The examined species of Burmannia are characterized by few- to several-flowered inflorescences. The above-ground shoot bears several bracts inserted following a 1/3 spiral, and ends in a terminal flower (Figures 1B–H). The outer tepals of the terminal flower alternate with the three uppermost bracts. The two uppermost bracts are placed close to the terminal flower, thus referred here to as involucral bracts; the other bracts (= stem leaves) are divided from them and from each other by long internodes. The involucral bracts are morphologically similar to the stem leaves, but in photosynthetic species they are usually smaller. The involucral bracts have either monochasial cymes (cincinni) or individual flowers in their axils. In addition, solitary flowers are usually present in axils of several stem leaves below the involucre; however, we never observed these flowers to be fully developed. In B. disticha, some stem leaves subtend severely underdeveloped cymes.

Several-flowered inflorescences were observed in B. disticha (with more than ten flowers, excluding those in axils of stem leaves; Figures 1B,E) and B. lutescens (with up to eight flowers; Figures 1B,F); inflorescence of these species is a thyrsoid (here and below, terminology after Endress, 2010). The involucral bracts have monochasial cymes in their axils. Within a monochasial cyme, each flower possesses a single floral prophyll (bracteole), which is arranged transversally or nearly transversally. A lateral flower of the next order developed in the axil of the floral prophyll thus has a transversal position. Along the cyme, there is an alternation of right and left positions of the floral prophylls, which leads to an apparently two-rowed arrangement of the flowers. The cyme with such a branching pattern corresponds to a cincinnus (Buys and Hilger, 2003). Burmannia disticha shows the longest cincinni: they often have more than six orders of branching, i.e., bear more than six flowers. In the first flower of the upper cyme, the floral prophyll is usually in an anodic position, i.e., closer to the end of the ontogenetic spiral (see Korn, 2006 for terminology). In the lower cyme, in contrast, the floral prophyll of the first flower is in a cathodic position. Only in some inflorescences of B. lutescens, the floral prophyll of the first flower is in anodic position in both cymes. In all lateral flowers, one of the inner tepals lies in the same radius with the floral prophyll; this allows compact packaging of the flowers, placing them side by side, with wings of the neighboring flowers more or less parallel to each other (Figure 1).

In few-flowered species, the inflorescences consist of up to four flowers (excluding flowers in axils of stem leaves) in B. coelestis (Figures 1C,G), up to three flowers in B. itoana and B. oblonga, up to two flowers in B. chinensis (Figures 1D,H). In all these species, the axils of the involucral bracts usually bear solitary flowers; in B. chinensis and B. itoana, only the lower involucral bract subtends a lateral flower, whereas the upper one lacks flowers in its axil. Such an inflorescence can be described as a botryoid; it is structurally similar to the thyrsoid found in the several-flowered species, differing in the presence of single lateral flowers instead of cincinni. In some individuals of B. coelestis, and rarely also in the other species, flowers of an additional order were observed, with the inflorescences approaching a thyrsoid. The arrangement of the floral prophylls in the few-flowered species, except for B. oblonga, is the same as in the several-flowered species. In B. oblonga, the floral prophylls are nearly adaxial median (at a slight angle to a strictly median position). The flower is oriented in such a way that one of its inner tepals occupies the radius of the floral prophyll, i.e., uniformly with the other species with respect to the floral prophyll, but differently from the other species with respect to the flower-subtending bract. In addition, B. oblonga differs from the other species studied in the involucral bracts inserted on stem slightly below the ovary base (with the flower thus being evidently pedicellate), and in a subopposite arrangement of the involucral bracts.

Floral vasculature of eight species studied here shows the same groundplan. Below we provide a detailed description of the floral vasculature in B. disticha (Figures 3, 4). It has the most extensive vascular system of all species examined. This is followed by a description of the recognized interspecific differences.

Seven vascular bundles enter the receptacle from the pedicel (Figure 3A). Of them, one bundle occupies the central position (referred here to as a central bundle), and six bundles are arranged in two alternating rings with three bundles in each ring lying in the radii of corresponding tepals. The bundles of the outer ring occur in the radii of the outer tepals (and also the perianth wings and the styles); they are referred here to as outer bundles (OBs). The bundles of the inner ring occupy the radii of the inner tepals (and also the stamens and the ovary septae); they are referred here to as inner bundles (IBs).

The central bundle enters the ovary base. At the level of the bases of placentas in the synascidiate zone, the central bundle divides into three branches which become the three synventral bundles, located in the radii of septae (i.e., occupy intercarpellary, = heterocarpellary, positions) (Figure 3B). Each synventral bundle gives off two centrifugal branches (placental bundles). Each branch innervates a corresponding part of the placenta and runs along margins of two neighboring carpels (Figures 3C,D). The ventral bundles (three synventral and six placental ones) reach the ovary roof where they terminate (Figure 3F).

Within the ovary wall, each OB gives off a centripetal radial branch (dorsal carpellary bundle) (Figure 3D), which passes through the ovary roof (Figure 4A) and innervates the style (Figures 4B–E). The IBs continue directly into the floral tube. Ultimately, the floral tube receives six bundles: three OBs alternating with three IBs (Figure 4B).

Below the level of stamen attachment, each IB gives off three branches: two tangential bundles (referred to as lateral inner bundles, one at each side of the median inner bundle) and one centripetal radial bundle (stamen bundle) (Figures 4C,D). Each stamen bundle enters the stamen which occupies the same radius (Figure 4E). Thus, above the level of stamen attachment the floral tube contains 12 bundles, of which three groups of inner bundles (one median and two lateral ones in each group) alternate with three OBs. All of them run into the tepal lobes (Figure 4E).

Each outer tepal lobe receives an OB, which becomes its median trace, and one lateral inner bundle at each side (lateral traces). The inner tepal lobes are single-traced; each of them receives a median inner bundle (Figure 4E). In both outer and inner tepal lobes, branching of bundles takes place. In the outer tepal lobes, each lateral bundle gives off two tangential branches, resulting in seven bundles per tepal. The bundles of the inner tepal lobes trifurcate tangentially (Figures 4F,G).

Burmannia longifolia differs from B. disticha only in having unbranched bundles of the inner tepal lobes (Figure 5J).

Burmannia championii, B. chinensis and B. coelestis show identical floral vasculature. They differ from B. disticha in having unbranched bundles of both outer and inner tepal lobes (Supplementary Figures 1–3).

In B. lutescens, the median inner bundles terminate near the bases of inner tepal lobes; the inner tepal lobes are thus unvascularized. Bundles of the outer tepal lobes are unbranched (Supplementary Figures 5G,H).

In B. itoana, the differences from B. disticha are found in the innervation of the upper part of the floral tube and the perianth. The inner bundles give off only the stamen bundles within the floral tube (Supplementary Figure 4H), and end blindly below the bases of the inner tepal lobes. Thus, B. itoana has single-traced outer tepal lobes and unvascularized inner tepal lobes. The bundles of the outer tepal lobes are unbranched (Supplementary Figures 4J,K).

Burmannia oblonga, which is characterized by the absence of the inner tepal lobes, differs from B. disticha in the vascular system of the gynoecium and the upper part of the floral tube. First, the receptacle receives only the six peripheral bundles from the pedicel, whereas the central bundle is lacking (Supplementary Figure 6A). Each IB gives off two centripetal radial branches. At the bottom of the ovary, these branches merge pairwise forming three ventral bundles in the radii of the ovary locules (i.e., in homocarpellary positions) (Supplementary Figures 6B,C). Toward the upper part of the ovary, each ventral bundle bifurcates into placenta bundles (without prolongation beyond the point of branching). Second, below the level of stamen attachment, each IB gives off two branches in tangential direction (i.e., the lateral inner bundles) and continues into the stamen which occupies the same radius (Supplementary Figures 6F,G). Each pair of the lateral inner bundles continues into the two neighboring (outer) tepal lobes, and the OBs become the median traces of the tepal lobes, just as in B. disticha (Supplementary Figure 6H). The bundles of the tepal lobes are unbranched.

The floral development of the six species of Burmannia studied here is essentially uniform. Below we provide a detailed description for B. disticha, as we obtained the most complete data for this species. We then describe the additional observations made for the other species.

In early development of perianth of B. chinensis (Figure 12A), B. coelestis (Figure 12B), and B. itoana (Figures 13B–D), the outer tepal lobes are often prominently unequal in size. This asymmetry seems not to be related to the position of the flower-subtending bract, although our material does not allow to verify it in details. Apparently, the asymmetry disappears at middle or late stages. In B. itoana, the stage of the drastic elongation of the floral tube (synchronized with elongation of the common style) was recognized to correspond to the diameter of flower between 1.3 and 1.8 mm (Figures 14C,D), i.e., far after all the main floral parts are completely formed.

Floral development of B. oblonga was traced since the stage of initiation of the inner tepal lobes and stamens from their common primordia, when the flower is c. 300 μm in diameter (Figures 15A,B). At this stage, the outer tepal lobes alternate with slightly elevated sectors of the floral tube. We consider these elevations to represent the inner tepal lobes. They are evident at least till the outer tepal lobes begin to cover the flower (which becomes c. 700 μm in diameter), but completely vanish by the stage of a preanthetic flower.

For B. coelestis and B. lutescens, cross sections of flowers at early developmental stages were investigated (Figure 8). In a flower of B. lutescens c. 450 μm in diameter, a very short, apparently just initiated synascidiate zone of gynoecium was detected (Figure 8B). The symplicate zone is unilocular: the septae (bearing parietal placentas) are evidently free from each other (Figure 8C). In a flower of B. coelestis c. 700 μm in diameter, a synascidiate zone is still lacking. The ventral slits in the symplicate zone are already closed; the ovary is thus secondarily trilocular and the placentation is axile in this part. The developing flowers of B. coelestis are comparable in size with those of B. disticha at corresponding stages, being only slightly smaller (Figure 12B, 335 μm vs. Figure 10B, 350 μm; Figure 12C, 420 μm vs. Figure 10C, 500 μm; Figure 12D, 515 μm vs. Figure 10D, 650 μm). The same is true for B. chinensis (Figure 12A). This allows to assume the flowers of the same size to represent similar developmental stages in Burmannia.

Development of ovules was studied in B. lutescens (Supplementary Figure 7). In the symplicate zone, the ovules cover the entire surface of the two placenta lobes except for their sides that face each other (i.e., along the boundary between the neighboring carpels). The ovules are anatropous, bitegmic; the inner integument initiates before the outer one. The ovules are nearly sessile at early stages, and later a long funiculus develops.

The inflorescences of Burmannia with more or less numerous flowers are thyrsoids with two lateral monochasial cymes. The botryoids of the few-flowered species possibly evolved from the thyrsoids as a result of reduction of monochasia; this idea is consistent with the available phylogenetic data (Merckx et al., 2008; Zhao et al., 2021). The presence of floral prophylls, their number, position and thus the inflorescence type are usually conserved at the genus or family level in monocots (Remizowa et al., 2013). The occurrence of a floral prophyll provides the possibility of further branching in its axil and leads to formation of a monochasium (instead of a solitary flower in the axil of a flower-subtending bract on the primary inflorescence axis). Thus, in taxa with floral prophylls present, the racemose inflorescences are structurally close to thyrses (see also Nuraliev et al., 2016, 2020).

An evolutionary transition from a raceme to a thyrse (and vice versa) is possible, but the presence of prophylls on the floral pedicels represents a key condition for such a transition. All members of Dioscoreales constantly develop floral prophylls, but their number and position are variable. For example, in Dioscoreaceae and Nartheciaceae, the single floral prophyll is in transversal or nearly transversal position. Among Nartheciaceae, Aletris L., Metanarthecium Maxim. and Narthecium Huds. possess racemes, whereas Lophiola Ker Gawl. and Nietneria Benth. are characterized by thyrses with monochasia (Remizowa et al., 2006a,2008, 2013; see also Tobe et al., 2018). In genera with racemes, the floral prophyll is inserted either at the right or left side of the flower within the same inflorescence. In species with thyrses, the position of the floral prophyll is more precise. In Dioscorea L. (Dioscoreaceae), the female flowers are often in spikes and the male flowers are arranged in thyrses (Remizowa et al., 2010a). The branched male inflorescences of Dioscorea demonstrate alternation of right and left floral prophylls, which results in development of cincinni. The inflorescences of Thismia Griff. (Thismiaceae) are thyrsoids. Thismia is unusual in the variable number of floral prophylls (2 or 3) and preferred branching positions (Nuraliev et al., 2021). In T. annamensis K. Larsen and Aver. with three floral prophylls, the branching occurs in the axil of the abaxial median prophyll, and the monochasium is therefore a drepanium. In other species of Thismia, branching occurs in transverse plane in the axil of a lateral prophyll, and the inflorescence is a bostryx.

Species of Burmannia examined here constantly produce a single floral prophyll, but differ by its position and intensity of branching. Most of the species examined constantly produce transversal floral prophyll in a very precise position, resulting in formation of cincinni in species with several orders of branching. Burmannia oblonga is the only species with a nearly median (adaxial) floral prophyll. If there were flowers in the axil of this floral prophyll, this would lead to formation of another type of monochasial cyme: a rhipidium. As in the other monocots with a single floral prophyll, the flower orientation is dependent on position of the floral prophyll in Burmannia. The floral prophyll is inserted between the two outer tepals and in the same radius with an inner tepal.

The floral vasculature is remarkably uniform in all the species of Burmannia studied to date (Pai, 1966; Rao, 1969; Rübsamen, 1986, this study). Certain variation is found in the vascular system at the flower base. Pai (1966) reported six central bundles at base (instead of a single bundle found in most of the species), which fuse with each other pairwise to form the synventral gynoecium bundles. The synventral bundles are common for all the species except for B. oblonga; they are thus formed in different ways. In B. oblonga, there are no central bundles at the ovary base, and the ventral bundles appear as branches of the peripheral bundles. Most important, the three ventral bundles of B. oblonga occupy the positions in the radii of the ovary locules, in contrast to the positions of the synventral bundles (alternating with the locules) of the other species.

The vasculature of the perianth and the distal part of the floral tube is most diverse. The outer bundles of the floral tube uniformly supply the outer tepal lobes forming their median traces. The inner bundles in most of the species continue up to the inner tepal lobes (forming their median traces) and give the lateral branches that form lateral traces of the outer tepal lobes. However, in B. oblonga, the inner bundles do not project beyond the point of departure of the stamen bundles, which is consistent with the absence of the inner tepal lobes in this species. In B. itoana, the inner bundles do not send the lateral branches, and therefore the outer tepal lobes receive no lateral traces. Then, in B. itoana and B. lutescens, the inner bundles do not reach the bases of inner tepals, which does not affect the groundplan of the floral vascular system (see “Perianth: inner tepal lobes”), but makes these tepals unvascularized. Finally, the species of Burmannia differ from each other in the degree of branching of the bundles within the tepal lobes.

As follows from the drawings provided by Rao (1969), he has observed neither lateral traces of the outer tepal lobes nor lateral inner bundles in the floral tube in B. disticha. At the same time, the lateral bundles of outer tepal lobes are drawn by Rao (1969) for B. nepalensis and B. pusilla. To date, single-traced outer tepal lobes are known only in B. itoana studied here. Given the rarity of this feature in Burmannia and the three-traced nature of the outer tepal lobes in our specimen of B. disticha, we suppose that Rao overlooked these bundles in the latter species.

A tube in a flower that bears all the floral elements except for the gynoecium fits the idea of a hypanthium (Leins and Erbar, 2010; Ronse De Craene, 2010). In most cases, it is impossible to prove the axial or appendicular nature of such a tube, and for this reason the morphological nature is currently not taken into account in definition of a hypanthium (Sokoloff et al., 2018). In Burmannia, the stamens are usually attached distinctly below the apex of the floral tube, and therefore only a part of the floral tube corresponds to the hypanthium (i.e., from the tube base to the level of stamen attachment). The rest of the floral tube (above the level of stamen attachment) bears only the free tepal lobes, and thus corresponds to the perianth tube. The perianth tube in Burmannia was overlooked by Caddick et al. (2000) who considered the entire floral tube to be formed by a hypanthium.

In this paper, we use the term “floral tube” for clarity, assuming that it consists of a hypanthium and a perianth tube, the latter being vanishingly short in some species of Burmannia. We have not observed any additional structural features that delimit hypanthium and perianth tube, apart from the level of stamen attachment and the pattern of branching of the vascular bundles related to the departure of the stamen bundles. Accordingly, the so-called perianth ribs or wings in Burmannia are morphologically parts of the ovary wall at their bases, hypanthium in the middle and perianth tube at the apex. There is no reason to consider the wings as expansions of the outer tepals, as was suggested by Pai (1966).

Both outer and inner tepal lobes in Burmannia are characterized by either single or double margins. Although these characters are widely used for species delimitation and identification in the genus (Jonker, 1948; Zhang, 1999; Wu et al., 2010), they have never been described for some of the species. We here supply the description of the species under study with information on tepal margins based on careful analysis of spirit material, SEM investigations and the cross sections (see Table 2). In particular, we report the shape of tepal margins for the first time for B. longifolia (outer tepals slightly double-margined at base, inner tepals single-margined) and B. lutescens (outer tepals minutely double-margined at base, inner tepals single-margined).

In tepals with double margins, the border between the adaxial and abaxial sides is not obvious. As follows from our observations on B. chinensis and B. coelestis (not shown), the double nature is a result of development of a concavity in a thick tepal margin. Thus, the surface of the double margin belongs neither to the adaxial nor to the abaxial side of tepal.

The most intriguing feature of perianth in Burmannia is the apparent absence of the inner tepal lobes in some cases. Some species, including B. cryptopetala Makino and B. oblonga, are uniformly characterized by this feature, whereas in other cases this feature contributes to uncertainties in species delimitation. The most prominent example is the B. lutescens species complex, which was divided into two species by Zhang (1999): the widely distributed B. lutescens (including B. tridentata Becc.), polymorphic with respect to this and several other characters, and B. gracilis Ridl. endemic to the Malay Peninsula with the inner tepal lobes present. Jonker (1938, 1948), in contrast, accepted B. lutescens (including B. gracilis) as uniformly possessing the inner tepal lobes, and B. tridentata as a distinct species known from a single collection made in Sarawak that lacks the inner tepal lobes. Although the former view, already adopted by Govaerts et al. (2011), is followed here, we argue that the taxonomy of this species complex is far from being resolved, and reliable morphological data is wanted, including the results of developmental studies and the information on structural variation in various geographical areas.

In B. oblonga, we observed the inner tepal lobes at the middle stages of floral development. They were found to be extremely short, but quite broad, being only twice narrower than the outer tepal lobes. Preanthetic flowers are shown to lack any evidences of the inner tepal lobes, and there are no vascular bundles in the floral tube corresponding to bundles of the inner tepals.

The two Vietnamese specimens of B. lutescens studied here are geographically remote from the rest of the known distribution area of this species, representing the only documented extra-Malesian populations (Nuraliev et al., 2018). Anthetic flowers of both specimens show distinct inner tepal lobes, which are unvascularized, but the bundles that act as median traces of the inner tepal lobes in the other species of Burmannia (i.e., the median inner bundles of the perianth tube) terminate right at the bases of the inner tepal lobes. One can interpret the absence of vascularization as a trend toward reduction of the inner tepal lobes in B. lutescens; possibly, in some populations of this species they are short enough to be hardly distinct in mature flowers. The same vascular pattern is found in B. itoana, which has even shorter (semi-orbicular) inner tepal lobes. However, B. itoana has never been reported to lack the inner tepal lobes (Jonker, 1938; Ohashi et al., 2000; Wu et al., 2010; Tsukaya, 2016).

In total, our data indicate that the inner tepal lobes are present in all species of Burmannia at least developmentally. Since most of the species possess well-developed and vascularized inner tepal lobes, and the species with a seeming absence of the inner tepal lobes (in maturity) or their vascular bundles are nested deep within the phylogeny of Burmannia (Merckx et al., 2006, 2008; Zhao et al., 2021), we assume that the deviations from this common pattern are cases of evolutionary reduction in the genus. The variation of morphology of the inner tepal whorl in the genus is likely to be quantitative (continuous) rather than binary, with several intermediate variants already recognized. It remains unclear, whether a considerable variation of degree of this reduction occurs at infraspecific level in Burmannia, or the presence/absence of the inner tepal lobes in maturity (and their vascular supply) is a species-specific character.

A tendency to perianth reduction in Burmannia mirrors a situation in Thismiaceae, another family of Dioscoreales. In some lineages of Thismia, there is a tendency to reduction of the outer tepals which are much smaller than the inner ones and do not contribute to the flower protection. Species of Thismia possess a rather uniform floral vasculature, with most prominent variation observed in the innervation of tepals. In T. mucronata Nuraliev (Nuraliev et al., 2021) and T. puberula Nuraliev (Yudina et al., unpublished), the small but well-visible outer tepals are non-vascularized, whereas the outer tepals of other species of Thismia studied are uniformly three-traced.

Interestingly, the evidences of reduction in the outer tepal whorl do not affect the number of stamens in Thismia: flowers of Thismia uniformly possess six stamens. In the development, stamens of Thismia appear as protrusions at bases of tepal primordia after the carpels are initiated (Nuraliev et al., 2021). This pattern resembles late division of common tepal-stamen primordia. In contrast, Burmannia demonstrates reduction in both perianth and androecium. Surprisingly, irrespective of larger outer tepals, the outer stamens are lacking in Burmannia. This pattern of stamen whorl reduction is apparently due to initiation of inner tepals and stamens via common primordia (which was not necessarily the case for outer tepals and outer stamens in a putative ancestor of Burmannia). We speculate that organs developing via common primordia cannot be modified independently in the course of evolution as they are united in a solid module. Indeed, the inner tepals are always initiated in Burmannia (although not always well-detectable in mature flowers).

Caddick et al. (2000) assumed the family Burmanniaceae to be the only lineage of Dioscoreales with the inner stamen whorl present and the outer one lacking. Most of Dioscoreales are characterized by two-whorled androecium, and examples of the taxa with a single stamen whorl being the outer one are provided by Caddick et al. (2000). In fact, the androecium structure found in Burmanniaceae is also characteristic of Oxygyne Schltr., a member of Thismiaceae (Shepeleva et al., 2020).

In this study, the entire process of androecium initiation and development is documented. No traces of initiation of the outer stamens were recognized. This allows to state that the outer stamens are structurally completely absent in Burmannia. However, the orientation of the inner androecium whorl and the gynoecium in Burmannia corresponds to that of a typical pentacyclic monocot flower, i.e., the inner stamens are in the radii of the inner tepals, and the carpels in the radii of the outer tepals. It is likely that the outer stamens do not develop morphologically, but act as a source of positional information governing the orientation of the subsequent floral whorls. This hypothesis corresponds to the idea of ablasted organs (Stützel and Gansser, 1995; Choob, 2010).

The common style in Burmannia is a complex structure, consisting of three postgenitally fused styles (belonging to three carpels), as has already been shown by Zhang (1999) and Caddick et al. (2000); it thus represents the asymplicate zone of gynoecium. In some other Dioscoreales, such as Thismiaceae (Nuraliev et al., 2021), some Nartheciaceae (Remizowa et al., 2006b,2008; Tobe et al., 2018), Taccaceae (Igersheim et al., 2001) the columnar portion of the gynoecium above the ovary is formed by the distal part of the symplicate zone, and usually bears free styles at its apex. It can be called a stylar column. We find it appropriate to apply distinct terms for these morphologically different structures, despite some authors use “style” also for the common style (Maas et al., 1986; Maas-van de Kamer, 1998; Merckx et al., 2013).

The identification of the border between the synascidiate and symplicate zones of gynoecium (i.e., the cross zone) in mature flowers of Burmannia is complicated due to postgenial fusion accompanied by complete re-differentiation of contacting epidermises at the base of the plicate zone of each carpel. As a result, the lines of postgenital fusion are indiscernible. The symplicate zone is thus proximally very similar to the synascidiate zone, as it is trilocular without distinct lines of postgenital fusion in the ovary center (in cross section). We based on the following evidences to identify the zones. (1) Shape of placentas: we assume that in the synascidiate zone, there are no morphogenetic reasons for development of the bilobed placentas (in cross section). Therefore, placentas of this structure indicate symplicate zone, and the entire (rounded) placentas indicate the synascidiate zone. (2) Developmental data: at the stage of already elongated ovary with no postgenital fusion yet occurred (e.g., Figure 10E), most of the ovary length is unilocular, thus representing the symplicate zone.

We conclude that all the studied species of Burmannia have uniform structure of the ovary, comprising a long symplicate zone and a much shorter synascidiate zone. The placentas occupy the entire symplicate zone, where each placenta in the secondarily trilocular part has a bilobed shape with the lobes corresponding to the fused carpel margins of the same carpel, and in the unilocular part (if present) it has a form of two separate strands. In the synascidiate zone, the placentation takes place only in a short distal part; since the placentas have hanging portions below the level of their attachment, they are found in the cross sections of the synascidiate zone for some length further down.

Pai (1966) stated the entire trilocular portion of the ovary in B. pusilla to be “a result of the intruding placentae meeting in the center,” i.e., belonging to the symplicate zone. However, Figure 2 of his work clearly shows the basal part of the ovary with round (and not bilobed) placentas in the cross section, indicative of the synascidiate zone. Pai’s observation therefore fits well our general conclusion on the structure of the ovary in Burmannia.

The presence of the common style formed by postgenitally fused individual styles (i.e., the asymplicate zone right above the ovary) correlates with presence of septal nectaries. Septal nectaries of Burmannia are confined to the plicate carpel zone and located in the distal region of the ovary above the level of ovule insertion; they open via slits at the ovary roof. Similar supralocular septal nectaries are also reported for Dioscorea (Caddick et al., 2000). In Natheciaceae (Aletris and Metanarthecium), septal nectaries are more pronounced; they are interlocular, occupying almost the entire length of the plicate carpel zone along the ovary. The level of nectary orifices depends on ovary position. In Aletris, characterized by semi-inferior ovary, the opening slits are located at the ovary roof as in Burmannia. In Metanarthecium, the ovary is superior and nectaries open via special ducts at the ovary base (Remizowa et al., 2008; Tobe et al., 2018).

Our study revealed synorganization of the stamens and the gynoecium in Burmannia which involves diverse structural interactions. In some of the species, the interaction is limited to the tight arrangement of the stamens around the common style (without organ fusion), with the thecae placed just below the broad stigmas, and the apical crests of the connectives projecting between the stigmas. In the others, we observed the connectives to be postgenitally fused to the common style, with their central longitudinal ribs fitting the grooves between the carpels. In the latter case, the stamen-style complex corresponds to the idea of a gynostegium earlier considered to be unique for some lineages of Apocynaceae (Endress, 2006, 2016). In the species of Burmannia with gynostegia, the stamens can hardly be detached from the common style without disruption of tissues in at least a part of the area of contact (e.g., Figure 11E). The gynostegia of Burmannia vary in the degree of the fusion, which is manifested in evidently or hardly visible boundary between the fused organs in the cross sections.

The fusion between the stamens and the common style has been reported by Pai (1966) for B. pusilla, who stated that “the stamens are connected with the style by three broad bands of parenchymatous tissue.” Rao (1969) failed to find any such connection in B. disticha and B. nepalensis. It is difficult to compare their observations with our results, because no appropriate illustrations were provided by these authors.

All the structures described above are ultimately involved in the process of seed production. The complex floral construction and the reward offered via septal nectaries suggest that Burmannia is pollinated by insects. However, only scarce information about pollination and reproduction of this genus is currently available. Several species of Burmannia are known to perform pollinator-independent autogamy (mostly via cleistogamy), including B. capitata (Walter ex J.F.Gmel.) Mart. (Wood, 1983), B. championii (Ernst and Bernard, 1912), B. lutescens [as B. candida (Blume) Engl., nom. illeg.] (Ernst and Bernard, 1912) and B. wallichii (Miers) Hook.f. (Zhang and Saunders, 2000). Burmannia coelestis is reported to be apomictic (Ernst, 1909). The only direct indications of entomophily in Burmannia are those published by Kato (1996) and Momose et al. (1998). These authors observed B. lutescens in Sarawak, Borneo, visited by two genera of mosquitoes (Culicidae, Diptera) whose proboscises correspond to the width of the openings between stigma and stamens and to the depth of the floral tube (Kato, 1996). However, it remains unclear, whether mosquitoes are the regular pollinators of this species, and do the pollinators facilitate xenogamy, geitonogamy or autogamy in Burmannia.

Our study indicates a need for broad-scale investigations of reproductive biology in Burmanniaceae. First, we failed to reveal any structural characters of flowers and inflorescences correlating with the mode of nutrition (mycoheterotrophic vs. photosynthetic lifestyle). Our data therefore suggest uniform pattern of pollination biology in Burmannia, which would contradict the idea of a “mycoheterotroph floral syndrome” (Waterman et al., 2013, p. 281), i.e., a putative set of adaptive floral features common to mycoheterotrophic angiosperms due to their common ecological preferences (Leake, 1994; Waterman et al., 2013). This question is to be resolved by direct observations of pollination process in species with different mode of nutrition. Second, it appears worthwhile to compare the patterns of reproductive ecology in the species of Burmannia with and without a gynostegium, including the diversity of floral visitors and their interactions with flowers, as the presence/absence of a gynostegium is possibly of high adaptive significance.