A Phylogenetic and Morphological Evolution Study of Ribes L. in China Using RAD-Seq

Ribes L. belongs to the Grossulariaceae family and has important edible, medicinal, ornamental, and landscaping values. Taxonomic classification within this genus is difficult due to its large variety of species, wide distribution, large morphological variations, and presence of two complex taxonomic groups with bisexual or unisexual flowers. Our study aims to clarify the phylogenetic relationships of Ribes L. taxa in China, and further, to provide a reference for a revised global classification of it. The phylogenetic analysis of 52 Ribes L. samples from 30 species was constructed based on restriction site-associated DNA sequencing and single nucleotide polymorphisms. Afterward, two important taxonomic characters were selected for ancestral state reconstruction over the molecular phylogeny. The results showed that the 52 samples could be divided into six branches, i.e., six subgenera, which caused some controversy regarding the morphological classification of Ribes L. in China. The molecular phylogeny supported the separation of subg. Coreosma from subg. Ribesia and subg. Hemibotrya from subg. Berisia and validated the rationale for recognizing subg. Grossularia as an independent subgenus, the rationality of which was further verified by the reconstruction of ancestor traits. Gene flow among Ribes L. was identified and further confirmed our results.


Introduction
There are approximately 200 species of Ribes L. in the world, and they are primarily distributed in East Asia, North America, and the Andes of South America [1]. China is a distribution centre of Ribes L., with approximately 59 species and 30 varieties [2], including the main taxa in phylogenetic development. The economic value of this genus is high enough, as its fruits can be eaten raw or used to produce fruit wine, beverages, candies, and jams since they are rich in various vitamins, sugars, and organic acids [3,4].
Meanwhile, the roots and seeds of some species can also be used for medicinal purposes, while some species have high ornamental value due to their bright flowers and attractive fruits [5]. Therefore, clarifying the phylogenetic relationship within Ribes L. would be beneficial for the protection and rational use of the plant resources in this genus.
Linnaeus initially established the genus Ribes L. into eight species in 1753. Throughout the years, the classification of taxa within this genus remains a hot topic of debate [6][7][8][9][10]. Janczewski [6] classified Ribes L. gathered across the world into six subgenera: Grossularioides, Parilla, Berisia, Coreosma, Ribesia, and Grossularia, according to the characteristics of the flower, including sexuality, the texture of the bud scales as well as the presence or absence of a pedicel, pedicel joints, and thorns on the branches. Rehder [11] categorized this genus into four subgenera and 15 sections; in his classification, subg. Coreosma and subg. Parilla were placed in subg. Ribesia and subg. Berisia, respectively, which were different from the classifications by Janczewski. Berger [8] divided Ribes L. into eight subgenera: Grossularioides, Parilla, Berisia, Coreosma, Ribesia, Calobotrya, Heritiera, and Symphocalyx and

RAD-Tag Sequencing and SNP Discovering
In this study, the raw data from 52 Ribes L. samples of 26 species and 4 varieties were filtered, with a total of 86.98 G of clean data obtained. We acquired 590.36 million clean reads by using Illumina HiSeq4000, after removing the low-quality reads (Q score < 20), and ambiguous reads with incorrect barcodes ( Table 1). The sequencing quality scores of 20 (Q 20 ), which represent an error rate of 1 in 100, with a corresponding call accuracy of 99%, with all samples at more than 97.72%, indicating that the sequencing quality was good. Of these high-quality reads, the highest reads (22.03 million reads) were detected in Ribes nigrum (D5), and the lowest reads (0.7 million reads) were found in R. pseudofasciculatum (B5), with an average read number of 11.35 million per accession.

Phylogenetic Trees and Morphological Characteristics
Two phylogenetic trees of Ribes L. were constructed based on neighbor-joining and maximum likelihood analyses, while almost the same results were obtained. It showed that the 52 samples were clearly divided into two major branches and six subgenera, which correlated with the morphological characteristics ( Figure 1). The first branch contained subg. Ribesia, Ribes griffithii Hook. f. et Thoms., subg. Coreosma, and subg. Grossularia, and the second branch contained subg. Berisia, and subg. Hemibotrya. A total of 23 samples representing nine species were clustered into subg. Ribesia, including R. moupinense, R. setchuense, R. altissimum, R. himalense, R. longiracemosum, R. mandshuricum, R. palczewskii, R. triste, and R. atropurpureum. The common morphological characteristics of subg. Ribesia species were having a bisexual flower and a raceme. Previous studies have also classified R. griffithii into subg. Ribesia [2,7]; however, our phylogeny showed that R. griffithii formed a monophyletic clade.
The subg. Coreosma included Ribes nigrum, R. procumbens, and R. fragrans, representing a total of three species (five samples), with common morphological characteristics, such as having bisexual flowers, racemes, glands on the abaxial leaf epidermis, and purple or black fruits. The subg. Grossularia contained R. aciculare, R. burejense, and R. stenocarpum, representing a total of three species (five samples), with common morphological characteristics, such as having thorny branches and inverted sepals.
A total of 15 samples representing 12 species were clustered into subg. Berisia, including Ribes kialanum, R. tenue, R. takare var. desmocarpum, R. pseudofasciculatum, R. hunanense, R. davidii, R. laurifolium, R. laurifolium var. yunnanense, R. heterotrichum, R. giraldii var. polyanthum, R. alpinum, and R. komarovii, with common morphological characteristics such as unisexual flowers and racemes. Among them, the samples of sect. Davidia were clustered on the same subbranch, with common morphological characteristics, such as being evergreen, having unisexual flowers, and having drooping inflorescences. This result was consistent with the clustering of sect. Davidia species into a section of subg. Berisia by Lu [2]. The other branch contained two species of the subg. Hemibotrya, namely R. fasciculatum and R. fasciculatum var. chinense, with common morphological characteristics such as unisexual flowers and umbels.

Principal Component Analysis of Ribes L.
PCA, using the first and second eigenvectors, identified six groups, i.e., subg. Hemibotrya, subg. Berisia, subg. Grossularia, subg. Coreosma, Ribes griffithii and subg. Ribesia, which were consistent with the phylogenetic clades ( Figure 2A). Each group was represented by a different color. The 52 Ribes L. samples were clearly divided into two major categories. The PCA plot illustrated that the subg. Grossularia were more disperse than the other groups ( Figure 2A).

Principal component analysis of Ribes L.
PCA, using the first and second eigenvectors, identified six groups, i.e., subg. Hemibotrya, subg. Berisia, subg. Grossularia, subg. Coreosma, Ribes griffithii and subg. Ribesia, which were consistent with the phylogenetic clades ( Figure 2A). Each group was represented by a different color. The 52 Ribes L. samples were clearly divided into two major categories. The PCA plot illustrated that the subg. Grossularia were more disperse than the other groups ( Figure 2A).  Figure 2B). The PCA analysis indicated that the genetic relationship between subg. Hemibotrya and subg. Berisia was relatively close. R. griffithii, subg. Coreosma, and subg. Ribesia formed three separate groups. The results showed that R. griffithii and subg. Ribesia were relatively separate.

Population Genetic Structure of Ribes L.
Each group is represented by a different color. The 52 Ribes L. samples were clearly divided into two major categories. The PCA plot illuminated that subg. Grossularia was more dispersed than the other groups ( Figure 2A). Through population structure analysis, we could clarify how the Ribes L. species were clustered and, thus, we could understand the individual ancestries of the different species ( Figure 3). For K = 2, the 52 Ribes L. samples in the figure were clustered into two categories, and the taxa with a blue background included all the samples from subg. Ribesia. Specifically, Ribes griffithii displayed an admixture of subg. Berisia, while subg. Hemibotrya. R. griffithii had a 5% brown background hybridization, supporting the classification of R. griffithii as a monophyletic clade. The taxa with brown backgrounds were subg. Coreosma, subg. Grossularia, subg. Berisia, and subg. Hemibotrya.  Figure 2B). The PCA analysis indicated that the genetic relationship between subg. Hemibotrya and subg. Berisia was relatively close. R. griffithii, subg. Coreosma, and subg. Ribesia formed three separate groups. The results showed that R. griffithii and subg. Ribesia were relatively separate.

Population Genetic Structure of Ribes L.
Each group is represented by a different color. The 52 Ribes L. samples were clearly divided into two major categories. The PCA plot illuminated that subg. Grossularia was more dispersed than the other groups ( Figure 2A). Through population structure analysis, we could clarify how the Ribes L. species were clustered and, thus, we could understand the individual ancestries of the different species (Figure 3). For K = 2, the 52 Ribes L. samples in the figure were clustered into two categories, and the taxa with a blue background included all the samples from subg. Ribesia. Specifically, Ribes griffithii displayed an admixture of subg. Berisia, while subg. Hemibotrya. R. griffithii had a 5% brown background hybridization, supporting the classification of R. griffithii as a monophyletic clade. The taxa with brown backgrounds were subg. Coreosma, subg. Grossularia, subg. Berisia, and subg. Hemibotrya. Plants 2023, 12, x FOR PEER REVIEW 7 of 18  When K = 3, subg. Berisia and subg. Hemibotrya showed different genetic backgrounds from subg. Coreosma and subg. Grossularia, and all the samples were clearly clustered into four groups. At K = 4, hybridization occurred in some samples. For K = 5, Ribes moupinense and R. setchuense showed different pedigree compositions from those of other species in subg. Ribesia. When the K value increased, subg. Coreosma and subg. Berisia were separated, showing an independent genetic background. The optimal value of the population genetic structure analysis was at K = 7, and the ancestral compositions of R. himalense, R. mandshuricum, and R. longiracemosum differed from those of R. altissimum and R. atropurpureum, indicating that the former three species had closer genetic relationships. The results showed that subg. Ribesia has more interspecific hybridization. For K = 7, subg. Hemibotrya was separated from subg. Berisia with the hybridization in contrast to the observations at K = 4. At different K values, subg. Berisia contained a single color, confirming its independent genetic background.

Ancestral Character Reconstruction
Two key morphological characters of Ribes L. were selected to reconstruct their ancestral characters. As for the glands, we found that the lower leaf surface, calyx, ovary, and fruit of the three species of subg. Coreosma were densely covered with yellow glands, while other subgenus had no glands on the surface of plant bodies. The ancestral state reconstructions ( Figure 4A) identified glandless as the most probable ancestral state for each subgenus of Ribes L. (subg. Ribesia: p = 0.99; Ribes griffithii: p = 0.99; subg. Grossularia: p = 0.99; subg. Berisia: p = 0.99; subg. Hemibotrya: p = 0.99). However, the gland state was reconstructed as the present state in the clade of subg. Coreosma (p = 0.99), which was clearly distinguished from the other subgenera. Specifically, glandular morphology of subg. Coreosma transitioned from glandless to gland, while there was no reversion to a glandless state.
The morphological characteristics of subg. Hemibotrya were significantly different from those of subg. Berisia. In terms of inflorescence type ( Figure 4B), raceme was reconstructed as the ancestral state of Ribes L. (p = 0.99). Specifically, the inflorescences of three plants in the clade of subg. Grossularia changed from raceme (ancestral state) to short raceme independently (p = 0.98). However, the inflorescences of two plants of subg. Hemibotrya changed from raceme to umbel differently (p = 0.99). Neither subg. Hemibotrya nor subg. Grossularia reverted into the raceme. This indicates that each genus has had a high degree of independence in their own evolutionary history.

Phylogenetic Relationships and Taxonomy of Ribes L.
The subgenus classification of Ribes L. has been controversial. Janczewski [6] classified Ribes L. into six subgenera based on the comprehensive analysis of some global morphological traits of Ribes L. Indeed, Berger [8] classified Grossulariaceae into two genera, namely, Ribes L. and Grossularia, and further divided Ribes L. into eight subgenera and Grossularia into four subgenera. Rehder [11] merged subg. Coreosma into subg. Ribesia and subg. Hemibotrya into subg. Berisia, thereby dividing Ribes L. into four subgenera. Weigend [7] classified Ribes L. into seven subgenera based on the morphological and micromorphological characteristics of this genus. Lu [2] clustered Ribes L. in China into four subgenera according to plant morphological characteristics. Huang [27] also divided Ribes L. into four subgenera based on an analysis of the pollen traits for 21 taxa of the genus. In contrast  The subgenus classification of Ribes L. has been controversial. Janczewski [6] classified Ribes L. into six subgenera based on the comprehensive analysis of some global morphological traits of Ribes L. Indeed, Berger [8] classified Grossulariaceae into two genera, namely, Ribes L. and Grossularia, and further divided Ribes L. into eight subgenera and Grossularia into four subgenera. Rehder [11] merged subg. Coreosma into subg. Ribesia and subg. Hemibotrya into subg. Berisia, thereby dividing Ribes L. into four subgenera. Weigend [7] classified Ribes L. into seven subgenera based on the morphological and micromorphological characteristics of this genus. Lu [2] clustered Ribes L. in China into four subgenera according to plant morphological characteristics. Huang [27] also divided Ribes L. into four subgenera based on an analysis of the pollen traits for 21 taxa of the genus. In con-trast to previous studies, we divided 30 representative Ribes L. species in China into six branches based on phylogenetic trees by RAD-seq ( Figure 1). In summary, for the relationships between these species, all three analyses, shown above, demonstrate similar patterns (Figures 2, 3, and 5). Ribes L. was seen to have two large complex taxonomic groups with unisexual or bisexual flowers. In Figure 1, the species with unisexual flowers are clustered into two branches, i.e., subg. Hemibotrya and subg. Berisia, which differ from the conclusions of Lu [2] and Huang [27] since these are based on morphology or micromorphology. We support the viewpoints of Berger [8] and Soltis [1]. We found that three samples of subg. Hemibotrya showed a hybrid genetic background from subg. Berisia and subg. Coreosma, although subg. Hemibotrya was not clustered with subg. Berisia on the phylogenetic trees or in the PCA diagram (Figures 2 and 5). Ribes fasciculatum var. chinense and R. fasciculatum exhibited a different genetic background from other species of subg. Berisia. When K = 5, species of subg. Hemibotrya began to show hybridization. Therefore, we support the classification of subg. Hemibotrya as independent subgenus separate from subg. Berisia.
Huang [27] and Lu [2] allocated the Ser. Nigra species, namely, Ribes procumbens, R. nigrum, and R. fragrans in subg. Ribesia, thereby clustering Ribes L. into four subgenera. Berger [8] and Messinger [13] support placing these three species in subg. Coreosma, an independent subgenus. In addition, we found these three species of Ribes L. were not clustered in the same branch as other subg. Ribesia species (Figures 1 and 2). The genetic structure analysis showed that with the increase in the K value, these three Ribes L. species showed an independent genetic background distinguished from that of other subg. Ribesia species. Based on the results of the phylogenetic analysis, we support placing these three Ribes L. species in subg. Coreosma instead of subg. Ribesia. Lu [2] classified R. griffithii as subg. Ribesia. We found it differed to subg. Ribesia in terms of some important morphological characteristics, including large changes in bract morphology, an oval, or lingual line to lanceolate, bracteoles, nectaries in flowers. Furthermore, we found that R. griffithii was formed by the hybridization of two ancestral populations (Figure 3), which was significantly different from subg. Ribesia. The taxonomic status of R. griffithii in Ribes L. needs to be further studied by expanding the sampling range.
Previous studies have classified gooseberries into the genus Grossularia or subg. Grossularia due to their unique morphological characteristics [7,8]. Our phylogenetic trees showed that Ribes burejense, R. stenocarpum, and R. aciculare in subg. Grossularia have relatively close phylogenetic relationships with subg. Ribesia, R. griffithii, and subg. Coreosma. In the PCA diagram, the five gooseberry samples (C1-C5) exhibited a scattered distribution, with distances far from the samples of other subgenera. The analysis of the population genetic structure showed that the genetic backgrounds of the five samples became independent when K = 5. This result confirmed a very close phylogenetic position and the relationship between them and each subgenus. Based on the above considerations, we are in favor of placing gooseberry plants into subg. Grossularia rather than considering them as an independent genus of Grossularia.

Gene Exchange Analysis of Ribes L.
China is a primary distribution center of Ribes L. in East Asia, with approximately 59 species and 30 varieties [2]. Its species abundance, with its many endemic species and large-scale overlap of the distribution areas of some species, has led to interspecific gene exchange in plants of the same genus [28]. We found that hybridization in subg. Ribesia is more complex (Figures 3 and 5). When the K value increases, Ribes altissimum and R. atropurpureum had genetic backgrounds that differed from that of other subg Ribesia species and were not hybridized. For K = 5, R. triste and R. palczewskii exhibited a closer phylogenetic relationship, indicating the same ancestral origin. R. mandshuricum occupied three colors: blue, brown, and cyan, indicating a complex pedigree and more gene exchange with other species of Ribes L. ples of subg. Hemibotrya showed a hybrid genetic background from subg. Berisia and subg. Coreosma, although subg. Hemibotrya was not clustered with subg. Berisia on the phylogenetic trees or in the PCA diagram (Figures 2 and 5). Ribes fasciculatum var. chinense and R. fasciculatum exhibited a different genetic background from other species of subg. Berisia. When K = 5, species of subg. Hemibotrya began to show hybridization. Therefore, we support the classification of subg. Hemibotrya as independent subgenus separate from subg. Berisia.  Regarding the Ribesia species, in the previous classification results, Ribes griffithii was placed in subg. Ribesia [2,7]. In Figure 1, R. griffithii and other subg. Ribesia plants were divided into two branches. Population genetic structure showed that hybridization consistently occurred in R. griffithii (Figures 3 and 5). When K = 7, R. griffithii exhibited a histogram composed of two colors, cyan and yellow, and the genetic background of both subg. Berisia (20%) and subg. Ribesia (80%). Therefore, the sample size must be expanded for further analysis of the taxonomic position of R. griffithii.

Revision of subg. Hemibotrya
The views on the classification of Ribes fasciculatum and R. fasciculatum var. chinense seemed to differ widely among researchers [2,11]. Initially, they were divided into subg.
Berisia, yet were also clearly distinguished from the other species of subg. Berisia because of their morphological characteristics of umbels. Moreover, some researchers had suggested that these two Ribes should be subg. Parilla, separately [6,29]. However, in our study we found that R. fasciculatum and R. fasciculatum var. chinense did not cluster with other species of subg. Berisia (Figures 1 and 2). The genetic structure analysis also showed the interspecific hybridization between these two species, which was different from other species of subg. Berisia (Figures 3 and 5). Furthermore, Figure 4 also shows that the inflorescence type of these two plants had changed from racemes to umbels (p = 0.99).
In addition, Weigend thought that dividing Ribes fasciculatum distributed in East Asia into the subg. Parilla would cause a mixture of species in this subgenus; therefore, suggested that R. fasciculatum should be separated from subg. Parilla [15]. Soltis analyzed the ITS sequences of 66 species of plants in 12 subgenera of Ribes L. and found that R. fasciculatum did not cluster with other species of subg. Berisia [1], which was consistent with our results. Therefore, Soltis supported the independence of R. fasciculatum as a sister species to the other species of subg. Berisia. Combined with the results of this study, we are in favor of the classification of R. fasciculatum and R. fasciculatum var. chinense as a separate subgenus, rather than classifying them into subg. Berisia or subg. Parilla.

Plant Materials and DNA Isolation
For this study, we collected 52 representative samples of 30 Ribes L. species from 15 provinces in China ( Figure 6 and Table 2). It is worth mentioning that although some species are remote and difficult to collect, the 30 species we collected and analyzed depict nearly all the representative species of each branch in China. The plant names in this study follow the nomenclature of Flora of China [30]. Genomic DNA was extracted using the cetyltrimethylammonium bromide (CTAB) method [31]. Following quality assessment, the DNA concentration was adjusted to 100 ng/µL for RAD-seq library preparation.

RAD Library Preparation and Sequencing
A reduced representation restriction-associated DNA (RAD) sequencing method was used for library construction following the protocol previously outlined in Zhang et al. [32]. In brief, genomic DNA (1 μg) was digested by EcoRI (New England Biolabs, Ipswich, MA, USA), which recognizes the 5′-GAATTC-3′ sequence. An Illumina P1 adapter containing specific nucleotide barcodes 4-8 bp long was ligated onto the digested DNA. Then, the products from different samples were pooled and randomly fragmented by Covaris E210 (Covaris, Woburn, MA, USA) and selected on an agarose gel of 300-500 bp. The products were purified using a QIAquick PCR Purification Kit. The fragments were endrepaired with an End Repair mix, and then, purified. The repaired DNA was combined with an A-Tailing Mix, then, the Illumina P2 adapters were ligated to the adenylate 3′ ends DNA and followed by purification of the products. Several rounds of PCR amplification with a PCR Primer Cocktail and PCR Master Mix were performed to enrich the adapterligated DNA fragments. The PCR products were selected by agarose gel electrophoresis with target fragments and, afterward, purified. The library was qualified using the Agilent Technologies 2100 bioanalyzer and ABI StepOnePlus RealTime PCR System. The qualified libraries were pair-end sequenced on the HiSeq 4000 System (Illumina).

Quality Filtering and SNP Discovery
Raw sequence reads were segregated by barcodes assigned to individuals and lowquality reads, and those that lacked a correct barcode were removed [33]. The reads were first assigned to each individual by the unambiguous barcodes, and the reads without the unique barcodes were discarded. Reads were quality-filtered by removing the adapter sequences and the reads containing greater than 40% low-quality bases (quality value < 20) [34]. All reads were pooled and used for a de novo assembly and SNP calling in ustacks

RAD Library Preparation and Sequencing
A reduced representation restriction-associated DNA (RAD) sequencing method was used for library construction following the protocol previously outlined in Zhang et al. [32]. In brief, genomic DNA (1 µg) was digested by EcoRI (New England Biolabs, Ipswich, MA, USA), which recognizes the 5 -GAATTC-3 sequence. An Illumina P1 adapter containing specific nucleotide barcodes 4-8 bp long was ligated onto the digested DNA. Then, the products from different samples were pooled and randomly fragmented by Covaris E210 (Covaris, Woburn, MA, USA) and selected on an agarose gel of 300-500 bp. The products were purified using a QIAquick PCR Purification Kit. The fragments were end-repaired with an End Repair mix, and then, purified. The repaired DNA was combined with an A-Tailing Mix, then, the Illumina P2 adapters were ligated to the adenylate 3 ends DNA and followed by purification of the products. Several rounds of PCR amplification with a PCR Primer Cocktail and PCR Master Mix were performed to enrich the adapter-ligated DNA fragments. The PCR products were selected by agarose gel electrophoresis with target fragments and, afterward, purified. The library was qualified using the Agilent Technologies 2100 bioanalyzer and ABI StepOnePlus RealTime PCR System. The qualified libraries were pair-end sequenced on the HiSeq 4000 System (Illumina).

Quality Filtering and SNP Discovery
Raw sequence reads were segregated by barcodes assigned to individuals and lowquality reads, and those that lacked a correct barcode were removed [33]. The reads were first assigned to each individual by the unambiguous barcodes, and the reads without the unique barcodes were discarded. Reads were quality-filtered by removing the adapter sequences and the reads containing greater than 40% low-quality bases (quality value < 20) [34]. All reads were pooled and used for a de novo assembly and SNP calling in ustacks (STACKS v2.55 software pipeline) [35]. We set a minimum stack size of five reads (−m) and a maximum distance between stacks (−M) within a locus as two.

Phylogenetic Analysis
To construct the phylogenetic trees, the genetic distances between the different accessions were calculated based on the high-confidence SNPs extracted from the RAD data. The p-distance, defined as D ij , between two accessions (i and j), was calculated using the following equation [36]: where L is the length of regions where high-quality SNPs were identified, and d ij was defined as d ij = 1 if the genotypes at position l were AA and CC. Neighbor-joining and maximum likelihood phylogenetic trees were constructed by Treebest software, and bootstrap replicates were set to 1000 [28].
The maximum likelihood (ML) tree was constructed using MEGA 5.0 [37] with ultrafast bootstrap analysis of 1000 replicates. After comparing all the models, JC model was selected as the best-fit substitution model by model selection implemented in MEGA [38].

Principle Component Analysis
Principle component analysis (PCA) was performed using EIGENSOFT based on the SNP dataset [39]. The decomposition of the eigenvectors from the covariance matrix was performed with the R function Eigen, and the significances of the eigenvectors were further investigated with Tracy-Widom tests, using the twstats program in the Eigensoft package.

Genetic Structure Analysis
We identified the population genetic structures using ADMIXTURE 1.3 [40]. We first assumed that the value of ancestral origin is K [41]. The difference in the K value indicated whether different individuals have the same genetic background and whether the ancestral components are consistent between them. In the analysis of the population genetic structure, each individual was a column of histograms in the figure, and color differences were used to indicate the ancestral composition of different individuals. Individuals with the same color evolved from the same genetic background, and different color ratios of two individuals indicated that the pedigree compositions of the two individuals were different. If a sample always comprised a single color, then, it indicated that there was no hybridization in the sample. Conversely, if a sample comprised multiple colors, then this individual was likely to be a hybrid of several ancestral subpopulations [42]. The individual ancestry proportion was calculated 10,000 times from a given number of inferred populations (K) based on the maximum likelihood algorithm. The K values were set from two to seven. ADMIXTURE 1.3 is a calculation method based on the Bayesian model [43]. Crossvalidation error will be generated for each K value simulation result. The minimum value of cross-validation error corresponds to the optimal K value, and the fitting result at this time is closest to the real situation of the population [44].

Ancestral State Reconstruction of Morphological Characters
In this study, we selected two important taxonomic characters of interest, reported in previous studies [2,29], for ancestral state reconstruction over molecular phylogeny. We traced the evolution of two important characters of Ribes L.: 1. Gland: (0) absent (1) present; 2. Inflorescence: (0) raceme (1) umbel; (2) short raceme. The character states were optimized onto the tree generated from the RAD-seq dataset in Mesquite v. 2.73 [45] using the maximum likelihood criterion with the Markov k-state one-parameter (Mk1) model [46].

Conclusions
Our study revealed the genetic relationships of Ribes L. species in China by RAD-seq, with a total of 2,451,454 SNPs obtained from 30 species of Ribes L. Thus, these 30 species were divided into six branches, i.e., six subgenera, which were not fully consistent with the traditional phenetic sectional division of the genus. Our study was in favor of the separation of subg. Coreosma from subg. Ribesia and subg. Hemibotrya from subg. Berisia. Moreover, the results supported subg. Grossularia as an independent subgenus. Likewise, the subg. Grossularia, subg. Hemibotrya, subg. Coreosma, and subg. Berisia were verified as monophyletic groups by analyzing the genetic relationships and backgrounds of each subgenus. The subg. Ribesia had a more complex genetic background, with frequent interspecific hybridizations. Ribes griffithii, which originally belonged to subg. Ribesia, had a hybrid genetic background. However, the sample size must be expanded for further validation of the rationale to upgrade R. griffithii to an independent subgenus.

Institutional Review Board Statement:
The study did not require ethical approval.

Informed Consent Statement: Not applicable.
Data Availability Statement: Not applicable.