Asplenium danxiaense sp. nov. (Aspleniaceae, Aspleniineae), a new tetraploid fern species from Guangdong, China, based on morphological and molecular data

The Asplenium coenobiale complex is distributed in Eastern Asia and Southeast Asia with its distribution center in southwestern China. In this study, we carried out a detailed morphological, cytological, and phylogenetic study by adding two samples from Danxia landform in Guangdong. The sequences of fi ve chloroplast markers and one nuclear marker of the A. coenobiale complex were analyzed with maximum likelihood and Bayesian inference, respectively. The morphological and phylogenetic analyses support the recognition of a new species (A. danxiaense K.W.Xu sp. nov.) of the A. coenobiale complex from a cave of Danxia mountain, Guangdong province, southern China. This new species can be distinguished from A. coenobiale and A. pulcherrimum by having scales narrowly triangular to lanceolate, apex ending in a short apical tail, basal basiscopic pinnule usually largest, fertile segment scarce, and exospore length usually more than 50 μm and shows signifi cant molecular differences from other species in this complex. A detailed description and illustrations are presented.


Introduction
Asplenium L. (Aspleniaceae) is an approximately globally distributed genus comprising over 700 extant species (Kramer & Viane 1990). Reticulate evolution caused by hybridization and polyploidy has generated various species complexes that pose diffi culty with species delimitation in the genus, e.g., the A. monanthes complex (Dyer et al. 2012), the A. normale complex (Chang et al. 2018), the A. pekinense complex, and the A. varians complex (Liang et al. 2021). The A. coenobiale complex belongs to A. ser. Variantia Ching & S.H.Wu (Wu 1999). Members in this complex are characterized by the presence of small fi nely dissected leaves, and shiny castaneous to purplish black and threadlike stipes, and petioles.
Asplenium coenobiale complex is mainly distributed in the tropical and subtropical regions of southern and southwestern China, Malaysia, southern Japan, and northern Vietnam (Lin & Viane 2013). Members within this complex are morphologically variable and species delimitation is diffi cult. Wu (1999) recognized four species in total within this complex, including A. coenobiale Hance, A. fuscipes Bak., A. subtoramanum Ching ex S.H. Wu, and A. toramanum Makino. Lin & Viane (2013) treated the latter three species as synonyms of A. coenobiale and believed that this species has variable frond division probably depending on growing conditions with large and more divided plants often growing on rocks in forests. However, Lin & Viane (2013) retained Asplenium pulcherrimum (Baker) Ching ex Tardieu, which had been confused with A. coenobiale , and the similar diploid species A. cornutissimum X.C. Zhang & R.H.Jiang. Most recently, Liang et al. (2019) made an integrative taxonomic study using evidence of morphology, cytology, and molecular phylogeny to resolve the species delimitation of the complex. On the basis of treatment by Lin & Viane (2013), Liang et al. (2019) described one autotetraploid novel species ( A. maguanense S.Q.Liang, R.Wei & X.C.Zhang) in addition to the species recognized by Lin & Viane (2013).
However, the previous studies on the integrative taxonomy of the A. coenobiale complex mainly focused on samples from China's Guizhou, Guangxi, and Yunnan provinces, as well as a small number of representatives from China's Taiwan Island, Japan, and Vietnam (Liang et al. 2019;Xu et al. 2020). Their taxonomic analyses of this complex did not involve samples from South China's Guangdong province. By examinations of specimens from Danxia mountain of Guangdong province and the comparative analyses of the whole complex, we hypothesized that these samples from Danxia mountain, Guangdong, represented a cryptic new species. In the present study, we reconducted a taxonomic analysis by adding samples from Danxia mountain, Guangdong, which is one of the famous examples of Danxia landform characterized by red-colored sandstones and conglomerates of largely Cretaceous age (Peng 2009). We aimed to detect the presence of the new cryptic species of A. coenobiale complex in this special habitat of Danxia landform.

Material and taxon sampling
To detect the presence of cryptic species in Asplenium coenobiale complex in Danxia landform, we newly generated DNA sequence data including three plastid genes and one nuclear gene of the two samples of the complex from a cave of Danxia mountain, Guangdong province. We also included accessions of the A. coenobiale complex from the previous studies of Liang et al. (2019) and Xu et al. (2020). A total of fi ve plastid genes of 53 accessions and one nuclear gene of 33 accessions as the ingroup were incorporated into the molecular analyses (Supplementary fi le 1). The selection of outgroup taxa refl ected the phylogenetic relationships between the A. coenobiale complex and its closely related species (Xu et al. 2020).

Spore observation and cytotype determination
Scanning Electron Microscope (SEM) images were taken of the spores of members in the A. coenobiale complex. Spores obtained from herbarium specimens, were mounted on specimen tabs and then coated with platinum in a sputter coater. Observations were conducted using an ESEM-Quanta 200 (FEI, Hillsboro, Oregon, USA) with 15 Kv at Nanjing Forestry University, Nanjing, China. The ImageJ software (Pérez & Pascau 2013) was used for measurement on SEM micrographs. For plants of the A. coenobiale complex from Danxia mountain, DNA ploidy levels were determined by measuring DNA-C values using a BD FACScalibur, and Zea mays L. was employed as the internal standard. The obtained data were then compared to the previously recorded genome size of members in this complex (Liang et al. 2019).

DNA extraction, PCR amplifi cation, and sequencing
Total genomic DNA was extracted from silica-gel-dried leaves using the modifi ed 2 ×CTAB procedure of Doyle & Doyle (1987). Three plastid markers ( rbcL , rps4 , and the rps4 -trnS intergenic spacer) and one nuclear gene ( pgiC ) were selected for amplifi cation and sequencing. The rbcL gene was amplifi ed with primers ESRBCL1F and ESRBCL1361R (Schuettpelz & Pryer 2007). The rps4 gene and the rps4 -trnS intergenic spacer were amplifi ed using the forward primer 5'-ATG TCMCGTTAYCGAGGRCCTCGT-3' and the reverse primer 5'-TACCGAGGGTTCGAATC-3' (Schneider et al. 2005). The nuclear gene pgiC was amplifi ed using primers 14F and 16R (Ishikawa et al. 2002). The PCR mixtures and cycling parameters of plastid genes followed those of Ohlsen et al. (2015). Amplifi ed fragments were purifi ed with TIANquick mini purifi cation kits (TIANGEN, Beijing, China). Purifi ed PCR products were sequenced by Sangon Biotech (Shanghai, China). For the nuclear gene, the purifi ed nuclear DNA products were ligated into a pUCm-T Vector. Ten positive clones for each individual were randomly selected for sequencing.

Sequence alignment and phylogenetic analysis
The newly generated s equences were assembled and edited using Sequencher ver. 4.14 (GeneCodes Corporation, Ann Arbor, Michigan). Subsequently, all sequences were initially aligned with MAFFT ver. 7 (Katoh & Standley 2013) and manually adjusted in BioEdit (Hall 1999). Alignments of six cpDNA regions were concatenated using PhyloSuite (Zhang et al. 2020). Independent phylogenetic analyses of the plastid and nuclear datasets were conducted using maximum likelihood (ML) and Bayesian inference (BI) on the Cipres web server (Miller et al. 2010), respectively. The ML tree searches were performed using RAxML-HPC2 on XSEDE with 1000 bootstrap replicates. The model TPM1uf+G for combined plastid DNA dataset and HKY for nuclear dataset in maximum likelihood analyses was used based on results of jModelTest2 (Darriba et al. 2012) on XSEDE (2.1.6) using the Bayesian information criterion (BIC). The BI was conducted using MrBayes ver. 3.2.7a (Ronquist & Huelsenbeck 2003) with temperature parameter set to 0.2, and other priors set to their default values. Two independent runs, each with four chains (one cold, three heated), were conducted, each beginning with a random tree and sampling one tree every 1000 generations of 10 000 000 generations. Convergence among runs and stationarity were assessed using Tracer ver. 1.4 (Rambaut & Drummond 2007), and a burnin was discarded. The remaining trees were used to calculate a 50% majority-rule consensus topology and posterior probabilities (PP).

Molecular analysis
The 64 aligned cpDNA sequences are 5105 bp in length, with 525 parsimony informative sites in total and the 78 aligned nuclear sequences are 717 bp in length, with 20 parsimony informative sites. The present result from the combined plastid dataset shows that the two samples of Asplenium from Danxia mountain form an independent clade that is clearly separated from all the other species in the phylogenetic tree (Fig. 1), while the nuclear dataset resolved the three pgiC alleles of collection XKW674 into two well-supported clades (Fig. 2).  rigid and threadlike stipe and rachis, fi nely dissected frond, and lophate (cristate-alate) perispore, but the former has rhizome scales narrowly triangular to lanceolate, with a short apical tail, basal-most basiscopic secondary pinnae usually largest, fertile segment scarce, and exospore length usually more than 50 μm. In contrast, A. coenobiale and A. pulcherrimum have scales narrowly triangular to linearsubulate, with a long apical tail, basal acroscopic pinnule usually largest, fertile segment abundant, and exospore length usually less than 50 μm.

Etymology
Based on the mountain name, Danxia, in northern Guangdong, China, and the Latin suffi x, -ense, of origin, referring to the type locality of the species.

Distribution and habitat
Asplenium danxiaense sp. nov. is only known from Guangdong, China, growing in a cave of Danxia mountain in evergreen broad-leaved forest, at the elevation of 100-300 m.

Vernacular name
We propose a Chinese name, Dānxiáshān tiějiǎojué ( ) to refl ect the type locality of the new species.

Conservation Status
We provisionally assess Asplenium danxiaense sp. nov. as Critically Endangered based on criterion D of IUCN (2012). Only one population and fewer than 50 individuals of A. danxiaense sp. nov. are known from Danxia mountain. It was not seen by one of us during separate expeditions to nearby sites on Danxia mountain.

Discussion
Members of the Asplenium coenobiale complex are mainly distributed in karst mountains of south and southwestern China's Guangxi, Guizhou, and Yunnan provinces, also extending to adjacent regions including Guangdong, Hunan, and Taiwan Island of China, Japan, and Vietnam (Lin & Viane 2013). Four species within the complex have been recognized by recent study (Liang et al. 2019). During our fi eld investigation in Danxia mountain of Guangdong province, we found some plants of the A. coenobiale complex in a cave. These plants are morphologically close to A. coenobiale and A. pulcherrimum , but they could be obviously distinguished based on the morphological and micromorphological characters. The rhizome scales of the plants from Danxia mountain are narrowly triangular to lanceolate, with a short apical tail, while those of A. coenobiale and A. pulcherrimum are narrowly triangular to linear-subulate, with a long apical tail (Fig. 3). In the plants from Danxia mountain, the basal-most basiscopic secondary pinnae are larger than the basal-most acroscopic secondary pinnae (Fig. 4), while in A. coenobiale and A. pulcherrimum , the basal-most acroscopic secondary pinnae are usually larger than the basalmost basiscopic secondary pinnae. In addition, plants from Danxia mountain usually have less fertile segments per frond than those of A. coenobiale and A. pulcherrimum . The spore ornamentation of plants from Danxia mountain is lophate (cristate-alate) perispore, which is similar to that of A. coenobiale and A. pulcherrimum (Fig. 5), while the exospore length of plants from Danxia mountain (usually more than Fig. 6. Estimation of Asplenium danxiaense K.W.Xu sp. nov. genome size by fl ow cytometry. The internal control Zea mays L. cv. B73 has 1C = 2.3Gbp. 50 μm) is longer than the exospore length of species A. coenobiale and A. pulcherrimum (usually less than 50 μm). Thus, a new species, A. danxiaense K.W.Xu, sp. nov., is proposed here. This new species also resembles species A. cornutissimum and A. maguanense . However, it differs from A. cornutissimum by its plants more strong, tetraploid, laminae fi rmly herbaceous, widest at base, and exospore length usually more than 50 μm, and differs from A. maguanense by its laminae 3-pinnate, ultimate segment apex with long teeth, and exospore length usually more than 50 μm (Jiang et al. 2011;Liang et al. 2019).
The genome size of the new species A. danxiaense sp. nov. is estimated as 1C = 7.06 Gb, namely 14.8 pg/2C (1 pg≈978 Mb; Dolezel et al. 2003), which is close to the value previously estimated for tetraploid taxa ( A. coenobiale , A. maguanense and A. pulcherrimum ) in the A. coenobiale complex ( Fig. 6; Lin & Viane 2013;Liang et al. 2019). Moreover, the spore size of the new species is somewhat larger than that of other tetraploid taxa in the A. coenobiale complex. Based on the genome size and spore size, we speculated that the new species from Danxia mountain is also a tetraploid species.
Phylogenetically, the new species was well resolved as a distinct clade, which is sister to a clade containing accessions of A. coenobiale and A. pulcherrimum based on phylogeny reconstructed using maternally inherited plastid genes (Fig. 1). In contrast, a close relationship between A. danxiaense , A. coenobiale and A. pulcherrimum was also supported by their sharing of similar gene copies in the nuclear phylogeny (Fig.  2). The three pgiC alleles present in A. danxiaense sp. nov. , are nested within two well-supported clades (Fig. 2). Plastid genes confi dently resolved relationships of the A. coenobiale complex into fi ve major clades corresponding to fi ve species recognized in this complex, while the nuclear pgiC gene resolved members of this complex into four major clades. Confl icts between the plastid and low-copy nuclear phylogenies at species level were clearly observed in this complex. This discordance might be caused by reticulation events involving hybridization and allopolyploidization, as has been frequently documented in Asplenium (Dyer et al. 2012;Fujiwara et al. 2017;Chang et al. 2018;Liang et al. 2019Liang et al. , 2021.