Background

Elm (Ulmus L.) is a representative genus of Ulmaceae including about 40 species, which is mainly distributed in the temperate regions of the Northern Hemisphere [1,2,3]. China has been regarded as one of the original and divergent centers of Ulmus owning more than 25 species. In addition, there are at least 10 species of Ulmus restricted to small regions and almost half of them are endemic to China [1]. For instance, U. elongata is regarded as the ancient Tertiary species and extremely small populations, and U. macrocarpa, U. lamellosa as well as U. lanceifolia are listed as nationally protected and endangered species in China [4]. Elm trees have the characteristics of high ornamental value, fast growth, and wide adaptability due to their strong resistance to abiotic stress, such as drought and cold [5]. In addition, its bark, leaf and fruit have been widely used as medicine to calm nerves and diuretics, with immeasurable development prospects of raw material for the chemical industry [6]. According to flowering phase, inflorescence, and samara type, a complete infrageneric classification of the world Ulmus species was built recently, in which Ulmus was divided into five sections including sect. Blepharocarpa Dumert, sect. Chaetoptelea Schneid, sect. Microptelea Benth, sect. Trichoptelea Schneid, and sect. Ulmus [1,2,3, 7, 8]. Specifically, sect. Microptelea and sect. Trichoptelea are distinguished from other species by flowering phase in the fall to winter rather than spring. The sect. Blepharocarpa is named by its samara pubescent in the margin, while peduncles slender of sect. Chaetoptelea are elongated and particularly pendulous. The sect. Ulmus with more than 20 species is extremely diverged, which is comprised of four series (Ser. Lanceifoliae, Ser. Nitentes, Ser. Glabrae, and Ser. Villosa) [8]. However, a wide range, similar ecological affinities and morphological characters, and recent diversification caused it is difficult to discriminate Ulmus species based on morphological variations independently. Thus, an intensification of taxonomic and systematic research is still needed for Ulmus species. Previous molecular evidence based on several fragments (e.g., ITS, atpB-rbcL) or limited cp genomes has largely enhanced our understanding of phylogeny in Ulmus [8,9,10,11]. Wiegrefe et al. (1994) established the phylogenetic backbone of the Ulmus based on restriction sites, dividing five sections into two subgenera: subg. Oreoptelea (sects. Blepharocarpa, Chaetoptelea, and Trichoptelea) and subg. Ulmus (sects. Microptelea and Ulmus) and assigning U. lanceifolia to sect. Lanceifoliae. Latterly, based on DNA fragments and cp genomes, resolution of phylogeny in Ulmus was largely improved, but the relationships among sections and within the sect. Ulmus were still unresolved [9, 10, 12]. Until now, species delimitation and phylogenetic relationships within this genus have been improperly addressed due to the rapid radiations and hybridization [13]. Therefore, a further study based on genomic information is crucial to explore the phylogenetic evolution of Ulmus and accelerate the effective conservation and utilization of Ulmus germplasm.

Ulmus mianzhuensis is an endemic species from Ulmus which is distributed in Sichuan, China [14]. Based on the morphological trait, U. mianzhuensis is assigned to the sect. Microptelea as flowering phase of this section usually appears in autumn. Currently, sect. Microptelea is composed of two species, U. crassifolia and U. parvifolia. Geographically, U. mianzhuensis can be easily discriminated from the other two, because U. parvifolia is mainly distributed in East Asia, and U. crassifolia is restricted to North America [2]. The variability of bark color, petioles, and pedicel length have been recognized as the main keys of taxonomy among the different species. Specifically, the bark is grey and light brown in U. parvifolia and U. crassifolia respectively, but it is dark gray in U. mianzhuensis. Petioles length is very short (less than 2 mm) in U. crassifolia, but it is 2–6 mm and 3–5 mm in U. parvifolia and U. mianzhuensis respectively. Besides, the pedicel length of U. parvifolia is over 8 mm, which is short in other species. A previous study suggested that U. mianzhuensis is highly similar to U. parvifolia in identical flowering phase, pale pubescent at the stigma of samara and seed locating the middle to upper of samara. Whereas, molecular systematics is still sparse to determine the taxonomic status of U. mianzhuensis and the relationship between U. mianzhuensis and relative species remains uncertain.

Due to uniparental inheritance, and large copy numbers in plant cells without recombination, chloroplast genome has been widely applied in phylogenomic studies in the recent years, resolving deep relationships of particularly recalcitrant lineages undergoing recent radiations [15,16,17]. Previous studies demonstrated that the structural variation and abundant phylogenetic information of cp genomes are useful to solve the evolutionary relationship of complicated species [18, 19]. Although the cp genome size and gene structure are usually conserved, it can also provide insights into the molecular evolutionary patterns to help with species discrimination. For example, comparative studies showed that structural variation also occurred at the genus level and presented specific phylogenetic signals, such as Amphilophium, Corylus and Epimedium [20,21,22]. Except for structural variations among species, the adequate evolutionary information in cp genome sequences can significantly improve the resolution of deep phylogenetic relationships, such as the backbone tree of major angiosperms, and the identification of closely related species in Rhododendron and Pimelea [23, 24]. Therefore, combining structural variation and phylogenetic information in cp genome provides an important approach for the interspecific phylogenetic relationship and identification. In the present study, we sequenced the complete chloroplast genome of U. mianzhuensis and the objectives of this study were as follows: (1) to investigate the interspecific structural variation in the cp genome of U. mianzhuensis and other related species from the genus Ulmus; (2) to screen highly variable hotspot regions and simple sequence repeats candidate sequences for species identification and genetics resources; (3) to resolve the systematic position of U. mianzhuensis and reconstruct phylogenetic relationships between U. mianzhuensis and relative species among Ulmus based on cp genomes. These results will enhance our understanding of the evolution of the genus Ulmus and its close relatives.

Results

Organization and features of chloroplast genome

Our results found that U. mianzhuensis and all the relative species of Ulmus possessed a typical quadripartite structure consisting of a pair of identical IRs separated by LSC and SSC regions (Fig. 1). The genome size of the 31 Ulmus species ranged from 158,742 bp (U. lanceifolia) to 159,795 bp (U. microcarpus). The largest LSC, SSC, IR were found in U. microcarpus (88,408 bp), U. macrocarpa (19,038 bp) and U. mianzhuensis (26,546 bp), while the smallest LSC, SSC and IR were found in U. lanceifolia (87,170 bp), U. laevis (18,650 bp) and U. pumila (26,288 bp), respectively (Table S1). Besides, all the cp genomes from Ulmus shared the same GC content of 35.5%. A total of 113 unique genes, comprising 79 protein-coding genes, 30 tRNA genes, and 4 rRNA were enrolled in Ulmus species (Fig. 1). Among the 113 distinct genes, 15 genes (trnK-UUU, rps16, atpF, rpoC1, trnL-UAA, ndhB, trnA-UGC, ndhA, trnG-GCC, trnV-UAC, trnI-GAU, rpl2, rpl16, petB, petD) contained one intron and three genes contained two introns (ycf3, clpP, rps12) in the Ulmus cp genomes.

Fig. 1
figure 1

Physical maps of Ulmus mianzhuensis chloroplast genome. The direction of transcription is indicated by arrows. The gene function is color-coded as shown in the legend. The darker gray in the inner circle shows the GC content, while the lighter gray shows the AT content. LSC, large single-copy region; SSC, small single-copy region; IRa, IRb, inverted repeat A and B, respectively

Full size image

Comparison of chloroplast genome among Ulmus species

Based on our results, gene content and order were relatively conserved among the Ulmus cp genomes, and no rearrangement occurred in gene organization (Fig. 2). The divergence of cp genomes plotted using the mVISTA program presented lower sequence divergence in IR regions than in SC regions (Fig. S1). When compared to the diversity of nucleotide divergence between U. mianzhuensis and U. parvifolia, our result showed a relatively low-level nucleotide variation of the cp genome (0.0003). The nucleotide divergence for all Ulmus cp genomes ranged from 0 to 0.02876, and the hotspots variation region with the divergence over 0.025 was found in ndhC-trnV-UAC, ndhF-rpl32, psbI-trnS-GCU (Fig. 3A). Meanwhile, the average nucleotide diversity in the SC region is higher than that in the IR region, which is same as shown in mVISTA (Fig. 3B). We further found the OG groups had the highest median value of nucleotide variation from the functional groups for all protein-coding genes, while genes associated with ATP synthase (ATP), photosystems I (PI) and photosystems II (PII) had a lower nucleotide diversity (Fig. 3C). Comparison of adjacent genes among the Ulmus species uncovered genes ycf1 was located in the junction of IRb/SSC and rps19 was located in IRa/LSC boundary. Nevertheless, the genes in the IR border position presented slight variations among these Ulmus species (Fig. 2). The ycf1 gene occurring on the SSC/IRb boundary weakly varied with 1483–1490 bp located in the IRb region across sections of Ulmus. Similarly, the rps19 occurring on the IRa/LSC boundary had 30–148 bp located in the IRa region.

Fig. 2
figure 2

Chloroplast region (IR/SSC/LSC) junctions of eight representative Ulmus chloroplast genomes from five sections and different series. JSA: junction between SSC and IRa; JSB: junction between SSC and IRb; JLA: junction between LSC and IRa; JLB: junction between LSC and IRb

Full size image
Fig. 3
figure 3

(A) Sliding-window analysis of the chloroplast genomes for Ulmus species. The window length is set as 600 bp with the step size as 200 bp. The X-axis corresponds to the midpoint position of a window and Y-axis shows the nucleotide diversity (Pi) of each window; (B) The nucleotide diversity (Pi) in chloroplast regions (IR/SSC/LSC); (C) The nucleotide diversity (Pi) of different functional groups and the detailed information for functional groups could be found in Table S3

Full size image

Simple sequence repeats and long repeats sequences analysis

A total of 129 simple sequence repeats (SSRs) were identified in U. mianzhuensis mainly composed of four types of SSRs (Fig. 4). Among these SSRs, the majority of SSRs were mononucleotides (86.82%), eight were dinucleotide repeats (6.20%), eight were tetranucleotide (6.20%), whereas only one was trinucleotide repeat (0.08%). We also identified different types of repeats sequences within U. mianzhuensis, about 41.02% were forward repeats, 34.62% were palindromic repeats, 19.23% were reverse repeats, and only 5.13% were complement repeats screened in U. mianzhuensis (Fig. 4B). Furthermore, the SSRs distribution among species of representative sections showed the same trend in types and number of SSRs (Fig. 4C). A total of 120 ~ 133 SSRs were identified among these species, and the highest and lowest number of SSRs were found in U. bergmanniana/U. glaucescens and U. alata respectively. Among these SSRs, over 80% were mononucleotides, and hexanucleotides were only found in U. serotina and U. lanceifolia.

Fig. 4
figure 4

(A) The type and distribution of SSRs in the U. mianzhuensis chloroplast genome. (B) The proportion of four repeats types in U. mianzhuensis chloroplast genome; (C) Number of identified SSR repeat types in the eight representative Ulmus chloroplast genomes from five sections and different series

Full size image

Phylogenomic analysis and PCGs substitution rates within Ulmus

Base on the 32-taxon data matrix, we obtained a well-resolved and highly consistent phylogeny including the two main clades of Ulmus via ML and BI methods (BS = 100, PP = 1; Fig. 5). From our phylogeny, five Ulmus sections could not be formed to the monophyletic groups, and similar results exhibited in the series level. For the clade I, there were three subclades including two species of sect. Microptelea and almost species of sect. Ulmus with U. lanceifolia in the basal group (BS = 100, PP = 1). Notably, our target specie U. mianzhuensis was clustered into subclade II together with U. parvifolia in sect. Microptelea (BS = 100, PP = 1). Within the subclade III of sect. Ulmus, the species were separated into two clusters, which were distributed irregularly at the series level. For the clade II, U. elongata from sect. Chaetoptelea was evolved independently from other two species (U. thomasii and U. alata; BS = 100, PP = 1), of which U. thomasii was clustered to U. serotina and U. crassifolia. The sect. Blepharocarpa did not appear monophyletic group with U. rubra embedded in it. For 79 concatenated protein-coding genes, the partitioned scheme has obtained highly congruent phylogeny with cp genome of strongly supported bootstrap values (Figs. 5 and 6). Whereas, the systematic position for U. alata, U. davidiana and U. lamellosa is incongruent with the cp genome phylogeny (Fig. 6). When compared the concatenated tree to species tree from ASTRAL-III, we found they had highly consistent phylogenetic backbone of Ulmus, with U. lanceifolia resolved as an isolated clade (Fig. 6). There were still a few alternative topologies among species including U. davidiana, U. microcarpus, U. pumila, U. lamellosa and U. wallichiana with low support. The analyses of the selection pressure on the 79 protein-coding genes within Ulmus plastomes by CODEML indicated that two genes atpF and rps15 were under positive selection (adaptive selection) (Fig. 7A). Except these genes, the other genes were probably under neutrality or purifying selection. The unknown (CSF) genes, DNA-dependent RNA polymerase genes (RPO), other genes (OGs) and small subunit of ribosome genes (RPS) had higher median values of dN/dS (Fig. 7B).

Fig. 5
figure 5

Phylogram of Maximum likelihood (ML) tree of Ulmus based on chloroplast genomes. Only the posterior probabilities (PP) and likelihood bootstrap (BS) values were separated by a slash and only the support values less than 100/1.00 was shown at nodes

Full size image
Fig. 6
figure 6

Comparison of the topology of Ulmus based on plastid protein-encoding genes. The left is a phylogenetic tree based on concatenated genes based on ML and BI methods. Only the maximum likelihood bootstrap (BS) values and the posterior probabilities (PP) less than 100/1.00 was shown at nodes, ‘-’ indicates the support less than 50/0.50, the tip labels with a blue star showing the topology conflicts between chloroplast genome and protein-encoding genes; (b) phylogeny from the coalescent method in ASTRAL-III, and only local posterior probabilities (LPP) below 1.00 was shown in the tree. Conflicted taxa are highlighted in red font and connected by black lines between two phylogenetic trees

Full size image
Fig. 7
figure 7

(A) The nonsynonymous (dN), synonymous (dS) substitution rates and dN/dS value of all plastid protein-coding genes (PCGs) based on PAML; (B) dN/dS value of different functional protein-coding genes (PCGs) groups

Full size image

Discussion

The cp genome variations of Ulmus and its utility in DNA barcoding

In the current study, the cp genomes of gene content and order are highly conserved among the Ulmus species and all the Ulmus genomes shared the typical quadripartite architecture and the same GC content (35.5%). The junction genes located in SC/IRs boundaries were stable for Ulmus species with rps19/ycf1 genes lying in LSC/IR and SSC/IR boundaries respectively that are identical with most angiosperm plants [25]. Previous studies found that IR expansion/contraction could alter the structural conservatism of the cp genomes, which partly explain the absence of rearrangement events and variations of the Ulmus [26, 27]. Significant differences in IR expansion/contraction have been reported in other genera (e.g., Anemopaegma, Pedicularis) that could be useful signals for species identification at an interspecific level [22, 28]. Moreover, this stability of Ulmus cp genomes might be due to its recent diversification, or related to the conservative ecological niches within the genus [1]. Variations in length occurring on the junctions of IR/SC regions between different Ulmus sections were still detected, just as we found the variation of nearly 1 kb between the smallest (sect. Lanceifoliae, U. lanceifolia) and largest (sect. Ulmus, U. microcarpus) genomes, respectively (Fig. 2). The cp genomes size is usually closely associated with the contraction/expansion of the IR region or gene losses as shown in previous studies [29]. For instance, extreme IR expansion has caused a higher genome size in Pelargonium [30] and lots of gene losses had shrunk chloroplast genome size in some parasitic plants and conifers [31,32,33]. Our result confirmed that IR contraction of SSC regions was responsible for the relatively smaller size of U. lanceifolia (sect. Lanceifoliae) and IR expansion of SSC regions was partly result in the larger size in U. bergmanniana (sect. Ulmus).

Complete cp genome sequences have been regarded as super barcodes and are increasingly used to solve taxonomic problems among closely related groups. Both structural variation and valuable evolutionary information from cp genome are sufficient to discriminate genetically close species [34,35,36]. Although the quadripartite architecture in our current study was highly conserved within Ulmus, abundant information in cp genome could provide sufficient and full resolution for species discrimination. Regardless of the fact that trnH-psbA, rbcL and matK have been recommended as the core barcodes in land plants, such as Calycanthaceae [37], Saxifragales [38] and ferns [39, 40]. Our analyses found three noncoding regions (psbI-trnS-GCU, ndhC-trnV-UAC, ndhF-rpl23) with high levels of genetic divergence between sections and species of Ulmus were effective to species distinguishing than the universal barcodes (Fig. 3). Thus, compared to the universal DNA barcoding, these variations hotspots exhibited higher solution for taxonomy in the Ulmus with priority to be used as future barcodes for species identification. Furthermore, cp genome data revealed great potential to resolve the phylogenetic relationships in the genus Ulmus with full resolution.

Signature of positive selection on plastid genes

The evolutionary rate estimation based on dN/dS is widely used in phylogenetic and evolutionary studies. In most protein-coding genes, nucleotide substitutions of synonymous occurred more frequently than non-synonymous, with most of them were under neutral and purifying selection [41]. Overall, Ulmus species were subjected to a purifying selection, which helps to eliminate the disadvantageous mutations in populations evolution as the almost other genes (Fig. 7). However, our analyses also found that two genes were significantly under positive selection (dN/dS > 1) among the protein-coding genes within Ulmus (i.e., atpF, rps15), The atpF gene is one of photosystem subunit genes and participates in the encoding of the H+-ATP subunits, with essential roles in photosynthetic processes [42]. This gene has experienced positive selection in other land plants including Cucurbitaceae [43], Liliaceae [44] and Zingiberaceae [45], which functioned as the main source of energy for living cells and multicellular organisms. The rps15 is a small ribosomal protein which involved in the regulation of chloroplast translation. Empirical test found that knockout of the rps15 gene in tobacco have caused a definite reduction in small 30 S ribosomal subunits [46]. The positive pressure of rps15 is also identified in Araliaceae [47], Fabaceae [48] and Rhizophoraceae [49], suggesting the important role to adapt to their living environment. The Ulmus species are widely distributed in East Asia and North America, and have been undergoing adaptive evolution in response to stressful environments in heterogeneous habitats. Aa a result, these positive selection genes will enrich the Ulmus variety and adaptability during long term evolution.

The systematic position of U. mianzhuensis and reference to the phylogenomic relationship within Ulmus

Based on our phylogenomic results from cp genomes and protein-coding datasets, U. mianzhuensis was resolved as a sister group to U. parvifolia (Fig. 5; BS = 100, PP = 1), which demonstrated its systematic position in sect. Microptelea. Notably, nucleotide diversity between U. mianzhuensis and U. parvifolia showed low level variation of the cp. genome, suggesting that U. mianzhuensis might be merged into U. parvifolia and regarded as a subspecies of U. parvifolia. According to recent taxonomic systems [2, 8], sect. Microptelea is comprised of two species (U. crassifolia and U. parvifolia) and of which shared fascicle-cyme, equal pedicel and flowers appearing in fall. However, our phylogenomic analyses found that these species were not a monophyletic group, with U. crassifolia clustered to U. serotina (sect. Trichoptelea) (Fig. 5; BS = 100, PP = 1). Both U. crassifolia and U. serotina are mainly distributed in Southeastern North America, although they are differentiated in length and type of inflorescences. Therefore, our result suggested U. crassifolia should be separated from the sect. Microptelea, and merged to sect. Trichoptelea, which is consistent with previous molecular evidence [10]. Besides, the conflict between morphological taxonomy and phylogenetic topology was also found in sect. Ulmus, which was the most diverse group with the largest number of species. The particular species, U. lanceifolia has been basal species from the sect. Ulmus, which was only an evergreen tree and limited to Southeast Asia and South regions of China (Fig. 5; BS = 100, PP = 1). The position of U. lanceifolia has been slightly resolved in previous results, as Wiegrefe et al. (1994) and Fu (1980) has assigned it to sect. Lanceifoliae and sect. Ulmus successively. Our results revealed that U. lanceifolia should be independent from sect. Ulmus and supported the resumption of sect. Lanceifoliae rather than one of series of sect. Ulmus. We also found that U. rubra from sect. Ulmus was close to two species (U. americana and U. laevis) from sect. Blepharocarpa (Fig. 5; BS = 100, PP = 1). Morphologically, U. rubra is highly similar to U. americana, including a ciliate samara margin or pubescent samara body, which is totally different from sect. Ulmus. Besides, three species from sect. Chaetoptelea formed a polyphyletic group, in which U. alata and U. thomasii was clustered to sect. Blepharocarpa and sect. Trichoptelea respectively, and U. elongata was independent from them.

A satisfactory agreement for the backbone of Ulmus is found between the gene tree and species tree (Fig. 6). However, there are still several significant incongruences, which are mostly distributed in the shallow nodes with low or mid support (Fig. 4). Given that the backbone of Ulmus phylogeny is highly consistent between different dataset (i.e. cp genome, PCGs and noncoding regions) and methods (i.e. BI, ML) from cp genome, plastomic data is still sufficient for phylogenetic resolution of this genus. The prevalent conflicts have been focused on the five species from sect. Ulmus (Fig. 6), which may be caused by incomplete lineage sorting, network evolution and polyploidy. Within Ulmus, the extent of hybridization has been reported from population genetics [50, 51], and also inferred by inconsistent phylogeny between nuclear and cp genomes in the previous studies [9, 10]. The chromosomes number of Ulmus is relatively stable with 2n = 28 in most of species, which might rule out as the cause of ancient polyploidy [8]. In summary, phylogenomic analysis from nuclear datasets including genomes and transcriptomes are urgently needed for better understanding the species diversity and evolution for Ulmus in the future studies.

Conclusions

Our current study provided comprehensive insights into the whole cp genome organization and content of U. mianzhuensis. Although the structure and gene content evolution of cp genomes are highly conserved within Ulmus species, we still found structural variations in IR/SC boundary at the section levels. Furthermore, important genetic characteristics including repetitive sequences, SSRs, and sequence divergence were explored to population genetics and DNA barcoding. We also reconstructed the phylogenomic relationship within Ulmus and provided an important foundation for further revision and adaptive evolution of Ulmus. In future, we will focus on conflicts between the gene tree and species tree with extensive samplings and nuclear datasets to verify the potential mechanism for the recent radiations in Ulmus.

Materials and methods

Ethical statement

For the collection of samples for this study, no special licenses were needed. The relevant Chinese laws were followed as this research was conducted.

Taxon sampling, sequencing and assembly

Fresh leaves of U. mianzhuensis were collected from Central Forest Tree Nursery in Sichuan, China and stored in silica gel prior. The voucher specimen (Deng11468) was deposited in the Herbarium of Henan Agricultural University. For this species, 50 mg dried leaves were ground and high-quality genomic DNAs were extracted using a Plant Genomic DNA Extraction Kit (Tiangen, Beijing, China), and then were subsequently sent to Novogene (http://www.novogene.com, China) for short insert (350 bp) library construction and next-generation sequencing by Illumina Hiseq 4000 genome analyzer platform (Illumina, San Diego, CA) by Novogene, Beijing, China. Raw reads from the paired-end for quality were filtered with the NGSQC ToolKit by removing adapter sequences and low-quality reads with Q value B20 [52]. Then the clean data were assembled by NOVOPlasty 2.6.3 [53]and annotated by GeSeq [54], and the results were manually checked and verified in Geneious v.9.1 according to Ulmus parvifolia (MT165940) [55]. The visualization of the cp genome map was performed in OGDRAW [56].

Comparisons analyses of cp genomes in Ulmus species

We obtained 31 cp genomes from NCBI representing all the five sections of Ulmus species (Table S1). The genome size GC content, gene content and number were firstly compared based on Geneious v.9.1 and manual correction [55]. Since the structure and organization of cp. genomes are highly conservative within sections, eight cp genomes were selected representing the five sections and different series to explore the genome variations among Ulmus species. We initially aligned all the complete cp genomes using MAFFT v7.0 and further checked by manual [57]. The cp genomes with annotation were compared by using the program mVISTA in shuffle-LAGAN mode with U. parvifolia as a reference [58]. We employed the IRscope script to generate and compared the variation of inverted-repeat (IR) and single-copy (SC) borders of the cp genomes from five sections. [59]. For simple sequence repeats (SSRs), we used MISA by setting the minimum number of repeats to 10, 6, 5 for mono-, di-, tri- repeats, and 3 for tetra-, penta-, and hexanucleotides repeats, respectively [60]. REPuter were used to identify and locate the repeat sequences with 30 bp minimum repeat size, 90% or greater sequence identity [61].

Evaluation of variation hotspots

In addition, DnaSP v5.0 was employed to analyze nucleotide diversity (Pi) for variation hotspots among the Ulmus species [62]. A sliding window was conducted to compare Pi value based on a 600 bp window length with a 200 bp step size. We further divided the all protein-coding genes (PCGs) into 11 groups and also evaluated the Pi value at a functional group level for all protein-coding genes (PCGs) and quadripartite structure to detect nucleotide heterogeneity (Table S3).

Phylogenomic analysis and adaptive evolution

A total of 32 species were used for our phylogenetic analysis including U. mianzhuensis and 30 relatives representing almost Ulmus species across the world, with Zelkova schneideriana (Ulmaceae) as outgroup (Table S1). For concatenated analysis, both cp genome and 79 protein-coding genes were compiled into a single file of the 32-taxon data set respectively and aligned with MAFFT v7.0 for phylogenetic inferences [57]. Maximum likelihood (ML) analyses were conducted using IQ-TREE [63]and the best substitution model and partitioning scheme were simultaneously implemented in ModelFinder [37] under the Bayesian information criterion (BIC) [64]. The Bayesian inference (BI) analysis was carried out using MrBayes v3.2 [65], and the Markov chain Monte Carlo analysis was executed for 100,000,000 generations, with four chains (one cold and three heated) sampled at every 1000 generations. The first 25% of the trees were discarded as burn-in, and the remaining trees were used to construct majority-rule consensus trees. The convergence of runs was estimated by Tracer v1.5 [66]. To further estimate a coalescent-based species tree, we first inferred individual gene trees using RAxML v8.2.11 under The GTRGAMMA model and with 1000 bootstrap replicates. The resulting gene trees were collapsed to infer a species tree with ASTRAL-III v5.6.3 [67] using local posterior probabilities to assess clade support. To detect the selective pressure of protein-coding genes in Ulmus, the sequences for each gene were aligned separately and the maximum likelihood phylogenetic tree from the protein-coding gene was used as a constraint tree. The synonymous (dS), non-synonymous (dN) nucleotide substitution rates and the dN/dS ratio (ω) were calculated using the codeml program of PAML v4.9 [68]. The pairwise dN and dS substitution rates between different taxa were calculated based on the custom selection model by setting CodonFreq prior as F3 × 4 model [69]. The dN/dS ratio was then calculated and compared at the 11 functional groups of all PCGs to detect evolutionary rate heterogeneity.