Delimitation of Iranian species of Scorzonera subg. Podospermum and S. subg. Pseudopodospermum (Asteraceae, Cichorieae) based on morphological and molecular data

Scorzonera L. is represented by 57 species in Iran including three subgenera: S. subg. Scorzonera, S. subg. Podospermum and S. subg. Pseudopodospermum. Species of S. subg. Podospermum and S. subg. Pseudo­ podospermum in Iran are morphologically similar, which limits species delimitation. In order to clarify intersubgeneric and interspecific delimitation in Iran, we carried out extensive sampling of the two subgenera in Iran. We conducted phylogenetic analyses based on the nuclear Internal Transcribed Spacer (nrITS), detailed morphological studies, and we evaluated the systematic value of achene features. Our results showed that Scorzonera s.l. is polyphyletic, and both S. subg. Podospermum and S. subg. Pseudopodospermum are monophyletic. The monophyly of S. subg. Podospermum morphologically corresponds to a combination of characters containing pinnatifid leaves, phyllaries with black corniculate projections, and the presence of a swollen carpopodium on the achenes. A comparison of the topology observed in the nrITS phylogeny with achene features indicates that a sculptured achene wall surface in members of S. subg. Pseudopodospermum provides a synapomorphy for this lineage. This study supports a broader circumscription of S. subg. Pseudopodospermum with the addition of S. calyculata (S. sect. Incisae), S. ovata, S. papposa and S. paradoxa (S. sect. Papposae). Finally, we provide a taxonomic treatment, including an identification key and species diagnoses and distributions, with nomenclature of Iranian species.

Previous molecular phylogenetic studies have sampled broadly across subtribe Scorzonerinae. The genus Scorzonera was resolved as polyphyletic, based on the nuclear ribosomal Internal Transcribed Spacer (nrITS; Mavrodiev & al. 2004) and a combined nrITS and external transcribed spacer (ETS) analysis and Amplified Fragment Polymorphisms (AFLPs) by Winfield & al. (2006). Intergeneric nodes in those analyses were, however, statistically unsupported. Both studies revealed that the "Lasiospora clade", named after the Scorzonera segregate Lasiospora Cass. based on S. hirsuta (Gouan) L., represents a lineage that is far from the core of Scorzo nera, in the sense of its type S. humilis L. Mavrodiev & al. (2004) showed that the Lasiospora clade can also be distinguished from Scorzonera based on chromosome number (6 and 7, respectively; see also Nazarova 1977;Diaz De La Guardia & Blanca 1987;Martin & al. 2012). The morphological distinction of Lasiospora species from Scorzonera is, however, unclear. Importantly, both studies confirmed that S. subg. Podospermum is monophyletic, but with S. purpurea L. resolved as sister to the Podospermum clade. Zaika & al. (2020) recently provided a taxonomic reassessment of Scorzonera s.l. based on broad taxonomic sampling, carpological (including anatomical) data, and nrITS and two plastid markers (partial rbcL and matK) molecular phylogenetic analyses (Zaika & al. 2020). That study confirmed the polyphyly of Scorzonera and proposed a revised classification of the subtribe. As a result of their analyses, the following seven genera were confirmed: Gelasia Cass.; Pseudopodospermum (Lipsch. & Krasch.) Kuth.; Pterachaenia (including S. codringtonii Rech. f.); Scorzonera (including four major clades: Po dospermum, Scorzonera s. str., S. albicaulis Bunge and S. purpurea); Takhtajaniantha; and the newly described Lipschitzia Zaika & al. (S. divaricata Turcz. clade) and Ramaliella Zaika & al. (S. polyclada Rech. f. & Köie clade). Therefore, the authors proposed a narrow circumscription of Scorzonera (containing the Podospermum clade) and accepted Pseudopodospermum as a separate genus. At present, we are uncertain of the most appropriate taxonomic concept for the clades Podospermum and Pseudopodospermum. Therefore, for the purpose of this study, we follow the wider circumscription of genus Scor zonera that recognizes these lineages at subgeneric rank: S. subg. Podospermum, and S. subg. Pseudopodosper mum. This taxonomic concept is in accordance with the following studies: Lipschitz (1964), Rechinger (1977), Safavi (2013), and Coşkunçelebı & al. (2015). In contrast to Zaika & al. (2020), who investigated generic-level relationships within Scorzonera s.l. Our study focuses on the shallower taxonomic levels and aims to clarify the morphological delimitation among Iranian species within S. subg. Podospermum and S. subg. Pseudopodospermum and compare it with the nrITS tree.
Currently, the genus Scorzonera is represented by 57 species in Iran; of which 33% (20 species) are considered endemic to the country (Rechinger 1977;Safavi 2013;Safavi 2016;Safavi 2019). Of the 57 species in Iran, 36 belong to S. subg. Scorzonera (137 worldwide), nine to S. subg. Pseudopodospermum (20 worldwide) and 12 (21 worldwide) to S. subg. Podospermum (Kamelin & Tagaev 1986;Rechinger 1977;Safavi 2013;Safavi 2016;Safavi 41 Willdenowia 50 -2020 2019). In Iran, S. subg. Podospermum and S. subg. Pseu dopodospermum contain closely related species that are currently difficult to distinguish based on morphological characters, which limits species identification. With the exception of leaf-anatomical studies of Iranian species of S. subg. Pseudopodospermum and S. subg. Podospermum (Norouzi & al. 2016), and karyological analyses of a limited number of species from Iran (Safavi 1999;Bord bar & al. 2019;Hatami & al. 2019), no comprehensive study has been carried out to date to clarify the nomenclature and intergeneric and interspecific diagnostic characters of these subgenera in Iran. In order to clarify the circumscription of S. subg. Podospermum and S. subg. Pseu dopodospermum and give stronger insight into these lineages in Iran, we conducted extensive sampling of the two subgenera in Iran for detailed morphological studies. We also sampled representatives of all major clades across Scorzonerinae for phylogenetic analysis based on nrITS. Therefore, the aims of the present study are to: 1. Conduct phylogenetic analyses using nrITS sequence data to test the monophyly of Scorzonera subg. Po dospermum and S. subg. Pseudopodospermum in Iran and evaluate the phylogenetic relationship between species. We sampled representatives from across subtribe Scorzonerinae and the two subgenera with a focus on Iranian species. 2. Investigate morphological characters of species of Scorzonera subg. Podospermum and S. subg. Pseu dopodospermum to understand the morphological boundaries and diagnostic characters both between the subgenera and among their species in Iran. 3. Assess the systematic value of achene features for distinguishing between taxa and to compare the patterns of achene features with the inferred topology from phylogenetic analyses of the nrITS region. 4. Provide a taxonomic treatment, including diagnoses and distributions, with a clarified circumscription and nomenclature of all species of Scorzonera subg. Podospermum and S. subg. Pseudopodospermum in Iran. We also provide an identification key for the two subgenera and their species in Iran.

Material and methods
Sampling -For nrITS analyses, the sampling approach for Scorzonera subg. Podospermum and S. subg. Pseu dopodospermum was guided by treatments in Flora iranica (Rechinger 1977). Refer to Appendix 1 for a list of all samples included in the nrITS analyses with voucher information and GenBank numbers for newly generated sequences in this study. See Table 1 for classifications of Scorzonera species based on previous studies. In order to include representatives from the major clades across Scorzonerinae in the molecular analyses, we also incorporated already published nrITS sequence data of members outside of the focus subgenera. Therefore, the ingroup comprised extensive sampling from Iranian S. subg. Po dospermum and S. subg. Pseudopodospermum (Rechinger 1977), and sampling outside of those subgenera: species from S. subg. Scorzonera and from the genera Epilasia, Koelpinia, Pterachaenia and Tragopogon guided by the clades in subtribe Scorzonerinae in the nrITS analyses of Mavrodiev & al. (2004;Appendix 1 Phylogenetic reconstruction -Genomic DNA was extracted from leaf material using the DNeasy Plant Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer's protocol. For amplification of the nrITS region, primers ITS-A and ITS-B were used (Blattner 1999 (Seoul, Korea) using the same PCR primers. Sequences were initially aligned using MAFFT v. 6.0 (Katoh & Toh 2008) and checked manually using the program PhyDE v. 0.9971 (Müller & al. 2005). Indels were coded as binary characters using the simple indel coding approach, according to Simmons & Ochoterena (2000) in SeqState v. 1.4.1 (Müller 2005). Phylogenetic analyses were conducted using Maximum Parsimony (MP), Maximum Likelihood (ML) and Bayesian Inference (BI). Maximum Parsimony analyses were performed using heuristic searches in PAUP* v. 4.0b10 (Swofford 2003) in combination with parsimony Ratchet (Nixon 1999) in PRAP (Müller 2004). Ratchet settings included 200 iterations with 25 % of the positions randomly unweighted (weight = 2), and 100 random additional cycles. Jackknife Table 1. Scorzonera species included in this study and their subgeneric classification based on previous literature (in column 2 : Lipschitz 1964;Rechinger 1977;Kamelin & Tagaev 1986;Mavrodiev & al. 2014;Safavi 2013;Zaika & al. 2020) compared to results of phylogenetic analyses (Fig. 1) and the taxonomic treatment of this study (in column 3).
Macro photographs of achenes from species belonging to Scorzonera subg. Podospermum and S. subg. Pseu dopodospermum were made using a stereomicroscope (Olympus SZX16) equipped with DP72 (a 12.5 megapixel digital colour camera), connected to cellSens Standard programme with an extended focus imaging function, at the Botanic Garden and Botanical Museum Berlin.

Results
Molecular analyses -The aligned nrITS data matrix comprised of 55 sequences and 826 characters including 126 coded indels, 291 parsimony informative sites and 156 parsimony uninformative sites (see alignment in Appendix 3). Maximum Parsimony analyses resulted in 16 most parsimonious trees with a length of 1096, a consistency index of 0.615, and a retention index of 0.863.
Bayesian Inference, ML, and MP analyses of the nrITS dataset produced identical topologies. The MrBayes 50 % majority-rule consensus tree is presented in Fig. 1. Here, we report statistical support values that are well-supported (>0.95 posterior probability [PP], and >80 % BS and JK) in parentheses (Fig. 1). Subtribe Scorzonerinae received full statistical support (1 PP, 100 JK, 100 BS; Fig.  1). The Lasiospora clade was resolved as monophyletic (1 PP, 100 JK, 100 BS) and as sister to a clade (1 PP, 99 JK, 97 BS) containing a polytomy that consisted of the rest of subtribe Scorzonerinae. This polytomy comprised five clades; one clade (1 PP, 92 BS) included in two subclades that contained the genera Epilasia (1 PP, 100 JK, 100 BS) and Tragopogon (1 PP, 100 JK, 100 BS), respectively. A second lineage in the polytomy corresponded to a single accession representing the monotypic genus Morphology -Images of achenes of 22 species are presented in this study ( Fig. 2; Fig. 3). A summary of the achene surface pattern, pubescence, carpopodium and pappus bristles of species in Scorzonera subg. Podosper mum and S. subg. Pseudopodospermum from Iran is provided in Table 2. Our observations revealed that, in S. subg. Podospermum, outer achenes are smooth, subterete to sulcate, glabrous or lanate, with a conspicuous carpopodium, and the pappi are apically scabrous with plumose bristles for most of the length. In contrast, in S. subg. Pseudo podospermum, the outer achenes are lamellate, muricate, Willdenowia 50 -2020 or verrucose, with or without a conspicuous carpopodium, glabrous, and the pappi either consist of bristles that are plumose for the entire length or for most of the length, and are apically scabrous, sometimes with five obvious longer scabrous bristles. More detailed characteristics of achenes for each species are given in Table 2.
In addition to achene morphology, a range of morphological features was examined for the identification key and taxonomic treatment below. Voucher information for specimens examined are provided for all species in Appendix 3; and the summary of morphological features (life cycle, plant height, root, stem, leaves, flowering   Fig. 1. Bayesian 50 % majority-rule consensus tree inferred from the nuclear Internal Transcribed Spacer dataset. Values above nodes indicate posterior probability (bold) and jack-knife support (italic), and values below nodes indicate bootstrap support. Tip names correspond to species names and GenBank numbers or DNA numbers for newly generated sequences; see Appendix 1 for specimen details and Appendix 3 for the alignment). Clade names are indicated to the right of the phylogenetic tree. The star corresponds to the monophyletic Scorzonera subg. Podospermum and the circle corresponds to S. subg. Pseudopodospermum. Squares next to two tip names correspond to accessions of species names that are newly synonymized with S. szowitzii in this study (LAC540 S. leptophylla; LAC529 S. stenocephala; see the taxonomic treatment for details).

Discussion
Relationships within subtribe Scorzonerinae -The topology inferred in our phylogenetic analyses ( Fig.  1) confirms the results of previous studies that showed Scorzonera s.l. is polyphyletic (Mavrodiev & al. 2004;Winfield & al. 2006;Zaika & al. 2020). Based on our molecular analysis, the Lasiospora clade is well-resolved as a lineage that is remote from the core of Scorzonera, in the sense of its type S. humilis, in accordance with Mavrodiev & al. (2004), Winfield & al. (2006) and Zaika & al. (2020). However, Zaika & al. revealed that S. villosa Scop. is included in this lineage, and they proposed the oldest generic name: Gelasia. Our molecular analysis showed a sister-group relationship between the genus Koel pinia (K. linearis Pall. and K. macrantha C. Winkl.) and the S. intricata clade (S. intricata and S. tortuosis sima here) from S. sect. Intricatae (Boiss.) Lipsch. within S. subg. Scorzonera. Our findings are in agreement with Zaika & al (2019), who revealed a sister relationship between the S. polyclada clade (including S. intricata, S. longipapposa Rech. f. and S. polyclada from S. sect. In tricatae of S. subg. Scorzonera) and Koelpinia proposing the generic name Ramaliella for S. polyclada. The sister relationship between Epilasia and Tragopogon in Fig. 1 is also congruent with previous studies (Mavrodiev & al. 2006;Zaika & al. 2020).
Scorzonera subg. Podospermum -The monophyly of Scorzonera subg. Podospermum (Fig. 1) morphologically corresponds to a combination of morphological characters containing pinnatifid leaves, phyllaries with a black corniculate projection (Fig. 4A), and the pre sence of a swollen carpopodium ( Fig. 2A -J; Fig. 4B). Species of S. subg. Podospermum are mostly distributed in the Euro-Siberian region in the north of Iran, and in the Irano-Turanian region, particularly the Kurdo-Zagrosian zone in the west of Iran (Zohary 1973;Rechinger 1977;Safavi 2013). This subgenus is most common in highland regions (>1000 m a.s.l) of more or less wet meadows, ruderal sandy or gravelly soils, stony slopes, or wet corners of agricultural fields (based on observations by E.H. and M.M.; Fig. 4D -F).
All species of Scorzonera subg. Podospermum that are distributed in Iran (Rechinger 1977;Safavi 2013) are included in our phylogenetic analyses, with the exception of S. lachnostegia, a rare species, which was sampled by Zaika & al. (2020) and resolved in a polytomy within the Podospermum clade in their nrITS tree. In spite of sampling multiple accessions of different species in our phylogenetic studies, the interspecific relationships were poorly resolved based on nrITS sequences, similar to Zaika & al. (2020) (Fig. 1). However, morphological characters, particularly achene features, were informative for clarifying the interspecific circumscription of this subgenus; see taxonomic treatment below (Table 2; Fig. 2A -J).
Within the Scorzonera songorica clade (Fig. 1), S. meshhedensis (Rech. f.) Rech. f. and S. songorica are morphologically similar to each other in possessing basal or cauline simple leaves (see Fig. 4E for S. meshhedensis). The pappus of these two species can be easily detached; a characteristic that is unique to this lineage among all Scorzonera species sampled in this study. The type of S. subg. Podospermum is S. laciniata L. (Lipschitz 1964). It is morphologically distinctive in having the longest carpopodium (6 -7 mm) with the highest ratio of carpopodium length to total achene length among all species of S. subg. Podospermum (Table 2; Fig. 2E, F). The S. cana clade received strong statistical support (0.99 PP, 91 JK, 80 BS), but its internal relationships are unresolved due to the low variation of the nrITS region. However, species within the S. cana clade can be distinguished based on diagnostic morphological characters (Fig. 4C, D; Appendix 4); see taxonomic treatment below.
Previous treatments placed Scorzonera purpurea in S. subg. Podospermum (Candolle 1838;Kamelin & Tagaev 1986), whereas others treated it as a member of S. subg. Scorzonera sect. Purpureae Lipsch. (Lipschitz 1964;Zaika & al. 2020). The achenes of S. purpurea possess pale and tube-like swollen bases (Lipschitz 1964;Chater 1976), which is morphologically similar to the carpopodium of species in S. subg. Podospermum (Fig. 2K, L). However, the swollen bases are entirely fertile in S. pur purea compared to infertile carpopodia in species of S. subg. Podospermum. Furthermore, members of S. subg. Podospermum differ from S. purpurea in having pinnatifid leaves and phyllaries with corniculate projections (Fig. 4A); in contrast, S. purpurea has simple leaves and no corniculate projections. Although a sister relationship between S. purpurea and S. subg. Podospermum is well-supported in our nrITS analyses (95 PP; Fig. 1), we propose to maintain the treatment of this species outside of S. subg. Podospermum based on morphological differences observed here and in accordance with previous studies (Lipschitz 1964;Zaika & al. 2020).
Scorzonera subg. Pseudopodospermum -Our nrITS analyses included all species of S. subg. Pseudo podospermum distributed in Iran (Rechinger 1977; Safavi 2013; except S. syriaca and S. turkeviczii); as well as S. mollis, which represents the type of this subgenus (Lipschitz 1964). Members of this subgenus are widely distributed in dry to humid areas of the Irano-Turanian region (Zohary 1973;Rechinger 1977;Safavi 2013) and mostly grow in highland regions (>1000 m a.s.l) on sandy, clay or gravelly steppe hills, stony and rocky slopes, limestone hills in grassy steppes, stony semi-deserts, and open grassland (based on observations by E.H. and M.M.). Scorzonera subg. Pseudopodospermum is monophyletic based on our nrITS phylogeny (Fig. 1) and the morphological observations in this study suggest that a sculptured (as opposed to smooth) achene-wall surface provides an exclusive synapomorphy for this lineage (Table 2). Within S. subg. Pseudopodospermum, the type of sculpturing on the achene surface is variable among species, and can be muricate, tuberculate, lamellate or verrucose (Table 2; Fig. 2M -T). Moreover, we include sequences of species from Iran that also have sculptured achenes from S. sect. Incisae and S. sect. Papposae to clarify their phylogenetic position and compare their morphological characters with members of S. subg. Pseudopodospermum. The S. mollis clade in our nrITS tree ( Fig. 1) represents S. subg. Pseudopodospermum in its traditional circumscription (Lipschitz 1964;Rechinger 1977; Fig. 1). Based on our nrITS tree, S. calycu lata (S. subg. Scorzonera sect. Incisae) and S. papposa (S. subg. Scorzo nera sect. Papposae) are sister to the S. mollis clade with full statistical support and together they form a monophyletic clade. In Iran, S. sect. Incisae is represented by S. calyculata, and S. sect. Papposae by S. ovata Trautv., S. papposa and S. paradoxa Fisch. & C. A. Mey. ex DC.; we include all species in our morphological studies and the taxonomic treatment (see below). Morphological comparisons revealed that the sculptured achene surface supports a close relationship between the S. mollis clade and S. calyculata and all members of S. sect. Papposae in Iran, in support of the close relationship observed in the nrITS tree (Table 2; Fig. 1). We therefore propose a broader circumscription of S. subg. Pseudo podospermum to include members of S. sect. Incisae and S. sect. Papposae (S. calyculata and S. papposa, S. ovata and S. paradoxa; Table 1). Members of S. sect. Incisae and S. sect. Papposae outside of Iran also have sculptured achene surfaces (Lipschitz 1964;Rechinger 1977); it would therefore be beneficial incorporate all members of these sections and S. subg. Pseudopodospermum into future phylogenetic studies. Based on previous treatments (Lipschitz 1964;Rechinger 1977), members of S. sect. Incisae and S. sect. Papposae were morphologically distinguishable within S. subg. Scorzonera according to the absence of a carpopodium. Furthermore, we found that the carpopodium is only sometimes swollen among members of the S. mollis clade. The newly circumscribed S. subg. Pseudopodospermum in our study contains species either without a carpopodium or with a carpopodium that may be conspicuous or inconspicuous. Our study therefore suggests that the presence or absence of carpopodia is not a diagnostic character for intersubgeneric classification within the genus Scorzonera, in accordance with Haque & Godward (1984) and Zaika & al. (2020).
The Scorzonera mollis clade within S. subg. Pseu dopodospermum (Fig. 1) contains a number of wellsupported subclades, which we now discuss. In the nrITS analyses, we include sequences of an accession that we identified as S. szowitzii (LAC530) and accessions that corresponded of S. leptophylla (DC.) Krasch. & Lipsch. (LAC540) and S. stenocephala Boiss. (LAC529) according to species descriptions in Flora iranica ( Fig. 1; Rechinger 1977). All three accessions were resolved in a strongly supported clade (0.99 PP; Fig. 1). We also studied morphological characters of multiple populations of S. leptophylla, S. stenocephala and S. szowitzii from several geographic regions in the field and herbaria (Appendix 1; Appendix 2). We observed extreme phenotypic plasticity among these species in vegetative characters including plant height, leaf width, possessing an entire or undulate leaf margin, and glabrous or tomentose indumentum (Appendix 4). All species are indistinguishable according to capitula length, number of florets in each capitulum, and achene and pappus features ( Table 2; Appendix 4). Based on overlapping morphological characters and nomenclatural priority, we treat them as S. szowitzii (see taxonomic treatment below).
Three accessions of Scorzonera raddeana in the nrITS analyses show intraspecific variation, which is consistent with morphological variation; length and width of fruiting capitula and width of leaves are variable between individuals; LAC540 and LAC534 were distinct from LAC513 (see taxonomic treatment below). However, all individuals of S. raddeana samples in the nrITS tree have the typical S. raddeana achene and pappus type (Table 2). Scorzonera phaeopappa is resolved as sister to S. semicana with high statistical support (Fig. 1). Based on our observations, a combination of the following characters represents the synapomorphy for the S. phaeopappa and S. semicana subclade: achenes with a swollen carpopodium and five conspicuously long scabrous bristles in the pappus (Table 2; (Table  2; Appendix 4). We consider them as separate species based on differences in achene and phyllary characters (see Notes under S. mucida). Hatami & al. (2019) recently found that chromosome number and ploidy are also different between these species: tetraploid and 2n = 28 in S. mucida in contrast to diploid and 2n = 14 in S. tunicata. The placement of the diploid S. tunicata in a clade with the tetraploid S. mucida in the nrITS tree may suggest that S. tunicata represents a parent of S. mucida. Further studies are required to determine if S. mucida is an auto-or allotetraploid. It was not possible to include other species from S. subg. Scorzonera that have sculptured achenes in our molecular studies because they are rare in Iran, including S. helodes Rech. f. and S. lim nophila Boiss. (S. sect. Dimophopapposae Lipsch.) and S. nivalis Boiss. & Hausskn. (S. sect. Foliosae (Boiss.) Lipsch.) (Rechinger 1977;Kamelin & Tagaev 1986). It would be beneficial to include these rare Scorzonera species that also have sculptured achenes in future studies, in order to explore their relationships with members of S. subg. Pseudopodospermum.

Taxonomic treatment of Scorzonera subg. Podospermum and S. subg. Pseudopodospermum in Iran
Habit -Most taxa in Scorzonera subg. Podospermum and S. subg. Pseudopodospermum are perennial, possessing a taproot or tuberous root with developed lateral roots (Appendix 4). Morphology and placement of tubers are not used here as diagnostic features because they are highly variable within species depending on the ecological conditions. Tubers may be cylindric or spherical and can be deep underground or near to the surface. Some species within S. subg. Podospermum possess a caudex, which is the persisting woody axis of the (former) rosette shoot (Beentje 2010) that may be branched or unbranched, characterized by densely set leaf scars or leaf remains; it is often dark brown and transversely rough with numerous dry and membranous scales at the apex. With the exception of two species in S. subg. Podospermum, all species of S. subg. Podospermum and S. subg. Pseudopodospermum are perennial herbs with either a caespitose, subcaespitose or caulescent habit ( Fig. 4: Fig. 5). In caespitose perennials, the flowering stems arise from radical rosettes and are often scape-like, thus leafless or bearing few reduced leaves or bracts (Fig. 4C, F; Fig. 5A, D, E). Caulescent perennials usually have developed cauline leaves and a branched flowering stem ( Fig. 4D; Fig. 5B). Two species (S. laciniata and S. songorica from S. subg. Podosper mum) are biennial with a thin taproot, typically lacking remains of previously withered leaves.
Leaves -Basal and cauline leaves vary from undivided to deeply pinnatisect among species of Scorzonera subg. Podospermum and S. subg. Pseudopodospermum ( Fig.  4; Fig. 5). Undivided leaves may be linear, lanceolate or ovate and the leaf margin may be flat or undulate (Fig. 5F, G). Pinnatifid leaves are pinnately divided but not all the way down to the rachis, whereas pinnatisect leaves (Fig.  4C, D) are deeply divided reaching the rachis (Allaby 1992). Leaf segments may be linear, oval-lanceolate or orbicular. The leaves can be sessile or with a long petiolelike portion usually with an enlarged base.
Capitula and phyllaries -The length and width of fullflowering and fruiting capitula were examined from herbarium samples. Length of capitulum was measured from the base of the longest innermost phyllary to the apex (Appendix 4). Phyllaries are always herbaceous, usually with a scarious margin; the width of the margin varies depending on the species. Small black spiny appendages occur on the apex of phyllaries only in members of Scor zonera subg. Podospermum (Fig. 4A).
Achenes -Achenes are sometimes ribbed and the ribbing depth varies ranging from subterete to sulcate. The ribs are either smooth (Scorzonera subg. Podospermum; Fig. 2A  Species in S. subg. Pseudopodospermum always have glabrous achenes, however in Podospermum they can be glabrous or hairy (Fig. 2). The achene features are often not fully expressed in the innermost achenes of a capitulum; therefore, we only examine the outermost achenes in this study. Carpopodium -The carpopodium refers to the basal prolongation of the achene wall forming a hollow tubelike sterile foot, which may be swollen or not in comparison to the fertile portion (Mukherjee & Nordenstam 2004). Species within Scorzonera subg. Podospermum almost always have a conspicuous carpopodium ( Fig.  2A -H). In contrast, within S. subg. Pseudopodosper mum, the carpopodium may be absent (e.g. S. calyculata [ Fig. 2S, T], S. ovata, S. papposa [Fig. 2M, N] and S. paradoxa), or inconspicuous to conspicuous (Fig. 2O -R;  Fig. 3K, L). The size and presence of the carpopodium is important for interspecific delimitation but not informative at intersubgeneric level.
Pappus -Within Scorzonera subg. Podospermum and S. subg. Pseudopodospermum, the setaceous pappus is typically persistent, but in two cases (S. meshhedensis and S. songorica) the bristles can be easily detached (S. son gorica in Fig. 2J). Pappus bristles are entirely plumose in S. ovata (Fig. 3K, L) and S. papposa (Fig. 2M), in contrast to pappus bristles that are plumose in the proximal part and naked or scabrous at the distal part for all other species in both subgenera (Table 2; Fig. 2; Fig. 3). The pappus bristles on a single achene can be of equal or unequal lengths. Some species have only five bristles that are conspicuously longer and darker than the others (e.g. S. phaeopappa; Fig. 2O). Diagnosis -Scorzonera calyculata can be distinguished from other species of S. subg. Pseudopodospermum in Iran based on a combination of the presence of pinnatifid leaves, ligules that are mostly yellow, but the ligule base and entire tube are black-purple (Fig. 5C2), and verrucose achene surfaces without a carpopodium.

Key to species of
Distribution -Armenia, Iran (north, northwest, west, central, south) and Iraq.
Notes -Leaf shape within Scorzonera calyculata varies from undivided to pinnatisect in different populations or even in one individual (Fig. 5C). Achenes are sometimes with a swollen part due to insect galls. This species is morphologically similar to the following species, which do not occur in Iran: S. incisa DC., S. lacera Boiss. & Balansa and S. violacea D. F. Chamb. Based on Lipschitz (1935), the floret colour is a diagnostic character to distinguish S. calyculata. Individuals with entirely violet florets correspond to S. incisa, S. lacera and S. violacea whereas those with florets that are a combination of yellow and black-purple correspond to S. calyculata. The morphological differences between S. calyculata, S. incisa, S. lacera and S. violacea were not sufficiently resolved in Lipschitz (1935) and Chamberlain (1975). Further molecular and morphological studies are required to examine the delimitation of these species. ( Diagnosis -Scorzonera phaeopappa can be distinguished by its entirely violet florets based on our observations of live plants and herbarium specimens (Fig. 5B), which is in agreement with previous literature (Boissier 1875;Coşkunçelebı & al. 2015).
Notes -Specimens from different localities across the distribution of Scorzonera leptophylla, S. stenocephala and S. szowitzii, including Armenia, Iran, Palestine, Syria and Turkey, were examined in herbaria and in the field (in Iran). According to Flora iranica (Rechinger 1977) and Flora of Iran (Safavi 2013), individuals of S. lepto phylla, S. stenocephala and S. szowitzii can be differentiated from one another by leaf width, undulate or entire leaf margins and density of hairs of the indumentum. Our field and herbarium observations showed, however, that these characters were very variable, even within the same population, and overlap between species. By comparing the original descriptions of S. leptophylla, S. stenocephala and S. szowitzii (Boissier 1846;Lipschitz 1935), we noted that they also overlap considerably. Lipschitz (1935) also commented on the similarities between S. leptophylla and S. szowitzii, and he differentiated them by the more undulate leaf margin and the tuber located near the surface for individuals of S. leptophylla in contrast to the plane leaf margin and deeper position of the tuber in individuals of S. szowitzii. Based on our observations, these characters are in fact very plastic in nature. Therefore, we conclude that S. leptophylla and S. stenocephala are conspecific with S. szowitzii and they are synonymized here. The characters are poorly represented on the holotype of S. szowitzii at G (G00498242), but an isotype at LE (LE01051867) shows all relevant characteristics for its delimitation. Diagnosis -Scorzonera turkeviczii is similar to S. phaeo pappa in Iran in the presence of a swollen carpopodium on the achenes and a pappus with five conspicuous scabrous bristles, which are longer than the rest ( Distribution -Iran (north, northwest, southwest, west), Iraq, Palestine, Russia (northern Caucasus) and Turkey.

Scorzonera turkeviczii
Notes -In Flora of Turkey (Chamberlain 1975), Scor zonera turkeviczii was considered a synonym of S. semi cana. Chamberlain (1975) considered S. semicana an Willdenowia 50 -2020 Irano-Turanian element that varies markedly in the length of stem, width and shape of leaves; its morphological characteristics overlap with those of S. turke viczii. However, in Flora iranica (Rechinger 1977) and the Flora of Iran (Safavi 2013), only S. turkeviczii (not S. semicana) was included in the treatment of Scorzo nera. Therefore, there are inconsistencies between Floras from different countries. More field and herbarium observations, as well as molecular and morphological studies including samples from both species across the distribution range, are needed in order to compare populations of the species from Iran and Turkey and to clarify species delimitations. We did not include S. semicana in this study because its occurrence in Iran has not been reported, with the exception of Flora of Turkey (Chamberlain 1975 Distribution -Afghanistan, Iran (northeast, east, southeast, central), Pakistan, Tajikistan and Turkmenistan.
Notes -Scorzonera raddeana (Fig. 5A) is morphologically very variable (Appendix 4) and we observe geographic structuring of the morphological variation. The populations from eastern Iran are similar to the holotype of this species, which was collected in Turkmenistan to the northeast of Iran. Individuals from the eastern range of S. raddeana have linear basal leaves 1 -3(-5) mm wide with a bent apex and an entire margin, few or no cauline leaves, and fruiting capitula 2 -2.5 cm long and 1.5 -2 cm wide. In contrast, individuals in southwestern and central Iran have wider leaves (5 -8 mm wide) and wider and longer fruiting capitula (3.5 -4[-4.5] cm long and 2.5 -3[-4] cm wide) compared to individuals in the eastern populations. However, individuals across the distribution of this species are similar in achene and pappus features, they all have yellow florets with violet or red stripes, muricate achenes without a conspicuous swollen carpopodium, and a pappus with bristles of unequal lengths; see further discussion under S. mucida below (Appendix 4; Table 2; Fig. 2Q, R). Distribution -Afghanistan, Iran (east, northeast, southeast) and Pakistan.

Scorzonera tunicata
Notes -Based on our morphological observations, the height of the holotype of Scorzonera tunicata (c. 20 cm tall) is not typical compared to that of most individuals of this species, which are typically 5 -10 cm tall and rarely reach 20 cm tall, in accordance with Rechinger (1977) (Appendix 4).