stupické polní rané, 2C = 1.96 pg) was given as unit of fluorescence.The highest (>0.60) values of factor loadings are given in bold.See Tab. 2 for character abbreviations.stamenfilaments.The stigma was most commonly yellow in N. alba and N. candida (85.6% and 50.6%, respectively).In those individuals of N. alba which had not yellow stigma it was either yellow-orange (8.5%), orange (5.1%) or red (0.8%).In the case of N. candida the remaining individuals had stigmas red (29.5%), orange (12.7%) or yellow-orange (7.2%).Red stigma was observed in nearly all specimens of N. tetragona (95.6%), while remaining 4.4% of specimens had yellow stigma.A short, hemispherical central stigma projection was noted in 75.4% of individuals representing N. alba whereas it was mostly long and conical in the vast majority of N. candida (95.2%) and N. tetragona (100%) samples.A tetrahedrally-shaped cup base was observed in all specimens of N. candida.Similarly, all N. tetragona individuals had a cup base which was tetrahedrally-shaped with a prominent rib.It was round-shaped in only 33.1% plants of N. alba.
Values of Wilks' lambda, partial Wilks' lambda and P values for the four quantitative characters provided by discriminant analysis of individual plants.