Molecular phylogenetic analysis resolves Trisetum (Poaceae: Pooideae: Koeleriinae) polyphyletic: Evidence for a new genus, Sibirotrisetum and resurrection of Acrospelion

To investigate the evolutionary relationships among the species of Trisetum and other members of subtribe Koeleriinae, a phylogeny based on DNA sequences from four gene regions (ITS, rpl32‐trnL spacer, rps16‐trnK spacer, and rps16 intron) is presented. The analyses, including type species of all genera in Koeleriinae (Acrospelion, Avellinia, Cinnagrostis, Gaudinia, Koeleria, Leptophyllochloa, Limnodea, Peyritschia, Rostraria, Sphenopholis, Trisetaria, Trisetopsis, Trisetum), along with three outgroups, confirm previous indications of extensive polyphyly of Trisetum. We focus on the monophyletic Trisetum sect. Sibirica clade that we interpret here as a distinct genus, Sibirotrisetum gen. nov. We include a description of Sibirotrisetum with the following seven new combinations: Sibirotrisetum aeneum, S. bifidum, S. henryi, S. scitulum, S. sibiricum, S. sibiricum subsp. litorale, and S. turcicum; and a single new combination in Acrospelion: A. distichophyllum. Trisetum s.s. is limited to one, two or three species, pending further study.

Recent infrageneric classifications of Trisetum by Barberá et al. (2017aBarberá et al. ( , 2017bBarberá et al. ( , 2018a accepted four sections. Trisetum sect. Acrospelion (Besser) Pfeiff. (seven species) and T. sect. Sibirica (Chrtek) Barberá (six species) are mainly distributed in the Old World (Barberá et al., 2017a; and T. sect. Trisetaera Asch. & Graebn. (±20 species) and T. sect. Trisetum (14 species) are distributed worldwide (Finot et al., 2004(Finot et al., , 2005a(Finot et al., , 2005bFinot, 2010;Barberá et al., 2018a). The Mexican and Central American T. subg. Deschampsioidea (Louis-Marie) Finot (7 species) and several isolated species such as T. angustum Swallen, T. pringlei (Scribn. ex Beal) Hitchc., and T. filifolium Scribn. ex Beal systems (Finot et al., 2004(Finot et al., , 2005a(Finot et al., , 2005b, are unplaced in current sectional or generic treatments (Barberá et al., 2017a(Barberá et al., , 2018a. Saarela et al. (2010) identified two major clades within Koeleriinae in their internal transcribed spacer (ITS) phylogeny, referring to them as the "Old World Trisetum Alliance" and the "New World Trisetum Alliance." These two clades were later renamed "Koeleriinae clade A" and "Koeleriinae clade B", respectively (Saarela et al., 2017). Species of T. sect. Acrospelion, T. sect. Trisetaera, and Old World species of T. sect. Trisetum were resolved as part of "Koeleriinae clade A", together with species of Koeleria, Trisetaria, Avellinia, and Gaudinia; whereas T. subg. Deschampsioidea and New World species of T. sect. Trisetum were part of "Koeleriinae clade B" together with Leptophyllochloa, Peyritschia, Sphenopholis, Limnodea, Trisetopsis, and the Central and South American species of the Calamagrostis/Deyeuxia complex. However, relationships among T. sect. Sibirica and the "Koeleriinae clades A and B" in Saarela et al. (2017) were not resolved perhaps due to the small number of samples (the focus of the Saarela et al. analysis was on Calamagrostis s.l. and included only 17 of the 70 species of Trisetum). The North American T. cernuum Trin. and Graphephorum present discordant placements within the Koeleriinae in nuclear and plastid trees (Quintanar et al., 2007;Saarela et al., 2017;Wölk & Röser, 2017). To mitigate some of the problems of polyphyly within Trisetum noted above, and using our own unpublished results, Soreng et al. (2017) proposed to move the "Trisetum spicatum complex" (T. sect. Trisetaera) to Koeleria, as well as accepting generic status for T. sect. Acrospelion. A fuller solution is not readily apparent due to the complex taxonomic structure of Trisetum, particularly in the deep nodes (Barberá et al., in prep.). Barberá et al. (2017b) included the following seven taxa in Trisetum sect. Sibirica: T. aeneum (Hook. f.) R. R. Stewart, T. bifidum (Thunb.) Ohwi, T. henryi Rendle, T. scitulum Bor ex Chrtek, T. sibiricum Rupr. subsp. sibiricum, T. sibiricum subsp. litorale Rupr. ex Roshev., and T. turcicum Chrtek. The section ranges from Eastern Europe, eastward to Alaska and Yukon Territory, Canada, with a center of diversity in eastern Asia. The species are characterized in having goldish-brown spikelets, lemmas with a callus glabrous or with short hairs, and glabrous ovaries. Chrtek (1968) first differentiated this group within T. sect. Trisetum as the series Sibirica Chrtek, having recurved and nongeniculate awns, not clearly twisted below and included T. sibiricum, the Central Asian T. altaicum Stephan ex Roshev., and the Himalayan species T. aeneum and T. micans (Hook. f.) Bor. (the latter with some doubts). Tzvelev (1976), in his treatment of Trisetum for the Soviet Union did not discuss the series but included all ser. Sibirica taxa of USSR in T. sect. Trisetum. Probatova (1979) and Veldkamp & Van der Have (1983) recognized this assemblage as T. subsect. Sibirica (Chrtek) Probat. including T. sibiricum and T. turcicum with acute lemmas, a callus glabrous or with very short hairs, and short aristules on the teeth of the lemma. The widespread T. sibiricum and the southeastern Asian T. bifidum were sister to the remaining members of the Koeleriinae in the ITS and matK trees of Saarela et al. (2017) and this topology was found in the ITS and the nuclear gene topo6 tree of Wölk & Röser (2017). However, in the plastid trees of these two studies, T. bifidum and T. sibiricum were resolved in the large polytomy within the Koeleriinae. Further efforts to effectively characterize and delimit Trisetum and relatives are necessary (Saarela et al., 2017).
Evaluating relationships among genera in the Koeleriinae from morphology alone is challenging since there are few synapomorphies, and when using molecules since there is incongruence among markers for some individuals and sets of taxa. Past reticulation and convergence events were evoked to explain the topological incongruence among phylogenetic trees in the tribe Poeae R. Br. sensu lato (Soreng & Davis, 2000;Quintanar et al., 2007Quintanar et al., , 2010Gillespie et al., 2008). The objective of the present study is to present a more focused phylogeny of Trisetum and relatives using plastid (rps16-trnK, rps16, rpl32-trnL) and nuclear ribosomal (ITS) DNA regions exposing the extensive polyphyly of Trisetum and showing that species of sect. Sibiricum are phylogenetically isolated. We have increased the sampling of species in this section, including six of seven recognized taxa (all except the Chinese species T. henryi).
All the analyses were conducted on the CIPRES science Gateway (Miller et al., 2010). We applied maximum likelihood (ML) and Bayesian searches to infer the overall phylogeny. The combined datasets were partitioned in accordance with the number of markers used. We selected the models of molecular evolution for the cpDNA and nrDNA regions using Akaike´s information criterion, as implemented in MrModeltest v.2.3 (Nylander, 2004). The best fit models for the data partitions were SYM+G for each marker and the combined plastid and nuclear dataset, incorporating a gamma distribution for the combined plastid and ITS. ML analyses were performed using RAxML-HPC2 on XSEDE (Stamatakis, 2014), assuming a generalized time-reversible (GTR) model (default) using the rapid bootstrap algorithm option, and 1000 replicates for assessing branch support. In all analyses, gaps were treated as missing data. Bootstrapping was automatically halted based on default criteria. Bootstrap (BS) values of 90-100% were interpreted as strong support, 70-89% as moderate, and 50-69% as weak.
Bayesian analyses (Huelsenbeck & Ronquist, 2001;Ronquist & Huelsenbeck, 2003) were performed using Mr. Bayes v.3.2.6 (Ronquist et al., 2012). Two runs were executed each with eight Markov chain Monte Carlo (MCMC) chains for twenty million generations, sampling once per 1000 generations. The analysis was run until the value of the standard deviation of split sequences dropped below 0.01 and the potential scale reduction factor was close to or equal to 1.0. The fraction of the sampled values discarded as burn-in was set at 0.25. Posterior probability (PP) of 0.95-1.00 were considered to be strong support. Trees were visualized in FigTree v.1.4.3.

Analysis of ITS sequences
The phylogenetic tree derived from the ITS sequences ( Fig. 1

Analysis of combined plastid and ITS sequences
The overall topology of the combined phylogram (Fig. 3) is similar to that of the plastid-derived tree, even in the terminal branches. As in the plastid phylogram, Trisetum sect. Sibirica (Sibirotrisetum) is included in "Koeleriinae clade B" and sister to the remaining members. Some notable minor differences with the plastid phylogeny occur in "Koeleriinae clade A", where Avellinia festucoides clade is sister (with Fig. 3. Maximum-likelihood tree inferred from combined plastid (rpl32-trnL, rps16 intron, and rps16-trnK) and ITS sequences. Numbers above branches are bootstrap values; numbers below branches are posterior probabilities; color blue indicates species of Trisetum s.l.; color red indicates species now included in Sibirotrisetum; T indicates the type species of the genus; scale bar = 0.3% substitutions/site. weak support, BS = 66, PP = 0.8) to a Trisetaria ovata, Gaudinia fragilis, Trisetaria dufourei, and T. loeflingiana clade plus a clade of Trisetum spicatum, T. rosei, T. montanum, and the Koeleria species clade. Another difference with the combined plastid tree is the placement of Trisetum sect. Sibirica in "Koeleriinae clade B". One sample of Trisetum turcicum (Tatli 5331) is sister to the remaining species of this clade, also found in the ITS-derived phylogram.

Discussion
Our analyses show that Trisetum s.l., as traditionally circumscribed, is polyphyletic with representative species distributed in multiple clades that include the other genera of Koeleriinae (see Figs. 1-3, names in blue and red text). The species align in both Koeleriinae A (with Acrospelion, Avellinia, Gaudinia, Koeleria, Rostraria, Trisetaria s.s. and s.l., Trisetum s.s.), and Koeleriinae B (with Cinnagrostis, Leptophyllochloa, Limnodea, Peyritschia, Sphenopholis, and Trisetopsis). This is the first study to include the type species of all these genera (see Fig. 3 with labeled types). Our subdivision of the Koeleriinae into clades A and B is in agreement with Saarela et al. (2017) who employed a different set of plastid markers to confirm these two clades. Given this extensive polyphyly, if Trisetum monophyly is to be maintained, the genus will be limited to its type species, T. flavescens, and perhaps one or two others. If it is expanded at all beyond this, it would be supplanted in priority by the older name Trisetaria. Kellogg (2015), relying on earlier published DNA studies and sequences with more limited sampling within genera, proposed lumping most of the Koeleriinae into Trisetaria, while accepting Graphephorum, Limnodea and Sphenopholis, and not mentioning Acrospelion, Cinnagrostis or Trisetopsis. Deeper sampling in recent investigations has shed new light on the problem (Saarela et al., 2017;Barberá et al., 2018b and in prep.). For instance, we now have data indicating that most of the Central and South American Deyeuxia or Calamagrostis s.l. belong to Koeleriinae clade B. The oldest available name for these species appears to be Cinnagrostis, a little known genus described by Grisebach (1874) for a single dichogamous species, C. polygama (type), initially thought by the author to be near Cinna L. and Agrostis L. Now that we have evidence of the substantial phylogenetic structure within Koeleriinae, we are breaking up Trisetum s.l. into smaller genera as proposed by Soreng et al. (2017), yet some new genera and substantial realignments of species are needed. We have sampled nearly all of the 70 species of Trisetum s.l. (Barberá et al., 2018b), and there is strong evidence of reticulation among several other lineages. In our study, Sibirotrisetum appears as a phylogenetically isolated and strongly supported lineage.
All species of Sibirotrisetum align in a strongly supported monophyletic clade sister to the remaining species of "Koeleriinae clade B" in the plastid (Fig. 2) and combined plastid/ITS (Fig. 3) trees, whereas, in the ITS-derived tree, they are sisters with a weak support to the entire Koeleriinae. Previous molecular studies (Saarela et al., 2017;Wölk & Röser, 2017) published before the revision of the section by Barberá et al. (2017b) identified a lineage in the plastid and nuclear trees that included only T. sibiricum and T. bifidum, and their relationship within the Koeleriinae was unresolved. We have increased the species sampling here to include six of the seven taxa of T. sect. Sibirica (all except the Chinese species T. henryi), and propose a new genus for this group.
Trisetum turcicum, a species often confused with T. sibiricum but clearly differentiated by its longer anthers and geniculate awns (Probatova, 1979;Barberá et al., 2017b), is a member of T. sect. Sibirica. However, in our ITS and combined plastid/ITS trees, the Tatli 5331 sample (from Turkey) was sister to the remaining clade while the Soreng 7950 sample (from the Caucasus) appeared in a polytomy. Further study of this species is needed.
Trisetum bifidum and T. scitulum appear to be closely related and are sympatric in their distribution. Both species have morphological similarities that include very lax panicles and paleas much shorter than the lemma. These two species formed a clade together with the Himalayan species T. aeneum, a species with short anthers.
To test our subspecific ranking, we included seven samples from different Asian regions of T. sibiricum, the most polymorphic and widespread species of this section. All samples of this species form a clade in the ITS (strongly supported) and plastid/ITS trees. However, in the plastid tree, three samples of T. sibiricum subsp. sibiricum from Kyrgyzstan form a clade sister to T. bifidum, T. scitulum, and T. aeneum while the remaining samples of T. sibiricum subsp. litorale and T. sibiricum subsp. sibiricum from Mongolia, Russia, and Alaska form a polytomy. None of our topologies present support for a monophyletic origin of the two subspecies, perhaps further study using low copy nuclear genes might resolve their relationships.
Since Avena distichophylla, lectotype species designated by Pfeiffer (1871) for Trisetum sect. Acrospelion, has not yet been combined in Acrospelion, we make the new combination below. We previously accepted Acrospelion placing 14 species in the genus . More study is needed on this genus since the nuclear ITS marker is not congruent with the plastid signal, suggesting hybridization.

Taxonomy
Because our molecular analysis recognizes a monophyletic and morphologically cohesive Trisetum sect. Sibirica, isolated from Trisetum s.s., we elevate the section to generic rank and provide seven new combinations. We also provide a new combination in Acrospelion. The species preceded by an asterisk (*) was not included in our DNA analysis.
Acrospelion distichophyllum (Vill.) Barberá, comb. nov Diagnosis: The species of Sibirotrisetum differs from Trisetum flavescens (L.) P. Beauv. in having panicles with a glabrous rachis or slightly hairy on the upper part, goldenbrown, rarely pale yellowish spikelets, a glabrous or with short hairs up to 0.7 mm long callus, and a recurved or basally slightly twisted and rarely geniculate dorsal awn.