The specimens included in this study were collected in the field by the authors and dried in silica-gel, or sampled from herbaria. Vouchers for specimens collected by the authors are deposited in the National Herbarium of Canada, Canadian Museum of Nature (CAN), the United States National Herbarium, Smithsonian Institution (US), and/or Herbarium of the Institute of Botany, Polish Academy of Sciences (KRAM). We aimed for broad taxonomic and geographic coverage of taxa in Poeae chloroplast group 1, and also sampled taxa representative of major lineages (subtribes) of Poeae chloroplast group 2, given known intermixing of subtribes of Poeae chloroplast groups 1 and 2 in nrDNA trees. We obtained new DNA sequence data from 421 individuals, with 1 to 17 (mean = 2.03 ± 1.88) individuals sampled per species. Following the classification of Soreng et al. (2015b), our sampling represents one subfamily (Pooideae), one tribe (Poeae), Poeae chloroplast group 1, Poeae chloroplast group 2 and 10 subtribes (Agrostidinae, Anthoxanthinae, Aveninae, Brizinae, Calothecinae, Holcinae, Loliinae, Phalaridinae, Poinae and Torreyochloinae). Bromus vulgaris (Hook.) Shear (tribe Bromeae) was designated as the outgroup given the close relationship between Bromeae and Poeae (Saarela et al. 2015). Voucher information and GenBank accession numbers for all new sequences are given in Appendix 1 and Suppl. material 1. Sources of previously published sequences are given in Suppl. material 2; in the figures these are appended with their GenBank accession number. Identifications of newly sequenced collections were made or confirmed by JMS, PMP, RJS and/or BP, and a large subset of the South American Calamagrostis/Deyeuxia material sampled from US had been identified or confirmed by Z.E. Rúgolo de Agrasar. Species of Calamagrostis/Deyeuxia from South America are referred to by their names in Deyeuxia, following Rúgolo de Agrasar (2006), except in the few cases where combinations in Deyeuxia are not available. Asian species of Calamagrostis/Deyeuxia are referred to by the generic names under which they were identified. The remaining north temperate species are referred to by their names in Calamagrostis, as commonly recognized.

We extracted DNA from leaf material using a slightly modified version of the protocol outlined by Alexander et al. (2007). We sequenced the ITS and ETS regions of the nrDNA encoding ribosome subunits in eukaryotes. ITS includes the two internal transcribed spacer regions (ITS1 and ITS2) and the intervening 5.8S nrDNA locus. The following primers were used to amplify and sequence the ITS regions: ITS1, ITS2, ITS3, ITS4 (White et al. 1990); ITS_p2, ITS_p3, ITS_u2, ITS_u4 (Cheng et al. 2016); AB102, equivalent to 26SE from Sun et al. (1994); KRC (Torrecilla and Catalán 2002); and ITS5A (Stanford et al. 2000). The 3’-end of the ETS region was amplified and sequenced using primers RETS4-F (Gillespie et al. 2010) and 18S-R (Starr et al. 2003).

We sequenced five plastid regions, including (1) the ca. 841 bp central portion of the gene matK recommended for DNA barcoding; (2) the trnL–trnF region including a portion of the 5’-trnL(UAA) exon, the 3'-trnL(UAA) exon, the trnL(UAA) intron, the trnL(UAA)–trnF(GAA) intergenic spacer and the 3’-trnF(GAA) gene; (3, 4) two intergenic spacer regions (atpF–atpH, psbK–psbI); and (5) the region spanning trnH to psbA. In grasses, the rps19 gene is inserted between the trnH and psbA genes, so the widely sequenced “psbA–trnH intergenic spacer” comprises the psbA–rps19 intergenic spacer, the rps19 gene and the rps19–trnH intergenic spacer. For clarity, we refer to this region as psbA–rps19–trnH. matK was amplified and sequenced with matK-2.1F (Kress and Erickson 2007), matK-1326r (Cuénoud et al. 2002) and two new primers we designed: matK_po1F (5’-CGCTCTATTCATTCAATATTTC-3’) and matK_po3R (5’-CGTACCGTGCTTTTATGTTTACGAG-3’). matK_po1F has the same binding location as matK-390f (Cuénoud et al. 2002) but is modified by one nucleotide, and matK_po3R has the same binding location as MatK-3FKIM-r (Ki-Joong Kim, unpublished primer) but is modified by three nucleotides. For samples that would not amplify for the full matK fragment, internal primers matK_ag520F (5’-TGTTCGATATCAAGGAAAGGCA-3’) and matK_ag640R (5’-TCGCGGCTGAGTCCAAAAAG-3’) were designed to amplify and sequence the region in two overlapping fragments. The newly designed plastid primers were based on an alignment of Poeae chloroplast genomes (Saarela et al. 2015). trnL–trnF was amplified and sequenced with primers c, d, e and f developed by Taberlet et al. (1991) and five primers newly designed during this study: C_113f (5’-TCCTGAGCCAAATCCRTGTT-3’), F_1157rD (5’-AGCTATCCTGACCTTWTMTTRTG-3’), trnLF_181f (5’-AGGATAGGTGCAGAGACTCA-3’), trnLF_518f (5’- TGGATTAATCGGACGAGGACA-3’), and trnLF_808r (5’-TCTCTTCGCACTCCTTTGGG-3’). atpF–atpH and psbK–psbI were amplified and sequenced with the primers atpF, atpH, psbK and psbI (Fazekas et al. 2008). We also designed two new primers to amplify and sequence the psbK–psbI intergenic spacer: psbK_po1F (5’-TGGCAAGCTGCTGTAAGTTT-3’), psbI_po1R (5’-AAAGTTTGAGAGTAAGCAT-3’). psbA–rps19–trnH was amplified and sequenced with the primers psbAF (Sang et al. 1997) and trnH2 (Tate and Simpson 2003). Intron and exon boundaries for all plastid regions were determined by comparison with the complete plastid genome of Agrostis stolonifera L. (Saski et al. 2007).

PCR amplifications were performed in a 15 µl volume with 1X buffer, 1.5 mM of MgCl₂, 0.2 mM dNTP, 0.5 µM of each primer, 0.3 U Phusion High-Fidelity DNA Polymerase, and 1 µL of DNA template. The thermal profile was initial denaturing of 30 sec at 98 °C; 34 cycles of 10 sec at 98 °C, 30 sec at 56 °C, and 30 sec at 72 °C; and a final extension of 5 min at 72 °C. Sequencing products were generated using BigDye Terminator v3.1 Cycle Sequencing Kits (ThermoFisher Scientific, Waltham, MA, U.S.A.) with 0.5 µl of BigDye Ready Reaction Mix in a 10 µl reaction with 1 µL of PCR product as template, and the following thermal profile: initial denaturing of 3 min at 95 °C, 30 cycles of 30 sec at 96 °C, 20 sec at 50 °C, and 4 min at 60 °C. Sequencing reactions were analyzed via capillary electrophoresis using an Applied Biosystems 3130xl Genetic Analyzer. We performed base-calling and contig assembly using Sequencher 4.7 (Genes Code Corporation, Ann Arbor, Michigan) and Geneious version 8.1.8 (http://www.geneious.com) (Kearse et al. 2012). Sites in nrDNA sequences with polymorphic bases were scored as N. Alignments were generated using MUSCLE (Edgar 2004) and other alignment tools in Geneious, and then edited manually.

We compiled individual matrices for each of the seven DNA regions studied. New sequences were validated (quality control) throughout the data collection phase. A large proportion of the variable characters in the alignments, particularly those near the beginnings and ends of contigs and when we observed infraspecific variation (i.e., when multiple individuals of a species were sampled), were carefully checked on chromatograms and edited as necessary to ensure accuracy in base calling. This process was conducted iteratively for each matrix as new sequences were added. To check for putative contamination, misidentification and/or other errors, we generated neighbour joining trees for each of the seven separate matrices using the PAUP* plugin in Geneious. These trees were examined for individuals that clustered in different parts of the trees compared to congeneric and/or conspecific taxa. We re-examined the voucher specimens for these problematic samples and corrected misidentified specimens as necessary. Some previously published sequences were grossly misplaced in the ITS tree in preliminary analyses. We concluded these are erroneous (data not shown), probably reflecting mis-identifications or laboratory mix ups, and excluded them from subsequent analyses. Once the matrices were finalized, we concatenated the two nrDNA and five plastid regions into single matrices (Suppl. materials 12–19). A summary of the seven matrices is presented in Table 1, including the number of new and previously published sequences in each matrix, the length of each aligned matrix, and the average, minimum and maximum lengths of sequences in each matrix.

In this study, we generated 2425 new sequences, and the number of new sequences per DNA region ranges from 294 (ITS) to 379 (psbA–rps19–trnH) (Table 1). The ITS matrix is the largest and includes 1079 accessions. The ETS matrix includes 352 sequences, of which 328 are new. The combined ITS+ETS matrix includes 338 samples, with both regions complete for each. The psbA–rps19–trnH (380 sequences), atpF–atpH (356) and psbK–psbI (391) matrices comprise new data for all accessions except Bromus vulgaris. The matK matrix includes 928 sequences, of which 367 are new. The combined plastid matrix includes 383 accessions, each with data for at least three of the five plastid regions.

Two plastid matrices included small inversions, identified as the reverse complements of other individuals’ nucleotides in the same alignment positions and flanked by inverted repeats, similar to what has been found elsewhere (Kelchner and Wendel 1996, Graham et al. 2000). We replaced these inversions with their reverse complement sequences, as recommended by Whitlock et al. (2010). Inversions and the taxa in which they were observed are described in Appendix 1 and Suppl. material 1. A two base pair inversion in the rps19–trnH intergenic spacer region flanked by a six bp inverted repeat could not unambiguously be modified to the same inversion configuration because of a high level of intra and interspecific variation in the two bp; 9 of 12 possible permutations were present in the matrix (i.e., GG CC CG CT TG GA AG TC CA). We staggered this alignment region because it was impossible to simultaneously determine positional and inversion configuration homology.

All analyses were conducted on the CIPRES science Gateway (Miller et al. 2010). We conducted maximum likelihood (ML) and Bayesian inference (BI) analyses with the ITS+ETS and combined plastid matrices, and ML analyses with each individual plastid matrix and the ITS matrix, which was too large for BI analysis to reach convergence in the maximum available analysis time (168 hrs) on the CIPRES server (J.M. Saarela, pers. obs.). We determined models of evolution using the Akaike Information Criterion (AIC) in jModelTest2 (Darriba et al. 2012). The best fit models for the data partitions were General Time Reversible (GTR) incorporating a gamma distribution (GTR + G) for the ETS and the ITS+ETS matrices, and TVM + G for the combined plastid matrix. In all analyses, gaps were treated as missing data; we did not code indels as separate characters. We conducted BI analyses in MrBayes 3.2.6 (Ronquist et al. 2012), with default prior settings. For the plastid matrix, we used GTR+G, the model closest to TVM + G in MrBayes. The Markov chain Monte Carlo (MCMC) analysis was set to run for 2 × 50,000,000 generations with four chains and sampled every 1000 chains, and the analysis was set to stop early if the convergence diagnostic (average standard deviation of split frequencies) was less than 0.01. The combined plastid analysis reached convergence after 8,460,000 generations. ML analyses were performed using RAxML-HPC Black Box (Stamatakis 2014), with the GTR+G model (the only one available), and bootstrapping was automatically halted based on default criteria. Trees were visualized in FigTree v.1.4.2 (http://tree.bio.ed.ac.uk/software/figtree/).

We present phylograms of the ML trees in the main text, and report both ML bootstrap and BI posterior probabilities on the ITS+ETS and combined plastid ML trees. For each of the three analyses, we provide a summary tree in which major clades, often corresponding to subtribes, are collapsed to clearly show relationships among major lineages. We present the details of these trees in multiple figures, and on each summary tree note the subsequent figures in which detailed results of the tree are presented. The ITS+ETS tree is divided into six figures, the ITS tree into nine figures, and the plastid tree into six figures. A subset of the ITS tree (Airinae p.p., Holcinae p.p., Poinae, Miliinae and Coleanthinae) is not presented in the main text. All trees are provided in full in Suppl. materials 3–11, including all single-region ones. We use the terms ‘weak or poor’, ‘moderate’, and ‘strong’ in reference to clades that received bootstrap support values of <70%, 70–90% and 91–100%, respectively; and posterior probabilities <.8, .8–.94 and .95–1, respectively. We use the term ‘unsupported’ for clades with bootstrap support <50%.

Several clades corresponding to subtribes, subtribes in part and/or multiple subtribes are recovered with moderate to strong support in the ITS+ETS tree (Figs 1, 4–9, Suppl. material 3). The ITS tree (Figs 2, 10–18, Suppl. material 4) includes substantially greater taxon sampling than the ITS+ETS tree and the ETS tree (Suppl. material 5), and identifies many of the clades recovered in the ITS+ETS tree, but support across the tree is mostly weaker, especially for deeper branches.

Figure 1. 

Overview of the maximum likelihood phylogram inferred from ITS+ETS data. Major clades in the complete tree are collapsed. The corresponding figures showing details of subsections of the tree are indicated. ML bootstrap support (left) and BI poster probabilities (right) are recorded along branches. A dash indicates bootstrap support <50%. No support is shown for branches with bootstrap support <50% and posterior probability <.5. The ML tree is presented in its entirety in Suppl. material 3.

Aveninae s.str. is monophyletic in the ITS+ETS tree because Arrhenatherum, Avena and Helictotrichon form a moderately supported clade (bootstrap support = 80%, posterior probability = .87; Figs 1, 5). In the ITS tree, however, Aveninae s.str. is not monophyletic because the four sampled genera form two separate clades: a moderately supported clade comprises Avena (80; Figs 2, 14) and a weakly supported clade comprises Arrhenatherum, Helictotrichon s.str. and Tricholemma (54; Figs 2, 14). All species of Arrhenatherum form a maximally supported clade (Fig. 14), and Arrhenatherum and Tricholemma are weakly supported as sister taxa (68; Fig. 14). All species of Helictotrichon s.str. form a moderately supported clade (76; Fig. 14). Also in the ITS tree, an unsupported lineage corresponds to Sesleriinae (Figs 2, 14) and is divided into two moderately to strongly supported subclades, one comprising Oreochloa Link and Mibora Adans. (79), the other Echinaria Link and Sesleria (98) (Fig. 14).

Figure 2. 

Overview of the maximum likelihood phylogram inferred from ITS data. Major clades in the complete tree are collapsed. The corresponding figures showing details of subsections of the tree are indicated. ML bootstrap support is recorded along branches. No support is shown for branches with bootstrap support <50%. The ML tree is presented in its entirety in Suppl. material 4.

A clade corresponding to Koeleriinae is strongly supported in the ITS+ETS tree (100, 1; Figs 1, 5), and this clade is divided into two strongly supported subclades referred to here as Koeleriinae clade A (98, 1; Figs 1, 5) and Koeleriinae clade B (100, 1; Figs 1, 6). A clade corresponding to Koeleriinae is moderately supported in the ITS tree (85; Figs 2, 15, 16) and is divided into four lineages, including Koeleriinae clade A (79; Figs 2, 16) and Koeleriinae clade B (64; Figs 15, 16), both with weaker support than in the ITS+ETS tree. The two additional lineages comprise taxa not sampled in the ITS+ETS tree: Lagurus ovatus L., on a relatively long branch (Figs 2, 15), and a strongly supported clade corresponding to Trisetum subsect. Sibirica (Chrtek) Prob. (99; Figs 2, 16). Relationships among these four lineages of Koeleriinae in the ITS tree are unsupported.

Our analyses identify several lineages within Koeleriinae clade A in the ITS+ETS tree. One clade is strongly supported and comprises Trisetum cernuum Trin. (Trisetum sect. Trisetum) and Graphephorum wolfii J.M. Coult. (95, 1; Fig. 5), with T. distichophyllum P. Beauv. resolved as its sister group (90, .7; Fig. 5). This three taxon clade is sister to a large, weakly supported clade (59, .77; Fig. 5) including the following successively diverging lineages: (1) a maximally supported clade of T. flavescens (L.) P. Beauv. (Trisetum sect. Trisetum) and Rostraria pumila (Desf.) Tzvelev; (2) Avellinia michelii (Savi) Parl.; (3) Gaudinia fragilis (L.) P. Beauv.; and (4) a clade including species of Koeleria, Trisetum sect. Trisetaera Asch. & Graebn. and Trisetum sect. Trisetum p.p. (T. macbridei Hitchc., T. irazuense (Kuntze) Hitchc.). In the better-sampled ITS tree, Koeleriinae clade A similarly includes Avellinia michelii, Gaudinia fragilis, Graphephorum, Rostraria, Trisetaria and Trisetum sects. Trisetum and Trisetaera. However, most relationship among taxa in the clade are unsupported. Strongly supported lineages within Koeleriinae clade A in the ITS tree include clades of (1) Trisetaria dufourei (Boiss.) Paunero and T. loeflingiana (L.) Paunero (97; Fig. 16), and (2) Rostraria p.p. (three species) and Trisetum sect. Trisetum p.p. (three species, including T. flavescens) (99; Fig. 16). The latter clade corresponds to a more poorly sampled clade in the ITS+ETS tree.

Our analyses identify several lineages within Koeleriinae clade B in the ITS+ETS tree. Koeleriinae clade B is divided into three deep lineages that form a trichotomy (Fig. 6). One large, weakly to strongly supported clade includes Trisetum subg. Deschampsioidea (Louis-Marie) Finot, Calamagrostis/Deyeuxia p.p. (species from Mexico and South America) and Leptophyllochloa (56, 1; Fig. 6). Most of the species from Mexico (Calamagrostis, Trisetum subg. Deschampsioidea) form a clade. The four sampled species of Sphenopholis form a strongly supported clade (92, 1; Fig. 6). A weakly supported clade includes Peyritschia deyeuxioides (Kunth) Finot and five species of Calamagrostis/Deyeuxia from South America (53, -; Fig. 6). The ITS tree includes the same taxa as well as Trisetopsis (not sampled in the ITS+ETS tree), but relationships in the clade are more poorly resolved and supported (Fig. 15) than in the ITS+ETS tree. In the ITS tree, one clade of multiple species of Calamagrostis/Deyeuxia from South America is weakly supported (52; Fig. 15), the two sampled species of Peyritschia form a clade (63; Fig. 15), and the species of Trisetopsis form an unsupported clade (Fig. 15).

A large clade comprising taxa of Agrostidinae, Brizinae and Calothecinae is weakly supported in the ITS+ETS tree (53, .73; Figs 1, 7–9) and unsupported in the ITS tree (Figs 2, 10–12), and none of these subtribes are monophyletic in the nrDNA trees. This clade includes four main lineages in the ITS+ETS tree: (1) Agrostidinae p.p. (unsupported), including most taxa currently classified in the subtribe; (2) a strongly supported clade of Deyeuxia effusa (Agrostidinae p.p.) and Chascolytrum (Calothecinae p.p.) (99, 1; Figs 1, 9); (3) Brizinae; (4) a weakly supported clade of Calamagrostis coarctata Eaton, Echinopogon caespitosus C.E. Hubb. (both Agrostidinae p.p.) and Relchela panicoides Steud. (Calothecinae p.p.) (67, .78; Figs 1, 9). These same clades are not resolved in the ITS tree. Calothecinae is not monophyletic in the nrDNA trees because Chascolytrum and Relchela do not form a clade. Chascolytrum, however, is monophyletic: the two sampled species (C. subaristatum Desv. and C. monandrum (Hack.) Essi, Longhi-Wagner & Souza-Chies) form a moderately supported clade (70, .88; Fig. 9) in the ITS+ETS tree, and the multiple sampled species form an unsupported clade in the ITS tree (Fig. 12). Neither Brizinae nor Briza are monophyletic in the ITS tree, because Briza and Airopsis do not form a clade and two lineages of Briza are resolved. One lineage of Briza is represented by B. maxima L., which is part of weakly supported clade including Aveninae s.str., Koeleriinae and Sesleriinae (Figs 2, 16). The other lineage includes the four other species of Briza sampled, which form a strongly supported clade of unclear relationship relative to taxa of Agrostidinae and Calothecinae (99; Figs 2, 10). The relationship of Airopsis to Briza and taxa of Agrostidinae and Calothecinae is similarly unclear. We were not able to test the monophyly of Brizinae or Briza in the ITS+ETS analyses because only one species of Briza is sampled there.

Agrostidinae is not monophyletic in the nrDNA trees given the placements of Calamagrostis coarctata, Echinopogon and Deyeuxia effusa in the broader Agrostidinae + Brizinae + Calothecinae clade, and some species of Calamagrostis/Deyeuxia in a clade with Deschampsia P. Beauv. (see below). Moreover, even though most other genera and species traditionally recognized in Agrostidinae and sampled here are part of the Agrostidinae + Brizinae + Calothecinae clade, they do not resolve in a supported clade in the nrDNA trees (Figs 1, 2). However, the broader clade comprising Agrostidinae p.p., Deyeuxia effusa and Chascolytrum is weakly supported in the ITS+ETS tree (Figs 1, 8). There is also some clear phylogenetic structure among subsets of taxa of Agrostidinae p.p. Species of Agrostis, Lachnagrostis and Polypogon are intermixed in two strongly supported clades in the nrDNA trees (Figs 7, 11, 12). In the ITS+ETS tree, a maximally supported clade comprises all species of Agrostis except A. exarata Trin., and P. elongatus Kunth (Polypogon sect. Polypogonagrostis Asch. & Graeb.) (Fig. 7). A similar strongly supported clade is present in the ITS tree, and also includes three species of Lachnagrostis not sampled in the ITS+ETS tree and Chaetopogon fasciculatus (Link) Hayek. (also not sampled in the ITS+ETS tree) (94; Fig. 12). The other clade in the ITS+ETS tree is maximally supported and comprises the four sampled species of Polypogon sect. Polypogon, A. exarata and L. adamsonii (Fig. 7). The equivalent clade in the ITS tree is strongly supported and includes Polypogon sect. Polypogon, A. exarata and three species of Lachnagrostis, of which only L. adamsonii is sampled in the ITS+ETS tree (97, Fig. 11).

The species of Calamagrostis/Deyeuxia that are part of the Agrostidinae + Brizinae + Calothecinae, excluding the more distantly related Calamagrostis coarctata and Deyeuxia effusa, do not form a clade in the nrDNA trees. However, some smaller clades of Calamagrostis/Deyeuxia are resolved. Moreover, the two species of Ammophila are included in different clades with species of Calamagrostis/Deyeuxia: Ammophila is not monophyletic. Ammophila breviligulata Fernald and Calamagrostis porteri A. Gray form a clade in the ITS+ETS (76, 1; Fig. 7) and ITS trees (64, Fig. 11). The broader affinities of this two-taxon clade are unsupported in the ITS+ETS tree, whereas the clade is part of a broader weakly supported clade in the ITS tree also including C. pickeringii A. Gray, C. perplexa Scribn. and C. cainii Hitchc. (75; Fig. 11). Ammophila arenaria (L.) Link, one accession of ×Calammophila baltica (Flüggé ex Schrad.) Brand. and two Chinese species of Deyeuxia (D. nyingchiensis P.C. Kuo & S.L. Lu and D. sichuanensis (J.L. Yang) S.M. Phillips & W.L. Chen.) form a clade in the ITS+ETS tree (76, 1; Fig. 8), and these three taxa are part of a broader clade including a second accession of ×Calammophila baltica, C. arundinacea (L.) Roth p.p., C. × acutiflora (Schrad.) DC., C. emodensis Griseb., C. epigeios (L.) Roth p.p., C. pseudophragmites (Haller f.) Koeler, C. rivalis H. Scholz and C. varia (Schrad.) Host. (72, 1; Fig. 8). Ammophila arenaria and the same two Chinese species of Deyeuxia, along with C. coarctata, form an unsupported clade in the ITS tree that is part of a broader unsupported clade including Echinopogon and Relchela (Fig. 10). Other clades with two or more taxa of Calamagrostis/Deyeuxia in the ITS+ETS tree comprise (1) C. bolanderi Thurb. and C. foliosa Kearney (100, 1; Fig. 7); (2) C. nutkaensis (J. Presl) J. Presl ex Steud., C. arundinacea p.p., C. brachytricha Steud., C. distantiflora Luchnik, D. scabrescens (Griseb.) Munro ex Duthie and D. pulchella (Griseb.) Hook. f. (78, 1; Fig. 7); (3) D. diffusa Keng, D. mazzettii Veldkamp, D. tripilifera Hook. f. and D. nivicola Hook. f. (54, .96; Fig. 7); (4) C. stricta subsp. groenlandica (Schrank) Á. Löve p.p., C. purpurascens R. Br. and C. deschampsioides Trin. (60, .98; Fig. 8); (5) C. epigeios p.p., C. stricta (Timm) Koeler p.p., C. × gracilescens Blytt (70, .98; Fig. 8); (6) C. canescens (Weber ex F.H. Wigg.) Roth and C. villosa (Chaix) J.F. Gmelin (99, 1; Fig. 8); (7) C. anthoxanthoides Regel p.p. and C. holciformis Jaub. & Spach (71, .98; Fig. 8). Some similar clades are resolved in the ITS tree, but most aspects of relationship among species of Calamagrostis/Deyeuxia in the ITS tree are poorly supported (Figs 10–12). The ITS tree also includes four accessions of Dichelachne, which form a clade, and these species are allied with four species of Deyeuxia from Australia and New Zealand (Fig. 12). Gastridium and Triplachne are maximally supported as sister taxa (Fig. 11) in the ITS tree, but their broader affinities are unresolved; these genera are not sampled in the ITS+ETS tree.

The other sampled tribes of Poeae chloroplast group 1 include Torreyochloinae, Phalaridinae, Scolochloinae and Anthoxanthinae. Torreyochloinae is monophyletic and strongly supported in the ITS+ETS (99, 1; Figs 1, 9) and ITS (95; Figs 2, 10) trees. In the ITS+ETS tree, Torreyochloinae and the Agrostidinae + Brizinae + Calothecinae clade are weakly supported as sister groups (56, .75; Figs 1, 9). Within Torreyochloinae, Torreyochloa pallida (Torr.) G.L. Church is sister to a maximally supported Amphibromus clade in the ITS+ETS tree, (Fig. 9), whereas A. scabrivalvis Swallen (not sampled in the ITS+ETS tree) is sister to a weakly supported clade comprising T. pallida and the remainder of Amphibromus (Fig. 10) in the ITS tree. Amphibromus is not monophyletic in this tree. Phalaridinae (Phalaris) is maximally supported in the ITS+ETS (Fig. 1, 9) and ITS trees (Figs 2, 17). In the ITS tree, Phalaris sects. Digraphis Link and Caroliniana Voshell, Stephanie M., Baldini & Hilu are sister groups (63; Fig. 17), and relationships among this clade, Phalaris sects. Bulbophalaris Tzvelev + Heterachne Dumort. (intermixed in a clade) and Phalaris sect. Phalaris are unresolved (Fig. 17). In the ITS+ETS tree, Phalaridinae and Scolochloinae (Scolochloa) are part of a broader weakly supported clade with Torreyochloinae, Agrostidinae, Calothecinae and Brizinae. The two genera of Scolochloinae (Dryopoa Vickery and Scolochloa Link) are sampled in the ITS tree, and the subtribe is monophyletic (88; Fig. 17). Anthoxanthinae (Anthoxanthum) is maximally supported in the ITS+ETS (Figs 1, 9) and ITS trees (Figs 2, 13). Clades corresponding to Anthoxanthum sects. Anthoxanthum and Ataxia (R. Br.) Stapf are maximally supported in the ITS tree, and two other clades comprising the rest of the sampled species are resolved.

In addition to Sesleriinae and Scolochloinae, which are classified in Poeae chloroplast group 2 but closely related to taxa of Poeae chloroplast group 1 in nrDNA trees, we newly sampled exemplars representing five other subtribes of Poeae chloroplast 2: Airinae, Holcinae, Dactylidinae, Loliinae and Poinae. Unexpectedly, a subset of species of Calamagrostis/Deyeuxia from South America recognized in Deyeuxia sect. Stylagrostis (Mez) Rúgolo & Villav. form a strongly supported clade with Deschampsia in the ITS+ETS (100, 1; Figs 1, 4) and ITS trees (94; Figs 2, 18); the clade in the ITS tree also includes Scribneria bolanderi (not sampled in ITS+ETS analyses). We refer to this clade in the text as the “Deschampsia clade”. Affinities of the Deschampsia clade are unresolved in both nrDNA trees. Moreover, our analyses show that Holcinae, of which we sampled Deschampsia, Vahlodea and Holcus, is not monophyletic. A lineage corresponding to Holcinae p.p. in the ITS+ETS tree is represented by Vahlodea, which is included in a weakly supported clade with Poinae (Figs 1, 4), and in the ITS tree Holcus and Vahlodea form a strongly supported clade (Fig. 2, Suppl. material 4).

The combined plastid tree (hereafter referred to as the plastid tree except when comparing and contrasting the combined plastid and single plastid region trees) includes all samples with data for at least three of the five plastid regions (Figs 3, 19–24, Suppl. material 6) and taxon sampling comparable to the ITS+ETS tree. Relationships in the plastid tree are mostly congruent with and better resolved than in the ITS+ETS and ITS trees, and there are instances of incongruence between nrDNA and plastid trees. We consider a taxon’s placement to be incongruent or discordant if it is part of different moderately to strongly supported clades in nrDNA and plastid trees. We did not conduct an incongruent length difference (ILD) test to characterize incongruence statistically because we did not conduct analyses with combined nrDNA and plastid data.

Figure 3. 

Overview of the maximum likelihood phylogram inferred from combined plastid data (atpF–atpH, psbK–psbI, psbA–rps19–trnH, matK, trnL–trnF). Major clades in the complete tree are collapsed. The corresponding figures showing details of subsections of the tree are indicated. ML bootstrap support (left) and BI poster probabilities (right) are recorded along branches. No support is shown for branches with bootstrap support <50% and posterior probability <.5. The ML tree is presented in its entirety in Suppl. material 6.

The plastid tree recovers Poeae chloroplast groups 1 (99, 1; Figs 3, 24) and 2 (100, 1; Figs 3, 19) with strong support. Poeae chloroplast group 1 consists of Agrostidinae p.p., Anthoxanthinae, Aveninae s.str., Brizinae, Calothecinae, Koeleriinae, Phalaridinae and Torreyochloinae. Phalaridinae and Torreyochloinae are sister taxa (85, .98; Figs 3, 24). Torreyochloinae is monophyletic (Fig. 24). A moderately to strongly supported clade (74, .99) includes the following four successively-diverging lineages: (1) Calamagrostis pisinna (Agrostidinae p.p.); (2) Lagurus ovatus; (3) Aveninae s.str. (84, 1); and (4) Koeleriinae excluding L. ovatus (95, 1) (Figs 3, 20, 21). Lagurus ovatus, Aveninae s.str. and Koeleriinae form a clade corresponding to Aveninae s.l. (83, .99; Figs 3, 20), and L. ovatus is the sister taxon of Aveninae s.str. + Koeleriinae (64, .95; Figs 3, 20). This placement of L. ovatus is discordant with the ITS tree (the taxon is not sampled in the ITS+ETS tree). Within Aveninae s.str., Arrhenatherum and Avena form a clade (79, 1; Fig. 20). Koeleriinae (excluding Lagurus ovatus) is strongly supported (95, 1; Figs 3, 20) and divided into two clades: Koeleriinae clade A (72, .99; Figs 3, 20) and Koeleriinae clade B (95, 1; Figs 3, 21). Within Koeleriinae clade A, the following three lineages diverge successively: (1) Trisetum distichophyllum; (2) Avellinia michelii, T. flavescens and Rostraria pumila (88, 1), with T. flavescens and R. pumila forming a maximally supported clade; and (3) a maximally supported clade including Koeleria, Gaudinia fragilis and Trisetum sect. Trisetaera (Figs 20). Within the latter clade, all species of Trisetum sect. Trisetaera and three species of Koeleria form a clade (84, 1; Fig. 20) with little internal resolution. Another clade includes the other three species of Koeleria (83, 1; Fig. 20). Relationships among these two lineages and Gaudinia are unsupported.

Figure 4. 

A portion (Holcinae p.p., Agrostidinae p.p., Loliinae, Dactylidinae and Poinae) of the maximum likelihood phylogram inferred from ITS+ETS data. ML bootstrap support (left) and BI poster probabilities (right) are recorded along branches. A dash indicates bootstrap support <50%. No support is shown for branches with bootstrap support <50% and posterior probability <.5. The shaded area of the smaller tree on the left indicates the location in the overall tree of the portion shown. The branch subtending Dactylidinae, with double slashes, is shortened for presentation. Backbone branches represented by ellipses are shown only in Figure 1.

Koeleriinae clade B comprises Calamagrostis/Deyeuxia p.p. (species from Mexico and a subset of species from South America), Graphephorum, Leptophyllochloa, Peyritschia, Sphenopholis, Trisetum subg. Deschampsioidea and Trisetum sect. Trisetum p.p. (Fig. 21). Within Koeleriinae clade B, a large clade includes all but three species of Calamagrostis/Deyeuxia from South America that are part of Koeleriinae (51, .96; Fig. 21). Sphenopholis is the only genus resolved as monophyletic (93, 1; Fig. 21). Graphephorum wolfii and Trisetum cernuum (Trisetum sect. Trisetum) form a clade (70, .7; Fig. 21). Placement of this clade in Koeleriinae clade B conflicts with its placement in Koeleriinae clade A in the nrDNA trees. Trisetum macbridei and T. irazuense (Trisetum sect. Trisetum) are part of Koeleriinae clade B in the plastid tree, whereas they are part of Koeleriinae clade A in the ITS+ETS tree. A few lineages with more than one species receive some support in the plastid tree, but most relationships among taxa in Koeleriinae clade B are unresolved.

Figure 5. 

A portion (Sesleriinae, Aveninae s.str., Koeleriinae clade A) of the maximum likelihood phylogram inferred from ITS+ETS data. ML bootstrap support (left) and BI poster probabilities (right) are recorded along branches. A dash indicates bootstrap support <50%. No support is shown for branches with bootstrap support <50% and posterior probability <.5. The shaded area of the smaller tree on the left indicates the location in the overall tree of the portion shown. Placements of samples with asterisks (***) are incongruent in nrDNA and plastid trees. Two indels in ETS and one in ITS are mapped onto the phylogram.

A large clade comprising Agrostidinae p.p., Anthoxanthinae, Brizinae and Calothecinae is strongly supported (99, 1; Figs 3, 24). Anthoxanthinae and Brizinae are successively diverging lineages sister to a strongly supported clade comprising Agrostidinae and Calothecinae (91, 1; Figs 3, 22–24). However, neither Agrostidinae nor Calothecinae are monophyletic. The Agrostidinae + Calothecinae clade includes four main lineages, all strongly supported: (1) Agrostidinae p.p. (Calamagrostis coarctata, Echinopogon) + Calothecinae p.p. (Relchela) (99, 1; Figs 3, 24); (2) Calothecinae p.p. (Chascolytrum) (93, 1; Figs 3, 24); (3) Agrostidinae p.p. (Deyeuxia effusa) (99, 1; Figs 3, 24); and (4) Agrostidinae p.p. (96, 1; Figs 3, 22–24), including most genera traditionally included in the subtribe. Calothecinae is not monophyletic in the plastid tree because Chascolytrum and Relchela do not form a clade.

Figure 6. 

A portion (Koeleriinae clade B) of the maximum likelihood phylogram inferred from ITS+ETS data. ML bootstrap support (left) and BI poster probabilities (right) are recorded along branches. A dash indicates bootstrap support <50% or posterior probability <.5. No support is shown for branches with bootstrap support <50% and posterior probability <.5. The shaded area of the smaller tree on the left indicates the location in the overall tree of the portion shown.

There is no deep resolution within the large Agrostidinae p.p. clade in the plastid tree, although several clades of two or more species of Calamagrostis/Deyeuxia are identified. The branches that define each of these clades are very short. These clades and the multiple species of Calamagrostis/Deyeuxia not included in a clade form a polytomy along the Agrostidinae p.p. backbone. Furthermore, like in the nrDNA trees, Ammophila is not monophyletic. Ammophila breviligulata is part of a clade with multiple species of Calamagrostis/Deyeuxia, whereas A. arenaria is part of the polytomy. Multispecies clades of Calamagrostis/Deyeuxia in the plastid tree include (1) C. epigeios, C. arundinacea, C. varia, C. pseudophragmites, C. rivalis p.p., and C. × acutiflora (50, .97; Fig. 22); (2) Ammophila breviligulata, C. purpurascens, C. rubescens Buckley, C. foliosa, C. sesquiflora (Trin.) Kawano, C. pickeringii, C. scopulorum M.E. Jones, C. koelerioides Vasey, C. howellii Vasey, C. cainii and C. guatemalensis Hitchc. (75, .58; Fig. 22); (3) C. llanganatensis Laegaard and C. carchiensis Laegaard (84, 1; Fig. 22); and (4) C. lapponica (Wahlenb.) Hartm., C. stricta subsp. groenlandica, C. deschampsioides, C. emodensis, C. macrolepis Litv., C. perplexa, C. epigeios p.p., C. stricta subsp. stricta p.p., C. nutkaensis, C. stricta subsp. inexpansa (A. Gray) C.W. Greene, C. stricta, C. rivalis, C. villosa, C. × gracilescens and C. canescens (57, 1; Fig. 23). Another clade (51, 1; Figs 23, 24) includes 13 species of Calamagrostis (C. arundinacea p.p., C. brachytricha, C. distantiflora, C. canadensis (Michx.) P. Beauv., C. stricta subsp. inexpansa p.p., C. porteri, C. angustifolia Komarov, C. phragmitoides Hartman, C. chalybaea Fr., C. lapponica p.p., C. cf. purpurascens, C. rubescens and C. montanensis (Scribn.) Vasey) and a maximally supported clade including Agrostis, Polypogon, Calamagrostis bolanderi, Podagrostis aequivalvis (Trin.) Scribn. & Merr. and four other species of Calamagrostis/Deyeuxia (D. tripilifera, D. nivicola, D. diffusa, D. mazzettii) (Figs 22–24). The latter large clade includes four main lineages: (1) a strongly supported clade including C. bolanderi and Podagrostis aequivalvis (98, 1; Fig. 24); (2) a moderately supported clade including five species of Calamagrostis/Deyeuxia and Agrostis rosei Scribn. & Merr. (88, 1; Fig. 24); and (3) a large strongly supported clade including species of Agrostis and Polypogon (99, 1; Fig. 24). The Agrostis + Polypogon clade is divided into two maximally supported clades. One includes all species of Agrostis except A. capillaris L. p.p., A. gigantea Roth p.p. and A. rosei. The other includes three sublineages: (1) A. capillaris p.p. and A. gigantea p.p. (95, 1; Fig. 24); (2) Polypogon australis Brongn. and P. interruptus Kunth; and (3) P. elongatus, P. monspeliensis (L.) Desf. and P. viridis (Gouan) Breistr. (92, 1; Fig. 24).

Figure 7. 

A portion (part of Agrostidinae p.p.) of the maximum likelihood phylogram inferred from ITS+ETS data. ML bootstrap support (left) and BI poster probabilities (right) are recorded along branches. A dash indicates bootstrap support <50%. No support is shown for branches with bootstrap support <50% and posterior probability <.5. The shaded area of the smaller tree on the left indicates the location in the overall tree of the portion shown. Placements of samples with asterisks (***) are incongruent in nrDNA and plastid trees. One indel in ETS is mapped onto the phylogram.

The plastid tree includes exemplars from three subtribes of Poeae chloroplast 2: Holcinae, Loliinae and Poinae. As in the nrDNA trees, a subset of species of Calamagrostis/Deyeuxia from South America recognized in Deyeuxia sect. Stylagrostis are part of a strongly supported clade with Deschampsia (100, 1; Figs 3, 19). There is little deep structure within this clade. Moreover, Holcinae, of which we sampled Deschampsia, Vahlodea and Holcus, is not monophyletic because Holcus and Vahlodea form a maximally supported clade separate from the Deschampsia clade.

Figure 8. 

A portion (part of Agrostidinae p.p.) of the maximum likelihood phylogram inferred from ITS+ETS data. ML bootstrap support (left) and BI poster probabilities (right) are recorded along branches. A dash indicates bootstrap support <50%. No support is shown for branches with bootstrap support <50% and posterior probability <.5. The shaded area of the smaller tree on the upper left indicates the location in the overall tree of the portion shown. Placements of samples with asterisks (***) are incongruent in nrDNA and plastid trees. Two indels in ETS are mapped onto the phylogram.

Numerous small indels representing tandem repeats likely arose as a result of slipped-strand mispairing and were present in each plastid matrix except matK. These indels are highly homoplasious, thus we did not score them and do not discuss them further. Non-tandem repeat indels in the plastid matrices were also present. We did not score these as separate characters in the analysis, but summarize them briefly; we also mapped these onto the trees. Several unambiguous indels are present in the psbK–psbI intergenic spacer (Appendix 1, Suppl. material 1). One is a 247 bp deletion (298 bp in the aligned matrix gaps) present in 107 accessions of 59 species, including all species in Koeleriinae clade B, Deyeuxia pulchella, Calamagrostis pisinna and Avellinia michauxii (Figs 21, 22 Suppl. material 9). An 84 bp deletion (117 bp in the aligned matrix) is shared by six accessions of four species of Calamagrostis (C. macrolepis, C. stricta subsp. stricta, C. perplexa and C. epigeios p.p.) (Fig. 23, Suppl. material 9). A 10 bp deletion is shared by all accessions of Agrostis, C. bolanderi, Deyeuxia diffusa, D. mazzettii, D. nivicola, D. tripilifera, Podagrostis aequivalvis and all accessions of Polypogon (Agrostidinae p.p.; Fig. 24); Anthoxanthum odoratum L. (Anthoxanthinae; Fig. 24); Arrhenatherum elatius (L.) P. Beauv., Avena fatua L. and A. sativa L. (Fig. 20); Briza minor L. (Fig. 24, Suppl. material 9); Gaudinia fragilis, all accessions of Koeleria and Trisetum sect. Trisetaera, Rostraria pumila, T. flavescens and T. distichophyllum (Fig. 20). In the atpF-H intergenic spacer region, a 260–268 bp deletion (varying in length at the 5’-end) is shared by one individual of Calamagrostis anthoxanthoides and six genera of Poinae (Figs 19, 22, Suppl. material 9).

Figure 9. 

A portion (part of Agrostidinae p.p., Anthoxanthinae, Brizinae, Calothecinae, Phalaridinae, Scolochloinae and Torreyochloinae) of the maximum likelihood phylogram inferred from ITS+ETS data. ML bootstrap support (left) and BI poster probabilities (right) are recorded along branches. A dash indicates bootstrap support <50%. No support is shown for branches with bootstrap support <50% and posterior probability <.5. The shaded area of the smaller tree on the left indicates the location in the overall tree of the portion shown. One indel in ETS is mapped onto the phylogram.

The 3’-end of the ETS region sampled here includes relatively conserved 5’- and 3’-ends and more rapidly evolving middle regions. There are several unambiguous indels in the ETS alignment (Appendix 1, Suppl. material 1), including a 63–67 bp insertion present in Polypogon elongatus and all accessions of Agrostis except A. exarata, A. capillaris and A. gigantea (Fig. 7), and a 12 bp deletion in all taxa of Aveninae s.str. and Koeleriinae (Fig. 5). Presence of the latter indel in Helictotrichon, however, is unclear because the 12 bp deletion overlaps with a 26 bp insertion present in the two Helictotrichon samples. An 81 bp insertion is present in Gaudinia fragilis and Rostraria cristata (L.) Tzvelev (Fig. 5). Species of Anthoxanthum share an 86–192 bp insertion (Fig. 9). All taxa of the Agrostidinae + Brizinae + Calothecinae clade share a 48–56 bp insertion (73 bp in the alignment) (Fig. 8). The clade is also defined by a 107–159 bp deletion (excluding the outgroup from the alignment, and 190 bp in the alignment, including gapped sites) (Fig. 8). Including B. vulgaris, the indel is 472 bp (426 bp in B. vulgaris excluding gapped sites). Each of these latter two indels includes additional substructure we have not attempted to describe.

Figure 10. 

A portion (part of Agrostidinae p.p., Brizinae p.p., Calothecinae p.p. and Torreyochloinae) of the maximum likelihood phylogram inferred from ITS data. ML bootstrap support is recorded along branches when >50%. The shaded area of the smaller tree on the left indicates the location in the overall tree of the portion shown. Backbone branches represented by ellipses are shown only in Figure 2.

Aveninae s.l. is a subtribe of annual and perennial grasses with lemmas awnless, mucronate or with an abaxial awn, awns geniculate and hila short or linear (Kellogg 2015). Aveninae s.l. is not monophyletic in nrDNA trees because Sesleriinae is nested within it. Support for the Aveninae s.l. + Sesleriinae clade is stronger in the ITS+ETS tree than in the ITS tree, and is further supported by a 12 bp deletion in the ETS alignment (data are missing for Sesleria in this part of the ETS matrix). In the plastid tree, Aveninae s.l. is monophyletic and Sesleriinae is part of Poeae chloroplast group 2. Support for the Aveninae s.l. clade in the combined plastid tree is higher than in most trees based on fewer plastid regions (e.g., Quintanar et al. 2007, Schneider et al. 2009, Saarela et al. 2010).

Figure 14. 

A portion (Aveninae s.str., Sesleriinae) of the maximum likelihood phylogram inferred from ITS data. ML bootstrap support is recorded along branches when >50%. The shaded area of the smaller tree on the upper left indicates the location in the overall tree of the portion shown.

Aveninae s.l. is divided into two subclades. In ITS+ETS and plastid trees, one subclade includes Arrhenatherum, Avena and Helictotrichon s.str. (Aveninae s.str.); in the nrDNA trees, Sesleria (Sesleriinae) is part of this lineage. In the ITS tree, the equivalent subclade includes Arrhenatherum, Avena, Helictotrichon and Tricholemma (not sampled in the ITS+ETS tree). Within the clade in the ITS tree, Arrhenatherum, Avena and Helictotrichon s.str. are densely sampled, each genus is resolved as monophyletic, Arrhenatherum and Tricholemma are sister taxa, and Arrhenatherum + Tricholemma and Helictotrichon are sister groups, as in other studies (e.g., Quintanar et al. 2007, Wölk and Röser 2014). In the ITS+ETS tree, however, Avena and Helictotrichon are sister taxa. This different placement for Helictotrichon compared to the ITS tree may be due to the poorer taxon sampling in the ITS+ETS tree, conflicting signal in the ITS and ETS regions, or both. In the plastid tree, Avena and Arrhenatherum form a clade, a topology consistent with an earlier combined nrDNA and plastid tree (Winterfeld et al. 2009a), but conflicting with the ITS and ITS+ETS trees. In a recent plastid tree, relationships among Avena, Helictotrichon, Arrhenatherum and ×Trisetoptrichon Röser & A. Wölk (Helictotrichon × Trisetopsis) are mostly unresolved (Wölk and Röser 2017).

The second subclade of Aveninae s.l. is recovered in the ITS+ETS and plastid trees with moderate to strong support. This clade has been recovered in previous studies based on plastid and nuclear data (Saarela et al. 2010, Wölk and Röser 2017), and corresponds to Koeleriinae (type: Koeleria) sensuQuintanar et al. (2007, 2010). All taxa in this clade share an eight bp deletion in ITS 1, an informative indel first noted by Quintanar et al. (2007) in a subset of the taxa sampled here. Because a considerable portion of our results focus on the latter subclade, and because being able to refer to this lineage with a rank-based scientific name rather than an informal one facilitates clear communication, we accept Koeleriinae as a subtribe separate from Aveninae s.str., following Quintanar et al. (2007, 2010). Moreover, recognition of Aveninae s.str. and Koeleriinae as subtribes results in a classification consistent with both nrDNA and plastid trees. However, if Koeleriinae is included in Aveninae s.l., as in Soreng et al. (2015b), Aveninae s.l. is paraphyletic in the context of nrDNA trees given the placement of Sesleriinae in those trees.

Figure 15. 

A portion (Lagurus, Koeleriinae clade B) of the maximum likelihood phylogram inferred from ITS data. ML bootstrap support is recorded along branches when >50%. The shaded area of the smaller tree on the left indicates the location in the overall tree of the portion shown.

Relationships in Koeleriinae in the trees reported here are generally congruent with previous phylogenetic studies of the subtribe. Quintanar et al. (2007) sampled seven species of Koeleria, six of Rostraria, 14 of Trisetum and five of Trisetaria, and species of each genus were intermixed with each other and with species of Avellinia, Gaudinia and Graphephorum in ITS and plastid trees. Saarela et al. (2010) increased sampling of New World species of Koeleriinae, and identified two major clades of Koeleriinae in an ITS tree. They referred to these clades as the “Old World Trisetum Alliance” and the “New World Trisetum Alliance”, because species of Trisetum were present in each clade and most species in each clade were from either the Old or New World. These major clades are strongly supported in the current ITS+ETS and plastid trees, and we refer to them more simply as Koeleriinae clade A and Koeleriinae clade B, respectively. In Wölk and Röser (2017), Koeleriinae clade B is resolved in both ITS and plastid trees, but Koeleriinae clade A is resolved only in their ITS tree. In our analyses, Koeleriinae clade A includes Avellinia, Gaudinia, Koeleria, Rostraria, Trisetaria and Trisetum p.p. We did not sample the new genus Tzveleviochloa, with two species, recently described by Wölk and Röser (2017), which is also part of Koeleriinae clade A. Koeleriinae clade B includes many taxa of Calamagrostis/Deyeuxia sampled from Mexico, Central and South America, Leptophyllochloa, Peyritschia, Sphenopholis, Trisetum p.p., and Trisetopsis. A 247 bp deletion in the psbK-psbI intergenic spacer region is present in all taxa of Koeleriinae clade B except Deyeuxia tripilifera, and in two taxa not part of this clade, Avellinia michauxii and Calamagrostis pisinna (sister to Aveninae s.l.). The distribution of this indel indicates it may be symplesiomorphic in the clade comprising C. pisinna and Aveninae s.l. We also identify a lineage of Koeleriinae in the ITS and matK (Suppl. material 7) trees comprising two species classified in Trisetum subsect. Sibirica (Veldkamp and van der Have 1983): T. bifidum (Thunb.) Ohwi, from China, eastern Asia and Papuasia, and T. sibiricum Rupr., from Eurasia and northwestern North America. A similar lineage of these two species is identified in the ITS tree in Wölk and Röser (2017). Relationships among this lineage and Koeleriinae clades A and B are unresolved in the trees. Increased sampling of genes and taxa is needed to better resolve the placement of this lineage and to identify all the species that are part of it.

Figure 16. 

A portion (Koeleriinae clade A, part of Koeleriinae clade B, Trisetum subsect. Sibirica and Brizinae p.p.) of the maximum likelihood phylogram inferred from ITS data. ML bootstrap support is recorded along branches when >50%. The shaded area of the smaller tree on the left indicates the location in the overall tree of the portion shown.