Convergent Evolution of Fern-Specific Mitochondrial Group II Intron atp1i361g2 and Its Ancient Source Paralogue rps3i249g2 and Independent Losses of Intron and RNA Editing among Pteridaceae.

Mitochondrial intron patterns are highly divergent between the major land plant clades. An intron in the atp1 gene, atp1i361g2, is an example for a group II intron specific to monilophytes (ferns). Here, we report that atp1i361g2 is lost independently at least 4 times in the fern family Pteridaceae. Such plant organelle intron losses have previously been found to be accompanied by loss of RNA editing sites in the flanking exon regions as a consequence of genomic recombination of mature cDNA. Instead, we now observe that RNA editing events in both directions of pyrimidine exchange (C-to-U and U-to-C) are retained in atp1 exons after loss of the intron in Pteris argyraea/biaurita and in Actiniopteris and Onychium We find that atp1i361g2 has significant similarity with intron rps3i249g2 present in lycophytes and gymnosperms, which we now also find highly conserved in ferns. We conclude that atp1i361g2 may have originated from the more ancestral rps3i249g2 paralogue by a reverse splicing copy event early in the evolution of monilophytes. Secondary structure elements of the two introns, most characteristically their domains III, show strikingly convergent evolution in the monilophytes. Moreover, the intron paralogue rps3i249g2 reveals relaxed evolution in taxa where the atp1i361g2 paralogue is lost. Our findings may reflect convergent evolution of the two related mitochondrial introns exerted by co-evolution with an intron-binding protein simultaneously acting on the two paralogues.

Introduction

Group II introns can be found in mitochondrial and chloroplast genomes of plants, algae and fungi but also in the genomes of bacteria and archaea (Allet and Rochaix 1979; Dujon 1980; Michel and Dujon 1983; Bonen 2008; Simon et al. 2008; Lambowitz and Zimmerly 2011; McNeil et al. 2016). Group II introns may have originated in eubacteria and transmitted into eukaryotes via endosymbiosis from α-proteobacteria or cyanobacteria which developed into mitochondria or chloroplasts, respectively. Possibly, group II introns later migrated into the nucleus and gave rise to spliceosomal introns in eukaryotes (Cech 1986; Rogers 1990; Cavalier-Smith 1991; Doolittle 1991; Palmer and Logsdon 1991; Roger and Doolittle 1993; Logsdon 1998; Belshaw and Bensasson 2006).

The classification of group II introns is based on their characteristic secondary structures and the splicing mechanism. According to a widely accepted nomenclature proposal (Dombrovska and Qiu 2004), organelle introns are named after their host gene, the nucleotide position upstream of the insertion site (with the homologous liverwort Marchantia polymorpha sequence as reference) and the type of intron.

Group II introns feature six distinct domains which are positioned around a central wheel structure (Michel et al. 1989; Michel and Ferat 1995; Qin and Pyle 1998; Toor et al. 2001; Simon et al. 2008). The complex intron structure is based on highly conserved intra-intronic and intron-exon binding sites. Although some group II introns are known to be capable of self-splicing (Michel and Ferat 1995), plant organelle group II introns rely on nuclear co-factors or intron-encoded maturases for splicing (Matsuura et al. 2001; Brown et al. 2014). Maturases are generally encoded in domain IV and many of them possess reverse transcriptase domains (Michel and Lang 1985; Kennell et al. 1993; Matsuura et al. 2001). This allows group II introns to act as mobile genetic elements (Kennell et al. 1993; Lambowitz and Belfort 1993; Cousineau et al. 2000; Lambowitz and Zimmerly 2004, 2011; McNeil et al. 2016). Most plant organellar group II introns, however, do not contain maturase ORFs (Bonen and Vogel 2001).

The group II intron atp1i361g2 was discovered in an earlier study, which introduced the mitochondrial atp1 gene a new locus to investigate fern (monilophyte) phylogeny (Wikström and Pryer 2005). Intron atp1i361g2 was found in most ferns except Ophioglossales, Psilotales and Equisetales and Danaea elliptica (Wikström and Pryer 2005) and was later also found to be lacking in a subset of Pteridaceae (Rothfels and Schuettpelz 2014). Given its absence in Ophioglossales, Psilotales and Equisetales the most parsimonious explanation of the distribution of atp1i361g2 is a single gain in the joint clade of Marattiales and leptosporangiate ferns (Grewe et al. 2013; Knie et al. 2015).

The process of pyrimidine exchanges in RNA transcripts, termed RNA editing, is common in the organelles of land plants except the marchantiid liverworts as part of RNA maturation in chloroplasts and mitochondria (Chateigner-Boutin and Small 2011; Knoop 2011). In most cases, RNA editing restores conserved codons in mRNAs, but it has also been shown to be present in noncoding RNAs, including introns (Lippok et al. 1994; Carrillo and Bonen 1997; Carrillo et al. 2001; Castandet et al. 2010; Bégu et al. 2011; Farré et al. 2012; Oldenkott et al. 2014), rRNAs (Hecht et al. 2011) and tRNAs (Binder et al. 1994; Marchfelder et al. 1996; Maréchal-Drouard et al. 1996; Grewe et al. 2009, 2011). Intronic editing may restore critical nucleotides which are essential for the secondary structures and a prerequisite for splicing (Castandet et al. 2010; Bégu et al. 2011; Farré et al. 2012), although it should be noted that expected RNA editing to recreate highly conserved domain structures had not been observed in other cases (Carrillo and Bonen 1997)

The loss of RNA editing sites is a common phenomenon which can be explained by restoring point mutations or the recombination of matured cDNA into the organelle genomes. Intron losses are so far described as accompanied by a partial or complete loss of RNA editing sites in the vicinity of the insertion sites (Geiss et al. 1994; Itchoda et al. 2002; Lopez et al. 2007; Ran et al. 2010; Grewe et al. 2011; Oldenkott et al. 2014) and provide evidence for recombination events of in vivo retrotranscribed cDNA and genomic DNA.

In this study, we report on an extended investigation of the Pteridaceae finding at least four independent losses of atp1i361g2 within this family. Surprisingly, two of the atp1i361g2 losses did not cause a loss of the flanking RNA editing sites, hence in conflict with a model invoking a cDNA-mediated recombination event. We present a scenario postulating the more ancient intron rps3i249g2 as potential source of atp1i361g2 arising in the joint clade of Marattiales and leptosporangiate ferns. Finally, the detailed inspection of the two intron paralogues shows an intriguing co-evolution of secondary structure elements, notably their domains III, in the monilophyte clades in general and among Pteridaceae in particular.

Materials and Methods

Plant Material and Molecular Work

Plant material (table 1) was obtained from the Bonn Botanic Gardens, the Botanic Garden Marburg and the Botanic Garden München-Nymphenburg. Adiantum raddianum and Pteris biaurita were obtained from a commercial source (Exotic Plants, Mannheim: http://www.exotic-plants.de). Nucleic acids were extracted using the CTAB protocol (Doyle and Doyle 1990). Crude nucleic acid preparations were either treated with RNase A or DNase I (Thermo Scientific) to purify DNA or RNA, respectively. RNA was reverse transcribed into cDNA using the RevertAid First Strand cDNA Synthesis Kit (Thermo Scientific) and random hexamer primers (Roth).

Table 1

Taxa and Loci

Species	Botanic garden accession numbers	matK	rbcL	atpA	atpB	rps3	atp1	atp1 cDNA
Cheilanthes fraseri	PE-0-BONN-29933	–	KU744807	–	KU744792	–	KU744752	KU744740
Astrolepis sinuata	xx-0-BONN-34363	KF289577	EF452141	EF452079	EF452021	KU744816	KU744753	KU744739
Myriopteris microphylla	xx-0-BONN-17356	KU744798	KF961807	–	KU744791	–	KU744754	–
Doryopteris pedata	xx-0-BONN-34193	KU744799	U27206	KU744781	KU744797	–	KU744755	–
Pellaea falcata	xx-0-BONN-17727	–	GU136794	–	GU136775	–	KU744756	–
Adiantum capillus-veneris	xx-0-BONN-22292	AY178864	AY178864	AY178864	AY178864	KU744819	KU744760	KU744738
Adiantum raddianum	Commercial source	KJ605550	KU744803	EF452071	GU136769	KU744820	KU744764	–
Adiantum peruvianum	xx-0-B-0436988	KJ605530	EF452133	EF452070	EF452013	KU744822	KU744759	KU744742
Adiantum reniforme	ES-0-BONN-1097	KJ605551	KU744804	KJ742797	JF935364	KU744821	KU744763	–
Adiantum trapeziforme	MX-0-BONN-17375	KU744800	KU744805	–	KU744789	KU744823	KU744761	KU744743
Adiantum pedatum	US-0-BONN-33986	KJ605529	KU744802	EF452069	EF452012	KU744824	KU744762	–
Vittaria lineata	xx-0-BONN-17295	–	KU744813	KU744782	KJ716384	–	KU744757	–
Antrophyum mannianum	RW-0-BONN-31914	–	KU744814	KU744785	KU744790	KU744818	KU744758	KU744741
Pteris argyraea	xx-0-BONN-3977	KF289554	EF452169	EF452117	EF452054	KU744826	KU744770	KU744745
Pteris quadriaurita	xx-0-BONN-17387	KF289541	EF452173	EF452121	KU744795	–	KU744768	–
Pteris cretica	xx-0-BONN-17520	KF289524	KU744810	EF452118	EF452055	KU744825	KU744765	KU744747
Pteris biaurita	Commercial source	KF289546	KF289546	KU744783	–	–	KU744771	–
Pteris ryukyuensis	MB-1992/478	KF289492	AB574842	KU744780	HM582601	–	KU744767	KU744744
Pteris vittata	MB-1975/82	KF289512	KU744812	EF452123	KU744796	KU744827	KU744769	KU744748
Pteris umbrosa	MB-1979/1674	KF289518	KU744811	–	–	–	KU744766	KU744746
Pityrogramma sulphurea	AR-0-MB-2011/0060	–	KU744809	KU744786	KU744794	KU744828	KU744772	KU744749
Actiniopteris dimorpha	xx-0-M-1996/3147	KF289571	EF452130	EF452066	KU744788	–	KU744773	–
Onychium japonicum	xx-0-BONN-26126	KF289514	KU744808	EF452107	EF452045	KU744817	KU744774	KU744750
Onychium siliculosum	BG München	KF289580	KU744815	KU744784	–	–	KU744775	–
Ceratopteris thalictroides	xx-0-BONN-1069	KJ772645	KU744806	–	–	KU744830	KU744776	–
Acrostichum aureum	xx-0-BONN-1096	KF848284	KU744801	JF303991	KU744787	KU744829	KU744777	KU744751
Coniogramme japonica	xx-0-M-S/1230	JF303920	KC700111	JF303990	KU744793	–	KU744778	–
Outgroups
Polypodium cambricum	TR-0-BONN-15737	HE970727	FJ825703	JF832137 P. vulgare	EF463510 P. vulgare	KU744831	KJ944565	–
Woodwardia radicans	xx-0-BONN-3522	JF303937 W. japonica	AY137667	EF463623 W. virginica	EF463359 W. virginica	KU744832	KJ944567	–
Dicksonia antarctica	xx-0-BONN-20037	HM021802	U05919	AM176442	U93829	KU744834	AJ548853	–
Azolla filiculoides	xx-0-BONN-16921	–	U24185 A. caroliniana	DQ390547 A. pinnata	AY612689	KU744835	KJ944561	–
Vandenboschia radicans	PT-0-BONN-17818	JF303901	AF275650	DQ390581	AY612715	KU744836	KJ944574	–
Todea barbara	xx-0-BONN-20306	KM925082	AY612686	DQ390580	AY612714	KU744833	KJ944575	–
Angiopteris madagascariensis	SC-0-BONN-17551	DQ821119 A. evecta	EF463239 A. evecta	DQ390544 A. evecta	EF463485 A. evecta	KU744837	KJ944577	–
Ophioglossum petiolatum	xx-0-BONN-167	HF585134	AF313582 O. reticulatum	DQ390571 O. reticulatum	U93825 O. reticulatum	KU744838	KJ944580	–
Equisetum arvense	xx-0-BONN-3965	JN968380	JN968380	JN968380	JN968380	KU756282 E. giganteum	KJ944581 E. giganteum	–

Note.—Database accessions are given for all sequences investigated in this study. Accession numbers in bold indicate new sequences obtained in this study, “–” indicates that no data are available. Plant material was obtained from the Botanic Garden Bonn or from commercial sources. Accession numbers of plant material from the botanic garden are also shown. The respective epithet is indicated below the accession number when originating from a different species of the same genus.

Mitochondrial DNA and cDNA amplicons were amplified by Touchdown (RT)-PCR with the Go-Taq polymerase (Promega). A list of primers for PCR amplification is provided as supplementary table S1, Supplementary Material online. An initial annealing temperature of 55 °C was lowered to 45 °C in one degree centigrade steps per cycle. The elongation times were between 2 min and 2 min 30 s. After gel electrophoresis in 0.8% agarose gels the PCR products were cut out and purified with the NucleoSpin Extract II Kit (Macherey Nagel). The recovered PCR products were cloned into the pGEM-T Easy Vector (Promega) and transformed into Escherichia coli XLI blue cells. Plasmids were isolated using the BibDo method (Birnboim and Doly 1979). Plasmid insert sequencing was done at Macrogen Europe (Amsterdam, The Netherlands).

RNA Editing Analysis

Exon RNA editing sites were determined from at least two independent cDNA clones per taxon. Incongruencies in editing patterns, e. g. owing to partial editing, were evaluated by parallel direct sequencing of the respective RT-PCR products. Editing sites in introns were determined with two overlapping RT-PCRs using one primer binding within the intron and one to an edited exon sequences followed by merging of the separate sequences. All new sequences are deposited in GenBank (table 1). Editing analysis was aided by using PREPACT available under www.prepact.de (Lenz et al. 2010; Lenz and Knoop 2013). RNA editing sites are labelled following a standardizing nomenclature (Rüdinger et al. 2009; Lenz et al. 2010). Lacking cDNA sequences were supplemented by predictions with the alignment prediction tool of PREPACT and the sequence of Pteris umbrosa as reference. Using diverse alternative references, including those of liverworts, mosses or angiosperms, for editing prediction in the highly conserved atp1 gene had no significant impact. All detected exon editing sites are shown in supplementary table S2, Supplementary Material online. Intron RNA editing sites are labeled with the name of the intron, the nucleotide resulting from editing (eU or eC) and the intron nucleotide position. All detected intronic RNA editing sites are shown in supplementary table S3, Supplementary Material online.

Phylogenetic Analysis

Sequence handling and alignment was performed with MEGA 5.21 (Tamura et al. 2011) using the implemented ClustalW algorithm (Thompson et al. 1994) with subsequent manual corrections. Phylogenetic tree inference was done with IQ-Tree (Nguyen et al. 2015), model selection was set auto and bootstrapping (1,000 replicates) was done with the ultrafast bootstrapping algorithm (Minh et al. 2013). The best fitting model according to the Bayesian Information Criterion (Schwarz 1978) and the corrected Akaike Information Criterion (Burnham and Anderson 2004) was the GTR + Γ+I substitution model (Rodríguez et al. 1990) with four discrete gamma categories. Phylogenetic tree construction was independently done with PhyML using the SPR branch-swapping option and evaluating node support with 150 bootstrap replicates. The resulting tree was fully congruent with the one shown in figures 2 and 3. Significant bootstrap supports of at least 70% deviating by >7% only affected nodes placing Saccoloma inaequale as sister of Pteridaceae, Dennstaedtiaceae and Eupolypodiales (98–83), Davallia solida as sister of Nephrolepis cordifolia (100–91), Cryptogrammoids as sister of all other Pteridaceae (96–72), the monophyletic clade of Anetium citrifolium, the genera Vittaria and Antrophyum (100–78), uniting all taxa of the genus Pteris (100–78) and the node placing Botrychium lunaria sister to Helminthostachys zeylanica and Ophioglossum californicum (89–76).

F

Phylogeny of monilophytes infered with IQ-Tree (Nguyen et al. 2015), based on an alignment of chloroplast loci atpA, atpB, rbcL and matK. Bootstrapping was performed with the ultrafast bootstrap algorithm (Minh et al. 2013) with 1,000 replications. Only bootstrap node supports of at least 70 are shown. Shading highlights species investigated for the atp1 locus, indicating the presence (blue) or absence (red) of the atp1i361g2 intron. The likely gain of intron atp1i361g2 in the common ancestor of Marattiales and leptosporangiate ferns is shown with a filled triangle, its loss in Danaea elliptica is shown with an open triangle. The phylogeny of Pteridaceae is shown in detail in figure 3.

F

Phylogeny of the Pteridaceae and losses of the atp1i361g2 intron. The phylogram shown is an expanded view of the clade collapsed in figure 2. Shading highlights species investigated for the atp1 locus, indicating the presence (blue) or absence (red) of the atp1i361g2 intron. We could show that atp1i361g2 is absent in the AD/CH clade, the Ceraptopteridoids, Actiniopteris dimorpha, Onychium and Pteris argyraea and P. biaurita. According to this phylogeny, atp1i361g2 was lost at least 4 times in the Pteridaceae.

PCR amplification of the mitochondrial atp1 gene, here exemplarily shown for a DNA sampling of nine Pteridaceae species and the Pteris cretica cDNA. The different product sizes indicate variable presence of intron atp1i361g2, subsequently confirmed by cloning and sequencing.

Results

Phylogeny of Monilophytes Focusing on Pteridaceae: Five Independent Losses of atp1i361g2

Extending the taxon sampling for the mitochondrial atp1 gene among monilophytes (table 1), we observed a variability of PCR product sizes suggesting a variance in the presence of intron atp1i361g2 among Pteridaceae as exemplarily shown in figure 2. In order to elucidate the evolutionary history of the monilophyte-specific intron atp1i361g2 in monilophytes and particularly in Pteridaceae, we extended the original data set of Schuettpelz et al. (2007) with an expanded taxon sampling (table 1 and supplementary table S4, Supplementary Material online) and included matK for phylogenetic analysis, which has previously been used successfully in different studies of monilophyte and Pteridaceae phylogeny (Kuo et al. 2011; Rothfels et al. 2012; Chao et al. 2014; Knie et al. 2015). The resulting phylogenetic trees are shown in figures 2 and 3.

Recent insights on monilophyte backbone phylogeny placing horsetails (Equisetales) as sister to all remaining monilophytes (Knie et al. 2015; Rothfels et al. 2015) are used to root the tree in figure 2. The phylogenetic tree is fully congruent with a modern understanding of monilophyte phylogeny. Noteworthy are high bootstrap values supporting sister group relationships of Gleicheniales and Hymenophyllales and Cyatheales and Polypodiales, respectively. In our study, Pteridaceae are identified as sister group to a clade comprising Dennstaedtiaceae and Eupolypod clades I and II (fig. 2). Lindsaeaceae and Lonchitis are identified as a monophylum, sister to all other Polypodiales. The isolated taxon Saccoloma inequale alone is sister to the large clade comprising Dennstaedtiaceae, Pteridaceae and Eupolypods, the “core polypods”. Except for the Pteridaceae, which we will consider separately (fig. 3), the evolutionary history of the atp1i361g2 intron is most parsimoniously explained with a unique gain in the common ancestor of Marattiales and the leptosporangiate lineage, followed by an early loss in Danaea elliptica (fig. 2; Wikström and Pryer 2005; Grewe et al. 2013; Knie et al. 2015).

Within Pteridaceae, the clades consisting of Adiantoids and Cheilanthoids and the Pteridoids and Ceratopteridoids, respectively, are sister to each other (fig. 3). Cryptogrammoids are the sister group to all other Pteridaceae. The support for the positions of Cryptogrammoids as the earliest-branching taxon in Pteridaceae and Ceratopteridoids sister to Pteridoids in our analysis is significantly higher compared with the IQ-Tree derived supports using the Schuettpelz data set, which in our analysis receives support values of 93 and 81 (not shown). The genus Pteris is paraphyletic with Neurocallis, Ochropteris and Platyzoma considered as separate genera (Schuettpelz et al. 2007). The extended genus Pteris sensu lato (Zhang et al. 2015) is well supported in our analysis (fig. 3). The genus Onychium is identified to be monophyletic in contrast to an earlier study on Pteridaceae phylogeny (Chao et al. 2014).

Group II intron atp1i361g2 has previously been assumed to be absent in Adiantoids and Cheilanthoids (AD/CH) but present in Cryptogrammoids and Pteridoids (Rothfels and Schuettpelz 2014). We now find that atp1i361g2 is also absent in all investigated Ceratopteridoids, in the genus Onychium, in Actiniopteris dimorpha and in Pteris argyraea and P. biaurita. In particular, the absence of atp1i361g2 in P. biaurita and P. argyraea and in Actiniopteris dimorpha and Onychium but its retention in their respective sister species is striking and indicates (at least) two evolutionary recent losses on top of two deeper losses in the Ceratopteridoids and in the AD/CH stem lineage.

Searches for coexisting atp1 alleles using specific primers across exon–intron borders in species which lack the atp1i361g2 intron revealed an additional, intron-containing but pseudogenized atp1 gene copy only in Onychium japonicum, which is frameshifted and, as cDNA studies suggested, not transcribed. No evidence for such additional intron-containing alleles of atp1 was found for the other intron-lacking species.

RNA Editing in Pteridaceae atp1 and Its Correlation to Intron Losses

Plant organelle intron losses are often accompanied by loss of RNA editing events in the flanking exons, assumed to result from recombination of cDNA derived from matured transcripts. We expected that comparing RNA editing patterns could be particularly interesting for atp1 of the Pteridaceae, given that we postulate two deep and two shallow losses of atp1i361g2. Our RNA editing analyses of atp1 show that the numbers of RNA editing sites vary widely between the monophyletic subgroups of the Pteridaceae, most notably comparing species Antrophyum mannianum and Acrostichum aureum to the basal Cryptogrammoids (CR). Many RNA editing sites are conserved among the species of the different subclades, in particular among the early-branching Cryptogrammoids and Pteridoids (fig. 4).

F

Venn diagram showing the occurrence of RNA editing sites in the five subclades of Pteridaceae. Editing sites which are shared between the early-branching Cryptogrammoids and at least one of the other four subclades is considered to be an ancient plesiomorphic editing state among Pteridaceae. Most editing sites observed in the Pteridaceae are therefore plesiomorphic. The Adiantoid and Pteridoid clades feature 14 and 8 unique editing sites, respectively, adding to the more ancient editing sites shared with the other clades.

We consider edits shared between the CR subclade and at least one of the other subclades to represent a plesiomorphic editing state among Pteridaceae. In total, we count 124 different sites of RNA editing in the atp1 gene of Pteridaceae. Of these, we assume 88 edits to be ancient plesiomorphic editing events and consider 13 as clade-specific edits (which occur in at least two genera) and 23 as “unique” genus-specific sites.

Overall, it is evident that the AD and CH clades show particularly diverging editing patterns. A detailed investigation of the RNA editing sites in selected taxa of the different Pteridaceae clades makes this more obvious. In species, which have lost atp1i361g2, numbers of RNA editing sites are overall drastically reduced (fig. 5). Most notably, Acrostichum aureum lacking atp1i361g2 is nearly completely devoid of editing sites downstream of the previous intron insertion site as could be expected for a cDNA-mediated recombination extending far into the downstream exon. Surprisingly, however, such a correlation in the absence of intron and editing sites is not found for Pteris argyraea/biaurita and Actiniopteris dimorpha and Onychium (fig. 5). In these taxa, nearly all editing sites in comparison to the respective closest relatives, including the edits in close vicinity to the intron insertion site, are retained. This is particularly intriguing given the close relation of Pteris argyraea and P. biaurita and its intron-containing sister taxon P. quadriaurita.

F

Intron losses and RNA editing patterns in atp1 of selected taxa of the Pteridaceae. Phylogenetic relationships as determined (fig. 3) and the events of intron losses (open triangles) are shown with the simplified cladogram on top. Species without atp1i361g2 are shaded in grey. Acrostichum aureum has only one editing site downstream of the intron insertion site. In the intron-less species Pteris argyraea and Actiniopteris dimorpha no reduction of RNA editing sites is observed. Editing sites of Coniogramme japonica are predicted.

Although the majority of the editing sites in the intron-less species of the Cheilanthoids and Adiantoids can be considered to be plesiomorphic edits also found in the intron-containing species and intron-lacking species that are not affected by RNA editing loss, we also found at least 21 new editing sites. These unique edits, mostly found in the genus Adiantum, may have evolved after the loss of atp1i361g2 in the common ancestor of Cheilanthoids and Adiantoids. A particularly interesting case is the editing pattern of Adiantum trapeziforme given that exclusively U-to-C editing could be observed within the atp1 amplicon. To our knowledge, this is the first example of a plant organelle transcript region featuring exclusively U-to-C editing. However, other mitochondrial cDNAs of Adiantum trapeziforme, including the here investigated rps3 gene, show the usual mix of C-to-U and U-to-C editing typically identified in monilophytes.

Structural Analysis of atp1i361g2 and Its Likely Source Paralogue, rps3i249g2

Its unique presence in the monilophytes and its independent losses connected to, or disconnected from, the loss of flanking editing sites prompted us to investigate atp1i361g2 more closely. Where present, atp1i361g2 features conserved canonical group II intron structures, with sequence or structural differences pronounced mainly in domain IV among Marattiales and Polypodiales. The atp1i361g2 secondary structure is exemplarily shown for Pteris cretica (fig. 6).

F

Secondary structure model of atp1i361g2 in Pteris cretica and similarities with its paralogue rps3i249g2. The intron features all typically conserved structures in group II introns (Michel et al. 1989; Michel and Ferat 1995; Qin and Pyle 1998; Toor et al. 2001; Simon et al. 2008). The six intron domains are labelled with Roman numerals (I–VI) and the tertiary interaction sites with Greek letters. Exon binding sites (EBS) and corresponding intron binding sites (IBS) in the 5′-exon are indicated. Nucleotides which are identical to the rps3i249g2 intron are shown in green. Differences in nucleotide sequences that can be explained by transitions (A/G and C/U) are shown in yellow. Four experimentally verified C-to-U editing sites are shaded in blue and the three newly created A–U bindings in stem regions of domain I and domain VI are marked with black dots. Nucleotides in black circles highlight tertiary base-pairing interactions. Introns were manually folded and the drawing was made using the VARNA software (Darty et al. 2009).

We noted that atp1i361g2 shows significant primary sequence similarities to rps3i249g2, a mitochondrial group II intron conserved in Adiantum capillus-veneris, the lycophyte Phlegmariurus squarrosus and gymnosperms (Ran et al. 2010; Regina and Quagliariello 2010; Liu et al. 2012; Bonavita and Regina 2016). We targeted the rps3i249g2 insertion site for a phylogenetic wide sampling of monilophytes, finding it highly conserved and without evidence for intron losses (table 1). For closer comparison to atp1i361g2 we cloned and sequenced several monilophyte rps3 amplicons with a focus on the Pteridaceae. A secondary structure model of rps3i249g2 in Pteris cretica is given in supplementary figure S1, Supplementary Material online, for direct comparison to its atp1i361g2 counterpart in the same species (fig. 6). The most significant similarities between the two intron paralogues are found in the region encompassing the 3′-part of domain I up to the beginning of domain IV and in their near-identical domains V (fig. 6 and supplementary fig. S1, Supplementary Material online). Given the now observed wide conservation of rps3i249g2 among monilophytes and its presence also in lycophytes and gymnosperms, we conclude that an intron-copying event via reverse splicing of rps3i249g2 into atp1 in the joint stem lineage of Marattiales and leptosporangiate ferns may be a likely evolutionary scenario to explain the existence of atp1i361g2 in monilophytes. However, it should be noted that the present dissimilar exon binding sites of the two intron paralogues do not immediately support such a reverse splicing scenario. Possibly, an alternative intron folding within an ancestral rps3i249g2 domain I may have created novel EBS sites for interacting at the new location in atp1 (supplementary fig. S2, Supplementary Material online). An alternative scenario of large-scale genomic recombination appears less likely given that similarities are entirely restricted to the intron sequences and no conversion of flanking exon sequences is discernible.

Convergent Evolution of Domains III of atp1i361g2 and rps3i249g2

As our preliminary data suggested that the two paralogues would not decrease but rather even increase in similarity over evolutionary time, we extended the monilophyte sampling for atp1i361g2 and rps3i249g2 also outside of the Pteridaceae (table 1). The comparative analysis revealed a striking convergent evolution of the region most similar between the two intron paralogues, most notably with respect to the structure of their domains III. In fact, the domains III of the two intron paralogues within one species are extremely similar, even nearly or completely identical in Polypodiales, whereas the orthologous regions differ even between closely related taxa (figs. 7 and 8; supplementary fig. S3, Supplementary Material online). The independent unique losses of atp1i361g2 are interesting cases to investigate the evolutionary fate of the rps3i249g2 paralogue. The losses of atp1i361g2 indeed seem to allow domain III of rps3i249g2 to evolve more freely as reflected by higher sequence divergence among related species. The only exception in our sampling are rps3i249g2 domains III of Pteris agyraea and Pteris quadriaurita, two species which have split only very recently in evolution.

F

Convergent evolution of domains III of atp1i361g2 and rps3i249g2 in Polypodiales and Pteridaceae. Graphic display is as in figure 7. Convergent evolution of the structure of domain III is observed in species with both introns. When atp1i361g2 is lost (open triangles), domain III of rps3i249g2 undergoes structural changes. A conserved C-to-U editing site in a bulge of the stem region in Pteris cretica is highlighted in blue. Black circles indicate inserted nucleotides.

F

Convergent evolution of domains III of atp1i361g2 and rps3i249g2. Domains III of the two group II introns are compared in four selected fern species. Examples within the Polypodiales are displayed in detail in figure 8. Identical nucleotides in paralogues are shown in green and transitions between both domains are shown in yellow, analogous to figure 6. The paralogous domains III are nearly identical whereas orthologous domains are highly different in the analysed monilophyte taxa.

Intron RNA Editing

High numbers of RNA editing sites in plant organelle coding regions generally indicate that intron sequences might be affected by RNA editing, too. Given the high abundance of RNA editing identified in the atp1 coding sequences, we exemplarily checked for RNA editing in partially matured, unspliced pre-mRNAs. We identified five sites of C-to-U and two sites of U-to-C editing in Pteris cretica and P. vittata (see supplementary table S3, Supplementary Material online). One of these sites is of particular interest as it affects a bulged region close to the conserved proximal stem of domain III, which shows convergent evolution with its rps3i249g2 counterpart (fig. 8). The position of C-to-U editing identified in the two Pteris species is “pre-edited” with a T present on DNA level in this position in the atp1i361g2 orthologues of other taxa including most distantly related genera Angiopteris, Marattia and Todea. Intriguingly, this position is also pre-edited in the paralogue domain III of rps3i249g2, including the adiantoids Adiantum capillus-veneris and A. trapeziforme taxa where atp1i361g2 is lost.

Discussion

Here, we have described the evolutionary history of monilophyte-specific intron atp1i361g2 and its likely source paralogue rps3i249g2. Our data are discussed on the basis of an extended monilophyte phylogeny based on four chloroplast loci (atpA, atpB, rbcL and matK). The resulting phylogenetic trees (figs. 2 and 3) are fully congruent with recent insights on monilophyte phylogeny and even further contribute to open issues of fern phylogeny such as a likely sister grouping of Gleicheniales and Hymenophyllales, of Cyatheales and Polypodiales, and, among the latter, in supporting the serial sister group relationships of Dennstaedtiaceae, Pteridaceae, Saccoloma and Lindsaeaceae (and Lonchitis) to the Eupolypods. We have used this encompassing monilophyte phylogeny and the detailed phylogeny of the Pteridaceae as the framework to investigate the evolutionary fates of the two mitochondrial introns, which are of rather restricted occurrence among land plants. Importantly, critical nodes such as the position of the Ceratopteridoids are now perfectly supported. The well-supported phylogeny enables to identify an evolutionary scenario, in which intron atp1i361g2 has been lost 4 times independently in the Pteridaceae (fig. 3).

Introns in plant mitochondrial DNAs are strikingly different between the major land plant clades and have contributed to elucidate the embryophyte backbone phylogeny. Particularly intriguing, for example, are the intron patterns supporting the sister group relationships of liverworts and all other land plants (Qiu et al. 1998) and of hornworts and vascular plants (Groth-Malonek et al. 2005), respectively. Notably, not one single mitochondrial intron of altogether 100 recognized among bryophytes is shared between all of their three distinct clades, the liverworts, mosses and hornworts (Knoop 2010, 2013). In contrast to the intron gain and loss dynamic along the backbone of land plant phylogeny, mitochondrial introns are much more stable once a clade is established. For example, most of 27 introns in the mitochondrial DNA of the gymnosperm Cycas taitungensis (with the exception of rps3i249g2 and the chloroplast-derived trnVi36g2) are conserved among flowering plants. Exceptions of introns lost more frequently and independently among angiosperms are cox2i373g2, cox2i691g2 and nad4i976g2 (Unseld et al. 1997; Kubo et al. 2000; Itchoda et al. 2002; Kudla et al. 2002; Sloan et al. 2010; Hepburn et al. 2012). A recently identified unique exception to the overall stability of mitochondrial introns within a clade is the gymnosperm Welwitschia mirabilis, which has lost more than half of the ancient seed plant mitochondrial intron set (Guo et al. 2016).

The here investigated atp1i361g2 intron is unique to monilophytes. We now suggest that it originated in the joint stem lineage of Marattiales and leptosporangiate ferns from its more ancient counterpart rps3i249g2, which likely emerged much earlier with the origin of tracheophytes. Alternative scenarios, for example, invoking events of horizontal gene transfer (HGT) are less likely given that the phylogenies of both introns are fully congruent with the organismic phylogeny (supplementary fig. S4, Supplementary Material online). The rps3i249g2-to-atp1i361g2 intron copying event likely is a further molecular synapomorphy for the joint clade of Marattiales and leptosporangiate ferns, which is well supported from independent recent phylogenetic studies (Knie et al. 2015; Rothfels et al. 2015). Whereas rps3i249g2 has never been found in flowering plants and was likely lost in the angiosperm stem lineage, we here report that it is highly conserved among monilophytes.

The here identified independent losses of atp1i361g2 in a moderate sampling of 42 monilophyte taxa, four of which occur in the Pteridaceae alone, are unusual for a mitochondrial group II intron. Intriguingly though, nad5i1242g2 has also been found lost at least 3 times independently in a sampling of 27 monilophytes (Vangerow et al. 1999; Knie et al. 2015). Possibly, monilophytes may ultimately reveal more dynamic intron loss scenarios than seed plants.

Losses of organelle introns are usually accompanied by a loss of RNA editing sites in the closer or wider vicinity of the intron insertion site (Geiss et al. 1994; Itchoda et al. 2002; Lopez et al. 2007; Ran et al. 2010; Grewe et al. 2011; Oldenkott et al. 2014). The comparative studies suggest that the losses of RNA editing sites in the environment of the lost introns are within a range of 210 bp to >1,600 bp (Geiss et al. 1994; Itchoda et al. 2002; Ran et al. 2010; Grewe et al. 2011; Oldenkott et al. 2014). Integration of matured and reverse transcribed RNA into the genomic DNA is assumed to cause such a simultaneous loss of editing sites together with the intron. This scenario is also plausible for two of the observed deep losses of atp1i361g2 in the CE and in the joint AD/CH clade. In the AD/CH clade, we see a sporadic re-occurrence of what we classified as plesiomorphic, ancient editing sites. We assume that the common ancestor of the AD/CH clade has lost atp1i361g2 and most flanking RNA editing sites in a single, ancient cDNA recombination event. The nuclear-encoded RNA editing factors responsible for the individual RNA editing sites, however, initially remain present and will turn into pseudogenes and ultimately disappear only if not needed over a longer period of time. A re-appearing pyrimidine transition at an ancient editing site can immediately be addressed, when the respective editing factor is still present to correct the mutation and evolutionary pressure on its maintenance is regained. Notably, we recently found that editing factors are more strongly maintained in the course of angiosperm evolution when addressing multiple sites on different transcripts in parallel (Hein et al. 2016), a scenario that is very likely for the abundant editing observed in monilophytes.

Pteris argyraea, its close relative Pteris biaurita, as well as the genus Onychium and Actiniopteris dimorpha are intriguing cases, because these taxa lost the atp1i361g2 intron without an apparent effect on editing patterns in the vicinity of the former insertion site. A scenario as described earlier with a complete regain of all editing sites by pyrimidine transitions appears very unlikely. A possible explanation for the observations could be efficient splicing of unedited transcripts giving rise to cDNA substrates for recombination in the ancestors of the taxa lacking the intron but retaining RNA editing. However, for partially matured atp1 transcripts in intron-containing species, we rather observed partially or even completely edited but unspliced variants in our cDNA analyses. It appears more likely that the Pteris argyraea/biaurita and Onychium/Actiniopteris dimorpha cases represent examples of recombination mediated through very short stretches of homology between the most proximal editing sites flanking the intron insertion site, here 18-bp downstream and 42-bp upstream. Such a concept of recombination across very short nucleotide stretches was previously also proposed to explain intron and editing site losses in Silene noctiflora (Sloan et al. 2010).

Monilophyte-specific intron atp1i361g2 shows significant similarity to rps3i249g2, its likely source paralogue. Particularly intriguing are the structural similarities in domains III, which do not decrease but rather increase over evolutionary time after creation of the atp1i361g2 paralogue (figs. 7 and 8; supplementary fig. S3, Supplementary Material online). Additional support for a scenario of convergent evolution comes from the relaxed evolution of rps3i249g2 domain III once its paralogue atp1i361g2 is lost and from an editing event further increasing similarity in domain III. Domain III plays a key role in positioning domain I and V during splicing (Fedorova and Pyle 2005). Micro-recombinational events of cDNA across short sequence stretches, as assumed here for the loss of introns without the loss of flanking editing sites and also proposed previously (Sloan et al. 2010), could be the cause for the extraordinary sequence similarities of the intron domain III paralogues. However, if mediated by cDNA, this would require a significant amount of unspliced immature RNA as a source for cDNA synthesis. Moreover, the lack of extensive sequence similarities downstream of domain III that could serve as an anchoring guide for homology-mediated recombination makes this mechanism less likely in our eyes. The frequent recombination in plant mtDNA on genomic level generally relies on much larger and perfect repeat sequences that create reciprocal exchanges of the flanking unique sequences. We rather assume a splicing-factor or an otherwise RNA-binding protein affecting both intron paralogues simultaneously to be the underlying cause for our observations. Numerous such splicing factors affecting individual or multiple organelle introns simultaneously have already been identified in model taxa like Arabidopsis thaliana, Physcomitrella patens, rice or maize (Ostersetzer et al. 2005; Keren et al. 2009, 2012; Brown et al. 2014; Cohen et al. 2014). If such a factor pre-existed for the more ancient rps3i249g2 to assist splicing, it likely also serves the new atp1i361g2 intron paralogue. Unfortunately, the restricted occurrence of rps3i249g2 only in the lycophyte Phlegmariurus, in gymnosperms and, as here shown, also in ferns, all of which are clades not represented by prominent model organisms, does hitherto not provide hints to a relevant splicing factor.

Supplementary Material

Supplementary tables S1–S4 and Supplementary figures S1–4 are available at Genome Biology and Evolution online (http://www.gbe.oxfordjournals.org/).