The study was carried out in the Dendrobium germplasm resource garden in Ruili City, Yunnan Province, China (97°51′4.79″–97°51′14.64″ E, 24°01′9.42″–24°01′14.00″ N; 770–779 msl). The Dendrobium germplasm resource garden was situated on the Mengxiu Mountain, which has a high temperature tropical monsoon climate in Tongbiguan Nature Reserve. Since 2002, D. chrysotoxum was mostly propagated using plant division, cuttage, and tissue culture, and then it was planted on the stem of trees by imitating wild conditions. Tissue culture uses seeds, the product of sexual reproduction of D. chrysotoxum, as explants to obtain many individuals. More importantly, different individuals are labeled.

From March to May 2020, D. chrysotoxum’s floral phenology was monitored to note flowering period’s beginning, peak, and finish. On 40 inflorescences in the bud stage (40 plants), 100 unopened flowers were marked randomly and then examined. When a flower’s dorsal sepal parted from an extended labellum, it was considered to be open. A flower was considered closed when the labellum wilted and crumpled. From the opening of the first bloom until the collapse and withering of the last flower, the flowering time of each individual flower and the flowering duration of each population were observed.

The following parameters were measured with vernier calipers (0.01 mm) to examine the floral morphology of 30 fresh flowers on 30 additional unopened inflorescences (30 plants). These measurements included the length and width of the petal, the labellum, the central sepal, the lateral sepal, the height of the flower passage, the width of the flower passage [11,12] and the depth of the flower passage (the distance between the spur mouth and the middle of the stigmatic surface).

In order to determine if the flowers produced nectar, 30 flowers from 10 individual D. chrysotoxum plants were randomly chosen in late April of 2020, and their spurs were examined using a stereoscopic microscope (Leica M165 C, Germany).

SONY digital cameras (Sony ILCE-7M2, Thailand) were utilized to capture visitors in the research locations throughout D. chrysotoxum’s flowering periods in 2019 and 2020, as well as to note their behavior. The observation times every day were from 6:00 until 18:00. By observing the disappearance of pollinia in the mornings, the nightly visitors were assessed. Behavior and species of insect visitors were observed and recorded by camera, with the record of daily weather conditions.

The ethyl acetate fumes from ethyl acetate were used to catch and kill the insects that were observed visiting D. chrysotoxum in a jar. Insect specimens were identified by X. Gao and R. Zhang of Yunnan Agriculture University. Voucher specimens are deposited at the Dehong Tropical Agriculture Research Institute of Yunnan. On the caught insects, the following dimensions were recorded: the height of the insect’s thorax, the width of the insect body, and the length between the midpoint of the thorax and the endpoint of the proboscis.

3000 flowers of D. chrysotoxum that were planted on the stems of trees in 2019 and 2020 made up the sampling. Each flower was marked and labeled. We checked that all labelled flowers contained pollinia at the column apex but no pollinia deposited in the stigmatic cavity. Success of natural pollination conditions was estimated as capsules setting. In 2019 and 2020, artificial pollination treatments comprised untouched flowers as direct autogamy, self-pollination, and cross-pollination. This sampling included 400 flowers on 40 inflorescences (40 plants). The labellum of each flower expanded and opened before any insect activity happened on the first day for all selected flowers in each treatment. Flowers were chosen at random and divided into three treatments. (1) Direct autogamy. Flowers were bagged in white paper bags before anthesis till senescence. (2) Self-pollination. A second flower on the same plant provided the stigma with whole pollinia. (3) Cross-pollination. Each stigma received whole pollinia from another plant’s flower. At the end of the flowering period, the ovary’s swelling was noticed, and the number of capsule setting was counted. At 24, 48, 72, and 96 h following pollination, pollen tube growth was observed. The observed samples were fixed, softened, and stained [13], then a rotary microtome (Leica CM3050S, Germany) was used to cut sample into longitudinal slices (90 μm), and fluorescence microscopy (Leica DM5000 B, Germany) was used to observe the sample.

Comparison of the lengths of dorsal sepals, petals, and labellums of D. chrysotoxum was assessed using the Duncan test while comparing the morphological traits of D. chrysotoxum flowers and the pollinator was carried out using the independent sample T test by SPSS version 19.

Flowering of D. chrysotoxum began in early March, peaked in mid-April, and finished in mid-May. An individual flower had a life span of about 12 days under natural conditions. Fertilized flowers would wither in approximately 3 days after pollination. The inflorescence of D. chrysotoxum was a raceme with 10 to 20 flowers (Fig. 1A). Furthermore, the flowers of D. chrysotoxum often emitted a mild fragrant scent when exposed to direct sunlight or high temperatures during the day. The flowers of D. chrysotoxum were both zygomorphic and hermaphroditic (Figs. 1C, 1F and 1G). The sepals were golden and oval. The petals were yellow and ovate. The labellums were large oval, yellow, and had a sorrel-colored core with trichomes (Fig. 1B). The mean length and width of lateral sepals, central sepal, petals and labellum were 21.13 mm (SD = 1.2, N = 30) and 9.91 mm (SD = 0.6, N = 30), 21.30 mm (SD = 1.3, N = 30) and 10.18 mm (SD = 0.5, N = 30), 24.99 mm (SD = 1.2, N = 30) and 15.49 mm (SD = 0.8, N = 30), and 25.90 mm (SD = 1.1, N = 30) and 28.97 mm (SD = 1.9, N = 30), respectively. A short tissue connected the top of the yellow anther cap to the gynostemium (Figs. 1I–1K). The pollinia were ceraceous and yellow under the anther cap (Fig. 1L). The flowers of D. chrysotoxum possessed a single nectar gland at the bottom of the spur (Figs. 1D, 1E and 1H). Base on the fact that nectar secretions were found on every single one of the 30 flowers analyzed, the spur dissection experiment results revealed that D. chrysotoxum generates nectar.

We discovered that hymenopterans were the most abundant insect visitors based on the observed visitations. It was also noted that members of the Coleoptera and Diptera visited the flowers of D. chrysotoxum, but they did not engage in typical pollinator behavior; rather, the majority of them only simply landed or rested on the flowers. Although the thrip Frankliniella occidentalis and the ant Camponotus japonicus couldn’t get nectar and could not carry pollinia due to their small bodies, occasionally nectar-seeking behavior was observed. In the garden, we counted 29 species of visiting insects, although the majority were not actual pollinators (Table 1). Only one species of the visiting insects was observed to pollinate, depositing pollinia and removing them from the scutum of Ctenoplectra davidi Valhalla (Hymenoptera: Apidae) (Figs. 2A and 2B). The pollinators of D. chrysotoxum were exceptionally rare, hardly observable in the garden. On sunny or warm days, C. davidi usually appeared. C. davidi visited the flowers most frequently between the hours of 13:00 and 15:00 one day in April. They frequently flew quickly and landed on the flower of D. chrysotoxum. Then C. davidi crawled toward the sorrel center of the labellum with trichomes and probed into the spur cavity. The C. davidi was prevented from entering by the labellum’s protrusion. They held labellum with their fore and middle legs and inserted their mouthparts deeply into the spur to find and suck nectar at the base of the passage with bitingsucking mouthparts. By pushing back of the labellum with a flexible joint, a retreating C. davidi of appropriate size obtained smearing of stigmatic liquid on its scutum, its scutellum sticked anther cap and pushed back the anther cap, coupling with the anther cap falling (Fig. 2C). ‘Tipping-like’ anther cap delivered the pollinia precisely to the scutum of C. davidi. After C. davidi has left the flower, the pollinia could be positioned on the scutum in the middle of its thorax. If a C. davidi carrying pollinia on its scutum entered a second flower of D. chrysotoxum that lacked pollinia on its receptive stigma. Then C. davidi leaved the flower, pollinia was placed on the stigma precisely by pushing back the labellum (Fig. 3).

Some more Dendrobium species were blooming at the same time as D. chrysotoxum, such as D. thyrsiflorum, D. fimbriatum, and D. nobile. Their gynostemia were intact, and the anther cap had not been pushed away. So, their pollinia were not lost and never removed by C. davidi. As a result, it was certain that the pollen carried by C. davidi originated from D. chrysotoxum. This finding agreed with C. davidi being the pollinator of D. chrysotoxum. Therefore, we could surmise from the observation that C. davidi was the only pollinator of D. chrysotoxum. A comparison of the pollinator and D. chrysotoxum flower sizes provided evidence to support this hypothesis (Table 2). The mean height of C. davidi thorax was 2.46 mm (SD = 0.12, N = 11), the mean width of C. davidi body was 3.03 mm (SD = 0.14, N = 11), the mean length between the midpoint of the thorax and the endpoint of the proboscis was 4.54 mm (SD = 0.23, N = 11). Then the flower passage, which had a mean height of 2.75 mm (SD = 0.24, N = 30), a mean width of 3.44 mm (SD = 0.26, N = 30), a mean depth of 4.53 mm (SD = 0.26, N = 30), fitted the pollinator needs perfectly. The chances that C. davidi would remove pollinia and ensure that the pollinia were transferred to another flower were increased by these morphological changes to the flowers. These results indicated that the floral structure of D. chrysotoxum was well adapted to its pollinators. When C. davidi was probing the spur cavity, the curved gynostemium of the flower exactly fitted the scutum and scutellum between the mesothorax and the abdomen of C. davidi. The protrusion of labellum made the C. davidi arch its body, and thus the hind edge of the mesothorax rose up and reached a position towarded the anther cap and stigma when the C. davidi lowered its head to probe deep into the spur cavity (Figs. 4A–4C). The pollinia changed their direction passively and pointed towards the stigma as C. davidi left and entered the flower, and pollination was completed.

Only 0.17% of capsules setting were naturally pollinated in 2019, which was quite low. No capsule setting of D. chrysotoxum was formed under natural pollination conditions in 2020. No capsule setting of D. chrysotoxum was formed by direct autogamy in 2019 and 2020. The capsules setting of flowers treated by self-pollination was 17.75%, 5.50% in 2019 and 2020, respectively, while the capsules setting of flowers treated by cross-pollination was 85.00% and 99.25% (Table 3). The data showed very low capsules setting at natural pollination conditions, especially there was no capsule setting in the absence of pollinators. Whereas artificial pollination was known to yield high capsules setting rates, both when the flowers were treated by self-pollination or cross-pollination, proving that they were cross-compatible and partially self-compatible. During pollen tube growth of self-pollinated individuals, the pollen tube grew very slowly and was short (Figs. 5A–5D), black spots were present after 48 h of self-pollination (Fig. 5B), and the zone of black spots became larger until the pollen tube grew into the ovary (Fig. 5D). This might be due to the degradation of callose in the pollen tube which cannot make the pollen tube fluorescent, leading to the partly Self-incompatibility phenotype. Inversely, the pollen tube bundle of cross-pollinated individuals grew faster, reaching the ovaries in 72 h or 96 h after cross-pollination (Figs. 5E–5H).