Combining nuclear genome, chloroplast fragments and morphology to solve the species delimitation of Aquilegia incurvata (Ranunculaceae) Authors: Fei Gao, Jiao-Jie Li, Cheng Xue, Fang-Dong Geng, Lei Huang, Jian-Qiang Zhang, Xiao-Hui Zhang, Yi Ren, Ju-Qing Kang Published: 2026 ABSTRACT AbstractHybridization can be widespread among closely related plant species, and the resulting intermediate traits may complicate species delimitation and identification. Such taxonomic uncertainty can lead to misinformed conservation actions or poor resource management. Aquilegia incurvata is a Chinese endemic species restricted to the Qinba Mountains. Despite numerous field surveys in its reported distribution area, over a five-year period, no specimens fitting its description in “Flora Tsinlingensis” have been found, casting doubts on the validity of this species. To determine whether A. incurvata truly exists, we examined the type specimens of A. incurvata (including holotype and paratypes), as well as fresh material from four populations of Aquilegia occurring at the type locality. We also conducted morphological and genetic analyses to explore species boundaries between A. incurvata and four closely related species (A. ecalcarata, A. kansuensis, A. yabeana and A. yangii). Additionally, we integrated morphological and genetic data from 12 previously reported populations of these four closely related species. We performed morphological clustering via principal component analysis (PCA), as well as phylogenetic and population genetic analyses. Eleven floral traits from six type specimens and four populations were quantified for interspecific differentiation. Strict filtering of whole-genome resequencing data yielded 9,073 high-quality SNPs and 20 cpDNA loci from A. incurvata’s type specimens and its closely related species. Our study revealed two key findings: (1) The type specimens of A. incurvata are hybrids between A. kansuensis, A. yangii and A. ecalcarata, with genetic components of all three species and floral morphology intermediate between A. kansuensis and A. yangii; (2) Samples from the locus classicus and the provenances of the paratype specimens of A. incurvata cluster into two groups—One corresponds to A. yabeana and the other one to A. kansuensis. Our study confirms A. incurvata is not an independent species but a local hybrid. This paper is dedicated to the memory of Prof. Yi Ren (1959–2019), who passed away shortly before the submission of this manuscript. We will always remember him for his optimism, bravery and wisdom. Introduction Hybridization is widespread in plants, presenting substantial challenges for species delimitation (Naciri and Linder 2015; Hong 2016; Niu et al. 2018). For instance, hybridization can cause numerous continuous traits to overlap between species, blurring the boundaries between them. Additionally, a random collection of specimens may sometimes represent only the extremes within these continuous traits, or some traits can be hardly distinguishable and controversial, which leads to confusion in the recognition of distinct taxa. Fortunately, advances in sequencing technologies have facilitated the fast production of abundant molecular data, facilitating the use of neutral genetic markers to answer questions around species delimitation. In particular, species concepts such as the “integrative species concept” and “gen-morph species concept” have been proposed, providing a theoretical framework for comprehensive verification of species relationships using multiple types of data (e.g., morphological, molecular and geographical) (Hong 2016, 2020; Liu 2016; Niu et al. 2018). It has been known for decades that hybridization in Aquilegia is common (Grant 1952). Previous studies have demonstrated that postzygotic reproductive isolation among species in Aquilegia is weak, promoting frequent hybridization (Prażmo 1961; Groh and Cronk 2020). For example, Groh and Cronk (2020) have reported hybrid individuals with flower color variation in sympatric populations of the North American taxa A. flavescens and A. formosa (Groh and Cronk 2020). Hybridization has also been reported between Chinese species of Aquilegia, where frequent gene flow indicates that reproductive isolation has not yet been fully established (Geng et al. 2024; Liu et al. 2025). The Chinese endemic Aquilegia incurvata P.K.Hsiao was first described in “Flora Tsinlingensis” by Hsiao in 1974 (Instituto Botanico Boreali-Occidentali 1974). This purple-flowered species is restricted to the Qinba Mountains (comprising the Qinling Mountains and Daba Mountains) at altitudes of 1000–2800 m. This area is considered a biodiversity hotspot (Myers et al. 2000). The holotype for this species (i.e. WUK 0122218) was collected from Ningshan County, Shaanxi Province, located in the Qinba Mountains. In recent years, this area has become a research hotspot for Aquilegia (e.g. Xue et al. 2020; Xue et al. 2021; Huang et al. 2022; Geng et al. 2022; Liu et al. 2025), but few studies have focused on A. incurvata. Over the years, the occurrence of at least four other Aquilegia species with similar purple flowers has been confirmed in the Qinba Mountains (Instituto Botanico Boreali-Occidentali 1974; Luo et al. 2018): A. ecalcarata Maxim., A. yabeana Kitag., A. kansuensis (Brühl) Erst and A. yangii Luo & Li. Field surveys conducted over a five-year period, however, revealed no population of Aquilegia fully matching the morphological description of A. incurvata. Additionally, examination of morphological descriptions of A. incurvata and its four closely related species in the “Flora of China” indicates that multiple floral traits of these species show numerical overlap and continuity, which makes delimitation and taxonomic identification of these species difficult (Suppl. material 1: table SS1). For example, the sepal length of A. incurvata overlaps with A. kansuensis, A. yangii, A. yabeana and A. ecalcarata, the spur length of A. incurvata overlaps with A. kansuensis, and the length of its follicles overlaps with A. kansuensis and A. yabeana (Suppl. material 1: table SS1). Given the absence of natural populations in the wild conforming to the morphological description of A. incurvata, coupled with weak reproductive isolation between different species in Aquilegia, as well as the sympatric distribution of multiple species in the Qinba Mountains (Wang et al. 2012; Xu et al. 2019), we hypothesized that A. incurvata might be a hybrid taxon. To test this hypothesis, we employed an integrative approach combining nuclear genomic data, chloroplast loci and morphological data to investigate species delimitation of A. incurvata. Our aims were: (1) to determine whether genetics and morphological evidence support A. incurvata as an independent species; (2) to clarify the taxonomic identity of the type specimens (including holotype and paratypes) and fresh specimens collected from the type locality. Materials and methods Taxon sampling We undertook a detailed morphological study of the holotype specimen of A. incurvata and eight paratypes (Table 1; Suppl. material 1: fig. S1). Floral traits were measured for the holotype (S_Typus) and five paratypes (S_01, S_04, S_05, S_06 and S_07), for which floral trait data were available and sufficient. Concurrently, with permission from Northwest A&F University herbarium (WUK), we removed small leaf fragments from the holotype (S_Typus) and two paratypes (S_01 and S_02) for genetic analysis, ensuring there was no compromise to the specimens’ integrity or future studies. Table 1.Collection details for the type material of A. incurvata included in this study at the herbarium CODE.No.Collection informationElevation/mCollection DateType Specimens StatusVoucherGenome sequenceTraits measurementS_TypusGuankou, Ningshan, Shaanxi, China1050May 14, 1959HolotypeWUK 0122218√√S_01Dongan, Chengkou, Sichuan, China1250July 2, 1959ParatypeWUK 0423362√√S_02Taohe, Ankang, Shaanxi, China1920July 29, 1959ParatypeWUK 0423366√–S_03Sijihe, Langao, Shaanxi, China1950July 18, 1959ParatypeWUK 0133927––S_04Huangguan, Shiquan, Shaanxi, China1020June 19, 1959ParatypeWUK 0123903–√S_05Jiangkou, Shiquan, Shaanxi, China1700June 13, 1959ParatypeWUK 0122785–√S_06Taohe, Ankang, Shaanxi, China1800July 22, 1959ParatypeWUK 0137389–√S_07Longgangwan, Chengkou, Sichuan, China1900July 16, 1959ParatypeWUK 0137370–√S_08Gaojiaba, Foping, Shaanxi, China1300July 14, 1952ParatypeWUK 0059584–– Collection details for the type material of A. incurvata included in this study at the herbarium CODE. In 2018, we conducted field trips to the localities from where the holotype and paratypes of A. incurvata originated and collected fresh samples for morphological and genetic analyses from four populations (SN61, SN65, SN66 and SN67) (Tables 2, 3; Figs 1, 2). Five to fifteen flowers were collected per population (50 total) and fixed in FAA solution (70% alcohol: formalin: acetic acid = 90: 5: 5 by volume) for later examination. To ensure a comprehensive sampling of the four populations, we collected samples at the same altitudes recorded for the type specimens and across an extended altitudinal range. Based on morphological traits and description in “Flora of China”, plants from populations SN61 and SN65 were identified as A. yabeana and samples from the populations SN66 and SN67 as A. kansuensis (Fig. 1). The floral and androecium morphology of A. yabeana, A. yangii, A. kansuensis and A. ecalcarata, and samples from four populations (SN61, SN65, SN66 and SN67) found at the Qinba Mountains, where the holotype (S_Typus) and paratypes of A. incurvata were collected in 2018. Distribution of the type specimens of A. incurvata and populations of its closely related species A. yabeana, A. yangii, A. kansuensis and A. ecalcarata. The labels in black font represent extant populations, and those in red font indicate the type specimens of A. incurvata. The topographic map is from http://srtm.csi.cgiar.org/. Table 2.Sample information for 16 populations of Aquilegia yabeana, A. yangii, A. kansuensis and A. ecalcarata included in this study.No.Population(s)TaxonLocalityLatitude, LongitudeAltitude (m)Number of individuals1SN61 A. yabeana Ningshan County, Shaanxi 33°23'18"N, 108°24'46"E 1934*32SN65 A. yabeana Ningshan County, Shaanxi 33°36'14"N, 108°23'4"E 1356*53SN01 A. yabeana Mei County, Shaanxi 34°5'16"N, 107°42'28"E 118254SN11 A. yabeana Ningshan County, Shaanxi 33°26'6"N, 108°26'49"E 158255QH03 A. yangii Tu Autonomous County of Huzhu, Qinghai 36°55'8"N, 102°24'57"E 251756SN05 A. yangii Feng County, Shaanxi 34°12'0"N, 106°35'45"E 157857GS11 A. yangii Zhang County, Gansu 34°38'2"N, 104°28'26"E 245158SN23 A. kansuensis Nanzheng District, Shaanxi 32°48'0"N, 106°39'16"E 156259HB01 A. kansuensis Xingshan County, Hubei 31°26'2"N, 110°21'46"E 1608512CQ04 A. kansuensis Chengkou County, Chongqing 32°2'34"N, 108°50'9"E 2245511SN21 A. kansuensis Zhenping County, Shaanxi 32°1'1"N, 109°18'36"E 2079510SN66 A. kansuensis Taibai County, Shaanxi 33°43'54"N, 107°28'6"E 1131*513SN67 A. kansuensis Langao County, Shaanxi 32°06'22"N, 108°51'37"E 1630*514SN03 A. ecalcarata Mei County, Shaanxi 34°0'46"N, 107°48'32"E 2770515SN08 A. ecalcarata Hu County, Shaanxi 33°49'22"N, 108°36'18"E 2439516SN12 A. ecalcarata Ningshan County, Shaanxi 33°28'18"N, 108°29'46"E 21675 Sample information for 16 populations of Aquilegia yabeana, A. yangii, A. kansuensis and A. ecalcarata included in this study. Table 3.Principal component analysis (PCA) for the floral traits.Eigen value6.7752.1061.227Proportional variation (%)61.58919.14811.159Cumulative variation (%)61.58980.73691.895 Traits PC1 PC2 PC3 Petal blade width 0.949 0.120.198Sepal length 0.941 0.2370.097Petal length 0.93 0.3250.039Sepal width 0.922 -0.270.15Petal blade length 0.888 -0.162-0.281Spur aperture 0.843 0.4080.231Spur arc length 0.825 0.543 0.09Petal blade length/width -0.537 -0.501 -0.528 Sepal length/width-0.099 0.944 -0.141Petal length/spur length-0.288 -0.809 -0.31Color of dehiscent anthers-0.0020.006 0.951 The bold values are the highest absolute values among 4 principal components (PCs) for each. Principal component analysis (PCA) for the floral traits. The bold values are the highest absolute values among 4 principal components (PCs) for each. To clarify the species delimitation of A. incurvata and its relatives, we also incorporated samples from another 12 populations of the four closely related species (Figs 1, 2; Table 2), whose distributions overlap with the documented distribution of A. incurvata. Floral morphology and genomic data for these 12 populations, along with genomic data from the four populations sampled in 2018 (SN61, SN65, SN66 and SN67), were sourced from Xue et al. (2020) and our own contributions (Geng et al. 2022). Genome sequencing, assembly and gene annotation DNA extraction method and genome skimming approach followed Zeng et al. (2018). This generated approximately 8 Gb of sequencing data per sample with an average depth of 8×. Raw data processing adhered to the pipeline reported in Geng et al. (2022). For populations SN61, SN65, SN66 and SN67, single-nucleotide polymorphisms (SNPs) were called from raw genomic data using the bioinformatics pipeline described in Geng et al. (2022). For chloroplast DNA data, we used A. ecalcarata (GenBank accession: MK569475) (Zhai et al. 2019) as the reference to assemble sequences, with gaps repaired using MITObim v1.9 (Hahn et al. 2013). Annotation was performed in Geneious 10.2.2 (http://www.geneious.com/). Due to incomplete genetic information from type specimens, complete plastid genome data could not be assembled; thus, we screened for fragments that could yield full-length sequences. Twenty chloroplast DNA fragments (i.e., psbE-petL, psaI-accD, ndhC-trnV, ndhJ-trnF, atpH-atpI, rps16-trnQ, trnK-rps16, ndhA intron, rpl32-trnL, trnS-trnG, trnC-rpoB, trnD-psbM, ycf6-psbM, trnC-ycf6, trnD-trnT, trnS-trnG, rps4-trnF, rps12-rpl20, psbB-psbH, and matK) (Fior et al. 2013) were obtained by sequence alignment in Geneious 10.2.2. Sequence alignments were conducted in MAFFT 7.3 using default parameters (Katoh and Standley 2013). For the nuclear genome sequence data of all samples, we used A. oxysepala var. kansuensis as a reference (Xie et al. 2020). This species has been identified as A. yangii by Luo et al. (2018), with a genome size of 293.21 Mb and seven chromosomes. Clean reads were then mapped to this reference genome using BWA-MEM v.0.7.15 with default parameters (Li and Durbin 2009). After alignment, mapped reads were sorted and duplicates removed using SAMtools v.1.3.1 (Li 2011), with SNPs called and converted to the Variant Call Format. Analyzing the SNPs filtered with different parameter combinations, we determined that only the combination ‘--maf 0.01 --max-meanDP 30 --minDP 5 --minQ 25 --max-missing 1 --remove-indels’ (where “maf” denotes minor allele frequency) yielded SNPs that constructed the most robust phylogenetic tree with the highest support values at branch nodes. Due to incomplete genetic data generated from type specimens, long contiguous and complete sequences were not obtained. Finally, a total of 9,073 SNPs were retained from the initial 16,200,122 SNPs. Phylogenetic and population genetic analyses Three genetic datasets were used in this study, as follows: Multi-locus cpDNA dataset of Aquilegia , including 20 highly variable chloroplast loci (psbE-petL, psaI-accD, ndhC-trnV, ndhJ-trnF, atpH-atpI, rps16-trnQ, trnK-rps16, ndhA intron, rpl32-trnL, trnS-trnG, trnC-rpoB, trnD-psbM, ycf6-psbM, trnC-ycf6, trnD-trnT, trnS-trnG, rps4-trnF, rps12-rpl20, psbB-psbH and matK) (Fior et al. 2013) from 82 individuals of Aquilegia genus and Semiaquilegia adoxoides (DC.) Makino. These individuals included the holotype (S_Typus) of A. incurvata, a representative individual of A. kansuensis (CQ0428), as well as 80 individuals of Aquilegia species used in Fior et al. (2013) and outgroup S. adoxoides. The dataset was constructed to resolve the phylogenetic relationship of A. incurvata within Aquilegia. Given that cpDNA is maternally inherited (Jeffrey et al. 2018), haploid and non-recombined, these sequences were concatenated for all individuals before analyses. Multi-locus cpDNA dataset of Aquilegia , including 20 highly variable chloroplast loci (psbE-petL, psaI-accD, ndhC-trnV, ndhJ-trnF, atpH-atpI, rps16-trnQ, trnK-rps16, ndhA intron, rpl32-trnL, trnS-trnG, trnC-rpoB, trnD-psbM, ycf6-psbM, trnC-ycf6, trnD-trnT, trnS-trnG, rps4-trnF, rps12-rpl20, psbB-psbH and matK) (Fior et al. 2013) from 82 individuals of Aquilegia genus and Semiaquilegia adoxoides (DC.) Makino. These individuals included the holotype (S_Typus) of A. incurvata, a representative individual of A. kansuensis (CQ0428), as well as 80 individuals of Aquilegia species used in Fior et al. (2013) and outgroup S. adoxoides. The dataset was constructed to resolve the phylogenetic relationship of A. incurvata within Aquilegia. Given that cpDNA is maternally inherited (Jeffrey et al. 2018), haploid and non-recombined, these sequences were concatenated for all individuals before analyses. Multi-locus cpDNA dataset ofA. incurvataand its relatives, including 20 highly variable chloroplast loci of the holotype (S_Typus) of A. incurvata and 79 individuals from its relatives (Table 2), which included six populations of A. kansuensis, three populations of A. yangii, four populations of A. yabeana and three populations of A. ecalcarata, as well as the published plastid genome of A. coerulea (GenBank accession: MK569474), which was used as the outgroup. Because the data quality of the chloroplast locus of two A. incurvata specimens were poor (i.e. S_01 and S_02), their data were not included in this data set. Multi-locus cpDNA dataset ofA. incurvataand its relatives, including 20 highly variable chloroplast loci of the holotype (S_Typus) of A. incurvata and 79 individuals from its relatives (Table 2), which included six populations of A. kansuensis, three populations of A. yangii, four populations of A. yabeana and three populations of A. ecalcarata, as well as the published plastid genome of A. coerulea (GenBank accession: MK569474), which was used as the outgroup. Because the data quality of the chloroplast locus of two A. incurvata specimens were poor (i.e. S_01 and S_02), their data were not included in this data set. Genome-wide SNPs dataset, including 9,073 genome-wide SNPs from three type specimens (S_Typus, S_01 and S_02) of A. incurvata and 79 individuals of its relatives, which included six populations of A. kansuensis, three populations of A. yangii, four populations of A. yabeana, three populations of A. ecalcarata and A. oxysepala (JL01), which was used as the outgroup. Genome-wide SNPs dataset, including 9,073 genome-wide SNPs from three type specimens (S_Typus, S_01 and S_02) of A. incurvata and 79 individuals of its relatives, which included six populations of A. kansuensis, three populations of A. yangii, four populations of A. yabeana, three populations of A. ecalcarata and A. oxysepala (JL01), which was used as the outgroup. Phylogenetic consensus trees of the two multi-locus cpDNA and the genome-wide SNPs dataset were constructed by the maximum-likelihood (ML) method and applying a Bayes Test (Anisimova et al. 2011) using IQ-TREE 1.6.8 (Nguyen et al. 2015). ML analyses employed 50,000 replicates with the SH-like approximate likelihood ratio test (SH-aLRT) (Guindon et al. 2010). Based on the Bayesian Information Criterion (BIC) score calculated by ModelFinder (Kalyaanamoorthy et al. 2017) in PhyloSuite v1.1.16 (Zhang et al. 2020), the best-fit models for the three datasets were revealed: TVM+R3+F (multi-locus cpDNA dataset of Aquilegia), F81+I+F (multi-locus cpDNA dataset of A. incurvata and its relatives) and TVM+R3+F (Genome-wide SNPs dataset). For population genetic inference, we used Plink v1.07 (Purcell et al. 2007) to generate input files for ADMIXTURE (Alexander et al. 2009). The species complex structure was investigated in ADMIXTURE with four independent runs per K value (K = 2-10). Log likelihood values were used to assess convergence across runs, retaining the run with the lowest log likelihood. The optimal number of genetic clusters (K) was determined by selecting the value with the lowest cross-validation (CV) error, as this best describes the population structure (Alexander et al. 2009). Barplots of the genetic assignments of the different individuals were illustrated using an R script (v3.6.3). Morphological measurement and analysis To explore the morphological variation between A. incurvata and its relatives, floral traits were measured from six type specimens (S_Typus, S_01, S_04, S_05, S_06 and S_07) of A. incurvata, two populations of A. yabeana (SN61 and SN65, n = 5 and 15) and two of A. kansuensis (SN66 and SN67, n = 15 and 15), which were collected from the locality where the holotype (S_Typus) and paratypes of A. incurvata were gathered, respectively. A total of 11 floral traits were measured per sample, including ten quantitative traits: sepal length, sepal width, sepal length/width ratio, petal length, petal length/spur length ratio, petal blade length, petal blade width, petal blade length/width ratio, spur arc length and spur aperture; one qualitative trait: color of dehiscent anthers (Suppl. material 1: table SS2) (Data S1). Floral trait measurements followed the protocol described in our previous study (Xue et al. 2020). Concurrently, we retrieved the data of the 11 floral traits of other Aquilegia species used in this study (Xue et al. 2020). Before analyses, Z-score was used to normalize the data and a principal component analysis (PCA) was conducted using SPSS software (SPSS v22.0 for Windows, SPSS Inc., Chicago, IL, USA). The 3D-PCA plot and the violin plots were generated using 3D Scatter tools and Violin tools in Hiplot (https://hiplot.com.cn). Result Genome resequencing and chloroplast loci features The combined length of 20 chloroplast loci from the holotype (S_Typus) of A. incurvata was 20,602 bp (GC content: 33.8%). In the multi-locus cpDNA dataset of Aquilegia samples, sequence lengths ranged from 18,136 to 20,602 bp (GC content: 33.6–33.9%), and the length of the aligned dataset was 21,254 bp. For the 20 loci cpDNA dataset of A. incurvata and its relatives, the sequence lengths ranged from 20,272 to 20,602 bp (GC content: 33.7–33.8%) and the aligned matrix measured 20,421 bp. From an initial set of 16,200,122 single-nucleotide polymorphisms (SNPs) generated via whole-genome resequencing, a total of 9,073 high-quality SNPs were retained after stringent filtering. Phylogenetic relationships of cpDNA datasets To resolve the phylogenetic position of A. incurvata within the genus Aquilegia, we reconstructed the phylogenetic tree using the multi-locus cpDNA dataset (Suppl. material 1: fig. S2). The topology of the inferred tree was consistent with the phylogenetic tree reported by Fior et al. (2013). The holotype (S_Typus) of A. incurvata clustered within a monophyletic group alongside five species endemic to China (A. yabeana, A. yangii, A. kansuensis, A. ecalcarata, and A. rockii). Four out of these five species (excluding A. rockii) exhibited overlapping distributions with A. incurvata in Shaanxi Province, Chongqing City and Gansu Province, belonging to the Qinba Mountains. To further resolve the phylogenetic relationship of A. incurvata and its sympatric relatives, we inferred a phylogenetic tree using the 20 loci cpDNA dataset comprising the type specimen of A. incurvata and 78 individuals from its four relatives, employing ML/aBayes methods (Suppl. material 1: fig. S3). Phylogenetic analysis revealed that A. incurvata and its relatives diverged into two clades. One clade consisted of A. yangii and a single A. kansuensis individual (SN66-18), mainly distributed in the Qilian Mountains. The remaining individuals, including the type specimen of A. incurvata, formed the other clade, mainly distributed in the Qinba Mountains and the northern mountainous region of the Wuling Mountains. None of the five species formed a monophyletic group. The chloroplast haplotypes of A. incurvata haplotype (S_Typus) and some individuals of A. kansuensis (SN67, SN23 and HB01) and A. ecalcarata (SN08) clustered into a strongly supported monophyletic clade. Phylogenetic relationships and population structure of Genome-wide SNP datasets The ML/aBayes phylogenetic tree inferred from the 9,073 SNPs dataset of the nuclear genome showed that the type specimens of A. incurvata and its relatives diverged into six clades and one independent branch containing only the A. incurvata holotype (S_Typus) (Fig. 3). Clade I and Clade II each included all of the individuals of A. ecalcarata and A. yangii, respectively. Clade III included two populations of A. kansuensis (SN23 and SN66). Clade IV included two paratypes of A. incurvata (S_01 and S_02) and four additional populations of A. kansuensis. Clade V and Clade VI included all populations of A. yabeana. In the phylogeny, S_Typus was located between Clade III and Clade IV (Fig. 3). Phylogenetic relationships of A. incurvata and its closer relatives A. yabeana, A. yangii, A. kansuensis and A. ecalcarata based on a genome-wide SNPs dataset. The numbers by the branches represent ML bootstrap values / aBayes values. ADMIXTURE cross-validation errors (CV) suggested that K = 4 had the lowest CV error, designating it as the optimal K value (Suppl. material 1: fig. S4). At K = 2, all populations of A. yabeana clustered into a single group, whose genetic composition was dominated by the genetic component represented by the yellow-colored segment in the ADMIXTURE plot (Fig. 4); this group corresponded to Clade V and Clade VI in Fig. 3. All other samples formed a separate cluster with a distinct genetic component. At K = 3, A. ecalcarata formed an independent cluster, whose primary genetic component was indicated by the purple-colored segment in the plot and corresponded to Clade I in Fig. 3. Meanwhile, the three type specimens (S_Typus, S_01 and S_02) of A. incurvata exhibited a genetic composition consisting of the two components marked by purple and blue segments in the plot—this profile was similar to the genetic makeup of most individuals of A. kansuensis (Fig. 4). When K = 4 (the optimal value), a new genetic component (derived from the one previously marked by the blue segment) emerged, which was visualized as the brown-coloured segment in the plot (Fig. 4). This brown-marked component corresponded to Clade II in Fig. 3 and represented the unique genetic signature of A. yangii (except for one individual from QH03, whose genetic composition contained a large proportion of the component characteristic of A. ecalcarata). At this optimal K value, the four relatives of A. incurvata (A. kansuensis, A. yangii, A. yabeana and A. ecalcarata) were fully distinguished by their distinct genetic component profiles (Fig. 4). However, the genetic composition of the A. incurvata type specimens was similar to that of one population (SN66) of A. kansuensis, with both containing three genetic components (shown as purple, blue and brown segments in the plot). At K = 5, only intraspecific genetic structure increased—specifically, the A. yabeana population (SN01) separated into a distinct cluster (Fig. 4). Higher K values thereafter generated redundant clusters that explained minimal additional genetic variation. Population genetic structure analysis of A. incurvata type material and affin species A. yabeana, A. yangii, A. kansuensis and A. ecalcarata at K = 2–5 on a genome-wide SNPs dataset. The optimal K value is 4. Each vertical bar represents an individual. Statistical analyses of the morphological dataset The most informative principal components (PCs) with eigenvalue > 1.0 and the vector loadings of 11 floral traits of A. incurvata and its relatives explained this percentage change (Table 3). The first three PCs cumulatively explained 91.895% of the total variation. PC1 (61.589% of variation) was mainly comprised of petal blade width, sepal length, petal length, sepal width, petal blade length, spur aperture, spur arc length and petal blade length/width. PC2 (19.148% of variation) was mainly comprised of spur arc length, petal blade length/width, sepal length/width, and petal length/spur length. PC3 (11.159% of variation) was mainly composed of the color of dehiscent anthers. PC1 was interpreted as reflecting floral size; PC2 was floral morphological proportions; PC3 was the color of dehiscent anthers. The 3D-PCA plot, excluding the type specimens of A. incurvata, revealed that A. kansuensis, A. yangii, A. yabeana and A. ecalcarata were clustered into four groups, respectively, each forming a separate group with notable floral trait differences (Fig. 5). However, the six type specimens (S_Typus, S_01, S_04, S_05, S_06 and S_07) of A. incurvata did not cluster together. They were instead scattered around A. kansuensis and A. yangii, linking the two species to form a large group. Furthermore, the holotype (S_Typus) of A. incurvata showed greater similarity to A. kansuensis. When compared to A. kansuensis and A. yangii, the flower size of the holotype was between the two species on PC1; on PC2, its floral morphological proportions did not differ substantially from those of A. kansuensis, A. yangii and A. yabeana; and on PC3, the holotype was the most similar to A. kansuensis (Fig. 5). That is, the color of dehiscent anthers of A. incurvata was black, which was different from the yellow of A. yangii. Four populations (SN61, SN65, SN66 and SN67) sampled from the locus classicus and the provenances of the paratype specimens of A. incurvata in the Qinling Mountains were identified as: SN61 and SN65 belonged to A. yabeana, whereas SN66 and SN67 belonged to A. kansuensis. This classification was consistent with their phylogenetic traits and morphological results (Fig. 3 and Fig. 5). To further validate the morphological similarity between A. incurvata type specimens and their relatives, independent samples t-tests were conducted. The results revealed significant differences in most floral traits between the compared groups (Fig. 6). Interestingly, the group of A. incurvata type specimens showed no significant differences in eight quantitative traits when compared to A. yangii, except for the color of dehiscent anthers (Fig. 6). For two key quantitative traits (sepal length/width and petal blade length/width), no statistically significant differences were detected when the A. incurvata specimen group was compared against three distinct groups: the A. kansuensis group excluding SN66 and SN67, the A. kansuensis group restricted to SN66 and SN67, and the A. ecalcarata group (Fig. 6). In addition, dehiscent anthers were consistently black in color across all these groups (Fig. 7). For the holotype of A. incurvata, only petal blade length was shorter, whereas all other quantitative traits fell within or between the range of those in A. kansuensis (except SN66 and SN67) and A. yangii. Furthermore, eight quantitative traits showed no significant differences between A. kansuensis (SN66 and SN67) and other A. kansuensis populations, and the color of dehiscent anthers was black in both (Figs 1, 6). Conversely, four quantitative traits did not differ significantly between the group of A. yabeana (SN61 and SN65) and other A. yabeana populations, and the color of dehiscent anthers was yellow in both (Figs 1, 6). 3D-PCA plot of Principal Component Analysis (PCA) results based on the morphological measurements of 22 individuals representing A. incurvata and related species A. yabeana, A. yangii, A. kansuensis and A. ecalcarata. Violin plots of ten quantitative traits of A. incurvata and related species A. yabeana, A. yangii, A. kansuensis and A. ecalcarata. ‘ns’ means no significant difference. Anther colour in the type of A. incurvata and related species A. yabeana, A. yangii, A. kansuensis and A. ecalcarata observed from specimens at the Plant Herbarium of Northwest Agriculture and Forestry University (NWFC). The specimens of A. yabeana, A. yangii, A. kansuensis and A. ecalcarata were all collected in the same year or at least 20 years before the holotype of A. incurvata. The color of the dehiscent anthers of these observed specimens was consistent with the field observations, indicating that the color of the dehiscent anthers of the herbarium specimens was not affected by the year. The color of the dehiscent anthers of the holotype of A. incurvata is black. Discussion This study utilized multiple datasets (nuclear genome data, chloroplast loci and morphological data) to investigate species delimitation of A. incurvata. When specimens representing sites recorded in the holotype and type specimens of A. incurvata were excluded, nuclear genome SNP phylogenetic analysis clearly resolved A. kansuensis, A. yangii, A. yabeana and A. ecalcarata. Moreover, PCA analysis of floral traits demonstrated pronounced morphological differences among these four related species. When including the type specimens of A. incurvata and the four populations (SN61, SN65, SN66 and SN67) collected from the sampling sites of A. incurvata specimens, genetic and morphological analysis revealed the complex genetic and morphological relationships between A. incurvata and its four relative species. Given that the type specimen is the most basic and most important basis for the delimitation of a species in the taxonomic concept, the holotype and paratypes of A. incurvata were initially subjected to detailed study. Admixture analysis revealed that the genetic compositions of S_Typus, S_01 and S_02 remained similar and admixed across all K values but K = 2 (Fig. 4). Especially at the optimal K value of 4, these three type specimens of A. incurvata were each composed of the genetic compositions from A. kansuensis, A. yangii and A. ecalcarata. Populations of the latter three species exhibit admixed genetic compositions, and interspecific hybridization has also been reported in previous studies (Geng et al. 2024; Liu et al. 2025; Peng et al. 2025; Xue et al. 2021). In this study, it was found that the three type specimens of A. incurvata did not form a monophyletic group but instead exhibited a polyphyletic relationship with A. kansuensis in the phylogenetic analysis of nuclear genome SNPs. Meanwhile, the phylogenetic analysis of the chloroplast loci showed that the holotype of A. incurvata shared the closest ancestral haplotype with certain populations of A. kansuensis and A. ecalcarata (Suppl. material 1: fig. S3). Therefore, genetic evidence indicated that the type specimens of A. incurvata likely represent hybrid individuals among A. kansuensis, A. yangii and A. ecalcarata. Furthermore, PCA analysis of floral traits showed that the quantitative floral morphological variation of holotype and paratypes of A. incurvata fell within the range of A. kansuensis and A. yangii, exhibiting an intermediate state between the two species. Additionally, these type specimens showed no significant differences in most floral quantitative traits when compared to A. kansuensis, A. yangii and A. ecalcarata, supporting the hypothesis that the type specimens of A. incurvata represent hybrid individuals. The continuous floral morphological variation likely reflects the manifestation of genetic admixture in floral traits. Results of molecular and morphological analyses did not cluster samples from the populations SN61, SN65, SN66 and SN67 together with the type specimens of A. incurvata, but separated them into two groups. Group 1 (SN61 and SN65) was genetically and morphologically identified as A. yabeana. Nuclear genome SNPs analysis showed genetic composition identical to A. yabeana, while the 3D-PCA plot of floral traits clustered these populations together, with dehiscent anthers consistently yellow. Group 2 (SN66 and SN67) was genetically and morphologically assigned to A. kansuensis, exhibiting genetic similarity to A. kansuensis in SNP analyses, clustering together in floral trait 3D-PCA and displaying black dehiscent anthers. The main distribution area of A. incurvata is the Qinling Mountains and Bashan Mountains, which constitute a biodiversity hotspot (Myers et al. 2000). The Qinba Mountains served as refugia for numerous species during the Quaternary glacial period and functioned as corridors for species migration and convergence (Hoorn et al. 2013; Qiu et al. 2011; Zhou et al. 2010), and different species frequently come into contact here, facilitating hybrid events and generating numerous hybrid offspring with admixed genetic backgrounds. It is well known that postzygotic reproductive isolation among species of Aquilegia is very weak (Prażmo 1961; Groh and Cronk 2020). Aquilegia ecalcarata and A. chrysantha Gray, which diverged long ago and are distributed in Asia and North America, can be artificially crossed (Prażmo 1961). In our study, A. kansuensis and A. yangii co-occur in the Qinba Mountains, enabling hybridization and backcrossing in their contact zones. Meanwhile, A. ecalcarata is also found in the Qinba Mountains, further facilitating interspecific hybridization. Furthermore, pollination studies of Aquilegia in the Qinba Mountains found that bumblebees serve as pollinators for spurred species (A. incurvata and A. yabeana) (Tang et al. 2007), suggesting that spurred species (A. kansuensis and A. yangii) in this region likely share similar pollinators. Thus, we hypothesize that the combined factors of weak postzygotic reproductive isolation, potentially shared pollinators and sympatric distribution promote extensive hybridization among A. kansuensis, A. yangii and A. ecalcarata in the Qinba Mountains, giving rise to numerous transitional forms. The holotype and paratypes of A. incurvata likely represent such a hybrid example, being similar to the natural populations exhibiting admixed genetic compositions of A. kansuensis, A. yangii and A. ecalcarata (Geng et al. 2024; Liu et al. 2025; Peng et al. 2025; Xue et al. 2021). In summary, multiple pieces of evidence from this study show that these morphological traits, once used as diagnostic criteria, are not exclusive to A. incurvata; instead, they represent phenotypic reflections of introgression between three sympatric congeneric species—A. kansuensis, A. yangii and A. ecalcarata. This finding reveals the key limitation of relying solely on single morphological traits for species delimitation in taxa with frequent hybridization, such as the Aquilegia: the intermediate morphologies previously regarded as “unique to new species” may actually be the result of continuous gene flow among existing species (Sheidai et al. 2018; Chang et al. 2022; Webster et al. 2025). For taxonomic practice, this work suggests that when describing new species within Aquilegia or other taxa with weak postzygotic reproductive isolation (Prażmo 1961; Groh and Cronk 2020), a comprehensive delimitation approach should be adopted by integrating phylogenetic relationships, population-level sampling data and morphological traits (Liu 2016; Hong 2020). This integrative taxonomic method not only avoids the issue of “taxonomic inflation” caused by misidentifying hybrid individuals as new species but also ensures that taxonomic treatment results are consistent with the evolutionary nature of species boundaries (Young et al. 2019; Mathews et al. 2025). Thereby, it guarantees that conservation efforts (e.g., priority conservation of endangered species) are not misdirected toward transient hybrid taxa.