The complete chloroplast genome sequence of Helwingia himalaica (Helwingiaceae, Aquifoliales) and a chloroplast phylogenomic analysis of the Campanulidae

Complete chloroplast genome sequences have been very useful for understanding phylogenetic relationships in angiosperms at the family level and above, but there are currently large gaps in coverage. We report the chloroplast genome for Helwingia himalaica, the first in the distinctive family Helwingiaceae and only the second genus to be sequenced in the order Aquifoliales. We then combine this with 36 published sequences in the large (c. 35,000 species) subclass Campanulidae in order to investigate relationships at the order and family levels. The Helwingia genome consists of 158,362 bp containing a pair of inverted repeat (IR) regions of 25,996 bp separated by a large single-copy (LSC) region and a small single-copy (SSC) region which are 87,810 and 18,560 bp, respectively. There are 142 known genes, including 94 protein-coding genes, eight ribosomal RNA genes, and 40 tRNA genes. The topology of the phylogenetic relationships between Apiales, Asterales, and Dipsacales differed between analyses based on complete genome sequences and on 36 shared protein-coding genes, showing that further studies of campanulid phylogeny are needed.


INTRODUCTION
Complete chloroplast genome sequences (plastomes) have been very useful for understanding phylogenetic relationships in angiosperms at the family level and above, and have been used to resolve previously recalcitrant nodes (Barrett et al., 2016). However, there are currently large gaps in the coverage of orders and families. Within the 'very large, very old, and very widespread' subclass Campanulidae (Beaulieu, O'Meara & Donoghue, 2013; also known as 'Asterids II'), complete chloroplast genome sequences are currently Here, we first explore the structure of the chloroplast genome in the Helwingiaceae using H. himalaica. We then investigate the phylogenetic relationships in the Campanulidae by using the complete genome sequences and the protein-coding genes shared between H. himalaica and other published genomes.

MATERIALS AND METHODS
Helwingia himalaica is distributed from Nepal through northern India to southwestern China. Plants materials used in this study were intact, fresh, young leaves collected in Bingzhongluo county of Yunnan province (28.015306 N, 98.607944 E). The specimen has been deposited in the herbarium of the Xishuangbanna Tropical Botanical Garden, Chinese Academy of Sciences (HITBC). Total genomic DNA was extracted from fresh leaves using a modified CTAB method (Doyle, 1987;Yang, Li & Li, 2014). Each amplification was performed in 25 mL of a reaction mixture containing 1ÂPrimeSTAR GXL buffer (10 mM Tris-HCl (pH 8.2), 1 mM MgCl2, 20 mM NaCl, 0.02 mM EDTA, 0.02 mM DTT; 0.02% Tween 20, 0.02% Nonidet P-40, and 10% glycerol); 1.6 mM of dNTPs, 0.5 mM of each primer; 1.25 U of Prime-STAR GXL DNA polymerase (TAKARA BIO INC., Dalian, China), and 30-100 ng of DNA template. The amplification was conducted using 94 C for 1 min, 30 cycles of 98 C for 10 s and 68 C for 15 min, followed by a final extension step at 72 C for 10 min. The purified Polymerase chain reaction (PCR) product was fragmented and used for constructing the short-insert (500 bp) libraries according to the manufacturer's manual (Illumina). DNA of each sample was then indexed by tags and pooled together in one lane in an Illumina Hiseq 2000 to sequence (Yang, Li & Li, 2014).
Raw reads were filtered by quality control software NGSQCToolkit v2.3.3 (Patel & Jain, 2012) to obtain high quality Illumina data (cut-off value for percentage of read length = 80, cut-off value for PHRED quality score = 30) and vector-and adaptorfree reads. Filtered reads were assembled into contigs in CLC Genomics Workbench v.8 (http://www.clcbio.com) by the de novo method using a k-mer of 63 and a minimum contig length of 1 kb. Outputted contigs were aligned with the chloroplast genome of the asterid Camellia yunnanensis (GenBank accession number: KF156838), which was the most similar genome identified via BLAST (http://blast.ncbi.nlm.nih.gov/), and ordered according to the reference genome. Genes in the assembled chloroplast genome were predicted using Dual Organellar GenoMe Annotator (DOGMA) (Wyman, Jansen & Boore, 2004). The chloroplast genome was assembled using aligned contigs in Geneious v. 8.1.7 (http://www.geneious.com, Kearse et al., 2012). Junctions between large single-copy (LSC)/inverted repeats (IRs) and small single-copy (SSC)/inverted repeats (IRs) were validated by Sanger sequencing of PCR-based products (Table S1).
A matrix of chloroplast genome sequences, including H. himalaica, 36 other campanulid species, and Coffea arabica as an outgroup (EF044213 in GenBank) (Table 1), was aligned using MAFFT (Katoh & Standley, 2013) and manually edited where necessary. These 37 campanulid species represent all families and major clades within the Campanulidae that had complete chloroplast genome sequences in GenBank. Unambiguously aligned DNA sequences were used for phylogeny construction. Phylogenies were constructed by maximum likelihood (ML), Bayesian Inference analyses (BI), and maximum parsimony (MP).
ML analyses were conducted in RAxML version 8.2.8 (Stamatakis, 2014), using the GTACAT approximation. Convergence of the bootstrap was tested in RAxML using a posteriori bootstrapping analysis. BI analysis was conducted using MrBayes version 3.2.6 (Ronquist et al., 2012) and the best substitution model ('TVM+G') tested by AIC in jModelTest version 2.1.10 (Darriba et al., 2012). Four independent Markov Chain Monte Carlo algorithms were calculated for 10,000,000 generations and sampled every 1,000 generations. Potential Scale Reduction Factor (PSRF) values were used to determine convergence in BI using MrBayes version 3.2.6. All PSRF values were 1, indicating that these analyses converged. The first 25% of calculated trees was discarded as burn-in and a consensus tree constructed using the remaining trees. MP analysis was conducted PAUP version 4.0a150 (http://people.sc.fsu.edu/~dswofford/paup_test/), using the heuristic searches with tree bisection-reconnection (TBR) branch swapping and the 'Multrees' option in effect. Bootstrap analysis was conducted with 1,000 replicates with TBR branch swapping.
In addition, 36 protein-coding genes (Table 1) shared across all the 37 campanulid species were selected to build the phylogeny. ML analyses were conducted in RAxML version 8.2.8 (Stamatakis, 2014), using the GTACAT approximation. Convergence of the bootstrap was tested in RAxML using a posteriori bootstrapping analysis. BI analysis was conducted using MrBayes version 3.2.6 (Ronquist et al., 2012) and the best substitution model ('GTR+I+G') tested by AIC in jModelTest version 2.1.10 (Darriba et al., 2012). Methods for phylogeny construction using the 36 proteincoding genes follow the description above. PSRF values were used to determine convergence in BI using MrBayes version 3.2.6. All PSRF values were 1, indicating that these analyses converged. The first 25% of calculated trees was discarded as burn-in and a consensus tree constructed using the remaining trees. MP analysis was conducted in PAUP version 4.0a150 (http://people.sc.fsu.edu/~dswofford/paup_test/), using the heuristic searches with TBR branch swapping and the 'Multrees' option in effect. Bootstrap analysis was conducted with 1,000 replicates with TBR branch swapping.

Genome features
The total length of the chloroplast genome is 158,362 bp. Its quadripartite structure includes an LSC with 87,810 bp and SSC with 18,560 bp, separated by a pair of IR regions with lengths of 25,996 bp (Fig. 1). The GC content is 37.7% (see Table 1 in Yao et al., 2016). A total of 102 unique genes were detected in the chloroplast genome, of which 20 were duplicated in IR regions. Totally, 94 protein-coding genes (76 unique) encode proteins acting in processes related to photosynthesis, the genetic system, and some currently unknown functions (e.g. ycf). In addition, 40 genes (26 unique) encode for tRNAs and eight genes for rRNAs (Table 2). All eight rRNA genes are in IR regions. One ycf1 gene is a functional pseudogene as it is on the border between the SSC and IRa region. Gene rps19 is outside the IRb region at the LSC-IRb junction and rpl2 is fully included in the IRa region. Five genes (atpF, rpoC1, rpl2, ndhB and ndhA) have one intron and two genes have two introns (ycf3, clpP and rps12).

Repeated sequences and SSR
Thirty repeated sequences were detected, with lengths ranging from 18 to 43 bp and sequence identity more than 90% (Table 3). Among them, 19 repeated sequences were Table 2 List of genes in the chloroplast genome of Helwingia himalaica.

Genome rearrangement in the Campanulidae
Genome alignment among seven species from the seven campanulid families with known chloroplast genomes revealed massive gene rearrangement, especially in the LSC (Fig. 2). Moreover, all four Campanulaceae species had longer genomes and LSCs, and shorter SSCs, compared with other campanulid species (Table 1). The IR in the chloroplast genome of some Apiaceae (Angelica acutiloba, Foeniculum vulgare, Ostericum grosseserratum and Petroselinum crispum) was contractive (Table 1; Fig. 2). Even though the lowest number of coding genes in any campanulid species was 57 (Parthenium argentatum), only 36 coding genes were shared across all the campanulid families (  Figure 2 Gene arrangement map of chloroplast genome alignment of seven representative species from seven campanulid families and Coffea arabica (as a reference) determined by Mauve software (Darling et al., 2004). The polyline in the blocks indicates sequence similarity among these eight species. Line linking gene blocks among the eight species with same colour indicates ortholog. Gene blocks above are transcribed clockwise and those below are transcribed counterclockwise. The coding genes in the 14 main gene blocks are listed under the figure.

Phylogenetic analyses of the Campanulidae
The phylogeny produced from the analysis of 37 complete chloroplast genomes is well-supported, but while the results from ML and BI are congruent, the phylogeny from MP is not (Figs. 3A and 3B). With ML and BI, Aquifoliales are basal, Asterales are the next branch, and the Dipsacales are sister to the Apiales. The six families with multiple species are all well-supported (Fig. 3A). In the MP phylogeny, however, the Dipsacales are sister to the Asterales, and the Apiales are the next branch (Fig. 3B).
The phylogeny based on 36 shared protein-coding genes has a consistent family-level topology in analyses with BI, ML, and MP. The Aquifoliales are still basal, but the Dipsacales are the next branch, and the Asterales are sister to the Apiales (Fig. 4). However, within the Asteraceae, the topology from BI is different from those from ML and MP, and MP also did not resolve the relationships of Ilex wilsonii and Ilex szechwanensis (Fig. 4C).

DISCUSSION
The only other published chloroplast genomes in the order Aquifoliales are for seven species of Ilex in the family Aquifoliaceae (Yao et al., 2016). The length of the Helwingia genome (158,362 bp) is similar to that of Ilex (157,610-157,918 bp) (see Table 1 in Yao et al., 2016). Helwingia (94 genes) has two fewer protein coding genes than Ilex (96) while both have the same number of tRNA (40) and rRNA (eight) genes.
Minor differences among the published chloroplast genomes are common, including gene loss or gain events, but these do not separate H. himalaica from the others. Both H. himalaica and Helianthus annuus have ycf15 (Timme, 2009), but Guizotia abyssinica does not (Dempewolf et al., 2010). H. himalaica and L. sativa have trnE (Kanamoto et al., 2004) but H. annuus and G. abyssinica do not. Gene rps16 has one intron in G. abyssinica, H. annuus, and L. sativa, but none in H. himalaica, while gene ycf3 has two introns in H. himalaica, G. abyssinica, and H. annuus, but none in L. sativa. Gene ycf15 is between ycf2 and trnL-CAA in H. himalaica, but between rps7 and trnV-GAC in H. annuus. Gene ndhF is in the IRb-SSC junction in both L. sativa and H. himalaica, but in the SSC-IRa junction in G. abyssinica and H. annuus. The lengths of the published chloroplast genomes for the Campanulaceae range from 162,321 bp (T. caeruleum, Haberle et al., 2008) to 171,724 bp (Adenophora remotiflora, Kim et al., 2016) and are longer than those of other campanulid species. Kim et al. (2016) attribute this longer length to expansion occurring in the IR and LSC regions as well as the gene arrangements. The many mononucleotide SSRs identified in H. himalaica are potentially useful for studies of the evolutionary history of populations (Khadivi-Khub et al., 2014;Chae et al., 2014). The dominance of A/T in mononucleotide SSRs in Helwingia is similar to    other published studies (Huang et al., 2014;Kuang et al., 2011). It has been suggested that repeated sequences play an important role in genomic rearrangement and sequence variation in chloroplast genomes (Huang et al., 2014;Yang et al., 2013). Approximately 63% of repeats were found in intergenic regions which are often also divergent hotspot regions (e.g. Yao et al., 2015;Huang et al., 2014), showing the potential of these regions for the development of new phylogenetic markers for species identification in Helwingia and related genera in the Aquifoliales. Massive rearrangements in the chloroplast genome have been identified in the Campanulaceae in comparison with other campanulid families (Fig. 2). Except for gene block 1 and 2, most gene blocks in the LSC have been rearranged, including changes in gene order and transcribing direction (Fig. 2). The chloroplast gene rearrangement in Campanulaceae was first identified in T. caeruleum, and inferred as the effects of recombination of repeats or tRNA genes (Haberle et al., 2008). Comparing with other angiosperm chloroplast genomes, more repeats and tRNA genes occurred near rearrangement endpoints in this species. The positive connection between rearrangement and repeated sequences has also been found in other plants, like Arbutus unedo (Martínez-Alberola et al., 2013), Geraniaceae (Weng et al., 2013), Vaccinium macrocarpon (Fajardo et al., 2013) and cupressophytes (Wu & Chaw, 2014). However, the effects of these chloroplast gene rearrangements on plant physical functions still need more study.
The phylogenetic trees based on complete chloroplast genomes are incongruent with those from the protein-coding genes. Aquifoliales are basal in all phylogenetic analyses, but the phylogenetic relationships among the Asterales, Apiales and Dipsacales differ in different analyses (Figs. 3 and 4). The phylogeny based on complete chloroplast genomes using BI and ML methods found that the Apiales are sister to the Dipsacales (Fig. 3A), which agrees with recent phylogenies for this subclass based on other markers (Beaulieu, Tank & Donoghue, 2013;Wikström et al., 2015;Chen et al., 2016;The Angiosperm Phylogeny Group, 2016). However, using the MP method with the same data resulted in a phylogeny with the Asterales sister to the Dipsacales (Fig. 3B). The phylogenies based on protein-coding genes found that the Apiales are sister to Asterales with all three methods, although the topology within the Asteraceae differed between BI and the other two methods (Fig. 4). Three orders (of seven) and 22 families (of 29) in the subclass Campanulidae could not be included in our analyses because there are no published complete chloroplast genomes for these clades, which emphasizes the need for increased coverage of angiosperm orders and families in future studies of chloroplast genomes.

CONCLUSION
We report the chloroplast genome of H. himalaica as the first in the Helwingiaceae and the second genus in the Aquifoliales. It has the typical quadripartite circular structure, including an LSC with 87,810 bp and an SSC with 18,560 bp, separated by a pair of IR regions with 25,996 bp. In total, 142 genes were detected in this genome, consisting of 94 protein-coding genes, 40 tRNA, and eight rRNA. Repeated sequences are mainly distributed in intergenic regions. Comparisons among the available chloroplast genomes within the campanulids reveal massive chloroplast gene rearrangement in the Campanulaceae. The phylogenetic relationships among Apiales, Asterales and Dipsacales were incongruent between phylogenetic results produced from complete chloroplast genomes and the 36 shared protein-coding genes. The topology within Asteraceae also varied, which shows that further studies are still needed in these three orders. The results of this study will facilitate understanding of not only the family Helwingiaceae and its relationships with other taxa in the Aquifoliales, but also phylogenetic relationships within the angiosperms at higher levels.