Comparative and phylogenetic analyses of six Kenya Polystachya (Orchidaceae) species based on the complete chloroplast genome sequences

Background Polystachya Hook. is a large pantropical orchid genus (c. 240 species) distributed in Africa, southern Asia and the Americas, with the center of diversity in Africa. Previous studies on species of this genus have not obtained the complete chloroplast genomes, structures and variations. Additionally, the phylogenetic position of the genus in the Orchidaceae is still controversial and uncertain. Therefore, in this study, we sequenced the complete plastomes of six Kenya Polystachya species based on genome skimming, subjected them to comparative genomic analysis, and reconstructed the phylogenetic relationships with other Orchidaceae species. Results The results exhibited that the chloroplast genomes had a typical quadripartite structure with conserved genome arrangement and moderate divergence. The plastomes of the six Polystachya species ranged from 145,484 bp to 149,274 bp in length and had an almost similar GC content of 36.9–37.0%. Gene annotation revealed 106–109 single-copy genes. In addition, 19 genes are duplicated in the inverted regions, and 16 genes each possessd one or more introns. Although no large structural variations were observed among the Polystachya plastomes, about 1 kb inversion was found in Polystachya modesta and all 11 ndh genes in the Polystachya plastomes were lost or pseudogenized. Comparative analysis of the overall sequence identity among six complete chloroplast genomes confirmed that for both coding and non-coding regions in Polystachya, SC regions exhibit higher sequence variation than IRs. Furthermore, there were various amplifications in the IR regions among the six Polystachya species. Most of the protein-coding genes of these species had a high degree of codon preference. We screened out SSRs and found seven relatively highly variable loci. Moreover, 13 genes were discovered with significant positive selection. Phylogenetic analysis showed that the six Polystachya species formed a monophyletic clade and were more closely related to the tribe Vandeae. Phylogenetic relationships of the family Orchidaceae inferred from the 85 chloroplast genome sequences were generally consistent with previous studies and robust. Conclusions Our study is the initial report of the complete chloroplast genomes of the six Polystachya species, elucidates the structural characteristics of the chloroplast genome of Polystachya, and filters out highly variable sequences that can contribute to the development of DNA markers for use in the study of genetic variability and evolutionary studies in Polystachya. In addition, the phylogenetic results strongly support that the genus of Polystachya is a part of the tribe Vandeae. Supplementary Information The online version contains supplementary material available at 10.1186/s12870-022-03529-5.


Background
The genus Polystachya Hook (1824: 103; Orchidaceae) comprises approximately 240 species, of which most species are distributed in Africa, a few ones extend to tropical and subtropical Americas, and only three species in Asia [1][2][3][4][5][6]. It is difficult to give an exact number of species of the genus for while some species are widespread, others are narrow endemics. Furthermore, the taxonomic nomenclature of the genus is also complex since there are about 500 names available associated with the currently recognized species [2,7]. Representatives of the genus are usually epiphytic, occasionally lithophytic or terrestrial perennial herbs [8,9] (http:// powo. scien ce. kew. org/ taxon/ 325981-2; http:// www. thepl antli st. org/ tpl1.1/ search? q= Polys tachya). Most species of Polystachya are compact plants that take up little space with long-lasting and scented flowers, worth cultivating [8]. The ploidy levels of the genus range from diploids (2n = 2x = 40) to hexaploid (2n = 6x = 120), where it is generally considered to have evolved in the Neotropics, Madagascar and Reunion [10][11][12]. Previous studies indicated that there exists a close relationship between Polystachyinae and the tribe Vandeae. Based on morphological data analysis, Polystachyinae was placed within a larger Vandeae [13][14][15]. However, Carlsward et al. (2006b) maintained a stricter definition of Vandeae in their studies, citing several morphological characters of Vandeae, including monopodial habit, loss of mucilage and tilosomes, and the presence of spherical silica bodies in leaf sclerenchyma, to differentiate it from Polystachyinae [16]. A study by Li et al. (2019), using only one sample of Polystachya found that this genus belongs to Vandeae based on plastid genome sequence analysis. Nevertheless, the results of mitochondrial genome sequence analysis indicated that it may belong to Malaxideae, which is also supported by their shared morphological features such as distinctive pseudobulbs, terminal inflorescences, floral mentum and waxy pollinia [17].
The monophyly of the genus Polystachya has been reported in several studies, but it is uncertain whether its infrageneric taxonomic units are monophyletic [6,18]. The genus was usually divided into 15 sections based on morphological characteristics, but these natural groupings are not fully supported in currently available molecular studies [6,19,20]. In these molecular studies, all the sections are either polyphyletic or paraphyletic, except for sect. Isochiloides (Russell et al., 2010b). A well-resolved phylogenetic hypothesis could help clarify the infrageneric classification of the genus and might be used to redefine sections as a step towards a muchneeded generic revision. The latest taxonomic work to attempt an account of the entire genus was A Monograph of the Subtribe Polystachyinae Schltr. (Orchidaceae) edited by Mytnik-Ejsmont (2011) [18]. The species of the genus Polystachya are reclassified into 13 sections in the book. The classification system divided the sections of Affines (only Polystachya affinis), Isochiloides, Dendrobianthe (including Polystachya dendrobiiflora), and Polystachya longiscarpa (originally in Dendrobianthe) into separate genera. Kermesina and Polystachate were assigned to subsection levels. At present, Polystachya dendrobiiflora is an acceptable name, while Dendrobianthe dendrobiiflora is treated as its synonym, thus Polystachya dendrobiiflora is still used in this study. Apart from the sect. Polystachya which is pantropical, all sections are confined within Africa and Madagascar with two being endemic to Madagascar and the Indian Ocean islands. Although DNA sequence analysis has been informative in studies of the genus as a whole, relationships between members of this pantropical species group seem to play a lesser role. This is probably inseparable from the fact that the genus has a notoriously complicated taxonomy, with several sections that are widely used but probably not monophyletic [6,12,18,19]. In addition, individuals from different geographical regions with duplicate samples may also not be monophyletic groups [21]. There has been some work on the genus, but their results rarely support the infrageneric taxonomic relationships currently delineated by Mytnik-Ejsmont, even with contradictions. Additionally, taxonomical and phylogenetic uncertainties remain in some sections or subsections or species because of poor internal resolutions, low bootstrap support, and incongruent plastid and nuclear gene phylogenies. It is generally believed that hybridization, polyploidization, ambiguous species definition, low sequence divergence level, reticulate evolution, incomplete lineage sorting, and plastid capture may complicate the issue for phylogenetic reconstruction of genus Polystachya [6,12,22]. Moreover, naming in Polystachya species is also complicated due to the influence of factors such as erroneous species identification, multiple synonyms, and highly variable morphological characters. Therefore, the development of more effective genetic resources on the basis of increasing samples is necessary studies in Polystachya. In addition, the phylogenetic results strongly support that the genus of Polystachya is a part of the tribe Vandeae.
Keywords: Polystachya, Complete chloroplast genomes, Comparative genomics, Phylogenetic analysis for further phylogenetic studies of Polystachya. In recent years, an increasing number of researchers have focused on the cp genome to develop genetic markers for phylogenies. The cp genome sequences have been successfully used to evaluate relationships of different taxonomic units and yielded better results in handling the phylogenetic relations of many difficult groups [17,[23][24][25][26][27][28][29][30][31]. Accordingly, it is considered to be an informative and valuable resource for phylogenetic analysis in plants at multiple taxonomic levels.
Chloroplast (cp) is an important semi-autonomous organelle in plant cells with a complete genetic system and its genetic information is called chloroplast genome. The cp genomes in general are inherited uniparentally, and maternally in most angiosperms species at a slower evolutionary rate of change compared to nuclear genomes [32]. The typical plastomes in angiosperms have a highly conserved quadripartite circular structure comprised of a pair of inverted repeat regions (IRs, about 20-28 kb) and two single copy regions (large single copy region, LSC, 80-90 kb; small single copy region, SSC, 16-27 kb) [33]. These genomes range in sizes from 120 to 160 kb and usually encode 120-130 genes, with an overall GC content normally in the order of 30-40% [34][35][36]. Variation in genome size can be mainly attributed to IR expansion/contraction or even loss. The structure, length, GC content, gene type, gene content and order of cp genomes are generally conserved. When studying plant origin and phylogenetic relationships, the plant genome is made up of three parts: nuclear, mitochondrial and chloroplast genomes. Compared with nuclear and mitochondrial genomes, the chloroplast genomes are small, less prone to recombination, and have low rates of nucleotide substitutions, hence can provide distinct genetic information for phylogenetic studies [37,38]. Furthermore, the phylogenetic reconstruction based on complete cp genome sequences may reduce errors and uncertainty resulting from insufficient sampling of DNA sequences [39].
With the rapid development of next-generation sequencing (NGS) technology, it is now more convenient and cheaper to obtain cp genome sequences, feasible to address various phylogenetic questions at the different taxonomic levels. Currently, over 6000 cp genomes of plants are available in the National Center for Biotechnology Information (NCBI) organelle genome database (https:// www. ncbi. nlm. nih. gov), among which about 500 complete cp genome sequences (c. 370 species) of Orchidaceae have been released by NCBI (2021/5/20). Up to now, there is only one report on cp genome sequences of the genus Polystachya [21]. However, our reassembly of the original data from the report revealed that the cp genome sequences reported in that paper were both incomplete. Therefore, it is necessary to perform a comprehensive cp genomic comparison and phylogenetic analysis in the Polystachya.
In this study, we newly sequenced the complete chloroplast genomes of six Polystachya species and conducted comparative genomic analyses. Another 79 published cp genome sequences (77 from Orchidaceae; Allium cepa from Allioideae and Iris gatesii from Iridoideae were chosen as outgroups) downloaded from the NCBI database were used to construct phylogenetic trees. Our sampling Orchidaceae materials covered 5/5 subfamilies, 16/22 tribes and 20/49 subtribes of Orchidaceae. The objectives of this study are: (1) to report the first complete cp genome of the genus Polystachya and to reveal the structure and sequence variation of plastomes within Polystachya, (2) to reconstruct a more comprehensive and better-resolved phylogenetic tree for exploring the phylogenetic position of Polystachya in Orchidaceae and (3) to screen potential DNA markers in cp genomes that can be used for phylogenetic analysis and classification of Polystachya.

Chloroplast genomes structure and features
We obtained the six complete chloroplast genomes of Polystachya species, and these cp genomes ranged in sizes from 145,484 bp (P. tenuissima) to 149,274 bp (P. dendrolliflora). Like most angiosperms, all newly sequenced Polystachya plastomes displayed a typical quadripartite structure consisting of a pair of inverted repeats IR regions (IRA and IRB; 25,049-25,716 bp) separated by one large single copy region (LSC; 82,104-83,848 bp) and one small single copy region (SSC; 11,894-14,822 bp) (Fig. 1, Table 1). The six cp genomes were all AT-rich, overall GC content ranged from 36.9 to 37.0%, and the GC content of IR region (43.2-43.3%) was always higher than that of LSC and SSC regions (34.3-34.5% and 28.8-29.5%, respectively) ( Table 1). This high GC percentage in the IR regions could be due to the presence of eight ribosomal RNA (rRNA) sequences in these regions. Previous studies have also shown similar results, with the high GC content in IR regions being related to the presence of all rRNA genes in this region [40]. By comparing all sequenced Polystachya cp genomes generated in this study, we found that they had highly conserved gene content, gene number, orientation and intron number.
A total of 106-109 unique genes were identified in the six cp genomes, including a shared 68 protein-coding genes (PCGs), 30 transfer RNA (tRNA) genes, and four ribosomal RNA (rRNA) genes ( Table 1, Table S2). The difference in gene numbers were due to the variations in the number of ψndh among the six species. The gene distribution in these six cp genomes was exactly the same: Fig. 1 Chloroplast genome map of Polystachya species. The full circles represent common genes annotated for the six Polystachya species in this study, and the red empty box indicates differences in the genes annotated in this region of these plastomes. The bottom of the figure exhibits the differences in the genes annotated in this region for the plastomes. Genes shown outside the circle are transcribed in the counter counterclockwise direction, while those inside the circle are transcribed in the clockwise direction. The colored bars indicate genes belonging to different functional groups. The darker gray inside the inner circle denotes GC content while the lighter gray corresponds to the AT content of the genome. The ψ signifies pseudogenes the LSC regions encoded 59 protein-coding genes and 22 tRNA genes, and the SSC regions contained five proteincoding genes and one tRNA gene. Moreover, 19 genes were duplicated in IR regions, including six PCGs (rpl2, rpl23, ycf2, rps19, rps7 and rps12), eight tRNA genes (trnH -CUG , trnI -CAU , trnL -CAA , trnV -GAC , trnI -GAU , trnA -UGC , trnR -ACG and trnN -GUU ), four rRNA genes (rrn16, rrn23, rrn4.5 and rrn5) and one pseudogene (ψndhB) ( Table S2). The remaining non-genic regions include introns, intergenic spacers (IGS), and pseudogene (ψ). Sixteen genes possessed introns: 13 genes (seven PCGs: rps16, atpF, rpoC1, petB, petD, rpl16, rpl2; six tRNAs: trnK -UUU , trnG -GCC , trnL -UAA , trnV -UAC , trnI -GAU and trnA -UGC ) have only one intron, while another three PCGs (rps12, ycf3 and clpP) contain two introns (Table S2 and S3). Among the 16 intron-containing genes, 13 were present in LSC region and three were duplicated in the IR regions. At the same time, we found that the exon length was almost the same in the above 16 intron-containing genes, but the length of introns changed in all these genes. Interestingly, among the 11 plastid genes encoding the subunits of the NAD(P) H dehydrogenase complex (ndh genes), some genes were lost and some were pseudogenized (Fig. 1). The trnK -UUU gene contains the longest intron, matK gene is located within trnK -UUU intron. Rps12 was a special trans-splicing gene, whose first exon is located in the LSC region, while the second and third exons reside in IR regions.

Codon usage analyses
Codon usage frequency and relative synonymous codon usage (RSCU) were calculated based on protein-coding genes. All the 74 protein-coding genes were composed of 19,308-19,373 codons and encoded 20 amino acids in the chloroplast genomes of the six Polystachya species. (Fig. 2, Table S4). Among these amino acids, leucine (Leu; 1941-1955, 10.05-10.11%) is the most frequently used, cysteine (Cys; 209-216, 1.08-1.12%) is the least universal amino acid in the cp genomes of these species. The RSCU value analysis showed that almost all amino acids are encoded by 2-6 synonymous codons, except methionine (Met) and tryptophan (Trp), this strategy could protect protein mutations in biosynthesis. Relative synonymous codon usage is 1 for methionine (Met) and tryptophan (Trp). About half of the codons have RSCU > 1 (30/64), and most (29/30, 96.7%) end with the base A or T. Similarly, about half of the codons have RSCU < 1 (32/64), with the majority (29/32, 90.6%) ending with the base C or G. In nearly all of the protein-coding genes in Polystachya species had the standard ATG/CAT start codon (RSCU = 1), but rpl2 started with ATA/TAT. ATA/ TAT as an initiation codon has been reported in other cp genomes [41,42]. All three stop codons were present, with TAA being the most frequently used among the six plastomes (Table S4).

IR expansion and contraction
The contraction and expansion of IR borders are common evolutionary events and are the major reason for size differences between chloroplast genomes [43,44]. We conducted a comparative analysis to investigate the expansion/contraction of IR among the six species, and found that the cp genomes of Polystachya species are greatly conserved. However, some structural variations were present in the four boundaries (LSC/IRB, IRB/ SSC, SSC/IRA, IRA/LSC) (Fig. 4). Our results showed that the genes rpl22-rps19-psbA and trnN-rpl32-ycf1 were located in the junctions of the LSC/IR and SSC/ IR regions. In all of the six plastomes, the rpl22 gene spans the LSC/IRB junction region and extends to the IRA region for 23-66 bp. Rps19 is duplicated in the IR regions, the distance between the rps19 gene located in the IRB region and the rpl22 gene is 138-149 bp, and the distance between the rps19 gene located in the IRA region and the IRA/LSC boundary is 167-212 bp. IRB/ SSC region is situated in the intergenic regions between trnN -GUU and rpl32, the distance from trnN -GUU to IRB/ SSC region is 328-681 bp. In addition, the distance of

Structure comparison and divergence hotspot identification analysis
Mauve comparison found a reversal of about 1 kb in the IRB/SSC boundary region of the cp genome of P. modesta, in which the rpl32 gene was contained, resulting in the reversal of the rpl32 gene (Fig. 5). Sequence identity plots of the six Polystachya species were generated, with the annotation of P. dendrobiiflora chloroplast genome as a reference (Fig. 6). LSC and SSC regions were more divergent than IRs regions. Whereas the coding regions were more conserved than the non-coding regions, the highly divergent non-coding regions among the six cp genomes appeared in the intergenic regions (IGS), such as trnS -GCU -trnG -GCC , atpH-atpI, petA-psbJ, trnN -GUU -rpl32, rpl32-trnL -UAG and psaC-rps15. The trnI -GAU intron was also relatively divergent. On the other hand, all the rRNA genes were highly conserved and were similar to other plants' cp genomes [31].   , with an average of 0.02158, which was 3.88-fold higher than that in coding genes.

Positive selection analysis
The ratio of non-synonymous (dN) to synonymous substitutions (dS), dN/dS (ω), has been widely used to evaluate the natural selection pressure and evolution rates of nucleotides in genes [45,46]. The ω ratio > 1 specifies positive selection (adaptive evolution), while ω ratio < 1 signifies negative selection (purifying evolution). We compared the ratio of non-synonymous (dN) and synonymous (dS) substitution for 68 shared protein-coding genes among six Polystachya species. Likelihood ratio tests (M7 vs. M8) supported the presence of positively selected codon sites (p < 0.05, p < 0.01) (Table S7). According to the M8 (beta & ω > 1) models, a total of thirteen genes are under significant positive selection in the Bayes Empirical Bayes (BEB) method, in which eight genes (atpH, clpP, psbA, rbcL, rpl14, rpoA, rps3 and ycf1) harbored one significant positive selection site, three genes (ccsA, matK and rpl16) possessed two significant positive selection sites, while the rpoC2 gene had five sites under positive selection ( Table 2,  Table S8). In addition, it was found that the ycf2 gene located in the IR region had the highest number of positive selection sites, including 13 significant positive selection sites and one extremely significant positive selection site.

Phylogenetic analyses
To investigate the phylogenetic position of Polystachya in the Orchidaceae family and the relationship among the Polystachya species, two datasets were extracted from 85 complete chloroplast genome sequences and used in this study. The phylogenetic trees generated from the two datasets were consistent except for the tribes of Vandeae, Cymbidieae and Epidendrea (Fig. 8, Table S10). In addition, the Maximum Likelihood (ML) and Bayesian Inference (BI) trees constructed from the same data set produce similar topologies. Therefore, phylogenetic trees constructed based on 79 CDS were selected for discussion in this study, and ML bootstrap (BS) and posterior probabilities (PP) values are given above branches (Fig. 8, Table S10). In all sampled species, Polystachya species formed a monophyletic clade with strong support ( Fig. 8 and Table S10, BS, PP =100%, 1.00, respectively), where P. dendrolliflora is located at the base of this clade (Fig. 8, Table S10). Furthermore, the six Polystachya species and other all species of tribe Vandeae clustered together forming a monophyletic clade with strongly supported values (100%, 1.00). Based on the current data, our results yield a relatively stable phylogenetic relationship for most of the tribes and subtribes in Orchidaceae, with the exception of the tribes of Vandeae, Epidendrea and Cymbidieae. Additionally, the phylogenetic relationships of Orchidaceae at the subfamily level are fairly stable, with (Apostasioideae (Vanilloideae (Cypripedioideae (Orchidoideae, Epidendroideae)))) (Fig. 8, Table S10).

Sequence variation
In this study, we collected six species of Polystachya and obtained their complete chloroplast genome sequences. In most angiosperm, the plastome is generally maternal inherited, with little recombination, and highly conserved structure [32,47]. We found that the structure, gene content and gene order of the six Polystachya species were also highly conserved, and ranged in size from 145,484 bp (P. tenuissima) to 149,274 bp (P. dendrolliflora), containing 68 PCGs, four rRNA genes, and 30 tRNA genes ( Table 1, Fig. 1). The length of the LSC, SSC, and IRs varied in the range of 84,046-89,021 bp, 16,914-18,821 bp, and 23,902-25,914 bp, respectively. Chloroplast genome size variation among different species, or even within different individuals of the same species, which has been reported in other species, such as Camptotheca acuminata [48,49], Eucommia ulmoides [50], Rosa rugosa [51][52][53] and Calanthe davidii [54]. This, besides polyploidy of plant material, has been shown to be due to the expansion/contraction of IR [55]. The GC content in six Polystachya species was almost similar (36.9-37.0%). Although the Polystachya cp genomes were AT-rich, the higher GC content in the IR region is most likely due to the presence of rrn4.5, rrn5, rrn16, and rrn23 [56][57][58], which is consistent with the previously published Orchidaceae cp genomes [59,60]. Non-coding regions, especially introns may have accumulated mutations more rapidly than coding regions, hence having an influence at gene expression level [61]. In this study, 16 genes possess introns among the six Polystachya species, and the lengths of their exons were almost the same, whereas the lengths of introns varied in all these genes. These changes in length may affect the size of the cp genome and gene expression. Moreover, trnK -UUU gene containing the longest intron, and rps12 are considered as a trans-splicing gene (Table S2), this identical phenomenon was consistent with those of most other members of Orchidaceae species [54,62]. Inverted repeat (IR) expansion/contraction and sequence inversion represent important mechanisms of chloroplast genome rearrangement and show the diversity of cp genome structure in plants [63]. A detailed comparison of four IR/SC junctions of the six Polystachya plastomes showed that the border structures were highly similar to one another (Fig. 4). Although the boundary regions of the cp genome sequences of the Polystachya species were relatively stable, we found expansion in IR regions, where rpl22 gene in the LSC region was expanded by 23-66 bp to the IRB region. In addition, the ycf1 gene in P. adansoniae was located exclusively in the SSC region with 2 bp away from the SSC/IRA border, whereas the ycf1 gene in the SSC region expanded 5-359 bp into the IRA region in five other Polystachya species. The pseudogenization of ycf1 gene is common in dicots since the IR/SC boundary is within the ycf1 gene [26,64], but only a short fragment of ycf1 expanded into the IR region in this study, hence we did not annotate it as a pseudogene. An inversion of approximately 1 kb, within the IRB/SSC boundary region, was observed in the cp genome of P. modesta. However, the region corresponding to the inversion region was located in the SSC region in the remaining five species. This inversion caused the rpl32 gene of P. modesta to reverse direction.
As a result of the high of mutation, the SSR markers were widely used for studies of genetic diversity, population structure and species authentication [65][66][67]. Moreover, repeat sequences were also considered as one of the important reasons affecting gene duplication, expansion, and genome rearrangement [68,69]. A total of 58-73 SSRs and 11-37 long repeat sequences were identified, which were vastly distributed in the IGS region of LSC. The repeats in the coding regions were mainly located in the exons of ycf1, ycf2, accD, ccsA, clpP, rpoA and rpoC2. Most of the SSRs types are mononucleotide repeats, and A/T (no C/G) is the only mononucleotide SSRs type in the six Polystachya species. Previous research suggested that polyA and polyT have a more stable framework compared to polyC and polyG [70]. The large variation of long repeats in closely related species may reflect a certain degree of evolutionary flexibility [71]. The mVISTA percent identity plot and sliding window analysis showed that the most divergent regions were located in the trnS -GCU -trnG -GCC , atpH-atpI, petA-psbJ, trnI -GAU intron, trnN -GUU -rpl32, rpl32-trnL -UAG and psaC-rps15 regions in the Polystachya plastomes. A comparative analysis of the six Polystachya species revealed that IR regions were mostly conserved and non-coding regions were more highly divergent than coding regions.

NDH complex coding genes lost or pseudogenized
The chloroplast NAD(P)H-dehydrogenase-like (NDH) complex is located in the thylakoid membrane and plays an important role in mediating photosystem I cyclic electron transport (PSI-CET) and facilitating chlororespiration. A chloroplast genome usually contains a total of 113 genes, comprising 6 atp, 11 ndh, 6 pet, 9 rpl, 4 rpo, 12 rps, 4 rrn, 5 psa, 15 psb, 30 trn and 11 ungrouped genes. However, we found that 11 cp-ndh genes were lost or pseudogenized in the six Polystachya cp genomes (Fig. 9) and gradual pseudogenization, fragmentation and loss of ndh genes can be observed by comparative analyses of plastomes from the genus (Table S9). It is noteworthy that all ndh genes were also lost or pseudogenized in the plastomes of the sampled other species in the tribe Vandeae. Generally, ndh genes loss or pseudogenization is a common phenomenon in cp genomes of Orchidaceae, which is widespread in subfamilies Epidendroideae and Vanilloideae, and also occurs in subfamilies Cypripedioideae and Apostasioideae, but rare in the subfamily Orchidoideae. In addition, ndh genes are more frequently lost or pseudogenized in epiphytes than in terrestrial orchid plants [28,72]. Considering, moreover, that most orchid species in which ndh gene non-functionalization has occurred retain photosynthetic capacity, it is inferred that the function of this gene class may be affected by the nuclear or mitochondrial genomes. It has been suggested that the gene function may be maintained by RNA editing after pseudogenization, but the possibility cannot be accurately assessed for the time being because most pseudogenizations involve not only base changes but also indels [28]. In general, ndh genes are more easily lost in saprophytic and lithophytic orchids than in epiphytic and terrestrial orchids. Nevertheless, there was no significant difference in the loss and pseudogenization of ndh genes between the saprophytic Cymbidium macrorhizon and the other two epiphytic orchids in the genus Cymbidium. Additionally, all ndh genes lost in the epiphytic Oberonioides microtatantha, Corybas taliensis and lost or pseudogenized in Neottia cordata. The ndh genes of the epiphytic Agrostophyllum callosum and Pholidota articulata were normal. However, we could not accurately determine whether the lifestyle of orchids is directly related to ndh gene evolution. Some studies have suggested that ndh gene loss in mycoheterotrophic orchid lineages (including leafy and photosynthetic orchids) may have not been so disadvantageous for the lineages that live in low-light canopy habitats as epiphytes, or in dark, understory habitats [73][74][75]. Furthermore, the loss and pseudogenization of the ndh gene is prevalent in heterotrophic plants [76] and is also widespread in other autotrophic plants such as conifers [77][78][79][80], Gnetales [80], Circaeasteraceae [81], Corydalis (Papaveraceae) [82], and Erodium (Geraniaceae) [83], which perhaps indicates that the loss of ndh genes might be a random phenomenon in photosynthetic lineage. The loss of plastome genes may be due to the transfer into the nucleus, substitution of a nuclear encoded mitochondrial targeted gene or substitution of a nuclear gene for a plastid gene [84][85][86]. It is not clear whether the ndh genes in Polystachya have been transferred to the nucleus or whether their loss represents the complete loss of the NDH complex.

Divergent hotspots and adaptive evolution
The divergent regions as molecular markers could provide abundant valuable information for DNA barcoding and phylogenetic studies, and numerous phylogenetic reconstructions researches using divergent hotspots [66,87]. We found that the nucleotide sequence diversity of the non-coding regions was higher than that of the coding regions, which is generally consistent with most previous studies on the chloroplast genomes of angiosperms [88]. In our study, we compared interspecific chloroplast diversity in six Polystachya species, which indicated that IR regions were mostly conserved and non-coding regions in the cp genome had most of the variation compared to protein-coding regions. This finding is generally consistent with most previous studies of the cp genomes of angiosperms [83]. Among the seven IGS regions (rps19-psbA, trnS -GCU -trnG -GCC , trnG -GCC -trnR -UCU , atpH-atpI, psbT-psbN, rpoA-rps11 and trnL -UAG -ccsA) have high nucleotide diversity values (Pi > 0.05) which were most divergent regions identified (Fig. 7). No significant mutations existed in IR regions. Based on the calculated Pi values for each gene region, we have conducted a comprehensive comparative analysis of all published chloroplast genomic data in Polystachya, and observed that IR region had a lower nucleotide diversity than LSC and SSC regions, trnS -GCU -trnG -GCC with the highest variability values. These regions could potentially serve as universal candidate DNA barcodes for Polystachya species marker identification studies. It is noteworthy that the chosen divergent regions and the value of Pi are related to species selected, different collections were used, and different molecular markers will be selected out consequently. Thus, more suitable and accurate barcodes need to be further explored according to different sample categories. The dN/dS (ω) ratios have been widely used to infer the evolutionary dynamics and identify adaptive signatures among species [89]. It has been indicated that the variation rates of the chloroplast genomes can be influenced under different environmental pressures, which may be the main reason for the differences in the number of genes in the cp genomes of some genera or species [90]. In our study, a total of 33 positively selected sites (including 32 significant and 1 extremely significant sites) were detected based on the BEB method, which were distributed in atpH, ccsA, clpP, matK, psbA, rbcL, rpl14, rpl16, rpoA, rpoC2, rps3, ycf1 and ycf2 genes ( Table 2, Table S8). Of these loci, ycf2 gene has the largest number of significant positively selected sites, and although it is the longest plastid gene reported in angiosperms and its function is yet largely unknown [91]. Despite this, ycf2 provides a low-cost alternative to comprehensive multigene or genome datasets for investigating angiosperm relationships [92,93]. Positive selection has also been found in ycf2 in Bulbophyllum [94] and Malvaceae [95]. In addition, the genes ccsA, clpP, matK, rbcL, rpl14, rpl16, rpoA, rpoC2, ycf1 and ycf2 were identified in this study are also the targets of positive selection in many other flowering plants [95][96][97].

Phylogenetic relationship
The chloroplast genomes have been widely employed as important data to resolve lineages within species phylogenetic analysis. Especially, the whole cp genomes or protein-coding regions have been considered more effective and accurate than the use of single gene sequences or several fragments [98,99]. Consensus trees from BI and ML were almost similar, but the BI tree had higher resolution. Polystachya species formed a clade with strong support (BS, PP =100%, 1.00, respectively) based on the protein-coding sequences of cp genomes. The monophyly of the Polystachya has been similarly inferred in previous studies [6]. P. tenuissima (sect. Cultriformes) was sister to P. bennettiana (sect. Caulescentes), whereas P. modesta (sect. Polystachya) and P. concreta (sect. Polystachya) were sister taxa. These two clades are more closely related to each other than to P. adansoniae (sect. Polychaete). P. dendrolliflora represents the earliest extant lineage to diverge from the rest of the genus. Mytnik-Ejsmont (2011) proposed to divide P. dendrolliflora into an independent genus Dendrobianthe. Chase (2015) argued that there were no convincing arguments to split up the monophyletic and easily recognized genus Polystachya into smaller genera. Thus, Dendrobianthe dendrobiiflora is still treated as a synonym of P. dendrobiiflora (in TPL, POWO, IPNI). P. modesta is morphologically the most similar to pantropical tetraploid (including P. concreta) accessions and could be one of the parent species [12]. Morphological homology among species may be caused by introgression, plastid capture, hybridization and other factors, which makes it difficult for species delimitations in the genus Polystachya. Furthermore, the six Polystachya species and other all species of tribe Vandeae clustered together forming a monophyletic clade with strongly supported values (BS, PP = 100%, 1.00), which substantiated the definitive classification of this genus into Vandeae [13][14][15]. The phylogenetic relationships of all the sampled Orchidaceae species above subtribes are basically consistent with the results of Jin's (2019) study based on cpCDS, except that the phylogenetic positions among the tribes of Vandeae, Epidendrea and Cymbidieae are slightly different [17]. Although all phylogenetic relationships cannot be resolved by using only the complete cp genomes, our results suggest that plastome-wide analysis will provide higher a resolution for some disputed taxonomic relationships. In addition, the genus has a notorious taxonomic problem that has not been well resolved below the genus level [6,12,18]. Previous phylogenetic studies of this genus were mainly based on a few plastids and nuclear gene fragments, and although these results were able to roughly locate most taxonomic units, the phylogenetic relationships constructed by selecting different fragments were biased and many branches had low support. The chloroplast genome sequences with higher resolution provided by our study may provide a promising perspective for further elucidating the phylogeny and evolution within the Polystachya genus.

Conclusions
In this study, we obtained the complete chloroplast genomes of six Polystachya species (Polystachya adansoniae, P. bennettiana, P. dendrobiiflora, P. tenuissima, P. modesta and P. concreta) and revealed that the overall structure and gene content of the plastomes of the six species are relatively conserved, with only some variations in genome sizes, gene content, GC contents, introns, repeats and IR borders. It is interesting to note that all ndh genes were lost or pseudogenized in the plastomes of Polystachya, and were also observed in sampled other species in the tribe Vandeae. We also identified 13 positive selection genes and 16 variable regions, which provide a reference for developing DNA markers and adaptive evolution to further studies of Polystachya species. Phylogenetic analysis based on the current data identified the genus Polystachya in the tribe Vandeae and largely established the phylogenetic relationships of most taxonomic units in tribe and above tribe of Orchidaceae. These findings not only strengthen our understanding of the Polystachya plastomes, but also contribute to our further appreciation of the phylogenetic position of Polystachya in the family Orchidaceae.  Fig. 10). Young and fresh leaves were sampled and immediately preserved using silica gel [100]. The voucher specimens were deposited in the East African Herbarium (EA) at the National Museums of Kenya and the Herbarium of Wuhan Botanical Garden, CAS (HIB). Total genomic DNA was extracted using a modified cetyltrimethylammonium bromide (CTAB) method [101]. DNA integrity was examined by electrophoresis in 1% (w/v) agarose gel, their purity was determined using a NanoDrop spectrophotometer 2000 (Thermo Scientific; Waltham, MA, USA) at 260 and 280 nm, and precisely quantify DNA concentration with Qubit 2.0 (Life Technologies, CA, USA).

High throughput sequencing, genome assembly and annotation
The purified high-quality genomic DNA was used to construct paired-end (PE) libraries by shearing the genomic DNA into short fragments of approximately 350 bp before sequencing in 150 bp paired-end mode was implemented on an Illumina HiSeq 2500 platform (Illumina, Inc., San Diego, CA, USA) at Novogene Company (Tianjin, China). Genomes assembly were performed using GetOrganelle v1.7.1 [102] and CLC Genomic Workbench v10 (CLC Bio., Aarhus, Denmark) with the default parameters. The quality of the newly assembled genomes was evaluated on read level basis by aligning the trimmed raw reads to the de novo assemblies using Geneious mapper, Geneious v8.0.2 with medium-to low-sensitivity option and iteration up to five times [103,104]. The resulting complete chloroplast genomes were automatically annotated using the Perl script Plastid Genome Annotator (PGA) and the Annotation of Organellar Genomes (GeSeq) (https:// chlor obox. mpimp-golm. mpg. de/ geseq. html) [105,106]. Further annotation confirmation was performed with published genomes the cp genomes of the tribe Vandeae, Cymbidieae and Epidendreae sampled species in this study were used as the reference sequences. According to the annotation results of the two softwares, manual corrections of start/stop codons and intron/exon boundaries were performed in Geneious v8.0.2. All transfer RNA (tRNA) genes were proofread with the web server tRNAscan-SE v2.0 (http:// lowel ab. ucsc. edu/ tRNAs can-SE/) [107]. A gene was classified as a pseudogene if its reading frame was truncated (incl. Due to a premature stop codon) or frameshifted compared with Orchidaceae species [108]. Physical maps of the circular plastomes were visualized using the Organelle Genome DRAW (OGDRAW) (http:// ogdraw. mpimp-golm. mpg. de/) [109]. All annotated complete cp genome sequences were deposited into GenBank under the accession numbers listed in Table 1 and Table S1.

Comparison and divergence hotspot identification analysis
To investigate the structure difference exists, the expansion/contraction of the IR regions was assessed by comparing the positions of SC/IR junctions and their adjacent genes using IRscope [114], complete chloroplast genomes were aligned with MAFFT v7 [115]. The online mVISTA program (http:// genome. lbl. gov/ vista/ mvista/ submit. shtml) [116] was employed to compare the whole cp genome divergence within six Polystachya species in Shuffle-LAGAN mode, and with Polystachya dendrobiiflora as the reference. Genome rearrangement, inversions and gene synteny was detected using MAUVE [117] with default settings. All protein-coding genes of each genome sequence were extracted in Geneious v8.0.2 to examine the relative synonymous codon usage (RSCU) using MEGA v7.0 [118]. The codon usage frequency and relative synonymous codon usage (RSCU) of the six species were conducted based on 68 PCGs using MEGA v7.0. The nucleotide diversity (Pi) of each plastome was implemented to evaluate by DnaSP v6 [119].

Positive selection analysis
To detect the protein-coding genes under selection in the six Polystachya species, we calculate the nonsynonymous (dN) and synonymous (dS) substitution rates utilizing the CodeML algorithm implemented in EasyCodeML [120]. Each single-copy CDS was aligned separately using codon mode under the MAFFT v7 [115], followed by checking each single gene matrix and manually adjust abnormal codon alignments. Subsequently, these alignments were concatenated into a matrix using PhyloSuite v1.1.15 [121]. The ML tree was constructed using IQ-TREE [122] as an input tree. In preset mode, we used site model to test for natural selection, and eight codon substitution models described as M0 (one-ratio), M1a (nearly neutral), M2a (positive selection), M3 (discrete), M7 (beta), M8 (beta & ω > 1) and M8a (beta & ω = 1) were investigated. This model allowed the ω ratio to vary among sites with a fixed ω ratio in all branches in order to test for site-specific evolution in the gene phylogeny. The likelihood ratio test (LRT) was performed to detect positively selected sites with four pairs of site-specific models: M0 vs M3, M1a vs. M2a, M7 vs. M8, and M8a vs. M8. Comparing the four pairs of site-specific models, M7 vs. M8 was calculated to identify positive selection sites based on both ω and LRTs values. The Bayes empirical Bayes (BEB) method was used to statistically identify sites under positive selection with posterior probabilities ≥0.95 [123].

Phylogenetic analysis
To ascertain the phylogenetic position of Polystachya within Orchidaceae, a total of 85 species were analyzed, of which six Polystachya species were newly sequenced in this study and other 77 cp genome sequences of Orchidaceae species downloaded from NCBI database. Moreover, these sampled Orchidaceae materials represented 5/5 subfamilies, 16/22 tribes and 20/49 subtribes of Orchidaceae. According to previous studies [17,28], Allium cepa (MK335926, Allioideae) and Iris gatesii (KM014691, Iridoideae) were chosen as outgroups. GenBank accession numbers and detailed information of all samples used in this study are listed in Table S1. Each protein-coding sequence (CDS) matrix alignment was performed using MAFFT v7 plugin integrated into PhyloSuite v1.1.15 [121]. An incongruence length difference (ILD) test was performed in PAUP v4.0a168 [124] to determine whether data from different genes can be combined, and this test indicated that these datas are not homogeneous (P < 0.01). All alignments were eventually concatenated into one supermatrix utilizing PhyloSuite v1.1.15 [121]. Substitutional saturation was assessed using DAMBE v 7.0.68 [125] for the concatenated matrix. Subsequently, the 79 CDSs and the first and second codon positions of 79 CDSs were used, and the phylogenetic trees were constructed by Maximum Likelihood (ML) and Bayesian Inference (BI) algorithms. ML phylogenies were conducted using RAxML v8.2.12 [126] with 1000 bootstrap replicates and the GTRGAMMA model. BI phylogenies were inferred using MrBayes 3.2.6 [127] with the best-fitting substitution model by ModelFinder [128]. The Markov Chain Monte Carlo (MCMC) algorithm was run for 2,000,000 generations, sampling every 1000 generations, in which the initial 25% of sampled data are discarded as burn-in. Samples were combined and convergence of chains was checked in Tracer v1.7.2 [129]. Figtree v1.4.4 (http:// tree. bio. ed. ac. uk/ softw are/ figtr ee/) was used to visualize and annotate trees.