Leveraging plastomes for comparative analysis and phylogenomic inference within Scutellarioideae (Lamiaceae)

Scutellaria, or skullcaps, are medicinally important herbs in China, India, Japan, and elsewhere. Though Scutellaria is the second largest and one of the more taxonomically challenging genera within Lamiaceae, few molecular systematic studies have been undertaken within the genus; in part due to a paucity of available informative markers. The lack of informative molecular markers for Scutellaria hinders our ability to accurately and robustly reconstruct phylogenetic relationships, which hampers our understanding of the diversity, phylogeny, and evolutionary history of this cosmopolitan genus. Comparative analyses of 15 plastomes, representing 14 species of subfamily Scutellarioideae, indicate that plastomes within Scutellarioideae contain about 151,000 nucleotides, and possess a typical quadripartite structure. In total, 590 simple sequence repeats, 489 longer repeats, and 16 hyper-variable regions were identified from the 15 plastomes. Phylogenetic relationships among the 14 species representing four of the five genera of Scutellarioideae were resolved with high support values, but the current infrageneric classification of Scutellaria was not supported in all analyses. Complete plastome sequences provide better resolution at an interspecific level than using few to several plastid markers in phylogenetic reconstruction. The data presented here will serve as a foundation to facilitate DNA barcoding, species identification, and systematic research within Scutellaria, which is an important medicinal plant resource worldwide.

As currently defined, Scutellarioideae includes approximately 380 species in five genera [1]: Holmskioldia, Renschia, Wenchengia, Scutellaria, and Tinnea. The former three are monotypic genera. The genus Holmskioldia, comprising the single species H. sanguinea Retz., is native to the subtropical Himalayan region but is widely grown as an ornamental in warm climates and has become naturalized throughout the Old and New Worlds [12]. The monotypic Renschia, represented by R. heterotypica (S. Moore) Vatke, is narrowly endemic to the Ahl Mountains in northern Somalia [13], and its systematic position within Scutellarioideae remains unclear. The placement of Wenchengia in Scutellarioideae was resolved by Li et al. [9] based on the rediscovery of the extremely rare species, W. alternifolia C.Y. Wu & S. Chow. This genus was long thought to be endemic to Hainan Island in southern China [14,15], but recently it was also reported from Vietnam [16]. With 19 species recognized to date, Tinnea is the second largest genus in Scutellarioideae, occurs mainly in fire-prone grassland, woodland, and scrub vegetation, and is endemic to Africa [17].
Scutellaria, containing approximately 360 species and commonly known as skullcaps, is the largest genus in Scutellarioideae [18]. The genus is distributed nearly worldwide and occurs in various habitats, but is mostly found in tropical montane and temperate regions [5,19]. Most species are herbaceous perennials or small shrubs. The calyx of Scutellaria consists of two undivided lips and bears an appendage on the upper lip, which is described as a scutellum and is the most distinctive character of the genus; this feature is the basis for the common name skullcap. Many Scutellaria species possess medicinal uses, and some species are of economic importance. For example, S. baicalensis Georgi (baical skullcap or Chinese skullcap; 'Huangqin' in Chinese) is a traditional Chinese medicinal herb that was first recorded in Shen Nong Ben Cao Jing in ca. 100 BC [20], and is widely used to treat hepatitis, jaundice, tumor, leukemia, hyperlipaemia, arteriosclerosis, diarrhea, and inflammatory diseases [21].
The chloroplast is an essential organelle in angiosperms because it provides energy for plant cells [44]. This uniparentally inherited plastid is characterized by a circular double-stranded DNA molecule between 120,000-160,000 base pairs in length, multiple copies per cell, and a quadripartite structure that includes two identical regions in opposite orientations called the inverted repeat (IR), flanked by large single copy (LSC) and small single copy (SSC) regions [45]. With increasingly rapid and less expensive next generation sequencing (NGS) technologies continually developing, ever-increasing numbers of non-model species plastid genome are being sequenced and successfully used for resolving phylogenetic and taxonomic problems in flowering plants at various ranks [46][47][48]. However, using cp genomes to resolve phylogenetic questions within the mint family has been rare [49], and plastomes of only two species, Scutellaria baicalensis and S. indica L. var. coccinea S. Kim & S. Lee, have been published from Scutellarioideae [50,51]. Sequences of S. insignis Nakai and S. lateriflora L. were uploaded to GenBank without any related publication or analyses. Consequently, little is known regarding plastome structure variation within Scutellaria.
In this study, we sequenced 12 plastomes from 11 species representing four of the five genera of Scutellarioideae. In addition, three previously released plastomes of Scutellaria (S. baicalensis, S. insignis and S. lateriflora) were downloaded from GenBank and included for comparative analyses. The species S. indica var. coccinea was exclude in this study because the sequence was unavailable. With these data, we aim to: 1) characterize and compare the structure and gene organization of plastid genomes within Scutellarioideae; 2) identify candidate molecular markers for future phylogenetic and/or population genetic studies within Scutellaria; and 3) reconstruct the phylogeny of Scutellarioideae using complete chloroplast genome sequences. The data presented in this study will provide abundant information for further studies about phylogeny, taxonomy, species identification, and population genetics of Scutellaria, and will also be helpful for exploration, utilization, and conservation of plant genetic resources of this important medicinal plant resources.

Taxon sampling, DNA extraction, and sequencing
Plastomes of 12 samples, including eight species of Scutellaria, one species each of Holmskioldia and Tinnea, and two individuals of Wenchengia alternifolia, were newly generated for this study. Voucher information is listed in Table 1 and all voucher specimens were deposited at the Herbarium of Kunming Institute of Botany (KUN), Chinese Academy of Sciences. In addition, three complete plastomes of Scutellaria from GenBank, S. baicalensis (MF521633), S. insignis (KT750009), and S. lateriflora (KY085900), were included for comparative analyses (Table 1).
Total genomic DNA was extracted from 150 mg fresh or silica-gel dried leaves using the CTAB method [52]. The DNA samples were sheared into fragments of about 300 bp to construct libraries according to manufacturer's instructions (Illumina, San Diego, CA, USA). Paired-end (PE) sequencing of 150 bp was conducted on an Illumina Hiseq-2500 platform (Illumina Inc.) at BGI-Wuhan.
Quality control of raw sequence reads was carried out using FastQC toolkit (http://www. bioinformatics.babraham.ac.uk/projects/fastqc; [53]) with the parameter set as Q � 25 to acquire high-quality clean reads for downstream analyses. De novo assembling of the plastomes was implemented in the GetOrganelle pipeline [54]. The filtered de Bruijn graphs file "gfa" was visualized in Bandage v. 0.8.1 [55] and the complete chloroplast sequence paths were manually selected, with the minimum depth of contigs above 100 × and the minimum length > 300 bp. Then all PE reads were mapped to the assembled plastomes using the Bowite2 [56] plugin in Geneious v.11.0.4 [57] to verify quality and correct assembly errors.
Plastome annotation was first performed using the online programs Dual Organellar Genome Annotator (DOGMA) [58] and Ge-seq [59]. We then inspected and curated all annotation manually with comparisons to the published plastome of S. baicalensis (MF521633) in Geneious v.11.0.4 [57]. The tRNAs were verified using the online tRNAscan-SE service with default parameters [60]. The resulting circular plastome maps were drawn using the Organel-larGenomeDRAW tool [61].

Characterization of simple sequence repeats and repeat structure
The simple sequence repeats (SSRs) in plastomes were identified using MISA perl script (http://pgrc.ipk-gatersleben.de/misa). Thresholds for the minimum repeated size were set as follows: � 10 for mono-nucleotide, � 5 for di-nucleotide, � 4 for tri-nucleotide, and � 3 for tetra-nucleotide, penta-nucleotide, and hexa-nucleotide repeats. The location and size of the repeating sequences (forward, reverse, palindromic and complement) were visualized in REPuter [62] with the parameter set as with a hamming distance of 3 and a minimum repeat size of 30 bp following the procedure outlined in Jiang et al. [50].

Comparative plastome and sequence divergence analysis
Comparative analyses of 15 plastomes of Scutellarioideae were carried out using the Mauve v.2.3.1 [63] plugin in Geneious v.11.0.4 [57]. We applied mVISTA [64] to visualize the results and evaluate the similarity among different plastomes, using default parameters to align plastomes under the LAGAN model and the annotations of S. baicalensis (MF521633) as a reference. In order to investigate the IR contraction or expansion, we also compared the boundaries between IR and SC regions in Geneious v.11.0.4 [57]. Two data sets (alignments of all 15 samples from Scutellarioideae and 11 species of Scutellaria) were used for the sliding window analysis to evaluate the intergeneric and intrageneric nucleotide sequence variabilities (Pi). Sequences were aligned using MAFFT v.7.221 [65] and misaligned regions were manually adjusted in Geneious v.11.0.4. [57]. DnaSP v.6 [66] was then used to calculate the Pi. The step size was set to 200 bp, with a 600 bp window length.

Phylogenetic analysis based on complete plastome sequences
In addition to the previously published plastomes of Scutellaria, plastomes of 31 species from within other subfamilies of Lamiaceae (12 Nepetoideae, 15 Lamioideae, two Ajugoideae, and one each from Premnoideae and Tectonoideae) were also included in the analyses to evaluate the utility of complete plastome sequences for resolving broad relationships within Scutellarioideae. Based on previous studies [1], Callicarpa americana (assembly from the WGS data under the SRR6940059) from Callicarpoideae was selected as the outgroup. GenBank accession numbers are provided in S1 Table. Alignments were initially performed using MAFFT v.7.221 [65] with default settings, and subsequently manually adjusted in Geneious v.11.0.4 [57]. Ambiguously aligned regions (e.g. characters of uncertain homology among taxa and single-taxon insertions) were excluded before phylogenetic analyses. Since the plastid genome is uniparentally inherited and does not undergo recombination [67], we combined all sequences and constructed three matrices: (i) combined coding regions (dataset CR); (ii) combined non-coding regions (dataset NCR); (iii) combined whole plastome sequences (dataset CPG). In order to reduce the overrepresentation of duplicated sequences, only the IRa region was included in all data sets. In addition, in order to evaluate the efficacy of the complete plastome sequences for phylogeny reconstruction within Scutellarioideae, we also created two additional datasets for phylogenetic analyses and comparison. One was a combined dataset of hyper-variable regions (16VAR) detected in this study, the other dataset consisted of six commonly used DNA regions (6CP) from previous studies [9,41,68].
Maximum likelihood (ML) and Bayesian inference (BI) analyses were performed on the Cyberinfrastructure for Phylogenetic Research Science (CIPRES) Gateway (http://www.phylo. org/; [69]. ML analyses were conducted using RAxML HPC2 v.8.2.9.0 [70] with the general time reversible (GTR) + G model and 1000 bootstrap replicates. BI analyses were carried out using MrBayes v.3.2.6 [71]. The best substitution model for each data set was determined using jModelTest2 [72] on the CIPRES Gateway, under the Bayesian information criterion (BIC) [73]. Four Markov Chain Monte Carlo (MCMC) chains (one cold and three heated) were run for 20 million generations. Convergence of the MCMC runs and estimated sample size (ESS) were analyzed by Tracer v.1.7.0 [74]. The first 25% of trees discarded as burn-in, and the remaining trees were summarized to construct the 50% majority-rule consensus tree.

Genome assembly, features, and gene content across scutellarioideae
Illumina paired-end sequencing generated 16 Table 2). The GC content was similar among different species of Scutellarioideae and the average GC content was 38.3% (Table 2). In general, the GC content in the IR regions (43.4-43.6%) was higher than in the LSC (36.3-36.5%) and SSC (32.4-32.8%) regions, and the GC content within non-coding regions (35.0%) was lower than within coding regions (40.5%).
Intraspecific plastome polymorphisms can be evaluated among multiple individuals from the same species. The sequence identity between the two samples of Wenchengia alternifolia was 98.6%, with only two large indels (> 100 bp), within the intergenic psbE-petL (344 bp) and psbM-trnD (GUC) (226 bp) regions, detected. The plastome maps of Holmskioldia sanguinea, W. alternifolia HN, Tinnea aethiopica, and Scutellaria przewalskii are presented as representatives of Scutellarioideae (Fig 1), while maps of the remaining species are provided in supplementary materials (S1 Fig). All newly sequenced and annotated plastomes were submitted to the National Center for Biotechnology Information (NCBI) database under accession numbers MN128378-MN128389 (Table 2).
When duplicated genes in IR regions were counted only once, each of the plastomes included 114 unique genes (80 protein-coding genes, 30 tRNAs and four rRNAs; Table 2) that

PLOS ONE
were arranged in the same order. A total of 18 genes exist in duplication within the IR region, including seven protein-coding genes, seven tRNAs and four rRNAs ( Table 3). Ten of the protein-coding genes and six of the tRNA genes contained one intron, and two genes (ycf3 and clpP) contained two introns. Among those newly sequenced samples, protein-coding regions accounted for 52.1-53.5% of the length of the whole genome, while tRNA and rRNA regions accounted for 1.78-1.92% and 5.9-5.96%, respectively (S2 Table). The remaining regions were non-coding sequences, including intergenic spacers, introns, and pseudogenes. All of the gene functions and groups were shown in Table 3. Large chain of rubisco rbcL Transfer RNA genes 30 tRNA genes (6 contain one intron, 7 are duplicated in the IR region)

SSRs and repeat structure
In total, 590 SSRs were identified in the 15 plastomes of Scutellarioideae, of which 483 SSRs (81.86%) were in the LSC region, 65 SSRs (11.02%) were in the SSC region, and 42 SSRs (7.12%) were in the IR region (Fig 2, S3 Table). The number of SSRs (or microsatellite loci) ranged from 31 (Scutellaria altaica) to 48 (Wenchengia alternifolia HN) among species of Scutellarioideae (Fig 2). The mononucleotide represents the highest variability with the repeat number ranging from 15 (S. altaica) to 35 (W. alternifolia HN), while the number of dinucleotide, trinucleotide, and tetranucleotide repeats showed no significant difference among the 15 samples. The number and frequency of each repeat type within the 15 plastomes of Scutellarioideae is shown in Fig 2 and S3 Table. When the cyclic queues and reverse complements were regarded as the same SSRs, the 590 SSRs can be classified into 17 different repeat types. The mononucleotide repeat unit (A/T); dinucleotide repeat unit (AT/AT), trinucleotide repeats unit (AAG/CTT) and tetranucleotide repeat unit (AAAG/CTTT, AAAT/ATTT) were shared in all the 15 samples (Fig 3). The mononucleotide repeat unit (G/C) was absent in Scutellaria calcarata. Within the trinucleotide repeat, the repeat unit (AAC/GTT) was unique to Wenchengia, and the repeat unit (AAT/ATT)

PLOS ONE
were detected in both individuals of W. alternifolia and in Tinnea aethiopica, while the hexanucleotide repeats were only found in S. baicalensis. The distribution of the 17 repeat types among the 15 plastomes and their relationships is shown in Fig 3. In total, 489 long repeats including forward, reverse, and palindromic were detected in the 15 plastomes (Fig 4). The most abundant type were the palindromic repeats, which accounted for 54.26% of the total repeats, followed by forward repeats (44.91%). The reverse repeats were rare and accounted for only 0.83% of the total repeats (Fig 4). Most repeats were located in the non-coding regions (77.96%; Fig 4). The length of the repeats ranged from 30 bp to 136 bp, and most of the repeat sequences were 30 bp, 32 bp, 39 bp, 41 bp, and 60 bp long (Fig 4, S4 Table).

Comparative analysis of plastomes of Scutellarioideae
The Mauve results showed that the organization of the plastomes in Scutellarioideae is highly conserved; neither translocations nor inversions were detected. However, differences in the size of the plastomes were detected. For example, the plastome of Scutellaria przewalskii was the shortest (151,675 bp), while that of Holmskioldia sanguinea (153,272 bp) was longer than the other species (S2 Fig). Results from the analyses by mVISTA showed that the two IR regions were less divergent than the LSC and SSC regions. Moreover, the non-coding regions and the intergenic spacers exhibited a higher divergence than the coding regions ( Fig 5). In all species, the IRa/LSC junctions were located within the rps19 gene, with a 41-74 bp protrusion of the rps19 gene into the IRa region that resulted in a part of the rps19 gene (ψrps19) present in the IRb region. In Wenchengia alternifolia and Tinnea aethiopica, the ndhF gene was completely located in the SSC region while in H. sanguinea and all species of Scutellaria a small fragment of the ndhF gene extended into the IRa region with (29 bp in H. sanguinea and 25-45 bp among species of Scutellaria). The IRb/SSC boundary was within the ycf1 gene, with between 771 and 1,184 bp in the IRb region. An equal length ycf1 pseudogene (ψycf1) was detected in the IRa region. The IRb/LSC boundary was located between the pseudogene rps19 (ψrps19) and trnH-GUG across the 15 plastomes. The distance between trnH-GUG and the IRb/LSC boundary for all species varied from 0 to 3 bp (Fig 6).

Characteristics of the datasets and phylogenetic relationships within Scutellarioideae
After the exclusion of ambiguously aligned sites, the total length of the complete aligned dataset (CPG) was 144,120 bp, of which 36,934 bp were variable (25.63%). The length of the CR  Table 4. Topologies obtained from both ML and BI analyses for all three datasets were identical, thus the ML topology resulting from the analysis of the CPG dataset (Fig 8) is presented here for subsequent discussion of phylogenetic relationships.

General characteristics of the plastomes of Scutellarioideae
Prior to this study, three plastomes of Scutellaria were available on GenBank, but two of them were without any related publication or analysis; only S. baicalensis was formally published [50]. The species S. indica var. coccinea has since been published, but the sequences were not PLOS ONE yet available [51]. Here, we report on 12 complete plastomes representing 11 species from four genera of Scutellarioideae for the first time. In total, 15 plastomes were included for comparative analysis.
The length of plastomes of the 15 taxa from Scutellarioideae ranged from 151,675 bp to 153,272 bp, with the variation mainly caused by large indels (insertions/deletions) in the noncoding regions. The plastomes of Scutellarioideae are highly conserved in structure, gene  order, and content. All the 15 plastomes encode 114 unique genes in the same gene order and display the typical quadripartite structure, including a pair of IR regions separated by the LSC and SSC regions (Fig 1 and S1 Fig). Lee and Kim [51] have recently identified 115 genes from the plastome of S. indica var. coccinea. In comparison with the present study, one extra tRNA

PLOS ONE
gene was identified. Because sequences and annotation information of this plastome have not been released, we could not include it for comparative analysis. The average GC content of Scutellarioideae plastomes in our study was38.3%, very similar to other species in Lamiaceae [50,51,[75][76][77]. The complete aligned sequences indicate that the 15 plastomes of Scutellarioideae are conserved, with the sequence identity among genera higher than 95% and no major structural rearrangements or gene losses discovered. The location of the IR boundaries, especially as this pertains to IR contraction and expansion, can be exploited for phylogenetic purposes as small expansions or contractions tend to have similar endpoints in closely related species [78]. We find that the variation in the IR boundaries in Scutellarioideae, however, is not as extensive as reported in previous studies [79].
Chen et al. [79] reported that the LSC/IR regions within Lamiales can be divided into four different types: type I, with the LSC/IR regions being located in the intergenic rpl2-rps19; type II, with the rps19 pseudogene at the LSC/IR border; type III, with the ycf2 pseudogene at the IR/LSC border; and type IV, with the IR extending to include the trnH gene and a truncated psbA pseudogene at the IR/LSC border. Subsequently, Gao et al. [48] detected a new type where the IR/LSC border was found in the intergenic rpl2-rps19. In our study, the LSC/IR junction of all 15 species of Scutellarioideae belongs to type II, and the boundary of the SSC and IRa regions in Wenchengia alternifolia and Tinnea aethiopica is aberrant, with an expansion that involved the complete ndhF gene being included in the SSC region (Fig 6).
SSRs are widely used in molecular identification, genetic diversity, and population genetics studies [80]. Studies have shown that A/T mononucleotides are often very rich in SSRs [50,76,77]. Our analyses also show that SSRs in Scutellarioideae are generally composed of short polyadenine (poly A) or polythymine (poly T) repeats and rarely contain tandem guanine (G) and/ or cytosine (C). In this study, a total of 455 SSRs are made up of A or T bases, accounting for approximately 77% of the total SSRs. In addition, most mononucleotide repeats were detected in the non-coding regions (S3 Table). A potential reason for the higher frequencies of the AT repeats is the strand separation for ATs is relatively easier than GCs during plastome replication, which increases slipped-strand mispairing. There is a tendency for SSRs to occur in the non-coding region of the chloroplast genome of higher plants [81]. The molecular processes that give rise to repeats are more likely to be preserved in non-coding regions because there is strong selection against them in coding regions. In addition, because the non-coding regions are so AT rich, there is an expectation that repeats will be biased towards AT content, especially in the single copy regions. In general, the structure and organization of plastomes is conserved and SSRs primers are transferable across species or genera. Thus, the new SSRs detected in this study are potential resources for estimating the genetic diversity of some important medicinal species of Scutellaria, and for phylogenetic study among species and genera.
It has been demonstrated that short dispersed repeats are a major factor promoting plastome rearrangements in land plants [82], but within the unrearranged plastid sequence the function of these repeats remains unknown [76]. Our study reveals three types of repeats (forward, reverse, and palindromic) in the 15 plastomes of Scutellarioideae. As has been reported in other species of Lamiales [79,83], most of these repeats are located in the intergenic spacers and introns, but several also occur in the coding regions. In total, 22.04% of the repeats occur in four protein coding regions (psaB, psaA, ycf1, and ycf2; S4 Table). The genes ycf1 and ycf2 have been demonstrated to be associated with repeat events [84]. In our study, the richest repeats are found in the ycf2 gene, similar to other studies [48,79,83]. However, only one palindromic repeat, in the ycf1 gene of Wenchengia alternifolia VN was detected. The absence of the dispersed repeats from the ycf1 gene in this study is partially because the plastomes from closely related species are highly similar and lack of variation.

Potential DNA barcodes for Scutellaria
Genomic comparative analyses of complete plastome sequences have become necessary for developing variable DNA barcodes, especially for finding mutation "hotspot" regions for novel DNA barcodes in addition to the set of widely used DNA markers (matK, rbcL, psbA-trnH, and nrITS [85][86][87]).
Though Scutellaria is the second largest genus within Lamiaceae and has medicinally important [88], DNA barcoding research within the genus is wanting. Guo et al. [68] attempted to distinguish the most widely used medicinal species, S. baicalensis, from its congeners, S. amoena, S. rehderiana Diels, and S. viscidula Bunge. However, this study had sparse sampling and only three DNA regions were used (matK, rbcL, and psbA-trnH). In previous studies, the cpDNA markers rps16 (as part of the trnK-rps16 intron), ndhF, rps15-ycf1, and ycf1 were used to resolve the systematic position of some genera within Lamiaceae [89,90], and fragments of psbA-trnH, rpl32-trnL, rps15-ycf1, and ycf1 were applied to infer the intrageneric relationships [91,92]. Some fragments, such as petN-psbM and petA-psbJ have been commonly used in seed plant phylogenetic studies [93,94], but never have been used to resolve phylogenetic relationships in Lamiaceae. The intergenic spacer rbcL-accD and petB-petD intron have been identified as highly variable regions in other plants [95,96]. The 10 highly variable regions (psbA-trnH, trnK-rps16 intron, petN-psbM, rbcL-accD, petA-psbJ, petB-petD intron, ndhF, rpl32-trnL, rps15-ycf1, and ycf1; Fig 7) identified here could be used as potential barcodes for species identification and phylogenetic study of Scutellaria. Although further research is needed to investigate the reliability and effectiveness of using these regions and/or complete plastome sequences for DNA barcodes in Scutellaria, the results obtained here could be a reference for future studies on global genetic diversity assessment, phylogeny, and population genetics.

Phylogenetic relationships within Scutellarioideae
Our study is the first to use complete plastome sequences to reconstruct the phylogeny of Scutellarioideae. The phylogenetic tree obtained here is largely consistent with previous studies based on the plastid DNA markers [1,9,97,98]. However, some phylogenetic relationships within Lamiaceae differ from recent nuclear trees [99]. Such incongruence between plastid and nuclear phylogenies emphasizes a need for phylogenetic inferences based on both plastome sequences and nuclear data, which can together both robustly resolve relationships and point to potential ancient hybridization events.
The monophyly of Scutellarioideae is confirmed based on the analyses of all datasets (Fig 8,  S3-S8 Figs), and the major splits determined in this study for Scutellarioideae agree with previous studies [1,9]. This study confirmed that the monotypic genus Wenchengia is sister to the remainder of Scutellarioideae (Fig 8). This relationship has been reported in a previous study using two DNA markers (i.e. rbcL and ndhF; [9]). The accession of W. alternifolia from Vietnam was recovered in a clade with an accession of W. alternifolia from Hainan, China in our analyses. The genus has long been thought to be endemic to Hainan Island in China and was only recently reported from Vietnam. As suggested by Paton et al. [16], the distribution of Wenchengia in Vietnam indicates that the Hainan populations are probably relicts of a once more widely distributed W. alternifolia. The discovery of living plants in Vietnam offers the opportunity for population genetic and biogeographic studies of Wenchengia in future.
The African genus Tinnea is sister to Scutellaria, as reported by Wagstaff et al. [8] and Li et al. [1,9]. Although Renschia has never been included in a molecular analysis, morphological characters, e.g. ciliate anthers, well-developed nectar disk, bilabiate calyx with entire, rounded lips, and the closing of the calyx during fruit maturation [6]), suggest a close relationship among Renschia, Tinnea, and Scutellaria. Renschia is probably most closely related to Tinnea based on distribution (both genera are distributed in Africa; Renschia is endemic to North Somalia and Tinnea to tropical Africa) and morphology. Vatke [100] established Renschia based on Tinnea heterotypica S. Moore, and distinguished Renschia from Tinnea by its protruding stamens, the short and basal areoles of nutlets, and the indistinct nervation of calyces.
A total of 11 species of Scutellaria were sampled from both subgenera sensu Paton [5]. The monophyly of Scutellaria is supported here as in other studies [1,9,18,43], but the infrageneric classification of Scutellaria as proposed by Paton [5] is not supported by the present study (Fig 8). As shown in Fig 8, in our sampling Scutellaria is comprised of two subclades: Subclade I included five taxa from subg. Scutellaria and two taxa from subg. Apeltanthus; Subclade II consists of six species from subg. Scutellaria sect. Scutellaria. Species from sect. Scutellaria are recovered in both subclades, thus the monophyly of subgenus Scutellaria and sect. Scutellaria is not supported by the plastome sequences in this study or nuclear ribosomal sequences in previous studies [18,43]. With only one species of sect. Anaspis sampled here, it is premature to assess its monophyly. Though a recent study by Safikhani et al. [18] revealed that sect. Anaspis is a well-supported group, only four representatives of the section from Iran were included in their study. Subgenus Apeltanthus is well supported in all studies [18,43]. The two sections of subg. Apeltanthus, sect. Apeltanthus and sect. Lupulinaria, are shown to be monophyletic in our study as in Zhao et al. [43]. However, based on a broader sampling, Safikhani et al. [18] revealed that neither of the two sections is supported. Further phylogenetic study of subg. Apeltanthus is needed based on a more comprehensive sampling and more DNA markers.
Despite the limited sampling, our study, based on complete plastomes, presents a more resolved and better supported phylogeny of Scutellarioideae than previous studies [1,9,18,43,98]. All the phylogenetic trees inferred from the complete plastome sequences have higher resolution (Fig 8)