Molecular and morphological analyses clarify species delimitation in section Costatae and reveal Betula buggsii sp. nov. (sect. Costatae, Betulaceae) in China

Abstract

Background and Aims

Delineating closely related and morphologically similar species is difficult. Here, we integrate morphology, genetics, ploidy and geography to resolve species and subspecies boundaries in four trees of section Costatae (genus Betula): Betula ashburneri, B. costata, B. ermanii and B. utilis, as well as multiple subspecies and polyploid races.

Methods

We genotyped 371 individuals (20–133 per species) from 51 populations at 15 microsatellite markers, as well as a subset of individuals, using restriction-site associated DNA sequencing and nuclear internal transcribed spacers. We determined the ploidy level of eight individuals using flow cytometry and characterized leaf variation for a subset of 109 individuals by morphometric analysis.

Key Results

Integration of multiple lines of evidence suggested a series of revisions to the taxonomy of section Costatae. Betula costata and B. ermanii were found to be valid. Molecular and leaf morphology analyses revealed little differentiation between diploid B. albosinensis and some samples of B. utilis ssp. utilis. By contrast, other B. utilis ssp. utilis samples and ssp. albosinensis formed a morphological continuum but differed based on genetics. Specifically, B. utilis ssp. albosinensis was divided into two groups with group I genetically similar to B. utilis ssp. utilis and group II, a distinct cluster, proposed as the new diploid species Betula buggsii sp. nov. Phylogenomic analysis based on 2285 620 single nucleotide polymorphisms identified a well-supported monophyletic clade of B. buggsii. Morphologically, B. buggsii is characterized by elongated lenticels and a distinct pattern of bark peeling and may be geographically restricted to the Qinling–Daba Mountains.

Conclusions

Our integrated approach identifies six taxa within section Costatae: B. ashburneri, B. buggsii, B. costata, B. utilis ssp. utilis, B. utilis ssp. albosinensis and B. ermanii. Our research demonstrates the value of an integrative approach using morphological, geographical, genetic and ploidy-level data for species delineation.

INTRODUCTION

Species delineation based on morphology may be confounded by intraspecific variation among populations and limited differentiation between closely related species (Whittall et al., 2004; Leliaert et al., 2009; Wang et al., 2014b; Lissambou et al., 2019). Where species co-occur, this may be further exacerbated by hybridization and introgression (Bacon et al., 2012; Andújar et al., 2014), or may be impossible based on morphology alone due to cryptic speciation (Bickford et al., 2007; Fišer et al., 2018). Despite advances in phylogenetic methods, this has meant that many species-rich genera have remained unresolved, hindering our understanding of species ecology and evolution as well as limiting our ability to deliver effective conservation management.

Betula L. (Betulaceae) is such a genus with many taxonomic challenges. Comprising approximately 65 species and subspecies (Ashburner & McAllister, 2016), several span very broad latitudinal and longitudinal ranges and encompass substantial morphological variation (Wang et al., 2014b; Ashburner & McAllister, 2016). For example, B. platyphylla ranges from the Himalayas to Siberia and varies substantially in bark colour across its geographical distribution (Ashburner & McAllister, 2016). Similarly, species such as B. michauxii and B. nana are morphologically convergent but distantly related (Wang et al., 2016 , 2021). Analysis of Betula taxonomy is further complicated by frequent interspecific hybridization, introgression and polyploidy (Anamthawat-Jónsson & Tómasson, 1999; Thórsson et al., 2010; Wang et al., 2014a; Zohren et al., 2016; Tsuda et al., 2017; Bona et al., 2018; Ding et al., 2021). Betula species from different subgenera appear able to hybridize readily, such as between B. alleghaniensis and B. papyrifera (Thomson et al., 2015). Polyploids account for over half of the described taxa, ranging from diploid to dodecaploid, with cytotypes observed for some species, such as B. chinensis (6x and 8x) (Ashburner & McAllister, 2016).

As a result of these challenges, taxonomic treatments based solely on morphology or genetics may be misleading (Dayrat, 2005). For example, some researchers emphasize genetic distinctness among ‘species’ as a prerequisite for species delimitation (de Queiroz, 2005, 2007). However, genetic differentiation often exists among populations, which can elevate signals of structure when only portions of the species range are surveyed (Maliouchenko et al., 2007). Alternatively, species delineation may be based on visible and heritable morphological characters. However the degree of morphological variation used to define a species is subjective (Agapow et al., 2004), species with similar morphological characters can be genetically distinct (Sistrom et al., 2013; Fišer et al., 2018) and it is difficult to ascribe morphological characters to neutral processes or adaptive differentiation. To address these challenges an increasing number of studies have used multiple lines of evidence in species delineation, termed an integrative approach (Su et al., 2015; Balao et al., 2020; Newton et al., 2020).

In this study, we apply an integrated approach, combining multiple morphological and genetics methods, to resolve taxonomic uncertainty in the genus Betula, section Costatae (Table 1). Section Costatae includes two diploids: B. ashburneri discovered from south-east Tibet and reported to have distributions in north-west Yunnan and western Sichuan (McAllister & Rushforth, 2011), and B. costata distributed in northern and north-eastern China, Japan and Russian Far East (Ashburner & McAllister, 2016). Section Costatae also includes two tetraploids: B. utilis (subdivided into ssp. utilis and ssp. albosinensis) occurring from the Himalayas to north China without clear geographical and morphological intraspecific boundaries, and B. ermanii from north-eastern China, Japan and Russian Far East. Several varieties of these tetraploid species have also been named based on a limited number of herbarium specimens, such as B. utilis var. prattii, B. albosinensis var. septentrionalis and B. ermanii var. lanata (Ashburner & McAllister, 2016), though their taxonomic validity is unclear.

Table 1.

Detailed information on taxa of section Costatae used in the present study.

Species	Variety	Ploidy	Distribution	Reference
B. utilis	ssp. utilis (B. utilis [AM])	4x	From the Himalayas to the Hengduan Mountains	Ashburner and McAllister (2016)
	ssp. utilis (B. utilis [FC])	4x	The Qinling Mountains and the Hengduan Mountains	Li and Skvortsov (1999)
	var. prattii	4x	Kangding, western Sichuan (the Hengduan Mountains)	Ashburner and McAllister (2016)
	ssp. albosinensis	4x	the Qinling Mountains, the Hengduan Mountains and the Taihang Mountains	Ashburner and McAllister (2016)
	ssp. albosinensis (B. albosinensis [DA])	2x	the Qinling Mountains	Hu et al. (2019)
B. albosinensis	ssp. septentrionalis	4x	Western Sichuan (the Hengduan Mountains)	Ashburner and McAllister (2016)
B. ermanii	ssp. ermanii	4x	North-east China (mainly the Changbai Mountains)	Ashburner and McAllister (2016)
	var. lanata	4x	Russia: from the eastern shores of Lake Baikal, eastward to the Pacific coast	Ashburner and McAllister (2016)
B. costata	NA	2x	North-east China	Ashburner and McAllister (2016)
B. ashburneri	NA	2x	South-east Tibet	McAllister (2011)

NA, not applicable.

We integrate morphological, molecular [microsatellite genotyping, internal transcribed spacer (ITS) sequencing and restriction-site associated DNA sequencing (RAD-seq)] and ploidy analysis (flow cytometry) to resolve taxonomic issues within section Costatae. For example, microsatellite makers are powerful in detecting recent population differentiation cost-effectively and RAD-seq is powerful in inferring phylogenetic relationships of recently diverged species (Wang et al., 2021). Ploidy level can serve as a good indicator for reproductive isolation, especially for diploids and tetraploids. Based on multiple lines of evidence, we present a revised treatment of species boundaries within section Costatae.

MATERIALS AND METHODS

Study species

Section Costatae includes Betula ashburneri, B. costata, B. ermanii, B. utilis ssp. albosinensis and B. utilis ssp. utilis. Betula utilis ssp. utilis was described to have distributions only in south-western China according to the monograph of Ashburner & McAllister (2016) but was recorded to have distributions in central China according to the Flora of China (Li & Skvortsov, 1999) (Table 1). Recently, a cryptic diploid species of B. utilis ssp. albosinensis, namely ‘diploid’ B. albosinensis, was reported (Hu et al., 2019). To avoid confusion, we refer to B. utilis ssp. utilis described in Ashburner and McAllister’s monograph, B. utilis ssp. utilis described in Flora of China and ‘diploid’ B. albosinensis as B. utilis [AM], B. utilis [FC] and B. albosinenesis [DA], respectively (Table 1; Supplementary Data Fig. S1). Morphological differences among B. utilis [AM], B. utilis [FC] and B. utilis ssp. albosinensis are briefly described in Supplementary Data Section 1.

Sampling

Initial sample identification was based on location and micromorphology. All taxa were collected from naturally occurring woodlands, with between four and 15 populations for each assigned taxon (Fig. 1). Twig samples were collected between May and September of 2018–21, with each sample spaced by ~20 m. A GPS system (UniStrong) was used to record the location of each population. Detailed species and population sampling information is provided in Supplementary Data Table S1.

Fig. 1.

The distribution of samples used in the present study.

Morphometric analyses

For analyses of leaf shape among these species, we selected 6–48 individuals per taxon and sampled 371 leaves with one to five leaves per individual. Leaves were scanned individually using a Hewlett-Packard printer (LaserJet Pro MFP M128fn) with a resolution of 600 dpi. Thirteen landmarks were selected from each scanned leaf according to the protocols of Liu et al. (2018) and Hu et al. (2019). The 13 landmarks were converted to a configuration of 26 cartesian coordinates using ImageJ (Abràmoff et al., 2004). A Generalized Procrustes Analysis (GPA) was performed using the procGPA function in the R package ‘shapes’ (Dryden, 2019). Eigenleaves were visualized using the ‘shapepca’ function, and principal component scores, percentage variance and Procrustes-adjusted coordinates were obtained from procGPA object values.

DNA extraction and microsatellite genotyping

High-quality DNA was extracted from cambial tissues following a modified 2× CTAB (cetyltrimethylammonium bromide) protocol (Wang et al., 2013). Extracted DNA was assessed with 1.0 % agarose gels. Fifteen microsatellite loci developed for B. maximowicziana (Tsuda et al., 2009a), B. platyphylla var. japonica (Wu et al., 2002; Tsuda et al., 2009b), B. pendula (Kulju et al., 2004) and B. pubescens ssp. tortuosa (Truong et al., 2005) were used to genotype our samples (Supplementary Data Table S2), with the 5′ terminus of the forward primers labelled with FAM, HEX or TAM fluorescent probes. These microsatellite loci have a good cross-compatibility in multiple Betula species. Each microsatellite locus was amplified individually and was manually combined into four multiplexes. The PCR protocol followed Hu et al. (2019). Microsatellite alleles were scored using GENEMARKER 2.4.0 (Softgenetics) and checked manually. All the 371 individuals had no more than three missing loci and were retained for further analyses.

ITS sequencing

We detected a distinct genetic cluster of B. utilis ssp. albosinensis (group II, see below) and amplified nuclear ribosomal (nr)ITS for 53 individuals of this genetic cluster using primers ITS4 (White et al., 1990) and ITSLeu (Baum et al., 1998). Reactions were performed following Hu et al. (2019). PCR products were outsourced to Tsingke Company (Qingdao, China) for purification and sequencing. ITS sequences were deposited at NCBI with GenBank accession numbers OK050306–OK050358.

RAD-seq

A subset of 34 DNA samples were selected for RAD-seq using an Illumina HiSeq 2500 and 150-bp pair-end sequencing with the restriction enzyme PstI (Personalbio, Shanghai, China). These were combined with eight additional samples of section Costatae previously sequenced, using the same restriction enzyme in Wang et al. (2021). These samples represented six B. costata, six B. utilis [AM], six B. ermanii, 12 B. utilis ssp. albosinensis, seven B. utilis [FC] and one each of B. albosinensis [DA], B. ashburneri, B. ermanii var. lanata, B. albosinensis var. septentrionalis and B. utilis var. prattii (Supplementary Data Table S3). The raw data were trimmed using Trimmomatic (Bolger et al., 2014) in paired-end mode. Reads with a quality of below 20 within the sliding-window of 5 bp and reads shorter than 40 bp after trimming were discarded. Filtered reads of each sample were aligned to the whole genome sequence of B. pendula (Salojärvi et al., 2017) using the BWA-MEM v.0.7.17-r1188 algorithm in BWA (v.0.7.17) with default parameters (Li & Durbin, 2009). Non-specific mapped reads were discarded. All subsequent analyses were performed using SAMtools v.1.8 (Li et al., 2009) and GATK v.4.1.4 (McKenna et al., 2010; DePristo et al., 2011). These analyses include conversion of alignments into indexed binary alignment map (BAM) files, marking duplicates, calling genotypes and filtering single nucleotide polymorphisms (SNPs) (McKenna et al., 2010; DePristo et al., 2011). SNPs within a 50-kb window with r² > 0.5 and a minimum allele frequency (MAF) < 0.01 were removed to reduce linkage disequilibrium using BCFtools v.1.10.2 (Li, 2011). Prior to population structure analysis, we retained only sites with no missing data, resulting in 82 137 SNPs for downstream analyses. Sequences were deposited in the NCBI-Sequence Read Archive (SRA) repository (BioProjectID PRJNA761548).

Population structure analysis

A principal coordinate analysis (PCoA) was performed on microsatellite data of B. utilis [AM], B. utilis [FC], B. utilis ssp. albosinensis, B. albosinensis [DA], B. costata and B. ermanii using POLYSAT (Clark & Jasieniuk, 2011) implemented in R 4.0.2 (R Core Team, 2020), based on Bruvo’s genetic distances (Bruvo et al., 2004). For nucleotide SNPs, a principal component analysis (PCA) was carried out using the ‘adegenet’ R package 2.1.1 (Jombart, 2008).

Microsatellite data were analysed in STRUCTURE (Pritchard et al., 2000) to identify the most likely number of genetic clusters (K) with a ploidy of four. Ten replicates were performed with 1000 000 iterations and a burn-in of 100 000 for each run at each value of K from 1 to 10. We used the admixture model, with an assumption of correlated allele frequencies among populations. Individuals were assigned to clusters based on the highest membership coefficient averaged over the ten independent runs. The number of genetic clusters was estimated using the ‘Evanno test’ (Evanno et al., 2005) implemented in Structure Harvester (Earl & vonHoldt, 2012). Replicate runs were grouped based on a symmetrical similarity coefficient of > 0.9 using the Greedy algorithm in CLUMPP (Jakobsson & Rosenberg, 2007) and visualized in DISTRUCT 1.1 (Rosenberg, 2004).

The filtered SNPs were analysed in ADMIXTURE v.1.3.0, a model-based approach to assessing population structure in a maximum-likelihood (ML) framework (Alexander & Lange, 2011). We ran ADMIXTURE for K = 1–10 with 20 replicates for each K value and performed cross-validation error estimation in order to assess the most suitable value of K (Alexander & Lange, 2011). Replicate runs were aligned and visualized in pong v.1.4.9 with the greedy algorithm (Behr et al., 2016).

Phylogenetic analyses

To provide an additional line of evidence for the phylogenetic position of B. utilis ssp. albosinensis group II, we generated ITS sequence- and SNP-based phylogenies.

Sixty-nine additional ITS sequences from Betulaceae (Wang et al., 2016) were included to infer the phylogenetic position of B. utilis ssp. albosinensis group II. In total, 122 sequences were aligned using BioEdit v.7.0.9.0 (Hall, 1999) with default parameters.

We collated RAD-seq data of 20 Betula taxa representing genus-wide diploid species (Wang et al., 2021). The identity of the 20 sequenced Betula taxa was initially inferred via ITS sequences and genome size estimates (Wang et al., 2016). In addition, we included RAD-seq data of 17 samples generated in the present study. Alnus inokumae was selected as the outgroup (Supplementary Data Table S3). SNPs of a total of 38 taxa were concatenated into a supermatrix for phylogenetic analysis. SNPs with > 50 % missing data were excluded, resulting in 2285 620 SNPs.

We analysed the ITS alignment and the matrix of SNPs separately, with a rapid bootstrap analysis under a GTR + GAMMA nucleotide substitution model, with 100 bootstraps and ten searches using the ML in RAxML v.8.1.16 (Stamatakis, 2006). The phylogenetic tree based on the ITS alignment and the matrix of SNPs were visualized in FigTree v.1.3.1.

Ploidy determination

We conducted flow cytometry for eight samples of B. utilis ssp. albosinensis group II and one B. utilis [FC]. Fresh cambial tissues were co-chopped with internal standards Solanum lycopersicum in 0.5 mL of Extraction Buffer (Cystain PI absolute P, Partec GmbH, Germany) and then filtered into a tube containing 1 mL of Staining Solution (Cystain PI absolute P, Partec GmbH) with 50 µL of diluted propidium iodide (PI). Samples were incubated at room temperature in the dark for ~5 min. We analysed one replicate per sample with > 5000 nuclei measured using flow cytometry (Beckman Coulter CytoFLEX). The resulting histograms were analysed with CytExpert v.2.4.0.28. We also inferred the ploidy of the samples using the method described in Zohren et al. (2016). Briefly, we plotted the distribution of allele ratios from read counts at heterozygous sites with a minimum coverage of 30. We would expect a peak around 0.50 for a diploid and peaks close to 0.25, 0.50 and 0.75 for a tetraploid.

Taxonomic description of B. utilis ssp. albosinensis group II.

The taxonomic study is based on herbarium specimens and observations of live material by N.W. All specimens cited have been seen. Herbarium citations follow Index Herbariorum (Thiers, 2021), nomenclature follows Turland et al. (2018) and binomial authorities follow the International Plant Name Index (IPNI, 2021). Material of the suspected new species was compared morphologically with material of all other Chinese Betula taxa at Kew. Herbarium material was examined with a Leica Wild M8 dissecting binocular microscope fitted with an eyepiece graticule measuring in units of 0.025 mm at maximum magnification. The drawing was made with the same equipment using a Leica 308700 camera lucida attachment. The description follows the standard of Ashburner & McAllister (2016). The extinction risk assessment was made using the IUCN (2012) standard.

RESULTS

Morphometric analyses

Landmarks were first aligned using a GPA and then a PCA was conducted to visualize the major sources of shape variance of leaves from B. albosinensis [DA], B. utilis ssp. albosinensis, B. utilis [AM], B. utilis [FC], B. costata and B. ermanii. PC1 and PC2 produced largely overlapping clusters among B. albosinensis [DA], B. utilis ssp. albosinensis, B. utilis [AM] and B. utilis [FC], but B. costata and B. ermanii overlapped to a much lesser extent (Fig. 2A). The shape variance, represented by PC1, is mainly influenced by leaf length and leaf width (Fig. 2B).

Fig. 2.

(A) Principal component analysis (PCoA) of leaves of section Costatae species. (B) ‘Eigenleaves’ showing leaf morphs represented by principal components (PCs) at ± 3 s.d. and shape variance explained by each PC. Each dot represents a leaf.

Population genetic structure

PCoA based on microsatellite markers revealed five clusters, with the first three axes accounting for 43.8 % of the total variation (Fig. 3A). Betula utilis ssp. albosinensis forms two groups: group I overlaps substantially with B. utilis [AM] and B. ermanii whereas group II is separate from all the other species on coordinate 1 (Fig. 3A). Betula utilis [FC] and B. albosinensis [DA] overlap substantially whereas B. costata is separate from the remaining species on coordinates 2 and 3 (Supplementary Data Fig. S2A). Betula ermanii is separate from B. utilis [AM] on coordinate 3 with B. utilis ssp. albosinensis group I intermediate (Fig. S2A).

Fig. 3.

Principal coordinates analysis (PCO) and STRUCTURE results of section Costatae species at 15 microsatellite markers (A) and principal component analysis (PCoA) of section Costatae species at 82 137 SNPs (B), at K = 5 and 6 based on 15 microsatellite markers (C) and admixture analysis of section Costatae at K = 5 and 6 at the 82 137 SNPs (D). Squares with solid or dashed lines represent samples with RAD-seq data from Wang et al. (2021).

For the sequenced individuals, between 12 234 848 and 28 155 092 reads were retained for each individual (mean 18 862 242) after trimming and filtering (Supplementary Data Table S3). The number of variable sites of the sequenced individuals ranged from 5520 333 to 9735 507. Consistent with PCoA based on microsatellite markers, a PCA based on genotype calls for 82 137 SNPs supports the division of B. utilis ssp. albosinensis into the same two groups, with B. utilis ssp. albosinensis group II and B. costata separate from the remaining species and from each other (Fig. 3B). Betula utilis [AM] forms a cluster with the previously sequenced B. utilis ssp. albosinensis whereas B. ermanii and B. ermanii var. lanata form a cluster when PC1 and PC2 are considered (Fig. S2B). Betula albosinensis [DA] forms a cluster with one accession of B. utilis [FC] whereas the remaining accessions of B. utilis [FC] form another cluster. The two individuals of B. utilis ssp. albosinensis group I position between B. ermanii and four individuals of B. utilis ssp. albosinensis group I. Betula ashburneri forms a continuum with B. utilis [AM] and B. utilis [FC] on PC1 and PC3 (Fig. S2B).

STRUCTURE analyses based on microsatellite markers identified four clusters: (1) B. utilis [AM], B. utilis ssp. albosinensis group I and B. ermanii, (2) B. albosinensis [DA] and B. utilis [FC], (3) B. costata and (4) B. utilis ssp. albosinensis group II (Fig. 3C; Supplementary Data Figs S3 and S4). Admixture analysis based on SNPs showed that cross-validation error was smallest at K = 5 and slighter higher at K = 6. Four out of 20 replicates have an average pairwise similarity of 0.98 at K = 5 and 14 out of 20 replicates have an average pairwise similarity of 0.98 at K = 6 (Figs S5 and S6). At K = 6, B. utilis ssp. albosinensis group I is separate from B. utilis [AM] and two individuals of B. utilis ssp. albosinensis group I have genetic admixture from B. ermanii (Fig. 3D). Betula utilis ssp. albosinensis group II is separate from the remaining species at K = 3 and above (Fig. S6). Interestingly, B. albosinensis var. septentrionalis, B. utilis var. prattii and B. utilis ssp. albosinensis group I are genetically similar whereas B. utilis ssp. albosinensis and B. utilis ssp. utilis are genetically similar to B. utilis [AM] (Fig. 3D).

Ploidy level

The 2C genome sizes of B. utilis ssp. albosinensis group II and B. utilis [FC] ranged between 400.4 and 448.5 Mbp, indicating their diploidy. Bar charts of allele ratios at heterozygous sites of B. utilis ssp. albosinensis group II, B. utilis [FC], B. albosinensis [DA] and B. costata showed a peak around 0.50, and of B. utilis ssp. albosinensis group I, B. utilis [AM] and B. ermanii showed peaks close to 0.25, 0.50 and 0.75 (Supplementary Data Fig. S7).

Applying an integrated approach to identify Betula buggsii sp. nov.

Integrating molecular data (ITS, microsatellite data and SNPs), ploidy and a further examination of morphological characters, we confidently establish B. utilis ssp. albosinensis group II as a new species, namely Betula buggsii.

The phylogenetic tree based on ITS showed that B. buggsii samples formed a cluster, despite the low support value (Fig. 4A). The phylogenetic tree based on a matrix of 2285 620 SNPs showed that B. buggsii formed a monophyletic clade with 100 % support; this clade was basal to a clade of B. costata, B. ashburneri, B. albosinensis [DA] and B. utilis [FC] (Fig. 4B). Individuals of B. ashburneri, B. albosinensis [DA] and B. utilis [FC] intermixed and together formed a monophyletic clade with 100 % support (Fig. 4B).

Fig. 4.

Phylogenetic tree from the maximum-likelihood analysis of B. buggsii using ITS sequences (A) and using the supermatrix approach based on data from 2285 620 SNPs (B). The scale bars below indicate the mean number of nucleotide substitutions per site. Species were classified according to Ashburner and McAllister (2016). Values above branches are bootstrap percentages of > 50 %.

Taxonomic treatment

Betula buggsii Nian & Cheek sp. nov. (Fig. 5).

Fig. 5.

Betula buggsii. A, habit, leafy stem with female catkins; B, stem from current year’s growth showing elliptical lenticels; C, stem from previous season showing orbicular and elliptic lenticels; D, axillary bud; E, base of leaf-blade, abaxial surface; F, apex of leaf-blade, abaxial surface; G, detail of marginal teeth from F; H, apex of leaf-blade, adaxial surface; I, male catkins, immature; J, female catkin; K, fruiting catkin scale, outer face; L, side view of K; M, fruit.

Type:

China, Shaanxi Province, Mian County, 1736 m alt. 33.37°N, 106.57°E, fl. 20 May 2021, Nian Wang in DEDL006 (holotype K000593326; isotypes K000593327, SDAU, SDAU, PE).

Diagnosis:

Similar to Betula utilis D. Don, especially B. utilis ssp. albosinensis (Burkill) Ashburner & McAll., differing in the combination of characters of: the highly persistent peeling red-brown bark of the trunk and major branches; the presence of elliptic as well as circular lenticels on the current-year twigs; the long, parallel-sided tips of the leaf-blade marginal teeth and the fruiting catkin scales with sessile, suborbicular, patent lateral lobes which are broader than long. Single trunked tree, 4–30 m high. Trunk up to 60 cm in diameter, bark red-brown, peeling along lines of transverse lenticels into short, persistent, tattered sheets, festooning the trunk and branches giving a ‘bearded’ appearance. Twigs of current-year growth densely covered in translucent resin glands, lenticels yellow-white, longitudinally elliptic, slightly raised, 0.5–0.6 × 0.25 mm, epidermis red-purple, hairs sparse, white, straggling, 0.5 mm long, glabrescent; stem of previous seasons with internodes 20–36(–46) × 1.2–2 mm, lacking glands, glabrous, lenticels elliptic and orbicular, white, epidermis purple-black. Buds elliptic, 8 × 2 mm, scales brown, glabrous. Stipules caducous, oblong, 4.5–4.7 × 1.8–2.5 mm, apex triangular-acute, base truncate, outer surface moderately densely hairy with minute, white, appressed hairs ~0.02 mm long, margin with hairs 0.15 mm long. Leaf-blades ovate-oblong, often slightly asymmetric at base, (50–)52–65(–70) × (26–)28–38(–40) mm, acumen 4–9 mm long, base obtuse-truncate and abruptly shortly decurrent to the petiole, lateral nerves (7–)8–10(–11) on each side of the midrib, arising at ~45°, straight, at the margin ascending, fading, terminating in marginal teeth; tertiary nerves scalariform and as prominent as the tertiary nerves, forming a fine reticulum, cells 0.2–0.3 mm in diameter, axils of lateral nerves with cluster of ~10 white, randomly orientated, filamentous hairs 0.25–0.75 mm long, midrib and lateral nerves with scattered white hairs of the same length, spreading, more densely hairy bands parallel between the lateral veins, abaxial blade with scattered colourless resin glands; margin crenate, lateral nerve-ending teeth largest; interstitial teeth 2–5, increasing in number distally, ~1.25 mm long, proximal part of tooth shallow crenate, distil part nipple-like, ± parallel-sided, 0.75 × 0.3–0.4 mm, apex rounded. Petioles canaliculate, (8–)9–14(–15) mm × 0.7–0.8 mm, margins sparsely straggly-white hairy, blade/petiole length ratio ~5 : 1. Male catkins in clusters of 2–3, pendulous, 8–10 × 0.3 cm, length/breadth ratio ~30 : 1.

Female catkins green, erect at anthesis, in terminal pairs, pedicels unequal in a pair, 8–13 mm long, bracts absent, catkins 20–26 mm long, 2 mm broad excluding the patent scales, 4.5–5.5 mm broad including the scales.

Fruiting catkins:

Erect at maturity ~30 × 5 mm, scales with mid-lobes patent, 5 × 4.5 mm, margins glabrous or sparsely hairy, hairs ~0.1 mm, long, patent, otherwise ± glabrous, base thick, oblong ~1.5 × 1.3 cm; mid-lobes thick, orbicular to transversely elliptic, broader than long ~1.2 × 1.5 mm, apex rounded, base broad 8–9 mm wide, sessile, margin with sessile glands; central lobe narrowly oblong, 2.8 × 1 mm, tapering gradually and slightly to the rounded-apiculate apex, apiculus ~0.1 mm long. Seeds (immature) transversally elliptic, 0.75 × 1.1 mm, nutlet ~0.6 mm broad, wings 0.25 mm broad, nutlet and wings hairy at distal margin, hairs 0.15 mm long, spreading; styles 0.5 × 0.15 mm, not contiguous with the wings.

Range and habitat:

Chongqing city and Shaanxi and Hubei Provinces, China (Supplementary Data Fig. S8). Betula buggsii occurs in the Qinling–Daba Mountains in mixed forests with B. utilis [FC] and (or) B. luminifera at an altitude of between 1500 and 2400 m. We only discovered it from five localities with eight to 29 individuals confirmed from each locality. Given this situation, B. buggsii should be ranked as ‘Endangered’.

Etymology:

Betula buggsii is named after Prof. Richard J. A. Buggs, an evolutionary biologist from the Royal Botanic Gardens, Kew, and Queen Mary University of London, for his devotion to research on hybridization, phylogenetics and conservation of the genus Betula. The Chinese name of B. buggsii is ‘秦巴桦’ (qín bā huà).

DISCUSSION

In this study, we present a framework for species delimitation in section Costatae of the genus Betula, combing population genetic data, morphological variation, geographical distribution and ploidy information. In addition, we propose a new species of section Costatae using the integrated approach.

Advantage of an integrated approach in species delimitation

Morphological data alone may be insufficient in species delimitation (Hofmann et al., 2010). For example, we misidentified both B. utilis [FC] and B. utilis [AM] as B. utilis ssp. albosinensis group I in natural stands based on morphology. Our results also indicate it may be preferable to use different types of molecular markers, particularly where they differ in mutation rate and are therefore indicative of processes occurring over different timescales [e.g. SNPs and simple sequence repeats (SSRs)] (Bradbury et al., 2015; Borrell et al., 2018; Bohling et al., 2019). In our study, B. utilis ssp. albosinensis group I formed a genetic continuum with B. ermanii based on SSRs but distinct clusters based on genome-wide SNPs (Fig. 3CD). The two tetraploid species were formed by B. ashburneri and B. costata (Wang et al., 2021) and have adjacent distributions. That means their possible recent speciation history accompanied by continuous gene flow make a smaller number of SSRs probably insufficient in distinguishing the two species.

Apart from morphological and molecular data that are increasingly used for species delimitation and establishing new species, ploidy level could be used more widely, particularly in plant groups (Levin, 2019), such as Sorbus (Robertson et al., 2010; Pellicer et al., 2012) and Salix (Wagner et al., 2020) and presumably others which have not been surveyed. Different ploidy was used to separate species with a morphological continuum, such as silver birch and downy birch (Brown & Tuley, 1971), which were once regarded as one species (Linnaeus, 1753; Brown & Williams, 1984). Here we show that ploidy is important in distinguishing species of section Costatae from sympatric populations, such as the sympatric B. costata and B. ermanii, B. ashburneri and B. utilis [AM], and B. utilis [FC] and B. utilis ssp. albosinensis group I. Although measuring ploidy in the field is challenging, our study shows that both the number of heterozygous alleles at each microsatellite marker and the distribution of read counts covering heterozygous SNPs are robust indicators of the ploidy of section Costatae; and these methods can apply to other taxa.

Finally, geographical and altitudinal information is also helpful in assigning species of section Costatae. We found that both B. ermanii and B. costata have non-overlapping distributions with B. ashburneri, B. utilis ssp. albosinensis group I, B. utilis [AM] and B. utilis [FC], whereas the former two occur in north-eastern China. For sympatric species, we observed that B. ermanii occupied a higher altitude than B. costata, and B. utilis [FC] occupied a higher altitude than B. utilis ssp. albosinensis group I (N.W., pers. obser.). Similarly, B. ashburneri also occupied a higher altitude than B. utilis [AM] (McAllister, 2011).

A framework for species delimitation within section Costatae.

Our results suggest recognition of six taxa with an unexpected and previously unidentified taxon we name B. buggsii. The six taxa are B. ashburneri (referred to as B. albosinensis [DA] or B. utilis [FC]), B. costata, B. utilis ssp. utilis (referred to as B. utilis [AM]), B. utilis ssp. albosinensis (referred to as B. utilis ssp. albosinensis group I), B. buggsii (referred to as B. utilis ssp. albosinensis group II) and B. ermanii (Fig. 6AB).

Fig. 6.

A schematic illustration of species delineation within section Cosatate (A) and various sources of information used to distinguish species (B). Photos of each species are placed below names. (Note: the merging of B. albosinensis [DA], B. ashburneri and B. utilis [FC] under the name of B. ashburneri is merely for convenience, without taxonomic implications.).

(1) B. ashburneri

Here, we merge B. albosinensis [DA], B. ashburneri and B. utilis [FC] under the name B. ashburneri for convenience without taxonomic implications. Microsatellite markers and SNPs show an overlapping genetic cluster of B. albosinensis [DA] and B. utilis [FC] (Fig. 3A, B) and phylogenomic analysis including B. ashburneri shows a fully supported monophyletic clade of B. albosinensis [DA], B. ashburneri and B. utilis [FC] (Fig. 4B). Admixture analysis of SNPs shows genetic similarity of B. ashburneri to B. utilis [AM]. This was unsurprising given that B. ashburneri is a diploid parent of B. utilis [AM] (Wang et al., 2021). Ploidy level further supports the grouping of the three taxa. Betula ashburneri is a diploid based on chromosome number and genome size analysis (Ashburner & McAllister, 2016; Wang et al., 2016), consistent with the observation that B. albosinensis [DA] and B. utilis [FC] are diploids based on flow cytometry. The description of B. utilis [FC] as a tetraploid according to Flora of China should be taken with caution (Li & Skvortsov, 1999).

(2) B. costata

Betula costata is genetically differentiated from other species of section Costatae (Fig. 3). Betula costata and B. ermanii co-occur in some populations, but B. costata occupies a lower altitude and has different fruit, leaf and bark colour and ploidy level.

(3) B. utilis ssp. utilis

Despite occupying a morphological continuum with B. utilis ssp. albosinensis, molecular results support it as a genetically distinct unit. Betula utilis ssp. utilis is described from the Himalayas, north-western Yunnan and with an extension into western Sichuan where it is sympatric with B. utilis ssp. albosinensis.

(4) B. utilis ssp. albosinensis

Molecular analyses indicate that B. utilis ssp. albosinensis forms a distinct cluster with B. utilis ssp. utilis (Fig. 3D). Two individuals of B. utilis ssp. albosinensis (XLA01 and XLA32), collected from its northern distribution, show a genetic admixture between B. utilis ssp. albosinensis and B. ermanii (Fig. 3D), indicating their hybrid origin. The two individuals were close to the southern distribution of B. ermanii, suggesting hybridization is possible due to long-distance pollen dispersal.

(5) Betula ermanii

Betula ermanii should be recognized as a species based on morphological characters and distribution. Morphologically, B. ermanii shows apparent differences in fruit, leaf shape and bark colour with B. utilis ssp. albosinensis. Geographically, B. ermanii is distributed around the Changbai Mountains and in north-east China where B. utilis ssp. albosinensis is absent.

(6) Betula buggsii (discussed below)

Betula buggsii, a new species of section Costatae.

Our genetic analyses revealed a distinct cluster of B. utilis ssp. albosinensis (group II). Based on multiple lines of evidence, this justifies the description of a new species, which we describe as B. buggsii. Betula buggsii meets several criteria of species concepts: obvious morphological difference, molecular differentiation and phylogenetic monophyly. Phylogenetic analysis based on ITS sequences shows that B. buggsii samples cluster together (Fig. 4A) and phylogenomic analysis shows a fully supported monophyletic clade, which was placed within section Costatae (Fig. 4B). Betula buggsii is a diploid based on flow cytometry, number of heterozygous microsatellite alleles and distribution of SNPs at heterozygous sites, and it shows clear difference with B. utilis ssp. albosinensis in the patterns of bark exfoliation (Fig. 6A). Bark of Betula buggsii exfoliates along the elongated lenticels in stripes while that of B. utilis ssp. albosinensis exfoliates in large sheets or flakes (Fig. 6A). Like other Betula species (Tarieiev et al., 2019; Jadwiszczak et al., 2020), bark colour in B. buggsii varies substantially. Betula buggsii is unlikely to be of hybrid origin, for the following reasons. First, its sympatric species include diploid B. utilis [FC] as was revealed in the flow cytometry analysis, and tetraploid B. luminifera. We would expect B. buggsii to be triploid if hybridization occurs between B. utilis [FC] and B. luminifera. By contrast, B. buggsii is a diploid. Second, B. buggsii does not possess intermediate morphological characters between B. utilis [FC] and B. luminifera.

Misidentification of taxa within section Costatae.

Our research indicates misidentification of two B. utilis ssp. albosinensis accessions as B. albosinensis var. septentrionalis and B. utilis var. prattii and misidentification of one B. utilis ssp. utilis accession as B. utilis ssp. albosinensis (Supplementary Data Fig. S6). Betula albosinensis var. septentrionalis and B. utilis var. prattii were described by Ashburner and McAllister (2016) based on a very limited number of provenances in cultivation in the UK, making misidentification likely. Moreover, the three probably misidentified taxa are from Sichuan province where B. utilis ssp. utilis and B. utilis ssp. albosinensis co-occur. Hence, hybridization may occur, generating intermediates with considerable morphological variations.

Finally, for the challenging tetraploids within section Costatae, we propose that the most practical taxonomy is to treat populations in north-west Yunnan and the eastern and central Himalaya as B. utilis ssp. utilis; those from the Qinling Mountains as B. utilis ssp. albosinensis; and those from north-eastern China (e.g. Changbaishan) as B. ermanii. Populations collected from the region between north-western Yunnan and the Qinling Mountains may be hybrids between B. utilis ssp. utilis and B. utilis ssp. albosinensis, and those collected from between the north and north-east China may be hybrids between B. utilis ssp. albosinensis and B. ermanii. Further research is needed to characterize patterns of genetic admixture between these species and to guide future management of genetic diversity.

Conclusions

Species delimitation has shifted from studies using a single criterion towards an integrative approach using multiple lines of evidence or multiple species concepts. With the rapid emergence of advanced sequencing techniques, genomic data are powerful in resolving complex species relationships (Huang et al., 2020), but even for molecular data, multiple markers (i.e. multi-loci-based DNA barcodes, microsatellite markers and SNPs, as well as ploidy) are preferable. Our integrative analyses using different types of molecular markers together with spatial and morphological metadata identified six differentiated clusters, representing six taxa within section Costatae. We argue that with this approach, these six taxa may represent less biased species units. In addition, we identify a new species, B. buggsii, which is morphologically and genetically distinct and monophyletic based on phylogenomic analysis. We suggest that broader application of integrative approaches may have significant advantages in both taxonomy and conservation.

SUPPLEMENTARY DATA

Supplementary data are available at Annals of Botany online and consist of the following. Figure S1: A schematic illustration of the approximate range of each taxon of section Costatae in China. Figure S2: Principal coordinate analysis of section Costatae species at 15 microsatellite markers and principal component analysis of section Costatae species at 82 137 SNPs. Figure S3: The best number of clusters inferred using the ‘Evanno test’ method. Figure S4: STRUCTURE results of section Costatae at K values from 2 to 6 based on 15 microsatellite markers. Figure S5: The cross-validation error for each K value from 1 to 10. Figure S6: Admixture results at K values from 2 to 10 based on 82 137 SNPs. Figure S7: Plots of read count ratios for heterozygous sites covered by at least 30 reads. Figure S8: A distribution map of B. buggsii. Table S1: Detailed information on populations used in the present study. Table S2: Details of microsatellite primers used in the present study. Table S3: Detailed information of samples used for ITS and RAD sequencing.

FUNDING

This work was funded by the National Natural Science Foundation of China (31770230 and 31600295) and Funds of Shandong ‘Double Tops’ Program (SYL2017XTTD13).