Discussion
In this study, we carried out molecular phylogenetic analyses using ITS and trnL-trnF alignments, as well as the combination of these datasets to evaluate intra-generic relationships of Tulipaspecies growing wild in Kosovo. Our data revealed the feasibility of ITS and trnL-trnF sequences for the phylogeny of Tulipaspecies, confirming previous findings for successfully use of ITS (Christenhusz et al., 2013; Fay et al., 2006; Turktas et al., 2013; Yanagisawa et al., 2012) and trnL-trnF sequences (Peterson et al., 2008). Phylogenetic trees obtained from ITS sequences provided better resolution compared with those generated from trnL-trnFsequences, which are in accordance with previous reports (Peterson et al., 2008; Sang et al., 2015; Turktas et al., 2013). The generated trees from the combined ITS + trnL-trnF dataset showed more similarities with trees obtained from ITS sequences, but in general with stronger support.
The analysed sequences of Tulipa species grouped into two main clades, one composed by specimens of the subgenus Eriostemones (T. sylvestris ) and the second composed by specimens of the subgenusTulipa (T. albanica, T. kosovarica, T. luanica, T. scardica, T. serbica, T. ulophylla, T. tschimganica, T. suaveolens, T. julia andT. gesneriana ). The subgenera Eriostemones andTulipa showed strong support for the monophyly generated by all applied methods, which agreed with previous findings (Christenhusz et al., 2013; Turktas et al., 2013). All obtained phylogenetic trees based on ITS, as well as combined ITS and trnL-trnF datasets, grouped the analyzed species of subgenus Tulipa , including sectionSpiranther (T. tschimganica ), section Tulipanum(T. julia and T. ulophylla ) and section Tulipa (T. albanica, T. kosovarica, T. luanica, T. scardica and T. serbica ) confirming the previous classification of those species by Christenhuszet al. (2013). Phylogenetic analyses based on trnL-trnFsequences fully congruent with those provided by ITS.
Within the Tulipa clade, generated by the ITS and combined ITS+ trnL-trnF datasets, the most distinct species from the newly generated sequence was T. tschimganica, which belongs to section Spiranthera , followed by T. julia and T. ulophylla (section Tulipanum ), while T. suaveolens andT. gesneriana were the closest to the newly sequenced species as all of these species belongs to section Tulipa et al. 2012; Zonneveld, 2016). Furthermore, grouping of the species T. scardica, T. serbica, T. albanica, T. kosovarica andT. luanica in one subclade (section Tulipa ), clear evidenced a close relationship between those taxa and confirmed that these species belong to a complex of species (T. scardicacomplex) distributed in the southern Balkans (Christenhusz et al., 2013), which cannot be clearly resolved by molecular methods. However, in most of the cases intra-specific variation was detected within theT. scardica complex indicating the presence of polymorphism in the gene pool. The polymorphism within populations of the Tulipaspecies was reported previously too (Booy and Raamsdonk, 1998; Christenhusz et al., 2013; Mayer Ernest, Micevski, 1970; Zonneveld, 2009).
Our phylogenetic analyses did not show species-specific resolution between newly sequenced specimens of the T. scardica complex (T. albanica, T. scardica, T. serbica, T. kosovarica and T. luanica ), except T. albanica , which was moderately supported (Figure 1A) or moderately to strongly supported (Figure 3A).
Between those species, except molecular similarities the morphological similarities exist too. Because of their similarities T. scardicasometimes have been synonymized as T. gesneriana (Christenhusz et al., 2013; “The Plant List,” 2013; Zonneveld, 2016), not accepted by Flora Eurepea (Tutin et al. 1980), but accepted as a species by theWorld Checklist of Selected Plant Families (Govaerts, 2019). Our findings based on ITS sequences confirm that T. gesneriana was erroneously used as a synonym for T. scardica. Tulipa scardicawas the first species described as a new species in the T. scardica complex (Bornnullejr, 1923), (Mayer Ernest, Micevski, 1970). Individuals of this species show great variation in several morphological characters such as leaf form, flower colour, length of filaments and anthers in different areas of its distribution (Mayer Ernest, Micevski, 1970). Tulipa serbica, also belonging to this complex (Christenhusz et al., 2013), was the first time recorded at Mt Rogozna (Pavlovic 1962) and described as T. scardica , but was later revised and described as the new species T. serbica (Tatić and Krivošej, 1997). Both species (T. scardica and T. serbica ) are considered to be closely related to each other. Tulipa serbica differs from T. scardica in its paler, unspotted periapt segments, pale (not blackish) staminal filaments, dull violet (not yellowish) and acute anthers (Tatić and Krivošej, 1997). Our phylogenetic results generated from ITS and ITS + trnL-trnFsequences did not support the separation of T. serbica as independent species, while results obtained by the trnL-trnFsequences weakly supported this opinion. Based on those findingsT. serbica could not be confirmed as independent species. Specimens of T. kosovarica collected for the first time along Mrasori river (Mirusha region) at the foot of Mt Kozniku in 2010, were described in that time as T. scardica (Shuka et al., 2010). In 2012 the material was revised and described as T. kosovarica(Shuka et al. 2012). Later, this species was recorded in some other location such as Guriç, Llapushnik, Qafë - Prush, Devë (Millaku et al., 2018). Tulipa kosovarica shares morphological similarities with T. scardica, T . serbica and T. albanica (Shuka et al., 2010). It differs from T. scardica by its white or whitish perianth base that is sometimes masked by obtrullate patches of maroon and violet, while T. albanica differs from this species by its combination of yellow perianth bases without black blotches (Shuka et al. 2012). Phylogenetic analyses obtained by the ITS,trnL-trnF and ITS + trnL-trnF dataset did not show the divergence of T. kosovarica sequences from other taxa of theT. scardica complex. Thus phylogenetic results did not support the separation of the T. kosovarica as an independent species.Tulipa luanica is the most recent species described as member of the T. scardica complex (Millaku and Elezaj, 2015). According to Millaku and Elezaj (2015) T. luanica shares many morphological characters with T. gesneriana, T. albanica , T .kosovarica and T . serbica , but also differs in several characters, including the substrate (T. luanica grows exclusively on limestone, while other species grow only on serpentine). Based on our sequence analyses there is no genetic difference betweenT. luanica and T. scardica . T. albanica, another species of the T. scardica complex (Christenhusz et al., 2013).Tulipa luanica Millaku was recorded as a new species in Northeast Albania for the first time, but it was recently found in Kosovo too (Millaku et al., 2018). Tulipa albanica shows great variation in several morphological characters, for example its campanulate flowers exist in two colour forms, yellow to golden-yellow or carmine-scarlet turning deep reddish maroon, with a dominance of the golden-yellow flowers (Shuka et al., 2010). Furthermore some individuals have an intermediate colour of yellow to reddish maroon. Tulipa albanicashares many morphological similarities with T. scardica, T. serbica, T . kosovarica and T. luanica, but it differs from them by its combination of yellow perianth bases without black blotches, yellow filaments and violet-purple pollen (Shuka et al. 2012, Millaku and Elezaj, 2015). Our phylogenetic results based on ITS,trnL-trnF and ITS + trnL-trnF datasets showed weak to moderate supported divergences between T. albanica and other taxa of the T. scardica complex, what does not necessarily confirmT. albanica as independent species, but indicate the presence of polymorphism in the gene pool of this taxon.
Our phylogenetic analyses showed that the unidentified Tulipa species (sample T9, Table 1) obtained from herbarium material of the Herbarium of the University Prishtina, belongs to a taxon of theT. scardica complex.
In the morphological analysis, the flower colour was one of the main characters used to discriminate the species of the T. scardicacomplex, but the flower colour appeared to be very variable within one species (Eker et al., 2014; Mayer Ernest, Micevski, 1970; Millaku and Elezaj, 2015; Shuka et al., 2010; Shuka et al. 2012; Zonneveld, 2009). Hence, it seems not to be very suitable for the classifications of Tulipa species (Christenhusz et al., 2013). For example, the flower colour even within one population of T. albanica was reported to be form yellow/golden-yellow to carmine-scarlet turning deep reddish maroon (Shuka et al., 2010), including individuals with intermediates colour. Later individuals with intermediate flower colour were reported and explained as natural hybrids of different species in sympatric distribution areas (Millaku et al. 2018). For example, individuals with mixed characters like half of the perigon base in yellow colour (an inherited trait from T. gesneriana ) and the other half of the perigon base in white colour (an inherited trait from T. kosovarica ) were recorded, or intermediate individuals with yellow perigon base (as it is in T. gesneriana ), while the rest of the perigon was pink (like to T. luanica ) (Millaku et al., 2018). According to Raamsdonk and Vries (1995) the flower colour within species may differ in two aspects, first the blotch and the blotch margins may show differences in size and in colour intensity and secondly, within some species anthocyanidins are lacking in certain accessions resulting in yellow or very light colours. Experiments based on selection of accessions obtained from natural provenances, as well as mutation experiments with radiation showed that blotch margin and flower colour can easily be influenced (Christenhusz et al., 2013).
Genome size analyses (2C) of Tulipa species revealed that 2C ofT. albanica was 54.15 pg (Shuka et al., 2010) or 43.86 pg (Osmani, 2018), 45.71 pg for T. kosovarica and 47.49 pg forT. luanica (Osmani 2018) and 69 pg for T. scardica(Zonneveld, 2009). The incongruent results for T. albanicareported by Shuka et al. (2010) and Osmani (2018) were explained by the origin of the plant material by Osmani (2018). Osmani (2018) used leaves collected from plants in blooming time, from wild populations, while Shuka et al. (2010) used adult leaves germinated from seeds collected from natural populations, what seems to be an unconvincing explanation. Such differences of genome sizes within species could be correlated with differences in the habitat (Jakob et al., 2004), plant phenotype (Beaulieu et al., 2005), or probably caused by technical artefacts (Obermayer and Greilhuber, 2005). However, DNA content and cytogenetic analyses were not carried out in all of the species present in Kosovo to provide information about relationships of species within T. scardica complex.