Introduction

The trnT–trnF region is located in the large single-copy region of the chloroplast (cp) genome. Together with its flanking genes (Fig. 1A), it forms a cistron that is conserved from mosses to seed plants (Taberlet et al. 1991). Even where chloroplast genomes have undergone extreme structural changes, such as in Pinus thunbergii (Wakasugi et al. 1994) (Fig. 1B), which has pseudogenized ndhF genes located immediately downstream from the trnT–trnF region, the region itself is conserved. Recent large-scale analyses have shown that the cistron comprised of the three tRNA genes is a uniquely shared characteristic of land plants (Quandt et al., 2004); in green algae, the trnF gene is located farther away from the other two genes. The trnT and trnL genes are lost, and the trnL–trnF region is disrupted, in a few nonphotosynthetic flowering plants, such as Epifagus virginiana (Orobanchaceae [Wolfe et al. 1992]) (Fig. 1C).

Figure 1
figure 1

Organization of the trnT–trnL–trnF cistron in the large single-copy region of the chloroplast genome. Open arrows indicate postulated directions of transcription; filled arrows, orientation of genes. Primers used to amplify the trnT–trnF region are shown as half-arrows. A Map representing gene order in mosses, ferns, and angiosperms. Genes flanking the trnT–trnF region are conserved across Marchantia (liverwort), Anthoceros (hornwort), Psilotum (whisk fern), and the angiosperm genera Amborella, Calycanthus, Oryza sativa, Triticum aestivum, Zea mays, Lotus japonica, Spinacia oleracea, Nicotiana tabacum, Atropa belladonna, and Arabidopsis thaliana. B Map of Pinus thunbergii (black pine). Note the pseudogenized ndhK and ndhC and the lack of ndhJ compared to mosses, ferns, and angiosperms. C Map of Epifagus virginiana, a nonphotosynthetic angiosperm, modified from Wolfe et al. (1992). Note the loss of trnT, trnL, and ndh genes and the pseudogenization of trnS, atpB, and rbcL.

Full size image

Because of its conserved gene order, its high variability, and the adequate size of its intergenic spacer (IGS) and intron, the trnT–trnF region is widely sequenced for phylogenetic analyses at the species and genus level in seed plants (Taberlet et al. 1991; Bakker et al. 2000; Borsch et al. 2003; Quandt et al., 2004; Shaw et al. 2005). Throughout green plants, tRNA genes are either monomeric, with putative promoter elements located upstream of the genes, or polymeric, containing other tRNAs, rRNAs, and mRNAs (Martin 1995). Examples of monomeric tRNAs are tRNASer (Gruissem et al. 1986; Wu et al. 1997) and tRNA Met2 (Gruissem and Zurawski 1985). Examples of polymeric tRNAs are tRNAArg, which is part of an operon encompassing 5′-16S-tRNAIle-tRNAAla-23S-4.5S-5S-tRNAArg-3′ and is located within the inverted repeat region (Leal-Klevezas et al. 2000), and the trnE operon of tRNAGlu, tRNATyr, and tRNAAsp, which is located in the large single-copy region. The transcripts of such operons are processed to release the individual tRNA molecules (Ohme et al. 1985).

Chloroplast genes, such as the tRNA genes, usually contain upstream and downstream signals facilitating their expression. Termination of transcription is signaled by hairpin structures and T-rich regions downstream of the operon (Bogorad 1991). Steinmetz et al. (1983) compared the 5′-flanking regions of the tRNA Ser3 , tRNAPhe, tRNA Thr2 , tRNA Leu2 , and tRNA Metm genes and discovered nucleotide stretches that resembled the “−35” and “−10” promoter regions of bacterial genes, probably because of the bacterial origin of the chloroplast (see also Hanley-Bowdoin and Chua 1987; Bogorad 1991; Inokuchi and Yamao 1995). Analysis of these conserved regions together with ones seen in other plastid genes yields the consensus sequences ATTGANA at “−35” and TAAGAT at “−10.” A study of the function of these prokaryote-type promoter regions in psbA, rbcL, and atpB showed that they are essential for the proper expression of these genes (Gruissem and Zurawski 1985). Other tRNA genes, however, lack prokaryote-type promoter elements and instead have internal promoter elements that are highly conserved (Galli et al. 1981; Gruissem et al. 1986; Cheng et al. 1997). Putative promoter elements upstream of trnT as well as trnF (Steinmetz et al. 1983; Kanno and Hirai 1993; van Ham et al. 1994) suggest that these genes can be expressed independently, although no experimental study has been done. Independent expression is also indicated by the opposite orientation of the tRNAThr(UGU) gene relative to the tRNALeu(UAA) and tRNAPhe(GAA) genes (Figs. 1A and B). This suggests that the tRNA genes may comprise two operons, one consisting of the trnT and rps4 genes and the other of trnL and trnF. Chloroplast tRNA genes or operons are transcribed with 5′ leader and 3′ trailer sequences, and the precursor-tRNAs are posttranscriptionally modified to form mature tRNAs. The processing events consist of 5′ end maturation, formation of a 3′ end, CCA addition, intron removal, and base modification (Martin 1995). The endonucleolytic cleavage of the 5′ leader sequence and the 3′ trailer sequence has been attributed, respectively, to endonucleases RNase P (Gegenheimer 1996; Frank and Pace 1998) and RNase Z (Frank and Pace 1998; Schiffer et al. 2001, 2002). The spacers are not needed for the processing of tRNAs, leading Bakker et al. (2000) to the conclusion that they may evolve neutrally. By contrast, the group I intron contained in the trnL gene maintains its secondary structure and sequence elements critical for its self-splicing from precursor-RNAs (Cech 1988; Michel and Westhof 1990; Simon et al. 2003). These sequence elements are conserved from eubacteria to chloroplasts (Kuhsel et al. 1990; Xu et al. 1990; Besendahl et al. 2000).

Here we analyze the cp trnT–trnF region in the three genera of Gnetales, an ancient lineage of seed plants for which well-supported species-level phylogenies are available (Ickert-Bond and Wojciechowski 2004; Huang et al. 2005; Won and Renner 2005a). The sequence characteristics discovered, together with the typical function of enzymatic systems involved in cp tRNA processing, lead us to suggest a model for the expression and processing of the genes tRNAThr(UGU), tRNALeu(UAA), and tRNAPhe(GAA).

Materials and Methods

Taxon Sampling

The samples, voucher information, and GenBank accession numbers obtained in this study are listed in Appendix 1 (Supplementary Material). Most samples were collected in the field by H.W., but a few are from botanical gardens or herbarium material.

Gene Sequencing

We extracted DNA from silica gel-dried leaves, using QIAGEN Plant DNeasy minikits. Concentration and quality of extracted DNAs were checked by 1% agarose gel electrophoresis with a (λHindIII/EcoRI size marker. The chloroplast tRNALeu(UAA) intron and adjacent IGSs were amplified using primers designed by Taberlet et al. (1991) (Fig. 1 shows their placement). The PCR reaction volume was 25 μl and contained 2.5 μl of Taq 10× reaction buffer, 1.5 mM MgCl2, 0.4 mM primer, a 0.2 mM concentration of each dNTP, 0.625 unit of Taq DNA polymerase (Promega), and 0.5–2 μg of template DNA. PCR cycles consisted of 30 cycles of 1 min at 97°C for template denaturation, 1 min at 53°C for primer annealing, and 1 min 20 s at 72°C for primer extension, followed by 7 min at 72°C for completion of primer extension. After electrophoresis of PCR products in 1% agarose gels, PCR products were purified with QIAquick PCR product extraction kits (QIAGEN) and directly sequenced using the ABI Prism Big Dye Terminator Cycle Sequencing Ready Reaction Kit (Perkin Elmer). Sequencing reactions were electrophoresed in an ABI 377 sequencer (Perkin Elmer). We were able to sequence/read both strands except for a stretch in the middle of the trnL–F spacer, where signals after the repeat of ‘AT’ started to overlap and fade out. This region was thus mainly read in only one direction. The same problem occurred after the ‘A’ and ‘G’ repeats in the middle of the Gnetum trnL intron. Sequences were aligned and cleaned up using Seqman II (DNASTAR).

Sequence Analysis

We obtained 41 complete sequences from about 25 species of Gnetum, 1 sequence from Welwitschia, and 3 from Ephedra (Appendix 1, Table 1). For Ephedra, we only sequenced the trnT–trnL IGS and the trnL intron. A data matrix for sequences of each chloroplast region was constructed using Se-Al software (version 2.0a11; http://evolve.zoo.ox.ac.uk/). Length and G+C content were calculated from the aligned sequences. Sequence divergences were calculated using Kimura’s (1980) two-parameter model. Chloroplast tRNALeu(UAA) intron and adjacent IGS sequences were compared to the sequences of tRNALeu(UGU), tRNAThr (UAA), and tRNAPhe(GAA) of Pinus thunbergii, the chloroplast of which has been sequenced (Wakasugi et al. 1994), to determine the boundaries of the genes, intron, and spacer. To predict the secondary structure of the region, especially of palindromic (inverted repeated) sequences found in the trnL–trnF IGS, we used MFOLD (version 3.1; http://www.bioinfo.rpi.edu/applications/mfold/ [Zuker 2003]), which estimates the secondary structure of DNA by thermodynamics. To detect promoter elements of the tRNA genes, we examined the IGS sequences for “–35” and “–10” promoter element homologies. The data matrices of Borsch et al. (2003) and Quandt et al. (2004), which comprise trnL intron and spacer sequences of numerous land plants, were also compared to that of the Gnetales sequenced here.

Table 1 Length variation (bp) and G+C content (percentage) of the chloroplast trnT–trnL IGS, trnL intron, and trnL–trnF IGS in Gnetales
Full size table

Phylogenetic Analysis

The flanking spacers of the tRNALeu gene showed sufficient variation in length among the three genera to make alignment difficult (see Results). We therefore designated the South American Gnetum group as a functional outgroup for the intra-Gnetum analysis, based on results of a phylogenetic analysis of Gnetales rbcL and matK sequences (Won and Renner 2003). We used PAUP* 4.0b10 (Swofford 2002) and Mr. Bayes (version 3.0b4 [Huelsenbeck and Ronquist 2001]) for phylogenetic analyses. The data matrix of the trnL intron and trnT–trnL and trnL–trnF spacers was analyzed by parsimony, neighbor-joining, and Bayesian inference. Parsimony analyses were run using heuristic searching, with 100 random taxon-addition replicates, holding 100 trees at each step, tree-bisection-reconnection (TBR) branch swapping, the MulTrees, Collapse, and Steepest Descent options, and no upper limit for trees held in memory. Neighbor joining used K-2-P distances and ignored gaps. Bayesian probabilities were obtained under the GTR + Γ + Pinv model, with four Markov chain Monte Carlo chains run for 3 million generations, using random trees as starting point, and sampling every 10,000th generation. The trees sampled before the saturation of maximum likelihood estimates were discarded as burn-in. Nonparametric bootstrap support for parsimony and neighbor joining was obtained by resampling the data 1000 times with the same search options and model.

Results

Length and G+C Contents

The length of the trnL intron was 319–384 bp; that of the trnT–trnL spacer, 292–340 bp; and that of the trnL–trnF spacer, 132–169 bp (Table 1). These sizes are similar to Welwitschia (317, 313, and 250 bp) but significantly longer than Ephedra (290, 156, and 112 bp), especially in the trnT–trnL spacer. Of the multiple accessions of Gnetum, only the four of G. gnemon showed distinct variation in length (one was excluded from the data matrix). Two of them, Suksathan s.n. and RBGE19480112, had shorter spacer sequences due to autapomorphic gaps at base positions 355–365 of the trnT–trnL spacer and positions 1104–1113 of the trnL–trnF spacer compared to the other two accessions, Won 514 and Won 568.

G+C contents of the trnL intron and spacers showed opposite trends among the three genera (Table 1). In the trnL intron, Gnetum had a higher G+C content than Welwitschia (34.8–39.3% vs. 30.4%) but was similar to Ephedra (36.6–37.2%). In the trnT–trnL and trnL–trnF spacers, Gnetum was comparable to Welwitschia (30.4–34.2% vs. 30.4% and 33.7–45.7% vs. 33.2%, respectively) but had a higher G+C content than Ephedra (30.4–34.2% vs. 28.8% and 33.7–45.7% vs. 26.7%, respectively).

Sequence Comparison, Alignment, and Divergence

Due to the large length and sequence variation, we could not align the spacer sequences of three genera, nor could we find homologous motifs in their spacer regions. Like Borsch et al. (2003) and Quandt et al. (2004), we failed to detect specific homologies of Gnetales with other seed plant sequences. The trnL intron sequences, by contrast, contained several highly conserved motifs (Fig. 2), especially the P, Q, R, and S sequence elements and the P1, P2, and P3 elements (Figs. 2 and 3A). The P5, P6, P8, and P9 structural elements were variable. Comparison of gnetalean sequences with those of other land plants also revealed the expected conserved tRNALeu group I intron structure. By comparing our data matrix with the basal angiosperm matrix of Borsch et al. (2003), we detected a highly conserved “–35”-like element (TTGACA) and “–10”-like element (TAGGAT) between base positions −45 and −16 in the upstream of the tRNAPhe gene. However, we could not detect any promoter elements in the trnT–trnL or trnL–trnF spacer regions of Gnetales.

Figure 2
figure 2

A Sequence alignment of the chloroplast tRNALeu(UAA) gene and intron (a group I intron). Open boxes indicate the tRNALeu 5′ and 3′ coding regions and shaded boxes indicate conserved sequence elements (P, Q, R, S) in the intron. Nine stem-and-loop structures (P1–P9) are marked by brackets. See Cech (1988) for the secondary structure nomenclature of group I introns. Sequences shown are those of Gnetum africanum, Welwitschia mirabilis, and Ephedra trifurca. B Secondary structure model of the chloroplast trnL intron in Gnetales. Intron splicing sites are marked with two arrows. Locations of the A8-13G4-14 motif in the P6 element, the 39-bp indel in the P8 element, and the 44-bp indel in the P9 element of Gnetum are indicated. Boxed bases indicate variation within Gnetum; shaded bases, variation among the three genera of Gnetales. Variable bases are in boldface. The secondary structure model is modified from Cech et al. (1994) and Quandt et al. (2004)

Full size image
Figure 3
figure 3

Secondary structure model for Gnetum chloroplast tRNAs and inferred mode of tRNA transcription and processing. A Proposed secondary structure for Gnetum chloroplast tRNAThr(UGU), tRNALeu(UAA), and tRNAPhe(GAA). Solid lines indicate transcribed regions; dashed lines indicate nontranscribed regions upstream or downstream of the tRNA transcription start/termination points. Filled arrows indicate the transcription start/termination point; open arrows indicate additional putative transcription termination/start points for the independent transcription model. The secondary structure model for trnL intron is modified from that of Kuhsel et al. (1990).The structure of the rps4 gene is arbitrary. B Proposed transcription of Gnetum trnT, trnL, and trnF genes. TrnT and rps4 are transcribed in the opposite direction from trnL and trnF. Both trnT and rps4, and trnL and trnF, could be cotranscribed or independently transcribed. C Proposed order of tRNA processing.

Full size image

Alignment of the trnL intron and IGSs of Gnetum required 1-, 4-, 5-, 6-, 8-, 9-, 10-, 15-, 39-, 40-, and 44-bp-long noninformative indels and 11 informative indels (Appendix 2). The P8 and P9 structural elements of G. paniculatum and G. africanum contained 39- and 44-bp-long indels (Fig. 2B), and the trnT–trnL IGS region of G. aff. latifolium Beaman 9408 and G. klossi contained 40- and 15-bp-long indels. The P6 structural element of the intron also contained an A8-13G4-14 motif with varying repeat numbers of ‘A’ and ‘G’ nucleotides (Fig. 2B), and the trnL–trnF spacer a G6-16 motif at base positions 977–994 and palindromic sequences with AT repeats at base positions 1008–1061 (Fig. 3A; see Palindromic Structures in the trnL–trnF IGS Region, below). There was intraspecific variation in these mono- and dinucleotide repeats.

Sequence divergences among the gnetalean genera and among Gnetum species are specified in Appendix 3).

Palindromic Structures in the trnL–trnF IGS Region

The palindromic sequences in the trnL–trnF spacer were first identified by sequence comparison and then verified by looking at the secondary structure predicted by MFOLD. They are listed in Table 2. Palindromes were found in all Gnetum sequences. They formed hairpin structures, with tandem repeats of ‘AT’ making up the stem of the hairpin (Fig. 3A). The secondary structure of the trnL–trnF spacer suggests that the G6-16 motif is also involved in a conserved hairpin immediately upstream of the palindromic sequences, with UCGGGGGG//AGCUCCUU forming a stem (Fig. 3A). Prediction of the secondary structure of the Welwitschia trnL–trnF spacer detected several hairpins, although the sequence itself does not have palindromic sequences. Palindromic structures found in Gnetum coincided with geographical clades; Asian and African palindromic sequences started with AAGTT and ended in AACTT, which form a stem by complementary base-pairing. The AAGTT motif was conserved throughout Gnetum, while the AACTT motif was absent from the South American species (Table 2; Fig. 4, gap I). However, although the AAGTT motif was present in South America, it was not involved in hairpin formation there. Asian species had AT repeats except for G. gnemonoides, which had AT plus AAAT repeats, perhaps as a result of mutation from ATAT, with compensatory downstream changes of ATAT to ATTT. Gnetum africanum had two consecutive palindromic structures, referred to as “doubly palindromic” in Table 2, separated by five bases (TTTTG). The length of the palindromes 5′-AAGTT-(AT) n -AACTT-3′ ranged from 18 bp (n = 4) to 30 bp (n = 10), and there were several cases with substitutions in the AT repeats (Table 2). The C’s and G’s in the middle of AT repeats of the South American and African sequences probably result from compensatory base changes in the stem region of hairpin structures.

Table 2 Palindromic sequences in the Gnetum chloroplast tRNALeu-tRNAPhe IGS region
Full size table
Figure 4
figure 4

Phylogeny of Gnetum obtained from Bayesian analysis of the chloroplast trnL intron and spacer sequences. Values next to nodes are Bayesian posterior probabilities, followed by bootstrap values under parsimony (CI = 0.55, RI = 0.45, excluding uninformative characters) and neighbor joining. Base positions 142–181, 246–260, 297–301, 576–606, 706–745, 810–853, 976–993, and 1012–1070 were excluded from the analyses and 11 informative gaps (A–K) were binary-coded (see Appendix 2).

Full size image

Phylogenetic Analyses

Alignment of chloroplast trnL intron and adjacent spacer sequences required 1 to 44 bp indels as described above; number of aligned characters, parsimony-informative sites, and gaps are shown in Table 3, and the data file has been submitted to TreeBase. Sequences of Gnetum could not be aligned with those of Welwitschia and Ephedra except in the trnL intron region (Fig. 2). One of the tree topologies estimated by Bayesian analysis (in four repeat runs) of the complete intron and spacer sequences is shown in Fig. 4. Parsimony analysis yielded 71 trees on three islands. Accessions sorted geographically into a South American clade, an African clade, and two Asian clades (labeled Asia I and Asia II in Fig. 4) that were well supported except for the monophyly of the Asian group. Within the Asia II clade, two subclades were resolved, a G. cuspidatum clade and a G. hainanense s.l. clade (Fig. 4). Except for two homoplasious gaps marked A and G in Fig. 4, gaps (B–K) were informative. A neighbor-joining tree from just the trnL intron sequences also identified the South American species as the basal branch and sister to the Asian and African clades (data not shown).

Table 3 Number of characters, parsimony-informative sites, and phylogenetically informative indels in the chloroplast tRNALeu intron and intergenic spacers in Gnetum
Full size table

Discussion

Behavior of the trnL Intron and Adjacent Spacers in Gnetum

Topologies obtained from the trnL sequences were congruent with phylogenies from separate and combined analyses of nuclear sequences (rRNA Internal Transcribed Spacer [ITS] and the second intron of LEAFY), chloroplast matK and rbcL genes, and the conserved region of the mitochondrial nad1 second intron and partial exon sequences (Won and Renner 2005a, b). This is in agreement with other studies that show that fast evolving regions can sometimes provide cost-efficient phylogeny reconstruction even of deep divergences (Hershkowitz and Lewis 1996; Källersjö et al. 1998; Borsch et al. 2003; Hilu et al. 2003; Quandt et al. 2004). Based on fossil-calibrated molecular clocks, the split between Gnetum and Welwitschia had already occurred by 115 million years ago (Won and Renner 2003). This great age partly explains the high sequence variation in the two spacers; in addition, the absence of conserved promoter elements for tRNA genes from both spacers suggests that neither is under functional constraints as suggested earlier for the trnL–trnF spacer but not the trnT–trnL spacer (Bakker et al. 2000; Quandt et al. 2004).

We found several palindromes in the Gnetum trnL–trnF IGS region (Table 2). These sequences were involved in hairpin formation by intrastrand base-pairing. In addition, two more hairpin structures were commonly formed between the 3′ end of the tRNALeu gene and the palindromic structure (Fig. 3A). Such stem-and-loop structures can be substrates for structure-specific nucleases and/or mismatch repair enzymes (Leach 1994) and can cause double-strand breakage during DNA replication (Nasar et al. 2000). Slippage during replication is also the likely cause of the many ‘A,’ ‘G,’ and ‘AT’ repeats in the intron and spacer regions (Levinson and Gutman 1987; Kelchner 2000; Graham et al. 2000). So far, no enzymes that would recognize the hairpin structures in the tRNA spacers have been reported. Instead, the enzymes that process precursor tRNAs, such as RNase P, RNase Z, and ATP (CTP):tRNA nucleotidyl transferase, recognize only the tRNA’s tertiary structure (Martin 1995; Frank and Pace 1998; Kunzmann et al. 1998; Schiffer et al. 2002), suggesting that the nucleotide sequence between the tRNA genes may be almost irrelevant to tRNA enzymatic maturation. Circumstantial evidence, however, indicates that the palindromes in the trnL–trnF spacer serve two functions: By forming thermodynamically stable hairpins, they may reduce the chance that stem-loops interfere with tertiary structure formation of adjacent tRNAs. By forming hairpin structures, they may also function as transcription termination signals as has been found in operons originating from bacteria, including chloroplast operons (Bogorad 1991). Hairpin structures/inverted repeated sequences have been reported from the downstream of maize tRNAHis(GUG) (Schwarz et al. 1981) and tRNAPhe (UGU) (Steinmetz et al. 1983), spinach tRNASer (UGA) (Holschuh et al. 1984), tobacco tRNAGlu (UUC)-tRNATyr(GUA)-tRNAAsp(GUC) (Ohme et al. 1985), and the Brassica napus rRNAs-tRNAArg (ACG) operon (Leal-Klevezas et al. 2000).

In many seed plants, the trnL–trnF spacer appears to harbor a bacterial-type promoter for the tRNAPhe gene (Steinmetz et al. 1983). Indeed, we detected the “−35” and “−10” promoters in several conifers and basal angiosperms, and Quandt et al. (2004) found them in most land plant lineages except lycophytes and leptosporangiate ferns. In Gnetales, however, we detected no promoters in either of the two spacers. Studies on the function of tRNA upstream sequences have shown that either they contain prokaryote-type promoters (Gruissem et al. 1983; Steinmetz et al. 1983; Wu et al. 1997) or there are eukaryote-type internal promoters directly in the tRNA (Galli et al. 1981; Gruissem et al. 1986; Cheng et al. 1997). The loss of promoter elements from the nonphotosynthetic Epifagus tRNAGlu(UUC) (Morden et al. 1991) suggests, however, that prokaryote-type promoters are not always required for transcription of tRNA genes, and a few studies of tRNA transcription in lineages with functioning chloroplasts also suggest this (Gruissem et al. 1986; Jahn 1992; Wu et al. 1997). We therefore interpret the lack of promoter elements in the upstream regions of Gnetales tRNAThr, tRNALeu, and tRNAPhe genes as indicating that these genes have internal promoters, relieving their spacer regions from functional constraints that would come from harboring functional promoters. Release from involvement in the tRNA promoting/processing mechanism may explain why the trnT–trnL and trnL–trnF intergenic spacers contain so many point mutations and indels.

In contrast to the two spacers, the trnL intron maintains highly conserved structural motifs that allow it to form the characteristic secondary structure of group I introns of land plants (Kuhsel et al. 1990; Xu et al. 1990; Lambowitz et al. 1999; Belfort et al. 2002; Quandt et al., 2004) (Fig. 2). Unlike the self-splicing introns of Anabaena and Tetrahymena (Zaug et al. 1993), chloroplast trnL introns of land plants require helper proteins for splicing (Xu et al. 1990; Lambowitz et al. 1999; Simon et al. 2003), in this regard resembling group II introns in tRNA genes (Vogel et al. 1999). The three gaps that we detected in the trnL intron all belonged to the P8 stem-and-loop, one of the autapomorphic indels also belonged to the P8 element, and another occurred in the P9 element. Borsch et al. (2003) similarly found that 68% of the characters in their basal angiosperm trnL intron matrix were contributed by the P6 and P8 stem-loops, while the P, Q, R, and S structural elements were highly conserved.

Expression and Processing of the cp tRNAThr, tRNALeu, and tRNAPhe Genes: A Model

Based on our results, the secondary structure of the cp trnT–trnF region is as shown in Fig. 3A. As discussed above and shown in Figs. 1 A and B and Fig. 3, the cp tRNAThr gene and the tRNALeu–tRNAPhe genes are transcribed in opposite directions. Two earlier studies have presented indirect evidence for cotranscription of the tRNALeu and tRNAPhe genes: Bonnard et al. (1984) saw the absence of a termination signal in the Vicia faba trnL–trnF spacer as such evidence, and Koch et al. (2005) the presence of duplicated tRNAPhe genes between the two bacterial promoter elements present in the Brassicaceae they studied. However, the structural aspects of the trnL–trnF spacer in Gnetum, especially the presence of palindromic sequences and hairpin structures, favor an interpretation of independent transcription (previous section). Regardless of whether tRNA cistrons or operons including tRNA(s) are cotranscribed or independently transcribed, the processing machinery of the chloroplast has to successfully mature each tRNA (Steinmetz et al. 1983; Ohme et al. 1985; Marion-Poll et al. 1988; Christopher and Hallick 1990; Tonkyn and Gruissem 1993; Leal-Klevezas et al. 2000). The tRNA processing pathway includes (1) 5′ end maturation by excision of 5′ leader sequence by RNase P, (2) splicing of the trnL intron (group I intron), (3) excision of the 3′ trailer sequence by RNase Z, (4) terminal ‘CAA’ addition by ATP (CTP):tRNA nucleotidyl transferase, and (5) RNA editing. The order of steps 1, 2, and 3 in the processing of the tRNALeu gene has not been studied, but the splicing of group II introns in the tRNAIle(GAU), tRNAVal(UAC), tRNAAla(UGC), and tRNAGly(UCC) genes indicates that 5′ endonucleolytic cleavage precedes intron splicing and 3′ end maturation (Vogel and Hesse 2001; but see Delp et al. 1991). Because group I and II intron splicing mechanisms in plant chloroplasts have common features, such as mediation by intron RNA and a requirement for helper proteins (Belfort et al. 2002) or chloroplast ribonucleoproteins (cpRNPs [Nakamura et al. 1999]), it is likely that processing of the tRNALeu gene also follows the above pathway. In tRNAs without introns, 5′ maturation also usually precedes 3′ maturation (Wang et al. 1988; Martin 1995; Gegenheimer 1996). The processing and splicing pathway proposed here for the cp trnL intron now needs to be tested via experiments and comparison across a much broader taxonomic range.