**4. Discussion**

Previously designed primers for ETS amplification for some of the Poaceae genera and species could amplify a fragment of ~500–900 bp [30,45,46], but we aimed to obtain a shorter amplicon for a broad spectrum of Poaceae species of diverse genera that could be sequenced entirely using HTS and is suitable for metabarcoding analysis. Unfortunately, we could not find a region for a universal forward primer for all Poaceae species in the study due to the lack of long consecutive conservative regions. However, we have found a region that allowed us to design 7 primers, an equimolar mix of which proved to be efficient for specific amplification of all species in the study. New Poaceae-specific primers (degenerate 5--ETS forward and universal 18S reverse) amplify the 5--ETS fragment less

than 600 bp, which is ~300 bp shorter than other published primers could amplify for the same species in this study.

The effectiveness of the protocols for sample preparation for HTS highly depends, among other things, on the quality and quantity of the DNA. Various methods for pollen DNA extraction involve using commercial solutions such as column-based and DNA binding with magnetic beads purification methods after preliminary homogenization of the pollen sample with metal beads [47,48]. We propose a protocol based on a classical CTABlysis extraction method [49] with modifications that achieve results similar to commercial kits for pollen DNA extraction. The addition of a small amount of SDS, which helped increase the DNA extraction efficiency from fossil pollen of *Abies* spp. from Pleistocene peat [42], showed increased extraction efficiency from Poaceae pollen as well. The DNA yield from the samples is associated with the lysis efficiency of the pollen grains. In different plant species, the structure of the shells of the pollen grains can vary greatly. The use of methods of mechanical destruction, such as grinding with metal balls [47] or the use of a bullet blender [24], increases the DNA yield. In this case, the yield becomes comparable with the one we have achieved using tubes with fine sand for grinding pollen.

*Festuca* and *Lolium* genera form a phylogenetic complex where *Lolium* is a subgroup of the *Festuca* genus according to several phylogenetic studies that employed restriction fragment length polymorphism (RFLP), random amplified polymorphic DNA (RAPD), as well as rDNA (ITS region) and cpDNA sequences data for analysis [50,51]. It was also pointed out that *Festuca pratensis* is the most closely related to *Lolium* species in *Festuca*/*Lolium* complex, and suggested that the closeness of *Festuca pratensis* ITS to *Lolium* ITS sequences could represent a reticulate evolutionary event [50,51]. The closeness of these species is also supported by the fact that *Festuca* species readily cross with *Lolium* species in nature or synthetically form *Festulolium* hybrids (e.g., *F. pratensis* × *L. perenne*, or *L. multiflorum* × *F. pratensis*) [52]. Furthermore, species of these genera display a high level of sequence similarity for orthologous genes (>91% identity) and conservation of gene family content, as showed by the transcriptome analysis [53]. Several plastome barcodes have been used to untangle the relationships within complex and construct phylogenetic trees [54,55], though nuclear barcode ITS2 showed better results than plastome barcodes [56]. We have found that the 5--ETS barcode has 6 SNPs, ITS1 has 3 SNPs, and ITS2 has 8 SNPs between *Festuca pratensis* and *Lolium perenne* sequences. In contrast, *trnL-F* 5--ETS has only 2 short insertions (4 and 5 bp long) and no SNP in *Lolium perenne* sequences compared to *Festuca pratensis*. Thus, nuclear barcodes resolve these species better than the *trnL-F* plastome barcode, and ITS2 shows the least errors in distinguishing these species due to more SNPs than other nuclear barcodes have.

Plant pollen taxon identification using the *trnL* barcode showed promising results and a pollen-to-read quantitative correlation [22]. However, it was also shown that this barcode could give incorrect taxon predictions, e.g., *Lolium/Festuca* and *Arrenatherum/Poa* [24]. In this study, we have assessed the taxon identification capabilities of the adjacent plastome region *trnL-F* intergenic spacer, but it has also shown a high error rate in resolving *Lolium* and *Festuca* species. Moreover, we have observed a low amplification efficiency of this barcode for pollen-extracted DNA. The low efficiency of amplification of the plastome area may be caused by the fact that during the development of the pollen grain, chloroplasts, which can be found in both vegetative and generative cells, are destroyed, and thus cpDNA can be severely damaged [57].
