Phylogeny of Maleae (Rosaceae) Based on Multiple Chloroplast Regions: Implications to Genera Circumscription

Maleae consists of economically and ecologically important plants. However, there are considerable disputes on generic circumscription due to the lack of a reliable phylogeny at generic level. In this study, molecular phylogeny of 35 generally accepted genera in Maleae is established using 15 chloroplast regions. Gillenia is the most basal clade of Maleae, followed by Kageneckia + Lindleya, Vauquelinia, and a typical radiation clade, the core Maleae, suggesting that the proposal of four subtribes is reasonable. In the core Maleae including 31 genera, chloroplast gene data support that the four Malus-related genera should better be merged into one genus and the six Sorbus-related genera would be classified into two genera, whereas all Photinia-related genera should be accepted as distinct genera. Although the phylogenetic relationships among the genera in Maleae are much clearer than before, it is still premature to make a formal taxonomic treatment for these genera.


Introduction
Rosaceae or rose family, consisting of approximately 4,828 species in 91 genera [1], is of great economic and ecological importance. Many species are cultivated for their fruits or as ornamentals. The monophyly of the family is implied by the presence of unique floral structures and strongly supported by rbcL phylogeny [2]. However, the rose family displays a considerable diversity in morphology and anatomy, and it had been generally subdivided into four subfamilies, that is, Rosoideae, Spiraeoideae, Amygdaloideae (incorrectly Prunoideae), and Maloideae. Such a subdivision was recently challenged by molecular phylogenies of matK and trnL-F [3], six nuclear and four chloroplast regions [4], and hundreds of nuclear genes [5]. A formal three-subfamily classification was proposed: Dryadoideae, Rosoideae, and Amygdaloideae (incorrectly Spiraeoideae in [4]). Dryadoideae was separated from Rosoideae, and Spiraeoideae and Maloideae were merged with Amygdaloideae (incorrectly Spiraeoideae).
The origin of core Maleae ( = 17) has long been considered an example of allopolyploidization between the species with = 9 in traditional Spiraeoideae and the species with = 8 in traditional Amygdaloideae [14][15][16][17]. However, recently discovered genomic data suggested an origin via autopolyploidization followed by aneuploidization around 50 million years ago [18,19]. In contrast, the allopolyploid nature of core Maleae was confirmed by GBSSI, which had four copies in the core Maleae but only two copies in other groups [6].
Polyploidization affects systematics at both generic and specific levels. It is unlikely to resolve the polytomy of ancestral populations that have just a few closely related species involved in historical speciation and subsequent diversification, owing to the lack of phylogenetically informative signals and incomplete lineage sorting. The merging of several genomes into one species enriches the pool of available genetic combinations and the survival of recombinants is overcome by apomixis. Maleae is one such species-rich tribe of genera but is difficult to classify.

Sequence Data Preparation and Evaluation.
The newly generated sequences were checked and assembled using Sequencer 4.7 (Gene codes Corporation, Ann Arbor, Michigan, USA). The resulting sequences (Supplementary Table  S2) were combined with the published sequences of Maleae species downloaded from GenBank (Supplementary Table  S1), aligned with Clustal X [47,48], and then manually adjusted with Se-Al 2.0 [49]. Each individual gene dataset was then subjected to several rounds of phylogenetic evaluations to select reliable sequences. Sequences that were misidentified or exhibited large amounts of missing data were excluded from the datasets. To predict the phylogenetic performance of individual gene partitions, the variability of genes was parameterized by DnaSP 5.0 [50] and PAUP 4.0b10 [51] using nucleotide diversity, the number of polymorphic sites, and so forth. Prior to concatenating the dataset of each marker, incongruence length difference (ILD) tests were performed on all 15 datasets. The datasets were finally concatenated Table 1: Representative species of Maleae, other major lineages of Amygdaloideae, and an outgroup used in this study, together with sample vouchers and sampling localities.
into one final data matrix using SequenceMatrix [52] for phylogenetic analyses.

Phylogenetic Analysis.
Phylogenetic analyses were performed on single gene datasets and the concatenated dataset, using PAUP 4.0b10 [51] for maximum parsimony (MP), RAxML v.8.1.24 [53] for maximum likelihood (ML), and MrBayes 3.2.2 [54] for Bayesian inference (BI). The MP analysis used a heuristic search that treats all characters as equally weighted and unordered, obtaining the starting trees  The concatenated data matrix of the 41 taxa reached an aligned length of 17,554 bp with 1,266 parsimony-informative characters. MP searches yielded one best tree with a consistency index (CI) of 0.779, a retention index (RI) of 0.637, and a tree length of 5,974.

Phylogenetic Relationships of the Basal Maleae.
Polytomies were observed in all the 15 best trees based on each chloroplast region. However, the tree topologies were similar; we thus concatenated them to build best resolved phylogenetic trees using MP, ML, and BI methods. The consensus trees from the MP, ML, and BI analyses showed substantially identical topologies and the monophyly of Maleae was recovered (Figure 1). A long branch leads to the crown taxa of Maleae, including Gillenia, Lindleya, Kageneckia, and Vauquelinia (Figure 1(A)), and this branch is strongly supported (Figure 1(B)) (PB = 100, BP = 100, PP = 1).
An earliest and remarkable divergence of Gillenia is clearly indicated in Figure 1. Lindleya and Kageneckia form a monophyletic clade and they diverged slightly earlier than Vauquelinia. The divergence among Lindleya + Kageneckia, Vauquelinia, and the core Maleae happened within a very short span of time, as indicated by the very short branches. All the three branches are well supported. The basal branching pattern of Maleae may serve as the foundation of subtribal division within the tribe if necessary.

Phylogenetic Relationships within the Core Maleae.
The monophyly of the core Maleae is clearly indicated (Figure 1(A)) and well supported (Figure 1(B)). The genera in the core Maleae seem to be from the second radiation event. Nevertheless, three multigenus clades within the core Maleae are recognizable. Clade I consists of Amelanchier Medik., Crataegus L., Cydonia Mill., Malacomeles (Decne.) Decne., Mespilus L., and Peraphyllum Nutt. and is well supported

Taxa and Gene Sampling Strategies.
It is often a challenge to reconstruct the phylogeny of taxa from recent radiation events due to unclear relationships among subdivisions and low resolution of markers. Maleae is such a taxon and that is why early studies failed to establish solid phylogenetic relationships among its subdivisions. In "A checklist of the subfamily Maloideae (Rosaceae)" [21], only 23 genera were accepted. Six Sorbus-related entities, that is, Aria, Chamaemespilus, Cormus, Micromeles, Sorbus, and Torminalis, were proposed to be merged. To test the distinctness of these taxa, we included all of them in this study. Malus-related and Photiniarelated entities have similar taxonomic problems. Inclusion of representative species from other lineages of Amygdaloideae was to test the monophyly of Maleae. Considering that the monophyly of the genera in the narrowest sense has been well verified, 35 entities with one representative species each were sampled for being the most inclusive and economically affordable.
Sampling chloroplast regions as molecular markers for Maleae is another challenge. The tribe was found to be quite young and the core Maleae was even younger. Chloroplast markers have showed very low resolutions in previous studies [2,4,7,9,42]. Therefore, the most variable regions suggested by Dong et al. [56] were used together with the four conventional DNA barcodes, matK, psbA-trnH, rbcL, and ycf1 [46]. By doing so, the major groups in Maleae were well resolved.

Phylogenetic Relationships among Major Groups of
Maleae. The phylogeny (Figure 1) strongly suggests inclusion of Gillenia, Kageneckia, Lindleya, and Vauquelinia in Maleae (incorrectly Pyrodae in [7] and in [4]). Gillenia seemed to diverge slightly early, and Kageneckia + Lindleya, Vauquelinia, and the core Maleae are quite probably from the first radiation event. Although the inclusion of Kageneckia, Lindleya, and Vauquelinia into Maleae seems disagreeable with respect to fruit types, their basal chromosome number ( = 15 or 17) suggests that they had probably experienced similar speciation events. They bridge the gap between the core Maleae and true spiraeoid Gillenia. The inclusion of Gillenia in Maleae verifies that the core Maleae is from spiraeoid members within the tribe or other tribes in Amygdaloideae. Unfortunately, transcriptome data of nuclear genes did not provide more information for the issue of origin of the core Maleae because Figure 1 is substantially similar to the tree topologies based on transcriptome data in [5]. Given that Gillenia is the only diploid member in Maleae, the ancestor of Gillenia or its close relatives must be the maternal parent of the extant Maleae.

Phylogenetic Relationships within the Core Maleae.
The core Maleae is a natural group with several synapomorphic characters, such as syncarpous ovaries, epigynous flowers, and fleshy fruits derived from the hypanthial ovary [57][58][59]. It could be better classified as subtribe Malinae (incorrectly Pyrinae) if necessary. The Malinae diverged into genera by radiation which is the second radiation event in Malese. The genera are so closely related that it is unnecessary to subdivide this subtribe further.

Generic Pairs and Their Taxonomic Status.
There are four generic pairs, for which the taxonomic status needs to be clarified. Firstly, a close genetic relationship between Crataegus and Mespilus has been revealed by Lo and Donoghue [9] in detail, which is confirmed here again. If Mespilus were to be accepted as a distinct genus, Crataegus would become a paraphyletic group. Although Mespilus was merged with Crataegus, its distinction was still recognized by giving it the taxonomic rank of section in Crataegus [9]. Secondly, reduction of Pseudocydonia to Chaenomeles was done by Gu and Spongberg [60], and their treatment is supported by Campbell et al. [7]. However, their close relationship does not receive high support in this study (BP = 52, PP = 1). They have diverged for some time because the clade does not show remarkable branch length. They had better be considered distinct genera. Thirdly, Stranvaesia is morphologically similar to both Photinia and Cotoneaster and sometimes submerged into Photinia [61]. Phylogeny based on chloroplast regions done by Campbell et al. [7] indicated a close relationship between Stranvaesia and Pyracantha. However, our data support a close relationship between Stranvaesia and Cotoneaster, a relationship also supported by GBSSI-1B [7]. Fourthly, although very close morphological and genetic relationships have been revealed in almost all studies involving Eriobotrya and Rhaphiolepis, there has been no suggestion yet to merge them into one, despite the existence of their hybrids [62]. Genetic divergence between them is very recently because their clade has a relatively long branch.

Sorbus-Related Genera.
So far, no well-sampled phylogeny of all Sorbus-related taxa is available, owing to the genetic complexity of the group and difficulties in sampling. There are ca. 250 species belonging to mainly temperate areas in the Northern Hemisphere [21], and six narrowly defined genera or subgenera under Sorbus in the broadest sense are usually accepted: Aria, Chamaemespilus, Cormus, Micromeles, Sorbus, and Torminalis [21, 37-39, 43, 58, 63-72]. These taxa fall into two clades in this study (Figure 1). Sorbus sensu stricto and Cormus with compound leaves nested in Clade I, and Aria, Chamaemespilus, and Torminalis with lobed or unlobed simple leaves nested in Clade III. Micromeles, a genus created to contain the species without persistent calyx lobes, is not a natural entity and is, therefore, unacceptable [8]. Micromeles folgneri in this study is a synonym of Sorbus folgneri. Although Cormus had been considered a synonym of Sorbus, its taxonomic position remains to be determined. Aria, Chamaemespilus, and Torminalis in Clade III are closely related and they had better merge into one separate from Sorbus.
Sorbus-related entities are notorious in taxonomy due to complexity in phenotypes resulting from interspecific hybridization and facultative agamospermy [39,69,70]. Apomictic microspecies confound systematic resolution of agamic complexes using nuclear markers [71]. Phylogenies based on nuclear genes may suffer from paralogue problems and chloroplast markers would work better at the very beginning when no clear ideas are available for classification.

Photinia-Related Genera.
Four genera in their narrowest sense, Aronia, Heteromeles, Pourthiaea, and Stranvaesia, are considered related to Photinia [24,25,29,31,37]. This study demonstrates that all the five genera are superficially similar but actually distinct groups. Heteromeles possesses a soft pyrene, while Photinia possesses a soft core [16,37,58]. Stranvaesia is distinguishable from Photinia by the dehiscent fruits at maturity. Although Guo et al. [73] stated with quite certainty that Stranvaesia must be merged into Photinia, we believe that further studies are needed for a reliable conclusion. Pourthiaea belongs to Clade III instead of Clade II as other genera. It is readily separable from other genera by characters such as deciduous habit, warty peduncles and pedicels, and a pulp structure in the fruits [41]. The distinction of Pourthiaea is further supported by leaf epidermis and wood anatomy [74,75]. Aronia is not a sister group of Pourthiaea because the clade is poorly supported, a conclusion that is similar to the earlier studies ( [8,9,73]).

Malus-Related
Genera. The four genera, Docynia, Docyniopsis, Eriolobus, and Malus, really have very close relationships among them, especially between Docyniopsis and Malus. Some species have been transferred among the four genera by different authors. The monophyly of a clade comprising the four genera is indicated by our chloroplast data (Figure 1(B)). There are only two species in Docynia, two species in Docyniopsis, and one species in Eriolobus. Considering that these species are nested within Malus [18], these genera could better be merged with Malus.

Conclusions
This molecular phylogenetic study was conducted on all the genera of Maleae, by using 15 variable chloroplast gene regions. Four major clades are well resolved and the branches are well supported. These clades could be classified into four subtribes. The first radiation event gave birth to Lindleya + Kageneckia, Vauquelinia, and the ancestor of the core Maleae, and the second radiation event triggered the divergence of the genera in the core Maleae within a short time period. Within the core Maleae the four Malus-related genera should better be merged into one genus; the six Sorbus-related genera would be better classified into at least two genera; and all Photinia-related genera should be accepted as distinct genera. For a more reliable classification, phylogeny based on the whole chloroplast genomes of representative species from each genus should be used. Such a strategy has been practiced on many seed plants (e.g., [76][77][78][79][80][81][82][83]). Besides, the nuclear genes, especially the single copy nuclear genes in diploid species such as starch-branching enzyme (Sbe; [84]), should be considered with priority to document the origins of Maleae, as the maternally transmitted markers can only tell one aspect of the whole story. Application of nuclear genome information in phylogenetic reconstruction of Maleae also seems feasible, because the genomes of several species in the tribe have been sequenced, and the resequencing of many species is underway. We are expecting a completely resolved phylogeny of Maleae, and that is why we do not provide a formal taxonomic treatment in this study.

Conflicts of Interest
The authors declare that there are no conflicts of interest regarding the publication of this paper.