Ophiorrhiza japonica is a hermaphroditic perennial herb (Institute of Botany, the Chinese Academy of Sciences, 1983). We observed that both red- and white-flowered individuals coexist in populations on Shibing Yuntai Mountain, located in the South China Karst World Heritage Site (26°55′13″N 108°07′47″E). Field experiments were conducted from 2016 to 2019.

To assess whether the two flower color morphs differed in traits relevant to pollination, we compared various plant traits between pink and white morphs. We measured these traits using tape measure and vernier calipers. Specifically, plant height, leaf length, and leaf width were measured with an accuracy of 1 mm using tape measure, while tube length, tube diameter, flower length, flower width, stamen length, anther length, and pistil length were measured with an accuracy of 0.01 mm using caliper micrometers. More than 30 random selected individuals of each color morph were measured. These measurements were taken in April and December 2018. Due to the complexity of the terrain, the use of the sample plot survey method proved inadequate for accurately reflecting the distribution of O. japonica. Therefore, we opted for a sample line survey approach. The sampling route followed the valley from the bottom to the top, beginning at the first identified location. We compared the distribution of pink and white morphs across these populations. We obtained 15 observation locations in spring(March, 2019) and 14 locations in winter(December, 2018) where we recorded the number of plants with pink and white morphs and calculated their ratio. Each O. japonica over 50-meter interval was recorded as a location, and all plants within a 5-meter radius of the initial population were counted.

To test whether pollinators exhibited a preference between two color morphs, we conducted observations of visitor groups on pink and white color morphs. To ensure the accuracy of our results, we conducted observations at three locations with both pink and white morphs, five locations with only pink morph, and five locations with only white morph. We recorded the number of open flowers of each morph in these locations, with each location containing more than 50 individuals. Observations were made at 15-minute intervals between 10:00 a.m. and 5:00 p.m. in daytime, during periods of active foraging by pollinators in favorable weather conditions. Using a night camera (Ruiyucheng R-71) for two consecutive nights during the winter months when weather conditions are favorable. Additionally, in spring, random observations of 1 to 2 hours will be conducted each night between 7 pm to10 pm accumulating more than 5 hours cumulative to determine the presence of nocturnal pollinators. We recorded visitors as pollinators if their bodies came into contact with anthers and noted their visit numbers. Additionally, we tracked pollinators moving among individuals to quantify visit frequencies to two color morphs (visits per flower per hour). Observations were conducted during several discrete intervals: December 12–24, 2016; December 1–10, 2017; January 17–22, 2018; April 1–10, 2018; December 24, 2018–January 10, 2019; and March 19–28, 2019. The total cumulative time reaches 60 hours.

To distinguish color differences between pink and white flowers under spatial vision in insects, we employed Ocean Optics equipment to measure the reflectance spectra of both flower morphs. We recorded the diffuse spectral reflectance (300-700 nm) of individual flowers using a spectrometer (RPH-1) relative to a white reflectance standard (WS-1) under a deuterium/tungsten light source (Ocean Optics DH-2000-BAL), following the methods described by Chittka and Kevan (2005) and Ohashi et al. (2015).

To compare reproductive success between two flower morphs, we randomly selected at least 30 healthy flowers from each treatment group for each morph to undergo bagging and performed artificial pollination when these flowers were fertile. Each flowering plant was marked according to the treatment. We then evaluated the differences in fruit set by open pollination, intra-morph pollination, and inter-morph pollination for both pink and white morphs. Open pollination was performed without any additional treatment. Intra-morph pollination involved the application of pollen from individuals of the same color morph, whereas inter-morph pollination involved the application of pollen from the opposite color morph. Flowers were bagged with mesh until they reached the female phase, and the different treatments were marked with cotton threads of various colors. Each pollination type was replicated approximately 30 times. After pollination, flowers were re-bagged to exclude pollinators. Three weeks later, we collected all labeled capsules, counted seeds, and undeveloped ovules within each capsule. These procedures were conducted in both spring (March 2018 and April 2017) and winter (December 2017 and January 2018) of both 2017 and 2018.

To investigate whether color dimorphism is associated with various environmental conditions, we random measured humidity and light intensity in over 20 observing locations using a hygrometer (KIMO HD100) and a digital luxmeter (Field Scout Foot-candle Meter, Spectrum Technologies, Inc.). To standardize light measurements, we calculated the light application ratio, defined as the ratio of actual illuminance to natural illuminance in the absence of shelter. We employed a sampling survey method to locate pure white and pure red plants along the valley, from the bottom to the top. Starting at the first identified site, we recorded the color of any pure-colored flowers encountered and established a 3-meter by 3-meter plot, ensuring that all positions were spaced at least 50 meters apart. After measuring the humidity and light intensity within the plot, we counted the number of blooming flowers.

To assess the temperature differences experienced by pink and white flowers in the same wild environment, we used a thermal infrared imager (FLIR T650sc) to measure flower temperatures. FLIR T650sc is an intelligent instrument that monitors the surface temperature of objects in real time. By capturing infrared images and transferring them to a computer for professional software processing (FLIR Thermal Studio Suite 1.6.10), real-time temperature can be obtained with great accuracy, capable of sensing tiny temperature differences. Temperature is analyzed and compared in degrees Celsius.

To ensure consistency, we randomly selected pink and white flowers from mixed-morph populations. To account for environmental variability, measurements were conducted daily between 10:00 a.m. and 4:00 p.m. on several warm and cold days in both spring (March) in 2018 and winter (January) in 2019.The measurement sites refer to locations where flower color statistics are collected. Each location includes at least two flower color phenotypes simultaneously, with measurements taken hourly. During each measurement, the temperatures of the center and four sides of the flower and leaf are recorded, along with the background environmental temperature. Each season, measurements are conducted at a minimum of 10 sites for the two different flower colors.

To compare the response of the two morphs to fluctuating temperatures, we employed an artificial climate box to control environmental conditions. In 2019, we transplanted six to eight individuals from the wild into an indoor setting. The plants were cultivated indoors for several days until they were well-established and prepared for the experiment. We assessed temperature responses at -4°C, 0°C, 4°C, 8°C, 12°C, 16°C, 20°C, and 24°C using the thermal infrared imager. Each temperature was tested three times in an artificial climate box (SRG-D400B), with each trial including more than three flowers per morph.

We compared fruit set, plant traits, and visit frequency between the two morphs using a general linear model (GLM) in SPSS version 20. Temperature between the two morphs was also assessed using SPSS version 20. Additionally, excel 2020 was used to model and calculate the effects of pollinator vision on morphs.

To assess the temperature differences experienced by pink and white flowers, we employed a general linear model (GLM) in SPSS to evaluate the effects of three variables—light application ratio, humidity, and color—on flowering quantity. Additionally, we analyzed the interaction effects among these three variables.

To examine whether pink and white morphs look different to pollinators, we compared the spectral reflectance of leaflets, petals in O. japonica. Eight flowers from each morph were examined at 300–700 nm range with an Ocean Optics JAZ-EL200 (Ocean Optics Inc., Dunedin, Fla.) spectrometer, with a fiber optic reflection probe (QR400-7-SR) held at 45° to the petal and leaflet surface (Johnson and Andersson, 2002).

Color patterns perceived by pollinators were represented by plotting the reflectance of the petal of O. japonica (pink and white morph) in a bee color hexagon model, divided into six sectors, blue (B), blue-green (BG), green (G), ultraviolet-green (UG), ultraviolet (U), and ultraviolet-blue (UB) and one uncolored central circle (Chittka, 1992), in accord with the colors perceived by bees (Backhaus and Menzel, 1987; Jersáková et al., 2006).

Throughout the flowering period in Shibing Karst, all flowers were either pink or white ( Figure 1 ). Significant differences in plant traits were observed between the two morphs. White morph individuals were taller than pink morph (P < 0.01) in winter, and their leaves were larger (P < 0.01) than those of pink morph ( Table 1 ). Conversely, in spring, the traits of the pink morph were larger than those of the white morph ( Table 1 ). Petal size also varied seasonally: white flowers were longer and wider in winter than pink flowers but shorter in spring.

Analysis of color morph ratios revealed that, in spring, data from 15 locations and 3,775 flowers indicated that the proportion of pink morphs was 41.4% higher than that of white morphs. In winter, data from 14 locations and 255 flowers showed that the proportion of white morphs was 39.3% higher than that of pink morphs ( Figure 2 ).

We observed visitor activitiesbetween the two flower color morphs during both winter and spring. In winter, visitor activity was minimal in daytime, with no visitors recorded during nighttime observations. In spring, bumblebees were the primary visitors to both pink (0.312 ± 0.04 visit/hour/flower) and white (0.278 ± 0.04 visit/hour/flower) morphs, and no significant difference in visit frequency was found between the two morphs. Honeybees (pink, 0.002 ± 0.02 visit/hour/flower; white, 0.041 ± 0.03 visit/hour/flower) and hoverflies (pink, 0.006 ± 0.03 visit/hour/flower; white, 0.004 ± 0.02 visit/hour/flower) were observed infrequently. Moths were seen with a high frequency of visits and rapid movement and visitation during the day; however, their overall occurrence was rare, with low total frequencies (pink, 0.037 ± 0.03 visit/hour/flower; white, 0.017 ± 0.02 visit/hour/flower). No significant difference (P > 0.05) in visit frequency was observed between the two morphs ( Figure 3 ). The reflectance in the blue and green spectra differed markedly between the two morphs ( Figure 4 ). We assessed the ability of bees to recognize white and pink flowers against leaf backgrounds. The results indicated that both flower morphs were poorly distinguishable, with no significant difference in detectability by bees ( Figure 5 ).

In winter, the fruit set percentages for the three pollination treatments in white morphs were as follows: inter-morph pollination (63.6 ± 15.2%), intra-morph pollination (53.3 ± 13.3%), and open pollination (63.2 ± 8.0%). No significant differences were observed between the treatments. For pink morphs in winter, the fruit sets were: inter-morph pollination (42.9 ± 13.7%), intra-morph pollination (26.7 ± 11.8%), and open pollination (47.4 ± 6.7%). Similarly, no significant differences were found between the treatments (P > 0.05). Additionally, within the same season, there were no significant differences in fruit set between white and pink morphs for any treatment ( Figure 6 ).

In spring, the fruit sets for the three pollination treatments in white morphs were: inter-morph pollination (55.6 ± 17.6%), intra-morph pollination (70.6 ± 11.4%), and open pollination (85.7 ± 14.3%). No significant differences were detected between the treatments. For pink morphs in spring, the fruit set percentages were: inter-morph pollination (70.0 ± 15.2%), intra-morph pollination (77.4 ± 7.6%), and open pollination (72.2 ± 10.9%). No significant differences were observed between the treatments. Similarly, within the same season, there were no significant differences in fruit set between the white and pink morphs for any treatment ( Figure 6 ).

We assessed various environmental factors for the two color morphs at 20 measurement locations. The average illuminance in biotopes without shelter was 37.95 ± 1.92 lux, compared to 11.29 ± 0.84 lux in biotopes with vegetation. Given the significant variation in illumination due to weather conditions, we examined the correlation between light application ratios and color polymorphism, which was found to be low ( Table 2 ). The average humidity ranged from 80% to 90% and showed a low correlation with color polymorphism ( Table 2 ). Under high-temperature conditions, pink flowers were warmer than white flowers, whereas under low-temperature conditions, white flowers were warmer than pink flowers ( Figure 7 ). These results suggest that pink flowers exhibit higher activity at elevated temperatures and lower activity at cooler temperatures than white flowers. We also measured temperature differences between the two morphs across various temperature gradients. At -4°C, pink flowers were cooler than white flowers, but from 0°C to 20°C, pink flowers were warmer than white flowers ( Figure 8 ).

Flexibility in pollinator color preferences can significantly affect the evolution of floral traits. Variations in pollinator preferences result in increased floral visitation, which in turn drives selection for favored flower types (Waser and Price, 1981; Levin and Brack, 1995; Schemske and Bradshaw, 1999; Aldridge and Campbell, 2007).

Previous research has revealed that pollinator-mediated selection is unexpectedly limited (Trunschke et al., 2021). In a phenotypic selection study, leafcutting bees were found to select flowers based on hue or flower color, whereas bumblebees chose flowers based on chroma or darkness (Brunet et al., 2021). Wassink and Caruso (2013) observed significant net selection for color saturation in Lobelia siphilitica. Flower pigment significantly influenced both pollinator visitation and resistance to seed predators, resulting in substantial net selection of flower color (Carlson and Holsinger, 2012; Veiga et al., 2015). In a population of Iris pumilla, a significant increase in anthocyanin concentration was detected in the blue-flowered morph, whereas no such selection was observed in the purple morph (Souto-Vilarósa et al., 2018).

In our study, neither the frequency of visitors nor the visual perception of bees showed significant differences between the two color morphs, and there were significant changes in the morphological traits of the flowers. This suggests that color alone is unlikely to influence visitor preference between the two morphs. Additionally, fruit set data indicated that flower color did not affect reproductive success. Thus, color polymorphism in Ophiorrhiza japonica may not be an adaptation to pollinators. However, the number of pink morphs was greater than that of the white morphs. We also investigated the interaction between flower color and light application ratio but found no significant effect. In contrast to our previous studies on other species in the Karst region, which indicate that flower color dimorphism may be maintained by pollinator-mediated selection, this study examined the case of Allium wallichii (Tang et al., 2020). Therefore, the potential role of pollinators in maintaining color polymorphism in other regions should be carefully considered.

Many enzymes involved in anthocyanin synthesis are crucial for the production of other flavonoid compounds. These compounds affect not only flower color but also various ecological and physiological traits in plants (Shirley, 1996). Consequently, several researchers have proposed that the evolution of flower color may be influenced more by the selection on these pleiotropic effects than by pollinator-driven selection, or that both factors may contribute (Rausher and Fry, 1993; Simms and Bucher, 1996; Fineblum and Rausher, 1997; Armbruster, 2002; Irwin et al., 2003).

Furthermore, pigments responsible for flower color may play a role in resistance to abiotic stressors, such as high temperatures, low precipitation, high ultraviolet radiation, and low productivity environments (Schemske and Bierzychudek, 2001; Chalker-Scott, 2008; Arista et al., 2013; Koski and Ashman, 2015). Flavonoid development can be influenced by temperature (Ben-Tal and King, 1997; Jaakola and Hohtola, 2010). Evidence indicates that temperature can differentially impact the fitness of flower morphs with low versus high color saturation (Coberly and Rausher, 2003, 2008). If anthocyanins provide heat stress resistance, then higher temperatures are expected to favor flowers with greater chromatic contrast than those at lower temperatures.

Ophiorrhiza japonica exhibits a prolonged flowering period that spans both spring and winter, and is susceptible to significant temperature fluctuations. Observations indicate that pink flowers are warmer than white flowers at high temperatures, but cooler at low temperatures, suggesting that pink flowers are more active under warmer conditions. Temperature tests revealed that the temperature response between the two morphs varied across different temperature gradients, implying that temperature may influence flower color polymorphism. Generally, plant temperature reflects its activity level (Kim et al., 2016). Within the 0-20°C range, pink morphs demonstrated higher activity compared to white morphs, while white morphs exhibited greater activity at -4°C. No significant differences were observed at temperatures greater than 24°C. In spring, most characteristics of pink individuals are significantly larger than those of white individuals; however, in winter, the trend reversed, suggesting that pink morphs might possess an adaptive advantage to heat. These findings indicate that flower color in O. japonica is associated with heat stress, with pink morphs showing better fitness under such conditions. Thus, flower color polymorphism may be primarily associated with heat stress rather than honeybee preference or other factors affecting color and morphology. These findings align with those of studies on Gentiana leucomelaena, which exhibited different morphs adapted to varying temperature conditions (Mu et al., 2010). Our previous research also supports these results, demonstrating that temperature also plays a crucial role in maintaining flower color polymorphism in Geranium nepalense (Tang et al., 2016).

Coberly and Rausher (2003) demonstrated that flower color, regulated by the flavone gene, can alleviate heat stress. High temperature negatively impacts the fertilization success of both male and female homozygous Ipomoea purpurea individuals. Additionally, individuals with the white morph, resulting from a loss-of-function mutation, exhibit reduced fertilization success at high temperatures compared with pigmented individuals, although this effect is not observed at low temperatures. These results align with observations and further support the role of flavones in regulating flower color in Ophiorrhiza japonica (e.g. Sun et al., 2019, 2021).

The MADS-box transcription factor FLOWERING LOCUS M (FLM) has been implicated in phenotypic variation related to plant growth and color (Lee et al., 2013). Leaf temperature in flm-3 mutant plants was lower than that in wild-type plants. To quantify the adaptive effects of FLM alleles in natural environments, common garden experiments should be conducted at locations with varying temperature fluctuations and precipitation levels (Ågren and Schemske, 2012). FLM is a well-known transcription factor whose splicing variation in response to temperature change modulates flowering time (Scortecci et al., 2001; Balasubramanian et al., 2006). In the present study, the pink morph produced more individuals than the white morph in spring, whereas the white morph produced more individuals than the pink morph in winter. At higher temperatures, pink flowers were warmer than white flowers, while white flowers were warmer than pink flowers under low-temperature conditions. Thus, the pink morph is more adaptive to warmer weather conditions, whereas the white morph is better suited to cooler conditions.