Complete chloroplast genome sequences of Hordeum vulgare, Sorghum bicolor and Agrostis stolonifera, and comparative analyses with other grass genomes

Abstract

Comparisons of complete chloroplast genome sequences of Hordeum vulgare, Sorghum bicolor and Agrostis stolonifera to six published grass chloroplast genomes reveal that gene content and order are similar but two microstructural changes have occurred. First, the expansion of the IR at the SSC/IRa boundary that duplicates a portion of the 5′ end of ndhH is restricted to the three genera of the subfamily Pooideae (Agrostis, Hordeum and Triticum). Second, a 6 bp deletion in ndhK is shared by Agrostis, Hordeum, Oryza and Triticum, and this event supports the sister relationship between the subfamilies Erhartoideae and Pooideae. Repeat analysis identified 19–37 direct and inverted repeats 30 bp or longer with a sequence identity of at least 90%. Seventeen of the 26 shared repeats are found in all the grass chloroplast genomes examined and are located in the same genes or intergenic spacer (IGS) regions. Examination of simple sequence repeats (SSRs) identified 16–21 potential polymorphic SSRs. Five IGS regions have 100% sequence identity among Zea mays, Saccharum officinarum and Sorghum bicolor, whereas no spacer regions were identical among Oryza sativa, Triticum aestivum, H. vulgare and A. stolonifera despite their close phylogenetic relationship. Alignment of EST sequences and DNA coding sequences identified six C–U conversions in both Sorghum bicolor and H. vulgare but only one in A. stolonifera. Phylogenetic trees based on DNA sequences of 61 protein-coding genes of 38 taxa using both maximum parsimony and likelihood methods provide moderate support for a sister relationship between the subfamilies Erhartoideae and Pooideae.

Introduction

Chloroplasts are the most noticeable feature of green cells in leaves and, excluding the vacuole, probably constitute the largest compartment within mesophyll cells (Lopez-Juez and Pyke 2005). Plastids are multifunctional and are used by the plant for critical biochemical processes other than photosynthesis, including starch synthesis, nitrogen metabolism, sulfate reduction, fatty acid synthesis, DNA and RNA synthesis (Zeltz et al. 1993). The chloroplast genome generally has a highly conserved organization (Palmer 1991; Raubeson and Jansen 2005) with most land plant genomes composed of a single circular chromosome with a quadripartite structure that includes two copies of an inverted repeat (IR) that separate the large and small single copy regions (LSC and SSC). The size of this circular genome varies from 35 to 2,217 kb but among photosynthetic organisms the majority are between 115 and 165 kb (Jansen et al. 2005).

Our knowledge of the organization and evolution of chloroplast genomes has been expanding rapidly because of the large numbers of completely sequenced genomes published in the past decade. The use of information from chloroplast genomes is well established in the study of the evolutionary patterns and processes in plants (Avise 1994; Raubeson and Jansen 2005). Genetic markers derived from organelle genomes generally show simple, uniparental modes of inheritance, which makes them invaluable for the purposes of population genetic and phylogenetic studies (Bryan et al. 1999; Provan et al. 2001) and this feature also facilitates transgene containment (Daniell 2002).

Sorghum, with 25 species, is a member of the family Poaceae and tribe Andropogoneae (Garber 1950). Recent molecular phylogenetic analyses indicated that the genus may be paraphyletic (Spangler et al. 1999), and that it is comprised of three distinct lineages, Sorghum, Sarga and Vacoparis (Spangler 2003). The genus Sorghum was redefined to include three species, Sorghum bicolor, Sorghum halepense, and Sorghum nitidum. Sorghum bicolor, grain sorghum, is the third most important cereal crop in the United States and the fifth most important crop in the world (Crop Plant Resources 2000). Sorghum is well known for its capacity to tolerate conditions of limited moisture and to produce during periods of extended drought, in circumstances that would impede production in most other grains (Crop Plant Resources 2000). Sorghum is used for human nutrition and feed grain for livestock throughout the world (Carter et al. 1989). A more recent use of Sorghum is the production of ethanol, with one bushel producing the same amount of ethanol as one bushel of corn (National Sorghum Producers 2006). Some Sorghum varieties are rich in anti-oxidants and all varieties are gluten-free, an attractive alternative for those allergic to Triticum aestivum (US Grains Council 2006).

Of the various cereals, Hordeum vulgare L. (barley) is a major food, feed and malt crop. In 2005, H. vulgare ranked fourth in quantity produced and in area of cultivation of cereal crops in the world (http://faostat.fao.org/faostat/) demonstrating its broad consumption and wide adoption in a variety of climates, from sub-arctic to sub-tropical. According to the USDA/NASS, H. vulgare is the third major feed grain crop produced in the United States, after Zea mays (maize) and Sorghum bicolor. Production is concentrated in the Northern Plains and the Pacific Northwest. The United States is the eighth largest producer of H. vulgare in the world with current production estimated at 4.9 million acres. It is a short-season, early maturing crop grown on both irrigated and dry land production areas in the United States. Whole grain H. vulgare contains high levels of minerals and important vitamins, including calcium, magnesium, phosphorus, potassium, vitamin A, vitamin E, niacin and folate.

Among the non-food grasses, Agrostis stolonifera L. (creeping bentgrass) has attracted great attention in both academia and the biotech industry due to its social and economic importance. A. stolonifera is a wind-pollinated, highly outcrossing perennial grass used on golf courses worldwide. It can also enhance the natural beauty of the environment and increase the value of residential and commercial property, and provide many environmental benefits including preventing soil erosion, filtering water and trapping dust and pollutants (Bonos et al. 2006). It has been extensively used, covering millions of acres globally making it an economically valuable grass crop. Due to its aforementioned importance, transgenic A. stolonifera was produced conferring the herbicide resistance trait by engineering the CP4 EPSPS gene, which is one of the first transgenic, perennial, wind-pollinated crops intending to be grown outside of agricultural fields (i.e., on golf courses). Unfortunately, pollen-mediated transgene flow has been reported in several studies (Wipff and Fricker 2001; Watrud et al. 2004; Reichman et al. 2006) limiting its commercialization and demonstrating the requirement of effective containment strategies to protect the environment and to engineer this plant with environmentally friendly approaches like chloroplast engineering or cytoplasmic male sterility.

The agronomic, economic and/or social importance of H. vulgare, Sorghum bicolor and A. stolonifera has made them the focus of numerous studies attempting to improve these crop species. Much of this work has been restricted to investigations of nuclear genomes of these species (USDA 2006, Cheng et al. 2004). This has resulted in very limited information on the organization and evolution of chloroplast genomes of H. vulgare, Sorghum bicolor and A. stolonifera. Therefore, the current study could enhance our understanding of the chloroplast genome organization of grasses facilitating the improvement of those crops by chloroplast genetic engineering. The plastid transformation approach has been shown to have a number of advantages, most notably with regard to its high transgene expression levels (De Cosa et al. 2001), capacity for multi-gene engineering in a single transformation event (De Cosa et al. 2001; Lossl et al. 2003; Ruiz et al. 2003; Quesada-Vargas et al. 2005; Daniell and Dhingra 2002), and ability to accomplish transgene containment via maternal inheritance (Daniell 2002). Moreover, chloroplasts appear to be an ideal compartment for the accumulation of certain proteins, or their biosynthetic products, which would be harmful if they accumulated in the cytoplasm (Daniell et al. 2001; Lee et al. 2003; Leelavathi and Reddy 2003; Ruiz and Daniell 2005). In addition, no gene silencing has been observed in association with this technique, whether at the transcriptional or translational level (De Cosa et al. 2001; Lee et al. 2003; Dhingra et al. 2004). Because of these advantages, the chloroplast genome has been engineered to confer several useful agronomic traits, including herbicide resistance (Daniell et al. 1998), insect resistance (McBride et al. 1995; Kota et al. 1999), disease resistance (DeGray et al. 2001), drought tolerance (Lee et al. 2003), salt tolerance (Kumar et al. 2004a), and phytoremediation (Ruiz et al. 2003). The chloroplast genome has also been utilized in the field of molecular farming, for the expression of biomaterials, human therapeutic proteins, and vaccines for use in humans or other animals (Guda et al. 2000; Staub et al. 2000; Fernandez-San et al. 2003; Leelavathi et al. 2003; Molina et al. 2004; Vitanen et al. 2004; Watson et al. 2004; Koya et al. 2005; Grevich and Daniell 2005; Daniell et al. 2005a, b; Kamarajugadda and Daniell 2006; Chebolu and Daniell 2007; Arlen et al. 2007; Ruhlman et al. 2007; Daniell et al. 2004a, b).

In this article, we present the complete sequences of the chloroplast genomes of H. vulgare, Sorghum bicolor and A. stolonifera. One goal is to compare the genome organization of H. vulgare, Sorghum bicolor and A. stolonifera with six other completely sequenced grass chloroplast genomes; Oryza sativa, O. nivara, Saccharum hybrid, Saccharum officinarum, T. aestivum, and Z. mays. In addition to examining gene content and gene order, we determined the distribution and location of repeated sequences among these genomes, including potential microsatellite markers. A second goal is to compare levels of DNA sequence divergence of non-coding regions. Intergenic spacer (IGS) regions have been examined to identify ideal insertion sites for transgene integration, and to assess the utility of these regions for resolving phylogenetic relationships among closely related species (Kelchner 2002; Shaw et al. 2005, 2007; Saski et al. 2005; Daniell et al. 2006; Timme et al. 2007). A third goal of this paper is to examine the extent of RNA editing in the H. vulgare, Sorghum bicolor and A. stolonifera chloroplast genomes by comparing the DNA sequences with available expressed sequence tag (EST) sequences. RNA editing is a co- or post-transcriptional process that occurs in organelles and changes the coding information in mRNAs (Kugita et al. 2003; Wolf et al. 2004; Peeters and Hanson 2002). Most of our knowledge about the frequency of this process in crop plants comes from studies in Z. mays (Maier et al. 1995) and Nicotiana tabacum (Hirose et al. 1999), and additional comparative studies are needed in other plant species to understand the extent of RNA editing in chloroplast genomes. A final goal is to assess phylogenetic relationships between H. vulgare, Sorghum bicolor, A. stolonifera and other completely sequenced angiosperm chloroplast genomes.

Materials and methods

DNA sources

Bacterial artificial chromosome (BAC) libraries of H. vulgare cv Morex and Sorghum bicolor cv BTX623 were constructed by ligating size fractionated partial HindIII digests of total cellular, high molecular weight DNA with the pINDIGOBAC536 vector. The average insert size of H. vulgare (HV_MBa) and Sorghum bicolor (SB_BBc) libraries was 106 and 120 kb, respectively. BAC related resources for these public libraries can be obtained from the Clemson University Genomics Institute BAC/EST Resource Center (www.genome.clemson.edu).

Bacterial artificial chromosome clones containing chloroplast genome inserts were isolated by screening the library with a soybean chloroplast DNA probe. The first 96 positive clones from screening were pulled from the library, arrayed in a 96 well microtitre plate, copied and archived. Selected clones were then subjected to HindIII fingerprinting and NotI digests. End-sequences were determined and localized on the chloroplast genome of Arabidopsis thaliana to deduce the relative positions of the clones; then clones that covered the entire chloroplast genomes of H. vulgare and Sorghum bicolor were chosen for sequencing.

Preparation of intact chloroplasts and rolling circle amplification

The A. stolonifera L. cultivar Penn A-4 was supplied by HybriGene, Inc. (Hubbard, OR, USA). Prior to chloroplast isolation, plants were kept in dark for 2 days to reduce levels of starch. Chloroplasts from young leaves were isolated using the sucrose step gradient method of Palmer (1986) as modified by Jansen et al. (2005). About 10 g of leaf tissue was homogenized in Sandbrink isolation buffer using pre-chilled tissue blender bursts at high speed for 5 s to get sufficient quantities of chloroplasts. The homogenate was filtered using four layers of cheesecloth and one layer of miracloth (Calbiochem, catalog number 474855) without squeezing. The filtrate was transferred to pre-chilled centrifuge tubes and centrifuged at 1,000 g for 15 min at 4°C. Pellets were resuspended in 7 ml of ice-cold wash buffer and gently loaded over the step gradient consisting of 18 ml of 52% sucrose, over-layered with 7 ml of 30% sucrose. The sucrose step gradient was centrifuged at 25,000 rpm for 30–60 min at 4°C in a SW-27 rotor (Beckman). The chloroplast band from the 30–52% interface was removed using a wide bore pipette, diluted with ten volumes wash buffer, and centrifuged at 1,500 g for 15 min at 4°C. Purified chloroplast pellets were resuspended in a final volume of 2 ml. The entire chloroplast genome was amplified by Rolling Circle Amplification (RCA) using the Repli-g RCA kit (Qiagen, Inc.) following the methods described in (Jansen et al. 2005). RCA was performed at 30°C for 16 h; the reaction was terminated with final incubation at 65°C for 10 min. Digestion of the RCA product with the restriction enzymes BstXI, EcoRI and HindIII verified successful genome amplification, as well as DNA quality for sequencing.

DNA sequencing and genome assembly

The nucleotide sequences of the BAC clones and RCA product were determined by the bridging shotgun method. The purified BAC DNA or RCA product was subjected to hydroshearing, end repair and then size-fractionated by agarose gel electrophoresis. Fractions of approximately 3.0–5.0 kb were eluted and ligated into the vector pBLUE-SCRIPT IIKS+. The libraries were plated and arrayed into 40 96-well microtitre plates for the sequencing reactions.

Sequencing was performed using the Dye-terminator cycle sequencing kit (Perkin Elmer Applied Biosystems, USA). Sequence data from the forward and reverse priming sites of the shotgun clones were accumulated. Sequence data equivalent to eight times the size of the genome was assembled using Phred-Phrap programs (Ewing et al. 1998).

Gene annotation

Annotation of the Sorghum bicolor, H. vulgare and A. stolonifera chloroplast genomes was performed using DOGMA (Dual Organellar GenoMe Annotator, Wyman et al. 2004, http://bugmaster.jgi-psf.org/dogma/). This program uses a FASTA-formatted input file of the complete genomic sequences and identifies putative protein-coding genes by performing BLASTX searches against a custom database of previously published chloroplast genomes. The user must select putative start and stop codons for each protein-coding gene and intron and exon boundaries for intron-containing genes. Both tRNAs and rRNAs are identified by BLASTN searches against the same database of chloroplast genomes.

Molecular evolutionary comparisons

Comparisons of gene content and gene order

Gene content comparisons were performed with Multipipmaker (Schwartz et al. 2003). Comparisons included nine genomes: O. sativa (NC_001320, Hiratsuka et al. 1989), O. nivara (NC_005973, Shahid-Masood et al. 2004), Saccharum officinarum (NC_006084, Asano et al. 2004), Saccharum hybrid (NC_005878, Calsa et al. unpublished), T. aestivum (NC_002762, Ogihara et al. 2000), Z. mays (NC_001400, Maier et al. 1995), H. vulgare (NC_008590, current study), Sorghum bicolor (NC_008602, current study) and A. stolonifera (NC_008591, current study) using O. sativa as the reference genome. Gene orders were examined by pair-wise comparisons between the above genomes using PipMaker (Elnitski et al. 2002).

Examination of repeat structure

Shared and unique repeats were identified for H. vulgare, Sorghum bicolor and A. stolonifera genomes and compared to other grass genomes using Comparative Repeat Analysis (CRA, N. Holtshulte and S. K. Wyman, unpublished, http://bugmaster.jgi-psf.org/repeats/). This program filters the redundant output of REPuter (Kurtz et al. 2001) and identifies shared repeats among the input genomes. For repeat identification, the following constraints were set in CRA: a minimum repeat size of 30 bp and a Hamming distance of 3 (i.e., a sequence identity of ≥90%). Oryza sativa was used as the reference genome. Blast hits 30 bp and longer with a sequence identity of ≥90% were identified to determine the shared repeats among the seven genomes examined. To detect SSRs we used a modified version of the Perl script SSRIT (Temnykh et al. 2001). The modified script, CUGISSR (Jung et al. 2005), was used to search for SSRs ranging from di-to penta-nucleotide repeats.

Comparison of intergenic spacer regions

Intergenic spacer regions from seven grass chloroplast genomes were compared using MultiPipMaker (Schwartz et al. 2003, http://pipmaker.bx.psu.edu/pipmaker/tools.html). MultiPipMaker has a suite of software tools to analyze relationships among more than two sequences. We used a program known as ‘all_bz’ that iteratively compares a pair of nucleotide sequences at a time until all possible pairs from all species have been examined. However, this program processes only one set of IGS regions at a time. For genome-wide comparisons of corresponding intergenic regions from all species, we developed two programs written in PERL. The first iteration creates a set of input files containing corresponding intergenic regions from each species and compares them using ‘all_bz’ program, until all the intergenic regions in the chloroplast genome are processed. The second program parses the output from the above comparisons, calculates percent identity by using the number of identities over the length of the longer sequence, and generates results in tab-delimited tabular format.

Variation between coding sequences and cDNAs

Each of the genes from the H. vulgare, Sorghum bicolor and A. stolonifera chloroplast genomes were used to perform a BLAST search of expressed sequence tags (ESTs) from the NCBI Genbank. The retrieved EST sequences from A. stolonifera, H. vulgare and Sorghum bicolor were then aligned with the corresponding annotated gene for each species separately, using Clustal X. The aligned sequences were then screened and nucleotide and amino acid changes were detected using the Megalign software and the plastid/bacterial genetic code. Due to variation in length between an EST and the corresponding gene, the length of the analyzed sequence was recorded.

Phylogenetic analyses

The 61 genes included in the analyses of Goremykin et al. (2003a, 2004a, 2005), Leebens-Mack et al. (2005), Chang et al. (2006), Lee et al. (2006a, b), Jansen et al. (2006) and Ruhlman et al. (2006) were extracted from the chloroplast genome sequence of A. stolonifera, H. vulgare and Sorghum bicolor using DOGMA (Wyman et al. 2004). The same set of 61 genes was extracted from chloroplast genome sequences of 35 other sequenced genomes (see Table 1 for complete list). All 61 protein-coding genes of the 38 taxa were translated into amino acid sequences, aligned using MUSCLE (Edgar 2004) followed by manual adjustments, and then nucleotide sequences of these genes were aligned by constraining them to the aligned amino acid sequences. A Nexus file with character sets for phylogenetic analyses was generated after nucleotide sequence alignment was completed. The complete nucleotide alignment is available online at Chloroplast Genome Database (Cui et al. 2006, http://chloroplast.cbio.psu.edu).

Table 1.

Taxa included in phylogenetic analyses with GenBank accession numbers and references

Taxon	GenBank accession numbers	Reference
Gymnosperm outgroups
Pinus thunbergii	NC_001631	Wakasugi et al. 1994
Ginkgo biloba	NC_008788	Leebens-Mack et al. 2005
Basal angiosperms
Amborella trichopoda	NC_005086	Goremykin et al. 2003a
Nuphar advena	NC_008788	Leebens-Mack et al. 2005
Nymphaea alba	NC_006050	Goremykin et al. 2004
Magnoliids
Calycanthus floridus	NC_004993	Goremykin et al. 2003b
Drimys granatensis	NC_008456	Cai et al. 2006
Liriodendron tulipifera	NC_008326	Cai et al. 2006
Piper coenoclatum	NC_008457	Cai et al. 2006
Monocots
Acorus americanus	DQ069337-DQ069702	Leebens-Mack et al. 2005
Agrostis stolonifera	NC_008591	Current study
Hordeum vulgare	NC_008590	Current study
Oryza sativa	NC_001320	Hiratsuka et al. 1989
Phalaenopsis aphrodite	NC_007499	Chang et al. 2006
Saccharum officinarum	NC_006084	Asano et al. 2004
Sorghum bicolor	NC_008602	Current study
Triticum aestivum	NC_002762	Ogihara et al. 2000
Typha latifolia	DQ069337-DQ069702	Leebens-Mack et al. 2005
Yucca schidigera	DQ069337-DQ069702	Leebens-Mack et al. 2005
Zea mays	NC_001666	Maier et al. 1995
Eudicots
Arabidopsis thaliana	NC_000932	Sato et al. 1999
Atropa belladonna	NC_004561	Schmitz-Linneweber et al. 2002
Citrus sinensis	NC_008334	Bausher et al. 2006
Cucumis sativus	NC_007144	Plader et al. unpublished
Eucalyptus globulus	NC_008115	Steane 2005
Glycine max	NC_007942	Saski et al. 2005
Gossypium hirsutum	NC_007944	Lee et al. 2006a
Lotus corniculatus	NC_002694	Kato et al. 2000
Medicago truncatula	NC_003119	Lin et al. unpublished
Nicotiana tabacum	NC_001879	Shinozaki et al. 1986
Oenothera elata	NC_002693	Hupfer et al. 2000
Panax schinseng	NC_006290	Kim and Lee 2004
Populus trichocarpa	NC_008235	Unpublished
Ranunculus macranthus	NC_008796	Leebens-Mack et al. 2005
Solanum lycopersicum	DQ347959	Daniell et al. 2006
Solanum bulbocastanum	NC_007943	Daniell et al. 2006
Spinacia oleracea	NC_002202	Schmitz-Linneweber et al. 2001
Vitis vinifera	NC_007957	Jansen et al. 2006

Phylogenetic analyses using maximum parsimony (MP) and maximum likelihood (ML) were performed with PAUP* version 4.10b10 (Swofford 2003) and GARLI version 0.942 (Zwickl 2006, http://www.bio.utexas.edu/grad/zwickl/web/garli.html), respectively. Phylogenetic analyses excluded gap regions to avoid alignment ambiguities in regions with variation in sequence lengths. All MP searches included 100 random addition replicates and TBR branch swapping with the Multrees option. Non-parametric bootstrap analyses (Felsenstein 1985) were performed for MP analyses with 1,000 replicates with TBR branch swapping, one random addition replicate, and the Multrees option. Modeltest 3.7 (Posada and Crandall 1998) was used to determine the most appropriate model of DNA sequence evolution for the combined 61-gene dataset. Hierarchical likelihood ratio tests and the Akaike information criterion were used to assess which of the 56 models best fit the data, which was determined to be GTR + I + Γ by both criteria. For ML analyses in GARLI two independent runs were performed using the default settings (see Garli manual at http://www.bio.utexas.edu/grad/zwickl/web/garli.html). Non-parametric bootstrap analyses (Felsenstein 1985) were performed in GARLI for ML analyses using default settings.

Results

Size, gene content and organization of the H. vulgare, S. bicolor and A. stolonifera chloroplast genomes

The complete sizes of the H. vulgare, Sorghum bicolor and A. stolonifera chloroplast genomes are 136,462, 140,754 bp and 136,584 bp, respectively (Fig. 1). The genomes include a pair of IRs of 21,579 bp (H. vulgare), 22,782 bp (Sorghum bicolor) and 21,649 bp (A. stolonifera) separated by a small single copy region of 12,704 bp (H. vulgare), 12,502 bp (Sorghum bicolor) and 12,740 bp (A. stolonifera) and a large single copy region of 80,600 bp (H. vulgare), 82,688 bp (Sorghum bicolor) and 80,546 bp (A. stolonifera).

Fig. 1.

Gene map of Hordeum vulgare, Sorghum bicolor and Agrostis stolonifera chloroplast genomes. The thick lines indicate the extent of the inverted repeats (IRa and IRb), which separate the genome into small (SSC) and large (LSC) single copy regions. Genes on the outside of the map are transcribed in the clockwise direction and genes on the inside of the map are transcribed in the counterclockwise direction

The H. vulgare, Sorghum bicolor and A. stolonifera chloroplast genomes contain 113 different genes, and 18 of these are duplicated in the IR, giving a total of 131 genes (Fig. 1). There are 30 distinct tRNAs, and 7 of these are duplicated in the IR. Sixteen genes contain one or two introns, and six of these are in tRNAs. The H. vulgare chloroplast genome consists of 56.7% coding regions that includes 48% protein coding genes, 8.7% RNA genes and 43.3% non-coding regions, containing both IGS regions and introns. The Sorghum bicolor chloroplast genome is composed of 52.1% coding regions that includes 43.4% protein coding genes, 8.7% RNA genes and 47.9% non-coding regions. The A. stolonifera chloroplast genome is composed of 53.6% coding regions that includes 44.7% protein coding genes, 8.9% RNA genes and 46.4% non-coding regions. The overall GC and AT content of the H. vulgare, Sorghum bicolor and A. stolonifera chloroplast genomes are 38.31% (H. vulgare), 38.50% (Sorghum bicolor), 38.45% (A. stolonifera) and 61.69% (H. vulgare), 61.50% (Sorghum bicolor) and 61.55% (A. stolonifera), respectively.

Gene content and gene order

Gene content and order of the H. vulgare, Sorghum bicolor and A. stolonifera chloroplast genomes are similar to the other six sequenced grass chloroplast genomes (O. sativa, O. nivara, Saccharum hybrid, Saccharum officinarum, T. aestivum, and Z. mays). Like other grass chloroplast genomes, the IR in H. vulgare, Sorghum bicolor and A. stolonifera has expanded to include rps19. However, the extent of the IR at the SSC/IRa boundary differs between two of the genomes with the IR of H. vulgare and A. stolonifera expanded to duplicate a portion of ndhH, a feature that is shared with the T. aestivum chloroplast genome (Ogihara et al. 2000). This expansion includes 207 bp (69 amino acids) in H. vulgare, 174 bp (58 amino acids) in A. stolonifera, and 96 bp (32 amino acids) in T. aestivum. The H. vulgare, Sorghum bicolor and A. stolonifera genomes also share the loss of introns in clpP and rpoC1 with other grasses. There are insertions and deletions (indels) of nucleotides within several coding sequences. For example, CAAAAC is uniquely present within matK of Sorghum bicolor, but absent in the rest of the grasses examined (Supplementary Figure 1). There is also a 6 bp deletion in the ndhK gene in H. vulgare, A. stolonifera, T. aestivum and both species of Oryza (Supplementary Figure 1).

Repeat structure

Repeat analyses identified 19–37 direct and IRs 30 bp or longer with a sequence identity of at least 90% among the nine chloroplast genomes examined (Fig. 2). With one exception of a 91 bp repeat, all other repeats range in size between 30 and 60 bp, and 78.4% are in the direct orientation while 21.6% are inverted. The longest repeats other than the IRs found in H. vulgare and Sorghum bicolor are 540 and 524 bp, respectively. BlastN comparisons of the O. sativa repeats against the chloroplast genomes of the eight other grasses identified 26 shared repeats ≥30 bp with a sequence identity ≥90% (Table 2). H. vulgare and T. aestivum share four repeats (31, 32, 36, and 38 bp) not found in any other genomes. Both Oryza species share 41 and 59 bp repeats. Zea mays has the most repeats with 37 and A. stolonifera has the fewest with 19. Seventeen of the 26 repeats are found in all eight chloroplast genomes and all of these are located in the same genes or IGS regions.

Fig. 2.

Histogram showing the number of repeated sequences ≥30 bp long with a sequence identity ≥90% in nine grass chloroplast genomes

Table 2.

Oryza sativa repeats blasted against all eight chloroplast genomes

Repeat number	Size (bp)	Number of hits	Orientation	Location	Genomes
1	30	2	Direct	IGS—(trnN-GUU-rps15)	Sb, So, Sh, On, Zm
2	30	2	Direct	rps3	Sb, On, Ta, Hv, Sh, So, Zm, As
3	30	2	Direct	IGS—(trnM-CAU-trnG-UCC), trnM-CAU	Sb, On, Ta, Hv, Sh, So, Zm, As
4	30	2	Direct	Intron—(ndhB)	Sb, On, Hv, Sh, So, Zm, As
5	31	3	Direct	IGS—(trnG-GCC—trnM-CAU), IGS—(trnM-CAU—rps14)	Sb, On, Ta, Hv, Sh, So, Zm, As
6	31	2	Direct	rpoC2	Sb, On, Sh, So, Zm, As
7	32	2	Inverted	trnS-UGA	Sb, On, Ta, Hv, Sh, So, Zm, As
8	32	3	Inverted	rpl23	Sb, On, Ta, Hv, Sh, So, Zm, As
9	32	3	Inverted	rpl23	Sb, On, Ta, Hv, Sh, So, Zm, As
10	33	2	Inverted	trnT-GGU	Sb, On, Ta, Hv, Sh, So, Zm, As
11	34	2	Direct	psaB, psaA	Sb, On, Ta, Hv, Sh, So, Zm, As
12	34	2	Direct	rpoC2	Sb, On, Ta, Hv, Sh, So
13	34	2	Direct	trnfM-CAU	Sb, On, Ta, Hv, Sh, So, Zm, As
14	36	3	Inverted	Intron—(ycf3 Exon1—ycf3 Exon2), IGS—(trnV-GAC—rps12_3end)	Sb, On, Ta, Hv, Sh, So, Zm, As
15	36	3	Direct	rpoC2	Sb, On, Ta, Hv, Sh, So, Zm, As
16	36	2	Inverted	trnS-GCU	Sb, On, Ta, Hv, Sh, So, Zm, As
17	37	2	Direct	rpoC2	Sb, On, Ta, Hv, Sh, So, Zm, As
18	45	3	Direct	rps8	Sb, On, Ta, Hv, Sh, Zm, As
19	45	2	Direct	rpoC2	Sb, On, Ta, Sh, So, Zm, As
20	47	2	Direct	IGS—(trnG-GCC—trnfM-CAU), Intron—(trnfM-CAU—trnG-UCC	On, Ta
21	50	3	Inverted	IGS—(psbE—petL), Intron—(rps12_3end—rps7)	Sb, On, Ta, Hv, Sh, So, Zm, As
22	52	2	Direct	IGS—(trnN-GUU-rps15)	Sb, On, Ta, Hv, Sh, So, Zm, As
23	52	4	Inverted	IGS—(ndhB-trnL-CAA)	Sb, On, Ta, Hv, Sh, So, Zm, As
24	56	2	Direct	rps18	Sb, On, Sh, So, Zm, As
25	59	2	Inverted	IGS—(psaI-rpl23)	On
26	91	3	Inverted	rp123 (69 bp)—IGS (rp123—accD), rp123 (79 bp)—IGS (rp123—rp12)	Sb, On, Ta, Hv, Sh, So, Zm, As

Includes blast hits at least 30 bp in size, a sequence identity ≥90%, and a bit-score of great than 40

Sb Sorghm bicolor, On Oryza nivara, Ta Triticum aestivum, Hv Hordeum vulgare, Sh Saccharum hybrid, So Saccharum officinarum, Zm Zea mays, As Agrostis stolonifera

Previous studies of grass chloroplast genomes identified three inversions relative to the established consensus chloroplast gene order identical to that found in tobacco (Hiratsuka et al. 1989, Doyle et al. 1992, Palmer and Stein 1986). Because inversions are often associated with repeated sequences (Palmer 1991) we examined inversion endpoint regions for repeats. We located shared repeats flanking the endpoints of the largest 28 kb inversion of grasses. Repeat analyses identified a 21 bp direct repeat in O. sativa that contains the motif GTGAGCTACCAAACTGCTCTA and flanks the inversion endpoints. This repeat has a Hamming distance of 2, and is shared by all the other grasses examined. Repeat analyses at the endpoints of the two other grass inversions failed to identify any shared repeats at the settings used in this analysis.

Our analyses identified 16–21 SSRs per genome and these are composed of di-to penta-nucleotide repeating units (Supplementary Table 3). Nearly 50% of all SSRs are tetra-nucleotide repeats with no common motif. The next most common SSR consists of di-nucleotide repeats and accounts for 30% of the SSRs with a predominant motif of TA or AT. The remaining 20% of the SSRs are composed of tri- and penta-nucleotide repeats. Of the SSRs identified, the same dinucleotide repeat (AT) is located within the coding region of the gene rpoC2 in all chloroplast genomes examined.

Intergenic spacer regions

We analyzed the similarity and divergence of IGS regions from seven grass chloroplast genomes including A. stolonifera, H. vulgare, Z. mays, O. sativa, Sorghum bicolor, Saccharum officinarum and T. aestivum. The results of these analyses are presented in Tables 3 and 4, Figs. 3 and 4, and in Supplementary Tables 1 and 2. These species were subdivided into two groups for comparative analyses based on their position in phylogenetic trees (Figs. 5, 6). The first group includes O. sativa, T. aestivum, H. vulgare and A. stolonifera and the second group contains Z. mays, Saccharum officinarum and Sorghum bicolor.

Table 3.

Analysis of intergenic spacer regions of O. sativa, T. aestivum, H. vulgare and A. stolonifera

Intergenic region	A. stolonifera/H. vulgare	O. sativa/H. vulgare	T. aestivum/H. vulgare	A. stolonifera/O. sativa	A. stolonifera/T. aestivum	O. sativa/T. aestivum
trnA-UGC:trnA-UGC	100	99	99	99	98	98
trnH-GUG:rpl2	100	91	100	91	100	91
trnA-UGC:trnI-GAU	100	94	91	92	91	91
rpl23:trnI-CAU	97	97	100	97	97	97
trnI-CAU:rpl23	97	97	100	97	97	97
rrn4.5:rrn23	92	94	100	89	92	94
rrn23:rrn4.5	91	94	100	88	92	94
trnE-UUC:trnY-GUA	89	92	100	90	89	92
trnN-GUU:trnR-ACG	88	85	100	94	88	85
trnR-ACG:trnN-GUU	88	85	100	94	88	85
rps12_5end:clpP	86	80	100	78	86	80
ndhB:rps7	98	95	95	95	95	100
rps7:ndhB	98	94	94	94	94	100
trnQ-UUG:psbK	92	91	91	91	91	100
rps16:trnQ-UUG	40	36	36	56	56	100

Intergenic spacer regions that are 100% identical in at least two of the four species are shown

Table 4.

Analysis of intergenic spacer regions of Z. mays, S. officinarum and S. bicolor

Intergenic spacer region	Z. mays/S. officinarum	Z. mays/S. bicolor	S. officinarum/S. bicolor
ndhD:psaC	100	100	100
psbJ:psbL	100	100	100
psbN:psbH	100	100	100
rrn23:trnA-UGC	100	100	100
trnA-UGC:rrn23	100	100	100
ndhB:trnL-CAA	100	99	99
trnL-CAA:ndhB	100	99	99
rps19:trnH-GUG	100	96	96
trnH-GUG:rps19	100	96	96
ndhB:ndhB	99	100	99
rps12:trnV-GAC	99	99	100
trnA-UGC:trnA-UGC	99	99	100
trnV-GAC:rps12	99	99	100
rrn16:trnV-GAC	98	98	100
trnN-GUU:trnR-ACG	98	98	100
trnR-ACG:trnN-GUU	98	98	100
trnV-GAC:rrn16	98	98	100
rpl23:trnI-CAU	97	97	100
rps2:atpI	97	97	100
rps7:rps12	97	97	100
rrn4.5:rrn5	97	97	100
trnI-CAU:rpl23	97	97	100
petG:trnW-CCA	96	96	100
ndhI:ndhA	95	100	95
psbC:trnS-UGA	95	95	100
rrn4.5:rrn23	95	95	100
rpl22:rps19	94	94	100
rpl36:infA	94	94	100
trnM-CAU:atpE	93	93	100
trnE-UUC:trnY-GUA	92	92	100
cemA:petA	91	91	100
ndhJ:ndhK	90	90	100
rps3:rpl22	89	89	100
trnA-UGC:trnI-GAU	86	86	100
psbT:psbN	69	69	100
rps12:rps7	9	9	100

Intergenic spacer regions that are 100% identical in at least two of the three species are shown below

Fig. 3.

Histogram showing pairwise sequence divergence of the intergenic spacer regions of rice (Oryza sativa), wheat (Triticum aestivum) barley (Hordeum vulgare) and bentgrass (Agrostis stolonifera) chloroplast genomes. Comparisons of 19 most variable intergenic regions with less than 80% average sequence identity. The values plotted in this histogram come from Supplementary Table 1, which shows percent sequence identities for all intergenic spacer regions. The plotted values were converted from percent identity to sequence divergence on a scale from 0 to 1 and included on the Y-axis. Asterisk indicates regions that are in the top 25 most variable intergenic spacer regions in Solanaceae (adapted from Daniell et al. 2006), plus indicates regions that are in the top 25 most variable intergenic spacer regions in Asteraceae (adapted from Timme et al. 2007)

Fig. 4.

Histogram showing pairwise sequence divergence of the intergenic spacer regions of maize (Zea mays), sugarcane (Saccharum officinarum) and sorghum (Sorghum bicolor) chloroplast genomes. Comparisons of the nine most variable intergenic spacer regions with less than 80% average sequence identity. The values plotted in this histogram come from Supplementary Table 2, which shows percent sequence identities for all intergenic spacer regions. The plotted values were converted from percent identity to sequence divergence on a scale from 0 to 1 and included on the Y-axis. Asterisk indicates regions that are in the top 25 most variable intergenic spacer regions in Solanaceae (adapted from Daniell et al. 2006), plus indicates regions that are in the top 25 most variable intergenic spacer regions in Asteraceae (adapted from Timme et al. 2007)

Fig. 5.

Phylogenetic tree of 38 taxa based on 61 plastid protein-coding genes using maximum parsimony. The tree has a length of 62,437, a consistency index of 0.407 (excluding uninformative characters) and a retention index of 0.627. Numbers above node indicate number of changes along each branch and numbers below nodes are bootstrap support values. Ordinal and higher level group names follow APG II (2003). Taxa in red are the new genomes reported in this paper

Fig. 6.

Phylogenetic tree of 38 taxa based on 61 plastid protein-coding genes using maximum likelihood. The tree has a ML value of −lnL = 348086.2268. Numbers at nodes are bootstrap support values 50%. Ordinal and higher level group names follow APG II (2003). Taxa in red are the new genomes reported in this paper

Five IGS regions (ndhD:psaC, psbJ:psbL, psbN:psbH, rrn23:trnA-UGC, trnA-UGC:rrn23) have 100% sequence identity among Z. mays, Saccharum officinarum and Sorghum bicolor, whereas no spacer regions are identical among O. sativa, T. aestivum, H. vulgare and A. stolonifera despite of their close phylogenetic relationship. Divergence among Z. mays, Sorghum bicolor and Saccharum officinarum chloroplast genomes is much less because there are only nine IGS regions with less than 80% average sequence identity versus 19 among O. sativa, T. aestivum, H. vulgare and A. stolonifera (Figs. 3, 4). Only three of the intergenic regions in the two sets of comparisons have more than 80% average sequence divergence (rpl16:rps3, psbH:petB, and rps12_3end:rps7; compare Figs. 3, 4). Some spacer regions have indels resulting in extremely low sequence identity. For example, in Z. mays, deletion of a 558 bp intergenic region between rps12 3′ end and rps7 IGS has resulted in only 9% sequence identity between Z. mays:Sorghum bicolor and Z. mays:Saccharum officinarum comparisons. Nevertheless, this region shows 100% identity between Sorghum bicolor and Saccharum officinarum (see Supplementary Table 2). Regions marked with asterisks or plus signs in Figs. 3 and 4 are in the top 25 most variable IGSs in Solanaceae (Daniell et al. 2006) and Asteraceae (Timme et al. 2007), respectively.

Variation between coding regions and cDNAs

Alignment of EST sequences and DNA coding sequences identified 15 nucleotide substitution differences in the Sorghum bicolor chloroplast genome (Table 5), 25 in the H. vulgare genome (Table 6) and 1 in A. stolonifera (not shown). Sorghum bicolor has six C–U conversions, five of which result in amino acid changes. H. vulgare also has six C–U conversions, all of which result in amino acid changes. Of these substitutions, 11 are non-synonymous and 4 are synonymous in Sorghum bicolor. In H. vulgare, 17 substitutions are non-synonymous and eight are synonymous. Sorghum bicolor experienced 1–2 substitutions per gene while H. vulgare has 1–5 variable sites per identified gene. H. vulgare and Sorghum bicolor share three variable positions in the rpoC2, psaA and atpB genes (Tables 5, 6). At the time of the analysis of A. stolonifera, there were only 9018 EST sequences available to analyze potential RNA editing sites. Comparing the coding regions of the A. stolonifera chloroplast genome to available ESTs reveals only one potential editing site. This site is located within the psbZ gene at position 54 and suggests a C–U change, which does not result in a change in the amino acid. There are 89 ESTs that show support for a C–U change, and five that don’t show the edit.

Table 5.

Differences observed by comparison of S. bicolor chloroplast genome sequences with EST sequences obtained by BLAST search of NCBI GenBank

Gene	Gene size (bp)	Sequence analyzeda	Number of variable sites	Variation type	Position(s)b	Amino acid change
atpA	1,523	1069–1523	1	C–U	1148	S–L
ndhK	746	1–297	1	C–U	128	P–L
rpoC2	4,562	2728–3143	1	C–U	2753	S–L
psaA	2,284	893–1281	1	T–G	968	L–W
atpB	1,496	551–1488	2	T–G	535	H–Q
				A–G	1466	I–V
psbJ	122	1–122	2	T–A	35	L–Q
				T–C	60	L–L
psbD	1,061	306–1061	1	G–A	741	M–I
psbC	1,421	534–1065	1	T–G	1047	G–G
psaB	2,204	95–587	1	T–G	99	S–R
ndhA	1,089	1023–1089	1	C–U	1070	S–F
rp12	843	1–511	2	C–U	14	T–M
				A–G	405	G–G
ndhI	543	1–543	1	C–U	513	I–I

^aSequence based on the gene sequence, considering the first base of the initiation codon as 1

^bVariable position is given in reference to the first base of the initiation codon of the gene sequence

Table 6.

Differences observed by comparison of H. vulgare chloroplast genome sequences with EST sequences obtained by BLAST search of NCBI Gen-Bank

Gene	Gene size	Sequence analyzeda	No of variable sites	Variation type	Nucleotide position(s)b	Amino acid change
rpoB	3231	1–2150	4	T–A	241	Y–N
				G–C	2,048	S–T
				G–U	2,050	E–L
				A–U	2,051	E–L
clpP	651	265–651	5	G–A	337	A–T
				A–U	417	E–D
				T–C	508	S–P
				A–G	598	K–E
				G–A	630	P–P
rpl2	390	1–390	1	C–U	2	T–M
psaA	2,253	117–894	3	G–C	81	A–A
				T–G	138	I–S
				C–A	396	F–L
ycf4	558	38–376	3	T–C	319	W–R
				T–C	342	R–R
				T–C	347	V–A
atpB	1,497	1–670	3	C–U	490	R–C
				A–G	663	V–V
				T–C	669	N–N
ycf3	228	1–228	1	T–A	23	N–I
rpoC2	4,434	3640–4315	1	C–U	4,025	S–L
psaJ	129	1–129	1	T–G	72	G–G
petA	963	821–963	4	T–C	870	P–P
				C–U	883	R–C
				C–U	917	S–F
				C–U	949	V–I

^aSequence based on the gene sequence, considering the first base of the initiation codon as 1

^bVariable position is given in reference to the first base of the initiation codon of the gene sequence

Phylogenetic analyses

The data matrix comprises 61 protein-coding genes for 38 taxa, including 36 angiosperms and two gymnosperm out-groups (Pinus and Ginkgo, Table 1). The aligned sequences include 46,188 nucleotide positions but when the gaps are excluded to avoid ambiguities due to insertion/deletions there are 39,574 characters. MP analyses resulted in a single most-parsimonious tree with a length of 62,437, a consistency index of 0.407 (excluding uninformative characters) and a retention index of 0.627 (Fig. 5). Bootstrap analyses indicate that 26 of the 35 nodes have bootstrap values ≥95%, five nodes have 80–94%, and four nodes have 50–79%. ML analysis results in a single tree with a ML value of −lnL = 348,086.2268 (Fig. 6). Support is very strong for most clades in the ML tree with 32 of the 35 nodes with ≥95% bootstrap values and 3 with 60–69% support. The ML and MP trees only differ in the relationships among the rosids (compare Figs. 5, 6), although this difference is not strongly supported in the ML tree (63% bootstrap value). In the MP tree the eurosid II clade is sister to a clade that includes both members of eurosid I and Myrtales, whereas in the ML tree the eurosid II clade is sister to a clade that includes the Myrtales and one member of the eurosid I (Cucurbitales).

Discussion

Significance of transgene integration into grass chloroplast genomes

Although plastid transformation has been accomplished via organogenesis in a number of eudicots, two major obstacles have been encountered to extend plastid transformation technology to crop plants that regenerate via somatic embryogenesis: (1) the expression of transgenes in non-green plastids, in which gene expression and gene regulation systems are quite distinct from those of mature green chloroplasts, and (2) our current inability to generate homoplastomic plants via subsequent rounds of regeneration, using leaves as explants. Despite these limitations, plastid transformation has recently been accomplished via somatic embryogenesis in several eudicot crops, including Glycine max L. Merr. (soybean), Daucus carota L. (carrot) and Gossypium hirsutum L. (cotton, Dufourmantel et al. 2004, 2005; Kumar et al. 2004a, b) and foreign genes have been expressed in high levels in non-green plastids, including proplastids and chromoplasts (Kumar et al. 2004a). Breakthroughs in plastid transformation of recalcitrant crops, such as G. hirsutum and G. max, have raised the possibility of engineering plastid genomes of other major crops via somatic embryogenesis. To date, only fragmentary data were reported for O. sativa plastid transformation (Khan and Maliga 1999). However, a promising step toward stable plastid transformation in O. sativa has been reported recently (Lee et al. 2006b). Transplastomic O. sativa plants generated in this study exhibited stable integration and expression of the aadA and sgfp transgenes in their plastids. Moreover, the transplastomic O. sativa plants generated viable seeds, which were confirmed to transmit the transgenes to the T1 progeny. Unfortunately, conversion of the transplastomic O. sativa plants to homoplasmy was not successful, even after two generations of continuous selection. Thus, tissue culture and selection of transformed events continues to be a major challenge.

The success of chloroplast genetic engineering of crop plants is dependent, at least in part, on access to conserved spacer regions for inserting transgenes. The availability of sequences of complete chloroplast genomes for multiple crop plants in the grass family should facilitate plastid genetic engineering. Several studies have demonstrated that the use of IGS regions that have low sequence identities between the target genome and the flanking sequences in the chloroplast transformation vectors can result in substantially lower frequencies of transformants (Nguyen et al. 2005; Ruf et al. 2001; Sidorov et al. 1999). Given the low number of intergenic sequences that have high sequence identities among the seven sequenced chloroplast genomes (Tables 3, 4) it is unlikely that a single, highly conserved IGS region will be appropriate throughout the grass family. Among Solanaceae chloroplast genomes, only four spacer regions have 100% sequence identity among all sequenced genomes and three of these regions are within the IR region (Daniell et al. 2006). Five IGS regions have 100% sequence identity among Z. mays, Saccharum officinarum and Sorghum bicolor chloroplast genomes. Thus the variation in the IGS region is quite similar between solanaceae and grass chloroplast genomes. However, not a single IGS region is identical among O. sativa, T. aestivum and H. vulgare chloroplast genomes. Thus, conservation of IGS regions is not uniform even within the same family. However, it is noteworthy that the same IGS regions have very low sequence identity within Poaceae, Solanaceae and Asteraceae, as discussed below.

Genome organization and evolutionary implications

Organization and evolution of grass chloroplast genomes

The organization of chloroplast genomes is highly conserved in most land plants but alterations in gene content and order have been identified in several lineages (Raubeson and Jansen 2005). Notable rearrangements are known in two families with many crop species, a single 51-kb inversion common to most papilionoid legumes (Palmer et al. 1988; Doyle et al. 1996; Saski et al. 2005) and three inversions in the grasses (Quigley and Weil 1985; Howe et al. 1988; Hiratsuka et al. 1989; Doyle et al. 1992; Katayama and Ogihara 1996). The H. vulgare, Sorghum bicolor and A. stolonifera chloroplast genomes contain all three of the inversions present in grasses.

Gene order and content of the sequenced grass chloroplast genomes are similar. However, two microstructural changes have occurred. First, the expansion of the IR at the SSC/IR boundary that duplicates a portion of the 5′ end of ndhH is restricted to the three genera of the subfamily Pooideae (Agrostis, Hordeum and Triticum). These three genera form a monophyletic group in the phylogenetic trees based on DNA sequences of protein-coding genes (Figs. 5, 6) but the extent of the IR expansion differs in each of the three genera (32, 69 and 58 amino acids in wheat, barley and bentgrass, respectively). Thus, it is not possible to determine if there have been three independent expansions or a single expansion followed by two subsequent contractions. Second, a 6 bp deletion in ndhK (Supplementary Figure 1) is shared by Agrostis, Hordeum, Oryza and Triticum, and this event supports the sister relationship between the subfamilies Erhartoideae and Pooideae (Figs. 5, 6).

Other than the IR, repeated sequences are considered to be relatively uncommon in chloroplast genomes (Palmer 1991). The analysis of the repeated sequences of grass chloroplast genomes revealed 26 groups of repeats shared among various members of the family (Table 2, Fig. 2). Furthermore, 17 of the 26 repeats are shared among all eight of the chloroplast genomes examined suggesting a high level of conservation of repeat structure among grasses. Examination of the location of these repeats suggests that all of them occur in the same location, either in genes, introns or within IGS regions. This high level of conservation of both sequence identity and location suggests that these elements may play a functional role in the genome, although we cannot rule out the possibility that this conservation may simply be due to a common ancestry. Because organellar genomes are often uniparentally inherited, chloroplast DNA polymorphisms have become a marker of choice for investigating evolutionary issues such as sex-biased dispersal and the directionality of introgression (Willis et al. 2005). They are also invaluable for the purposes of population-genetic and phylogenetic studies (Bryan et al. 1999; Raubeson and Jansen 2005). Also, knowledge of mutation rates is important because they determine levels of variability within populations, and hence greatly influence estimates of population structure (Provan et al. 1999). Based on our mining for SSRs, we identified 16–18 SSRs within the nine genomes examined. These initial findings indicate a potential to test and utilize SSRs to rapidly analyze diversity in germplasm collections.

Previous studies of grass chloroplast genomes have identified three inversions in the family (Quigley and Weil 1985; Howe et al. 1988; Hiratsuka et al. 1989; Doyle et al. 1992; Katayama and Ogihara 1996). Our analysis of the inversion endpoints indicate that there are shared repeats flanking the endpoints of the largest 28 kb inversion. This first inversion has endpoints between trnG-UCC and trnR-UCU at one end and rps14 and trnfM-CAU at the other creating an intermediate form of the chloroplast genome prior to the second inversion when compared to N. tabacum (Hiratsuka et al. 1989; Doyle et al. 1992). Repeat analyses identified a 21 bp direct repeat in O. sativa that flanks the inversion endpoints, and this repeat is shared by all other grasses examined. It is likely that the shared repeat facilitated this large inversion by intramolecular recombination. Two additional inversions, one largely overlapping the 28 kb event, subsequently gave rise to the gene order observed in O. sativa and T. aestivum (Hiratsuka et al. 1989). The endpoints of the second inversion (ca 6 kb) occur between trnS and psbD on one end and trnG-UCC and trnT-GGU on the other (Doyle et al. 1992). The third inversion has endpoints between trnG-UCU and trntT-GGU and trnT-GGU and trnE-UUC. This inversion is quite small and accounts for the inverted orientation of trnT-GGU (Hiratsuka et al. 1989). Our repeat analyses found no shared repeats that may have played a role in these two inversions. Chloroplast genome organization is also known from other monocots based on both gene mapping and complete genome sequencing (de Heij et al. 1983; Chase and Palmer 1989; Chang et al. 2006). Four non-grass monocots Spirodela oligorhiza (Lemnaceae), two orchids (Oncidium excavatum and Phalaenopsis aphrodite), and members of the Alliaceae (Allium cepa), Asparagaceae (Asparagus sprengeri) and Amaryllidaceae (Narcissus × hybridus) have the same gene order as tobacco. Thus, the inversions in H. vulgare, Sorghum bicolor and A. stolonifera reported here are confined to the grass family as was previously suggested by Doyle et al. (1992).

Comparisons of DNA and EST sequences for H. vulgare, Sorghum bicolor and A. stolonifera identified many differences (Tables 5, 6), most of which are not likely due to RNA editing. Previous investigations of RNA editing in chloroplast genomes in the angiosperms N. tabacum (Hirose et al. 1999) and Atropa (Schmitz-Linneweber et al. 2002) and in the fern Adiantum (Wolf et al. 2004) indicated that RNA edits only result in C–U changes. In the case of H. vulgare, Sorghum bicolor and A. stolonifera, only seven differences in the DNA and EST sequences were C–U changes. Thus, these are the only changes that may be the result of RNA editing. The other 9 differences in Sorghum bicolor and 19 differences in H. vulgare are likely due to either polymorphisms resulting from the use of different plants or cultivars or sequencing errors. In the case of A. stolonifera, only one C–U change was found. This could be attributed to the lack of available expression information since only 9,018 EST sequences were available for A. stolonifera when the analysis was performed, suggesting a need for more comprehensive investigations into the chloroplast and nuclear transcriptomes.

Several recent comparisons of DNA and EST sequences for other crop species including G. hirsutum (Lee et al. 2006a), Vitis vinifera (Jansen et al. 2006), Citrus sinensis L. (Bausher et al. 2006), carrot (Ruhlman et al. 2006), Lactuca and Helianthus (Timme et al. 2007) and Solanum lycopersicum and Solanum bulboscastanum (Daniell et al. 2006) have identified both putative RNA editing sites and possible sequencing errors. The much greater depth of coverage in the chloroplast genome sequences (generally 4-20X coverage) suggests that most of the differences other than changes from C to U are likely due to errors in EST sequences.

Phylogenetic utility of intergenic spacer regions

Phylogenetic studies at the inter- and intraspecific levels in plants have relied extensively on IGS regions of chloroplast genomes because the coding regions are generally too highly conserved at these lower taxonomic levels (Kelchner 2002; Raubeson and Jansen 2005; Jansen et al. 2005; Shaw et al. 2005, 2007). There have been many efforts to identify the most divergent IGSs for phylogenetic comparisons at lower taxonomic levels with the hope that some universal regions could be found for angiosperms (Shaw et al. 2005, 2007, Daniell et al. 2006; Timme et al. 2007). Only two previous studies have performed genome-wide comparisons among multiple, sequenced genomes in the families Asteraceae (Timme et al. 2007) and Solanaceae (Daniell et al. 2006). Comparison of our results in the Poaceae with these earlier studies indicates that there are considerable differences regarding which IGS regions are most variable in these three families (see asterisks and plus signs in Figs. 3, 4). Only three (Fig. 4) to five (Fig. 3) of the 25 most variable regions of Solanaceae are among the most variable IGSs in grasses. The overlap in the regions with high sequence divergence between the Asteraceae and grasses is higher, with three (Fig. 4) to nine (Fig. 3) of the most variable IGS regions in the Poaceae among the 25 most variable regions in the Asteraceae. Overall, genome-wide comparisons among these three families indicate that there may be few universal IGS regions across angiosperms for phylogenetic studies at lower taxonomic levels. Thus, it will likely be necessary to identify variable IGS regions in chloroplast genomes for each family to locate the most appropriate markers for phylogenetic comparisons.

Phylogenetic relationships of angiosperms

During the past three years there has been a rapid increase in the number of studies using DNA sequences from completely sequenced chloroplast genomes for estimating phylogenetic relationships among angiosperms (Goremykin et al. 2003a, b, 2004, 2005; Leebens-Mack et al. 2005; Chang et al. 2005; Lee et al. 2006a; Jansen et al. 2006; Ruhlman et al. 2006; Bausher et al. 2006; Cai et al. 2006). These studies have resolved a number of issues regarding relationships among the major clades, including the identification of either Amborella alone or Amborella + Nymphaeales as the sister group to all other angiosperms, strong support for the monophyly of magnoliids, monocots and eudicots, the position of magnoliids as sister to a clade that includes both monocots and eudicots, the placement of Vitaceae as the earliest diverging lineage of rosids, and the sister group relationship between Caryophyllales and asterids. However, some issues remain unresolved, including the monophyly of the eurosid I clade and relationships among the major clades of rosids. The phylogenetic analyses reported here (Figs. 5, 6) with expanded taxon sampling are congruent with these earlier studies so our discussion will focus on relationships among grasses.

Our study has added complete chloroplast genome sequences for three genera of grasses representing two subfamilies (Pooideae and Erhartoideae, sensu Grass Phylogeny Working Group 2001). This expands the number sequenced grass genera to seven from three different subfamilies, Panicoideae, Pooideae and Erhartoideae. Our phylogenetic trees (Figs. 5, 6) indicate that the Erhartoideae is sister to the Pooideae with weak to moderate bootstrap support (60 or 81% in ML and MP trees, respectively). The sister relationship of these subfamilies is also supported by a 6 bp deletion in ndhK (Supplementary Figure 1). This result is congruent with phylogenetic trees based on sequences of six genes (four chloroplast and two nuclear, Grass Phylogeny Working Group 2001). This multigene tree, which included 68 genera of grasses, also provided only moderate bootstrap support (71%) for a close phylogenetic relationship between these two subfamilies. Furthermore, the clade including Pooideae and Erhartoideae also contained members of the Bambusioideae. Clearly, many additional chloroplast genome sequences are needed from the grasses to provide sufficient taxon sampling to generate a family-wide phylogeny based on whole genomes.