Biogeography and genome size evolution of the oldest extant vascular plant genus, Equisetum (Equisetaceae)

Abstract

Background and Aims

Extant plant groups with a long fossil history are key elements in understanding vascular plant evolution. Horsetails (Equisetum, Equisetaceae) have a nearly continuous fossil record dating back to the Carboniferous, but their phylogenetic and biogeographic patterns are still poorly understood. We use here the most extensive phylogenetic analysis to date as a framework to evaluate their age, biogeography and genome size evolution.

Methods

DNA sequences of four plastid loci were used to estimate divergence times and investigate the biogeographic history of all extant species of Equisetum. Flow cytometry was used to study genome size evolution against the framework of phylogenetic relationships in Equisetum.

Key Results

On a well-supported phylogenetic tree including all extant Equisetum species, a molecular clock calibrated with multiple fossils places the node at which the outgroup and Equisetum diverged at 343 Mya (Early Carboniferous), with the first major split among extant species occurring 170 Mya (Middle Jurassic). These dates are older than those reported in some other recent molecular clock studies but are largely in agreement with a timeline established by fossil appearance in the geological record. Representatives of evergreen subgenus Hippochaete have much larger genome sizes than those of deciduous subgenus Equisetum, despite their shared conserved chromosome number. Subgenus Paramochaete has an intermediate genome size and maintains the same number of chromosomes.

Conclusions

The first divergences among extant members of the genus coincided with the break-up of Pangaea and the resulting more humid, warmer climate. Subsequent tectonic activity most likely involved vicariance events that led to species divergences combined with some more recent, long-distance dispersal events. We hypothesize that differences in genome size between subgenera may be related to the number of sperm flagellae.

INTRODUCTION

Equisetum has long been a focus of attention for botanists and palaeontologists because, given its long and well-documented fossil record, it is considered a key element in trying to understand the evolution of vascular plants (Scott, 1900; Browne, 1908; Eames, 1936; Boureau, 1964; Hauke, 1963, 1978; Bierhorst, 1971; Rothwell, 1999; Pryer et al., 2001; Karol et al., 2010; Husby, 2013; Elgorriaga et al., 2018). The number of extant species has recently been reported to be 18 (Christenhusz et al., 2019; Table 1) based on phylogenetic and taxonomic studies, but it is clear from fossil data that the genus was much more diverse in the ancient past (Elgorriaga et al., 2018; Clark et al., 2019).

Table 1.

Overview of the classification and distribution of Equisetum including information on the samples used in this study

Classification (Christenhusz et al., 2019)	Distribution	Collection (herbarium) [GenBank numbers rbcL, trnL-F, rps4]
Subgenus Paramochaete Christenh. & Husby
Equisetum bogotense Kunth	Andes, Galapagos, Costa Rica	Husby 19 (FTG) [MH750045, MH750162, MH750115]
Subgenus Equisetum
E. arvense L.	Panboreal	McBeath 1622 (E) [MH750038, MH750155, MH750108]
E. braunii J.Milde	Western North America	Christenhusz 6233 (H) [MH750087, MH750208, MH750138]
E. diffusum D.Don	East Asia	Husby 26 (FTG) [MH750050, MH750167, MH750118]
E. fluviatile L.	Panboreal	Christenhusz 6025 (H) [MH750051, MH750168, MH750119]
E. palustre L.	Panboreal	Christenhusz 6736 (K) [MH750072, MH750192, MH750129]
E. pratense Ehrh.	Panboreal	Väre 19931 (H) [MH750073, MH750193, MH750130]
E. sylvaticum L.	Panboreal	Fay 427 (K) [MH750083, MH750204, –]
E. telmateia Ehrh.	Western and southern Europe, West Asia	Christenhusz 6009 (H) [MH750090, MH750211, MH750140]
Subgenus Hippochaete (J.Milde) Baker
E. giganteum L.	South America, Central America, Caribbean	Husby 8 (FTG) [MH750054, MH750172, MH750121]
E. hyemale L.	Eurasia	McAllister s.n. (E) [MH750062, MH750180, MH750123]
E. laevigatum A.Br.	North America	Husby 15 (FTG) [MH750065, MH750183, MH750125]
E. myriochaetum Schltdl. & Cham.	Central America and Mexico	Husby 30 (FTG) [MH750070, MH750189, MH750127]
E. praealtum Raf.	North America, northeast Asia	Hauk 2011-001 (DEN) [MH750060, MH750178, MH750122]
E. ramosissimum var. huegelii (J.Milde) Christenh. & Husby	Tropical Asia, Pacific	Chase 24579 (K) [MH750077, MH750197, MH750131]
E. ramosissimum Desf. var. ramosissimum	Warm-temperate Europe, West Asia, East Africa	Christenhusz 5838 (H) [MH750079, MH750199, MH750133]
E. scirpoides Michx.	Panboreal	Long 13004 (H) [MH750082, MH750202, MH750134]
E. variegatum Schleich.	Panboreal	Husby 10 (FTG) [MH750092, MH750213, MH750142]
E. xylochaetum Mett.	Atacama (Chile)	Husby 1 (FTG) [MH750107, MH750228, –]
Outgroup (Marattiaceae)
Marattia alata Sw.	Jamaica, Cuba	Christenhusz 3266 (TUR) [–, EF463240, –]

This study uses molecular estimates of divergence dates, biogeographic approaches and flow cytometry to understand the major diversification events among extant Equisetum species and genome size divergence within the genus. Equisetum was previously divided in two subgenera, although recently this has increased to three (reviewed in Christenhusz et al., 2019). Although the genus is most diverse in the Northern Hemisphere, extant species extend into tropical South America and the Galapagos Islands, and one species occurs in tropical Africa, the Mascarenes and tropical Asia east to New Caledonia and Fiji (Hauke, 1963, 1978; Christenhusz et al., 2019). The genus is not now native to Australia or New Zealand, but Mesozoic fossils are known from Australasia (e.g. Gould, 1968; Pole and McLoughlin, 2017). The modern distribution of Equisetum has thus been suggested to be at least partly relictual (Christenhusz et al., 2019). However, Testo and Sundue (2016) found nearly equal and low rates of extinction and diversification for the family, which is similar to the rate found for many other fern families, and lower overall rates for both, so extinction may not have a great effect on biogeographic studies of the extant species. The fossil history of Equisetum (or Equisetaceae, depending on the generic treatment of fossil taxa) probably dates back to the Carboniferous (Good, 1971; Stewart and Rothwell, 1993; Taylor and Taylor, 1993; Elgorriaga et al., 2018; Clark et al., 2019), and because there has been little morphological change the genus can be considered a ‘living fossil’ (Arnold, 1947; Herrera-Flores et al., 2017). Indeed, it has been suggested to be the oldest genus of vascular plants (Vanneste et al., 2015), although this depends on taxonomic circumscription, which is a difficult issue.

Equisetum has been thought to be a member of the ‘sphenophytes’, Equisetopsida or Sphenopsida, which originated in the Devonian (Stewart and Rothwell, 1993; Taylor and Taylor, 1993; Clark et al., 2019). In the Carboniferous, arborescent genera like Calamites exhibited secondary growth and were an important element of the forests of that time. Modern classifications (e.g. Christenhusz and Chase, 2014) based on molecular phylogenetic analyses (e.g. Pryer et al., 2001; Testo and Sundue, 2016; Shen et al., 2018) have included Equisetum among the ferns, giving a slightly unexpected twist to the story of vascular plant evolution and raising questions about the relationships of ferns (including Equisetum) to some fossil sphenopsids (already questioned by Bower, 1908).

The first reliable insights into the cytology of Equisetum were made by Manton (1950), who analysed several species and reported that they all possessed a somatic chromosome number of 2n = 216. Because all meiotic studies exhibited 108 regular bivalents, a somatic chromosome count of 216 is considered to be cytologically diploid. Since then, several other studies have confirmed that 2n = 216 is widely conserved in the genus (Ninan, 1955; Mehra and Bir, 1959; Löve and Löve, 1961), with chromosome counts now available for all but the three giant species, E. giganteum, E. myriochaetum and E. xylochaetum (Rice et al., 2015). Indeed, the only reported variation from this is by Bennert et al. (2005), who identified several triploid populations of hybrid origin in the Rhine Valley with 2n = 3x = 324.

Consistent with observed subgeneric differences in chromosome size, investigations into the genome size diversity of Equisetum identified significant differences between the two subgenera recognized at that time (Obermayer et al., 2002; Clark et al., 2016). The mean 1C value (the total DNA amount in an unreplicated gametic cell) for five species in subgenus Hippochaete (1C = 25.9 pg) was observed to be approximately double that for three species in subgenus Equisetum (1C = 13.4 pg). Vanneste et al. (2015) looked at the incidence of polyploidy within extant Equisetum and found all to have the same ploidy. To complete phylogenetic coverage of all currently recognized species and subspecies, we measured genome sizes in 28 populations, comprising 18 species (one with a subspecies), including seven reported for the first time. In addition, by combining these data with those reported in previous publications (Grime et al., 1988; Obermayer et al., 2002; Bennert et al., 2005; Bainard et al., 2011; Bai et al., 2012; Clark et al., 2016), genome size changes are modelled on a new robust phylogenetic framework to provide further insights into the dynamics of genome size evolution in Equisetum. Although Clark et al. (2019) also studied genome size evolution, they relied upon values collated from the Plant DNA C-values Database (https://cvalues.science.kew.org/), many of which are based on unvouchered material and/or could be incorrectly identified. This is also the case for the DNA sequence data available in GenBank that was used by Clark et al. (2019) to create the phylogenetic framework for their study. In contrast, our study is underpinned by taxonomically well studied and documented accessions for all known extant Equisetum (Christenhusz et al., 2019).

Earlier phylogenetic studies (Des Marais et al., 2003; Guillon, 2004, 2007; Saslis-Lagoudakis et al., 2015) only studied subsets of these taxa, and most were focused on North American accessions. Here, we use a selected subset of the Christenhusz et al. (2019) accessions to provide a global representative phylogenetic framework for exploring biogeography, divergence estimates and genome size evolution of extant Equisetum. We have used fossils for calibration, but because these fossils are single data points their overall distribution at that time is unknown and therefore could not be used for biogeographic inference.

MATERIALS AND METHODS

Taxonomic sampling and DNA sequencing

The samples used are listed in Table 1 with associated voucher information and GenBank accession numbers for the DNA sequences of four plastid regions: rps4, the trnL intron, trnL-F spacer and rbcL. These are a subset of those used in Christenhusz et al. (2019), which were selected to represent all extant species in our molecular clock and biogeographic study. Previous papers in which DNA sequences have been analysed (e.g. Elgorriaga et al., 2018; Clark et al., 2019) relied mostly on data from GenBank largely produced by Guillon (2004, 2007) and Des Marais et al. (2003). Terminals with multiple plastid sequences in the earlier studies were often mixtures of accessions in GenBank generated by independent studies, some of which did not supply voucher information and possibly included misidentified samples. In contrast, here we have used a DNA sequence matrix with accessions of all species collected worldwide that are fully documented and reliably identified (Christenhusz et al., 2019).

Phylogenetic analyses

Three methods were used in our phylogenetic analyses: maximum parsimony (MP) searches were implemented in PAUP v.4.10b (Swofford, 2002), Bayesian inference (BI) in MrBayes 3.2.6 (Ronquist et al., 2012) and maximum likelihood (ML) in RAxML 8.2.3 (Stamatakis, 2014). Both BI and ML analyses were performed using the CIPRES cluster (Miller et al., 2010) and details of each are described in Christenhusz et al. (2019), including model selection and other optional procedures, which were the same for this reduced matrix. Marattia was selected as outgroup because it is one of the major subclades within the ferns and a potential sister clade of Equisetum (Pryer et al., 2001; Testo and Sundue, 2016). It should be noted that in several studies Equisetum is sister to the rest of the ferns (Shen et al., 2018), so the node at which Marattia and Equisetum diverge could then be considered the stem node of the ingroup. Nonetheless, we do not use that term for this node, which would include Equisetaceae, but perhaps also all sphenopsids. Clades with bootstrap percentages of 75–100 (both MP and ML) and posterior probabilities of 0.95–1.00 (MrBayes) are here considered supported.

Divergence time estimates using fossil calibrations

Despite the great number of known fossils of Equisetaceae, most are impression fossils revealing little internal structure. Of the few anatomically preserved fossils reported (Boureau, 1964; Brown, 1975), most cannot be attributed to any extant clade of Equisetum. Nevertheless, attempts have been made to match specific fossils to one subgenus; for instance, Gould (1968) compared Triassic and Jurassic Equisetum fossils (including E. laterale) with both subgenera Hippochaete and Equisetum, McIver and Basinger (1989) placed a Palaeocene fossil (E. fluviatoides) in subgenus Equisetum and Brown (1975) attributed Eocene E. clarnoi to subgenus Hippochaete. The Triassic fossil E. laterale from Australia was shown to share characters with subgenus Paramochaete, suggesting a possible Triassic link between Australian and South American Equisetaceae (Gould, 1968; Elgorriaga et al., 2018).

For the current work, we reviewed the literature on Equisetum fossils (Gould, 1968; Brown, 1975; Thomasson, 1980; Watson, 1983; McIver and Basinger, 1989; Falaschi et al., 2009; Stanich et al., 2009; Schwendemann et al., 2010; Channing et al., 2011; Contreras and Lutz, 2014; Elgorriaga et al., 2015, 2018; Clark et al., 2019) for comparison with extant material, and we selected those that could be used for the molecular clock analysis because they could be assigned to specific nodes (e.g. they shared characteristics with all or most species above that node, and thus were robust with respect to their identification and interpretation of relationships to extant species/subgenera of Equisetum, as described in the above-cited papers; Table 2). Our application of dated fossils to nodes in our phylogenetic tree also parallels their applications in Elgorriaga et al. (2015, 2018).

Table 2.

Fossils used for nine calibration nodes of the molecular clock. The four with dashes in the last three columns were not used because they are redundant with (and sometimes younger than) the ones used

Calibration node*	Fossil assigned to	Fossil (reference)	Location	Period	Age	OFFSET	LOGMEAN	LOGSTDV
21	All taxa	Calamostachys binneyana (Good, 1971)	USA, Kentucky, Lewis Creek and Shack Branch	Lower Pennsylvanean	323–359	323	3	1
22	All Equisetum	Equisetum thermale (Channing et al., 2011)	Argentina, Patagonia	Middle to Upper Jurassic	152–164	–	–	–
22	All Equisetum	Equisetites lyelii (Watson, 1983)	UK, Sussex	Lower Cretaceous	125–150	–	–	–
22	Subgenus Paramochaete	Equisetites minimus (Falaschi et al., 2009)	Argentina, Sta Cruz Prov.	Middle Jurassic	174–164	164	2	1
22	Subgenus Paramochaete	Equisetum laterale (Gould, 1968)	Australia, Queensland	Middle Jurassic	145–174	–	–	–
25	E. palustre, E. pratense	Equisetum ‘pratense’ (Zhang et al., 2007)	China, Yunnan	Late Tertiary, Miocene	5.3–23	5.3	0.5	1
27	E. telmateia, E. braunii	Equisetum maximum (Reed, 1971)	Upper Silesia (Poland)	Miocene	5–23	5	0.5	1
28	E. sylvaticum, E. diffusum, E. arvense, E. fluviatile	Equisetum fluviatoides (McIver and Basinger, 1989)	Canada, Saskatchewan	Palaeocene	56–66	56	1	1
31	Subgenus Hippochaete	Equisetum perlaevigatum (Reed 1971)	USA, Colorado	Late Cretaceous	66–72	66	1	1
32	E. variegatum, E. ramosissimum, E. praealtum, E. hyemale, E. giganteum, E. xylochaetum, E. laevigatum, E. myriochaetum	Equisetum clarnoi (Brown, 1975)	USA, Oregon	Middle Eocene	37.8–47.8	37.8	1	1
34	E. giganteum, E. xylochaetum, E. laevigatum, E. myriochaetum	Equisetum sp. (Thomasson, 1980)	USA, Nebraska	Late Tertiary	0–5.3	–	–	–
34	E. giganteum, E. xylochaetum, E. laevigatum, E. myriochaetum	Equisetites sp. (Contreras and Lutz, 2014)	Argentina, Formosa, Bermejo River	Lower Holocene	2.6–11.6	2.6	0.5	1
38	E. praealtum, E. hyemale	Equisetum hyemale (Reed, 1971)	Hungary	Upper Miocene	11.6–13.8	11.6	0.5	1

*Numbers correspond to nodes shown in Fig. 2

We provide a brief explanation here for some of our assignments of fossil ages (Table 2). Equisetum thermale (Channing et al., 2011) is an Equisetum without specificity and thus should be assigned to the stem node of Equisetum, but it cannot go to node 21 because this is the stem node of all Equisetaceae, so it must go to the next node in the tree, node 22, which is the crown node of Equisetum. We do not have a stem node for all Equisetum in this tree because this would involve including fossils in the phylogenetic analysis, which, because it is based on DNA data, is not possible. Because we re-ran the analysis removing each fossil calibration one by one (see below), we know that removing any single fossil has little effect. In fact, E. thermale is redundant with Equisetites minimus (Falaschi et al., 2009), which we did use, and with Equisetum laterale (Gould, 1968) and both are applied to the stem node of subgenus Paramochaete. Equisetum perlaevigatum is somehow related to the E. variegatum/laevigatum/hyemale group, excluding E. scirpoides, so we assigned this to node 31. Equisetum clarnoi (Brown, 1975) is in either the E. giganteum or E. hyemale subclade, excluding E. variegatum, so we placed it on the node of that larger group, node 32. We also included some dates based on some Mesozoic and Cenozoic fossils, which Reed (1971) suggested could refer to sets of specific extant taxa, and additional data from the Paleobiology Database (http://fossilworks.org) that were specific enough to use. Our most important criterion for selection was the degree to which the fossil exhibited enough morphological detail so that it could be attributed to taxa in our DNA tree. The phylogenetic position of E. bogotense did present some problems in the interpretation of the patterns of morphological evolution (Elgorriaga et al., 2015, 2018), but application of the fossil dates selected avoided these problems by being conservatively placed (i.e. they applied to most or all taxa above a given node).

Out of a selected set of 13 reliable fossils (Table 2), we used only nine (due to the redundancy of four, we used the oldest applicable fossils) as primary calibration points by applying a minimum age that corresponded to the upper period for which each fossil is assigned, and they were attributed to the stem nodes of their respective clades. Age calibration points correspond to the geological periods assigned to the date of each fossil and a lognormal prior distribution was applied (Ho and Phillips, 2009). Absolute divergence times were estimated using a Bayesian lognormal relaxed clock approach implemented in BEAST version 1.8.4 (Drummond and Rambaut, 2007), applying the birth–death process model after being selected based on the log marginal likelihood values in path and stepping-stone sampling methods using a chain length of 1 000 000 and 100 path steps. We ran the analyses by sampling solely from the priors afterwards. In most cases, the specified priors matched the effective prior distributions with data. To determine the influence of each calibration point, we ran BEAST repeatedly, excluding one fossil in each iteration (Supplementary Data Table S1). As expected, the main changes occurred in the nodes that were calibrated, and there were minor impacts (<5 % change) on the estimated dates for the other nodes of these trees. Two Monte Carlo Markov chains were run for 10⁸ generations and sampled every 10 000th generation. All parameters showed effective sample size values >200 in TRACER version 1.6.0. Post-burn-in trees were summarized into a maximum clade credibility tree with mean values and 95 % highest posterior density (HPD) confidence intervals for nodal ages using TREEANNOTATOR version 1.8.4. The selection of an uncorrelated relaxed clock rather than a strict clock was supported by the coefficient of variation of ρ = 1.2537 (95 % HPD 0.6282–1.6359), and there was no evidence of rate autocorrelation between neighbouring branches (e.g. marginal posterior probability of rate covariance: mean = −0.139; 95 % HPD interval −0.3737 to 0.1391).

Even though many fossil taxa used for calibration here are the same as those used by Elgorriaga et al. (2018) and Clark et al. (2019), our method is different from both. In Elgorriaga et al. (2018) the tree was dated by constraining it to a geological timeline based on the ages of the fossil terminals included. In contrast, here we assign the calibration point based on fossil ages as a constraint on the node deeper than where it is placed morphologically, except for E. thermale (all Equisetum, node 22). Clark et al. (2019) used estimates from a molecular species divergence analysis as priors on nodes and fossil tip ages from a uniform distribution across the geographical distribution of the species and a correlated rates clock model (Thorne and Kishino, 2002) across the phylogenetic tree, which constrained node ages to a greater degree than our method and that of Elgorriaga et al. (2018). Clark et al. (2019) applied the fossil data to calibrate tips except for that of the basal node of Equisetum, whereas we assigned fossils to nodes based on morphological traits and then calibrated the corresponding node to include all extant and extinct taxa (Magallón and Sanderson, 2001). We prefer our method because (1) it is less constrained and (2) it includes all extant and extinct taxa that exhibit the characteristic traits.

Ancestral range reconstruction analyses

Five operational areas were defined representing the current ranges of extant Equisetum taxa and palaeogeographic history (Buerki et al., 2011): A, North America; B, Central and South America; C, Europe and Asia; D, Madagascar, India and Southeast Asia; and E, Australasia and Oceania. The likelihood dispersal–extinction–cladogenesis (DEC) model implemented by Lagrange version 20130526 (Ree et al., 2005; Ree and Smith, 2008) was used to estimate global extinction and dispersal rates and reconstruct the ancestral range of each node using the maximum clade credibility tree from BEAST. All extant taxa were assigned to a maximum of two operational areas, and therefore the maximum number of areas in the analysis was set to two. We assessed two alternative dispersal models, an unconstrained dispersal model (M0) in which dispersal had the same probability among operational areas through time, and a constrained model (M1) in which dispersal probabilities varied through time in four time slices: Early Miocene to Present (34–0 Mya), Early Eocene to Early Oligocene (55–34 Mya), Late Cretaceous to Early Eocene/Ypresian (55–90 Mya) and Jurassic to Late Cretaceous (201–90 Mya). Operational area connectivity in each time slice varies (Viruel et al., 2016, a procedure modified from that in Buerki et al., 2011; Supplementary Data Table S2).

Genome size estimation using flow cytometry

Plants used for genome size estimation were obtained from a variety of sources (Table 3), often from material in cultivation with specific provenance data. Genome sizes were estimated using flow cytometry based on the methods outlined in Pellicer et al. (2012). To examine genome size diversity and evolution within a phylogenetic framework, a mean value for each species was calculated (Table 3). We reconstructed the ancestral genome sizes using a Brownian motion model throughout the maximum credibility tree of Equisetum using functions ace and fastAnc and plotting them onto the phylogenetic tree with the function contMap of the library phytools (Revell, 2012) in R (R Core Team, 2019). A Brownian motion model is appropriate because there is no change in ploidy within extant Equisetum (Vanneste et al., 2015), and we excluded the genome size of the outgroup because between Marattia and extant species of Equisetum there is at least one whole-genome duplication (Vanneste et al., 2015).

Table 3.

Genome size data available for Equisetum species including new estimates reported in this paper and those from the literature. Details of plant material (where available) are also given. N.B. All genome size estimates listed in the table were determined by flow cytometry (using propidium iodide as the fluorochrome), except for the three estimates reported by Grime et al. (1988), who used Feulgen microdensitometry

Species	Voucher (herbarium)	Country of origin	1C DNA amount (pg)	2C DNA amount (pg)	SD	Calibration standard*	Source
Subgenus Paramochaete
E. bogotense	Smrzova 2005.00915 (H)	Peru	21.13	42.27	0.18	Allium 1	This paper
E. bogotense	Husby 28 (FTG)	Chile	21.28	42.56	0.37	Allium 1	This paper
		Mean	21.05
Subgenus Equisetum
E. pratense	Väre 19931 (H)	Finland	11.9	23.8	0.17	Allium 1	This paper
E. palustre	Acock 17 (-)	UK	15.48	30.96	0.04	Allium 1	This paper
E. palustre	Acock s.n. (-)	UK	15.47	30.94	0.11	Pisum 4	This paper
E. palustre	Basil s.n. (-)	UK	15.3	30.6	0.5	Allium 1	This paper
E. palustre	Christenhusz 6736 (K)	UK	16.37	32.74	0.37	Pisum 4	This paper
E. palustre	Christenhusz 6793 (K)	UK	15.65	31.3	0.25	Allium 1	This paper
E. palustre	Christenhusz 6866 (PG)	UK	15.01	30.02	0.49	Pisum 4	This paper
E. palustre	Christenhusz 7128 (K)	Netherlands	14.69	29.38	0.1	Pisum 4	This paper
E. palustre	Fay 09-07-2015 (K)	UK	14.75	29.5	0.25	Pisum 4	This paper
E. palustre	Bai et al. (WIS)	USA (WI)	14.2	–	–	Pisum 3	Bai et al. (2012)
		Mean	15.41
E. telmateia	Christenhusz 6724 (H)	Spain	13.71	27.42	0.46	Pisum 5	This paper
E. telmateia	Christenhusz 6792 (K)	UK	14.31	28.62	0.21	Pisum 5	This paper
E. telmateia		Mean	14.01
E. braunii	UConn 200202437		13.17	26.34	0	Allium 1	This paper
E. braunii	UConn 1998800141		13.1	26.2	0.09	Allium 1	This paper
		Mean	13.14
E. sylvaticum	Väre 19923 (H)	Finland	12.37	24.74	0.12	Allium 1	This paper
E. sylvaticum	Fay s.n. (K)	UK	13.04	26.07	0.3	Pisum 4	This paper
E. sylvaticum	Bainard et al. (OAC)	Canada	12.89	25.78	0.54	Allium 2	Bainard et al. (2011)
E. sylvaticum	Bai et al. (WIS)	USA (WI)	12.7	–	–	Pisum 4	Bai et al. (2012)
		Mean	12.65
E. diffusum	Sekerka & Chvosta 2008.10498 (H)	China	13.83	27.76	0.2	Allium 1	This paper
E. fluviatile	Christenhusz 6025 (H)	Finland	13.95	27.9	0.03	Allium 1	This paper
E. fluviatile	No data		13.5	–	–	Allium 1	Grime et al. (1988)
		Mean	13.73
E. arvense	Christenhusz 6735 (K)	Finland	13.79	27.58	0.78	Allium 1	This paper
E. arvense	No data		14.2	–	–	Allium 1	Grime et al. (1988)
E. arvense	Bainard et al. (OAC)	Canada	14.65	29.29	0.25	Pisum 5	Bainard et al. (2011)
E. arvense	Bai et al. (WIS)	USA (WI)	13.7	–	–	Pisum 4	Bai et al. (2012)
		Mean	14.09
Subgenus Hippochaete
E. scirpoides	RBG, Kew Acc 1934–97905 (K)		21.25	42.5	0.04	Allium 1	Obermayer et al. (2002)
E. scirpoides	Lubienski 41 (BOCH)	Sweden	21.1	42.2	0.05	Allium 1	Bennert et al. (2005)
E. scirpoides	Lubienski 66 (BOCH)	Norway	21.7	43.4	0.62	Allium 1	Bennert et al. (2005)
E. scirpoides	Bai et al. (WIS)	USA (WI)	21.8	–	–	Vicia 3	Bai et al. (2012)
E. scirpoides	Firsov s.n. (H, K)	Russia (Kamchatka)	21.46	42.92	0.64	Allium 1	This paper
		Mean	21.46
E. variegatum	RBG, Kew Acc. No. 1977-882 (K)		30.35	60.7	0.19	Allium 1	Obermayer et al. (2002)
E. variegatum	Lubienski 36 (BOCH)	Germany	31.1	62.2	0.7	Allium 1	Bennert et al. (2005)
E. variegatum	Lubienski 63 (BOCH)	Norway	32.6	65.1	0.27	Allium 1	Bennert et al. (2005)
E. variegatum	Lubienski SP 73/93 (BOCH)	Switzerland	31.3	62.6	0.19	Allium 1	Bennert et al. (2005)
		Mean	31.34
E. ramosissimum var. huegelii	RBG, Kew Acc. No 1974-1881 (K)	Sri Lanka	26.2	52.5	0.24	Allium 1	Obermayer et al. (2002)
E. ramosissimum var. ramosissimum	Christenhusz 6700 (BEI, K, H)	Lebanon	26.71	53.42	0.06	Allium 1	This paper
E. ramosissimum var. ramosissimum	Lubienski 33 (BOCH)	Germany	28.2	56.4	0.06	Allium 1	Bennert et al. (2005)
E. ramosissimum var. ramosissimum	Lubienski 35 (BOCH)	France	28.1	56.1	0.05	Allium 1	Bennert et al. (2005)
E. ramosissimum var. ramosissimum	Lubienski 91 (BOCH)	France	28.3	56.5	0.36	Allium 1	Bennert et al. (2005)
		Mean	27.83
E. hyemale	Lubienski 28 (BOCH)	Germany	26.3	52.5	0.71	Allium 1	Bennert et al. (2005)
E. hyemale	Lubienski 29 (BOCH)	Germany	26.7	53.3	0	Allium 1	Bennert et al. (2005)
E. hyemale	Lubienski 70 (BOCH)	Germany	26.2	52.4	0.28	Allium 1	Bennert et al. (2005)
		Mean	26.4
E. praealtum	Christenhusz 6232 (H)	USA (CA)	26.73	53.46	0.2	Allium 1	This paper
E. praealtum	Bai et al. (WIS)	USA (MI)	26.5	–	–	Vicia 3	Bai et al. (2012)
E. praealtum	Bainard et al. (OAC)	Canada	26.9	53.8	0.605	Pisum 5	Bainard et al. (2011)
		Mean	26.71
E. laevigatum	Christenhusz 6161 (H)	USA (CA)	25.75	51.49	0.39	Allium 1	This paper
E. laevigatum	Bai et al. (WIS)	USA (MI)	25.7	–	–	Pisum 4	Bai et al. (2012)
		Mean	25.73
E. myriochaetum	RBG, Kew Acc. No. 1994–3384 (K)		25.65	51.3	0.4	Allium 1	Obermayer et al. (2002)
E. giganteum	Husby 3 (FTG)	Peru	23.4	46.8	0.08	Allium 1	This paper
E. giganteum	Husby 20 (FTG)	Argentina	23.42	46.84	0.06	Allium 1	This paper
E. giganteum	MBC 20100220A (FTG)	Peru	23.59	47.18	0.07	Allium 1	This paper
		Mean	23.47
E. xylochaetum	UConn 200400108 = Husby 1 (FTG)	Chile	23.59	47.18	0.16	Allium 1	This paper

BOCH, Herbarium, Fakultät für Biologie, Ruhr-Universität Bochum, Germany; FTG, Herbarium, Fairchild Tropical Botanic Garden, Florida, USA; H, Herbarium, Finnish Museum of Natural History, Helsinki University, Finland; K, Herbarium, Royal Botanic Gardens, Kew, UK; MBC,– Montgomery Botanical Centre, Florida, USA; OAC, Bio Herbarium, University of Guelph, Canada; RBG, Kew, Living Collections, Royal Botanic Gardens, Kew, UK; WIS, Herbarium, University of Wisconsin, Madison WI, USA; UConn, Department of Ecology & Evolutionary Biology, University of Connecticut, USA.

*Calibration standards used and their assumed 2C-values for converting relative units into absolute DNA amounts: 1Allium cepa ‘Ailsa Craig’ 2C = 33.5 pg; 2Allium cepa ‘Alice’ 2C = 34.89 pg; 3Vicia faba ‘Bell Bean’ 2C = 26.53 pg; 4Pisum sativum ‘Dwarf Gray Sugar’ 2C = 8.92 pg; 5Pisum sativum ‘Citrad’ 2C = 9.09 pg.

RESULTS

Phylogenetic analyses

The model of molecular evolution and phylogenetic relationships (Fig. 1) between the taxa analysed here are the same as those found by Christenhusz et al. (2019). It might be expected that, with the reduced sampling, relationships and model selection might have been differently estimated, but this is not the case here.

Fig. 1.

Bayesian tree for Equisetum constructed with combined data from four plastid regions (rbcL, trnL intron, trnL-F intergenic spacer, rps4) with an accession of each accepted taxon in extant Equisetum. Support is shown above branches: Bayesian posterior probabilities/bootstrap percentages from maximum likelihood analyses/bootstrap percentages from maximum parsimony analyses.

Molecular clock analysis using fossil calibrations

The calibrations applied had differing effects on the results obtained (Supplementary Data Fig. S1). For nodes 21 and 22, a lack of prior calibration would result in wide age distributions, with ranges of 184.0–1875.4 and 107.7–347.3 Mya (95 % HPD intervals; Supplementary Data Fig. S1), respectively. For other nodes, fossil calibrations returned older ages than those estimated without prior information (nodes D, E, F, G). According to BEAST analyses (Fig. 2), divergence between our outgroup Marattia and ingroup Equisetum probably occurred during the Viséan age (Middle Mississippian of the Early Carboniferous) ~342 Mya (95 % HPD 323.6–426.8 Mya), whereas the age reconstructed for the node of all extant Equisetum was 175 (164.6–209.7) Mya during the Toarcian of the Early Jurassic, concomitant with the divergence of the ancestor of E. bogotense (subgenus Paramochaete) from the main clade of Equisetum. We cannot date the crown age of Equisetum because to do this requires inclusion of fossils, and our data are only DNA sequences. The subsequent split between the other two subgenera, Equisetum and Hippochaete, probably occurred 135 (95.8–177.0) Mya in the Valanginian (Early Cretaceous), whereas the crown nodes of each subgenus are dated to the Late Cretaceous: 89 (62.8–128.1) Mya for Equisetum and 72 (66.2–87.3) Mya for Hippochaete. Equisetum subgenus Equisetum recovered older splits among its main clades than those in subgenus Hippochaete, with two splits before the Eocene in Equisetum, whereas the main splits in Hippochaete occurred during the Eocene to Miocene. Nevertheless, most divergences among extant Equisetum taxa likely occurred relatively recently, in the Middle to Late Miocene and into the Pliocene.

Fig. 2.

BEAST maximum clade credibility tree of Equisetum based on four plastid regions, showing mean ages and 95 % HPD intervals for node ages (grey bars in Myr). Node numbers are indicated below branches in italics. Vertical dashed grey lines separate the four time slices (TSI to TSIV) used in the stratified ancestral range reconstruction DEC analysis (M1). Branch colours represent the estimated ancestral genome sizes (GS, in pg/1C value; Supplementary Data Table S3) reconstructed using Phytools based on the mean 1C values obtained from extant species, which range from 11.90 to 31.34 pg/1C (Table 3) as shown in the legend. We also provide the estimated genome sizes for the MRCA of all extant Equisetum species and those for the MRCA of subgenera Equisetum and Hippochaete (pg/1C value; Supplementary Data Table S3).

Ancestral range reconstruction analyses

The two biogeographical DEC models (M0 and M1) implemented in Lagrange gave highly similar results, although M1 (Fig. 3) produced a better –ln likelihood (41.92) than M0 (47.10; Supplementary Data Fig. S2). Nevertheless, although the dispersal rate differed considerably between the two models (M0, 0.01125; M1, 0.03538), the extinction rate was similar (M0, 0.001655; M1, 0.00161). Both models were consistent in predicting a long-term persistence in Laurasia (corresponding to extant operational areas A and C) throughout the evolution of Equisetum, with an early split into Gondwana (extant operational area B) during the Jurassic. However, the biogeographic scenario that explains the divergence of E. bogotense differs between models. The unconstrained M0 model reconstructed an ancestral area AC for the ancestor of Equisetum, with the origin of subgenus Paramochaete explained by a dispersal from Laurasia (AC) to Gondwana (B) (Supplementary Data Fig. S2). In contrast, M1 reconstructed a higher probability of an ancestry for Equisetum in BC (i.e. Gondwana + Eurasia), with a vicariance event predicted to explain the subsequent divergence of E. bogotense. This uncertainty in both models for reconstructing the ancestral distribution of Equisetum could hint at an expanded distribution of the common ancestor of Equisetum throughout the still connected post-Pangaean subcontinents during the Jurassic. Dispersals from Eurasia (C) to India (D) and Australasia (E) and a second colonization of the Neotropics (B) from North America (A) were inferred by both models, with the frequency of these long-distance dispersal events inferred to be higher in subgenus Hippochaete than subgenus Equisetum.

Fig. 3.

Spatiotemporal evolution of Equisetum as inferred by the stratified DEC model (M1) (results for mode M0 are shown in Supplementary Data Fig. S2). Pie charts at nodes indicate the relative probabilities of ancestral ranges. Vertical dashed grey lines separate the four time slices (TSI to TSIV). Inset maps represent the palaeogeographic configuration of land masses at each time slice, with the main inferred dispersal events represented with arrows. Inferred dispersal (x→y), vicariance (x/x) and peripheral isolate speciation (x\xy) events are represented. RL, relative likelihood.

Genome size analysis of Equisetum

Analysis of Equisetum by flow cytometry resulted in high-resolution flow histograms (e.g. see Supplementary Data Fig. S3), with the 2C peaks of both the internal standard and Equisetum samples having low coefficients of variation (1.5–5.4 %, mean 3.0 %). For two species, E. giganteum and E. bogotense, this study analysed two geographically well-separated populations, one from a tropical habitat in Peru and the other from a more temperate habitat in Argentina (E. giganteum) and Chile (E. bogotense), but there were only small differences in the estimates (Table 3). In contrast, our 1C values for four populations of E. palustre L. (three of which are from the southern UK) varied from 15.30 to 16.37 pg, and, if the data from Bai et al. (2012) were also included (1C = 14.20 pg), this species infraspecifically exhibited 1.15-fold variation in genome size.

Overall genome sizes in Equisetum range 2.7-fold from 11.90 pg/1C in E. pratense to 32.6 pg/1C in one accession of E. variegatum (Table 3). Comparing genome sizes of subgenera Equisetum and Hippochaete showed distinct differences in their profiles, with subgenus Hippochaete (21.46–31.34 pg/1C; mean 25.80 pg/1C) having a mean genome size that was nearly double that of subgenus Equisetum (11.90–15.41 pg/1C; mean 13.60 pg/1C). The intermediate genome size estimate for the ancestor of both subgenera (node 23, 19.20 pg/1C; 14.98–23.42) is similar to that inferred for the crown node of Equisetum (node 22, 19.54 pg/1C; 14.18–24.91). The ancestral genome sizes reconstructed for each node shown in Fig. 2 are given in Supplementary Data Table S3.

DISCUSSION

Origin, age and diversification of Equisetum

Our molecular clock estimates of divergence place the origin of the Equisetum lineage in the Early Carboniferous (~342 Mya). This age estimate almost certainly represents Equisetaceae, not just Equisetum, but in order to identify the line of Equisetum as distinct from the other genera of Equisetaceae we would need to include fossils. Regardless of this circumscription issue (see below), Equisetum is clearly one of the most evolutionarily isolated and possibly the oldest extant vascular plant genus (Husby, 2013; Vanneste et al., 2015). We do not use the term ‘stem node’ here for the Marattia/Equisetum divergence because we do not evaluate which portion of the fern clade is sister to Equisetum. Among published analyses to date, Equisetum has been inconsistent in position, making any member of the fern clade an effective outgroup in our analyses. However, given that Equisetum has been shown to be sister to all other ferns in some analyses (e.g. Testo and Sundue, 2016; Shen et al., 2018), this split is similar to if not coincident with the stem node of Equisetum/Equisetaceae. Because we are not focusing on the topology within the fern clade and have used only one of the possible outgroups, we cannot claim that the Marattia/Equisetum divergence is the stem node. In any case, this node and its significance is not a major focus of this paper and estimating the stem node age is best left to studies that sample the complete set of fern clades and include fossils.

Elgorriaga et al. (2018), who used morphology and the ages of fossils to establish a timeline for Equisetum and related fossil taxa, estimated a Middle Carboniferous (Mississippian) split of the clade that includes Equisetum (= Equisetaceae + Neocalamitaceae) from Calamitaceae. Fossil Neocalamitaceae resemble Equisetum in most characters apart from having secondary growth like Calamitaceae (Bell et al., 1979; Elgorriaga et al., 2018), and therefore with a minor adjustment to the circumscription Equisetaceae could include fossil taxa such as Equisetites, Neocalamostachys, Neocalamites and Spaciinodum. Admittedly, the definition of a human construct like a genus is debatable (see Schuettpelz et al., 2018 versus Christenhusz and Chase, 2018). Certainly, at a deep time level it is a difficult concept to agree upon, but it is indisputable that extant Equisetum species represent living members of an ancient lineage with the same morphology. Whereas placement of Equisetites and Spaciinodum in Equisetum is still a matter of opinion rather than science, they are clearly morphologically similar (Osborn et al., 2000; Schwendemann et al., 2010; Clark et al., 2019). Generic circumscription of Equisetum is not the focus here, and our estimated divergence times fit well into the timeline established with fossils by Elgorriaga et al. (2018).

Diversifying during the Triassic and Jurassic, Equisetum comprises three extant clades recognized as morphologically well-defined subgenera by Christenhusz et al. (2019). Our age estimates indicate that Equisetum subgenus Paramochaete diverged from the rest of Equisetum 175 Mya, during the Middle Jurassic, which was followed in the Cretaceous (~90 Mya) by continuation of the Andean uplift in northern South America (Orme, 2007), where the subgenus still occurs. Its extant distribution also includes isolated mountain peaks in Costa Rica and the Galapagos Islands, showing its potential for long-distance dispersal (Christenhusz and Chase, 2013). The close similarity between extant E. bogotense and the Australian Mesozoic fossil E. laterale (Gould, 1968) raises the possibility that species belonging to subgenus Paramochaete occur in South America through Gondwanan vicariance. Although such fossils could be argued to indicate that extant Equisetum emerged in the Southern Hemisphere, fossils are difficult to employ in biogeographic studies. If the distributions of fossil species are known, as in Wood et al. (2013), then fossils can be usefully employed in understanding biogeographic patterns, but if a fossil is only known from a single locality, as is the case here for E. laterale, its use is precluded except as a calibration point.

Our results closely mirror those of Elgorriaga et al. (2018), who predicted that extant Equisetum diversified in the early Jurassic. In contrast, our results differ from the mid-Cretaceous crown age of Equisetum estimated by Clark et al. (2019), who used a strict molecular clock and assumptions different from those used here to estimate node ages (see the Materials and methods section). We suggest our results are more robust because our data fit better with a relaxed molecular clock and align better with the fossil timeline (Elgorriaga et al., 2018).

Subgenera Equisetum and Hippochaete are estimated to have diverged 135 Mya, with Equisetum radiating ~89 Mya and Hippochaete initiating radiation slightly later at ~72 Mya. Nevertheless, since error margins of these estimates overlap, we can only be confident that these radiations occurred during the Late Cretaceous. Elgorriaga et al. (2018), using a different method of establishing the ages of the two subgenera, came to the same relative sequence of radiations, but placed the radiations of subgenera Equisetum and Hippochaete slightly more recently (at the very end of the Cretaceous and into the Eocene, respectively). Thus, overall, these two studies using different methods have arrived at roughly congruent results, whereas Clark et al. (2019) found more recent ages, which is not supported by the fossil record of the genus. Our molecular age estimates are well corroborated by fossil ages, contrasting markedly with molecular estimates of ages in angiosperms, for which nearly all studies (reviewed in Li et al., 2019) have indicated an origin long before the first unequivocal angiosperm fossils, representing the so-called ‘Jurassic gap’ of Li et al. (2019).

Biogeography of Equisetum

Figure 3 illustrates the possible migration routes of Equisetum over time. It suggests that the genus may have occurred across Pangaea during the Early Jurassic, with the first divergence ~175 Mya coinciding with the breakup of the Pangaean supercontinent, after which climates became warmer and more humid. This divergence separated subgenus Paramochaete in Gondwana (with South American E. bogotense and the associated Australian fossil E. laterale) from the remainder of the genus, which persisted in Laurasia. The Laurasian clade subsequently split into subgenera Equisetum and Hippochaete during the Early Cretaceous ~135 Mya, after which each subgenus diversified independently. Subgenus Equisetum diversified in temperate regions partly through vicariance, with a predicted long-distance dispersal to East Asia (where E. diffusum probably evolved), and another long-distance dispersal during the Miocene separating European–Mediterranean E. telmateia from E. braunii in western North America.

The biogeographic history of subgenus Hippochaete appears to be a little more complicated. It includes a grade with a few Arctic species (E. scirpoides, E. variegatum) that diverged independently in the Late Cretaceous–Palaeocene, following the evident cooling trend in the northern continental islands that then made up Laurasia (Williams et al., 2009). In the Miocene this was followed by migration from North to South America in which two scenarios are possible: (1) migration to South America once with a re-migration of E. laevigatum to North America, or (2) diversification to form a clade that gave rise to the giant horsetail clade, with E. giganteum and E. xylochaetum migrating to South America, followed, more recently, by E. myriochaetum also dispersing south to tropical Central America and into northern South America, while E. laevigatum persisted in North America.

The second clade within subgenus Hippochaete comprises E. hyemale, E. praealtum and E. ramosissimum. In contrast to the giant horsetail, E. ramosissimum shows much more recent dispersal events into Africa, India and Sri Lanka, across Malesia and the Philippines, and into the western Pacific Islands to New Caledonia and Fiji. Unexpectedly, none of the species in this clade has yet to get a foothold in Australia or New Zealand, even though suitable climates exist. Equisetum ramosissimum is subdivided into two varieties that are not yet morphologically completely distinct, with frequent intermediate forms arising where the two varieties meet (Zhang and Turland, 2013; Christenhusz et al., 2019).

The pattern of evolution in Equisetum presented here is contrary to the hypothesis proposed by Hauke (1963, 1978) that the giant species of South America (i.e. E. giganteum and E. myriochaetum) must be an ancestral lineage of Equisetum due to their similarity in size to giant fossil forms, and hence he proposed that the genus evolved in Gondwana and subsequently migrated to the Northern Hemisphere. As noted above, our results instead indicate that the giant horsetails evolved in South America from lineages that arrived from the north after the closure of the Isthmus of Tehuantepec. Therefore, the traditional scenario of a southern origin of the giant horsetails needs to be reconsidered.

In Elgorriaga et al. (2018) there is a grade of Southern Hemisphere fossils leading up to the extant subgenera, which are predominantly from the Northern Hemisphere. If these are endemic solely to the Southern Hemisphere and actually related to E. bogotense, then perhaps our findings are spurious, but we cannot be certain that these fossil species were in fact not more broadly distributed. Therefore, we cannot know that these are definitely related to E. bogotense, given the observed convergence of E. giganteum and the giant fossil species noted above. The Australian fossils of E. laterale attributed to subgenus Paramochaete show that that lineage is probably of Gondwanan origin and persisted in South America, but the lack of any other extant species of subgenus Paramochaete outside South America suggests that these died out in other Gondwanan continents, leaving few fossils behind.

Genome size diversity and evolution in Equisetum

From the genome size data presented in this paper together with additional C values in the literature, we now have complete taxonomic coverage for the genus (Table 3, Fig. 2). Comparing our data with previously published measurements (Table 3) shows a general agreement between studies, with small differences most likely arising from technical issues, such as different 2C values assumed for calibration standards (e.g. E. sylvaticum), use of different calibration standards (e.g. E. praealtum and E. sylvaticum) and/or employment of different techniques not following best practice recommendations (i.e. flow cytometry versus Feulgen microdensitometry, as in E. fluviatile and E. arvense; reviewed in Greilhuber and Temsch, 2001; Doležel et al., 2007; Suda and Leitch, 2010).

Superimposing the mean genome size data for each species onto the well-supported phylogenetic tree for Equisetum confirms that despite the consistency in chromosome number, the contrasting chromosome and genome sizes between the two subgenera noted for a limited number of species by Manton (1950), Obermayer et al. (2002) and Clark et al. (2019) holds across the whole genus (mean 1C values for subgenus Equisetum and Hippochaete are 15.62 and 23.56 pg, respectively; Supplementary Data Table S3). The somewhat intermediate genome size for E. bogotense (mean 1C value 21.05 pg), which is sister to the rest of the genus (Fig. 2), makes inferences of the direction of genome size changes in the genus problematic, although some insights are suggested from the reconstruction of ancestral genome size analyses based on data for all extant species. An ancestral 1C value of 19.20 pg/1C for the most recent common ancestor (MRCA) of subgenera Equisetum and Hippochaete followed by ancestral 1C values of 15.62 and 23.56 pg/1C in the ancestors of each subgenus, respectively (Supplementary Data Table S3), suggest contrasting trends for each clade.

In subgenus Equisetum, genome size appears to be fairly conserved, with values ranging just 1.30-fold (mean 1C values 11.90–15.41 pg) and an overall trend towards genome contraction with respect to 15.62 pg/1C (node 24, Fig. 2, Supplementary Data Table S3) predicted for the subgenus Equisetum MRCA. In contrast, genome sizes are slightly more variable in subgenus Hippochaete, with 1C values varying 1.50-fold (mean 1C values 21.1–31.34 pg), and they have clearly undergone moderate genomic expansions relative to the predicted ancestral genome size of the MRCA of both these subgenera (i.e. 19.20 pg/1C; node 23, Fig. 2, Supplementary Data Table S3). Despite contrasting approaches and datasets, it is notable that there is overall similarity in the size of the reconstructed 1C values across Equisetum and the trends in genome size change between the results presented here and those of Clark et al. (2019). Vanneste et al. (2015) found no evidence of ploidal change within Equisetum, but their estimated age of this last polyploid event was 92 Mya, which is too recent according to our dating as well as that of others for the crown age of Equisetum. Clark et al. (2019) noted that although there was evidence of two whole-genome duplication events prior to the origin of Equisetaceae (dated to have occurred in the Middle to Late Carboniferous, 329–307 Mya, and between the Late Permian and Late Triassic, 253–233 Mya), they also did not detect evidence of more recent whole-genome duplication events. Thus, the predicted expansion and contraction of genome sizes within subgenera Hippochaete and Equisetum, respectively, in their recent evolutionary history has most likely been driven by changes in the abundance of non-coding repetitive DNA sequences, especially given the constancy in chromosome number of 2n = 216 reported for all extant species. However, the lack of complete-genome sequence data for any species of Equisetum precludes insights into the repeat dynamics underpinning the contrasting genome size profiles between the two main subgenera.

The reasons behind the striking differences in genome size between subgenera are still unclear, but one possible explanation may lie in the nucleotypic consequences of genome size variation on the mobility and size of the flagellate spermatozoids in Equisetum, which must swim from the archegonia of the female gametophyte to the antheridium of the male to effect fertilization. Several studies (Greilhuber and Leitch, 2013; Roddy et al., 2020) have shown that genome size can affect the phenotype of an organism independently of the genotype, a concept that has been termed the nucleotype (Bennett, 1972). Nucleotypic effects have been shown to operate at all levels from the nucleus (e.g. the durations of mitosis and meiosis are both positively correlated with genome size; Bennett, 1972) up to the whole-plant level, at which genome size can play a role in determining where and when a plant lives (reviewed in Greilhuber and Leitch, 2013 and Pellicer et al., 2018). In relation to spermatozoids, which are biflagellate in mosses and some lycopods and multiflagellate in ferns and two groups of gymnosperms (cycads and Ginkgo), genome size may be under selection due to its effect on sperm size, mobility, and hence ability to effect fertilization. This hypothesis arises because the nucleus contributes most of the cell mass in spermatozoids, and there is consequently a reasonable correlation between genome size and spermatozoid cell size (Renzaglia et al., 1995). Since the mobility of spermatozoids depends on their size, it could be envisaged that there might be selection against accumulating extra DNA because this would result in larger spermatozoa, making them less mobile and hence less efficient in fertilization. Renzaglia et al. (1995) were the first to suggest this as an explanation for the narrow range of small genome sizes encountered in mosses and lycopods that possess biflagellate spermatozoa. Nevertheless, it is also possible that in species with multiflagellate spermatozoa, those with fewer flagella may be under greater selection for smaller genome sizes than those with more flagella, given the predicted impact of a larger genome on cell size and spermatozoal mobility. Although we are fully aware that the data are limited in this context, it is noteworthy that the only study focusing on flagella number in three species of Equisetum (E. arvense, mean 1C 14.09 pg and E. telmateia, mean 1C 13.71 pg, in subgenus Equisetum; and E. hyemale, mean 1C 26.40 pg, in subgenus Hippochaete) shows that the species with fewer flagella are E. telmateia and E. arvense with ~40 and 56 flagella, respectively, compared with ~80 found in E. hyemale (Duckett and Bell, 1977; Renzaglia et al., 2002), which in turn also have smaller genomes. This correlation is somewhat supportive of our hypothesis. If this assumption holds true and is a general pattern in the genus, it would be tempting to speculate that one of the reasons for the larger genome sizes in subgenus Hippochaete may be relaxed selection due to their greater number of flagella. Notwithstanding, it should also be noted that the smaller genomes are found in deciduous subgenus Equisetum, whereas larger genome sizes are generally found in the (semi-)evergreen subgenus Hippochaete.

Conclusions

This study provides new perspectives on the evolution of horsetails. With a calculated age of 342 Mya, we place the origin of the ancestor of the extant members of Equisetaceae in the Early Carboniferous, with Equisetum evolving at some point after that, probably soon, but definitely before 175 Mya, making it possibly the oldest extant genus of vascular plants. We show that the ancestors of the three clades now treated as subgenera (Christenhusz et al., 2019) diverged successively between the Middle Jurassic and Early Cretaceous (175–135 Mya). Their biogeographical history is a combination of several long-distance dispersal events and vicariance. Geographical models are consistent with long-term persistence of the genus Equisetum in Laurasia. The larger mean genome size of subgenus Hippochaete is hypothesized to be related to more numerous sperm flagellae and most likely arose through increases in the non-coding repetitive DNA sequences rather than ancient whole-genome duplication events. Many questions regarding the evolution of horsetails and vascular plants remain, but this study aids our understanding of this ancient but dynamic living fossil genus.

SUPPLEMENTARY DATA

Supplementary data are available online at https://academic.oup.com/aob and consist of the following. Figure S1: density plots showing the marginal probability distribution of the data obtained in our analysis, without data and without calibration for each of the nodes where a calibration was applied. Figure S2: spatiotemporal evolution of Equisetum as inferred by the M0 DEC model. Figure S3: flow histograms of Equisetum laevigatum and E. pratensis run with Allium cepa as the internal calibration standard. Table S1: mean and height 95 % HPD values obtained in BEAST with all fossils and excluding one each time. Table S2: dispersal rate matrices reflecting the palaeogeographic connectivity among the study areas in each historical scenario. Table S3: genome size estimates for each node in Fig. 2.