Arabis watsonii (P.H.Davis) F.K.Mey.: An overlooked cruciferous species from eastern Anatolia and its phylogenetic position

Abstract Arabis watsonii (P.H.Davis) F.K.Mey. was initially reported as Thlaspi watsonii P.H.Davis in Flora of Turkey. Although F.K.Meyer transferred this species to Arabis L., this species has been overlooked and treated as Thlaspi L. in relevant literature for Flora of Turkey. In this study this species was evaluated using molecular (nuclear ITS and plastidic trnL-F sequences) and morphological data. Results clearly show that Arabis watsonii is sister to the Arabis hirsuta aggregate and its relatives. In conclusion, our results increased the number of known Arabis species in Turkey to 23. Furthermore, detailed description and distribution of the species are given and a new IUCN threat category for Arabis watsonii is proposed.

Arabis watsonii (P.H.Davis) F.K.Mey. (Fig. 1) (Meyer 2006), and he transferred this species to Arabis since due to the presence of siliques (rather than silicles which are typical for Th laspi) and leaves with bifi d hairs. Ever since this taxon either has been overlooked (Al-Shehbaz et al. 2007, Özhatay et al. 2009, Mutlu and Erik 2015 or still treated as Th laspi L. (Güner et al. 2014).
To verify Meyers's treatment of Arabis watsonii, plant material was collected from type locality and nearby areas. Collected specimens were evaluated morphologically and molecularly to analyse 1) is A. watsonii a member of Arabis and 2), which main clade does it belong? We carried out morphological and molecular phylogenetic studies of A. watsonii plus representatives of Arabis and other Arabideae and combined this data with climatic and biogeographic data.

Sampling
Th e present study includes for the fi rst time sequences of nuclear ribosomal ITS1, ITS2 and 5.8 S rRNA (hereafter ITS) and trnL(UAA) intron/trnL-trnF intergenic spacer sequence data (hereafter trnL-F) for Arabis watsonii (Voucher: M. Fırat 32513 at HUB). All other sequences of the 88 Arabis and other Arabideae species were taken from ITS and trnL-F data sets of Karl and Koch (2013). In addition, sequences of the related A. hirsuta aggregate (A. stelleri DC. and A. takesimana Nakai) were taken from GenBank (http://www.ncbi.nlm.nih.gov/genbank) and incorporated into the above mentioned data sets. To determine the phylogenetic placement of A. watsonii, we reduced the above mentioned data as follows: All Arabis species were added along with two species from Aubrietia Adans., two Draba L. species and one species each from other small genera including Sinoarabis, Arcyosperma O.E. Schulz, Baimashania Al-Shehbaz, Acirostrum, Botschantzevia Nabiev, Dendroarabis (C.A. Mey.) D. German & Al-Shehbaz, Pachyneurum Bunge, Pseudodraba Al-Shehbaz, D. German & M. Koch, Scapiarabis, Tomostima Raf. Pseudoturritis turrita (L.) Al-Shehbaz was used as the outgroup. Genbank accessions of species included in this study are at terminal nodes of phylogenetic trees (Figs 3, 4). 50 specimens belonging to fi ve populations were used for extending description of A. watsonii. Th e vouchers were deposited at Hacettepe University Herbarium (HUB) and private herbarium of M. Fırat.

DNA extraction, amplification and sequencing
Total genomic DNA was isolated using DNeasy Plant Mini Kit (Qiagen, Hilden, Germany) following the manufacturer's instructions. ITS and trnL-F regions were amplifi ed using primers ITS1 and ITS 4 (White et al. 1990) and C and F of Taberlet et al. (1991), respectively. Amplifi cation of ITS and trnL-F followed the protocol in Warwick et al. (2004) and Ansell et al. (2007), respectively. Purifi cation and sequencing were performed by BIOEKSEN (İstanbul, Turkey).

Data analysis
Firstly, to determine whether A. watsonii belongs to tribe Arabideae, the phylogenetic tool in Brassibase Kiefer et al. 2013) was used. ITS and trnL-F sequences data were edited with Codon Code Aligner (CodonCode Corporation) and directly incorporate into the alignment fi les of Karl and Koch (2013). Both data sets were analysed using a Bayesian approach as implemented in the software BEAST ver. 1.8 (Drummond et al. 2012).
Sequence evolution models were selected by the Akaike information criterion (AIC) implemented in MEGA v.6 (Tamura et al. 2013).
Th e GTR + G + I model was selected for ITS and GTR + I for trnL-F and a Yule process of speciation was used as the tree prior. Two independent Markov Chain Monte Carlo (MCMC) runs for each data set were conducted with 10 million generations and sampled every 1000 generations. Each run was checked using TRACER v1.6 (http://beast.bio.ed.ac.uk/Tracer) and then log and trees fi les were combined in LOGCOMBINER (Drummond et al. 2012).
A total of 20000 trees were obtained and 10% (2000) of these were discarded as burn-in. 18000 post-burn-in trees were used in the program TREEANNOTATER v.1.7.5 to obtain a single posterior probability and maximum clade credibility tree as visualized using FIGTREE v1.3.1.
Field observations and records taken from relevant herbaria indicate that A. watsonii has two distinct populations (Fig. 2). A large population growing around Bahçesaray district and a second population occurring in the Gürpınar district especially around the Güzeldere pass. Th erefore total distribution areas for these populations were estimated separately. Th e area around Bahçesaray was calculated as 108.99 km 2 and the second area around Gürpınaras 2.69 km 2 . In summary, to propose IUCN threat categories of A. watsonii, these two population groups and calculated areas were considered. Th e occupancy area (AOO) of A. watsonii was calculated as 111.68 km 2 in which about 1000 individuals in each population were estimated to occur. Overgrazing and reaping activities by the local people were observed in fi eld studies. Th erefore, in accordance with the criteria of the IUCN (2016), A. watsonii is assessed here as "Vulnerable" [(VU) (B2a, C2a(i))], because distribution area of the species is severely fragmented and the species is currently known from no more than ten localities occupying less than 2,000 km 2 ,( although it was considered "Endangered" (EN) according to Ekim et al. (2000)).
Th e basic climatic requirements of A. watsonii, annual main temperature and annual precipitation were calculated as 5.7 °C and 583 ml respectively.
Phylogeny. Th e aligned ITS and trnL-F data matrices included 91 species. Th e ITS data set was 642 bp, of which 236 were variable and 168 parsimony informative, whereas the trnL-F data set incorporates 855 bp, of which 181 were variable and 108 parsimony informative. Th e query of ITS sequences of A. watsonii in BrassiBase (version 1.1) supported its phylogenetic placement within tribe Arabideae and clearly matching Arabis. Th e outcome of Bayesian phylogenetic analyses using ITS and trnL-F data sets (Figs 3, 4) were congruent with each other in regard to the placement of A. watsonii. In both analyses A. watsonii falls into the main Arabis clade as sister to A. hirsuta (L.) Scop. aggregate and its relatives outlined in Karl and Koch (2014). Whereas A. watsonii forms a monophyletic lineage with A. cretica (Bayesian posterior probability (pp) = 0.98) in ITS analysis (Fig. 3), this sister relationship was not supported by trnL-F analysis (Fig.  4) and chloroplast data shows that A. watsonii, the A. hirsuta aggregate, its relatives plus non -European Arabis species are linked to this aggregate (including A. georgiana R.M.Harper, A. pycnocarpa M.Hopkins and A. borealis DC.) forming a monophyletic clade (pp=1.00).

Discussion
In this study, we used evidence from nuclear ITS and plastidic trnL-F sequences to determine the phylogenetic and taxonomic position of the overlooked Anatolian endemic A. watsonii. In addition, morphological and climatic data were used to better understand ecological and evolutionary relationships of A. watsonii with representatives of the well-defi ned A. hirsuta aggregate and its relatives.
Th e diff erences in the phylogenetic placement of A. watsonii in relation to its sister position to A. cretica, according to ITS and trnL-F analyses, indicates possible chromosome capture /ancient hybridization. Th ese processes are well known in the A. hirsuta aggregate and its relatives (Koch et al. 2010, Karl and. A recent study of A. hirsuta aggregate recognised eight European species including A. hirsuta (L.) Scop., A. sagittata (Bertol.) DC., A. planisiliqua (Pers.) Rchb., A. nemorensis (Wolf ex Hoff m.) W.D.J. Koch, A. allionii DC., A. sudetica Tausch, A. sadina (Samp.) Coutinho, and A. juressi Rothm. Th e historical defi nition and circumscription of such an aggregate depends on diff erent authors (see Karl and Koch (2014)) and because relationships between the A. hirsuta aggregate and its European/non-European relatives have already been discussed in detail before, we will not repeat this discussion here.
As indicated above, the ITS phylogeny supports a clear monophyly between A. watsoni and Greece endemic A. cretica Boiss. & Heldr. Whereas this relationship does not supported by trnL-F, both species seems to be at a basal position for A. hirsuta aggregate and its relatives. Th is results is somewhat expected because the Western Irano-Turanian and the East Mediterranean regions have already been suggested as diversity centres for almost all Arabideae clades (Jordon-Th aden et al. 2010, Karl and Koch 2013).
Morphologically A. watsonii is smilar to A. hirsuta and A. sagittata, therefore specimens were treated under these names in some herbaria. However A. watsonii diff ersfrom A. hirsuta and A. sagittata in having glabrous stems and relatively large petals. Meyer (2006) argued that A. watsonii is related to A. abietina Bornm. from the Ilgaz Mountain (Turkey) since this species is also characterized by glabrous stems and a similar petal length. Th e latter species has been treated as a synonym of A. suedica in Karl and Koch (2014) based on Jalas and Souminen (1994). Despite the relatively large geographic gap between A. abietina and A. suedica, ITS sequences of the taxa were identical and trnL-F sequences diff er only in one single nucleotide position. Th us, A. abietina was not included in the current study, although it is treated as a valid species according to the Flora of Turkey (Cullen 1965) and the actual check-list (Güner et al. 2012). Apart from geographic isolation, branched trichomes on leaf margins and rela-tively small fruits are the main diagnostic characters distinguishing A. watsonii from both A. abietina and A. allionii, which are also members of the A. hirsuta aggregate. In summary, more comprehensive studies are needed to clarify the validity of A. abietina. With the proper assignment of A. watsoni in the current study, the total number of the Turkish Arabis species increased from 22 to 23 (25 taxa) Finally, distribution in the alpine zone and perennial life cycle of A. watsonii is concordant with general trends of the tribe Arabideae (Karl and Koch 2013). Estimated climatic conditions for A. watsonii, including annual mean temperature, annual precipitation and other bioclimatic variable (not provided here) refl ect a continental climate also described for other members of the A. hirsuta aggregate including A. sudeica, A. hirsuta etc. In conclusion, all environmental parameters and life cycle strategies of A. watsonii are in agreement with the genetic affi liation to the A. hirsuta aggregate and its relatives within the main Arabis clade.