What is Atraphaxis L. (Polygonaceae, Polygoneae): cryptic taxa and resolved taxonomic complexity instead of the formal lumping and the lack of morphological synapomorphies

Background: The recently proposed recircumscription of the genus Atraphaxis (incl. Atraphaxis section Ovczinnikovia O.V. Yurtseva ex. S. Tavakkoli and Polygonum sect. Spinescentia Boissier (=A. sect. Polygonoides S. Tavakkoli, Kaz. Osaloo & Mozaff.) makes this genus fairly heterogeneous and therefore almost undefinable based on morphology. A critical comprehensive reappraisal of the group is necessary. Methods: Using the DNA sequence data (ITS1&2 regions of nrDNA and combined trnL intron + trnL–F IGS and rpl32–trnL(UAG) IGS regions of plastid genome), Maximum Likelihood (ML) and Bayesian analyses (BI) were applied for phylogenetic reconstructions of the tribe Polygoneae with special attention to Atraphaxis, and related taxa. Maximum parsimony reconstructions of the evolution of perianth morphology and sporoderm ornamentation in the tribe Polygoneae were also performed. Life history, morphology of shoots, leaf blades, ocreas, perianth and achene morphology, ultrasculpture of achene surface, and pollen morphology were compared, and SEM and LM images were provided. Principal findings: The genera Atraphaxis and Polygonum were found to be widely polyphyletic. The rarest and morphologically remarkable endemic of Tian-Shan and Pamir Atraphaxis ovczinnikovii (Atraphaxis sect. Ovczinnikovia O.V. Yurtseva ex. S. Tavakkoli) was confirmed to be a sister of the clade (Atraphaxis + Polygonum sect. Spinescentia) in plastid topology. The genus Bactria (=Atraphaxis sect. Ovczinnikovia), which circumscribes two species, is newly established as a result of this analyses. Morphological data confirm the originality of the taxon. Discussion: We are arguing for a narrow delimitation of Atraphaxis with petalloid segments and striato-perforate sporoderm ornamentation as morphological synapomorphies. The recently proposed inclusion of Polygonum sect. Spinescentia in Atraphaxis is fairly questionable from a morphological standpoint. The rank of Polygonum sect. Spinescentia requires further clarification. The generic composition of the tribe Polygoneae also requires future reappraisals.

The majority of authors supported the division of Atraphaxis into two groups based on flower merosity, the first of which was comprised of species with dimeric perianths and lenticular achenes, and the second-of species with pentamerous perianths and trigonous achenes (Table S2).
Another remarkable dwarf shrub has been collected in Tien Shan (Kyrgyzstan, Naryn reg., Dzhumgal distr., 7 July 2006, Lazkov 24) and also initially assigned as Polygonum ovczinnikovii. It was shown to be a sister to Atraphaxis in phylogenetic reconstructions based on ITS data (Yurtseva et al., 2012a;Yurtseva et al., 2012b).
In other words, the genus Atraphaxis emend S. Tavakkoli combines the taxa with different ornamentation of sporoderm and also with fairly different perianth morphology, varying from the perianth common for Atraphaxis (Jaubert & Spach, 1844-1846, to the campanulate or urceolate perianth with five equal-sized segments (typical for Polygonum sect. Avicularia sensu Meisner, 1857).
To summarize, from traditional standpoint, the most significant difference between the genus Atraphaxis and Polygonum is the morphology of the perianth with equal-sized segments not enlarged in fruiting stage in Polygonum, and with the inner two or three segments accrescent and surrounding the achene in Atraphaxis (used, for example, by Brandbyge (1993) and Li et al. (2003)).
Therefore, it seems clear that the recent formal proposal by Tavakkoli et al. (2015) makes the technically monophyletic genus Atraphaxis extremely heterogeneous and almost undefinable based on morphological grounds.
Atraphaxis sect. Ovczinnikovia O.V. Yurtseva ex. S. Tavakkoli was based solely on position of the single accession A. ovczinnikovii from Tien Shan in the ITS-based phylogenetic reconstruction by Tavakkoli et al. (2015). However, a more careful comparison of the accession A. ovczinnikovii from Tien Shan (Kyrgyzstan) with the specimens of A. ovczinnikovii from Pamir-Alay (Tajikistan), corresponding to P. ovczinnikovii, showed prominent differences in their morphology, clearly demonstrating, that they represent two different taxa.
Also, if the recently described Atraphaxis sect. Polygonoides will be treated as part of Polygonum (as it is supposed to be based on the morphology of the perianth), then the genus Atraphaxis, as circumscribed by Tavakkoli et al. (2015), will appear to be polyphyletic, with the sistership of A. sect. Ovczinnikovia to the clade (Atraphaxis s.str. + Polygonum sect. Spinescentia (=A. sect. Polygonoides)).
Therefore, the details of the phylogenetic placement of A. sect. Ovczinnikovia must be clarified, and, as a result, a more accurate taxonomical approach to the genus Atraphaxis must be established as a frame for future studies of the whole complex.
Our initial aim was to recover the phylogenetic placement of different accessions of Atraphaxis ovczinnikovii from Pamir-Alay and Tien-Shan in the tribe Polygoneae based on the analyses of combined cp DNA (trnL intron + trnL-F IGS and rpl32-trnL (UAG) IGS) and nuclear ITS sequences. Later, our aim was to compare the morphological characteristics of the species, which a) had been traditionally placed in Atraphaxis and b) had been included recently.

Plant Material
The morphological study involved 20 specimens of Atraphaxis and Polygonum from field collections, and ca. 1,000 specimens stored in the herbaria of V.L. Komarov Botanical Institute RAS, St. Petersburg, Russia (LE); Lomonosov Moscow State University, Moscow, Russia (MW); Tsitsin Main Botanical Garden RAS, Moscow, Russia (MHA); and Main Botanical Garden, National Academy of Science, Bishkek, Kyrgyzstan (FRU). For scanning electron microscopy (SEM), 26 specimens of 19 species of Atraphaxis and Polygonum were used (Data S1). For light microscopy (LM) 18 specimens of 12 species of Atraphaxis and Polygonum were used (Data S2).
The molecular study involved all of the genera of the tribe Polygoneae, as circumbscribed by Schuster, Reveal & Kron (2011b): genera Atraphaxis s.l. (two accessions A. ovczinnikovii collected in locus classicus (Pamir-Alay), as well as the single accession from Tien Shan (which had also been analyzed), Polygonum s.l. (incl. Polygonum sect. Spinescentia), Polygonella, Duma, Muehlenbeckia, Fallopia, Reynoutria, and Knorringia. Based on the results of Schuster, Reveal & Kron (2011b), Knorringia has been chosen as an outgroup for the a posteriori rooting of the molecular topologies (Figs. 1 and 2). Table S3 contains information on the taxa and GenBank accession numbers used in the study.

DNA isolation and amplification
DNA was extracted from the herbarium specimens using a NucleoSpin Plant Extraction Kit (Macherey-Nagel, Germany) with the yield of DNA ranged from 0.005-0.1 mg per 0.1 g of plant material.
The PCR was performed in 0.02 ml of a mixture contained 10-20 ng DNA, 5 pmol of each primer and MaGMix (Dialat LTD, Russia), contained 0.2 mM of each dNTP, 2.0 mM MgCl 2 , 2.5 units of Smart Taq polymerase. 1.0 mM of DMSO was included for amplification of nrDNA regions with high CG content.
Purification of PCR products and DNA sequencing Amplification products were purified by electrophoresis (Sambrook, Fritsch & Maniatis, 1989), one-band DNA fragments were extracted from the gel and purified using the GFX TM PCR DNA, Gel Band Purification Kit (GE HealthCare, Little Chalfont, UK), or Evrogene Cleanup Mini (Russia), and then used as a template in sequencing reactions with the ABI Prism BigDye Terminator Cycle Sequencing Ready Reaction Kit (Applied Biosystems, Foster City, CA, USA) following the standard protocol provided for 3100 Figure 1 ITS phylogeny of the tribe Polygoneae. (A) the best tree from ML analysis of the ITS dataset (-log likelihood: 7894.75825). (B) the same tree with the branch lengths. Numbers above the branches indicate the aLRT support values equals or more than 0.8 from ML analysis/posterior probabilities equals or more than 0.9 from the BI of the same Matrix. Images: E. Mavrodiev. Avant Genetic Analyzer (Applied Biosystems, Foster City, CA, USA). Fragment sequences were determined at the Genom Center (Engelhardt Institute of Molecular Biology RAS, Moscow, Russia).

Cloning
Purified ITS amplificates from three plants of A. tortuosa were ligated into the pBluescript KS+ vector and were cloned in Escherichia coli NM 522 cells. The lysed colonies containing Figure 2 Plastid phylogeny of the tribe Polygoneae. (A) the best tree from ML analysis of combined plastid dataset: trnL intron + trnL-F IGS and rpl32-trnL (UAG) IGS regions of cpDNA (-log likelihood: 12644.7663). (B) the same tree with the branch lengths. Numbers above the branches indicate the aLRT support values equals or more than 0.8 from ML analysis/posterior probabilities equals or more than 0.9 from the BI of the same Matrix. Images: E. Mavrodiev. recombinant plasmids were used in the amplification reaction with M13/pUC sequencing primers and the resulting amplificates were sequenced with the original ITS1&2 PCR-primers.

Sequences used in phylogenetic analyses and the alignment strategy
Voucher information, current and GenBank numbers are presented in Table S3.

Phylogenetic analyses
The Maximum Likelihood analysis was performed with the PhyML v. 3.0 (Guindon & Gascuel, 2003;Guindon et al., 2010) following the automatic Smart Model Selection (SMS) option, with an estimated proportion of invariable sites and empirical nucleotide equilibrium frequencies. We took a BioNJ tree as a starting tree, and defined the strategy of the tree topology search as "best of NNIs and SPRs" following with ten random starts. Branch supports were calculated with the approximate likelihood-ratio test (aLRT) (reviewed in Guindon et al. (2010)). The GTR + G model was selected by PhyML-SMS as the best choice based on both Akaike and Bayesian information criteria for both plastid and ITS data matrices.
Following the general assumptions of the same model, the Bayesian analyses of the ITS and plastid matrices were conducted with the MrBayes (v. 3.1.2) (Ronquist & Huelsenbeck, 2003) as implemented in CIPRES (Miller, Pfeiffer & Schwartz, 2010). Two runs with four chains each (three heated and one cold) were run for 10 million generations; the chains were sampled every 1,000 generations with a default parameter.

Morphological analysis
In total, 8 morphological characters are discussed below. Specific attention has been paid to the most important diagnostic traits of Polygoneae (the perianth morphology and the ornamentation of the sporoderm).
Eventually the observed perianth morphology of Polygoneae was formalized using the following five character states: 0-campanulate divided to 1/2 in 5 equal-sized segments, with a short tube; 1-campanulate divided to 2/3-3/4 in 5 equal-sized segments, with a short tube; 2-campanulate divided to 4/5-5/6 in 5 equal-sized segments, with a short tube; 3-campanulate divided to 8/10-9/11 in 5 equal-sized segments, with a short tube; 4-divided in 4-5 segments, the outer ones smaller than the inner ones, with a long filiform tube; 5-divided in 4-5 segments, the outer segments larger than the inner ones, with a long filiform tube.

The optimization of morphological traits
Using topologies that resulted from the ML analysis of the a. combined plastid and b. the ITS matrices, the selected morphological traits (the perianth morphology and the ornamentation of sporoderm) were optimized in the most parsimonious way, as implemented in Mesquite v. 3.01 (Maddison & Maddison, 2011) treating all character states as "unordered" (reviewed and summarized in Kitching et al. (1998) and Maddison & Maddison (2011)).

Scanning electron microscopy (SEM) and light microscopy (LM)
Dry material and pollen samples were placed onto aluminum stubs, coated with gold or an alloy of platinum and palladium using a JFC-1100E sputter coater and studied under scanning electronic microscopes, Camscan-S2 and JEOL JSM-6380LA at 15-20 kV. SEM investigation was performed in the Laboratory of Electron Microscopy of M.V. Lomonosov Moscow State University, Faculty of Biology. Describing the achene surface, we followed the terminology of Ronse De Craene & Akeroyd (1988), Barthlott (1981) and Barthlott (1984). Palynological traits were described as suggested by Punt et al. (2007) and Hesse et al. (2009). LM-images were made with the stereoscopic microscope Stemi 2000-C Carl Zeiss (Zeiss, Oberkochen, Germany) using the camera Axiocam-MR and programm AxioVision V. 4.8 free edition.

Journal nomenclatural statement
The electronic version of this article in Portable Document Format (PDF) will represent a published work according to the International Code of Nomenclature for algae, fungi, and plants (ICN), and hence the new names contained in the electronic version are effectively published under that Code from the electronic edition alone. In addition, new names proposed in this work have been issued with identifiers by IPNI, and will eventually be made available to the Global Names Index. The IPNI-LSIDs can be resolved and the associated information viewed through any standard web browser by appending the LSID contained in this publication to the prefix "http://ipni.org/." The online version of this work is archived and available from the following digital repositories: PeerJ, PubMed Central, and CLOCKSS.

ITS phylogeny
The results of ML and BI of the ITS matrix of Polygoneae ( Atraphaxis also appeared as polyphyletic. The clade corresponding to Atraphaxis s.str. is well supported (0.96/0.99). However, A. ovczinnikovii from Pamir (two accessions) was recognized as an immediate sister to the clade (Atraphaxis s.str. + Polygonum sect. Spinescentia), and A. ovczinnikovii from Tien-Shan appeared as a non-supported sister to the clade (A. ovczinnikovii (Pamir, two accessions) + Polygonum sect. Spinescentia + Atraphaxis s.str.). Therefore, A. sect. Ovczinnikovia had been found as paraphyletic (Fig. 1).
The members of the sections Atraphaxis, Physopyrum and Tragopyrum, as well as the former members of Polygonum (A. toktogulica, A. tortuosa, A. atraphaxiformis, and A. ariana) are intermixed within the clade Atraphaxis s.str.

Combined chloroplast phylogeny
In general, the results of the ML and BI analyses of the combined chloroplast data matrix are similar to the results of phylogenetic analyses of the ITS matrix (Fig. 2).
The majority of the species from Polygonum sect. Spinescentia (P. aridum, P. dumosum, P. salicornioides, P. khajeh-jamali, P. spinosum) formed a highly supported clade (0.98/1.0) that appeared as a sister of well-supported Atraphaxis s.str. clade (Fig. 2). Tavakkoli et al. (2015) reported the conflicting phylogenetic placement of the rarest and poorly known Polygonum botuliforme, confirmed in own analyses (Figs. 1 and 2). This issue requires future reappraisals and at the moment is excluded from future discussion.
Atraphaxis s.l. appeared as polyphyletic: two accessions A. ovczinnikovii from Pamir (both grouped together with a high level of support) and the single accession from Tien Shan (appeared as a strongly supported sister to the Pamirian subclade) form a highly supported clade (0.95/1.00), sister to the well supported (Polygonum sect. Spinescentia + Atraphaxis s.str.) clade (0.99/0.99).
Stressing the remarkable phylogenetic placement of A. sect. Ovczinnikovia, that emerged far away from the rest of the Atraphaxis s.str., both ML and BI analyses of the combined plastid matrix argue for the strong monophyly of the sect. Ovczinnikovia, that appeared to be paraphyletic as a result of the phylogenetic analyses of the ITS data set (see above.) Morphology of Atraphaxis ovczinnikovii, Polygonum sect. Spinescentia, and Atraphaxis s.str.
We compared main morphological characteristics of the species that had been traditionally placed in Atraphaxis s.str. and had been included recently. The latters include the Pamirian and Tien-Shanian accessions Atraphaxis ovczinnikovii, remarkable by their distant positions from Atraphaxis s.str. in phylogenetic reconstructions, Polygonum sect. Spinescentia, which is sister to Atraphaxis s.str., and A. toktogulica, A. tortuosa, A. atraphaxiformis, and A. ariana, which are nested within Atraphaxis s.str. clade. Morphological characteristics of the taxa are summarized in Tables 1 and S4.
Life history and general architecture of shoot system The taxa under study demonstrate some distinctions in the structure of vegetative organs and inflorescences. Both Atraphaxis ovczinnikovii from Pamir (Tajikistan) and A. ovczinnikovii from Tien Shan (Kyrgyzstan) can be described as divaricately branched dwarf shrubs 10-30 cm tall, with elongated leafy annual shoots, and generative shoots terminated by frondulose thyrses with 3-7 axillary cymes of 1-2 flowers (Figs. S1 and S2). The second-year shoots are covered with a gray fibrously disintegrated bark. The plants from Tian Shan have foxy-colored ribbed annual shoots, the plants from Pamir have greenish-gray rounded annual shoots, and in both taxa they are shortly puberulent.
The majority of species have light-green or creamy annual shoots. The second-year shoots are usually covered with a light-gray bark, rarely they are yellowish-gray (A. laetevirens and A. caucasica), or foxy-brown (A. pungens and A. muschketowi). The annual shoots are densely covered with 0.2 mm long papillae, or glabrous.

The ochreae in thyrses and vegetative shoots
Atraphaxis ovczinnikovii from Pamir in thyrses has the ochreae 2-4 mm long, broadly ovate, cup-form inflated under the petiole, at base greenish-brown, herbaceous, densely shortly puberulent, above membranous, semitransparent, without visible veins, entire, bidentate (if a leaf blade is reduced fully), or cleft in two shortly acuminate lacinulae without veins at both sides of reduced leaf blade (Fig. 4A). In vegetative shoots the ochreae are similar in size and consistence, but lanceolate-tubulate, later cleft in two lanceolate lacinulae.
Atraphaxis ovczinnikovii from Tien Shan has similar ochreae, but brownish at base, transparent above, puberulent along the keel and later split in two lanceolate lacinulae, each with a single reddish vein.
However, in vegetative shoots of A. ariana, A. toktogulica, A. atraphaxiformis, A. tortuosa the ochreae are 7-10 mm long, tubulate, herbaceous and puberulent at the base, membranous and transparent above, with two long linear-lanceolate aristate lacinulae at both sides of the leaf blade, each with a single vein, and a short finely serrate-incised middle part at the side opposite to the leaf blade. Other species (A. kopetdaghensis, A. avenia, A. seravschanica, A. frutescens, A. pungens, A. bracteata, A. virgata) have the ochreae similar in morphology and consistence, but 3-7 mm long, oblique-tubulate. Only A. spinosa has the ochreae 2-3 mm long, lacerate in two linearlanceolate lacinulae with hardly visible veins, and a short middle part.

Perianth morphology
A total of 25 characteristics are listed in Table 1, however, the perianth shape, the perianth partition, and the segment consistence are the most prominent for the discrimination of Tien-Shanian and Pamirian accessions Atraphaxis ovczinnikovii from Polygonum sect. Spinescentia and the rest Atraphaxis s.str. within Polygoneae (Table S5). Characteristics of the perianth, the perianth and achene size of the taxa under study are given in Table S4. Some brief notes on the perianth shape are still necessary.
All accessions of Atraphaxis ovczinnikovii from Tien Shan and Pamir have the campanulate perianth and five equal-sized segments with stomata along the midveins and papillae at the edge of the segments (Figs. 4-6).
Pamirian accessions have the perianth 3.0-4.0 mm long (4.0-5.0 mm by fruiting), divided almost to the base (8/10) in segments lanceolate, gradually acuminate, greenpurple, coriaceous, with a narrow pinkish margin. The outer segments are concave, narrowly keeled and slightly cucullate, the inner segments are slightly keeled and almost flat. The perinath tube of Pamirian plants is funnel-form, short, suddenly narrowed to a filiform base at the place of articulation with a pedicel (Figs. 4, 5 and 16A; Tables 1 and S4).
Tien Shanian accession has the perianth 2.0-2.5 mm long (2.5-3.0 mm by fruiting), divided to 4/5-5/6 in segments elliptical or broadly ovate, petalloid, herbaceous-green along the midveins, with wide semitransparent pink margin ( Fig. 6; Tables 1 and S4). The outer segments are slightly acuminate, cucullate, without keel, the inner segments are obtuse and flat. The tube is cup-shaped or sacciform at base and joined to a pedicel with articulation, not narrowed to the place of articulation.
The perianths of the majority of Atraphaxis species undergo transformation by the fruiting stage, the inner segments becoming much larger, then the outer segments, and enclosing the achene. The inner segments are broadly-elliptical, broadly ovate, orbiculate or reniform. The filiform basal part of the tube is up to 2.0-7.0 mm, the expanded upper part of tube is either wedge-or cup-shaped (Figs. 13B, 13D, 14C, 14D and 14F; Table S4).
Atraphaxis badghysi, A. bracteata, A. aucherii, A. angustifolia, A. grandiflora demonstrate a transitional perianth type divided to 1/2-2/3 into five subequal rotundate, rhomboidelliptical or broadly-elliptical segments, and the tube with a filiform base 1.5-3.0 mm long (Figs. 12C, 12D and 16G; Table S4). The tube of A. badghysi bears papillae (Fig. 12E).  The majority of the species in Atraphaxis sect. Tragopyrum have three inner segments greatly accrescent by fruiting, and two small segments reflected to a pedicel (Figs. 13B and 13D). The members of the section Atraphaxis have the perianth with two reniform inner segments accrescent by fruiting and two small outer segments (Figs. 14C, 14D and 14F). Atraphaxis teretifolia (sect. Physopyrum) has the perianth with three inner segments orbiculate, concave, spherically surrounding the achene, and two outer segments small and reflected to a pedicel (Figs. 15C, 15F, 15G and 17G; Table S4). The filiform basal part of the tube, 0.5-0.7 mm long, joins to a pedicel 5-6 mm long.
Depending on relative size of outer and inner segments, the shape of the segments (oblong, ovate, rounded, or reniform, flat, or concave), the edge of the segments (entire, undulate, crenulate, smooth or papillate), flower merosity (the number of outer and inner segments), the length of the filiform basal part of the tube, the shape of the upper part of the perianth tube (wedge-shaped, funnel-form, or cup-shaped), several perianth types are distinguishable within Atraphaxis s.str. (Table S4). However, all of the species of Atraphaxis are sharing the thin petalloid segments that are expanded at the top, white or brightly colored.
The analysis of the perianth morphology of Polygoneae (Tables 1 and S5; Fig. 20) shows that the ancestral character state of the clade (Atraphaxis ovczinnikovii + Polygonum sect. Spinescentia + Atraphaxis s.str.) is a campanulate perianth with equal-sized segments and a short tube. More specialized perianth with accrescent inner segments and a long filiform tube is a synapomorphy of Atraphaxis s.str.
Atraphaxis toktogulica, A. ariana, A. atraphaxiformis, and A. tortuosa (all with equal-sized perianth segments) do not form a separate subclade within Atraphaxis s.str., however, A. angustifolia and A. grandiflora, A. angustifolia and A. aucheri, all with subequal segments, are grouped together (Figs. 1 and 2). Within Atraphaxis s.str., the members of the sections Atraphaxis, Physopyrum and Tragopyrum with accescent inner segments are intermixed with the species with equal-sized segments (Fig. 20). In other words, the specialized perianth with accrescent inner segments evolved homoplastically.

Achene morphology
Achene sizes are given in Table S4. Atraphaxis ovczinnikovii from Pamir has ovoid, gradually acuminate larger achenes (4.0-5.0 Â 2.5-2.8 mm) with three styles connate at the base and forming a triquetrous stub at the top of the achene, each inflated at the base and filiform under the capitate stigma. The achenes with distinct ribs and equal concave faces are light-brown, smooth and glossy, either slightly exserted, or enclosed by the perianth (Fig. 5C).
Atraphaxis teretifolia from A. sect. Physopyrum has lanceolate trigonous achenes 2.5 Â 1.5 mm, black and glossy, with concave faces, the ribs distinct near the top and obtuse below, and three short free styles with small capitate stigmae (Figs. 15D, 17H and 17I).

Pollen morphology
The majority of taxa in the tribe Polygoneae have prolate to subprolate, or spheroidal pollen grains, ellipsoidal in equatorial view, trilobed-circular in polar view, tri-colporate (rarely loxocolporate or syncolporate); the colpi are long, narrow, sometimes anastomosing at the poles, with well-developed elliptical ora (Table S6). The taxa of the tribe Polygoneae differ mainly in sporoderm ornamentation.
Atraphaxis ovczinnikovii from Pamir differs by microreticulate-foveolate sporoderm surface (Figs. 19A and 19B): lumina with 4-6-angular pits are sharply defined at the edges; pits are rarely perforated at the bottom; perforations are few and small (0.1-0.2 mm in diameter), singular at the lumina (see also Yurtseva, Severova & Bovina, 2014). The plants from Tien Shan have sporoderm ornamentation that varies from microreticulatefoveolate, peculiar to the plants from Pamir,, with rounded pits smoothened at the edges and rare perforations 0.5-1.5 mm in diameter, single at the lumina, or two, divided by thin bridges.
The clade Atraphaxis s.str. includes the taxa with striate-perforate sporoderm ornamentation : smoothened or distinct striae are divided by deep or shallow grooves with rows of perforations in the bottom (more in Hong, 1995;Yurtseva, Severova & Bovina, 2014). That type of sporoderm ornamentation is obviously a synapomorphy (Figs. 20A and 20B). However, Atraphaxis toktogulica, the most basal in the ITS topology (Figs. 1 and 20A), has the pollen surface with a transitional type of sporoderm ornamentation from reticulate-perforate to striate-perforate one: polygonal shallow perforated lumina are arranged in rows of 2-3 (4), which are oriented meridionally and divided laterally by hardly visible smoothened striae (Figs. 19C and 19D).
Therefore, the accessions A. ovczinnikovii from Pamir and Tien Shan share palynotype with microreticulate-foveolate sporoderm ornamentation. It is noteworthy, that A. ovczinnikovii from Tien Shan demonstrates variability of the pollen surface, which in some pollen grains is foveolate-perforate with large single or double perforations. Both variants are fairly different from that of Polygonum sect. Spinescentia, which is reticulateperforate, or tectate-perforate, and in all the taxa distinct from striato-perforate sporoderm surface of Atraphaxis s.str. (Figs. 20A and 20B).  (Maddison & Maddison, 2011) of the perianth morphology (A) and the ornamentation of sporoderm (B) using topology resulted the ML analysis of the ITS data set (A) and plastid data set (B). The description of the character states is given in "Materials and Methods" (see above). All character states were treated as "unordered." Images: E. Mavrodiev.

DISCUSSION
The results of ML and Bayesian analyses based on nrDNA ITS1&2 sequences and combined cpDNA trnL intron (UAA) + trnL-F IGS and rpl32-trnL (UAG) IGS sequences ( Figs. 1 and 2) confirmed the division of the tribe Polygoneae in two major sister clades (RFM and ADP) previously recognized by Schuster, Reveal & Kron (2011b).
Compared to the results of Schuster, Reveal & Kron (2011b), the genus Fallopia (as sampled) appeared as polyphyletic (ITS topology, Fig. 1): the accession Fallopia baldshuanica is grouped with Reynoutria, and F. convolvulus and F. dumetorum form a separate clade. In plastid topology, Fallopia still remains monophyletic (Fig. 2). Contrary to the results of Schuster, Reveal & Kron (2011b), the genus Duma appeared to be the next clade of Polygoneae after the RFM clade in the ITS topology (Fig. 1), but is grouped with Polygonum s.l. in plastid topology (Fig. 2), which agrees with the findings of Schuster, Reveal & Kron (2011b).
Genus Polygonum appears as highly polyphyletic (Figs. 1 and 2). Polygonella and the majority of current sections of the genus Polygonum (Polygonum, Pseudomollia, Duravia), excepting P. sect. Spinescentia, form a moderately supported Polygonum s.l. clade in both chloroplast and nuclear trees. Within Polygonum s.l. clade, in ITS topology (Fig. 1), Polygonella and P. sect. Duravia Watson (1873) from North America are monophyletic and form a clade sister to the rest Polygonum s.str. In the combined plastid topology (Fig. 2), in the absence of P. sect. Duravia, Polygonella falls sister to Polygonum s.str. That partly agrees with the results of Schuster, Reveal & Kron (2011b), and better agrees with the results of Schuster et al. (2015).
The division of the Polygonum s.str. in two subclades (part of P. sect. Polygonum from Eurasia and North America, and part of P. sect. Polygonum from SW and Central Asia) confirmes the ITS-based topologies obtained previously (Yurtseva et al., 2010;Schuster, Reveal & Kron, 2011b). The members of both subclades share palynotype Avicularia (Hedberg, 1946;Hong, Oh & Ronse De Craene, 2005). The latter of two subclades, however, includes P. molliiforme and P. bornmuelleri from P. section Pseudomollia (Boissier, 1879;Komarov, 1936), sharing palynotype Pseudomollia (defined by Hong, Oh & Ronse De Craene, 2005), therefore, the part of Polygonum sect. Polygonum from SW and Central Asia was recircumscribed by Schuster, Reveal & Kron (2011b) as P. sect. Pseudomollia Boissier (1879). Our analysis showed high variability in habit and perianth morphology in the members of both subclades of Polygonum s.str., but no special features were found to distinguish the members of two subclades (Yurtseva et al., 2010).

Polygonum sect. Spinescentia and Atraphaxis
Polygonum sect. Spinescentia is very distant from the rest of the Polygonum. The samenamed clade appeared as the immediate sister of Atraphaxis s.str. clade (Figs. 1 and 2). This group resembles Polygonum in perianth shape and partition (Meisner, 1857;Boissier, 1879), but is fairly distinct from the other knotgrasses in habit, ochreae (Tavakkoli et al., 2015), and also possesses another type of sporoderm ornamentation (Fig. 20), which is never reticulate-perforate or tectate-perforate in Polygonum (see more in Table 1). The rank of Polygonum sect. Spinescentia clearly requires future clarification.
Despite variability of perianth morphology within the clade Atraphaxis (see Results) deep similarity in life history, habit, inflorescence structure, pollen shape and size, and striato-perforate sporoderm ornamentation suggests a narrow delimitation of Atraphaxis corresponding to the clade Atraphaxis s.str. in our own plastid and ITS-based reconstructions. Petalloid segments and striato-perforate sporoderm ornamentation of pollen are unique morphological synapomorphies appeared in the genus Atraphaxis.
The campanulate perianth with equal-sized segments present in some Atraphaxis species looks as ancestral state for the more advanced perianth with accrescent inner segments and a filiform tube as long as a pedicel. The latter perianth type appears and predominates only in the clade Atraphaxis s.str., being correlated with compact abracteose or bracteose thyrses of congested cymes of flowers, making the floral units compact and showy.
Due to the flower buds and premature flowers of all Atraphaxis species have the campanulte perianth with equal-sized segments and rather short filiform base of tube, the perianth with enlarged inner segments and long filiform tube seems to have appeared through ontogenetic transformation of the perianth by the elongation of the basal part of the tube and by the accrescence of the inner segments. Within the clade Atraphaxis s.str., mainly basal positions of the species with presumably ancestral character state (A. toktogulica, A. ariana, A. tortuosa) and distal positions of the species with transitional perianth type (A. anfustifolia, A. grandiflora, A. bracteata, A. badghysi), or advanced type allow to suggest homoplastic origin of the specialized perianth widespread in Atraphaxis s.str.
The advanced perianth type provides more effective pollination in compact thyrses with congested cymes of flowers and better dispersal of the achenes. Long filiform basal part of a tube joined to an equally long pedicel puts out a flower from the thyrse and rises up the swing amplitude of the fruit, increasing the chance of far-distant dispersal. Accrescent inner segments serve for better protection of the ovary and facilitate dispersal of fruits by wind in open communities. Atraphaxis teretifolia has the shortest filiform tube ( Fig. 18; Table S4), but the longest pedicel (6-7 mm). This species from sandy deserts of Kazakhstan realizes another way of dispersal of the achenes hidden in a spherical perianth, by rolling on the surface, while the shortest tube does not prevent slipping along sandy surface.
Papillae found at the perianth tube of A. toktogulica, A. atraphaxiformis, and A. ariana are present as well at the perianth of some Polygonum s.str., P. sect. Spinescentia, Oxygonum, and Fagopyrum (Hong, Ronse De Craene & Smets, 1998), far distant from Atraphaxis and Polygonum in phylogenetic reconstructions (Sanchez et al., 2011;Schuster, Reveal & Kron, 2011b). Along with protective functions, papillae at the perianth surface are considered as an adaptation to insect-pollination mechanism (Hong, Ronse De Craene & Smets, 1998), playing a tactil role for recognition by pollinators (Whitney et al., 2009;Ojeda, Francisco-Ortega & Cronk, 2009). Being present at the perianths of Polygonum sect. Spinescentia, A. toktogulica, A. atraphaxiformis, and A. ariana, the papillae are absend from more specialized perianths of the majority of Atraphaxis species characterized by accrescent inner segments and filiform basal part of tube, compact thyrses of congested cymes of flowers, and striato-perforate sporoderm ornamentation, a morphological complex, that was possibly caused by another pollination mode predominating in Atraphaxis.
The subclades recovered in the clade Atraphaxis s.str. in the ITS-based phylogeny do not correspond to the sections Atraphaxis, Tragopyrum, and Physopyrum, which previously were described in the genus Atraphaxis (Table S2). However, in the plastid phylogeny, the members of the section Atraphaxis with tetrameric perianth and dimeric gynoecium form a separate subclade.
Flower merosity was traditionally the most important characteristic discriminating the section Atraphaxis (A. spinosa, A. fischeri, A. canescens, A. replicata, A. karataviensis, A. compacta) from the section Tragopyrum. According to Krasnov (1888), A. muschketowi, A. laetevirens, and A. variabilis (=A. billardierei) from the section Tragopyrum with a pentamerous perianth, in arid conditions have smaller flowers with four segments and dimeric gynoecium. Variable flower merosity in some taxa might be a result of intrageneric hybridization. Finally, tetrameric perianth and dimeric gynoecium peculiar for the section Atraphaxis might be useful for the achene and fruit dispersal in open sandy deserts of Central Asia. The lenticular achenes hidden between two flat papery inner segments possibly have higher windage, than the fruits with three inner segments hiding triquetrous achenes. Sharp and flat ribs present at the lenticular achenes of the section Atraphaxis possibly facilitate dispersal of the achenes by wind in open steppe and semi-desert communities.
Atraphaxis sect. Ovczinnikovia The major focus of our research is Atraphaxis sect. Ovczinnikovia O.V. Yurtseva ex S. Tavakkoli. We already demonstrated that the morphologically remarkable Polygonum ovczinnikovii from Pamir is distinct from the rest members of Atraphaxis in pollen morphology (Yurtseva et al., 2012a;Yurtseva, Severova & Bovina, 2014). The recent and well-supported phylogenetic placement of Polygonum section Spinescentia as a sister to Atraphaxis s.str. (Tavakkoli et al., 2015, see also Figs. 1 and 2) followed by the sistership of A. ovczinnikovii from Tien Shan to the clade (Atraphaxis s.str. + Polygonum sect. Spinescentia) (Figs. 1 and 2) makes the recent inclusion of A. ovczinnikovii from Tien Shan in Atraphaxis (Tavakkoli et al., 2015) very questionable.
Our analysis of combined plastid matrix showed, that Pamirian and Tian-Shanian accessions of Atraphaxis ovczinnikovii are not intermixed with each other, but appeared as well supported sister groups: one includes the Pamirian plants, with the Tian Shanian accession as an immediate sister (Fig. 2). Moreover, Atraphaxis sect. Ovczinnikovia was recognized as paraphyletic based on the results of the phylogenetic analysis of the ITS matrix, with Tien-Shanian accession fell as a sister to Atraphaxis s.l. clade (incl. A. sect. Ovczinnikovia and Polygonum sect. Spinescentia), as recently circumscribed by Tavakkoli et al. (2015) (Fig. 1).
These results may be arguing for the hybrid origin of the Tien Shanian accession of Atraphaxis ovczinnikovii. However, much more investigation is necessary due to the wellknown issues with ITS marker (summarized in Álvarez & Wendel (2003), Baldwin et al. (1995) and Poczai & Hyvönen (2010)). This may be the subject of future considerations, but here we are arguing for the recognition of the Pamirian and Tian-Shanian samples of Atraphaxis ovczinnikovii as a new genus that seems to be well distinguishable from both Atraphaxis s.str. and Polygonum sect. Spinescentia based on morphological or phylogenetic standpoints (Figs. 1,2,(4)(5)(6)16A,16B,19A,19B,19K,19L and 20 4/5 (B. lazkovii) and to 8/10 (B. ovczinnikovii) in 5(6) equal-sized petalloid segments with papillae at the segment edge.
The pollen sample of Bactria lazkovii includes the pollen with the sporoderm ornamentation peculiar for Bactria ovczinnikovii, and the pollen with foveolato-perforate ornamentation of the sporoderm, with small rounded foveolae and rare single or double perforations 0.5-1.5 mm in diameter unique in Polygoneae (Figs. 19A and 19B), as well as some transitional types. This findings need additional study.
4. Our results are arguing for the narrow delimitation of Atraphaxis characterized by petalloid segments of perianth and striato-perforate sporoderm ornamentation of pollen as main morphological synapomorphies.
5. Due to the polyphyly of the genus Polygonum s.l., the generic composition of the tribe Polygoneae may also requires future reappraisals, but at the moment much more work is necessary to resolve the problem.
The key to the genera was constructed using own observations and reference data (Tables 1 and S4-S5).
Phenology: fl. June-July, fr. July-August. Distribution and habitat: endemic of Central Tien Shan (Kyrgyzstan), at rooks. Etymology: The species is named for G.A. Lazkov, the collector of the type specimen. Affinity: The species is similar to Bactria ovczinnikovii in habit, terminal frondulose thyrses, papillate covering of annual shoots, ochreae, leaves, but differs in foxy-brown slightly ribbed annual shoots, oblong-ovate leaf blades puberulent only at midvein abaxially, obtuse petalloid segments, cup-shaped or sacciform base of tube, flat achene sides and free styles not connate at base.