A taxonomic study of Quercus langbianensis complex based on morphology and DNA barcodes of classic and next generation sequences

Abstract The taxonomy of Quercus langbianensis and its relatives in Vietnam and Cambodia have been revised based on evidence obtained from field observations, morphological comparison of herbarium specimens and molecular analyses using both classic and next generation DNA markers. Based on Bayesian inference using rbcL, matK and ITS regions and Neighbour-joining tree using genome-wide sequences amplified with multiplexed inter-simple sequence repeat (ISSR) primers (MIG-seq), the authors recognised ten species in the complex in Vietnam and Cambodia, three of which are newly described in this paper: Q. baolamensis sp. nov., Q. bidoupensis sp. nov. and Q. honbaensis sp. nov. These new species are all phenotypically similar to Q. langbianensis s. str. in having lanceolate to oblanceolate leaf shape, upper 4–5/6–serrated leaf margin, acute or acuminate leaf apex and bracts of cupule arranged in 5–9 rings but distinguished both morphologically and phylogenetically. In molecular phylogenetic reconstructions, Q. bidoupensis is not close to any other species. In the Bayesian tree, Q. honbaensis is sister to both Q. blaoensis and Q. camusiae that are found in the same locality but morphologically distinct and those three species are sister to Q. langbianensis s. str., while Quercus baolamensis is not sister to Q. langbianensis s. str. in both the Bayesian tree and MIG-seq tree. In addition, Q. cambodiensis and Q. baniensis previously reduced to Q. langbianensis s. lat. have been recognised as distinct species. Six species were in need of lectotypification and that is undertaken herein.

A taxonomic study of Quercus langbianensis complex based on morphology and DNA barcodes of classic and next generation sequences

Introduction
The genus Quercus L., with 400-500 species, is the largest genus in the family Fagaceae (Nixon 1993, Valencia-A et al. 2016).The genus is widely distributed in the northern Hemisphere including tropical montane forests in South East Asia and often dominant in temperate deciduous forests in East Asia, Europe and North America and desert scrubs in the Mediterranean (Nixon 1993, Hubert et al. 2014, Valencia-A et al. 2016).In Vietnam, 45 species of the genus Quercus have been recognised (Ban 2003, Ho 2003, Binh et al. in press) but taxonomic identities of some species remain to be revised.One of them is Quercus langbianensis Hickel & A.Camus (1921), described from Mt. Langbian of Lam Dong Province, southern Vietnam.Following previous studies including Deng et al. (2010), The Plant List (2013) adopted a broad concept of this species by treating the following seven names as synonyms: Q. baniensis A. Camus and Cyclobalanopsis faadoouensis Hu from mainland of China.However, the authors' recent comparison based on the collections of Q. camusiae and Q. cambodiensis from their type localities revealed that both Q. camusiae and Q. cambodiensis are distinct species from Q. langbianensis s. str.This finding triggered the re-examination of the taxonomy of Q. langbianensis s. lat.hereafter designated as "Q.langbianensis complex" and its similar species such as Q. auricoma A.Camus in which Q. cambodiensis was recently included (Tagane et al. 2017).Deng et al. (2010) studied the relationship of Q. camusiae, Q. cambodiensis and Q. langbianensis and concluded that the three species are phenotypically indistinguishable.However, their study was based on the comparison of a limited number of herbarium specimens.
In this study, specimens of the Q. langbianensis complex were observed and collected more widely: Mt.Hon Ba of Khanh Hoa Province (the type locality of Q. camusiae), some localities of Lam Dong Province (near the type locality of Q. langbianensis s. str.), Mt.Ba Na (the type locality of Q. baniensis) and Mt.Bokor of Cambodia (the type locality of Q. cambodiensis).In Mt.Hon Ba, Q. camusiae was found at the higher elevation whereas two additional morphologically similar but distinct species were found at the lower elevation.Observations in the field revealed that two neighbouring provinces of southern Vietnam, Khanh Hoa Province and Lam Dong Province, harbour the highest diversity of the Q. langbianensis complex including three unknown species.However, those species are phenotypically very similar to each other and evidence based on molecular analyses is needed to elucidate their identities and relationships.
Recently, molecular studies of the genus Quercus have succeeded in elucidating phylogenetic relationships within the genus by using multiple gene markers (Hubert et al. 2014, Simeone et al. 2016) or RAD-seq (Hipp et al. 2014, Cavender-Bares et al. 2015, Fitz-Gibbon et al. 2017).In this study, both classic multiple gene markers (rbcL, matK and ITS) and genome-wide markers have been employed using the next generation sequencing platform (MIG-seq; Suyama and Matsuki 2015) to clarify relationships of the species within Q. langbianensis complex.As in RAD-seq, MIG-seq provides genetic markers of relatively short sequence reads determined by the next generation sequencer, but it is obtained with a PCR-based procedure without restriction enzyme digestion steps and is widely applicable to field samples even with low-quality DNA and/or small quantities of DNA (Suyama and Matsuki 2015).
The purpose of this paper is to revise the taxonomy of the Q. langbianensis complex based on evidence obtained from field observations, morphological studies and molecular data from both classic and next generation DNA markers.In conclusion, 10 species in the Q. langbianensis complex, including the seven species treated as synonyms of Q. langbianensis (The Plant List 2013) and the remaining three undescribed species have been distinguished.The latter three species are described as Q. baolamensis, sp.nov., Q. bidoupensis sp.nov.and Q. honbaensis sp.nov.

Observations and collections in the field
The field surveys were carried out in 13 conservation areas (national parks, nature reserves and conservation areas) in Vietnam and one national park in Cambodia (Fig. 1).In Hon Ba Nature Reserve, Khanh Hoa Province, southern Vietnam, eight rectangular plots of 100 m × 5 m were placed at various locations from 225 m to 1,498 m altitude and girth and height for all the individual trees above 4 m tall within the plots were recorded (Table 1).The authors also recorded trees in the same way for the following localities: two plots at 1,553 m and 1,807 m in Bidoup-Nui Ba National Park, Lam Dong Province; three plots at 1,850 m, 2,225 m and 2,933 m in Hoang Lien National Park (Mt.Fan Si Pan), Lao Cai Province; four plots at 86 m, 650 m, 1,420 m and 1,720 m in Vu Quang National Park, Ha Tinh Province; two plots at 450 m and 1,274 m in Bach Ma National Park, Thua Thien Hue Province; and three plots at 833 m, 1,070 m and 1,376 m in Ngoc Linh Nature Reserve, Kon Tum Province.In these localities, general collections of vascular plants outside of the plots were also made, with particular attention to the species of Fagaceae.In addition to the above 6 conservation areas, general sampling of Fagaceae was undertaken in the following 6 conservation areas: Ba Na Nature Reverse, Da Nang Province; Ba Vi National Park, Bokor, Cambodia, 20 plots from 266 m to 1,048 m altitude were established and Q. cambodiensis was sampled (Zhang et al. 2016, Tagane et al. 2017).For each specimen collected, photographs were taken in the field and samples of silica gel-dried leaf pieces for DNA isolation were gathered.
Amongst the collections of Quercus, the authors regarded species having the following traits as members of the Q. langbianensis complex: mature leaves are 12-17 cm long, 3-5 cm wide, serrated along the upper 5/6 to 1/3 margin (although young leaves of Q. camusiae are often almost entire), acute or acuminate at apex, cuneate at base and hairy when young but almost glabrous when mature; cupule obconical or bowlor cup-shaped, bracts of cupule arranged in 5-9 rings and covers 1/4 to 2/3 of a nut that is ovoid or subglobose to ellipsoid.Q. auricoma in the Q. langbianensis complex was not included because mature leaves have entire margin and smaller size (5.5-7cm long, 2-2.7 cm wide, from E. Poilanei 13098 (P)).

DNA extraction
DNA was isolated from each silica-gel dried sample by the CTAB method (Doyle and Doyle 1987) with the following modifications: dried leaf material was milled by QIAGEN TissueLyser to obtain fine powder and washed three times in a 1 ml buffer (including 0.1 M HEPES, pH 8.0; 2% Mercaptoethanol; 1% PVP; 0.05 M Ascorbic acid) as in Toyama et al. (2015).

Next generation DNA sequencing -MIG-seq
For 105 samples, thousands of short sequences (loci) were amplified from each genome using primers designed for "multiplexed ISSR genotyping by sequencing" (MIG-seq, Suyama and Matsuki 2015) and presence/absence of each sequence (amplicon) were used in each sample for phylogenetic tree reconstruction regardless of whether it has SNP or not, as sequence-based dominant markers.The experimental standard conditions were performed following Suyama and Matsuki (2015).The 1st PCR step was performed to amplify ISSR regions from genomic DNA with MIG-seq primer set-1.The products of the 1st PCR were diluted 50 times for each 1st PCR product with deionised water.The 2nd PCR step was conducted independently to add individual indices to each sample using indexed primers.Then, 3 µl of each 2nd PCR product was pooled as a single mixture library.The mixture was purified and fragments in the size range 350-800 bp were selected by a Pippin Prep DNA size selection system (Sage Science, Beverly, MA, USA).Finally, the concentration of size-selected library was measured by using a SYBR green quantitative PCR assay (Library Quantification Kit; Clontech Laboratories, Mountain View, CA, USA) with approximately 10 pM of libraries that were used for sequencing on an Illumina MiSeq Sequencer (Illumina, San Diego, CA, USA), using a MiSeq Reagent Kit v3 (150 cycle, Illumina).

Phylogenetic analyses
For classical phylogenetic analyses, a phylogenetic tree was constructed by combining nucleotide sequences of the three DNA regions comprising rbcL, matK and ITS for 30 samples of 29 Quercus species and one Trigonobalanus verticillatus (as an outgroup).
All DNA sequences were newly generated in this study.The sequences were aligned by MEGA v7.0 (Kumar et al. 2016).For reconstructing phylogeny, a Bayesian method implemented in the programme BEAST v1.8.4 (Drummond et al. 2012) was used.The GTR + γ model of molecular evolution and an uncorrelated lognormal (UCLN) relaxed-clock model were selected to infer relative divergence times.In computation, the programme was started with a random tree and a tree prior that was useful for species-level was set according to a Yule process (Drummond and Rambaut 2007).Five independent chains of 100 million generations each were run with sampling every 10,000 generations.The first 1,000 trees were discarded as burn-in from each run.The remaining trees from each run were combined by using LogCombiner v 1.6.1 (Drummond and Rambaut 2007).Amongst the posterior distribution of 9,000 trees, the maximum clade credibility tree was identified using TreeAnnotator v 1.6.1 (Drummond and Rambaut 2007) with a posterior probability limit of 0.5 and median node heights.The congruence amongst rbcL, matK and ITS trees was tested using the incongruence length difference test (Farris et al. 1994) implemented in PAUP* 4.0b10 (Swofford 2003).As the incongruence was rejected (p=0.06), a combined tree using concatenated sequences was constructed.
For MIG-seq, raw data were pretreated from 105 samples and quality control was completed following Suyama and Matsuki (2015).The programme 'fastx_trimmer' in the FASTX-Toolkit (http://hannonlab.cshl.edu/fastx_toolkit/)was used to trim read 2 sequences including 12 bases of SSR region and two bases of anchor sequences in the 1st primers.The authors used option 'quality_ filter' of FASTX-Toolkit to select reads in which 40% or more sequences had quality scores Q30 or more.Then the TagDust programme (Lassmann et al. 2009) was used to remove the reads derived from extremely short library entries and to trim read 1 and read 2 sequences.Then, loci were assembled from the quality-filtered reads data with the de novo map pipelines (ustacks, cstacks, sstacks) in Stacks software package version 1. 35 (Catchen et al. 2011) and then a table prepared of presence/absence of loci in each individual from the outputs of the populations pipeline of Stacks 1.35.Using ustacks, homologous sequences (loci) were assembled in each individual with the following settings: minimum depth of coverage (m) = 10, maximum distance allowed between stacks (M) = 1, maximum distance allowed to align secondary reads to primary stacks (N) = 1 and maximum gaps = 2. Using cstacks, a catalogue of consensus loci was built for all the individuals by assembling loci in each individual assembled using ustacks, with the number of mismatches allowed between sample loci (n) = 2. Using sstacks, IDs of loci were associated in each individual with IDs of the consensus loci.Finally, presence/absence of loci were determined in each individual from a haplotypes list obtained using the populations pipeline.The populations pipeline output file haplotypes.tsvprovides genotypes of individuals at each locus.For each individual, the authors recorded a locus that had genotype information as "1" and a locus that had no genotype information as "0".The authors obtained a list of loci that were detected in at least one individual (1/105 = 0.01) with the following settings: all samples belong to the same population and threshold frequency of haplotype count in a population (r) = 0.001, a threshold one-order higher than 0.01.Using presence/absence (1/0) data of loci, the authors computed distance matrix, constructed a neighbour-joining (NJ) tree and examined the reliability of tree topology by bootstrapping with 1000 replicate using PHYLIP ver.3.695 (Shimada and Nishida 2017) as follows; 1000 times re-sampling with Seqboot, distance computation with Restdist, tree construction with Neighbour and consensus tree construction with Censense.The resulted tree was visualised with FigTree v1.4.3 (http://tree.bio.ed.ac.uk/software/figtree/).A phylogenetic analyses was first made for 105 samples including more Quercus species and then the sample size reduced to 31 by focusing on the Q. langbianensis complex.A total of 16,809 loci were used for the final phylogenetic tree.

Morphological and taxonomic comparison
The collections contain considerable numbers of sterile specimens including those from young trees that are often morphologically different from adult trees.Thus, after phylogenetic trees were obtained, morphological traits of leaves and shoots were carefully re-examined as well as reproductive organs if available and species were distinguished.If two OTUs are morphologically distinguishable and also not monophyletic on phylogenetic trees, these were regarded as two distinct species.Then, these were identified by a thorough literature review and comparisons with type specimen images available online (e.g.JSTOR Global Plants, http://plants.jstor.org/).In Q. langbianensis complex, lectotypification was needed for Q. baniensis, Q. blaoensis, Q. cambodiensis, Q. camusiae, Q. dilacerata and Q. donnaiensis.One of the co-authors, J.S. Strijk, examined specimens at P for lectotypification; selected for each species, was one of the specimens cited in the original description, which best represents the diagnostic traits of each species.

Observation in the field
In Hon Ba Nature Reverse, tree diversity was examined in eight plots of 100 m × 5 m and four species of Quercus (Table 1) were found including Q. poilanei and three species of the Q. langbianensis complex: Q. blaoensis, Q. camusiae and an undescribed species, Q. honbaensis.Quercus camusiae was found in the two plots at 1,336 m and 1,498 m altitude and one of canopy trees in the latter.Quercus honbaensis was found in three plots at 225 m, 400 m and 617 m altitude and occurred sympatrically with Q. blaoensis in the plot at 225 m altitude.Quercus honbaensis was one of the canopy trees at both 225 m and 400 m altitude (Table 1).In late February of 2014, Q. honbaensis had mature fruits and Q. blaoensis had young fruits.Two species were distinct in pubescence on young shoots (Q.honbaensis has long, very thin and curly hairs vs. Q.blaoensis has short, thicker and straight hairs).Quercus camusiae was distinct from Q. honbaensis and Q. blaoensis in that shoots and leaves were golden tomentose when young.
In Bidoup-Nui Ba National Park, approximately 100 km west of Mt.Hon Ba, tree diversity was examined in two plots at 1,553 m and 1,807 m altitude and Q. langbianensis s. str.was found at 1,553 m altitude.Quercus langbianensis s. str.was similar to Q. camusiae in having golden tomentose cupules, but different in distinctly toothed leaves and longer nuts (vs.almost entire or with only a few low teeth in Q. camusiae).The flora was surveyed above 800 m altitude in Bidoup-Nui Ba National Park and Q. camusiae and Q. honbaensis were not found.On the other hand, two additional and unknown species of the Q. langbianensis complex were found: Q. bidoupensis and Q. donnaiensis.Quercus bidoupensis was distinct from Q. langbianensis s. str. in having oblong-lanceolate leaves, acuminate and slightly caudate at apex and undulate and distinctly serrate along the upper half of the margin.Quercus donnaiensis was similar to Q. bidoupensis in leaf shape but differs in its margin not being undulate, serrated only near the apex and with 3-5 teeth.From the general collection in Lam Dong Province, three species of the Q. langbianensis complex were collected: Q. bidoupensis and Q. donnaiensis in Lam Tranh District and another undescribed species, Q. baolamensis, in Bao Lam District.
In Ba Na Nature Reserve and Son Tra Natural Conservation Area, central Vietnam, Q. baniensis of the Q. langbianensis complex and Q. poilanei and Q. auricoma of non-Q.langbianensis complex were found.
In the top plateau of Mt.Bokor, Cambodia, Q. cambodiensis of the Q. langbianensis complex and Q. augustinii of non-Q.langbianensis complex were collected.
A phylogenetic tree combining three DNA regions (rbcL, matK, and ITS) A total of 2,034 bases consisting of three DNA regions (657 bp for rbcL, 834 bp for matK and 543 bp for ITS) included 142 variable sites, amongst which 56 bases were parsimony-informative (Table 2).According to the Bayesian tree combining the three regions (Fig. 2), two major clades were supported by posterior probabilities higher than 80%: Clade 1 with 85 % posterior probability consists of five species of non-Quercus langbianensis complex (Q.poilanei, Q. kerrii, Q. austrocochinchinensis, Q. helferiana and Q. braianensis) and Clade 2 with 82 % posterior probability including seven species of the Q. langbianensis complex (Q.cambodiensis and six Vietnamese species) and five species of non-Quercus langbianensis complex (Q.neglecta nested with the Q. langbianensis complex and Q. annulata, Q. auricoma, Q. djiringensis and Q. macrocalyx).In Clade 2, Q. cambodiensis was sister to Q. neglecta with 81% posterior probability and clearly separated from the Vietnamese species of the Q. langbianensis complex (Q.langbianensis s. str., Q. baniensis, Q. blaoensis, Q. honbaensis, Q. baolamensis and Q. camusiae).Quercus langbianensis s. str.was sister to Q. blaoensis, Q. camusiae and Q. honbaensis with a strong branch support (PP = 1.00).Quercus camusiae was sister to Q. blaoensis with a high branch support (PP = 0.99).Quercus baolamensis and Q. baniensis were clustered together, but with weak branch support (PP = 0.64).Trees based on single gene sequences gave lower resolution but the ITS tree (see Suppl.material 1: Figure 1) supported the following points: (1) A clade, consisting of seven species of the Q. langbianensis complex and Q. neglecta, was supported by 100 % PP. (2) Quercus bidoupensis was not clustered with the other seven species of the Q. langbianensis complex.(3) A clade, including five species of non-Q.langbianensis complex, was supported by 70 % PP.In the cpDNA tree (see Suppl.material 1: Figure 2), neither the seven species of the Q. langbianensis complex nor the five species of non-Q.langbianensis complex was monophyletic.Neither Q. poilanei, Q. austrocochinchinensis, Q. helferiana and Q. braianensis was monophyletic whereas those four species were monophyletic in the ITS tree.
A phylogenetic tree using MIG-seq A neighbour-joining (NJ) tree based on MIG-seq for 31 samples of Quercus recognised three major clades excluding an outgroup of Trigonobalanus (Fig. 3).Clade M1 includes single species, Q. bidoupensis.Clade M2 with a 100 % bootstrap value consists of five species of non-Q.langbianensis complex (Q.poilanei, Q. kerrii, Q. austrocochinchinensis, Q. helferiana and Q. braianensis).Clade M3 with 100 % bootstrap value includes Q. neglecta, Q. macrocalyx, Q. auricoma and eight species of the Q. langbianensis complex.Within this clade, Q. cambodiensis was sister to Q. neglecta with a 74 % bootstrap value.Quercus honbaensis and Q. baolamensis were monophyletic with a bootstrap value of 100 %.Quercus donnaienis and Q. camusiae were also monophyletic with a bootstrap value of 75 %.Quercus blaoensis and Q. langbianensis s. str. of the Q. langbianensis complex are clustered with Q. baniensis, Q. auricoma, Q. macrocalyx, forming a clade with 82 % bootstrap value.

Discussion
The results of the three gene tree (Bayesian tree) and MIG-seq tree (NJ tree) were mostly consistent.First, five species of non-Q.langbianensis complex (Q.poilanei, Q. kerrii, Q. austrocochinchinensis, Q. helferiana and Q. braianensis) formed a highly supported clade, Clade 1 or Clade M2.This clade was supported also in the ITS tree.Second, three gene and MIG-seq trees matched in Clade 2 and Clade M3.Third, the species of the Q. langbianensis complex except Q. bidoupensis formed a highly supported clade (also in ITS tree) and Q. auricoma, Q. macrocalyx and Q. neglecta of non-Q.langbianensis complex were included in this clade (Fig. 4).Fourth, Q. cambodiensis was sister to Q. neglecta and separated from the Vietnamese species of the Q. langbianensis complex by relatively high supports (posterior probability 0.81 and bootstrap probability 74%).Fifth, Q. bidoupensis was placed in Clade 2 or Clade M1 and not close to the other species of the Q. langbianensis complex (also in ITS tree).The cpDNA tree did not support monophilies of Q. poilanei, Q. austrocochinchinensis, Q. helferiana and Q. braianensis that were monophyletic in the ITS tree, three gene tree and MIG-seq tree and thus the cpDNA tree alone provides less reliable evidence.
The consistent topology of three gene and MIG-seq trees (Figs 2-4) provided reliable evidence to resolve taxonomy of "species" currently treated as synonyms of Q. lang- bianensis s. lat.(Q.camusiae, Q. blaoensis, Q. cambodiensis and Q. baniensis).First, Q. cambodiensis is separated as a species because it is sister to Q. neglecta that is morphologically distinct in linear leaves and small nuts and has been treated as a distinct species in the Flora of China (Huang et al. 1999 as Cyclobalanopsis neglecta).Amongst the others, both Q. camusiae and Q. blaoensis are native in the Hon Ba Nature Reserve where Q. camusiae occurs at the higher elevation and Q. blaoensis occurs at the lower elevation (Table 1).As Q. camusiae and Q. blaoensis are sister to each other in the MIG-seq tree and not sympatric but paratactic in the distribution, those can be treated as two infraspecific taxa (varieties or subspecies) or two different species.Considering the morphological distinction described above, the latter treatment has been adopted.Quercus blaoensis co-occurs with another undescribed species: Q. honbaensis.As Q. blaoensis and Q. honbaensis are sympatric and morphologically distinct, those are recognised as different species.The monophyly of Q. honbaensis and Q. baolamensis was strongly supported in the MIG-seq tree with a bootstrap value of 100%.While Q. honbaensis occurs at an elevation lower than 617 m in Hon Ba Nature Reserve, Khanh Hoa Province, Q. baolamensis is collected at 1000 m from the Lam Dong province.Considering this distinction in cupule and nut morphology (Figs 6, 12), they are treated as two distinct species.While Q. baniensis is found in Da Nang Province of central Vietnam, the other five species (Q.baolamensis, Q. blaoensis, Q. camusiae, Q. honbaensis and Q. langbianensis s. str.) occur in Khanh Hoa or Lam Dong Province of southern Vietnam.From phylogenetic trees and morphological observations, it is difficult to relate Q. baniensis to any of the five species.In particular, a sister relationship between Q. langbianensis s. str.and Q. baniensis is not strongly supported (Fig. 4).Thus, Q. baniensis is treated as a distinct species.
Although the topologies of the three gene and MIG-seq tree are mostly consistent, there are some notable differences, particularly in Clade 2 and Clade M3 containing the Q. langbianensis complex (Fig. 4).In the Bayesian tree, based on the three regions of rbcL, matK and ITS, the monophyly of the Vietnamese species of the Q. langbianensis complex was only weakly supported (PP = 0.28), whereas it was strongly supported in MIG-seq tree (bootstrap value 100 %).This higher support in MIG-seq tree was obtained because MIG-seq provided more informative sites for constructing phylogenetic relationships amongst the species in the Q. langbianensis complex.Quercus auricoma and Q. macrocalyx were included in a clade of the langbianensis complex in the MIGseq tree but clustered with Q. annulata in the Bayesian tree.Further studies using more gene markers are needed to derive a conclusion on the placement of these two species.
A comparison, based on morphological characters both in the field and from dried specimens of the herbarium and the molecular evidence for the Q. langbianensis complex, revealed that Q. baniensis, Q, blaoensis, Q. cambodiensis, Q. camusiae and Q langbianensis s. str.are all distinct species (Table 3).In addition, it is concluded that three species amongst the Q. langbianensis complex are undescribed and below they are described  as Q. baolamensis, Q. bidoupensis and Q. honbaensis.In this study, only sterile specimens of Q. donnaiensis were collected and the sequence of ITS for Q. donnaienis could not be determined due to low DNA quality.In the MIG-seq tree, Q. donnaienis and Q. camusiae were monophyletic with a bootstrap value of 64 %.In vegetative traits, Q. donnaiensis is distinguished from Q. camusiae in having distinct serrations on the upper 1/3 of the leaf margin (vs.almost entire in Q. camusiae).Additional materials having fruits need to be examined to conclude whether those two are distinct species or infraspecific taxa.In the following taxonomic section, they are tentatively treated as two species.Amongst species treated as synonyms of Q. langbianensis s. lat.(Plant List 2013), Q. dilacerata is morphologically distinct as described in the following taxonomic section, but DNA samples of this species could not be obtained.Further studies using phylogenetic analyses are required to clarify the identity of Q. dilacerata.While Q. cambodiensis is treated as a synonym of Q. auricoma by Tagane et al. (2017), those two species are not sister to each other in both Bayesian and MIG-seq trees.The treatment of Tagane et al. (2017) is based on the broad concept of Q. auricoma adopted in the Flora of Thailand (Phengklai 2008) in which a species morphologically similar to Q. cambodiensis in northern and north-eastern Thailand is treated as Q. auricoma.However, after examining the collection of Q. auricoma from Son Tra (V3135, V3138) that is morphologically identical with the species of the type specimen of Q. auricoma, it is concluded that the species treated as Q. auricoma in the Flora of Thailand (Phengklai 2008) is different from genuine Q. auricoma, in that leaves are serrate along the upper 1/2-1/3 margin (vs.completely entire in Q. auricoma), nuts ovoid to oblong (vs.suborbicular) and cupules densely hairy (vs.less hairy).As far as is known, Q. auricoma is endemic to Vietnam.Further studies are needed to elucidate the identity of the species called "Q. auricoma" in Thailand.Distribution and habitat.VIETNAM.Da Nang Province: Ba Na Nature Reserve.In this study, this species was found along the roadside and edge of evergreen forest, at 707 and 789 m altitude.

Key to the species of
Additional specimens examined.VIETNAM.Ba Na Nature Reserve, 16°00'07.30"N,108°01'33.90"E,alt.707 m, 29 May 2015, Tagane S., Toyama H., Nguyen N., Nguyen C. Note.Camus (1935) illustrated Quercus baniensis in Chênes Atlas 2 (Pl.231) and later Camus (1936) effectively described this species based on the specimen Clemens 3455 collected from mountain Bani, Vietnam.The authors examined six specimens of Clemens 3455 in P, BM, MICH, U and US directly or by using digitised images on the web.Amongst them, Clemens 3455 in P [P00753998] was selected as the lectotype of Q. baniensis because the trait of a nut is well represented in this specimen.
Phenology.Fruiting specimens were collected in June.Distribution and habitat.VIETNAM.Lam Dong Province: Bao Lam District.At present, this species is known only from the type locality.Only one individual was found along the roadside and edge of evergreen forest, at 1,000 m altitude.
Etymology.The specific epithet is derived from the name of its type locality, Bao Lam District.Diagnosis.Similar to Quercus langbianensis s. str. in leaf shape, the number of secondary veins and basal scar of the nut convex, but distinguished in having bud oblong to ellipsoid (vs.globose to broadly ovoid), undulate and distinctly serrate leaf margin along the upper half (vs.regularly distinctly serrate in the upper 1/3), obconical cupules (vs.cup-shape), bracts of cupule arranged in 5-6 rings (vs.6-9 rings), and nut ovoid (vs.obovoid to ellipsoid).
Phenology.Unknown.Fallen fruits were collected in February.Distribution and habitat.VIETNAM.Lam Dong Province: Bidoup-Nui Ba National Park.At present, this species is known only from the type locality.
Etymology.The specific epithet "bidoupensis" is derived from its type locality.
Note.Camus (1935) described Quercus blaoensis based on the specimen Poilane 22372 from Vietnam.The authors examined specimens of Poilane 22372 in P ([P00754000], [P00753999]) and the digitised images of the specimens in K ([K000832201, K000832202, K000832203, K000832204]) and G [G00358072].Amongst them, only two specimens in P are fertile and only P00754000 represents diagnostic traits of a nut with cupule.Thus, the specimen Poilane 22372 deposited in P [P00754000] was selected as the lectotype for Q. blaoensis.nam: Pres de Nha-trang, massif de Honba") to replace the name Q. geminate Hickel & Camus (1923) without any collection number.Later, Camus (1938) Hickel and Camus (1929).Here, the specimen [P00753996] with more nuts was designated as the lectotype for Q. dilacerata.
Phenology.Fruiting specimens were collected in February.

Figure 1 .
Figure 1.Collection sites in Vietnam and Cambodia in this study, including eight national parks, four nature reserves and two conservation areas.

Figure 2 .
Figure 2. Bayesian phylogeny of 29 samples of Quercus and one Trigonobalanus (outgroup) based on rbcL, matK and ITS sequences.Braches are labelled with posterior probabilities.

Figure 3 .
Figure 3. NJ tree of 31 samples of Quercus and one Trigonobalanus (outgroup) based on presence/absence data of 16,809 MIG-seq loci.Branches are labelled with bootstrap supports (% of 1000 replicates).

Figure 5 .
Figure 5. Quercus baniensis A.Camus.A Leafy twig B Abaxial side of mature leaf C Infructescence and young fruits D Dried specimen.Materials: A, B from Hoang T.S. & Tagane S. V6922 C, D from Tagane et al.V3089.

Figure 6 .
Figure 6.Quercus baolamensis Binh & Ngoc.A Leafy twig B Abaxial side of mature leaf C Mature fruit D Inside of cupule E Nut.Materials: A-E from Ngoc et al.V3191.

Figure 7 .
Figure 7. Quercus bidoupensis Binh & Ngoc.A Leafy twig B Abaxial side of mature leaf C, D Side view and base view of the cupule, respectively E Inside of cupule F Nut. Materials: A-F from Tagane et al.V4328.

Figure 8 .
Figure 8. Quercus blaoensis A.Camus A. Branch with fruits B Young fruit C Dried specimen Materials: A-C from Toyama et al.V1366.

Figure 9 .
Figure 9. Quercus cambodiensis Hickel & A.Camus.A Leafy twig B Abaxial side of mature leaf C Infructescence and fruits D Nut E Basal scar of the nut.Materials: A-E from Tagane et al. 4302.

Figure 10 .
Figure 10.Quercus camusiae Trel.ex Hickel & A.Camus.A Branch with young fruit, B. Infructescence and young fruits C, D Abaxial side of young and mature leaf, E. Dried specimen.Materials: A-D from Tagane et al.V342 E from Toyama et al.V2173.

Figure 11 .
Figure 11.Quercus donnaiensis A.Camus.A Leafy twig B Infructescence, young fruits and abaxial side of mature leaf C Dried specimen.Materials: A, B from Tagane S., Wai J. V4398 C from Ngoc et al.V3208.

Figure 12 .
Figure 12.Quercus honbaensis Binh, Tagane & Yahara.A Leafy twig B Infructescence and mature fruits, C, D Side view and base view of the cupule, respectively, E. Inside of cupule, F. Nut.Materials: A-F from Toyama et al.V1378.

Figure 13 .
Figure 13.Quercus langbianensis Hickel & A.Camus.A Leafy twig B Abaxial side of mature leaf C Infructescence and mature fruits D Apex of the nut E Basal scar of the nut F Inside of cupule.Materials: A, B from Tagane et al.V 4165 C-F from Tagane et al.V4166.

Table 1 .
Altitudinal distribution of Quercus spp.found in Mt.Hon Ba.

Table 2 .
Summary statistics of datasets used for phylogenetic inference comprising rbcL, matK and ITS sequences.
cited specimen Chevalier 38650 from Vietnam as the type specimen of Q. camusiae.Two specimens of Chevalier 38650 were found in P comprising [P00379252] and [P00379253], between which only one specimen [P00379252] represents the diagnostic traits of nuts and cupules.Thus, specimen [P00379252] was selected as the lectotype of Q. camusiae.VIETNAM.Lao Cai Province: Lo Qui Ho Pass, Chapa.Note.In the original publication of Quercus dilacerata, Hickel and Camus (1929) cited the specimen collected by E. Poilane from Tonkin, Km. 8 du col de Lo qui Ho près de Chapa, Vietnam without any collection number.Two specimens of Quercus collected by Poilane from Tonkin, Km. 8 du col de Lo qui Ho près de Chapa, Vietnam were found in P with collector's number 12645 (P [P00753996], [P00753997]).Both specimens are fertile and consistent with the description of