Introduction

Malvaceae Juss. is a family constituting the monophyletic core Malvales with Bombacaceae, Sterculiaceae and Tiliaceae (Alverson et al. 1998, 1999; Judd and Manchester 1997). However, recent cladistic analyses of the morphological, anatomical, palynological and chemical characteristics (Judd and Manchester 1997) and molecular evidences (Alverson et al. 1998, 1999; Bayer et al. 1999a, b; Hernández-Gutiérrez and Magallón 2019; Richardson et al. 2015; Wang et al. 2021) have suggested that Tiliaceae, Sterculiaceae, and Bombacaceae are non-monophyletic and that only the Malvaceae is likely monophyletic. In addition, Tiliaceae, Sterculiaceae, and Bombacaceae were merged into the expanded Malvaceae (Judd and Manchester 1997). With sequence analyses of the plastid genes atpB and rbcL, the expanded Malvaceae was divided into nine subfamilies, i.e., Grewioideae, Byttnerioideae, Tilioideae, Helicteroideae, Dombeyoideae, Brownlowioideae, Sterculioideae, Malvoideae and Bombacoideae (Bayer et al. 1999a; Byng et al. 2016; Stevens 2001).

Traditionally, the genus Reevesia was placed into tribe Helictereae under family Sterculiaceae (Bentham and Hooker 1862; Cronquist 1988; Hsue 1984a; Hutchinson 1967; Takhtajan 1997; Thorne 1992; Wu et al. 2003). When it was established by Lindley (1827), it was assigned to Byttneriaceae (including Sterculiaceae) and presumed that it is similar to Pterospermum Schreb. in having petals and capsules and Sterculia L. in having calyx and stamens. However, due to recent research, Reevesia was merged into the expanded Malvaceae with Sterculiaceae (Byng et al. 2016; Stevens 2001). Accordingly, the genus Reevesia was assigned to tribe Helictereae under subfamily Helicteroideae. Reevesia is easily distinguished from other genera by having the large terminal inflorescences, the bisexual flowers with the presence of white corolla and conspicuous long androgynophore, woody capsules and membranous-winged seeds (Bayer and Kubitzki 2003; Tang et al. 2007).

Reevesia Lindl. is a genus representing an example of well-known eastern Asian-eastern North American floristic disjunction (Tiffney 1985; Wen 1999; Wen et al. 2010; Wolf 1980). The genus contains 25 species, of which 23 species occur in East and Southeast Asia and two species in Mexico and Nicaragua (Bayer and Kubitzki 2003; Dorr 2017; Mabberley 2017; Solheim 1991; Xu and Hsue 2001). The genus was considered to be restricted to East Asia from Northeast India to Southeast China and Indochina for a long time (Hsue 1984a; Long et al. 1985; Mabberley 1987; Phengklai et al. 2001; Wu 1991; Wu et al. 2003). The American genus Veeresia Monach. & Moldenke was considered to resemble Asia Reevesia and was named so by an anagram of the latter’s name, but differs by the presence of the staminodes (Monachino 1940). However, further studies revealed that both genera have minute tooth-like staminodes at the summit of the anther column, and therefore, Veeresia was treated as the synonym of Reevesia (Bayer and Kubitzki 2003; Dorr 2017; Machuca 2017; Solheim 1991; Tang et al. 2007; Terada and Suzuki 1998).

Some species of Reevesia are of significant economic value as ornamental, timber, fiber and medicine (Bayer and Kubitzki 2003; Chang et al. 2013; Hsue 1984b). The species of Reevesia are well-known as “nine-layer barks”, “oil hemp tree”, “hard shell fruit tree”, etc. (Hsue 1984b). Some species have been cultivated as ornamental garden plants for their large inflorescence and white fragrant flowers. The plants of Reevesia are rich in fiber, which can be used as raw material for weaving sacks, rope and paper making (Bayer and Kubitzki 2003; Hsue 1984b).

The species delimitation of Reevesia depends on vegetative features including indumentum, petiole length, shape and size of leaf blades (Anthony 1926; Feng et al. 2022; Hsue 1984b; Solheim 1991; Tang et al. 2007). However, this is not accurate. Long (1985) thought that using leaf blade shapes to distinguish species in Reevesia with the same flower morphology characteristics is unreasonable. Due to the lack of significant differences in reproductive characteristics, the challenge of species delimitation is often enhanced. For example, R. siamensis Craib (1924) was treated as a variety of R. pubescens Mast. by Anthony (1926). However, this treatment was not supported by studies of pollen morphology. Reevsia siamensis has pentacolporate pollen grains, while tetracolporate pollen grains are found in R. pubescens (Long et al. 1985). However, the phylogenetic relationships within Reevesia have not been determined due to the lack of plastome.

The plastid organelle is a remarkable feature in plants, and is considered to have evolved from cynobacterium since tens of millions of years ago (Timmis et al. 2004). Compared with the nuclear genome, the plastid genome (plastome) has a simple structure with a low molecular weight and multiple copies. Previous studies have revealed that the plastome, whose size ranges from 120 to 170 kb and includes 120–130 genes under normal conditions, consists of a characteristic circular quadripartite structure that includes a large single copy (LSC) and small single copy (SSC) separated by a pair of inverted repeats (Aldrich et al. 1985, 1988; Bendich 2004). However, there are some exceptions to the normal structure, such as IRLC (IR-lacking clade) in Fabaceae (Moghaddam et al. 2022), three tandem repeats with the same direction in Euglena gracilis and IR regions with direct duplication in Porphyra purpurea (Xing and Liu 2008). Due to its low substitution rates of nucleotides and relatively conserved variation in genomic structures, plastomes are extensively utilized for phylogenetic relationship inference, DNA barcoding and molecular markers development (Du et al. 2021; Luo et al. 2021; Ren et al. 2020; Tian et al. 2021). With the development of high-throughput sequencing technologies, it has been demonstrated that the plastomes are remarkable at resolving phylogenetic relationships at different taxonomic levels (Barrett et al. 2016; Ruhsam et al. 2015; Wang et al. 2016; Wu and Chaw 2016; Xie et al. 2019; Zhang et al. 2022b). However, up to date, the plastomes in Reevesia have been sequenced for four species, R. botingensis Hsue, R. pycnantha Y. Ling, R. rotundifolia Chun, and R. thyrsoidea Lindl., and are available for public usage (Quan et al. 2020; Wang et al. 2021; Zhang et al. 2022b).

In the present study, eleven plastomes of Reevesia species were newly sequenced. The aims are to: (1) compare the structural variation, examine variations of simple sequence repeats (SSRs), and calculate the nucleotide diversity for future population genetic and phylogenetic studies; (2) identify the hypervariable regions of these plastomes as DNA barcodes for species delimitation and identification; (3) test for the presence of adaptive evolution in all genes located in the two single copy regions and one of the two IR regions, using selective pressure analysis; and (4) resolve phylogenetic relationships within the genus and among genera within the family Malvaceae s.l.

Materials and methods

Taxon sampling and DNA extraction

Eleven samples representing ten Reevesia species were newly sequenced in the present study. Among them, eight species were sequenced for the first time. Samples were collected in their native habitats during several botanical surveys, or obtained from Nanjing Botanical Garden MEM, Sun Yat-sen or Shanghai Jiao Tong University. Leaf samples were preserved with silica gel and the voucher specimens were deposited at the Herbarium of Nanjing Forestry University (NF). A list of all the newly sequenced samples with geographical origin, collection history, herbarium voucher information and GenBank accession numbers were included in Table 1. Total genomic DNA was extracted from silica gel-dried leaf material. DNA was extracted using the CTAB protocol modified from Doyle and Doyle (1987) and Super Plant Genomic DNA Kit (Polysacchardes & Polyphenoilics-rich; TIANGEN, Beijing).

Table 1 Basic information and NCBI GenBank accession numbers of 11 Reevesia samples used in this study
Full size table

Plastome sequencing, assembly, and annotation

The DNA concentration was quantified using a Qubit Fluorometer or microplate reader and visualized in a 1% agarose-gel electrophoresis. The genomic DNA was sheared to produce 270 bp fragments with Covaris, and was purified using AxyPrep Mag PCR clean up Kit. The selected fragments were amplified after suffering from end repair and adding polyA tail and adaptor ligation. The processed fragments were heat denatured to single strand after purification. The single strands were circularized, and single strand circular DNA was obtained as the final library. Then, the DNA extractions were sent to BGI (The Beijing Genomics Institute, China) for library preparation and sequencing.

DNA was sequenced with 101 bp paired-end reads using the Illumina HiSeq 4000 platform at BGI. Raw reads were first filtered to obtain high-quality data by removing adapter sequences and low-quality reads using the program SOAPnuke (Chen et al. 2018). Bases were discarded in case of quality < Q20 of read ends. The remaining reads were trimmed if average quality in a 5-bp window was < Q20. After trimming, reads were removed if their length fell below 36 bp. The number and quality of clean reads obtained from each Reevesia sample were evaluated with FastQC v.0.11.9 (Andrews 2010), and the details were provided in Table 1.

The mapped plastid reads were assembled into contigs using the GetOrganelle v.1.7.5.3 with the suitable parameter setting for Reevesia (wordsize = 102, k-mer = 75,85,105,115,127) by mapping the reads to reference genomes obtained from GenBank (Durio zibethinus Moon, NC_036829.1, Gossypium tomentosum Nutt. ex Seem., HQ325745.1 and R. thyrsoidea, NC_041441.1). Bandage v0.8.1 software was used to map all reads to the assembled plastid sequence for visualization processing and obtaining accurate plastomes (Wick et al. 2015). Relative position and direction of contigs were manually adjusted according to the reference genome using Geneious v.9.0.2 (Kearse et al. 2012).

Successfully assembled plastomes were annotated using the PGA program (Qu et al. 2019) with Durio zibethinus, R. rotundifolia as references. The start/stop codons and the exon/intron boundaries of genes were manually adjusted in Geneious v.9.0.2 (Kearse et al. 2012). The plastomes were visualized using OGDRAW v.1.3.1 (https://chlorobox.mpimp-golm.mpg.de/OGDraw.html) (Greiner et al. 2019). The complete plastid sequences and gene annotation of the 11 newly assembled Reevesia Samples were submitted to NCBI database (https://www.ncbi.nlm.nih.gov) under the accession numbers listed in Table 1.

Genome comparison and structural analysis

For each plastome, the LSC, SSC, IRa and IRb regions were plotted with boundary positions being compared using IRscope online software (https://irscope.shinyapps.io/irapp/; Amiryousefi et al. 2018).

All protein-coding genes (CDSs) for each Reevesia plastome were extracted using Geneious v.9.0.2. The amount of codon and relative synonymous codon usage (RSCU) ratio was calculated using CodonW v.1.4.2 software (http://codonw.sourceforge.net/) with default parameters.

Repeated sequences analysis

Repeat structure in 15 Reevesia plastomes was examined and compared. SSRs were identified using the MISA online software (https://webblast.ipk-gatersleben.de/misa/). The thresholds of repeats were 12 for mono-, 6 for di-, 4 for tri-, and 3 for tetra-, penta- and hexa-nucleotides.

Sequence divergence analysis

Genomic similarity analysis was performed using the Shuffle-LAGAN mode in mVISTA (Brudno et al. 2003; Frazer et al. 2004), with R. botingensis (OP856555) as the reference. The nucleotide diversity (Pi) for all protein-coding and noncoding (intergenic spacer and intron) regions was calculated using sliding window analysis to detect the most divergent regions (i.e., mutation hotspots) in DnaSP v.5.10.01 (Librado and Rozas 2009; Rozasas and Sánchez 2003).

Selective pressure analysis

Selective pressures were analyzed for 80 common CDSs of 15 Reevesia plastomes included in the phylogenetic analysis. The values of nonsynonymous (dN) and synonymous nucleotide (dS) substitution rates and the ratio (ω = dN/dS) for each protein-coding gene were calculated using the Codeml program in PAML4.9 with the site-specific model (seqtype = 1, model = 0, NSsites = 0, 1, 2, 3, 7, 8; Yang 2007; Yang and Nielsen 2002). The likelihood ratio tests (LRTs) were used to identify positively selected sites in comparisons of M0 vs. M3, M1 vs. M2, M7 vs. M8 (Yang and Nielsen 2002). The codon frequencies were set by the F3 × 4 model. The Bayes Empirical Bayes (BEB) method was used to identify codons under positive selection (Yang et al. 2005). BEB values higher than 0.95 indicate sites that are potentially under positive selection. For CDSs that were detected under positive selection from the site-specific model, the branch model was used to detect signatures of positive selection along specific lineages by using three models (one-ratio, free-ratio and two-ratio model). The one-ratio model (m0), assumes the same ω ratio for all branches in the phylogeny. This model was compared to the free-ratio model (m1), which assumes that an independent ω ratio for each branch. The m0 was then compared to the two-ratio model (m2), which assumes that the Reevesia clade (set as the foreground branch) has a ω value different from the ratio of the other lineages (set as the background branch). Similarly, the LRT was used to identify positively selected branches in comparisons of m0 vs. m1 and m0 vs. m2 (Whelan and Goldman 1999). In general, genes were considered to be under strong positive selection when ω > 1, under neutral evolution when ω = 1 and under higher pressure of negative selection when ω < 1.

Phylogenetic reconstruction

Outgroups included representatives in eight subfamilies of Malvaceae s.l. and one species of the sister family Dipterocarpaceae. All the samples available for Pterospermum (Dombeyoideae, Malvaceae) and Sterculia (Sterculioideae, Malvaceae) were included because preliminary morphological investigations suggested the affinities to Reevesia (Lindley 1827).

Datasets were separately aligned using MAFFT v.7.037 (Katoh and Standley 2013) as implemented in Geneious v.9.0.2 using default settings and manually adjusted. Phylogenetic relationships were inferred using maximum likelihood (ML) and Bayesian inference (BI). Best-fit nucleotide substitution models were determined using MrModeltest v.2.3 (Nylander 2004) based on the Akaike Information Criterion (AIC). ML analyses were performed using RAxML v.8.2.10 (Stamatakis 2014) with GTR + GAMMA model, and they comprised 100 runs and 1,000 thorough bootstrap (BS) replicates. BI was conducted in MrBayes v.3.2.7 (Ronquist and Huelsenbeck 2003). The Markov chain Monte Carlo (MCMC) was carried out for 6,000,000 generations sampling trees every 1000th generation. The initial 25% of the sampled trees were discarded as burn-in, and the two runs were combined. Both ML analyses and BI analyses were visualized by the FigTree v.1.4.3 (http://tree.bio.ed.ac.uk/software/figtree).

Results

Plastome features of Reevesia

The number of clean reads in the 11 newly sequenced Reevesia plastomes ranged from 6.6 to 17.4 million. Mean read coverage across these accessions ranged from 101 to 248x (Table 1). Assembled plastomes ranged in size from 161,532 (R. botingensis; OP856555) to 161,945 bp (R. glaucophylla). All plastomes show the typical quadripartite structure of angiosperms, which consists of an LSC with a length between 90,225 and 90,665 bp, an SSC with a length between 20,289 and 20,317 bp; and a pair of IRs with a length between 25,465 and 25,495 bp (Table S1, Fig. 1). The total GC content is nearly similar, 36.8% for 10 samples and 36.9% for R. botingensis (OP856555; Table S1). The GC content in the IR regions (42.9–43.0%) is noticeably above that of the LSC (34.6–34.7%) and SSC (31.4–31.5%) regions in each plastome (Table S1).

Fig. 1
figure 1

Gene map of 11 Reevesia plastomes. The outer circle shows the genes at each locus, and inverted repeat regions are indicated with thicker lines. Genes on the outside of the outer circle are transcribed in a counterclockwise direction, while genes on the inside of the outer circle are transcribed in a clockwise direction. The inner circle indicates the range of the LSC, SSC, and IRs, and also shows a GC content graph of the genome. In the GC content graph, the dark gray lines indicate GC content, while light gray lines indicate the AT content at each locus

Full size image

Each of the sequences encoded 80 protein-coding genes, 30 transfer RNA (tRNA) genes, and 4 ribosomal RNA (rRNA) genes. Seven protein-coding genes (rps12, rpl2, rpl23, ycf2, ycf15, ndhB, and rps7), seven tRNAs (trnI-CAU, trnI-GAU, trnL-CAA, trnN-GUU, trnR-ACG, trnV-GAC, and trnA-UGC), and all four rRNAs (rrn4.5, rrn5, rrn16 and rrn23) had two copies located at the IR regions (Table S2). Eighteen genes contained one (atpF, ndhA, ndhB, petB, petD, rpl2, rpl16, rpoC1, rps12, rps16, trnA-UGC, trnG-GCC, trnI-GAU, trnK-UUU, trnL-UAA and trnV-UAC) or two (clpP and ycf3) introns, and six of these were the tRNA genes (Table S2, Fig. 1). The genomes consist of 56.2–56.3% coding regions (48.9–49.0% protein coding genes and 7.3% RNA genes) and 43.7–43.8% non-coding regions, including both intergenic spacers and introns (Table S1).

The boundaries between the IR and LSC/SSC regions of the 12 Reevesia species (15 individuals) and the three species in other genera in Malvaceae s.l. were compared (Fig. 2). At the IRb/LSC boundary, the IR ends within the rps19 gene so that 6 bp of the 5’ end of the gene is duplicated in 14 Reevesia plastomes. In R. botingensis (MN533972.1), Pterospermum kingtungense C.Y. Wu ex H.H. Hsue and Sterculia lanceolata Cav., the IRb/LSC boundary shifted into the IGS between the rps19 and rp12 genes, and in Durio zibethinus, the boundary further shifted into the IGS between rpl23 and trnI. The IRa/LSC boundary is located between the rpl2 and trnH genes in all species except in Durio zibethinus, which is located between the trnI and trnH genes. The IRa/SSC boundary in all 15 Reevesia samples, Durio zibethinus and Sterculia lanceolata are located between ycf1 and trnN, while in R. botingensis (MN533972.1) and Pterospermum kingtungense, the boundary is located within ycf1. The IRb/SSC boundary of all the 17 plastomes are located between trnN and ndhF genes.

Fig. 2
figure 2

Comparisons of the LSC, SSC and IRs boundaries within Reevesia and among three other Malvaceae s.l. plastomes

Full size image

Codon preference analysis

Codon usage patterns of protein coding sequences for the 15 Reevesia plastomes were calculated based on the relative synonymous codon usage (RSCU) value. The total sequence sizes of the coding genes for codon analysis were 72,906–79,164 bp and 26,302–26,388 codons were encoded (Table S3). Leucine encoded with the highest number of codons ranging from 2,577 to 2,785, followed by isoleucine, with the number of codons encoded between 2,088 and 2,263. Cysteine is the least abundant (286–305). The RSCU values varied slightly among the 15 Reevesia sequences. Thirty codons are used frequently with RSCU > 1 and 32 codons used less frequently with RSCU < 1. UUA showed a preference in all the 15 plastomes. The frequency of use for the start codons AUG and UGG, encoding methionine and tryptophan, showed no bias (RSCU = 1). Codons with A (31.84%) or U (37.92%) in the third position are all 69.76%, thus the codon usage is biased toward A or U at the third codon position.

Repeated sequences analysis

The total number of SSRs in 15 Reevesia plastomes ranged from 35 to 45 (Table S4). Six kinds of SSRs were detected in the fifteen species, namely mono-, di-, tri-, tetra-, penta- and hexa- nucleotides. The most common SSRs are mononucleotides, which range from 14 to 24, followed by tetranucleotides ranging from 7 to 10, and dinucleotides ranging from 4 to 8. The number of trinucleotides and pentanucleotides varied from 2 to 7 and 1 to 3, respectively (Fig. 3a). SSRs are mainly distributed in the intergenic regions (65.0–76.2%), with much lower quantities distributed in the intron regions (15.8–27.5%) and exon regions (6.7–10.8%; Fig. 3b). All mononucleotide and dinucleotide SSRs belong to AT type and the majority of trinucleotide, tetranucleotide, and pentanucleotide SSRs are especially rich in AT type.

Fig. 3
figure 3

Distribution of repeats in Reevesia plastomes. a. Distribution of SSRs types. b. Distribution of SSRs among intergenic spacer, intron and exon regions

Full size image

Polymorphic variation and hypervariable regions

Comparative plastome analysis revealed that noncoding regions are generally more divergent than coding regions and LSC/SSC regions are more divergent than IR regions (Fig. 4). Comparison of nucleotide diversity in the LSC, SSC and IR regions indicated that the SSC region shows the highest nucleotide diversity (Pi = 0.00374), followed by LSC (Pi = 0.00269) and the IR (Pi = 0.00171) regions. For the 80 protein-coding regions (CDSs), Pi values for each locus ranged from 0 to 0.00540, with an average of 0.00110, whereas 15 regions (i.e. infA, ycf15, rps16, psbZ, petD, ndhF, rpl32, ycf1, rps15, petN, ndhH, ccsA, rpl20, psbI and rpl36) had relatively high values (Pi > 0.002; see Fig. 5a and Table S5). Across all 133 noncoding regions, Pi values ranged from 0 to 0.04063, with an average of 0.00363. Likewise, thirteen of those regions had relatively high values (Pi > 0.007; see Fig. 5b and Table S5), i.e. rps19-rpl2, ccsA-ndhD, psbZ-trnG, rps2-rpoC2, infA-rps8, ndhH-rps15, trnH-psbA, petN-psbM, ndhD-psaC, rpl33-rps18, trnG-trnR, trnD-trnY and trnM-atpE.

Fig. 4
figure 4

Comparison of plastomes of 15 Reevesia samples. Purple blocks indicate conserved genes, while red blocks indicate noncoding sequences (CNS). White blocks represent regions with sequence variation among the 15 plastomes. Gray arrows indicate the direction of gene transcription

Full size image
Fig. 5
figure 5

Comparison of nucleotide variability (Pi) values in Reevesia plastomes. a. Pi values among protein-coding genes (CDS). b. Pi values among intergenic spacer (IGS) regions

Full size image

Selective pressure analysis

Analysis of 80 protein-coding sequences across 15 Reevesia plastomes indicated that all CDSs were subjected to purifying selection (ω<1; Table S6). Fifty-five genes are under strong purifying selection with the dN/dS ratio of 0.000. No coding sequences with positively selected sites were identified, even in the seven genes with relatively high dN/dS ratios (ω>0.5): ndhD (ω = 0.968), petD (ω = 0.882), accD (ω = 0.829), rpoA (ω = 0.721), ycf4 (ω = 0.658), rpl20 (ω = 0.574) and matK (ω = 0.509).

Phylogenetic analysis

Thirty-nine sequences were used in the phylogenetic analyses. Eleven out of 39 sequences were newly generated (Table S7). The topologies of the trees resulting from ML and BI analyses based on whole plastid genomes are congruent. There are only slight differences in support values between the phylogenetic trees. Therefore, only the ML topology is shown here with the BI support values added at each node (Fig. 6).

Fig. 6
figure 6

Phylogenetic tree based on plastomes resulting from the maximum likelihood analysis (ML) with Bayesian inference (BI) value at nodes. Names of taxa newly sequenced in Reevesia are in bold

Full size image

The whole plastid genome analysis strongly supported the monophyly of Malvaceae s.l. and its two major phylogenetic lineages, i.e., Byttneriina and Malvadendrina. Byttneriina includes subfamilies Byttnerioideae and Grewioideae, near the base of the plastome phylogeny, and appears as sister to the remaining subfamilies. Six monophyletic branches that correspond to Malvoideae, Bombacoideae, Dombeyoideae, Helicteroideae, Sterculioideae and Tilioideae, nested in one lineage Malvadendrina.

Reevesia forms a well-supported monophyletic group (ML = 100, PP = 1.0). The genus is sister to Durio zibethinus and nested well within subfamily Helicteroideae (Fig. 6). Within Reevesia, three full-supported clades are resolved. Clade A comprises two samples of R. botingensis, which were resolved as sisters to the remainder of the genus. Clade B includes the type of the genus, R. thyrsoidea, together with R. longipetiolata and R. tomentosa. Clade C consists of eight taxa, i.e., R. glaucophylla, R. lofouensis Chun & Hsue, R. membranacea Hsue, R. orbicularifolia Hsue, R. pubescens, R. xuefengensis (C.J. Qi) C.J. Qi, R. pycnantha and R. rotundifolia.

Discussion

Plastome structure comparisons

The comparative analysis of plastomes provides new insights into sequence variation and molecular evolutionary patterns. In the present study, 15 Reevesia plastomes representing 12 species were compared to 23 plastomes reported in Malvaceae s.l. to clarify phylogenetic relationships and resolve taxonomic uncertainties.

In terms of plastome size, the 11 newly sequenced Reevesia samples (161,532–161,945 bp) are slightly larger than most plastomes reported in Malvaceae s.l. (Fig. 1; Cai et al. 2015; Cheon et al. 2017; Heckenhauer et al. 2019; Lee et al. 2006; Wang et al. 2021). All the newly sequenced Reevesia plastomes contain 132 genes, which is consistent with plastomes reported previously in the genus (Quan et al. 2020; Wang et al. 2021; Zhang et al. 2022b). Indeed, the gene content and order of the Reevesia plastomes are highly conserved. The GC content of the plastomes newly sequenced (36.8–36.9%) are in accordance with that of other reported Reevesia plastomes and other Malvales (Heckenhauer et al. 2019; Wang et al. 2021).

The IR regions are resistant to re-combinational loss and therefore help in the stabilization of the plastome (Perry and Wolfe 2002). The IR boundaries are highly conserved among species in the genus Reevesia which all possess 18 completely duplicated genes. However, shifts in IR boundaries were detected in different genera of Malvaceae s.l. The IR/LSC of Reevesia have expanded relative to the other three genera Durio, Pterospermum and Sterculia to include a small portion of rps19, which is consistent with the results of recent studies (Wang et al. 2021).

Codon usage preference is closely related to gene expression, mutation pressure on DNA sequences and natural selection (Lyu and Liu 2020; Pfitzinger et al. 1987; Zhou et al. 2013). Analysis of the Reevesia plastomes revealed that leucine was the most abundant and cysteine was the least abundant amino acid, which has also been frequently reported in the chloroplast genomes of other angiosperms (Ren et al. 2022; Somaratne et al. 2020; Wen et al. 2021). The bias for high representation of A and U at third codon position is observed in this study and is also shown in other land plant plastomes (Moghaddam et al. 2022; Wen et al. 2021).

SSRs are short stretches which have tandemly repeated DNA motifs of 1–6 nucleotides and are often used as genetic markers to identify related species, to explore population relationships and to explore evolutionary history due to their high polymorphism levels (Li et al. 2019; Powell et al. 1995; Zhang et al. 2012). Six types of SSRs were detected in Reevesia plastomes, of which A/T mononucleotides were the most common as reported in previous studies (Ren et al. 2022). SSRs in Reevesia are mainly restricted to intergenic spacers in LSC regions. Furthermore, the similarity across the twelve species is supported by SSRs alignment.

All the protein-coding genes are under purifying selection in the Reevesia plastomes compared. However, eight of them (ndhD, petD, accD, rpoA, ycf4, rpl20 and matK) evolved rapidly with relatively high dN/dS ratio. Among these genes, positively selected sites were detected in accD by Wang et al. (2021) and in petD by Wu et al. (2018) in Malvaceae s.l. plastomes. Purifying selection is responsible for genomic sequence conservation across long evolutionary timescales (Cvijović et al. 2018) and can be attributed to morphological similarities and habitat homogeneity among Reevesia species.

The nucleotide diversity was higher in the non-coding regions than the coding regions, which is generally consistent with most previous studies of the chloroplast genomes of angiosperms (Cao et al. 2022; Ye et al. 2018). At present, data from non-coding regions is the most commonly used tool for phylogeographic and phylogenetic studies at low taxonomic levels of plants (Mapaya and Cron 2021; Shaw et al. 2014). The rps19-rpl2, ccsA-ndhD, psbZ-trnG, rps2-rpoC2, infA-rps8, ndhH-rps15, trnH-psbA, petN-psbM, ndhD-psaC, and rpl33-rps18 are identified as the top ten variable regions in the non-coding regions at the species level within Reevesia (Figs. 3 and 4). Among them, two, rps19-rpl2 and ccsA-ndhD were identified as the best possible choices for low-level phylogenetic studies of Malvaceae s.l. (Wang et al. 2021). However, to determine whether these variable regions could serve as powerful DNA barcodes for species delimitation and identification in the genus, a broad sampling of well-dispersed species across the genus need to be included for further studies.

The cp. genome sequences of three species with two samples were compared in detail. R. thyrsoidea and R. membranacea were more similar in the IR/LSC boundary pattern, and R. botingensis obviously different between the two samples (Fig. 2). R. botingensis and R. membranacea exhibited the distinct SSRs patterns, and these distinctions may be helpful for population genetic analysis in Reevesia (Fig. 3). All codons of R. thyrsoidea and R. membranacea had identical RSCU values, 53 of which were different from those of the two R. botingensis samples (Table S3). Discrepancies in annotation protocols may eventually explain all the differences in the two chloroplast genomes of R. botingensis.

Phylogenetic analysis

Whole plastid genomes are used widely for inferring phylogenetic relationships at the generic (Saski et al. 2005; Simmonds et al. 2021), supra-generic (Heckenhauer et al. 2019; Wang et al. 2021; Ye et al. 2018), interspecific and intraspecific species levels (Song et al. 2022; Wicke et al. 2011). In this phylogenetic analysis of Reevesia, the monophyly of the genus and the relationship with other genera in Malvaceae s.l. were investigated. The relationships among species within the genus and the delimitation of some species were also examined.

In consistent with previous studies based on molecular (Alverson et al. 1999; Nyffeler et al. 2005; Wang et al. 2021) morphological (Judd and Manchester 1997) and biogeographical data (Bayer et al. 1999), the 39 species analyses of whole plastid genomes strongly support the monophyly of the Malvaceae s.l. Within Malvaceae s.l., two major lineages, Byttneriina and Malvadendrina, and the eight subfamilies included in this study were found to be monophyletic.

The genus Reevesia is characterized by a number of vegetative and reproductive characteristics, such as longitudinal fissured bark, flowers with conspicuous long androgynophores, 5-angled capsules and winged seeds. The monophyly of Reevesia is supported here and in previous molecular studies based on whole plastid genomes (Wang et al. 2021) or a few sequences (Bayer et al. 1999; Nyffeler and Baum 2000).

When the genus Reevesia was established by Lindley (1827), it was compared with Pterospermum and Sterculia. However, in our analysis, Reevesia is sister to Durio zibethinus in Durioneae, and did not nest in the same lineage with Pterospermum and Sterculia. It is in accordance with the result of Wang et al. (2021) based on whole plastid genomes data. Previous studies based on a few molecular sequences (Alverson et al. 1999; Bayer et al. 1999; Hernández-Gutiérrez and Magallón 2019; Nyffeler and Baum 2000) have demonstrated that Reevesia was sister to the monotypic genus Ungeria Schott & Endl. and more closely related to Helicteres L., Mansonia J.R. Drumm. ex Prain, Triplochiton K. Schum. and Durio rather than to Pterospermum and Sterculia. Traditionally, Pterospermum was included in Helictereae (Bentham and Hooker 1862; Hutchinson 1967), or elevated to a separate tribe Pterospermeaeae C.Y. Wu et Y. Tang (Tang 1992), although it is now shown to be distantly related to Reevesia and is placed in Dombeyoideae (Alverson et al. 1999; Bayer et al. 1999). Sterculia, the type of Sterculiaceae, is also distantly related to Reevesia and is included in Sterculioideae.

Within Reevesia, three full-supported clades are resolved. Clade I includes Reevesia botingensis, which is the first diverging species among the genera. This species is endemic to China, occurring in the Hainan islands, and it can be easily distinguished from other species by having oblanceolate and glabrous leaves. Clade II consists of three species, R. longipetiolata, R. thyrsoidea and R. tomentosa formed a full supported clade B, and the remaining eight taxa were included in clade C. In terms of vegetative and reproductive characters, clade B and C are rather heterogenous and it is difficult to come up with good and consistent synapomorphies. Thus, it remains unclear if this indeed represents the basal split in Reevesia.

Three species with two samples, Reevesia botingensis, R. thyrsoidea and R. membranacea, are all supported as monophyletic. However, the recent treatment of two species, R. pubescens by Hsuee (1984b) and Tang et al. (2007)d thyrsoidea by Feng et al. (2022), are not supported here. Both Hsuee (1984b) and Tang et al. (2007) treated R. membranacea as a synonym of R. pubescens and R. xuefengensis as a variety of R. pubescens. In our analysis, both R. membranacea and R. xuefengensis are not clustered with R. pubescens. Reevesia pubescens located at the base of clade C, while R. membranacea was sister to R. xuefengensis and located at the distal end of clade C. When published, R. membranacea was considered to be different from other species in the genus in having the membranous leaves and biauriculate protuberances at the middle part of petals (Hsue, 1963). Reevesia xuefengensis was first described as a variety of R. pubescens by Qi (1984). In 2000, Qi himself elevated it to a species rank. It is obviously different from R. pubescens by having the deciduous and lately glabrous leaves. Recently, Feng et al. (2022) merged four species, i.e., R. botingensis, R. lofouensis, R. longipetiolata and R. pycnantha, with R. thyrsoidea, all of which were recognized by Hsuee (1984b) and Tang et al. (2007) as independent species. Our results supported treating them as independent species. Morphologically, R. botingensis differs from R. thyrsoidea in having the oblanceolate leaves; R. lofouensis differs in the present of densely yellowish stellate pubescent on the branchlets, R. longipetiolata differs in having cuneate leaf base and R. pycnantha differs in having papery and deciduous leaves. Furthermore, these five species can be distinguished from each other in blooming time: R. botingensis flowers from late December to February; R. longipetiolata flowers in early March; R. lofouensis flowers in mid-May while R. thyrsoidea flowers in mid-April in the evergreen broad-leaved forest of Luofu Mountain; and R. pycnantha flowers from May to July. Therefore, we propose to treat them as distinct species.

Conclusions

In summary, 15 plastomes representing 12 Reevesia species were analyzed here. Although plastomes of Reevesia show a conservative quadripartite structure, the top ten hypervariable regions identified in noncoding regions, and the long repeated sequenced and SSR screened are informative for species delimitation and identification, phylogenetic and population genetic studies in this genus. A phylogenetic analysis based on 39 plastomes supported the monophyly of Reevesia, the eight subfamilies of Malvaceae s.l. included and Malvaceae s.l. Moreover, the analysis indicated that Reevesia forms a natural group and is sister to Durio within subfamily Helicteroideae. Redefining of R. pubescens and R. thyrsoidea is proposed for the two species newly delimitated are not monophyletic. Overall, the results of this study provide better understanding of species delimitation and the phylogenetic position of the genus Reevesia.