|Home | About | Journals | Submit | Contact Us | Français|
Conceived and designed the experiments: XZ SH ISAM JY. Performed the experiments: YF YY LP. Analyzed the data: YF HW TZ. Contributed reagents/materials/analysis tools: MY XY. Wrote the paper: YF HW TZ.
Based on next-generation sequencing data, we assembled the mitochondrial (mt) genome of date palm (Phoenix dactylifera L.) into a circular molecule of 715,001 bp in length. The mt genome of P. dactylifera encodes 38 proteins, 30 tRNAs, and 3 ribosomal RNAs, which constitute a gene content of 6.5% (46,770 bp) over the full length. The rest, 93.5% of the genome sequence, is comprised of cp (chloroplast)-derived (10.3% with respect to the whole genome length) and non-coding sequences. In the non-coding regions, there are 0.33% tandem and 2.3% long repeats. Our transcriptomic data from eight tissues (root, seed, bud, fruit, green leaf, yellow leaf, female flower, and male flower) showed higher gene expression levels in male flower, root, bud, and female flower, as compared to four other tissues. We identified 120 potential SNPs among three date palm cultivars (Khalas, Fahal, and Sukry), and successfully found seven SNPs in the coding sequences. A phylogenetic analysis, based on 22 conserved genes of 15 representative plant mitochondria, showed that P. dactylifera positions at the root of all sequenced monocot mt genomes. In addition, consistent with previous discoveries, there are three co-transcribed gene clusters–18S-5S rRNA, rps3-rpl16 and nad3-rps12–in P. dactylifera, which are highly conserved among all known mitochondrial genomes of angiosperms.
The widely-accepted hypothesis about the origin of the mitochondrion assumes that it descended from an endosymbiontic event involving an α-proteobacterium-like organism and the common ancestor of eukaryotes . Evolving from algae to land plants, including bryophytes and angiosperms, plant mitochondrial (mt) genomes have increased their sizes, especially in the non-coding region. Among land plants, bryophytes, i.e., liverworts, mosses, and hornworts, represent the basal forms. They have similar gene order, genome size, and a fraction of non-coding sequences , . As evolution continues, land plants gained new mechanisms to facilitate frequent gene exchange between mitochondrial and chloroplast genomes as well as between mitochondrial and nuclear genomes , . For instance, mitochondrial genomes of angiosperms have long been known for their slow evolutionary rate , existence of subgenomic circles in addition to a master genomic circle , extraordinarily large and highly variable genome sizes , trans-splicing of group II introns , high density of RNA editing , , divergent non-coding sequences , and frequent gene transfer . The inter-genomic gene transfer, together with the continuing increase of non-coding DNA sequences, leads to a broad size range in angiosperm mt genomes, which as of today is from ~200 to 2400 kb based on the known mt sequences and experimental estimations , ; up to date (July, 2011), there have been more than 40 plant mt genomes sequenced, including 22 angiosperm mt genomes (http://www.ncbi.nlm.nih.gov/genomes/GenomesGroup.cgi?taxid=33090&opt=organelle).
Phoenix dactylifera L., also known as date palm, is economically the most important plant in the Middle East and North Africa , and it is estimated to have more than 450 cultivars or varieties in the Kingdom of Saudi Arabia and nearly 2,000 varieties around the world . Therefore, sequencing its mitochondrial genome, together with its nuclear  and chloroplast genomes , is of essence in improving its agricultural, horticultural, and nutritional values. In this study, combining data from two next-generation sequence platforms, pyrosequencing (Roche GS FLX) and ligation-based sequencing (Life Technologies SOLiD), we assembled P. dactylifera mt-genome (cultivar Khalas, Al-Hasa Oasis, Saudi Arabia) –the first from the Arecaceae family. In addition, analysis of the mt genome sequence and transcriptomic data are of importance in revealing mechanisms underlying mitochondrial genome evolution and the unique evolutionary status of P. dactylifera among angiosperms. Furthermore, based on the data from three commonly-grown cultivars, we also investigated RNA editing sites and SNPs within the species.
We assembled the P. dactylifera master mt chromosome into a 715,001 bp circular molecule (Figure 1; the assembling details are described in Materials and Methods) with an average GC content of 45.1%; it is now the fourth largest mt genome sequenced after those of Cucumis sativus (1,555,935 bp) , Cucurbita pepo (982,833 bp), and Vitis vinifera (773,279 bp). Its protein coding sequence is composed of only 6.5% of the genome (46,770 bp) and this gene content is similar to other published angiosperm mt-genomes (Table S1 and Table S2). The rest, also the majority (93.5%) of the genome, is composed of non-coding (the cp-derived regions are also considered as non-coding in this regard), which harbors 0.33% tandem and 2.3% long repeats (the repeat lengths are greater than 50 bp). RNA genes and intron sequences are 1.1% and 4.3% of the mt genome, respectively. This mt genome also contains the second highest proportion (10.3%) of cp-derived sequences among the sequenced mt genomes to date, of which several intact genes, such as petA, petG, petL, psaJ, psbT, rpl20, rpl33, and rps8 are identified. Since the age of the cp-derived sequences or time when the sequences inserted into mt genomes varies greatly , we are unable to prove whether these genes are actually transcribed or active since we extracted the total RNA (contains both nuclear and organellar transcripts) from each tissue for constructing transcriptomic libraries among which the expression data of the cp-derived sequences and authentic cp sequences are impossible to separate (see Materials and Methods for more details). Most of P. dactylifera mt genome is rather diverged from other angiosperms. For example, only ~21% of P. dactylifera mt genome sequence is shared (over 70% identity) by Vitis, Oryza and Bambusa, and even less by Zea (~15%) and Arabidopsis (~11%). In addition, consistent with results from previous studies, we observed that three co-transcribed gene clusters, 18S-5S rRNA, rps3-rpl16, and nad3-rps12, are conserved in other angiosperm mt genomes . We summarized general mt genome features including size variations, AT content, and intron types of 15 non-redundant sequenced plant mt genomes (including 12 higher plants and three lower plants) in Table 1. Our phylogenetic analysis based on 22 concatenated conserved genes among 15 selected mt genomes (Figure 2) revealed that P. dactylifera appears to be the more basal among monocots.
The P. dactylifera mt genome contains at least 38 protein-coding genes and five complete ORFs, most of these genes encode proteins of the electron transport chain, such as nine subunits of nicotinamide adenine dinucleotide dehydrogenase (complex I), apocytochrome b (complex III), three subunits of cytochrome c oxidase (complex IV), five subunits of ATP synthase F1 (complex V) and four cytoplasmic membrane proteins required for cytochrome c maturation (Table 2).
We compared these protein coding genes to 11 other sequenced angiosperm mt genomes (Table S2). First, P. dactylifera mt genome does not have the genes encoding respiratory chain complex II, such as sdh3 and sdh4, which are only found in two dicots, Nicotiana tabacum and V. vinifera. Second, our assembly is similar to V. vinifera, and both contain one copy of RNA polymerase gene harboring a conserved domain characteristic of pfam00940 superfamily of polymerases . Third, rps14 present in Brassica napus and V. vinifera is also found here, whereas rps11, another ribosomal protein gene, is exclusively detected in our assembly. Both genes have full open reading frame (ORF) and are likely functional in date palm, though in many other known angiosperm mt genomes they are either pesudogenes or transferred into nuclear genomes , . Fourth, the recently identified rpl10 gene, being identified as orf-bryo1 in vascular plants and charophycean green algae  and orf168-related sequences in bryophytes and angiosperms , seems to be interrupted in our assembly and possibly because of a frame shift event. Fifth, we found several pseudogenes in our assembly, which appear intact in other mt genomes, such as orf99-b (as orf100-ψ in our gene list) in Zea mays and cp-derived gene psbT in V. vinifera. In addition, some of the universally expressed ribosomal genes, including three rRNA genes (5S, 18S, and 26S ribosomal RNA genes), are also unambiguously identified ; 5S and 18S rRNA genes are also closely related and distant from 26S rRNA gene in date palm mt genome.
A genome-wide screening yielded 30 full-length tRNA sequences (Table S3) in our assembly; among them, 12 seem to be cp-derived, which exhibit higher sequence identity (>98%) to their chloroplast counterparts than their mitochondrial counterparts , and three predicted tRNAs seem to have introns. Among these 30 tRNA genes five amino acids (A, L, R, T, and V) are not encoded, although tRNAs for 20 amino acids are necessary for protein synthesis in mitochondria. In addition, having compared the date palm tRNA gene content to those of seven other plants mt and cp genomes (Figure 3), we conclude that there are 10 tRNA genes, among which nine encoding tRNAs for the five amino acids, are actually lost after the divergence of liverworts from seed plants. These results suggest that the missing tRNAs are supplied by either the chloroplast or nuclear genomes. In addition, we found that four mt tRNA genes of higher plants are gradually lost and replaced by cp-derived tRNA . The reason why mt tRNA genes are replaced by both cp-derived and nuclear counterparts remains an open question. There is also another possibility–all mt tRNA genes may eventually be replaced and what we observed here is only an intermediate and dynamic process.
Chloroplast and mitochondrial genomes are known to share sequences due to frequent gene transfer events , , . Frequent DNA transfer from cpDNA to mtDNA occur as far back as the common ancestor of the extant gymnosperms and angiosperms, about 300 MYA (million-years-ago) . Our assembly contains more than 100 fragments of chloroplast origin (over 80% identity), ranging from 50 to 6,521 bp in length (Table S4). The total fraction of chloroplast DNA sequences present in P. dactylifera mt genome is 73,691 bp, corresponding to 10.3% of the whole mt genome, and 46.5% of P. dactylifera cp-genome. The proportion of cp-derived sequences in our mt genome assembly is comparable to the two large sequenced plant mt genomes, V. vinifera (8.8%) and Cucurbita pepo (11.6%) , but larger than those of the other known plant mt genomes (Table 1). These results suggest that chloroplast DNA sequence insertion is an important mechanism for plant mitochondrial genome size expansion and sequence diversity.
Most of chloroplast sequence insertions in P. dactylifera mt genome are unique, as evident from the observation that only nine out of 44 insertions (over 200 bp) have full-length homologous sequences shared by other known mt genomes (>90% length coverage, >70% identity) (Table 3). Among the nine cp-derived homologous sequences, six and five of them are also found in Vitis and Bambusa, respectively, whereas none is found in A. thaliana. These nine insertions tend to have higher GC content, resembling that of mt genomes as compared to the unique and possibly new insertions (Figure S1), which suggests that these cp-derived sequences did, in some extent, gradually increase their GC content to become similar to their host mtDNAs.
We identified 23 group II introns in 10 protein-coding genes, including four trans-spliced introns of nad1 and nad5, and 20 cis-spliced introns in ccmFc, cox2, nad1, nad2, nad4, nad5, nad7, rpl2, and rps3. No group I intron was discovered in our assembly. In general, the functional mitochondrial rRNA and tRNA genes of the sequenced angiosperm mt genomes do not possess introns, but we found three intron-containing tRNA genes in our assembly: trnK-TTT, trnN-ATT and trnSup-CTA, and we have yet to validate if they are functional or not.
Mitochondrial RNA editing is essential for functional protein synthesis since nearly all plant mt mRNAs are edited , ,  and it modifies amino acids and generates new start or stop codons , , , , , and it has been documented in most plants except algae and mosses. It suggests that this cellular process is ancient arisen in early land plants after they split from Bryophyta . We predicted nearly 600 putative RNA editing sites (Table S5) using PREP-Mt –an effective tool identifying C-U editing sites. We found that the nad4 gene contains the most edited sites (59). In addition, our comparative analysis revealed that 305 (51.5%) and 278 (47.0%) C-U editing sites in date palm are shared by O. sativa and A. thaliana, respectively (Figure 4). Experimental examination confirmed 40 of 41 predicted C-U editing sites in five randomly chosen genes (atp1, atp4, atp9, rpl116 and rps19) using cDNA sequences (Table S6) and additional nine sites not detected by PREP-Mt were identified. We also compared their tissue disparity between mRNA transcripts extracted from yellow and green leaves, but no obvious tissue-specific RNA editing patterns are yet identified among these five genes, although reports in the literature indicated that the extent of atp6 editing is significantly different among tissue types . Therefore we assume the tissue-specific RNA editing patterns may be only detectable in certain genes, cell types, and developmental stages .
Plant cells usually possess hundreds to thousands mitochondria or copies of mt genomes that can be regarded as a population when genetic heterogeneity is to be investigated. High throughput next-generation sequence technologies provide us the opportunity to survey single nucleotide polymorphisms in the same or different species (subspecies or cultivars) by mapping reads to a reference sequence and to each cultivar. The polymorphisms within the same cultivar genome (intra-varietal SNPs) and among different cultivars genomes (inter-varietal SNPs), discovered by high-coverage of reads, can also be separated into major and minor genotypes based on simple read counts. Here, we use three runs of SOLiD fragment data from each of the three cultivars (Khalas, Fahal, and Sukry) sequenced in our study for intra-varietal (Table 4) and inter-varietal SNP identification (Table 5). We identified 651, 703, and 731 intra-varietal SNPs in cultivar Khalas, Fahal, and Sukry, respectively, estimated to have a polymorphism rate of one in 1,000 bp, which is about two times higher than that of date palm chloroplast  but is only about one tenth of rice mt genome . We should be cautious here since different SNP analysis methods are applied because of the distinct sequencing strategies used in sequencing these mt or cp genomes. The rates of each variation type among these intra-varietal SNPs of the three cultivars are very similar except the types (such as A to T or G to C and vice versa) that do not change GC contents are less represented. These SNPs can also be separated into transition and transversion types, and as a result, there are 297, 325, 287 transitions and 354, 378, 347 transversions for Khalas, Fahal, and Sukry, respectively. The rate of transversions is slightly higher than that of transitions, though in chloroplast transversion (52) is 2× that of transition (26) .
All together, there are 120 candidate SNP sites identified among the three cultivars (Table 5), with an inter-varietal polymorphism rate of 0.017%, similar to that of subspecific (between subspecies) polymorphisms between rice cultivar 93-11 and PA64S, ~0.02% . The inter-varietal SNPs are predominantly found in non-coding regions, only seven SNPs were found in coding sequences (all are located in 26S and 18S rRNA genes; Table S7): two between Khalas and Fahal, six between Khalas and Sukry, and six between Fahal and Sukry (Table 5). As to the remaining 113 inter-varietal SNPs residing in non-coding regions (Table S8), 79 of them are between Khalas and Fahal, 91 between Khalas and Sukry, and 50 between Fahal and Sukry (Table 5). Such a distribution implies that Fahal and Sukry are more related than either one of them to Khalas.
P. dactylifera mt genome has much less repetitive sequences as compared to those of other known angiosperms . Only one long palindromic sequence with repeat unit longer than 1000 bp was identified (Table S9) and no inverted repeats were found. Overall, long repeats (>50 bp) only account for 2.3% of the genome, even lower than that of Vitis (2.9%) and Vigna radiata (2.7%) , which contain the lowest repeat contents among sequenced angiosperm mt genomes, whereas Tripsacum and Oryza contains 36.4% and 30.4% long repetitive sequences, respectively. This situation also applies to tandem repeats, which constitute only 0.33% of the genome (Table S10). Among the examined 15 plant mt genomes, whose tandem repeat contents range from 0.08% (N. tabacum) to 6.13% (Cycas taitungensis), only three mt genomes, those from N. tabacum, O. sativa and Chara vulgaris, contain tandem repeats lower than date palm (Figure 5). It is well known that plant mitochondria are exceptionally flexible in genome size and structure, and the accumulation of repetitive sequences often results in high sequence divergence. For instance, Cucurbita mt genome contains 38% of short repeats (19–621 bp in length) that make it the largest reported mt genome so far , whereas maize expanded its mt genome size by duplication of large sequences . Therefore, it is rather unusual that date palm mitochondrial genome is both lower in tandem repeat content and rare in large duplications. It seems that larger mt genomes of angiosperms tend to have shorter repeat lengths when long repeats are compared. For instance, mt genomes of Cucurbita (982 kb), Vitis (773 kb), and P. dactylifera (715 kb) have their largest repeat lengths of 621 bp, 651 bp, and 1,171 bp, respectively.
The mt-genome is transcribed by a phage-type RNA polymerase encoded in the nuclear genome . The transcription process is rather complex characterized by splicing, editing, terminus processing, and multiple promoters . In addition, mitochondrial genome transcription is reported to be capable of adapting to specific regulation . Here, in order to better understand tissue-specific mt gene regulation and the contribution of mt genes to the development of different tissues, we performed a thorough transcriptome survey across eight date palm tissues (Figure 1 and Figure S2) using high-performance next-generation sequencers. We discovered that ~30.8% regions of our assembly are transcribed (Table 6), slightly lower than that of the rice (~48.5%) , with an average sequence coverage of ~44× calculated based on 40 conserved house-keeping genes (Figure 6). On the one hand, our whole genome level gene expression profiling indicated that two tissues, green leaf and fruit, have the most abundant transcripts (Figure S2) but have the lowest gene expression level in terms of RPKM value (reads per kilobase of exon model per million mapped reads)  (Figure 6). Male and female flowers, root, and bud, on the other hand, tend to have higher gene expression levels but less in transcript abundance than the leaves. We assume that developing tissues, such as yellow leaf, bud, and root, need more energy than the relatively mature tissues, such as green leaf and fruit. By the same token, the highly expressed genes in female and male flowers are possibly related to flower development that not only depends on a set of nuclear genes but also on the coordinated action of mitochondrial genes . It is possible that the variable expression of mt genome-encoded genes is relevant to the copy number variation of mt genomes (similar to the number of mitochondria per cell) ,  or its changing status in tissue development. In addition, several other obvious tissue-specific gene expression patterns can be observed. First, consistent with a previous study that atp1 gene prefers to express in pollen mother cells , we also detected that the transcript of atp1 is obviously more abundant in male flower than in other tissues examined. There is another gene matR that encodes a maturase-related protein also expressed in a relative higher level. Previous study revealed that this gene suffers from modest RNA editing in maize and soybean and was predicted to be functional . Our results here indicate that this gene should have utmost importance in male flower development. Second, the maturation process of yellow to green leaves seems to involve the suppression of about half of the 40 mt house-keeping genes, and the fact is further confirmed that developing tissues are more affected by mt gene expression. Third, during seed maturation, half of the genes were found to be up-regulated when compared to the mt gene expression pattern in fruit. Fourth, interestingly, we found that the gene expression patterns of yellow leaves and seeds are quite distinct–down-regulated genes in one tissue tend to be highly regulated in the other tissue–except that of ccmFn, cox2, rps1, rps3, and rps19 which have no obvious differences between these two tissues. Fifth, two genes, rps1 and rps19, are found clearly highly expressed in root as compared to other tissues. The functional roles they play in root development still need further experimental confirmation. Sixth, consistent with previous studies, rRNA gene transcripts are found to be more abundant, ~9–13 fold than protein coding genes , but our large-scale transcriptomic analysis reveals a much higher transcription level changes ~50–400 fold than the average level of protein-coding genes according to RPKM values, and the order of expression levels for the ribosomal RNAs is 5S rRNA >26S rRNA >18S rRNA.
As the first of the palm family plants, P. dactylifera mt genome displays several unique features. First, it positions at the root of the known monocot mt genomes. Second, it has a very low level of repeat content and shows abundant RNA editing events. Third, it exhibits a high level of chloroplast sequence insertions as compared to other known angiosperm mt genomes. Furthermore, our large-scale transcriptome analysis revealed that ~30.8% of its sequences are transcribed and show obvious tissue-specific gene regulation patterns, among which both female and male flowers, root, and bud exhibit higher gene expressions than other sampled tissues. Our complete mt genome sequence assembly represents a new addition to the growing number of plant mt genomes in the public databases and paves a way for further investigations on mitochondrial biology of seed plants.
We used three domestic P. dactylifera cultivars, Khalas (male and female), Sukry (female), and Fahal (male), for this study. Tissue samples from adult date palm trees grown in Al-Hasa Oasis of Kingdom of Saudi Arabia are harvested, including soft bud, flower (male and female), fruit, root, yellow leaf (young), and green leaf (old). We disinfected the samples with 75% ethanol and froze them in liquid nitrogen immediately. For longer term storage, they are stored in −80°C freezer until use.
The P. dactylifera mt genome sequences are produced as part of the Date Palm Genome Project (DPGP, a joint effort between KACST and CAS). Genomic DNA was extracted from 50 g soft bud tissues according to the CTAB-based method. We used 5 µg purified DNA for shearing and constructing fragment libraries following the GS FLX Titanium general library preparation protocol. The ssDNA libraries were amplified with emulsion-PCR and enriched, and the samples were sequenced on Roche/454 GS FLX platform.
SOLiD long mate pair (LMP) libraries of the three cultivars were constructed by following SOLiD Library Preparation Guide (SOLiD 4.0) and at least 20 µg genomic DNA was used depending on different insert sizes (600–6000 bp). After emulsion PCR and beads enrichment (EZ beads system, AB), template beads of each LMP library were deposited to 2 quarter of slide and then loaded onto a SOLiD 4.0 instrument.
For transcriptomic study, tissue samples were grinded into fine powder followed by CTAB-based RNA extraction, and 2.5 M LiCl was used to remove polysaccharides. 0.5 µg rRNA-depleted total RNA (RiboMinus Plant Kit, Invitrogen) were used to construct transcriptomic libraries according to the instruction from SOLiD Total RNA-Seq Kit.
We separated candidate mt genome reads from eight Roche/454 GS FLX runs based on 40 published plant mt genome sequences (identity ≥80% and E-value ≤10−5). About 1.5 millions reads were obtained and assembled by using Newbler (version 2.3 with default parameters)–a de novo sequence assembly software provided by Roche. As a result, we obtained 29 mt genome contigs (total ~438 kb) with an average length of 15 kb. These contigs were extended to 662 kb by adding additional Roche/454 reads. Subsequently, SOLiD mate-pair data (2×50 bp libraries) with insertion sizes of 1–2 kb and 3–4 kb were used to construct scaffold (50-nt overlap cutoff and less than 2-nt mismatch). A total of 3,918 homopolymers with repeat unit ranging from 5 to 11 were verified and revised based on SOLiD fragment data using BFAST program (version 0.6.4d) . At last, 715,001 bp complete mt genome was assembled with an average sequence depth 130×. The final genome sequence was validated by SOLiD LMP data with insertion sizes of 3–4 kb, 4–5 kb, and 5–6 kb in a 2 kb sliding window with variable step sizes; we show the result from an analysis using 5–6 kb insert size in 15-kb step size in Figure 1.
The complete sequence of the date palm mt genome was deposited to GenBank (accession number JN375330).
A preliminary annotation was carried out by mapping final genome sequence with BLAST (identity >90% and overlap>90%) ,  hits to known mitochondrial genes, and subsequently by testing for consistency of the ORFs using NCBI online tool the ORF finder (http://www.ncbi.nlm.nih.gov/projects/gorf/, the standard genetic code was applied). The exact gene and exon boundaries were determined by alignment of homologous genes from several common mt genomes (Table S2) and verified based on transcriptomic data. The tRNA genes were identified by using a local chloroplast and mitochondrial tRNA database, BLAST search tools , , and the help of tRNAscan-SE program (version 1.4 and default parameters were used) . Both group I and group II introns were predicted by using an online software Rfam (version 10.1; http://rfam.sanger.ac.uk/; default parameters) . Homology search using BLAST ,  was carried out to identify chloroplast-derived regions in the mt genome assembly (over 80% sequence identity; E value ≤1e-5; length >50 bp).
We predicted putative RNA editing sites in protein-coding genes using the PREP-mt web-based program (http://prep.unl.edu/) . To achieve a balanced tradeoff between the number of false positive and false negative sites, the cutoff score (C-value) was set to 0.6 as suggested by the author. All other parameters are set to default values. We also verified some of the RNA editing sites in five genes (atp1, atp4, atp9, rpl116, and rps19) across the two leaf tissues (yellow and green leaves) using cDNA data from Roche/454 GS FLX system (NCBI accession number SRA045434.3). The five genes are chosen randomly, whose cDNA sequences are full-length and better in quality.
We carried out both intra-varietal and inter-varietal SNP analysis across three cultivars: Khalas, Fahal, and Sukry. Three runs of SOLiD LMP reads for each cultivar (about 60 Gb) were mapped to the reference mt genome (Khalas) by using BioScope software (version 1.3). The mapping results were then used for SNP identification based on a Bayesian algorithm according to the BioScope Software User Guide.
We identified repetitive sequences, including forward, palindromic, reverse, and complemented repeats, using the REPuter (version 2.74; with a minimal length of 50 bp and 3 mismatches) . We removed overlapped repeats manually and obtained information on tandem repeats using a tandem repeat finder (http://tandem.bu.edu/trf/trf.html; default parameters were used) .
We used 22 protein-coding genes (atp1, atp4, atp6, atp8, atp9, ccmC, ccmFn, cob, cox1, cox2, cox3, nad1, nad2, nad3, nad4, nad4L, nad5, nad6, nad9, rps3, rps4, and rps12) common to 15 plant mt genomes for our phylogenetic analysis. We aligned the sequences using Clustalw2 (default parameters were used) , removed ambiguously aligned regions based on Gblocks (version 0.91b; minimum number of sequences for a conserved position and flanking position is set to 10 and 15, respectively; no more than eight contiguous non-conserved positions and no gap are allowed) , and concatenated the sequences. The maximum-likelihood tree was constructed by using PHYML (version 3.0)  under HKY85+Γ4 model (C. vulgaris is used as outgroup). The bootstrap value was set to 100. All other parameters are set as default.
We used transcriptome data from bud, root, seed, fruit, male and female flowers, yellow and green leaves of cultivar Khalas. On average, ~700,000 SOLiD reads (50 bp with 3 mismatches or less) are used from the libraries. RPKM values are measured (reads per kilobase of exon model per million mapped reads)  and used to estimate gene expression, which are calculated according to:
GC content variations between new and old chloroplast-derived sequences. We defined chloroplast-derived sequences unique to P. dactylifera mitochondrial genome as “New" and those shared by other plant mt genomes as “Old".
Transcriptome analysis across eight tissues. FF, female flower (~422,000 reads); MF, male flower (~589,000 reads); F, fruit (~1,048,000 reads); S, seed (~179,000 reads); B, bud (~457,000 reads); GL: green leaf (~2,388,000 reads); YL, yellow leaf (~606,000 reads); R, root (~545,000 reads); P, genes on the positive strand; N, genes on the negative stand; and CP, chloroplast-derived regions. Their RPKM values (transformed using log10) range from 0 to 9 for genes on the positive strand and 0 to −9 for genes on the negative strand.
15 plant mt genomes used in this study.
Genes in 12 angiosperm mt genomes.
tRNA gene content of P. dactylifera mt genome.
Chloroplast-derived sequences in P. dactylifera mt genome.
RNA editing sites in three different plant mt genomes.
RNA editing validation of five genes in two tissues based on GS FLX reads.
Inter-varietal SNPs in coding regions among the three cultivars.
Inter-varietal SNPs in non-coding regions among the three cultivars.
Long repeats (repeat unit >50 bp) in P. dactylifera mt genome.
Tandem repeats in P. dactylifera mt genome.
We thank Jiucheng Liu, Dawei Huang, Guiming Liu, Quanzheng Yun, Sun Zhang, Guangyu Zhang, Duojun Zhao, Jixiang Wang, and Gaoyuan Sun for their excellent technical supports and contributions to this project. We also thank Professor Jun Tan for his helpful comments and discussions.
Competing Interests: The authors have declared that no competing interests exist.
Funding: The project is supported by a project grant from King Abdulaziz City for Science and Technology, Riyadh, Saudi Arabia [428-29]. The funder had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.