The complete chloroplast genomes of Tetrastigma hemsleyanum (Vitaceae) from different regions of China: molecular structure, comparative analysis and development of DNA barcodes for its geographical origin discrimination

Tetrastigma hemsleyanum is a valuable traditional Chinese medicinal plant widely distributed in the subtropical areas of China. It belongs to the Cayratieae tribe, family Vitaceae, and exhibited significant anti-tumor and anti-inflammatory activities. However, obvious differences were observed on the quality of T. hemsleyanum root from different regions, requiring the discrimination strategy for the geographical origins. This study characterized five complete chloroplast (cp) genomes of T. hemsleynum samples from different regions, and conducted a comparative analysis with other representing species from family Vitaceae to reveal the structural variations, informative markers and phylogenetic relationships. The sequenced cp genomes of T. hemsleyanum exhibited a conserved quadripartite structure with full length ranging from 160,124 bp of Jiangxi Province to 160,618 bp of Zhejiang Province. We identified 112 unique genes (80 protein-coding, 28 tRNA and 4 rRNA genes) in the cp genomes of T. hemsleyanum with highly similar gene order, content and structure. The IR contraction/expansion events occurred on the junctions of ycf1, rps19 and rpl2 genes with different degrees, causing the differences of genome sizes in T. hemsleyanum and Vitaceae plants. The number of SSR markers discovered in T. hemsleyanum was 56–57, exhibiting multiple differences among the five geographic groups. Phylogenetic analysis based on conserved cp genome proteins strongly grouped the five T. hemsleyanum species into one clade, showing a sister relationship with T. planicaule. Comparative analysis of the cp genomes from T. hemsleyanum and Vitaceae revealed five highly variable spacers, including 4 intergenic regions and one protein-coding gene (ycf1). Furthermore, five mutational hotspots were observed among T. hemsleyanum cp genomes from different regions, providing data for designing DNA barcodes trnL and trnN. The combination of molecular markers of trnL and trnN clustered the T. hemsleyanum samples from different regions into four groups, thus successfully separating specimens of Sichuan and Zhejiang from other areas. Our study obtained the chloroplast genomes of T. hemsleyanum from different regions, and provided a potential molecular tracing tool for determining the geographical origins of T. hemsleyanum, as well as important insights into the molecular identification approach and and phylogeny in Tetrastigma genus and Vitaceae family.


Introduction
Tetrastigma hemsleyanum Diels et Gilg (T. hemsleyanum) is a unique and valuable Chinese medicinal herb belonging to the tribe cayratieae of family Vitaceae. It is mainly distributed around the central, eastern, southern and southwestern provinces of China [1]. As a customary Chinese medicine, T. hemsleyanum has been recorded in the Zhejiang Provincial Standards of Processing Chinese Crud Drugs (2015) with the therapeutic effects of heat-clearing, toxicity-removing, promoting blood circulation and pain relief. The roots of T. hemsleyanum have been used traditionally to treat high fever, pneumonia, hepatitis, and multiple types of cancers alone or in combination with other herbal medicines [2]. Recent pharmacological investigations have demonstrated that T. hemsleyanum possesses antiinflammatory, anti-virus, anti-tumor and immunomodulatory effects, which could be attributed to the active components of flavonoids, polysaccharides, terpenoids and alkaloids [3]. Total flavones of T. hemsleyanum could inhibit the proliferation and induce apoptosis of breast cancer cells MDA-MB-468 and MCF-7 by inhibiting the expression of p-p42/44 and blocking MAPK signaling pathway [4]. In addition, dietary flavone from T. hemsleyanum vine was also found to trigger human lung adenocarcinoma apoptosis via autophagy [5]. The extracts from T. hemsleyanum leaves exhibited protective effects against acrylamide induced toxicity both in HepG2 cells and Caenorhabditis elegans via regulating DAF-16/FOXO signaling pathway [6]. Furthermore, the polysaccharide isolated from T. hemsleyanum enhanced the immune responses in both OVA-immunized C57BL/6 mice and Lewis lung carcinoma bearing mice through activating TLR4 signaling pathway [7]. The phenolic contents of T. hemsleyanum leaves exhibited significant antioxidant activities, indicating the possibility of developing T. hemsleyanum leaves as functional foods [8]. Moreover, the herbal formula Hua Shi Xuan Fei mixture mainly composed of T. hemsleyanum has been approved by Zhejiang Provincial Drug Administration for clinical treatment of COVID-19, which played important roles in fighting the epidemic [9]. The high medicinal and economic value of T. hemsleyanum makes its further genetic and phylogenetic investigation necessary.
The roots of T. hemsleyanum were one of the most popular Chinese folk medicines in herbal market with the common name Sanyeqing. The genus Tetrastigma contained more than 100 species of climbers basically distributed in the tropics and subtropics of Asia, and a few species extending to Australia [10]. The Tetrastigma climbers have also attracted a ton of attention due to its unique host-parasite association with Rafflesiaceae, which possess the largest flowers in the world [11]. The great number of Tetrastigma species provided rich resources of root tubers for the potential counterfeits for T. hemsleyanum. Due to the lack of typical identification characteristics, it is difficult to distinguish roots of T. hemsleyanum from those adulterants of climbers in the genus of Tetrastigma and other closely related species. The increased number of counterfeit products and substitutes of T. hemsleyanum in the herbal medicine market have seriously harmed the clinical safety and effectiveness of T. hemsleyanum and significantly threatened its healthy development [12]. For instance, the root tubers of toxic Aconitum carmichaeli have been often sold as adulterants of T. hemsleyanum in medicine markets owing to high profit in Zhejiang Province, which caused many serious poisoning incidents. Furthermore, significant differences have been indicated in chemical compositions and therapeutic values of medicinal plants from different regions, which could be attributed to genetic varieties and growth environment [13]. T. hemsleyanum was widely distributed in the tropical to subtropical areas of China with multiple varieties and large gaps in the yield and quality of root tubers. The molecular investigations of T. hemsleyanum based on ISSR and SRAP analysis revealed the high genetic diversity at the level of species and low diversity in populations [14]. The T. hemsleyanum samples from Zhejiang and Fujian Province exhibited the highest contents of total flavonoids and strongest inhibitory activities against HepG2 cells, further confirming the crucial role of the geographical origins [15]. Numerous different expressed transcripts and multiple differentially accumulated metabolites mainly involved in phenylpropane and flavonoid biosynthesis have been found between two ecotypes of T. hemsleyanum from different regions [16]. In addition, the starch characterizations of T. hemsleyanum tuber roots from different origins showed significant differences on the granule diameter, amylose content and peak gelatinization temperature, from Zhejiang and Guangxi province [17]. Although the unique growth environment of different regions has great and profound influence on species, the diversity of T. hemsleyanum germplasm resources is beyond doubt, and their interaction determines the quality of T. hemsleyanum. The relationship of geographical difference and genetic difference of germplasm is of great significance. Therefore, authenticity and traceability of geographical origin of T. hemsleyanum were imperative and imminent for its quality and medicinal values to prevent mislabeling. The rapid and accurate discrimination of geographical origin of root from T. hemsleyanum would be conducive to the breeding and brand establishment of superior varieties. The growing demand of identifying the provenances of decoction pieces of T. hemsleyanum required effective approaches for geographical origin determination, which would significantly contribute to the further development and clinical applications of T. hemsleyanum.
The chloroplast (cp), which performs photosynthesis as well as harbors lots of other metabolic pathways, is an important organelle in plants and is generally nonrecombinant and uniparentally inherited [18,19]. In most angiosperms, chloroplast genome is generally double stranded and circular and the size varies from 120 kb to 160 kb [20]. The structure of chloroplast genome is highly conservated, consisting of a small single copy region (SSC), a large single copy region (LSC) and a pair of inverted repeat regions (IR) [21]. The cp genome harbors many different gene loci and non-coding regions containing relatively large amount of DNA sequence information, which has been widely considered as a powerful tool for phylogenetic analysis and further development of species identification and restoration strategies [22]. A great number of cp genomes have been determined and applied in phylogenetic reconstructions from population genetics to investigate the sequence evolution and perform deep divergence analysis at a genera and family level. Comparative chloroplast genome analysis of six Impatiens species reconstructed the taxonomic relationship and provided detailed information about nucleotide diversity hotspots, which could facilitate the systematic evolution research of the Balsaminaceae species [23]. Furthermore, various chloroplast genome regions and DNA barcodes have been considered as useful molecular markers in systematic and population genetic studies to distinguish its closely related species and adulterants [24]. For instance, four chloroplast single nucleotide polymorphism (SNP) variants were identified as powerful markers to differentiate rubber dandelion species from weedy relatives [25]. The DNA barcodes rps16 and trnQ-UUG were designed based on the comparative analysis of complete chloroplast genome sequences of Conyza bonariensis, which could successfully separate three predominant Conyza species [26]. More importantly, DNA barcodes from chloroplast genome exhibited the potential as a powerful analytical tool for determination of geographical origins. The haplotype map generated by four chloroplast DNA markers identified 10 informative intra specifically variable sites, providing useful molecular tool for tracking Merbau timber originating from peninsular Malaysia [27]. The cp DNA marker trnL-F was successfully applied to investigate phylogeographical pattern of 27 populations of Chrysanthemum indicum across the southwest to northwest areas in China, which revealed fifteen haplotypes and correlative high genetic differentiation among populations [28]. The DNA barcodes from chloroplast genomes would be promising approach to distinguish the geographical origins of T. hemsleyanum, as well as rapidly differentiate the genuine and adulterated crude drugs. Therefore, it is highly important and essential to investigate the complete chloroplast genomes of T. hemsleyanum from different regions and develop typical molecular markers to discriminate their provenances for its further clinical application and development.
In the current study, we have sequenced and assembled the cp genomes of T. hemsleyanum collected from five different regions using Illumina sequencing platform. These sequences were further compared with other known chloroplast genomes from Vitaceae species to reveal the conserved and different features on basic genome structure, codon usage bias, repetitive structure characteristics and IR expansion. The phylogenetic relationship between Tetrastigma species and other closely related taxa from Vitaceae were reconstructed to infer the taxonomic status of Tetrastigma climbers within the families. Finally, we designed two specific DNA barcodes to identify the geographical origins of T. hemsleyanum, which successfully divided the T. hemsleyanum climbers from different regions into four clades. The genomic and marker resources described here expanded our understanding of the diversity of chloroplast genomes and their taxonomic relationships within Vitaceae species, and provided an efficient molecular approach for the geographical origin identification of T. hemsleyanum.

Morphological features and chloroplast genome characteristics of Tetrastigma hemsleyanum
The samples of Tetrastigma hemsleyanum were collected from Zhejiang, Fujian, Jiangxi, Guangxi and Sichuan provinces, which have been indicated as the main producing zones of T. hemsleyanum crude drugs (Fig. 1A). The external morphology and microstructure of the root tubers of T. hemsleyanum from different regions were analyzed to identify the differences on morphological features. As shown in Fig. 1B, all of the root tubers from different regions exhibited similar morphological characters, including the root tuber size, the shape of elliptical or spindle, and the epidermis with tan. Moreover, most of root tubers of T. hemsleyanum showed smooth appearance, while a few of them presented folds and lenticel-like protuberances, as well as depressions (Fig. 1B). In addition, the microscopic features of the powder revealed that the cork cells, brown patches, needle crystals of calcium oxalate, starch granules and marginal orifice catheters were abundant in root tubers of T. hemsleyanum from five different regions, while cluster crystals of calcium oxalate were rarely observed in the tubers (Fig. 1C). However, the pharmacognostical analysis failed to identify significant differences on external morphology and microstructure of root tubers of T. hemsleyanum from different regions, requiring the development of alternative strategies to discriminate its geographical origin. Therefore, we sequenced the complete cp genome of T. hemsleyanum from five different regions and conducted a comparative analysis to establish and develop Fig. 1 The geographical distributions and morphological characteristics of Tetrastigma hemsleyanum from different regions. A Regional distribution map of Tetrastigma hemsleyanum sample collection (B) The microscopic characters of the dried roots from Tetrastigma hemsleyanum: 1. brown patches; 2. marginal orifice catheter; 3. cluster crystals of calcium oxalate; 4. needle crystals of calcium oxalate; 5. starch granules; 6. cork cell. C Comparison of macroscopic characters of the root tubers of Tetrastigma hemsleyanum from five different regions potential molecular approach for geographical origins traceability of T. hemsleyanum.
The complete chloroplast sequences of T. hemsleyanum from five different regions of Zhejiang, Fujian, Guangxi, Sichuan and Jiangxi Provinces, have been deposited in the GenBank database with the accession No. of MW375707 ~ MW375711. The size of the whole chloroplast genomes of T. hemsleyanum varied from 160,124 bp to 160,518 bp, with the smallest and largest T. hemsleyanum cp genome from Jiangxi and Zhejiang Province, respectively. All of the five T. hemsleyanum cp genomes exhibited a typical angiosperm circular chloroplast structure containing four regions: large single-copy region (LSC; 88,131 bp-89,298 bp), small single-copy region (SSC; 18,962 bp-18,965 bp), and a pair of inverted repeats (IR; 26,126 bp-26,517 bp) (Fig. 2). A total of 112 genes, including 80 protein-coding genes, 28 tRNAs, and 4 rRNAs were identified from each genome of T. hemsleyanum from different regions ( Table 1). The cp genomes showed high similarity in terms of gene contents, orders and orientations. Specifically, the overall GC contents of T. hemsleyanum from five regions revealed almost the same results in five regions, among which medicinal plant from Zhejiang, Fujian and Guangxi exhibited a GC content of 37.50%, while that from other two regions showed a result of 37.52% (Table 2). No significant differences on protein coding genes were identified in the T. hemsleyanum cp genomes from different regions, with a total length of 80,022 bp. There were 18 duplicated genes identified in the IR regions of T. hemsleyanum cp genome including 8 protein coding genes (rpl2, rpl23, ycf1, ycf2, ycf15, ndhB, rps12 and rps7), 7 tRNA genes (trnA-UGC , trnI-CAU , trnI-GAU , trnL-CAA , trnN-GUU , trnR-ACG and trnV-GAC ) and 4 rRNA genes (rrn4.5, rrn5, rrn16, rrn23). (Table 2). Furthermore, 18 distinct genes were indicated as intron-containing genes in the cp genome of T. hemsleyanum, including 13 protein coding genes and 5 tRNA genes. All these genes exhibited a single intron, except for rps12, clpP and ycf3 which contained two introns. Moreover, it is intriguing that the location and the intron aera of rpl2 gene were diverse in T. hemsleyanum cp genomes from different genomes. The rpl2 gene of Guanxi and Zhejiang T. hemsleyanum cp genomes possessed two introns and across the junction of IRA and LSC region, which occupied in LSC region with 149 bp and 223 bp respectively. While the rpl2 gene of T. hemsleyanum cp genome from other three regions showed only one intron and located in IRA completely. The above results indicated the cp genomes of T. hemsleyanum from different regions were slightly different, but it is highly conserved on basic structure, genome size, gene number and total GC content.
To further determine the conserved and variable structures of cp genome in family Vitaceae, we conducted a comparative analysis between T. hemsleyanum plant and other species from tribe Cayratieae (Tetrastigma planicaule), tribe Ampelopsideae (Ampelopsis japonica) and tribe Viteae (Vitis vinifera). The structure of the chloroplast genome appeared to be largely conserved across the family Vitaceae, with little differences on the total genome length, gene number and GC content ( Table 2). The size of the chloroplast genome varied from 160,323 bp in T. planicaule to 161,430 bp in A. japonica, and the overall GC content ranged from 37.32% (A. japonica) to 37.49% (T. planicaule). However, the types and numbers of genes coded in the cp genomes of T. planicaule, A. japonica and V. vinifera were not identical with that of T. hemsleyanum. The lack of ycf15 gene resulted in a decrease in the number of protein coding genes of T. planicaule and V. vinifera, while the protein coding gene number of A. japonica was consistent with that of T. hemsleyanum. In addition, the ycf1 gene of V. vinifera completely located in IRB region and was indicated as a pseudogene copy. Compared with the protein coding genes, more significant differences were identified on the the tRNA genes among the cp genomes from the four Vitaceae plants.
The trnS-GCU gene was uniquely encoded by T. hemsleyanum, while trnG-GCC , trnG-UCC , trnV-GAU were solely encoded by other three Vitaceae species, which led to the uniqueness of T. hemsleyanum.

Comparative analyses of chloroplast genome
As a link between the nucleic acids and proteins, the genetic code plays an important role in the transmission of genetic information in organisms [29]. Therefore, we analyzed the codon distribution among the protein coding genes in cp genome of T. hemsleyanum from different regions and performed a comparison analysis. The cp genomes of T. hemsleyanum from five regions exhibited almost identical protein-coding sequences, which represented a total of 26,674 codons. All of these codons belonged to 64 codon types and encoded 20 amino acids ( Supplementary Fig. 1). However, the numbers of amino acid and the bias of codon usage of T. hemsleyanum cp genomes from different regions exhibited a slight disparity. Leucine was the most abundant amino acid (2774 ~ 2776 codons, 10.40% ~ 10.41% of the total), whereas Cysteine (320 ~ 322 codons, 1.20% ~ 1.21% of the total) showed the least abundance in the cp genome of T. hemsleyanum. Regardless of stop codons, the most commonly applied codon was AUU (1117 ~ 1118), encoding isoleucine and the least one was UGC (89 ~ 91), encoding cysteine (Supplementary Table 3). The single most striking observation to emerge from the data comparison in Supplementary Table 3 was that the codon usage patterns of T. hemsleyanum from five different regions could be divided into three types. According to the data in Supplementary Table 3, the codon usage bias of T. hemsleyanum from Jiangxi and Sichuan were exactly the same, while those from Fujian and Guangxi completely displayed the same bias, and the special one from Zhejiang exhibited a unique pattern of codon usage differing from the other regions. Further comparative analysis revealed that a total of 28 variants sites in 21 protein-coding genes of T. hemsleyanum from different regions, which  Table 4). What stands out in the Supplementary Table 4 was that, the protein coding genes of atpB, ccsA, ycf2 and ycf1 exhibited two variable sites while accD gene displayed 3 mutation sites. The more surprising correlation was the variant sites of accD gene in T. hemsleyanum from Jiangxi and Sichuan provinces resulted in the encoding of lysine, which was obviously distinguished with methionine, glutarnine and asparagine encoded in cp genome from other three regions. Comparing with T. hemsleyanum of Sichuan, Jiangxi and Zhejiang regions, one base variation was identified in the ndhD and ycf2 genes of that from Fujian and Guangxi regions, which led to the preference of GGG and GGC, respectively (Supplementary Table 4). Taken together, these results provided important insights into the understanding of protein adaptive evolution and strategy development of identification geographical origin of T. hemsleyanum.
Previous reports have indicated that codon usage bias of chloroplast genome may be affected by selection, mutation and random drift [30,31]. Further comparing  , indicating an obvious codon bias in the amino acids. Most (29 codons) of these preferred codons among eight Vitaceae plants species ended with the nucleotide of A or U. Therefore, the investigation on codon preferences is conducive to understand the exogenous gene expression and the molecular evolution mechanisms of T. hemsleyanum in Vitaceae. Contraction and expansion of the IR region is a common phenomenon known as ebb and flow, which could be used as effective tool for phylogenetic relationship and classification research of medicinal plants [32]. A comparison of five T. hemsleyanum plants and three Vitaceae species for borders was performed between the IRs and two single copies regions in detail. The length of the IR regions was similar among the eight Vitaceae species ranging from 26,126 bp in T. hemsleyanum (Zhejiang) to 26,523 bp in T. planicaule, with certain expansion and contraction (Fig. 3). Particularly, some notable differences were found at the boundaries among cp genomes of T. hemsleyanum from different regions. The LSC-IRb border was located within the rps19 gene in T.  (Fig. 3). However, the rps19 gene was completely encoded in LSC region and exhibited 290 and 216 bp distance to the junction of the LSC-IRB region in the cp genome of T. hemsleyanum from Zhejiang and Guangxi, respectively (Fig. 3). In contrast, the locating position of ycf1 gene was highly conserved at the boundary of IR/SSC region among Vitaceae plants except that of V. vinifera, which exhibited a pseudogene gene with the size of 1030 bp completely locating in IRB region (Fig. 3). The contraction and expansion of ycf1 gene at the boundary of LSC-IRB within five T. hemsleyanum plants were identical, which occupied 1140 bp and 29 bp in IRB and SSC regions, respectively. Another interesting observation is that the overlap between ycf1 gene and the IRA region was significantly longer of other three species than that of T. hemsleyanum, which showed 1144 bp for T. planicaule and 1116 bp for A. japonica and V. vinifera, respectively. However, the overlap length was only 34 bp in the cp genome of T. hemsleyanum, which could be considered to be one of the reasons for the length change among these Vitaceae cp genomes.

Repeat sequences analysis and RNA editing sites identification
Long repeats are significant genetic resources, playing a crucial role in genome rearrangement and intermolecular recombination [33]. As shown in Supplementary Fig. 2, the long repeat sequences detected in T. hemsleyanum cp genomes of Jiangxi and Sichuan revealed the identical results, while specimens from Zhejiang, Fuajin and Guangxi regions exhibited slightly different types and number of repeat sequences. Within the five T. hemsleyanum plants, the long repeats analysis revealed the most abundant repeats were length of 30-39, with the largest number in type of forward repeats (27)(28), followed by palindrome (18)(19), complemented (2) and reverse (2) repeats. These results further confirmed the high similarities on the type of repeats and certain slight variations on terms of the number and length among cp genomes of T. hemsleyanum from different regions (Supplementary Fig. 2A). However, the comparison of T. hemsleyanum with the other three Vitaceae species displayed an obvious distinction. A total of 49, 48, 40 long repeats were identified in the cp genomes of T. planicaule, A. japonica and V. vinifera respectively. In contrast with the cp genome of T. hemsleyanum, no complemented repeats were determined in the cp genome of other three Vitaceae plants. In addition, the type of reverse repeats was also lost in the cp genome of T. planicaule. Among these Vitaceae plants, most of the repeat units were short, ranging from 30 to 59 bp ( Supplementary Fig. 2B).
Simple sequence repeats (SSRs) play an essential role in plant taxonomy and population genetics studies for the high polymorphism and codominance [34]. In total, 56 SSRs were identified in the cp genomes of T. hemsleyanum plants from four regions, while the species from Guangxi exhibited a SSR number of 57. The majority of SSR sequences were mononucleotide repeats (42)(43), followed by dinucleotides (11) and tetranucleotides (3) ( Table 3). The cp genome of T. hemsleyanum of Jiangxi and Sichuan exhibited the identical results on SSR types and numbers. However, the distinctions of SSRs in T. hemsleyanum cp genomes from the other regions were embodied in SSRs count of mononucleotide repeats (Fig. 4A). Specifically, the numbers of A/T repeats in the cp genomes of T. hemsleyanum plants from Jiangxi, Zhejiang, Fujian, Guangxi and Sichuan were 42, 41, 41, 42 and 42, respectively. In addition, the samples from Jiangxi and Sichuan showed no C/G SSR repeats in the cp genomes. These results further indicated that SSR might be a useful molecular marker for species determination of geographical origins of T. hemsleyanum. In addition, a comparative SSRs analysis conducted with three Vitaceae species revealed 55, 69 and 54 SSRs were detected in the cp genomes of T. planicaule, A. japonica and V. vinifera, respectively (Table 3). It is must mentioned that T. planicaule from Tetrastigma genus showed identical SSRs types with slight distinctions on SSR quantities ( vinifera also revealed the discrepancy of SSR loci among different genus, which might also provide a basis for the identification of the Tetrastigma genus. Among all Vitaceae species, the number of SSRs composed by A/T were significantly greater than that containing G or C, indicating that the base composition of SSRs was biased toward the application of A/T bases, which was consistent with A-T enrichment in complete chloroplast genomes [35]. Taken together, these results provided important insights into understanding intrageneric and intergeneric variations within T. hemsleyanum and its relatives in Vitaceae species. The RNA editing process is an essential maturation mechanism to avoid incorrect RNA mutations and is widespread in the chloroplast genome of plants [36]. In total, 71 potential RNA editing sites have been predicted in 24 protein-coding genes of the cp genome of T. hemsleyanum, which displayed no distinction in numbers of RNA editing sites and conversions of amino acids in cp genome of T. hemsleyanum from different regions. (Table 4). Among the 71 RNA editing sites, 17 codons were observed to be edited at the first nucleotide position, whereas 54 codons were identified to be edited at the second nucleotide position, and no codons were edited at both of the first and second nucleotide. All of the identified codon changes in the cp genomes of T. hemsleyanum showed C to T conversions. Especially, the ndhB gene showed the largest number of RNA editing sites (11 editing sites), followed by ndhD (8 editing sites) and ndhF (7 editing sites), while nine genes (accD, atpI, atpF, ccsA, clpP, psbE, psbF, psbL and rpl20) exhibited only one editing site in T. hemsleyanum ( Table 4). The RNA edition on protein gene resulted in a total of 11 kinds of amino acid conversions in the cp genome of T. hemsleyanum. The conversions of H to Y, L to F, P to S, R to W, R to C were due to codons edited at the first nucleotide position, while the S to L, P to L, S to F, T to M, A to V, T to I conversions were caused by codons edited at the second nucleotide position (Supplementary Table 5). The conversion of serine to leucine (S to L) was the most abundant kind of conversion, accounting for 42.3%, while arginine to tryptophan (R to W) and arginine to cysteine (R to C) were the least conversion, accounting for 1.4% merely (Supplementary Table 5). Furthermore, the predicted RNA editing sites in the cp genomes of T. planicaule, A. japonica and V. vinifera showed similar results with that of T. hemsleyanum, with the RNA editing sites number of 71, 72 and 70 respectively. The slight difference of the number of RNA editing sites were observed in accD, ndhB and ndhF genes among these Vitaceae plants, which led to the distinctions of amino acid conversions (Supplementary Table 5). Since the close correlation between RNA editing sites and nucleotide substitution of protein coding genes, we performed further analysis to investigate the synonymous substitutions (Ks) and nonsynonymous substitutions (Ka) of protein coding genes with abundant RNA editing sites. The Ka/Ks ratios of most genes (22/24) in T. hemsleyanum were less than 0.5 expect the matK (0.5534) and rps16 (0.5687), suggesting an obvious purifying selection pattern. Particularly, the clpP, psbL and psbF genes even exhibited a Ka/Ks value of 0, showing the three genes were possibly under strong purifying selection pressure (Table 5).

Phylogenetic analysis
The previous reports by molecular and morphological data indicated the family of Vitaceae could be classified into five major clades, including the tribe of Ampelopsideae, Cisseae, Cayratieae, Parthenocisseae, and Viteae [37]. However, the deep phylogenetic relationship of Vitaceae still needs further explorations to reveal the evolutionary characters and genetic status of grape species. Therefore, we constructed phylogenetic tree of Viteae family based on the 70 protein-coding gene datasets by maximum likelihood (ML) and maximum parsimony (MP) method, respectively. These grape plants contained 4 species from tribe Viteae, 3 species Table 3 The number and types of SSR in five T. hemsleyanum plants and three Vitaceae species from tribe Ampelopsideae and 6 plants from tribe Cayratieae. Melaleuca alternifolia and M. cajuputi were chosen as the outgroups for phylogenetic analysis. As shown in Fig. 5, nearly all of the nodes received moderate to high support values in the ML and MP tree analysis. However, several topological differences have been occurred in relationships within the five T. hemsleyanum species and the tribe of Viteae between the ML and MP tree results (Fig. 5). The phylogenetic analysis among the five T. hemsleyanum plants revealed a stable monophyletic group with high bootstrap values, which exhibited a stable sister relationship with T. planicaule, indicating a close genetic relationship within the genus of Tetrastigma (Fig. 5). In addition, the ML analysis indicated that the samples of T. hemsleyanum from Fujian and Guangxi regions clustered together to form a combined group with a bootstrap score of 68, which subsequently gathered together with T. hemsleyanum species from other three regions (Fig. 5A). These results indicated certain subtle protein coding differences of T. hemsleyanum cp genomes from different regions, providing potential molecular tools for distinguishing the geographical origins of T. hemsleyanum. Furthermore, the Ampelopsideae species and Viteae plants combined together to form a clade with strong statistical support, which combined with six Tetrastigma plants to form a robust monophyletic clade, which was consist with the previous classification of the tribes of Ampelopsideae, Viteae and Cayratieae in Vitaceae.

Nucleotide diversity analysis and development of molecular marker for geographical origin discrimination
The complete cp genomes with high variable levels provides potential molecular marker for species identification and geographical origin determination. In order to assess the sequences divergence level within the Vitaceae species, the complete cp genomes have been multiple aligned and applied DnaSP to calculate nucleotide variability (Pi). As shown in Fig. 6A, the sliding window analysis revealed 5 highly variable regions with Pi values ranging from 0.06194 to 0.10611 across four complete cp genomes of Vitaceae species, including 4 intergenic regions (rps16-trnQ, psbM-trnD, psbZ-trnfM and ycf3-trnS) and one protein coding genes (ycf1) (Fig. 6A). Among the five mutational hotspot loci, four highly variable hotspots were located in the LSC region, and ycf1 gene with the Pi value of 0.06194 was identified in the SSC region. However, none of the hypervariable loci were determined in the IR region, further confirmed that the IR regions were highly conserved in the cp genomes among the Vitaceae species. The rps16-trnQ gene exhibited the highest Pi value of 0.10611, followed by psbZ-trnfM and ycf3-trnS with the Pi values of 0.10083 and 0.10056, respectively. Besides, a comparative analysis was carried out to determine the numbers of SNP sites and Gaps to further explore the characteristics of five hypervariable regions among four Vitaceae plants. The five mutational hotspots in the cp genome of T. hemsleyanum from Zhejiang province ranged from 892 bp (psbZ-trnfM) to 1139 bp (rps16-trnQ) in length (Table 6). Apparently, the high variable sequences of T. planicaule from Tetrastigma genus exhibited a small number of SNP sites (3)(4)(5)(6)(7)(8) and Gaps (0-19) than that of T. hemsleyanum from Jiangxi Province except psbM-trnD region, which contained the SNP site and Gaps of 79 and 39 in the cp genome of T. planicaule, respectively. However, a great deal of variable sites was displayed in the 5 mutational hotspots of A. japonica and V. vinifera. For instance, the hypervariable regions of psbZ-trnfM showed 104 and 115 SNP sites in A. japonica and V. vinifera, respectively, which was significantly higher than that of T. hemsleyanum of Zhejiang. All these discrepancies led to variable mutational hotspot lengths in the Vitaceae plants eventually, and also provided potential molecular markers to resolve the difficulties in species identification of Vitaceae species.  To determine the potential of variable sequences in cp genome for geographical origin discrimination, we further evaluate the the sequences divergence level of T. hemsleyanum from different regions. The results demonstrated that the intraspecific differences of T. hemsleyanum was much lower than interspecific differences among Vitaceae species (Fig. 6). A total of 5 mutational hotspots with relative high Pi value (≥0.004) have been screened out in T. hemsleyanum plants, including 2 hypervariable regions (trnL-CAA and trnN-GUU ) located in IRs and one intergenic region located in SSC (ndhD-psaC) with the Pi value of 0.009 (Fig. 6B). Accordingly, these hypervariable regions with abundant intraspecific variable sites could be developed as potential DNA barcodes to discriminate the geographical origins of T. hemsleyanum. Interestingly, we found that both the SSC and IR regions were more variable than the LSC region in the chloroplast genomes of T. hemsleyanum from different regions. This result was significant different with the general observations in other species, where the IR regions usually exhibited lower variability than the LSC and SSC regions. One important reason was conjectured that the intraspecific variation among individuals was influenced by genes of NADH dehydrogenase mostly distributed in SSC region and transfer RNAs in IR regions, while the interspecific evolution among different species was driven by genes of Photosystem I/II, ATP synthase and Ribosomal proteins located in LSC. This result also indicated that more attention should be focused on the dissimilarity of SSC region in cp genomes for breeding of excellent species. Our study designed five DNA barcodes (accD, trnL-CAA , trnN-GUU , ndhD-psaC and ndhC-trnV) based on hypervariable regions for PCR amplification of T. hemsleyanum medicinal materials in the Zhejiang region (Fig. 7A). The single bright band in agarose gel electrophoresis implied amplification of accD, trnL-CAA and trnN-GUU , while the trnL-CAA and trnN-GUU showed higher PCR amplification efficiency and sequence diversity. As a result, the two DNA barcodes were amplified with DNA of T. hemsleyanum samples from six different regions in batches to further analyze the efficiency of geographical origin discrimination. The detailed sequence information of the two PCR products is shown in Table 7.  (Table 7). Notably, that T. hemsleyanum plants from Jiangxi, Zhejiang, Fujian, Guangxi and Guangdong regions harbored the identical trnL haplotype of A1, indicating the genetic variation of A1 was the main variety distributed in China due to its strong environmental adaptability. Additionally, the trnN sequences of T. hemsleyanum from different origins showed an identical GC content of 43.50%. However, these trnN sequences exhibited four variable bases at the position of 164 bp, 165 bp,166 bp and 167 bp, generating 2 haplotypes among different regions (Supplementary Table 6). Interestingly, the trnN sequence from Sichuan region showed a unique haplotype of B2, while that from other regions exhibited the same haplotype of B1 (Table 7). These results demonstrated that the intraspecies discrepancy of T. hemsleyanum plants      among different regions, further confirming the availability and necessity of geographical origin identification strategy based on molecular markers of chloroplast genome. This study also explored the genetic distance of intraspecific and interspecific variation within the trnL and trnN sequences of the T. hemsleyanum medicinal materials from different regions (Figs. 7B, C, D). The K2P distance of both trnL and trnN sequences among the 53 T. hemsleyanum samples ranged from 0.000 to 0.004, suggesting a significant barcoding gap among plants from different regions. For instance, the divergence value of trnL was highest (0.004) between the Jiangxi and Zhejiang regions and lowest (0.001) between the Jiangxi and Sichuan regions. The intraspecific genetic distance of trnN sequence between Jiangxi and the Sichuan region had a K2P value of 0.004, suggesting a barcoding gap. However, both of the two cp molecular markers failed to generate barcoding gap among species from Jiangxi, Fujian, Guangxi and Guandong province, indicating the inability of discriminating geographical origin from these regions by trnL and trnN. Moreover, the combination of trnL and trnN sequences revealed a lower intraspecific distance among different geographical origin of T. hemsleyanum than the single molecular marker (Fig. 7D). The intraspecific genetic distances based on trnL+ trnN sequences between Jiangxi and Zhejiang, Jiangxi and Sichuan, and among Jiangxi, Fujian, Guangxi, Guangdong and Zhejiang were 0.002, 0.003 and 0, respectively (Fig. 7D). The interspecific distance was greater among the Vitaceae species than that intraspecific distance of T. hemsleyanum samples, suggesting the developed DNA barcodes could be successfully applied for T. hemsleyanum species identification from other Vitaceae plants (Fig. 7D). The NJ tree analysis of the trnL and trnN barcodes revealed a clear distinction clearly among the different geographical origins of T. hemsleyanum plants (Figs. 7E, F). The trnL-based NJ tree generated four groups with different geographical origins, while trnNbased NJ tree only provided two clades of T. hemsleyanum plants. The Clade I of trnL-based NJ tree consisted of all T. hemsleyanum samples from Jiangxi, Fujian, Guangxi and Guangdong areas, seven samples from Zhejiang (ZT-01 ~ 03 and ZT-07 ~ 10) and one sample from Genbank (NCBI) with the bootstrap support value of 51. Clade II included T. hemsleyanum samples from Sichuan and T. planicaule samples. The rest of T. hemsleyanum samples from Zhejiang Province formed the Clade III (ZT-04 ~ 06) and Clade IV (ZT-11 ~ 15), respectively (Fig. 7E). The Clade II of the trnN-based NJ tree included all samples from Sichuan, while Clade I consisted of the samples from the other five regions and the sample (NCBI) from the Genbank. Although trnL barcode is more powerful in discriminating geographical origins of T. hemsleyanum than the trnN barcode, it failed to distinguish other Vitaceae species from T. hemsleyanum (Fig. 7E). Finally, we constructed the NJ tree based on the combination sequence of trnL+ trnN to determine the identification accuracy for T. hemsleyanum plants from different regions. Interestingly, this NJ tree divided the T. hemsleyanum plants into four groups and the dendrograms showed clear clustering pattern of geographically distribution. The six samples from Zhejiang province of ZT-04 to ZT-06，ZT-11 to ZT-15, and Sichuan samples of CT-01 to CT-03 each formed a separate group, while the other samples from Zhejiang province (ZT-01 ~ 03, ZT-07 ~ 10) and samples from other regions formed the fourth group (Fig. 8). Furthermore, the position of each Vitaceae species in the NJ tree based on a combination barcode was similar to phylogenetic trees based on the 70 protein-coding gene datasets (Figs. 5, 8). The NJ tree clearly showed that the Tetrastigma genus (T. planicaule and T. hemsleyanum samples) formed the main branch (bootstrap score, 56). They were significantly distinguished from the representative species of the Viteae tribe and Ampelopsideae tribe. This is an indication of the species identification potential of the combined molecular markers. These results demonstrate that the combined barcode strategy of trnL+ trnN derived from comparative chloroplast genomes is a potential molecular tool for the geographical origin discrimination of T. hemsleyanum in China.

Chloroplast genome features of Tetrastigma and Vitaceae plants
The chloroplast genomes of the five T. hemsleyanum plants with different geographical origins as well as other representative species of Tetrastigma and Vitaceae exhibited a typical quadripartite structure containing a pair of IR regions, a single LSC region, and a single SSC region, which was similar to that of most vascular plants [38,39]. The total length of the chloroplast genomes in this study ranged from 160,124 bp (T. hemsleyanum from Jiangxi) to 161,430 bp (A. japonica), that was consistent with the reported cp genomes from multiple other plants in Vitaceae, such as A. grossedentata (162,147 bp) and V. davidii (160,950 bp) [40,41]. In spite of highly conserved cp genomes of angiosperms, gene loss and gain events continually occurred in certain species [42]. In our observations, the ycf15 gene was only encoded uniquely in T. hemsleyanum and A. japonica, while it absented in T. planicaule and V. vinifera (Table 2). According to the previous studies regarding 10 Chinese wild Vitis species [43], the uniform loss of ycf15 gene was indicated as one of the essential characteristics in Vitis plants.
However, other reports from Ampelopsis species showed that the ycf15 gene was disabled in A. brevipedunculata [44], and A. humulifolia [45]. These results further suggested relative complex regulatory functions of ycf15 gene in the evolutionary of Vitaceae plants. Additionally, the trnS-GCU gene was merely encoded by T. hemsleyanum, while trnG-GCC , trnG-UCC and trnV-GAU were solely encoded by other three Vitaceae species, evidently reflecting the uniqueness of T. hemsleyanum in the family of Vitaceae. It is worth noting that the four special tRNA genes were all encoded by Ampelopsis brevipedunculata [44], resulting in a total of 31 distinct tRNA genes. The above results indicated that gene losses are not always dependable markers for phylogenetic relationships and further explorations focused on gene functions ought to be implemented and investigated urgently. With regards to the GC contents, chloroplasts from genus Tetrastigma (37.49-37.52%) showed slightly higher ratio than that  [44,45] and Vitis species (37.05-37.40%), except that of V. romanetii (38.28%) exhibiting an unusually large genome in size (232,020 bp) [43]. It is common knowledge that GC base pair is more stable than the AT. Accordingly, the increase of GC content in Tetrastigma species could potentially improve the stability of chloroplasts, consequently contributing to their adaptation to some harsh growing environments such as rocks crevices.
The expansion and contraction of the IR boundary is one of the main driving forces of changes in chloroplast genome size [46]. Except for the rpl2 gene, there was no significant variation among the T. hemsleyanum chloroplasts from different regions. In addition, only slight IR expansions and contractions were found in every border of T. hemsleyanum chloroplasts, further confirming their conserved traits of IR boundary (Fig. 3). However, the analyzed results indicated the location and the intron number of rpl2 gene were diverse in T. hemsleyanum cp genomes from different regions. The rpl2 gene of T. hemsleyanum cp genome from Sichuan, Fujian and Jiangxi T. hemsleyanum cp genomes showed one intron and located in IRA completely, which was consistent with the reports of previous researches on T. planicaule [47] as well as other Vitaceae species [45,48]. Remarkably, T. hemsleyanum chloroplasts from Guangxi and Zhejiang provinces possessed two introns that located across the border of JLA, which was not common in Vitaceae plants (Fig. 3). Nevertheless, several lineages of dicotyledons, including Saxifragaceae and Convolvulaceae, have even been reported to lose the intron of rpl2 gene independently [49], which was regarded as the main characteristic of core members of Caryophyllales [50]. Moreover, shrinkage and expansion of the IR boundary could also trigger the duplication of genes or the generation of pseudogenes in angiosperms chloroplast genome [51]. Among the analyzed Vitaceae species in this study, only the cp genome of V. vinifera possessed a pseudogene ycf1 located in IRB region completely. The pseudogenizations of ycf1 gene were also documented in other Vitis species [43]. However, the similar event was not observed in Ampelopsis and Tetrastigma plants. Previous studies have suggested that ycf1, a well-known gene with the most variable sites in cp genome, could be a promising DNA barcode with better performance than these current universal barcodes [52]. Nevertheless, our results showed that ycf1 region was not the marked hotspot with the most variation sites during the cp genome comparison analysis among T. hemsleyanum plants and other four Vitaceae species. Therefore, the function of ycf1 gene and its specific role in phylogenetic relationship of Vitaceae need to be further elucidated with more cp genomes.
The adaptive evolution of cp genome genes represented valuable tool for exploring the variation of gene function, structural change and evolutionary of species [53]. The pairwise Ka/Ks values have been extensively used as an efficacious indicator to reveal positive selection pressure and adaptive evolution rate of species [54]. For the protein-coding genes with RNA editing sites in T. hemsleyanum chloroplasts in our study, the majority (91.67%) of Ka/Ks values exhibited a range from 0 to 0.5, which was in accordance with the previous researches on some Vitaceae plants [44]. The lowest Ka/Ks value (0) was observed within genes encoding subunits of photosystem II (psbF and psbL) and protease (clpP). While the most salient Ka/Ks values happens on matK and rps16, which encoded maturase and small subunit of ribosome. The higher Ka/Ks values of matK and rps16 indicated that these two genes might be more sensitive to the environment changes. The combination approach of matK and rps16 sequences has been successfully applied to distinguish the primitive species of Polygonatum species [55]. In addition, the matK gene was also considered to be necessary for the survival of plant cells, and its expression required seriously strict regulation to prevent pernicious effects [56]. The previous data have suggested that the evolution of matK region was considered as the fastest gene in chloroplast genome, providing a powerful marker for genetic classification of jewel orchid accessions in Vietnam [57]. Herein, the matK gene with high Ka/Ks value could also be conjectured to be potential for species identification and genetic relationship study among Vitaceae plants.

Phylogenetic analysis and taxonomic implications within family Vitaceae
The grape family Vitaceae is an economically important family of angiosperms containing 16 genera and about 950 species. The Phylogenetic analysis based on the chloroplast genome sequences strongly supported the position of the Vitaceae as the earliest diverging lineage of rosids [58]. In addition, the Vitaceae family could be divided into five tribes by both nuclear and chloroplast genomic data, including Ampelopsideae, Cayratieae, Cisseae, Parthenocisseae, and Viteae, respectively [59]. Our phylogenomic analysis recovered three well supported lineages within Vitaceae (Fig. 8) that correspond to the three tribes reported in previous study [37]. The evolutionary relationships constructed in our study were consistent with those of previous researches [37,60,61], further confirming the advances of whole chloroplast genome analysis on the phylogenetic resolution in certain lineages. The molecular phylogenetic analysis of Vitaceae was initially investigated utilizing several plastid genes, including the trnL-F, atpB-rbcL, trnC-petN, trnH-psbA, spacer and rps16 intron [61,62]. A phylogenetic tree of Vitaceae with 417 single-copy nuclear genes was reconstructed from transcriptomes of 15 Vitaceae species, providing robust support for the deep relationships of the grape family and indicating the phylogenetic utility of transcriptome data for plants over a time scale [63]. Subsequently, the deep relationships of Vitaceae family was explored by skimming the chloroplast and mitochondrial genomes for 27 taxa, supporting the division of five major clades of the family: the Vitis-Ampelocissus clade, the Parthenocissus-Yua clade, the core Cissus clade, the Cayratia-Cyphostemma-Tetrastigma (CCT) clade, and the Ampelopsis-Rhoicissus clade [37,64]. It is interesting to note that the plants of tribe Cayratieae, including genera Causonis, Cayratia, Cyphostemma, Pseudocayratia and Tetrastigma, exhibited larger genomes than other four tribes in grape family [65]. Specially, five chloroplast molecular markers were employed to the phylogeny and biogeography of Cayratia, confirming the monophyly of the CCT clade and further suggesting its geographical origin of continental Africa in the late Cretaceous [60]. Furthermore, phylogenetic relationships within 72 Tetrastigma species recognized six strongly supported clades on the basis of ten plastid DNA regions, which do not completely correspond to their geographical distributions [66]. Nevertheless, the genus of Cayratia has been proved to be divided into three branches, relating to their geographical distribution [67]. The genus of Cayratia in Africa formed an independent group, while the other two groups were consisted of samples from Asia and Australia, which reflected the great impacts of different geographical origins on phylogenetic classification [68]. In this study, we determined the complete cp genomes of T. hemsleyanum from five provinces and constructed a comparative analysis with representing plants from tribe Cayratieae, which enriched the genome database of T. hemsleyanum and provided the basic data for improving the phylogenetic relationships among Tetrastigma species with better resolution. However, the sample sources in our research were all from China in Asia. Studies including a wider sampling area should be carried out across the genus to further elucidate the deeper relationships within Tetrastigma and Cayratieae.

Development of molecular markers for efficient species classification in Vitaceae
The complete chloroplast genomes have proven to be powerful tool to resolve evolutionary relationships among plant species, and provide valuable information for molecular species authentication [69]. Compared to the potential markers from nuclear genome, the chloroplast genome derived markers harbored rich gene copies in plants and sufficient inter-specific divergence coupled with low intra-specific variations [70]. The identified highly variable regions of cp genomes could be developed as efficient DNA barcodes and used extensively in rapid species identification and large-scale phylogenetic analysis [71,72]. The previous reports have confirmed that seven DNA barcodes selected based on the basis of the cp genomes of Pterocarpus species displayed better discernibility when comparing with the universal barcodes of rbcL, matK, trnH-psbA and trnL-F [73,74]. In our study, a total of five hypervariable regions across four complete cp genomes of Vitaceae species were identified, of which four were intergenic regions (rps16-trnQ, psbM-trnD, psbZ-trnfM and ycf3-trnS) and one was protein coding gene (ycf1) (Fig. 6A). In addition, the IR regions of Vitaceae plants exhibited significant lower sequence divergence than the SSC and LSC regions, which was consistent with the results from other plants [75]. It is a very common phenomenon in the cp genomes of most angiosperms, where the intergenic spacers contained more sequence variants than the coding genes. The developed molecular markers with higher inter-specific levels have been widely reported in other studies and exhibited excellent discriminating ability in investigating taxonomy and molecular phylogeny, such as Pulsatilla of Ranunculaceae [76] and Rhodiola of Crassulaceae [77]. It is worth noting that the three intergenic regions, rps16-trmQ [78], psbM-trnD [40] and psbZ-trnfM [79] found in our results were also reported to be candidate DNA barcodes for plant identification and phylogenetic relationships analysis in Echinacanthus, Myrtales and Zygophyllum species, respectively. The deep phylogenetic relationship within the tribe Cayratieae and family Vitaceae still remains uncertain due to high morphological similarity and less molecular data. Furthermore, the dried root from T. hemsleyanum bears a striking morphological resemblance to that from closely related species of Cayratieae and Vitaceae, which caused the emergence of adulterant and counterfeit drugs in the market and led to potential hazard to health and clinical efficiency. The molecular phylogeny based on the combination of three chloroplast markers suggested the split of genus Vitis into three clades and supported a relatively recent and intense gene flow in species from different regions [80]. The universal DNA barcode ITS2 was reported to distinguish T. hemsleyanum from its adulterants, providing an effective and accurate identification strategy for this endangered herb [12]. However, we failed to sequence and amplify the ITS2 gene with the DNA template extracted from the processed roots of T. hemsleyanum, indicating the potential destruction on nuclear genomes during the processing of crude drugs (data not shown). The destroyed nuclear DNA brought the limitations of universal nuclear DNA barcode sequences for distinguishing processed medicinal plants. Contrast with the nuclear barcode ITS2, the developed chloroplast genome markers of trnL and trnN could be successfully amplified with the genomic DNA from the dried root of T. hemsleyanum, suggesting the chloroplast genome was more stable than the nuclear DNA during the processing of medicinal plant. Therefore, the identified mutational hotspots regions and according primer set were believed to help to distinguish the taxa in the genus level, which would provide a credible approach to identify related species and assess the interspecific phylogenetic relationships among Vitaceae plants.
The hypervariable region worthy of special attention was ycf1, which exhibited the highest Pi value among the coding genes in this study (Fig. 6A). The DNA barcode of ycf1 has been confirmed the excellent ability in identification of Fritillaria species [81] and phylogeny reconstruction of Primula species [82]. The comparative analysis of the ycf1 gene in four Vitaceae plants revealed that fewer SNP sites and Gaps were shown between T. hemsleyanum and T. planicaule while more variations were presented between A. japonica and V. vinifera (Table 6). These results indicated that the ycf1 was more reliable for the research on phylogenetic relationships in Vitaceae plants than the species identification of related plants within Tetrastigma. In addition, the PCR amplified length of the currently determined three hypervariable regions of psbM-trnD, psbZ-trnfM and ycf1 were less than 1000 bp, which could result in the high success rates of amplification and sequencing. Therefore, these three diversity regions presented candidate barcoding sequence, which might be helpful to plant identification, systematic investigation of Vitaceae and evaluating the phylogenetic relationships among the tribe of Cayratieae.

Geographical origin discrimination strategy for medicinal plants
The quality of medicinal plants depends on various factors with significant contributions of genetic impacts and geographic location. The development of geo-authentic Chinese medicinal material was closely associated with cultivated outstanding genetic mutants and eligible local environmental impacts, leading to the production of famous crud drugs with higher quality and price in the market [83]. With the increasing global demand for plant medicines, great numbers of medicinal species have been cultivated in different areas to generate multiple genetic populations with similar morphological features [84]. Due to the significant advantages of genuine medicinal material from geo-authentic producing areas, it was more likely to be adulterated with lower-price counterparts from other different regions [26]. However, the geographical origin of herbal medicines has been indicated as a crucial factor influencing the quality and potential treatment efficacy of the medicinal materials, which could be attributed to the variations in the environmental conditions and genetic reasons [85]. Zhang et al. [86] confirmed that there were obvious differences in chemical components of dandelions from four different geographical regions by metabolomics analysis, especially in phenolic compounds. To ensure the health benefits and clinical effectiveness of herbal medicine to consumers, it is necessary to develop strategies for the recognition of geographic origin crud drugs. Since the obvious differences on the content of chemical constituent and plant genetic populations, several analytical approaches of DNA and chromatographic analysis have been extensively applied to determine the geographical origin of medicinal plants and foods [87]. The HPLC similarity analysis and content of alkaloids was indicated as valuable tools for differentiating the geographical origin of the Fuzi samples [88]. The excitation-emission matrix fluorescence and chemometric strategies have also been considered as promising methods for distinguishing the geographical origin of Gastrodia elata [89]. Compared with the chemotaxonomical investigations, the DNA analysis based on either unique sequence regions or DNA polymorphism from genetic markers represented alternative approaches to identify plant populations and authenticate plant species. The amplification of unique gene fragment exhibited a rapid and easy method to identify the geographical authenticity of Scrophularia ningpoensis [67]. The SSR markers were indicated as suitable tool for assessing genetic diversity and population structure of spinach germplasm, which clearly separated the accessions with different geographical origins [90]. Recently, the combination of DNA molecular markers and chemical analytical techniques has been successfully used to distinguish the geographical origins of traditional Chinese medicines. The integration of microsatellite markers and chemical analysis could discriminate the Moutan Cortex from different sources and geographical origins [91]. The ISSR fingerprinting combined with FTIR spectrum analysis established a rapid and efficient approach to determine the Cassia tora populations with different eco-geographical origins [92]. The development of efficient geographical tracing system of medicinal plants would significantly contribute to the protection of genuine plant genetic resources and improvement for quality control of herbal drugs.
Since the significantly differences on the crude drug quality and clinical efficiency of T. hemsleyanum from different regions in China, a variety of identification approaches have been reported for discriminating geographical origins of T. hemsleyanum, including determining strategies based on macroscopic analysis, spectroscopic technology, chromatographic fingerprint and bio-activity evaluation. Previous study has reported that the root tuber of T. hemsleyanum from Zhejiang and Guangxi province could be successfully distinguished by analyzing the external characteristics of tuber and the typical micro-structures of powder, thus providing an intuitive and simple approach for plant origin determination [93]. However, pharmacognosy-based identification cannot accurately identify the processed decoction pieces of T. hemsleyanum from the above two aspects. Li et al. [94] effectively distinguished the T. hemsleyanum samples from Zhejiang, Yunnan and Guizhou Province using a combined identification approach based on HPLC fingerprints and the random forest (RF) algorithm analysis. Machine learning algorithm has been widely used for spectral data processing to discriminate medicinal herbs from different habitats. A dual-mode microscopic hyperspectral imager (DMHI) system has been developed using the combined dataset of RMHI and FMHI modes for hyperspectral detection of the origins and varieties of T. hemsleyanum, obtaining a prediction accuracy as high as 97.5% of both origins and varieties [95]. Besides, the near-infrared spectroscopy (NIRS) combined with deep learning models also exhibited potential capability to distinguish the medicinal plant T. hemsleyanum among different origins [96]. However, the processing of crude drugs and growth years and harvesting seasons of medicinal materials can significantly affect the geographical origin identification methods based on chemical contents and compositions, thus generating inaccurate results. Furthermore, deep learning is complex, time-consuming, and associated with a low signal-to-noise ratio (SNR), instability, and spectrum peaks overlap, thus limiting its application. In contrast, the origin discrimination approaches based on DNA markers could target the direct carrier of genetic information in plant populations, thus providing an alternative strategy with higher stability than those based on chemical analysis. Therefore, this study provides an efficient method for distinguishing the geographical origins of T. hemsleyanum based on DNA barcodes from the cp genome from different regions. The developed specific DNA barcodes and their combination divided the 53 T. hemsleyanum samples from six provinces into four haplotypes and successfully classified a sample from Genbank into Clade I. The specie of T. hemsleyanum exhibited genetic patterns characteristic of long-term separation in multiple refugia and lower levels of interpopulation gene flow, indicating that the genetic population divergence largely driven by mutation or drift, further supporting the genetic stability of T. hemsleyanum samples in this study [1]. Previous study on lineage diversification reported that T. hemsleyanum was consisted of two major cpDNA lineages, Southwest (SW) and Central-South-East (CSE) China, consistent with our grouping results [97]. Herein, the T. hemsleyanum samples from CSE China were gathered into a gigantic branch, while those from Sichuan Province in SW China were grouped in another cluster. Furthermore, Besides the DNA molecular barcoding strategy, other molecular markers of RAPD and SRAP also displayed the potential capacity in determining the geographical origins of T. hemsleyanum. For instance, Yin et al. [98] have sifted out 10 pairs of RAPD primers for PCR amplification of 64 samples of T. hemsleyanum from 14 provinces in China. They revealed abundant genetic diversities of T. hemsleyanum germplasm resources and significant complexity of geographical distribution. However, the RAPD cluster analysis was inconsistent with the geographical distance of the provenance, requiring further revisions for primers for geographical origin determination. Notably, SRAP markers divided the same 64 T. hemsleyanum samples into nine groups, which exhibited certain different results with that from RAPD analysis [99]. Consortium for the Barcode of Life [100] has confirmed that combining barcodes has a better identification efficiency than a single barcode. Herein, the combination of trnL + trnN barcode divided the T. hemsleyanum plants into four groups, thus effectively identifying the genetic populations of T. hemsleyanum samples from Zhejiang and Sichuan provinces. It is worth noting that the samples of T. hemsleyanum in Zhejiang Province exhibited three haplotypes. One of them was consistent with the samples from Fujian, Guangdong, Guangxi and Jiangxi provinces, and the other two haplotypes were unique to Zhejiang Province. Besides, all samples from Sichuan Province were clustered into a particular branch. However, our results failed to distinguish the strains in each of other provinces concretely, generating a clade with T. hemsleyanum samples from Zhejiang, Jiangxi, Fujian, Guangxi, Guangdong and Genbank database. A promising approach for the accurate traceability of T. hemsleyanum from different regions should be established urgently.
Multiple universal barcode markers have been proposed for recognizing species at a genera and family level, such as matK, rbcL [100], ITS2 [101], trnH-psbA [102] and trnL-F [103] with the length of DNA sequences between 400 and 1000 bp. Similarly, it is also applicable to trace various geographical origins of species with universal barcodes. The method based on universal DNA barcode ITS2 suggested that T. hemsleyanum from Zhejiang province had a unique genetic status, exhibiting potentiality for the plant population and geographical origin discriminating between Zhejiang and other provinces [12]. Nevertheless, there is no doubt that low discriminatory power is inevitable for universal DNA barcodes, especially in tracing intraspecific geographical origins. Increasing number of case studies have indicated that the universal DNA barcodes have lower divergence and poor discriminatory power [12]. The mutation events in the chloroplast genome are not universally randomly distributed within the sequence and are concentrated in certain regions forming the "hotspot" regions [104]. Comparison of the chloroplast genome sequences is an effective strategy to identify the mutation hotspots and these highly variable regions can be used as the specific DNA barcodes. This view could be explained by the fact that universal barcodes possessed lower nucleotide diversity (Pi) values (0.00075-0.0025) while the higher Pi values (0.003) presented in specific barcodes of trnL-CAA and trnN-GUU (Supplementary Fig. 3). This is also the reason that we performed the work to identify more informative DNA regions for geographic identification of Tetrastigma hemsleyanum.
SSRs have become a new molecular marker technology and extensively applicated in plants genetic diversity, gene mapping and variety identification with the characteristics of strong polymorphism, co-dominance, high universality and good stability [105]. The distribution of chloroplast SSR characteristics exhibited taxon specificity in Cyatheaceae species, which provided valuable phylogenic information at the genus level [106]. Furthermore, the SSR primers designed based on different Gracilaria tenuistipitata chloroplast genomes from various regions could separate the samples into two main geographical regions, which significantly contributed to the mass cultivation of seaweeds with high economic potential [107]. The developed program, ChloroSSRdb focused on the application of chloroplastic SSRs from Viridiplantae, thus providing useful resources in developing genetic markers and phylogenetic analysis [108]. The ISSR molecular markers divided the germplasm resources of T. hemsleyanum from the main distribution areas of China into 4 Clades, among which Zhejiang samples were all clustered in the Clade I [109]. Another report showed that ISSR and SRAP markers could cluster the wild accessions of T. hemsleyanum into four groups (similarity coefficient level, 0.75) [14]. All the wild populations from Zhejiang were highly distinct for ISSR polymorphism and formed a separate cluster, while those from the other three clusters consisted of T. hemsleyanum samples from Guangxi, Jiangxi and Hunan. This study also found abundant SSRs in the cp genomes of T. hemsleyanum plants. Furthermore, significant differences were identified on SSR numbers and types in the T. hemsleyanum cp genomes among five samples from different regions, mainly distributed in quantities of mononucleotide repeats and types of repeat unit (Fig. 4A). Meanwhile, a comparison among four Vitaceae species including T. hemsleyanum from Jiangxi has revealed great discrimination in SSR counts and types between different tribes, such as tribe Cayratieae, Ampelopsideae and Viteae, suggesting the potentiality of SSRs in species discrimination and classification (Fig. 4B). However, most of current studies focused on the interindividual genetic variation of T. hemsleyanum, thus limiting the use of SSR markers in determining the genetic populations from different regions. Therefore, the SSRs fingerprints should be systematically constructed for geographical origin identification of T. hemsleyanum based on the present comparative analysis of cp genomes from different regions.

Plant material and DNA extraction
The plant materials of T. hemsleyanum were collected from six different provinces in China (Supplementary Table 1 Table S1. Total genomic DNA was extracted from the fresh and healthy leaves of T. hemsleyanum using a modified cetyltrimethylammonium brofmide (CTAB) method [110]. The final DNA integrity and concentration were assessed by electrophoresis on 1.0% agarose gel and Nanodrop 2000 Spectrophotometer (Thermo Fisher Scientific, United States).

Sequencing, genome assembly and annotation
The paired-end (150 bp) sequencing of the DNA libraries was conducted on the Illumina HiSeq 2500 platform, generating about 2.4 GB of raw data for each sample. Then the quality of paired-end Illumina reads was assessed with FastQC, and the low-quality reads were removed using Fastp. The filtered reads were assembled de novo using metaSPAdes with the complete cp genome of T. hemsleyanum (NC_029339) as reference and the proteincoding genes, mRNA genes, tRNA genes were annotated by GeSeq annotation tool [111]. The CPGAVAS2 software also used to annotate protein-coding genes [112]. BLAST was further used to correct the annotation of chloroplast genome manually. The circular chloroplast genome map of T. hemsleyanum collected from Jiangxi Province were drawn by OrganellarGenomeDRAW (OGDRAW) tool [113]. Finally, the fully annotated cp genomes were deposited at the GenBank database (Supplementary Table 1).

Comparative analysis of chloroplast genomes and identification of hypervariable regions
MEGA 7.0 [114] was used to analyze the genome feature and Codon W software was used to investigate the distribution of codon usage using the RSCU value [115]. The IR/SC boundary locations in five samples of T. hemsleyanum and three representative Vitaceae species were compared using IR scope [116]. As for the repeats analysis, long repeats of four different type (forward (F), palindromic (P), reverse (R), and complementary (C)) were identified by REPuter, with hamming distance 3, minimal repeats 30 and maximum computed repeats 50 [117]. Simple sequence repeats in Vitaceae species were detected by MISA, setting parameters as 10 for mononucleotide SSRs, 5 for dinucleotide SSRs, 4 for trinucleotide SSRs, 3 each for tetranucleotide, pentanucleotide and hexanucleotide SSRs [118]. To predict the number of RNA editing sites, the PREP-Cp program was employed with a cutoff value 0.8 [119]. For the identification of hypervariable regions within five samples of T. hemsleyanum and among four representative species of Vitaceae, we aligned cp genome sequences using MAFFT [120] and evaluated the sequence divergence among Vitaceae species through a sliding window analysis in DnaSP v6 [121]. The parameters of sliding window analysis were set as window length for 800 sites and the step size of 200 sites.

Ka/Ks and positive selection analyses
In order to analyzed the Ka and Ks substitution rates and Ka/Ks ratio, Melaleuca cajuputi was compared with T. hemsleyanum (Jiangxi Province) in 24 protein coding genes. The alignment was carried out by MAFFT v7.037b [120], and the calculation of the value of Ka/Ks was implemented by DnaSP v6 [121].

Primier design and PCR amplification
Based on conserved nucleotide sequences at both ends of mutation hotspots, 5 pairs of specific primers were designed by Primer Premier 5 (Supplementary Table 2). PCR was performed in 50-μL reactions consisting of 3 μL of genomic DNA, 5 μL of dNTP Mix and 10× LA PCR Buffer, 2.5 μL of forward and reverse primers (25 μmol/L), 0.5 μL of LA Taq (Takara) and ddH2O supplemented to 50 μL. PCR amplification was carried out by Veriti Thermal Cycler (Applied Biosystems) with the following program: 5 minutes at 94 °C for initial denaturation; denaturation 94 °C, 45 seconds; 33 cycles consisting, annealing temperature 53 °C -58 °C, 30 seconds, extension temperature 72 °C, 45 seconds, final 10 minutes extension at 72 °C, 4 °C low temperature save. PCR products were examined by electrophoresis on 1.0% Agarose Gel and visualized with 4S GelRed Nucleic Acid Stain (Sangon Biotech, China). Finally, optimum annealing temperature of trnL and trnN sequences was determined as 56 °C. The synthesis of primers and Sanger sequencing were conducted by Sangon Biotechnology and Shanghai Sunny Biotechnology Co., Ltd., respectively.

Multiple sequence alignment and data analysis
The sequencing results were performed using BioEdit 7.0 [122] and aligned by MAFFT 7.0 [120] to quantify the sequence length and base composition. To count the variant information, the aligned sequences were analyzed by MEGA 7 [114]. Based on the single and combination sequences, the intraspecific genetic distance of T. hemsleyanum between Jiangxi Province and other regions, as well as the interspecific genetic distance among representative species of Vitaceae were calculated using MEGA 7 with Kimura 2-parameter (K2P) distance model.

Phylogenetic analysis and effectiveness of marker discriminatory
The maximum likelihood (ML) tree and maximum parsimony (MP) tree were constructed using 70 conservative protein-coding genes of five T. hemsleyanum complete cp genomes sequenced in this study and 10 additional publicly available sequences that we downloaded from the NCBI. Among these 15 species, Melaleuca alternifolia and Melaleuca cajuputi were chosen as outgroups. The ML tree was constructed based on the K2P model with 500 bootstrap replications. Similarly, MP tree was obtained from MEGA 7 [114] under default parameters with 500 bootstrap replications. Based on the K2P distance model, the single and combination trnL or trnN sequences of Vitaceae species were used to construct the NJ trees with 1000 bootstrap replications.

Conclusions
This study provides 5 complete chloroplast genome sequences of Tetrastigma hemsleyanum with different geographical origins in China, and presents a comparative analysis of cp genomes with other representing species from family Vitaceae. The chloroplast genome structure of T. hemsleyanum samples from different regions and other Vitaceae plants was highly conserved. However, IR expansion and contraction was observed among cp genomes of T. hemsleyanum from different areas, resulting in cp genomes of different sizes. In addition, significant differences in SSR types and numbers were identified in the cp genomes of different T. hemsleyanum samples as well as Vitaceae species, providing valuable genetic information for the development of species identification strategy and geographical origin determining system. Phylogenetic analysis revealed the five T. hemsleyanum plants clustered together to form a stable monophyletic group, exhibiting sister relationship with T. planicaule to compose the tribe of Cayratieae. A total of five highly variable regions with significant differences between T. hemsleyanum and other Vitaceae species were identified and may be applied as potential markers for species identification and further phylogenetic