Introduction

Variation in chloroplast genomes (plastomes) is a source of information for studies of widely different scope. At one extreme, the conservation of plastome structure has advanced our understanding of broad angiosperm relationships (APG II 2003; Chase et al. 1993; Duvall et al. 1993a; Goremykin et al. 2003, 2005; Moore et al. 2007; Graham et al. 2000; Saarela et al. 2007; Shaw et al. 2005, 2007 etc.). Plastomes also include fine-scale variation for the resolution of relationships in species-rich radiations, such as the graminoid Poales (Bremer 2002). Historically, molecular studies of Poaceae targeted a few specific regions—including conserved genes (ndhF—Clark et al. 1995; rbcL—Duvall and Morton 1996), variable regions (rpoC2 insert—Barker et al. 1999; Duvall et al. 2001), or introns and other noncoding sequence, such as rpl16 and trnL-F (Cialdella et al. 2007; Jakob and Blattner 2006; Kelchner and Clark 1997; Yang et al. 2007). These studies identified two major lineages of Poaceae referenced by acronyms composed of the initial letters of the included subfamilies: the “PACMAD” clade (panicoid, arundinoid, chloridoid, micrairoid, aristidoid, and danthonioid grasses—recently revised from PACCAD, Duvall et al. 2007) and the “BEP” clade (bambusoid, ehrhartoid, and pooid grasses—Kellogg 2000). Although the subfamilial interrelationships are largely resolved, many intertribal and intergeneric phylogenetic issues remain.

Plastome sequences are available for nine genera (ten species) of Poaceae. In this paper we report a draft plastome for Joinvillea plicata and a complete sequence for Coix lacryma-jobi. Coix lacryma-jobi is an andropogonoid grass native to tropical Asia that is widely adventive and considered invasive (Mito and Uesugi 2004; Mosango et al. 2001; Shluker 2003; Townsend and Newell 2006; Villaseñor and Espinosa-Garcia 2004). Deliberate introductions are likely since C. lacryma-jobi is sometimes cultivated (Arber 1965; Barrau 1965; Clayton and Renvoize 1986; Watson and Dallwitz 1992). Production of this grain has declined in Asia with the introduction of other cereals (Roder 2006). Coix lacryma-jobi is also used as an herbal remedy in China (Ruan et al. 2006). Extracts from C. lacryma-jobi, thought to enhance the effectiveness of chemotherapy in the treatment of cancer, were the first traditional Chinese herbal remedies to be approved for clinical trials in the U.S. (Normile and Yimin 2003).

The addition of a plastome from C. lacryma-jobi better balanced the sampling of existing plastomes between the PACMAD and BEP lineages. Higher-resolution investigations involving microstructural mutations could also be performed because plastome sequences from three other Andropogoneae are published. The sequences from J. plicata are useful as outgroup data for plastome-scale research. These results are analyzed to indicate the scope of research applications with plastomes.

Methods

The general approach was to produce overlapping amplicons at known positions around the grass plastome for direct sequencing (following Dhingra and Folta 2005; Mardanov et al. 2008).

Primer Design

The complete plastome sequences of Saccharum officinarum (Asano et al. 2004; Calsa et al. 2004, NC006084), Zea mays (Maier et al. 1995, NC001666), Oryza sativa (NC001320), and Triticum aestivum (Ogihara et al. 2002, NC002762) were aligned (Clustal X—Thompson et al. 1997). Primers were designed manually by choosing highly conserved regions among the four species with an approximate spacing of one kilobase (kb). Primers were positioned to have amplicons overlap by 100–200 base pairs (bp). When possible, priming sites were identical for all species, but preference was ultimately given to Z. mays and S. officinarum, which are more closely related to C. lacryma-jobi (Clayton and Renvoize 1986). Primers were ideally at least 25 bp in length, with a T m of approximately 55°C, and without significant self-dimer formation (∆G < −10). In regions with large indels, S. officinarum and Z. mays were used as references to maintain the 1 kb-spacing. The amplification of the IR used 40 of 54 IRB primers provided to us (Dhingra and Folta 2005). The remaining 227 primers are listed (Tables 1, S1–S3).

Table 1 Sequences of alternate graminoid-specific primers used in the LSC region of J. plicata
Full size table

Taxa

Sequences from C. lacryma-jobi (voucher: M. Duvall s. n. 26 May 2006, DEK) and J. plicata (L. Thien 84, NO) were newly determined and analyzed with those of the four plastomes aligned during primer design plus six others: Agrostis stolonifera (Saski et al. 2007; GenBank accession NC_008591), Brachypodium distachyon (Bortiri et al. 2008; EU325680), Hordeum vulgare (Saski et al. 2007; NC_008590), Lolium perenne (Diekmann et al. 2008; AM777385), Oryza nivara (Shahid-Masood et al. 2004; Tang et al. 2004; NC_005973), and Sorghum bicolor (Saski et al. 2007; NC_008602).

DNA Preparation

Total DNA was isolated from fresh leaf tissue of C. lacryma-jobi using the Qiagen DNeasy Kit (Qiagen Inc., Valencia, CA). The extract of J. plicata was obtained for a previous study. The amount of DNA used for PCR reactions was optimized by testing different dilutions with primer pair IRB19, which amplified the highly conserved 16 S rrn locus (Dhingra and Folta 2005).

PCR Conditions and Sequencing Strategy

Touchdown PCR was performed following “Round I” conditions of Dhingra and Folta (2005) with elongation times extended to 40–150 s to insure complete amplification. All amplicons were generated using Pfu Turbo DNA polymerase (Stratagene Inc., Carlsbad, CA), which reduced PCR enzyme generated errors.

Amplicons were prepared for capillary sequencing using Wizard SV PCR clean-up kits (Promega Corp., Madison, WI). Automated capillary sequencing was performed by Macrogen Inc. (Seoul, South Korea) on each amplicon in both directions giving 2× coverage within the amplicon overlap and the overlapping two adjacent amplicons. Only sequences with Macrogen QV scores (equivalent to Phred scores) of at least 20 were retained. Sequences were manually inspected in Sequencher vers. 4.14 (GenCodes Corp., Ann Arbor, MI). Contigs were formed from overlapping sequences and base-calling discrepancies were resolved by inspection of the electropherograms.

Fragments that initially failed to amplify were sequentially treated as follows. First, the forward primer of the upstream successful amplicon was used with the unsuccessful amplicon’s reverse primer and the downstream successful amplicon’s reverse primer with the unsuccessful forward primer. Second, Coix specific primers were designed in the sequenced regions flanking the unsuccessful amplicon and used to amplify the missing sequence. Third, the amplification was repeated with the previous primer combinations using FideliTaq (USB Corp., Cleveland, OH) following the manufacturer’s standard protocol except that the durations of the extension steps were increased to 3.0 min.

Phylogenetic Analysis

A set of 61 protein coding genes was analyzed following the precedent in other studies (Cai et al. 2006; Goremykin et al. 2003; Leebens-Mack et al. 2005; Moore et al. 2007). ndhF, which was avoided in other whole plastome studies because of its lack of conservation among angiosperms, was used here for its relatively greater conservation and phylogenetic utility in grasses (Clark et al. 1995). This set of loci was extracted from the completed plastomes and from the draft plastome of J. plicata, aligned, and concatenated (Gene Inspector vers. 1.6, Textco BioSoftware, Inc., West Lebanon, NH). Gaps introduced by the alignment were excluded. The maximum parsimony (MP) method, implemented in PAUP* (vers. 4.0b10 Swofford 2003), was used for branch and bound analyses. Nonparametric MP bootstrap analysis was also performed with 1000 pseudoreplicates (Felsenstein 1985).

Small Inversion Analysis

Kim and Lee (2005) identified hairpin structures in the plastomes of grasses and other angiosperms. These structures were characterized by inverted repeats in the stem-forming regions ranging from 11 to 24 bp and had high stability as determined by estimates of free energy. The stem-loop structures were found to be associated with inversions ranging from 5 to 50 bp. Inversions of the loop regions took place when recombination occurred between the flanking stem-forming regions. Kim and Lee (2005) found stem-loop structures to be widely conserved in angiosperms including species of Oryza and Triticum. Homologous regions were thus readily identified by position in the grass plastome and by sequence similarity.

Inversion analysis was conducted on the 11 complete plastomes. We selected six regions, in which the hairpin structure was preserved and for which two or more taxa showed the inverted loop. For each inversion, the loop orientation was determined and arbitrarily coded as “0” (uninverted) or “1” (inverted) for 11 Poaceae. In four cases, because of incomplete sequencing in the outgroup, inversions were scored as missing. Saccharum officinarum was used as a reference to obtain coordinates. These binary coded data were included in a combined branch and bound MP bootstrap analysis.

Indel Analysis

Indel mutations were identified in 11 aligned complete plastomes excluding J. plicata. Each potential indel was evaluated on the basis of three criteria. (1) The indel could be unambiguously attributed to a slipped-strand mispairing event, identified by the presence of a perfect or near-perfect repeated sequence. Indels were accepted when some copies of the repeat were imperfect due to subsequent mutations. (2) The indel was two or more nucleotides long, readily distinguished from putative single-base sequencing artifacts. (3) Both character states (insertion and deletion) of the indel were shared by at least two of the sequences. Indel states were coded as “0” (deletion) or “1” (insertion) and concatenated onto the data matrix for MP analysis. If an indel region was lost because of a larger deletion, or not sequenced, e.g., in J. plicata, it was scored as missing (Table S4).

Results

Size, Gene Content, and Organization of the Coix lacryma-jobi Plastome

The complete C. lacryma-jobi plastome was 145,745 bp long (GenBank accession FJ261955). The two IRs were 22,715 bp, a small single copy region was 12,523 bp, and a large single copy region was 82,792 bp. The positions of the boundaries of the IR and the gene content and order were identical to those in S. officinarum.

An estimated 70% of the J. plicata plastome was completed from which 50 of the 62 protein coding genes were banked as either partial or complete cds (GenBank accessions FJ486219–FJ486269, Leseberg (2009). Note that three other sequences, rbcL (L01471), ndhF (U21973), and rpoC2 (AF001864) were already available. Of the total aligned sequence data, 11.6% were missing for J. plicata. While the coding regions were readily aligned with those of grasses, alignments of noncoding regions were more ambiguous and so only the conserved regions of J. plicata were reported and analyzed here.

Indel and Small Inversion Analysis

Seventy-eight indels meeting the specified criteria were found among the 11 Poaceae ranging in size from 2 to 26 bp (mean length, 5 bp; see Supplementary Table S4). Indels were found primarily in intergenic sequences (IGS; 57 indels), 14 were found in introns, and 7 were in coding regions (Table 2).

Table 2 Locations of 78 Indels identified among 11 Poaceae
Full size table

Six loop regions could be identified on the basis of position, flanking stem sequences, and homologous loop regions in 11 Poaceae except for two instances—the inversion at coordinate position 63 275 in S. officinarum was apparently lost as part of a larger deletion in both Triticum and Hordeum (Table 3).

Table 3 Character coding and locations for small inversions in regions originally identified by Kim and Lee (2005) as stable hairpin-loops
Full size table

Phylogenetic Analysis

Four branch and bound maximum parsimony (MP) analyses were performed with a series of increasingly inclusive data sets. The datasets included [1] a set of 61 protein-coding genes, [2] the ndhF coding sequence, [3] 78 indels, and [4] six inversions. MP analyses were performed on subsets [1], [1,2], [1–3], and [1–4] and the results are presented (Table 4). The tree and bootstrap values presented in Fig. 1 were produced from the total combined datasets [1–4]. This analysis resulted in a single MP tree of 8169 steps with a consistency index (excluding uninformative characters) of 0.7639 and a retention index of 0.8513 (Table 4). The ranges of total and informative characters were 42478–44627 and 2681–2985, respectively. The ranges of total summed internal branch lengths for the entire tree and the Andropogoneae were 3557–3972 and 682–776, respectively. All nodes were supported with 100% bootstrap values with the exception of the BEP clade. The support for this internal node increased with the inclusion of more datasets with values of 90, 96, 96, and 97 for datasets [1], [1,2], [1–3], and [1–4], respectively (trees not shown). Within the Andropogoneae, S. officinarum and S. bicolor were sister to C. lacryma-jobi, and this clade of three species was sister to Z. mays.

Table 4 Cumulative statistics for maximum parsimony analyses with successively increasing character data
Full size table
Fig. 1
figure 1

Branch and bound maximum parsimony phylogram. Analysis included 61 conserved protein-coding sequences, the ndhF coding sequence, 6 inversions, and 78 indels from 12 graminoid plastomes. The BEP clade (subfamilies Pooideae and Ehrhartoideae) and the Andropogoneae, are indicated. Branches are proportional to the number of changes along the branch. Numbers along the branches are MP bootstrap values

Full size image

Discussion

Recent whole plastome phylogenetic studies have been largely restricted to analyses of point mutations in 61 conserved protein-coding genes (Cai et al. 2006; Chang et al. 2006; Goremykin et al. 2003; Leebens-Mack et al. 2005; Moore et al. 2007; Saski et al. 2007; Wu et al. 2009). Here we show in the graminoid radiation the effects of combining this data partition with the ndhF coding sequence, hairpin-loop inversions, and indel characters found in coding and noncoding regions. Since maximum likelihood models for indel and inversion data are incompletely developed, the analyses here were confined to MP analyses. The combination of all these types of data increased characters and support at both deep and shallow nodes in parsimony analyses. This is especially relevant for intensively sampled taxa where the number of informative substitutions in the conserved genes becomes limiting.

The phylogenetic analyses performed here were largely in agreement with each other and with recent multi-gene analyses (e.g., Bouchenak-Khelladi et al. 2008; Duvall et al. 2007). With the exception of the BEP node in the MP analyses, the levels of support for all nodes in the MP analysis were 100%. The BEP clade has historically been rather weakly supported since the study that first identified it (Clark et al. 1995). In this study, successively adding data from different partitions increased the MP bootstrap value for the BEP clade from 90 to 97% when all data were included (Fig. 1). Character support for the BEP node, as approximated by branch length (ACCTRAN optimization), increased correspondingly with values of 481, 525, 538, and 540 steps for data partitions [1], [1,2], [1–3], and [1–4], respectively. Note that maximum support for the BEP clade in MP trees was obtained when a representative bambusoid was included in the analyses (Cryptochloa strictiflora—Grennan and Duvall, unpublished), so that improved sampling retrieves this clade with stronger support.

Coix lacryma-jobi was sister to a clade comprising S. officinarum and S. bicolor in our analysis consistent with nuclear gene studies (tb1—Lukens and Doebley 2001; PHYB, GBBSI—Mathews et al. 2002). The position of C. lacryma-jobi in our MP tree is in agreement with a broadly delimited Andropogoneae (GPWG 2001; Spangler et al. 1999), but contradicts other studies that classified C. lacryma-jobi in a putative sister tribe, Maydeae, with Z. mays (e.g., Kellogg and Birchler 1993; Watson and Dallwitz 1992). Here, Maydeae is paraphyletic with the remaining Andropogoneae (Fig. 1).

Results for the position of C. larcryma-jobi exemplify how data from complete plastomes can be used in comparisons of relatively closely related taxa such as the four Andropogoneae in this study. Maximum parsimony bootstrap values for all nodes in Andropogoneae were at a maximum in the 61-locus tree (not shown). However, the number of parsimony informative characters and the lengths of internal branches were increased by including data from other character partitions. For example, character support for the (S. officinarum, S. bicolor) clade increased with values of 22, 24, 30, and 31 steps for data partitions [1], [1,2], [1–3], and [1–4], respectively. Similarly, character support for the position of C. lacryma-jobi increased with values of 39, 45, 49, and 50 steps for the same series of data partitions. The number of characters supporting the monophyly of Andropogoneae likewise increased from 621, 661, 694, to 695. The use of the other data partitions increased parsimony informative characters by 11.3% across Andropogoneae. Internal branch lengths in Andropogoneae were increased by 7.0, 5.9, and 0.39%, respectively (Table 4). Given the short branches in this clade in the 61-locus tree (mean value, 30.5), the addition of further Andropogoneae is predicted to reduce support and/or decrease phylogenetic resolution when only conserved protein coding loci are analyzed, an apparent limit to plastome-scale phylogenetic analyses as they have been typically conducted. However, the abundance of little-used characters in complete plastomes can add significantly to the information supporting the relationships of closely related genera.

As more plastomes become available, characters that are currently noninformative autapomorphies are expected to become informative in the same way that the addition of the C. lacryma-jobi plastome changed autapomorphies in Z. mays into synapomorphies for subsets of Andropogoneae. Indel character 75 is one example (Table S4). Without the C. lacryma-jobi sequence the deletion of one four nucleotide direct repeat in Z. mays is autapomorphic, and so would have been rejected by our indel criterion number 3, as uninformative. However, since the C. lacryma-jobi sequence also has the deletion, this condition can be interpreted to be symplesiomorphic for Andropogoneae in the context of our MP tree, with a subsequent synapomorphic insertion in the (S. officinarum, S. bicolor) clade. Indels 5, 28, 30, and 51 become similarly informative only after the addition of the C. lacryma-jobi sequence to the other sampled Poaceae.

Conclusion

In this study a novel set of primers is specified, which were designed to create overlapping amplicons in the grass plastome and tested in two graminoids (Tables S1–S3 and LSC-specific primers for J. plicata in Table 1). This set of primers has already been successfully used for other grasses including species of Anomochloa, Chasmanthium, Cryptochloa, Microcalamus, Pharus, Puelia, and another outgroup, Ecdeiocolea (Ecdeiocoleaceae) (Duvall et al. unpublished). Now that this significant investment of obtaining graminoid-specific primers is accomplished, future sequencing of graminoid plastomes will be facilitated.

Joinvillea plicata was chosen as an outgroup for draft plastome sequencing because of the close relationship between Joinvilleaceae and Poaceae (Campbell and Kellogg 1987; Doyle et al. 1992; Duvall et al. 1993a, b; Linder and Rudall 1993; Kellogg 2000; Marchant and Briggs 2007; Michelangeli et al. 2003). This set of conserved protein coding gene sequences is a useful resource for outgroup rooting of Poaceae in phylogenetic studies.

The 78 indels (Table 2) and six inversions (Table 3) in this study were identifiable only on a whole plastome level and 91% were located in non-coding regions. The indel dataset proved quite robust with a total of 75 out of the 78 capable of being scored unambiguously for all 11 species (Table S4). The indel and inversion characters also had low homoplasy. The portion of the data matrix with sequences from 62 loci had a consistency index (excluding uninformative characters) of 0.6441 while that for the 78 indels and six inversions (all characters informative) was 0.8442.

The use of the substitutions in the 61 conserved gene dataset clearly increases phylogenetic information compared to single and multi-gene analyses of the past. However, it is shortsighted to rely solely on these characters in studies of the graminoid radiation when substantial molecular evolution occurs in non-coding regions identifiable from whole plastome sequences. Other data, such as entire intergenic and intron sequences as well as pseudogenes may also be exploited for potential divergence among closely related species in species-rich evolutionary radiations. The complete plastome can provide considerable character state data to support key nodes in a wide range of evolutionary studies.