Infrageneric Plastid Genomes of Cotoneaster (Rosaceae): Implications for the Plastome Evolution and Origin of C. wilsonii on Ulleung Island

Cotoneaster is a taxonomically and ornamentally important genus in the family Rosaceae; however, phylogenetic relationships among its species are complicated owing to insufficient morphological diagnostic characteristics and hybridization associated with polyploidy and apomixis. In this study, we sequenced the complete plastomes of seven Cotoneaster species (C. dielsianus, C. hebephyllus, C. integerrimus, C. mongolicus, C. multiflorus, C. submultiflorus, and C. tenuipes) and included the available complete plastomes in a phylogenetic analysis to determine the origin of C. wilsonii, which is endemic to Ulleung Island, Korea. Furthermore, based on 15 representative lineages within the genus, we carried out the first comparative analysis of Cotoneaster plastid genomes to gain an insight into their molecular evolution. The plastomes were highly conserved, with sizes ranging from 159,595 bp (C. tenuipes) to 160,016 bp (C. hebephyllus), and had a GC content of 36.6%. The frequency of codon usage showed similar patterns among the 15 Cotoneaster species, and 24 of the 35 protein-coding genes were predicted to undergo RNA editing. Eight of the 76 common protein-coding genes, including ccsA, matK, ndhD, ndhF, ndhK, petA, rbcL, and rpl16, were positively selected, implying their potential roles in adaptation and speciation. Of the 35 protein-coding genes, 24 genes (15 photosynthesis-related, seven self-replications, and three others) were found to harbor RNA editing sites. Furthermore, several mutation hotspots were identified, including trnG-UCC/trnR-UCU/atpA and trnT-UGU/trnL-UAA. Maximum likelihood analysis based on 57 representative plastomes of Cotoneaster and two Heteromeles plastomes as outgroups revealed two major lineages within the genus, which roughly correspond to two subgenera, Chaenopetalum and Cotoneaster. The Ulleung Island endemic, C. wilsonii, shared its most recent common ancestor with two species, C. schantungensis and C. zabelii, suggesting its potential origin from geographically close members of the subgenus Cotoneaster, section Integerrimi.


Introduction
The genus Cotoneaster Medik. is one of the most taxonomically challenging lineages in the family Rosaceae because of apomixes, hybridization, polyploidy, and unclear species circumscription [1][2][3][4]. The genus, comprising approximately 150 species, is mainly distributed in the northern hemisphere excluding Japan, and the most important center of diversity is the Himalayas and two neighboring southwestern provinces of China, Yunnan and Sichuan [1,2,[5][6][7][8]. Owing to many complex species groups, insufficient diagnostic morphological features, and complex evolutionary processes, the infrageneric classification of Cotoneaster has been debated over the past 130 years. Koehne [9] recognized two subgenera primarily based on the petal characteristics: subgenus Chaenopetalum (spreading or rarely The chloroplast genome of angiosperms usually encodes 110-130 genes with a size range of 120-160 kb and is generally recognized as a valuable genetic resource for phylogenetic and population genetic studies [18][19][20][21][22][23]. High-throughput sequencing technologies have allowed rapid accumulation of complete plastome sequences in various plant lineages, providing opportunities for comparative analyses to gain insights into plastome organization and evolution. Indeed, comparative analyses of plastomes at various taxonomic levels have revealed their basic genomic structure, gene content, gene order, and mutation hotspots, and helped improve the present understanding of intracellular gene transfer, photosynthetic evolution in parasitic plants, insular plant evolution, and plant adaptation [24][25][26][27][28][29][30][31][32][33][34][35][36]. Since the first report on the complete plastome of Fragaria [37] and Malus [38], numerous plastome sequences in the family Rosaceae have been characterized, including Prunus L. [39,40], Pyrus L. [41], Rosa L. [30], and Rubus [27,29,34,42]. Although the plastomes of Cotoneaster wilsonii and numerous other congeneric species have been characterized and utilized for phylogenetic analysis, respectively, no attempt has been The chloroplast genome of angiosperms usually encodes 110-130 genes with a size range of 120-160 kb and is generally recognized as a valuable genetic resource for phylogenetic and population genetic studies [18][19][20][21][22][23]. High-throughput sequencing technologies have allowed rapid accumulation of complete plastome sequences in various plant lineages, providing opportunities for comparative analyses to gain insights into plastome organization and evolution. Indeed, comparative analyses of plastomes at various taxonomic levels have revealed their basic genomic structure, gene content, gene order, and mutation hotspots, and helped improve the present understanding of intracellular gene transfer, photosynthetic evolution in parasitic plants, insular plant evolution, and plant adaptation [24][25][26][27][28][29][30][31][32][33][34][35][36]. Since the first report on the complete plastome of Fragaria [37] and Malus [38], numerous plastome sequences in the family Rosaceae have been characterized, including Prunus L. [39,40], Pyrus L. [41], Rosa L. [30], and Rubus [27,29,34,42]. Although the plastomes of Cotoneaster wilsonii and numerous other congeneric species have been characterized and utilized for phylogenetic analysis, respectively, no attempt has been made to understand their genomic structure, gene order, gene contents, mutation hotspots, and positively selected plastid genes within the genus [4,28]. In this study, we characterized the complete chloroplast genome sequence of seven Cotoneaster species (C. dielsianus, C. hebephyllus, C. integerimus, C. mongolicus, C. multiflorus, C. submultiflorus, and C. tenuipes) and included them as part of a broader phylogenetic framework within the genus. To gain insights into plastome organization and evolution within Cotoneaster, we selected the major lineages within the genus and conducted comparative analyses, including codon usage, positive selection, RNA editing sites, and mutation hotspots. We further explored the phylogenetic relationship of C. wilsonii relative to other congeneric species to determine its origin and evolution on Ulleung Island, Korea. Based on the plastid phylogenomic and comparative analyses, this study provides new insights into the origin and evolution of the insular endemic C. wilsonii, as well as the overall plastome evolution within the taxonomically challenging and horticulturally important genus, Cotoneaster.

Comparative Plastome Analysis
To gain insights into plastome evolution within the genus Cotoneaster, we selected 15 taxa and compared their genomic features using the Shuffle-LAGAN mode [47] of mVISTA [48]. The 15 taxa included seven species that we newly sequenced, Ulleung Island endemic C. wilsonii (NC046834), and seven species representing two major lineages within the genus: the Chaenopetalum group (C. soongoricus, C. vandelaarii, and C. conspicuous) and Cotoneaster group (C. microphyllus, C. foveolatus, C. horizontalis, and C. franchetii) [4]. Sequences of the 15 Cotoneaster plastomes were aligned using the back-translation approach with MAFFT v7.490 [49] and were manually edited using Geneious R10 [45]. Using DnaSP v6.10 [50] sliding window analysis was performed with a step size of 200 bp and window length of 800 bp to determine the nucleotide diversity (Pi) of the plastomes. Codon usage frequency was calculated using MEGA v7 [51] based on the relative synonymous codon usage (RSCU) value [52], which is a simple measure of non-uniform usage of synonymous codons in a coding sequence. The DNA code used by bacteria, archaea, prokaryotic viruses, and chloroplast proteins was used [53]. Protein-coding genes were run using the PREP suite [54] with 35 reference genes and a cut-off value of 0.8 to predict the possible RNA editing sites in 15 Cotoneaster plastomes. To evaluate the natural selection pressure on the protein-coding genes of 15 Cotoneaster plastomes, a site-specific model was developed using EasyCodeML [55] with the CODEML algorithm [56]. Seven codon substitution models (M0, M1a, M2a, M3, M7, M8, and M8a) were constructed and compared to detect the positively selected sites based on the likelihood ratio test (LRT).

Codon Usage
The frequency of codon usage in the 15 plastomes of Cotoneaster, representing the seven newly sequenced and eight major lineages within the genus, was calculated based on the sequences of protein-coding and tRNA genes. The results revealed that the average codon usage among the 15 species ranged from 25,868 (C. submultiflorus) to 26,586 (C. horizontalis) (Supplementary Table S2 Table S2).

Comparative Analysis of Chloroplast Genome Structure
The plastomes of 14 Cotoneaster species (i.e., C. conspicuous, C. dielsianus, C. foveolatus, C. franchetii, C. hebephyllus, C. horozontalis, C. integerimus, C. microphyllus, C. mongolicus, C. multiflorus, C. soongricus, C. submultiflorus, C. vandelaarii, and C. tenuipes) were plotted using mVISTA, using the annotated C. wilsonii plastome as a reference (Figure 3). The results indicated that the LSC region was the most divergent, whereas the two IR regions were highly conserved. Furthermore, the non-coding regions were found to be more divergent and variable than the coding regions. Sliding window analysis performed using the DnaSP program revealed highly variable regions in the plastomes of 15 Cotoneaster species (Figure 4). Comparison of these 15 plastomes revealed that the average value of nucleotide diversity (Pi) over the entire chloroplast genome was 0.001345, with the most variable region (Pi = 0.01076) being the trnG-UCC/trnR-UCU/atpA intergenic region. One additional

Identification of Genes under Positive Selection
Positive selection analysis allowed us to identify positively selected genes among the 15 Cotoneaster plastomes (Table 2). Among the conserved genes, eight genes with positively selected sites were identified with an effectively significant LRT p-value (Table 2). These genes included the c-type cytochrome synthesis gene (ccsA), maturase K gene (matK), three NADH dehydrogenase subunit genes (ndhD, ndhF, and ndhK), cytochrome f precursor gene (petA), Rubisco gene (rbcL), and mitochondrial ribosomal protein L16 gene (rpl16), and based on the M8 model, all eight genes had one positive site. However, most of the genes, i.e., 68 of 76 genes had an average Ka/Ks ratio below 1, indicating that these genes have been subjected to the strong purifying selection in the Cotoneaster chloroplast.

Identification of Genes under Positive Selection
Positive selection analysis allowed us to identify positively selected genes among the 15 Cotoneaster plastomes (Table 2). Among the conserved genes, eight genes with positively selected sites were identified with an effectively significant LRT p-value (Table 2). These genes included the c-type cytochrome synthesis gene (ccsA), maturase K gene (matK), three NADH dehydrogenase subunit genes (ndhD, ndhF, and ndhK), cytochrome f precursor gene (petA), Rubisco gene (rbcL), and mitochondrial ribosomal protein L16 gene (rpl16), and based on the M8 model, all eight genes had one positive site. However, most of the genes, i.e., 68 of 76 genes had an average Ka/Ks ratio below 1, indicating that these genes have been subjected to the strong purifying selection in the Cotoneaster chloroplast.
As for the phylogenetic relationships of several conspecific plastomes between this study and Meng et al. [4], we found some congruences as well as incongruences. For example, C. submultiflorus (208-2000-A; originally from Gansu) sequenced in this study was closely related to C. multiflorus (1134-A; originally from Shaanxi) and C. multiflorus (Yunnan) [4]. Cotoneaster submultiflorus (Xinjiang) [4] was a sister to C. mongolicus (1007-86-A, originally from Guangdong). All these accessions formed a clade with 100% BS support in clade B. Although these accessions were part of the monophyletic group, other conspecific plastomes showed drastically different positions. For example, C. hebephyllus (sect. Multiflori) (Tibet) [4] represented an early diverged lineage within Chaenopetalum (Clade B), but the accession sampled in this study (68-85-A; Arnold Arboretum) was a sister to the clade containing C. perpusillus, C. dielsianus, C. wilsonii, C. schantungensis, and C. zabelii (100% BS). Furthermore, C. dielsianus (sect. Franchetioides) sampled in this study (Hubei; 13428-B, Arnold Arboretum) was closely related to C. perpusillus (sect. Adpressi; 98% BS) in subclade A1, whereas the other accession (Yunnan) [4] was a sister to C. huahongdongensis (sect. Franchetioides, 100% BS) in subclade A2. These two accessions, C. dielsianus (Yunnan) and C. huahongdongensis, were sampled from Yunnan, with a direct distance of <70 km (Supplementary Table S2 of Meng et al. [4]). Given the wide disjunct distribution of C. dielsianus in Sichuan and Hubei, it is uncertain whether C. dielsianus sampled from Hubei in this study represents a different taxon. We unwittingly sequenced C. tenuipes (sect. Adpressi; 7276-C, Arnold Arboretum), the same accession that was sequenced by Meng et al. [4], and found that they have identical sequences, ruling out the possibility of sequencing mistakes between the two studies.

Chloroplast Genome Structure and Evolution in Genus Cotoneaster
In this study, we assembled and characterized seven additional species of Cotoneaster (C. dielsianus, C. hebephyllus, C. integerrimus, C. mongolicus, C. multiflorus, C. submultiflorus, and C. tenuipes) and added them to the existing chloroplast genome database of the genus [4]. Furthermore, for the first time, we performed several comparative analyses of plastomes based on seven newly sequenced and seven major lineages within the genus, including C. wilsonii on Ulleung Island, to gain an insight into plastome evolution. The complete chloroplast genome size in Cotoneaster ranged from 159,521 bp (C. acutifolius) to 160,016 bp (C. hebephyllus), and the largest plastome from Meng et al. [4] belonged to C. melanocarpus (159,970 bp). Thus, we have characterized and added the largest plastome found within genus Cotoneaster to date. There were differences of less than 500 bp in the complete length of the plastomes, indicating their conservation within Cotoneaster. Given the highly conserved nature of plastomes, no structural variation or gene content rearrangement was found within the genus ( Table 1). As expected, the LSC region was the most divergent, whereas the two IR regions were highly conserved. Furthermore, non-coding regions were found to be more divergent and variable than the coding regions. These findings are consistent with the patterns observed in common angiosperms [22,27,29,30,39,58]. The GC content of the complete plastomes in the 15 representative Cotoneaster species was identical (36.6%), and this high GC content could be attributed to the high GC content in the IR regions [59].
As we compared the 15 representative plastomes of Cotoneaster in this study, we revealed the retention of an intron in atpF belonging to group II introns [60]. Intron loss or gain in the plastome can be an evolutionarily significant event as introns are highly conserved among land plants [61]. Loss of the atpF intron has been reported in several Rosaceae genera, such as Fragaria, Rosa, Potentilla, and Rubus [34,37,62]. These genera belong to subfamily Rosoideae, and thus it seems that the loss of introns within atpF genes has occurred once within this subfamily. In contrast, several other genera of Rosaceae belonging to subfamily Amygdaloideae, such as Cotoneaster, Alchemilla, Malus, Prunus, Pyrus, and Sorbus, retain introns within atpF genes [62,63]. As suggested, it is unclear whether intron loss has occurred in the species of subfamily Rosoideae, genera in the family Rosaceae, and families in the order Rosids, and whether this has phylogenetic significance, utility in the classification system, and forms a potentially resourceful evolutionary maker in angiosperms.

The Codon Usage Pattern in the Cotoneaster Chloroplast Genome
The frequency of codon usage in the 15 Cotoneaster plastomes was determined based on the sequences of protein-coding genes (Supplementary Table S2). The preferential use of codons during gene translation (i.e., specific codons are used more often than others) is known as codon usage bias, and codon usage values are described by the relative synonymous codon usage (RSCU) [43]. RSCU is the ratio between the expected frequency of use and the actual frequency usage of a particular codon. RSCU values less than 1 indicate lower frequency usage than expected, whereas values greater than 1 indicate a higher usage frequency [52]. The codon usage bias and any intraspecific or interspecific codon usage variation are indicative of selective constraints on codon choice. We found the highest RSCU value in the usage of the UUA codon for leucine (1.92-1.94) followed by that of GCU for alanine (1.83-1.84) and AGA for arginine (1.82-1.84), whereas the lowest value was found in the usage of AGC for serine (0.38-0.39) and GAC for aspartic acid (0.37-0.38) (Supplementary Table S2). Codons AUG (M) and UGG (W) encoding methionine and tryptophan showed no bias (RSCU = 1). This pattern is consistent with that of the genus Malus, belonging to the same subfamily Amygdaloideae [58]. Similar to other Rosaceae species (Potentilla L. and Spiraea, L. [62]; Alchemilla L., [63]; Malus, Cho et al., [58]), we found that codon usage was biased toward a high RSCU value for U and A at the third codon position in genus Cotoneaster.

The Characteristic of RNA Editing Sites in the Cotoneaster Chloroplast Genome
Although previous studies have demonstrated variable numbers of RNA editing sites among higher taxonomic ranks of land plants and between the two organellar genomes [64][65][66], the extent of variation among closely related species or multiple genera of the same family is known to be sporadic. In organellar genomes (chloroplast and mitochondria), conversion from C (cytidine) to U (uridine) has been shown to be the most prevalent [67,68]. Regarding the RNA editing sites in the 15 Cotoneaster plastomes, all species shared RNA editing sites in 14 photosynthesis-related genes, seven self-replication genes, and three other functional genes (Supplementary Table S3). All 15 Cotoneaster species also shared the same 10 genes without any RNA editing sites. Although RNA editing sites are highly conserved among closely related species, we found that C. microphyllus, belonging to Cotoneaster Clade A, subclade A2 (Figure 3), exceptionally, contains one additional gene, rps8: with RNA editing site conversion from ACC (T, threonine) to ATC (I, isoleucine). This is in contrast to the Malus plastomes from East Asia, which showed that the rps8 gene did not have an RNA editing site [40]. Three other functional genes (accD, clpP, and matK) and seven self-replication genes contained RNA editing sites and were common between Malus and Cotoneaster; however, we found that the petG gene, reported to have an RNA editing site in Malus, did not have an RNA editing site in the Cotoneaster species surveyed in this study. As shown in previous studies [40,62,65,69], the highest number of potential editing sites were found in the NADH dehydrogenase genes, with the ndhB gene harboring 12 sites, and ndhD gene harboring 8 sites. Similar to Malus, we found that the highest conversions in the editing sites were represented by changes from serine (S) to leucine (L) (average confidence score of 23.81) followed by proline (P) to leucine (L) (average confidence score of 8.86).

Positively Selected Genes in Cotoneaster Chloroplast Genomes
Most plastome genes have evolved under purifying selection because of functional limitations throughout chloroplast genome evolution [76][77][78][79]. As synonymous nucleotide substitutions occur more frequently than non-synonymous substitutions, Ka/Ks values are usually less than 1 [80]. In the 15 plastomes of genus Cotoneaster, most of the genes have been under strong purifying selection; 68 of the 76 genes have an average Ka/Ks ratio below 1. However, among the 15 representative species selected in Cotoneaster, eight genes, that is, ccsA, matK, ndhD, ndhF, ndhK, petA, rbcL, and rpl16, have undergone selective pressure. Positive selection of several functional genes has been previously reported in several studies. For example, the rbcL gene, which encodes the large subunit of RuBisCO, plays an important role in photosynthesis and is often under positive selection in various plant groups including Fragaria [76], Gossypium L. [81], Panax L. [78], Paulownia [79], Poaceae grass after the C3-C4 photosynthetic transition [82], and Rubus [34]. Based on the current analysis, it is also likely that the rbcL gene was the target of selection during the evolution of Cotoneaster. The matK gene was identified to be under positive selection in Cotoneaster.
The matK gene has also been shown to be under positive selection in several lineages within Allium L., suggesting its role in adaptation to a wide range of environments [36]. Positive selection of the matK gene has been reported in various other plant lineages, such as PACMAD grasses (Poaceae, [82]), Chrysosplenium L. [83], Symplocarpus Salisb. ex W.P.C.Barton [69], and Rubus [34]. As shown in several other plant groups (e.g., [36,84]), our study also revealed that three genes from the ndh family, that is, ndhD, ndhF, and ndhK, were under positive selection. Of the ndh gene family, ndhK has been shown to be positively selected in species adapted to different altitudinal habitats [25] and in shade-tolerant and sun-loving plants [85]. In addition, ndhF evolved under positive selection because of its involvement in the adaptation to hot and dry climates [86]. Therefore, these ndh gene family members likely contributed to adaptation to high light intensity during the evolution of Cotoneaster. It is also likely that the ribosomal protein-coding gene, rpl16, was selected to maintain the integrity of the protein synthesis machinery under various environmental stresses [87]. Overall, we hypothesized that these positively selected genes in different categories of the chloroplast genome, including subunits of cytochrome (ccsA and petA), are results of their important adaptive roles in diverse environmental conditions during the evolutionary radiation of the genus from the late Miocene to today.

Phylogenetic Position of Cotoneaster wilsonii on Ulleung Island
The origin and evolution of C. wilsonii on Ulleung Island have been problematic given their unusual geographic distribution in Korea. Cotoneaster wilsonii, which occurs very narrowly on Ulleung Island as a critically endangered species, represents the easternmost range of the entire genus Cotoneaster. Without natural distribution in the Japanese archipelago, only one additional species of Cotoneaster, C. integerrimus, is known to occur in North Korea. This species is also known to occur in a few isolated limestone areas in Gangwondo Province of South Korea (Samcheok city, Yeongwol-gun, and Jeongseon-gun), but its species identity and relationship with C. wilsonii are yet to be determined. Thus, considering its narrow geographic distribution in the oceanic Ulleung Island, which was formed approximately 1.8 million years ago, and the lack of a broad phylogenetic framework of the genus, the origin and phylogenetic relationships of C. wilsonii relative with other congeneric species has been a matter of speculation. With the broad scale phylogenomic study by Meng et al. [4] and our current study, we, for the first time, assessed the phylogenetic position of C. wilsonii. Although several cases of incongruences between nuclear and chloroplast phylogeny in Cotoneaster caused by hybridization and incomplete lineage sorting were revealed, a tentative conclusion about the phylogenetic position of C. wilsonii can be suggested based on the congruence in the clade of our interest ( Figure 4 of Meng et al. [4]).
Based on morphological similarity and shared flavonoid profiles, Chang and Jeon [14] suggested that C. multilforus would be the closest continental sister species of C. wilsonii, or that they are conspecific. Cotoneaster multiflorus and C. hebephyllus contain flavone Oglycosides identical to those in C. wilsonii, suggesting their possible role in the origin of C. wilsonii on Ulleung Island. Cotoneaster multiflorus and several related species belong to sect. Cotoneaster series Multiflori sensu Yü et al. [5]. Our current study strongly suggests that C. wilsonii shares its most recent common ancestor with two species, C. schantungensis and C. zabelii, which belong to sect. Cotoneaster series Integerrimi sensu Yü et al. [5] (Figure 5). Unlike C. multiflorus, which has 5-21 flowers per inflorescence and spreading petals, the two most closely related species, C. schantungenis and C. zabelii, tend to have fewer flowers, 3-6 or 3-10 (or more), respectively [7]. In addition, both C. zabelii and C. schantungensis have erect petals, whereas C. wilsonii has 4-17 flowers (average of 10 flowers per corymb) and spreading petals. While the clade containing C. wilsonii belongs to Cotoneaster (Clade A), C. multiflorus and related species all belong to sect. Cotoneaster series Multiflori, which belongs to a different clade, i.e., Chaenopetalum (Clade B). Therefore, it is less likely that C. multiflorus and related species are involved in the origin of C. wilsonii, implying that their morphological similarities and similar flavonoid profiles are most likely convergent features or symplesiomorphy. Furthermore, the phylogeny of 203 low-copy nuclear genes also suggested that C. multiflorus and C. hebephyllus are not closely related to the clade containing C. wilsonii [4], thus corroborating our current results. Furthermore, the chromosome number of C. wilsonii is known to be diploid (2n = 34) [88], whereas C. multiflorus is tetraploid (2n = 68) [7]. The accession of C. hebephyllus (68-85-A) from the Arnold Arboretum with its wild origin unknown is related to the C. wilsonii-containing clade, but the wild origin accession (Tibet) by Meng et al. [4] is distantly related to C. wilsonii. In fact, the C. hebephyllus accession from Tibet is a sister to the clade containing the species of sect. Multiflori and other sections sensu Fryer and Hylmö [1]. Therefore, we are uncertain about the species identity of the C. hebephyllus accession at Arnold Arboretum. Taking this caveat together with the current phylogeny obtained, we can safely rule out the possibility of sect. Cotoneaster series Multiflori sensu Yü et al. [5] being involved in the origin of C. wilsonii. Rather, it is highly likely that the species in sect. Cotoneaster series Integerrimi was involved in the origin of C. wilsonii. The chloroplast phylogenomic tree suggests that the sect. Cotoneaster series Integerrimi sensu Yü et al. [5] is not monophyletic ( Figure 5). Of several species from series Integerrimi, it seems likely that a common ancestor shared with species such as C. schantungensis and C. zabelii was involved in the origin of C. wilsonii on Ulleung Island. Cotoneaster schantungensis is endemic to Shandong Province, which is geographically close to the Korean Peninsula, just across the Yellow Sea. Furthermore, C. zabelii occurs quite broadly in western Qinghai, northeastern Nei Mongol, and eastern Shandong [7].
As part of the clade containing C. wilsonii, the potential involvement of C. dielsianus in the origin of Ulleung Island is also plausible. C. dielsianus belongs to the same sect. Cotoneaster series Integerrimi sensu Yü et al. [5] or sect. Franchetioides sensu Fryer and Hylmö [1] but is morphologically distinct from C. wilsonii by having 3-7 small (6-7 mm in diameter) flowers, abaxially villous hypanthium, erect petals, and 3 (rarely 5) styles [7]. In contrast, C. wilsonii has 4-17 (average of 10) large (8-12 mm) flowers, an abaxially glabrous hypanthium, spreading petals, and 2 (rarely 3) styles. The accession (13428-B) of C. dielsianus from the Arnold Arboretum, which was originally collected from western Hubei, contained a very different plastome compared to the one sequenced by Meng et al. [4], which was sampled from Yunnan. Cotoneaster dielsianus occurs somewhat broadly, ranging from central to southwestern China, and without examining the voucher specimen of the Yunnan accession, it is difficult to determine whether these two accessions represent distinct taxa or infraspecific variation within C. dielsianus. The highly polyphyletic sect. Cotoneaster series Integerrimi sensu Yü et al. [5] or sect. Francheotioides sensu Fryer and Hylmö [1] in the Cotoneaster phylogeny further complicate the resolution of this issue.
The plastome phylogenetic position of C. perpusillus as part of a clade containing C. wilsonii and related species (subclade A1) seems unusual given its morphology and sectional/serial assignment. Cotoneaster perpusillus belongs to the sect. Uniflos sensu Yü et al. [5] and sect. Adpressi sensu Fryer and Hylmö [1]. It is currently recognized as C. horizontalis var. perpusillus, known to occur in central China (Guizhou, Hubei, Shaanxi, and Sichuan), with characteristics of having only one or two flowers, smaller leaves (<1 cm), and erect petals [7]. With the exclusion of C. hebephyllus, this is the only species of sect. Adpressi or Uniflos, placed in the clade of sect. Cotoneaster sensu Fryer and Hylmö [1] and sect. Cotoneaster series Integerrimi sensu Yü et al. [5]. The accession of C. perpusillus sequenced by Meng et al. [4] was collected from Yunnan, which is a neighboring Sichuan Province. Although it is a part of subclade A1, C. perpusillus is a sister to C. harrysmithii (albeit weakly supported, with a Bayesian posterior probability of 0.88, and Ultrafast bootstrap support value < 50%) in the species tree based on 203 low-copy nuclear genes [4]. Cotoneaster harrysmithii, which occurs rather narrowly in western Sichuan and southeastern Xizang, belongs to sect. Uniflors sensu Yü et al. [5], and sect. Adpressi sensu Fryer and Hylmö [1], which is the same sectional assignment as C. perpusillus. Thus, it is highly likely that C. perpusillus experienced hybridization events with C. dielsianus, which also occurs in central and southwestern provinces (including Yunnan and Sichuan), and sub-sequently captured the chloroplast of C. dielsianus, its sister species in the chloroplast phylogenomic tree.
Based on the broad phylogenomic framework and molecular dating of Cotoneaster, we can also gain an insight into the timing of C. wilsonii on Ulleung Island, Korea. Meng et al. [4] suggested that the crown node age for the subclade A1, including C. wilsonii and related species, is estimated to be 6.25 million years (MY) old. In addition, the clade containing C. perpusillus, C. schantungensis, and C. zabelii, is estimated to be 0.72 MY, whereas the clade containing all these species plus C. schangsiensis is 2.41 MY old. As C. wilsonii is sister to the clade of C. schantungensis and C. zabelii, the most recent common ancestor of C. wilsonii and C. schantungensis, C. zabelii, should be younger than 0.72 MY. This suggests that C. wilsonii may have originated very recently, long after the formation of Ulleung Island, which is slightly less than 2 MY old. Although nearly 40 vascular endemic species occur on Ulleung Island, little is known about their timing of origin. Thus, further investigation based on a robust and well-resolved phylogenetic framework and molecular dating is required to better understand the temporal scale of these endemic assemblages on the island.   Data Availability Statement: The datasets generated and/or analyzed during this study can be found in GenBank, National Center for Biotechnology Information (http://www.ncbi.nlm.nih.gov/ genbank/, (8 January 2022) under accession numbers MZ475328 and MZ475334.