Isolation and staining of nuclei and analysis of nuclear DNA content

Some of the species tested in this study were analysed using different, established methods at the Rutgers and Vienna laboratories, representing identical plant material. This seems reasonable because of the possibility, when measuring ultrasmall genome sizes, of laboratory-specific errors that may be associated with the equipment used, the stage of the tissue or secondary metabolites originating from the size standard plant material that interact with DNA or interfere with DNA staining. Additionally, different internal size standards were applied for the flow cytometric measurements in the two laboratories, to minimize the negative contribution of cytosolic compounds from the standard plant tissue to the variation in gene size measurements. At Rutgers, nuclei were isolated then mixed to reduce the effect of polysaccharides and polyphenols. Due to the abundance of polysaccharides and polyphenols in G. aurea tissue (Płachno et al., 2007; Fleischmann and Heubl, 2009), the protocol for measurement of nuclear DNA content by flow cytometry was modified (Peterson et al., 2000; Doležel et al., 2007). Briefly, 10 mg of fresh Genlisea plant tissue was placed in 1 ml of ice-cold 2-methyl-2,4-pentanediol (MPD)-based extraction buffer (Peterson et al., 2000) and immediately chopped into very fine slices, and the homogenate was filtered through a 30-μm nylon mesh. The nuclei were pelleted at 9000 rpm for 2 min, the supernatant was carefully removed and the nuclei were resuspended in 500 μl of Galbraith's buffer (Galbraith et al., 1983). The same steps were conducted for the internal controls Spirodela polyrhiza ‘7498’ (1C = 0·16 pg or 158 Mbp; Wang et al., 2011), Brachypodium distachyon ‘Bd21’ (1C = 0·31 pg or 300 Mbp; Bennett and Leitch, 2005), Selaginella apoda (1C = 0·10 pg or 88 Mbp; Little et al., 2007) and Arabidopsis thaliana ‘Columbia’ (1C = 0·16 pg or 157 Mbp; Bennett et al., 2003). The Genlisea sample and corresponding internal standard were combined (the ratio was determined empirically with respect to the concentration of nuclei in each plant, so that their G1 peaks were of similar height on the histograms of DNA content), 50 μl of stock solution (1 mg/ml) of propidium iodide (Sigma) and 5 μl of stock solution (5 mg/ml) of RNase A (Sigma) were added and the sample was incubated on ice before flow cytometry. Propidium iodide-stained nuclei were analysed for DNA content with a Coulter Cytomics FC500 Flow Cytometer (Beckman Coulter). In all experiments, the fluorescence of at least 3000 G1-phase nuclei was measured. DNA content of each target sample was calculated by comparing its mean nuclear fluorescence with an internal standard. At least two independent replicates for each sample were analysed on different days to obtain the mean DNA content, and the exact number of replicates is specified in Table 1.

The absolute DNA content of a sample was calculated on the basis of the G1 peak means: 1C nuclear DNA content of the sample = (sample G1 peak mean)/(standard G1 peak mean) × 1C DNA content of standard (Mbp).

At Vienna, ~25 mg of fresh leaves was co-chopped as described by Galbraith et al. (1983) with the internal standard material A. thaliana (1C = 0·16 pg or 157 Mbp; Bennett et al., 2003), Solanum pseudocapsicum (1C = 1·29 pg or 1266 Mbp; Temsch et al., 2010), Pisum sativum ‘Kleine Rheinländerin’ (1C = 4·42 pg or 4322·8 Mbp; Greilhuber and Ebert, 1994), Raphanus sativus ‘Saxa’ (1C = 0·61 pg or 596·6 Mbp; Doležel et al., 1998) and Zea mays ‘CE-777’ (1C = 2·73 pg or 2670 Mbp; Doležel et al., 1998) in Otto's buffer I (Otto et al., 1981), and the preparation was filtered and incubated at 37 °C for 30 min with RNase (RNase A; Sigma). Subsequently, the isolates were stained in a propidium iodide (50 mg/L) solution containing Otto's buffer II (Otto et al., 1981). The measurement was performed with a Cyflow ML flow cytometer (Partec, Münster, Germany) equipped with a green laser (100 mW, 532 nm; Cobolt Samba, Cobolt, Stockholm, Sweden). For each preparation, three runs were done with a total of 15 000 particles, recording the fluorescence intensity and side scatter for each particle. The coefficients of variation of the G1 nuclei peaks were usually <3 %; if higher, up to five runs were measured. The 1C values of the Genlisea taxa were calculated for each run according to the formula described above. The run results were averaged to obtain the preparation's C value.

A few taxa (G. tuberosa, G. aurea var. minor from Itacambira; Table 1) that yielded interesting preliminary results with flow cytometry were included for ‘size estimations, based on one or two runs.

Chromosome preparation from rhizophylls

The apical tips of developing rhizophylls [before the apical end forked (Fig. 1C) or at a stage just after the two arms branched dichotomously, but before they began twisting helically (Fig. 1D)] were taken from cultivated plants grown under artificial light, in the late afternoon, and pre-treated with 0·002 m 8-hydroxyquinoline for 15 h at 4 °C to achieve mitotic arrest. The tips were then fixed in ethanol:glacial acetic acid (3:1 v/v) and stored at 4 °C. Fixed rhizophyll tips were hydrolysed in 2 N hydrochloric acid at 60 °C for 30 min, then rinsed with distilled water and squashed in 45 % acetic acid on glass slides. The prepared rhizophyll tip meristems were air-dried after removing the cover slip by the dry ice technique, and stained with aceto-orcein (Orcein; Roth) or DAPI (4′-6-diamidino-2-phenylindole; Merck, 2 μg/ml in McIlvaine's buffer, pH 7·0). Slides for DAPI staining were pre-incubated in McIlvaine's buffer for 30 min at room temperature before staining. DAPI-stained slides were incubated in the dark at room temperature for at least 15 min before use.

Chromosome preparation from Genlisea flower buds for species with very small chromosomes

Floral buds at an early stage of development (at a size when the sepal tips are still apically touching; Fig. 1B) were taken from young inflorescences of cultivated plants and fixed in freshly prepared ethanol:glacial acetic acid (3:1 v/v) at room temperature overnight and kept at –20 °C. Fixed anthers were prepared from the dissected flower buds and washed in distilled water, and the thecae were dissected in a drop of 45 % acetic acid and squashed. Coverslips were removed after freezing and preparations were air-dried at room temperature.

To access the DNA for staining, we used the first steps of the fluorescence in situ hybridization protocol modified by Sousa et al. (2013) before staining slides with DAPI. Firstly, slides were pre-fixed in a solution of 3:1 (v/v) ethanol:glacial acetic acid for 15 min, dehydrated for 5 min in a 70 % and 100 % ethanol series at room temperature, and incubated for 30 min at 60 °C. After cooling the slides at room temperature for ~10 min, they were pre-treated with 100 μg/ml RNase A (Sigma) in 2× saline–sodium citrate (SSC) buffer for 1 h at 37 °C in a wet chamber and washed three times for 5 min in 2× SSC. Then, they were treated with 10 μg/ml pepsin (Sigma) in 0·01 N HCl for 20 min at 37 °C in a wet chamber, washed twice for 5 min in 2× SSC, post-fixed in 4 % formaldehyde solution (Roth) for 5 min at room temperature, washed again three times for 5 min in 2× SSC, dehydrated for 5 min in a 70 % and 100 % ethanol series and air-dried for at least 1 h at room temperature before being stained. The chromosomes were stained with DAPI (2 μg/ml) and mounted in Vectashield (Vector).

Chromosome counting

Chromosome counts were made using a fluorescence light microscope (DMR RXE; Leica) under 1000× magnification, and slides were documented photographically using a CCD camera (Kappa) and by camera lucida drawings, which were drawn combining different focus layers. The digital zoom of the Kappa camera was used for the very small chromosomes of most species, and digital images were optimized by manually improving colour contrast, brightness and noise rendering using Photoshop CS5 version 10·0 (Adobe).

Chromosome staining

Aceto-orcein stained chromosomes of Genlisea very well (Fig. 2), equivalent to Feulgen staining, which was used by Greilhuber et al. (2006). However the chromosomes of most species (all except members of subgenus Tayloria) were too small for accurate counting under a light microscope at 1000× magnification, as their size range of 0·2–0·8 μm was close to the maximum size resolution of light microscopes (the Abbé limit). Furthermore, the metaphase chromosomes were usually found clustered in a bundle in the nucleus, so that adjacent chromosomes could not be reliably differentiated. When chromosome preparations of species possessing very small chromosomes were stained with classical dyes (e.g. Giemsa, Feulgen and orcein), the cytoplasm was also stained by these dyes and a complete rounded cell was seen, and differentiation of the very small chromosomes was not possible.

Open in a separate window

Fig. 2.

Light microscope photographs and camera lucida drawings of orcein-stained mitotic prophase plates from rhizophyll tip meristems of Genlisea spp. (A, B) G. uncinata, 2n = 16. (C, D) G. metallica, 2n = 16. (E, F) G. violacea from Couto de Magalhães, 2n = 16. (G, H) G. flexuosa, 2n = 16. (J, K) G. hispidula, 2n = 32. (L, M) G. subglabra, 2n = 32. (N, O) G. guianensis, 2n ~40. Scale bars = 3 μm.

The fluorochrome DAPI, which preferentially stains DNA regions rich in AT bases, did not stain the chromosomes of Genlisea species well without additional pre-treatment to remove DNA background noise as described above; we believe that Genlisea species have a high density of cytoplasmic components that are stained by these dyes and thus are only accessible after treatment with RNase A (which degrades RNA into smaller components) and pepsin. The GC content in Genlisea is not exceptionally high, but differs strongly between species (Veleba et al., 2014). The species with ultrasmall genomes, as studied by Greilhuber et al. (2006) and Ibbarra-Laclette et al. (2013), had DNA of a relatively high GC content, which could make them generally less receptive to staining. Nucleoli of G. aurea could be easily identified as DAPI-negative regions surrounded by more DAPI-dense DNA (Fig. 3).

Open in a separate window

Fig. 3.

Metaphase chromosomes from pollen mother cells of Genlisea spp., stained with DAPI. (A–C) G. margaretae from Zambia, n = 18, 19. (C) Two meiotic metaphase cells. (D–F) G. aurea var. minor from Itacambira, pre-meiotic mitosis cells, 2n = 46. Note the smaller chromosome size compared with G. margaretae. (F) DAPI-stained interphase nuclei. Nucleoli (*) can be identified as DAPI-negative regions surrounded by more DAPI-positive DNA staining. Scale bars = 10 μm.

Genome sizes and karyotypes

The known Genlisea genome sizes, chromosome numbers and chromosome size ranges (previously published or newly presented here) are summarized in Table 1.

Chromosome counts made for selected species of Genlisea, representing different phylogenetic clades, displayed variation between 2n = 16 and 2n~46. Species with 2n = 16 were restricted to Genlisea subgenus Tayloria (Figs 2 and ​and4)4) and displayed larger chromosomes (1–4 μm) than species with higher chromosome numbers (all <1 μm) (Table 1). In Genlisea section Africanae, the two studied species showed 2n = 32, likely resulting from polyploidization, while the consecutive sister clade, section Recurvatae, exhibited dysploid series of 2n = 36 and 38 (Figs 2–4). The single member of section Genlisea with ultrasmall chromosomes for which chromosome counts could be performed was G. aurea var. minor with 2n~46 (Table 1, Fig. 3).

Karyotype evolution

All members of Genlisea subgenus Tayloria that were analysed were diploids, with 2n = 16, and possessed comparatively large chromosomes 1–4 μm in length. In contrast, the sister group Genlisea subgenus Genlisea seemed to be derived from a tetraploid lineage with 32 small chromosomes. We suggest that the haploid number n = 8 represents the ancestral chromosome number for the genus Genlisea. Due to the lack of cytological data for the sister genus Utricularia, however, inference of the ancestral chromosome number of Genlisea is difficult.

Reported base numbers for Utricularia are x = 6, 7, 9, 10 and 11 (Subramanyam and Kamble, 1968; Casper and Manitz, 1975); however, only ~10 % of the species of the large genus Utricularia have been studied cytologically thus far, while those in the common sister genus Pinguicula have x = 6, 8, 9, 11 and 14 (Casper and Stimper, 2009). Common evidence of n = 8 was found in many species of Genlisea; however, cytogenetic support for a common ancestral base number of x = 8 is still lacking for the sister genus Utricularia. Some support for the hypothesis of x = 8 in the clade is given by the fact that the supposed primary basic number in the common sister genus Pinguicula is also x = 8 according to Casper and Stimper (2009). As the cytology of Genlisea and its sister genus Utricularia is poorly studied and few chromosome counts are known at present, the estimated haploid number n = 8 and its evolutionary meaning obviously may change as more cytogenetic studies become available and are interpreted in the phylogenetic context (Cusimano et al., 2012).

Tetraploids with 2n = 4x = 32 can be found in the basal branching Genlisea section Africanae, here represented by G. hispidula and G. subglabra (Fig. 4). In the derived sections Recurvatae and Genlisea, likely dysploidy led to chromosome numbers of 2n = 36–38. Greilhuber et al. (2006) assumed that differences in chromosome number were a result of fission–fusion rearrangements, but not polyploidy; however, this hypothesis could not be tested in the focus of the present study. At least in the basalmost branching lineages of Genlisea subgenus Genlisea, we found what could be the true tetraploid group, with 2n = 32.

In Genlisea, the chromosome number was found to be negatively correlated with chromosome size. We observed that species with fewer chromosomes (diploids of Genlisea subgenus Tayloria with 2n = 16 and the tetraploids G. hispidula and G. subglabra with 2n = 32) have comparatively large chromosomes (especially evident in members of Genlisea subgenus Tayloria if compared with the rest of the genus; Fig. 2) than the remainder of Genlisea species of subgenus Genlisea, which possess very small chromosomes (dysploids with 2n = 36, 38 and ~46). The euploid species with larger chromosomes also displayed large genome sizes, while ultrasmall genomes could only be found in the dysploid species with very small chromosomes (Table 1, Fig. 4).

In G. aurea, flow cytometry measurements performed in Vienna for four accessions showed two DNA levels: one shared by three accessions yielded a 1C DNA content of ~65 Mbp and the other accession 117 Mbp (1C). In the Rutgers laboratory the values were similar but somewhat higher, at 73–83 and 119–131 Mbp, respectively. The value of 131 Mbp has been confirmed independently for G. aurea by Veleba et al. (2014) – the accession they used was G. aurea from Itacambira (only variety minor occurs in that area), but their value does not correspond to our results for the same taxon from that region. Feulgen densitometry (Greilhuber et al., 2006) had revealed 63·6 Mbp for G. aurea. A later Feulgen test in material from Gatersleben yielded 150 Mbp (1C) (I. Schubert, IPK Gatersleben, Germany, pers. comm.), and thus was similar to the flow cytometry data in Gatersleben and about twice the value measured previously by Feulgen staining, and even higher than that measured by flow cytometry in the present study. As genome size is not generally correlated to ploidy levels (Leitch and Bennett, 2007; Leong-Škorničková et al., 2007), this does not necessarily indicate the occurrence of two ploidy levels (e.g. as put forward by Veleba et al., 2014), but can rather be explained by the inclusion of two infraspecific taxa of G. aurea in the study (see below).

Genome size evolution

Interestingly, the ploidy shift from diploids (2n = 16 in Genlisea subgenus Tayloria) to likely tetraploids (2n = 32 at least in the basal-branching G. hispidula of subgenus Genlisea) does not directly correspond to a notable decrease in genome size (995–1722 Mbp in subgenus Tayloria compared with ~1500 Mbp in G. hispidula), as repeatedly observed in several angiosperm groups, where polyploidy often goes along with genome miniaturization (e.g. Leitch and Bennett 2004; Weiss-Schneeweiss et al., 2006). However, in the case of the two Genlisea subgenera we are dealing with two immediate sister clades but not a phylogenetic grade, which makes the reconstruction of the ancestral karyotype difficult. Thus, it is best to examine both subgenera independently, as previously done with contrasting morphological traits (Fleischmann et al., 2010).

In Genlisea subgenus Tayloria an apparent evolutionary increase in genome size is evident (Figs 4 and ​and5),5), ranging from 995 Mbp in the basalmost branching G. uncinata to ~1600–1700 Mbp in the more derived species G. violacea and G. lobata. This increase in 1C DNA content is consistent with the shift from perennial to annual life history (Fleischmann et al., 2010). This correlation of genome size with generation time is remarkable, as this tendency is opposite to what occurs in the majority of angiosperms, in which annual taxa generally show smaller genome sizes than perennial taxa (Bennett, 1972; Bennett and Leitch, 2011).

In the sister group Genlisea subgenus Genlisea, a similar situation was not observed, as annual and ephemeral taxa (such as G. oxycentron, 1C ~75 Mbp) do not show any significant difference in genome size compared with their perennial sisters (e.g. G. nigrocaulis, 1C = ~ 86 Mbp; Vu et al., 2012). However, the taxon sampling for 1C DNA content of annual species of Genlisea (e.g. G. stapfii, G. barthlottii and G. filiformis) was rather deficient in this study due to difficulties in obtaining enough plant material of these delicate short-lived species.

Within G. aurea, two distinct groups of genome size range were evident (one with ultrasmall genomes of 63·5–83 Mbp and another group displaying an almost 2-fold size range, of 117–131 Mbp). These two groups correlate with the two morphologically distinct, geographically more or less separated varieties, G. aurea var. aurea (ultrasmall genomes) and G. aurea var. minor (small genomes; Fleischmann, 2012). The wide span of genome sizes observed in G. aurea populations was assumed by Albert et al. (2010) to result from DNA double-strand breaks caused by reactive oxygen species (ROS) as by-products of increased oxidative phosphorylation due to a unique cytochrome c oxidase modification found in Lentibulariaceae (Jobson et al., 2004; Laakkonen et al., 2006). This ROS hypothesis was also used by Albert et al. (2010) to explain the small genome sizes and the very high nucleotide substitution rates observed in Genlisea and Utricularia (Jobson and Albert, 2002; Müller et al., 2004), as a result of error-prone DNA repair mechanisms and chromatin breaks. The same mutagenic action of self-produced ROS was postulated by Ibarra-Laclette et al. (2011a, b, 2013) for Utricularia gibba, likewise a species with a minimal genome. This would be consistent with the shorter non-coding sequences and introns and less repetitive sequences that characterize the ultrasmall genomes of G. aurea (Leushkin et al., 2013) and U. gibba (Ibarra-Laclette et al., 2013).

It is generally recognized that in plants, but also other organisms, genome sizes increase due to the proliferation of long terminal repeat (LTR) retrotransposons (Kellogg and Bennetzen, 2004). In contrast, the mechanisms of genome size reduction are less well characterized, but evidence is emerging from whole-genome analyses in Arabidopsis, rice (Oryza) and recently the miniature genomes of G. aurea and U. gibba. In a comparison of two closely related Arabidopsis species, A. thaliana and A. lyrata, the smaller genome of the former could be explained by many small deletions (Hu et al., 2011). In contrast, genome size differences in rice are explained by both LTR retrotransposon expansion and DNA loss as a result of double-strand break repair through non-homologous end joining (Ma and Bennetzen, 2004; Chen et al., 2013). Utricularia gibba has few, if any, full-length and presumably active LTRs, consistent with its reduction in genome size (Ibarra-Laclette et al., 2013). Interestingly, U. gibba has undergone three rounds of whole-genome duplication since ancestry with tomato (Solanum), and its genome has decreased to one-ninth the size while maintaining the standard number of genes for a plant (Ibarra-Laclette et al., 2013). It is compelling to speculate that there is a force actively removing LTR retrotransposons or excess DNA from whole-genome duplication, resulting in miniature genome sizes. The mechanism driving a genome to lose DNA is still unknown, and more in-depth studies of Genlisea genomes could therefore provide clues about the underlying mechanisms.

Further correlations between geography and genome size are evident in Genlisea. Accessions of G. repens from Brazil south of the Amazon (from Uberaba, Minas Gerais state, and from the Chapada dos Veadeiros in Goiás state) have relatively large genomes (Table 1, Fig. 5) compared with accessions from lower latitudes and high-altitude mountain summits of Venezuelan tepuis north of the Amazon basin (from Roraima tepui in Bolívar state and Cerro Aracamuni and Cerro Avispa in Amazonas state). This geographical difference in genome size is also mirrored in phylogenetic reconstructions (Fleischmann et al., 2010). Interestingly, G. repens with the larger genomes are phylogenetically close to G. oxycentron and G. nigrocaulis, species that exhibit ultrasmall genomes.

A geographical correlation of horizontal (latitudinal) and vertical (altitude) distribution with infraspecific DNA content was previously noticed in some plant species (Ohri and Khoshoo, 1986; Rayburn and Auger, 1990; Reeves et al., 1998; Bottini et al., 2000; Leong-Škorničková et al., 2007; Díez et al., 2013), while other studies found no such correlation (summarized in Wang et al., 2011). It is assumed that variation in genome size might have importance for the adaptation of plants to different ecological and environmental conditions (Bottini et al., 2000). All three accessions of G. repens from high-altitude tepui summits in Venezuela (from ~1350 m on Cerro Avispa and Aracamuni to 2700 m on Roraima tepui) had smaller genomes than the two accessions from upland savannah of central Brazil (altitude ~1000 m); this trend, however, is opposite to an observed increase in DNA content in Zea mays growing at increasing altitudes (Rayburn and Auger, 1990). Regarding the correlation of genome size with chromosome number, more data, particularly on chromosome counts, are needed to clearly establish a putative relationship between chromosome number and genome size.

Genlisea aurea and G. tuberosa have the smallest genomes, not G. margaretae

The smallest genome sizes currently known in angiosperms were obtained in the Feulgen densitometry measurements of the two sister species G. aurea var. aurea (1C values ranged from 63·4 to 67·2 Mbp; flow cytometry results were slightly higher, 1C ~73–83 Mbp) and G. tuberosa (1C ~65 Mbp, size estimation was based on a single run, but the low range was confirmed by flow cytometric results of 1C ~61 Mbp). The ultrasmall genome size of 63·4 Mbp reported from Feulgen measurements by Greilhuber et al. (2006) for G. margaretae from Madagascar (from a cultivated plant in Bonn Botanical Gardens) actually belongs to a specimen of G. aurea, probably even the same accession that yielded an identical value in this study. The ultrasmall 1C value of G. margaretae could not be reproduced with Feulgen densitometry or flow cytometric measurements, either for material of G. margaretae from Bonn or for any of the other accessions of the same species used in the present study (vouchers LE309 and LE260). All accessions of G. margaretae independently had an ~3-fold genome size of ~180–195 Mbp with both flow cytometry and Feulgen measurements. This was confirmed by flow cytometric measurements of material of identical strains of G. margaretae sent to the IPK Gatersleben (Vu et al., 2012; I. Schubert, pers. comm.). A later test made by J. Greilhuber with Feulgen densitometry on G. margaretae (LE309) revealed the same larger value, combined with a microscopic appearance of relatively dense euchromatin in the interphase nuclei even in differentiated tissue. A young trap fixed in ethanol-acetic acid (3:1) sent to Vienna yielded 185 and 189 Mbp (1C) in meristem and differentiated tissues, respectively.

Genlisea aurea and G. margaretae have a very similar vegetative morphology, with dense rosettes consisting of narrowly spathulate leaves, and are difficult to tell apart when not in flower. Although no herbarium vouchers are available, it seems likely that plants were mixed up in cultivation in Bonn, and material of G. aurea was erroneously used as ‘G. margaretae’ by Greilhuber et al. (2006). Potted plants in Bonn Botanic Garden labelled as the respective voucher ‘11400’ from Madagascar were studied personally by the first author in spring 2011, and these did in fact represent G. aurea. A photo-voucher made in Vienna by Eva Temsch from the material sent from Bonn for Feulgen measurements, for the published value from 2006, could clearly also be identified as G. aurea by the first author and Fernando Rivadavia.

Kamil Pasek, Thomas Carow, Matt Hochberg, Markus Welge and Matthias Teichert are thanked for providing additional plant material and Ingo Schubert, Jörg Fuchs, Lubomír Adamec and Adam Veleba for helpful correspondence. Two anonymous reviewers and the handling editor are thanked for valuable comments on the manuscript. This study was in part supported by the Deutsche Forschungsgemeinschaft, DFG Grant MU2875/2 to K.F.M. and by the DOE Plant Feedstock Genomics for Bioenergy Program (DE-FG02-08ER6430) to T.P.M.