Tragopogon abbreviatus (Asteraceae): a little-known species inferred from morphological and molecular analysis

: Tragopogon porrifolius occurs in Turkey with 3 taxonomically confused subspecies: T. porrifolius subsp. eriospermus, T. porrifolius subsp. longirostris , and T. porrifolius subsp. abbreviatus . In the present paper, T. porrifolius subsp. abbreviatus endemic to Turkey is raised at specific level based on morphological and molecular analysis. After detailed assessment of original herbarium specimens and literature studies, we found that this name is not typified yet. Thus, a lectotype is provided together with an emended diagnosis; synonym, distribution map, and conservation status for the first time. Additionally, a preliminary phylogenetic position of this little-known taxon is given on several newly sequenced data.


Introduction
Tragopogon L. (Asteraceae) includes about 150 species native to Eurasia and known as salsify worldwide (Bremer, 1994;Soltis et al., 2004) and 22 species (25 taxa) in Turkey (Coşkunçelebi et al., 2017). Members of this genus are known as Yemlik or Tekesakalı in Turkish  and consumed by local people as a source of food (Baytop, 2000). According to Borisova (1964), Blanca and Diaz De La Guardia (1997), and Sukhorukov and Nilova (2015), the members of the Tragopogon are almost impossible to be identified without mature achenes, phyllaries, and ligules colours. However, most of the Turkish taxa including T. porrifolius L. and T. longirostris Sch. Bip. were identified on incomplete specimens in the Flora of Turkey (Matthews, 1975).
Tragopogon longirostris was treated under T. longirostris var. abbreviatus Boiss. and T. longirostris var. brachyphyllum Boiss. by Boissier (1875). Although Boissier (1875) did not give a detailed description in the protologue of T. longirostris var. abbreviatus, he reported that it differs from T. longirostris var. brachyphyllum of achene features. Boissier (1975) also listed T. dshimilensis K.Koch as a synonym under the T. longirostris var. abbreviatus in the protologue of T. longirostris var. abbreviatus. Later Coşkunçelebi et al. (2017) reported that T. dshimilensis is an accepted taxon endemic to Turkey due to differences in the ligule colour, presence of the leaf remains at the base of the stem and plant habit. Although Matthews (1975) followed Boissier's view in the Flora of Turkey during the preparation of the Turkish Tragopogon account, this account does not always coincide with current published accounts of the genus Tragopogon (Feinbrun-Dothan, 1978;Dimopoulos et al., 2016). Greuter (2003) transferred T. longirostris under T. porrifolius as subspecies (T. porrifolius subsp. longirostris (Sch. Bip.) Greuter), however he ignored the names T. longirostris var. abbreviatus in his paper about the Euro+Med treatment of Cichorieae (Compositae).
During a taxonomic revision of the Turkish Tragopogon, the authors collected several specimens easily keying out T. longirostris var. abbreviatus according to the Flora of Turkey (Matthews, 1975). Further comparisons based on several morphological features that are important for the genus Tragopogon (Blanca and Diaz De La Guardia, 1997;Sukhorukov and Nilova, 2015) revealed that these samples distinctly differ from T. porrifolius subsp. longirostris (=T. longirostris var. longirostris) and T. porrifolius subsp. eriospermus (Ten.) Greuter (=T. porrifolius var. brachyphyllus Boiss). Although T. longirostris and T. porrifolius (=T. porrifolius var. brachyphyllus Boiss.) were treated as separate species by Matthews (1975) in the Flora of Turkey, Coşkunçelebi and Gültepe (2012) transferred T. longirostris var. abbreviatus under the T. porrifolius as a subspecies (T. porrifolius subsp. abbreviatus (Boiss.) Coşkunç and M.Gultepe) contrary to Greuter (2003). Besides Gültepe et al. (2016) reported, contrary to Greuter (2003), that T. porrifolius and T. longirostris should be treated as separate species based on molecular data. Correspondingly, the aim of the present study is to identify the authentic material of the name T. longirostris var. abbreviatus and ascertain its taxonomy and provide a preliminary phylogenetic position of this little-known species.

Morphological studies
The current study was mainly performed on authors' own specimens preserved in Herbarium of Biology Department at Karadeniz Technical University (KTUB) as well as herbarium specimens from the following herbaria (ANK, B, E, EGE, G, GAZI, HUB, ISTE, ISTF, ISTO, KATO, KNYA, LE, TBI, VANF). Detailed locality information of the examined specimens is given in the Appendix. Herbarium acronyms are given according to Index Herbariorum (Thiers, 2019).
All specimens were identified according to the Flora of Turkey (Matthews, 1975), and nomenclature is given according to Coşkunçelebi and Gültepe (2012). Morphological characters were measured from specimens according to Boisser (1875), Matthews (1975), Gültepe (2014), and relevant literature (Borisova, 1964;Richardson, 1976;Rechinger, 1977). All measurements and observations were performed from at least 3 specimens. Outer mature achenes, phyllaries, fruiting, and flowering capitula obtained from herbarium specimens stored at KTUB were drawn and/or photographed under the stereo microscope.

Red list assessment studies
The threat category has been assessed according to the red list criteria of IUCN (2012) using area of occupancy (AOO), the extent of occurrence (EOO), population numbers, and field observations. A distribution map, calculation of the EOO (km²), and AOO (km²) were carried out by using Geo-Cat (Bachman et al., 2011).

Molecular studies
Genomic DNAs were extracted from silica-dried leaves following the modified extraction procedure of Doyle and Doyle (1987). Amplification and sequence of nrDNA ITS regions were carried out according to Gültepe et al. (2010). Phylogenetic analyses were performed on the ITS data set consisting of 87 accessions (Table 1), of which 15 were newly generated sequences, 72 sequences were obtained from GenBank and 10 were used as an outgroup provided from Gültepe et al. (2016). For each taxon, multiple samples were used whenever possible to observe the infra-specific DNA sequence variation and construct a robust phylogenetical analysis. All sequences (85 accessions) were aligned with Bioedit v. 7.0 (Hall, 1999). Phylogenetic relationships were reconstructed using Maximum Parsimony (MP) and Bayesian Inference (BI) analyses. MP analyses were run in PAUP* version 4.0b10 (Swofford, 2003) using heuristic search with the following parameters: all characters have equal weight, gaps are regarded as 'missing' , simple addition of sequences, TBR branching swapping, maxtrees setting to 100 and autoincreased by 100, one nonbinary starting tree arbitrarily dichotomized before branch swapping, only one tree was saved. A majority rule consensus tree was calculated from the most parsimonious trees. Jackknife (JK) support values for the nodes found in the MP analysis were calculated in PAUP* applying the optimal jackknife parameters according to Farris et al. (1996) and Müller (2005) with the following parameters: 10,000 jackknife replicates using the TBR branch swapping algorithm with 36.788% of characters deleted and one tree held during each replicate. Prior to BI analyses, the nucleotide substitution model that best fits the datasets was determined for ITS with MrModeltest 2.3 (Nylander, 2004), following the Akaike Information Criterion (AIC). BI analyses in MrBayes 3.2 (Rounquist et al., 2012) were performed with 4 simultaneous runs of Metropolis-coupled Markov Chains Monte Carlo (MCMCMC), each with 4 parallel Markov chains. Each chain was performed for 20 million generations and, starting with a random tree, one tree was saved every 1000th generation. For other parameters, the default settings of the program were left unchanged. A conservative burn-in of 0.2 (i.e. discarding the first 20% of the trees) was applied. The remaining trees were used to generate a majority-rule consensus tree, and visualized in TreeGrap v.2 (Stöver and Müller, 2010). Lectotype (designed here): Turkey/B2 Manisa: Alaşehir ("Tmoli subra Philadelpbiam") Boisser s.n (G00330030!).
Phenology: Flowering in April-July, fruiting in May-July Habitat: T. abbreviatus grows in cultivated fields, roadside, graveyards, hill side, and under Pinus brutia Ten. Forest between 170 and 1150 m above sea level.
Chorotype: Endemic to Turkey Distribution: North, west and south-west of Turkey (Figure 2). Vernacular name: Çayır yemliği (Güner at al., 2012) Lectotypification After critical examinations on specimens and literature study, we found that the name T. longirostris var.
abbreviatus has not been typified up to now according to ICN, Article 9.2 (McNeill et al., 2012). Although Boissier (1875) included 7 specimens in the protologue, further examination revealed that 4 of them belong to T. longirostris var. abbreviatus, 1 specimen belongs to T. dshimilensis and 1 specimen to T. longirostris var. brachyphyllum. In the present study, the sheets stored at G are being taken into account while typifying T. longirostris var. abbreviatus (=T. abbreviatus). Authentic specimens stored at G are now available online via JSTOR Global Plants (JSTOR, 2019). All characteristics of Boissier's specimens (G00330029!, G00330030!, G00330032!, G00330034!) coincide with the description in the protologue. Also, there are special labels on the sheets written by either collectors or Boissier (1875). Thus, all these samples can be considered as authentic materials according to ICN, Article 9.3 (McNeill et al., 2012). However, for the specimen (G00330030!) collected from B2 Manisa: Alaşehir ("Tmoli subra Philadelpbiam"), Boisser s.n is more complete and in a better conservation status. Therefore, it has been selected as the lectotype in the present study (Figure 3).
Key to Tragopogon abbreviatus and related species T. abbreviatus and related taxa may be inserted in the key adapted from the Flora of Turkey and the East Aegean Islands (Matthews, 1975).

Conservation status
T. abbreviatus was recorded from more than 14 localities in Turkey (Figure 2). Each population consists of sufficient number of mature plant individuals. There is no specific threat in the habitat; the habitat is qualitatively good and far from human activities. Additionally, the AOO and EOO are 27,500 km² and 290,157 km², respectively. Thus, it should be assessed as the least concern (LC) in the present paper.

Molecular analysis
The aligned data matrix of the ITS region of rDNA (ITS1, 5.8S, ITS2) consists of 687 characters, of which 332 are constant and 205 are parsimony-informative sites including outgroups. Specimens belonging to T. abbreviatus have C and A at the position of 512 and 516, respectively. However, the rest examined Tragopogon taxa including T. porrifolius and T. coelesyriacus clades consisting of nucleotides A and G at the same sites. Bayesian phylogram (majority rule consensus tree) of Tragopogon based on ITS with JK support values is presented in Figure 4. According to this phylogram, members of Tragopogon were separated from out-group (PP = 1, JK = 100) and fell in 4 distinct subgroups (PP = 1.00, JK = 100). The examined specimens treated under T. porrifolius before this study were grouped in the same clade in Subgroup I with good supports (PP = 1.00, JK = 64). However, individuals belonging to T. porrifolius subsp. eriospermus, T. abbreviatus and T. porrifolius subsp. longirostris were discriminated from each other in the phylogenetic tree.

Discussion
T. abbreviatus was firstly described as a variety of T. longirostris based on specimens of Boissier, Bourgeau, Calvert, and Koch collected from different parts of Turkey. Although Boisser (1875) did not specify collection details, i.e. collection date, collector number and herbarium, in the protologue, he noted that the beak of T. longirostris var. abbreviatus was scarcely longer ("rostrum achenio vix longius"), not well 2 times longer ("vel duplo longius") as in the T. longirostris var. longirostris. The authentic material of the name is currently stored at G, but a critical examination of these specimens revealed that 4 of them (G 00330029!, G 00330030!, G 00330032!, G 00330034!) clearly belong to T. abbreviatus due to achene traits and general habitus. However, specimen collected from A8 Erzurum, Tortum ("in Armenia circa Turtum", Calvert s.n., G [G00330031!]) differs from the rest of the cited specimens due to unbranched stem and the number of phyllaries (6 digit), so it was identified as T. longirostris var. longirostris (=T. porrifolius subsp. longirostris) in the present study. The specimens stored at P (P 00720169 photo!) and B (B 100093737!) collected from A8 Rize: Cimil (Ponto Lazico) listed under name of T. longirostris var. abbreviatus by Matthews (1975) were identified as  T. dshimilensis K.Koch by Coşkunçelebi et al. (2017). Additional specimens of P.H. Davis (D. 44523 (E!)) and Trelawny (Trelawny 1003 (E!)) listed under the name of T. longirostris var. abbreviatus by Matthews (1975) in the Flora of Turkey were also identified in the present study as T. reticulatus Boiss. & Huet and T. pterocarpus DC, respectively.
Tragopogon abbreviatus is morphologically related to T. longirostris var. longirostris and T. porrifolius subsp. eriospermus. Although the length of achene beak of T. abbreviatus is shorter than the closely related taxa, achene beak of T. abbreviatus (Figure 1) and T. porrifolius subsp. eriospermus ( Figure 5) is straight and stout contrary to T. porrifolius subsp. longirostris ( Figure 6) characterized  by a thinner and sometimes curved beak as well (Table  2). Additionally, the achene beak is typically sulcate in T. longirostris var. longirostris and T. abbreviatus ( Figure  1, Figure 6), and not sulcate in T. porrifolius subsp. eriospermus ( Figure 5). More detailed morphological comparisons are given in Table 2.
According to Gültepe et al. (2015), the chromosome number of T. abbreviatus is 2n = 2x = 12 based on 4 different specimens. However, detailed morphological investigations of these specimens stored at KTUB revealed that none of them belong to T. abbreviatus. Pollen grains with oblate spheroidal shape were reported in T. abbreviatus by Gültepe et al. (2018) based on specimens collected from Muğla: Seki, Coşkunçelebi & M. Gültepe 345! (Appendix).
The present phylogenetic analysis is a preliminary study examining a large number of Tragopogon accessions from Turkey. The molecular phylogenetic analysis based on the ITS data set with multiple sampling supported the monophyly of the genus Tragopogon as indicated in recent molecular studies (Mavrodiev et al., 2004;2005). ITS dataset also confirmed that the genus Tragopogon is Group V, but shorter achene length Group V Group V, but beak not angled Group VII and Group VIII distinct from the closely related genus Gerepogon L. and other members of Subtribe Scorzonerinae as stated by Mavrodiev et al. (2004). As seen in Figure 4, 4 subgroups appeared in the phylogenetic tree. All sub-groups consist of yellow and/or purple-flowered species. Similarly, both (all) sub-groups are not discriminated on features of peduncles below the capitula (all sub-groups consist of swollen peduncles). According to the phylogenetic tree, 5 accessions belonging to the T. abbreviatus are clustered with strong support (PP = 1.00, JK = 93). Thus, present phylogenetic tree inferred from the dataset of our own and GenBank support the status of T. abbreviatus as a distinct species, and also distinguish T. longirostris var. longirostris from T. porrifolius as previously suggested by Greuter (2003). Gültepe et al. (2016) also made similar taxonomical comment about T. longirostris var. longirostris. As seen in molecular phylogenetic tree (Fig.  4), T. longirostris and T. coelesyriacus (Boissier, 1849) are sister taxa so 2 of them should be treated under the name of T. coelesyriacus takes priority over that the names of T. longirostris (Feinbrun-Dothan, 1978;Dimopoulos et al., 2016). Besides, morphological examination of the holotype of T. coelesyriacus stored at G-Boiss (G00330023 photo!) shows that it belongs to T. longirostris as stated by Boissier (1875). However, T. coelesyriacus clade consists of 2 recently accepted taxa T. porrifolius subsp. australis, T. porrifolius subsp. longirostris and T. krascheninnikovii (Nikitin, 1933) proposed as a synonym to T. longirostris by Rechinger (1977). It looks that the T. coelesyriacus clade is a monophyletic complex including several unresolved taxa so further studies need to solve the infraspecific taxonomy of this clade (Fig. 4). Additionally, as seen in the phylogenetic tree ( Figure  4), some representatives of T. porrifolius were aggregated into Subgroup I and rest of accessions are appeared in polytomy. This situation coincides with the view of Mavrodiev et al. (2007), who indicated that T. porrifolius is a polyphyletic complex based on combined ITS and ETS datasets. Present findings also supplied additional molecular evidence for T. dshimilensis, previously accepted as a synonym of T. abbreviatus by Boissier (1875). As seen in Figure 4, T. dshimilensis is located in the Subgroup II with strong support (PP = 1.00, JK = 61) rather than T. abbreviatus. In conclusion, the present study has shown that T. abbreviatus, T. porrifolius subsp. longirostris and T. coelesyriacus are not the members of T. porrifolius complex contrary to what stated in Greuter (2003). Besides, this complex includes in Turkey only T. porrifolius subsp. eriospermus contrary to what reported by Coşkunçelebi and Gültepe (2012).