A DNA Barcode Reference Library of Native Seed Plants in Chinese Dryland Regions Authors: Feng Song, Jie Lian, Zhan-Lin Liu, Xiao-Jun Li, Wen-Da Huang, Xin Zhao, Xia Yan, Lu Jin, Guang-zhao Jin, Xue-Yong Zhao, Xue-Jun Ge Published: 2025 ABSTRACT AbstractApproximately 6.6 million km2 of China’s territory consists of drylands at high desertification risk. Despite ecological vulnerability, these regions remain understudied in biodiversity research. DNA barcoding provides an efficient tool for rapid species identification, offering significant potential for biodiversity conservation and ecological surveys in drylands. However, comprehensive reference data for China’s drylands are currently lacking, underscoring the urgent need for a reliable, extensive, and traceable DNA barcode reference library. To address this gap, we systematically surveyed plant diversity across nearly all Chinese drylands from 2013 to 2021. Our study documented 1,140 species, representing 434 genera, 76 families, and 29 orders. Based on this effort, we constructed the most comprehensive DNA barcode reference library for native dryland seed plants in China, comprising 13,246 sequences across five markers: rbcL (3,056), matK (3,332), ITS (3,358), psbA-trnH (1,770), and trnL-F (1,730). This resource establishes a critical foundation for integrated taxonomic research, biodiversity monitoring, and environmental impact assessments. Moreover, it supports initiatives in desertification control, plant conservation, and sustainable development in China’s dryland regions.Subject terms: Plant sciences, Ecology Background & Summary Drylands cover approximately 41% of Earth’s land area1, forming highly fragile ecosystems vulnerable to climate change. With global population growth, desertification threatens biogeochemical cycles and human livelihoods, demanding urgent international action2,3. China exemplifies these challenges, with 6.6 million km² of drylands at risk of expansion due to increasing aridity—potentially impacting 580 million people4. This degradation worsens socioeconomic risks, highlighting the need for sustainable solutions. Dryland flora serve as critical ecological stabilizers in dryland regions, providing essential ecosystem services such as windbreak and sand fixation. China’s dryland harbor approximately 1,425 documented plant species5–7, including over 80 endemics and 50–60 rare or endangered taxa (e.g., Tugarinovia mongolica, Cistanche deserticola, Ammopiptanthus mongolicus, and Helianthemum songaricum). These species constitute both the ecological foundation of these fragile ecosystems and vital biological barriers against desert expansion8. While species richness in dryland is markedly lower than in traditional biodiversity hotspots, dryland plants have evolved sophisticated adaptive mechanisms through prolonged evolutionary processes. They possess unique functional gene arrays that confer resistance to multiple abiotic stresses, including drought, temperature extremes (both hyperthermia and hypothermia), high salinity, and nutrient-deficient soils6,9,10. These characteristics make them invaluable genetic resources for studying extremophile adaptations. Beyond their ecological significance, dryland flora hold substantial socioeconomic value as sources of medicinal compounds (e.g., C. deserticola), biofuel sources (e.g., Haloxylon ammodendron), forage plants (e.g., Caragana korshinskii), and other commercial products. Keystone species like Calligonum mongolicum and Nitraria tangutorum exemplify this dual ecological-economic importance11,12. Consequently, conserving dryland plant biodiversity represents a strategic imperative for both desertification control and sustainable regional development. A systematic documentation of plant resource characteristics and their geographical distributions in dryland regions is fundamental for achieving sustainable utilization and effective conservation. DNA barcoding has emerged as a transformative approach for this purpose, employing standardized DNA fragments to enable rapid and accurate species identification through nucleotide sequence variation13. This technique addresses critical limitations of traditional taxonomic methods, including their time-consuming nature, requirement for specialized expertise, and dependence on specimen type, integrity, and developmental stage14–17. Furthermore, the methodology has been propelled by sequencing advances, transitioning from Sanger sequencing to next-generation sequencing (NGS) techniques18. This shift enables the cost-effective, rapid and high-throughput retrieval of barcode sequences from complex samples, such as mixtures or environmental DNA (eDNA), thereby significantly enhancing the efficiency of DNA barcoding research19. However, the accurate taxonomic assignment of sequences obtained through barcoding methods relies entirely upon precise reference databases. As a result, the development of such reference databases has become a critical bottleneck, constraining the effectiveness of field surveys and conservation evaluations18,19. Building such comprehensive databases is therefore essential to overcoming this limitation and directly enhancing the efficiency and reliability of biodiversity monitoring and conservation applications. In response to this challenge, numerous national and international initiatives have established region- or taxon-specific DNA barcoding libraries, with prominent examples from Australia, Canada, Japan, Norway, and the United Kingdom20–24. Following global trends, China has seen significant development of DNA barcode databases, evidenced by 261,609 records in BOLD (of which only ∼12% are from plants as of October 2025) and specialized libraries for mangroves, timber, and tropical/subtropical evergreen broadleaf forests25–27. Nevertheless, a pronounced geographic bias remains: most barcode data are concentrated in southern, biodiversity-rich regions, while arid ecosystems are notably underrepresented7,28. This specific gap hinders the rapid and accurate identification of plants in dryland ecosystems, which in turn undermines our ability to accurately assess ecological patterns and biodiversity in these vulnerable areas. To bridge this critical knowledge gap, we established the first comprehensive DNA barcode database for desert flowering plants across Northwest China’s arid land in 20237. Using the standard barcode system for land plants (rbcL, matK, and ITS/ITS2)29, we generated 5,064 high-quality sequences from 1,816 specimens representing 890 species, 385 genera, and 72 families; these comprised 1,804 rbcL, 1,717 matK, and 1,599 ITS sequences, collectively designated as Dataset “DS-DPNC” in this study. To ensure comprehensive coverage of China’s dryland flora, we expanded our sampling framework to include not only typical desert ecosystems but also high-altitude desert regions, such as Tibetan Plateau, Alxa Plateau, Ordos Plateau, and semi-arid sandy lands in central and eastern Inner Mongolia. Additionally, to maximize taxonomic resolution, we supplemented the core barcodes with two chloroplast markers (trnL-F and psbA-trnH). To this end, this study was designed with two primary objectives: (1) to construct a comprehensive DNA barcode database for China’s dryland flora through extensive sampling; (2) to evaluate the effectiveness of core DNA barcodes in resolving species identities within these highly adapted floras. This publicly accessible reference library represents a significant advancement in dryland plant genomics, offering both theoretical and practical value by: (1) enabling key applications such as invasive species monitoring and the protection of endangered plants; (2) elucidating dryland vegetation patterns; and (3) optimizing ecosystem management strategies, ultimately facilitating the conservation and sustainable use of China’s dryland biological resources. Methods Two elements of the data pertinent to this study are tabulated in Table 1 according to the categories of data in the library: newly available data from desert plant community surveys (named “DPC”); and (“DS-DPNC”) published data. The locations of the survey sites included in this study are illustrated in Fig. 1. Details of the library construction are provided below. A complete data source is available in Figshare30. Table 1Summary of the Core DNA barcode for desert plant in China.DatasetSequence informationCollected dataDNA barcodesrbcLmatKITSpsbA-trnHtrnL-FRMI/RMIPFDS-DPNCNo. sampled individuals1,8161,8041,7161,675——1,604 (88%)DPCNo. sampled individuals1,9541,8981,6601,7251,7701,7301,238 (63%)TotalNo. sampled species1,1401,1261,0661,0754504431,009 (89%)No. sampled genus434432414420227231403 (93%)No. sampled family76767574535473 (96%)No. sampled individuals3,7243,0563,3323,3581,7701,7303,005 (81%)RMI represents the combination of rbcL, matK, and ITS; RMIPF represents the combination of rbcL, matK, ITS, psbA-trnH and trnL-F. Summary of the Core DNA barcode for desert plant in China. RMI represents the combination of rbcL, matK, and ITS; RMIPF represents the combination of rbcL, matK, ITS, psbA-trnH and trnL-F. Map of desert plant collection sites in China. Red dots represent dataset DS-DPNC. Blue triangles indicate dataset DPC. The map was generated using ArcGIS 10.1. Sampling and identification This study established the DPC dataset through systematic field collections conducted across China’s dryland regions from 2017 to 2021. The sampling strategy encompassed five key biogeographic zones: (1) Tarim-Junggar Basin deserts, (2) Qinghai-Tibetan Plateau alpine desert, (3) Alxa Plateau-Hexi Corridor desert, (4) Western Ordos Plateau-Yinshan northern foothills desert, and (5) semi-arid sandy lands in central and eastern Inner Mongolia (Fig. 1). Standardized field protocols were implemented at each sampling site: (1) Comprehensive collection of all observed plant species; (2) Documentation of habitat characteristics, GPS coordinates, and elevation; (3) Immediate preservation of leaf tissue in silica gel for DNA analysis; and (4) Preliminary morphological identification in situ. All voucher specimens were deposited in the following herbaria: the South China Botanical Garden (IBSC), the Xinjiang Institute of Ecology and Geography (XJBI), the Turpan Eremophytes Botanical Garden (TEBG), the Naiman Desertification Research Station and the Urat Desert-grassland Research Station of Northwest Institute of Eco-Environment and Resources. To ensure taxonomic accuracy, professional taxonomists at each institution conducted detailed morphological examinations to verify the initial field identifications. The geographical distribution of all sampling locations is presented in Fig. 1. The DS-DPNC dataset primarily comprises samples collected from Xinjiang Province and selected areas of Inner Mongolia, Gansu, and Ningxia within China’s northwestern arid regions (Fig. 1). Specimens were obtained during field campaigns conducted between 2013 and 2016. As documented by Song et al.7, complete sample metadata - including voucher specimens, georeferenced collection data (GPS coordinates), photographic records, and DNA barcode sequences - have been made publicly accessible. To maintain taxonomic consistency across datasets, we standardized all species names according to the Checklist of Plant Species in China (2022 Edition). For angiosperms, family-level classification follows the APG IV system31, while gymnosperm taxonomy adheres to the classification proposed by Christenhusz et al.32. DNA Extraction and Barcoding Protocol Total genomic DNA was isolated using the CTAB method33. Five established DNA barcode markers were amplified: two plastid protein-coding genes (rbcL and matK); two plastid intergenic spacers (trnL-F and psbA-trnH); the nuclear ribosomal internal transcribed spacer region (ITS, including ITS1/ITS2). PCR amplification was performed in 25 μL reaction volumes containing: 2.5 μL 10 × PCR buffer (100 mM Tris-HCl, 500 mM KCl, 15 mM MgCl₂), 0.5 μL each primer (10 μM), 2.0 μL dNTPs (2.5 mM each), 0.5 μL DNA template (20–30 ng), 0.2 μL rTaq polymerase (5 U/μL−1), 18.8 μL ddH₂O. Amplification conditions and primer sequences are detailed in “Primer Information” at Figshare30. All amplicons were sequenced on an ABI 3730xl DNA analyzer (Applied Biosystems). To ensure high base-calling accuracy, the ITS, matK, trnL-F, and psbA-trnH regions were bidirectionally sequenced. In contrast, the rbcL region was sequenced in the forward direction only, as its high sequence conservation and low ambiguity enable reliable assembly and accurate base-calling from a single read. Raw sequence reads were assembled and quality-trimmed using Geneious Prime v 11.0.234, with consensus sequences generated from forward-reverse alignments where applicable. Data verification Subsequent verification of taxonomic matching (family/genus level) employed tree-based methods. Sequence alignment was performed using MAFFT v7.47135 followed by manual adjustments in Geneious Prime v11.0.234 when required. All rbcL and matK sequences were aligned simultaneously, whereas ITS, trnL-F and psbA-trnH were aligned at order-level. Maximum-likelihood (ML) trees were constructed per alignment using RAxML-HPC2 (8.2.12) under the GTRGAMMA model, with nodal support evaluated through 1,000 bootstrap replicates. Monophyletic grouping of conspecific sequences was assessed using the R package ‘spider’36. Data Records The newly compiled DPC dataset comprises 1,954 specimens, representing 497 species from 244 genera, and 57 families. A total of 8,783 DNA barcode sequences were generated from five plastid and nuclear markers: rbcL (1,898 sequences; 97% success), matK (1,660; 85%), psbA-trnH (1,770; 91%), trnL-F (1,730; 89%), ITS (1,725; 88%). Notably, 1,238 specimens (63% of total samples) yielded complete barcode coverage across all five-marker. The five-marker barcode sequences have been deposited in GenBank under the following accession numbers: PV921672-PV923569 (rbcL), PV923570-PV925229 (matK), PV899998-PV901722 (ITS/ITS2), PV86836-PV870090 (trnL-F), and PV897716-PV899485 (psbA-trnH). All associated metadata - including taxonomic identifications, voucher specimen details, georeferenced collection data (GPS coordinates and elevation), barcode sequences and GenBank accession numbers - have been deposited in Figshare30 (“DPC”; “Sequences”). This study incorporates the established DS-DPNC DNA barcode reference library. This comprehensive dataset, publicly accessible via the BOLD Systems through the database DS-DPNC37 (10.5883/DS-DPNC), comprises 5,064 curated DNA barcode sequences (rbcL: 1,804; matK: 1,717; ITS: 1,599). These sequences were generated from 1,816 voucher specimens representing 890 species, 385 genera, and 72 families. The dataset includes specimen vouchers, georeferenced collection data (GPS coordinates), digital images, and sequence files. Sample information and sequence files are additionally archived on Figshare30 (“DS-DPNC”; “Sequences”). Data Overview The merger of the DS-DPNC and DPC datasets resulted in an integrated dataset of 3,770 samples. Of these, 3,724 (98.8%) were successfully identified to the species level, representing 1,140 species, 434 genera, 76 families, and 29 orders, while the remaining 46 (1.2%) were identified only to the genus level (35 genera) (Fig. 2; see “Total sampling” in Figshare30). This integrated dataset constitutes the most comprehensive DNA barcode reference library for native dryland seed plants in China, containing a total of 13,246 sequences from five markers: rbcL (3,056), matK (3,332), ITS (3,358), psbA-trnH (1,770), and trnL-F (1,730). Sampling composition of dominant plant families (A), characteristics of plant genera (B) and the maximum likelihood (ML) tree using the combined RMI dataset (C) in Chinese desert floras. The species-level ML tree “RMI-tree” was stored in Figshare30. Taxonomic analysis revealed (Fig. 2A) that the ten most speciose families accounted for 64% of genera and 63% of species: Asteraceae (66 genera/171 species), Fabaceae (25/109), Amaranthaceae (34/87), Brassicaceae (43/86), Poaceae (39/70), Rosaceae (17/46), Lamiaceae (23/44), Polygonaceae (11/38), Caryophyllaceae (13/37), Ranunculaceae (8/29). Genus-level patterns (Fig. 2B) demonstrated that thirteen hyper-diverse genera (Artemisia [34 spp.], Astragalus [27], Caragana [21], Allium [15], Calligonum [14], Tamarix [14], Silene [13], Stipa [13], Zygophyllum [12], Lappula [11], Pedicularis [11], Lepidium [10], and Potentilla [10]) collectively represented just 18% of total species (Fig. 2B). 62% of species belonged to genera containing < 5 species. 20% were from genera with 6–9 species. Technical Validation For technical validation, ITS, matK, and rbcL sequences from the integrated dataset were extracted and assessed for barcode resolution using three methods following Song et al.38: the barcoding gap (based on K2P intra- and inter-specific genetic distances), Best Close Match (BCM, taxonDNA), and tree-based method (requiring conspecific sequences to form a monophyletic clade). Only species with more than one individual were included in the data analysis to minimize the effects of missing data. This process involved 2,374 samples representing 477 species and 257 genera. Subsequently, alignments of the three loci were merged into a supermatrix (RMI). In terms of species identification capacity (Table 2), ITS demonstrated the highest recognition rate among single loci (BCM method: 71% to Barcode gap method: 60%), yet all were surpassed by the three-locus combination RMI (74% to 63%), consistent with our earlier study on desert flowering plants in northwestern China7. Based on the tree-based method results, nearly all families (rbcL/RMI: 100%; matK/ITS: 97%) and most genera (from RMI: 93% to rbcL: 77%) were monophyletic, regardless of whether the trees were constructed from single loci or the RMI combination. However, at the species level, only 63% of the species were monotypic even with the RMI combination. Many non-monophyletic genera and species predominantly occurred in species-rich families (e.g., Asteraceae, Fabaceae, Amaranthaceae, Brassicaceae and Poaceae), which is consistent with previous reports7 of inadequate barcode discrimination in such clades and in paraphyletic/polyphyletic genera (e.g., Salsoleae s.l., Gentianella, Aster, Polygonum, and Isatis)39–43. These results collectively demonstrate that core barcodes alone cannot achieve precise species-level identification for certain desert plants. This recurrent challenge highlights the necessity for more effective approaches to species delimitation in these complex lineages. Since the flora of China’s dryland is dominated by mono- and oligo-specific genera6,7 (as shown in Fig. 2B), species identification within these groups is readily achievable through morphology and biogeography once genus assignment via DNA barcoding is accomplished. Looking forward, the development of specialized reference databases utilizing next-generation DNA barcodes (e.g., plastome and nuclear ribosomal DNA) or genomic-scale data offers a promising path to enhance resolution in these difficult clades7,44. By providing substantially more informative sites, such resources have repeatedly been shown to improve species discrimination where standard barcode regions fall short. Table 2Species discrimination rates for core barcodes based on three methods for desert plant in China.Barcode typeBCMBarcodes gapTree-basedCorrectAmbiguousIncorrectwith gapSpecies/Genus/Family identificationsrbcL50%48%2%45%50%/77%/100%matK61%34%5%52%59%/88%/97%ITS71%18%11%54%60%/89%/97%RMI74%10%16%—63%/93%/100%RMI represent the combination of rbcL, matK, and ITS2. Species discrimination rates for core barcodes based on three methods for desert plant in China. RMI represent the combination of rbcL, matK, and ITS2. Usage Notes We present the first comprehensive, open-access DNA barcode reference library for seed plants inhabiting China’s dryland regions. This integrated resource, simultaneously archived in GenBank, Figshare (10.6084/m9.figshare.29589746), and BOLD Systems (dataset DS-DPNC: 10.5883/DS-DPNC), provides standardized molecular identifiers for 1,140 species; georeferenced specimen records across China’s major desert ecosystems; and multi-locus barcode coverage (rbcL, matK, ITS, psbA-trnH, trnL-F). The library serves as an essential tool for: taxonomic revisions and molecular specimen identification; biogeographic and evolutionary studies in extreme environments; biodiversity monitoring and ecological assessments; and conservation prioritization and habitat management. By enabling precise species delimitation and distribution mapping, this resource directly supports: science-based desertification control strategies, endangered species protection initiatives, and sustainable resource utilization policies in China’s ecologically vulnerable dryland zones. Acknowledgements We extend our sincere gratitude to the China Desert Plant DNA Barcode Group for their invaluable contributions to the fieldwork and sample collection. We are also grateful to the professional taxonomists who have helped with the morphological identification of the species. This study was financially supported by the Special Project for National Science and Technology Basic Resource Survey of the Ministry of Science and Technology of China (Grant No. 2017FY100200) and the Hunan Provincial Natural Science Foundation of China (Grant No. 2025JJ60200). Author contributions X.-Y.Z. and X.-J.G. conceived and designed the study; Z.-L.L., X.-J.L., W.-D. H., X.Z., G.-Z.J. and X.Y. collected and identified plant materials and carried out the DNA barcode experiment; F.S. and J.L. analyzed data and wrote the manuscript. X.-Y.Z. and X.-J.G., revised and finalized the manuscript. Data availability The data supporting this study are openly available in GenBank (individual accession numbers are provided in the “DPC” and “DS-DPNC” on Figshare), Figshare (10.6084/m9.figshare.29589746), and the BOLD Systems dataset DS-DPNC (10.5883/DS-DPNC). Code availability All analyses employed existing software without custom code development for this study. Monophyly assessment was conducted using the R package ‘spider’36, with source code accessible at: https://CRAN.R-project.org/package=spider. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Contributor Information Xue-Yong Zhao, Email: zhaoxy@lzb.ac.cn. Xue-Jun Ge, Email: xjge@scbg.ac.cn. References Associated Data Data Citations Data Availability Statement The data supporting this study are openly available in GenBank (individual accession numbers are provided in the “DPC” and “DS-DPNC” on Figshare), Figshare (10.6084/m9.figshare.29589746), and the BOLD Systems dataset DS-DPNC (10.5883/DS-DPNC). All analyses employed existing software without custom code development for this study. Monophyly assessment was conducted using the R package ‘spider’36, with source code accessible at: https://CRAN.R-project.org/package=spider.