﻿The molecular phylogenetic position of Harpagocarpus (Polygonaceae) sheds new light on the infrageneric classification of Fagopyrum

﻿Abstract In the context of the molecular phylogeny of Polygonaceae, the phylogenetic positions of most genera and their relationships have been resolved. However, the monotypic genus Harpagocarpus has never been included in any published molecular phylogenetic studies. In the present study, we adopt a two-step approach to confirm the phylogenetic placement of Harpagocarpus using two datasets: (1) a concatenated dataset of three chloroplast DNA (cpDNA) regions (matK, rbcL and trnL-F) for Polygonaceae and (2) a combined cpDNA dataset of five sequences (accD, matK, psbA-trnH, rbcL and trnL-F) for Fagopyrum. Our analyses confirm the previous hypothesis based on morphological, anatomical and palynological investigations that Harpagocarpus is congeneric with Fagopyrum and further reveal that H.snowdenii (≡ F.snowdenii) is sister to the woody buckwheat F.tibeticum. Within Fagopyrum, three highly supported clades were discovered and the first sectional classification was proposed to accommodate them: sect. Fagopyrum comprises the two domesticated common buckwheat (F.esculentum and F.tataricum) and their wild relatives (F.esculentumsubsp.ancestrale, F.homotropicum and F.dibotrys) which are characterised by having large corymbose inflorescences and achenes greatly exceeding the perianth; sect. Tibeticum, including F.snowdenii and F.tibeticum, is characterised by the achene having appurtenances along the ribs, greatly exceeding the perianth and the perianth accrescent in fruit; sect. Urophyllum contains all other species of which the achenes were completely enclosed in the perianth. This study is very helpful to understand the phylogeny of the Fagopyrum and sheds light on the future study of taxonomy, biogeography, diversification and character evolution of the genus.

The genus Harpagocarpus was established on the basis of its distinct fruit morphology (Hutchinson and Dandy 1926) and contains the sole species, H. snowdenii Hutch. & Dandy, which was originally recorded only in Uganda, but now has been reported from Kenya, Tanzania, Rwanda and Cameroon (Ayodele 2003). Jacques-Félix (1946) described Fagopyrum ciliatum Jacq.-Fél. from Cameroon, but according to Graham (1958), it is merely a synonym of H. snowdenii. Due to its unique appurtenances growing along the achene ribs, which are long purple setae with the radially arranged retrorse barbs at the tip of each seta (Fig. 1), H. snowdenii is a distinctive species in Polygonaceae (Hutchinson and Dandy 1926).
In the protologue, Harpagocarpus was morphologically compared to Polygonum L. and Fagopyrum Mill., but it was thought to resemble the latter considerably more on its broad cotyledons, large and obviously exerted fruits and the shape and venation of the leaves (Hutchinson and Dandy 1926). On the basis of anatomical similarities, Haraldson (1978) hypothesised that Harpagocarpus may be closely related to Fallopia Adans. However, Ronse Decraene and Akeroyd (1988) argued against this hypothesis and pointed out that Harpagocarpus and Fagopyrum share considerable similarity in the morphology of floral characteristics. Hong (1988) further reduced Harpagocarpus to a synonym of Fagopyrum and proposed the new combination F. snowdenii (Hutch. & Dandy) S.P.Hong for H. snowdenii after concluding from additional palynological research. Though this treatment has been followed in some literature (e.g. Brandbyge (1993); Friis and Vollesen (1998);de Klerk et al. (2015)), it was, nonetheless, recommended that molecular data be used to confirm the phylogenetic position of Harpagocarpus (Schuster et al. 2015).
In the present study, we obtained a few precious pieces of leaf materials of H. snowdenii from the specimen Marshall A.R. WK 374 (detailed information available from: http://legacy.tropicos.org/image/100427626), which provided us an invaluable opportunity to investigate the phylogenetic position of Harpagocarpus, based on additional molecular data. We adopted two steps of phylogenetic analyses to infer the generic and specific affinities of H. snowdenii. Firstly, we used three chloroplast DNA (cpDNA) markers (matK, rbcL and trnL-F) to present the backbone phylogeny of Polygonaceae and affirmed the position of Harpagocarpus in Fagopyrum. Subsequently, based on five cpDNA regions (accD, matK, psbA-trnH, rbcL and trnL-F), we further reconstructed the phylogeny of Fagopyrum and clarified the accurate specific relationships of F. snowdenii within Fagopyrum.

Taxon sampling, choice of markers and datasets
We employed matK, rbcL and trnL-F sequences, which have been extensively used in previous studies (e.g. Lamb-Frye and Kron (2003); Sanchez and Kron (2008); ; Burke et al. (2010); Schuster et al. (2015)), to generate a concatenated cpDNA dataset (D1) for reconstructing the backbone phylogeny of Polygonaceae. The ingroup taxa were selected from the entire family to cover all recognised tribal clades Kempton 2012;Schuster et al. 2015) with at least one representative of each genus. A total of 37 genera and 77 species were sampled. Plumbago auriculata Lam. from Plumbaginaceae, which is the sister family of Polygonaceae (Yao et al. 2019;Li et al. 2021), was selected as the outgroup taxon. The source publications or voucher information for all sequences were listed in Suppl. material 1: table S1.
As the analyses of the D1 dataset demonstrated that Harpagocarpus is nested within Fagopyrum, we designed another dataset (D2) using five cpDNA regions (accD, matK, psbA-trnH, rbcL and trnL-F), with an expanded sampling of Fagopyrum aiming for a more accurate placement of H. snowdenii (= F. snowdenii). The ingroups of D2 dataset included 33 taxa of Fagopyrum covering most of the recognised species in the genus and the outgroup taxon was set as Pteroxygonum giraldii Damm. et Diels according to the results presented in Schuster et al. (2015). Voucher information and GenBank accession numbers for taxa used in the D2 dataset are provided in Suppl. material 1: table S2.

DNA extraction, amplification and sequencing
Total genomic DNA was extracted from fresh or silica gel dried leaves following the manufacturer's specifications of the DNEasy Plant Mini Kit (Qiagen, Valencia, CA, USA). After extraction, the DNA was resuspended in double-distilled water and kept at -40 °C for polymerase chain reaction (PCR). The PCR reactions and amplification protocol followed Schuster et al. (2011a). The amplified products were purified using a PCR Product Purification Kit (Shanghai SBS, Biotech Ltd., China). Sequencing reactions were conducted with the forward and reverse PCR primers using the DYEnamic ET Terminator Cycle Sequencing Kit (Amersham Biosciences, Little Chalfont, Buckinghamshire, U.K.) with an ABI PRISM 3730 automatic DNA sequencer (Shanghai Sangon Biological Engineering Technology & Services Co., Ltd., Shanghai, China). Both strands of the DNA were sequenced with overlapping regions to ensure that each base was unambiguous. Electropherograms were assembled and consensus sequences were generated with Geneious Prime 2022.0.2 platform.

Phylogenetic analysis
Sequencher version 5.4.6 (Gene Codes Corporation 2021) was used to evaluate chromatograms for base confirmation and editing contiguous sequences. All DNA sequences were initially aligned using Clustal X version 2.1 (Larkin et al. 2007) and adjusted manually in BioEdit Sequence Alignment Editor version 7.2.1 (Hall 1999).
Phylogenetic analyses were conducted, based on the combined cpDNA dataset D1 and D2. The cpDNA regions were supposedly safe to be combined in phylogenetic analyses (Olmstead and Sweere 1994) because the plastid genome is mostly uniparentally inherited (Soltis and Soltis 1998). The datasets were analysed separately using the methods of Maximum Likelihood (ML) and Bayesian Inference (BI).
ML and BI analyses were carried out using RAxML-HPC2 version 8.2.9 (Stamatakis 2014) and MrBayes version 3.2.2 (Ronquist et al. 2012) as implemented on the CIPRES Science Gateway (Miller et al. 2010), respectively. The ML analysis was performed under the GTRGAMMA model with the bootstrap iterations (-# | -N) set to 1000. The BI analysis was executed with most of the default parameters, but manually setting the following: the best substitution types (Nst) and rate distribution models (rates) that were determined by the jModelTest version 2.1.7 (Darriba et al. 2012), sampling one tree every 3000 generations for 100 million generations, stop early if the convergence diagnostic falls below the stop value 0.001 and show tree probabilities on the 50% majority-rule consensus tree with simple output format.
The ingroup (Polygonaceae) is well supported as monophyletic ( Fig. 2; BS = 100%, PP = 1.00; all support values follow this order hereafter). Within Polygonaceae, the first branch, represented by Symmeria paniculata Benth., is Symmerioideae which is Maximum Likelihood phylogram of Polygonaceae as inferred from analysis of the combined cpDNA dataset of matK, rbcL and trnL-F. Support values ≥ 50% BS or 0.90 PP are displayed above the branches, respectively. The tribal classification of Eriogonoideae followed Sanchez and Kron (2008) and Kempton (2012) and that of Polygoniodeae followed Schuster et al. (2015). The green star indicates the position of Harpagocarpus snowdenii. of Pterostegieae is shown to be nested within Eriogoneae in our analyses. In Polygonoideae, all seven tribes are fully supported as monophyletic (Fig. 2) with Persicarieae, Oxygoneae, Fagopyreae, Pteroxygoneae, Calligoneae and Rumiceae successively sister to the rest. With the inclusion of Harpagocarpus, Fagopyrum obtained high support values ( Fig. 2; 100, 1.00).

Phylogenetic analyses of Fagopyrum
The combined dataset D2 has 31 aligned sequences and comprises 6378 characters (1425 bp for accD, 2278 bp for matK, 513 bp for psbA-trnH, 1278 bp for rbcL and 883 bp for trnL-F), of which 735 are variable (11.52%) and 428 are parsimony-informative (6.71%). ML and BI trees generated from the D2 dataset yielded similar topologies (Suppl. material 1: figs S3, S4); thus, only the ML tree is shown (Fig. 3). In both of the analyses, the monophyly of Fagopyrum was strongly supported and three monophyletic subclades were recovered: the first subclade comprises Harpagocarpus is distinct from all other genera of Polygonaceae in having a remarkable fruit that has long setae growing along the ribs with the radially arranged retrorse barbs at the tip of each seta (Hutchinson and Dandy 1926) (Fig. 1). The current study is the first to include the genus Harpagocarpus in molecular phylogenetics and it demonstrates in all analyses, based on the cpDNA datasets D1 and D2, that H. snowdenii, the sole species of the genus, is deeply nested within Fagopyrum (Figs 2, 3). The additional molecular evidence undoubtedly confirms the previous hypothesis that Harpagocarpus and Fagopyrum are congeneric in respect of morphological, anatomical and palynological investigations (Hong 1988; Ronse Decraene and Akeroyd 1988; Hong et al. 1998).
Morphologically, H. snowdenii has sagittate to ovate-triangular leaf blades, seven palmate veins, large and clearly exerted fruits from the persistent tepals and broad cotyledons, which are very similar to those traits presented in Fagopyrum species (Hutchinson and Dandy 1926). Anatomical studies showed that H. snowdenii and Fagopyrum species share a series of similar floral characteristics, such as nectaries which are present as receptacular mamillae behind the stamens, inner stamens which are always linked with two lateral nectaries and cells of the inner tepal epidermis which are rectangular Figure 3. Maximum Likelihood phylogram of Fagopyrum as inferred from analysis of the combined cpDNA dataset of accD, matK, psbA-trnH, rbcL and trnL-F. Support values ≥ 50% BS or 0.90 PP are displayed above the branches, respectively. The two black boxes covered the cymosum group and the urophyllum group as defined in Yasui and Ohnishi (1998a), respectively. The green box indicates the third group, namely the tibeticum group, as revealed in the present study. The representative photos in the circles showing the inflorescences and the fruits of urophyllum group, tibeticum group and cymosum group are F. urophyllum, F. tibeticum and F. dibotrys, respectively. All photos were taken by Bo Li. to elongated (Hong 1988;Ronse Decraene and Akeroyd 1988). Palynologically, Ronse Decraene and Akeroyd (1988) emphasised that they observed an identical pollen structure between Harpagocarpus and Fagopyrum and Hong (1988) further noted that it is hardly possible to find any pollen morphological differences between Harpagocarpus and the species of Fagopyrum. The pollen of Fagopyrum is ovate, tricolpate with narrow furrows and a reticulate surface pattern. The pollen grains of Harpagocarpus are slightly smaller than those of Fagopyrum species, but they undoubtedly belong to the same pollen type (Hong 1988;Ronse Decraene and Akeroyd 1988). Considering this evidence, Ronse Decraene and Akeroyd (1988) suggested that H. snowdenii should probably be included within Fagopyrum, perhaps as a separate section and Hong (1988) formally combined H. snowdenii as F. snowdenii.
It is noteworthy to point out that our molecular analyses not only supported the amalgamation of Harpagocarpus with Fagopyrum, but also clarified the accurate specific relationships of F. snowdenii within Fagopyrum, which was stably supported to be a sister of F. tibeticum using cpDNA sequences (Figs 2, 3). F. tibeticum was originally described in the monotypic genus Parapteropyrum A.J.Li as P. tibeticum A.J.Li, which is a shrub, endemic to the central Qinghai-Tibetan Plateau of China and is characterised by having fascicled leaves, terminal raceme-like inflorescences, five unequally lobed tepals with the outer two smaller, perianth persistent and accrescent in fruit, three free styles with capitate stigmas and trigonous achenes with broad wings along ribs (Li 1981). P. tibeticum was considered to be most similar to Pteropyrum Jaub. & Spach in gross morphology (Li 1981), but surprisingly tested to be a member of Fagopyrum in molecular phylogenetic studies Tavakkoli et al. 2010;Tian et al. 2011) and, thus, formally combined in Fagopyrum as F. tibeticum . The origin of the woody F. tibeticum was thought to be a consequence of the large-scale uplift of the Qinghai-Tibetan Plateau which not only promoted continental species radiation, but also the secondary feature of woodiness in a few herbaceous lineages in response to strong selection pressures (Tian et al. 2011).
The inclusion of F. tibeticum in Fagopyrum has updated our knowledge of morphology in the genus, but now, the sister relationships between F. snowdenii and F. tibeticum, revealed in our molecular analyses, would not only further expand the morphological variation of Fagopyrum, but also shed light on the thinking of the biogeographical origin of the genus, because F. snowdenii is the only species of Fagopyrum distributed in Africa, while all other congeneric taxa occur mainly in East Asia. Jacques-Félix (1946) suggested that Fagopyrum perhaps entered Africa via a Middle Asian pathway during the Quaternary-periglacial period, just like other genera with both Afromontane and Central Asian representatives, such as Cicer L. and Colutea L. (Chapman andWhite 1970). However, de Klerk et al. (2015) stated that long-distance transport of pollen grains of F. snowdenii from Asia to Africa seems unlikely, but alternatively, they found out there are indications from pollen and macrofossils that a wild Fagopyrum ancestor may have been widespread in western Eurasia during the Late Tertiary and the Pleistocene Ice-Ages and became extinct afterwards. F. snowdenii may represent the only surviving African lineage that split from the wild widespread Fagopyrum ancestor.

Infrageneric relationships within Fagopyrum
Fagopyrum is a small genus comprised of ca. 25 species according to the most updated classification (Ohsako and Li 2020). The genus is economically important and well known for containing two domesticated common buckwheat, i.e. F. esculentum and F. tataricum which have been widely cultivated in Australia, Asia, Europe and North America for producing gluten-free grains (Li and Hong 2003). Geographically, most of the wild species of Fagopyrum are mainly distributed in mountainous regions of southwest China, a few are endemic to the south-eastern edge of the Qinghai-Tibetan Plateau (Ohnishi and Matsuoka 1996;Ohnishi 1998;Li and Hong 2003) and only the F. snowdenii confirmed in the present study is occurring in Africa (Hutchinson and Dandy 1926;Ayodele 2003). Eastern Tibet to western Sichuan of China was indicated to be the birthplace of the two cultivated common buckwheat in the AFLP (amplified fragment length polymorphism) analysis (Konishi et al. 2005). Taxonomically, Fagopyrum was separated from the large and heterogenous Linnaeus's genus Polygonum L. (Miller 1754) and has long been treated as a section of Polygonum (e.g. Meisner (1856); Samuelsson (1929); Steward (1930)) or considered to be an independent genus, but closely related to Polygonum (e.g. Dammer (1894); Gross (1913); Hedberg (1946); Haraldson (1978); Ronse Decraene and Akeroyd (1988)). In the context of the molecular phylogeny of Polygonaceae, Fagopyrum was not only supported as a monophyletic genus, but also indicated to represent an isolated tribal clade in the subfamily Polygonoideae Schuster et al. 2015). Morphologically and anatomically, Fagopyrum could be distinguished from other genera of Polygonoideae by having large conduplicate cotyledons and/or embryos in the central region in achene (Dammer 1894;Gross 1913;Nakai 1926;Chapman and White 1970;. Within Fagopyrum, two groups have been recognised in classical taxonomy, based on the morphology of inflorescence and the achene size: one group was mainly represented by F. cymosum (Trevir.) Meisn. (= F. dibotrys), F. esculentum and F. tataricum and characterised by having corymbose inflorescences with many branching and dense flowers and the achene greatly exceeding the perianth, while the other group is composed of other species (including F. urophyllum (Bureau & Franch.) H.Gross) having racemelike inflorescences with sparse flowers and the achene completely enclosed in perianth (Gross 1913;Roberty and Vautier 1964;Ohnishi and Matsuoka 1996) (Fig. 3). These two groups are mostly concordant with the cymosum group and the urophyllum group defined by Yasui and Ohnishi (1998a) in molecular phylogenetic analyses using DNA sequences of the nuclear internal transcribed spacer (nrITS) and cpDNA region rbcL-accD. Other molecular studies, no matter using isozyme variability and RFLP (Ohnishi and Matsuoka 1996), cpDNA sequences (Yasui et al. 1998;Ohsako et al. 2001;Jin et al. 2018), nuclear genes or regions (Yasui and Ohnishi 1998b;Nishimoto et al. 2003) and complete plastomes (Fan et al. 2021;Li et al. 2022), all clearly indicated that the cymosum group and the urophyllum group are both monophyletic clades.
In our present analyses, the above-mentioned two clades were recovered too, but the third clade, formed by F. snowdenii and F. tibeticum, was discovered, which is sister to the 'Urophyllum' clade (Fig. 3). We failed to generate any nuclear sequences from the specimen sample of F. snowdenii; thus, we could not test the sister relationships between F. snowdenii and F. tibeticum, as well as the sister relationships between F. snowdenii +F. tibeticum clade and the 'Urophyllum' clade in nuclear analysis. However, when only F. tibeticum was included in the ITS analysis, the topology of the phylogenetic tree is similar to that yielded from the combined cpDNA dataset, in which F. tibeticum is sister to the 'Urophyllum' clade clade (Tian et al. 2011). Considering the sister relationships between F. snowdenii and F. tibeticum could be additionally supported by morphological and palynological evidence, such as racemelike inflorescences, unequal tepals with the outer two smaller, perianth accrescent in fruit, large achenes greatly exceeding the perianth, special appurtenances (either wings or setae) growing along the fruit ribs and smaller pollen grains than the other Fagopyrum species (Hutchinson and Dandy 1929;Ronse Decraene and Akeroyd 1988;Hong 1995), we believe that F. snowdenii and F. tibeticum represent a separate clade in Fagopyrum. Future analyses, based on more comprehensive sampling and using nuclear sequences data, may further confirm or update the infrageneric relationships of Fagopyrum as inferred in this study. As far as the current results are concerned, a sectional classification for Fagopyrum is here proposed, based on the differentiation of gross morphology in the three clades, which is the first infrageneric classification of the genus.