We examined plant material of F. esculentum, F. homotropicum, F. tataricum and F. urophyllum cultivated in the Skolkovo Institute of Science and Technology (Moscow, Russia) and the All-Russia Research Institute of Grain Legumes and Groat Crops (Orel, Russia). At least 15 individual plants of each accession were examined. The following cultivars of F. esculentum subsp. esculentum were studied: Batyr, Chatyr Tau, Dasha, Dozhdik, and Temp. In addition, an accession of the wild subspecies F. esculentum subsp. ancestrale was examined (C9015, seeds received from the collection of Kyoto University, originally from Yongsheng district of Yunnan province, China) as well as an accession derived from its cross with cv. Dasha (FAD_F2_2018). Three accessions of F. tataricum were used: K17 is a cultivated line (received from the collection of Federal Research Center N.I. Vavilov All-Russian Institute of Plant Genetic Resources, St. Petersburg), C9119 is a wild accession from Lixian district of Sichuan province, China (received from Kyoto University) and Zhd001 comes from a seed collection made in a ruderal location near a railway in Moscow Province, Russia (55.587° N, 36.716° E). The accession of F. homotropicum examined here (С9139) comes from Yongsheng district of Yunnan province, China (received from Kyoto University). The accession of F. urophyllum (С9069) comes from southern China (received from Kyoto University).

Whole plants and young shoot tips were fixed in 70% ethanol. SEM work on material of the tlb mutant of F. esculentum was conducted at the Royal Botanic Garden, Kew (UK). The material was dissected in 70% ethanol, dehydrated through absolute ethanol and critical-point dried using an Autosamdri-815B CPD (Tousimis Research, Rockville, MD, USA). Material was mounted on SEM stubs, coated with platinum using an Emitech (Hailsham, East Sussex, UK) K550 sputter coater and examined using a Hitachi cold-field emission SEM S-4700-II (Hitachi, Japan). SEM work on material other than the tlb mutant was conducted at the Laboratory of Electron Microscopy at the Biological Faculty of Moscow University. The material was dissected in 70% ethanol and transferred to 100% acetone using the following series: 96% ethanol (twice for 30 min), 96% ethanol: 100% acetone (1:1 v/v, 30 min), 100% acetone (three times for 30 min). The material was critical-point dried using a Hitachi HCP-2 critical-point dryer (Hitachi, Japan), then coated with gold and palladium using a Eiko IB-3 ion-coater (Tokyo, Japan) and observed using a CamScan S-2 (Cambridge Instruments, London, UK) and a JSM-6380LA SEM (JEOL, Tokyo, Japan).

Specimens of F. tataricum (42 sheets) and F. cymosum (10 sheets) were examined in the Herbarium of Moscow University (MW) and the Herbarium of the Main Botanic Garden of the Russian Academy of Sciences, Moscow (MHA). In addition, available online resources of herbarium collections were examined. Inflorescence morphology of F. urophyllum was examined using the Chinese Virtual Herbarium (https://www.cvh.ac.cn/index.php). Fragments of herbarium material of F. tataricum were studied using high-resolution X-ray computed tomography (HRXCT). Dried herbarium material was rehydrated by soaking in boiling water and, transferred into an infiltration medium (1% phosphotungstic acid [PTA] in 70% ethanol) for a week to increase its contrast (Staedler et al., 2013). The PTA solution was exchanged four times. Then the samples were transferred in to 100% acetone using the following series: 1% PTA in 96% ethanol (twice for 30 min), 1% PTA in 96% ethanol: 100% acetone (1:1 v/v, 30 min), 1% PTA 100% acetone (three times for 30 min). The material was critical point-dried using a Hitachi HCP-2 critical point dryer (Hitachi, Japan) and then individually mounted to brass sample holder using carbon adhesive disc. The scans were performed on a SkyScan 1272 microtomograph (Bruker, Billerica, USA) equipped with a Hamamatsu L10101-67 source (Hamamatsu Photonics, Hamamatsu, Japan) and a Ximea xiRAY16 camera (Ximea GmbH, Münster, Germany) at the Biological Faculty of Moscow University. Source voltage and current were set to 35 kV and 175 μA, respectively, while exposure time was set to 550 ms, and no X-ray filter was used. The sample was rotated 180° around the vertical axis with a rotation step of 0.07. One tomographic acquisition was performed, resulting in an image stack of 2669 reconstructed micro-CT slices at 2 μm pixel size. Visualisation from the scanning data was performed using CTVox (Bruker, Billerica, USA). Images were manually coloured to show different phyllomes.

Comparative and evolutionary studies of angiosperm inflorescence morphology are hindered by complexity and inconsistent use of descriptive terminology (Kusnetzova, 1988; Weberling, 1989; Bell and Bryan, 2008; Prenner et al., 2009; Endress, 2010). Unfortunately, available descriptions of inflorescence characters of buckwheat, either in taxonomic literature or publications on breeding, cannot be readily converted into homology-based terminology. Here, we attempt to clarify the use of basic descriptive terms with respect to buckwheat, in order to highlight the organ homologies.

In Fagopyrum and many other Polygonaceae, the thyrse represents an important inflorescence module. In a thyrse, the primary axis never terminates in a flower and there are lateral cymes with successively developing flowers of several orders. Each cyme is located in the axil of a cyme-subtending bract, which is also the subtending bract of the first flower (Figure 1). With very few exceptions (see below), cyme-subtending bracts are scale-like green phyllomes with no traces of even a reduced leaf lamina. Cyme-subtending bracts are spirally arranged along the thyrse axis (Figure 2A). Thyrses with clockwise and anticlockwise spirals are both found, without any noticeable regularity. Each flower of a cyme is enclosed together with the next-order flower by a thin, membranous and transparent bilobed sheath interpreted here as two congenitally fused prophylls (α-prophyll and β-prophyll). We use the term ‘prophyllar sheath’ for this structure. The β-prophyll is a subtending bract of the next-order flower of the cyme (Figure 1). The position of the next-order flower is intermediate between transversal and abaxial relative to the cyme-subtending bract (i.e. it is closer to the subtending bract than to the thyrse axis). The next-order flower and its prophyllar sheath repeat the morphology of those of the first order but are mirror-shaped. There is a change of handedness with every next order of branching in the cyme. As a result, successive flowers of a cyme form a zig-zag pattern (Figures 1, 3P). The cyme of Fagopyrum can be classified as a monochasium because branching always takes place in the axil of a single prophyll, and more precisely as a cincinnus because of the zig-zag pattern of flower arrangement. Two mirror-shaped types of cincinni can be recognized that differ in the left- or right-hand positions of the α- and β-prophylls of the first flower and consequently of the second flower. For example, the β-prophyll of the first flower and the second flower situated in its axil are in right-hand positions in Figures 1, 2J, K, 3C and in left-hand positions in Figures 2I, 3B.

Flowers of Fagopyrum share with several other Polygonaceae the occurrence of five tepals, five outer-whorl stamens, three inner-whorl stamens and three united carpels (Figure 1). There is a single ovule that is attached at the base of the unilocular ovary and shared by all three carpels. The tepals can be numbered according to their inferred sequence of (pre)patterning, though the sequence of their actual appearance is very rapid. This numbering agrees with tepal aestivation (i.e., the mode of their overlapping) that can be seen at the latest developmental stages (not shown). Tepals 1 and 2 are in transversal positions. Tepal 1 is in the same radius as the α-prophyll (Figure 1). Tepal 4 is in the adaxial (=posterior) part of the flower (the part that is closer to the thyrse axis or to the previous flower of the cyme). Tepals 3 and 5 are in the abaxial (anterior) part of the flower (the part that is closer to the subtending bract). Tepal 3 is closer to tepal 1 and tepal 5 is closer to tepal 2 (Figure 1). The next-order flower is located in the sector of tepal 5 (Figures 1, 3E, H, J, K, M). The five outer-whorl stamens could be viewed as alternating with the five tepals, but they are not equally spaced early in development. Two stamens adjacent to tepal 1 are closely spaced and form a pair. Similarly, there is another pair of stamens adjacent to tepal 2 (Figures 1, 3D, E, G-I). Of the three inner-whorl stamens, one is adaxial and two are transversal-abaxial. The inner-whorl stamens alternate in radii with the three carpels, of which one is abaxial and two are adaxial-transversal.

The following stages of early flower development can be recognized in F. esculentum. We found no differences in flower development among observed cultivars of the Common buckwheat (Dasha, Temp, Chatyr Tau, Batyr, Dozhdik) and its accession FAD_F2_2018.

Stage 1 (Figures 2A, B). Flower primordium. The primordium is elliptic in outline. There is no evidence of prophyll or tepal initiation at this stage.

Stage 2 (Figures 2C–E). Prophyll(s) are initiated, but tepals cannot be recognized. It is likely that the α-prophyll initiates before the β-prophyll, but Figures 2C, D cannot be taken as robust proof of this statement, because the β-prophyll may be hidden in these views.

Stage 3 (Figures 2F–H). The two prophylls are clearly united into a sheathing tube by congenital fusion. Tepals 1 and 2 are initiated. There is no evidence of stamen initiation at this stage.

Stage 4 (Figures 2I–L). Tepal 3 is initiated. Outer-whorl stamens initiate in pairs (sometimes as common primordia)? in sectors of tepals 1 and 2. The fifth outer-whorl stamen cannot be traced at this stage.

Stage 5 (Figures 3A–C). All tepals and all outer-whorl stamens are initiated. Inner-whorl stamens are present, though one of them may be yet weakly defined (the one that is closer to the next-order flower, Figure 3C). Therefore, it is likely that the inner-whorl stamens initiate in a rapid sequence. No evidence of gynoecium initiation so far.

Stage 6 (Figures 3D–F). The gynoecium can be recognized as a dome-shaped primordium between the three inner-whorl stamens. Stamens, especially those of the inner whorl, are larger than in Stage 5, but thecal structure is not yet pronounced. At this stage, the next-order flower cannot be observed without dissection of the prophyllar sheath.

Stage 7 (Figures 3G–I). The gynoecium is triangular in top view. Thecal structure can be recognized at least in the median-adaxial inner-whorl stamen (the only stamen unlabelled in Figures 3G–I).

Stage 8 (Figures 3J, K). The ovule can be seen at this stage, surrounded by the shallow rim of the gynoecium wall with the tips of the three carpels. Thecae are well-pronounced in all stamens. The inner-whorl stamens are longer than the outer-whorl stamens and possess short filaments. Their anther bases are below the level of the gynoecium rim.

Stage 9 (Figures 3L–O). The gynoecium rim has increased in height and the carpel tips are facing upwards and exceed the ovule. The inner-whorl stamens much exceed the outer-whorl ones. By filament elongation, their anthers extend above the gynoecium.

The mutant differs from the wild type in the tepal-like nature of the cyme prophylls. In particular, conical cells can be found on the abaxial surface of the prophylls. However, expression of the tepal-like features is unstable; in the cyme illustrated in Figure 4J, the α prophyll has conical cells but the β prophyll lacks them. The two prophylls that are always united to form a sheath in the wild type tend to be separate in the mutant, though the degree of separation varies; the two prophylls can be completely free from each other (Figure 4K), very shortly united at both sides (Figures 4I, J), united at the abaxial side only (Figures 4E, F) or united to form a proper sheath (Figure 4D). There is a tendency for elongation of axes of all orders below and between the prophylls. When the prophylls are united, this causes the cyme base to be crooked (Figure 4H). Our data on early flower development showed great similarity to the wild type (Figures 4A-F). Organ number and their positions are the same as in the wild type. Developmental data suggest the sporadic development of the α and β prophylls as two free organs rather than the sporadic appearance of yet another phyllome in the cymes. The two almost-free phyllomes visible in Figure 4B can be readily identified as α and β prophylls.

All species examined possessed thyrses of similar structure as described above for F. esculentum. Cyme-subtending phyllomes were scale-like, unless directly specified below.

The basic inflorescence module of F. urophyllum is a bracteose thyrse. Some specimens in collections of F. urophyllum possess inflorescences of the same type as described above for F. cymosum (for example, BNU 0026870, BNU 0026869, PE 01859341, PE 01859340, see https://www.cvh.ac.cn/). The main axis apparently has foliage leaves only (frondose structure), though the uppermost leaves are retarded in development. Structures in the axils of these leaves are groups of two or three thyrses. Branching on these second-order axes takes place in the axils of scale-like leaves (bracteose structure). The main inflorescence axis lacks a terminal thyrse, at least at developmental stages available for investigation, but the second-order inflorescence axes always possess a terminal thyrse (Figure 15B). The accession of F. urophyllum grown for the present study (Supplementary Figure 3) was similar to that described above, but tended to produce groups of more than three bracteose thyrses in the axils of the foliage leaves of the main axis. Branching in these groups of thyrses was either bracteose or frondo-bracteose.

Many herbarium specimens of F. urophyllum possess inflorescences with the axis terminated by a thyrse (Figure 15A). There are also lateral thyrses situated in the axils of the leaves of the main axis below the terminal thyrse. Lower lateral thyrses develop in the axils of foliage leaves, but upper ones are subtended by scale-like leaves (frondo-bracteose inflorescence structure). Third-order thyrses are normally absent. The terminal thyrse is usually longer than the uppermost lateral thyrse and equals the lower lateral thyrses (for example, PE 01062852, https://www.cvh.ac.cn/). Rarely, the terminal thyrse is short and, in contrast with most other specimens, anthesis is delayed in terminal and uppermost lateral thyrses as compared to the long lower lateral thyrses (PE 01859339). The terminal and lateral thyrses form what can be interpreted as flowering unit. Some specimens show lateral units of this sort that can be interpreted as developing on paracladia (for example, there is one paracladium in PE 00167717). Since F. urophyllum is a perennial and often robust plant, the complete branching pattern not always can be inferred from the herbarium material. Observations in nature are necessary to obtain a complete picture.

In cymes of Fagopyrum (tribe Fagopyreae) and many other Polygonaceae (members of tribes Polygoneae and Persicarieae, Sanchez et al., 2011), our data support the hypothesis that the bilobed sheath associated with each flower is composed of two congenitally fused prophylls, one of which subtends the next-order flower (Payer, 1857; Gross, 1913; Bauer, 1922; Galle, 1977; Ronse De Craene, 2022). This issue is far from trivial. The two prophylls (α and β) are most likely initiated as two distinct primordia, but their common sheathing part becomes conspicuous at very early developmental stages (Sattler, 1973). Photographs provided by Sattler (1973) are not fully convincing in the free nature of the two prophylls at their initiation, because their common tubular part could be hidden by the flower-subtending bract and main inflorescence axis. In a detailed study of Bistorta officinalis Delarbre (=Polygonum bistorta L.), Schumann (1890) disagreed with the double nature of the prophyllar sheath at its initiation and suggested that it is formed by a single prophyll. He contended that in rare instances when two prophylls are formed, they are free from each other from initiation, each subtending a next-order flower. He criticized Payer’s (1857) illustrations of early developmental stages in Fagopyrum cymosum, and suggested the occurrence of a single phyllome that forms the prophyllar sheath. Our Figure 2C shows convincing evidence of a developmental stage preceding the formation of the tubular sheath encircling the floral apex, at least on its adaxial side. Note however that a shallow abaxial connection can be already recognized between the two lobes (α and β) in a rather young developmental stage shown in Figure 2E.

In theory, initiation as two free primordia, taken in isolation from all other arguments, cannot be regarded as a necessary or sufficient condition of recognition of two congenitally united organs. In so-called early congenital fusion, free parts of united organs become visible after the initiation of their common part (Sokoloff et al., 2018). There are even cases when one and the same organ initiates as two distinct primordia. For example, in the flower of Groenlandia densa (L.) Fourr. (Potamogetonaceae), each lateral stamen arises as two separate primordia, whereas each median stamen initiates as a single primordium (Posluszny and Sattler, 1973).

Yurtseva (2006) re-interpreted the two lobes of the prophyllar sheath in cymes of Fagopyreae, Polygoneae and Persicarieae as two stipules of one and the same prophyll. Indeed, the ocrea of Polygonaceae can be viewed as a fusion-product of the two stipules and in the absence of a leaf blade one could easily imagine a sheathing phyllome composed of a pair of stipules. In contrast with Schumann (1890); Yurtseva (2006) concluded that the next-order flower is not subtended by the prophyll, but its subtending phyllome is completely suppressed. According to this interpretation, the single bilobed prophyll has an adaxial position relative to the flower-subtending bract (which is a cyme-subtending bract in the case of the first flower of a cyme). The two lobes (inferred stipules of the prophyll) are located in almost transversal positions (Yurtseva, 2006). The empirically observed position of the next-order flower is towards the abaxial (rather than adaxial) side of the first flower (Sattler, 1973; Yurtseva, 2006; present study). This represents the primary reason for not interpreting the next-order flower as occurring in the axil of a reportedly adaxial prophyll (Yurtseva, 2006).

Sattler’s (1973) diagram shows the two lobes of the prophyllar sheath as symmetrically arranged in transversal positions. Therefore, the next-order flower is out of the median of the β-lobe in Sattler’s diagram. Our data on the earliest developmental stages reveal differences in the shape and position of the lobes. The β-lobe is shifted towards the abaxial side rather than being strictly transversal; it is also wider than the α-lobe (Figure 2E). Some images taken at the mid stages of flower development still capture apparent differences in shape and position of the α- and β-lobe (for example, Figure 2L). Therefore, if the α- and β-lobes are assumed as two distinct prophylls, the position of the next-order flower can be safely interpreted as axillary with respect to the β-prophyll. Late development of the β-lobe is somewhat one-sided, so that its summit acquires a transversal position that fits Sattler’s diagram. Furthermore, our data on branching in the paracladia of F. esculentum provide examples of axillary bud displacement relative to the midline of its subtending leaf (for example, b2-3 and l2-3 in Figure 13C). An important argument against interpretation of the prophyllar sheath as formed by a single adaxial prophyll is the absence of adaxial prophylls in any other instances of branching in Fagopyrum. The two first phyllomes of all paracladia and the lateral thyrses are always in transversal positions. Therefore, it is logical to assume a transversal (rather than adaxial) position of the prophylls in the cymes.

Our interpretation of the prophyllar sheath as two fused prophylls agrees with other examples of pairwise leaf fusion in Fagopyrum. The most obvious example is the formation of the cotyledonary tube (Figure 14A). In all instances when we observed occasional whorled (opposite) leaf arrangement of foliage leaves in Fagopyrum, these leaves were congenitally united to form a sheathing tube (Figures 14B, C: at the first node of the main axis above the cotyledons; Figure 14D: at the first node of a lateral axis).

Plants of the tlb mutant of F. esculentum sometimes develop two free phyllomes associated with each flower, of which one phyllome clearly subtends a next-order flower (Fesenko et al., 2005; Yurtseva, 2006; Logacheva et al., 2008; present study). Yurtseva (2006) interpreted the sporadic occurrence of the two phyllomes as evidence for regain of the reportedly suppressed second prophyll that subtends the next-order flower. Our data show that the two occasionally free phyllomes in the tlb mutant are homologous to the two lobes of the prophyllar sheath found in wild-type plants. The homology is supported by the conserved relative positions of the two phyllomes/lobes relative to tepals 1–5 of the first-order flower of the cyme. Even more important, our data show a series of transitional forms between entirely free and strongly fused prophylls, including the condition of one-sided fusion (Figures 4E, F). The main morphogenetic effect of the tlb mutation is the acquisition of tepal-like features by the prophyllar sheath (Fesenko et al., 2005; Logacheva et al., 2008). Tepals of F. esculentum differ from prophylls in the occurrence of epidermal papillae, white or pink colouring and a narrow base. Tepals of Fagopyrum are largely free from each other, although located at the same level on the receptacle. The tendency for splitting of the prophyllar sheath into two distinct structures should be viewed as a manifestation of its tepal-like features. Wild-type plants of F. esculentum have a long internode (pedicel) below the tepals, but no stalk below the prophyllar sheath. With the acquisition of tepal-like features, tlb plants tend to elongate the stem region below the prophyllar sheath and between the two prophylls. Sometimes a mosaic of features can be seen in mutant plants, including a situation when stem elongation takes place between two still fused prophylls, thus producing a distorted prophyllar sheath (Figure 4H). To summarize, the data on the tlb mutant support rather than contradict our preferred interpretation of the prophyllar sheath as congenitally fused α- and β-prophylls.

The occurrence of two transversal prophylls in cymes of Polygonaceae is entirely consistent with the presence of the same condition in the sister family, Plumbaginaceae (Ronse De Craene, 2022).

Flowers of the tribes Fagopyreae, most Polygoneae and Persicarieae possess five tepals arranged with quincuncial aestivation, in which two tepals are completely outside the other tepals in the floral bud, two are completely inside and one tepal is intermediate (Ronse De Craene, 2022). The intermediate tepal has one margin covered by the adjacent outer tepal and the other margin covering the adjacent inner tepal. Some other Polygonaceae, such as Rumex and Rheum, possess a monocot-like condition of three outer plus three inner tepals, so the morphological interpretation of flowers with five tepals, as in Fagopyrum, Polygoneae and Persicarieae, has been extensively discussed. Assuming that the condition in Rumex and Rheum is trimerous and two-whorled, flowers with five tepals could be regarded as two-whorled and intermediate between dimerous and trimerous (2½-merous, Yurtseva and Choob, 2005). Then the intermediate tepal could be interpreted as belonging partly to the outer and partly to the inner whorl. However, phylogenetic placement of Polygonaceae among taxa with commonly pentamerous flowers suggests that flowers of Fagopyrum are pentamerous and one-whorled with quincuncial aestivation, a condition found in many other eudicots belonging to various orders and families (Ronse De Craene, 2022).

Eudicots with quincuncial aestivation of tepals or sepals often display sequential organ initiation in the perianth or calyx that could be interpreted as a 2/5 spiral. Two outer organs are initiated first, followed by the intermediate organ and the inner organs are the last ones to be initiated. This pattern has been most commonly reported for members of Polygonaceae with five tepals (Payer, 1857; Bauer, 1922; Sattler, 1973; Galle, 1977), but not properly documented using SEM. Schumann (1890) highlighted technical difficulties in making observations on the sequence of floral organ initiation in Polygonaceae and emphasised the need for theory-neutral descriptions. He concluded that the sequence of tepal initiation in Bistorta officinalis and apparently other Polygonaceae does not follow a 2/5 spiral. According to his data, following initiation of the two outer tepals in transversal positions, the third tepal to initiate is the (inner) adaxial one, followed by the two abaxial tepals (see also Yurtseva, 2000). However, our data on F. esculentum as well as those of Sattler (1973) do not support the ideas of Schumann (1890). After initiation of the two transversal tepals, the third one is initiated in an abaxial and not adaxial position (Sattler, 1973; present study). Our data do not contradict the idea of tepal initiation along a 2/5 spiral, but we were unable to find precise evidence for sequential initiation of tepals 1 and 2 as well as tepals 3 and 4. We believe that the epi-illumination light microscopy photographs of Sattler (1973), like our SEM images, do not show a stage with only one tepal initiated and a stage with only four tepals initiated. Thus, our numbering of tepals is based partly on theoretical grounds. Tepal 1 appears to be larger than tepal 2 in Figures 2F and 2H, so that their initiation in a very rapid sequence remains a plausible possibility.

Even though the initiation of the five outermost floral organs along a 2/5 spiral in Fagopyrum is shared with many other eudicots, the relative arrangement of the tepals and floral prophylls (bracteoles) in buckwheat and other Polygonaceae merits special attention. Eichler (1875) proposed that the arrangement of the outer floral organs (such as sepals) is to a large extent determined by the position of the floral prophylls. When five sepals (or tepals) have sequential patterning, the sequence should be initiated by the floral prophylls. The first sepal thus normally appears either in an adaxial-transversal or in an abaxial-transversal position, in both instances closer to the α-prophyll than to the β-prophyll (see Figure 13 in Eichler, 1875). The first tepal of Fagopyrum is located in almost exactly the same radius as the α-prophyll, a condition that differs from the two possibilities predicted by Eichler (1875). Note that Engler’s (1875) diagrams of Bistorta officinalis and Coccoloba nitida Kunth (C. guianensis Meisn.) do not show a superposition of the α-prophyll and the first of the five tepals. In this respect, these diagrams differ from our diagram of F. esculentum and the diagram of Persicaria lapathifolia in Ronse De Craene (2022). Further studies will determine if any variation in tepal positions occurs in Polygonaceae.

The non-trivial superposed position of the first tepal and the α-prophyll could be related to the asymmetric position of the two prophylls in Fagopyrum, though the prophylls of Persicaria were illustrated as symmetric by Ronse De Craene (2022). Another peculiar feature of Fagopyrum and other Polygonaceae is the shape of the floral primordium and the young flower, which is pronouncedly elliptical rather than circular in outline. Some images even suggest a crescent-shaped outline of the young flower tightly packed between the cyme-subtending bract and the thyrse axis. Refined mathematical modelling of early flower development in Polygonaceae may shed new light on the problem, especially when considering aspects of pre-patterning as well as mechanical pressure and the peculiar shape of the floral apex (Choob and Yurtseva, 2007; Ronse De Craene, 2018; Bull-Hereñu et al., 2022).

In the framework of the inhibitory field theory, the superposed position of the α-prophyll and the first tepal could be conditioned by an inhibitory influence from the β-prophyll that much exceeds that of the α-prophyll. Such differences between the two prophylls would be intriguing, because they are located at the same level and fuse with each other to form the prophyllar sheath. In all other instances of pairwise phyllome fusion observed in the present study (between cotyledons or vegetative foliage leaves), no evidence was found for similar superposition with the next-formed organs. For example, in Figure 14D, the two first leaves of a shoot are basally united and the third leaf occupies a position between their radii. It seems that in contrast to the situation shown in Figure 14D, the α- and β-prophylls fuse despite their sequential pre-patterning.

According to Sattler (1973), immediately after inception of the fifth tepal primordium (or even simultaneously with it) a pair of outer-whorl stamen primordia is initiated opposite each of the first two tepal primordia in F. esculentum. According to our data, the first evidence of initiation of these stamens can be seen at even earlier stages, before initiation of the fourth and fifth tepal (Figure 2J). The timing of initiation of the outer stamens found in the present study agrees with observations in members of Polygoneae and Persicarieae (Galle, 1977). Leins and Erbar (2010) summarized other examples of asynchronous organ initiation in different floral sectors and highlighted the possible sectorial differentiation in activity of genetic regulatory networks determining organ identity.

Our data on inflorescence morphology in Fagopyrum fit well with the concept of flowering unit (Maresquelle, 1970; Sell, 1976; Kusnetzova, 1988; Acosta et al., 2009). As in other angiosperms with a pronounced flowering unit, in Fagopyrum the lower boundary of the terminal flowering unit is determined by an abrupt change in fate and speed of development of the axillary branches along the main axis. Lateral axes belonging to the terminal flowering unit show an acropetal or almost simultaneous development, whereas the paracladia situated below the terminal flowering unit are developmentally retarded and tend to have a basipetal sequence of flowering. It is remarkable that Fesenko et al. (2004), using another terminology, provided a diagram of plant architecture in buckwheat that perfectly fits the concept of the terminal flowering unit and paracladia of various orders bearing additional flowering units. Yurtseva (2006) and Kostina and Yurtseva (2021) used the concept of flowering unit to describe inflorescence diversity at the level of Polygonaceae and in the genus Atraphaxis.

As highlighted by Sell (1976), the flowering unit provides taxon-specific characters of flower arrangement and represents a simplest possible pattern of inflorescence structure in a given species. Indeed, the occurrence of paracladia is somewhat optional and partly determined by environmental conditions, whereas the terminal flowering unit does develop in any plant that has achieved transition to anthesis. Of course, the degree of development of the paracladia of the first and subsequent orders is not solely determined by the environment, but depends on an interplay with the genetic background. For example, the wild Common buckwheat (F. esculentum subsp. ancestrale) develops more paracladia than cultivars of F. esculentum subsp. esculentum in common garden experiments (Fesenko et al., 2004). Some cultivars tend to develop the terminal flowering unit only. The present study showed that paracladia do initiate in the axils of all foliage leaves below the terminal flowering unit, even if they are subsequently arrested in development.

Comparisons at the level of the flowering unit provide a clear summary of inflorescence diversity in Fagopyrum. The following characters can be proposed. 1. Presence (Figures 15D, F–H) or absence (Figure 15E) of a terminal thyrse. 2. Presence (Figures 15A–G) or absence (Figure 15E) of lateral thyrses. 3. Presence (Figures 15B–D) or absence (Figures 15A, E–G) of bracteose branching of the stalks of lateral thyrses. 4. Bracteose (Figures 15A, D, F) or frondo-bracteose (Figures 15G, H) structure of the terminal thyrse. 5. Frondose (Figures 15B–G) or frondo-bracteose (Figure 15A) structure of the entire flowering unit (terminal thyrse not considered).

Our data show that the determinate growth pattern of cultivars Dozhdik, Temp and Dasha of F. esculentum is conditioned by the presence of the terminal thyrse in flowering units. Plants of F. esculentum with indeterminate growth (including those of ssp. ancestrale) as well as the closely related wild species F. homotropicum lack a terminal thyrse. These differences were already correctly described for F. esculentum by Fesenko (1968), though he incorrectly used the term ‘raceme’ for thyrse.

The present study highlights, apparently for the first time, important differences between wild accessions of F. tataricum and F. esculentum, with respect to the presence vs. absence of a terminal thyrse. Thus, our data suggest different evolutionary patterns in the two species. Selection towards fixation of the determinate growth pattern conditioned by the occurrence of a terminal thyrse took place in domesticated F. esculentum. In contrast, the cultivated accession of F. tataricum studied here (K17) is characterized by a delayed ontogenetic transition to terminal thyrse formation, compared with other accessions of this species. In theory, further heterochronic shifts in this direction may lead to permanent elimination of the terminal thyrse. It is tempting to suggest heritability of the striking differences in the timing of ontogenetic transition to terminal thyrse observed here among accessions of F. tataricum. Indeed, in a common garden experiment, at the stage when terminal thyrses were fully formed after 10-11 nodes of the flowering unit in C9119, no indication of their initiation was visible in K17. Furthermore, the ruderal accession Zhd001 initiated a terminal thyrse as early as starting from the third node above the cotyledons. Importantly, the occurrence of a terminal thyrse does not guarantee a determinate growth pattern in F. tataricum. The plant of F. tataricum C9119 illustrated in Figure 7 has a terminal thyrse, but does not meet the conditions of the determinate growth pattern. Note that the fruits are shed in lower thyrses of the terminal flowering unit whereas the upper ones are still in anthesis and fruit development. The accession K17 is even more problematic in this respect because in our experiments it developed at least 19 nodes with lateral thyrses within the flowering unit before formation of the terminal thyrse.

The concept of determinate growth pattern is not equivalent to the concept of determinate inflorescences. The former is related to the applied problem of more synchronized flowering/fruit set and early cessation of inflorescence growth, whereas determinate inflorescences are defined morphologically as those bearing a terminal flower (Weberling, 1989). The basic inflorescence unit of Polygonaceae is a thyrse that by definition lacks a terminal flower (Endress, 2010). Thyrsoids (which differ in the presence of a terminal flower, Endress, 2010) are extremely rare in Polygonoideae (Yurtseva, 2006) and not recorded in Fagopyrum. Therefore, determinate inflorescences do not occur in this group. The presence versus absence of the terminal thyrse is related to the concept of axiality, which is a more general concept than that of determinate/indeterminate inflorescences (Weberling, 1989). Axiality measures the minimum number of axis orders formed before the axes terminate in flowers (Weberling, 1989). Thus, plants bearing a terminal thyrse are diaxial and those with only lateral thyrses are triaxial. Use of the concept of axiality much simplifies descriptions of branching patterns and allows large-scale comparisons. It can be employed even in angiosperm-wide data sets. Axiality has been successfully employed in genus-level studies of various angiosperms (Notov and Kusnetzova, 2004; Sokoloff et al., 2015). Interestingly, the phenotypic effects of mutants of TFL homologues in various angiosperms can be generalized as a decrease in the level of axiality. For example, wild-type Arabidopsis is diaxial, the tfl mutant is monoaxial; wild-type Pigeon Pea is 4-axial, the Dt1 mutant is 3-axial (Bradley et al., 1997; Singer et al., 1999; Teeri et al., 2006; Sinjushin, 2013; Benlloch et al., 2015; Saxena et al., 2017).

As highlighted by Yurtseva (2006), thyrses of various Polygonaceae-Polygonoideae show either determinate or indeterminate growth patterns despite possessing the same basic groundplan of flower arrangement. Our study revealed that the terminal thyrse is extremely variable in its length and apparent duration of growth among examined material of F. tataricum. Some specimens had a long, frondo-bracteose terminal thyrse approaching an indeterminate condition. In F. tataricum, the morphological difference is striking between a morphotype with a long terminal frondo-bracteose thyrse and no lateral thyrses on the one hand, and one with a short terminal and many lateral thyrses, on the other. An environmental component in these differences should be considered, and the possibility of a trade-off between formation of lateral thyrses and lateral cymes, along with possible heritability of the observed differences (see Geber, 1990). One could suggest that cymes occurring in the axils of foliage leaves develop as a result of homeotic replacement of lateral thyrses, thus giving rise to a long frondo-bracteose terminal thyrse. However, testing this hypothesis is problematic. Another potential phenomenon is extension of the foliage leaf developmental program into the proximal region of the terminal thyrse. Indeed, the developmental conditions of the terminal thyrse differ from those of lateral thyrses in F. tataricum and F. esculentum because the terminal thyrse continues a shoot with foliage leaves. A tendency for foliage leaf development in the first node of the terminal thyrse was observed here in both F. tataricum and F. esculentum.

Kusnetzova (1988) developed the earlier ideas of Maresquelle (1970); Sell (1976) and other authors on the possible patterns of evolutionary transformation of angiosperm inflorescences. These ideas, especially the unidirectional nature of the proposed transformations, are sensitive to criticism that is beyond the scope of the present paper. More recent studies proposed other ideas on inflorescence evolution (e.g. Claßen-Bockhoff and Bull-Hereñu, 2013; Stützel and Trovó, 2013). However, an important point is the occurrence of certain ‘privileged’ patterns (Figures 15I–L) of arrangement of individual flowers and flower aggregations that are repeated across various angiosperm groups (Kusnetzova, 1988). Interestingly, the pattern of thyrse arrangement described here for Fagopyrum cymosum does not apparently fit any of these patterns. Yurtseva (2006) did not mention this pattern while discussing the inflorescence diversity in Polygonaceae. Therefore, it merits special attention. Fagopyrum cymosum has complex lateral aggregations of thyrses and no terminal thyrse. The complexity of these lateral aggregations increases towards the distal part of the entire inflorescence.

Inflorescences essentially similar to those of F. cymosum (Figure 15C) are found, along with other patterns, in three other species of the genus, F. urophyllum (Figure 15B, Supplementary Figure 3), F. tataricum (Figure 15D, though with a delayed development of the terminal thyrse) and F. esculentum (Figure 9C). This re-iteration could represent an example of Vavilov’s (1922) Law of Homologous Series in variation. The inflorescence variation found in F. urophyllum is noteworthy because the two patterns (Figures 15A, B) are morphologically contrasting and as far as can be inferred from herbarium material strongly differ in their frequencies. One of the conditions found in F. urophyllum (Figure 15A), is interesting in its frondo-bracteose structure: distal lateral thyrses are located in the axils of the scale-like leaves of the main inflorescence axis. This structure differs from members of the Cymosum group of Fagopyrum, where the main axis has foliage leaves, at least until the terminal thyrse, if one is present (Figures 15B–G). Field observations on patterns of flower arrangement in various localities of F. urophyllum are needed to get a complete picture of diversity of this species, which consists of two phylogenetically distinct lineages (Kawasaki and Ohnishi, 2006). Fagopyrum cymosum (like F. urophyllum) is a perennial species and the only perennial member of the Cymosum group. Its phylogenetic position does not preclude a plesiomorphic condition for (some) its morphological traits, including inflorescence morphology. If F. tataricum is derived from F. cymosum (Yasui and Ohnishi, 1998; Yamane et al., 2003; Ohnishi, 2016), the pattern found in F. cymosum could be plesiomorphic for F. tataricum. However, given the occurrence of infraspecific variation and complex evolutionary patterns in Fagopyrum that involve polyploidy, use of conventional approaches of character evolution are problematic here. It is more appropriate to develop studies in evolutionary developmental genetics of buckwheat in the context of microevolution, speciation, and natural and artificial selection.