1 | INTRODUCTION
The tribe Triticeae (Poaceae) represented an important gene pool for genetic improvement of cereal crops and forage grasses. It included approximately 450 species that distributed in a wide range of ecological habitats over the temperate and subtropical and tropic alpine regions (Dewey, 1984). The majority of species in Triticeae were allopolyploids, with ploidy levels ranging from diploid (2n=2x) to dodecaploid (2n=12x). Natural hybridization between different genera or species oftenoccurred in the Triticeae. As early as 1926, the natural hybrid ofTriticum - Aegilops - Secale was reported (Von Tschermak & Bleier, 1926). Stebbins et al. reported a series of natural hybrids in the Triticeae: natural hybrids of E. condensatus andE. triticoides (Stebbins & Walters, 1949); triploid hybrids ofAgropyron and Elymus (Stebbins & Singh, 1950); natural hybrids of Elymus and Sitanion (Stebbins & Vaarama, 1954). At the same time, they performed artificial hybridization, artificial and natural hybrids for morphology, chromosome pairing, seed set and other aspects of comparison, speculated that the possible origin of natural hybrids (Stebbins & Walters, 1949; Stebbins & Singh, 1950; Stebbins & Vaarama, 1954).
Roegneria C. Koch was a relatively large perennial genus in Triticeae, and includes approximately 130 species, most of which were tetraploid with StY -genome, nearly 70 of which were found in China (Yang et al., 2008). Roegneria species not only provided genetic material for the improvement of forage crops but could also be used as potential contributors of genes for cereal crops (Keng, 1959). Predecessors have reported some studies on the hybrids of Roegneria , such as a hybrid of Roegneria and Hordeum (Zhou et al., 1995), a hybrid of R. ciliaris and Leymus multicaulis (Zhang et al., 2008). These hybrids were created by the artificial hybridization and could not replace the value of natural hybrids. In recent years, researchers had discovered some natural hybrids of Roegneria . For example, Zeng et al. (2012) had discovered the pentaploid natural hybrids (StStYYP ) between Roegneria (StY ) andKengyilia (StYP ).
In the process of plant system evolution, hybridization was the direct cause of the formation of diploid and polyploid species and the production of reticulate evolution (Rieseberg, 1995; Soltis & Soltis, 1993). However, it was not easy to accurately identify whether a species was a hybrid and to explore origin of hybrids (Rauscher et al., 2002). Early identification of hybridization was mainly based on morphological characteristics, and it was often based on the morphology intermediate of the parents to infer whether a plant came from a hybridization. However, the reliability of morphological markers was low, and morphology intermediate was not always related to hybridization. It may also be caused by convergent evolution or environment. Therefore, morphological markers could not be used alone to identify hybrids (Rieseberg, 1995). Cytological markers have been used as important criteria for hybridization, including karyotype analysis, meiotic pairing analysis, Genomic in situ hybridization (GISH) and Fluorescence in situ hybridization (FISH), which could be used to identify and analyze natural hybrids (Han et al., 2004; Mao et al., 2017). For example, Using FISH and GISH techniques, the two parents ofElytrigia ×mucronata could be studied through signal sites on chromosomes (Paštová et al., 2019). However, due to the high parental chromosome homology of interspecific hybrids, it was difficult to explore origin of hybrids by FISH and GISH. Therefore, even if a species has been determined from morphology or cytogenetics to be a hybrid or hybrid origin, it still needs to be verified with some other evidence (Soltis et al., 1992). Phylogenetic analysis could not only reflect the genetic relationship between hybrids and parents, but also overcome the shortcomings of non-dominance and insufficient repetitiveness of other molecular markers. The method was also the first choice for identifying natural hybrids (Quijada et al., 1997; Sang et al., 1995). Since genes at different sites in the genome of diploid or polyploid hybrids derived from different parent species, these genes have different evolutionary processes. This was the basis for detecting hybridization by phylogenetic analysis (Yu et al., 2011). Single- or low-copy nuclear genes, which were less susceptible to concerted evolution, could serve as useful markers for studies of phylogenetic relationships (Lei et al., 2018; Sha et al., 2010). Among the available nuclear sequences, DNA meiotic recombinase 1 (DMC 1) gene sequences have been used to examine hybridization events (Tang et al., 2017). The chloroplast DNA (cp DNA) is maternally inherited in grasses (Smith et al., 2006). Among the available chloroplast sequences, ribosomal protein S16 (rps 16) were used to identify the maternal donor of genera in Triticeae (Yan et al., 2014).
The experimental field of Sichuan Academy of Grassland Science (SAGS) located on Northwest Sichuan Plateau, China (Hongyuan county, Sichuan Province, 31°51′ to 33°33′ N, 101°51′ to 103°22′ W) at altitude 3500 m. Two species ofRoegneria [Roegneria stricta Keng and Roegneriaturczaninovii (Drob.) Nevski] were planted very close in SAGS. We harvested the seeds of the two species and then individual planting. After three years growing, we found that 23 putative hybrids (5.75%, 23/400) randomly distributed in R. stricta field and 54 putative hybrids (16.36%, 54/330) randomly distributed in R. turczaninovii field (Figures 2 a-c). These putative hybrids grew stronger than around plants and their seed set was very low (0.23%-5.59%, Figure 3). They showed intermediate morphological characters of R. stricta and R. turczaninovii , such as pubescence of leaf, basal leaf sheath and stem node (Figures 1 d-o).
Both of R. stricta and R. turczaninovii were tetraploid perennial species (2n=4x=28) with the StY genome. R. stricta come from Luhuo County, Sichuan, China while R. turczaninovii origins from Linxi County, Inner Mongolia, China. Because these two species had the same ploidy and genome constitutions, natural hybridizations might occur between them if they were grown together for a long time. In the current study, we hypothesized that the sterile plants were hybrids of R. stricta and R. turczaninovii . To determine if this is indeed the case, we conducted different methods including morphological analysis, fertility analysis, karyotype, meiotic pairing analysis, in situ hybridization and DNA sequence analysis in these putative hybrids and their accompanying plants. The results provided useful resources for origin and formation mechanism of natural hybrids, species evolution in the Triticeae, and laid material foundation for breeding new varieties.