Background

The chloroplast genome (cp genome) encodes a set of proteins for photosynthesis and other house-keeping functions that are essential to plant development [1]. Cp genomes are often used for research on plant evolution. Furthermore, cp genomes are predominantly uniparentally inherited [2], have highly conserved gene content and quadripartite organisation, and consist of a large single copy (LSC), a small single copy (SSC) and two inverted copies (IRs). Therefore, cp genome is widely used to trace species history [3]-[6]. In the past several years, there has been a dramatic increase in the numbers of complete chloroplast genomes from higher plants [7]-[12]. To date, there have been 437 complete chloroplast genomes of plants deposited in the NCBI database, along with the emergence of next-generation sequencers. These database resources provide information to better understand cp genome evolution in land plants. The `living fossil’ Nelumbo Adans is a small genus of angiosperms with long evolutionary history. They are perennial aquatic plants that flourished during the middle Albian [13],[14]. Now, there are only two surviving species, Nelumbo nucifera Gaertn. and Nelumbo lutea Willd, respectively. The former is mainly distributed in Asia and northern Australia, and the latter is mainly found in North and South America [15]. Nelumbo are economically important aquatic crops with ornamental, edible and medicinal properties. In 1795, Linnaeus placed the Nelumbo in Nymphaea Nelumbo Linn. In the intervening 200 years, Nelumbo has been considered a member of Nymphaeales (water lilies) and was then established as a single family belonging to the Nymphaeales [16].

During the past two decades, DNA sequences have been used to reevaluate the systematic position of Nelumbo. The traditional view has been challenged by non-molecular studies [17],[18] and rbcL sequence data [19]. To date, five different coding genes and several non-coding sequences have been used to reconstruct the relationships of Nelumbo [19]-[25]. Besides the nuclear genome [26], the complete cp genome of N. nucifera should be performed to elucidate the genomic evolution of N. nucifera. An accurate cp genome map is essential to study the phylogenesis, evolution and resource conservation of N. nucifera.

Obtaining an accurate cp genome is a prerequisite to understand its biological function and evolution for higher plants. At the beginning, most of the plant cp genomes were de novo assembled from the traditional Sanger sequencing [27]-[31]. This method is slow, expensive, laborious and low-throughput. More recently, next-generation sequencers such as Illumina, known for being high-throughput and cost-effective, have been used to assemble genomes based on a related reference genome. Because of its short read lengths, it cannot resolve a genome assembly with long repeats or low-GC regions, leading to gaps [32]. Single Molecule, Real-Time sequencing technology (SMRT) is the third-generation sequencing technology developed by Pacific Biosciences (PacBio). The process is as follows: first, DNA-imbedded DNA polymerases are attached to the bottom of 50 nm-wide wells, termed zero-mode waveguides (ZMWs), second, polymerases synthesise DNA using γ-phosphate fluorescently labelled nucleotides in the ZMWs, third, the width of the ZMWs cannot allow light to propagate through the waveguide, but energy can penetrate a short distance and excite the fluorophores incorporated into the growing DNA molecules in the vicinity of the polymerase at the bottom of the well. Compared to Sanger and Illumina platforms, PacBio can generate average read lengths of approximately 3,000 bp, with some reads reaching up to 30 kb with the current PacBio RS II platform. There have been some concerns about accuracy rates and insertion/deletion (indel) events caused by incorporation events or intervals-undetected events, but these can be improved by increased throughput in multiple SMRT cells [33],[34]. The optimisation of the assembly method [32],[35] and elevation of the accuracy rates make this platform have a great promise for genome sequence finishing. Currently, the PacBio platform is widely applied to de novo sequencing for various organisms, including human [36],[37], microorganisms [38],[39] and plants [40].

In this study, three goals were reached: first, N. nucifera (a representative of Nelumbo) was selected as the material to evaluate and compare cp genomes , including the accuracy rates and sequence sizes from three types of sequencing platforms, Sanger, Illumina MiSeq, and PacBio RS II. Second, we de novo assembled, annotated and analysed the cp genome of N. nucifera using PacBio RS II data. Third, to evaluate the systematic position and divergence of Nelumbo, as well as other basal eudicots, the cp genome of N. lutea was also sequenced by the Sanger platform with an average depth of coverage of 6×. We constructed a large phylogenetic tree that included 133 species from 56 orders (Additional file 1). Finally, we also estimated the divergence time of the basal eudicots, and compared the cp genomic structures to illustrate IR expansions/contractions among these early-diverged eudicots.

Results and discussion

De novoassembly from Illumina MiSeq and PacBio RS II platforms

Using the HGAP method [32], the PacBio RS II data was de novo assembled to one 163,600 bp contig with 105× depth of coverage (Figure 1) using Celera Assembler 7.0. The Illumina MiSeq data were de novo assembled to one 163,747 bp contig (Additional file 2: Figure S1) with 712× depth of coverage using Celera Assembler 7.0. The 163,307 bp contig (Additional file 3: Figure S2) for Sanger was assembled with Sequencher software. The sequence gaps marked with NNN from the Illumina MiSeq platform and Sanger technology were filled in using PCR. Data statistics and assembles from Illumina MiSeq and PacBio RS II were summarised in Table 1. Using the Sanger data, we found that the sequence identities among the three sequences were extremely similar, but the lengths of the three contigs from Sanger, Illumina MiSeq and PacBio RS II sequencing platforms were different. Using ClustalW alignment and PCR confirmation, we found a 282 bp deletion in the region of ndhA intron using the Sanger platform and a 152 bp insertion in the inverted repeats using the Illumina MiSeq platform (Additional file 2: Figure S1 and Additional file 3: Figure S2). These errors may be caused by low-throughput in the Sanger sequencing reads and short read lengths in the next-generation sequencing methods [41]. Erroneous insertions and deletions caused by sequencing technologies often lead to incorrect analyses of genome features. Additionally, low-throughput techniques and short read lengths are not ideal for reaching certain regions with highly repetitive sequences. However, in some small genomes, such as those of microorganisms, such repeated sequences appear to provide critical insights into the distinctions among bacterial strains [42].

Figure 1
figure 1

Gene map of N. nucifera chloroplast genome from PacBio RS II platform. The inverted repeats are indicated by thick lines. Asterisks indicate genes containing introns. Genes on the outside of the circle are transcribed in a clockwise direction and genes on the inside of the circle are transcribed in a counter-clockwise direction.

Full size image
Table 1 Statistics of the N. nucifera chloroplast genome sequencing data from Illumina MiSeq and PacBio RS II
Full size table

Currently, the advance in plastid sequencing is largely promoted by next-generation sequencing technologies. There have been related reports comparing next-generation sequencing platforms for plastid sequencing, such as the GS20 system (454 Life Sciences Corporation) [43] and the Illumina GA II platform [44], and the improved conventional Sanger method [45]. Additionally, the first comparison of next-generation technology (Illumina) with third-generation technology (PacBio) was performed in the last year [40]. However, the comprehensive comparison of the pros and cons of the three representatives of sequencing eras, Sanger, Illumina and PacBio, has not been determined. Especially for the accuracy-challenged PacBio platform, newly developed assembled methods and upgraded chemistry (from C1 to C2) will improve the accuracy rates and throughput [32]. We applied three independent sequencing platforms to evaluate the cp genome of N. nucifera. The results confirmed by PCR amplification showed that the de novo assembly genome from PacBio RS II platform was the most intact, reaching 100% coverage. Given the sufficient depth (105×), SMRT sequencing by PacBio RS II provides a highly accurate cp genome of N. nucifera, as it is highly unlikely that the same error will be randomly observed multiple times [34]. Deep sequencing coverage and additional Illumina library containing large fragments are essential to obtain the accurate structure of genome for Sanger and Illumina Miseq, respectively. Despite small differences among the three de novo assembled genomes, the accurate cp genomic structures may have more important roles than the cp genomic sequences in plant development. Meanwhile, the incomplete cp genomic information of N. nucifera caused by deletion and insertion from Sanger and Illumina, respectively, cannot reflect a genuine structure of cp genome in vivo. Furthermore, the run time of PacBio is very short at only 2 hours [34], which can save considerable time for researchers. Therefore, PacBio RS II platform, characteristics of the long reads and lack of bias in coverage of AT-rich regions, is promising for highly accurate `finished’ genomes.

General features and codon usage of N. nuciferacp genome

The final chloroplast circular map of N. nucifera from PacBio data corrected with Illumina Miseq data was 163,600 bp. In terms of structure and coding capacity, the cp genome of N. nucifera resembles those of eudicots, with minor length variations caused by lineage-specific insertions and deletions. This genome showed the typical quadripartite structure with a large single copy region (LSC, 91,846 bp) and a small single copy region (SSC, 19,626 bp) separated by two copies of an inverted repeat (IR, 26,064 bp) (Figure 1). The cp genome of N. nucifera contains the most complete 113 different genes, including four distinct rRNAs (16S, 23S, 4.5S and 5S), 30 distinct tRNAs and 79 distinct peptide-coding genes (including four ycfs). Four rRNAs, seven tRNAs and six peptide-coding genes (including rps12) are duplicated in the IR region, yielding a total of 130 genes (Table 2).

Table 2 List of genes present in the chloroplast genome of N. nucifera
Full size table

Start codon usage of N. nucifera was compared to those of eight other basal eudicots (Table 3). In these basal eudicots, ACG, GTG, or ATA appeared to be used as an alternative to ATG as the start codon. Among the changes of start codons, rpl2 and rps19 were found in all of the surveyed basal eudicots, but ndhB and ycf2 were only present in Ranunculus. Among the 79 distinct chloroplast protein-coding genes of N. nucifera, only three genes (psbL, rpl2 and rps19) used an alternative to ATG as the start codon: ACG for psbL and rpl2, and GTG for rps19. An ACG to AUG editing site in the ndhD, psbL and rpl2 transcripts is present in most angiosperm plastids [46],[47], but we only detected two RNA editing sites (psbL and rpl2) in the start codon region. Loss of such an editing site in ndhD transcripts may be caused by a very slow rate of evolution during the last 160 million years of Nelumbonaceae or back-mutation from C to T in the ndhD start codon. This loss of alternative start codons, ACG in ndhD may drastically impair the accumulation of the NDH complex in the leaves [48]. Furthermore, we examined codon usage patterns of the 79 distinct chloroplast protein-coding genes in N. nucifera. A total of 22,902 codons comprise the 79 different chloroplast protein-coding genes of N. nucifera. Overall codon usage in the N. nucifera is generally similar to that reported from other genomes, such as Panax [22] and Lotus [49]. Relative Synonymous Codon Usage (RSCU) analyses suggested that codons from the N. nucifera cp genome with the third position nucleotide of A or U were used more frequently than those ending with G or C (Table 4), as observed in most cp genomes of land plants [30]. For example, of the four codons coding for valine, the RSCUs of GUU and GUA were 1.43 and 1.5, but those of GUC and GUG were only 0.49 and 0.58, respectively.

Table 3 Alternative start codon usage in the sequenced basal eudicots
Full size table
Table 4 Relative synonymous codon usage for 79 distinct chloroplast protein-coding genes in N. nucifera
Full size table

During the evolution of angiosperms, the sizes of the most sequenced cp genomes range from approximately 120 kb to 160 kb in length. However, there are some exceptions for parasitic plants with unique lifestyles, of which the sizes of cp genomes were beyond the scope of 120 kb to 160 kb, such as Conopholis americana, with the smallest plastome of 45 kb of land plants [50]. Additionally, the numbers of genes in the cp genomes were present variously in different lineages, such as the losses of ndh genes. The events of ndh gene losses occurred in most non-photosynthetic plants, such as Cuscuta reflexa [51] and the parasitic plants, such as Epifagus virginiana [52], and in some non-parasitic, photosynthetic plants, such as Phalaenopsis aphrodite [53] and Geraniaceae [54]. The mechanism of the ndh gene losses may be explained for that either the genes are transferred to nuclear or they do not participate in the critical life development for the specific lineages [55]. In addition to the ndh genes, there were other independent gene losses in different lineages, including infA, rpl, rps, pet, psb and so on (Additional file 1). For example, the rpl21 gene loss of the cp genomes in the ancestral clades of gynosperms and angiosperms was compensated by the gene from the mitochondrial genome. The independent loss of infA in angiosperms (including almost all Rosaceae) was the result of transfer events from chloroplast to nuclear [56]. The cp genome of N. nucifera retained a complete set of genes data, suggesting these genes may be critical to its development.

Alternative start codons of cp genomes widely occurred in land plants, such as pteridophytes [30]. This editing pattern of the initiation codon seems to have occurred independently across the evolution of land plants, which does not correlate with the phylogenetic tree of the plant kingdom. Overall codon usage in the N. nucifera cp genome is similar to those of other reported cp genomes [30],[57] and mitochondrial genomes [58]. These codon usage patterns may be driven by the composition bias of the high proportion of A/T.

Phylogenetic and molecular dating analyses of the basal eudicots

Using three data matrices, maximum likelihood (ML) phylogenetic analyses were conducted using 79 protein-coding genes from 56 orders of seed plants. After searching the 56 models with Modeltest 3.7, the general time reversible (GTR) model with rate variations among sites and invariable sites (GTR + G + I) were selected as the best fit for the three data matrices. The phylogenetic trees inferred from the three data matrices showed the same topology. Additionally, the resulting topology, consistent with results from the Angiosperm Phylogeny Group (APG) [59], suggested that the phylogenetic tree was reliable. As sisters to Meliosma, Nelumbo and Platanus form a clade with 100% bootstrap values. This result confirmed that N. nucifera is a stem eudicot, supported by the morphological evidence of tricolpate pollen grains [21]. As a result of convergent evolution [60] in the same aquatic environment, a similar morphology has led to the misidentification of N. nucifera as a relative of Nymphaea alba. A phylogenetic analysis of 133 taxa from 56 orders indicated that Nelumbo was the sister clade to Platanus which is a genus of tall land trees (Figure 2).

Figure 2
figure 2

Phylogenetic tree of the 133 taxa based on 79 chloroplast protein-coding genes. The ML tree has a -lnL of −1601140.821388 with support values for ML provided at the nodes. Asterisks indicate ML BS =100%. Taxa in blue are the two new genomes sequenced in this study.

Full size image

The eudicots comprise the vast majority of the extant angiosperms, with an estimated 200,000 species. The clades can divided into basal eudicots and core eudicots [61],[62]. To date, plastid genomes have been completely sequenced for eight basal eudicots, including Buxus [63], Megaleranthis [64], Nandina [43], Platanus [43], Ranunculus [30], Trochodendron [65], Tetracentron [65] and Nelumbo (in this study). The addition of the un-sampled basal eudicot cp genome of N. nucifera will lead to a better understanding of the evolution of basal eudicots.

In the phylogenetic trees obtained in our study, the analysed basal eudicots, including Ranunculales (Nandina, Berberis, Megaleranthis, Ranunculus), Sabiaceae, Proteales (Platanus, Nelumbo), Trochodendrales (Trochodendron, Tetracentron) and Buxales (Buxus) formed separate clades (Figure 2). To estimate the divergence time in these clades, MCMCTree of PAML4.7 was used with the approximate likelihood calculation method [66]. This analysis dated Ranunculales, Sabiaceae, Proteales, Buxales and Trochodendrales to 197, 189, 189, 185 and 182 million years (Myr) ago, respectively. Nelumbo has an age of 177 Myr, and the splitting between the only two extant species, N. lutea and N. nucifera, is estimated to have occurred approximately 2 Myr ago (Figure 3).

Figure 3
figure 3

Posterior estimates of divergence time of 133 taxa on the phylogenetic tree. The values at the nodes represent mean ages in a 95% highest posterior density (HPD) analysis. Estimations were performed with MCMCTree using the IR (independent rate) model.

Full size image

In recent years, along with the released chloroplast data from NCBI, researchers used these cp genomes for plant evolution [67],[68]. In our study, we carefully selected different taxa from the NCBI database, of which the cp genomes were potentially published. Additionally, long-branch attraction will mislead to a wrong phylogenetic tree. To avoid long-branch attraction [69], the taxa uniformly distributed in species trees were selected. We controlled the numbers of taxa (no more than 8) in the same order. The saturation of substitution rates of codon sites, especially the third site, affects the topology of phylogenetic tree [28]. In our test, the phylogenetic topology from the matrix containing all three sites of each codon consistent with the results of the other two matrices (1st and 2nd sites, and 3rd site of each codon) verified there was no saturation of substitution rates in our analysed taxa.

Here, 133 taxa uniformly covering 56 orders were adopted to perform the phylogenetic analyses and estimation of divergence time for the basal eudicots. The phylogenetic analyses from three matrices (all three sites, 1st and 2nd sites , and 3rd site of each codon) of 79 chloroplast protein-coding genes supported the phylognesis of N. nucifera as a basal eudicot, sister to Platanus. Estimations of divergence time showed that Nelumbo and Platanus began to diverge approximately 177 Myr ago [66]. Additionally, the divergence of the basal eudicots (including Nelumbo) from Nympahea (the `early-diverging’ angiosperm), was approximately 255 Myr ago (Figure 3). The morphological similarity between Nelumbo and Nympahea caused by convergent evolution is typically contradictory to the similarity of molecular sequences among the three taxa Nelumbo, Nymphaea and Platanus. Therefore, the phenotypes of these species are determined by the combination of their molecular sequences and living environments.

The structural evolution within the basal eudicots

In angiosperms, frequent contractions and expansions at the junctions of SSC and LSC with IRs contributed to the size variations of cp genomes. Therefore, contractions and expansions of these junctions have been recognised as evolutionary markers for illustrating the relationships among taxa [70]. We were interested in the structural variations of N. nucifera and other basal eudicots. The structure of N. nucifera cp genome was compared to those of the seven basal eudicots (Trochodendron, Tetracentron, Platanus, Ranunculus, Buxus, Megaleranthis, and Nandia). Unlike the other six species, the largest expansions were found in the LSC/IRb boundary of Trochodendron and Tetracentron, up to 30 kb. The LSC/IRb conjunction of Trochodendron and Tetracentron expanded into the region between infA and rps8. However, the junctions of other 6 species appeared to be conserved with only minor expansions (Figure 4). The IRb of Platanus, Megaleranthis and Nandia expanded into the 3′ portion of rpl19 by 23 bp, 104 bp and 62 bp, respectively. The LSC/IRb boundaries of Ranunculus, Buxus and Nelumbo were located in the intergenic space regions downstream of rps19 (Figure 4). These data showed that various borders existed in these basal eudicots, even within the same order, such as Nelumbo (Proteales) and Platanus (Proteales). We speculated that the location of IR/LSC boundaries may not correlate to their positions of phylogenesis.

Figure 4
figure 4

Comparison of the boundaries of LSC, IR and SSC among eight chloroplast genomes of basal eudicots.

Full size image

N. nucifera is a member of land plants [71], which flourished during the Cretaceous. When Quaternary glaciations occurred, N. nucifera became trapped in water areas in response to environmental stress [72]. Previous reports noted that expansions of IR occurred more progressively in monocots than non-monocot angiosperms, and two hypotheses were proposed to explain IR expansions in the monocots [70]. The IR boundaries of 17 surveyed vascular plants vary among these cp genomes, even between closely related genera of the same family [22]. We wonder whether this clade of basal eudicots maintains the conservative IR boundaries. In this study, expansions and contractions of IR boundaries also varied in these basal eudicots, which was not related to the phylogeny of the lineages. For example, the IR boundaries of Nelumbo (Proteales) were more similar to that of Buxus (Buxales) than that of its closely related taxa, Platanus (Proteales). Despite that fact that each IR of Nelumbo was nearly 1 kb longer than that of Platanus (Figure 4), the former did not contain the portion of rps19 as did the latter. We found that variations of IRs were contributed by IR expansion to LSC or by an independent insertion of DNA fragments in IR regions. How the independent insertion occurred is still to be elucidated in future studies.

Conclusions

We first applied three sequencing platforms to evaluate the cp genome of N. nucifera. Using PacBio RS II data, Illumina MiSeq data and Sanger data, we de novo assembled, annotated and analysed the cp genome of N. nucifera. The precise cp genome of N. nucifera is a circular molecule of 163,600 bp with a typical quadripartite structure, containing a LSC region (91,846 bp) and a SSC region (19,626 bp) separated by IR regions (26,064 bp) with a total of 130 genes. The ML trees of 79 combined chloroplast protein-coding genes of 133 taxa confirmed that N. nucifera was a member of basal eudicots, sister to Platanus. Estimating the divergence time in MCMCTree with an approximate likelihood calculation showed that basal eudicots diverged at 197 Myr, and Nelumbo was 177 Myr. The splitting between N. lutea and N. nucifera was estimated to have occurred approximately 2 Myr. A structural comparison showed that the IR boundaries of basal eudicots occur in various border positions and an independent insertion of IR occurred in Nelumbo. This study showed that the PacBio platform will be useful for de novo assembly of genomes and the cp genome of N. nucifera provided new insight into the evolution of the basal eudicots. We believe that with the appearance of new PacBio sequencing platform, more accurate cp genomes will be obtained to understand the evolution of angiosperms at both the sequence and structural level.

Methods

Materials

The materials (Nelumbo nucifera Gaertn.) used in the experiment is maintained by Wuhan Vegetable Scientific Research Institute, Wuhan National Field Observation & Research Station for Aquatic Vegetables (30°12′N, 111°20′E).

Chloroplast genome DNA extraction

High quality DNA was obtained as follows: the unfolded tender leaves of N. nucifera were harvested and stored at 4°C in the dark to eliminate starch from the tissue. Chloroplast was isolated using the method of discontinuous sucrose gradient centrifugation, with DNase I digestion [73]. All steps must be conducted at 4°C unless otherwise specified. The chloroplast solutions were gently lysed by adding one-fifth volume of lysis buffer and one-twentieth Proteinase K to a final concentration of 200 μg/ml. The tubes were then gently inverted and mixed once every 15 min during 30-minute water baths at 37°C and then 50°C. After adding cold NH4Ac to a final concentration of 0.8 M, the nucleic acids were separately extracted with an equal volume of Tris-saturated phenol/chloroform/isoamyl alcohol (25:24:1) once and chloroform/isoamyl alcohol once (24:1).

Chloroplast genome DNA (cpDNA) was precipitated in two volumes of 100% ethanol overnight at −20°C and centrifuged at 18,000 × g for 20 min. The crude cpDNA was washed twice with 70% ethanol and re-suspended in TE buffer with RNase digestion for 30 min at 37°C. Finally, the values of OD260/280 and OD260/230 were 1.9 and 1.97, respectively.

Library construction, sequencing and de novoassembly

  1. (1)

    Sanger library construction

The entire chloroplast sequence was amplified using a long-range PCR technique with cpDNA. PCR primers were developed from the alignment of known eudicotyledon chloroplast genomic sequences (Additional file 4). We covered the two chloroplast genomes (N. lutea and N. nucifera) with PCR products, which showed lengths of between 8 and 15 kb. PCR products were eluted using electrophoresis on low-melting-point agarose gels, followed by column purification (Axgene). The purified products were sheared into random fragments of approximately 1.5 kb and cloned into the pMD18-T vector (TAKARA). The recombinant plasmids were transformed into TOP10 cells and sequenced using an ABI 3730 DNA sequencer.

Sequence masking and assembly were performed using Sequencher software (Gene Codes Corp., Ann Arbor, MI). Each nucleotide has approximately 7 × coverage in N. nucifera and 6 × in N. lutea. Gaps and low coverage regions were filled using PCR to make sure each nucleotide was covered by at least three reads.

  1. (2)

    Illumina MiSeq library construction

Approximately 5 μg cpDNA was used to construct a library containing a 400 bp insert size. Genome DNA sequencing libraries were constructed using TruSeq DNA LT Sample Preparation Kit V2 (Illumina), following the manufacturer’s protocol. The genome DNA was sheared to short fragments using Covaris S220. DNA fragments were adenylated at 3′ ends after end repair. Specific adapters were ligated to both ends of the DNA fragments, and the barcode sequence was included in one of the adaptors. Targeted size DNA fragments were selected by gel-cutting and amplified by 10-cycle PCR using universal primers (Illumina). After purification, quantification, and validation, the validated DNA libraries were sequenced on an Illumina MiSeq Sequencing System following the manufacturer’s standard workflow.

The raw data were filtered and the obtained clean data were joined by FLASH. The joined reads were aligned to cpBase with bwa-0.7.3a. The aligned reads were then selected for de novo assembly with a Celera Assembler 7.0. The final contigs were assembled into one scaffold with SSPACE and lastZ. The two gaps were filled using PCR amplification.

  1. (3)

    PacBio library construction

A sample of 20 μg pure and high-molecular-weight DNA is required to prepare size- selected approximately 20 kb SMRTbell templates. The cpDNA concentration was measured using both a NanoDrop spectrophotometer and a Qubit fluorometer, and approximately 200 ng of cpDNA was run on a field-inversion gel. The subsequent steps are based on the PacBio Sample Net-Shared Protocol, which is available at http://pacificbiosciences.com/

The primitive reads from Pacbio RS II were corrected with SMRT Analysis 2.1, yielding a total of 9,165 high quality reads, up to 42,623,117 bp. The average length of the reads was 4,651 bp. These reads were assembled using the Celera Assembler 7.0. The reads were aligned to the assembled contigs with bwa-0.7.3a. Eventually, 4 contigs with more than 10x coverage depth were assembled to one contig using the lastZ tool. Finally, to generate the best consensus sequence of the cp genome sequence, Quiver (a new multiread consensus algorithm) was used to correct the error regions [32].

Genome annotation

The entire chloroplast genome sequences of N. nucifera from three platforms were annotated using DOGMA (Dual Organellar GenoMe Annotator) [74]. The genes of N. lutea from the Sanger platform were also annotated. For some genes with very short exons, such as petB and petD, manual annotation was performed. We used tRNAscan-SE to corroborate tRNA boundaries [75]. Physical maps were drawn using GenomeVx [76].

Codon usage

RSCU (relative synonymous codon usage) [77] of the cp genome of N. nucifera was calculated from coding sequences of 79 protein-coding genes. Genes in IR regions were counted only once.

Phylogenetic analysis

Seventy-nine distinct peptide-coding genes were used for phylogenetic analyses. In addition to the two chloroplast genomes sequenced in the present research, 131 taxa from 56 orders were selected to construct the ML tree. All 79 protein-coding gene sequences were translated into amino acid sequences, which were aligned in ClustalW [77] and manually adjusted. Nucleotide sequences of these genes were aligned by constraining them to the amino acid sequence alignments. The more quickly evolving regions, which are difficult to align, were excluded from the analyses. A FASTA file concatenating 67,862 nucleotides (Additional file 5) was generated, the translated protein-coding sequences of which were seen in Additional file 6. We extracted the first and second sites of codons, and the third site of codons to form two other new matrices. Finally, three matrices were used for phylogenetic analyses. These were established in a DOS platform using our Perl scripts.

The nucleotide file containing 133 concatenated sequences was tested prior to inferring the phylogenetic trees. First, the saturation for substitutions was tested in DAMBE [78] across the three matrices. The results, Iss < Iss.c at p = 0.0000, suggested that there were no saturated sites in these sequences. Second, we used Modeltest 3.7 [79] to determine the most appropriate model of DNA sequence evolution. After searching the 56 models, the general time reversible (GTR) model, with rate variation among sites and invariable sites (GTR + G + I), was selected. To infer phylogenetic trees from three nucleotide matrices, we applied the maximum likelihood method with rapid bootstrapping of 1000 replicates in RAxML-HPC v7.2.8 available on http://www.phylo.org/ [80].

Estimation of divergence times

The concatenated 79 gene sequences from 133 species and the phylogenetic tree were used for molecular dating analyses. The divergence time was estimated using the Bayesian method implemented in MCMCTree of PAML4.7 [81]. Eight fossil calibrations were incorporated through the previous time (Additional file 7). Using the approximate likelihood calculation method, the gradient g and Hessian H with BASEML using the GTR + Γ5 substitution model were calculated [82]. The independent rate (IR) model [83] for the molecular clock and the GTR + Γ5 model for nucleotide substitutions were set in the mcmctree.ctl control file. To ascertain whether convergence was achieved, two independent MCMC analyses with 106 steps, following a discarded burn-in of 105 steps, were simulated [84],[85].

Availability of supporting data

The precise chloroplast genome of N. nucifera sequenced by PacBio RS II and corrected by Illumina MiSeq has been submitted to GenBank (accession KM655836). The chloroplast genomes of N. nucifera and N. lutea sequenced by Sanger are available on National Center for Biotechnology Information (accession NC_015610 and NC_015605). Other data sets supporting the results of this article are included within the article and its additional files.

Authors’ contributions

YD and ZHW conceived and designed the experiments. ZHW, STG, ZWQ, LP, SZW and DQL performed the experiments. WDK provided the plant material. ZHW and YD wrote the paper. All authors read and approved the final manuscript.

Additional files