Background

Patterns and processes of organic evolution are reflected in the structure and sequences of organisms' genomes. Although we are only starting to accumulate sufficient data to compare nuclear genomes of plants, more data are available for the smaller plastid genomes (plastomes). Comparative work on plastomes began in the early 1980's using restriction site mapping and hybridization with heterologous probes to generate phylogenetically informative data within small clades, generally below the family level [1, 2] as well as using comparative mapping to examine differences among more distantly related groups [3–6]. By the 1990's the emphasis shifted to nucleotide sequences from targeted regions (genes) with the result that studies with broad genomic sampling were few. However, a reduction in sequencing cost and concomitant development of analytical tools over the past 10 years has resulted in resurgence of comparative plastid genomics.

The plastome is highly conserved in overall structure and this provides the basis for comparative studies [4]. However, there is still sufficient variation to identify rare genomic events that mark critical branches in land plant evolution, thereby elucidating early evolutionary modifications to the plastid genome [e.g., [5, 7]]. Plastomes of Marchantia polymorpha [8] and Nicotiana tabacum [9] were the first to be sequenced and within 19 years there were 45 complete plastome sequences for green plants (Viridiplantae) in GenBank [10]. Acceleration of efforts yielded 146 plastome sequences by the end of 2009, with many others in progress.

Complete genome sequences offer several advantages over restriction site maps and nucleotide sequences of targeted regions. In addition to substantially increasing the number of gene sequences for comparison, plastome sequences provide vast structural and evolutionary information that includes gene order, genome rearrangements, patterns of base pair composition, codon usage, mechanisms of gene duplication and gene loss (e.g., pseudogenization), patterns of nucleotide insertion and deletion (indels), and the occurrence of noncoding regions (such as plastome microsatellites and introns). We now have sufficient sampling from most major land plant clades to begin comparative analyses of their plastomes in earnest.

Despite considerable progress over the last 15 years in our understanding of green plant phylogeny, several nodes remain poorly resolved. One example is the monilophytes, which include five major lineages: leptosporangiate ferns, horsetails, marattioid ferns, ophioglossoid ferns and psilophytes. The monilophytes seem to be well-supported as a group, and are generally accepted as sister to seed plants [11–14]. But relationships among the major monilophyte lineages remain unclear. Here we present the complete plastome for the horsetail Equisetum arvense L., representing one of the last major monilophyte lineages to have a representative complete plastome available. For further resolution of land plant relationships and plastome evolution we also sequenced the plastome of Isoetes flaccida Shuttlw. ex A. Braun, the last of the three major lycophyte lineages to be sampled. We use data for 49 genes from 43 green plant taxa to infer phylogenetic relationships of land plants and compile information on the distribution of gene translocations, genomic inversions, gene content and indels to augment the phylogenetic signal from gene variation. We also examine codon usage and base composition. Patterns of plastome architecture are compared across early land plant lineages that diverged 400 to 500 million years ago.

Results and Discussion

Our analyses included representatives of all major land plant lineages as well as sampling of charophycean green algae for appropriate phylogenetic context. We do not provide detailed comparative plastome analyses for charophycean algae and seed plants because these are presented elsewhere: charophycean green algae [15, 16], gymnosperms [17–20] and angiosperms [21–23].

Plastome structures and composition

Gene maps for plastomes of Equisetum arvense and Isoetes flaccida are shown in Fig. 1. The complete plastome of E. arvense is 133,309 base pairs (bp) and includes a 93,542 bp large single-copy (LSC) region, a 19,469 bp small single-copy (SSC) region and two 10,149 bp inverted repeats (IRA and IRB). The complete plastome of I. flaccida is 145,303 bp and includes a 91,862 bp LSC, a 27,205 bp SSC and two 13,118 bp IRs. The overall G/C content is 33.36% for E. arvense and 37.94% for I. flaccida. Both annotated plastomes have been deposited in GenBank (Table 1).

Figure 1
figure 1

Gene maps of the Equisetum arvense and Isoetes flaccida plastomes. The inverted repeats (IRA and IRB) which separate the genome into the large (LSC) and small (SSC) single copy regions are indicated on the inner cycle along with the nucleotide content (G/C dark grey, A/T light grey). Genes shown on the outside of the outer circle are transcribed clockwise and those on the inside counter clockwise. Gene boxes are color coded by functional group as shown in the key. An asterisk (*) denotes genes with introns and a psi (ψ) denotes pseudogenes.

Full size image
Table 1 List of plastomes analyzed in this study.
Full size table

Gene content of Isoetes and Equisetum

Several minor differences in gene content between Equisetum arvense and Isoetes flaccida were identified (Table 2). We detected trnS-CGA only in E. arvense, and pseudogenes of rps16, tufA and an extra fragment of ndhB only in I. flaccida. Because of difference in the IR boundaries E. arvense has one copy of rps7 whereas I. flaccida has two copies. The position of the IR boundary also results in a small fragment of ycf2 in the IR of I. flaccida. The gene ycf66 is in both taxa, but appears to be pseudogenized in E. arvense. The genes accD, infA and rps2 are also in both taxa but appear to be pseudogenized in I. flaccida.

Table 2 List of plastid genes and pseudogenes annotated for Equisetum arvense and Isoetes flaccid a.
Full size table

clpP introns

In our newly described plastomes, clpP contains only one intron in Equisetum arvense and two introns in Isoetes flaccida. The distribution of introns in clpP is variable across plastomes sampled to date: all bryophytes have two introns, lycophytes have either two introns (Huperzia lucidula and I. flaccida) or one intron (Selaginella moellendorffii and S. uncinata), and all monilophytes except E. arvense have two introns. Among gymnosperms, Cycas taitungensis has two introns and other gymnosperms have either lost both introns (Pinus thunbergii, Ephedra equisetina and Welwitschia mirabilis) or have lost clpP entirely (Gnetum parvifolium). Most flowering plant clpP genes have two introns, except the monocot Agrostis calamus. In the charophycean algae, Chara vulgaris and Chaetosphaeridium globosum each have one clpP intron, whereas other charophycean green algae have no introns in clpP. It appears that the second clpP intron is unique to land plants, although there have been several lineage-specific losses (e.g., Selaginella spp., E. arvense, most gymnosperms, and at least one lineage of monocots).

tufA in land plant plastomes

The functional tufA gene is encoded in the plastid genome of most green algae but has been transferred to the nuclear genome in land plants and certain charophycean green algae [24–26]. A tufA-like fragment was identified in Anthoceros formosae, cycads and Gingko biloba plastomes situated between psbE and petL [20]. We found a 1,364-bp region between psbE and petL in Isoetes flaccida that shares 41% nucleotide similarity with the plastid encoded tufA gene found in Chara vulgaris. The I. flaccida tufA homologue lacks the expected start and stop codons and contains 29 internal stop codons within the reading frame. Furthermore, when this region was aligned with the C. vulgaris homologue, 38 indels were identified (Additional File 1). Taken together, the I. flaccida tufA-like region is most likely a pseudogene. These relictual fragments of a once functional plastid gene now identified in four distantly related land plant lineages (i.e., hornworts, lycophytes and two gymnosperms) offer an interesting opportunity to study patterns of plastome gene loss as well as the process of gene decay.

Putative RNA editing

Reliable inferences on RNA editing require comparisons of genomic sequence to that of mature mRNA or cDNA. However, a fraction of editing sites are in start or stop codons, or result in premature stop codons, thus providing a hint of RNA editing from genomic sequences. We found such evidence in two genes in Equisetum arvense and 28 genes in Isoetes flaccida (Table 2). The former had an undetermined start codon for rps11 and undetermined start and stop codons for accD. In I. flaccida we found 21 loci with undetermined start codons, two with undetermined stop codons, and seven internal stop codons. Based on the sequence of trnR-CCG, we also infer that the second position of the anticodon is RNA edited (U- >C).

ycf1

The region between chlN and rps15 in plastomes presents an interesting challenge. In certain taxa this region contains two putative protein-coding genes, both read from the same DNA strand, but in different reading frames. Other taxa, including Equisetum arvense and Isoetes flaccida, contain a single large open reading frame (orf) commonly annotated as ycf1. Recent interpretation of the interruption of ycf1 has been either to recognize the entire region as a ycf1 pseudogene [e.g., [27]], or to recognize the larger of the two orfs as ycf1 while leaving the smaller unannotated [e.g., [28]]. The function of these orfs (or orf) is unknown and there are at least two processes that explain the differences in putative orfs in this region of the plastome. One possibility is that the large orf could yield two separate protein products through a post-transcription edit that adds a stop codon. Another possibility is that the two-orf pattern is a result of a frame shift that can be read through during translation, yielding a single protein product [e.g., [29]]. However, transcripts of this region have been determined for Anthoceros formosae with two-orfs and Adiantum capillus-veneris with a single large orf [30, 31]. These transcript data were consistent with the original DNA annotation. That is, the reading frame was not 'corrected' in Anthoceros formosae and an internal stop codon was not detected in the larger orf in Adiantum capillus-veneris. Therefore at both the DNA and the RNA transcript levels there appears to be two alternative patterns.

Gene order

Comparative plastome maps for representative early diverging land plant lineages are shown in Fig. 2. The liverwort Marchantia polymorpha was not included in this figure because it shares similar gene order with the moss Syntrichia ruralis and the hornwort Anthoceros formosae.

Figure 2
figure 2

Comparison of plastome maps of early diverging land plant lineages showing major structural changes. Two mosses (Pp = Physcomitrella patens and Sr = Syntrichia ruralis), a hornwort (Af = Anthoceros formosae), four lycophytes (Hl = Huperzia lucidula, If = Isoetes flaccida, Sm = Selaginella moellendorffii and Su = Selaginella uncinata) and a monilophyte (Ea = Equisetum arvense) are compared. The liverwort Marchantia polymorpha shares a similar gene order with Sr, Af and Hl (not shown). Inverted repeat regions are depicted with grey boxes. Inversions and translocations are shown with arrows and solid or dashed lines respectively. The monilophyte-specific inversion shown for Ea is detailed in Fig. 3. Gene colors follow the key in Fig. 1.

Full size image

Bryophytes

Previous reports of plastomes representing exemplar bryophytes (liverwort Marchantia polymorpha, mosses Physcomitrella patens and Syntrichia ruralis, hornwort Anthoceros formosae) have shown that plastome content and gene order are highly conserved [8, 30, 32, 33]. This is true even in Aneura mirabilis, the only non-photosynthetic bryophyte [28]. One exception was reported in P. patens, involving a large inversion (~71 kb) encompassing nearly the entire LSC relative to a sample of four other mosses [33]. Goffinet et al.[34] surveyed the phylogenetic extent of the P. patens inversion and found it was restricted to the moss order Funariales. Oliver et al.[32] found that the moss S. ruralis does not have this large inversion but instead has the ancestral land plant plastome gene order.

Lycophytes

The plastomes of Isoetes flaccida and Huperzia lucidula (lycophytes) share similar gene order with bryophytes. However, we identified two unique structural changes in I. flaccida (Fig. 2). First, ycf2 in I. flaccida, which normally resides in the LSC in most land plant plastomes, has been translocated to the SSC with the 5' end now incorporated into the IR. Second, the chlL/chlN gene cluster has been inverted in I. flaccida and now resides adjacent to ycf2 rather than ycf1 as in H. lucidula. Neither the ycf2 translocation nor the chlL/chlN inversion occurs in either of the Selaginella plastomes.

Both Selaginella plastomes are considerably different in gene order from typical land plant plastomes (Fig. 2) [[35], Banks et al. personal communications]. A large region (~14-kb) has been translocated from the LSC to the IR/SSC in both plastomes, although the genes included in this translocation differ slightly. In both Selaginella plastomes, rps4 marks one endpoint of this translocated segment and this gene now resides in the IR. The other endpoint resides in the SSC and is marked by psbD in S. moellendorffii and by three additional genes (trnE-UUC, trnY-GUA and trnD-GUC) in S. uncinata. These tRNA genes remain in the LSC adjacent to ycf2 in S. moellendorffii, as is the case in most land plant plastomes. Several unique features were identified in the plastome of S. uncinata including a ~20-kb LSC inversion (psbI to rpoB-trnC-GAC), duplication of psbK and trnQ-UUG, and translocation of petN from the LSC to the SSC [35].

Euphyllophytes

Raubeson and Jansen [5] identified a ~30-kb LSC inversion marking the euphyllophytes as a monophyletic group to the exclusion of the lycophytes. This inversion is confirmed in all monilophyte (including Equisetum arvense, Fig. 2) and seed plant plastomes sequenced and does not occur thus far in bryophyte or lycophyte plastomes (including Isoetes flaccida, Fig. 2).

Monilophytes

Taking into account lineage specific gene losses and differences in IR gene content, Angiopteris evecta, Equisetum arvense and Psilotum nudum share identical gene order. Two large inversions associated with the IR and two smaller inversions in the LSC characterize the plastome of Adiantum capillus-veneris [36, 37]. The tree fern Alsophila spinulosa shares these inversions [38], indicating that they probably occurred in the ancestor to a clade that includes most extant leptosporangiate ferns. Comparative gene-order analyses identified one of the LSC inversions (involving trnG-GCC to trnT-GGU) in Psilotum nudum, Angiopteris evecta and leptosporangiate ferns [27, 38, 39]. Here we identified the same inversion in Equisetum arvense, further supporting the hypothesis that this inversion is shared by all extant monilophyte taxa and can serve as a reliable molecular marker for the monophyly of this lineage (Fig. 3). Neither of the two IR inversions nor the remaining LSC inversion found in Adiantum capillus-veneris and Alsophila spinulosa was identified in other monilophyte plastomes.

Figure 3
figure 3

Monilophyte-specific inversion. Detailed view of an inversion found only in monilophyte taxa (solid black lines and grey shadow) and a second inversion found only in Adiantum capillus-veneris and Alsophila spinulosa (dashed black lines), which together represent a large clade comprising about 90% of extant fern species diversity. Additional gene loss and movements are shown with arrows (SSC = gene translocated to small single copy). The lycophyte gene order and orientation in this region is similar to that of bryophytes and seed plants (not shown). Note that the ~5 kb region between trnE and trnY unique to Equisetum arvense contains no significant open reading frames when subjected to ORF Finder (http://www.ncbi.nlm.nih.gov/gorf/gorf.html). Gene colors follow the key in Fig. 1.

Full size image

The inverted repeat boundaries

Most plastome IRs have similar gene content that primarily includes rRNA and tRNA genes [16, 22, 40]. This is the case even in the ferns that have a reorganized IR [37, 38]. Variation from the typical IR gene content has been explained by movement ("ebb and flow") of the IR boundaries into, and out of, the LSC and SSC regions [41]. In our broad sample of land plant plastomes we found several taxa with unique IR boundaries that differ from the basic theme (Fig. 4). In some cases, distantly related taxa have very similar IR boundaries. The liverwort and the two mosses are identical in gene content at both ends of the IR. The two distantly related ferns are also identical to each other. Similarities at only one end of the IR are seen in a wider range of taxa. One end of the LSC is bounded by trnI-CAU in all seed-free land plants in our study, except Selaginella spp. (because they have lost many or their tRNA genes including trnI-CAU) and Angiopteris evecta. The SSC is bounded by ndhF at one end and chlL at the other, in most taxa. This suggests that whereas the ends of the IR clearly ebb and flow in some lineages, in other lineages they appear to be rather stable.

Figure 4
figure 4

Genes found in the inverted repeat of seed-free land plants. A liverwort (Mp = Marchantia polymorpha), two mosses (Pp = Physcomitrella patens and Sr = Syntrichia ruralis), a hornwort (Af = Anthoceros formosae), four lycophytes (Hl = Huperzia lucidula, If = Isoetes flaccida, Sm = Selaginella moellendorffii and Su = Selaginella uncinata) and five monilophytes (Ac = Adiantum capillus-veneris, Ae = Angiopteris evecta, As = Alsophila spinulosa, Ea = Equisetum arvense, Pn = Psilotum nudum) are compared for inverted repeat (IR) gene content, including new plastomes reported here. The figure is organized so that IRA, the small single copy (SSC) and IRB are presented top to bottom. Grey boxes denote genes found in the IR and white boxes denote genes found in the large single copy or SSC. Genes in square brackets are those partially encoded in the IR. Genes in bold emphasize that gene order within the leptosporangiate ferns have been reorganized relative to other plants [37, 38].

Full size image

Phylogenetic Analyses

We analysed four datasets that included 49 protein-coding genes and each was subjected to Maximum Likelihood (ML) and Bayesian Inference (BI). The datasets differed by the number of taxa included: one data set included all taxa listed in Table 1, another excluded both Selaginella species, another excluded all three gnetophyte taxa (Ephedra equisetina, Gnetum parvifolium and Welwitschia mirabilis) and the final excluded both Selaginella spp. and all gnetophyte taxa. Selaginella spp. and the gnetophytes were selectively included or excluded because of their unusual evolutionary patterns (e.g., plastomes with relatively high G/C content in Selaginella spp. and accelerated divergence rates in gnetophytes; [[18, 19, 35], Banks et al. personal communications], which can be problematic for phylogenetic reconstruction). Nearly identical topologies were recovered using nucleotide data regardless of analytical method (ML and BI) or taxon set (exclusion of Selaginella spp. and/or gnetophytes) with two exceptions discussed below. A summary including alternate topologies and support values is shown in Fig. 5A-C. Phylogenetic results for angiosperms are shown in Additional file 2.

Figure 5
figure 5

Phylogenetic results using nucleotide data. A) Cladogram based on ML analysis (-ln = 473022.24372) of 49 gene sequences from all 43 plastomes sampled (Table 1). Mapped non-homoplastic indels are shown in yellow circles. Numbers in grey circles are indels identified with a more restrictive taxon sampling that were difficult to score across all charophycean green algae and land plants. B) Phylogram based on ML analysis (-ln = 442145.83514) of gene sequences, excluding the two Selaginella species. C) Phylogram based on ML analysis (-ln = 436696.62372) of gene sequences, excluding the three gnetophyte taxa (Ephedra equisetina, Gnetum parvifolium and Welwitschia mirabilis). ML phylogenetic results (-ln = 405904.24872) that excluded Selaginella spp. and gnetophytes were similar in topology to b (not shown). Numbers above the branches are ML bootstrap proportions and numbers below are Bayesian posterior probabilities in this order: all taxa included/Selaginella spp. excluded/gnetophytes excluded/Selaginella spp. and gnetophytes excluded. The asterisk (*) in c indicates the ML bootstrap proportion for Marchantia polymorpha sister to the remaining land plants found in bootstrap analyses that included all taxa, though the best ML tree recovered a sister relationship of M. polymorpha with the mosses Physcomitrella patens and Syntrichia ruralis. Boxes and dashed lines indicate the relative position of monilophytes (light grey) and lycophytes (dark grey).

Full size image

Phylogenetic relationships among the major bryophyte lineages have been controversial with nearly every possible topology being reported, often with strong support. However, recent convergence in results supports hornworts as sister to vascular plants (as seen here) and liverworts sister to all other land plants [[7, 42] and references therein]. Monophyly of the bryophytes has also been proposed using plastome gene data. However this relationship was only found when inferred amino acid sequences or protein plus DNA sequences corrected for A-T bias were analyzed [[15, 43, 44], Additional file 3]. A sister relationship between mosses and liverworts was proposed previously based on morphological and molecular data. Renzaglia and Garbary [45] concluded that characters related to sperm cell development were compelling evidence for the monophyly of mosses plus liverworts, a clade they referred to as Setaphytes. In a recent analysis that included seven genes from all three DNA compartments (plastid, mitochondria and nuclear) for 192 land plant and green algal taxa, the liverworts were found with strong support as sister to all remaining land plants [12].

In the nucleotide analyses presented here, regardless of taxon set or analytical method, monophyly of land plants was confirmed, as was the sister relationship between the hornwort Anthoceros formosae and tracheophytes (e.g., lycophytes, monilophytes, gymnosperms and angiosperms; Fig. 5). More problematic is the position of liverworts among bryophytes as our results show two taxon-dependent placements of the liverwort Marchantia polymorpha. In ML and BI analyses that included all taxa or excluded Selaginella spp., with or without gnetophytes, M. polymorpha was sister to the mosses Physcomitrella patens and Syntrichia ruralis (Fig. 5A, 5B). The alternative topology with liverworts sister to the remaining land plants was supported in the present study only when the gnetophytes were excluded (Fig. 5C). ML bootstrap proportions were generally weak for this part of the tree regardless of taxon exclusion and BI posterior probabilities were low with the analysis that included all taxa and the analysis that only excluded the gnetophytes. Increased taxon sampling is critical in resolving the pattern of these ancient divergences.

The lycophytes (Huperzia lucidula, Isoetes flaccida, with or without Selaginella spp.) and monilophytes (Adiantum capillus-veneris, Alsophila spinulosa, Angiopteris evecta, Equisetum arvense and Psilotum nudum) were each strongly supported monophyletic groups (Fig. 5). However, the relationships of these two clades to each other and to seed plants varied across our analyses. The monilophytes were sister to the seed plants only when Selaginella spp. were excluded, with or without gnetophytes (Fig. 5B). Alternatively, the lycophytes were supported (weakly in most analyses) as sister to seed plants when Selaginella spp. were included (Fig. 5A, 5C). Strong support was recovered for the monilophyte + seed plant relationship by Qiu et al.[12] and this relationship is further supported by the large plastid LSC inversion identified in monilophytes and seed plants and not found in lycophytes (Fig. 2 discussed above) [5, 39]. Within the lycophyte clade, the homosporous H. lucidula was consistently sister to the heterosporous taxa I. flaccida and Selaginella spp. (or sister to just I. flaccida in analyses where Selaginella spp. were excluded).

Strongly supported resolution among the major monilophyte lineages has been elusive, and it was not clear if lack of support was a function of insufficient phylogenetic signal, rapid radiation that is indistinguishable from a true polytomy, or ancient extinctions resulting in long-branch taxa that make phylogenetic reconstruction challenging. Our analyses that used 43,491 nucleotide characters (and several subpartitions of these) revealed a moderately supported sister relationship between Psilotum nudum and Equisetum arvense. Although these groups are usually not found as sister lineages, this topology is reasonable given the lack of sampling of sister taxa to both genera. More typically, P. nudum is found in a lineage that is sister to ophioglossoid ferns [12, 13, 46], a lineage not sampled in this study, and these two lineages are found sister to the remaining monilophytes (marattioid ferns, horsetails and leptosporangiate ferns). However, a sister relationship between Psilotaceae and Equisetaceae was recently reported in a phylogenetic analysis of morphological characters across vascular plants where five unambiguous character state changes for the two groups were identified [47]. These were primarily characters associated with leaves and organization of the shoot system. Schneider et al.[47] did not find the same relationship from an analysis of the same taxa using nucleotide sequences, and they cautioned that the morphological evidence for this sister relationship could well be a function of homoplasy and structural simplification reverting to a pleisiomorphic state, thus causing "concern for a possible misleading bias." Failure of so many studies to resolve relationships among the main lineages of monilophytes is consistent with a rapid radiation (i.e., very short internal branches) rather than lack of phylogenetic signal. Additional data from Tmesipteris (a close relative of Psilotum), ophioglossoid ferns, and additional Equisetum species may help to resolve these nodes.

Some of the deep branching events in the land plant tree are still poorly resolved even with full plastome sequence data now available from most of the major clades. It is likely this is because the intervals of shared history among these major clades were relatively short as compared to their subsequent separate histories, giving rise to extreme asymmetry in branch lengths which causes known problems with inferring the correct tree - particularly using data with a small number of character states such as nucleic acid sequence data [48]. Rare genome structural changes such as gene-order rearrangements have the potential for resolving such short internal branches [49]. Indeed a phylogenetic analysis of plastome gene-order data found liverworts sister to all other land plants and monilophytes sister to the seed plants [50]. However, in that study the lycophytes were not resolved as a monophyletic group as they are here. This topological convergence suggests that a future combined analysis with nucleotide, structural and morphological data may be able to resolve some of the short internal branches.

Indels

Many genes contained three base-pair indels with apparent phylogenetic signal. We examined all aligned gene sequences for potential indels across the entire taxon sample, scoring characters as 1 for presence or 0 for absence of a sequence stretch. We scored 152 indels (see Additional files 4 and 5); 40 support the phylogenetic tree (synapomorphic), 99 are unique to a terminal taxon (autapomorphic) and 13 contradict the tree (homoplastic). The synapomorphic and autapomorphic indels are presented in Fig. 5A; the 13 homoplastic indels are not mapped. Indels alone, analyzed by MP, did not provide a well-resolved tree (not shown). However, several nodes were supported by uncontradicted indels; land plants (2 indels), lycophytes (2 indels), Selaginella spp. + Isoetes flaccida (1 indel), Selaginella spp. (12 indels), Angiopteris evecta + Adiantum capillus-veneris + Alsophila spinulosa (1 indel), seed plants (1 indel), gnetophytes (11 indels) and angiosperms (1 indel). Angiosperm-specific indels are shown in Additional file 2. Interpretation of indels is dependent on taxon sampling. When subsets of the data were examined more closely we could score additional indels that were difficult to interpret at the broader level. Several such indels were identified; focusing on the monilophytes we found two indels (in atpA and rps4) uniting the monilophytes; the rps4 indel has been reported previously [11]. Five indels in psbT, petG, atpA, rpl16 and rpoB uniting Adiantum capillus-veneris + Alsophila spinulosa, and one in rpl16 uniting Equisetum arvense and Psilotum nudum were also identified (Fig. 5A).

Nucleotide composition and codon usage

Codon usage bias is often correlated with genomic nucleotide composition bias but may also be correlated with gene expression, selection and drift [51, 52]. The overall G/C content of the genes used in our phylogenetic analyses ranged from 34% in Marchantia polymorpha to 54% in Selaginella uncinata. Both S. uncinata and S. moellendorffii were remarkable for their high (greater than 50%) G/C content. Among land plants included in this study, only M. polymorpha and the mosses contain less than 35% G/C. The nucleotide composition bias found in many plastomes may result in codon usage biases as well.

The effective number of codons (ENc) is often used as a measure of the amount of codon bias within a gene or genome [53] with values ranging from 20 (very biased) to 61 (very unbiased). Most vascular plants in our study had a value between 46 and 50 (Table 1). Less bias was detected in Adiantum capillus-veneris (ENc = 55.12), Alsophila spinulosa (52.99), Selaginella uncinata (58.30) and S. moellendorffii (57.73). More bias was found in bryophytes and charophycean algae, notably Marchantia polymorpha (35.25), Physcomitrella patens (36.12) and Syntrichia ruralis (35.46). These differences may have ramifications for the estimation of phylogenetic relationships among these groups.

Conclusions

This study illustrates the advantage of whole plastome sequences for studies of plastome architecture and land plant evolution. Nucleotide sequence data from targeted regions have contributed greatly to our understanding of plant phylogeny. However short sequence data can have limitations because they can miss important aspects of genome structure, and nucleotide substitution rates are not always applicable to solving specific evolutionary questions. Restriction site mapping also is limiting because it usually requires a series of cross comparisons every time a taxon is added. Complete plastome sequences provide data that can be used for nucleotide-based analyses, but also can add to a growing database of structural genomic data, including inversions, gene content, and indels. New taxa can be added relatively easily without additional experimental comparisons. Also, changes that might first appear to characterize a large clade might turn out to be more restricted once additional taxa are sampled. For example, some of the unique structural attributes first identified in Selaginella uncinata plastome [35] appear to be unique (so far), whereas others are shared with a second plastome sequence of S. moellendorffii (Banks et al. personal communications). Another important aspect of this type of data is that changes that first appear autapomorphic in small datasets might later emerge as key synapomorphies as additional taxa are sampled. This is illustrated by the series of large inversions in the IR first identified in Adiantum capillus-veneris [36, 37] but later shown to be shared with the distantly related tree fern Alsophila spinulosa [38], and by inference shared among more than 90% of leptosporangiate fern species.

Most of the currently available plastome sequences are from seed plants, with very few available from the presumed sister clade, the monilophytes. Here we have added an additional monilophyte plastome sequence as well as a lycophyte plastome sequence that represent two critical lineages that were not previously sampled (i.e., Equisetaceae and Isoetaceae). We examine multiple aspects of these plastomes to reveal patterns of evolution across land plants. We demonstrate that comparative plastome analysis can provide valuable information about evolutionary processes at the nucleotide, gene and genome scale in early land plants. These data are phylogenetically informative at many levels. Gene rearrangement and indels are informative at deep nodes. Plastome sequences also provide valuable nucleotide data that can be used in studies of ancient and more recent divergences.

Now that plastome sequence data are available for most major lineages of early land plants additional sampling can focus on resolution of poorly supported nodes by increased taxon sampling within these lineages. Although we have one or two representatives of each major group, these often represent lineages containing hundreds or thousands of species. Additional sampling within species-rich lineages may resolve some areas of phylogenetic conflict. As DNA sequencing costs continue to decrease it will become possible to improve sampling and gain further insight into early land plant evolution as well as the patterns and processes that shape the evolution of plastomes.

Methods

Taxa

We compared and analyzed complete plastid genomes of 43 taxa, including 37 land plants and six charophycean algae (Table 1). Two of the land plant plastomes are new to this study (herbarium vouchers at WTU): Equisetum arvense and Isoetes flaccida. For E. arvense and I. flaccida, tissue preparation, plastid isolation, DNA extraction, cloning, sequencing, and assembly followed Wolf et al.[39]. Annotation was carried out with DOGMA [54] and tRNAscan [55].

Plastome analyses

Gene maps for representative plastomes were drawn with OGDraw and compared for gene content, gene order and IR boundaries [56]. Codon usage and nucleotide frequencies were determined using CodonW [57].

DNA alignment

Nucleotide sequences for the 49 protein-coding genes found in all study taxa were extracted from DOGMA-annotated plastomes or from those found in GenBank (Table 1). The hornwort Anthoceros formosae and the fern Adiantum capillus-veneris both have extensive RNA editing [30, 36, 37]. Therefore, we used the cDNA sequences rather than genomic sequences for phylogenetic analysis. Each gene was aligned with MacClade v4 [58] and the resulting individual gene alignments were assembled into a single data matrix. A Nexus block was written that identified all unalignable DNA regions, overlapping gene regions, and inferred RNA edited stop codons for exclusion in subsequent phylogenetic analyses. Indels were scored in a separate data matrix as binary characters.

The resulting 49-gene alignment including all codon positions (see Additional file 5) was partitioned into four data sets, 1) one that included all 43 taxa and 30,018 characters, 2) one that excluded both Selaginella species (41 taxa and 29,961 characters), 3) one that excluded the gnetophytes Ephedra equisetina, Gnetum parvifolium and Welwitschia mirabilis (40 taxa and 29,871 characters), and 4) one that excluded both Selaginella species and all gnetophytes (38 taxa and 29,814 characters). Altering the taxon set changes the number of characters because of lineage specific indels. Nucleotide alignments were translated into amino acids using MacClade assuming the universal code [58]. Ambiguous amino acids resulting from polymorphic nucleotides were treated as missing.

Phylogenetic Analyses

Maximum Likelihood (ML) and Bayesian Inference (BI) analyses were performed on each of the four nucleotide data sets using PAUP* v4.0b10 [59] and MrBayes v3.1.2 [60–62], respectively. MrModeltest v2 [63] was used to determine the best fitting model of DNA substitution. With the best fitting model (GTR+I+Γ with four rate catagories) an iterative procedure described in Swofford et al.[64] was used to converge on the best fitting model parameters. The resulting model parameters were then fixed for ML analyses and a heuristic search was performed with random taxon addition using TBR branch swapping. Two hundred bootstrap replicates [65] were performed for each data set, each replicate with a single random taxon addition and NNI branch swapping. For BI analyses (also using GTR+I+Γ) two runs each with four markov chains running five million generations were performed. The heated markov chain was sampled every 100th generation. The -ln likelihood scores were plotted against generation number and all samples collected prior to the markov chain reaching stationarity were discarded as burn-in. The remaining samples were summarized using the sumt command in MrBayes. The insertion/deletion matrix was analyzed using the maximum parsimony (MP) criterion in PAUP* as unordered characters with 10 random taxon additions and 200 bootstrap replicates. BI analyses of amino acid data sets were executed in a fashion similar to the nucleotide analyses. We used a fixed model, cpREV+I+Γ, estimated specifically for phylogenetic estimation of plastid-encoded proteins [66]. ML bootstrap analyses of amino acid data were performed in RAxML also using cpREV+I+Γ [67].