Field investigation of Burmannia disticha was performed in northern Vietnam (Lao Cai province, Bat Xat district, Bat Xat Nature Reserve); the general information on Bat Xat Nature Reserve is provided by Bui et al. (2020). The study site is located around the point 22° 37.539’ N, 103° 37.658’ E. Voucher herbarium specimens were collected under the collector’s number Vislobokov 19039 and deposited at MW (barcode MW0757926) and at HN. The field investigations took place during 4–13 June 2019. The time of sunrise was 5:20 and time of sunset was 18:48 at the study site during this period, as estimated using the Solar Calculator available at the ESRL (2023) website.

Flowers of Burmannia disticha (illustrated in Figures 1 , 2 ) are most often uniformly blue to violet outside except for yellow or greenish perianth lobes (Jonker, 1938; Jonker, 1948; Zhang, 1999). The floral tube is 3–8 mm long. The perianth wings are conspicuous, 1–3.5 mm wide, running from the pedicel along the ovary and floral tube to the outer tepal lobes, continuing into abaxial crest-like keels of the outer tepal lobes (Wu et al., 2010). The inner tepal lobes are well developed in mature flowers. The stamen connectives are postgenitally fused to the common style to form a gynostegium. The stigmas are placed right above the thecae and occlude the flower entrance (Yudina et al., 2022).

For observations of successive stages of anthesis, inflorescences were labeled individually and checked daily. Thirty-one inflorescences were monitored in total (each inflorescence on a separate above-ground shoot). Observations of floral visitors were performed during day time and night time on 4–13 June 2019. Observations were performed in a series of overlapping time intervals, i.e., different time intervals were covered on different days, so that altogether they cover all the day periods except for 2:00–6:00. The total observation time was 40 hours. Visual observations were performed as well as video recording and photography using HC-VX1EE-K (Panasonic, Japan), GZ-R495BEU (JVC, Japan) and EOS 70D (Canon, Japan) digital cameras. Insects visiting flowers of B. disticha were captured by a net, killed by ethyl acetate fumes, and fixed individually in 95% ethanol. The records of visits were based on captured visitors, direct observations and video records. Terminology related to plant reproduction follows Cardoso et al. (2018).

For analysis of pollen grains attached to the insect bodies, in Moscow State University, each insect was washed individually in 95% alcohol. Several drops of glycerol were added to the washing, after which the washing was mixed and centrifuged for 5 min at 3000 rpm. After that, 50 µl of sediment was taken from each sample, placed on a slide and investigated using a Micromed-2 light microscope (Ningbo Sheng Heng Optics & Electronics, China). All pollen grains were counted in each sample; pollen of B. disticha and pollen of all the other seed plants was recorded separately. Fourteen insects were sampled in total. Identification of pollen grains of B. disticha was performed by their comparison against pollen from anthers of this species. Since B. disticha is the only species of Burmanniaceae observed in the study area, misidentification of its pollen was unlikely.

Counting of pollen and ovules in flowers of B. disticha was performed in Moscow State University, using flowers with undehisced thecae fixed in 70% ethanol. Thecae and ovaries were dissected in 70% ethanol under an Olympus SZX7 (Olympus Corp., Japan) stereomicroscope. All pollen grains from one theca were placed on a slide and counted. All ovules from one placenta were placed on a slide and counted. Slides with pollen and ovules were examined using a Micromed-2 light microscope. Six thecae and six placentas from six different flowers (each flower taken from a separate plant) were investigated in total. Pollen-ovule ratio for the species was estimated by the following formula: P/O = 6p/3o, where p is the mean number of pollen grains in one theca and o is the mean number of ovules in one placenta.

For estimation of length of floral tube, 10 flowers were dissected in 70% ethanol under an Olympus SZX7 stereomicroscope and measured by a ruler and a vernier caliper. Similarly, length of proboscises of insects (fixed and stored in 95% ethanol) were measured by a ruler under an Olympus SZX7 stereomicroscope.

Scanning electron microscopy (SEM) was employed to investigate the mutual arrangement of organs inside the floral tube, and also to describe the fine characters of shape and surface of flower organs. The flowers fixed in 70% ethanol were dissected under an Olympus SZX7 (Olympus Corp., Japan) stereomicroscope, dehydrated through 80% ethanol, 96% ethanol and a mixture of ethanol (96%) and acetone (100%) at a 1:1 proportion followed by absolute acetone. The material was then critical-point dried in a Hitachi HCP-2 critical-point drier (Hitachi, Japan) using liquid carbon dioxide. Dried samples were mounted onto stubs using double-sided sticky tape, coated with gold using an Eiko IB-3 ion coater (Eiko Engineering Co. Ltd, Japan), and observed using a CamScan S2 SEM (CamScan, UK) at Moscow State University. Terminology for floral structure follows Yudina et al. (2022).

In June, massive flowering of B. disticha was recorded in Bat Xat Nature Reserve. In each inflorescence, the terminal flower opens first, followed by the two lateral flowers (i.e. the first flowers of the lateral cincinni) ( Figure 1A ). A flower of the next order opens when signs of senescence appear in the currently anthetic flower, or 1–2 days after that. Within a cincinnus, the lifespans of two adjacent flowers never overlap. Thus, there is normally a single blooming flower in each cincinnus at a given moment, and not more than two flowers blooming simultaneously in an inflorescence. A flower opens between 10:00 and noon (probably depending on weather). Tepal lobes are greenish yellow in flower bud becoming bright yellow right before flower opening. Tepal lobes are tightly appressed to each other in bud ( Figures 2A, F ); at flower opening they separate from each other and maintain the erect (upright) position during anthesis ( Figure 1H ). The inner tepal lobes are arranged close to the gaps between the neighboring style branches ( Figures 1B–E ). Anthers of all three stamens dehisce simultaneously at flower opening ( Figures 1F, H–J , 2E ). No evidence of dichogamy was observed. The stamen connective possesses a pair of pronounced divergent apical crests and a thick pendent basal spur; all three appendages are distally covered by long papillae ( Figures 1D, H–J , 2E ). The floral tubes of anthetic flowers were 5.3–8.1 mm long. Septal nectaries produce nectar, which is accumulated in the form of a film at the bottom of the floral chamber ( Figures 1B, C, G ). The stigma is fan-shaped, with a thin membranous transverse folding, covered with dense brush-like papillae ( Figures 2B–D ). In the anthetic flower, the stamens and stigmas are enclosed in the floral chamber which is hardly accessible from the outside: the orifice of the floral tube is obstructed by the apical crests of stamens and the style branches ( Figures 1E, F , 2B ). An individual flower remained anthetic for three days; when the tepal lobes changed their colour to violet and dull black ( Figure 1A , central flowers), the anthesis was considered to be over.

Flowers of B. disticha were visited by flies (Diptera), bees and bumblebees (Hymenoptera), butterflies and moths (Lepidoptera) and occasionally by orthopterans and spiders (Araneae). The floral visitors were predominantly recorded during day time; night visitors were rare, represented by nocturnal orthopterans only. The number and time of visits are summarized in Table 1 .

Flies (Diptera) from the families Muscidae and Syrphidae were recorded. They visited the flowers from 9:00 to 16:00. The flies landed on the tepal lobes and apparently intended to reach the floral chamber with their proboscises, but could not penetrate beyond the stigmas (and under the outer tepal lobes if they visited flowers that were not yet fully open).

Hymenopteran visitors were represented predominantly by bumblebees and also by a single bee. These insects were observed on the flowers from 11:00 to 16:00. They landed on the floral tube and tepal lobes and inserted their proboscises into the flower by bending the distal part of the transversally folded stigma down, accessing the floral chamber. Pollen attached to their heads is thus expected to be deposited on the brush-like papillae of the stigmas. All the nine observed bumblebees are members of the Bombus trifasciatus species complex, a group with complicated species delimitation (Hines and Williams, 2012). The captured insects are most close to Bombus burmensis Williams, 2020 ( Figure 3 ). The captured bumblebees (seven individuals, see Table 2 ) have proboscises 7–11 mm long, which is long enough to reach nectar from the bottom of the floral chamber, whereas the recorded bee (belonging to the genus Coelioxys Latreille, 1809) has a proboscis about 3 mm long, i.e. significantly shorter than the floral tube in B. disticha.

Lepidopterans were represented by butterflies from the families Hesperiidae and Satyridae, and moths belonging to Pyralidae. They visited flowers of B. disticha from 12:00 to 17:00 ( Figure 4 ). Hesperiidae were the most frequent of them. The lepidopterans landed on the floral tube and tepal lobes and inserted their proboscises between the transverse folding of the stigma and the floral tube (slightly bending the stigma), i.e. in a manner similar to that of the hymenopterans. The lengths of the proboscises of captured insects were: 20–23 mm in Hesperiidae, about 6 mm in Pyralidae, and 7 mm in Satyridae, which allowed all these lepidopterans to reach the bottom of the floral chamber.

Orthopteran insects and spiders occasionally visited the flowers during day and night. All of them set on the outer surface of the floral tube and ovary never touching the tepal lobes or trying to get inside the floral chamber.

Among the visitors, pollen of B. disticha was detected only in the washings from the hymenopterans and the pyralid moths ( Table 2 ). The largest number of pollen grains was detected on bumblebees: 365 pollen grains per insect in average. Pollen grains of other seed plants were detected on the bodies of dipterans, hymenopterans, lepidopterans and Araneae.

In B. disticha, number of pollen grains in one anther theca ranges from 1638 to 2563 (1990 in average). Number of ovules in one placenta ranged from 488 to 767 (582 in average). Pollen-ovule ratio for B. disticha was estimated as 6.84 ( Table 3 ).

The entire genus Burmannia is likely to possess a prominent entomophilous floral syndrome, as defined by Faegri and van der Pijl (1979). This idea has already been proposed for various species of the genus (Spitmann, 1975; Maas et al., 1986; Zhang and Saunders, 2000). Flowers of Burmannia are brightly coloured; in B. disticha, yellow tepal lobes contrast with the violet floral tube and ovary. The tepal lobes possibly serve as a visual cue signaling the entrance to the floral tube (where the nectar is presented), or imitating anthers for pollen-collecting insects. In particular, yellow coloration of the inner part of the floral display is known to be similar to the pollen coloration for Apoidea (Hymenoptera), where bees and bumblebees belong (Heuschen et al., 2005). It is likely that the flowers of B. disticha have no cues of any other types, since we found them to be odorless, at least to humans. According to Maas et al. (1986), there are no records of scented flowers in the genus; however, Zhang (1999) noted fragrance in flowers of B. coelestis. The other features of the entomophilous syndrome in Burmannia include adaxially papillose outer tepal lobes and nectar secretion via septal nectaries (e.g. Yudina et al., 2022). The presence of nectar was directly observed only in B. lutescens (Kato, 1996) and in B. disticha (present study). Thus, the flowers of Burmannia possess a primary attractant, i.e. a reward (nectar) and secondary attractants (coloration pattern and, in certain species, odor). In our study, attraction of visitors (insects) by the flowers of B. disticha is shown for the first time. The presence of floral visitors together with the modes of the flower-insect interactions reported here is in concordance with the idea of entomophily of this species.

Flower of B. disticha exhibits the following traits that limit the access to the pollen, stigmas and nectar. The stamens and stigmas are enclosed in the floral chamber which is hardly accessible from the outside. The connectives are fused with the common style into a gynostegium (Yudina et al., 2022). The inner tepal lobes of the fully open flower possibly also contribute to the limited accessibility of the floral chamber, since they are arranged close to the gaps between the neighboring style branches. Then, the stigma is fan-shaped, so that the potential visitors (first of all, insects) are expected to be able to bend the stigma margin down and get inside the floral chamber via the gap between the floral tube, two stamens and the common style. Finally, the nectar is accumulated at the bottom of a long chamber, which further complicates access to this reward.

Two logical ways exist for the potential visitors to reach the nectar (apart from the nectar robbing): (1) by an animal small enough to enter the floral chamber; such an animal would be able to move within the chamber freely; (2) by an animal with a proboscis thin enough to enter the floral chamber and long enough to reach the bottom of the chamber. In our study, not a single tiny visitor was detected, and therefore the implementation of the first way has not been demonstrated. Among the recorded visitors, the bumblebees and the lepidopterans were observed to apparently get access to the floral chamber successfully. These insects penetrated the flower entrance with their proboscises, and most of them have the proboscises at least as long as the floral chamber. Thus, the second way of interaction between the flower and the visitor is likely to be employed by B. disticha.

In total, our study demonstrates that despite the “closed” appearance, the flower of B. disticha is well accessible for certain visitors, contrary to the views of e.g. Schoch (1920) who considered such a floral construction as an argument in favour of self-pollination.

Among the hymenopterans, the bumblebees (Bombus burmensis) and a bee (Coelioxys sp.) visited the flowers of B. disticha. They inserted their proboscises inside the floral tube. At the same time, all of them bore pollen grains of B. disticha, and the bumblebees bore it in large amounts. Our results thus suggest that the bumblebees and bees act as pollinators of B. disticha. Since the bumblebees had proboscises 7–11 mm long, they were likely able to suck the nectar from the bottom of the floral chamber that was estimated as 5.3–8.1 mm deep. On the other hand, they could also possibly collect pollen, because bumblebees are known to use both honey and pollen for food and storage (Michener, 2007; Williams, 2020) and were even shown to have preferences for two-coloured flowers with a centre colour similar to that of pollen (Heuschen et al., 2005). The bee, in contrast, had a short proboscis about 3 mm long, probably reaching only anthers but not nectar. Coelioxys is a genus of cleptoparasitic bees, i.e. they lay their eggs in the nests of host species. Their hosts are other species of bees, mainly belonging to Megachile Latreille, 1802 and various Apidae; the parasite larva feeds on the food that had been provided for the host larvae (Michener, 2007). Accordingly, adults of Coelioxys bees lack any pollen-manipulating or pollen-carrying structures, and only forage for their own needs as the brood is cared for by the host species (Martin et al., 2000). Parasitic bees usually take only nectar from flowers, which they use as their own food. They only carry pollen that deposits passively on them. Parasitic bees are often not very hairy and thus probably play a less significant role in pollination than the pollen-collecting bees (as well as hairy bees) do (Michener, 2007). Thus, the bee captured on B. disticha possibly visited the flowers accidentally, while browsing bright flowers for foraging, and cannot be considered as a regular pollinator of B. disticha.

Among the lepidopterans, another group of frequent visitors of B. disticha in our study, only Pyralidae carried pollen of B. disticha, and only in a small amount. Though, the amount of pollen in the washings from the bodies of lepidopterans was possibly underestimated: numerous scales from the lepidopteran wings appeared in the washings together with pollen, which complicated detection of the pollen grains. All the recorded lepidopterans (especially Hesperiidae) had proboscises long enough to reach the nectar. The majority of lepidopterans (including the visitors of B. disticha) are anthophilous in adult stages, and nectar is known to be their principal nutrient source (Gilbert and Singer, 1975; Krenn, 2010), whereas for only a few lepidopteran species adaptations to pollen-feeding have been described (Romeis et al., 2005). At the same time, the majority of species likely benefit from pollen-derived nutrients when consuming pollen-contaminated nectar (Baker and Baker, 1975; Romeis et al., 2005). Pollen of Burmannia (including B. disticha) was indicated as starch-rich but lipid-poor (Zhang, 1999; Zhang and Saunders, 2000), which is typical of plants pollinated by Lepidoptera (Baker and Baker, 1983). Although Zhang and Saunders (2000) stated that “the small size of Burmannia flowers and the inaccessible anthers preclude wind-pollination or pollination by Lepidoptera or birds”, our data allow to refute this idea at least regarding Lepidoptera. Considering all the evidence, we argue that lepidopterans might play a significant role in pollination of B. disticha.

Although flies were relatively frequent visitors of flowers of B. disticha, they did not get access to the floral tube and therefore did not reach anthers and stigmas. Accordingly, pollen of B. disticha was not found on their bodies (though there was pollen of other flowering plants). They might have been attracted by the yellow tepal lobes resembling anthers, as pollen is included in the diet of some flies, especially Sirphidae. Spiders and orthopterans visited flowers occasionally and bore virtually no pollen grains on their bodies. All these groups of visitors evidently do not take part in pollination of B. disticha; instead, the spiders may be floral visitor predators, and the orthopterans may be florivores (but we have not detected any traces of florivory).

Three types of mating systems are distinguished in angiosperms according to the mode of gene transfer from one generation to the next through sexual reproduction: xenogamy, geitonogamy and autogamy (including cleistogamy) (Cardoso et al., 2018). Asexual reproductive systems, i.e. apomixis, which is currently known for a single species of Burmannia (Ernst, 1909), cannot be excluded for B. disticha, but unlikely occur in this species given the open anthetic flowers and visitations by insects.

Geitonogamy (which is equal to autogamy from the genetic and evolutionary viewpoint, but resembles xenogamy with respect to the pollination mechanism) is unlikely for the entire genus Burmannia for the reason of shoot structure in the genus. Indeed, in B. disticha, an individual develops a single inflorescence, and within an inflorescence the flowers open sequentially and not more than two of them bloom simultaneously. The cymose inflorescence structure can therefore be viewed as an adaptation against fertilization of an individual’s ovules by its own pollen (that would be performed in case of geitonogamy).

As demonstrated here, B. disticha is characterized by an entomophilous floral syndrome and shows adaptations for pollination by insects with long proboscises. Moreover, we revealed pollen transport by insects (predominantly bumblebees) in this species. Based on these findings, we conclude that B. disticha possesses a set of adaptations to xenogamy.

It should be taken into account that entomophily does not necessarily preclude from autogamous pollen transfer, i.e. the so-called facilitated autogamy (Lloyd, 1992; Cardoso et al., 2018). In B. disticha, this mechanism is possible, because the insects are expected to touch both stigmas and anthers of the same flower. However, facilitated autogamy was argued to be only rarely beneficial, being mainly a by-product of adaptations to cross-pollination (Lloyd, 1992). Therefore, entomophily can be considered a solid indicator of a xenogamous mating system.

Pollen-ovule ratio estimated in this study for B. disticha (6.84) evidences in favor of cleistogamy or obligate autogamy according to the analysis by Cruden (1977), whereas the ratios found by him in xenogamous angiosperms were above 20. In line with this finding, self-pollination as well as self-compatibility was assumed previously for a number species of Burmannia including B. disticha (Schoch, 1920; Fryxell, 1957; Wood, 1983; Zhang, 1999). Thus, a certain contribution of autogamy to the mating system of B. disticha is likely to take place. This inference does not contradict the idea of predominance of xenogamy in this species. With respect to the pollen-ovule ratio, as it has already been argued by Yudina et al. (2021) for some other genera of Dioscoreales (Tacca J.R.Forst. & G.Forst., Taccaceae, and Thismia Griff., Thismiaceae), the low values indicate high efficiency of pollen transfer that is not necessarily facilitated by autogamy. Moreover, a possibility of a versatile pollination strategy in a species should be considered. Comparing our results with the observations by Zhang (1999), one can infer that both insect-mediated cross-pollination and autogamous seed set via cleistogamy occur in B. disticha, possibly employed by different populations or depending on environmental conditions.