Pollinator activity and pollination success of Medicago sativa L. in a natural and a managed population

Abstract Medicago sativa L. is an important cash crop in the arid region of northwest China. Pollinator activity is an essential aspect of pollination success, but the relationships between pollinator visitation rate and seed set still need further study of M. sativa. We investigated the following characteristics of M. sativa in natural and managed populations: floral traits, pollinator activity, and breeding system. Our results indicated the management could affect the number of flowers produced; however, there was no detectable effect on the seed set per flower. We found the percentage of seeds among pollinated flowers in the managed population was significantly higher than that in the natural population. Moreover, the increase in the proportion of pollinated flowers could significantly increase seed set per flower, and pollinator visitation rate was the important limiting factor for seed set in both populations. Andrena lebedevi Popov was found to be the most frequent pollinator in both populations. Outcrossing was dominant in the breeding system and insect pollination played an important role in outcrossing. Our study suggested that proper management (artificial selection) could promote pollination success of M. sativa.

. Kishore, Shukla, Babu, Sarangi, and Patanayak (2012) also indicated the function of floral traits may not only to facilitate pollination by the primary pollinator but also to restrict other potential pollinators.
Pollinator visits are likely to be less frequent in stressful habitats (Neto, 2013). More specifically, it has been demonstrated that pollinators respond strongly to the local abundance of flowering plants, but habitat fragmentation may reduce the necessary resources to support resident pollinators (Kunin, 1997;Wagenius & Lyon, 2010).
Fragmented habitats result in isolation and edge effects, and disrupting species interactions, such as plant-animal mutualisms. Steffan-Dewenter, Klein, Gaebele, Alfert, and Tscharntke (2006) indicated pollinator abundance and activity decline with fragmented habitats owing to the reduction in floral rewards or habitats that could not meet the nesting requirements of pollinators. Similar pattern of pollinator activity was also documented between the wild and managed populations of Myrtillocactus schenckii (Cactaceae) (Fernando, Stoner, Pérez-Negrón, & Casas, 2010). Recent studies have also shown that management may affect flowering patterns, pollinator foraging behavior, and pollination processes (Chen, Zhao, & Zuo, 2015;Quesada et al., 2004).
Medicago sativa is an important cash crop and valuable member of the plant community in the arid region of northwest China.
Medicago sativa has great potential in terms of forage and medicinal uses (Jiang, Bi, He, & Zhang, 2003). Communities of this plant play a critical role in sand fixation and vegetation productivity.
This study aimed to test the effect of human management on

| Species
Medicago sativa is usually 0.3-1.0 m in height and mainly distributed throughout Gansu, Ningxia and western Inner Mongolia provinces in China. There are three species, one subspecies in China. Medicago sativa is a perennial plant that varies in racemes (inflorescences) per stem and open flowers per raceme, creating large variation in floral display size among plants (Pedersen, 1953). In the present study, we selected purple flowers of M. sativa, and this species has four petals and 10 stamens (Jiang et al., 2003). In addition, M. sativa have a tripping mechanism, and the flower remains open following tripping, with the stigma and anthers exposed. In order to collect pollen and nectar, pollinators must trip a flower to release the anthers and the stigma.

| Study area and experimental layout
The study area was at the Urat Desert-grassland in western Inner Mongolia of China (between 41°06′-41°25′N and 106°59′-107°05′E), and the annual mean rainfall is approximately 153.6 mm.
This study was carried out from April 2013 to October 2017. The experimental layout consisted of two studied patches and six plots in total, three natural plots (30 × 30 m) and three large managed plots (30 × 30 m). For this study, the studied plots occurred in an area that was originally dry, arid land. In the natural patch, plants survives grew naturally without any artificial management, and the average density of M. sativa was 15 individuals per 100 m 2 , and the two patches were separated by 100 m in the study area. In natural population, there were some general plants, such as Reaumuria songarica (Pall.) Maxim and Salsola passerina Bunge. Moreover, the flowering time of these species were no overlap with M. sativa. In the managed population, the average density of M. sativa was also 15 individuals per 100 m 2 , and other species but M. sativa were cleared (five times per year).
We also remove vegetation and provide water and nutrient in the managed patch. The managed population is nearly 300 m away from the natural plots, which had similar landforms to the natural population ( Figure 1).

| Phenology
Ten individual M. sativa plants per population were labeled and observed for phenological characters. The flower buds, flowers in anthesis and fruits on each branch were counted throughout the entire reproductive season (from April to September) for each plant. We then calculated the proportion of flower buds, flowers in anthesis and mature fruits.

| Floral traits
We marked 20 plants per population while they were still budding to facilitate the assessment of the flowers when they were completely open. Moreover, the timing of the following events was also recorded. Video filming was conducted continually throughout anthesis in 3-4 flowers in marked plants. In addition, we measured the structure of inflorescences, the length of flower (from petal tip to flower base), calyx and flower stalk.

| Managed experiment
In study area, we marked 36 flowering plants ( (Rowan et al., 2008). We collected the flower number and seeds set per flower produced by the control, water, and fertilizer treatments. Moreover, we counted the number of seeds and ovules per flower according to the following equation:

| Pollinator observations
To determine the identities and quantities of pollinators, we conducted surveys of pollinators from May to July. In each population.
We labeled 20 racemes (10 flowers per raceme) and repeatedly observed throughout the complete process of anthesis. We used fixed video cameras to assess the duration of each pollinator visit, including pollinators that were collecting pollen and nectar. We carefully analyzed the presence or absence of pollen grains adhering to the bodies of the pollinators and determined whether they contacted stamens and stigmas. In addition, images of the pollinators were used for identification in the laboratory. The flower visitation frequency of pollinators was calculated according to the following equation (Goverde, Schweizer, Baur, & Erhardt, 2002):

| Effect of pollinator visitation rate on pollination
We marked 18 flowering plants (nine natural and nine managed plants) while they were still budding. For each plant, 20 flower buds were randomly chosen and marked with tags. We noted the flowering stages and growth progress of the marked flowers throughout video filming.
To determine how pollinator visitation rates affected pollination success, we investigated the proportions of flowers anthesis, pollinator visitation, pollination, and seed production at 2-week intervals from May to September when all seeds were mature.
Fruits produced were collected in early August and the length of fruits (pots) and the number of seed were examined in the laboratory. Ovary enlargement was a valid criterion for assessing fertilization of ovules (Garwood & Horvitz, 1985;Suzuki, 2000). In these bagged and not pollinated flowers, we measured the length of ovaries in these un-pollinated flowers when other flowers were fruiting. The average length of ovaries was 5 ± 0.39 mm (n = 20, Mean ± SE). Flowers with ovaries (pots) longer than 5 mm were classified as successfully pollinated. Conversely, even if flowers were tripped open, these flowers in which the length of ovary was less than 5 mm were classified as visited by pollinators but not as successfully pollinated. The percent of fresh flowers and mature seeds were recorded and calculated according to the following equation (Suzuki, 2000): where P is the proportion of pollinated flowers, V is the proportion of visited flowers, and S is the proportion of seeding per flower on marked plants. Percentage of seeds among pollinated flowers = S P × 100%

| Breeding system
The experimental layout consisted of two studied patches and 6 plots in total, three natural plots (30 × 30 m) and three large managed plots (30 × 30 m). Natural plots were mirrored in a symmetrical arrangement and surrounded by undisturbed vegetation (gray area).
In the managed patch, other species but Medicago sativa were cleared (white area). We remove vegetation and provide water and nutrient in the managed patch experimental treatments were conducted prior to anthesis, in which we sampled 360 flowers (10 flowers per raceme, six racemes per plant) in each population. Each treatment has 60 flowers, and these flowers were on the same plant. Experimental treatments were as follows: Natural pollination (control): the flowers were marked and maintained under natural conditions.
Manual cross-pollination: samples of flowers received additional pollen (hand-pollination) collected from individual plants more than 20 m away, and flowers that received the pollen were emasculated before pollen liberation.
Nonmanipulated cross-pollination: the flowers were without any manipulation, and these flowers were also emasculated.
Manual self-pollination: flowers were pollinated with their own pollen and then bagged to prevent insects and wind pollination.
Nonmanipulated self-pollination: the inflorescences were bagged without any manipulation of the flowers.
Emasculation and netting: the stamens of the flowers were removed before pollen liberation, and nets of 1 mm 2 mesh were used to prevent insect pollination.
In October, we counted the number of seeds and ovules per flower for each treatment.

| Self-compatibility Index
We calculated the self-compatibility index (SCI) according to the following equation: Self-compatibility index values of ≤0.2 indicate selfincompatibility, whereas values >0.2 show self-compatibility (Zapata & Arroyo, 1978). We used ANOVA to compare the production of flowers in anthesis and mature fruits between natural and managed populations.

| Data analyses
All analyses were performed using the statistical software package SPSS 19.0 for Windows (SPSS Inc., Chicago, IL, USA).

| Phenology
In both populations, flowering of M. sativa typically occurs from late May until late July. However, the period of peak flowering was different between the natural and managed populations, with the latter experiencing a longer flowering period (Figure 2).
Our results indicated that mature fruits were available from June to September, though fruit production peaked at different times among the populations. The natural population reached its peak at mid-August, while the managed population peaked during the last week of July ( Figure 2). In addition, the average number of mature fruits per branch in the managed population was Self-compatibility index = The seed set of self-pollination The seed set of manual cross-pollination

| Managed experiment
In natural population, our results indicated that water-added increased the proportion of flowers in anthesis and differed significantly between the water-added and control (F = 38.26, p < 0.05). In addition, the proportion of flowers in anthesis was similar between the water-and fertilizer-added (F = 0.20, p > 0.05; Figure 3). We also found strong effects of both water-and fertilizer-added on the proportion of flowers in anthesis in the both populations.
In the natural population, the seed set per flower did not differ significantly between the control and fertilizer-added flowers, at 31.2 ± 4.3% in the control and 33.5 ± 5.6% in the fertilizer-added flowers. The seed set per flower also did not significantly differ between the control and the water-added flowers (32.3 ± 4.7%; F = 1.21, p > 0.05). Moreover, the control, water-and fertilizer-added plants produced similar the mean seed set per flower in the managed plots, indicating that water-and fertilizer-added had no detectable effect on the mean seed set per flower.

| Pollinators and pollinator activity
In the studied patches, M. sativa were visited by Andrena lebedevi

| Pollinator visitation rate
In the natural population, 49.21% of flowers were visited (V) at least once by effective pollinators, and 38.56% of flowers were  Figures 6 and 7). These outcomes showed that higher pollinator visitation rates resulted in a higher percentage of seeds. In both populations, our results indicated that the pollinator visitation rate (49.21% in the natural and 57.35% in the managed) was the important limiting factor for seed set.

| Breeding systems and self-compatibility index
The seed set obtained in each pollination treatment is shown in Figure 8. The seed set of manual cross-pollination was 56.2 ± 6.1% in the natural population, and 67.1 ± 6.3% in the managed population. We found the seed set of the managed population was significantly higher than that of the natural population (p < 0.05). In addition, our results indicated that the natural pollination seed set was 30.5 ± 3.5% for the natural population and 38.7 ± 4.3% for the managed population. In both populations, the seed set of manual cross-pollination was significantly higher compared with the seed set of natural pollination (natural and managed population: p < 0.01).
We found that outcrossing was dominant in the breeding system of M. sativa.
The nonmanipulated self-pollination treatment resulted in 4.1 ± 0.4% seed set in the natural population, and 4.9 ± 0.5% in the managed population. Under the emasculation and netting treatment, the seed set was only 10.3 ± 1.2% for the natural population and 11.6 ± 1.5% for the managed population. In both populations, the seed set of nonmanipulated cross-pollination was significantly In the natural population, 5.7% of the seed set under manual selfpollination and 56.2% in the manually cross-pollinated set. The SCI value was 0.10, which indicated this species was self-incompatible.

| Effect of floral traits and pollinator behavior on seed set
Floral traits might have evolved in order to attract and utilize more groups of pollinators in most flowering plants (Fenster et al., 2004).
The relationship between the proportion of pollinated flowers and the pollinator visitation rate for Medicago sativa F I G U R E 7 The relationship between the proportion of seeds per flower and the pollinator visitation rate for Medicago sativa F I G U R E 8 Seed set of Medicago sativa in different treatments Sandring and Agren (2009) have pointed out that the evolution of floral display is commonly attributed to pollinator-mediated selection in most animal-pollinated plants. Kishore et al. (2012) indicated the function of floral traits may not only to facilitate pollination by the primary pollinator but also to restrict other potential pollinators, and these traits could be better transferred pollen. In this study, the period of peak flowering of M. sativa were longer in the managed than that in the natural population. Moreover, the flowering peak is the highest offer of pollen and increases the possibility of insect pollination at this moment in M. sativa. In many species, floral differences between both populations have been associated with environmental heterogeneity, which would be magnified by human disturbance (Chen et al., 2015). The difference of management suggested that water-and fertilizer-added could affect the number of flowers produced, but had no detectable effect on the mean seed set per flower.
In M. sativa, the stigma and style emerge and protrude from the keel, and pollinators usually visit the lower side of the petals.
When the insects leave the flower, the stigma of tripped flowers pressed against the banner petal, providing a signal to pollinators that the flower has been visited and increase the efficiency of insect pollination. Thus, the tripping mechanism of M. sativa allows these flowers to achieve efficient insect pollination, and outcrossing also favors the production of a diverse population. Moreover, Strickler and Freitas (1999)

| Pollinator visitation rates limit pollination success
Pollination success is related to pollinator type, as different pollinators vary in pollination effectiveness (Gómez, Bosch, Perfectti, Fernández, & Abdelaziz, 2007). Goverde et al. (2002) reported that a managed environment could affect pollinator diversity and abundance in the natural population. In our study, the single flowering period in the managed flowers was longer than that in the natural flowers. Moreover, the total flower availability per day was significantly higher in the managed population than that in the natural population, which could explain the difference of pollinator visitation rates between both populations.
Many researchers have studied the relationships between pollinator visitation rates and plant reproduction, but how pollinator visitation rates affect plant reproductive success still need further study (Bauer, Clayton, & Brunet, 2017;Sletvold & Agren, 2010;Stephenson, 1981;Zimmerman & Pyke, 1988). In present study, we were able to identify whether flowers had been visited by effective pollinators because of the tripping mechanism of M. sativa.
A similar pattern of pollinator visitation also documented in Cytisus scoparius (Leguminosae), reaffirming the effect of pollinator visitation rates on plant reproduction (Suzuki, 2000). Bauer et al. (2017) suggest that the increase in the proportion of pollinated flowers combined significantly increased seed set per raceme. Our outcomes also support this view that both the regressions between the proportion of pollinated flowers and seed set per flower was statistically significant, indicating that the increase in the proportion of pollinated flowers could significantly increase seed set per flower and the visitation rate is an important limiting factor for pollination success.

| Pollinator activity and breeding system in different populations
Pollination by animals is largely considered a co-adaptive process in which plants evolve traits to attract certain pollinators, whereby pollinators improve the efficiency of their activities to better exploit the floral resources of plants (Sharma, Shaanker, Leather, Vasudeva, & Shivanna, 2011;Sunnichan, Mohan Ram, & Shivanna, 2004). In M. sativa, the period between 08:00 and 12:00 appeared to be crucial for pollination of most flowers since this was when the flowers had the largest opening and there was overlap in pollen release. This period coincided with the most active time for A. lebedevi. When A. lebedevi contacted the anthers, a large amount of pollen was collected on its legs and because of its suitable body size. The visitation activity of A. lebedevi in the managed population was more frequent than that in the natural population likely because pollinators preferred to visit areas with greater resource availability. Therefore, the pollination success of M. sativa in the managed population was higher than in the natural population.
In many species, plant fitness may be lower if pollinator activities are missing or reduced (Pavlik, Ferguson, & Nelson, 1993).
Previous studies indicated that reduced pollinator activity could disrupt plant-pollinator interactions, and thus reduce seed set and gene flow of plant populations (Jennersten, 1988;Robertson, Kelly, Ladley, & Sparrow, 1999). Empirical evidence suggests that changes in pollinator activity might strongly influence plant pollination success (Goverde et al., 2002). Our results showed M. sativa is self-incompatible, and insect pollination played an important role in outcrossing. Many researchers have pointed out that floral display size is a good indicator of reward, it is common that pollinators selection on floral display size (Bauer et al., 2017;Worley & Barrett, 2000). The environment may affect floral display size, and it also plays an important role in pollinator activity for a reward. In the managed population, plants had greater resource availability to attract more pollinators because human management provided water and mown. This may support the assertion that environmental heterogeneity affects pollinator activity and pollination success.

CO N FLI C T O F I NTE R E S T S
All authors declare that we do not have any competing financial interests.

AUTH O R CO NTR I B UTI O N S
MC designed the experiment. MC wrote the manuscript; XYZ and XAZ provided editorial advice. All the authors have read and approved the manuscript. We thank Urat Desert-grassland Research Station and Naiman Desertification Research Station for all the help and support during this study.