Genome evolution of ferns: evidence for relative stasis of genome size across the fern phylogeny
Summary
- The genome evolution of ferns has been considered to be relatively static compared with angiosperms. In this study, we analyse genome size data and chromosome numbers in a phylogenetic framework to explore three hypotheses: the correlation of genome size and chromosome number, the origin of modern ferns from ancestors with high chromosome numbers, and the occurrence of several whole-genome duplications during the evolution of ferns.
- To achieve this, we generated new genome size data, increasing the percentage of fern species with genome sizes estimated to 2.8% of extant diversity, and ensuring a comprehensive phylogenetic coverage including at least three species from each fern order.
- Genome size was correlated with chromosome number across all ferns despite some substantial variation in both traits. We observed a trend towards conservation of the amount of DNA per chromosome, although Osmundaceae and Psilotaceae have substantially larger chromosomes. Reconstruction of the ancestral genome traits suggested that the earliest ferns were already characterized by possessing high chromosome numbers and that the earliest divergences in ferns were correlated with substantial karyological changes.
- Evidence for repeated whole-genome duplications was found across the phylogeny. Fern genomes tend to evolve slowly, albeit genome rearrangements occur in some clades.
Introduction
Ever since the first comprehensive studies on the genetics of ferns, the evolution of fern genomes has been considered paradoxical owing to the conservation of high chromosome numbers in taxa with demonstrated diploid gene expression (Haufler, 1987, 2002, 2014). This paradox led to the hypothesis that ferns underwent multiple cycles of polyploidy (whole-genome duplications (WGDs)) accompanied by subsequent diploidization involving gene silencing, but without apparent chromosome loss, so high chromosome numbers were retained (Haufler, 2002, 2014). Support for this hypothesis has been provided by observations that polyploidy contributes to c. 31% of speciation events in ferns compared with c. 15% in angiosperms (Wood et al., 2009). Recurrent WGD events without subsequent reduction in chromosome number and genome size may hence explain several characteristics of fern genomes, including the 80-fold variation in chromosome number (ranging from 2n = 18 in Salvinia natans to 2n = 1440 in Ophioglossum reticulatum), 94-fold genome size variation in ferns (ranging from 1C = 0.77 pg in Azolla microphylla to 1C = 72.68 pg in Psilotum nudum), the highest chromosome number of any plant known to date (2n = 1440), and the average chromosome number in homosporous ferns (n = 57.05), greatly exceeding the average chromosome number in angiosperms (n = 15.99) (Klekowski & Baker, 1966; Leitch & Leitch, 2012, 2013; Barker, 2013; Henry et al., 2015) (the estimated average chromosome number for ferns is updated later). Haufler's hypothesis predicted a correlation between genome size and chromosome number that has so far been found in ferns but not in angiosperms or gymnosperms (Nakazato et al., 2008; Bainard et al., 2011; Barker, 2013; Leitch & Leitch, 2013).
Based on these observations, fern genomes are considered to have shown greater stability in their chromosome structure over the last 400 million yr compared with the sister lineage of ferns – the seed plants (Leitch & Leitch, 2012; Haufler, 2014). This hypothesis appears to be consistent with the recent proposal of ‘static genomes’ in the royal ferns (Bomfleur et al., 2014; Schneider et al., 2015) and arguments suggesting that chromosome size expansion via accumulation of repeats plays only a minor role in the evolution of fern genomes compared with seed plants (Wagner & Wagner, 1980; but see Dyer et al., 2013). The prediction is also consistent with the relatively small number of studies providing evidence for single chromosome gains and losses in ferns (Lovis, 1977), including recent studies that have incorporated phylogenetic evidence in their analyses, for example Hymenophyllum (Hennequin et al., 2010), Lepisorus (Wang et al., 2010) and the Loxoscaphe complex in Asplenium (Bellefroid et al., 2010).
The hypothesis of recurrent cycles of hybridization, WGD and conservation of chromosomes has been challenged owing to the lack of strong evidence in expression sequence tag (EST) data in the polypod ferns Ceratopteris richardii and Pteridium aquilinum (Nakazato et al., 2008; Barker & Wolf, 2010; Barker, 2013). Nevertheless, recent transcriptome data for the horsetail Equisetum giganteum is not inconsistent with multiple WGD events contributing to the high chromosome numbers in this species (2n = 216; Vanneste et al., 2015).
It is possible to argue that artefacts created by sampling biases, for example, low taxonomic coverage of existing genome size measurements, could mislead some of these interpretations. In particular, the critical observation of the correlation between genome size and chromosome number in ferns may be affected by the current taxonomic sampling comprising < 1% of extant ferns compared with other land plants, especially angiosperms (1.8% coverage according to Leitch & Leitch, 2013), and the absence of genome size data for some phylogenetically important clades of ferns. To address this issue, we designed a study to test several of the major predictions derived from the existing hypotheses, including the repeated cycles of polyploidization, high chromosome numbers in the ancestors of extant ferns, and the conservation of chromosome size (Soltis & Soltis, 1987; Haufler, 2002, 2014) by increasing the number of reliable genome size measurements to cover > 2.5% of the taxonomic diversity of ferns. This included at least two species from each of the 11 orders of ferns currently recognized (Smith et al., 2006; Christenhusz & Chase, 2014), plus a comprehensive sampling of the tree fern lineage (Cyatheales) and basal polypod (Polypodiales) families (e.g. Dennstaedtiaceae, Lindsaeaceae, Lonchitidaceae, Saccolomataceae). To achieve this, we generated new genome size data using best-practice techniques and combined these with existing measurements available in the Pteridophyte DNA C-values database (Bennett & Leitch, 2012). We also assembled a comprehensive database of all chromosome numbers published for ferns. This strategy provided us with the evidence to test the outlined hypotheses using a phylogenetic framework based on our current understanding of the fern phylogeny (Wickett et al., 2014). The analyses confirm the predicted positive correlation between chromosome number and genome size in ferns, demonstrate that the main lineages of extant ferns originated from ancestors having already high chromosome numbers and a chromosome size that is generally conserved through the phylogeny of ferns, and show repeated cycles of WGD and karyotypic changes, which have affected not only the early diverging fern clades but also the more derived lineages of leptosporangiate ferns.
Materials and Methods
Obtaining new genome size data
New genome size measurements were made using freshly collected specimens from the living collections of three institutes: Fairy Lake Botanical Garden at Shenzhen (China), The Royal Botanic Gardens, Kew (UK), and The Royal Botanic Garden Edinburgh (UK) together with additional samples from the private collection of M. Christenhusz, Kingston-upon-Thames (UK). Vouchers were deposited at the corresponding herbaria of these institutes (PI (Institute of Botany of the Chinese Academy of Science, Beijing, China), K (Royal Botanic Gardens, Kew, UK), E (Royal Botanic Gardens Edinburgh, Edinburgh, UK)). To obtain genome size measurements, we selected the most appropriate of four available reference standards for each fern species: Petroselinum crispum ‘Champion Moss Curled’ (2C = 4.50 pg; Obermayer et al., 2002), Pisum sativum ‘Ctirad’ (2C = 9.09 pg; Doležel et al., 1992), Vicia faba ‘Inovec’ (2C = 26.90 pg; Doležel et al., 1992), and Allium cepa ‘Ailsa Craig’. The genome size for Allium cepa was recalibrated using both Vicia faba and Pisum sativum, and a value of 2C = 34.89 pg was determined, which is larger than the 2C-value of 33.55 pg (Bennett et al., 2000) but identical to that of Doležel et al. (1998). Based on the peak quality in the flow histograms, one out of four nuclei isolation buffers (Doležel et al., 1989; Ebihara et al., 2005; Loureiro et al., 2007; Supporting Information Table S1) was selected. The nuclear DNA content was measured using the one-step procedure (Doležel et al., 2007) involving the co-chopping of a leaf fragment of the fern with leaf material of the selected reference standard directly into 2 ml of the chosen buffer in a Petri dish over ice, filtering and staining with propidium iodide. For each sample, c. 5000 particles were measured using a PAII or CyFlowSL Partec flow cytometer (Partec GmbH, Goettingen, Germany) fitted with a 100 W high-pressure mercury lamp or a 100 mW green (532 nm solid-state Cobalt Samba laser; Cobolt AB, Solna, Sweden), respectively. The resulting flow histograms were analysed using the Partec software for flow cytometry FloMax 2.7. Where possible, nine measurements were made per sample. Mean 2C-values and standard deviations were calculated for each taxa based on the ratio of fluorescence between the fern and calibration standard (Pellicer & Leitch, 2014).
As extant ferns, especially homosporous species, often have multiple cytotypes (e.g. Ophioglossum reticulatum has individuals with 2n = 240, 480, 720, 960, 1440; Khandelwal, 1990) which differ in holoploid 2C-values because of differences in the number of monoploid chromosome sets, insights into the genome size stability of the monoploid chromosome set can be obtained by calculating the monoploid genome size (= 1Cx-value; see Greilhuber et al., 2005). To do this requires identification of the base chromosome number x, which has been recognized as a challenge in ferns (Duncan & Smith, 1978). Here, we accepted the lowest gametophytic chromosome number (n) reported for a genus (including any counts that could be explained by single chromosome gains and losses) to be the most likely x for the investigated genus and, where possible, family (Manton & Vida, 1968). In many cases, chromosome counts were not available for the actual individual used for genome size estimation, so there is the potential that the 1Cx-value is over- or underestimated if an incorrect n or x value is assumed. To take this into account, several approaches were taken: wherever possible, chromosome counts were made for the plants studied; chromosome counts/ploidy levels were assumed based on previous reports in the literature; genome sizes were estimated for the same individual which had previously been reported by other authors; and the modelling of genome size and chromosome evolution (see later) were conducted multiple times using different x values to investigate the effect of this uncertainly on the reconstruction of ancestral states. In addition, to avoid the ambiguity introduced by the selection of x, we also calculated a further value as a proxy for the average chromosome size of a species. This was determined by dividing the 2C-value by the somatic chromosome number 2n. This provides an insight into the extent of variation in the average chromosome size across ferns. Nevertheless, it is also recognized that this value may be incorrectly calculated for species where an actual chromosome count was not made for the individual used for genome size estimation. The same approaches were used as earlier to take into account these uncertainties and hence increase the robustness of the analyses.
Integration of new and existing data
We assembled the following data for each corresponding taxon: 1C-values, chromosome number (2n) and plastid DNA sequences (see Table S1). We checked and assembled all published 1C-values of ferns using the Pteridophyte DNA C-values database (Bennett & Leitch, 2012) or recent publications (Ekrt et al., 2009; Bainard et al., 2011; Nitta et al., 2011; Williams & Waller, 2012; Bou Dagher-Kharrat et al., 2013; Chang et al., 2013; Dyer et al., 2013; Pustahija et al., 2013). The chromosome numbers of ferns were obtained from online databases (http://ccdb.tau.ac.il and http://chromosomes.binoz.uj.edu.pl/chromosomes in August 2015, plus the unpublished database comprising all fern chromosome counts assembled by H.S.). To estimate taxonomic coverage, we accepted the estimates of species diversity published in Smith et al. (2006).
Phylogenetic framework
The phylogenetic framework was generated by assembling a data matrix of at least one of three plastid genome regions (atpA, atpB, rbcL) accessible in GenBank (Table S1) for taxa with genome size data available. In some cases, DNA sequences of closely related species were used, as DNA sequences were not available for the sampled species.
The molecular dataset was assembled and aligned using Mesquite v.2.75 (Maddison & Maddison, 2014) and the best-fitting model of molecular evolution was determined using jModeltest v.2.1.4 (Darriba et al., 2012) to generate the selection statistics and the Bayesian information criterion. The phylogenetic framework was then estimated using MrBayes v.3.2 (Ronquist et al., 2012) with two independent runs for three million generations with the convergence verified using Tracer v.1.6 (Rambaut et al., 2014). To test the impact of the uncertain phylogenetic placement of horsetails, we also used alternative phylogenetic hypotheses including Equisetales sister to all other ferns, Equisetales sister to the clade comprising Ophioglossales and Psilotales, Equisetales sister to Marattiales, and Equisetales sister to Polypodiales (Pryer et al., 2001; Kim et al., 2014; Wickett et al., 2014). When required, polytomies were artificially resolved according to current consensus on fern phylogeny (Lehtonen, 2011). The phylogenetic signal of chromosome numbers (2n), base number (x), holoploid genome size (1C), monoploid genome size (1Cx), and the average chromosome size (2C/2n) was analysed with the traits module in Phylocom v.4.2 (Webb et al., 2008) with significance tested via randomization of trait values across the Bayesian trees. Furthermore, the phylogenetic signal of each trait was explored by applying phylogenetic dependent tests alongside phylogenetic independent tests, such as Spearman's rank correlation as implemented in the Hmisc package in R 3.14 (Harrell, 2014). Phylogenetic dependent tests comprised the calculation of phylogenetic independent contrasts (PICs; Felsenstein, 1985) and Felsenstein's contrast correlation between positivized contrasts (Garland et al., 1993) using the Pdap:pdtree package 1.16 (Midford et al., 2011) in Mesquite 2.75 with the trees obtained from the Bayesian analyses. Ancestral character states of holoploid genome size (a1C), monoploid genome size (a1Cx), and average chromosome size (a2C/2n) were estimated using BayesTraits v.2.0 (Pagel et al., 2004). The latter analyses were performed with 100 trees randomly drawn from the posterior distribution of the phylogenetic reconstruction and values were estimated from a Bayesian Markov chain Monte Carlo analysis of two million generations sampling every 1000th generation and discarding the first 10 000 generations. Reconstructed trait values were averaged over the posterior distribution with a 95% confidence interval calculated. The base chromosome number was inferred using ancestral state reconstruction in ChromEvol v.2.0 (Mayrose et al., 2010). The ancestral diploid chromosome number (a2n) was estimated using a guide tree derived from the Bayesian consensus tree with taxa with uncertain chromosome number set to an equal likelihood. Character states were optimized using the model assuming constant rate of chromosome gain, loss and duplication along with an estimated rate of demiduplication because this model was selected based on the output of the initial analyses with 10 models of chromosome evolution with the Akaike information criterion (Table S2).
Results
The taxonomic coverage of ferns with genome size data was increased to 2.8% by the addition of genome size measurements for 110 newly studied species, corresponding to first records for 47 genera, and one order (Gleicheniales) (Tables 1, S1). Each order of ferns now includes genome data for at least two distinct species, with taxon coverage ranging from 1.0% for Hymenophyllales to 93.3% for Equisetales. The data also include the first measurements for crucial genera such as Tmesipteris, the sister genus of Psilotum (Table S1), and three genera belonging to the early diverging lineage of the species-rich order Polypodiales, that is, Lindsaea, Lonchitis and Saccoloma. The new measurement of Tmesipteris (2C = 144.1 pg) is similar to the extremely large genome previously recorded in the whisk fern Psilotum nudum (2C = 145.2 pg), which has the largest genome size of any fern so far studied (Tables 1, S1). Nevertheless, our new data show that large genomes are not restricted to early-diverging ferns, as shown by the discovery of considerably larger genomes in some Polypodiales than previously reported. These were found in species nested in a clade within eupolypods I, that is, Dracoglossum plantagineum (2C = 85.5 pg) and Mickelia nicotianifolia (2C = 121.0 pg), but also within the early diverging lineages of Polypodiales, that is, Saccoloma domingense (2C = 77.5 pg). Relatively large genomes were also recorded in Bolbitis (2C = 50.7 and 55.3 pg) and Elaphoglossum (2C up to 67.0 pg), which form a clade together with Mickelia. Thus large genomes may be restricted to a few clades in the derived ferns.
- a Total species numbers based on Smith et al. (2006).
- SN, species number; TC, taxon coverage.
The smallest mean genome sizes were found in the leptosporangiate orders Gleicheniales and Salviniales, the latter including the smallest genome size of any fern so far reported (i.e. Azolla microphylla; 1C = 0.77 pg, Obermayer et al., 2002). However, whether these small mean values reflect the limited sample sizes (with only three estimates for each order) remains to be determined. Certainly the more extensive range of chromosome numbers encountered in each group (Salviniales 2n = 18–120; Gleicheniales 2n = 40–232) suggests that larger genome sizes may be encountered in these orders as data increase.
Based on the collation of chromosome data from different sources, chromosome counts are now available for 2639 species corresponding to 28.9% of all ferns, with numbers in the range 2n = 18–1440, and an estimated mean gametic chromosome number of n = 60.5 for all ferns included (Table 1; according to the database incorporating all accessible chromosome counts: n = 63.5 for all ferns, n = 63.5 for homosporous ferns and n = 58.4 for heterosporous ferns). The 11 orders of ferns show distinct differences in the extent of variation in genome size and chromosome number (Table 1; Fig. 1).
Overall there was a significant positive relationship between homoploid genome size (1C) and chromosome number (Table 2; Fig. 2). Both phylogenetic and nonphylogenetic tests found significant support for a correlation between monoploid genome size (1Cx) and estimates of base chromosome number (P < 0.001; Table S1). However, it is notable that, based on available data, there was no clear relationship between the range in chromosome sizes and genome size across the phylogeny of ferns (P < 0.01 for range-1C/range-2n; Table 1). For example, while the constant chromosome number of 2n = 44 coincides with a low variation in genome size (1.56-fold; 1C = 13.46–21.01 pg) in Osmundales, genome sizes in Equisetales range over 2.37-fold (1C = 12.78–30.35 pg) and yet their chromosome number is also highly conserved, with all species analysed to date having 2n = 216 (Fig. 1; Table 1).
| No. taxa | Nonphylogenetic test Spearman's correlation coefficient | Phylogenetic test Felsenstein contrast correlation | ||||
|---|---|---|---|---|---|---|
| ρ | P-value | No. contrasts | r | P-value | ||
| All ferns | 185 | 0.44 | < 0.001 | 166 | 0.50 | < 0.001 |
| Nonleptosporangiate fern grade | 38 | 0.61 | < 0.001 | 27 | 0.27 | 0.16 |
| Leptosporangiate ferns | 147 | 0.22 | 0.006 | 138 | 0.61 | < 0.001 |
| Leptosporangiate ferns without Osmundales | 136 | 0.42 | < 0.001 | 129 | 0.61 | < 0.001 |
| Core leptosporangiate ferns | 125 | 0.44 | < 0.001 | 119 | 0.60 | < 0.001 |
| Polypodiales | 111 | 0.46 | < 0.001 | 104 | 0.60 | < 0.001 |
| Eupolypod ferns | 90 | 0.44 | < 0.001 | 86 | 0.61 | < 0.001 |
| Eupolypods I | 38 | 0.39 | 0.02 | 34 | 0.54 | < 0.001 |
| Eupolypods II | 52 | 0.40 | 0.003 | 51 | 0.69 | < 0.001 |
Discussion
Prediction 1: There is a positive correlation between chromosome number and genome size
The considerably increased sampling of genome size and chromosome data from 231 taxa across the phylogenetic tree of ferns, analysed using either the phylogeny-independent Spearman's correlation coefficient (SCC) or the phylogeny-dependent Felsenstein contrast correlation (FCC) (Table 2) clearly supports the hypothesis that genome size and chromosome number are correlated in ferns. The correlation was also supported for most of the major subgroups when analysed separately. The only exceptions were found within the nonleptosporangiate fern grade and the eupolypods I (Polypodiaceae s.l.) clade where FCC failed to support a positive correlation, even though SCC did (Table 2). Thus, while the increase of chromosome numbers, through either polyploidy or other chromosomal processes, does not seem to be accompanied by extensive loss of DNA in the majority of fern lineages (a situation that contrasts strongly with observations in angiosperms; Leitch & Bennett, 2004), this configuration is not universal amongst ferns. However, the recovery of the same correlation between the monoploid genome size 1Cx and the base chromosome number x (P < 0.001) rejects the argument that this pattern is caused by neopolyploidy alone. Future work is certainly needed to focus within the nonleptosporangiate fern grade and eupolypods I to identify whether the lack of support for the correlation using FCC is a sampling artefact, or the result of genuine distinctive genomic processes operating within these lineages as suspected (see Prediction 3 later).
Prediction 2: The main lineages of extant ferns originated from ancestors with high chromosome numbers and chromosome size is broadly conserved throughout the phylogeny of ferns
The results presented here reconstruct high ancestral chromosome numbers (a2n) for each of the main extant lineages of ferns, with estimated a2n ranging from 40 in Schizaeales to 216 in Equisetales and an estimate of a2n = 44 for the ancestors of all ferns (Table 3). Given the high a2n of nonleptosporangiate ferns (a2n = 44–216), together with the estimated a2n = 48 for leptosporangiate ferns (both with and without Osmundales; Table 3), the data suggest that the main lineages of extant ferns originated from ancestors having already high chromosome numbers (Table 3). The presence of even higher a2n in some of the derived lineages of leptosporangiate ferns such as Hymenophyllales (a2n = 72), Cyatheales (a2n = 68) and Polypodiales (a2n = 64), especially all eupolypods I (a2n = 82), suggest that further increases have taken place along their evolutionary history leading to the establishment of even higher chromosome numbers at the base of these lineages (see Prediction 3 later).
| a1C (pg) | a1Cx (pg) | a2n | a2C/2n | |
|---|---|---|---|---|
| All ferns | 28.67 ± 0.32 | 14.60 ± 0.08 | 44 | 0.449 ± 0.005 |
| Psilotales + Ophioglossales | 37.40 ± 0.34 | 18.84 ± 0.19 | 44 | 0.523 ± 0.005 |
| Psilotales | 67.88 ± 0.25 | 34.27 ± 0.14 | 104 | 0.621 ± 0.003 |
| Ophioglossales | 32.19 ± 0.34 | 11.59 ± 0.20 | 90 | 0.270 ± 0.004 |
| Equisetales | 19.63 ± 0.20 | 19.80 ± 0.11 | 216 | 0.187 ± 0.003 |
| Marattiales | 10.61 ± 0.19 | 9.62 ± 0.12 | 78 | 0.233 ± 0.003 |
| Leptosporangiate ferns | 15.80 ± 0.33 | 12.78 ± 0.21 | 48 | 0.424 ± 0.004 |
| Osmundales | 15.91 ± 0.18 | 15.44 ± 0.10 | 44 | 0.690 ± 0.002 |
| Leptosporangiate ferns (minus Osmundales) | 11.84 ± 0.31 | 10.04 ± 0.18 | 48 | 0.266 ± 0.005 |
| Hymenophyllales | 14.97 ± 0.29 | 14.04 ± 0.17 | 72 | 0.327 ± 0.005 |
| Gleicheniales | 10.24 ± 0.30 | 9.58 ± 0.17 | 48 | 0.250 ± 0.005 |
| Schizaeales | 11.60 ± 0.44 | 9.16 ± 0.25 | 40 | 0.251 ± 0.006 |
| Core leptosporangiate ferns | 9.67 ± 0.26 | 8.73 ± 0.16 | 46 | 0.142 ± 0.004 |
| Salviniales | 5.68 ± 0.35 | 4.81 ± 0.21 | 46 | 0.099 ± 0.005 |
| Cyatheales | 9.45 ± 0.16 | 8.91 ± 0.10 | 68 | 0.134 ± 0.002 |
| Polypodiales | 16.07 ± 0.30 | 7.34 ± 0.18 | 64 | 0.156 ± 0.004 |
| Dennstaedtiaceae | 10.78 ± 0.24 | 5.70 ± 0.14 | 58 | 0.159 ± 0.004 |
| Pteridaceae | 9.13 ± 0.25 | 5.13 ± 0.15 | 60 | 0.177 ± 0.003 |
| Eupolypods | 10.95 ± 0.19 | 7.44 ± 0.11 | 54 | 0.190 ± 0.002 |
| Eupolypods I | 13.24 ± 0.18 | 9.11 ± 0.10 | 82 | 0.228 ± 0.003 |
| Dryopteridaceae | 13.25 ± 0.18 | 9.11 ± 0.10 | 82 | 0.229 ± 0.002 |
| Nephrolepidaceae | 8.69 ± 0.12 | 8.90 ± 0.05 | 82 | 0.216 ± 0.002 |
| Tectariaceae | 12.02 ± 0.16 | 11.56 ± 0.09 | 82 | 0.281 ± 0.002 |
| Davalliaceae | 10.60 ± 0.15 | 10.24 ± 0.09 | 80 | 0.257 ± 0.002 |
| Polypodiaceae | 12.02 ± 0.14 | 10.86 ± 0.07 | 74 | 0.297 ± 0.002 |
| Eupolypods II | 9.90 ± 0.15 | 7.05 ± 0.08 | 54 | 0.179 ± 0.002 |
| Cystopteridaceae | 8.00 ± 0.17 | 4.98 ± 0.09 | 82 | 0.132 ± 0.003 |
| Thelypteridaceae | 10.09 ± 0.15 | 7.12 ± 0.09 | 50 | 0.195 ± 0.003 |
| Onocleaceae | 13.29 ± 0.16 | 12.78 ± 0.10 | 78 | 0.329 ± 0.003 |
| Blechnaceae | 12.06 ± 0.15 | 10.93 ± 0.09 | 70 | 0.306 ± 0.003 |
| Athyriaceae | 12.02 ± 0.13 | 8.61 ± 0.08 | 80 | 0.223 ± 0.002 |
| Woodsiaceae | 7.40 ± 0.17 | 5.33 ± 0.10 | 78 | 0.142 ± 0.003 |
| Aspleniaceae | 8.86 ± 0.16 | 5.97 ± 0.09 | 72 | 0.173 ± 0.002 |
With the exception of Marattiales, all nonleptosporangiate fern lineages are characterized by possessing large genome sizes with a mean 1C-value > 20.0 pg (Table 1) and an ancestral monoploid genome size (a1Cx) > 11.0 pg (Table 3).
Extremely large genomes with 1C > 35.0 pg (Table 1) were concentrated in the sister orders Ophioglossales and Psilotales, but were also found for the first time in genera of derived leptosporangiate ferns within Polypodiales (i.e. Dracoglossum, Mickelia and Saccoloma; see Table S1). Such data, together with the high a2n estimates, support the predicted repeated establishment of ferns with large genomes through polyploidization without subsequent genome downsizing in these lineages. However, denser taxonomic sampling may show that this general trend is not strictly conserved. In particular, three patterns may require special attention: lineages showing evidence for reduction of chromosome number (e.g. Hymenophyllales); increased chromosome number without substantial increase of genome size (e.g. Cyatheales); and frequent variation of the chromosome number unlinked with ploidy (e.g. clade comprising relatives of Blechnum, Onoclea and Woodwardia).
It has often been stated that fern chromosomes are generally characterized by being small and uniform in size, in contrast to the diversity encountered in seed plants (e.g. Wagner & Wagner, 1980; Nakazato et al., 2008). Indeed, based on available data, an analysis of the range of chromosome sizes (2n/2C) shows that while angiosperm chromosome sizes range over 3100-fold (2n/2C, 0.003–9.300 pg per chromosome), in ferns they range only 31-fold, from 0.035 to 0.955 pg per chromosome. The results presented here, in which an estimate of ancestral chromosome sizes have been reconstructed using a2C/2n, broadly support this predicted overall conservation of small chromosomes throughout fern evolution, with an inferred a2C/2n value for eight out of 11 fern orders falling within the standard deviation of the mean value (Table 3). Nevertheless, there are exceptions, with larger chromosomes in the two previously mentioned sister orders, Ophioglossales (a2C/2n = 0.270 pg) and Psilotales (a2C/2n = 0.621 pg); the earliest diverging order of leptosporangiate ferns, the royal ferns (Osmundales, a2C/2n = 0.690 pg); the clade comprising Blechnum, Onoclea, Woodwardia and relatives; and the clade comprising relatives of Elaphoglossum (Polypodiales subfam. Dryopteridaceae) (Tables 3, S1).
Within the leptosporangiate ferns, the uniqueness of the chromosomes of Osmundales has previously been recognized (Manton, 1950), and while the remaining lineages show evidence for rather limited chromosome size variation, as noted earlier, some clades do show distinct trends that deviate from this. For example, a trend towards smaller genomes and chromosomes in the heterosporous Salviniales (a1Cx = 4.81 pg and a2C/2n = 0.099 pg; Table 3) is detected, while a tendency towards larger genomes coinciding with larger chromosomes in Hymenophyllales (a1Cx = 14.04 pg and a2C/2n = 0.327 pg) and several clades of eupolypods I and II is also evident (Table S1).
In summary, the recovered results are consistent with the prediction that ancestors of the extant lineages of ferns shared high chromosome numbers, whereas the previously predicted stability in chromosome size is found in some, but not all, of the main fern lineages.
Prediction 3: Repeated cycles of WGD and chromosome composition changes in the ancestors of ferns contribute to the diversity and high chromosome numbers encountered in extant ferns
Recurrent WGDs may explain the observed accumulation of large genomes coinciding with high chromosome numbers found in early diverging lineages such as Equisetales (a2n = 216), Marattiales (a2n = 78), Ophioglossales (a2n = 90) and Psilotales (a2n = 104), based on an estimated a2n of 44 for all ferns and a2n of 44 for the common ancestor of the Psilotales + Ophioglossales clade. These results are consistent with the evidence for a WGD in the common ancestor of the extant horsetails detected in transcriptome data (Vanneste et al., 2015).
Based on an estimated a2n = 48 reconstructed at the base of all leptosporangiate ferns, there is also evidence for at least three independent putative WGDs within this species-rich fern clade; namely the ancestors of Hymenophyllales (filmy ferns; a2n = 72), the base of the sister lineages Cyatheales (tree ferns; a2n = 68) and Polypodiales (a2n = 64), and the ancestors of eupolypods I (a2n = 82) (see Fig. 1).
In Hymenophyllales, the extant crown group is predicted to have originated from ancestors with an inferred a2n = 72, and yet some extant species belonging to the derived Hymenophyllum clade are reported to have chromosome numbers as low as 2n = 22. It is suggested that these are likely to be the result of chromosome deletions following WGD (see Hennequin et al., 2010). Support for a WGD at the base of Hymenophyllales is also suggested from the ancestral monoploid genome size (a1Cx) of 14.04 pg, which is larger than the ancestral a1Cx of the other two precore leptosporangiate orders, Gleicheniales (a1Cx = 9.58 pg) and Schizaeales (a1Cx = 9.16 pg), and leptosporangiate ferns minus Osmundales clade (a1Cx = 10.04 pg). Nevertheless, the observed 1C-values estimated for the three Gleicheniales species studied (i.e. 1C = 4.9, 6.4, 6.5 pg, Table S1) suggest that an even lower a1C-value, closer to half of the current estimate of a1C = 9.58 pg may be more realistic for this order.
While the higher a2n reconstructed for Cyatheales (a2n = 68) and Polypodiales (a2n = 64) suggest a WGD in their most recent common ancestor, the increase in the a1C to 16.07 pg at the base of Polypodiales compared with a1C = 9.67 pg of core leptosporangiates and a1C = 9.45 pg of Cyatheales (Table 3) suggests that if a WGD did give rise to the higher a2n observed in Cyatheales, this may have been accompanied by some genome downsizing.
In this context, it is important also to note that core leptosporangiate ferns and their three orders (Salviniales, Cyatheales and Polypodiales) tend to show lower inferred a2C/2n values than other fern orders (Table 3), perhaps suggesting some reduction in genome size at the base of this lineage. However, such a pattern needs to be considered in the context of the observed variation of these values among Polypodiales, for example, increased 2C/2n values in some eupolypods. For example, the clade of eupolypods II comprising the relatives of Blechnum, Onoclea and Woodwardia shows evidence of karyological changes, as indicated by the a2C/2n values of 0.329 and 0.306 pg, respectively, compared with values found in the closely related clade, which all have 2C/2n < 0.230 pg (Table 3). Nevertheless, this clade does not show a correlation between chromosome number and genome size (P = 0.118). Clearly further sampling within the core leptosporangiates is required to even out the effects of limited sample size that may have resulted in a too low estimate of the a2n in this major clade of derived ferns a2n = 54 (Table 3).
In the eupolypods I, the clade comprising the two mainly epiphytic lineages, Davalliaceae/Davallioideae and Polypodiaceae s.s. Polypodioideae, plus its relatives the climbing Oleandraceae/Oleandroideae and mainly terrestrial Tectariaceae/Tectarioideae, show increased a2C/2n values compared with the mainly terrestrial Dryopteridaceae/Dryopteridoideae and related clades. These coincide with a reduction in the ancestral chromosome number from a2n = 82 (Dryopteridaceae/Dryopteridoideae) via a2n = 80 (in Davalliaceae/Davallioideae) to a2n = 74 (in Polypodiaceae s.s/Polypodioideae), suggesting chromosome rearrangements following WGD but not accompanied by extensive changes in genome size. By contrast, within Dryopteridaceae/Dryopteridoideae there is evidence of genome size increases in the clade comprising the genera Bolbitis-Elaphoglossum-Mickelia (including the largest genome size so far reported for any derived leptosporangiate fern, i.e. 2C = 121.0 pg in Mickelia nicotianifolia), whereas other clades show evidence for conservation of genome size. In this context it is worth noting that the estimated ancestral chromosome number is more variable among the clades of eupolypods II (a2n = 50–82) than among those of eupolypods I (a2n = 74–82). Such results indicate that substantial changes in genome size, chromosome number and chromosome size have taken place during the recent diversification of these derived clades of ferns and point to these genomes being more dynamic than perhaps hitherto recognized.
Overall the results support the prediction that the high chromosome numbers in ferns have resulted, in part, from repeated cycles of WGD and chromosome composition changes that have affected not only the early diverging fern clades but also the more derived lineages of leptosporangiate ferns.
Phylogenetic uncertainty and the robustness of the inferred results
To infer the robustness of the reported results, we explored the impact of alternative phylogenetic hypotheses regarding the relative positions of the five basal orders of ferns (Equisetales, Marattiales, Ophioglossales, Osmundales, Psilotales; Fig. S1). The reported results indicate little variation in the reconstructed genome size estimates caused by changes in the tree topology. For example the reconstructed genome size for the ancestor shared by the sister orders Ophioglossales-Psilotales ranges from a1C = 37.24–38.43 pg across the four topologies explored (Fig. S2). However, we anticipate that the variation for several orders will increase as the amount of genome size data improves, especially in Gleicheniales. Certainly, increased sampling is required to improve our understanding of genome size evolution in Hymenophyllales and the eupolypods. The still relatively low taxon coverage of 2.8% for all ferns, and especially of some species-rich clades such as Hymenophyllales (1.0%) and Polypodiales (2.4%), is a concern, because there is some tantalising evidence that genome reorganization, including genome downsizing and chromosome rearrangements, may have occurred in these lineages, but the data are currently too limited to draw firm conclusions. If such processes have occurred, they will only become recognizable when a denser sampling can provide sufficient sensitivity to detect these patterns.
Other potential limitations of the current data to the accuracy of the analyses include errors arising from genome size estimates, determination of chromosome counts, and violations of the assumption that chromosome number is generally stable in the majority of fern species. Nevertheless, while only a small percentage of specimens used for our genome size measurements had their chromosome numbers determined (< 10%), there was no evidence to support these concerns, because measurements of specifically collected specimens with recently published chromosome counts fitted well into the pattern we observed. It is noted that the effect of sampling density may be especially high in the context of the estimated ancestral diploid chromosome numbers of Psilotales and Polypodiaceae/eupolypods I.
Concluding remarks
In summary, the increased sampling of genome size measurements presented here has provided evidence to support earlier predictions based on hypotheses of repeated WGD events in the ancestors of extant ferns. This includes evidence for repeated WGDs during the diversification of leptosporangiate ferns, for example Hymenophyllales, Polypodiales and the eupolypods I clade of the derived ferns (Fig. 1). This hypothesis is not inconsistent with the results of investigations based on comparative analyses of the transcriptomes of the polypod ferns Pteridium (Dennstaedtiaceae) and Ceratopteris (Pteridaceae), which suggested a WGD linked with the divergence of Polypodiales, although the data were difficult to interpret (Barker & Wolf, 2010). Our data also suggest some genomic rearrangements in the common ancestor of Cyatheales and Polypodiales as indicated by the rise of the ancestral chromosome number from around a2n = 46 in core leptosporangiate ferns to a2n = 68 in Cyatheales and a2n = 64 in Polypodiales, and a rise in the ancestral genome size from a1C = 9.67 pg in core leptosporangiates to a1C = 16.07 pg in Polypodiales but not in Cyatheales with an a1C = 9.45 pg. Future studies could concentrate on the recovered evidence for changes in the ancestral chromosome number in Cyatheales and Polypodiales, as this may provide further insights into the nature of the evolutionary processes that have enabled these lineages to diversify in the shadow of angiosperms (Schneider et al., 2004). The limited data for Hymenophyllales suggest that the ancestors for the crown group underwent a WGD, while the low chromosome numbers in some derived Hymenophyllum species are a result of substantial chromosome restructuring, probably involving chromosome fusion following WGD. Together with the findings of increased genome sizes in some of the more derived fern clades that include epiphytes, the results may suggest a link between genome size evolution and the colonization of and adaptation to epiphytic habitats.
The new findings are broadly consistent with the generally accepted view that fern genomes are less dynamic than those of angiosperms (Leitch & Leitch, 2013). For example, the 2C/2n – DNA content per chromosome – appears to be broadly conserved throughout the evolution of ferns, with the notable exception of Osmundales and the Psilotales-Ophioglossales clade. Furthermore, our results are consistent with the observations on the fern genome spaces based on low-coverage whole-genome shotgun sequencing of six fern species, especially the absence of a correlation between genome size and repeats in ferns (Wolf et al., 2015). However, this does not mean that all fern genomes are ‘frozen’ in time. WGD events have clearly played an important role in the evolutionary history of ferns and this has been accompanied by changes in genome size, and in chromosome size and chromosome reorganization in some lineages. Certainly the data hint at substantial divergences in the eupolypod I clade comprising Bolbitis, Dracoglossum, Elaphoglossum and Mickelia, highlighting the need for denser sampling within eupolypods as noted earlier. Indeed, it seems possible that as sampling improves, the general trend of a correlation between chromosome number and genome size may not be recovered in some clades of ferns that have undergone recent changes in genome organization.
While an analysis based on chromosome numbers has noted that polyploidy may contribute more frequently to speciation events in ferns compared with angiosperms (see Wood et al., 2009), whether there are fundamental differences in the distribution, frequency and genomic consequences of WGD between ferns and angiosperms will have to wait until there is a sufficient amount of genomic data for ferns to enable meaningful comparisons to be made. Comparing our results with studies on angiosperms (Soltis et al., 2009; Jiao et al., 2011) reveals some striking similarities but also differences such as a lower frequency of WPGs in ferns than in angiosperms. This observation is in conflict with the higher rate of neopolyploidy in ferns compared with angiosperms (Wood et al., 2009). It is certainly felt that at present, the meagre amount of genomic and transcriptome data for ferns, together with the striking differences in species diversity and evolutionary ages between ferns and angiosperms, means that any apparent similarities or differences may be misleading, and at this stage it would be highly speculative to try to establish evolutionary trends solely based on numeric comparisons between these groups. Such discoveries and insights will have to wait until the advances being made in genomic sequencing technologies impact more extensively on the ferns.
Acknowledgements
The authors are grateful to the support by the gardens of RBG Edinburgh, RBG Kew, The Australian National Botanical Garden Canberra, the Botanische Staatssammlung Munich, and colleagues sharing material (S. Siljak-Yakovlev, S. Zona). The project was part of the MSc thesis by J.C. generated in the course ‘Taxonomy and Biodiversity’ run jointly by the Natural History Museum and Imperial College London. The research was supported by the Chinese Academy of Science senior visiting professorship to H.S., scholarship of the Chinese Scholarship Council to H.L., and BBSRC PhD project to J.C. We declare we have no competing interests.
Author contributions
H.S. and I.J.L. conceived of and coordinated the study; J.C. carried out statistics and phylogenetic analyses; J.C., O.H., J.P. and J.M. generated genome size data; H.L., M.C., Y.R., S.Z. and M.G. provided critical material and chromosome counts; H.S. drafted the manuscript with the contribution of all authors. All authors gave final approval for publication.
