Revision of series Gravesiana (Adiantum L.) based on morphological characteristics, spores and phylogenetic analyses

Since the adoption of some ambiguous and quantitative characters in Flora Republicae Popularis Sinicae 3(1), species identifications of the series Gravesiana have been in disarray, requiring clarification. Two hundred and fifty-nine individuals from 47 different populations were collected for the estimation of morphological characters and phylogenetic analyses. Spores of 26 populations were observed through scanning electron microscope. Our results were different from those of previous research: (1) six identifiable species, rather than five species observed previously, were confirmed in the series Gravesiana, they are A. gravesii, A. juxtapositum, A. mariesii, A. dentatum, A. longzhouensis and A. obovatum, of which the latter three are newly recognized species. (2) Thirteen characters were measured and estimated through the program Mesquite v. 2.71. The character whether the pinna stalks were 1/3-1/2 times longer than the pinna was used to distinguish A. gravesii and A. lianxianense previously and was found to be unreliable here, whereas such characters as the height of the plant (H), pinna aligned forms (FP), number of pinna (NP), pinna margin (M), number of veins flabellate at base (NV), sori number and shape per pinna (NSS), pinna texture (T), and powder-covered or not on the abaxial surface of the pinna (P) are estimated to be stable and reliable characters useful for identification. Descriptions of new species and their retrieve keys are also listed. (3) Surface ornamentations and spore sizes are helpful for us to distinguish species in series Gravesiana.


Introduction
Pteridaceae is a large family containing approximately 50 genera and 950 species, with most of them existing in tropical and arid regions [1]. There are 20 genera and 233 species (89 a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 juxtapositum has three to many sori per pinna [1,7,8]. Unfortunately, few specimens of A. chienii have been collected and kept except for its typus. The morphological characters of these five species are shown in Table 1. A. chienii, A. gravesii and A. juxtapositum share a common character-the length of the pinna stalks is less than 1/5 times that of the pinna. However, when we examined each of their specimens (including typus Figs 1 and 2), pinna stalks from different parts of a lamina were of different length; some reached up to 1/5 of the pinna while others did not. The biggest distinction of these three species is that A. gravesii is alternipinnate and each pinna has only 1 or 2 nephroid or transversally linear sori with a sinus in the middle, whereas A. chienii and A. juxtapositum possess opposite pinna and each pinna has 1-to-many sori with apex truncate. However, some new individuals (doubtful taxa, short for DT) with alternating pinna and 3-tomany kidney-shaped sori per pinna were found during our field investigation. We found some individuals from different populations with similar new characters never encountered before (DT). In addition, many individuals have highly diverse characters in one population. Hence, Table 1. Main morphological characters of the five species in series Gravesiana according to Shing and Wu (1990).

A. juxtapositum
do these variable individuals belong to one species or to different species? To elucidate the identification of these five species, we investigated a total of 47 populations from Guangxi, Guizhou, Guangdong, Hunan, Jiangxi and Fujian provinces of China (Fig 3) and collected 259 individuals. Such characters as the height of the plant, leaf stalk, pinna stalk, pinna size and shape, veins, sori number per pinna, and scales were measured before DNA extraction. Combined analyses of microscopic spore observations, morphological and molecular data will be carried out to resolve the phylogenetic relationship of the series Gravesiana here.  [11], and another 33 Adiantum species were downloaded from GenBank. Vittaria flexuosa Fée was selected as outgroup. Details on the collection of the 78 individuals of the series Gravesiana including collection locality and date, longitude, latitude and altitude are shown in S1 Table, and the information for all other  taxa are listed in S2 Table. DNA sequencing and phylogenetic analyses Total DNA was extracted from silica-gel-dried leaf materials using a modified CTAB DNA extraction protocol [12]. Six pairs of primers "ESATPF412F and ESTRNR46F", "ESATB172F and ESATPE45R", "1F and 1379R", "f and pl", "trnS and rps4.5", "Adn matK fIHS Ã and FER matK rAGK" were used to amplify the chloroplast gene regions atpA, atpB, rbcL, trnL-F, rps4-trnS, and matK, respectively [13,14,15,16,17,18,19,20]. PCR reactions were performed in 30 μL reaction volumes, including 1.0-2.4 μL of each primer (5p), 17-60 ng sample DNA, 1.5 U of Taq DNA polymerase, 10 × buffer (including Mg 2+ ), 0.25 mmolÁL -1 dNTP, and ultrapure water. The PCR products were purified and sequenced with an ABI 3730XL by Majorbio Company.

Materials and methods
The sequences were assembled with Sequencher v. 4.14, aligned using the program Clustal X v. 2.0 [21] and then edited manually through Bioedit v.7.1.3 [22]. Phylogenetic trees of each marker and the combined markers (atpA, atpB, rbcL, trnL-F, rps4-trnS and matK) were constructed using maximum parsimony (MP) and Bayesian Markov chain Monte Carlo inference (BI). The maximum parsimony analyses were performed with PAUP Ã 4.0b10 [23], treating gaps as missing data and using the heuristic search options with 1000 random replicates and tree-bisection-reconnection (TBR) branch swapping. All characteristics were unordered and equally weighted. Through MrModeltest2 v. 2.3 [24], GTR+I+G was selected as the best fit molecular evolution model for the MP and Bayesian analyses. For Bayesian inference, trees were generated for 1,000,000 generations with sampling every 100 generations. Four chains were used with a random initial tree. For each of the individual data partitions and the combined dataset, the first 2500 sample trees were discarded as burn-in to ensure that the chains reached stationarity. Nodes receiving bootstrap support (BS) of < 70% in the MP analyses or PP of < 0.95 in the BI analyses were not considered to be well supported.  [26] and Plant identification terminology: An illustrated glossary [27] was followed.

Nomenclature
The electronic version of this article in Portable Document Format (PDF) in a work with an ISSN or ISBN will represent a published work according to the International Code of Nomenclature for algae, fungi, and plants, and hence the new names contained in the electronic publication of a PLOS ONE article are effectively published under that Code from the electronic edition alone, so there is no need to provide printed copies. IPNI LSIDs (Life Science Identifiers) for new species herein have been resolved and can be available at http://ipni.org, once the paper is published. The online version of this work is archived and available from the following digital repositories: PubMed Central, LOCKSS.

Phylogenetic analyses
The topologies derived from the analyses of single dataset and the combined dataset were congruent; thus, we adopted the topology from the combined dataset here. The combined 5-marker (atpA, atpB, rbcL, trnL-F and rps4-trnS) phylogenetic tree of Chinese Adiantum (in 132 accessions) comprised 6,497 nucleotides, of which 2,364 were variable (36.4%) and 1,772 were phylogenetically informative (27.3%). The MP analysis based on this dataset yielded one maximally parsimonious tree of 5,052 steps with a consistency index (CI) of 0.6033 and a retention index (RI) of 0.9198. The tree obtained from the BI analysis had a similar topology to the MP strict consensus tree (Fig 4). The monophyly of the series Gravesiana was also strongly supported here (100/1.0); however, all of its individuals were clustered into nine clades with a high support value (Fig 4).  The trees of the series Gravesiana constructed with the combined 6-markers (atpA, atpB, rbcL, trnL-F, trnS and matK) included 78 taxa and were 7,140 nucleotides in length, of which 1,137 were variable (15.9%) and 490 were phylogenetically informative (6.9%). 1,384 steps were run to generate a maximally parsimonious tree; the consistency index (CI) was 0.8779, and the retention index (RI) was 0.9609. Nine main clades were identified in Fig 5, which was similar to the topology of the combined 5-marker tree but with high supported value. Twentyone individuals of A. gravesii from twelve different populations were clustered into clade A, and 16 individuals of A. mariesii from 7 different populations were clustered into clade B. Seven individuals of the doubtful taxon "DT1" from four populations formed a clade (clade The bootstrap values were shown above the lines, and the Bayesian posterior probabilities were shown below the lines. Nine clades were labeled in columns and their own morphological characters were stated at right. For G1-3, GDY1-1, JW2-8, GJK2-2, . . .., front alphabets are the short names of different populations of taxa and the latter numbers represent single individuals as shown in S1 Table. https://doi.org/10.1371/journal.pone.0172729.g005 Revision of series Gravesiana (Adiantum L.) C). Three samples of "DT2" from two populations formed a new clade (clade D), and six individuals of "DT3" from three populations were clustered into clade E. All individuals of A. juxtapositum from four populations were clustered together (clade I).

Morphological characters analyses
In the 13 characters, scales among all individuals were similar; thus, we abandoned this character. The mapping to the phylogeny tree of the twelve other characters is visualized in Fig 6, and all characters of each clade are concluded in Table 2 to test whether the morphological characters of the species were consistent with the gene trees. It was clear that characters such as the height of the plant (H), pinna aligned forms, number of pinna (NP), pinna shape (SP), pinna margin (M), number of veins flabellate at base (NV), sori number and shape per pinna (NSS), pinna texture (T), and powder-covered or not on the abaxial surface of the pinna (P) are stable and reliable and can be distinguished through Table 2 and Fig 6. The size of the pinna was much relevant to its shape and the height of plants. Twenty-one individuals from thirteen different populations were clustered into clade A, and they shared some common traits: height usually longer than 5 cm, pinna alternate or opposite rarely, vein numbers flabellate at base greater than 4(5), sori 1, reniform or transversally linear, and with a notch at false indusia termination (Figs 5 and 6).
Clade B shares common characters: all individuals were no more than 5 cm, with pinna alternate and sub-round and NV were less than 4(5), sori 1 for each pinna, and orbicular. All individuals in clade C can be easily identified by their much shorter height, NV less than 4 and dentate pinna margin. Individuals in clade D were featured by their membranous pinna texture and obovate pinna shape. For clade E, as shown in Figs 5 and 6, all samples included were large with pinna obdeltoid or flabellate, sori 1 to many (often 2-4), reniform, and NV 6-9. All samples of clade I were clustered together for their shared opposite pinna and 1-to-many sori per pinna and transversally linear indusia, truncate at false indusia termination. We did not analyze clade F considering its insufficient samples.

SEM observation results
The spore shapes of all taxa in series Gravesiana are similar in polar and equatorial views, but their surface ornamentations and spore sizes are clearly different. All spores are actinomorphic and trilete with polar surface triangle, and the equatorial surface is semicircular or super-semicircular. There are three kinds of surface ornamentations in all: psilate, rugate and verrucate (Fig 7). All spores of clade A are verrucate, while that of clade I are rugate, and that of the population "HMM1" in clade H are psilate. Interestingly, spore ornamentations of clade B are unapparent verrucate. Spores of clade C are rare and abortive. Although clade E share same spore ornamentations with clade A, sizes of its spores are bigger than that of clade A. Mean spore sizes of different populations in clade E and clade I are the biggest, whose length of equatorial axes varies from 52.0±5.0 μm to 61.87±2.7 μm and length of polar axes varies from 49.5 ±2.9 μm to 57.7 ±2.3 μm. Mean spore sizes of clade B are the second biggest, which are bigger than mean sizes of clade A. All spore sizes were presented in Table 3.
Based on above molecular, morphological and spore analyses, clade A was considered to be A. gravesii, clade B was A. mariesii and clade H was A. juxtapositum without any doubt compared with their own type specimens (Figs 1 and 2). Clade C, D and E were treated as three new species, and we named them Adiantum dentatum A. H. Wang . Fig 8  A. dentatum differs from others for its fewer pinnae, obdeltoid or subovate pinna with prominent dentate margin from the broadest middle to apex and its variable orbicular or transversally linear sori.

IUCN Conservation assessment.
Although it grows well in the type locality, because of its restricted distribution region, small population size and low number of individuals in very small or restricted populations, A. dentatum should be considered endangered in accordance with the IUCN Red List criteria [28].  The obdeltoid pinna and 1-many kidney-shaped sori per pinna of this species distinguish it from A. juxtapositum and A. gravesii.

Adiantum obovatum
Plants epilithic, 26.5 cm tall. Rhizomes short, erect, covered with lanceolate and blackbrown scales, scales margin entire. Stipes tufted, 5-10 together, hard and stong, 5.6-11 cm long, ca. 1.0 mm thick, dark castaneous. Lamina 1-pinnate, odd-pinnate or paripinnate, 9-16 cm  IUCN Conservation assessment. Its restricted distribution region, small population size and low number of individuals in very small or restricted populations make A. longzhouensis the endangered scale in accordance with the IUCN Red List criteria [28].

Discussion
A. obovatum can be easily misidentified as A. lianxianense based on the description in Lin (1980) if we do not observe the type specimen of A. lianxianense carefully (Fig 1). A. lianxianense is coriaceous, and its sori are kidney-shaped rather than orbicular (Fig 1). Its similar individual "JW1-1" (Fig 11) clustered in clade H suggests that A. lianxianense is perhaps a variant of A. gravesii, which cannot be verified until we obtain DNA evidence of the type specimen. Samples of A. chienii have not been collected in the type locality, but similar ones were found in the Renhua Danxia mountain such as "RZ4-3" (Fig 11), which was clustered in A. juxtapositum (clade I). So it is possible that A. chienii and A. juxtapositum is of the same species. For clade G in Fig 5, the most remarkable difference from A. longzhouensis is that individuals in clade G are covered with white powder on the abaxial surface of the pinna while A. longzhouensis is not. However, we did not treat it as a new species considering its mixed characters: soris reniform or transversally linear with a truncate terminal at false indusium.
Height of the plant (H), pinna shape (SP) and number of veins flabellate at base (NV) of the four individuals "JE1-1" "JE1-2" "HB1-2" "HMM1-9" in clade H were same as A. mariesii, while their NSS (sori number and shape per pinna) were much more similar than A. gravesii (see Fig 12). Besides, morphological characters of plants in population SDB were similar to A. juxtapositum but their sori shapes were mixed characters: reniform or transversally linear with a truncate terminal at false indusium. Spores of "JE1-1" "JE1-2" "HB1-2" "HMM1-9" were scarce and seemed abortive, suggesting that gene exchange and hybridization may exist among A. gravesii, A. juxtapositum and A. mariesii considering the above morphological characters and spores results of clade G and H. It is interesting that three species exist in different microhabitats of the same cave in Yangzidong, Fengshan county ("HF1-1", "HF1-3" and "HF1-4" in Fig 5). "HF1-1" grows in the chalk soil near the entrance to the cave, where drips are falling down. "HF1-3" lives in the cave without adequate sunshine and water. "HF1-4" exists in the young limestone within the cave accompanied by many moist green mosses. Then, what has caused the differentiation of these three sympatric species? We did not find any spores of "HF1-2" and "HF1-3", spores of DT1 were also rare and abortive, so the doubtful species DT1 and DT2 are possibly hybrids. Or it is the result of species adapting to the soil, water, or sunshine or the interaction of these factors. At any rate, this interesting phenomenon has provided a good condition for us to study the mechanism of speciation in "terrestrial islands" later.