Molecular Phylogenetics and Micromorphology of Australasian Stipeae (Poaceae, Subfamily Pooideae), and the Interrelation of Whole-Genome Duplication and Evolutionary Radiations in This Grass Tribe

The mainly Australian grass genus Austrostipa (tribe Stipeae) comprising approximately 64 species represents a remarkable example of an evolutionary radiation. To investigate aspects of diversification, macro- and micromorphological variation in this genus, we conducted molecular phylogenetic and scanning electron microscopy (SEM) analyses including representatives from most of Austrostipa’s currently accepted subgenera. Because of its taxonomic significance in Stipeae, we studied the lemma epidermal pattern (LEP) in 34 representatives of Austrostipa. Plastid DNA variation within Austrostipa was low and only few lineages were resolved. Nuclear ITS and Acc1 yielded comparable groupings of taxa and resolved subgenera Arbuscula, Petaurista, and Bambusina in a common clade and as monophyletic. In most of the Austrostipa species studied, the LEP was relatively uniform (typical maize-like), but six species had a modified cellular structure. The species representing subgenera Lobatae, Petaurista, Bambusina as well as A. muelleri from subg. Tuberculatae were well-separated from all the other species included in the analysis. We suggest recognizing nine subgenera in Austrostipa (with number of species): Arbuscula (4), Aulax (2), Austrostipa (36), Bambusina (2), Falcatae (10), Lobatae (5), Longiaristatae (2), Petaurista (2) and the new subgenus Paucispiculatae (1) encompassing A. muelleri. Two paralogous sequence copies of Acc1, forming two distinct clades, were found in polyploid Austrostipa and Anemanthele. We found analogous patterns for our samples of Stipa s.str. with their Acc1 clades strongly separated from those of Austrostipa and Anemanthele. This underlines a previous hypothesis of Tzvelev (1977) that most extant Stipeae are of hybrid origin. We also prepared an up-to-date survey and reviewed the chromosome number variation for our molecularly studied taxa and the whole tribe Stipeae. The chromosome base number patterns as well as dysploidy and whole-genome duplication events were interpreted in a phylogenetic framework. The rather coherent picture of chromosome number variation underlines the enormous phylogenetic and evolutionary significance of this frequently ignored character.

The tribe Stipeae includes approximately 530 species in 33 genera and has almost worldwide distribution (Peterson et al., 2019) . It has major centers of radiation in Eurasia (especially in Stipa L. s.str., Piptatherum P.Beauv., A chnatherum P.Beauv.) and the Americas [in E riocom a Nutt., Nassella (Trin. Austrostipa is the third largest genus in the Stipeae and encompasses approximately 64 species, most o f which occur in Australia and Tasmania (Vickery et al., 1986;Jacobs and Everett, 1996;Everett et al., 2009;Williams, 2011). Only one species, A . stipoides (Hook.f.) S.W.L.Jacobs & J.Everett, is considered native to New Zealand; it is also present in southeastern Australia. Other species of Australian origin are naturalized in New Zealand (Jacobs et al., 1989;Edgar and Connor, 2000) . One species, A . scabra (Lindl.) S.W.L.Jacobs & J.Everett, is present on Easter Island; it is thought to have been introduced there after 1860 (Everett and Jacobs, 1990) . Ecologically, Austrostipa is adapted to the warm-to hot-summer Mediterranean climate of SW , S and SE Australia and the more Oceanic climate of SE Australia and Tasmania. This rather isolated southern outpost of Stipeae is widely separated from temperate Eurasia and the Americas, both regions with high diversity in Stipeae genera. Austrostipa displays a tremendous morphological diversity in, for example, habit, growth form, the size and form o f individual structures such as spikelets, glumes, etc. Moreover, the rich evolutionary diversification developed sympatrically, primarily in a comparatively narrow coastal strip of S Australia. This region has Mediterranean-type to steppe-like climate with open vegetation, similar to characterizing other areas of stipoids diversity. Many Austrostipa species seem to be edaphically specialized, being restricted to specific soil types (Everett et al., 2009;Williams, 2011) .
Characters of the lemma, visible even under low magnification, are taxonomically important in Poaceae and frequently used in species identification. The taxonomic value of micromorphological characters of the lemma epidermis is also substantial in many genera of grasses (for example, Thomasson, 1978Thomasson, , 1981Thomasson, , 1986Terrell and Wergin, 1981;Barkworth and Everett, 1987;Valdes-Reyna and Hatch, 1991;Snow, 1996;Acedo and Llamas, 2001;Terrell et al., 2001;Mejia Saules and Bisby, 2003;Ortunez and de la Fuente, 2010;Nobis, 2013) . Thomasson (1978Thomasson ( , 1981 was the first to use lemma epidermal characters in the Stipeae, demonstrating the value of such features as the presence of hooks, the shape o f the long cells and the presence of silica cells in elucidating the phylogeny of the tribe. More recently, Romaschenko et al. (2010Romaschenko et al. ( , 2012 have described two major lemma epidermal patterns in the tribe: Stipa-like, also called saw-like, dominated by long fundamental cells and hooks, and A chn atheru m -like, also called maize-like, dominated by short fundamental cells and paired with silica cells. Several authors have shown out that, even though LEP is relatively uniform within a genus, it may still be useful in identifying particular species as well as in delineating relationships among and between different subgenera or sections (Ortunez and de la Fuente, 2010;Nobis, 2013;Olonova et al., 2016;Nobis et al., 2019b), but lemmas of relatively few Austrostipa species had been studied prior to Bustam's (2010;2012) work.
Most research (Barkworth et al., 2008;Hamasha et al., 2012;Romaschenko et al., 2012) supports Jacobs and Everett (1996) in recognizing Austrostipa as separate from, and only distantly related to, Stipa s.str. Morphologically, Austrostipa has several floret characteristics (e.g., long, sharp calluses, lemmas are often dark and have tough margins, glabrous and prow-tipped paleas) that, although not individually unique to the genus, in combination distinguish it from other genera, including the rather similar and poorly understood genus A chnatherum (Jacobs and Everett, 1996). The closest extant relatives of Austrostipa within the Stipeae, however, have not yet been unequivocally identified. Analyses ofmorphological and anatomical data placed A ustrostipa in a clade together with A chnatherum and Ptilagrostis Griseb. (Jacobs and Everett, 1996;Jacobs et al., 2000, Figure 2 ). Previous molecular phylogenetic studies showed Austrostipa forming a clade together with the main part of A chnatherum , the American genera N assella, Jarav a, and several smaller genera (for example, A m elichloa Arriaga & Barkworth, Celtica F.M. Vazquez & Barkworth, Stipellula Röser & Hamasha), which represented one of the well supported major lineages within the tribe (Barkworth et al., 2008;Romaschenko et al., 2008Romaschenko et al., , 2010Romaschenko et al., , 2012Cialdella et al., 2010;Hamasha et al., 2012). Most studies sampled only one or a few species of A ustrostipa, making it hard to assess the monophyly o f this genus. One species of A ustrostipa was sampled for the internal transcribed spacer (ITS) regions of nrDNA by Hsiao et al. (1999), 13 for ITS and five plastid DNA regions by Romaschenko et al. (2008Romaschenko et al. ( , 2010Romaschenko et al. ( , 2012, six for ITS1 and seven for four plastid DNA regions by Barkworth et al. (2008), two for four plastid DNA regions by Cialdella et al. (2010), five for ITS and two for one plastid DNA region by Hamasha et al. (2012) as well as 25 for ITS and one plastid DNA region by Winterfeld et al. (2015). The ITS studies of Jacobs et al. (2000Jacobs et al. ( , 2007 encompassed 15 and 37 species, respectively. While monophyly of Austrostipa was supported by the former study, sequences o f some species o f A chn atherum , N assella, and Stipa were interspersed in the Austrostipa clade of the latter. In both studies, the New Zealand endemic A n em anthele was included in the A ustrostipa clade, but its position was unstable. The most comprehensive study conducted so far included 31 taxa for ITS and 52 for two plastid DNA regions (Syme et al., 2012) .
Overall variation between individual A ustrostipa ITS sequences was low and the differences between sequences from different accessions of the same species was often not much smaller than between sequences of different species (Jacobs et al., 2007) . This overall low variation made it difficult to compare their results with classification o f Austrostipa into 13 subgenera (Table 1; Jacobs and Everett, 1996;Everett et al., 2009) . The main characters employed were growth form, branching of the culms, characters of the spikelets (glumes, lemmas, awns, paleas) and the formation of dispersal units (whole panicle or florets). Some of the subgenera were reflected in the ITS data (for example, subg. Falcatae S.W.L.Jacobs & J.Everett), whereas others were mixed up (for example, subg. Austrostipa and subg. Tuberculatae S.W.L.Jacobs & J.Everett or subg. Arbuscula S.W.L.Jacobs & J.Everett and subg. B am bu sin a S.W.L.Jacobs & J.Everett, respectively), or were entirely unresolved (Jacobs et al., 2007;Syme et al., 2012) . The plastid DNA analyses resolved two main clades, neither of which corresponded to the recognized subgenera, and further resolution was low (Syme et al., 2012). By using a combination of morphological and molecular approaches, this study addresses the main phylogenetic and evolutionary problems regarding Austrostipa, namely its monophyly and its internal phylogenetic structure. The num b e r o f species sam pled relative to the total n um b e r o f species (in brackets) according to Ja cob s and Everett (1996), Everett e t al. (2009), and Williams (2011) is show n fo r the three DNA regions analyzed.
These questions are treated using on a broader sample of Austrostipa taxa by generating a taxonomically overlapping set of nr ITS and plastid DNA sequences of the 3'trnK region. Both molecular markers are frequently utilized and well-established in molecular phylogenetic studies (Baldwin et al., 1995;Liang and Hilu, 1996), although ITS from the repetitive 18S-26S nrDNA can be polymorphic in individual genomes for several reasons. This may lead to paralogous sequence relationships that can potentially confound phylogenetic reconstruction (Buckler et al., 1997;Alvarez and Wendel, 2003;Bailey et al., 2003;Razafimandimbison et al., 2004;Bayly and Ladiges, 2007;Nieto Feliner and Rosselló, 2007;Schneider et al., 2009Schneider et al., , 2011. Nonetheless, ITS is a nuclear marker useful to investigate. As a second nuclear marker we studied the single-copy gene Acc1 encoding plastid acetyl-CoA carboxylase 1 (Huang et al., 2002;Fan et al., 2007Fan et al., , 2009Sha et al., 2010;Hochbach et al., 2015) . The 3'trnK region, comprising the 3'part o f the chloroplast m atK gene with following intron and 3'trnK exon, was selected as sequence marker from the plastid DNA mainly because o f its comparatively high substitution rate. Moreover, these sequences are straightforward to align and are already available in many potential outgroup taxa from within Stipeae and neighboring tribes (Döring et al., 2007;Schneider et al., 2009Schneider et al., , 2011Schneider et al., , 2012Hamasha et al., 2012;Blaner et al., 2014;Röser, 2014, 2017;Hochbach et al., 2015Hochbach et al., , 2018Tkach et al., 2020). The sequence data from the nuclear and the plastid genome are used to examine the potential role of hybridization, reticulation and the origin of polyploidy in Austrostipa. The results of the phylogenetic analyses were further used to discuss the cytogenetic characteristics of this genus and other stipoids. To this end, we conducted an up-to-date survey of chromosome numbers in the Stipeae and discussed the chromosome base number(s), dysploid variation and the evolutionary role of wholegenome duplications in this tribe.

Plant Material
The sample for the molecular phylogenetic study included 51 species and subspecies of A ustrostipa. Geographic origin, collector or seed exchange locality and herbarium vouchers for the taxa used in this study are listed in Supplementary Appendix 1. For half of the species more than one specimen was included. Sampling density among the 13 subgenera for the analyzed DNA regions 3'trnK (3'part of the chloroplast m atK gene with the following intron and 3'trnK exon), ITS and A cc1 is summarized in Table 1. The dataset of the 3'trnK region encompassed 43, that o f ITS 48 A ustrostipa species, representing 71 and 75% o f the total species number, respectively. All subgenera except for subg. Lan tern a S.W.L.Jacobs & J.Everett were represented by at least half their species. For the analysis of the nuclear single-copy gene Acc1 sequence data we selected at least one specimen of each subgenus and studied a total of 22 (33%) species. In addition, we included representatives of eight other stipoid genera [A chnatherum , A nem anthele, Celtica, Nassella, N eotrinia (Tzvelev) M.Nobis, P.D. Gudkova & A.Nowak, Oloptum Röser & Hamasha, Stipa, Stipellula] that previous studies have shown to be most closely related to Austrostipa (Jacobs et al., 2000(Jacobs et al., , 2007Barkworth et al., 2008;Romaschenko et al., 2008Romaschenko et al., , 2010Romaschenko et al., , 2012Hamasha et al., 2012) . Genera from the tribes Bromeae (Brom us L.), Duthieeae (Anisopogon R.Br.) and Triticeae (H en rardia C.E.Hubb., H ordeu m , Secale L.) were chosen as outgroups for phylogenetic reconstructions based on studies of phylogenetic relationships within subf. Pooideae (for example, Catalan et al., 1997;Hilu et al., 1999;Mathews et al., 2000;Soreng and Davis, 2000;GPWG (Grass Phylogeny Working Group), 2001;Davis and Soreng, 2007;Döring et al., 2007;Soreng et al., 2007;Schneider et al., 2009Schneider et al., , 2011Saarela et al., 2015Saarela et al., , 2018 . For the Acc1 dataset, data for selected outgroup species of Bromeae (Brom us inermis Leyss.) and Triticeae (H enrardia persica (Boiss.) C.E.Hubb., H ordeu m chilense Roem. & Schult., H . vulgare L.) as well as some 3'trnK and ITS sequences were taken from ENA/GenBank (Supplem entary Appendix 1).
Most plant material used in this study for DNA extraction was collected in the field in 2007 by SWLJ and Mary E. Barkworth (Logan, UT, United States) along with herbarium specimens and duplicates, which have been subsequently distributed to various herbaria (Supplem entary Appendix 1). The leaf samples were preserved in saturated NaCl/CTAB buffer solution prepared according to Storchova et al. (2000). Further leaf material for DNA extraction was collected from living pot plants grown from seeds stored at the Millennium Seed Bank (Wakehurst Place, Royal Botanic Gardens, Kew, United Kingdom). These caryopses were collected from natural populations with verified identifications and voucher specimens deposited at K and, in some instances, at PERTH (Supplementary Appendix 1). The pot plants were cultivated in the greenhouses o f the Botanical Garden of the University Halle-Wittenberg (vouchers at HAL). Leaves for DNA extractions were silica gel-dried (Chase and Hills, 1991) . These living plants were also used for cytogenetic studies by Winterfeld et al. (2015).

DNA Extraction, PCR Amplification and Sequencing
For DNA extraction, leaves preserved in NaCl/CTAB buffer were removed from the solution, rinsed in water, immersed in liquid nitrogen, and then ground to fine powder using mortar and pestle. Silica gel-dried fresh leaves were shredded in a FastPrep FP 120 bead mill homogenizer (Qbiogene, Heidelberg, Germany). The ready-to-use NucleoSpin Plant Kit (Macherey-Nagel, Düren, Germany) was used for extraction.
The ITS and 3'trnK region were amplified and sequenced as in our previous studies with primers listed in Table 2 (Schneider et al., 2009(Schneider et al., , 2011(Schneider et al., , 2012Hamasha et al., 2012;Winterfeld et al., 2015) . The amplification of Acc1 (exons 6 -1 3 and intervening introns) was carried out using primers also listed in Table 2.
An overview o f the gene Acc1 is shown in Figure 1 together with the locations, directions and designations of the primers used in this study.
For DNA samples, which were obtained by extraction from leaves preserved in saturated NaCl/CTAB buffer solution, the PCR reaction was performed with 3 min at 94°C, followed by 35 cycles of 30 s at 94°C, 30 s-2 min at 50°C, 5 min at 68°C, and a final extension for 20 min at 68°C.
The DNA extracted from silica gel-dried leaf material was amplified by the following PCR program: 3 min at 94°C, followed by 35 cycles o f 30 s at 94°C, 30 s-2 min at 50°C, 2 min at 72°C, and a final extension for 20 min at 72°C. PCR products of Acc1 were column-purified with the NucleoSpin Extract II Kit (Macherey-N agel).
Due to the presence of different Acc1 copies in Austrostipa species, A n em an thele lessoniana (Steud.) Veldkamp and other polyploids (Supplementary Appendix 2;Winterfeld et al., 2015), Acc1 amplicons were cloned into the pGEM -T Easy Vector (Promega, Mannheim, Germany) according to the manufacturer's protocol. In the next step 10-30 individual white colonies containing the insert were picked. The isolation of plasmid DNA was performed with the Wizard Plus SV Minipreps DNA Purification System (Promega). The insert of the purified plasmid DNA was sequenced using the standard primers T 7 and SP6. The sequencing was performed by StarSEQ GmbH (Mainz, Germany) or Eurofins M W G Operon (Ebersberg, Germany).
TA BLE 2 | Primers used to amplify and sequence the plastid 3 'trn K region, nuclear ITS1-5.8S gene-ITS2 and the A cc1 gene (exons 6 -13 and intervening introns). D N A re g io n a n d p rim e r n a m e 5 ' -P r im e r s e q u e n c e -3 ' R e fe re n c e s 3 'tr n K re g io n  W e identified few double peaks in chromatograms of the ITS dataset already documented in our previous study (Winterfeld et al., 2015) . It was possible to edit these single nucleotide positions by IUPAC code and include all obtained ITS sequences. All clone-derived sequences of the Acc1 dataset were visually checked for the presence o f chimerical sequences or PCR artifacts (see Brassac et al., 2012) . Furthermore, we tested the protein sequence of the exon regions (696 bp) for each clone and compared the translation to the Acc1 sequence o f the diploid outgroup, B rom us inerm is, taken from ENA/GenBank (Supplem entary Appendix 1). We excluded chimerical sequences and clones different from that of Brom us inerm is in more than 20 amino acid positions of the exon regions. To reduce the number of singletons in the alignment, we summarized for each specimen highly similar Acc1 sequences of the remaining individual clones to consensus sequences. Sequences of ITS and 3'trnK region as well as the individual A cc1 clones used for assembling consensus sequences were submitted to ENA/GenBank under the accession numbers LR989057-LR 989267 (Supplem entary Appendix 1).
All DNA sequence datasets were analyzed using the phylogenetic approaches of maximum likelihood (ML), maximum parsimony (M P), and Bayesian inference (BI) following Tkach et al. (2019Tkach et al. ( , 2020. The trees were visualized with FigTree v.1.4.32. Support values are cited in the text in the following sequence: ML bootstrap support/MP bootstrap support/Bayesian posterior probability (PP).

Morphological Analyses
W e scored 65 species and subspecies o f Austrostipa for nine morphological characters commonly used in identification keys in 1 -5 specimens each or gathered the information from morphological descriptions of the taxa (Everett et al., 2009) . The characters studied were: mean length of ligules of the culm leaves; surface of inflorescence branches (glabrous, with prickles or 2http://tree.bio.ed.ac.uk/softw are/figtree/ Anisopogon avenaceus (Duthieeae) used as outgroup. Reduced datasets with each taxon represented by a single representative accession. M L and MP bootstrap support values > 50% as well as Bayesian PP > 0.5 are indicated on the branches. Clades w ith M L support < 50% are collapsed. The asterisked clade in the ITS tree w as recovered also in the Acc1 gene tree of F ig u re 3 (copy types A and B) as well as the clade with diam ond (copy type A). The taxonom ic groupings of the Austrostipa species according to different treatm ents are marked by different colors or num bers in colum ns 1-3. The LEPs studied are represented in colum n 4. N umbers in brackets after taxon names refer to individual accessions as listed in the S u p p le m e n ta ry A p p e n d ix 1. A., Austrostipa. macrohairs); mean length o f glumes, calli, lemmas, lemma lobes and awns; mean length ratio lemma:palea; shape of palea apex (without or with 2 -4 teeth) ( Table 3, Supplementary Table 1, and Supplementary Appendix 1) . These characters were chosen to evaluate morphological groupings of A ustrostipa taxa proposed by Jacobs and Everett (1996) and Everett et al. (2009).

Lemma Micromorphology
The ultrastructure of the lemma epidermis was studied in 34 taxa (species and subspecies) o f Austrostipa (Supplementary Appendix 1) . For scanning electron microscopy (SEM), dry samples were coated with a thin layer of gold using a JFC-1100E ion sputter (JEOL), then observed and photographed on a Hitachi S-4700 scanning electron microscope. Four diagnostic micromorphological characters, namely fundamental cells, silica cells, cork cells and hooks were recorded. W e examined the middle part of the abaxial lemma surface as being the least variable. It differs from the upper part, in which a variable admixture of hooks, prickles and macrohairs is usually observed.

Numerical Analyses
The numerical analyses were performed on the same 34-taxa set based on (1) four above-mentioned micromorphological characters, and (2) a combination of four micromorphological with eight macromorphological characters ( Table 3 and  Supplementary Table 2 ) . Each taxon was treated as an Operational Taxonomic Unit (O TU ), in accordance with the methods used in numerical taxonomy (Sokal and Sneath, 1963). The similarities among OTUs were calculated using Gower's general similarity coefficient. Cluster analysis, using PAST software (Hammer et al., 2001), was performed on all OTUs to estimate morphological similarities among the species.

Chromosome Numbers in Stipeae
To address the significance of the cytogenetic data in Austrostipa and relatives in a phylogenetic context of Stipeae, we extensively surveyed the published chromosome numbers. We prepared a comprehensive up-to-date list of chromosome numbers of Stipeae taxa (164 species, 22 infraspecific taxa) with currently accepted taxon names and painstakingly regarded nomenclature and synonyms used in the original publications (Supplementary Appendix 2) . Because the secondary literature frequently had reported incorrect numbers or wrongly cited the actual authors of the chromosome counts, we checked more than 150 original references. A few original publications we could not examine are identified as such in the references list of Supplementary Appendix 2 .
To infer the evolutionary history of chromosome and genomic characters o f special interest utilizing a simple cladistics analysis, we mapped chromosome base numbers, occurrence of dysploidy and whole-genome duplications in the evolution of tribe Stipeae on a molecular phylogenetic cladogram simplified and modified from the plastid DNA tree of Romaschenko et al. (2012)

Molecular Phylogenetics
W e analyzed a dataset of 110 DNA sequences for the 3'trnK region and 111 for ITS, respectively. The Acc1 dataset comprised a total of 266 clone-derived sequences. After evaluation of all clones of the polyploid genera Austrostipa and A n em an thele, we created 61 consensus sequences for the final dataset. We obtained two or three distinct Acc1 consensus sequences for each Austrostipa species with the exception of A . breviglum is (J.M.Black) S.W.L.Jacobs & J.Everett, which had only one consensus sequence. For tetraploid Stipa capillata L. and S. tirsa Steven (both 2n = 44; Supplementary Appendix 2 ) we identified two different Acc1 copy types after analyzing the clone sequences. The topology of the trees inferred by ML, MP, and BI analyses were largely identical although their statistical supports differed slightly. Figure 2 shows trees with plastid and nuclear ITS DNA data reduced to a single accession per taxon. The complete phylograms with all studied accessions are presented in Supplementary Figures 1, 2 .

Plastid DNA Analysis -3'trnK Region
The plastid 3'trnK region DNA sequence dataset (sequence lengths 579-798 bp) for 63 taxa of the reduced dataset (each species or subspecies represented by only one accession) included TA BLE 3 | M orphological characters and character states. w ith M L support < 50% are collapsed. The asterisked clade within the copy type A and B clades w as recovered also in the ITS tree of F ig u re 2 as well as the clade w ith diam ond. The taxonom ic groupings of the Austrostipa species according to Jacobs and Everett (1996)

Nuclear DNA -ITS
The reduced nr ITS DNA sequence dataset for 68 taxa (each species or subspecies represented by only one accession) included 644 aligned positions (sequence lengths 5 00-627 bp), of which 263 (41%) were variable and 185 (29%) parsimony-informative.
The phylogram of the single-copy region Acc1 with Brom us inermis, H enrardia persica, H ordeu m chilense, and H. vulgare as outgroup showed species of Stipa s.str. sister to a clade comprising all other stipoid taxa (A chnatherum , A nem anthele, Austrostipa, Nassella) ( Figure 3 ). W e identified two different copy types of Acc1 for the tetraploids S. capillata and S. tirsa (Figure 3), which resulted in the formation of two separate clades that were not sister (both 100/100/1.00). One of these Stipa copy type clades was sister to the strongly supported clade with all Acc1 copy types of A chnatherum , A nem an thele and A ustrostipa (100/100/1.00). The Acc1 copy types of A n em an thele and Austrostipa segregated into two lineages, copy type A and B in Figure 3. The diploids of A chnatherum (A. p arad ox u m , A. sibiricum ) as well as polyploid Nassella trichotom a (see Supplementary Appendix 2) had only a single copy type of Acc1. They formed a basal grade to the wellsupported copy type A Australasian clade of A nem anthele and A ustrostipa (87/74/1.00).
Copy type A clade comprised three subclades in a polytomy, namely (1) A n em an thele lessoniana, A ustrostipa geoffreyi, and A. m acalpinei (67/65/0.98), (2) A. acrociliata, A. elegantissima, A. ram osissim a, A. setacea, and A. verticillata (92/89/1.00) and (3) a clade (96/93/1.00) with all six A. scabra accessions plus A. trichophylla (100/100/1.00) and another clade (94/93/100) with some well-supported minor lineages. Copy type B clade showed A. exilis sister to a larger polytomy encompassing A nem anthele lessoniana and the remaining species of Austrostipa, organized in several minor lineages. Austrostipa breviglumis, which had only one Acc1 clone sequence, was placed in the copy type B clade. In both clades (copy type A and B), the accessions o f A. scabra and A. trichophylla as well as A. ram osissim a and A. verticillata formed supported clades, respectively. These clades, however, were differently placed in the copy type A and B clades, whose general topology was not fully corresponding. Everett, for which there are no molecular data. Information on the studied specimens is contained in Supplementary Appendix 1. In A ustrostipa, four LEPs were encountered.

Morphological Analyses Lemma Epidermal Patterns
In 28 of the 34 taxa of Austrostipa examined, the LEP was relatively uniform (Figures 2, 4 a -u , 5a-d ,f) and typical of achnatheroid grasses as seen, for example, in A chn atherum , A nem anthele, Jarava or Stipellula (Figures 5 l-o ). This maize like LEP is characterized by wider than long, short or square to rectangular fundamental cells with undulate to almost straight side walls. Silica cells were very frequent, ovate to elongate, densely packed and regularly alternating with fundamental cells. In cell; c, co rk cell; h, hook; mh, macrohair. The list of specim ens studied is presented in S u p p le m e n ta ry A p p e n d ix 1 .
the LEP was distinctively different due to the presence of numerous hooks and longer fundamental cells, here termed 'prickly maize-like' LEP, and hence reminiscent of the LEP found in Old World genera such as Stipa, N eotrinia, O rthoraphium Nees and Ptilagrostis ( Figures 5 s -y ). The LEP observed in species of subgenera Petaurista S.W.L.Jacobs & J.Everett and B am busin a is characterized by short cells with hooks alternating with square or rectangular fundamental cells, ovate silica cells sometimes paired with cork cells, which, however, are generally sparse. Hooks were frequent in A. pubescens (subg. T u bercu latae; Figure 5e) but scattered in A. cam pylachne (subg. Austrostipa); Figure 4h, however, due to their short (wider than long) fundamental cells, we classified them to maize like LEP group.
The LEP with straight-to slightly sinuously walled fundamental cells observed in A. stipoides (subg. L obatae S.W.L.Jacobs & J.Everett; Figures 2, 5k) is characterized by its rectangular to elongated fundamental cells (1 .5 -4 times as long as wide), which often alternate with silica cells and cork cells as well as sometimes scattered hooks. The elongated silica cells were often associated with cork cells and frequently had 1-4 constrictions.
A characteristic LEP with dumbbell-shaped fundamental cells alternating with elongated silica cells was encountered only in A. m uelleri (Figure 5 f).

Combined Analysis of Micro-and Marcomorphological Characters
The species representing subgenera Bam busina, L obatae, and Petaurista as well as A. m uelleri were well separated from all other species in the cluster analysis (UPGMA), which was performed on a combined macro-and micromorphological 34-taxa dataset ( Figure 6 and Supplem entary Figure 2 ).
Similar results were obtained when the micro-and macromorphological characters were analyzed separately (34-and 65-taxa set, respectively; Supplementary Figures   3, 4 and Supplem entary Tables 1, 2 ). The unique LEP of A. stipoides and presence of distinct lobes on the top of the lemma (A. ju n cifolia, A. geoffreyi, A. petraea) separated these four species of subg. L ob atae from the remaining subgenera of Austrostipa (Figure 6 and Supplementary Figures 3,  4 ). A cluster was formed by representatives of subgenera Petaurista and B am bu sin a, which had prickly maize-like LEP (Figure 6 and Supplementary Figure 3 ). The unique macromorphology of the inflorescences characterized by long pilose branches, which occurred exclusively in A. elegantissima and A. tuckeri (subg. P etaurista), resulted in a clear separation from the other subgenera of Austrostipa (Supplementary Figure 4 ). Due to its long apical lemma lobes as well as its particular LEP, A. m uelleri was well-distinguished not only from the other representatives of subg. T uberculatae but from all other studied species with achnatheroid, maize-like LEP ( Figure 6 and Supplem entary Figure 4 ). The species of the remaining subgenera of A ustrostipa with maize-like LEP were grouped in several (sub)clusters in accordance to each of the three performed analyses, with rather weakly noticeable subgeneric ordination ( Figure 6 and Supplementary  Figures 3 ,4 ) . the data matrix evaluated. The taxonom ic groupings of the Austrostipa species according to Jacobs and Everett (1996) and this study are marked by different colors in colum ns 1 and 2. A ., A u stro stip a .

Chromosome Numbers and Whole-Genome Duplications in Stipeae
The chromosome numbers of species and genera in Stipeae are listed in Supplem entary Appendix 2 (bold print). For each of the 33 genera, the most frequently found chromosome numbers are underlined, if applicable. In six genera the chromosome numbers is yet unknown (O rtachne Nees, O rthoraphium , P sam m ochloa Hitchc., T horneochloa Romasch., P.M.Peterson & Soreng, Tim ouria Roshev., T rikeraia Bor). Monoploid chromosome numbers, chromosome base numbers (x = ) and ploidy levels deduced from the chromosome counts in the Supplem entary Appendix 2 were added to Figure 7. This figure represents a simplified phylogenetic tree (cladogram) portraying the genera of Stipeae, their approximate sizes and distribution (see section "Materials and Methods" and legend to Figure 7 for further explanation).
The prevailing chromosome base number in Stipeae is x = 11, marked as orange lines in Figure 7. It occurs in Australasian Austrostipa and A nem an thele, in the genera o f New World Clade B and the Main x = 1 1 clade but not consistently in four of their genera (E riocom a, Nassella, N eotrinia, Piptatheropsis Romasch., P.M. Peterson & Soreng; hatched lines in Figure 7) and except for Oryzopsis Michx. with x = 1 2 (Figure 7 ). The number of x = 12 (blue lines) is less frequent and occurs in eight genera but not consistently in three of them (hatched lines). Interestingly, the x = 1 1 genera A ustrostipa and A n em an thele are placed in a clade with mainly x = 12. B arkw orthia Romasch., P.M. Peterson & Soreng (1 species) and Stipellula (5 species) consistently have deviant numbers of x = 10 and x = 7(?), 9, respectively (Supplem entary Appendix 2 and Figure 7 ).
In some instances, the occurrence of whole genome duplications (W GD) could be labeled on the tree (filled arrows in Figure 7; see below section "Discussion"). Due to missing chromosome number information for T horneochloa and Tim ouria and the uncertain occurrence of diploids in Nassella and E riocom a (triploids as hybrids involving putative diploids), it is impossible to attach W GD in Clade B to particular nodes (open arrows in Figure 7 ).

Molecular Phylogenetic Delineation of Austrostipa
Monophyly o f Austrostipa was not clearly supported by any of the three DNA regions we investigated (plastid 3'trnK region: Figure 2 and Supplem entary Figure 1; nr ITS region: Figure 2 and Supplementary Figure 2; the single-copy locus A ccl: Figure 3 ). The plastid DNA trees showed Austrostipa as paraphyletic. Most species (36/43) belonged either to the large clade o f 'core A ustrostipa' or to the clade comprising A. drum m ondii, A. muelleri, A. nitida, A. pilata. These two clades formed a polytomy with the three remaining species of Austrostipa (A. m acalpinei, A. ram osissim a, and A. verticillata) and three Eurasian stipoid genera (A chnatherum , Oloptum, and Stipellula). A nem anthele, however, was not part o f this polytomy, but of the next lower one which also included the two representatives o f the primarily South American genus Nassella and the western Mediterranean Celtica gigantea.
Possible explanations for the failure of the plastid data to support, even weakly, the monophyly of A ustrostipa include incomplete lineage sorting (ILS) affecting the inheritance of plastids or genetic introgression from the Eurasian species into the three A ustrostipa species in the lowest Austrostipacontaining clade (A. m acalpinei, A. ram osissim a, A. verticillata). This last seems unlikely, given the present day distribution of the species involved. Higher support for monophyly of Austrostipa (86/NA/1.00) was recorded for a set o f 13 Austrostipa species using more than 6.600 aligned plastid bp (Romaschenko et al., 2012) and A nem anthele was weakly supported sister (52/NA/0.95).
Nuclear ITS grouped A ustrostipa (all species) and A n em an thele in a single clade but support was minimal for the relationship (see reduced dataset with each species represented only by a single accession in Figure 2  A ustrostipa and A n em an thele were alike in having two different copies of the nuclear gene A ccl (copy types A and B). These resolved together in two separate clades (Figure 3 ). The copies obtained for A chnatherum p arad ox u m , A. sibiricum and Nassella trichotom a were close to the copy type A clade; those for the two species of Stipa included (S. capillata, S. tirsa) divided likewise into two copy types, both of which were outside the two Austrostipa clades (Figure 3 ).
Our results, in failing to contradict or providing only weak support for the monophyly of Austrostipa and the closer relationship of Austrostipa to A nem an thele rather than non-Australasian stipoids, basically agrees with the findings of several previous studies regardless of taxon sampling (Jacobs et al., 2000(Jacobs et al., , 2007Barkworth et al., 2008;Romaschenko et al., 2010Romaschenko et al., , 2012Syme, 2011;Syme et al., 2012;Hamasha et al., 2012).
The odd results for A ustrostipa stipoides reported in two studies (Jacobs et al., 2007;Barkworth et al., 2008), which placed the species distant from other species of the genus, were from duplicate collections (Barkworth et al., 2008, p. 725) and were not corroborated by this study, in which three different collections were used (see Supplementary Appendix 1 ). Their plastid and nuclear DNA sequences clustered with those of other Austrostipa species (Figures 2, 3 and Supplementary Figures 1, 2 ), as did sequences from the specimens of A. stipoides studied by Syme (2011) and Syme et al. (2012).

Phylogenetic Differentiation in Austrostipa and Taxonomy
All but one of the subgenera of Austrostipa were represented by two representatives for at least one of the sequences we examined ( Table 1). The exception was subg. Lanterna, which means we cannot comment on its monophyly.

Comparison of Plastid and nr ITS Tree
The plastid and nr ITS trees showed slightly different placements of subgenera B am bu sin a and Longiaristatae S.W.L.Jacobs & J.Everett. Austrostipa ram osissim a and A. verticillata (subg. B am bu sina) and A. m acalpinei (subg. Longiaristatae) were placed in the plastid DNA tree in a polytomy with the remainder of Austrostipa and other genera of Stipeae (Achnatherum, A nem anthele, Oloptum, and Stipellula). This was not reflected in the ITS tree, where all studied Austrostipa subgenera were resolved in a weakly supported clade together with A n em an thele. Both subgenera (Bam busina, L on giaristatae) resolved as monophyletic considering also A. com pressa (subg. Longiaristatae), which was sampled only for ITS DNA. Sampling more DNA regions could improve overall resolution of the plastid DNA phylogenetic tree.
A small clade o f A ustrostipa species in the plastid DNA tree comprised species o f two different subgenera, namely three species o f the large subgenus Falcatae and A. m uelleri o f subg. Tuberculatae (see also below). Both subgenera were represented also in the 'core A ustrostipa' clade of the plastid DNA tree with the remaining Austrostipa species (Figure 2 and Supplementary Figure 1). The subgenus Falcatae, however, was resolved in the ITS tree as monophyletic, whereas T uberculatae were highly polyphyletic. In other words, subg. Falcatae is characterized by remarkable cytonuclear discordance, having at least two different chloroplast 'types.' Ancient polymorphism, hybridization and introgression may be its potential causes of such discordance as encountered in many groups of angio-and gymnosperms (Rieseberg and Soltis, 1991;Seehausen, 2004;Folk et al., 2017;Kawabe et al., 2018;Tkach et al., 2020).
The weakly supported clade marked by a diamond in the ITS tree of Figure 2 united species of three subgenera resolved as monophyletic: B am bu sin a, Arbuscula, and Petaurista. This diamond-marked clade, however, was not recovered in the plastid DNA tree, and altogether three different plastid types occurred in this instance.

Single-Copy Locus A c c l
The sequences analyses of the A c c l, a gene represented by a single copy per monoploid genome (see section "Results"), corroborated monophyly of subgenera B am bu sin a and Falcatae, whereas subgenera A rbuscula, A ustrostipa, L an cea, and L obatae were non-monophyletic (Figure 3 ). Within copy type A clade, the asterisked clade supported by 94/93/1.00 (Figure 3) comprised species belonging to subgenera A ustrostipa, Ceres, E rem ophilae, L a n cea , L an tern a, L obatae, and T uberculatae. This clade was largely reflected also in the copy type B topology (asterisked; 97/92/1.00). A ustrostipa exilis (accession shown to be tetraploid with 2n = 44; Supplementary Appendix 2; Winterfeld et al., 2015) and A. hem ipogon (accession shown to be hexaploid with 2 n = 66; Supplementary Appendix 2; Winterfeld et al., 2015) have additional copies of A ccl gene copy type B. That for A. exilis was placed external to all other Australasian stipoids in the tree (Figure 3 ). The asterisked clades in A ccl copy type A and B clades corresponded well with the asterisked clade supported by 80/71/1.00 in the ITS tree (Figure 2 ), thus there is consistent phylogenetic signal in both nuclear markers studied.
Polyploidy, whether allo-or autopolyploidy, is difficult to recognize in Stipeae by ITS analysis. The occurrence of different A ccl copies (labeled as A, B, C in Figure 3) belonging to two copy types in the specimens of A nem an thele (4x), Austrostipa (4x -6 x ), and Stipa (4x ) suggests consistent allopolyploidy of these genera. The presence of more than two A ccl copies in some tetraploids (e.g., Austrostipa exilis, A. oligostachya) rests presumably on duplicated gene loci. The data on the different gene copies provides molecular evidence o f allopolyploidy in the mentioned genera of Stipeae. Allopolyploidy as suggested by sequence analyses of a nuclear gene (At103) has also been reported in the East Asian/North American stipoid genus Patis Ohwi (Romaschenko et al., 2014) .

Phylogenetic Utility of Micromorphological Traits
In 32 of the 34 micromorphologically studied Austrostipa taxa, twelve o f which were investigated for the first time for lemma epidermal characters, the LEP was maize-like, typical for achnatheroid grasses, with dominance of silica cells and with fundamental cells shorter, as long as wide up to 2 -3 times longer than wide, in A. densiflora even (1 -)2 -4 times longer than wide (see also Bustam, 2012, Figure A 2-7). The prevalence of this maize-like LEP corroborates the previous results for Australian feathergrasses (Barkworth and Everett, 1987;Romaschenko et al., 2010Romaschenko et al., , 2012Bustam, 2012) .
A ustrostipa ram osissim a and A. verticillata (subg. B am busina) as well as A. elegantissim a and A. tuckeri (subg. Petaurista) have a large number of conspicuous hooks in the middle part o f lemma in addition to rectangular fundamental cells and rounded silica cells associated with cork cells (prickly maize-like LEP), which was not seen in the other A ustrostipa taxa characterized by typical maize-like LEP. In the upper part of the lemma, however, most A ustrostipa taxa have a mixture of hooks alternating with shorter to equal, rarely somewhat longer than wide fundamental cells in addition to prickles, bicellular hairs and macrohairs.
These four species together with A. acrociliata, A. breviglumis, and A. platychaeta (subg. A rbuscula) were placed by Barkworth and Everett (1987) in their taxonomic group 2 o f the Australian Stipeae (Figure 2 ), considering for classifications not only LEPs but also a set of macromorphological characters. This group 2 is reflected by the diamond-marked clade in our nr ITS tree (Figure 2 ). Based on extremely long fundamental cells, Barkworth and Everett (1987) distinguished their group 5 including two species A. setacea and A. feresetacea. This group (subg. A u lax) was corroborated as monophyletic based on the ITS data ( Figure 2) but was not resolved by the cluster analyses using morphological characters (Supplem entary Figure 4 ). According to Barkworth and Everett (1987), A. setacea and A. feresetacea should have fundamental cells 3 -4 times longer than silica cells, however, they were shorter in our studied specimens of A. setacea (only 1-3 times longer), as depicted also in Bustam (2012, Figures A 2-28). Unfortunately, A. feresetacea was not available for this study.
The LEPs o f A. stipoides and A. m uelleri were strikingly different from that of all other Austrostipa taxa. Having rather long fundamental cells with elongated silica cells associated with cork cells, the LEP of A. stipoides (SWF) was somewhat more similar to Ptilagrostis than to the other examined A ustrostipa taxa. The overall appearance resembles the saw-like LEP but the side walls of the fundamental cells were straight to slightly sinuate, not deeply sinuous as in the typical saw-like LEP (Barkworth and Everett, 1987;Romaschenko et al., 2010Romaschenko et al., , 2012Nobis et al., 2019b,c;Figures 5r,v,w,x). A ustrostipa stipoides was the only representative of subgenus L o ba ta e we studied for LEP. Two further species (A. geoffreyi and A. ju n cifolia) were studied by Bustam (2012), and their fundamental cells also seem to be two or more times longer than wide. However, the details of the lemma epidermis are hardly discernible on the photographs presented in this publication.
Austrostipa muelleri, characterized by a unique LEP with dumbbell-shaped fundamental cells and elongated silica cells, is the only species of traditional subgenus Tuberculatae with distinct apical lobes on the lemma apex, otherwise found only in subgenus L o b a ta e. This segregation seems to fit the placement of A. m uelleri distant to remainder of the subgenus Tuberculatae in the phylogenetic trees (Figure 2; see below).

Delineation and Relationship of Subgenera
Despite limited resolution achieved by the sequenced plastid and nr DNA loci as well as the combined macro-and micromorphological analysis, some conclusions can be drawn with respect to the infrageneric taxonomy o f Austrostipa and the validity of the altogether 13 subgenera presented in Vickery et al. (1986), Jacobs and Everett (1996), and Everett et al. (2009), all of which were included in this study.
(1) The small subgenera Longiaristatae (both species sampled, plastid DNA data missing for A. com pressa) and B am busina (both species sampled) belong to the early branching lineages within A ustrostipa considering the plastid DNA tree. Subg. B am bu sin a assembled together with subgenera Petaurista (both species sampled) and A rbuscula (three of four species sampled) in the same ITS and in copy type A clades of the A ccl gene analyses marked by diamonds (Figures 2, 3 ). Petaurista and A rbuscula were placed in the 'core A ustrostipa' clade of the plastid DNA tree distantly to the species of subg. B am busina. Maintenance of subgenera Petaurista and A rbuscula is neither explicitly supported nor contradicted by our data. Thus, we argue that these four subgenera should remain unchanged.
(2) A ustrostipa m uelleri was placed distantly from all other taxa of subg. Tuberculatae (see below), in which it was accommodated (Jacobs and Everett, 1996;Everett et al., 2009) . This deviating position was noted already previously (Jacobs et al., 2007, Figure 4;Syme et al., 2012) . W e propose placing A. m uelleri by itself in a new subgenus (see below New names and combinations).
(3) Subg. Falcatae (9 o f 10 species sampled, plastid DNA data missing for A. pycnostachya and A. tenuifolia) was supported because o f the ITS and A ccl DNA data (Copies A and B) but it disintegrated into two lineages of the plastid DNA phylogeny. One group of species possessed the 'core A ustrostipa' plastid, the other shared a deviant plastid type with A. m uelleri (Figure 2 and Supplem entary Figure 1). The placement of A. pycnostachya in the ITS clade o f subg. Falcatae (Figure 2) supports the transfer of this species from subg. A rbuscula, in which it was placed by Jacobs and Everett (1996), to subg. Falcatae (Everett et al., 2009) . (4) Subg. A u lax (both species sampled, plastid DNA data missing for A. feresetacea) and subg. L obatae (4 of 6 species sampled) could be maintained after excluding A. petraea from the latter (Figure 2 ). Segregation of A. petraea from the other species of subg. L obatae was noted also by Syme et al. (2012). W e found no support, however, for a placement of this species in subg. A ulax as suggested by the latter study (see Syme et al., 2012, Figure 1) but the taxonomic position o f this comparatively narrowly distributed species o f eastern South Australia should be reviewed in future investigations. (5) The high-support clades asterisked in the ITS and A ccl phylograms (Figures 2, 3) encompass, apart from A. petraea, the species of subgenera A ustrostipa (6 of 7 species sampled), Ceres (5 of 6 species sampled), E rem ophilae (5 of 6 species sampled), L an cea (six of seven species sampled), Lan tern a (1 of 3 species sampled) and T uberculatae (5 o f 7 species sampled). The asterisked clades showed several sister species relationships and minor lineages within and between subgenera (see above), but none of the subgenera mentioned was resolved as separate lineage, which is in agreement with the trees presented by Syme et al. (2012). For the time being it seems best to assign all these subgenera to a single, expanded and most likely monophyletic subgenus Austrostipa. This suggestion, however, should not be interpreted as attempt to supersede traditional morphology-based by molecular phylogenetic taxonomic concepts. It is rather a contribution to obtain monophyletic taxa, which can serve as reliable units addressing questions about character evolution and/or biogeography in A ustrostipa, which have been barely touched upon to date.
Some of our suggestions for classification are not new, having been made in previous molecular phylogenetic studies of A ustrostipa, for example, the maintenance of subgenera Falcatae (Jacobs et al., 2007;Bustam, 2010Bustam, , 2012Syme et al., 2012), Longiaristatae and L o ba ta e (Jacobs et al., 2007;Syme et al., 2012), the broadening of subg. Austrostipa to include also subgenera Tuberculatae (Jacobs et al., 2007;Syme et al., 2012) and E rem ophilae (Syme et al., 2012), but our data do not support combining subgenera Arbuscula and Bam bu sin a, a suggestion based on their similar habit (Jacobs et al., 2007) .

Austrostipa and Anemanthele
The somatic chromosome numbers of 2n = 44 and 2n = 66 were established in 18 and in seven A ustrostipa species, respectively, as well as 2n = 44 in A n em an thele lessoniana in our previous study on chromosome numbers and karyotypes (Winterfeld et al., 2015) . These results corroborated the earlier chromosome counts in Austrostipa stipoides (2n = 44;Murray et al., 2005) and A nem anthele lessoniana (2n = 4 0 -4 4 ; Dawson and Beuzenberg, 2000;Edgar and Connor, 2000) . In some accessions a certain degree of aneusomaty was noted, for example, 2 n = 65, 66, 68, 70 in A ustrostipa sem ibarbata, but usually the chromosome number showed less variation or was uniform in the metaphase plates of each accession studied. A ustrostipa and A nem anthele thus encompass consistently polyploids with a chromosome base number o f x = 11. Apart from the overall similarity of their karyotypes, this common base number supports a close relationship of both genera and makes a common ancestry of Austrostipa and A nem anthele likely, in addition to the relationship shown by the molecular phylogenetic data (Figures 2, 3) (Jacobs et al., 2007;Romaschenko et al., 2012) .

Clade A
Austrostipa and A nem an thele were placed in a clade, in which otherwise the chromosome base number of x = 12 prevails (Clade A in Figure 7 ). This supports recognizing x = 1 1 as a synapomorphic character of both genera in this clade. The base number of x = 12 was found in the likely sister of Austrostipa and A n em an thele, namely a lineage formed by A chnatherum (2n = 2x = 24; rarely 2 n = 28 and few polyploids; see Supplem entary Appendix 2) and Oloptum (usually 2n = 2x = 24), whereas Stipellula most likely deviates from x = 12. Various somatic chromosome numbers have been reported for S. capensis (2n = 18, ca. 34, 36; Supplementary Appendix 2 ), 2n = 36 being the most frequent in the whole Mediterranean (Supplem entary Appendix 2 ). 2n = 18 appears to be trustworthy for an accession from Gran Canaria, Canary Islands (Borgen, 1970 using the synonym Stipa retorta Cav.), making a derived monoploid chromosome number of x = 9 strongly conceivable for this species with annual life form, which is unusual in Stipeae. Moreover, 2n = 28, possibly pointing to x = 7, was reported in its congener Stipellula parviflora (Desf.) Röser & Hamasha (Supplem entary Appendix 2 ). The clade of A ustrostipa, A nem anthele, A chnatherum , Oloptum and Stipellula has highly polyploid, monospecific Celtica (usually 2n = 8x = 96; x = 12) as sister. Australian/New Zealand Austrostipa and A n em an thele therefore are related to a group of genera distributed in Eurasia, the Mediterranean and with few outliers in Tropical East and South Africa (Clayton, 1970(Clayton, , 1972Freitag, 1989;Fish et al., 2015).
Also lower monoploid numbers such as x = 6 suggested by Stebbins and Love ( 1941, p. 379) for E riocom a, and x = 7, 8 suggested by Barkworth (2007) for Nassella or x = 9 might occur in both genera, which means that accessions with 2 n = 26, 28, 32, 36, 38 would represent tetraploids or hexaploids. Given the branching order in the phylogenetic scheme of Figure 7, such hypothetical monoploid chromosome sets of Ericom a and Nassella with x = 7 -9 have originated secondarily from x = 11, the most likely original number o f Clade B. In this phylogenetic context they do not give evidence of a sometimes suggested low 'original' base chromosome number of Stipeae (see Stebbins and Love, 1941;Johnson, 1972;Tzvelev, 1977) . Numbers reported in A m elichloa (2n = 40, 44, 46), Jarava (2n = 36, 40, 44, 66), and P seu doeriocom a Romasch., P.M. Peterson & Soreng (2n = 44,46) seem to be based most likely on x = 11 if aneusomaty also plays some role here to explain the slightly varying chromosome numbers (Figure 7 ). Chromosome numbers are unknown in monospecific North American T horneochloa and in Timouria, a Central to East Asian outlier o f this otherwise American clade. In summary, we suggest a secondary reduction of chromosome numbers in Eriocom a and Nassella via aneusomaty, whereas the chromosome base number originally was x = 11 in Clade B and not lower (Figure 7 ). This supposed reductional dysploidy in Nassella would agree with the result that the species of Nassella with low chromosomes numbers [2n = 26, 28, 30  TA BLE 4 | Subgenera and species of A ustrostipa in this study and according to Jacobs and Everett (1996) supplem ented by Williams (2011) and informal groups suggested by Barkworth and Everett (1987).
T h is s tu d y J a c o b s a n d E v e re tt, 1996 B a rk w o rth a n d E v e re tt, 1987 have comparatively large chromosomes due to non-reciprocal translocations from chromosomes that finally became lost.

Transcontinental Stipeae clade
Both lineages of prevailingly x = 12 (Clade A) and x = 1 1 (Clade B), though with exceptions in Stipellula and species of E riocom a and Nassella, constitute one of the major clades in Stipeae, which was named 'Transcontinental Stipeae clade' in Hamasha et al. (2012) to denote its representation on all continents including Australia and New Zealand (Figure 7 ), and it is congruent with the 'achnatheroid clade' of Romaschenko et al. (2012).

M ain x = 11 clade
This clade represents the second main clade of Stipeae and includes Stipa s.str., by far the largest genus of this tribe ( Figure 7 ). It agrees with the 'Clade 1 or x = 1 1 clade' of Cialdella et al. (2007). Exceptions from x = 1 1 are seemingly scarce in this clade but were noted for North American monospecific Oryzopsis (only O. asperifolia Michx. with probably x = 12), monotypic Asian N eotrinia (uncertain x = 11 or 12) and some species of the North American genus Piptatheropsis (five species; Supplementary Appendix 1). In P. pungens (Torr. ex Spreng.) Romasch., P.M.Peterson & Soreng 2 n = 22 and 24 were found, the latter number possibly caused by aneusomaty, whereas 2 n = 2x = 20 was counted in mitotic and meiotic stages of two different accessions in P. shoshon ean a (Curto & Douglass M.Hend.) Romasch., P.M. Peterson & Soreng (Curto and Henderson, 1998), which implies x = 10, and represents the lowest chromosome number of Stipeae in the New World as noted already by these authors.

Clades C and D
The  Figure 7 ). There are only few exceptions since monospecific Oryzopsis, widely distributed in woodland of North America, is nested in Eurasian Clade C and Ptilagrostis occurs in mountainous to alpine landscapes of both Central Asia and western North America (Figure 7 ).

Stipeae Chromosome Base Number
The occurrence oftwo main clades in Stipeae, one clade primarily with x = 12 harboring also Austrostipa and A n em an thele, which are characterized by a derived number of x = 11, the other with primary x = 11 and few exceptions (see above Stipellula, Oryzopsis, species of Piptatheropsis and possibly also of E riocom a and Nassella; Figure 7 ), raises the question which one was the 'original' chromosome base number o f the whole tribe Stipeae. Due to the tree topology with monospecific M acrochloa Kunth as sister to the remainder of the tribe (Figure 7 ), this question cannot be reliably answered because in M acrochloa chromosome counts are equivocal, some suggesting x = 12 and others x = 11 or x = 10 (Supplem entary Appendix 2 ). W e regard x = 12, as firstly proposed by Avdulov ( 1931, p. 130) and accepted also by Romaschenko et al. (2012), a bit more probable as original chromosome base number of Stipeae than x =11. Interestingly, this is supported mainly by chromosome numbers represented in the presumably closely related tribes of Stipeae (see below). Further research into chromosome numbers of Stipeae, especially the re-examination o f questionable counts contained in the older karyological literature as cited in reference works of Darlington and Wylie (1956) and Fedorov (1969), particularly the reported low numbers, seems worthwhile. This problem is frequently encountered with older chromosome counts also in other plants groups, because counting was made using tissue sections, in which single chromosomes could easily become lost, instead of the nowadays employed and more reliable squashing technique.

The Lowest Chromosome Number in Stipeae
2n = 18 counted in a therefore diploid accession o f Stipellula capensis from the Canary Islands (Borgen, 1970) seems to represent the lowest reliably known chromosome number of the whole tribe Stipeae (Supplem entary Appendix 2 ). The chromosome number so far considered as lowest in Stipeae (Curto and Henderson, 1998;Barkworth, 2007) refers to a Crimean accession of A chnatherum brom oides with likewise 2n = 18 (Petrova, 1968), which was cited also in the reference works of Prokudin et al. (1977) and Agapova et al. (1993). This report appears to be questionable in view of the other chromosome counts available for A. brom oides, namely 2n = 24 (Ghukasyan, 2004) and repeatedly reported 2n = 28 (Vazquez and Devesa, 1996 and references therein). Stipellula capensis, however, is otherwise known from many tetraploid populations widespread in the Mediterranean (2n = 36; Supplementary Appendix 2) and has further chromosome numbers (see above and Supplementary Appendix 2 ), which are currently difficult to interpret (possible triploid hybrids, aneusomatic specimens, and partly probably erroneous counting).

Whole-Genome Duplications in Stipeae
Although chromosome numbers are still unknown for a number of small genera encompassing only eight species (see above and Supplementary Appendix 2 ), the enormous significance of whole-genome duplications (Johnson, 1972) is clearly obvious in many genera of Stipeae, for which chromosome numbers are available. Diploids are by far the minority in this tribe and only four o f 33 genera (12%) are consistently diploid and two further genera (6%) have diploid as well as polyploid species. It was pointed out already by Tzvelev (1977) that most extant Stipeae are polyploids and have hybrid origin as corroborated by our exemplary findings on the single-copy gene A ccl in Austrostipa and Stipa (Figure 3 ). Although no correlation between wholegenome duplication and diversification could be found in many tested clades of angiosperms (Clark and Donoghue, 2018), most, if not all o f the larger, speciose genera of Stipeae have consistently polyploid species as far as known, for example, Stipa (> 1 5 0 species) as delineated in the present (Stipa s.str.), Nassella (117), A ustrostipa (64), E riocom a (27), P appostipa (31), etc., taking into account that relative low chromosome numbers in some species of Nassella and E riocom a might be derived from polyploids (see above).

Biogeographic Relations and Origin of Austrostipa and Anemanthele
A ustrostipa and A n em an thele were previously considered as overall rather derived members of Stipeae (ITS analyses of Jacobs et al., 2000Jacobs et al., , 2007, having groups of American Stipeae like A m elichloa, Jarava, N assella, American A chnatherum , whose species have meanwhile been transferred to E riocom a and P seu doeriocom a (Peterson et al., 2019), and Eurasian A chnatherum as phylogenetically close relatives. On the other hand, considerable differences in lemma epidermal patterns both within Austrostipa and between some of these genera were noted. Definitely most A ustrostipa species have maize-like LEP as widespread among achnatheroids, occasionally showing variants such as the prickly maize-like or patterns with particularly shaped fundamental cells (see above), whereas some of the American genera have further maize-like types, for example, in A m elichloa, E riocom a, or P seu doeriocom a or strongly deviant LEPs as in Nassella with ladder-like LEP (Barkworth and Everett, 1987;Romaschenko et al., 2010Romaschenko et al., , 2012 . Considering A ustrostipa, Tzvelev (1977) discussed a migration of Stipeae from South America via Antarctica to Australia as more likely than migration of Stipeae from north to south, i.e., from Eurasia to Australia. Nevertheless, he argued that the Australian feathergrasses were morphologically closer to Eurasian sections o f Stipa than to sections of South American feathergrasses and cited among the examples also Stipa section Stipella Tzvelev which encompassed only Stipa capensis Thunb. ( = Stipellula capensis) in his view (Tzvelev, 2011) . Within the Transcontinental Stipeae clade, Austrostipa and A nem anthele are closely affiliated with Eurasian to Mediterranean genera, namely A chnatherum (including few eastern to South Africa species; see above), Celtica, Oloptum and also Stipellula, whereas they are distant to the American members of this clade (Figure 7 ). This implies that the ancestors of Austrostipa and A nem anthele with x = 11 came from the lineage with x = 12 (Clade A) and not the mainly American lineage with x = 1 1 (Clade B), and therefore acquired x = 11 in parallel to the American representatives of the Transcontinental Stipeae clade. The ancestors were most probably diploid like most if not all extant species o f A chnatherum except for tetraploid South African A. dregeanum , comb. nov., with 2n = 48 (Supplementary Appendix 2 ). Whole-genome duplication could have followed later but probably preceded the evolutionary radiation of A ustrostipa. Finally, the colonization o f Australia started likely from Central/East Asia, where A chnatherum species, for example, are well represented in the present and the precursors of Austrostipa might have occurred as well. Judging from the current distribution, with no occurrences of A chnatherum in subtropical and tropical southeastern mainland Asia and Indonesia/New Guinea, long-distance dispersal from Central/East Asia to Australia as pictured in Figure 8 seems plausible. It has parallels considering, for example, the sister relation between D uthiea Hack. ex Procop.-Procop., a genus of China and the Himalayas, and A nisopogon from southeastern Australia in the neighboring tribe Duthieeae (plastid DNA data of Schneider et al., 2011) .
New Zealand A nem anthele could have acquired and established its main apomorphic character, the occurrence of only a single stamen per floret, due to isolation and initial small population size (Veldkamp, 1985;Jacobs and Everett, 1996;Edgar and Connor, 2000). It bears morphological resemblance with A chnatherum rather than with Austrostipa as noted by Jacobs and Everett ( 1996, p. 582). It is not sure that the precursor of A n em an thele came from Australia as likely in A ustrostipa stipoides, the only autochthonous species of this genus in New Zealand. A chnatherum is not represented in Australia by any autochthonous species, whereas New Zealand harbors an endemic species of A chnatherum , A. petriei (Figure 8) (Edgar and Connor, 2000), the presumably only autochthonous representative of this genus in whole Australasia. Chromosome counts reported 2 n = 42 for this species (Supplementary Appendix 2 ), which is remarkable because otherwise 2 n = 24 is characteristic of A chnatherum (Supplem entary Appendix 2 ), seemingly except for A. brom oides, in which both 2 n = 24 and 2 n = 28 was reported (see above). The tetraploid number of A. petriei is most likely based on x = 11, if aneusomatic change from 44 to 42 is assumed, rather than on x = 12. A chnatherum petriei thus seems to share with A ustrostipa and A n em anthele the chromosome base number of x = 11 representing most likely a synapomorphy within Clade A, and shares the ploidy level of 4x, both in marked contrast to typical A chnatherum . It can be assumed therefore that A chnatherum petriei might be a close relative o f A n em an thele and Austrostipa pending further investigation. investigations. Stipa dregeana var. dregeana is considered as South African endemic (Fish et al., 2015), whereas Afromontane var. elongata (Nees) Stapf (syn. S. keniensis (Pilg.) Freitag subsp. keniensis) occurs in Ethiopia and Tanzania (Freitag, 1989) . Description: Inflorescence reduced to 1 -3 spikelets, lemma with maize-like epidermal pattern, with short dumbbell-shaped fundamental cells and frequent elongate to ovate silica cells, apex of lemma with two distinct, ca. 3 mm long lobes; spreading plants without basal tuft o f leaves.

DATA AVAILABILITY STATEMENT
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/Supplementary M aterial.

SUPPLEMENTARY MATERIAL
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2020. 630788/full#supplementary-material S u p p le m e n ta ry F ig u re 1 | Maxim um likelihood phylogram of all studied accessions o f Austrostipa species inferred from plastid 3 'trn K region DNA sequences w ith A nisopogon avenaceus (Duthieeae), B rom us erectus (Bromeae), H ordeum vulgare, and Secale sylvestre (both Triticeae) used as outgroup. M L and M P bootstrap support values > 50% as well as Bayesian PP > 0.5 are indicated on the branches. Clades with M L support < 50% are collapsed. The taxonom ic groupings of the Austrostipa species according to Jacobs and Everett (1996) and this study are marked by different colors in colum ns 1 and 2. A .,  Jacobs and Everett (1996) and this study are marked by different colors in colum ns 1 and 2. A ., Austrostipa.
S u p p le m e n ta ry F ig u re 3 | C luster analysis (UPGMA) perform ed on four m icrom orphological characters of 34 Austrostipa taxa. See S u p p le m e n ta ry Table 2 for the data matrix evaluated. The taxonom ic groupings of the Austrostipa species according to Jacobs and Everett (1996) and this study are marked by different colors in colum ns 1 and 2. A., Austrostipa.
S u p p le m e n ta ry F ig u re 4 | C luster analysis (UPGMA) perform ed on nine m acrom orphological characters of 65 Austrostipa taxa. See S u p p le m e n ta ry Table 1 for the data matrix evaluated. The taxonom ic groupings of the Austrostipa species according to Jacobs and Everett (1996) and this study are marked by different colors in colum ns 1 and 2. A., Austrostipa.
S u p p le m e n ta ry T a ble 1 | Data matrix used for m acrom orphological analysis. See Table 3 for measurements and character coding. A., Austrostipa.
S u p p le m e n ta ry T a ble 2 | Data matrix used for m icrom orphological and com bined macro-and m icrom orphological analyses. See Table 3 for measurements and character coding. A., Austrostipa.
S u p p le m e n ta ry A p p e n d ix 1 | Taxa studied for DNA sequences, m orphology and lemma m icromorphology. Taxon, geographical origin, voucher information with collectors and herbarium code and ENA/GenBank accession num bers for plastid m a tK g e n e -3 'trn K exon region; nuclear ribosomal ITS1-5.8S gene-ITS2 and nuclear single-copy gene A c c l . Sequences LR 989057-LR 989267 were newly generated for this study. A single asterisk (*) indicates sequences previously generated in our lab, tw o asterisks (**) indicate sequences taken from ENA/GenBank, a dash (-) missing data. The uppercase letters (A), (B) and (C) denote different A c c l sequence copies. BG, Botanical Garden; LEP, lemma epidermal pattern studied; LEP ill., LEP illustrated in F ig u re s 4, 5; morph, m orphological data for S u p p le m e n ta ry Table 2