Variation in genome size, cell and nucleus volume, chromosome number and rDNA loci among duckweeds

Duckweeds are small, free-floating, largely asexual and highly neotenous organisms. They display the most rapid growth among flowering plants and are of growing interest in aquaculture and genome biology. Genomic and chromosomal data are still rare. Applying flow-cytometric genome size measurement, microscopic determination of frond, cell and nucleus morphology, as well as fluorescence in situ hybridization (FISH) for localization of ribosomal DNA (rDNA), we compared eleven species, representative for the five duckweed genera to search for potential correlations between genome size, cell and nuclei volume, simplified body architecture (neoteny), chromosome numbers and rDNA loci. We found a ~14-fold genome size variation (from 160 to 2203 Mbp), considerable differences in frond size and shape, highly variable guard cell and nucleus size, chromosome number (from 2n = 36 to 82) and number of 5S and 45S rDNA loci. In general, genome size is positively correlated with guard cell and nucleus volume (p < 0.001) and with the neoteny level and inversely with the frond size. In individual cases these correlations could be blurred for instance by particular body and cell structures which seem to be linked to specific floating styles. Chromosome number and rDNA loci variation between the tested species was independent of the genome size. We could not confirm previously reported intraspecific variation of chromosome numbers between individual clones of the genera Spirodela and Landoltia.

A correlation between genome size evolution, frond size and neoteny level was observed by Wang et al. 13 when investigating 115 clones of 23 out of 37 duckweeds species. For some individual species Wang et al. 13 and Bog et al. 12 reported different genome sizes. These differences might be due to the use of different internal reference standards, to true differences between clones, or simply to random variation between measurements.
Interestingly, duckweed frond sizes vary from 1.5 cm to less than 1 mm in diameter accompanied by a nearly 12-fold genome size variation (from 160 Mbp to 1881 Mbp according to Wang et al. 13 ) and a successive reduction of morphological structures from Spirodela towards Wolffia species 9,12,13 . This potential correlation of genome size with morphological reduction and frond size evolution makes duckweeds an interesting subject for genome and karyotype evolution studies. A positive correlation between nuclear DNA content and nuclear and cell volume was recorded for some angiosperms 14 and for endosperm cells of Sorghum bicolor 15 . To elucidate whether also for duckweeds a correlation between genome size, cell and nuclear volume is valid, accessions of eleven representative species of the five duckweed genera were investigated. Additionally, we studied the chromosome number and genomic distribution of 5S and 45S rDNA loci of these species.

Results
Differences in morphology between duckweed genera. The phylogenetic position of the eleven studied duckweed species according to Les et al. 16 , the frond morphology and the corresponding genome size is shown in Fig. 1. Both Spirodela species have the lowest genome size and the largest fronds, while the genera Landoltia, Lemna, Wolffiella and Wolffia have larger genomes (and genome size variation) and progressively smaller fronds. As mentioned by Landolt 9 , duckweed stomata usually stay open and display a slightly higher osmotic value than normal epidermis cells. The open Spirodela stomata can close when treated with 3-(4-chlorophenyl)-1 1-dimethylurea, Carbonyl cyanide-4-(trifluoromethoxy) phenylhydrazone, valinomycin or nigericin, while these substances had no effect on Lemna stomata 9 . Stomata are largely absent in some fully submerged species. Our observation confirmed that not only the frond morphology differs between duckweed genera as described 9 , but also the shape of guard and epidermal pavement cells. Guard cells form spherical stomata in Spirodela and Lemna species, or elliptic ones as in Landoltia, Wolffiella and Wolffia species (Fig. 1C). Species of the latter two genera show additionally flattened tips of guard cells, compared to the more round ones in Landoltia. In all investigated duckweed species displaying stomata, these were usually open. Epidermis cell walls are rather straight in Wolffiella and Wolffia species, but look bent in Spirodela and undulated in Landoltia and in Lemna species ( Fig. 2A). Only very few stomata could be found in Wa. lingulata and Wo. columbiana, two largely submerged species (Fig. 2B). To avoid the confusing between Landoltia and Lemna as well as Wolffiella and Wolffia genera, we use a two-letter code to abbreviate the names for these genera.
Genome size variation. The obtained genome sizes varied from 160 Mbp in S. polyrhiza to 2203 Mbp in Wo. arrhiza resulting in a ~14-fold difference between duckweed species (Fig. 1). The largest variation in genome size (from 432 to 2203 Mbp) occurred within the genus Wolffia. Except for the two Spirodela species, our genome size measurements yielded up to 26% larger values than measured for the same clones by Wang et al. 13 (Fig. 2C). In detail, the S. polyrhiza genome revealed no difference, while a 9% higher value was observed for La. punctata (7260), 8% for Le. minor (8623), 17% for Wo. arrhiza (8872), and 26% for Wa. hyalina (8640). The differences might be due to different internal reference standards, an unusually low assumption for the genome size of A. thaliana by Wang et al. 13 (147 Mbp instead of 157 Mbp as measured by Bennett et al. 17 ) and the use of different flow cytometry equipment.

Correlation between genome size, nuclear and cell volume within and between duckweed genera.
Instead of pavement cells used by Jovtchev et al. 14 , we selected guard cells for measurements to investigate a potential correlation between cell parameters of duckweed species with different morphology and genome size. The reason behind is on the one hand the highly variable size and irregular shape of pavement cells ( Fig. 2A), that www.nature.com/scientificreports www.nature.com/scientificreports/ is a challenge for measuring of cell dimensions and for calculating and comparing cell volumes in duckweeds. On the other hand, the permanently open status of stomata in floating aquatic plants 10,18 yields a rather homogenous guard cell shape, more suitable for precise volume measurement 19 .
Our results show a moderate but, because of the large number of samples (252) highly significant positive correlation between genome size and cell and nuclear volume in duckweeds. In general, the higher the nuclear DNA content, the bigger are cells and nuclei ( Fig. 3 and Table 1). In detail, average cell volume and nuclear volume are 541.7 µm 3 and 17.1 µm 3 for S. polyrhiza (160 Mbp) and increase to 649.6 µm 3 and 50.3 µm 3 in Le. disperma (651 Mbp), and to 1826.8 µm 3 and 112 µm 3 in Wo. arrhiza (2203 Mbp) (Fig. 1B-D). Scatterplots (Fig. 3B) representing all measured data (n = 252) revealed: (i) cell volume and nuclear volume increase with increasing genome size  Table 1).
Additionally, we found unexpected features in some duckweed species: (i) Le. aequinoctialis (2018) revealed a considerable variation in guard cell size and shape (Fig. S1A). In the younger part of frond, guard cells form spherical stomata while in the older part they are elongated and larger. Besides that, cell and nuclear volume are larger than that of Le. disperma possessing a larger genome. Therefore, we investigated another Le. aequinoctialis clone (6746) to see whether the variable guard cell volume is specific for this species. Interestingly, this clone showed variation in guard cell size and a nearly doubled genome size (900 Mbp) and correspondingly larger cell and nuclear volumes (1313 µm 3 and 238 µm 3 , respectively). Thus, the two tested Le. aequinoctialis clones showed variation not only in guard cell shape, cell volume and nucleus volume, but surprisingly also regarding the genome size (Table S3) . Therefore, we wanted to test other Wolffiella species to see whether very large cell volume is specific for this genus. Interestingly, only one or two stomata per frond were present in the Wa. lingulata clone 7725. The same was true for Wo. columbiana clone 9356. Differences in floating style of Wo. columbiana with spherical fronds, having most of the surface submerged, and Wa. lingulata also with a frond shape which keeps most of the frond below the water surface 9 (Fig. 2B) could be the reason for the almost complete absence of stomata in these species. Thus, so far it remains unclear whether or not a large guard cell size is a typical feature of the genus Wolffiella. (iii) Wa. hyalina and Wo. australiana displayed an unusual distribution of nuclei between sister guard cells. We found in 26% of Wa. hyalina and in 8% of Wo. australiana guard cells two nuclei located in one sister cell and none in the other (Fig. 4B, C, E). In some cases (6.8% of Wo. australiana guard cells) it was even possible to find transient stages, suggesting that nuclei may post-mitotically migrate into the sister cell (Fig. 4F). This observation resembles cytomixis, a so far unexplained phenomenon which occurs during microsporogenesis in several higher plants (for review see 20 ). These findings, in particular the large variation of guard cell and genome size in Le. aequinoctialis, and the abnormal nuclei distribution between the sister guard cells are biological features of some duckweeds that deserve further studies.
Chromosome numbers. Chromosome numbers of duckweed species have been studied by several researchers since 1933 (for references see Tables 2 and S2). However, different chromosome numbers were reported for the same species and it remained unclear whether the discrepancies are due to variation of chromosome number between largely asexual clones within a species. www.nature.com/scientificreports www.nature.com/scientificreports/ Among 34 S. polyrhiza clones mentioned by Wang et al. 13 , the chromosome number of nine clones was not determined, for three clones (7652, 7657 and 7364) 2n = 30, and for the other clones 2n = 40 was reported (Table 2).
Our chromosome counting results are mainly similar to that of Geber 23 (Table 2  www.nature.com/scientificreports www.nature.com/scientificreports/

Intraspecific variation of genome size, chromosome number and guard cell parameters.
Different chromosome numbers were found in different clones of Le. aequinoctialis 23 , 42 chromosomes were counted for clones 7382, 7321, 7300 and 7737, while 84 chromosomes were counted for clones 6746 and 7384. Meanwhile, these clones (except 6746) were lost from international duckweed collections. We chose the Le. aequinoctialis clone 2018 instead for ploidy testing within this species. As described above, genome size varies from 452 Mbp (clone 2018) to 900 Mbp (clone 6746). These data suggest that clone 6746 is tetraploid. We investigated the correlation between genome size, cell and nuclear volume and counted chromosome number of the two Le. aequinoctialis clones (6746 and 2018). In parallel, two clones of La. punctata: clone 7260 (diploid) and clone 5562_A4 (a true artificial tetraploid) were included.
Both genome size measurement and chromosome counting suggest that Le. aequinoctialis clone 6746 is tetraploid with larger cell and nuclear volume, and clone 2018 is diploid with smaller cell and nuclear volume  www.nature.com/scientificreports www.nature.com/scientificreports/ (Table S3). Fig. S1B represents all measured data (n = 40, p < 0.001) and revealed a positive correlation between cell and nuclear volume (r = 0.593).
A similar result was obtained for the two clones of La. punctata clones 7260 and 5562_A4 (Table S3 and Fig. S1C). In addition, the tetraploid La. punctata clone 5562_A4 frequently showed elongated instead of round nuclei (Fig. 4G-I). Cell and nucleus volumes are significantly different (at least at p = 0.01 level) for diploid and tetraploid clones of both species. Therefore, the 95% confidence intervals do not overlap (Fig. S1B,C).

Location of 5S and 45S rDNA loci on duckweed chromosomes.
A remarkably low copy number of 45S rDNA (18S and 26S rDNA) but also of 5S rDNA was reported for S. polyrhiza 25 . A significant decrease in copy number of 45S rDNA has apparently occurred in S. polyrhiza (81 copies) compared to the 13-times smaller genome of Saccharomyces cerevisiae (~12.2 Mbp/1 C) with 150 copies 26 , or the similar-sized genome of Arabidopsis thaliana with 570 copies 27 . The locus of 45S rDNA is located on chromosome ChrS 01 and two loci of 5S rDNA on ChrS 13 and ChrS 06 with 60 and 12 copies, respectively 25,28 .
The number of 45S and 5S rDNA loci of the eleven studied duckweed species was determined by FISH (Table 1, Fig. 6). In detail, one locus of 45S and 5S rDNA each was detected in Le. minor, Le. disperma, Le. aequinoctialis, Wo. microscopica, while S. polyrhiza, S. intermedia, La. punctata, Wa. hyalina and Wo. australiana www.nature.com/scientificreports www.nature.com/scientificreports/ displayed one locus of 45S rDNA and two loci of 5S rDNA. In Wo. arrhiza, two loci of 45S rDNA and three loci of 5S rDNA were detected.
In Wa. rotunda (clone 9072), three loci of 5S rDNA were detected and two chromosome pairs displayed 45S rDNA loci. One pair of NORs was more extended and showed a distal satellite (Figs. 6 and S3B). Without rDNA FISH signals, the satellite distal to the NOR could erroneously be counted as a small pair of chromosomes. The www.nature.com/scientificreports www.nature.com/scientificreports/ strength of FISH signals reflected differences in copy number of 5S rDNA. For instance, the 5S rDNA probe often yielded in Wo. arrhiza (clone 8872) two strong, two medium and two weak FISH signals. Noticeably, a very low copy number of 5S rDNA could apparently prevent the detection by FISH, e.g. the 5S rDNA locus with only 12 copies on ChrS 06 of S. polyrhiza 25 . Weak signals of 5S rDNA loci (in S. polyrhiza, S. intermedia, La. punctata and Wo. arrhiza) could only be detected in a few metaphases (Fig. 6), and thus are at risk to be overlooked. Therefore, the number of 5S rDNA loci which were detected by FISH in other duckweed species than S. polyrhiza might underestimate the true number of loci as long as their genomes are not completely assembled.

Discussion
Our measurements of genome size in relation to frond and cell shapes, guard cell volume, nuclear volume, chromosome number and number of rDNA loci for eleven species, representative for the five duckweed genera, led to several conclusions or speculations, or pointed to further open questions: (i) Some duckweed species seem to have specific frond and cell structures which are suitable for different floating-styles (totally, largely or not submerged) and are not strongly affected by genome size. (ii) Genome size is known to correlate with a number of traits in angiosperms. DNA content and nuclear volume as well as nuclear and cell volume showed positive correlation at different endopolyploidy levels in epidermis cells of A. thaliana (from 2C to 32C), Barbarea stricta (from 2C to 16C) as well as between species that differ in genome size up to ~500 fold (from 0.32 pg in A. thaliana to 154.99 pg in Fritillaria uva-vulpis) 14 or between 14 herbaceous angiosperm species 29 . A correlation of cell parameters (DNA content, cell volume, nuclear volume, cell surface, nucleus surface) was also reported for Sorghum bicolor endosperm cells from 3C to 96C 15 . In this study, cell and nuclear volumes from guard cells of the eleven duckweeds species provided in total a significant positive correlation between genome size, nuclear and cell volume. However, this correlation is not as strong as for cells of different endopolyploidy levels within one species 14,15 . The weaker correlation is likely caused by the fact that individual duckweed species may have an own specific body and cell structure and size, and a range of intraspecific variation of these features which might blur the influence of genome size on nuclear and cell volume. (iii) Genome size differences between duckweed species rise the question to what degree frond size and neoteny level are correlated with the genome size, which was previously shown not to be correlated with an organisms' complexity 30,31 . In general, genome size (and genome size variation) increases with the reduced morphological differentiation in duckweeds. However, there are some exceptions: In spite of similar genome sizes of about 400 Mbp, frond size and neoteny level differ between La. punctata, Le. minor and Wo. australiana, while species, with similar neoteny level, may own different genome size, e.g. Le. minor (409 Mbp), Le. disperma (651 Mbp). The genome size variation between Le. aequinoctialis clones 2018 and 6746 (452 and 900 Mbp) might be due to WDG, because also the chromosome number is doubled in clone 6746, and is accompanied by larger nuclear and cell volumes (Fig. S1B). Whether the large genome size differences between duckweed genera as well as between species within the genera Lemna, Wolffiella and especially Wolffia are based on WGD or on a retroelement burst remains to be solved. It might also be that DNA double-strand break repair biased towards deletions or duplications 32,33 plays a role in genome size variation, e.g. between Wolffia species. It also remains unclear why at all genome size increases with decreasing organismic complexity and decreasing frond size of duckweeds and whether or not this correlation results in a lower (and possibly constant) cell number. (iv) Mitotic chromosome spreads of all tested species (Fig. 5) revealed that, as expected, genome size is not correlated with chromosome number. That means, genome size and chromosome number vary independently from each other. (v) No chromosome number variation was detected between the tested clones of Spirodela and Landoltia species. The reported high variation of chromosome number in the phylogenetically younger genera Lemna, Wolffia and Wolffiella (as summarized in Fig. 1, Tables 2 and S2) needs further investigation to be confirmed or disproved. In case of confirmation it will be of interest to elucidate the mechanisms behind. (vi) Ribosomal genes (rDNA) are characterized by conserved sequences and organized as tandem repeat units in eukaryotic genomes. Variations regarding number and chromosomal distributions of 5S and 45S rDNA loci are informative markers for discriminating karyotypes of species, and in specific cases, for elucidating karyotype evolution, for instance in Brassicaceae 34,35 and in Anthemideae 36 . In the eleven tested duckweed species, the observed number of 5S and 45S rDNA loci revealed no correlation with chromosome number and/or genome size. Whether the extremely low copy number of rDNA sequences, as observed for S. polyrhiza, is typical for duckweeds has to be checked when complete sequences of further duckweed genomes will be available. Completely sequenced genomes will also reveal whether FISH experiments detected all 5S rDNA loci so far, or whether additional minor loci escaped from detection as was the case for the locus on chromosome 6 of S. polyrhiza with only 12 copies 25 .

Materials and Methods
Plant material and mitotic chromosome preparation. S These eleven species have been chosen because they cover the ranges of genome size variability between and within genera, are of different geographic origin and were available in the collections. La. punctata 5562 and its www.nature.com/scientificreports www.nature.com/scientificreports/ colchicine-induced tetraploid mutant 5562_A4 were obtained from M. Edelman, Rehovot, Israel. The fronds were grown in liquid nutrient medium 37 under 16 h white light of 100 µmol m −2 s −1 at 24 °C.
Spreading of mitotic chromosomes was carried out according to Cao et al. 38 with some modifications. In brief, healthy fronds were incubated in 2 mM 8-hydroxyquinoline at 37 °C and then fixed in fresh 3:1 absolute ethanol: acetic acid for at least 24 h. The samples were washed twice in 10 mM Na-citrate buffer, pH 4.6, for 10 min each before and after softening in 2 ml pectinase/cellulase enzyme mixture, prior to maceration and squashing in 60% acetic acid. After freezing on dry ice or in liquid nitrogen, the slides were treated with pepsin, post-fixed in 4% formaldehyde in 2xSSC (300 mM Na-citrate, 30 mM NaCl, pH 7.0) for 10 min, rinsed twice in 2xSSC, 5 min each, dehydrated in an ethanol series (70, 90 and 96%, 2 min each) and air-dried (Table S1).
Genome size measurement. Genome size measurements were performed according to Dolezel 39 . In total, for each species at least six independent measurements on two different days were performed.

Epidermis preparation, microscopic cell and nuclear volume measurements, and statistics.
Due to the small frond size, a single epidermis layer is difficult to obtain especially for species of the genus Wolffia (frond diameter ~1 mm). Therefore, we modified the epidermis preparation methods described [40][41][42] , by using domestic adhesive tape. Because stomata are located on the upper surface in floating plants 9,18 , duckweed fronds were placed with their upper side on the adhesive tape. Other parts of the fronds were carefully removed with a razor blade until only the transparent layer of epidermis stuck on the tape. Ten µl of DAPI (2 µg/ml) in Vectashield were dropped on slides before the adhesive tape with the epidermis layer was placed on slides and covered by a coverslip. Freshly prepared slides were used immediately to avoid the disintegration of the nuclei before imaging. Differential interference contrast (DIC) and fluorescence (excitation of DAPI with a 405 nm laser) image stacks were acquired using a Super-resolution Fluorescence Microscope Elyra PS.1 and the software ZEN (Carl Zeiss GmbH). The DIC image stacks were used to measure the x-y area A, and the z dimension of the guard cells via the ZEN software. Accordingly, the image stacks were used to measure the nuclei dimensions (Fig. 3A). These dimensions were applied to calculate the guard cell and nuclear volumes by the following formulae: Cell Volume A z cell = * = * * Nuclear volume 2/3 A z nucleus It means, the guard cells are considered as stacks with the base area A and the height z, while the nuclei are considered as ellipsoids.
The correlation values (Pearson product moment correlation coefficient) and the corresponding p values were calculated with the program SigmaPlot 12 (Systat Software, Inc.). The same program was used for the plot of the regression diagrams. At least 20 sister guard cells (ten stomata) with the corresponding nuclei were chosen for measurements per species. Telomere-specific probes were generated by PCR using tetramers of the Arabidopsis-type telomere repeats without template DNA according to Ijdo et al. 46 . PCR products were used as templates for PCR-labeling (5S rDNA) or nick-translation (18S, 26S rDNA and telomere sequences) to generate the corresponding FISH probes. The probes were labeled with Cy3-dUTP (GE Healthcare Life Science), Alexa Fluor 488-5-dUTP, Texas Red-12-dUTP, biotin-dUTP or digoxigenin-dUTP (Life Technologies) and precipitated as described in Hoang and Schubert 47 .
Probes were denatured at 95 °C for 5 min and chilled on ice for 10 min before adding 10 µl probe per slide (up to three different labeled probes simultaneously). Then, the mitotic chromosome preparations were denatured together with the probes on a heating plate at 80 °C for 3 min, followed by incubation in a moist chamber at 37 °C for at least 16 h. Post-hybridization washing and signal detection were carried out according to Lysak et al. 48 .
To analyze the ultrastructure and spatial arrangement of signals and chromatin at a lateral resolution of ~120 nm (super-resolution, achieved with a 488 nm laser), 3D structured illumination microscopy (3D-SIM) was applied using a Plan-Apochromat 63x/1.4 oil objective of an Elyra PS.1 microscope system and the software ZENblack (Carl Zeiss GmbH). Image stacks were captured separately for each fluorochrome using the 561, 488, and 405 nm laser lines for excitation and appropriate emission filters 49 . Maximum intensity projections of whole cells were calculated via the ZEN software. Zoom in sections were presented as single slices to indicate the subnuclear chromatin structures at the super-resolution level.